diff --git a/Grid/DisableWarnings.h b/Grid/DisableWarnings.h
index b3257c62..8ea219fb 100644
--- a/Grid/DisableWarnings.h
+++ b/Grid/DisableWarnings.h
@@ -30,8 +30,34 @@ directory
 #ifndef DISABLE_WARNINGS_H
 #define DISABLE_WARNINGS_H
 
+
+
+#if defined __GNUC__ && __GNUC__>=6
+#pragma GCC diagnostic ignored "-Wignored-attributes"
+#endif
+
  //disables and intel compiler specific warning (in json.hpp)
 #pragma warning disable 488  
 
+#ifdef __NVCC__
+ //disables nvcc specific warning in json.hpp
+#pragma clang diagnostic ignored "-Wdeprecated-register"
+#pragma diag_suppress unsigned_compare_with_zero
+#pragma diag_suppress cast_to_qualified_type
+
+ //disables nvcc specific warning in many files
+#pragma diag_suppress esa_on_defaulted_function_ignored
+#pragma diag_suppress extra_semicolon
+
+//Eigen only
+#endif
+
+// Disable vectorisation in Eigen on the Power8/9 and PowerPC
+#ifdef  __ALTIVEC__
+#define  EIGEN_DONT_VECTORIZE
+#endif
+#ifdef  __VSX__
+#define  EIGEN_DONT_VECTORIZE
+#endif
 
 #endif
diff --git a/Grid/GridCore.h b/Grid/GridCore.h
index 3f31701a..a48d2d49 100644
--- a/Grid/GridCore.h
+++ b/Grid/GridCore.h
@@ -38,16 +38,19 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #ifndef GRID_BASE_H
 #define GRID_BASE_H
 
-#include <Grid/GridStd.h>
 
+#include <Grid/DisableWarnings.h>
+#include <Grid/Namespace.h>
+#include <Grid/GridStd.h>
+#include <Grid/threads/Pragmas.h>
 #include <Grid/perfmon/Timer.h>
 #include <Grid/perfmon/PerfCount.h>
+#include <Grid/util/Util.h>
 #include <Grid/log/Log.h>
 #include <Grid/allocator/AlignedAllocator.h>
 #include <Grid/simd/Simd.h>
-#include <Grid/serialisation/Serialisation.h>
 #include <Grid/threads/Threads.h>
-#include <Grid/util/Util.h>
+#include <Grid/serialisation/Serialisation.h>
 #include <Grid/util/Sha.h>
 #include <Grid/communicator/Communicator.h> 
 #include <Grid/cartesian/Cartesian.h>    
@@ -57,5 +60,6 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #include <Grid/stencil/Stencil.h>      
 #include <Grid/parallelIO/BinaryIO.h>
 #include <Grid/algorithms/Algorithms.h>   
+NAMESPACE_CHECK(GridCore)
 
 #endif
diff --git a/Grid/GridQCDcore.h b/Grid/GridQCDcore.h
index 7f50761f..cae6f43f 100644
--- a/Grid/GridQCDcore.h
+++ b/Grid/GridQCDcore.h
@@ -38,5 +38,6 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #include <Grid/qcd/spin/Spin.h>
 #include <Grid/qcd/utils/Utils.h>
 #include <Grid/qcd/representations/Representations.h>
+NAMESPACE_CHECK(GridQCDCore);
 
 #endif
diff --git a/Grid/GridStd.h b/Grid/GridStd.h
index 097e62ab..16cfcf50 100644
--- a/Grid/GridStd.h
+++ b/Grid/GridStd.h
@@ -7,6 +7,7 @@
 #include <cassert>
 #include <complex>
 #include <vector>
+#include <array>
 #include <string>
 #include <iostream>
 #include <iomanip>
diff --git a/Grid/Grid_Eigen_Dense.h b/Grid/Grid_Eigen_Dense.h
index dd330dd9..f9bccf2d 100644
--- a/Grid/Grid_Eigen_Dense.h
+++ b/Grid/Grid_Eigen_Dense.h
@@ -1,14 +1,41 @@
+#include <Grid/GridCore.h>
 #pragma once
 // Force Eigen to use MKL if Grid has been configured with --enable-mkl
 #ifdef USE_MKL
 #define EIGEN_USE_MKL_ALL
 #endif
 
+
 #if defined __GNUC__
 #pragma GCC diagnostic push
 #pragma GCC diagnostic ignored "-Wdeprecated-declarations"
 #endif
+
+/* NVCC save and restore compile environment*/
+#ifdef __NVCC__
+#pragma push
+#pragma diag_suppress code_is_unreachable
+#pragma push_macro("__CUDA_ARCH__")
+#pragma push_macro("__NVCC__")
+#pragma push_macro("__CUDACC__")
+#undef __NVCC__
+#undef __CUDACC__
+#undef __CUDA_ARCH__
+#define __NVCC__REDEFINE__
+#endif
+
 #include <Grid/Eigen/Dense>
+#include <Grid/Eigen/unsupported/CXX11/Tensor>
+
+/* NVCC restore */
+#ifdef __NVCC__REDEFINE__
+#pragma pop_macro("__CUDACC__")
+#pragma pop_macro("__NVCC__")
+#pragma pop_macro("__CUDA_ARCH__")
+#pragma pop
+#endif
+
 #if defined __GNUC__
 #pragma GCC diagnostic pop
 #endif
+
diff --git a/Grid/Grid_Eigen_Tensor.h b/Grid/Grid_Eigen_Tensor.h
new file mode 100644
index 00000000..3a2120c5
--- /dev/null
+++ b/Grid/Grid_Eigen_Tensor.h
@@ -0,0 +1 @@
+#include <Grid/Grid_Eigen_Dense.h>
diff --git a/Grid/Namespace.h b/Grid/Namespace.h
new file mode 100644
index 00000000..29b229fa
--- /dev/null
+++ b/Grid/Namespace.h
@@ -0,0 +1,38 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/Namespace.h
+
+Copyright (C) 2016
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+#include <type_traits>
+#include <cassert>
+
+#define NAMESPACE_BEGIN(A) namespace A {
+#define NAMESPACE_END(A)   }
+#define GRID_NAMESPACE_BEGIN NAMESPACE_BEGIN(Grid)
+#define GRID_NAMESPACE_END   NAMESPACE_END(Grid)
+#define NAMESPACE_CHECK(x) struct namespaceTEST##x {};  static_assert(std::is_same<namespaceTEST##x, ::namespaceTEST##x>::value,"Not in :: at"  ); 
diff --git a/Grid/algorithms/CoarsenedMatrix.h b/Grid/algorithms/CoarsenedMatrix.h
index a6c6c030..a373bc0a 100644
--- a/Grid/algorithms/CoarsenedMatrix.h
+++ b/Grid/algorithms/CoarsenedMatrix.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -26,501 +26,487 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef  GRID_ALGORITHM_COARSENED_MATRIX_H
 #define  GRID_ALGORITHM_COARSENED_MATRIX_H
 
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  class Geometry {
-    //    int dimension;
-  public:
-    int npoint;
-    std::vector<int> directions   ;
-    std::vector<int> displacements;
+class Geometry {
+  //    int dimension;
+public:
+  int npoint;
+  std::vector<int> directions   ;
+  std::vector<int> displacements;
 
   Geometry(int _d)  {
   
-      int base = (_d==5) ? 1:0;
+    int base = (_d==5) ? 1:0;
 
-      // make coarse grid stencil for 4d , not 5d
-      if ( _d==5 ) _d=4;
+    // make coarse grid stencil for 4d , not 5d
+    if ( _d==5 ) _d=4;
 
-      npoint = 2*_d+1;
-      directions.resize(npoint);
-      displacements.resize(npoint);
-      for(int d=0;d<_d;d++){
-	directions[2*d  ] = d+base;
-	directions[2*d+1] = d+base;
-	displacements[2*d  ] = +1;
-	displacements[2*d+1] = -1;
-      }
-      directions   [2*_d]=0;
-      displacements[2*_d]=0;
+    npoint = 2*_d+1;
+    directions.resize(npoint);
+    displacements.resize(npoint);
+    for(int d=0;d<_d;d++){
+      directions[2*d  ] = d+base;
+      directions[2*d+1] = d+base;
+      displacements[2*d  ] = +1;
+      displacements[2*d+1] = -1;
+    }
+    directions   [2*_d]=0;
+    displacements[2*_d]=0;
       
-      //// report back
-      std::cout<<GridLogMessage<<"directions    :";
-      for(int d=0;d<npoint;d++) std::cout<< directions[d]<< " ";
-      std::cout <<std::endl;
-      std::cout<<GridLogMessage<<"displacements :";
-      for(int d=0;d<npoint;d++) std::cout<< displacements[d]<< " ";
-      std::cout<<std::endl;
-    }
+    //// report back
+    std::cout<<GridLogMessage<<"directions    :";
+    for(int d=0;d<npoint;d++) std::cout<< directions[d]<< " ";
+    std::cout <<std::endl;
+    std::cout<<GridLogMessage<<"displacements :";
+    for(int d=0;d<npoint;d++) std::cout<< displacements[d]<< " ";
+    std::cout<<std::endl;
+  }
   
-    /*
-      // Original cleaner code
-    Geometry(int _d) : dimension(_d), npoint(2*_d+1), directions(npoint), displacements(npoint) {
-      for(int d=0;d<dimension;d++){
-	directions[2*d  ] = d;
-	directions[2*d+1] = d;
-	displacements[2*d  ] = +1;
-	displacements[2*d+1] = -1;
-      }
-      directions   [2*dimension]=0;
-      displacements[2*dimension]=0;
-    }
-    std::vector<int> GetDelta(int point) {
-      std::vector<int> delta(dimension,0);
-      delta[directions[point]] = displacements[point];
-      return delta;
-    };
-    */    
-
+  /*
+  // Original cleaner code
+  Geometry(int _d) : dimension(_d), npoint(2*_d+1), directions(npoint), displacements(npoint) {
+  for(int d=0;d<dimension;d++){
+  directions[2*d  ] = d;
+  directions[2*d+1] = d;
+  displacements[2*d  ] = +1;
+  displacements[2*d+1] = -1;
+  }
+  directions   [2*dimension]=0;
+  displacements[2*dimension]=0;
+  }
+  std::vector<int> GetDelta(int point) {
+  std::vector<int> delta(dimension,0);
+  delta[directions[point]] = displacements[point];
+  return delta;
   };
+  */    
+
+};
   
-  template<class Fobj,class CComplex,int nbasis>
-  class Aggregation   {
-  public:
-    typedef iVector<CComplex,nbasis >             siteVector;
-    typedef Lattice<siteVector>                 CoarseVector;
-    typedef Lattice<iMatrix<CComplex,nbasis > > CoarseMatrix;
+template<class Fobj,class CComplex,int nbasis>
+class Aggregation   {
+public:
+  typedef iVector<CComplex,nbasis >             siteVector;
+  typedef Lattice<siteVector>                 CoarseVector;
+  typedef Lattice<iMatrix<CComplex,nbasis > > CoarseMatrix;
 
-    typedef Lattice< CComplex >   CoarseScalar; // used for inner products on fine field
-    typedef Lattice<Fobj >        FineField;
-
-    GridBase *CoarseGrid;
-    GridBase *FineGrid;
-    std::vector<Lattice<Fobj> > subspace;
-    int checkerboard;
+  typedef Lattice< CComplex >   CoarseScalar; // used for inner products on fine field
+  typedef Lattice<Fobj >        FineField;
 
+  GridBase *CoarseGrid;
+  GridBase *FineGrid;
+  std::vector<Lattice<Fobj> > subspace;
+  int checkerboard;
+  int Checkerboard(void){return checkerboard;}
   Aggregation(GridBase *_CoarseGrid,GridBase *_FineGrid,int _checkerboard) : 
     CoarseGrid(_CoarseGrid),
-      FineGrid(_FineGrid),
-      subspace(nbasis,_FineGrid),
-      checkerboard(_checkerboard)
-	{
-	};
+    FineGrid(_FineGrid),
+    subspace(nbasis,_FineGrid),
+    checkerboard(_checkerboard)
+  {
+  };
   
-    void Orthogonalise(void){
-      CoarseScalar InnerProd(CoarseGrid); 
-      std::cout << GridLogMessage <<" Gramm-Schmidt pass 1"<<std::endl;
-      blockOrthogonalise(InnerProd,subspace);
-      std::cout << GridLogMessage <<" Gramm-Schmidt pass 2"<<std::endl;
-      blockOrthogonalise(InnerProd,subspace);
-      //      std::cout << GridLogMessage <<" Gramm-Schmidt checking orthogonality"<<std::endl;
-      //      CheckOrthogonal();
-    } 
-    void CheckOrthogonal(void){
-      CoarseVector iProj(CoarseGrid); 
-      CoarseVector eProj(CoarseGrid); 
-      for(int i=0;i<nbasis;i++){
-	blockProject(iProj,subspace[i],subspace);
-	eProj=zero; 
-	parallel_for(int ss=0;ss<CoarseGrid->oSites();ss++){
-	  eProj._odata[ss](i)=CComplex(1.0);
-	}
-	eProj=eProj - iProj;
-	std::cout<<GridLogMessage<<"Orthog check error "<<i<<" " << norm2(eProj)<<std::endl;
-      }
-      std::cout<<GridLogMessage <<"CheckOrthog done"<<std::endl;
+  void Orthogonalise(void){
+    CoarseScalar InnerProd(CoarseGrid); 
+    std::cout << GridLogMessage <<" Gramm-Schmidt pass 1"<<std::endl;
+    blockOrthogonalise(InnerProd,subspace);
+    std::cout << GridLogMessage <<" Gramm-Schmidt pass 2"<<std::endl;
+    blockOrthogonalise(InnerProd,subspace);
+    //      std::cout << GridLogMessage <<" Gramm-Schmidt checking orthogonality"<<std::endl;
+    //      CheckOrthogonal();
+  } 
+  void CheckOrthogonal(void){
+    CoarseVector iProj(CoarseGrid); 
+    CoarseVector eProj(CoarseGrid); 
+    for(int i=0;i<nbasis;i++){
+      blockProject(iProj,subspace[i],subspace);
+      eProj=Zero(); 
+      thread_for(ss, CoarseGrid->oSites(),{
+	eProj[ss](i)=CComplex(1.0);
+      });
+      eProj=eProj - iProj;
+      std::cout<<GridLogMessage<<"Orthog check error "<<i<<" " << norm2(eProj)<<std::endl;
     }
-    void ProjectToSubspace(CoarseVector &CoarseVec,const FineField &FineVec){
-      blockProject(CoarseVec,FineVec,subspace);
-    }
-    void PromoteFromSubspace(const CoarseVector &CoarseVec,FineField &FineVec){
-      FineVec.checkerboard = subspace[0].checkerboard;
-      blockPromote(CoarseVec,FineVec,subspace);
-    }
-    void CreateSubspaceRandom(GridParallelRNG &RNG){
-      for(int i=0;i<nbasis;i++){
-	random(RNG,subspace[i]);
-	std::cout<<GridLogMessage<<" norm subspace["<<i<<"] "<<norm2(subspace[i])<<std::endl;
-      }
-      Orthogonalise();
+    std::cout<<GridLogMessage <<"CheckOrthog done"<<std::endl;
+  }
+  void ProjectToSubspace(CoarseVector &CoarseVec,const FineField &FineVec){
+    blockProject(CoarseVec,FineVec,subspace);
+  }
+  void PromoteFromSubspace(const CoarseVector &CoarseVec,FineField &FineVec){
+    FineVec.Checkerboard() = subspace[0].Checkerboard();
+    blockPromote(CoarseVec,FineVec,subspace);
+  }
+  void CreateSubspaceRandom(GridParallelRNG &RNG){
+    for(int i=0;i<nbasis;i++){
+      random(RNG,subspace[i]);
+      std::cout<<GridLogMessage<<" norm subspace["<<i<<"] "<<norm2(subspace[i])<<std::endl;
     }
+    Orthogonalise();
+  }
 
-    /*
+  /*
     virtual void CreateSubspaceLanczos(GridParallelRNG  &RNG,LinearOperatorBase<FineField> &hermop,int nn=nbasis) 
     {
-      // Run a Lanczos with sloppy convergence
-	const int Nstop = nn;
-	const int Nk = nn+20;
-	const int Np = nn+20;
-	const int Nm = Nk+Np;
-	const int MaxIt= 10000;
-	RealD resid = 1.0e-3;
+    // Run a Lanczos with sloppy convergence
+    const int Nstop = nn;
+    const int Nk = nn+20;
+    const int Np = nn+20;
+    const int Nm = Nk+Np;
+    const int MaxIt= 10000;
+    RealD resid = 1.0e-3;
 
-	Chebyshev<FineField> Cheb(0.5,64.0,21);
-	ImplicitlyRestartedLanczos<FineField> IRL(hermop,Cheb,Nstop,Nk,Nm,resid,MaxIt);
-	//	IRL.lock = 1;
+    Chebyshev<FineField> Cheb(0.5,64.0,21);
+    ImplicitlyRestartedLanczos<FineField> IRL(hermop,Cheb,Nstop,Nk,Nm,resid,MaxIt);
+    //	IRL.lock = 1;
 
-	FineField noise(FineGrid); gaussian(RNG,noise);
-	FineField tmp(FineGrid); 
-	std::vector<RealD>     eval(Nm);
-	std::vector<FineField> evec(Nm,FineGrid);
+    FineField noise(FineGrid); gaussian(RNG,noise);
+    FineField tmp(FineGrid); 
+    std::vector<RealD>     eval(Nm);
+    std::vector<FineField> evec(Nm,FineGrid);
 
-	int Nconv;
-	IRL.calc(eval,evec,
-		 noise,
-		 Nconv);
+    int Nconv;
+    IRL.calc(eval,evec,
+    noise,
+    Nconv);
 
-    	// pull back nn vectors
-	for(int b=0;b<nn;b++){
+    // pull back nn vectors
+    for(int b=0;b<nn;b++){
 
-	  subspace[b]   = evec[b];
+    subspace[b]   = evec[b];
 
-	  std::cout << GridLogMessage <<"subspace["<<b<<"] = "<<norm2(subspace[b])<<std::endl;
+    std::cout << GridLogMessage <<"subspace["<<b<<"] = "<<norm2(subspace[b])<<std::endl;
 
-	  hermop.Op(subspace[b],tmp); 
-	  std::cout<<GridLogMessage << "filtered["<<b<<"] <f|MdagM|f> "<<norm2(tmp)<<std::endl;
+    hermop.Op(subspace[b],tmp); 
+    std::cout<<GridLogMessage << "filtered["<<b<<"] <f|MdagM|f> "<<norm2(tmp)<<std::endl;
 
-	  noise = tmp -  sqrt(eval[b])*subspace[b] ;
+    noise = tmp -  sqrt(eval[b])*subspace[b] ;
 
-	  std::cout<<GridLogMessage << " lambda_"<<b<<" = "<< eval[b] <<"  ;  [ M - Lambda ]_"<<b<<" vec_"<<b<<"  = " <<norm2(noise)<<std::endl;
+    std::cout<<GridLogMessage << " lambda_"<<b<<" = "<< eval[b] <<"  ;  [ M - Lambda ]_"<<b<<" vec_"<<b<<"  = " <<norm2(noise)<<std::endl;
 
-	  noise = tmp +  eval[b]*subspace[b] ;
+    noise = tmp +  eval[b]*subspace[b] ;
 
-	  std::cout<<GridLogMessage << " lambda_"<<b<<" = "<< eval[b] <<"  ;  [ M - Lambda ]_"<<b<<" vec_"<<b<<"  = " <<norm2(noise)<<std::endl;
+    std::cout<<GridLogMessage << " lambda_"<<b<<" = "<< eval[b] <<"  ;  [ M - Lambda ]_"<<b<<" vec_"<<b<<"  = " <<norm2(noise)<<std::endl;
 
-	}
-	Orthogonalise();
-	for(int b=0;b<nn;b++){
-	  std::cout << GridLogMessage <<"subspace["<<b<<"] = "<<norm2(subspace[b])<<std::endl;
-	}
     }
-    */
-    virtual void CreateSubspace(GridParallelRNG  &RNG,LinearOperatorBase<FineField> &hermop,int nn=nbasis) {
+    Orthogonalise();
+    for(int b=0;b<nn;b++){
+    std::cout << GridLogMessage <<"subspace["<<b<<"] = "<<norm2(subspace[b])<<std::endl;
+    }
+    }
+  */
+  virtual void CreateSubspace(GridParallelRNG  &RNG,LinearOperatorBase<FineField> &hermop,int nn=nbasis) {
 
-      RealD scale;
+    RealD scale;
 
-      ConjugateGradient<FineField> CG(1.0e-2,10000);
-      FineField noise(FineGrid);
-      FineField Mn(FineGrid);
+    ConjugateGradient<FineField> CG(1.0e-2,10000);
+    FineField noise(FineGrid);
+    FineField Mn(FineGrid);
 
-      for(int b=0;b<nn;b++){
+    for(int b=0;b<nn;b++){
 	
-	subspace[b] = zero;
-	gaussian(RNG,noise);
+      subspace[b] = Zero();
+      gaussian(RNG,noise);
+      scale = std::pow(norm2(noise),-0.5); 
+      noise=noise*scale;
+	
+      hermop.Op(noise,Mn); std::cout<<GridLogMessage << "noise   ["<<b<<"] <n|MdagM|n> "<<norm2(Mn)<<std::endl;
+
+      for(int i=0;i<1;i++){
+
+	CG(hermop,noise,subspace[b]);
+
+	noise = subspace[b];
 	scale = std::pow(norm2(noise),-0.5); 
 	noise=noise*scale;
 
-	hermop.Op(noise,Mn); std::cout<<GridLogMessage << "noise   ["<<b<<"] <n|MdagM|n> "<<norm2(Mn)<<std::endl;
-
-	for(int i=0;i<1;i++){
-
-	  CG(hermop,noise,subspace[b]);
-
-	  noise = subspace[b];
-	  scale = std::pow(norm2(noise),-0.5); 
-	  noise=noise*scale;
-
-	}
-
-	hermop.Op(noise,Mn); std::cout<<GridLogMessage << "filtered["<<b<<"] <f|MdagM|f> "<<norm2(Mn)<<std::endl;
-	subspace[b]   = noise;
-
       }
 
-      Orthogonalise();
+      hermop.Op(noise,Mn); std::cout<<GridLogMessage << "filtered["<<b<<"] <f|MdagM|f> "<<norm2(Mn)<<std::endl;
+      subspace[b]   = noise;
 
     }
-  };
-  // Fine Object == (per site) type of fine field
-  // nbasis      == number of deflation vectors
-  template<class Fobj,class CComplex,int nbasis>
-  class CoarsenedMatrix : public SparseMatrixBase<Lattice<iVector<CComplex,nbasis > > >  {
-  public:
+
+    Orthogonalise();
+
+  }
+};
+// Fine Object == (per site) type of fine field
+// nbasis      == number of deflation vectors
+template<class Fobj,class CComplex,int nbasis>
+class CoarsenedMatrix : public SparseMatrixBase<Lattice<iVector<CComplex,nbasis > > >  {
+public:
     
-    typedef iVector<CComplex,nbasis >             siteVector;
-    typedef Lattice<siteVector>                 CoarseVector;
-    typedef Lattice<iMatrix<CComplex,nbasis > > CoarseMatrix;
+  typedef iVector<CComplex,nbasis >             siteVector;
+  typedef Lattice<siteVector>                 CoarseVector;
+  typedef Lattice<iMatrix<CComplex,nbasis > > CoarseMatrix;
 
-    typedef Lattice< CComplex >   CoarseScalar; // used for inner products on fine field
-    typedef Lattice<Fobj >        FineField;
+  typedef Lattice< CComplex >   CoarseScalar; // used for inner products on fine field
+  typedef Lattice<Fobj >        FineField;
 
-    ////////////////////
-    // Data members
-    ////////////////////
-    Geometry         geom;
-    GridBase *       _grid; 
-    CartesianStencil<siteVector,siteVector> Stencil; 
+  ////////////////////
+  // Data members
+  ////////////////////
+  Geometry         geom;
+  GridBase *       _grid; 
 
-    std::vector<CoarseMatrix> A;
+  CartesianStencil<siteVector,siteVector,int> Stencil; 
+
+  std::vector<CoarseMatrix> A;
 
       
-    ///////////////////////
-    // Interface
-    ///////////////////////
-    GridBase * Grid(void)         { return _grid; };   // this is all the linalg routines need to know
+  ///////////////////////
+  // Interface
+  ///////////////////////
+  GridBase * Grid(void)         { return _grid; };   // this is all the linalg routines need to know
 
-    RealD M (const CoarseVector &in, CoarseVector &out){
+  RealD M (const CoarseVector &in, CoarseVector &out){
 
-      conformable(_grid,in._grid);
-      conformable(in._grid,out._grid);
+    conformable(_grid,in.Grid());
+    conformable(in.Grid(),out.Grid());
 
-      SimpleCompressor<siteVector> compressor;
-      Stencil.HaloExchange(in,compressor);
+    SimpleCompressor<siteVector> compressor;
+    Stencil.HaloExchange(in,compressor);
+    auto in_v = in.View();
+    auto out_v = in.View();
+    thread_for(ss,Grid()->oSites(),{
+      siteVector res = Zero();
+      siteVector nbr;
+      int ptype;
+      StencilEntry *SE;
+      for(int point=0;point<geom.npoint;point++){
 
-      parallel_for(int ss=0;ss<Grid()->oSites();ss++){
-        siteVector res = zero;
-	siteVector nbr;
-	int ptype;
-	StencilEntry *SE;
-	for(int point=0;point<geom.npoint;point++){
-
-	  SE=Stencil.GetEntry(ptype,point,ss);
+	SE=Stencil.GetEntry(ptype,point,ss);
 	  
-	  if(SE->_is_local&&SE->_permute) { 
-	    permute(nbr,in._odata[SE->_offset],ptype);
-	  } else if(SE->_is_local) { 
-	    nbr = in._odata[SE->_offset];
-	  } else {
-	    nbr = Stencil.CommBuf()[SE->_offset];
-	  }
-	  res = res + A[point]._odata[ss]*nbr;
+	if(SE->_is_local&&SE->_permute) { 
+	  permute(nbr,in_v[SE->_offset],ptype);
+	} else if(SE->_is_local) { 
+	  nbr = in_v[SE->_offset];
+	} else {
+	  nbr = Stencil.CommBuf()[SE->_offset];
 	}
-	vstream(out._odata[ss],res);
+	auto A_point = A[point].View();
+	res = res + A_point[ss]*nbr;
       }
-      return norm2(out);
-    };
+      vstream(out_v[ss],res);
+    });
+    return norm2(out);
+  };
 
-    RealD Mdag (const CoarseVector &in, CoarseVector &out){
-      // // corresponds to Petrov-Galerkin coarsening
-      // return M(in,out);
+  RealD Mdag (const CoarseVector &in, CoarseVector &out){
+    // // corresponds to Petrov-Galerkin coarsening
+    // return M(in,out);
+    
+    // corresponds to Galerkin coarsening
+    CoarseVector tmp(Grid());
+    G5C(tmp, in);
+    M(tmp, out);
+    G5C(out, out);
+    return norm2(out);
+  };
 
-      // corresponds to Galerkin coarsening
-      CoarseVector tmp(Grid());
-      G5C(tmp, in);
-      M(tmp, out);
-      G5C(out, out);
-      return norm2(out);
-    };
+  void Mdir(const CoarseVector &in, CoarseVector &out, int dir, int disp){
+    
+    conformable(_grid,in.Grid());
+    conformable(in.Grid(),out.Grid());
+    
+    SimpleCompressor<siteVector> compressor;
+    Stencil.HaloExchange(in,compressor);
+    
+    auto point = [dir, disp](){
+      if(dir == 0 and disp == 0)
+	return 8;
+      else
+	return (4 * dir + 1 - disp) / 2;
+    }();
 
-    void Mdir(const CoarseVector &in, CoarseVector &out, int dir, int disp){
-
-      conformable(_grid,in._grid);
-      conformable(in._grid,out._grid);
-
-      SimpleCompressor<siteVector> compressor;
-      Stencil.HaloExchange(in,compressor);
-
-      auto point = [dir, disp](){
-        if(dir == 0 and disp == 0)
-          return 8;
-        else
-          return (4 * dir + 1 - disp) / 2;
-      }();
-
-      parallel_for(int ss=0;ss<Grid()->oSites();ss++){
-        siteVector res = zero;
-        siteVector nbr;
-        int ptype;
-        StencilEntry *SE;
-
-        SE=Stencil.GetEntry(ptype,point,ss);
-
-        if(SE->_is_local&&SE->_permute) {
-          permute(nbr,in._odata[SE->_offset],ptype);
-        } else if(SE->_is_local) {
-          nbr = in._odata[SE->_offset];
-        } else {
-          nbr = Stencil.CommBuf()[SE->_offset];
-        }
-
-        res = res + A[point]._odata[ss]*nbr;
-
-        vstream(out._odata[ss],res);
+    auto out_v = out.View();
+    auto in_v  = in.View();
+    thread_for(ss,Grid()->oSites(),{
+      siteVector res = Zero();
+      siteVector nbr;
+      int ptype;
+      StencilEntry *SE;
+      
+      SE=Stencil.GetEntry(ptype,point,ss);
+      
+      if(SE->_is_local&&SE->_permute) {
+	permute(nbr,in_v[SE->_offset],ptype);
+      } else if(SE->_is_local) {
+	nbr = in_v[SE->_offset];
+      } else {
+	nbr = Stencil.CommBuf()[SE->_offset];
       }
-    };
 
-    void Mdiag(const CoarseVector &in, CoarseVector &out){
-      Mdir(in, out, 0, 0); // use the self coupling (= last) point of the stencil
-    };
+      auto A_point = A[point].View();
+      res = res + A_point[ss]*nbr;
+      
+      vstream(out_v[ss],res);
+    });
+  };
 
-    CoarsenedMatrix(GridCartesian &CoarseGrid) 	: 
+  void Mdiag(const CoarseVector &in, CoarseVector &out){
+    Mdir(in, out, 0, 0); // use the self coupling (= last) point of the stencil
+  };
 
-      _grid(&CoarseGrid),
-      geom(CoarseGrid._ndimension),
-      Stencil(&CoarseGrid,geom.npoint,Even,geom.directions,geom.displacements),
-      A(geom.npoint,&CoarseGrid)
-    {
-    };
+  
+ CoarsenedMatrix(GridCartesian &CoarseGrid) 	: 
 
-    void CoarsenOperator(GridBase *FineGrid,LinearOperatorBase<Lattice<Fobj> > &linop,
-			 Aggregation<Fobj,CComplex,nbasis> & Subspace){
+    _grid(&CoarseGrid),
+    geom(CoarseGrid._ndimension),
+    Stencil(&CoarseGrid,geom.npoint,Even,geom.directions,geom.displacements,0),
+    A(geom.npoint,&CoarseGrid)
+  {
+  };
 
-      FineField iblock(FineGrid); // contributions from within this block
-      FineField oblock(FineGrid); // contributions from outwith this block
+  void CoarsenOperator(GridBase *FineGrid,LinearOperatorBase<Lattice<Fobj> > &linop,
+		       Aggregation<Fobj,CComplex,nbasis> & Subspace){
 
-      FineField     phi(FineGrid);
-      FineField     tmp(FineGrid);
-      FineField     zz(FineGrid); zz=zero;
-      FineField    Mphi(FineGrid);
+    FineField iblock(FineGrid); // contributions from within this block
+    FineField oblock(FineGrid); // contributions from outwith this block
 
-      Lattice<iScalar<vInteger> > coor(FineGrid);
+    FineField     phi(FineGrid);
+    FineField     tmp(FineGrid);
+    FineField     zz(FineGrid); zz=Zero();
+    FineField    Mphi(FineGrid);
 
-      CoarseVector iProj(Grid()); 
-      CoarseVector oProj(Grid()); 
-      CoarseScalar InnerProd(Grid()); 
+    Lattice<iScalar<vInteger> > coor(FineGrid);
 
-      // Orthogonalise the subblocks over the basis
-      blockOrthogonalise(InnerProd,Subspace.subspace);
+    CoarseVector iProj(Grid()); 
+    CoarseVector oProj(Grid()); 
+    CoarseScalar InnerProd(Grid()); 
 
-      // Compute the matrix elements of linop between this orthonormal
-      // set of vectors.
-      int self_stencil=-1;
-      for(int p=0;p<geom.npoint;p++){ 
-	A[p]=zero;
-	if( geom.displacements[p]==0){
-	  self_stencil=p;
-	}
+    // Orthogonalise the subblocks over the basis
+    blockOrthogonalise(InnerProd,Subspace.subspace);
+
+    // Compute the matrix elements of linop between this orthonormal
+    // set of vectors.
+    int self_stencil=-1;
+    for(int p=0;p<geom.npoint;p++){ 
+      A[p]=Zero();
+      if( geom.displacements[p]==0){
+	self_stencil=p;
       }
-      assert(self_stencil!=-1);
+    }
+    assert(self_stencil!=-1);
 
-      for(int i=0;i<nbasis;i++){
-	phi=Subspace.subspace[i];
+    for(int i=0;i<nbasis;i++){
+      phi=Subspace.subspace[i];
 	
-	std::cout<<GridLogMessage<<"("<<i<<").."<<std::endl;
+      std::cout<<GridLogMessage<<"("<<i<<").."<<std::endl;
 
-	for(int p=0;p<geom.npoint;p++){ 
+      for(int p=0;p<geom.npoint;p++){ 
 
-	  int dir   = geom.directions[p];
-	  int disp  = geom.displacements[p];
+	int dir   = geom.directions[p];
+	int disp  = geom.displacements[p];
 
-	  Integer block=(FineGrid->_rdimensions[dir])/(Grid()->_rdimensions[dir]);
+	Integer block=(FineGrid->_rdimensions[dir])/(Grid()->_rdimensions[dir]);
 
-	  LatticeCoordinate(coor,dir);
+	LatticeCoordinate(coor,dir);
 
-	  if ( disp==0 ){
-	    linop.OpDiag(phi,Mphi);
-	  }
-	  else  {
-	    linop.OpDir(phi,Mphi,dir,disp); 
-	  }
-
-	  ////////////////////////////////////////////////////////////////////////
-	  // Pick out contributions coming from this cell and neighbour cell
-	  ////////////////////////////////////////////////////////////////////////
-	  if ( disp==0 ) {
-	    iblock = Mphi;
-	    oblock = zero;
-	  } else if ( disp==1 ) {
-	    oblock = where(mod(coor,block)==(block-1),Mphi,zz);
-	    iblock = where(mod(coor,block)!=(block-1),Mphi,zz);
-	  } else if ( disp==-1 ) {
-	    oblock = where(mod(coor,block)==(Integer)0,Mphi,zz);
-	    iblock = where(mod(coor,block)!=(Integer)0,Mphi,zz);
-	  } else {
-	    assert(0);
-	  }
-
-	  Subspace.ProjectToSubspace(iProj,iblock);
-	  Subspace.ProjectToSubspace(oProj,oblock);
-	  //	  blockProject(iProj,iblock,Subspace.subspace);
-	  //	  blockProject(oProj,oblock,Subspace.subspace);
-	  parallel_for(int ss=0;ss<Grid()->oSites();ss++){
-	    for(int j=0;j<nbasis;j++){
-	      if( disp!= 0 ) {
-		A[p]._odata[ss](j,i) = oProj._odata[ss](j);
-	      }
-	      A[self_stencil]._odata[ss](j,i) =	A[self_stencil]._odata[ss](j,i) + iProj._odata[ss](j);
-	    }
-	  }
+	if ( disp==0 ){
+	  linop.OpDiag(phi,Mphi);
 	}
+	else  {
+	  linop.OpDir(phi,Mphi,dir,disp); 
+	}
+
+	////////////////////////////////////////////////////////////////////////
+	// Pick out contributions coming from this cell and neighbour cell
+	////////////////////////////////////////////////////////////////////////
+	if ( disp==0 ) {
+	  iblock = Mphi;
+	  oblock = Zero();
+	} else if ( disp==1 ) {
+	  oblock = where(mod(coor,block)==(block-1),Mphi,zz);
+	  iblock = where(mod(coor,block)!=(block-1),Mphi,zz);
+	} else if ( disp==-1 ) {
+	  oblock = where(mod(coor,block)==(Integer)0,Mphi,zz);
+	  iblock = where(mod(coor,block)!=(Integer)0,Mphi,zz);
+	} else {
+	  assert(0);
+	}
+
+	Subspace.ProjectToSubspace(iProj,iblock);
+	Subspace.ProjectToSubspace(oProj,oblock);
+	//	  blockProject(iProj,iblock,Subspace.subspace);
+	//	  blockProject(oProj,oblock,Subspace.subspace);
+	auto iProj_v = iProj.View() ;
+	auto oProj_v = oProj.View() ;
+	auto A_p     =  A[p].View();
+	auto A_self  = A[self_stencil].View();
+	thread_for(ss, Grid()->oSites(),{
+	  for(int j=0;j<nbasis;j++){
+	    if( disp!= 0 ) {
+	      A_p[ss](j,i) = oProj_v[ss](j);
+	    }
+	    A_self[ss](j,i) =	A_self[ss](j,i) + iProj_v[ss](j);
+	  }
+	});
       }
+    }
 
 #if 0
-      ///////////////////////////
-      // test code worth preserving in if block
-      ///////////////////////////
-      std::cout<<GridLogMessage<< " Computed matrix elements "<< self_stencil <<std::endl;
-      for(int p=0;p<geom.npoint;p++){
-	std::cout<<GridLogMessage<< "A["<<p<<"]" << std::endl;
-	std::cout<<GridLogMessage<< A[p] << std::endl;
-      }
-      std::cout<<GridLogMessage<< " picking by block0 "<< self_stencil <<std::endl;
+    ///////////////////////////
+    // test code worth preserving in if block
+    ///////////////////////////
+    std::cout<<GridLogMessage<< " Computed matrix elements "<< self_stencil <<std::endl;
+    for(int p=0;p<geom.npoint;p++){
+      std::cout<<GridLogMessage<< "A["<<p<<"]" << std::endl;
+      std::cout<<GridLogMessage<< A[p] << std::endl;
+    }
+    std::cout<<GridLogMessage<< " picking by block0 "<< self_stencil <<std::endl;
 
-      phi=Subspace.subspace[0];
-      std::vector<int> bc(FineGrid->_ndimension,0);
+    phi=Subspace.subspace[0];
+    std::vector<int> bc(FineGrid->_ndimension,0);
 
-      blockPick(Grid(),phi,tmp,bc);      // Pick out a block
-      linop.Op(tmp,Mphi);                // Apply big dop
-      blockProject(iProj,Mphi,Subspace.subspace); // project it and print it
-      std::cout<<GridLogMessage<< " Computed matrix elements from block zero only "<<std::endl;
-      std::cout<<GridLogMessage<< iProj <<std::endl;
-      std::cout<<GridLogMessage<<"Computed Coarse Operator"<<std::endl;
+    blockPick(Grid(),phi,tmp,bc);      // Pick out a block
+    linop.Op(tmp,Mphi);                // Apply big dop
+    blockProject(iProj,Mphi,Subspace.subspace); // project it and print it
+    std::cout<<GridLogMessage<< " Computed matrix elements from block zero only "<<std::endl;
+    std::cout<<GridLogMessage<< iProj <<std::endl;
+    std::cout<<GridLogMessage<<"Computed Coarse Operator"<<std::endl;
 #endif
       //      ForceHermitian();
       // AssertHermitian();
       // ForceDiagonal();
+  }
+
+  void ForceHermitian(void) {
+    for(int d=0;d<4;d++){
+      int dd=d+1;
+      A[2*d] = adj(Cshift(A[2*d+1],dd,1));
     }
-    void ForceDiagonal(void) {
-
-
-      std::cout<<GridLogMessage<<"**************************************************"<<std::endl;
-      std::cout<<GridLogMessage<<"****   Forcing coarse operator to be diagonal ****"<<std::endl;
-      std::cout<<GridLogMessage<<"**************************************************"<<std::endl;
-      for(int p=0;p<8;p++){
-	A[p]=zero;
-      }
-
-      GridParallelRNG  RNG(Grid()); RNG.SeedFixedIntegers(std::vector<int>({55,72,19,17,34}));
-      Lattice<iScalar<CComplex> > val(Grid()); random(RNG,val);
-
-      Complex one(1.0);
-
-      iMatrix<CComplex,nbasis> ident;  ident=one;
-
-      val = val*adj(val);
-      val = val + 1.0;
-
-      A[8] = val*ident;
-
-      //      for(int s=0;s<Grid()->oSites();s++) {
-      //	A[8]._odata[s]=val._odata[s];
-      //      }
-    }
-    void ForceHermitian(void) {
-      for(int d=0;d<4;d++){
-	int dd=d+1;
-	A[2*d] = adj(Cshift(A[2*d+1],dd,1));
-      }
-      //      A[8] = 0.5*(A[8] + adj(A[8]));
-    }
-    void AssertHermitian(void) {
-      CoarseMatrix AA    (Grid());
-      CoarseMatrix AAc   (Grid());
-      CoarseMatrix Diff  (Grid());
-      for(int d=0;d<4;d++){
+    //      A[8] = 0.5*(A[8] + adj(A[8]));
+  }
+  void AssertHermitian(void) {
+    CoarseMatrix AA    (Grid());
+    CoarseMatrix AAc   (Grid());
+    CoarseMatrix Diff  (Grid());
+    for(int d=0;d<4;d++){
 	
-	int dd=d+1;
-	AAc = Cshift(A[2*d+1],dd,1);
-	AA  = A[2*d];
+      int dd=d+1;
+      AAc = Cshift(A[2*d+1],dd,1);
+      AA  = A[2*d];
 	
-	Diff = AA - adj(AAc);
+      Diff = AA - adj(AAc);
 
-	std::cout<<GridLogMessage<<"Norm diff dim "<<d<<" "<< norm2(Diff)<<std::endl;
-	std::cout<<GridLogMessage<<"Norm dim "<<d<<" "<< norm2(AA)<<std::endl;
+      std::cout<<GridLogMessage<<"Norm diff dim "<<d<<" "<< norm2(Diff)<<std::endl;
+      std::cout<<GridLogMessage<<"Norm dim "<<d<<" "<< norm2(AA)<<std::endl;
 	  
-      }
-      Diff = A[8] - adj(A[8]);
-      std::cout<<GridLogMessage<<"Norm diff local "<< norm2(Diff)<<std::endl;
-      std::cout<<GridLogMessage<<"Norm local "<< norm2(A[8])<<std::endl;
     }
+    Diff = A[8] - adj(A[8]);
+    std::cout<<GridLogMessage<<"Norm diff local "<< norm2(Diff)<<std::endl;
+    std::cout<<GridLogMessage<<"Norm local "<< norm2(A[8])<<std::endl;
+  }
     
-  };
+};
 
-}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/algorithms/FFT.h b/Grid/algorithms/FFT.h
index ec558ad9..0d568c95 100644
--- a/Grid/algorithms/FFT.h
+++ b/Grid/algorithms/FFT.h
@@ -1,5 +1,5 @@
 
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,8 +24,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef _GRID_FFT_H_
 #define _GRID_FFT_H_
 
@@ -38,64 +38,64 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #endif
 
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  template<class scalar> struct FFTW { };
+template<class scalar> struct FFTW { };
 
 #ifdef HAVE_FFTW	
-  template<> struct FFTW<ComplexD> {
-  public:
+template<> struct FFTW<ComplexD> {
+public:
 
-    typedef fftw_complex FFTW_scalar;
-    typedef fftw_plan    FFTW_plan;
+  typedef fftw_complex FFTW_scalar;
+  typedef fftw_plan    FFTW_plan;
 
-    static FFTW_plan fftw_plan_many_dft(int rank, const int *n,int howmany,
-					FFTW_scalar *in, const int *inembed,		
-					int istride, int idist,		
-					FFTW_scalar *out, const int *onembed,		
-					int ostride, int odist,		
-					int sign, unsigned flags) {
-      return ::fftw_plan_many_dft(rank,n,howmany,in,inembed,istride,idist,out,onembed,ostride,odist,sign,flags);
-    }	  
+  static FFTW_plan fftw_plan_many_dft(int rank, const int *n,int howmany,
+				      FFTW_scalar *in, const int *inembed,		
+				      int istride, int idist,		
+				      FFTW_scalar *out, const int *onembed,		
+				      int ostride, int odist,		
+				      int sign, unsigned flags) {
+    return ::fftw_plan_many_dft(rank,n,howmany,in,inembed,istride,idist,out,onembed,ostride,odist,sign,flags);
+  }	  
     
-    static void fftw_flops(const FFTW_plan p,double *add, double *mul, double *fmas){
-      ::fftw_flops(p,add,mul,fmas);
-    }
+  static void fftw_flops(const FFTW_plan p,double *add, double *mul, double *fmas){
+    ::fftw_flops(p,add,mul,fmas);
+  }
 
-    inline static void fftw_execute_dft(const FFTW_plan p,FFTW_scalar *in,FFTW_scalar *out) {
-      ::fftw_execute_dft(p,in,out);
-    }
-    inline static void fftw_destroy_plan(const FFTW_plan p) {
-      ::fftw_destroy_plan(p);
-    }
-  };
+  inline static void fftw_execute_dft(const FFTW_plan p,FFTW_scalar *in,FFTW_scalar *out) {
+    ::fftw_execute_dft(p,in,out);
+  }
+  inline static void fftw_destroy_plan(const FFTW_plan p) {
+    ::fftw_destroy_plan(p);
+  }
+};
 
-  template<> struct FFTW<ComplexF> {
-  public:
+template<> struct FFTW<ComplexF> {
+public:
 
-    typedef fftwf_complex FFTW_scalar;
-    typedef fftwf_plan    FFTW_plan;
+  typedef fftwf_complex FFTW_scalar;
+  typedef fftwf_plan    FFTW_plan;
 
-    static FFTW_plan fftw_plan_many_dft(int rank, const int *n,int howmany,
-					FFTW_scalar *in, const int *inembed,		
-					int istride, int idist,		
-					FFTW_scalar *out, const int *onembed,		
-					int ostride, int odist,		
-					int sign, unsigned flags) {
-      return ::fftwf_plan_many_dft(rank,n,howmany,in,inembed,istride,idist,out,onembed,ostride,odist,sign,flags);
-    }	  
+  static FFTW_plan fftw_plan_many_dft(int rank, const int *n,int howmany,
+				      FFTW_scalar *in, const int *inembed,		
+				      int istride, int idist,		
+				      FFTW_scalar *out, const int *onembed,		
+				      int ostride, int odist,		
+				      int sign, unsigned flags) {
+    return ::fftwf_plan_many_dft(rank,n,howmany,in,inembed,istride,idist,out,onembed,ostride,odist,sign,flags);
+  }	  
     
-    static void fftw_flops(const FFTW_plan p,double *add, double *mul, double *fmas){
-      ::fftwf_flops(p,add,mul,fmas);
-    }
+  static void fftw_flops(const FFTW_plan p,double *add, double *mul, double *fmas){
+    ::fftwf_flops(p,add,mul,fmas);
+  }
 
-    inline static void fftw_execute_dft(const FFTW_plan p,FFTW_scalar *in,FFTW_scalar *out) {
-      ::fftwf_execute_dft(p,in,out);
-    }
-    inline static void fftw_destroy_plan(const FFTW_plan p) {
-      ::fftwf_destroy_plan(p);
-    }
-  };
+  inline static void fftw_execute_dft(const FFTW_plan p,FFTW_scalar *in,FFTW_scalar *out) {
+    ::fftwf_execute_dft(p,in,out);
+  }
+  inline static void fftw_destroy_plan(const FFTW_plan p) {
+    ::fftwf_destroy_plan(p);
+  }
+};
 
 #endif
 
@@ -104,203 +104,188 @@ namespace Grid {
 #define FFTW_BACKWARD (+1)
 #endif
 
-  class FFT {
-  private:
+class FFT {
+private:
     
-    GridCartesian *vgrid;
-    GridCartesian *sgrid;
+  GridCartesian *vgrid;
+  GridCartesian *sgrid;
     
-    int Nd;
-    double flops;
-    double flops_call;
-    uint64_t usec;
+  int Nd;
+  double flops;
+  double flops_call;
+  uint64_t usec;
     
-    std::vector<int> dimensions;
-    std::vector<int> processors;
-    std::vector<int> processor_coor;
+  Coordinate dimensions;
+  Coordinate processors;
+  Coordinate processor_coor;
     
-  public:
+public:
     
-    static const int forward=FFTW_FORWARD;
-    static const int backward=FFTW_BACKWARD;
+  static const int forward=FFTW_FORWARD;
+  static const int backward=FFTW_BACKWARD;
     
-    double Flops(void) {return flops;}
-    double MFlops(void) {return flops/usec;}
-    double USec(void)   {return (double)usec;}    
+  double Flops(void) {return flops;}
+  double MFlops(void) {return flops/usec;}
+  double USec(void)   {return (double)usec;}    
 
-    FFT ( GridCartesian * grid ) :
+  FFT ( GridCartesian * grid ) :
     vgrid(grid),
     Nd(grid->_ndimension),
     dimensions(grid->_fdimensions),
     processors(grid->_processors),
     processor_coor(grid->_processor_coor)
-    {
-      flops=0;
-      usec =0;
-      std::vector<int> layout(Nd,1);
-      sgrid = new GridCartesian(dimensions,layout,processors);
-    };
-    
-    ~FFT ( void)  {
-      delete sgrid;
-    }
-    
-    template<class vobj>
-    void FFT_dim_mask(Lattice<vobj> &result,const Lattice<vobj> &source,std::vector<int> mask,int sign){
-
-      conformable(result._grid,vgrid);
-      conformable(source._grid,vgrid);
-      Lattice<vobj> tmp(vgrid);
-      tmp = source;
-      for(int d=0;d<Nd;d++){
-	if( mask[d] ) {
-	  FFT_dim(result,tmp,d,sign);
-	  tmp=result;
-	}
-      }
-    }
-
-    template<class vobj>
-    void FFT_all_dim(Lattice<vobj> &result,const Lattice<vobj> &source,int sign){
-      std::vector<int> mask(Nd,1);
-      FFT_dim_mask(result,source,mask,sign);
-    }
-
-
-    template<class vobj>
-    void FFT_dim(Lattice<vobj> &result,const Lattice<vobj> &source,int dim, int sign){
-#ifndef HAVE_FFTW
-      assert(0);
-#else
-      conformable(result._grid,vgrid);
-      conformable(source._grid,vgrid);
-
-      int L = vgrid->_ldimensions[dim];
-      int G = vgrid->_fdimensions[dim];
-      
-      std::vector<int> layout(Nd,1);
-      std::vector<int> pencil_gd(vgrid->_fdimensions);
-      
-      pencil_gd[dim] = G*processors[dim];
-      
-      // Pencil global vol LxLxGxLxL per node
-      GridCartesian pencil_g(pencil_gd,layout,processors);
-      
-      // Construct pencils
-      typedef typename vobj::scalar_object sobj;
-      typedef typename sobj::scalar_type   scalar;
-      
-      Lattice<sobj> pgbuf(&pencil_g);
-      
-
-      typedef typename FFTW<scalar>::FFTW_scalar FFTW_scalar;
-      typedef typename FFTW<scalar>::FFTW_plan   FFTW_plan;
-      
-      int Ncomp = sizeof(sobj)/sizeof(scalar);
-      int Nlow  = 1;
-      for(int d=0;d<dim;d++){
-        Nlow*=vgrid->_ldimensions[d];
-      }
-      
-      int rank = 1;  /* 1d transforms */
-      int n[] = {G}; /* 1d transforms of length G */
-      int howmany = Ncomp;
-      int odist,idist,istride,ostride;
-      idist   = odist   = 1;          /* Distance between consecutive FT's */
-      istride = ostride = Ncomp*Nlow; /* distance between two elements in the same FT */
-      int *inembed = n, *onembed = n;
-      
-      scalar div;
-	  if ( sign == backward ) div = 1.0/G;
-	  else if ( sign == forward ) div = 1.0;
-	  else assert(0);
-      
-      FFTW_plan p;
-      {
-        FFTW_scalar *in = (FFTW_scalar *)&pgbuf._odata[0];
-        FFTW_scalar *out= (FFTW_scalar *)&pgbuf._odata[0];
-        p = FFTW<scalar>::fftw_plan_many_dft(rank,n,howmany,
-                                             in,inembed,
-                                             istride,idist,
-                                             out,onembed,
-                                             ostride, odist,
-                                             sign,FFTW_ESTIMATE);
-      }
-      
-      // Barrel shift and collect global pencil
-      std::vector<int> lcoor(Nd), gcoor(Nd);
-      result = source;
-      int pc = processor_coor[dim];
-      for(int p=0;p<processors[dim];p++) {
-        PARALLEL_REGION
-        {
-          std::vector<int> cbuf(Nd);
-          sobj s;
-          
-          PARALLEL_FOR_LOOP_INTERN
-          for(int idx=0;idx<sgrid->lSites();idx++) {
-            sgrid->LocalIndexToLocalCoor(idx,cbuf);
-            peekLocalSite(s,result,cbuf);
-	    cbuf[dim]+=((pc+p) % processors[dim])*L;
-	    //            cbuf[dim]+=p*L;
-            pokeLocalSite(s,pgbuf,cbuf);
-          }
-        }
-        if (p != processors[dim] - 1)
-        {
-          result = Cshift(result,dim,L);
-        }
-      }
-      
-      // Loop over orthog coords
-      int NN=pencil_g.lSites();
-      GridStopWatch timer;
-      timer.Start();
-      PARALLEL_REGION
-      {
-        std::vector<int> cbuf(Nd);
-        
-        PARALLEL_FOR_LOOP_INTERN
-        for(int idx=0;idx<NN;idx++) {
-          pencil_g.LocalIndexToLocalCoor(idx, cbuf);
-          if ( cbuf[dim] == 0 ) {  // restricts loop to plane at lcoor[dim]==0
-            FFTW_scalar *in = (FFTW_scalar *)&pgbuf._odata[idx];
-            FFTW_scalar *out= (FFTW_scalar *)&pgbuf._odata[idx];
-            FFTW<scalar>::fftw_execute_dft(p,in,out);
-          }
-        }
-      }
-      timer.Stop();
-      
-      // performance counting
-      double add,mul,fma;
-      FFTW<scalar>::fftw_flops(p,&add,&mul,&fma);
-      flops_call = add+mul+2.0*fma;
-      usec += timer.useconds();
-      flops+= flops_call*NN;
-      
-      // writing out result
-      PARALLEL_REGION
-      {
-        std::vector<int> clbuf(Nd), cgbuf(Nd);
-        sobj s;
-        
-        PARALLEL_FOR_LOOP_INTERN
-        for(int idx=0;idx<sgrid->lSites();idx++) {
-          sgrid->LocalIndexToLocalCoor(idx,clbuf);
-          cgbuf = clbuf;
-          cgbuf[dim] = clbuf[dim]+L*pc;
-          peekLocalSite(s,pgbuf,cgbuf);
-          pokeLocalSite(s,result,clbuf);
-        }
-      }
-      result = result*div;
-      
-      // destroying plan
-      FFTW<scalar>::fftw_destroy_plan(p);
-#endif
-    }
+  {
+    flops=0;
+    usec =0;
+    Coordinate layout(Nd,1);
+    sgrid = new GridCartesian(dimensions,layout,processors);
   };
-}
+    
+  ~FFT ( void)  {
+    delete sgrid;
+  }
+    
+  template<class vobj>
+  void FFT_dim_mask(Lattice<vobj> &result,const Lattice<vobj> &source,Coordinate mask,int sign){
+
+    conformable(result.Grid(),vgrid);
+    conformable(source.Grid(),vgrid);
+    Lattice<vobj> tmp(vgrid);
+    tmp = source;
+    for(int d=0;d<Nd;d++){
+      if( mask[d] ) {
+	FFT_dim(result,tmp,d,sign);
+	tmp=result;
+      }
+    }
+  }
+
+  template<class vobj>
+  void FFT_all_dim(Lattice<vobj> &result,const Lattice<vobj> &source,int sign){
+    Coordinate mask(Nd,1);
+    FFT_dim_mask(result,source,mask,sign);
+  }
+
+
+  template<class vobj>
+  void FFT_dim(Lattice<vobj> &result,const Lattice<vobj> &source,int dim, int sign){
+#ifndef HAVE_FFTW
+    assert(0);
+#else
+    conformable(result.Grid(),vgrid);
+    conformable(source.Grid(),vgrid);
+
+    int L = vgrid->_ldimensions[dim];
+    int G = vgrid->_fdimensions[dim];
+      
+    Coordinate layout(Nd,1);
+    Coordinate pencil_gd(vgrid->_fdimensions);
+      
+    pencil_gd[dim] = G*processors[dim];
+      
+    // Pencil global vol LxLxGxLxL per node
+    GridCartesian pencil_g(pencil_gd,layout,processors);
+      
+    // Construct pencils
+    typedef typename vobj::scalar_object sobj;
+    typedef typename sobj::scalar_type   scalar;
+      
+    Lattice<sobj> pgbuf(&pencil_g);
+    auto pgbuf_v = pgbuf.View();
+
+    typedef typename FFTW<scalar>::FFTW_scalar FFTW_scalar;
+    typedef typename FFTW<scalar>::FFTW_plan   FFTW_plan;
+      
+    int Ncomp = sizeof(sobj)/sizeof(scalar);
+    int Nlow  = 1;
+    for(int d=0;d<dim;d++){
+      Nlow*=vgrid->_ldimensions[d];
+    }
+      
+    int rank = 1;  /* 1d transforms */
+    int n[] = {G}; /* 1d transforms of length G */
+    int howmany = Ncomp;
+    int odist,idist,istride,ostride;
+    idist   = odist   = 1;          /* Distance between consecutive FT's */
+    istride = ostride = Ncomp*Nlow; /* distance between two elements in the same FT */
+    int *inembed = n, *onembed = n;
+      
+    scalar div;
+    if ( sign == backward ) div = 1.0/G;
+    else if ( sign == forward ) div = 1.0;
+    else assert(0);
+      
+    FFTW_plan p;
+    {
+      FFTW_scalar *in = (FFTW_scalar *)&pgbuf_v[0];
+      FFTW_scalar *out= (FFTW_scalar *)&pgbuf_v[0];
+      p = FFTW<scalar>::fftw_plan_many_dft(rank,n,howmany,
+					   in,inembed,
+					   istride,idist,
+					   out,onembed,
+					   ostride, odist,
+					   sign,FFTW_ESTIMATE);
+    }
+      
+    // Barrel shift and collect global pencil
+    Coordinate lcoor(Nd), gcoor(Nd);
+    result = source;
+    int pc = processor_coor[dim];
+    for(int p=0;p<processors[dim];p++) {
+      thread_for(idx, sgrid->lSites(),{
+          Coordinate cbuf(Nd);
+          sobj s;
+	  sgrid->LocalIndexToLocalCoor(idx,cbuf);
+	  peekLocalSite(s,result,cbuf);
+	  cbuf[dim]+=((pc+p) % processors[dim])*L;
+	  //            cbuf[dim]+=p*L;
+	  pokeLocalSite(s,pgbuf,cbuf);
+      });
+      if (p != processors[dim] - 1) {
+	result = Cshift(result,dim,L);
+      }
+    }
+      
+    // Loop over orthog coords
+    int NN=pencil_g.lSites();
+    GridStopWatch timer;
+    timer.Start();
+    thread_for( idx,NN,{
+        Coordinate cbuf(Nd);
+	pencil_g.LocalIndexToLocalCoor(idx, cbuf);
+	if ( cbuf[dim] == 0 ) {  // restricts loop to plane at lcoor[dim]==0
+	  FFTW_scalar *in = (FFTW_scalar *)&pgbuf_v[idx];
+	  FFTW_scalar *out= (FFTW_scalar *)&pgbuf_v[idx];
+	  FFTW<scalar>::fftw_execute_dft(p,in,out);
+	}
+    });
+    timer.Stop();
+      
+    // performance counting
+    double add,mul,fma;
+    FFTW<scalar>::fftw_flops(p,&add,&mul,&fma);
+    flops_call = add+mul+2.0*fma;
+    usec += timer.useconds();
+    flops+= flops_call*NN;
+      
+    // writing out result
+    thread_for(idx,sgrid->lSites(),{
+	Coordinate clbuf(Nd), cgbuf(Nd);
+	sobj s;
+	sgrid->LocalIndexToLocalCoor(idx,clbuf);
+	cgbuf = clbuf;
+	cgbuf[dim] = clbuf[dim]+L*pc;
+	peekLocalSite(s,pgbuf,cgbuf);
+	pokeLocalSite(s,result,clbuf);
+    });
+    result = result*div;
+      
+    // destroying plan
+    FFTW<scalar>::fftw_destroy_plan(p);
+#endif
+  }
+};
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/algorithms/LinearOperator.h b/Grid/algorithms/LinearOperator.h
index a1be48f4..c309ad41 100644
--- a/Grid/algorithms/LinearOperator.h
+++ b/Grid/algorithms/LinearOperator.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,153 +24,152 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef  GRID_ALGORITHM_LINEAR_OP_H
-#define  GRID_ALGORITHM_LINEAR_OP_H
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once 
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  /////////////////////////////////////////////////////////////////////////////////////////////
-  // LinearOperators Take a something and return a something.
-  /////////////////////////////////////////////////////////////////////////////////////////////
-  //
-  // Hopefully linearity is satisfied and the AdjOp is indeed the Hermitian conjugateugate (transpose if real):
-  //SBase
-  //   i)  F(a x + b y) = aF(x) + b F(y).
-  //  ii)  <x|Op|y> = <y|AdjOp|x>^\ast
-  //
-  // Would be fun to have a test linearity & Herm Conj function!
-  /////////////////////////////////////////////////////////////////////////////////////////////
-    template<class Field> class LinearOperatorBase {
-    public:
+/////////////////////////////////////////////////////////////////////////////////////////////
+// LinearOperators Take a something and return a something.
+/////////////////////////////////////////////////////////////////////////////////////////////
+//
+// Hopefully linearity is satisfied and the AdjOp is indeed the Hermitian Conjugateugate (transpose if real):
+//SBase
+//   i)  F(a x + b y) = aF(x) + b F(y).
+//  ii)  <x|Op|y> = <y|AdjOp|x>^\ast
+//
+// Would be fun to have a test linearity & Herm Conj function!
+/////////////////////////////////////////////////////////////////////////////////////////////
+template<class Field> class LinearOperatorBase {
+public:
 
-      // Support for coarsening to a multigrid
-      virtual void OpDiag (const Field &in, Field &out) = 0; // Abstract base
-      virtual void OpDir  (const Field &in, Field &out,int dir,int disp) = 0; // Abstract base
+  // Support for coarsening to a multigrid
+  virtual void OpDiag (const Field &in, Field &out) = 0; // Abstract base
+  virtual void OpDir  (const Field &in, Field &out,int dir,int disp) = 0; // Abstract base
 
-      virtual void Op     (const Field &in, Field &out) = 0; // Abstract base
-      virtual void AdjOp  (const Field &in, Field &out) = 0; // Abstract base
-      virtual void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2) = 0;
-      virtual void HermOp(const Field &in, Field &out)=0;
-    };
+  virtual void Op     (const Field &in, Field &out) = 0; // Abstract base
+  virtual void AdjOp  (const Field &in, Field &out) = 0; // Abstract base
+  virtual void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2)=0;
+  virtual void HermOp(const Field &in, Field &out)=0;
+};
 
 
-  /////////////////////////////////////////////////////////////////////////////////////////////
-  // By sharing the class for Sparse Matrix across multiple operator wrappers, we can share code
-  // between RB and non-RB variants. Sparse matrix is like the fermion action def, and then
-  // the wrappers implement the specialisation of "Op" and "AdjOp" to the cases minimising
-  // replication of code.
-  //
-  // I'm not entirely happy with implementation; to share the Schur code between herm and non-herm
-  // while still having a "OpAndNorm" in the abstract base I had to implement it in both cases
-  // with an assert trap in the non-herm. This isn't right; there must be a better C++ way to
-  // do it, but I fear it required multiple inheritance and mixed in abstract base classes
-  /////////////////////////////////////////////////////////////////////////////////////////////
+/////////////////////////////////////////////////////////////////////////////////////////////
+// By sharing the class for Sparse Matrix across multiple operator wrappers, we can share code
+// between RB and non-RB variants. Sparse matrix is like the fermion action def, and then
+// the wrappers implement the specialisation of "Op" and "AdjOp" to the cases minimising
+// replication of code.
+//
+// I'm not entirely happy with implementation; to share the Schur code between herm and non-herm
+// while still having a "OpAndNorm" in the abstract base I had to implement it in both cases
+// with an assert trap in the non-herm. This isn't right; there must be a better C++ way to
+// do it, but I fear it required multiple inheritance and mixed in abstract base classes
+/////////////////////////////////////////////////////////////////////////////////////////////
 
-    ////////////////////////////////////////////////////////////////////
-    // Construct herm op from non-herm matrix
-    ////////////////////////////////////////////////////////////////////
-    template<class Matrix,class Field>
-    class MdagMLinearOperator : public LinearOperatorBase<Field> {
-      Matrix &_Mat;
-    public:
-    MdagMLinearOperator(Matrix &Mat): _Mat(Mat){};
+////////////////////////////////////////////////////////////////////
+// Construct herm op from non-herm matrix
+////////////////////////////////////////////////////////////////////
+template<class Matrix,class Field>
+class MdagMLinearOperator : public LinearOperatorBase<Field> {
+  Matrix &_Mat;
+public:
+  MdagMLinearOperator(Matrix &Mat): _Mat(Mat){};
 
-      // Support for coarsening to a multigrid
-      void OpDiag (const Field &in, Field &out) {
-	_Mat.Mdiag(in,out);
-      }
-      void OpDir  (const Field &in, Field &out,int dir,int disp) {
-	_Mat.Mdir(in,out,dir,disp);
-      }
-      void Op     (const Field &in, Field &out){
-	_Mat.M(in,out);
-      }
-      void AdjOp     (const Field &in, Field &out){
-	_Mat.Mdag(in,out);
-      }
-      void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){
-	_Mat.MdagM(in,out,n1,n2);
-      }
-      void HermOp(const Field &in, Field &out){
-	RealD n1,n2;
-	HermOpAndNorm(in,out,n1,n2);
-      }
-    };
+  // Support for coarsening to a multigrid
+  void OpDiag (const Field &in, Field &out) {
+    _Mat.Mdiag(in,out);
+  }
+  void OpDir  (const Field &in, Field &out,int dir,int disp) {
+    _Mat.Mdir(in,out,dir,disp);
+  }
+  void Op     (const Field &in, Field &out){
+    _Mat.M(in,out);
+  }
+  void AdjOp     (const Field &in, Field &out){
+    _Mat.Mdag(in,out);
+  }
+  void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){
+    _Mat.MdagM(in,out,n1,n2);
+  }
+  void HermOp(const Field &in, Field &out){
+    RealD n1,n2;
+    HermOpAndNorm(in,out,n1,n2);
+  }
+};
 
-    ////////////////////////////////////////////////////////////////////
-    // Construct herm op and shift it for mgrid smoother
-    ////////////////////////////////////////////////////////////////////
-    template<class Matrix,class Field>
-    class ShiftedMdagMLinearOperator : public LinearOperatorBase<Field> {
-      Matrix &_Mat;
-      RealD _shift;
-    public:
-    ShiftedMdagMLinearOperator(Matrix &Mat,RealD shift): _Mat(Mat), _shift(shift){};
-      // Support for coarsening to a multigrid
-      void OpDiag (const Field &in, Field &out) {
-	_Mat.Mdiag(in,out);
-	assert(0);
-      }
-      void OpDir  (const Field &in, Field &out,int dir,int disp) {
-	_Mat.Mdir(in,out,dir,disp);
-	assert(0);
-      }
-      void Op     (const Field &in, Field &out){
-	_Mat.M(in,out);
-	assert(0);
-      }
-      void AdjOp     (const Field &in, Field &out){
-	_Mat.Mdag(in,out);
-	assert(0);
-      }
-      void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){
-	_Mat.MdagM(in,out,n1,n2);
-	out = out + _shift*in;
+////////////////////////////////////////////////////////////////////
+// Construct herm op and shift it for mgrid smoother
+////////////////////////////////////////////////////////////////////
+template<class Matrix,class Field>
+class ShiftedMdagMLinearOperator : public LinearOperatorBase<Field> {
+  Matrix &_Mat;
+  RealD _shift;
+public:
+  ShiftedMdagMLinearOperator(Matrix &Mat,RealD shift): _Mat(Mat), _shift(shift){};
+  // Support for coarsening to a multigrid
+  void OpDiag (const Field &in, Field &out) {
+    _Mat.Mdiag(in,out);
+    assert(0);
+  }
+  void OpDir  (const Field &in, Field &out,int dir,int disp) {
+    _Mat.Mdir(in,out,dir,disp);
+    assert(0);
+  }
+  void Op     (const Field &in, Field &out){
+    _Mat.M(in,out);
+    assert(0);
+  }
+  void AdjOp     (const Field &in, Field &out){
+    _Mat.Mdag(in,out);
+    assert(0);
+  }
+  void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){
+    _Mat.MdagM(in,out,n1,n2);
+    out = out + _shift*in;
 
-	ComplexD dot;	
-	dot= innerProduct(in,out);
-	n1=real(dot);
-	n2=norm2(out);
-      }
-      void HermOp(const Field &in, Field &out){
-	RealD n1,n2;
-	HermOpAndNorm(in,out,n1,n2);
-      }
-    };
+    ComplexD dot;	
+    dot= innerProduct(in,out);
+    n1=real(dot);
+    n2=norm2(out);
+  }
+  void HermOp(const Field &in, Field &out){
+    RealD n1,n2;
+    HermOpAndNorm(in,out,n1,n2);
+  }
+};
 
-    ////////////////////////////////////////////////////////////////////
-    // Wrap an already herm matrix
-    ////////////////////////////////////////////////////////////////////
-    template<class Matrix,class Field>
-    class HermitianLinearOperator : public LinearOperatorBase<Field> {
-      Matrix &_Mat;
-    public:
-    HermitianLinearOperator(Matrix &Mat): _Mat(Mat){};
-      // Support for coarsening to a multigrid
-      void OpDiag (const Field &in, Field &out) {
-	_Mat.Mdiag(in,out);
-      }
-      void OpDir  (const Field &in, Field &out,int dir,int disp) {
-	_Mat.Mdir(in,out,dir,disp);
-      }
-      void Op     (const Field &in, Field &out){
-	_Mat.M(in,out);
-      }
-      void AdjOp     (const Field &in, Field &out){
-	_Mat.M(in,out);
-      }
-      void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){
-	_Mat.M(in,out);
+////////////////////////////////////////////////////////////////////
+// Wrap an already herm matrix
+////////////////////////////////////////////////////////////////////
+template<class Matrix,class Field>
+class HermitianLinearOperator : public LinearOperatorBase<Field> {
+  Matrix &_Mat;
+public:
+  HermitianLinearOperator(Matrix &Mat): _Mat(Mat){};
+  // Support for coarsening to a multigrid
+  void OpDiag (const Field &in, Field &out) {
+    _Mat.Mdiag(in,out);
+  }
+  void OpDir  (const Field &in, Field &out,int dir,int disp) {
+    _Mat.Mdir(in,out,dir,disp);
+  }
+  void Op     (const Field &in, Field &out){
+    _Mat.M(in,out);
+  }
+  void AdjOp     (const Field &in, Field &out){
+    _Mat.M(in,out);
+  }
+  void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){
+    _Mat.M(in,out);
 	
-	ComplexD dot= innerProduct(in,out); n1=real(dot);
-	n2=norm2(out);
-      }
-      void HermOp(const Field &in, Field &out){
-	_Mat.M(in,out);
-      }
-    };
+    ComplexD dot= innerProduct(in,out); n1=real(dot);
+    n2=norm2(out);
+  }
+  void HermOp(const Field &in, Field &out){
+    _Mat.M(in,out);
+  }
+};
 
     //////////////////////////////////////////////////////////
     // Even Odd Schur decomp operators; there are several
@@ -183,13 +182,13 @@ namespace Grid {
       virtual  RealD Mpc      (const Field &in, Field &out) =0;
       virtual  RealD MpcDag   (const Field &in, Field &out) =0;
       virtual void MpcDagMpc(const Field &in, Field &out,RealD &ni,RealD &no) {
-      Field tmp(in._grid);
-      tmp.checkerboard = in.checkerboard;
+      Field tmp(in.Grid());
+      tmp.Checkerboard() = in.Checkerboard();
 	ni=Mpc(in,tmp);
 	no=MpcDag(tmp,out);
       }
       virtual void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){
-      out.checkerboard = in.checkerboard;
+      out.Checkerboard() = in.Checkerboard();
 	MpcDagMpc(in,out,n1,n2);
       }
       virtual void HermOp(const Field &in, Field &out){
@@ -216,20 +215,20 @@ namespace Grid {
       Matrix &_Mat;
       SchurDiagMooeeOperator (Matrix &Mat): _Mat(Mat){};
       virtual  RealD Mpc      (const Field &in, Field &out) {
-      Field tmp(in._grid);
-      tmp.checkerboard = !in.checkerboard;
+      Field tmp(in.Grid());
+      tmp.Checkerboard() = !in.Checkerboard();
 	//std::cout <<"grid pointers: in._grid="<< in._grid << " out._grid=" << out._grid << "  _Mat.Grid=" << _Mat.Grid() << " _Mat.RedBlackGrid=" << _Mat.RedBlackGrid() << std::endl;
 
 	_Mat.Meooe(in,tmp);
 	_Mat.MooeeInv(tmp,out);
 	_Mat.Meooe(out,tmp);
 
-      //std::cout << "cb in " << in.checkerboard << "  cb out " << out.checkerboard << std::endl;
+      //std::cout << "cb in " << in.Checkerboard() << "  cb out " << out.Checkerboard() << std::endl;
 	_Mat.Mooee(in,out);
 	return axpy_norm(out,-1.0,tmp,out);
       }
       virtual  RealD MpcDag   (const Field &in, Field &out){
-	Field tmp(in._grid);
+	Field tmp(in.Grid());
 
 	_Mat.MeooeDag(in,tmp);
         _Mat.MooeeInvDag(tmp,out);
@@ -247,7 +246,7 @@ namespace Grid {
       SchurDiagOneOperator (Matrix &Mat): _Mat(Mat){};
 
       virtual  RealD Mpc      (const Field &in, Field &out) {
-	Field tmp(in._grid);
+	Field tmp(in.Grid());
 
 	_Mat.Meooe(in,out);
 	_Mat.MooeeInv(out,tmp);
@@ -257,7 +256,7 @@ namespace Grid {
 	return axpy_norm(out,-1.0,tmp,in);
       }
       virtual  RealD MpcDag   (const Field &in, Field &out){
-	Field tmp(in._grid);
+	Field tmp(in.Grid());
 
 	_Mat.MooeeInvDag(in,out);
 	_Mat.MeooeDag(out,tmp);
@@ -275,7 +274,7 @@ namespace Grid {
       SchurDiagTwoOperator (Matrix &Mat): _Mat(Mat){};
 
       virtual  RealD Mpc      (const Field &in, Field &out) {
-	Field tmp(in._grid);
+	Field tmp(in.Grid());
 
 	_Mat.MooeeInv(in,out);
 	_Mat.Meooe(out,tmp);
@@ -285,7 +284,7 @@ namespace Grid {
 	return axpy_norm(out,-1.0,tmp,in);
       }
       virtual  RealD MpcDag   (const Field &in, Field &out){
-	Field tmp(in._grid);
+	Field tmp(in.Grid());
 
 	_Mat.MeooeDag(in,out);
 	_Mat.MooeeInvDag(out,tmp);
@@ -315,7 +314,7 @@ namespace Grid {
       double tMeo;
       double taxpby_norm;
       uint64_t ncall;
-    public:
+public:
       void Report(void)
       {
 	std::cout << GridLogMessage << " HermOpAndNorm.Mpc "<< tMpc/ncall<<" usec "<<std::endl;
@@ -333,17 +332,17 @@ namespace Grid {
         taxpby_norm=0;
 	ncall=0;
       }
-      virtual void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){
+  virtual void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){
 	ncall++;
 	tMpc-=usecond();
-	n2 = Mpc(in,out);
+    n2 = Mpc(in,out);
 	tMpc+=usecond();
 	tIP-=usecond();
-	ComplexD dot= innerProduct(in,out);
+    ComplexD dot= innerProduct(in,out);
 	tIP+=usecond();
-	n1 = real(dot);
-      }
-      virtual void HermOp(const Field &in, Field &out){
+    n1 = real(dot);
+  }
+  virtual void HermOp(const Field &in, Field &out){
 	ncall++;
 	tMpc-=usecond();
 	_Mat.Meooe(in,out);
@@ -352,135 +351,145 @@ namespace Grid {
 	taxpby_norm-=usecond();
 	axpby(out,-1.0,mass*mass,tmp,in);
 	taxpby_norm+=usecond();
-      }
-      virtual  RealD Mpc      (const Field &in, Field &out) {
-	tMeo-=usecond();
-	_Mat.Meooe(in,out);
-	_Mat.Meooe(out,tmp);
-	tMeo+=usecond();
-	taxpby_norm-=usecond();
-	RealD nn=axpby_norm(out,-1.0,mass*mass,tmp,in);
-	taxpby_norm+=usecond();
-	return nn;
-      }
-      virtual  RealD MpcDag   (const Field &in, Field &out){
-	return Mpc(in,out);
-      }
-      virtual void MpcDagMpc(const Field &in, Field &out,RealD &ni,RealD &no) {
-	assert(0);// Never need with staggered
-      }
-    };
-    template<class Matrix,class Field> using SchurStagOperator = SchurStaggeredOperator<Matrix,Field>;
+  }
+  virtual  RealD Mpc      (const Field &in, Field &out) 
+  {
+
+    Field tmp(in.Grid());
+    Field tmp2(in.Grid());
+
+    //    std::cout << GridLogIterative << " HermOp.Mpc "<<std::endl;
+    _Mat.Mooee(in,out);
+    _Mat.Mooee(out,tmp);
+    //    std::cout << GridLogIterative << " HermOp.MooeeMooee "<<std::endl;
+
+    tMeo-=usecond();
+    _Mat.Meooe(in,out);
+    _Mat.Meooe(out,tmp);
+    tMeo+=usecond();
+    taxpby_norm-=usecond();
+    RealD nn=axpby_norm(out,-1.0,mass*mass,tmp,in);
+    taxpby_norm+=usecond();
+    return nn;
+  }
+  virtual  RealD MpcDag   (const Field &in, Field &out){
+    return Mpc(in,out);
+  }
+  virtual void MpcDagMpc(const Field &in, Field &out,RealD &ni,RealD &no) {
+    assert(0);// Never need with staggered
+  }
+};
+template<class Matrix,class Field> using SchurStagOperator = SchurStaggeredOperator<Matrix,Field>;
 
 
-    /////////////////////////////////////////////////////////////
-    // Base classes for functions of operators
-    /////////////////////////////////////////////////////////////
-    template<class Field> class OperatorFunction {
-    public:
-      virtual void operator() (LinearOperatorBase<Field> &Linop, const Field &in, Field &out) = 0;
-      virtual void operator() (LinearOperatorBase<Field> &Linop, const std::vector<Field> &in,std::vector<Field> &out) {
-	assert(in.size()==out.size());
-	for(int k=0;k<in.size();k++){
-	  (*this)(Linop,in[k],out[k]);
-	}
-      };
-    };
-
-    template<class Field> class LinearFunction {
-    public:
-      virtual void operator() (const Field &in, Field &out) = 0;
-    };
-
-    template<class Field> class IdentityLinearFunction : public LinearFunction<Field> {
-    public:
-      void operator() (const Field &in, Field &out){
-	out = in;
-      };
-    };
-
-
-    /////////////////////////////////////////////////////////////
-    // Base classes for Multishift solvers for operators
-    /////////////////////////////////////////////////////////////
-    template<class Field> class OperatorMultiFunction {
-    public:
-      virtual void operator() (LinearOperatorBase<Field> &Linop, const Field &in, std::vector<Field> &out) = 0;
-    };
-
-    // FIXME : To think about
-
-    // Chroma functionality list defining LinearOperator
-    /*
-     virtual void operator() (T& chi, const T& psi, enum PlusMinus isign) const = 0;
-     virtual void operator() (T& chi, const T& psi, enum PlusMinus isign, Real epsilon) const
-     virtual const Subset& subset() const = 0;
-     virtual unsigned long nFlops() const { return 0; }
-     virtual void deriv(P& ds_u, const T& chi, const T& psi, enum PlusMinus isign) const
-     class UnprecLinearOperator : public DiffLinearOperator<T,P,Q>
-       const Subset& subset() const {return all;}
-     };
-    */
-
-  ////////////////////////////////////////////////////////////////////////////////////////////
-  // Hermitian operator Linear function and operator function
-  ////////////////////////////////////////////////////////////////////////////////////////////
-    template<class Field>
-    class HermOpOperatorFunction : public OperatorFunction<Field> {
-      void operator() (LinearOperatorBase<Field> &Linop, const Field &in, Field &out) {
-	Linop.HermOp(in,out);
-      };
-    };
-
-    template<typename Field>
-      class PlainHermOp : public LinearFunction<Field> {
-    public:
-      LinearOperatorBase<Field> &_Linop;
-      
-      PlainHermOp(LinearOperatorBase<Field>& linop) : _Linop(linop) 
-      {}
-      
-      void operator()(const Field& in, Field& out) {
-	_Linop.HermOp(in,out);
-      }
-    };
-
-    template<typename Field>
-    class FunctionHermOp : public LinearFunction<Field> {
-    public:
-      OperatorFunction<Field>   & _poly;
-      LinearOperatorBase<Field> &_Linop;
-      
-      FunctionHermOp(OperatorFunction<Field> & poly,LinearOperatorBase<Field>& linop) 
-	: _poly(poly), _Linop(linop) {};
-      
-      void operator()(const Field& in, Field& out) {
-	_poly(_Linop,in,out);
-      }
-    };
-
-  template<class Field>
-  class Polynomial : public OperatorFunction<Field> {
-  private:
-    std::vector<RealD> Coeffs;
-  public:
-    Polynomial(std::vector<RealD> &_Coeffs) : Coeffs(_Coeffs) { };
-
-    // Implement the required interface
-    void operator() (LinearOperatorBase<Field> &Linop, const Field &in, Field &out) {
-
-      Field AtoN(in._grid);
-      Field Mtmp(in._grid);
-      AtoN = in;
-      out = AtoN*Coeffs[0];
-      for(int n=1;n<Coeffs.size();n++){
-	Mtmp = AtoN;
-	Linop.HermOp(Mtmp,AtoN);
-	out=out+AtoN*Coeffs[n];
-      }
-    };
+/////////////////////////////////////////////////////////////
+// Base classes for functions of operators
+/////////////////////////////////////////////////////////////
+template<class Field> class OperatorFunction {
+public:
+  virtual void operator() (LinearOperatorBase<Field> &Linop, const Field &in, Field &out) = 0;
+  virtual void operator() (LinearOperatorBase<Field> &Linop, const std::vector<Field> &in,std::vector<Field> &out) {
+    assert(in.size()==out.size());
+    for(int k=0;k<in.size();k++){
+      (*this)(Linop,in[k],out[k]);
+    }
   };
+};
 
-}
+template<class Field> class LinearFunction {
+public:
+  virtual void operator() (const Field &in, Field &out) = 0;
+};
 
-#endif
+template<class Field> class IdentityLinearFunction : public LinearFunction<Field> {
+public:
+  void operator() (const Field &in, Field &out){
+    out = in;
+  };
+};
+
+
+/////////////////////////////////////////////////////////////
+// Base classes for Multishift solvers for operators
+/////////////////////////////////////////////////////////////
+template<class Field> class OperatorMultiFunction {
+public:
+  virtual void operator() (LinearOperatorBase<Field> &Linop, const Field &in, std::vector<Field> &out) = 0;
+};
+
+// FIXME : To think about
+
+// Chroma functionality list defining LinearOperator
+/*
+  virtual void operator() (T& chi, const T& psi, enum PlusMinus isign) const = 0;
+  virtual void operator() (T& chi, const T& psi, enum PlusMinus isign, Real epsilon) const
+  virtual const Subset& subset() const = 0;
+  virtual unsigned long nFlops() const { return 0; }
+  virtual void deriv(P& ds_u, const T& chi, const T& psi, enum PlusMinus isign) const
+  class UnprecLinearOperator : public DiffLinearOperator<T,P,Q>
+  const Subset& subset() const {return all;}
+  };
+*/
+
+////////////////////////////////////////////////////////////////////////////////////////////
+// Hermitian operator Linear function and operator function
+////////////////////////////////////////////////////////////////////////////////////////////
+template<class Field>
+class HermOpOperatorFunction : public OperatorFunction<Field> {
+  void operator() (LinearOperatorBase<Field> &Linop, const Field &in, Field &out) {
+    Linop.HermOp(in,out);
+  };
+};
+
+template<typename Field>
+class PlainHermOp : public LinearFunction<Field> {
+public:
+  LinearOperatorBase<Field> &_Linop;
+      
+  PlainHermOp(LinearOperatorBase<Field>& linop) : _Linop(linop) 
+  {}
+      
+  void operator()(const Field& in, Field& out) {
+    _Linop.HermOp(in,out);
+  }
+};
+
+template<typename Field>
+class FunctionHermOp : public LinearFunction<Field> {
+public:
+  OperatorFunction<Field>   & _poly;
+  LinearOperatorBase<Field> &_Linop;
+      
+  FunctionHermOp(OperatorFunction<Field> & poly,LinearOperatorBase<Field>& linop) 
+    : _poly(poly), _Linop(linop) {};
+      
+  void operator()(const Field& in, Field& out) {
+    _poly(_Linop,in,out);
+  }
+};
+
+template<class Field>
+class Polynomial : public OperatorFunction<Field> {
+private:
+  std::vector<RealD> Coeffs;
+public:
+  using OperatorFunction<Field>::operator();
+
+  Polynomial(std::vector<RealD> &_Coeffs) : Coeffs(_Coeffs) { };
+
+  // Implement the required interface
+  void operator() (LinearOperatorBase<Field> &Linop, const Field &in, Field &out) {
+
+    Field AtoN(in.Grid());
+    Field Mtmp(in.Grid());
+    AtoN = in;
+    out = AtoN*Coeffs[0];
+    for(int n=1;n<Coeffs.size();n++){
+      Mtmp = AtoN;
+      Linop.HermOp(Mtmp,AtoN);
+      out=out+AtoN*Coeffs[n];
+    }
+  };
+};
+
+NAMESPACE_END(Grid);
diff --git a/Grid/algorithms/Preconditioner.h b/Grid/algorithms/Preconditioner.h
index a26016be..65ac3cf9 100644
--- a/Grid/algorithms/Preconditioner.h
+++ b/Grid/algorithms/Preconditioner.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,24 +23,24 @@ Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_PRECONDITIONER_H
 #define GRID_PRECONDITIONER_H
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  template<class Field> class Preconditioner :  public LinearFunction<Field> { 
-    virtual void operator()(const Field &src, Field & psi)=0;
-  };
+template<class Field> class Preconditioner :  public LinearFunction<Field> { 
+  virtual void operator()(const Field &src, Field & psi)=0;
+};
 
-  template<class Field> class TrivialPrecon :  public Preconditioner<Field> { 
-  public:
-    void operator()(const Field &src, Field & psi){
-      psi = src;
-    }
-    TrivialPrecon(void){};
-  };
+template<class Field> class TrivialPrecon :  public Preconditioner<Field> { 
+public:
+  void operator()(const Field &src, Field & psi){
+    psi = src;
+  }
+  TrivialPrecon(void){};
+};
 
-}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/algorithms/SparseMatrix.h b/Grid/algorithms/SparseMatrix.h
index 6cc617a6..ffed7527 100644
--- a/Grid/algorithms/SparseMatrix.h
+++ b/Grid/algorithms/SparseMatrix.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,38 +23,38 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef  GRID_ALGORITHM_SPARSE_MATRIX_H
 #define  GRID_ALGORITHM_SPARSE_MATRIX_H
 
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  /////////////////////////////////////////////////////////////////////////////////////////////
-  // Interface defining what I expect of a general sparse matrix, such as a Fermion action
-  /////////////////////////////////////////////////////////////////////////////////////////////
-    template<class Field> class SparseMatrixBase {
-    public:
-      virtual GridBase *Grid(void) =0;
-      // Full checkerboar operations
-      virtual RealD M    (const Field &in, Field &out)=0;
-      virtual RealD Mdag (const Field &in, Field &out)=0;
-      virtual void  MdagM(const Field &in, Field &out,RealD &ni,RealD &no) {
-	Field tmp (in._grid);
-	ni=M(in,tmp);
-	no=Mdag(tmp,out);
-      }
-      virtual  void Mdiag    (const Field &in, Field &out)=0;
-      virtual  void Mdir     (const Field &in, Field &out,int dir, int disp)=0;
-    };
+/////////////////////////////////////////////////////////////////////////////////////////////
+// Interface defining what I expect of a general sparse matrix, such as a Fermion action
+/////////////////////////////////////////////////////////////////////////////////////////////
+template<class Field> class SparseMatrixBase {
+public:
+  virtual GridBase *Grid(void) =0;
+  // Full checkerboar operations
+  virtual RealD M    (const Field &in, Field &out)=0;
+  virtual RealD Mdag (const Field &in, Field &out)=0;
+  virtual void  MdagM(const Field &in, Field &out,RealD &ni,RealD &no) {
+    Field tmp (in.Grid());
+    ni=M(in,tmp);
+    no=Mdag(tmp,out);
+  }
+  virtual  void Mdiag    (const Field &in, Field &out)=0;
+  virtual  void Mdir     (const Field &in, Field &out,int dir, int disp)=0;
+};
 
-  /////////////////////////////////////////////////////////////////////////////////////////////
-  // Interface augmented by a red black sparse matrix, such as a Fermion action
-  /////////////////////////////////////////////////////////////////////////////////////////////
-    template<class Field> class CheckerBoardedSparseMatrixBase : public SparseMatrixBase<Field> {
-    public:
-      virtual GridBase *RedBlackGrid(void)=0;
+/////////////////////////////////////////////////////////////////////////////////////////////
+// Interface augmented by a red black sparse matrix, such as a Fermion action
+/////////////////////////////////////////////////////////////////////////////////////////////
+template<class Field> class CheckerBoardedSparseMatrixBase : public SparseMatrixBase<Field> {
+public:
+  virtual GridBase *RedBlackGrid(void)=0;
 
       //////////////////////////////////////////////////////////////////////
       // Query the even even properties to make algorithmic decisions
@@ -63,17 +63,17 @@ namespace Grid {
       virtual int    ConstEE(void)     { return 1; }; // Disable assumptions unless overridden
       virtual int    isTrivialEE(void) { return 0; }; // by a derived class that knows better
 
-      // half checkerboard operaions
-      virtual  void Meooe    (const Field &in, Field &out)=0;
-      virtual  void Mooee    (const Field &in, Field &out)=0;
-      virtual  void MooeeInv (const Field &in, Field &out)=0;
+  // half checkerboard operaions
+  virtual  void Meooe    (const Field &in, Field &out)=0;
+  virtual  void Mooee    (const Field &in, Field &out)=0;
+  virtual  void MooeeInv (const Field &in, Field &out)=0;
 
-      virtual  void MeooeDag    (const Field &in, Field &out)=0;
-      virtual  void MooeeDag    (const Field &in, Field &out)=0;
-      virtual  void MooeeInvDag (const Field &in, Field &out)=0;
+  virtual  void MeooeDag    (const Field &in, Field &out)=0;
+  virtual  void MooeeDag    (const Field &in, Field &out)=0;
+  virtual  void MooeeInvDag (const Field &in, Field &out)=0;
 
-    };
+};
 
-}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/algorithms/approx/Chebyshev.h b/Grid/algorithms/approx/Chebyshev.h
index b34fac7f..97e0e807 100644
--- a/Grid/algorithms/approx/Chebyshev.h
+++ b/Grid/algorithms/approx/Chebyshev.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,14 +25,14 @@ Author: Christoph Lehner <clehner@bnl.gov>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_CHEBYSHEV_H
 #define GRID_CHEBYSHEV_H
 
 #include <Grid/algorithms/LinearOperator.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 struct ChebyParams : Serializable {
   GRID_SERIALIZABLE_CLASS_MEMBERS(ChebyParams,
@@ -41,337 +41,339 @@ struct ChebyParams : Serializable {
 				  int, Npoly);
 };
 
-  ////////////////////////////////////////////////////////////////////////////////////////////
-  // Generic Chebyshev approximations
-  ////////////////////////////////////////////////////////////////////////////////////////////
-  template<class Field>
-  class Chebyshev : public OperatorFunction<Field> {
-  private:
-    std::vector<RealD> Coeffs;
-    int order;
-    RealD hi;
-    RealD lo;
+////////////////////////////////////////////////////////////////////////////////////////////
+// Generic Chebyshev approximations
+////////////////////////////////////////////////////////////////////////////////////////////
+template<class Field>
+class Chebyshev : public OperatorFunction<Field> {
+private:
+  using OperatorFunction<Field>::operator();
 
-  public:
-    void csv(std::ostream &out){
-      RealD diff = hi-lo;
-      RealD delta = (hi-lo)*1.0e-9;
-      for (RealD x=lo; x<hi; x+=delta) {
-	delta*=1.1;
-	RealD f = approx(x);
-	out<< x<<" "<<f<<std::endl;
-      }
-      return;
+  std::vector<RealD> Coeffs;
+  int order;
+  RealD hi;
+  RealD lo;
+
+public:
+  void csv(std::ostream &out){
+    RealD diff = hi-lo;
+    RealD delta = diff*1.0e-9;
+    for (RealD x=lo; x<hi; x+=delta) {
+      delta*=1.1;
+      RealD f = approx(x);
+      out<< x<<" "<<f<<std::endl;
     }
+    return;
+  }
 
-    // Convenience for plotting the approximation
-    void   PlotApprox(std::ostream &out) {
-      out<<"Polynomial approx ["<<lo<<","<<hi<<"]"<<std::endl;
-      for(RealD x=lo;x<hi;x+=(hi-lo)/50.0){
-	out <<x<<"\t"<<approx(x)<<std::endl;
-      }
-    };
-
-    Chebyshev(){};
-    Chebyshev(ChebyParams p){ Init(p.alpha,p.beta,p.Npoly);};
-    Chebyshev(RealD _lo,RealD _hi,int _order, RealD (* func)(RealD) ) {Init(_lo,_hi,_order,func);};
-    Chebyshev(RealD _lo,RealD _hi,int _order) {Init(_lo,_hi,_order);};
-
-    ////////////////////////////////////////////////////////////////////////////////////////////////////
-    // c.f. numerical recipes "chebft"/"chebev". This is sec 5.8 "Chebyshev approximation".
-    ////////////////////////////////////////////////////////////////////////////////////////////////////
-// CJ: the one we need for Lanczos
-    void Init(RealD _lo,RealD _hi,int _order)
-    {
-      lo=_lo;
-      hi=_hi;
-      order=_order;
-      
-      if(order < 2) exit(-1);
-      Coeffs.resize(order);
-      Coeffs.assign(0.,order);
-      Coeffs[order-1] = 1.;
-    };
-
-    void Init(RealD _lo,RealD _hi,int _order, RealD (* func)(RealD))
-    {
-      lo=_lo;
-      hi=_hi;
-      order=_order;
-      
-      if(order < 2) exit(-1);
-      Coeffs.resize(order);
-      for(int j=0;j<order;j++){
-	RealD s=0;
-	for(int k=0;k<order;k++){
-	  RealD y=std::cos(M_PI*(k+0.5)/order);
-	  RealD x=0.5*(y*(hi-lo)+(hi+lo));
-	  RealD f=func(x);
-	  s=s+f*std::cos( j*M_PI*(k+0.5)/order );
-	}
-	Coeffs[j] = s * 2.0/order;
-      }
-    };
-
-    
-    void JacksonSmooth(void){
-      RealD M=order;
-      RealD alpha = M_PI/(M+2);
-      RealD lmax = std::cos(alpha);
-      RealD sumUsq =0;
-      std::vector<RealD> U(M);
-      std::vector<RealD> a(M);
-      std::vector<RealD> g(M);
-      for(int n=0;n<=M;n++){
-	U[n] = std::sin((n+1)*std::acos(lmax))/std::sin(std::acos(lmax));
-	sumUsq += U[n]*U[n];
-      }      
-      sumUsq = std::sqrt(sumUsq);
-
-      for(int i=1;i<=M;i++){
-	a[i] = U[i]/sumUsq;
-      }
-      g[0] = 1.0;
-      for(int m=1;m<=M;m++){
-	g[m] = 0;
-	for(int i=0;i<=M-m;i++){
-	  g[m]+= a[i]*a[m+i];
-	}
-      }
-      for(int m=1;m<=M;m++){
-	Coeffs[m]*=g[m];
-      }
-    }
-    RealD approx(RealD x) // Convenience for plotting the approximation
-    {
-      RealD Tn;
-      RealD Tnm;
-      RealD Tnp;
-      
-      RealD y=( x-0.5*(hi+lo))/(0.5*(hi-lo));
-      
-      RealD T0=1;
-      RealD T1=y;
-      
-      RealD sum;
-      sum = 0.5*Coeffs[0]*T0;
-      sum+= Coeffs[1]*T1;
-      
-      Tn =T1;
-      Tnm=T0;
-      for(int i=2;i<order;i++){
-	Tnp=2*y*Tn-Tnm;
-	Tnm=Tn;
-	Tn =Tnp;
-	sum+= Tn*Coeffs[i];
-      }
-      return sum;
-    };
-
-    RealD approxD(RealD x)
-    {
-      RealD Un;
-      RealD Unm;
-      RealD Unp;
-      
-      RealD y=( x-0.5*(hi+lo))/(0.5*(hi-lo));
-      
-      RealD U0=1;
-      RealD U1=2*y;
-      
-      RealD sum;
-      sum = Coeffs[1]*U0;
-      sum+= Coeffs[2]*U1*2.0;
-      
-      Un =U1;
-      Unm=U0;
-      for(int i=2;i<order-1;i++){
-	Unp=2*y*Un-Unm;
-	Unm=Un;
-	Un =Unp;
-	sum+= Un*Coeffs[i+1]*(i+1.0);
-      }
-      return sum/(0.5*(hi-lo));
-    };
-    
-    RealD approxInv(RealD z, RealD x0, int maxiter, RealD resid) {
-      RealD x = x0;
-      RealD eps;
-      
-      int i;
-      for (i=0;i<maxiter;i++) {
-	eps = approx(x) - z;
-	if (fabs(eps / z) < resid)
-	  return x;
-	x = x - eps / approxD(x);
-      }
-      
-      return std::numeric_limits<double>::quiet_NaN();
-    }
-    
-    // Implement the required interface
-    void operator() (LinearOperatorBase<Field> &Linop, const Field &in, Field &out) {
-
-      GridBase *grid=in._grid;
-
-      // std::cout << "Chevyshef(): in._grid="<<in._grid<<std::endl;
-      //std::cout <<" Linop.Grid()="<<Linop.Grid()<<"Linop.RedBlackGrid()="<<Linop.RedBlackGrid()<<std::endl;
-
-      int vol=grid->gSites();
-
-      Field T0(grid); T0 = in;  
-      Field T1(grid); 
-      Field T2(grid);
-      Field y(grid);
-      
-      Field *Tnm = &T0;
-      Field *Tn  = &T1;
-      Field *Tnp = &T2;
-
-      // Tn=T1 = (xscale M + mscale)in
-      RealD xscale = 2.0/(hi-lo);
-      RealD mscale = -(hi+lo)/(hi-lo);
-      Linop.HermOp(T0,y);
-      T1=y*xscale+in*mscale;
-
-      // sum = .5 c[0] T0 + c[1] T1
-      out = (0.5*Coeffs[0])*T0 + Coeffs[1]*T1;
-      for(int n=2;n<order;n++){
-	
-	Linop.HermOp(*Tn,y);
-
-	y=xscale*y+mscale*(*Tn);
-
-	*Tnp=2.0*y-(*Tnm);
-
-	out=out+Coeffs[n]* (*Tnp);
-
-	// Cycle pointers to avoid copies
-	Field *swizzle = Tnm;
-	Tnm    =Tn;
-	Tn     =Tnp;
-	Tnp    =swizzle;
-	  
-      }
+  // Convenience for plotting the approximation
+  void   PlotApprox(std::ostream &out) {
+    out<<"Polynomial approx ["<<lo<<","<<hi<<"]"<<std::endl;
+    for(RealD x=lo;x<hi;x+=(hi-lo)/50.0){
+      out <<x<<"\t"<<approx(x)<<std::endl;
     }
   };
 
+  Chebyshev(){};
+  Chebyshev(ChebyParams p){ Init(p.alpha,p.beta,p.Npoly);};
+  Chebyshev(RealD _lo,RealD _hi,int _order, RealD (* func)(RealD) ) {Init(_lo,_hi,_order,func);};
+  Chebyshev(RealD _lo,RealD _hi,int _order) {Init(_lo,_hi,_order);};
 
-  template<class Field>
-  class ChebyshevLanczos : public Chebyshev<Field> {
-  private:
-    std::vector<RealD> Coeffs;
-    int order;
-    RealD alpha;
-    RealD beta;
-    RealD mu;
+  ////////////////////////////////////////////////////////////////////////////////////////////////////
+  // c.f. numerical recipes "chebft"/"chebev". This is sec 5.8 "Chebyshev approximation".
+  ////////////////////////////////////////////////////////////////////////////////////////////////////
+  // CJ: the one we need for Lanczos
+  void Init(RealD _lo,RealD _hi,int _order)
+  {
+    lo=_lo;
+    hi=_hi;
+    order=_order;
+      
+    if(order < 2) exit(-1);
+    Coeffs.resize(order);
+    Coeffs.assign(0.,order);
+    Coeffs[order-1] = 1.;
+  };
 
-  public:
-    ChebyshevLanczos(RealD _alpha,RealD _beta,RealD _mu,int _order) :
+  void Init(RealD _lo,RealD _hi,int _order, RealD (* func)(RealD))
+  {
+    lo=_lo;
+    hi=_hi;
+    order=_order;
+      
+    if(order < 2) exit(-1);
+    Coeffs.resize(order);
+    for(int j=0;j<order;j++){
+      RealD s=0;
+      for(int k=0;k<order;k++){
+	RealD y=std::cos(M_PI*(k+0.5)/order);
+	RealD x=0.5*(y*(hi-lo)+(hi+lo));
+	RealD f=func(x);
+	s=s+f*std::cos( j*M_PI*(k+0.5)/order );
+      }
+      Coeffs[j] = s * 2.0/order;
+    }
+  };
+
+    
+  void JacksonSmooth(void){
+    RealD M=order;
+    RealD alpha = M_PI/(M+2);
+    RealD lmax = std::cos(alpha);
+    RealD sumUsq =0;
+    std::vector<RealD> U(M);
+    std::vector<RealD> a(M);
+    std::vector<RealD> g(M);
+    for(int n=0;n<=M;n++){
+      U[n] = std::sin((n+1)*std::acos(lmax))/std::sin(std::acos(lmax));
+      sumUsq += U[n]*U[n];
+    }      
+    sumUsq = std::sqrt(sumUsq);
+
+    for(int i=1;i<=M;i++){
+      a[i] = U[i]/sumUsq;
+    }
+    g[0] = 1.0;
+    for(int m=1;m<=M;m++){
+      g[m] = 0;
+      for(int i=0;i<=M-m;i++){
+	g[m]+= a[i]*a[m+i];
+      }
+    }
+    for(int m=1;m<=M;m++){
+      Coeffs[m]*=g[m];
+    }
+  }
+  RealD approx(RealD x) // Convenience for plotting the approximation
+  {
+    RealD Tn;
+    RealD Tnm;
+    RealD Tnp;
+      
+    RealD y=( x-0.5*(hi+lo))/(0.5*(hi-lo));
+      
+    RealD T0=1;
+    RealD T1=y;
+      
+    RealD sum;
+    sum = 0.5*Coeffs[0]*T0;
+    sum+= Coeffs[1]*T1;
+      
+    Tn =T1;
+    Tnm=T0;
+    for(int i=2;i<order;i++){
+      Tnp=2*y*Tn-Tnm;
+      Tnm=Tn;
+      Tn =Tnp;
+      sum+= Tn*Coeffs[i];
+    }
+    return sum;
+  };
+
+  RealD approxD(RealD x)
+  {
+    RealD Un;
+    RealD Unm;
+    RealD Unp;
+      
+    RealD y=( x-0.5*(hi+lo))/(0.5*(hi-lo));
+      
+    RealD U0=1;
+    RealD U1=2*y;
+      
+    RealD sum;
+    sum = Coeffs[1]*U0;
+    sum+= Coeffs[2]*U1*2.0;
+      
+    Un =U1;
+    Unm=U0;
+    for(int i=2;i<order-1;i++){
+      Unp=2*y*Un-Unm;
+      Unm=Un;
+      Un =Unp;
+      sum+= Un*Coeffs[i+1]*(i+1.0);
+    }
+    return sum/(0.5*(hi-lo));
+  };
+    
+  RealD approxInv(RealD z, RealD x0, int maxiter, RealD resid) {
+    RealD x = x0;
+    RealD eps;
+      
+    int i;
+    for (i=0;i<maxiter;i++) {
+      eps = approx(x) - z;
+      if (fabs(eps / z) < resid)
+	return x;
+      x = x - eps / approxD(x);
+    }
+      
+    return std::numeric_limits<double>::quiet_NaN();
+  }
+    
+  // Implement the required interface
+  void operator() (LinearOperatorBase<Field> &Linop, const Field &in, Field &out) {
+
+    GridBase *grid=in.Grid();
+
+    // std::cout << "Chevyshef(): in.Grid()="<<in.Grid()<<std::endl;
+    //std::cout <<" Linop.Grid()="<<Linop.Grid()<<"Linop.RedBlackGrid()="<<Linop.RedBlackGrid()<<std::endl;
+
+    int vol=grid->gSites();
+
+    Field T0(grid); T0 = in;  
+    Field T1(grid); 
+    Field T2(grid);
+    Field y(grid);
+      
+    Field *Tnm = &T0;
+    Field *Tn  = &T1;
+    Field *Tnp = &T2;
+
+    // Tn=T1 = (xscale M + mscale)in
+    RealD xscale = 2.0/(hi-lo);
+    RealD mscale = -(hi+lo)/(hi-lo);
+    Linop.HermOp(T0,y);
+    T1=y*xscale+in*mscale;
+
+    // sum = .5 c[0] T0 + c[1] T1
+    out = (0.5*Coeffs[0])*T0 + Coeffs[1]*T1;
+    for(int n=2;n<order;n++){
+	
+      Linop.HermOp(*Tn,y);
+
+      y=xscale*y+mscale*(*Tn);
+
+      *Tnp=2.0*y-(*Tnm);
+
+      out=out+Coeffs[n]* (*Tnp);
+
+      // Cycle pointers to avoid copies
+      Field *swizzle = Tnm;
+      Tnm    =Tn;
+      Tn     =Tnp;
+      Tnp    =swizzle;
+	  
+    }
+  }
+};
+
+
+template<class Field>
+class ChebyshevLanczos : public Chebyshev<Field> {
+private:
+  std::vector<RealD> Coeffs;
+  int order;
+  RealD alpha;
+  RealD beta;
+  RealD mu;
+
+public:
+  ChebyshevLanczos(RealD _alpha,RealD _beta,RealD _mu,int _order) :
     alpha(_alpha),
-      beta(_beta),
-          mu(_mu)
-    {
-      order=_order;
-      Coeffs.resize(order);
-      for(int i=0;i<_order;i++){
-	Coeffs[i] = 0.0;
-      }
-      Coeffs[order-1]=1.0;
-    };
-
-    void csv(std::ostream &out){
-      for (RealD x=-1.2*alpha; x<1.2*alpha; x+=(2.0*alpha)/10000) {
-	RealD f = approx(x);
-	out<< x<<" "<<f<<std::endl;
-      }
-      return;
-    }
-
-    RealD approx(RealD xx) // Convenience for plotting the approximation
-    {
-      RealD Tn;
-      RealD Tnm;
-      RealD Tnp;
-      Real aa = alpha * alpha;
-      Real bb = beta  *  beta;
-      
-      RealD x = ( 2.0 * (xx-mu)*(xx-mu) - (aa+bb) ) / (aa-bb);
-
-      RealD y= x;
-      
-      RealD T0=1;
-      RealD T1=y;
-      
-      RealD sum;
-      sum = 0.5*Coeffs[0]*T0;
-      sum+= Coeffs[1]*T1;
-      
-      Tn =T1;
-      Tnm=T0;
-      for(int i=2;i<order;i++){
-	Tnp=2*y*Tn-Tnm;
-	Tnm=Tn;
-	Tn =Tnp;
-	sum+= Tn*Coeffs[i];
-      }
-      return sum;
-    };
-
-    // shift_Multiply in Rudy's code
-    void AminusMuSq(LinearOperatorBase<Field> &Linop, const Field &in, Field &out) 
-    {
-      GridBase *grid=in._grid;
-      Field tmp(grid);
-
-      RealD aa= alpha*alpha;
-      RealD bb= beta * beta;
-
-      Linop.HermOp(in,out);
-      out = out - mu*in;
-
-      Linop.HermOp(out,tmp);
-      tmp = tmp - mu * out;
-
-      out = (2.0/ (aa-bb) ) * tmp -  ((aa+bb)/(aa-bb))*in;
-    };
-    // Implement the required interface
-    void operator() (LinearOperatorBase<Field> &Linop, const Field &in, Field &out) {
-
-      GridBase *grid=in._grid;
-
-      int vol=grid->gSites();
-
-      Field T0(grid); T0 = in;  
-      Field T1(grid); 
-      Field T2(grid);
-      Field  y(grid);
-      
-      Field *Tnm = &T0;
-      Field *Tn  = &T1;
-      Field *Tnp = &T2;
-
-      // Tn=T1 = (xscale M )*in
-      AminusMuSq(Linop,T0,T1);
-
-      // sum = .5 c[0] T0 + c[1] T1
-      out = (0.5*Coeffs[0])*T0 + Coeffs[1]*T1;
-      for(int n=2;n<order;n++){
-	
-	AminusMuSq(Linop,*Tn,y);
-
-	*Tnp=2.0*y-(*Tnm);
-
-	out=out+Coeffs[n]* (*Tnp);
-
-	// Cycle pointers to avoid copies
-	Field *swizzle = Tnm;
-	Tnm    =Tn;
-	Tn     =Tnp;
-	Tnp    =swizzle;
-	  
-      }
+    beta(_beta),
+    mu(_mu)
+  {
+    order=_order;
+    Coeffs.resize(order);
+    for(int i=0;i<_order;i++){
+      Coeffs[i] = 0.0;
     }
+    Coeffs[order-1]=1.0;
   };
-}
+
+  void csv(std::ostream &out){
+    for (RealD x=-1.2*alpha; x<1.2*alpha; x+=(2.0*alpha)/10000) {
+      RealD f = approx(x);
+      out<< x<<" "<<f<<std::endl;
+    }
+    return;
+  }
+
+  RealD approx(RealD xx) // Convenience for plotting the approximation
+  {
+    RealD Tn;
+    RealD Tnm;
+    RealD Tnp;
+    Real aa = alpha * alpha;
+    Real bb = beta  *  beta;
+      
+    RealD x = ( 2.0 * (xx-mu)*(xx-mu) - (aa+bb) ) / (aa-bb);
+
+    RealD y= x;
+      
+    RealD T0=1;
+    RealD T1=y;
+      
+    RealD sum;
+    sum = 0.5*Coeffs[0]*T0;
+    sum+= Coeffs[1]*T1;
+      
+    Tn =T1;
+    Tnm=T0;
+    for(int i=2;i<order;i++){
+      Tnp=2*y*Tn-Tnm;
+      Tnm=Tn;
+      Tn =Tnp;
+      sum+= Tn*Coeffs[i];
+    }
+    return sum;
+  };
+
+  // shift_Multiply in Rudy's code
+  void AminusMuSq(LinearOperatorBase<Field> &Linop, const Field &in, Field &out) 
+  {
+    GridBase *grid=in.Grid();
+    Field tmp(grid);
+
+    RealD aa= alpha*alpha;
+    RealD bb= beta * beta;
+
+    Linop.HermOp(in,out);
+    out = out - mu*in;
+
+    Linop.HermOp(out,tmp);
+    tmp = tmp - mu * out;
+
+    out = (2.0/ (aa-bb) ) * tmp -  ((aa+bb)/(aa-bb))*in;
+  };
+  // Implement the required interface
+  void operator() (LinearOperatorBase<Field> &Linop, const Field &in, Field &out) {
+
+    GridBase *grid=in.Grid();
+
+    int vol=grid->gSites();
+
+    Field T0(grid); T0 = in;  
+    Field T1(grid); 
+    Field T2(grid);
+    Field  y(grid);
+      
+    Field *Tnm = &T0;
+    Field *Tn  = &T1;
+    Field *Tnp = &T2;
+
+    // Tn=T1 = (xscale M )*in
+    AminusMuSq(Linop,T0,T1);
+
+    // sum = .5 c[0] T0 + c[1] T1
+    out = (0.5*Coeffs[0])*T0 + Coeffs[1]*T1;
+    for(int n=2;n<order;n++){
+	
+      AminusMuSq(Linop,*Tn,y);
+
+      *Tnp=2.0*y-(*Tnm);
+
+      out=out+Coeffs[n]* (*Tnp);
+
+      // Cycle pointers to avoid copies
+      Field *swizzle = Tnm;
+      Tnm    =Tn;
+      Tn     =Tnp;
+      Tnp    =swizzle;
+	  
+    }
+  }
+};
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/algorithms/approx/Forecast.h b/Grid/algorithms/approx/Forecast.h
index 87eb84a6..67596272 100644
--- a/Grid/algorithms/approx/Forecast.h
+++ b/Grid/algorithms/approx/Forecast.h
@@ -26,127 +26,127 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 
 See the full license in the file "LICENSE" in the top level distribution directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 
 #ifndef INCLUDED_FORECAST_H
 #define INCLUDED_FORECAST_H
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  // Abstract base class.
-  // Takes a matrix (Mat), a source (phi), and a vector of Fields (chi)
-  // and returns a forecasted solution to the system D*psi = phi (psi).
-  template<class Matrix, class Field>
-  class Forecast
+// Abstract base class.
+// Takes a matrix (Mat), a source (phi), and a vector of Fields (chi)
+// and returns a forecasted solution to the system D*psi = phi (psi).
+template<class Matrix, class Field>
+class Forecast
+{
+public:
+  virtual Field operator()(Matrix &Mat, const Field& phi, const std::vector<Field>& chi) = 0;
+};
+
+// Implementation of Brower et al.'s chronological inverter (arXiv:hep-lat/9509012),
+// used to forecast solutions across poles of the EOFA heatbath.
+//
+// Modified from CPS (cps_pp/src/util/dirac_op/d_op_base/comsrc/minresext.C)
+template<class Matrix, class Field>
+class ChronoForecast : public Forecast<Matrix,Field>
+{
+public:
+  Field operator()(Matrix &Mat, const Field& phi, const std::vector<Field>& prev_solns)
   {
-    public:
-      virtual Field operator()(Matrix &Mat, const Field& phi, const std::vector<Field>& chi) = 0;
+    int degree = prev_solns.size();
+    Field chi(phi); // forecasted solution
+
+    // Trivial cases
+    if(degree == 0){ chi = Zero(); return chi; }
+    else if(degree == 1){ return prev_solns[0]; }
+
+    //    RealD dot;
+    ComplexD xp;
+    Field r(phi); // residual
+    Field Mv(phi);
+    std::vector<Field> v(prev_solns); // orthonormalized previous solutions
+    std::vector<Field> MdagMv(degree,phi);
+
+    // Array to hold the matrix elements
+    std::vector<std::vector<ComplexD>> G(degree, std::vector<ComplexD>(degree));
+
+    // Solution and source vectors
+    std::vector<ComplexD> a(degree);
+    std::vector<ComplexD> b(degree);
+
+    // Orthonormalize the vector basis
+    for(int i=0; i<degree; i++){
+      v[i] *= 1.0/std::sqrt(norm2(v[i]));
+      for(int j=i+1; j<degree; j++){ v[j] -= innerProduct(v[i],v[j]) * v[i]; }
+    }
+
+    // Perform sparse matrix multiplication and construct rhs
+    for(int i=0; i<degree; i++){
+      b[i] = innerProduct(v[i],phi);
+      Mat.M(v[i],Mv);
+      Mat.Mdag(Mv,MdagMv[i]);
+      G[i][i] = innerProduct(v[i],MdagMv[i]);
+    }
+
+    // Construct the matrix
+    for(int j=0; j<degree; j++){
+      for(int k=j+1; k<degree; k++){
+	G[j][k] = innerProduct(v[j],MdagMv[k]);
+	G[k][j] = conjugate(G[j][k]);
+      }}
+
+    // Gauss-Jordan elimination with partial pivoting
+    for(int i=0; i<degree; i++){
+
+      // Perform partial pivoting
+      int k = i;
+      for(int j=i+1; j<degree; j++){ if(abs(G[j][j]) > abs(G[k][k])){ k = j; } }
+      if(k != i){
+	xp = b[k];
+	b[k] = b[i];
+	b[i] = xp;
+	for(int j=0; j<degree; j++){
+	  xp = G[k][j];
+	  G[k][j] = G[i][j];
+	  G[i][j] = xp;
+	}
+      }
+
+      // Convert matrix to upper triangular form
+      for(int j=i+1; j<degree; j++){
+	xp = G[j][i]/G[i][i];
+	b[j] -= xp * b[i];
+	for(int k=0; k<degree; k++){ G[j][k] -= xp*G[i][k]; }
+      }
+    }
+
+    // Use Gaussian elimination to solve equations and calculate initial guess
+    chi = Zero();
+    r = phi;
+    for(int i=degree-1; i>=0; i--){
+      a[i] = 0.0;
+      for(int j=i+1; j<degree; j++){ a[i] += G[i][j] * a[j]; }
+      a[i] = (b[i]-a[i])/G[i][i];
+      chi += a[i]*v[i];
+      r -= a[i]*MdagMv[i];
+    }
+
+    RealD true_r(0.0);
+    ComplexD tmp;
+    for(int i=0; i<degree; i++){
+      tmp = -b[i];
+      for(int j=0; j<degree; j++){ tmp += G[i][j]*a[j]; }
+      tmp = conjugate(tmp)*tmp;
+      true_r += std::sqrt(tmp.real());
+    }
+
+    RealD error = std::sqrt(norm2(r)/norm2(phi));
+    std::cout << GridLogMessage << "ChronoForecast: |res|/|src| = " << error << std::endl;
+
+    return chi;
   };
+};
 
-  // Implementation of Brower et al.'s chronological inverter (arXiv:hep-lat/9509012),
-  // used to forecast solutions across poles of the EOFA heatbath.
-  //
-  // Modified from CPS (cps_pp/src/util/dirac_op/d_op_base/comsrc/minresext.C)
-  template<class Matrix, class Field>
-  class ChronoForecast : public Forecast<Matrix,Field>
-  {
-    public:
-      Field operator()(Matrix &Mat, const Field& phi, const std::vector<Field>& prev_solns)
-      {
-        int degree = prev_solns.size();
-        Field chi(phi); // forecasted solution
-
-        // Trivial cases
-        if(degree == 0){ chi = zero; return chi; }
-        else if(degree == 1){ return prev_solns[0]; }
-
-        RealD dot;
-        ComplexD xp;
-        Field r(phi); // residual
-        Field Mv(phi);
-        std::vector<Field> v(prev_solns); // orthonormalized previous solutions
-        std::vector<Field> MdagMv(degree,phi);
-
-        // Array to hold the matrix elements
-        std::vector<std::vector<ComplexD>> G(degree, std::vector<ComplexD>(degree));
-
-        // Solution and source vectors
-        std::vector<ComplexD> a(degree);
-        std::vector<ComplexD> b(degree);
-
-        // Orthonormalize the vector basis
-        for(int i=0; i<degree; i++){
-          v[i] *= 1.0/std::sqrt(norm2(v[i]));
-          for(int j=i+1; j<degree; j++){ v[j] -= innerProduct(v[i],v[j]) * v[i]; }
-        }
-
-        // Perform sparse matrix multiplication and construct rhs
-        for(int i=0; i<degree; i++){
-          b[i] = innerProduct(v[i],phi);
-          Mat.M(v[i],Mv);
-          Mat.Mdag(Mv,MdagMv[i]);
-          G[i][i] = innerProduct(v[i],MdagMv[i]);
-        }
-
-        // Construct the matrix
-        for(int j=0; j<degree; j++){
-        for(int k=j+1; k<degree; k++){
-          G[j][k] = innerProduct(v[j],MdagMv[k]);
-          G[k][j] = std::conj(G[j][k]);
-        }}
-
-        // Gauss-Jordan elimination with partial pivoting
-        for(int i=0; i<degree; i++){
-
-          // Perform partial pivoting
-          int k = i;
-          for(int j=i+1; j<degree; j++){ if(std::abs(G[j][j]) > std::abs(G[k][k])){ k = j; } }
-          if(k != i){
-            xp = b[k];
-            b[k] = b[i];
-            b[i] = xp;
-            for(int j=0; j<degree; j++){
-              xp = G[k][j];
-              G[k][j] = G[i][j];
-              G[i][j] = xp;
-            }
-          }
-
-          // Convert matrix to upper triangular form
-          for(int j=i+1; j<degree; j++){
-            xp = G[j][i]/G[i][i];
-            b[j] -= xp * b[i];
-            for(int k=0; k<degree; k++){ G[j][k] -= xp*G[i][k]; }
-          }
-        }
-
-        // Use Gaussian elimination to solve equations and calculate initial guess
-        chi = zero;
-        r = phi;
-        for(int i=degree-1; i>=0; i--){
-          a[i] = 0.0;
-          for(int j=i+1; j<degree; j++){ a[i] += G[i][j] * a[j]; }
-          a[i] = (b[i]-a[i])/G[i][i];
-          chi += a[i]*v[i];
-          r -= a[i]*MdagMv[i];
-        }
-
-        RealD true_r(0.0);
-        ComplexD tmp;
-        for(int i=0; i<degree; i++){
-          tmp = -b[i];
-          for(int j=0; j<degree; j++){ tmp += G[i][j]*a[j]; }
-          tmp = std::conj(tmp)*tmp;
-          true_r += std::sqrt(tmp.real());
-        }
-
-        RealD error = std::sqrt(norm2(r)/norm2(phi));
-        std::cout << GridLogMessage << "ChronoForecast: |res|/|src| = " << error << std::endl;
-
-        return chi;
-      };
-  };
-
-}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/algorithms/approx/MultiShiftFunction.cc b/Grid/algorithms/approx/MultiShiftFunction.cc
index 34befdde..310347f8 100644
--- a/Grid/algorithms/approx/MultiShiftFunction.cc
+++ b/Grid/algorithms/approx/MultiShiftFunction.cc
@@ -27,7 +27,8 @@ Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
     /*  END LEGAL */
 #include <Grid/GridCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
+
 double MultiShiftFunction::approx(double x)
 {
   double a = norm;
@@ -53,4 +54,4 @@ void MultiShiftFunction::csv(std::ostream &out)
   }
   return;
 }
-}
+NAMESPACE_END(Grid);
diff --git a/Grid/algorithms/approx/MultiShiftFunction.h b/Grid/algorithms/approx/MultiShiftFunction.h
index 56190573..58c75f6c 100644
--- a/Grid/algorithms/approx/MultiShiftFunction.h
+++ b/Grid/algorithms/approx/MultiShiftFunction.h
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #ifndef MULTI_SHIFT_FUNCTION
 #define MULTI_SHIFT_FUNCTION
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 class MultiShiftFunction {
 public:
@@ -63,5 +63,5 @@ public:
   }
 
 };
-}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/algorithms/approx/Remez.cc b/Grid/algorithms/approx/Remez.cc
index ca00a330..6fbaadc9 100644
--- a/Grid/algorithms/approx/Remez.cc
+++ b/Grid/algorithms/approx/Remez.cc
@@ -298,7 +298,7 @@ void AlgRemez::stpini(bigfloat *step) {
 // Search for error maxima and minima
 void AlgRemez::search(bigfloat *step) {
   bigfloat a, q, xm, ym, xn, yn, xx0, xx1;
-  int i, j, meq, emsign, ensign, steps;
+  int i, meq, emsign, ensign, steps;
 
   meq = neq + 1;
   bigfloat *yy = new bigfloat[meq];
@@ -306,7 +306,6 @@ void AlgRemez::search(bigfloat *step) {
   bigfloat eclose = 1.0e30;
   bigfloat farther = 0l;
 
-  j = 1;
   xx0 = apstrt;
 
   for (i = 0; i < meq; i++) {
diff --git a/Grid/algorithms/approx/Zolotarev.cc b/Grid/algorithms/approx/Zolotarev.cc
index 1aba7aa8..c2efd41c 100644
--- a/Grid/algorithms/approx/Zolotarev.cc
+++ b/Grid/algorithms/approx/Zolotarev.cc
@@ -58,8 +58,8 @@
 
 /* Compute the partial fraction expansion coefficients (alpha) from the
  * factored form */
-namespace Grid {
-namespace Approx {
+NAMESPACE_BEGIN(Grid);
+NAMESPACE_BEGIN(Approx);
 
 static void construct_partfrac(izd *z) {
   int dn = z -> dn, dd = z -> dd, type = z -> type;
@@ -516,7 +516,9 @@ zolotarev_data* higham(PRECISION epsilon, int n) {
   free(d);
   return zd;
 }
-}}
+
+NAMESPACE_END(Approx);
+NAMESPACE_END(Grid);
 
 #ifdef TEST
 
@@ -585,6 +587,7 @@ static PRECISION zolotarev_cayley_eval(PRECISION x, zolotarev_data* rdata) {
   return (ONE - T) / (ONE + T);
 }
 
+
 /* Test program. Apart from printing out the parameters for R(x) it produces
  * the following data files for plotting (unless NPLOT is defined):
  *
@@ -723,5 +726,5 @@ int main(int argc, char** argv) {
   return EXIT_SUCCESS;
 }
 
-
 #endif /* TEST */
+
diff --git a/Grid/algorithms/approx/Zolotarev.h b/Grid/algorithms/approx/Zolotarev.h
index f6ee56b1..800cf3c7 100644
--- a/Grid/algorithms/approx/Zolotarev.h
+++ b/Grid/algorithms/approx/Zolotarev.h
@@ -1,13 +1,13 @@
 /* -*- Mode: C; comment-column: 22; fill-column: 79; -*- */
 
 #ifdef __cplusplus
-namespace Grid {
-namespace Approx {
+#include <Grid/Namespace.h>
+NAMESPACE_BEGIN(Grid);
+NAMESPACE_BEGIN(Approx);
 #endif
 
 #define HVERSION Header Time-stamp: <14-OCT-2004 09:26:51.00 adk@MISSCONTRARY>
 
-
 #ifndef ZOLOTAREV_INTERNAL
 #ifndef PRECISION
 #define PRECISION double
@@ -83,5 +83,6 @@ void zolotarev_free(zolotarev_data *zdata);
 #endif
 
 #ifdef __cplusplus
-}}
+NAMESPACE_END(Approx);
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/algorithms/approx/bigfloat.h b/Grid/algorithms/approx/bigfloat.h
index a749fa81..f1e60563 100644
--- a/Grid/algorithms/approx/bigfloat.h
+++ b/Grid/algorithms/approx/bigfloat.h
@@ -10,10 +10,12 @@
 #ifndef INCLUDED_BIGFLOAT_H
 #define INCLUDED_BIGFLOAT_H
 
-
+#define __GMP_WITHIN_CONFIGURE
 #include <gmp.h>
 #include <mpf2mpfr.h>
 #include <mpfr.h>
+#undef  __GMP_WITHIN_CONFIGURE
+
 class bigfloat {
 
 private:
diff --git a/Grid/algorithms/iterative/AdefGeneric.h b/Grid/algorithms/iterative/AdefGeneric.h
index 2eff5871..ad053615 100644
--- a/Grid/algorithms/iterative/AdefGeneric.h
+++ b/Grid/algorithms/iterative/AdefGeneric.h
@@ -90,8 +90,8 @@ class TwoLevelFlexiblePcg : public LinearFunction<Field>
   void operator() (const Field &src, Field &psi){
   void operator() (const Field &src, Field &psi){
 
-    psi.checkerboard = src.checkerboard;
-    grid             = src._grid;
+    psi.Checkerboard() = src.Checkerboard();
+    grid             = src.Grid();
 
     RealD f;
     RealD rtzp,rtz,a,d,b;
diff --git a/Grid/algorithms/iterative/BlockConjugateGradient.h b/Grid/algorithms/iterative/BlockConjugateGradient.h
index 972aa608..0bd7cf51 100644
--- a/Grid/algorithms/iterative/BlockConjugateGradient.h
+++ b/Grid/algorithms/iterative/BlockConjugateGradient.h
@@ -27,11 +27,9 @@ See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
 /*  END LEGAL */
-#ifndef GRID_BLOCK_CONJUGATE_GRADIENT_H
-#define GRID_BLOCK_CONJUGATE_GRADIENT_H
+#pragma once
 
-
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 enum BlockCGtype { BlockCG, BlockCGrQ, CGmultiRHS, BlockCGVec, BlockCGrQVec };
 
@@ -154,12 +152,12 @@ virtual void operator()(LinearOperatorBase<Field> &Linop, const std::vector<Fiel
 void BlockCGrQsolve(LinearOperatorBase<Field> &Linop, const Field &B, Field &X) 
 {
   int Orthog = blockDim; // First dimension is block dim; this is an assumption
-  Nblock = B._grid->_fdimensions[Orthog];
+  Nblock = B.Grid()->_fdimensions[Orthog];
 /* FAKE */
   Nblock=8;
   std::cout<<GridLogMessage<<" Block Conjugate Gradient : Orthog "<<Orthog<<" Nblock "<<Nblock<<std::endl;
 
-  X.checkerboard = B.checkerboard;
+  X.Checkerboard() = B.Checkerboard();
   conformable(X, B);
 
   Field tmp(B);
@@ -334,11 +332,11 @@ void BlockCGrQsolve(LinearOperatorBase<Field> &Linop, const Field &B, Field &X)
 void CGmultiRHSsolve(LinearOperatorBase<Field> &Linop, const Field &Src, Field &Psi) 
 {
   int Orthog = blockDim; // First dimension is block dim
-  Nblock = Src._grid->_fdimensions[Orthog];
+  Nblock = Src.Grid()->_fdimensions[Orthog];
 
   std::cout<<GridLogMessage<<"MultiRHS Conjugate Gradient : Orthog "<<Orthog<<" Nblock "<<Nblock<<std::endl;
 
-  Psi.checkerboard = Src.checkerboard;
+  Psi.Checkerboard() = Src.Checkerboard();
   conformable(Psi, Src);
 
   Field P(Src);
@@ -478,7 +476,7 @@ void MaddMatrix(std::vector<Field> &AP, Eigen::MatrixXcd &m , const std::vector<
   for(int b=0;b<Nblock;b++){
     tmp[b]   = Y[b];
     for(int bp=0;bp<Nblock;bp++) {
-      tmp[b] = tmp[b] + (scale*m(bp,b))*X[bp]; 
+      tmp[b] = tmp[b] + scomplex(scale*m(bp,b))*X[bp]; 
     }
   }
   for(int b=0;b<Nblock;b++){
@@ -488,9 +486,9 @@ void MaddMatrix(std::vector<Field> &AP, Eigen::MatrixXcd &m , const std::vector<
 void MulMatrix(std::vector<Field> &AP, Eigen::MatrixXcd &m , const std::vector<Field> &X){
   // Should make this cache friendly with site outermost, parallel_for
   for(int b=0;b<Nblock;b++){
-    AP[b] = zero;
+    AP[b] = Zero();
     for(int bp=0;bp<Nblock;bp++) {
-      AP[b] += (m(bp,b))*X[bp]; 
+      AP[b] += scomplex(m(bp,b))*X[bp]; 
     }
   }
 }
@@ -517,7 +515,7 @@ void BlockCGrQsolveVec(LinearOperatorBase<Field> &Linop, const std::vector<Field
   std::cout<<GridLogMessage<<" Block Conjugate Gradient Vec rQ : Nblock "<<Nblock<<std::endl;
 
   for(int b=0;b<Nblock;b++){ 
-    X[b].checkerboard = B[b].checkerboard;
+    X[b].Checkerboard() = B[b].Checkerboard();
     conformable(X[b], B[b]);
     conformable(X[b], X[0]); 
   }
@@ -690,9 +688,7 @@ void BlockCGrQsolveVec(LinearOperatorBase<Field> &Linop, const std::vector<Field
   IterationsToComplete = k;
 }
 
-
-
 };
 
-}
-#endif
+NAMESPACE_END(Grid);
+
diff --git a/Grid/algorithms/iterative/CommunicationAvoidingGeneralisedMinimalResidual.h b/Grid/algorithms/iterative/CommunicationAvoidingGeneralisedMinimalResidual.h
index f0289683..6e206d59 100644
--- a/Grid/algorithms/iterative/CommunicationAvoidingGeneralisedMinimalResidual.h
+++ b/Grid/algorithms/iterative/CommunicationAvoidingGeneralisedMinimalResidual.h
@@ -34,6 +34,8 @@ namespace Grid {
 template<class Field>
 class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<Field> {
  public:
+  using OperatorFunction<Field>::operator();
+
   bool ErrorOnNoConverge; // Throw an assert when CAGMRES fails to converge,
                           // defaults to true
 
@@ -52,10 +54,10 @@ class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<
 
   Eigen::MatrixXcd H;
 
-  std::vector<std::complex<double>> y;
-  std::vector<std::complex<double>> gamma;
-  std::vector<std::complex<double>> c;
-  std::vector<std::complex<double>> s;
+  std::vector<ComplexD> y;
+  std::vector<ComplexD> gamma;
+  std::vector<ComplexD> c;
+  std::vector<ComplexD> s;
 
   CommunicationAvoidingGeneralisedMinimalResidual(RealD   tol,
                                                   Integer maxit,
@@ -76,7 +78,7 @@ class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<
 
     std::cout << GridLogWarning << "This algorithm currently doesn't differ from regular GMRES" << std::endl;
 
-    psi.checkerboard = src.checkerboard;
+    psi.Checkerboard() = src.Checkerboard();
     conformable(psi, src);
 
     RealD guess = norm2(psi);
@@ -86,7 +88,7 @@ class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<
     RealD ssq = norm2(src);
     RealD rsq = Tolerance * Tolerance * ssq;
 
-    Field r(src._grid);
+    Field r(src.Grid());
 
     std::cout << std::setprecision(4) << std::scientific;
     std::cout << GridLogIterative << "CommunicationAvoidingGeneralisedMinimalResidual: guess " << guess << std::endl;
@@ -142,11 +144,11 @@ class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<
 
     RealD cp = 0;
 
-    Field w(src._grid);
-    Field r(src._grid);
+    Field w(src.Grid());
+    Field r(src.Grid());
 
     // this should probably be made a class member so that it is only allocated once, not in every restart
-    std::vector<Field> v(RestartLength + 1, src._grid); for (auto &elem : v) elem = zero;
+    std::vector<Field> v(RestartLength + 1, src.Grid()); for (auto &elem : v) elem = Zero();
 
     MatrixTimer.Start();
     LinOp.Op(psi, w);
@@ -157,7 +159,9 @@ class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<
 
     gamma[0] = sqrt(norm2(r));
 
-    v[0] = (1. / gamma[0]) * r;
+    ComplexD scale = 1.0/gamma[0];
+    v[0] = scale * r;
+
     LinalgTimer.Stop();
 
     for (int i=0; i<RestartLength; i++) {
@@ -168,7 +172,7 @@ class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<
 
       qrUpdate(i);
 
-      cp = std::norm(gamma[i+1]);
+      cp = norm(gamma[i+1]);
 
       std::cout << GridLogIterative << "CommunicationAvoidingGeneralisedMinimalResidual: Iteration " << IterationCount
                 << " residual " << cp << " target " << rsq << std::endl;
@@ -194,11 +198,11 @@ class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<
     LinalgTimer.Start();
     for (int i = 0; i <= iter; ++i) {
       H(iter, i) = innerProduct(v[i], w);
-      w = w - H(iter, i) * v[i];
+      w = w - ComplexD(H(iter, i)) * v[i];
     }
 
     H(iter, iter + 1) = sqrt(norm2(w));
-    v[iter + 1] = (1. / H(iter, iter + 1)) * w;
+    v[iter + 1] = ComplexD(1. / H(iter, iter + 1)) * w;
     LinalgTimer.Stop();
   }
 
@@ -206,13 +210,13 @@ class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<
 
     QrTimer.Start();
     for (int i = 0; i < iter ; ++i) {
-      auto tmp       = -s[i] * H(iter, i) + c[i] * H(iter, i + 1);
-      H(iter, i)     = std::conj(c[i]) * H(iter, i) + std::conj(s[i]) * H(iter, i + 1);
+      auto tmp       = -s[i] * ComplexD(H(iter, i)) + c[i] * ComplexD(H(iter, i + 1));
+      H(iter, i)     = conjugate(c[i]) * ComplexD(H(iter, i)) + conjugate(s[i]) * ComplexD(H(iter, i + 1));
       H(iter, i + 1) = tmp;
     }
 
     // Compute new Givens Rotation
-    ComplexD nu = sqrt(std::norm(H(iter, iter)) + std::norm(H(iter, iter + 1)));
+    auto nu     = sqrt(std::norm(H(iter, iter)) + std::norm(H(iter, iter + 1)));
     c[iter]     = H(iter, iter) / nu;
     s[iter]     = H(iter, iter + 1) / nu;
 
@@ -221,7 +225,7 @@ class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<
     H(iter, iter + 1) = 0.;
 
     gamma[iter + 1] = -s[iter] * gamma[iter];
-    gamma[iter]     = std::conj(c[iter]) * gamma[iter];
+    gamma[iter]     = conjugate(c[iter]) * gamma[iter];
     QrTimer.Stop();
   }
 
@@ -231,8 +235,8 @@ class CommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<
     for (int i = iter; i >= 0; i--) {
       y[i] = gamma[i];
       for (int k = i + 1; k <= iter; k++)
-        y[i] = y[i] - H(k, i) * y[k];
-      y[i] = y[i] / H(i, i);
+        y[i] = y[i] - ComplexD(H(k, i)) * y[k];
+      y[i] = y[i] / ComplexD(H(i, i));
     }
 
     for (int i = 0; i <= iter; i++)
diff --git a/Grid/algorithms/iterative/ConjugateGradient.h b/Grid/algorithms/iterative/ConjugateGradient.h
index c1717e2a..398f578f 100644
--- a/Grid/algorithms/iterative/ConjugateGradient.h
+++ b/Grid/algorithms/iterative/ConjugateGradient.h
@@ -27,11 +27,11 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef GRID_CONJUGATE_GRADIENT_H
 #define GRID_CONJUGATE_GRADIENT_H
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 /////////////////////////////////////////////////////////////
 // Base classes for iterative processes based on operators
@@ -40,7 +40,10 @@ namespace Grid {
 
 template <class Field>
 class ConjugateGradient : public OperatorFunction<Field> {
- public:
+public:
+
+  using OperatorFunction<Field>::operator();
+
   bool ErrorOnNoConverge;  // throw an assert when the CG fails to converge.
                            // Defaults true.
   RealD Tolerance;
@@ -48,17 +51,18 @@ class ConjugateGradient : public OperatorFunction<Field> {
   Integer IterationsToComplete; //Number of iterations the CG took to finish. Filled in upon completion
   
   ConjugateGradient(RealD tol, Integer maxit, bool err_on_no_conv = true)
-      : Tolerance(tol),
-        MaxIterations(maxit),
-        ErrorOnNoConverge(err_on_no_conv){};
+    : Tolerance(tol),
+      MaxIterations(maxit),
+      ErrorOnNoConverge(err_on_no_conv){};
 
   void operator()(LinearOperatorBase<Field> &Linop, const Field &src, Field &psi) {
 
+    psi.Checkerboard() = src.Checkerboard();
 
-    psi.checkerboard = src.checkerboard;
     conformable(psi, src);
 
-    RealD cp, c, a, d, b, ssq, qq, b_pred;
+    RealD cp, c, a, d, b, ssq, qq;
+    //RealD b_pred;
 
     Field p(src);
     Field mmp(src);
@@ -70,7 +74,7 @@ class ConjugateGradient : public OperatorFunction<Field> {
 
     
     Linop.HermOpAndNorm(psi, mmp, d, b);
-
+    
     r = src - mmp;
     p = r;
 
@@ -127,10 +131,13 @@ class ConjugateGradient : public OperatorFunction<Field> {
       b = cp / c;
 
       LinearCombTimer.Start();
-      parallel_for(int ss=0;ss<src._grid->oSites();ss++){
-	vstream(psi[ss], a      *  p[ss] + psi[ss]);
-	vstream(p  [ss], b      *  p[ss] + r[ss]);
-      }
+      auto psi_v = psi.View();
+      auto p_v   = p.View();
+      auto r_v   = r.View();
+      accelerator_for(ss,p_v.size(), Field::vector_object::Nsimd(),{
+	  coalescedWrite(psi_v[ss], a      *  p_v(ss) + psi_v(ss));
+	  coalescedWrite(p_v[ss]  , b      *  p_v(ss) + r_v  (ss));
+      });
       LinearCombTimer.Stop();
       LinalgTimer.Stop();
 
@@ -143,12 +150,12 @@ class ConjugateGradient : public OperatorFunction<Field> {
         Linop.HermOpAndNorm(psi, mmp, d, qq);
         p = mmp - src;
 
-        RealD srcnorm = sqrt(norm2(src));
-        RealD resnorm = sqrt(norm2(p));
+        RealD srcnorm = std::sqrt(norm2(src));
+        RealD resnorm = std::sqrt(norm2(p));
         RealD true_residual = resnorm / srcnorm;
 
         std::cout << GridLogMessage << "ConjugateGradient Converged on iteration " << k << std::endl;
-        std::cout << GridLogMessage << "\tComputed residual " << sqrt(cp / ssq)<<std::endl;
+        std::cout << GridLogMessage << "\tComputed residual " << std::sqrt(cp / ssq)<<std::endl;
 	std::cout << GridLogMessage << "\tTrue residual " << true_residual<<std::endl;
 	std::cout << GridLogMessage << "\tTarget " << Tolerance << std::endl;
 
@@ -174,5 +181,5 @@ class ConjugateGradient : public OperatorFunction<Field> {
 
   }
 };
-}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/algorithms/iterative/ConjugateGradientMixedPrec.h b/Grid/algorithms/iterative/ConjugateGradientMixedPrec.h
index 4c2cf12f..08942097 100644
--- a/Grid/algorithms/iterative/ConjugateGradientMixedPrec.h
+++ b/Grid/algorithms/iterative/ConjugateGradientMixedPrec.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,13 +23,12 @@ Author: Christopher Kelly <ckelly@phys.columbia.edu>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_CONJUGATE_GRADIENT_MIXED_PREC_H
 #define GRID_CONJUGATE_GRADIENT_MIXED_PREC_H
 
-namespace Grid {
-
+NAMESPACE_BEGIN(Grid);
 
   //Mixed precision restarted defect correction CG
   template<class FieldD,class FieldF, 
@@ -67,98 +66,96 @@ namespace Grid {
       guesser = &g;
     }
   
-    void operator() (const FieldD &src_d_in, FieldD &sol_d){
-      TotalInnerIterations = 0;
+  void operator() (const FieldD &src_d_in, FieldD &sol_d){
+    TotalInnerIterations = 0;
 	
-      GridStopWatch TotalTimer;
-      TotalTimer.Start();
+    GridStopWatch TotalTimer;
+    TotalTimer.Start();
     
-      int cb = src_d_in.checkerboard;
-      sol_d.checkerboard = cb;
+    int cb = src_d_in.Checkerboard();
+    sol_d.Checkerboard() = cb;
     
-      RealD src_norm = norm2(src_d_in);
-      RealD stop = src_norm * Tolerance*Tolerance;
+    RealD src_norm = norm2(src_d_in);
+    RealD stop = src_norm * Tolerance*Tolerance;
 
-      GridBase* DoublePrecGrid = src_d_in._grid;
-      FieldD tmp_d(DoublePrecGrid);
-      tmp_d.checkerboard = cb;
+    GridBase* DoublePrecGrid = src_d_in.Grid();
+    FieldD tmp_d(DoublePrecGrid);
+    tmp_d.Checkerboard() = cb;
     
-      FieldD tmp2_d(DoublePrecGrid);
-      tmp2_d.checkerboard = cb;
+    FieldD tmp2_d(DoublePrecGrid);
+    tmp2_d.Checkerboard() = cb;
     
-      FieldD src_d(DoublePrecGrid);
-      src_d = src_d_in; //source for next inner iteration, computed from residual during operation
+    FieldD src_d(DoublePrecGrid);
+    src_d = src_d_in; //source for next inner iteration, computed from residual during operation
     
-      RealD inner_tol = InnerTolerance;
+    RealD inner_tol = InnerTolerance;
     
-      FieldF src_f(SinglePrecGrid);
-      src_f.checkerboard = cb;
+    FieldF src_f(SinglePrecGrid);
+    src_f.Checkerboard() = cb;
     
-      FieldF sol_f(SinglePrecGrid);
-      sol_f.checkerboard = cb;
+    FieldF sol_f(SinglePrecGrid);
+    sol_f.Checkerboard() = cb;
     
-      ConjugateGradient<FieldF> CG_f(inner_tol, MaxInnerIterations);
-      CG_f.ErrorOnNoConverge = false;
+    ConjugateGradient<FieldF> CG_f(inner_tol, MaxInnerIterations);
+    CG_f.ErrorOnNoConverge = false;
 
-      GridStopWatch InnerCGtimer;
+    GridStopWatch InnerCGtimer;
 
-      GridStopWatch PrecChangeTimer;
+    GridStopWatch PrecChangeTimer;
     
-      Integer &outer_iter = TotalOuterIterations; //so it will be equal to the final iteration count
+    Integer &outer_iter = TotalOuterIterations; //so it will be equal to the final iteration count
       
-      for(outer_iter = 0; outer_iter < MaxOuterIterations; outer_iter++){
-	//Compute double precision rsd and also new RHS vector.
-	Linop_d.HermOp(sol_d, tmp_d);
-	RealD norm = axpy_norm(src_d, -1., tmp_d, src_d_in); //src_d is residual vector
+    for(outer_iter = 0; outer_iter < MaxOuterIterations; outer_iter++){
+      //Compute double precision rsd and also new RHS vector.
+      Linop_d.HermOp(sol_d, tmp_d);
+      RealD norm = axpy_norm(src_d, -1., tmp_d, src_d_in); //src_d is residual vector
       
-	std::cout<<GridLogMessage<<"MixedPrecisionConjugateGradient: Outer iteration " <<outer_iter<<" residual "<< norm<< " target "<< stop<<std::endl;
+      std::cout<<GridLogMessage<<"MixedPrecisionConjugateGradient: Outer iteration " <<outer_iter<<" residual "<< norm<< " target "<< stop<<std::endl;
 
-	if(norm < OuterLoopNormMult * stop){
-	  std::cout<<GridLogMessage<<"MixedPrecisionConjugateGradient: Outer iteration converged on iteration " <<outer_iter <<std::endl;
-	  break;
-	}
-	while(norm * inner_tol * inner_tol < stop) inner_tol *= 2;  // inner_tol = sqrt(stop/norm) ??
-
-	PrecChangeTimer.Start();
-	precisionChange(src_f, src_d);
-	PrecChangeTimer.Stop();
-      
-	zeroit(sol_f);
-
-	//Optionally improve inner solver guess (eg using known eigenvectors)
-	if(guesser != NULL)
-	  (*guesser)(src_f, sol_f);
-
-	//Inner CG
-	CG_f.Tolerance = inner_tol;
-	InnerCGtimer.Start();
-	CG_f(Linop_f, src_f, sol_f);
-	InnerCGtimer.Stop();
-	TotalInnerIterations += CG_f.IterationsToComplete;
-      
-	//Convert sol back to double and add to double prec solution
-	PrecChangeTimer.Start();
-	precisionChange(tmp_d, sol_f);
-	PrecChangeTimer.Stop();
-      
-	axpy(sol_d, 1.0, tmp_d, sol_d);
+      if(norm < OuterLoopNormMult * stop){
+	std::cout<<GridLogMessage<<"MixedPrecisionConjugateGradient: Outer iteration converged on iteration " <<outer_iter <<std::endl;
+	break;
       }
-    
-      //Final trial CG
-      std::cout<<GridLogMessage<<"MixedPrecisionConjugateGradient: Starting final patch-up double-precision solve"<<std::endl;
-    
-      ConjugateGradient<FieldD> CG_d(Tolerance, MaxInnerIterations);
-      CG_d(Linop_d, src_d_in, sol_d);
-      TotalFinalStepIterations = CG_d.IterationsToComplete;
+      while(norm * inner_tol * inner_tol < stop) inner_tol *= 2;  // inner_tol = sqrt(stop/norm) ??
 
-      TotalTimer.Stop();
-      std::cout<<GridLogMessage<<"MixedPrecisionConjugateGradient: Inner CG iterations " << TotalInnerIterations << " Restarts " << TotalOuterIterations << " Final CG iterations " << TotalFinalStepIterations << std::endl;
-      std::cout<<GridLogMessage<<"MixedPrecisionConjugateGradient: Total time " << TotalTimer.Elapsed() << " Precision change " << PrecChangeTimer.Elapsed() << " Inner CG total " << InnerCGtimer.Elapsed() << std::endl;
+      PrecChangeTimer.Start();
+      precisionChange(src_f, src_d);
+      PrecChangeTimer.Stop();
+      
+      sol_f = Zero();
+
+      //Optionally improve inner solver guess (eg using known eigenvectors)
+      if(guesser != NULL)
+	(*guesser)(src_f, sol_f);
+
+      //Inner CG
+      CG_f.Tolerance = inner_tol;
+      InnerCGtimer.Start();
+      CG_f(Linop_f, src_f, sol_f);
+      InnerCGtimer.Stop();
+      TotalInnerIterations += CG_f.IterationsToComplete;
+      
+      //Convert sol back to double and add to double prec solution
+      PrecChangeTimer.Start();
+      precisionChange(tmp_d, sol_f);
+      PrecChangeTimer.Stop();
+      
+      axpy(sol_d, 1.0, tmp_d, sol_d);
     }
-  };
+    
+    //Final trial CG
+    std::cout<<GridLogMessage<<"MixedPrecisionConjugateGradient: Starting final patch-up double-precision solve"<<std::endl;
+    
+    ConjugateGradient<FieldD> CG_d(Tolerance, MaxInnerIterations);
+    CG_d(Linop_d, src_d_in, sol_d);
+    TotalFinalStepIterations = CG_d.IterationsToComplete;
 
+    TotalTimer.Stop();
+    std::cout<<GridLogMessage<<"MixedPrecisionConjugateGradient: Inner CG iterations " << TotalInnerIterations << " Restarts " << TotalOuterIterations << " Final CG iterations " << TotalFinalStepIterations << std::endl;
+    std::cout<<GridLogMessage<<"MixedPrecisionConjugateGradient: Total time " << TotalTimer.Elapsed() << " Precision change " << PrecChangeTimer.Elapsed() << " Inner CG total " << InnerCGtimer.Elapsed() << std::endl;
+  }
+};
 
-
-}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/algorithms/iterative/ConjugateGradientMultiShift.h b/Grid/algorithms/iterative/ConjugateGradientMultiShift.h
index 4c311bc1..7e8d01e2 100644
--- a/Grid/algorithms/iterative/ConjugateGradientMultiShift.h
+++ b/Grid/algorithms/iterative/ConjugateGradientMultiShift.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,147 +24,150 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_CONJUGATE_MULTI_SHIFT_GRADIENT_H
 #define GRID_CONJUGATE_MULTI_SHIFT_GRADIENT_H
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-    /////////////////////////////////////////////////////////////
-    // Base classes for iterative processes based on operators
-    // single input vec, single output vec.
-    /////////////////////////////////////////////////////////////
+/////////////////////////////////////////////////////////////
+// Base classes for iterative processes based on operators
+// single input vec, single output vec.
+/////////////////////////////////////////////////////////////
 
-  template<class Field> 
-    class ConjugateGradientMultiShift : public OperatorMultiFunction<Field>,
-                                        public OperatorFunction<Field>
-    {
+template<class Field> 
+class ConjugateGradientMultiShift : public OperatorMultiFunction<Field>,
+				    public OperatorFunction<Field>
+{
 public:                                                
-    RealD   Tolerance;
-    Integer MaxIterations;
-    Integer IterationsToComplete; //Number of iterations the CG took to finish. Filled in upon completion
-    int verbose;
-    MultiShiftFunction shifts;
 
-    ConjugateGradientMultiShift(Integer maxit,MultiShiftFunction &_shifts) : 
-	MaxIterations(maxit),
-	shifts(_shifts)
-    { 
-      verbose=1;
+  using OperatorFunction<Field>::operator();
+
+  RealD   Tolerance;
+  Integer MaxIterations;
+    Integer IterationsToComplete; //Number of iterations the CG took to finish. Filled in upon completion
+  int verbose;
+  MultiShiftFunction shifts;
+
+  ConjugateGradientMultiShift(Integer maxit,MultiShiftFunction &_shifts) : 
+    MaxIterations(maxit),
+    shifts(_shifts)
+  { 
+    verbose=1;
+  }
+
+  void operator() (LinearOperatorBase<Field> &Linop, const Field &src, Field &psi)
+  {
+    GridBase *grid = src.Grid();
+    int nshift = shifts.order;
+    std::vector<Field> results(nshift,grid);
+    (*this)(Linop,src,results,psi);
+  }
+  void operator() (LinearOperatorBase<Field> &Linop, const Field &src, std::vector<Field> &results, Field &psi)
+  {
+    int nshift = shifts.order;
+
+    (*this)(Linop,src,results);
+  
+    psi = shifts.norm*src;
+    for(int i=0;i<nshift;i++){
+      psi = psi + shifts.residues[i]*results[i];
     }
 
-void operator() (LinearOperatorBase<Field> &Linop, const Field &src, Field &psi)
-{
-  GridBase *grid = src._grid;
-  int nshift = shifts.order;
-  std::vector<Field> results(nshift,grid);
-  (*this)(Linop,src,results,psi);
-}
-void operator() (LinearOperatorBase<Field> &Linop, const Field &src, std::vector<Field> &results, Field &psi)
-{
-  int nshift = shifts.order;
-
-  (*this)(Linop,src,results);
-  
-  psi = shifts.norm*src;
-  for(int i=0;i<nshift;i++){
-    psi = psi + shifts.residues[i]*results[i];
+    return;
   }
 
-  return;
-}
+  void operator() (LinearOperatorBase<Field> &Linop, const Field &src, std::vector<Field> &psi)
+  {
+  
+    GridBase *grid = src.Grid();
+  
+    ////////////////////////////////////////////////////////////////////////
+    // Convenience references to the info stored in "MultiShiftFunction"
+    ////////////////////////////////////////////////////////////////////////
+    int nshift = shifts.order;
 
-void operator() (LinearOperatorBase<Field> &Linop, const Field &src, std::vector<Field> &psi)
-{
-  
-  GridBase *grid = src._grid;
-  
-  ////////////////////////////////////////////////////////////////////////
-  // Convenience references to the info stored in "MultiShiftFunction"
-  ////////////////////////////////////////////////////////////////////////
-  int nshift = shifts.order;
+    std::vector<RealD> &mass(shifts.poles); // Make references to array in "shifts"
+    std::vector<RealD> &mresidual(shifts.tolerances);
+    std::vector<RealD> alpha(nshift,1.0);
+    std::vector<Field>   ps(nshift,grid);// Search directions
 
-  std::vector<RealD> &mass(shifts.poles); // Make references to array in "shifts"
-  std::vector<RealD> &mresidual(shifts.tolerances);
-  std::vector<RealD> alpha(nshift,1.0);
-  std::vector<Field>   ps(nshift,grid);// Search directions
-
-  assert(psi.size()==nshift);
-  assert(mass.size()==nshift);
-  assert(mresidual.size()==nshift);
+    assert(psi.size()==nshift);
+    assert(mass.size()==nshift);
+    assert(mresidual.size()==nshift);
   
-  // dynamic sized arrays on stack; 2d is a pain with vector
-  RealD  bs[nshift];
-  RealD  rsq[nshift];
-  RealD  z[nshift][2];
-  int     converged[nshift];
+    // dynamic sized arrays on stack; 2d is a pain with vector
+    RealD  bs[nshift];
+    RealD  rsq[nshift];
+    RealD  z[nshift][2];
+    int     converged[nshift];
   
-  const int       primary =0;
+    const int       primary =0;
   
-  //Primary shift fields CG iteration
-  RealD a,b,c,d;
-  RealD cp,bp,qq; //prev
+    //Primary shift fields CG iteration
+    RealD a,b,c,d;
+    RealD cp,bp,qq; //prev
   
-  // Matrix mult fields
-  Field r(grid);
-  Field p(grid);
-  Field tmp(grid);
-  Field mmp(grid);
+    // Matrix mult fields
+    Field r(grid);
+    Field p(grid);
+    Field tmp(grid);
+    Field mmp(grid);
   
-  // Check lightest mass
-  for(int s=0;s<nshift;s++){
-    assert( mass[s]>= mass[primary] );
-    converged[s]=0;
-  }
+    // Check lightest mass
+    for(int s=0;s<nshift;s++){
+      assert( mass[s]>= mass[primary] );
+      converged[s]=0;
+    }
   
-  // Wire guess to zero
-  // Residuals "r" are src
-  // First search direction "p" is also src
-  cp = norm2(src);
-  for(int s=0;s<nshift;s++){
-    rsq[s] = cp * mresidual[s] * mresidual[s];
-    std::cout<<GridLogMessage<<"ConjugateGradientMultiShift: shift "<<s
-	     <<" target resid "<<rsq[s]<<std::endl;
-    ps[s] = src;
-  }
-  // r and p for primary
-  r=src;
-  p=src;
+    // Wire guess to zero
+    // Residuals "r" are src
+    // First search direction "p" is also src
+    cp = norm2(src);
+    for(int s=0;s<nshift;s++){
+      rsq[s] = cp * mresidual[s] * mresidual[s];
+      std::cout<<GridLogMessage<<"ConjugateGradientMultiShift: shift "<<s
+	       <<" target resid "<<rsq[s]<<std::endl;
+      ps[s] = src;
+    }
+    // r and p for primary
+    r=src;
+    p=src;
   
-  //MdagM+m[0]
-  Linop.HermOpAndNorm(p,mmp,d,qq);
-  axpy(mmp,mass[0],p,mmp);
-  RealD rn = norm2(p);
-  d += rn*mass[0];
+    //MdagM+m[0]
+    Linop.HermOpAndNorm(p,mmp,d,qq);
+    axpy(mmp,mass[0],p,mmp);
+    RealD rn = norm2(p);
+    d += rn*mass[0];
   
-  // have verified that inner product of 
-  // p and mmp is equal to d after this since
-  // the d computation is tricky
-  //  qq = real(innerProduct(p,mmp));
-  //  std::cout<<GridLogMessage << "debug equal ?  qq "<<qq<<" d "<< d<<std::endl;
+    // have verified that inner product of 
+    // p and mmp is equal to d after this since
+    // the d computation is tricky
+    //  qq = real(innerProduct(p,mmp));
+    //  std::cout<<GridLogMessage << "debug equal ?  qq "<<qq<<" d "<< d<<std::endl;
   
-  b = -cp /d;
+    b = -cp /d;
   
-  // Set up the various shift variables
-  int       iz=0;
-  z[0][1-iz] = 1.0;
-  z[0][iz]   = 1.0;
-  bs[0]      = b;
-  for(int s=1;s<nshift;s++){
-    z[s][1-iz] = 1.0;
-    z[s][iz]   = 1.0/( 1.0 - b*(mass[s]-mass[0]));
-    bs[s]      = b*z[s][iz]; 
-  }
+    // Set up the various shift variables
+    int       iz=0;
+    z[0][1-iz] = 1.0;
+    z[0][iz]   = 1.0;
+    bs[0]      = b;
+    for(int s=1;s<nshift;s++){
+      z[s][1-iz] = 1.0;
+      z[s][iz]   = 1.0/( 1.0 - b*(mass[s]-mass[0]));
+      bs[s]      = b*z[s][iz]; 
+    }
   
-  // r += b[0] A.p[0]
-  // c= norm(r)
-  c=axpy_norm(r,b,mmp,r);
+    // r += b[0] A.p[0]
+    // c= norm(r)
+    c=axpy_norm(r,b,mmp,r);
+  
+    for(int s=0;s<nshift;s++) {
+      axpby(psi[s],0.,-bs[s]*alpha[s],src,src);
+    }
   
-  for(int s=0;s<nshift;s++) {
-    axpby(psi[s],0.,-bs[s]*alpha[s],src,src);
-  }
- 
   ///////////////////////////////////////
   // Timers
   ///////////////////////////////////////
@@ -175,37 +178,37 @@ void operator() (LinearOperatorBase<Field> &Linop, const Field &src, std::vector
   GridStopWatch SolverTimer;
   SolverTimer.Start();
   
-  // Iteration loop
-  int k;
+    // Iteration loop
+    int k;
   
-  for (k=1;k<=MaxIterations;k++){
+    for (k=1;k<=MaxIterations;k++){
     
-    a = c /cp;
+      a = c /cp;
     AXPYTimer.Start();
-    axpy(p,a,p,r);
+      axpy(p,a,p,r);
     AXPYTimer.Stop();
     
-    // Note to self - direction ps is iterated seperately
-    // for each shift. Does not appear to have any scope
-    // for avoiding linear algebra in "single" case.
-    // 
-    // However SAME r is used. Could load "r" and update
-    // ALL ps[s]. 2/3 Bandwidth saving
-    // New Kernel: Load r, vector of coeffs, vector of pointers ps
+      // Note to self - direction ps is iterated seperately
+      // for each shift. Does not appear to have any scope
+      // for avoiding linear algebra in "single" case.
+      // 
+      // However SAME r is used. Could load "r" and update
+      // ALL ps[s]. 2/3 Bandwidth saving
+      // New Kernel: Load r, vector of coeffs, vector of pointers ps
     AXPYTimer.Start();
-    for(int s=0;s<nshift;s++){
-      if ( ! converged[s] ) { 
-	if (s==0){
-	  axpy(ps[s],a,ps[s],r);
-	} else{
-	  RealD as =a *z[s][iz]*bs[s] /(z[s][1-iz]*b);
-	  axpby(ps[s],z[s][iz],as,r,ps[s]);
+      for(int s=0;s<nshift;s++){
+	if ( ! converged[s] ) { 
+	  if (s==0){
+	    axpy(ps[s],a,ps[s],r);
+	  } else{
+	    RealD as =a *z[s][iz]*bs[s] /(z[s][1-iz]*b);
+	    axpby(ps[s],z[s][iz],as,r,ps[s]);
+	  }
 	}
       }
-    }
     AXPYTimer.Stop();
     
-    cp=c;
+      cp=c;
     MatrixTimer.Start();  
     //Linop.HermOpAndNorm(p,mmp,d,qq); // d is used
     // The below is faster on KNL
@@ -215,89 +218,89 @@ void operator() (LinearOperatorBase<Field> &Linop, const Field &src, std::vector
     MatrixTimer.Stop();  
 
     AXPYTimer.Start();
-    axpy(mmp,mass[0],p,mmp);
+      axpy(mmp,mass[0],p,mmp);
     AXPYTimer.Stop();
-    RealD rn = norm2(p);
-    d += rn*mass[0];
+      RealD rn = norm2(p);
+      d += rn*mass[0];
     
-    bp=b;
-    b=-cp/d;
+      bp=b;
+      b=-cp/d;
     
     AXPYTimer.Start();
-    c=axpy_norm(r,b,mmp,r);
+      c=axpy_norm(r,b,mmp,r);
     AXPYTimer.Stop();
 
-    // Toggle the recurrence history
-    bs[0] = b;
-    iz = 1-iz;
+      // Toggle the recurrence history
+      bs[0] = b;
+      iz = 1-iz;
     ShiftTimer.Start();
-    for(int s=1;s<nshift;s++){
-      if((!converged[s])){
-	RealD z0 = z[s][1-iz];
-	RealD z1 = z[s][iz];
-	z[s][iz] = z0*z1*bp
-	  / (b*a*(z1-z0) + z1*bp*(1- (mass[s]-mass[0])*b)); 
-	bs[s] = b*z[s][iz]/z0; // NB sign  rel to Mike
+      for(int s=1;s<nshift;s++){
+	if((!converged[s])){
+	  RealD z0 = z[s][1-iz];
+	  RealD z1 = z[s][iz];
+	  z[s][iz] = z0*z1*bp
+	    / (b*a*(z1-z0) + z1*bp*(1- (mass[s]-mass[0])*b)); 
+	  bs[s] = b*z[s][iz]/z0; // NB sign  rel to Mike
+	}
       }
-    }
     ShiftTimer.Stop();
     
-    for(int s=0;s<nshift;s++){
-      int ss = s;
-      // Scope for optimisation here in case of "single".
-      // Could load psi[0] and pull all ps[s] in.
-      //      if ( single ) ss=primary;
-      // Bandwith saving in single case is Ls * 3 -> 2+Ls, so ~ 3x saving
-      // Pipelined CG gain:
-      //
-      // New Kernel: Load r, vector of coeffs, vector of pointers ps
-      // New Kernel: Load psi[0], vector of coeffs, vector of pointers ps
-      // If can predict the coefficient bs then we can fuse these and avoid write reread cyce
-      //  on ps[s].
-      // Before:  3 x npole  + 3 x npole
-      // After :  2 x npole (ps[s])        => 3x speed up of multishift CG.
+      for(int s=0;s<nshift;s++){
+	int ss = s;
+	// Scope for optimisation here in case of "single".
+	// Could load psi[0] and pull all ps[s] in.
+	//      if ( single ) ss=primary;
+	// Bandwith saving in single case is Ls * 3 -> 2+Ls, so ~ 3x saving
+	// Pipelined CG gain:
+	//
+	// New Kernel: Load r, vector of coeffs, vector of pointers ps
+	// New Kernel: Load psi[0], vector of coeffs, vector of pointers ps
+	// If can predict the coefficient bs then we can fuse these and avoid write reread cyce
+	//  on ps[s].
+	// Before:  3 x npole  + 3 x npole
+	// After :  2 x npole (ps[s])        => 3x speed up of multishift CG.
       
-      if( (!converged[s]) ) { 
-	axpy(psi[ss],-bs[s]*alpha[s],ps[s],psi[ss]);
-      }
-    }
-    
-    // Convergence checks
-    int all_converged = 1;
-    for(int s=0;s<nshift;s++){
-      
-      if ( (!converged[s]) ){
-	
-	RealD css  = c * z[s][iz]* z[s][iz];
-	
-	if(css<rsq[s]){
-	  if ( ! converged[s] )
-	    std::cout<<GridLogMessage<<"ConjugateGradientMultiShift k="<<k<<" Shift "<<s<<" has converged"<<std::endl;
-	      converged[s]=1;
-	} else {
-	  all_converged=0;
+	if( (!converged[s]) ) { 
+	  axpy(psi[ss],-bs[s]*alpha[s],ps[s],psi[ss]);
 	}
-
       }
-    }
     
-    if ( all_converged ){
+      // Convergence checks
+      int all_converged = 1;
+      for(int s=0;s<nshift;s++){
+      
+	if ( (!converged[s]) ){
+	
+	  RealD css  = c * z[s][iz]* z[s][iz];
+	
+	  if(css<rsq[s]){
+	    if ( ! converged[s] )
+	      std::cout<<GridLogMessage<<"ConjugateGradientMultiShift k="<<k<<" Shift "<<s<<" has converged"<<std::endl;
+	    converged[s]=1;
+	  } else {
+	    all_converged=0;
+	  }
+
+	}
+      }
+    
+      if ( all_converged ){
 
     SolverTimer.Stop();
 
 
-      std::cout<<GridLogMessage<< "CGMultiShift: All shifts have converged iteration "<<k<<std::endl;
-      std::cout<<GridLogMessage<< "CGMultiShift: Checking solutions"<<std::endl;
+	std::cout<<GridLogMessage<< "CGMultiShift: All shifts have converged iteration "<<k<<std::endl;
+	std::cout<<GridLogMessage<< "CGMultiShift: Checking solutions"<<std::endl;
       
-      // Check answers 
-      for(int s=0; s < nshift; s++) { 
-	Linop.HermOpAndNorm(psi[s],mmp,d,qq);
-	axpy(tmp,mass[s],psi[s],mmp);
-	axpy(r,-alpha[s],src,tmp);
-	RealD rn = norm2(r);
-	RealD cn = norm2(src);
-	std::cout<<GridLogMessage<<"CGMultiShift: shift["<<s<<"] true residual "<<std::sqrt(rn/cn)<<std::endl;
-      }
+	// Check answers 
+	for(int s=0; s < nshift; s++) { 
+	  Linop.HermOpAndNorm(psi[s],mmp,d,qq);
+	  axpy(tmp,mass[s],psi[s],mmp);
+	  axpy(r,-alpha[s],src,tmp);
+	  RealD rn = norm2(r);
+	  RealD cn = norm2(src);
+	  std::cout<<GridLogMessage<<"CGMultiShift: shift["<<s<<"] true residual "<<std::sqrt(rn/cn)<<std::endl;
+	}
 
       std::cout << GridLogMessage << "Time Breakdown "<<std::endl;
       std::cout << GridLogMessage << "\tElapsed    " << SolverTimer.Elapsed()     <<std::endl;
@@ -307,16 +310,16 @@ void operator() (LinearOperatorBase<Field> &Linop, const Field &src, std::vector
 
       IterationsToComplete = k;	
 
-      return;
-    }
+	return;
+      }
 
    
+    }
+    // ugly hack
+    std::cout<<GridLogMessage<<"CG multi shift did not converge"<<std::endl;
+    //  assert(0);
   }
-  // ugly hack
-  std::cout<<GridLogMessage<<"CG multi shift did not converge"<<std::endl;
-//  assert(0);
-}
 
-  };
-}
+};
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/algorithms/iterative/ConjugateGradientReliableUpdate.h b/Grid/algorithms/iterative/ConjugateGradientReliableUpdate.h
index 13666f97..ab12a41f 100644
--- a/Grid/algorithms/iterative/ConjugateGradientReliableUpdate.h
+++ b/Grid/algorithms/iterative/ConjugateGradientReliableUpdate.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,234 +23,236 @@ Author: Christopher Kelly <ckelly@phys.columbia.edu>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_CONJUGATE_GRADIENT_RELIABLE_UPDATE_H
 #define GRID_CONJUGATE_GRADIENT_RELIABLE_UPDATE_H
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  template<class FieldD,class FieldF, typename std::enable_if< getPrecision<FieldD>::value == 2, int>::type = 0,typename std::enable_if< getPrecision<FieldF>::value == 1, int>::type = 0> 
-  class ConjugateGradientReliableUpdate : public LinearFunction<FieldD> {
-  public:
-    bool ErrorOnNoConverge;  // throw an assert when the CG fails to converge.
-    // Defaults true.
-    RealD Tolerance;
-    Integer MaxIterations;
-    Integer IterationsToComplete; //Number of iterations the CG took to finish. Filled in upon completion
-    Integer ReliableUpdatesPerformed;
+template<class FieldD,class FieldF, 
+	 typename std::enable_if< getPrecision<FieldD>::value == 2, int>::type = 0,
+	 typename std::enable_if< getPrecision<FieldF>::value == 1, int>::type = 0> 
+class ConjugateGradientReliableUpdate : public LinearFunction<FieldD> {
+public:
+  bool ErrorOnNoConverge;  // throw an assert when the CG fails to converge.
+  // Defaults true.
+  RealD Tolerance;
+  Integer MaxIterations;
+  Integer IterationsToComplete; //Number of iterations the CG took to finish. Filled in upon completion
+  Integer ReliableUpdatesPerformed;
 
-    bool DoFinalCleanup; //Final DP cleanup, defaults to true
-    Integer IterationsToCleanup; //Final DP cleanup step iterations
+  bool DoFinalCleanup; //Final DP cleanup, defaults to true
+  Integer IterationsToCleanup; //Final DP cleanup step iterations
     
-    LinearOperatorBase<FieldF> &Linop_f;
-    LinearOperatorBase<FieldD> &Linop_d;
-    GridBase* SinglePrecGrid;
-    RealD Delta; //reliable update parameter
+  LinearOperatorBase<FieldF> &Linop_f;
+  LinearOperatorBase<FieldD> &Linop_d;
+  GridBase* SinglePrecGrid;
+  RealD Delta; //reliable update parameter
 
-    //Optional ability to switch to a different linear operator once the tolerance reaches a certain point. Useful for single/half -> single/single
-    LinearOperatorBase<FieldF> *Linop_fallback;
-    RealD fallback_transition_tol;
+  //Optional ability to switch to a different linear operator once the tolerance reaches a certain point. Useful for single/half -> single/single
+  LinearOperatorBase<FieldF> *Linop_fallback;
+  RealD fallback_transition_tol;
 
     
-    ConjugateGradientReliableUpdate(RealD tol, Integer maxit, RealD _delta, GridBase* _sp_grid, LinearOperatorBase<FieldF> &_Linop_f, LinearOperatorBase<FieldD> &_Linop_d, bool err_on_no_conv = true)
-      : Tolerance(tol),
-        MaxIterations(maxit),
-	Delta(_delta),
-	Linop_f(_Linop_f),
-	Linop_d(_Linop_d),
-	SinglePrecGrid(_sp_grid),
-        ErrorOnNoConverge(err_on_no_conv),
-	DoFinalCleanup(true),
-	Linop_fallback(NULL)
-    {};
+  ConjugateGradientReliableUpdate(RealD tol, Integer maxit, RealD _delta, GridBase* _sp_grid, LinearOperatorBase<FieldF> &_Linop_f, LinearOperatorBase<FieldD> &_Linop_d, bool err_on_no_conv = true)
+    : Tolerance(tol),
+      MaxIterations(maxit),
+      Delta(_delta),
+      Linop_f(_Linop_f),
+      Linop_d(_Linop_d),
+      SinglePrecGrid(_sp_grid),
+      ErrorOnNoConverge(err_on_no_conv),
+      DoFinalCleanup(true),
+      Linop_fallback(NULL)
+  {};
 
-    void setFallbackLinop(LinearOperatorBase<FieldF> &_Linop_fallback, const RealD _fallback_transition_tol){
-      Linop_fallback = &_Linop_fallback;
-      fallback_transition_tol = _fallback_transition_tol;      
-    }
+  void setFallbackLinop(LinearOperatorBase<FieldF> &_Linop_fallback, const RealD _fallback_transition_tol){
+    Linop_fallback = &_Linop_fallback;
+    fallback_transition_tol = _fallback_transition_tol;      
+  }
     
-    void operator()(const FieldD &src, FieldD &psi) {
-      LinearOperatorBase<FieldF> *Linop_f_use = &Linop_f;
-      bool using_fallback = false;
+  void operator()(const FieldD &src, FieldD &psi) {
+    LinearOperatorBase<FieldF> *Linop_f_use = &Linop_f;
+    bool using_fallback = false;
       
-      psi.checkerboard = src.checkerboard;
-      conformable(psi, src);
+    psi.Checkerboard() = src.Checkerboard();
+    conformable(psi, src);
 
-      RealD cp, c, a, d, b, ssq, qq, b_pred;
+    RealD cp, c, a, d, b, ssq, qq, b_pred;
 
-      FieldD p(src);
-      FieldD mmp(src);
-      FieldD r(src);
+    FieldD p(src);
+    FieldD mmp(src);
+    FieldD r(src);
 
-      // Initial residual computation & set up
-      RealD guess = norm2(psi);
-      assert(std::isnan(guess) == 0);
+    // Initial residual computation & set up
+    RealD guess = norm2(psi);
+    assert(std::isnan(guess) == 0);
     
-      Linop_d.HermOpAndNorm(psi, mmp, d, b);
+    Linop_d.HermOpAndNorm(psi, mmp, d, b);
     
-      r = src - mmp;
-      p = r;
+    r = src - mmp;
+    p = r;
 
-      a = norm2(p);
-      cp = a;
-      ssq = norm2(src);
+    a = norm2(p);
+    cp = a;
+    ssq = norm2(src);
 
-      std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradientReliableUpdate: guess " << guess << std::endl;
-      std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradientReliableUpdate:   src " << ssq << std::endl;
-      std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradientReliableUpdate:    mp " << d << std::endl;
-      std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradientReliableUpdate:   mmp " << b << std::endl;
-      std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradientReliableUpdate:  cp,r " << cp << std::endl;
-      std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradientReliableUpdate:     p " << a << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradientReliableUpdate: guess " << guess << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradientReliableUpdate:   src " << ssq << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradientReliableUpdate:    mp " << d << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradientReliableUpdate:   mmp " << b << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradientReliableUpdate:  cp,r " << cp << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4) << "ConjugateGradientReliableUpdate:     p " << a << std::endl;
 
-      RealD rsq = Tolerance * Tolerance * ssq;
+    RealD rsq = Tolerance * Tolerance * ssq;
 
-      // Check if guess is really REALLY good :)
+    // Check if guess is really REALLY good :)
+    if (cp <= rsq) {
+      std::cout << GridLogMessage << "ConjugateGradientReliableUpdate guess was REALLY good\n";
+      std::cout << GridLogMessage << "\tComputed residual " << std::sqrt(cp / ssq)<<std::endl;
+      return;
+    }
+
+    //Single prec initialization
+    FieldF r_f(SinglePrecGrid);
+    r_f.Checkerboard() = r.Checkerboard();
+    precisionChange(r_f, r);
+
+    FieldF psi_f(r_f);
+    psi_f = Zero();
+
+    FieldF p_f(r_f);
+    FieldF mmp_f(r_f);
+
+    RealD MaxResidSinceLastRelUp = cp; //initial residual    
+    
+    std::cout << GridLogIterative << std::setprecision(4)
+	      << "ConjugateGradient: k=0 residual " << cp << " target " << rsq << std::endl;
+
+    GridStopWatch LinalgTimer;
+    GridStopWatch MatrixTimer;
+    GridStopWatch SolverTimer;
+
+    SolverTimer.Start();
+    int k = 0;
+    int l = 0;
+    
+    for (k = 1; k <= MaxIterations; k++) {
+      c = cp;
+
+      MatrixTimer.Start();
+      Linop_f_use->HermOpAndNorm(p_f, mmp_f, d, qq);
+      MatrixTimer.Stop();
+
+      LinalgTimer.Start();
+
+      a = c / d;
+      b_pred = a * (a * qq - d) / c;
+
+      cp = axpy_norm(r_f, -a, mmp_f, r_f);
+      b = cp / c;
+
+      // Fuse these loops ; should be really easy
+      psi_f = a * p_f + psi_f;
+      //p_f = p_f * b + r_f;
+
+      LinalgTimer.Stop();
+
+      std::cout << GridLogIterative << "ConjugateGradientReliableUpdate: Iteration " << k
+		<< " residual " << cp << " target " << rsq << std::endl;
+      std::cout << GridLogDebug << "a = "<< a << " b_pred = "<< b_pred << "  b = "<< b << std::endl;
+      std::cout << GridLogDebug << "qq = "<< qq << " d = "<< d << "  c = "<< c << std::endl;
+
+      if(cp > MaxResidSinceLastRelUp){
+	std::cout << GridLogIterative << "ConjugateGradientReliableUpdate: updating MaxResidSinceLastRelUp : " << MaxResidSinceLastRelUp << " -> " << cp << std::endl;
+	MaxResidSinceLastRelUp = cp;
+      }
+	  
+      // Stopping condition
       if (cp <= rsq) {
-	std::cout << GridLogMessage << "ConjugateGradientReliableUpdate guess was REALLY good\n";
-	std::cout << GridLogMessage << "\tComputed residual " << sqrt(cp / ssq)<<std::endl;
+	//Although not written in the paper, I assume that I have to add on the final solution
+	precisionChange(mmp, psi_f);
+	psi = psi + mmp;
+	
+	
+	SolverTimer.Stop();
+	Linop_d.HermOpAndNorm(psi, mmp, d, qq);
+	p = mmp - src;
+
+	RealD srcnorm = std::sqrt(norm2(src));
+	RealD resnorm = std::sqrt(norm2(p));
+	RealD true_residual = resnorm / srcnorm;
+
+	std::cout << GridLogMessage << "ConjugateGradientReliableUpdate Converged on iteration " << k << " after " << l << " reliable updates" << std::endl;
+	std::cout << GridLogMessage << "\tComputed residual " << std::sqrt(cp / ssq)<<std::endl;
+	std::cout << GridLogMessage << "\tTrue residual " << true_residual<<std::endl;
+	std::cout << GridLogMessage << "\tTarget " << Tolerance << std::endl;
+
+	std::cout << GridLogMessage << "Time breakdown "<<std::endl;
+	std::cout << GridLogMessage << "\tElapsed    " << SolverTimer.Elapsed() <<std::endl;
+	std::cout << GridLogMessage << "\tMatrix     " << MatrixTimer.Elapsed() <<std::endl;
+	std::cout << GridLogMessage << "\tLinalg     " << LinalgTimer.Elapsed() <<std::endl;
+
+	IterationsToComplete = k;	
+	ReliableUpdatesPerformed = l;
+	  
+	if(DoFinalCleanup){
+	  //Do a final CG to cleanup
+	  std::cout << GridLogMessage << "ConjugateGradientReliableUpdate performing final cleanup.\n";
+	  ConjugateGradient<FieldD> CG(Tolerance,MaxIterations);
+	  CG.ErrorOnNoConverge = ErrorOnNoConverge;
+	  CG(Linop_d,src,psi);
+	  IterationsToCleanup = CG.IterationsToComplete;
+	}
+	else if (ErrorOnNoConverge) assert(true_residual / Tolerance < 10000.0);
+
+	std::cout << GridLogMessage << "ConjugateGradientReliableUpdate complete.\n";
 	return;
       }
+      else if(cp < Delta * MaxResidSinceLastRelUp) { //reliable update
+	std::cout << GridLogMessage << "ConjugateGradientReliableUpdate "
+		  << cp << "(residual) < " << Delta << "(Delta) * " << MaxResidSinceLastRelUp << "(MaxResidSinceLastRelUp) on iteration " << k << " : performing reliable update\n";
+	precisionChange(mmp, psi_f);
+	psi = psi + mmp;
 
-      //Single prec initialization
-      FieldF r_f(SinglePrecGrid);
-      r_f.checkerboard = r.checkerboard;
-      precisionChange(r_f, r);
+	Linop_d.HermOpAndNorm(psi, mmp, d, qq);
+	r = src - mmp;
 
-      FieldF psi_f(r_f);
-      psi_f = zero;
+	psi_f = Zero();
+	precisionChange(r_f, r);
+	cp = norm2(r);
+	MaxResidSinceLastRelUp = cp;
 
-      FieldF p_f(r_f);
-      FieldF mmp_f(r_f);
-
-      RealD MaxResidSinceLastRelUp = cp; //initial residual    
-    
-      std::cout << GridLogIterative << std::setprecision(4)
-		<< "ConjugateGradient: k=0 residual " << cp << " target " << rsq << std::endl;
-
-      GridStopWatch LinalgTimer;
-      GridStopWatch MatrixTimer;
-      GridStopWatch SolverTimer;
-
-      SolverTimer.Start();
-      int k = 0;
-      int l = 0;
-    
-      for (k = 1; k <= MaxIterations; k++) {
-	c = cp;
-
-	MatrixTimer.Start();
-	Linop_f_use->HermOpAndNorm(p_f, mmp_f, d, qq);
-	MatrixTimer.Stop();
-
-	LinalgTimer.Start();
-
-	a = c / d;
-	b_pred = a * (a * qq - d) / c;
-
-	cp = axpy_norm(r_f, -a, mmp_f, r_f);
-	b = cp / c;
-
-	// Fuse these loops ; should be really easy
-	psi_f = a * p_f + psi_f;
-	//p_f = p_f * b + r_f;
-
-	LinalgTimer.Stop();
-
-	std::cout << GridLogIterative << "ConjugateGradientReliableUpdate: Iteration " << k
-		  << " residual " << cp << " target " << rsq << std::endl;
-	std::cout << GridLogDebug << "a = "<< a << " b_pred = "<< b_pred << "  b = "<< b << std::endl;
-	std::cout << GridLogDebug << "qq = "<< qq << " d = "<< d << "  c = "<< c << std::endl;
-
-	if(cp > MaxResidSinceLastRelUp){
-	  std::cout << GridLogIterative << "ConjugateGradientReliableUpdate: updating MaxResidSinceLastRelUp : " << MaxResidSinceLastRelUp << " -> " << cp << std::endl;
-	  MaxResidSinceLastRelUp = cp;
-	}
+	b = cp/c;
 	  
-	// Stopping condition
-	if (cp <= rsq) {
-	  //Although not written in the paper, I assume that I have to add on the final solution
-	  precisionChange(mmp, psi_f);
-	  psi = psi + mmp;
-	
-	
-	  SolverTimer.Stop();
-	  Linop_d.HermOpAndNorm(psi, mmp, d, qq);
-	  p = mmp - src;
-
-	  RealD srcnorm = sqrt(norm2(src));
-	  RealD resnorm = sqrt(norm2(p));
-	  RealD true_residual = resnorm / srcnorm;
-
-	  std::cout << GridLogMessage << "ConjugateGradientReliableUpdate Converged on iteration " << k << " after " << l << " reliable updates" << std::endl;
-	  std::cout << GridLogMessage << "\tComputed residual " << sqrt(cp / ssq)<<std::endl;
-	  std::cout << GridLogMessage << "\tTrue residual " << true_residual<<std::endl;
-	  std::cout << GridLogMessage << "\tTarget " << Tolerance << std::endl;
-
-	  std::cout << GridLogMessage << "Time breakdown "<<std::endl;
-	  std::cout << GridLogMessage << "\tElapsed    " << SolverTimer.Elapsed() <<std::endl;
-	  std::cout << GridLogMessage << "\tMatrix     " << MatrixTimer.Elapsed() <<std::endl;
-	  std::cout << GridLogMessage << "\tLinalg     " << LinalgTimer.Elapsed() <<std::endl;
-
-	  IterationsToComplete = k;	
-	  ReliableUpdatesPerformed = l;
+	std::cout << GridLogMessage << "ConjugateGradientReliableUpdate new residual " << cp << std::endl;
 	  
-	  if(DoFinalCleanup){
-	    //Do a final CG to cleanup
-	    std::cout << GridLogMessage << "ConjugateGradientReliableUpdate performing final cleanup.\n";
-	    ConjugateGradient<FieldD> CG(Tolerance,MaxIterations);
-	    CG.ErrorOnNoConverge = ErrorOnNoConverge;
-	    CG(Linop_d,src,psi);
-	    IterationsToCleanup = CG.IterationsToComplete;
-	  }
-	  else if (ErrorOnNoConverge) assert(true_residual / Tolerance < 10000.0);
-
-	  std::cout << GridLogMessage << "ConjugateGradientReliableUpdate complete.\n";
-	  return;
-	}
-	else if(cp < Delta * MaxResidSinceLastRelUp) { //reliable update
-	  std::cout << GridLogMessage << "ConjugateGradientReliableUpdate "
-		    << cp << "(residual) < " << Delta << "(Delta) * " << MaxResidSinceLastRelUp << "(MaxResidSinceLastRelUp) on iteration " << k << " : performing reliable update\n";
-	  precisionChange(mmp, psi_f);
-	  psi = psi + mmp;
-
-	  Linop_d.HermOpAndNorm(psi, mmp, d, qq);
-	  r = src - mmp;
-
-	  psi_f = zero;
-	  precisionChange(r_f, r);
-	  cp = norm2(r);
-	  MaxResidSinceLastRelUp = cp;
-
-	  b = cp/c;
-	  
-	  std::cout << GridLogMessage << "ConjugateGradientReliableUpdate new residual " << cp << std::endl;
-	  
-	  l = l+1;
-	}
-
-	p_f = p_f * b + r_f; //update search vector after reliable update appears to help convergence
-
-	if(!using_fallback && Linop_fallback != NULL && cp < fallback_transition_tol){
-	  std::cout << GridLogMessage << "ConjugateGradientReliableUpdate switching to fallback linear operator on iteration " << k << " at residual " << cp << std::endl;
-	  Linop_f_use = Linop_fallback;
-	  using_fallback = true;
-	}
-
-	
+	l = l+1;
       }
-      std::cout << GridLogMessage << "ConjugateGradientReliableUpdate did NOT converge"
-		<< std::endl;
+
+      p_f = p_f * b + r_f; //update search vector after reliable update appears to help convergence
+
+      if(!using_fallback && Linop_fallback != NULL && cp < fallback_transition_tol){
+	std::cout << GridLogMessage << "ConjugateGradientReliableUpdate switching to fallback linear operator on iteration " << k << " at residual " << cp << std::endl;
+	Linop_f_use = Linop_fallback;
+	using_fallback = true;
+      }
+
+	
+    }
+    std::cout << GridLogMessage << "ConjugateGradientReliableUpdate did NOT converge"
+	      << std::endl;
       
-      if (ErrorOnNoConverge) assert(0);
-      IterationsToComplete = k;
-      ReliableUpdatesPerformed = l;      
-    }    
-  };
-
-
+    if (ErrorOnNoConverge) assert(0);
+    IterationsToComplete = k;
+    ReliableUpdatesPerformed = l;      
+  }    
 };
 
 
+NAMESPACE_END(Grid);
+
+
 
 #endif
diff --git a/Grid/algorithms/iterative/ConjugateResidual.h b/Grid/algorithms/iterative/ConjugateResidual.h
index b6f99554..e0c3b69d 100644
--- a/Grid/algorithms/iterative/ConjugateResidual.h
+++ b/Grid/algorithms/iterative/ConjugateResidual.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,88 +24,90 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_CONJUGATE_RESIDUAL_H
 #define GRID_CONJUGATE_RESIDUAL_H
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-    /////////////////////////////////////////////////////////////
-    // Base classes for iterative processes based on operators
-    // single input vec, single output vec.
-    /////////////////////////////////////////////////////////////
+/////////////////////////////////////////////////////////////
+// Base classes for iterative processes based on operators
+// single input vec, single output vec.
+/////////////////////////////////////////////////////////////
 
-  template<class Field> 
-    class ConjugateResidual : public OperatorFunction<Field> {
-  public:                                                
-    RealD   Tolerance;
-    Integer MaxIterations;
-    int verbose;
+template<class Field> 
+class ConjugateResidual : public OperatorFunction<Field> {
+public:                                                
+  using OperatorFunction<Field>::operator();
 
-    ConjugateResidual(RealD tol,Integer maxit) : Tolerance(tol), MaxIterations(maxit) { 
-      verbose=0;
-    };
+  RealD   Tolerance;
+  Integer MaxIterations;
+  int verbose;
 
-    void operator() (LinearOperatorBase<Field> &Linop,const Field &src, Field &psi){
+  ConjugateResidual(RealD tol,Integer maxit) : Tolerance(tol), MaxIterations(maxit) { 
+    verbose=0;
+  };
 
-      RealD a, b, c, d;
-      RealD cp, ssq,rsq;
+  void operator() (LinearOperatorBase<Field> &Linop,const Field &src, Field &psi){
+
+    RealD a, b; // c, d;
+    RealD cp, ssq,rsq;
       
-      RealD rAr, rAAr, rArp;
-      RealD pAp, pAAp;
+    RealD rAr, rAAr, rArp;
+    RealD pAp, pAAp;
 
-      GridBase *grid = src._grid;
-      psi=zero;
-      Field r(grid),  p(grid), Ap(grid), Ar(grid);
+    GridBase *grid = src.Grid();
+    psi=Zero();
+    Field r(grid),  p(grid), Ap(grid), Ar(grid);
       
-      r=src;
-      p=src;
+    r=src;
+    p=src;
+
+    Linop.HermOpAndNorm(p,Ap,pAp,pAAp);
+    Linop.HermOpAndNorm(r,Ar,rAr,rAAr);
+
+    cp =norm2(r);
+    ssq=norm2(src);
+    rsq=Tolerance*Tolerance*ssq;
+
+    if (verbose) std::cout<<GridLogMessage<<"ConjugateResidual: iteration " <<0<<" residual "<<cp<< " target"<< rsq<<std::endl;
+
+    for(int k=1;k<MaxIterations;k++){
+
+      a = rAr/pAAp;
+
+      axpy(psi,a,p,psi);
+
+      cp = axpy_norm(r,-a,Ap,r);
+
+      rArp=rAr;
 
-      Linop.HermOpAndNorm(p,Ap,pAp,pAAp);
       Linop.HermOpAndNorm(r,Ar,rAr,rAAr);
 
-      cp =norm2(r);
-      ssq=norm2(src);
-      rsq=Tolerance*Tolerance*ssq;
-
-      if (verbose) std::cout<<GridLogMessage<<"ConjugateResidual: iteration " <<0<<" residual "<<cp<< " target"<< rsq<<std::endl;
-
-      for(int k=1;k<MaxIterations;k++){
-
-	a = rAr/pAAp;
-
-	axpy(psi,a,p,psi);
-
-	cp = axpy_norm(r,-a,Ap,r);
-
-	rArp=rAr;
-
-	Linop.HermOpAndNorm(r,Ar,rAr,rAAr);
-
-	b   =rAr/rArp;
+      b   =rAr/rArp;
  
-	axpy(p,b,p,r);
-	pAAp=axpy_norm(Ap,b,Ap,Ar);
+      axpy(p,b,p,r);
+      pAAp=axpy_norm(Ap,b,Ap,Ar);
 	
-	if(verbose) std::cout<<GridLogMessage<<"ConjugateResidual: iteration " <<k<<" residual "<<cp<< " target"<< rsq<<std::endl;
-
-	if(cp<rsq) {
-	  Linop.HermOp(psi,Ap);
-	  axpy(r,-1.0,src,Ap);
-	  RealD true_resid = norm2(r)/ssq;
-	  std::cout<<GridLogMessage<<"ConjugateResidual: Converged on iteration " <<k
-		   << " computed residual "<<sqrt(cp/ssq)
-	           << " true residual "<<sqrt(true_resid)
-	           << " target "       <<Tolerance <<std::endl;
-	  return;
-	}
+      if(verbose) std::cout<<GridLogMessage<<"ConjugateResidual: iteration " <<k<<" residual "<<cp<< " target"<< rsq<<std::endl;
 
+      if(cp<rsq) {
+	Linop.HermOp(psi,Ap);
+	axpy(r,-1.0,src,Ap);
+	RealD true_resid = norm2(r)/ssq;
+	std::cout<<GridLogMessage<<"ConjugateResidual: Converged on iteration " <<k
+		 << " computed residual "<<std::sqrt(cp/ssq)
+		 << " true residual "<<std::sqrt(true_resid)
+		 << " target "       <<Tolerance <<std::endl;
+	return;
       }
 
-      std::cout<<GridLogMessage<<"ConjugateResidual did NOT converge"<<std::endl;
-      assert(0);
     }
-  };
-}
+
+    std::cout<<GridLogMessage<<"ConjugateResidual did NOT converge"<<std::endl;
+    assert(0);
+  }
+};
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/algorithms/iterative/Deflation.h b/Grid/algorithms/iterative/Deflation.h
index 8df67b3c..626b60e3 100644
--- a/Grid/algorithms/iterative/Deflation.h
+++ b/Grid/algorithms/iterative/Deflation.h
@@ -33,7 +33,7 @@ namespace Grid {
 template<class Field>
 class ZeroGuesser: public LinearFunction<Field> {
 public:
-  virtual void operator()(const Field &src, Field &guess) { guess = zero; };
+    virtual void operator()(const Field &src, Field &guess) { guess = Zero(); };
 };
 template<class Field>
 class DoNothingGuesser: public LinearFunction<Field> {
@@ -60,14 +60,14 @@ public:
   DeflatedGuesser(const std::vector<Field> & _evec,const std::vector<RealD> & _eval) : evec(_evec), eval(_eval) {};
 
   virtual void operator()(const Field &src,Field &guess) {
-    guess = zero;
+    guess = Zero();
     assert(evec.size()==eval.size());
     auto N = evec.size();
     for (int i=0;i<N;i++) {
       const Field& tmp = evec[i];
       axpy(guess,TensorRemove(innerProduct(tmp,src)) / eval[i],tmp,guess);
     }
-    guess.checkerboard = src.checkerboard;
+    guess.Checkerboard() = src.Checkerboard();
   }
 };
 
@@ -90,15 +90,15 @@ public:
   
   void operator()(const FineField &src,FineField &guess) { 
     int N = (int)evec_coarse.size();
-    CoarseField src_coarse(evec_coarse[0]._grid);
-    CoarseField guess_coarse(evec_coarse[0]._grid);    guess_coarse = zero;
+    CoarseField src_coarse(evec_coarse[0].Grid());
+    CoarseField guess_coarse(evec_coarse[0].Grid());    guess_coarse = Zero();
     blockProject(src_coarse,src,subspace);    
     for (int i=0;i<N;i++) {
       const CoarseField & tmp = evec_coarse[i];
       axpy(guess_coarse,TensorRemove(innerProduct(tmp,src_coarse)) / eval_coarse[i],tmp,guess_coarse);
     }
     blockPromote(guess_coarse,guess,subspace);
-    guess.checkerboard = src.checkerboard;
+    guess.Checkerboard() = src.Checkerboard();
   };
 };
 
diff --git a/Grid/algorithms/iterative/FlexibleCommunicationAvoidingGeneralisedMinimalResidual.h b/Grid/algorithms/iterative/FlexibleCommunicationAvoidingGeneralisedMinimalResidual.h
index db857248..33bebcbb 100644
--- a/Grid/algorithms/iterative/FlexibleCommunicationAvoidingGeneralisedMinimalResidual.h
+++ b/Grid/algorithms/iterative/FlexibleCommunicationAvoidingGeneralisedMinimalResidual.h
@@ -34,6 +34,8 @@ namespace Grid {
 template<class Field>
 class FlexibleCommunicationAvoidingGeneralisedMinimalResidual : public OperatorFunction<Field> {
  public:
+  using OperatorFunction<Field>::operator();
+
   bool ErrorOnNoConverge; // Throw an assert when FCAGMRES fails to converge,
                           // defaults to true
 
@@ -53,10 +55,10 @@ class FlexibleCommunicationAvoidingGeneralisedMinimalResidual : public OperatorF
 
   Eigen::MatrixXcd H;
 
-  std::vector<std::complex<double>> y;
-  std::vector<std::complex<double>> gamma;
-  std::vector<std::complex<double>> c;
-  std::vector<std::complex<double>> s;
+  std::vector<ComplexD> y;
+  std::vector<ComplexD> gamma;
+  std::vector<ComplexD> c;
+  std::vector<ComplexD> s;
 
   LinearFunction<Field> &Preconditioner;
 
@@ -81,7 +83,7 @@ class FlexibleCommunicationAvoidingGeneralisedMinimalResidual : public OperatorF
 
     std::cout << GridLogWarning << "This algorithm currently doesn't differ from regular FGMRES" << std::endl;
 
-    psi.checkerboard = src.checkerboard;
+    psi.Checkerboard() = src.Checkerboard();
     conformable(psi, src);
 
     RealD guess = norm2(psi);
@@ -91,7 +93,7 @@ class FlexibleCommunicationAvoidingGeneralisedMinimalResidual : public OperatorF
     RealD ssq = norm2(src);
     RealD rsq = Tolerance * Tolerance * ssq;
 
-    Field r(src._grid);
+    Field r(src.Grid());
 
     std::cout << std::setprecision(4) << std::scientific;
     std::cout << GridLogIterative << "FlexibleCommunicationAvoidingGeneralisedMinimalResidual: guess " << guess << std::endl;
@@ -149,12 +151,12 @@ class FlexibleCommunicationAvoidingGeneralisedMinimalResidual : public OperatorF
 
     RealD cp = 0;
 
-    Field w(src._grid);
-    Field r(src._grid);
+    Field w(src.Grid());
+    Field r(src.Grid());
 
     // these should probably be made class members so that they are only allocated once, not in every restart
-    std::vector<Field> v(RestartLength + 1, src._grid); for (auto &elem : v) elem = zero;
-    std::vector<Field> z(RestartLength + 1, src._grid); for (auto &elem : z) elem = zero;
+    std::vector<Field> v(RestartLength + 1, src.Grid()); for (auto &elem : v) elem = Zero();
+    std::vector<Field> z(RestartLength + 1, src.Grid()); for (auto &elem : z) elem = Zero();
 
     MatrixTimer.Start();
     LinOp.Op(psi, w);
@@ -176,7 +178,7 @@ class FlexibleCommunicationAvoidingGeneralisedMinimalResidual : public OperatorF
 
       qrUpdate(i);
 
-      cp = std::norm(gamma[i+1]);
+      cp = norm(gamma[i+1]);
 
       std::cout << GridLogIterative << "FlexibleCommunicationAvoidingGeneralisedMinimalResidual: Iteration " << IterationCount
                 << " residual " << cp << " target " << rsq << std::endl;
@@ -206,11 +208,11 @@ class FlexibleCommunicationAvoidingGeneralisedMinimalResidual : public OperatorF
     LinalgTimer.Start();
     for (int i = 0; i <= iter; ++i) {
       H(iter, i) = innerProduct(v[i], w);
-      w = w - H(iter, i) * v[i];
+      w = w - ComplexD(H(iter, i)) * v[i];
     }
 
     H(iter, iter + 1) = sqrt(norm2(w));
-    v[iter + 1] = (1. / H(iter, iter + 1)) * w;
+    v[iter + 1] = ComplexD(1. / H(iter, iter + 1)) * w;
     LinalgTimer.Stop();
   }
 
@@ -218,13 +220,13 @@ class FlexibleCommunicationAvoidingGeneralisedMinimalResidual : public OperatorF
 
     QrTimer.Start();
     for (int i = 0; i < iter ; ++i) {
-      auto tmp       = -s[i] * H(iter, i) + c[i] * H(iter, i + 1);
-      H(iter, i)     = std::conj(c[i]) * H(iter, i) + std::conj(s[i]) * H(iter, i + 1);
+      auto tmp       = -s[i] * ComplexD(H(iter, i)) + c[i] * ComplexD(H(iter, i + 1));
+      H(iter, i)     = conjugate(c[i]) * ComplexD(H(iter, i)) + conjugate(s[i]) * ComplexD(H(iter, i + 1));
       H(iter, i + 1) = tmp;
     }
 
     // Compute new Givens Rotation
-    ComplexD nu = sqrt(std::norm(H(iter, iter)) + std::norm(H(iter, iter + 1)));
+    auto nu = sqrt(std::norm(H(iter, iter)) + std::norm(H(iter, iter + 1)));
     c[iter]     = H(iter, iter) / nu;
     s[iter]     = H(iter, iter + 1) / nu;
 
@@ -233,7 +235,7 @@ class FlexibleCommunicationAvoidingGeneralisedMinimalResidual : public OperatorF
     H(iter, iter + 1) = 0.;
 
     gamma[iter + 1] = -s[iter] * gamma[iter];
-    gamma[iter]     = std::conj(c[iter]) * gamma[iter];
+    gamma[iter]     = conjugate(c[iter]) * gamma[iter];
     QrTimer.Stop();
   }
 
@@ -243,8 +245,8 @@ class FlexibleCommunicationAvoidingGeneralisedMinimalResidual : public OperatorF
     for (int i = iter; i >= 0; i--) {
       y[i] = gamma[i];
       for (int k = i + 1; k <= iter; k++)
-        y[i] = y[i] - H(k, i) * y[k];
-      y[i] = y[i] / H(i, i);
+        y[i] = y[i] - ComplexD(H(k, i)) * y[k];
+      y[i] = y[i] / ComplexD(H(i, i));
     }
 
     for (int i = 0; i <= iter; i++)
diff --git a/Grid/algorithms/iterative/FlexibleGeneralisedMinimalResidual.h b/Grid/algorithms/iterative/FlexibleGeneralisedMinimalResidual.h
index efc8c787..cf108846 100644
--- a/Grid/algorithms/iterative/FlexibleGeneralisedMinimalResidual.h
+++ b/Grid/algorithms/iterative/FlexibleGeneralisedMinimalResidual.h
@@ -34,6 +34,8 @@ namespace Grid {
 template<class Field>
 class FlexibleGeneralisedMinimalResidual : public OperatorFunction<Field> {
  public:
+  using OperatorFunction<Field>::operator();
+
   bool ErrorOnNoConverge; // Throw an assert when FGMRES fails to converge,
                           // defaults to true
 
@@ -53,10 +55,10 @@ class FlexibleGeneralisedMinimalResidual : public OperatorFunction<Field> {
 
   Eigen::MatrixXcd H;
 
-  std::vector<std::complex<double>> y;
-  std::vector<std::complex<double>> gamma;
-  std::vector<std::complex<double>> c;
-  std::vector<std::complex<double>> s;
+  std::vector<ComplexD> y;
+  std::vector<ComplexD> gamma;
+  std::vector<ComplexD> c;
+  std::vector<ComplexD> s;
 
   LinearFunction<Field> &Preconditioner;
 
@@ -79,7 +81,7 @@ class FlexibleGeneralisedMinimalResidual : public OperatorFunction<Field> {
 
   void operator()(LinearOperatorBase<Field> &LinOp, const Field &src, Field &psi) {
 
-    psi.checkerboard = src.checkerboard;
+    psi.Checkerboard() = src.Checkerboard();
     conformable(psi, src);
 
     RealD guess = norm2(psi);
@@ -89,7 +91,7 @@ class FlexibleGeneralisedMinimalResidual : public OperatorFunction<Field> {
     RealD ssq = norm2(src);
     RealD rsq = Tolerance * Tolerance * ssq;
 
-    Field r(src._grid);
+    Field r(src.Grid());
 
     std::cout << std::setprecision(4) << std::scientific;
     std::cout << GridLogIterative << "FlexibleGeneralisedMinimalResidual: guess " << guess << std::endl;
@@ -147,12 +149,12 @@ class FlexibleGeneralisedMinimalResidual : public OperatorFunction<Field> {
 
     RealD cp = 0;
 
-    Field w(src._grid);
-    Field r(src._grid);
+    Field w(src.Grid());
+    Field r(src.Grid());
 
     // these should probably be made class members so that they are only allocated once, not in every restart
-    std::vector<Field> v(RestartLength + 1, src._grid); for (auto &elem : v) elem = zero;
-    std::vector<Field> z(RestartLength + 1, src._grid); for (auto &elem : z) elem = zero;
+    std::vector<Field> v(RestartLength + 1, src.Grid()); for (auto &elem : v) elem = Zero();
+    std::vector<Field> z(RestartLength + 1, src.Grid()); for (auto &elem : z) elem = Zero();
 
     MatrixTimer.Start();
     LinOp.Op(psi, w);
@@ -174,7 +176,7 @@ class FlexibleGeneralisedMinimalResidual : public OperatorFunction<Field> {
 
       qrUpdate(i);
 
-      cp = std::norm(gamma[i+1]);
+      cp = norm(gamma[i+1]);
 
       std::cout << GridLogIterative << "FlexibleGeneralisedMinimalResidual: Iteration " << IterationCount
                 << " residual " << cp << " target " << rsq << std::endl;
@@ -204,11 +206,11 @@ class FlexibleGeneralisedMinimalResidual : public OperatorFunction<Field> {
     LinalgTimer.Start();
     for (int i = 0; i <= iter; ++i) {
       H(iter, i) = innerProduct(v[i], w);
-      w = w - H(iter, i) * v[i];
+      w = w - ComplexD(H(iter, i)) * v[i];
     }
 
     H(iter, iter + 1) = sqrt(norm2(w));
-    v[iter + 1] = (1. / H(iter, iter + 1)) * w;
+    v[iter + 1] = ComplexD(1. / H(iter, iter + 1)) * w;
     LinalgTimer.Stop();
   }
 
@@ -216,13 +218,13 @@ class FlexibleGeneralisedMinimalResidual : public OperatorFunction<Field> {
 
     QrTimer.Start();
     for (int i = 0; i < iter ; ++i) {
-      auto tmp       = -s[i] * H(iter, i) + c[i] * H(iter, i + 1);
-      H(iter, i)     = std::conj(c[i]) * H(iter, i) + std::conj(s[i]) * H(iter, i + 1);
+      auto tmp       = -s[i] * ComplexD(H(iter, i)) + c[i] * ComplexD(H(iter, i + 1));
+      H(iter, i)     = conjugate(c[i]) * ComplexD(H(iter, i)) + conjugate(s[i]) * ComplexD(H(iter, i + 1));
       H(iter, i + 1) = tmp;
     }
 
     // Compute new Givens Rotation
-    ComplexD nu = sqrt(std::norm(H(iter, iter)) + std::norm(H(iter, iter + 1)));
+    auto nu = sqrt(std::norm(H(iter, iter)) + std::norm(H(iter, iter + 1)));
     c[iter]     = H(iter, iter) / nu;
     s[iter]     = H(iter, iter + 1) / nu;
 
@@ -231,7 +233,7 @@ class FlexibleGeneralisedMinimalResidual : public OperatorFunction<Field> {
     H(iter, iter + 1) = 0.;
 
     gamma[iter + 1] = -s[iter] * gamma[iter];
-    gamma[iter]     = std::conj(c[iter]) * gamma[iter];
+    gamma[iter]     = conjugate(c[iter]) * gamma[iter];
     QrTimer.Stop();
   }
 
@@ -241,8 +243,8 @@ class FlexibleGeneralisedMinimalResidual : public OperatorFunction<Field> {
     for (int i = iter; i >= 0; i--) {
       y[i] = gamma[i];
       for (int k = i + 1; k <= iter; k++)
-        y[i] = y[i] - H(k, i) * y[k];
-      y[i] = y[i] / H(i, i);
+        y[i] = y[i] - ComplexD(H(k, i)) * y[k];
+      y[i] = y[i] / ComplexD(H(i, i));
     }
 
     for (int i = 0; i <= iter; i++)
diff --git a/Grid/algorithms/iterative/GeneralisedMinimalResidual.h b/Grid/algorithms/iterative/GeneralisedMinimalResidual.h
index 10636234..0596e91e 100644
--- a/Grid/algorithms/iterative/GeneralisedMinimalResidual.h
+++ b/Grid/algorithms/iterative/GeneralisedMinimalResidual.h
@@ -34,6 +34,8 @@ namespace Grid {
 template<class Field>
 class GeneralisedMinimalResidual : public OperatorFunction<Field> {
  public:
+  using OperatorFunction<Field>::operator();
+
   bool ErrorOnNoConverge; // Throw an assert when GMRES fails to converge,
                           // defaults to true
 
@@ -52,10 +54,10 @@ class GeneralisedMinimalResidual : public OperatorFunction<Field> {
 
   Eigen::MatrixXcd H;
 
-  std::vector<std::complex<double>> y;
-  std::vector<std::complex<double>> gamma;
-  std::vector<std::complex<double>> c;
-  std::vector<std::complex<double>> s;
+  std::vector<ComplexD> y;
+  std::vector<ComplexD> gamma;
+  std::vector<ComplexD> c;
+  std::vector<ComplexD> s;
 
   GeneralisedMinimalResidual(RealD   tol,
                              Integer maxit,
@@ -74,7 +76,7 @@ class GeneralisedMinimalResidual : public OperatorFunction<Field> {
 
   void operator()(LinearOperatorBase<Field> &LinOp, const Field &src, Field &psi) {
 
-    psi.checkerboard = src.checkerboard;
+    psi.Checkerboard() = src.Checkerboard();
     conformable(psi, src);
 
     RealD guess = norm2(psi);
@@ -84,7 +86,7 @@ class GeneralisedMinimalResidual : public OperatorFunction<Field> {
     RealD ssq = norm2(src);
     RealD rsq = Tolerance * Tolerance * ssq;
 
-    Field r(src._grid);
+    Field r(src.Grid());
 
     std::cout << std::setprecision(4) << std::scientific;
     std::cout << GridLogIterative << "GeneralisedMinimalResidual: guess " << guess << std::endl;
@@ -140,11 +142,11 @@ class GeneralisedMinimalResidual : public OperatorFunction<Field> {
 
     RealD cp = 0;
 
-    Field w(src._grid);
-    Field r(src._grid);
+    Field w(src.Grid());
+    Field r(src.Grid());
 
     // this should probably be made a class member so that it is only allocated once, not in every restart
-    std::vector<Field> v(RestartLength + 1, src._grid); for (auto &elem : v) elem = zero;
+    std::vector<Field> v(RestartLength + 1, src.Grid()); for (auto &elem : v) elem = Zero();
 
     MatrixTimer.Start();
     LinOp.Op(psi, w);
@@ -166,7 +168,7 @@ class GeneralisedMinimalResidual : public OperatorFunction<Field> {
 
       qrUpdate(i);
 
-      cp = std::norm(gamma[i+1]);
+      cp = norm(gamma[i+1]);
 
       std::cout << GridLogIterative << "GeneralisedMinimalResidual: Iteration " << IterationCount
                 << " residual " << cp << " target " << rsq << std::endl;
@@ -192,11 +194,11 @@ class GeneralisedMinimalResidual : public OperatorFunction<Field> {
     LinalgTimer.Start();
     for (int i = 0; i <= iter; ++i) {
       H(iter, i) = innerProduct(v[i], w);
-      w = w - H(iter, i) * v[i];
+      w = w - ComplexD(H(iter, i)) * v[i];
     }
 
     H(iter, iter + 1) = sqrt(norm2(w));
-    v[iter + 1] = (1. / H(iter, iter + 1)) * w;
+    v[iter + 1] = ComplexD(1. / H(iter, iter + 1)) * w;
     LinalgTimer.Stop();
   }
 
@@ -204,13 +206,13 @@ class GeneralisedMinimalResidual : public OperatorFunction<Field> {
 
     QrTimer.Start();
     for (int i = 0; i < iter ; ++i) {
-      auto tmp       = -s[i] * H(iter, i) + c[i] * H(iter, i + 1);
-      H(iter, i)     = std::conj(c[i]) * H(iter, i) + std::conj(s[i]) * H(iter, i + 1);
+      auto tmp       = -s[i] * ComplexD(H(iter, i)) + c[i] * ComplexD(H(iter, i + 1));
+      H(iter, i)     = conjugate(c[i]) * ComplexD(H(iter, i)) + conjugate(s[i]) * ComplexD(H(iter, i + 1));
       H(iter, i + 1) = tmp;
     }
 
     // Compute new Givens Rotation
-    ComplexD nu = sqrt(std::norm(H(iter, iter)) + std::norm(H(iter, iter + 1)));
+    auto nu = sqrt(std::norm(H(iter, iter)) + std::norm(H(iter, iter + 1)));
     c[iter]     = H(iter, iter) / nu;
     s[iter]     = H(iter, iter + 1) / nu;
 
@@ -219,7 +221,7 @@ class GeneralisedMinimalResidual : public OperatorFunction<Field> {
     H(iter, iter + 1) = 0.;
 
     gamma[iter + 1] = -s[iter] * gamma[iter];
-    gamma[iter]     = std::conj(c[iter]) * gamma[iter];
+    gamma[iter]     = conjugate(c[iter]) * gamma[iter];
     QrTimer.Stop();
   }
 
@@ -229,8 +231,8 @@ class GeneralisedMinimalResidual : public OperatorFunction<Field> {
     for (int i = iter; i >= 0; i--) {
       y[i] = gamma[i];
       for (int k = i + 1; k <= iter; k++)
-        y[i] = y[i] - H(k, i) * y[k];
-      y[i] = y[i] / H(i, i);
+        y[i] = y[i] - ComplexD(H(k, i)) * y[k];
+      y[i] = y[i] / ComplexD(H(i, i));
     }
 
     for (int i = 0; i <= iter; i++)
diff --git a/Grid/algorithms/iterative/ImplicitlyRestartedLanczos.h b/Grid/algorithms/iterative/ImplicitlyRestartedLanczos.h
index 64f1e9a7..47dcee52 100644
--- a/Grid/algorithms/iterative/ImplicitlyRestartedLanczos.h
+++ b/Grid/algorithms/iterative/ImplicitlyRestartedLanczos.h
@@ -35,7 +35,7 @@ Author: Christoph Lehner <clehner@bnl.gov>
 //#include <zlib.h>
 #include <sys/stat.h>
 
-namespace Grid { 
+NAMESPACE_BEGIN(Grid); 
 
   ////////////////////////////////////////////////////////
   // Move following 100 LOC to lattice/Lattice_basis.h
@@ -52,26 +52,31 @@ void basisOrthogonalize(std::vector<Field> &basis,Field &w,int k)
 template<class Field>
 void basisRotate(std::vector<Field> &basis,Eigen::MatrixXd& Qt,int j0, int j1, int k0,int k1,int Nm) 
 {
+  typedef decltype(basis[0].View()) View;
+  auto tmp_v = basis[0].View();
+  std::vector<View> basis_v(basis.size(),tmp_v);
   typedef typename Field::vector_object vobj;
-  GridBase* grid = basis[0]._grid;
+  GridBase* grid = basis[0].Grid();
       
-  parallel_region
-  {
+  for(int k=0;k<basis.size();k++){
+    basis_v[k] = basis[k].View();
+  }
 
+  thread_region
+  {
     std::vector < vobj , commAllocator<vobj> > B(Nm); // Thread private
-       
-    parallel_for_internal(int ss=0;ss < grid->oSites();ss++){
+    thread_for_in_region(ss, grid->oSites(),{
       for(int j=j0; j<j1; ++j) B[j]=0.;
       
       for(int j=j0; j<j1; ++j){
 	for(int k=k0; k<k1; ++k){
-	  B[j] +=Qt(j,k) * basis[k]._odata[ss];
+	  B[j] +=Qt(j,k) * basis_v[k][ss];
 	}
       }
       for(int j=j0; j<j1; ++j){
-	  basis[j]._odata[ss] = B[j];
+	basis_v[j][ss] = B[j];
       }
-    }
+    });
   }
 }
 
@@ -80,16 +85,18 @@ template<class Field>
 void basisRotateJ(Field &result,std::vector<Field> &basis,Eigen::MatrixXd& Qt,int j, int k0,int k1,int Nm) 
 {
   typedef typename Field::vector_object vobj;
-  GridBase* grid = basis[0]._grid;
+  GridBase* grid = basis[0].Grid();
 
-  result.checkerboard = basis[0].checkerboard;
-  parallel_for(int ss=0;ss < grid->oSites();ss++){
-    vobj B = zero;
+  result.Checkerboard() = basis[0].Checkerboard();
+  auto result_v=result.View();
+  thread_for(ss, grid->oSites(),{
+    vobj B = Zero();
     for(int k=k0; k<k1; ++k){
-      B +=Qt(j,k) * basis[k]._odata[ss];
+      auto basis_k = basis[k].View();
+      B +=Qt(j,k) * basis_k[ss];
     }
-    result._odata[ss] = B;
-  }
+    result_v[ss] = B;
+  });
 }
 
 template<class Field>
@@ -119,7 +126,7 @@ void basisReorderInPlace(std::vector<Field> &_v,std::vector<RealD>& sort_vals, s
 
       assert(idx[i] > i);     assert(j!=idx.size());      assert(idx[j]==i);
 
-      std::swap(_v[i]._odata,_v[idx[i]]._odata); // should use vector move constructor, no data copy
+      swap(_v[i],_v[idx[i]]); // should use vector move constructor, no data copy
       std::swap(sort_vals[i],sort_vals[idx[i]]);
 
       idx[j] = idx[i];
@@ -150,6 +157,19 @@ void basisSortInPlace(std::vector<Field> & _v,std::vector<RealD>& sort_vals, boo
   basisReorderInPlace(_v,sort_vals,idx);
 }
 
+// PAB: faster to compute the inner products first then fuse loops.
+// If performance critical can improve.
+template<class Field>
+void basisDeflate(const std::vector<Field> &_v,const std::vector<RealD>& eval,const Field& src_orig,Field& result) {
+  result = Zero();
+  assert(_v.size()==eval.size());
+  int N = (int)_v.size();
+  for (int i=0;i<N;i++) {
+    Field& tmp = _v[i];
+    axpy(result,TensorRemove(innerProduct(tmp,src_orig)) / eval[i],tmp,result);
+  }
+}
+
 /////////////////////////////////////////////////////////////
 // Implicitly restarted lanczos
 /////////////////////////////////////////////////////////////
@@ -289,7 +309,7 @@ public:
   template<typename T>  static RealD normalise(T& v) 
   {
     RealD nn = norm2(v);
-    nn = sqrt(nn);
+    nn = std::sqrt(nn);
     v = v * (1.0/nn);
     return nn;
   }
@@ -321,8 +341,8 @@ until convergence
 */
   void calc(std::vector<RealD>& eval, std::vector<Field>& evec,  const Field& src, int& Nconv, bool reverse=false)
   {
-    GridBase *grid = src._grid;
-    assert(grid == evec[0]._grid);
+    GridBase *grid = src.Grid();
+    assert(grid == evec[0].Grid());
     
     GridLogIRL.TimingMode(1);
     std::cout << GridLogIRL <<"**************************************************************************"<< std::endl;
@@ -446,7 +466,7 @@ until convergence
       assert(k2<Nm);      assert(k2<Nm);      assert(k1>0);
 
       basisRotate(evec,Qt,k1-1,k2+1,0,Nm,Nm); /// big constraint on the basis
-      std::cout<<GridLogIRL <<"basisRotated  by Qt"<<std::endl;
+      std::cout<<GridLogIRL <<"basisRotated  by Qt *"<<k1-1<<","<<k2+1<<")"<<std::endl;
       
       ////////////////////////////////////////////////////
       // Compressed vector f and beta(k2)
@@ -454,7 +474,7 @@ until convergence
       f *= Qt(k2-1,Nm-1);
       f += lme[k2-1] * evec[k2];
       beta_k = norm2(f);
-      beta_k = sqrt(beta_k);
+      beta_k = std::sqrt(beta_k);
       std::cout<<GridLogIRL<<" beta(k) = "<<beta_k<<std::endl;
 	  
       RealD betar = 1.0/beta_k;
@@ -477,7 +497,7 @@ until convergence
 
 	std::cout << GridLogIRL << "Test convergence: rotate subset of vectors to test convergence " << std::endl;
 
-	Field B(grid); B.checkerboard = evec[0].checkerboard;
+	Field B(grid); B.Checkerboard() = evec[0].Checkerboard();
 
 	//  power of two search pattern;  not every evalue in eval2 is assessed.
 	int allconv =1;
@@ -515,7 +535,7 @@ until convergence
 	
   converged:
     {
-      Field B(grid); B.checkerboard = evec[0].checkerboard;
+      Field B(grid); B.Checkerboard() = evec[0].Checkerboard();
       basisRotate(evec,Qt,0,Nk,0,Nk,Nm);	    
       std::cout << GridLogIRL << " Rotated basis"<<std::endl;
       Nconv=0;
@@ -807,7 +827,7 @@ void diagonalize_QR(std::vector<RealD>& lmd, std::vector<RealD>& lme,
     
     // determination of 2x2 leading submatrix
     RealD dsub = lmd[kmax-1]-lmd[kmax-2];
-    RealD dd = sqrt(dsub*dsub + 4.0*lme[kmax-2]*lme[kmax-2]);
+    RealD dd = std::sqrt(dsub*dsub + 4.0*lme[kmax-2]*lme[kmax-2]);
     RealD Dsh = 0.5*(lmd[kmax-2]+lmd[kmax-1] +dd*(dsub/fabs(dsub)));
     // (Dsh: shift)
     
@@ -838,5 +858,6 @@ void diagonalize_QR(std::vector<RealD>& lmd, std::vector<RealD>& lme,
   abort();
 }
 };
-}
+
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/algorithms/iterative/LocalCoherenceLanczos.h b/Grid/algorithms/iterative/LocalCoherenceLanczos.h
index c771d1cb..9c945565 100644
--- a/Grid/algorithms/iterative/LocalCoherenceLanczos.h
+++ b/Grid/algorithms/iterative/LocalCoherenceLanczos.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,16 +24,15 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_LOCAL_COHERENCE_IRL_H
 #define GRID_LOCAL_COHERENCE_IRL_H
 
-namespace Grid { 
-
+NAMESPACE_BEGIN(Grid); 
 
 struct LanczosParams : Serializable {
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(LanczosParams,
 				  ChebyParams, Cheby,/*Chebyshev*/
 				  int, Nstop,    /*Vecs in Lanczos must converge Nstop < Nk < Nm*/
@@ -46,7 +45,7 @@ struct LanczosParams : Serializable {
 };
 
 struct LocalCoherenceLanczosParams : Serializable {
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(LocalCoherenceLanczosParams,
 				  bool, saveEvecs,
 				  bool, doFine,
@@ -59,7 +58,7 @@ struct LocalCoherenceLanczosParams : Serializable {
 				  RealD        , coarse_relax_tol,
 				  std::vector<int>, blockSize,
 				  std::string, config,
-				  std::vector < std::complex<double>  >, omega,
+				  std::vector < ComplexD  >, omega,
 				  RealD, mass,
 				  RealD, M5);
 };
@@ -83,14 +82,14 @@ public:
   };
 
   void operator()(const CoarseField& in, CoarseField& out) {
-    GridBase *FineGrid = subspace[0]._grid;    
-    int   checkerboard = subspace[0].checkerboard;
-      
-    FineField fin (FineGrid);     fin.checkerboard= checkerboard;
-    FineField fout(FineGrid);   fout.checkerboard = checkerboard;
+    GridBase *FineGrid = subspace[0].Grid();    
+    int   checkerboard = subspace[0].Checkerboard();
+
+    FineField fin (FineGrid);     fin.Checkerboard()= checkerboard;
+    FineField fout(FineGrid);   fout.Checkerboard() = checkerboard;
 
     blockPromote(in,fin,subspace);       std::cout<<GridLogIRL<<"ProjectedHermop : Promote to fine"<<std::endl;
-    _Linop.HermOp(fin,fout);             std::cout<<GridLogIRL<<"ProjectedHermop : HermOp (fine) "<<std::endl;
+    _Linop.HermOp(fin,fout);                   std::cout<<GridLogIRL<<"ProjectedHermop : HermOp (fine) "<<std::endl;
     blockProject(out,fout,subspace);     std::cout<<GridLogIRL<<"ProjectedHermop : Project to coarse "<<std::endl;
   }
 };
@@ -117,12 +116,12 @@ public:
   {  };
 
   void operator()(const CoarseField& in, CoarseField& out) {
-    
-    GridBase *FineGrid = subspace[0]._grid;    
-    int   checkerboard = subspace[0].checkerboard;
 
-    FineField fin (FineGrid); fin.checkerboard =checkerboard;
-    FineField fout(FineGrid);fout.checkerboard =checkerboard;
+    GridBase *FineGrid = subspace[0].Grid();    
+    int   checkerboard = subspace[0].Checkerboard();
+
+    FineField fin (FineGrid); fin.Checkerboard() =checkerboard;
+    FineField fout(FineGrid);fout.Checkerboard() =checkerboard;
     
     blockPromote(in,fin,subspace);             std::cout<<GridLogIRL<<"ProjectedFunctionHermop : Promote to fine"<<std::endl;
     _poly(_Linop,fin,fout);                    std::cout<<GridLogIRL<<"ProjectedFunctionHermop : Poly "<<std::endl;
@@ -133,7 +132,7 @@ public:
 template<class Fobj,class CComplex,int nbasis>
 class ImplicitlyRestartedLanczosSmoothedTester  : public ImplicitlyRestartedLanczosTester<Lattice<iVector<CComplex,nbasis > > >
 {
- public:
+public:
   typedef iVector<CComplex,nbasis >           CoarseSiteVector;
   typedef Lattice<CoarseSiteVector>           CoarseField;
   typedef Lattice<CComplex>   CoarseScalar; // used for inner products on fine field
@@ -142,7 +141,7 @@ class ImplicitlyRestartedLanczosSmoothedTester  : public ImplicitlyRestartedLanc
   LinearFunction<CoarseField> & _Poly;
   OperatorFunction<FineField>   & _smoother;
   LinearOperatorBase<FineField> &_Linop;
-  RealD                          _coarse_relax_tol;
+  RealD                             _coarse_relax_tol;
   std::vector<FineField>        &_subspace;
   
   ImplicitlyRestartedLanczosSmoothedTester(LinearFunction<CoarseField>   &Poly,
@@ -182,10 +181,10 @@ class ImplicitlyRestartedLanczosSmoothedTester  : public ImplicitlyRestartedLanc
   }
   int ReconstructEval(int j,RealD eresid,CoarseField &B, RealD &eval,RealD evalMaxApprox)
   {
-    GridBase *FineGrid = _subspace[0]._grid;    
-    int checkerboard   = _subspace[0].checkerboard;
-    FineField fB(FineGrid);fB.checkerboard =checkerboard;
-    FineField fv(FineGrid);fv.checkerboard =checkerboard;
+    GridBase *FineGrid = _subspace[0].Grid();    
+    int checkerboard   = _subspace[0].Checkerboard();
+    FineField fB(FineGrid);fB.Checkerboard() =checkerboard;
+    FineField fv(FineGrid);fv.Checkerboard() =checkerboard;
 
     blockPromote(B,fv,_subspace);  
     
@@ -305,11 +304,11 @@ public:
     int Nk = nbasis;
     subspace.resize(Nk,_FineGrid);
     subspace[0]=1.0;
-    subspace[0].checkerboard=_checkerboard;
+    subspace[0].Checkerboard()=_checkerboard;
     normalise(subspace[0]);
     PlainHermOp<FineField>    Op(_FineOp);
     for(int k=1;k<Nk;k++){
-      subspace[k].checkerboard=_checkerboard;
+      subspace[k].Checkerboard()=_checkerboard;
       Op(subspace[k-1],subspace[k]);
       normalise(subspace[k]);
     }
@@ -360,7 +359,11 @@ public:
 
     ImplicitlyRestartedLanczos<FineField> IRL(ChebyOp,Op,Nstop,Nk,Nm,resid,MaxIt,betastp,MinRes);
 
-    FineField src(_FineGrid); src=1.0; src.checkerboard = _checkerboard;
+    FineField src(_FineGrid); 
+    typedef typename FineField::scalar_type Scalar;
+    // src=1.0; 
+    src=Scalar(1.0); 
+    src.Checkerboard() = _checkerboard;
 
     int Nconv;
     IRL.calc(evals_fine,subspace,src,Nconv,false);
@@ -402,5 +405,5 @@ public:
   }
 };
 
-}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/algorithms/iterative/MinimalResidual.h b/Grid/algorithms/iterative/MinimalResidual.h
index fa1912cf..33b79ac2 100644
--- a/Grid/algorithms/iterative/MinimalResidual.h
+++ b/Grid/algorithms/iterative/MinimalResidual.h
@@ -33,6 +33,8 @@ namespace Grid {
 
 template<class Field> class MinimalResidual : public OperatorFunction<Field> {
  public:
+  using OperatorFunction<Field>::operator();
+
   bool ErrorOnNoConverge; // throw an assert when the MR fails to converge.
                           // Defaults true.
   RealD   Tolerance;
@@ -46,11 +48,11 @@ template<class Field> class MinimalResidual : public OperatorFunction<Field> {
 
   void operator()(LinearOperatorBase<Field> &Linop, const Field &src, Field &psi) {
 
-    psi.checkerboard = src.checkerboard;
+    psi.Checkerboard() = src.Checkerboard();
     conformable(psi, src);
 
-    Complex a, c;
-    Real    d;
+    ComplexD a, c;
+    RealD    d;
 
     Field Mr(src);
     Field r(src);
@@ -71,7 +73,6 @@ template<class Field> class MinimalResidual : public OperatorFunction<Field> {
     std::cout << std::setprecision(4) << std::scientific;
     std::cout << GridLogIterative << "MinimalResidual: guess " << guess << std::endl;
     std::cout << GridLogIterative << "MinimalResidual:   src " << ssq << std::endl;
-    std::cout << GridLogIterative << "MinimalResidual:    mp " << d << std::endl;
     std::cout << GridLogIterative << "MinimalResidual:  cp,r " << cp << std::endl;
 
     if (cp <= rsq) {
diff --git a/Grid/algorithms/iterative/MixedPrecisionFlexibleGeneralisedMinimalResidual.h b/Grid/algorithms/iterative/MixedPrecisionFlexibleGeneralisedMinimalResidual.h
index 04113684..d75fdb63 100644
--- a/Grid/algorithms/iterative/MixedPrecisionFlexibleGeneralisedMinimalResidual.h
+++ b/Grid/algorithms/iterative/MixedPrecisionFlexibleGeneralisedMinimalResidual.h
@@ -34,6 +34,9 @@ namespace Grid {
 template<class FieldD, class FieldF, typename std::enable_if<getPrecision<FieldD>::value == 2, int>::type = 0, typename std::enable_if< getPrecision<FieldF>::value == 1, int>::type = 0>
 class MixedPrecisionFlexibleGeneralisedMinimalResidual : public OperatorFunction<FieldD> {
  public:
+
+  using OperatorFunction<FieldD>::operator();
+
   bool ErrorOnNoConverge; // Throw an assert when MPFGMRES fails to converge,
                           // defaults to true
 
@@ -54,10 +57,10 @@ class MixedPrecisionFlexibleGeneralisedMinimalResidual : public OperatorFunction
 
   Eigen::MatrixXcd H;
 
-  std::vector<std::complex<double>> y;
-  std::vector<std::complex<double>> gamma;
-  std::vector<std::complex<double>> c;
-  std::vector<std::complex<double>> s;
+  std::vector<ComplexD> y;
+  std::vector<ComplexD> gamma;
+  std::vector<ComplexD> c;
+  std::vector<ComplexD> s;
 
   GridBase* SinglePrecGrid;
 
@@ -84,7 +87,7 @@ class MixedPrecisionFlexibleGeneralisedMinimalResidual : public OperatorFunction
 
   void operator()(LinearOperatorBase<FieldD> &LinOp, const FieldD &src, FieldD &psi) {
 
-    psi.checkerboard = src.checkerboard;
+    psi.Checkerboard() = src.Checkerboard();
     conformable(psi, src);
 
     RealD guess = norm2(psi);
@@ -94,7 +97,7 @@ class MixedPrecisionFlexibleGeneralisedMinimalResidual : public OperatorFunction
     RealD ssq = norm2(src);
     RealD rsq = Tolerance * Tolerance * ssq;
 
-    FieldD r(src._grid);
+    FieldD r(src.Grid());
 
     std::cout << std::setprecision(4) << std::scientific;
     std::cout << GridLogIterative << "MPFGMRES: guess " << guess << std::endl;
@@ -154,12 +157,12 @@ class MixedPrecisionFlexibleGeneralisedMinimalResidual : public OperatorFunction
 
     RealD cp = 0;
 
-    FieldD w(src._grid);
-    FieldD r(src._grid);
+    FieldD w(src.Grid());
+    FieldD r(src.Grid());
 
     // these should probably be made class members so that they are only allocated once, not in every restart
-    std::vector<FieldD> v(RestartLength + 1, src._grid); for (auto &elem : v) elem = zero;
-    std::vector<FieldD> z(RestartLength + 1, src._grid); for (auto &elem : z) elem = zero;
+    std::vector<FieldD> v(RestartLength + 1, src.Grid()); for (auto &elem : v) elem = Zero();
+    std::vector<FieldD> z(RestartLength + 1, src.Grid()); for (auto &elem : z) elem = Zero();
 
     MatrixTimer.Start();
     LinOp.Op(psi, w);
@@ -181,7 +184,7 @@ class MixedPrecisionFlexibleGeneralisedMinimalResidual : public OperatorFunction
 
       qrUpdate(i);
 
-      cp = std::norm(gamma[i+1]);
+      cp = norm(gamma[i+1]);
 
       std::cout << GridLogIterative << "MPFGMRES: Iteration " << IterationCount
                 << " residual " << cp << " target " << rsq << std::endl;
@@ -223,11 +226,11 @@ class MixedPrecisionFlexibleGeneralisedMinimalResidual : public OperatorFunction
     LinalgTimer.Start();
     for (int i = 0; i <= iter; ++i) {
       H(iter, i) = innerProduct(v[i], w);
-      w = w - H(iter, i) * v[i];
+      w = w - ComplexD(H(iter, i)) * v[i];
     }
 
     H(iter, iter + 1) = sqrt(norm2(w));
-    v[iter + 1] = (1. / H(iter, iter + 1)) * w;
+    v[iter + 1] = ComplexD(1. / H(iter, iter + 1)) * w;
     LinalgTimer.Stop();
   }
 
@@ -235,13 +238,13 @@ class MixedPrecisionFlexibleGeneralisedMinimalResidual : public OperatorFunction
 
     QrTimer.Start();
     for (int i = 0; i < iter ; ++i) {
-      auto tmp       = -s[i] * H(iter, i) + c[i] * H(iter, i + 1);
-      H(iter, i)     = std::conj(c[i]) * H(iter, i) + std::conj(s[i]) * H(iter, i + 1);
+      auto tmp       = -s[i] * ComplexD(H(iter, i)) + c[i] * ComplexD(H(iter, i + 1));
+      H(iter, i)     = conjugate(c[i]) * ComplexD(H(iter, i)) + conjugate(s[i]) * ComplexD(H(iter, i + 1));
       H(iter, i + 1) = tmp;
     }
 
     // Compute new Givens Rotation
-    ComplexD nu = sqrt(std::norm(H(iter, iter)) + std::norm(H(iter, iter + 1)));
+    auto nu = sqrt(std::norm(H(iter, iter)) + std::norm(H(iter, iter + 1)));
     c[iter]     = H(iter, iter) / nu;
     s[iter]     = H(iter, iter + 1) / nu;
 
@@ -250,7 +253,7 @@ class MixedPrecisionFlexibleGeneralisedMinimalResidual : public OperatorFunction
     H(iter, iter + 1) = 0.;
 
     gamma[iter + 1] = -s[iter] * gamma[iter];
-    gamma[iter]     = std::conj(c[iter]) * gamma[iter];
+    gamma[iter]     = conjugate(c[iter]) * gamma[iter];
     QrTimer.Stop();
   }
 
@@ -260,8 +263,8 @@ class MixedPrecisionFlexibleGeneralisedMinimalResidual : public OperatorFunction
     for (int i = iter; i >= 0; i--) {
       y[i] = gamma[i];
       for (int k = i + 1; k <= iter; k++)
-        y[i] = y[i] - H(k, i) * y[k];
-      y[i] = y[i] / H(i, i);
+        y[i] = y[i] - ComplexD(H(k, i)) * y[k];
+      y[i] = y[i] / ComplexD(H(i, i));
     }
 
     for (int i = 0; i <= iter; i++)
diff --git a/Grid/algorithms/iterative/NormalEquations.h b/Grid/algorithms/iterative/NormalEquations.h
index 55a84b4b..270b0115 100644
--- a/Grid/algorithms/iterative/NormalEquations.h
+++ b/Grid/algorithms/iterative/NormalEquations.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,38 +23,38 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_NORMAL_EQUATIONS_H
 #define GRID_NORMAL_EQUATIONS_H
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  ///////////////////////////////////////////////////////////////////////////////////////////////////////
-  // Take a matrix and form an NE solver calling a Herm solver
-  ///////////////////////////////////////////////////////////////////////////////////////////////////////
-  template<class Field> class NormalEquations : public OperatorFunction<Field>{
-  private:
-    SparseMatrixBase<Field> & _Matrix;
-    OperatorFunction<Field> & _HermitianSolver;
+///////////////////////////////////////////////////////////////////////////////////////////////////////
+// Take a matrix and form an NE solver calling a Herm solver
+///////////////////////////////////////////////////////////////////////////////////////////////////////
+template<class Field> class NormalEquations : public OperatorFunction<Field>{
+private:
+  SparseMatrixBase<Field> & _Matrix;
+  OperatorFunction<Field> & _HermitianSolver;
 
-  public:
+public:
 
-    /////////////////////////////////////////////////////
-    // Wrap the usual normal equations trick
-    /////////////////////////////////////////////////////
+  /////////////////////////////////////////////////////
+  // Wrap the usual normal equations trick
+  /////////////////////////////////////////////////////
   NormalEquations(SparseMatrixBase<Field> &Matrix, OperatorFunction<Field> &HermitianSolver) 
     :  _Matrix(Matrix), _HermitianSolver(HermitianSolver) {}; 
 
-    void operator() (const Field &in, Field &out){
+  void operator() (const Field &in, Field &out){
  
-      Field src(in._grid);
+    Field src(in.Grid());
 
-      _Matrix.Mdag(in,src);
-      _HermitianSolver(src,out);  // Mdag M out = Mdag in
+    _Matrix.Mdag(in,src);
+    _HermitianSolver(src,out);  // Mdag M out = Mdag in
  
-    }     
-  };
+  }     
+};
 
-}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/algorithms/iterative/PowerMethod.h b/Grid/algorithms/iterative/PowerMethod.h
index e85f258c..8a238ea6 100644
--- a/Grid/algorithms/iterative/PowerMethod.h
+++ b/Grid/algorithms/iterative/PowerMethod.h
@@ -14,7 +14,7 @@ template<class Field> class PowerMethod
 
   RealD operator()(LinearOperatorBase<Field> &HermOp, const Field &src) 
   { 
-    GridBase *grid = src._grid; 
+    GridBase *grid = src.Grid(); 
     
     // quickly get an idea of the largest eigenvalue to more properly normalize the residuum 
     RealD evalMaxApprox = 0.0; 
diff --git a/Grid/algorithms/iterative/PrecConjugateResidual.h b/Grid/algorithms/iterative/PrecConjugateResidual.h
index a149d093..b6178833 100644
--- a/Grid/algorithms/iterative/PrecConjugateResidual.h
+++ b/Grid/algorithms/iterative/PrecConjugateResidual.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,97 +23,97 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_PREC_CONJUGATE_RESIDUAL_H
 #define GRID_PREC_CONJUGATE_RESIDUAL_H
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-    /////////////////////////////////////////////////////////////
-    // Base classes for iterative processes based on operators
-    // single input vec, single output vec.
-    /////////////////////////////////////////////////////////////
+/////////////////////////////////////////////////////////////
+// Base classes for iterative processes based on operators
+// single input vec, single output vec.
+/////////////////////////////////////////////////////////////
 
-  template<class Field> 
-    class PrecConjugateResidual : public OperatorFunction<Field> {
-  public:                                                
-    RealD   Tolerance;
-    Integer MaxIterations;
-    int verbose;
-    LinearFunction<Field> &Preconditioner;
+template<class Field> 
+class PrecConjugateResidual : public OperatorFunction<Field> {
+public:                                                
+  RealD   Tolerance;
+  Integer MaxIterations;
+  int verbose;
+  LinearFunction<Field> &Preconditioner;
 
-    PrecConjugateResidual(RealD tol,Integer maxit,LinearFunction<Field> &Prec) : Tolerance(tol), MaxIterations(maxit),      Preconditioner(Prec)
-    { 
-      verbose=1;
-    };
+  PrecConjugateResidual(RealD tol,Integer maxit,LinearFunction<Field> &Prec) : Tolerance(tol), MaxIterations(maxit),      Preconditioner(Prec)
+  { 
+    verbose=1;
+  };
 
-    void operator() (LinearOperatorBase<Field> &Linop,const Field &src, Field &psi){
+  void operator() (LinearOperatorBase<Field> &Linop,const Field &src, Field &psi){
 
-      RealD a, b, c, d;
-      RealD cp, ssq,rsq;
+    RealD a, b, c, d;
+    RealD cp, ssq,rsq;
       
-      RealD rAr, rAAr, rArp;
-      RealD pAp, pAAp;
+    RealD rAr, rAAr, rArp;
+    RealD pAp, pAAp;
 
-      GridBase *grid = src._grid;
-      Field r(grid),  p(grid), Ap(grid), Ar(grid), z(grid);
+    GridBase *grid = src.Grid();
+    Field r(grid),  p(grid), Ap(grid), Ar(grid), z(grid);
       
-      psi=zero;
-      r  = src;
-      Preconditioner(r,p);
+    psi=zero;
+    r  = src;
+    Preconditioner(r,p);
 
       
 
-      Linop.HermOpAndNorm(p,Ap,pAp,pAAp);
-      Ar=Ap;
-      rAr=pAp;
-      rAAr=pAAp;
+    Linop.HermOpAndNorm(p,Ap,pAp,pAAp);
+    Ar=Ap;
+    rAr=pAp;
+    rAAr=pAAp;
 
-      cp =norm2(r);
-      ssq=norm2(src);
-      rsq=Tolerance*Tolerance*ssq;
+    cp =norm2(r);
+    ssq=norm2(src);
+    rsq=Tolerance*Tolerance*ssq;
 
-      if (verbose) std::cout<<GridLogMessage<<"PrecConjugateResidual: iteration " <<0<<" residual "<<cp<< " target"<< rsq<<std::endl;
+    if (verbose) std::cout<<GridLogMessage<<"PrecConjugateResidual: iteration " <<0<<" residual "<<cp<< " target"<< rsq<<std::endl;
 
-      for(int k=0;k<MaxIterations;k++){
+    for(int k=0;k<MaxIterations;k++){
 
 
-	Preconditioner(Ap,z);
-	RealD rq= real(innerProduct(Ap,z)); 
+      Preconditioner(Ap,z);
+      RealD rq= real(innerProduct(Ap,z)); 
 
-	a = rAr/rq;
+      a = rAr/rq;
 
-   	axpy(psi,a,p,psi);
-   cp = axpy_norm(r,-a,z,r);
+      axpy(psi,a,p,psi);
+      cp = axpy_norm(r,-a,z,r);
 
-	rArp=rAr;
+      rArp=rAr;
 
-	Linop.HermOpAndNorm(r,Ar,rAr,rAAr);
+      Linop.HermOpAndNorm(r,Ar,rAr,rAAr);
 
-	b   =rAr/rArp;
+      b   =rAr/rArp;
  
-	axpy(p,b,p,r);
-	pAAp=axpy_norm(Ap,b,Ap,Ar);
+      axpy(p,b,p,r);
+      pAAp=axpy_norm(Ap,b,Ap,Ar);
 	
-	if(verbose) std::cout<<GridLogMessage<<"PrecConjugateResidual: iteration " <<k<<" residual "<<cp<< " target"<< rsq<<std::endl;
-
-	if(cp<rsq) {
-	  Linop.HermOp(psi,Ap);
-	  axpy(r,-1.0,src,Ap);
-	  RealD true_resid = norm2(r)/ssq;
-	  std::cout<<GridLogMessage<<"PrecConjugateResidual: Converged on iteration " <<k
-		   << " computed residual "<<sqrt(cp/ssq)
-	           << " true residual "<<sqrt(true_resid)
-	           << " target "       <<Tolerance <<std::endl;
-	  return;
-	}
+      if(verbose) std::cout<<GridLogMessage<<"PrecConjugateResidual: iteration " <<k<<" residual "<<cp<< " target"<< rsq<<std::endl;
 
+      if(cp<rsq) {
+	Linop.HermOp(psi,Ap);
+	axpy(r,-1.0,src,Ap);
+	RealD true_resid = norm2(r)/ssq;
+	std::cout<<GridLogMessage<<"PrecConjugateResidual: Converged on iteration " <<k
+		 << " computed residual "<<sqrt(cp/ssq)
+		 << " true residual "<<sqrt(true_resid)
+		 << " target "       <<Tolerance <<std::endl;
+	return;
       }
 
-      std::cout<<GridLogMessage<<"PrecConjugateResidual did NOT converge"<<std::endl;
-      assert(0);
     }
-  };
-}
+
+    std::cout<<GridLogMessage<<"PrecConjugateResidual did NOT converge"<<std::endl;
+    assert(0);
+  }
+};
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/algorithms/iterative/PrecGeneralisedConjugateResidual.h b/Grid/algorithms/iterative/PrecGeneralisedConjugateResidual.h
index c723c4a9..43fbc340 100644
--- a/Grid/algorithms/iterative/PrecGeneralisedConjugateResidual.h
+++ b/Grid/algorithms/iterative/PrecGeneralisedConjugateResidual.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,8 +24,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_PREC_GCR_H
 #define GRID_PREC_GCR_H
 
@@ -36,206 +36,208 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 //NB. Likely not original reference since they are focussing on a preconditioner variant.
 //    but VPGCR was nicely written up in their paper
 ///////////////////////////////////////////////////////////////////////////////////////////////////////
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  template<class Field>
-    class PrecGeneralisedConjugateResidual : public OperatorFunction<Field> {
-  public:                                                
-    RealD   Tolerance;
-    Integer MaxIterations;
-    int verbose;
-    int mmax;
-    int nstep;
-    int steps;
-    GridStopWatch PrecTimer;
-    GridStopWatch MatTimer;
-    GridStopWatch LinalgTimer;
+template<class Field>
+class PrecGeneralisedConjugateResidual : public OperatorFunction<Field> {
+public:                                                
+  using OperatorFunction<Field>::operator();
 
-    LinearFunction<Field> &Preconditioner;
+  RealD   Tolerance;
+  Integer MaxIterations;
+  int verbose;
+  int mmax;
+  int nstep;
+  int steps;
+  GridStopWatch PrecTimer;
+  GridStopWatch MatTimer;
+  GridStopWatch LinalgTimer;
 
-   PrecGeneralisedConjugateResidual(RealD tol,Integer maxit,LinearFunction<Field> &Prec,int _mmax,int _nstep) : 
-      Tolerance(tol), 
-      MaxIterations(maxit),
-      Preconditioner(Prec),
-      mmax(_mmax),
-      nstep(_nstep)
-    { 
-      verbose=1;
-    };
+  LinearFunction<Field> &Preconditioner;
 
-    void operator() (LinearOperatorBase<Field> &Linop,const Field &src, Field &psi){
+  PrecGeneralisedConjugateResidual(RealD tol,Integer maxit,LinearFunction<Field> &Prec,int _mmax,int _nstep) : 
+    Tolerance(tol), 
+    MaxIterations(maxit),
+    Preconditioner(Prec),
+    mmax(_mmax),
+    nstep(_nstep)
+  { 
+    verbose=1;
+  };
 
-      psi=zero;
-      RealD cp, ssq,rsq;
-      ssq=norm2(src);
-      rsq=Tolerance*Tolerance*ssq;
+  void operator() (LinearOperatorBase<Field> &Linop,const Field &src, Field &psi){
+
+    psi=Zero();
+    RealD cp, ssq,rsq;
+    ssq=norm2(src);
+    rsq=Tolerance*Tolerance*ssq;
       
-      Field r(src._grid);
+    Field r(src.Grid());
 
-        PrecTimer.Reset();
-         MatTimer.Reset();
-      LinalgTimer.Reset();
+    PrecTimer.Reset();
+    MatTimer.Reset();
+    LinalgTimer.Reset();
 
-      GridStopWatch SolverTimer;
-      SolverTimer.Start();
+    GridStopWatch SolverTimer;
+    SolverTimer.Start();
 
-      steps=0;
-      for(int k=0;k<MaxIterations;k++){
+    steps=0;
+    for(int k=0;k<MaxIterations;k++){
 
-	cp=GCRnStep(Linop,src,psi,rsq);
+      cp=GCRnStep(Linop,src,psi,rsq);
 
-	std::cout<<GridLogMessage<<"VPGCR("<<mmax<<","<<nstep<<") "<< steps <<" steps cp = "<<cp<<std::endl;
+      std::cout<<GridLogMessage<<"VPGCR("<<mmax<<","<<nstep<<") "<< steps <<" steps cp = "<<cp<<std::endl;
 
-	if(cp<rsq) {
+      if(cp<rsq) {
 
-	  SolverTimer.Stop();
+	SolverTimer.Stop();
 
-	  Linop.HermOp(psi,r);
-	  axpy(r,-1.0,src,r);
-	  RealD tr = norm2(r);
-	  std::cout<<GridLogMessage<<"PrecGeneralisedConjugateResidual: Converged on iteration " <<steps
-		   << " computed residual "<<sqrt(cp/ssq)
-	           << " true residual "    <<sqrt(tr/ssq)
-	           << " target "           <<Tolerance <<std::endl;
-
-	  std::cout<<GridLogMessage<<"VPGCR Time elapsed: Total  "<< SolverTimer.Elapsed() <<std::endl;
-	  std::cout<<GridLogMessage<<"VPGCR Time elapsed: Precon "<<   PrecTimer.Elapsed() <<std::endl;
-	  std::cout<<GridLogMessage<<"VPGCR Time elapsed: Matrix "<<    MatTimer.Elapsed() <<std::endl;
-	  std::cout<<GridLogMessage<<"VPGCR Time elapsed: Linalg "<< LinalgTimer.Elapsed() <<std::endl;
-	  return;
-	}
+	Linop.HermOp(psi,r);
+	axpy(r,-1.0,src,r);
+	RealD tr = norm2(r);
+	std::cout<<GridLogMessage<<"PrecGeneralisedConjugateResidual: Converged on iteration " <<steps
+		 << " computed residual "<<sqrt(cp/ssq)
+		 << " true residual "    <<sqrt(tr/ssq)
+		 << " target "           <<Tolerance <<std::endl;
 
+	std::cout<<GridLogMessage<<"VPGCR Time elapsed: Total  "<< SolverTimer.Elapsed() <<std::endl;
+	std::cout<<GridLogMessage<<"VPGCR Time elapsed: Precon "<<   PrecTimer.Elapsed() <<std::endl;
+	std::cout<<GridLogMessage<<"VPGCR Time elapsed: Matrix "<<    MatTimer.Elapsed() <<std::endl;
+	std::cout<<GridLogMessage<<"VPGCR Time elapsed: Linalg "<< LinalgTimer.Elapsed() <<std::endl;
+	return;
       }
-      std::cout<<GridLogMessage<<"Variable Preconditioned GCR did not converge"<<std::endl;
-      assert(0);
+
     }
+    std::cout<<GridLogMessage<<"Variable Preconditioned GCR did not converge"<<std::endl;
+    assert(0);
+  }
 
-    RealD GCRnStep(LinearOperatorBase<Field> &Linop,const Field &src, Field &psi,RealD rsq){
+  RealD GCRnStep(LinearOperatorBase<Field> &Linop,const Field &src, Field &psi,RealD rsq){
 
-      RealD cp;
-      RealD a, b, c, d;
-      RealD zAz, zAAz;
-      RealD rAq, rq;
+    RealD cp;
+    RealD a, b;
+    RealD zAz, zAAz;
+    RealD rq;
 
-      GridBase *grid = src._grid;
+    GridBase *grid = src.Grid();
 
-      Field r(grid);
-      Field z(grid);
-      Field tmp(grid);
-      Field ttmp(grid);
-      Field Az(grid);
+    Field r(grid);
+    Field z(grid);
+    Field tmp(grid);
+    Field ttmp(grid);
+    Field Az(grid);
 
-      ////////////////////////////////
-      // history for flexible orthog
-      ////////////////////////////////
-      std::vector<Field> q(mmax,grid);
-      std::vector<Field> p(mmax,grid);
-      std::vector<RealD> qq(mmax);
+    ////////////////////////////////
+    // history for flexible orthog
+    ////////////////////////////////
+    std::vector<Field> q(mmax,grid);
+    std::vector<Field> p(mmax,grid);
+    std::vector<RealD> qq(mmax);
       
-      //////////////////////////////////
-      // initial guess x0 is taken as nonzero.
-      // r0=src-A x0 = src
-      //////////////////////////////////
-      MatTimer.Start();
-      Linop.HermOpAndNorm(psi,Az,zAz,zAAz); 
-      MatTimer.Stop();
 
-      LinalgTimer.Start();
-      r=src-Az;
-      LinalgTimer.Stop();
+    //////////////////////////////////
+    // initial guess x0 is taken as nonzero.
+    // r0=src-A x0 = src
+    //////////////////////////////////
+    MatTimer.Start();
+    Linop.HermOpAndNorm(psi,Az,zAz,zAAz); 
+    MatTimer.Stop();
+    
+    LinalgTimer.Start();
+    r=src-Az;
+    LinalgTimer.Stop();
+    
+    /////////////////////
+    // p = Prec(r)
+    /////////////////////
+    PrecTimer.Start();
+    Preconditioner(r,z);
+    PrecTimer.Stop();
 
-      /////////////////////
-      // p = Prec(r)
-      /////////////////////
-      PrecTimer.Start();
-      Preconditioner(r,z);
-      PrecTimer.Stop();
+    MatTimer.Start();
+    Linop.HermOp(z,tmp); 
+    MatTimer.Stop();
 
-      MatTimer.Start();
-      Linop.HermOp(z,tmp); 
-      MatTimer.Stop();
+    LinalgTimer.Start();
+    ttmp=tmp;
+    tmp=tmp-r;
+    LinalgTimer.Stop();
 
-      LinalgTimer.Start();
-      ttmp=tmp;
-      tmp=tmp-r;
-      LinalgTimer.Stop();
-
-      /*
+    /*
       std::cout<<GridLogMessage<<r<<std::endl;
       std::cout<<GridLogMessage<<z<<std::endl;
       std::cout<<GridLogMessage<<ttmp<<std::endl;
       std::cout<<GridLogMessage<<tmp<<std::endl;
-      */
+    */
+
+    MatTimer.Start();
+    Linop.HermOpAndNorm(z,Az,zAz,zAAz); 
+    MatTimer.Stop();
+
+    LinalgTimer.Start();
+    //p[0],q[0],qq[0] 
+    p[0]= z;
+    q[0]= Az;
+    qq[0]= zAAz;
+    
+    cp =norm2(r);
+    LinalgTimer.Stop();
+
+    for(int k=0;k<nstep;k++){
+
+      steps++;
+
+      int kp     = k+1;
+      int peri_k = k %mmax;
+      int peri_kp= kp%mmax;
+
+      LinalgTimer.Start();
+      rq= real(innerProduct(r,q[peri_k])); // what if rAr not real?
+      a = rq/qq[peri_k];
+
+      axpy(psi,a,p[peri_k],psi);         
+
+      cp = axpy_norm(r,-a,q[peri_k],r);
+      LinalgTimer.Stop();
+
+      if((k==nstep-1)||(cp<rsq)){
+	return cp;
+      }
+
+      std::cout<<GridLogMessage<< " VPGCR_step["<<steps<<"]  resid " <<sqrt(cp/rsq)<<std::endl; 
+
+      PrecTimer.Start();
+      Preconditioner(r,z);// solve Az = r
+      PrecTimer.Stop();
 
       MatTimer.Start();
-      Linop.HermOpAndNorm(z,Az,zAz,zAAz); 
+      Linop.HermOpAndNorm(z,Az,zAz,zAAz);
+      Linop.HermOp(z,tmp);
       MatTimer.Stop();
 
       LinalgTimer.Start();
-      //p[0],q[0],qq[0] 
-      p[0]= z;
-      q[0]= Az;
-      qq[0]= zAAz;
+      tmp=tmp-r;
+      std::cout<<GridLogMessage<< " Preconditioner resid " <<sqrt(norm2(tmp)/norm2(r))<<std::endl; 
 
-      cp =norm2(r);
-      LinalgTimer.Stop();
+      q[peri_kp]=Az;
+      p[peri_kp]=z;
 
-      for(int k=0;k<nstep;k++){
+      int northog = ((kp)>(mmax-1))?(mmax-1):(kp);  // if more than mmax done, we orthog all mmax history.
+      for(int back=0;back<northog;back++){
 
-	steps++;
+	int peri_back=(k-back)%mmax;   	  assert((k-back)>=0);
 
-	int kp     = k+1;
-	int peri_k = k %mmax;
-	int peri_kp= kp%mmax;
+	b=-real(innerProduct(q[peri_back],Az))/qq[peri_back];
+	p[peri_kp]=p[peri_kp]+b*p[peri_back];
+	q[peri_kp]=q[peri_kp]+b*q[peri_back];
 
-        LinalgTimer.Start();
-	rq= real(innerProduct(r,q[peri_k])); // what if rAr not real?
-	a = rq/qq[peri_k];
-
-	axpy(psi,a,p[peri_k],psi);         
-
-	cp = axpy_norm(r,-a,q[peri_k],r);
-        LinalgTimer.Stop();
-
-	if((k==nstep-1)||(cp<rsq)){
-	  return cp;
-	}
-
-	std::cout<<GridLogMessage<< " VPGCR_step["<<steps<<"]  resid " <<sqrt(cp/rsq)<<std::endl; 
-
-	PrecTimer.Start();
-	Preconditioner(r,z);// solve Az = r
-	PrecTimer.Stop();
-
-	MatTimer.Start();
-	Linop.HermOpAndNorm(z,Az,zAz,zAAz);
-	Linop.HermOp(z,tmp);
-	MatTimer.Stop();
-
-        LinalgTimer.Start();
-        tmp=tmp-r;
-	std::cout<<GridLogMessage<< " Preconditioner resid " <<sqrt(norm2(tmp)/norm2(r))<<std::endl; 
-
-	q[peri_kp]=Az;
-	p[peri_kp]=z;
-
-	int northog = ((kp)>(mmax-1))?(mmax-1):(kp);  // if more than mmax done, we orthog all mmax history.
-	for(int back=0;back<northog;back++){
-
-	  int peri_back=(k-back)%mmax;   	  assert((k-back)>=0);
-
-	  b=-real(innerProduct(q[peri_back],Az))/qq[peri_back];
-	  p[peri_kp]=p[peri_kp]+b*p[peri_back];
-	  q[peri_kp]=q[peri_kp]+b*q[peri_back];
-
-	}
-	qq[peri_kp]=norm2(q[peri_kp]); // could use axpy_norm
-        LinalgTimer.Stop();
       }
-
-      assert(0); // never reached
-      return cp;
+      qq[peri_kp]=norm2(q[peri_kp]); // could use axpy_norm
+      LinalgTimer.Stop();
     }
-  };
-}
+    assert(0); // never reached
+    return cp;
+  }
+};
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/algorithms/iterative/SchurRedBlack.h b/Grid/algorithms/iterative/SchurRedBlack.h
index 9f63e8c0..dd8a14b6 100644
--- a/Grid/algorithms/iterative/SchurRedBlack.h
+++ b/Grid/algorithms/iterative/SchurRedBlack.h
@@ -297,9 +297,9 @@ namespace Grid {
       /////////////////////////////////////////////////////
       // src_o = (source_o - Moe MeeInv source_e)
       /////////////////////////////////////////////////////
-      _Matrix.MooeeInv(src_e,tmp);     assert(  tmp.checkerboard ==Even);
-      _Matrix.Meooe   (tmp,Mtmp);      assert( Mtmp.checkerboard ==Odd);     
-      tmp=src_o-Mtmp;                  assert(  tmp.checkerboard ==Odd);     
+      _Matrix.MooeeInv(src_e,tmp);     assert(  tmp.Checkerboard() ==Even);
+      _Matrix.Meooe   (tmp,Mtmp);      assert( Mtmp.Checkerboard() ==Odd);     
+      tmp=src_o-Mtmp;                  assert(  tmp.Checkerboard() ==Odd);     
 
       _Matrix.Mooee(tmp,src_o); // Extra factor of "m" in source from dumb choice of matrix norm.
     }
@@ -317,17 +317,17 @@ namespace Grid {
       ///////////////////////////////////////////////////
       // sol_e = M_ee^-1 * ( src_e - Meo sol_o )...
       ///////////////////////////////////////////////////
-      _Matrix.Meooe(sol_o,tmp);        assert(  tmp.checkerboard   ==Even);
-      src_e = src_e-tmp;               assert(  src_e.checkerboard ==Even);
-      _Matrix.MooeeInv(src_e,sol_e);   assert(  sol_e.checkerboard ==Even);
+      _Matrix.Meooe(sol_o,tmp);        assert(  tmp.Checkerboard()   ==Even);
+      src_e = src_e-tmp;               assert(  src_e.Checkerboard() ==Even);
+      _Matrix.MooeeInv(src_e,sol_e);   assert(  sol_e.Checkerboard() ==Even);
      
-      setCheckerboard(sol,sol_e); assert(  sol_e.checkerboard ==Even);
-      setCheckerboard(sol,sol_o); assert(  sol_o.checkerboard ==Odd );
+      setCheckerboard(sol,sol_e); assert(  sol_e.Checkerboard() ==Even);
+      setCheckerboard(sol,sol_o); assert(  sol_o.Checkerboard() ==Odd );
     }
     virtual void RedBlackSolve   (Matrix & _Matrix,const Field &src_o, Field &sol_o)
     {
       SchurStaggeredOperator<Matrix,Field> _HermOpEO(_Matrix);
-      this->_HermitianRBSolver(_HermOpEO,src_o,sol_o);  assert(sol_o.checkerboard==Odd);
+      this->_HermitianRBSolver(_HermOpEO,src_o,sol_o);  assert(sol_o.Checkerboard()==Odd);
     };
     virtual void RedBlackSolve   (Matrix & _Matrix,const std::vector<Field> &src_o,  std::vector<Field> &sol_o)
     {
@@ -366,13 +366,13 @@ namespace Grid {
       /////////////////////////////////////////////////////
       // src_o = Mdag * (source_o - Moe MeeInv source_e)
       /////////////////////////////////////////////////////
-      _Matrix.MooeeInv(src_e,tmp);     assert(  tmp.checkerboard ==Even);
-      _Matrix.Meooe   (tmp,Mtmp);      assert( Mtmp.checkerboard ==Odd);     
-      tmp=src_o-Mtmp;                  assert(  tmp.checkerboard ==Odd);     
+      _Matrix.MooeeInv(src_e,tmp);     assert(  tmp.Checkerboard() ==Even);
+      _Matrix.Meooe   (tmp,Mtmp);      assert( Mtmp.Checkerboard() ==Odd);     
+      tmp=src_o-Mtmp;                  assert(  tmp.Checkerboard() ==Odd);     
 
       // get the right MpcDag
       SchurDiagMooeeOperator<Matrix,Field> _HermOpEO(_Matrix);
-      _HermOpEO.MpcDag(tmp,src_o);     assert(src_o.checkerboard ==Odd);       
+      _HermOpEO.MpcDag(tmp,src_o);     assert(src_o.Checkerboard() ==Odd);       
 
     }
     virtual void RedBlackSolution(Matrix & _Matrix,const Field &sol_o, const Field &src_e,Field &sol)
@@ -386,17 +386,17 @@ namespace Grid {
       ///////////////////////////////////////////////////
       // sol_e = M_ee^-1 * ( src_e - Meo sol_o )...
       ///////////////////////////////////////////////////
-      _Matrix.Meooe(sol_o,tmp);          assert(  tmp.checkerboard   ==Even);
-      src_e_i = src_e-tmp;               assert(  src_e_i.checkerboard ==Even);
-      _Matrix.MooeeInv(src_e_i,sol_e);   assert(  sol_e.checkerboard ==Even);
+      _Matrix.Meooe(sol_o,tmp);          assert(  tmp.Checkerboard()   ==Even);
+      src_e_i = src_e-tmp;               assert(  src_e_i.Checkerboard() ==Even);
+      _Matrix.MooeeInv(src_e_i,sol_e);   assert(  sol_e.Checkerboard() ==Even);
      
-      setCheckerboard(sol,sol_e); assert(  sol_e.checkerboard ==Even);
-      setCheckerboard(sol,sol_o); assert(  sol_o.checkerboard ==Odd );
+      setCheckerboard(sol,sol_e); assert(  sol_e.Checkerboard() ==Even);
+      setCheckerboard(sol,sol_o); assert(  sol_o.Checkerboard() ==Odd );
     }
     virtual void RedBlackSolve   (Matrix & _Matrix,const Field &src_o, Field &sol_o)
     {
       SchurDiagMooeeOperator<Matrix,Field> _HermOpEO(_Matrix);
-      this->_HermitianRBSolver(_HermOpEO,src_o,sol_o);  assert(sol_o.checkerboard==Odd);
+      this->_HermitianRBSolver(_HermOpEO,src_o,sol_o);  assert(sol_o.Checkerboard()==Odd);
     };
     virtual void RedBlackSolve   (Matrix & _Matrix,const std::vector<Field> &src_o,  std::vector<Field> &sol_o)
     {
@@ -437,12 +437,12 @@ namespace Grid {
       /////////////////////////////////////////////////////
       // src_o = Mdag * (source_o - Moe MeeInv source_e)
       /////////////////////////////////////////////////////
-      _Matrix.MooeeInv(src_e,tmp);     assert(  tmp.checkerboard ==Even);
-      _Matrix.Meooe   (tmp,Mtmp);      assert( Mtmp.checkerboard ==Odd);     
-      tmp=src_o-Mtmp;                  assert(  tmp.checkerboard ==Odd);     
+      _Matrix.MooeeInv(src_e,tmp);     assert(  tmp.Checkerboard() ==Even);
+      _Matrix.Meooe   (tmp,Mtmp);      assert( Mtmp.Checkerboard() ==Odd);     
+      tmp=src_o-Mtmp;                  assert(  tmp.Checkerboard() ==Odd);     
 
       // get the right MpcDag
-      _HermOpEO.MpcDag(tmp,src_o);     assert(src_o.checkerboard ==Odd);       
+      _HermOpEO.MpcDag(tmp,src_o);     assert(src_o.Checkerboard() ==Odd);       
     }
 
     virtual void RedBlackSolution(Matrix & _Matrix,const Field &sol_o, const Field &src_e,Field &sol)
@@ -463,12 +463,12 @@ namespace Grid {
       ///////////////////////////////////////////////////
       // sol_e = M_ee^-1 * ( src_e - Meo sol_o )...
       ///////////////////////////////////////////////////
-      _Matrix.Meooe(sol_o_i,tmp);    assert(  tmp.checkerboard   ==Even);
-      tmp = src_e-tmp;               assert(  src_e.checkerboard ==Even);
-      _Matrix.MooeeInv(tmp,sol_e);   assert(  sol_e.checkerboard ==Even);
+      _Matrix.Meooe(sol_o_i,tmp);    assert(  tmp.Checkerboard()   ==Even);
+      tmp = src_e-tmp;               assert(  src_e.Checkerboard() ==Even);
+      _Matrix.MooeeInv(tmp,sol_e);   assert(  sol_e.Checkerboard() ==Even);
      
-      setCheckerboard(sol,sol_e);    assert(  sol_e.checkerboard ==Even);
-      setCheckerboard(sol,sol_o_i);  assert(  sol_o_i.checkerboard ==Odd );
+      setCheckerboard(sol,sol_e);    assert(  sol_e.Checkerboard() ==Even);
+      setCheckerboard(sol,sol_o_i);  assert(  sol_o_i.Checkerboard() ==Odd );
     };
 
     virtual void RedBlackSolve   (Matrix & _Matrix,const Field &src_o, Field &sol_o)
diff --git a/Grid/allocator/AlignedAllocator.cc b/Grid/allocator/AlignedAllocator.cc
index 10b49f4b..ed27a8bd 100644
--- a/Grid/allocator/AlignedAllocator.cc
+++ b/Grid/allocator/AlignedAllocator.cc
@@ -1,11 +1,12 @@
 #include <Grid/GridCore.h>
 #include <fcntl.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 MemoryStats *MemoryProfiler::stats = nullptr;
 bool         MemoryProfiler::debug = false;
 
+#ifdef POINTER_CACHE
 int PointerCache::victim;
 
 PointerCache::PointerCacheEntry PointerCache::Entries[PointerCache::Ncache];
@@ -49,9 +50,9 @@ void *PointerCache::Insert(void *ptr,size_t bytes) {
 
 void *PointerCache::Lookup(size_t bytes) {
 
- if (bytes < 4096 ) return NULL;
+  if (bytes < 4096 ) return NULL;
 
-#ifdef _OPENMP
+#ifdef GRID_OMP
   assert(omp_in_parallel()==0);
 #endif 
 
@@ -63,7 +64,7 @@ void *PointerCache::Lookup(size_t bytes) {
   }
   return NULL;
 }
-
+#endif
 
 void check_huge_pages(void *Buf,uint64_t BYTES)
 {
@@ -90,7 +91,7 @@ void check_huge_pages(void *Buf,uint64_t BYTES)
       ++n4ktotal;
       if (pageaddr != baseaddr + j * page_size)
 	++nnothuge;
-      }
+    }
   }
   int rank = CartesianCommunicator::RankWorld();
   printf("rank %d Allocated %d 4k pages, %d not in huge pages\n", rank, n4ktotal, nnothuge);
@@ -106,20 +107,21 @@ std::string sizeString(const size_t bytes)
   double                 count = bytes;
   
   while (count >= 1024 && s < 7)
-  {
+    {
       s++;
       count /= 1024;
-  }
+    }
   if (count - floor(count) == 0.0)
-  {
+    {
       snprintf(buf, bufSize, "%d %sB", (int)count, suffixes[s]);
-  }
+    }
   else
-  {
+    {
       snprintf(buf, bufSize, "%.1f %sB", count, suffixes[s]);
-  }
+    }
   
   return std::string(buf);
 }
 
-}
+NAMESPACE_END(Grid);
+
diff --git a/Grid/allocator/AlignedAllocator.h b/Grid/allocator/AlignedAllocator.h
index b0f7e206..2aa7d82b 100644
--- a/Grid/allocator/AlignedAllocator.h
+++ b/Grid/allocator/AlignedAllocator.h
@@ -24,8 +24,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_ALIGNED_ALLOCATOR_H
 #define GRID_ALIGNED_ALLOCATOR_H
 
@@ -40,89 +40,95 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <mm_malloc.h>
 #endif
 
-namespace Grid {
+#define POINTER_CACHE
+#define GRID_ALLOC_ALIGN (2*1024*1024)
 
-  class PointerCache {
-  private:
+NAMESPACE_BEGIN(Grid);
 
-    static const int Ncache=8;
-    static int victim;
+// Move control to configure.ac and Config.h?
+#ifdef POINTER_CACHE
+class PointerCache {
+private:
 
-    typedef struct { 
-      void *address;
-      size_t bytes;
-      int valid;
-    } PointerCacheEntry;
+  static const int Ncache=8;
+  static int victim;
+
+  typedef struct { 
+    void *address;
+    size_t bytes;
+    int valid;
+  } PointerCacheEntry;
     
-    static PointerCacheEntry Entries[Ncache];
+  static PointerCacheEntry Entries[Ncache];
 
-  public:
+public:
 
 
-    static void *Insert(void *ptr,size_t bytes) ;
-    static void *Lookup(size_t bytes) ;
+  static void *Insert(void *ptr,size_t bytes) ;
+  static void *Lookup(size_t bytes) ;
 
-  };
-  
-  std::string sizeString(size_t bytes);
+};
+#endif  
 
-  struct MemoryStats
-  {
-    size_t totalAllocated{0}, maxAllocated{0}, 
-           currentlyAllocated{0}, totalFreed{0};
-  };
+std::string sizeString(size_t bytes);
+
+struct MemoryStats
+{
+  size_t totalAllocated{0}, maxAllocated{0}, 
+    currentlyAllocated{0}, totalFreed{0};
+};
     
-  class MemoryProfiler
-  {
-  public:
-    static MemoryStats *stats;
-    static bool        debug;
-  };
+class MemoryProfiler
+{
+public:
+  static MemoryStats *stats;
+  static bool        debug;
+};
 
-  #define memString(bytes) std::to_string(bytes) + " (" + sizeString(bytes) + ")"
-  #define profilerDebugPrint \
-  if (MemoryProfiler::stats)\
-  {\
-    auto s = MemoryProfiler::stats;\
-    std::cout << GridLogDebug << "[Memory debug] Stats " << MemoryProfiler::stats << std::endl;\
-    std::cout << GridLogDebug << "[Memory debug] total  : " << memString(s->totalAllocated) \
-              << std::endl;\
-    std::cout << GridLogDebug << "[Memory debug] max    : " << memString(s->maxAllocated) \
-              << std::endl;\
-    std::cout << GridLogDebug << "[Memory debug] current: " << memString(s->currentlyAllocated) \
-              << std::endl;\
-    std::cout << GridLogDebug << "[Memory debug] freed  : " << memString(s->totalFreed) \
-              << std::endl;\
-  }
+#define memString(bytes) std::to_string(bytes) + " (" + sizeString(bytes) + ")"
+#define profilerDebugPrint						\
+  if (MemoryProfiler::stats)						\
+    {									\
+      auto s = MemoryProfiler::stats;					\
+      std::cout << GridLogDebug << "[Memory debug] Stats " << MemoryProfiler::stats << std::endl; \
+      std::cout << GridLogDebug << "[Memory debug] total  : " << memString(s->totalAllocated) \
+		<< std::endl;						\
+      std::cout << GridLogDebug << "[Memory debug] max    : " << memString(s->maxAllocated) \
+		<< std::endl;						\
+      std::cout << GridLogDebug << "[Memory debug] current: " << memString(s->currentlyAllocated) \
+		<< std::endl;						\
+      std::cout << GridLogDebug << "[Memory debug] freed  : " << memString(s->totalFreed) \
+		<< std::endl;						\
+    }
 
-  #define profilerAllocate(bytes)\
-  if (MemoryProfiler::stats)\
-  {\
-    auto s = MemoryProfiler::stats;\
-    s->totalAllocated     += (bytes);\
-    s->currentlyAllocated += (bytes);\
-    s->maxAllocated        = std::max(s->maxAllocated, s->currentlyAllocated);\
-  }\
-  if (MemoryProfiler::debug)\
-  {\
-    std::cout << GridLogDebug << "[Memory debug] allocating " << memString(bytes) << std::endl;\
-    profilerDebugPrint;\
-  }
+#define profilerAllocate(bytes)						\
+  if (MemoryProfiler::stats)						\
+    {									\
+      auto s = MemoryProfiler::stats;					\
+      s->totalAllocated     += (bytes);					\
+      s->currentlyAllocated += (bytes);					\
+      s->maxAllocated        = std::max(s->maxAllocated, s->currentlyAllocated); \
+    }									\
+  if (MemoryProfiler::debug)						\
+    {									\
+      std::cout << GridLogDebug << "[Memory debug] allocating " << memString(bytes) << std::endl; \
+      profilerDebugPrint;						\
+    }
 
-  #define profilerFree(bytes)\
-  if (MemoryProfiler::stats)\
-  {\
-    auto s = MemoryProfiler::stats;\
-    s->totalFreed         += (bytes);\
-    s->currentlyAllocated -= (bytes);\
-  }\
-  if (MemoryProfiler::debug)\
-  {\
-    std::cout << GridLogDebug << "[Memory debug] freeing " << memString(bytes) << std::endl;\
-    profilerDebugPrint;\
-  }
+#define profilerFree(bytes)						\
+  if (MemoryProfiler::stats)						\
+    {									\
+      auto s = MemoryProfiler::stats;					\
+      s->totalFreed         += (bytes);					\
+      s->currentlyAllocated -= (bytes);					\
+    }									\
+  if (MemoryProfiler::debug)						\
+    {									\
+      std::cout << GridLogDebug << "[Memory debug] freeing " << memString(bytes) << std::endl; \
+      profilerDebugPrint;						\
+    }
 
-  void check_huge_pages(void *Buf,uint64_t BYTES);
+void check_huge_pages(void *Buf,uint64_t BYTES);
 
 ////////////////////////////////////////////////////////////////////
 // A lattice of something, but assume the something is SIMDized.
@@ -152,29 +158,45 @@ public:
     size_type bytes = __n*sizeof(_Tp);
     profilerAllocate(bytes);
 
-    _Tp *ptr = (_Tp *) PointerCache::Lookup(bytes);
-    //    if ( ptr != NULL ) 
-    //      std::cout << "alignedAllocator "<<__n << " cache hit "<< std::hex << ptr <<std::dec <<std::endl;
 
-    //////////////////
-    // Hack 2MB align; could make option probably doesn't need configurability
-    //////////////////
-//define GRID_ALLOC_ALIGN (128)
-#define GRID_ALLOC_ALIGN (2*1024*1024)
-#ifdef HAVE_MM_MALLOC_H
-    if ( ptr == (_Tp *) NULL ) ptr = (_Tp *) _mm_malloc(bytes,GRID_ALLOC_ALIGN);
+#ifdef POINTER_CACHE
+    _Tp *ptr = (_Tp *) PointerCache::Lookup(bytes);
 #else
-    if ( ptr == (_Tp *) NULL ) ptr = (_Tp *) memalign(GRID_ALLOC_ALIGN,bytes);
+    pointer ptr = nullptr;
 #endif
-    //    std::cout << "alignedAllocator " << std::hex << ptr <<std::dec <<std::endl;
-    // First touch optimise in threaded loop
-    uint8_t *cp = (uint8_t *)ptr;
-#ifdef GRID_OMP
-#pragma omp parallel for
-#endif
-    for(size_type n=0;n<bytes;n+=4096){
-      cp[n]=0;
+
+#ifdef GRID_NVCC
+    ////////////////////////////////////
+    // Unified (managed) memory
+    ////////////////////////////////////
+    if ( ptr == (_Tp *) NULL ) {
+      auto err = cudaMallocManaged((void **)&ptr,bytes);
+      if( err != cudaSuccess ) {
+	ptr = (_Tp *) NULL;
+	std::cerr << " cudaMallocManaged failed for " << bytes<<" bytes " <<cudaGetErrorString(err)<< std::endl;
+	assert(0);
+      }
     }
+    assert( ptr != (_Tp *)NULL);
+#else 
+    //////////////////////////////////////////////////////////////////////////////////////////
+    // 2MB align; could make option probably doesn't need configurability
+    //////////////////////////////////////////////////////////////////////////////////////////
+  #ifdef HAVE_MM_MALLOC_H
+    if ( ptr == (_Tp *) NULL ) ptr = (_Tp *) _mm_malloc(bytes,GRID_ALLOC_ALIGN);
+  #else
+    if ( ptr == (_Tp *) NULL ) ptr = (_Tp *) memalign(GRID_ALLOC_ALIGN,bytes);
+  #endif
+    assert( ptr != (_Tp *)NULL);
+
+    //////////////////////////////////////////////////
+    // First touch optimise in threaded loop 
+    //////////////////////////////////////////////////
+    uint64_t *cp = (uint64_t *)ptr;
+    thread_for(n,bytes/sizeof(uint64_t), { // need only one touch per page
+      cp[n]=0;
+    });
+#endif
     return ptr;
   }
 
@@ -183,133 +205,40 @@ public:
 
     profilerFree(bytes);
 
+#ifdef POINTER_CACHE
     pointer __freeme = (pointer)PointerCache::Insert((void *)__p,bytes);
+#else 
+    pointer __freeme = __p;
+#endif
 
-#ifdef HAVE_MM_MALLOC_H
+#ifdef GRID_NVCC
+    if ( __freeme ) cudaFree((void *)__freeme);
+#else 
+  #ifdef HAVE_MM_MALLOC_H
     if ( __freeme ) _mm_free((void *)__freeme); 
-#else
+  #else
     if ( __freeme ) free((void *)__freeme);
+  #endif
 #endif
   }
-  void construct(pointer __p, const _Tp& __val) { };
+
+  // FIXME: hack for the copy constructor, eventually it must be avoided
+  void construct(pointer __p, const _Tp& __val) { new((void *)__p) _Tp(__val); };
+  //void construct(pointer __p, const _Tp& __val) { };
   void construct(pointer __p) { };
   void destroy(pointer __p) { };
 };
 template<typename _Tp>  inline bool operator==(const alignedAllocator<_Tp>&, const alignedAllocator<_Tp>&){ return true; }
 template<typename _Tp>  inline bool operator!=(const alignedAllocator<_Tp>&, const alignedAllocator<_Tp>&){ return false; }
 
-//////////////////////////////////////////////////////////////////////////////////////////
-// MPI3 : comms must use shm region
-// SHMEM: comms must use symmetric heap
-//////////////////////////////////////////////////////////////////////////////////////////
-#ifdef GRID_COMMS_SHMEM
-extern "C" { 
-#include <mpp/shmem.h>
-extern void * shmem_align(size_t, size_t);
-extern void  shmem_free(void *);
-}
-#define PARANOID_SYMMETRIC_HEAP
-#endif
-
-template<typename _Tp>
-class commAllocator {
-public: 
-  typedef std::size_t     size_type;
-  typedef std::ptrdiff_t  difference_type;
-  typedef _Tp*       pointer;
-  typedef const _Tp* const_pointer;
-  typedef _Tp&       reference;
-  typedef const _Tp& const_reference;
-  typedef _Tp        value_type;
-
-  template<typename _Tp1>  struct rebind { typedef commAllocator<_Tp1> other; };
-  commAllocator() throw() { }
-  commAllocator(const commAllocator&) throw() { }
-  template<typename _Tp1> commAllocator(const commAllocator<_Tp1>&) throw() { }
-  ~commAllocator() throw() { }
-  pointer       address(reference __x)       const { return &__x; }
-  size_type  max_size() const throw() { return size_t(-1) / sizeof(_Tp); }
-
-#ifdef GRID_COMMS_SHMEM
-  pointer allocate(size_type __n, const void* _p= 0)
-  {
-    size_type bytes = __n*sizeof(_Tp);
-
-    profilerAllocate(bytes);
-#ifdef CRAY
-    _Tp *ptr = (_Tp *) shmem_align(bytes,64);
-#else
-    _Tp *ptr = (_Tp *) shmem_align(64,bytes);
-#endif
-#ifdef PARANOID_SYMMETRIC_HEAP
-    static void * bcast;
-    static long  psync[_SHMEM_REDUCE_SYNC_SIZE];
-
-    bcast = (void *) ptr;
-    shmem_broadcast32((void *)&bcast,(void *)&bcast,sizeof(void *)/4,0,0,0,shmem_n_pes(),psync);
-
-    if ( bcast != ptr ) {
-      std::printf("inconsistent alloc pe %d %lx %lx \n",shmem_my_pe(),bcast,ptr);std::fflush(stdout);
-      //      BACKTRACEFILE();
-      exit(0);
-    }
-    assert( bcast == (void *) ptr);
-#endif 
-    return ptr;
-  }
-  void deallocate(pointer __p, size_type __n) { 
-    size_type bytes = __n*sizeof(_Tp);
-
-    profilerFree(bytes);
-    shmem_free((void *)__p);
-  }
-#else
-  pointer allocate(size_type __n, const void* _p= 0) 
-  {
-    size_type bytes = __n*sizeof(_Tp);
-    
-    profilerAllocate(bytes);
-#ifdef HAVE_MM_MALLOC_H
-    _Tp * ptr = (_Tp *) _mm_malloc(bytes, GRID_ALLOC_ALIGN);
-#else
-    _Tp * ptr = (_Tp *) memalign(GRID_ALLOC_ALIGN, bytes);
-#endif
-    uint8_t *cp = (uint8_t *)ptr;
-    if ( ptr ) { 
-    // One touch per 4k page, static OMP loop to catch same loop order
-#ifdef GRID_OMP
-#pragma omp parallel for schedule(static)
-#endif
-      for(size_type n=0;n<bytes;n+=4096){
-	cp[n]=0;
-      }
-    }
-    return ptr;
-  }
-  void deallocate(pointer __p, size_type __n) {
-    size_type bytes = __n*sizeof(_Tp);
-
-    profilerFree(bytes);
-#ifdef HAVE_MM_MALLOC_H
-    _mm_free((void *)__p); 
-#else
-    free((void *)__p);
-#endif
-  }
-#endif
-  void construct(pointer __p, const _Tp& __val) { };
-  void construct(pointer __p) { };
-  void destroy(pointer __p) { };
-};
-template<typename _Tp>  inline bool operator==(const commAllocator<_Tp>&, const commAllocator<_Tp>&){ return true; }
-template<typename _Tp>  inline bool operator!=(const commAllocator<_Tp>&, const commAllocator<_Tp>&){ return false; }
-
 ////////////////////////////////////////////////////////////////////////////////
 // Template typedefs
 ////////////////////////////////////////////////////////////////////////////////
+template<class T> using commAllocator = alignedAllocator<T>;
 template<class T> using Vector     = std::vector<T,alignedAllocator<T> >;           
-template<class T> using commVector = std::vector<T,commAllocator<T> >;              
+template<class T> using commVector = std::vector<T,alignedAllocator<T> >;
 template<class T> using Matrix     = std::vector<std::vector<T,alignedAllocator<T> > >;
-    
-}; // namespace Grid
+
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/cartesian/Cartesian.h b/Grid/cartesian/Cartesian.h
index f3710a48..070cad95 100644
--- a/Grid/cartesian/Cartesian.h
+++ b/Grid/cartesian/Cartesian.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,8 +23,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_CARTESIAN_H
 #define GRID_CARTESIAN_H
 
diff --git a/Grid/cartesian/Cartesian_base.h b/Grid/cartesian/Cartesian_base.h
index 05a8a3da..76abe0ee 100644
--- a/Grid/cartesian/Cartesian_base.h
+++ b/Grid/cartesian/Cartesian_base.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,268 +25,266 @@
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_CARTESIAN_BASE_H
 #define GRID_CARTESIAN_BASE_H
 
+NAMESPACE_BEGIN(Grid);
 
-namespace Grid{
-
-  //////////////////////////////////////////////////////////////////////
-  // Commicator provides information on the processor grid
-  //////////////////////////////////////////////////////////////////////
-  //    unsigned long _ndimension;
-  //    std::vector<int> _processors; // processor grid
-  //    int              _processor;  // linear processor rank
-  //    std::vector<int> _processor_coor;  // linear processor rank
-  //////////////////////////////////////////////////////////////////////
-  class GridBase : public CartesianCommunicator , public GridThread {
+//////////////////////////////////////////////////////////////////////
+// Commicator provides information on the processor grid
+//////////////////////////////////////////////////////////////////////
+//    unsigned long _ndimension;
+//    Coordinate _processors; // processor grid
+//    int              _processor;  // linear processor rank
+//    Coordinate _processor_coor;  // linear processor rank
+//////////////////////////////////////////////////////////////////////
+class GridBase : public CartesianCommunicator , public GridThread {
 
 public:
-    int dummy;
-    // Give Lattice access
-    template<class object> friend class Lattice;
+  int dummy;
+  // Give Lattice access
+  template<class object> friend class Lattice;
 
-    GridBase(const std::vector<int> & processor_grid) : CartesianCommunicator(processor_grid) {};
-    GridBase(const std::vector<int> & processor_grid,
-	     const CartesianCommunicator &parent,
-	     int &split_rank) 
-      : CartesianCommunicator(processor_grid,parent,split_rank) {};
-    GridBase(const std::vector<int> & processor_grid,
-	     const CartesianCommunicator &parent) 
-      : CartesianCommunicator(processor_grid,parent,dummy) {};
+  GridBase(const Coordinate & processor_grid) : CartesianCommunicator(processor_grid) {}; 
 
-    virtual ~GridBase() = default;
+  GridBase(const Coordinate & processor_grid,
+	   const CartesianCommunicator &parent,
+	   int &split_rank) 
+    : CartesianCommunicator(processor_grid,parent,split_rank) {};
+
+  GridBase(const Coordinate & processor_grid,
+	   const CartesianCommunicator &parent) 
+    : CartesianCommunicator(processor_grid,parent,dummy) {};
+
+  virtual ~GridBase() = default;
 
 
-    // Physics Grid information.
-    std::vector<int> _simd_layout;// Which dimensions get relayed out over simd lanes.
-    std::vector<int> _fdimensions;// (full) Global dimensions of array prior to cb removal
-    std::vector<int> _gdimensions;// Global dimensions of array after cb removal
-    std::vector<int> _ldimensions;// local dimensions of array with processor images removed
-    std::vector<int> _rdimensions;// Reduced local dimensions with simd lane images and processor images removed 
-    std::vector<int> _ostride;    // Outer stride for each dimension
-    std::vector<int> _istride;    // Inner stride i.e. within simd lane
-    int _osites;                  // _isites*_osites = product(dimensions).
-    int _isites;
-    int _fsites;                  // _isites*_osites = product(dimensions).
-    int _gsites;
-    std::vector<int> _slice_block;// subslice information
-    std::vector<int> _slice_stride;
-    std::vector<int> _slice_nblock;
+  // Physics Grid information.
+  Coordinate _simd_layout;// Which dimensions get relayed out over simd lanes.
+  Coordinate _fdimensions;// (full) Global dimensions of array prior to cb removal
+  Coordinate _gdimensions;// Global dimensions of array after cb removal
+  Coordinate _ldimensions;// local dimensions of array with processor images removed
+  Coordinate _rdimensions;// Reduced local dimensions with simd lane images and processor images removed 
+  Coordinate _ostride;    // Outer stride for each dimension
+  Coordinate _istride;    // Inner stride i.e. within simd lane
+  int _osites;                  // _isites*_osites = product(dimensions).
+  int _isites;
+  int _fsites;                  // _isites*_osites = product(dimensions).
+  int _gsites;
+  Coordinate _slice_block;// subslice information
+  Coordinate _slice_stride;
+  Coordinate _slice_nblock;
 
-    std::vector<int> _lstart;     // local start of array in gcoors _processor_coor[d]*_ldimensions[d]
-    std::vector<int> _lend  ;     // local end of array in gcoors   _processor_coor[d]*_ldimensions[d]+_ldimensions_[d]-1
+  Coordinate _lstart;     // local start of array in gcoors _processor_coor[d]*_ldimensions[d]
+  Coordinate _lend  ;     // local end of array in gcoors   _processor_coor[d]*_ldimensions[d]+_ldimensions_[d]-1
 
     bool _isCheckerBoarded; 
 
 public:
 
-    ////////////////////////////////////////////////////////////////
-    // Checkerboarding interface is virtual and overridden by 
-    // GridCartesian / GridRedBlackCartesian
-    ////////////////////////////////////////////////////////////////
-    virtual int CheckerBoarded(int dim)=0;
-    virtual int CheckerBoard(const std::vector<int> &site)=0;
-    virtual int CheckerBoardDestination(int source_cb,int shift,int dim)=0;
-    virtual int CheckerBoardShift(int source_cb,int dim,int shift,int osite)=0;
-    virtual int CheckerBoardShiftForCB(int source_cb,int dim,int shift,int cb)=0;
-    virtual int CheckerBoardFromOindex (int Oindex)=0;
-    virtual int CheckerBoardFromOindexTable (int Oindex)=0;
+  ////////////////////////////////////////////////////////////////
+  // Checkerboarding interface is virtual and overridden by 
+  // GridCartesian / GridRedBlackCartesian
+  ////////////////////////////////////////////////////////////////
+  virtual int CheckerBoarded(int dim)=0;
+  virtual int CheckerBoard(const Coordinate &site)=0;
+  virtual int CheckerBoardDestination(int source_cb,int shift,int dim)=0;
+  virtual int CheckerBoardShift(int source_cb,int dim,int shift,int osite)=0;
+  virtual int CheckerBoardShiftForCB(int source_cb,int dim,int shift,int cb)=0;
+  virtual int CheckerBoardFromOindex (int Oindex)=0;
+  virtual int CheckerBoardFromOindexTable (int Oindex)=0;
 
-    //////////////////////////////////////////////////////////////////////////////////////////////
-    // Local layout calculations
-    //////////////////////////////////////////////////////////////////////////////////////////////
-    // These routines are key. Subdivide the linearised cartesian index into
-    //      "inner" index identifying which simd lane of object<vFcomplex> is associated with coord
-    //      "outer" index identifying which element of _odata in class "Lattice" is associated with coord.
-    //
-    // Compared to, say, Blitz++ we simply need to store BOTH an inner stride and an outer
-    // stride per dimension. The cost of evaluating the indexing information is doubled for an n-dimensional
-    // coordinate. Note, however, for data parallel operations the "inner" indexing cost is not paid and all
-    // lanes are operated upon simultaneously.
+  //////////////////////////////////////////////////////////////////////////////////////////////
+  // Local layout calculations
+  //////////////////////////////////////////////////////////////////////////////////////////////
+  // These routines are key. Subdivide the linearised cartesian index into
+  //      "inner" index identifying which simd lane of object<vFcomplex> is associated with coord
+  //      "outer" index identifying which element of _odata in class "Lattice" is associated with coord.
+  //
+  // Compared to, say, Blitz++ we simply need to store BOTH an inner stride and an outer
+  // stride per dimension. The cost of evaluating the indexing information is doubled for an n-dimensional
+  // coordinate. Note, however, for data parallel operations the "inner" indexing cost is not paid and all
+  // lanes are operated upon simultaneously.
   
-    virtual int oIndex(std::vector<int> &coor)
-    {
-        int idx=0;
-        // Works with either global or local coordinates
-        for(int d=0;d<_ndimension;d++) idx+=_ostride[d]*(coor[d]%_rdimensions[d]);
-        return idx;
-    }
-    virtual int iIndex(std::vector<int> &lcoor)
-    {
-        int idx=0;
-        for(int d=0;d<_ndimension;d++) idx+=_istride[d]*(lcoor[d]/_rdimensions[d]);
-        return idx;
-    }
-    inline int oIndexReduced(std::vector<int> &ocoor)
-    {
-      int idx=0; 
-      // ocoor is already reduced so can eliminate the modulo operation
-      // for fast indexing and inline the routine
-      for(int d=0;d<_ndimension;d++) idx+=_ostride[d]*ocoor[d];
-      return idx;
-    }
-    inline void oCoorFromOindex (std::vector<int>& coor,int Oindex){
-      Lexicographic::CoorFromIndex(coor,Oindex,_rdimensions);
-    }
+  virtual int oIndex(Coordinate &coor)
+  {
+    int idx=0;
+    // Works with either global or local coordinates
+    for(int d=0;d<_ndimension;d++) idx+=_ostride[d]*(coor[d]%_rdimensions[d]);
+    return idx;
+  }
+  virtual int iIndex(Coordinate &lcoor)
+  {
+    int idx=0;
+    for(int d=0;d<_ndimension;d++) idx+=_istride[d]*(lcoor[d]/_rdimensions[d]);
+    return idx;
+  }
+  inline int oIndexReduced(Coordinate &ocoor)
+  {
+    int idx=0; 
+    // ocoor is already reduced so can eliminate the modulo operation
+    // for fast indexing and inline the routine
+    for(int d=0;d<_ndimension;d++) idx+=_ostride[d]*ocoor[d];
+    return idx;
+  }
+  inline void oCoorFromOindex (Coordinate& coor,int Oindex){
+    Lexicographic::CoorFromIndex(coor,Oindex,_rdimensions);
+  }
 
-    inline void InOutCoorToLocalCoor (std::vector<int> &ocoor, std::vector<int> &icoor, std::vector<int> &lcoor) {
-      lcoor.resize(_ndimension);
-      for (int d = 0; d < _ndimension; d++)
-        lcoor[d] = ocoor[d] + _rdimensions[d] * icoor[d];
-    }
+  inline void InOutCoorToLocalCoor (Coordinate &ocoor, Coordinate &icoor, Coordinate &lcoor) {
+    lcoor.resize(_ndimension);
+    for (int d = 0; d < _ndimension; d++)
+      lcoor[d] = ocoor[d] + _rdimensions[d] * icoor[d];
+  }
 
-    //////////////////////////////////////////////////////////
-    // SIMD lane addressing
-    //////////////////////////////////////////////////////////
-    inline void iCoorFromIindex(std::vector<int> &coor,int lane)
-    {
-      Lexicographic::CoorFromIndex(coor,lane,_simd_layout);
-    }
+  //////////////////////////////////////////////////////////
+  // SIMD lane addressing
+  //////////////////////////////////////////////////////////
+  inline void iCoorFromIindex(Coordinate &coor,int lane)
+  {
+    Lexicographic::CoorFromIndex(coor,lane,_simd_layout);
+  }
 
-    inline int PermuteDim(int dimension){
-      return _simd_layout[dimension]>1;
-    }
-    inline int PermuteType(int dimension){
-      int permute_type=0;
-      //
-      // FIXME:
-      //
-      // Best way to encode this would be to present a mask 
-      // for which simd dimensions are rotated, and the rotation
-      // size. If there is only one simd dimension rotated, this is just 
-      // a permute. 
-      //
-      // Cases: PermuteType == 1,2,4,8
-      // Distance should be either 0,1,2..
-      //
-      if ( _simd_layout[dimension] > 2 ) { 
-        for(int d=0;d<_ndimension;d++){
-          if ( d != dimension ) assert ( (_simd_layout[d]==1)  );
-        }
-        permute_type = RotateBit; // How to specify distance; this is not just direction.
-        return permute_type;
-      }
-
-      for(int d=_ndimension-1;d>dimension;d--){
-        if (_simd_layout[d]>1 ) permute_type++;
+  inline int PermuteDim(int dimension){
+    return _simd_layout[dimension]>1;
+  }
+  inline int PermuteType(int dimension){
+    int permute_type=0;
+    //
+    // Best way to encode this would be to present a mask 
+    // for which simd dimensions are rotated, and the rotation
+    // size. If there is only one simd dimension rotated, this is just 
+    // a permute. 
+    //
+    // Cases: PermuteType == 1,2,4,8
+    // Distance should be either 0,1,2..
+    //
+    if ( _simd_layout[dimension] > 2 ) { 
+      for(int d=0;d<_ndimension;d++){
+	if ( d != dimension ) assert ( (_simd_layout[d]==1)  );
       }
+      permute_type = RotateBit; // How to specify distance; this is not just direction.
       return permute_type;
     }
-    ////////////////////////////////////////////////////////////////
-    // Array sizing queries
-    ////////////////////////////////////////////////////////////////
 
-    inline int iSites(void) const { return _isites; };
-    inline int Nsimd(void)  const { return _isites; };// Synonymous with iSites
-    inline int oSites(void) const { return _osites; };
-    inline int lSites(void) const { return _isites*_osites; }; 
-    inline int gSites(void) const { return _isites*_osites*_Nprocessors; }; 
-    inline int Nd    (void) const { return _ndimension;};
-
-    inline const std::vector<int> LocalStarts(void)             { return _lstart;    };
-    inline const std::vector<int> &FullDimensions(void)         { return _fdimensions;};
-    inline const std::vector<int> &GlobalDimensions(void)       { return _gdimensions;};
-    inline const std::vector<int> &LocalDimensions(void)        { return _ldimensions;};
-    inline const std::vector<int> &VirtualLocalDimensions(void) { return _ldimensions;};
-
-    ////////////////////////////////////////////////////////////////
-    // Utility to print the full decomposition details 
-    ////////////////////////////////////////////////////////////////
-
-    void show_decomposition(){
-      std::cout << GridLogMessage << "\tFull Dimensions    : " << _fdimensions << std::endl;
-      std::cout << GridLogMessage << "\tSIMD layout        : " << _simd_layout << std::endl;
-      std::cout << GridLogMessage << "\tGlobal Dimensions  : " << _gdimensions << std::endl;
-      std::cout << GridLogMessage << "\tLocal Dimensions   : " << _ldimensions << std::endl;
-      std::cout << GridLogMessage << "\tReduced Dimensions : " << _rdimensions << std::endl;
-      std::cout << GridLogMessage << "\tOuter strides      : " << _ostride << std::endl;
-      std::cout << GridLogMessage << "\tInner strides      : " << _istride << std::endl;
-      std::cout << GridLogMessage << "\tiSites             : " << _isites << std::endl;
-      std::cout << GridLogMessage << "\toSites             : " << _osites << std::endl;
-      std::cout << GridLogMessage << "\tlSites             : " << lSites() << std::endl;        
-      std::cout << GridLogMessage << "\tgSites             : " << gSites() << std::endl;
-      std::cout << GridLogMessage << "\tNd                 : " << _ndimension << std::endl;             
-    } 
-
-    ////////////////////////////////////////////////////////////////
-    // Global addressing
-    ////////////////////////////////////////////////////////////////
-    void GlobalIndexToGlobalCoor(int gidx,std::vector<int> &gcoor){
-      assert(gidx< gSites());
-      Lexicographic::CoorFromIndex(gcoor,gidx,_gdimensions);
+    for(int d=_ndimension-1;d>dimension;d--){
+      if (_simd_layout[d]>1 ) permute_type++;
     }
-    void LocalIndexToLocalCoor(int lidx,std::vector<int> &lcoor){
-      assert(lidx<lSites());
-      Lexicographic::CoorFromIndex(lcoor,lidx,_ldimensions);
+    return permute_type;
+  }
+  ////////////////////////////////////////////////////////////////
+  // Array sizing queries
+  ////////////////////////////////////////////////////////////////
+
+  inline int iSites(void) const { return _isites; };
+  inline int Nsimd(void)  const { return _isites; };// Synonymous with iSites
+  inline int oSites(void) const { return _osites; };
+  inline int lSites(void) const { return _isites*_osites; }; 
+  inline int gSites(void) const { return _isites*_osites*_Nprocessors; }; 
+  inline int Nd    (void) const { return _ndimension;};
+
+  inline const Coordinate LocalStarts(void)             { return _lstart;    };
+  inline const Coordinate &FullDimensions(void)         { return _fdimensions;};
+  inline const Coordinate &GlobalDimensions(void)       { return _gdimensions;};
+  inline const Coordinate &LocalDimensions(void)        { return _ldimensions;};
+  inline const Coordinate &VirtualLocalDimensions(void) { return _ldimensions;};
+
+  ////////////////////////////////////////////////////////////////
+  // Utility to print the full decomposition details 
+  ////////////////////////////////////////////////////////////////
+
+  void show_decomposition(){
+    std::cout << GridLogMessage << "\tFull Dimensions    : " << _fdimensions << std::endl;
+    std::cout << GridLogMessage << "\tSIMD layout        : " << _simd_layout << std::endl;
+    std::cout << GridLogMessage << "\tGlobal Dimensions  : " << _gdimensions << std::endl;
+    std::cout << GridLogMessage << "\tLocal Dimensions   : " << _ldimensions << std::endl;
+    std::cout << GridLogMessage << "\tReduced Dimensions : " << _rdimensions << std::endl;
+    std::cout << GridLogMessage << "\tOuter strides      : " << _ostride << std::endl;
+    std::cout << GridLogMessage << "\tInner strides      : " << _istride << std::endl;
+    std::cout << GridLogMessage << "\tiSites             : " << _isites << std::endl;
+    std::cout << GridLogMessage << "\toSites             : " << _osites << std::endl;
+    std::cout << GridLogMessage << "\tlSites             : " << lSites() << std::endl;        
+    std::cout << GridLogMessage << "\tgSites             : " << gSites() << std::endl;
+    std::cout << GridLogMessage << "\tNd                 : " << _ndimension << std::endl;             
+  } 
+
+  ////////////////////////////////////////////////////////////////
+  // Global addressing
+  ////////////////////////////////////////////////////////////////
+  void GlobalIndexToGlobalCoor(int gidx,Coordinate &gcoor){
+    assert(gidx< gSites());
+    Lexicographic::CoorFromIndex(gcoor,gidx,_gdimensions);
+  }
+  void LocalIndexToLocalCoor(int lidx,Coordinate &lcoor){
+    assert(lidx<lSites());
+    Lexicographic::CoorFromIndex(lcoor,lidx,_ldimensions);
+  }
+  void GlobalCoorToGlobalIndex(const Coordinate & gcoor,int & gidx){
+    gidx=0;
+    int mult=1;
+    for(int mu=0;mu<_ndimension;mu++) {
+      gidx+=mult*gcoor[mu];
+      mult*=_gdimensions[mu];
     }
-    void GlobalCoorToGlobalIndex(const std::vector<int> & gcoor,int & gidx){
-      gidx=0;
-      int mult=1;
-      for(int mu=0;mu<_ndimension;mu++) {
-        gidx+=mult*gcoor[mu];
-        mult*=_gdimensions[mu];
-      }
+  }
+  void GlobalCoorToProcessorCoorLocalCoor(Coordinate &pcoor,Coordinate &lcoor,const Coordinate &gcoor)
+  {
+    pcoor.resize(_ndimension);
+    lcoor.resize(_ndimension);
+    for(int mu=0;mu<_ndimension;mu++){
+      int _fld  = _fdimensions[mu]/_processors[mu];
+      pcoor[mu] = gcoor[mu]/_fld;
+      lcoor[mu] = gcoor[mu]%_fld;
     }
-    void GlobalCoorToProcessorCoorLocalCoor(std::vector<int> &pcoor,std::vector<int> &lcoor,const std::vector<int> &gcoor)
-    {
-      pcoor.resize(_ndimension);
-      lcoor.resize(_ndimension);
-      for(int mu=0;mu<_ndimension;mu++){
-        int _fld  = _fdimensions[mu]/_processors[mu];
-        pcoor[mu] = gcoor[mu]/_fld;
-        lcoor[mu] = gcoor[mu]%_fld;
-      }
-    }
-    void GlobalCoorToRankIndex(int &rank, int &o_idx, int &i_idx ,const std::vector<int> &gcoor)
-    {
-      std::vector<int> pcoor;
-      std::vector<int> lcoor;
-      GlobalCoorToProcessorCoorLocalCoor(pcoor,lcoor,gcoor);
-      rank = RankFromProcessorCoor(pcoor);
-      /*
-      std::vector<int> cblcoor(lcoor);
+  }
+  void GlobalCoorToRankIndex(int &rank, int &o_idx, int &i_idx ,const Coordinate &gcoor)
+  {
+    Coordinate pcoor;
+    Coordinate lcoor;
+    GlobalCoorToProcessorCoorLocalCoor(pcoor,lcoor,gcoor);
+    rank = RankFromProcessorCoor(pcoor);
+    /*
+      Coordinate cblcoor(lcoor);
       for(int d=0;d<cblcoor.size();d++){
-        if( this->CheckerBoarded(d) ) {
-          cblcoor[d] = lcoor[d]/2;
-        }
+      if( this->CheckerBoarded(d) ) {
+      cblcoor[d] = lcoor[d]/2;
       }
-      */
-      i_idx= iIndex(lcoor);
-      o_idx= oIndex(lcoor);
-    }
+      }
+    */
+    i_idx= iIndex(lcoor);
+    o_idx= oIndex(lcoor);
+  }
 
-    void RankIndexToGlobalCoor(int rank, int o_idx, int i_idx , std::vector<int> &gcoor)
-    {
-      gcoor.resize(_ndimension);
-      std::vector<int> coor(_ndimension);
+  void RankIndexToGlobalCoor(int rank, int o_idx, int i_idx , Coordinate &gcoor)
+  {
+    gcoor.resize(_ndimension);
+    Coordinate coor(_ndimension);
 
-      ProcessorCoorFromRank(rank,coor);
-      for(int mu=0;mu<_ndimension;mu++) gcoor[mu] = _ldimensions[mu]*coor[mu];
+    ProcessorCoorFromRank(rank,coor);
+    for(int mu=0;mu<_ndimension;mu++) gcoor[mu] = _ldimensions[mu]*coor[mu];
 
-      iCoorFromIindex(coor,i_idx);
-      for(int mu=0;mu<_ndimension;mu++) gcoor[mu] += _rdimensions[mu]*coor[mu];
+    iCoorFromIindex(coor,i_idx);
+    for(int mu=0;mu<_ndimension;mu++) gcoor[mu] += _rdimensions[mu]*coor[mu];
 
-      oCoorFromOindex (coor,o_idx);
-      for(int mu=0;mu<_ndimension;mu++) gcoor[mu] += coor[mu];
+    oCoorFromOindex (coor,o_idx);
+    for(int mu=0;mu<_ndimension;mu++) gcoor[mu] += coor[mu];
       
+  }
+  void RankIndexCbToFullGlobalCoor(int rank, int o_idx, int i_idx, int cb,Coordinate &fcoor)
+  {
+    RankIndexToGlobalCoor(rank,o_idx,i_idx ,fcoor);
+    if(CheckerBoarded(0)){
+      fcoor[0] = fcoor[0]*2+cb;
     }
-    void RankIndexCbToFullGlobalCoor(int rank, int o_idx, int i_idx, int cb,std::vector<int> &fcoor)
-    {
-      RankIndexToGlobalCoor(rank,o_idx,i_idx ,fcoor);
-      if(CheckerBoarded(0)){
-        fcoor[0] = fcoor[0]*2+cb;
-      }
-    }
-    void ProcessorCoorLocalCoorToGlobalCoor(std::vector<int> &Pcoor,std::vector<int> &Lcoor,std::vector<int> &gcoor)
-    {
-      gcoor.resize(_ndimension);
-      for(int mu=0;mu<_ndimension;mu++) gcoor[mu] = Pcoor[mu]*_ldimensions[mu]+Lcoor[mu];
-    }
+  }
+  void ProcessorCoorLocalCoorToGlobalCoor(Coordinate &Pcoor,Coordinate &Lcoor,Coordinate &gcoor)
+  {
+    gcoor.resize(_ndimension);
+    for(int mu=0;mu<_ndimension;mu++) gcoor[mu] = Pcoor[mu]*_ldimensions[mu]+Lcoor[mu];
+  }
 };
 
-
-}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/cartesian/Cartesian_full.h b/Grid/cartesian/Cartesian_full.h
index b6297d3d..c083817b 100644
--- a/Grid/cartesian/Cartesian_full.h
+++ b/Grid/cartesian/Cartesian_full.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,97 +23,96 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_CARTESIAN_FULL_H
 #define GRID_CARTESIAN_FULL_H
 
-namespace Grid{
+NAMESPACE_BEGIN(Grid);
     
 /////////////////////////////////////////////////////////////////////////////////////////
 // Grid Support.
 /////////////////////////////////////////////////////////////////////////////////////////
 
-
 class GridCartesian: public GridBase {
 
 public:
-    int dummy;
-    virtual int  CheckerBoardFromOindexTable (int Oindex) {
-      return 0;
-    }
-    virtual int  CheckerBoardFromOindex (int Oindex)
-    {
-      return 0;
-    }
-    virtual int CheckerBoarded(int dim){
-      return 0;
-    }
-    virtual int CheckerBoard(const std::vector<int> &site){
-        return 0;
-    }
-    virtual int CheckerBoardDestination(int cb,int shift,int dim){
-        return 0;
-    }
-    virtual int CheckerBoardShiftForCB(int source_cb,int dim,int shift, int ocb){
-      return shift;
-    }
-    virtual int CheckerBoardShift(int source_cb,int dim,int shift, int osite){
-      return shift;
-    }
-    /////////////////////////////////////////////////////////////////////////
-    // Constructor takes a parent grid and possibly subdivides communicator.
-    /////////////////////////////////////////////////////////////////////////
-    GridCartesian(const std::vector<int> &dimensions,
-		  const std::vector<int> &simd_layout,
-		  const std::vector<int> &processor_grid,
-		  const GridCartesian &parent) : GridBase(processor_grid,parent,dummy)
-    {
-      Init(dimensions,simd_layout,processor_grid);
-    }
-    GridCartesian(const std::vector<int> &dimensions,
-		  const std::vector<int> &simd_layout,
-		  const std::vector<int> &processor_grid,
-		  const GridCartesian &parent,int &split_rank) : GridBase(processor_grid,parent,split_rank)
-    {
-      Init(dimensions,simd_layout,processor_grid);
-    }
-    /////////////////////////////////////////////////////////////////////////
-    // Construct from comm world
-    /////////////////////////////////////////////////////////////////////////
-    GridCartesian(const std::vector<int> &dimensions,
-		  const std::vector<int> &simd_layout,
-		  const std::vector<int> &processor_grid) : GridBase(processor_grid)
-    {
-      Init(dimensions,simd_layout,processor_grid);
-    }
+  int dummy;
+  virtual int  CheckerBoardFromOindexTable (int Oindex) {
+    return 0;
+  }
+  virtual int  CheckerBoardFromOindex (int Oindex)
+  {
+    return 0;
+  }
+  virtual int CheckerBoarded(int dim){
+    return 0;
+  }
+  virtual int CheckerBoard(const Coordinate &site){
+    return 0;
+  }
+  virtual int CheckerBoardDestination(int cb,int shift,int dim){
+    return 0;
+  }
+  virtual int CheckerBoardShiftForCB(int source_cb,int dim,int shift, int ocb){
+    return shift;
+  }
+  virtual int CheckerBoardShift(int source_cb,int dim,int shift, int osite){
+    return shift;
+  }
+  /////////////////////////////////////////////////////////////////////////
+  // Constructor takes a parent grid and possibly subdivides communicator.
+  /////////////////////////////////////////////////////////////////////////
+  GridCartesian(const Coordinate &dimensions,
+		const Coordinate &simd_layout,
+		const Coordinate &processor_grid,
+		const GridCartesian &parent) : GridBase(processor_grid,parent,dummy)
+  {
+    Init(dimensions,simd_layout,processor_grid);
+  }
+  GridCartesian(const Coordinate &dimensions,
+		const Coordinate &simd_layout,
+		const Coordinate &processor_grid,
+		const GridCartesian &parent,int &split_rank) : GridBase(processor_grid,parent,split_rank)
+  {
+    Init(dimensions,simd_layout,processor_grid);
+  }
+  /////////////////////////////////////////////////////////////////////////
+  // Construct from comm world
+  /////////////////////////////////////////////////////////////////////////
+  GridCartesian(const Coordinate &dimensions,
+		const Coordinate &simd_layout,
+		const Coordinate &processor_grid) : GridBase(processor_grid)
+  {
+    Init(dimensions,simd_layout,processor_grid);
+  }
 
-    virtual ~GridCartesian() = default;
+  virtual ~GridCartesian() = default;
 
-    void Init(const std::vector<int> &dimensions,
-	      const std::vector<int> &simd_layout,
-	      const std::vector<int> &processor_grid)
-    {
-      ///////////////////////
-      // Grid information
-      ///////////////////////
+  void Init(const Coordinate &dimensions,
+	    const Coordinate &simd_layout,
+	    const Coordinate &processor_grid)
+  {
+    ///////////////////////
+    // Grid information
+    ///////////////////////
       _isCheckerBoarded = false;
-      _ndimension = dimensions.size();
+    _ndimension = dimensions.size();
 
-      _fdimensions.resize(_ndimension);
-      _gdimensions.resize(_ndimension);
-      _ldimensions.resize(_ndimension);
-      _rdimensions.resize(_ndimension);
-      _simd_layout.resize(_ndimension);
-      _lstart.resize(_ndimension);
-      _lend.resize(_ndimension);
+    _fdimensions.resize(_ndimension);
+    _gdimensions.resize(_ndimension);
+    _ldimensions.resize(_ndimension);
+    _rdimensions.resize(_ndimension);
+    _simd_layout.resize(_ndimension);
+    _lstart.resize(_ndimension);
+    _lend.resize(_ndimension);
 
-      _ostride.resize(_ndimension);
-      _istride.resize(_ndimension);
+    _ostride.resize(_ndimension);
+    _istride.resize(_ndimension);
 
-      _fsites = _gsites = _osites = _isites = 1;
+    _fsites = _gsites = _osites = _isites = 1;
 
-      for (int d = 0; d < _ndimension; d++)
+    for (int d = 0; d < _ndimension; d++)
       {
         _fdimensions[d] = dimensions[d];   // Global dimensions
         _gdimensions[d] = _fdimensions[d]; // Global dimensions
@@ -136,30 +135,30 @@ public:
 
         // Addressing support
         if (d == 0)
-        {
-          _ostride[d] = 1;
-          _istride[d] = 1;
-        }
+	  {
+	    _ostride[d] = 1;
+	    _istride[d] = 1;
+	  }
         else
-        {
-          _ostride[d] = _ostride[d - 1] * _rdimensions[d - 1];
-          _istride[d] = _istride[d - 1] * _simd_layout[d - 1];
-        }
+	  {
+	    _ostride[d] = _ostride[d - 1] * _rdimensions[d - 1];
+	    _istride[d] = _istride[d - 1] * _simd_layout[d - 1];
+	  }
       }
 
-      ///////////////////////
-      // subplane information
-      ///////////////////////
-      _slice_block.resize(_ndimension);
-      _slice_stride.resize(_ndimension);
-      _slice_nblock.resize(_ndimension);
+    ///////////////////////
+    // subplane information
+    ///////////////////////
+    _slice_block.resize(_ndimension);
+    _slice_stride.resize(_ndimension);
+    _slice_nblock.resize(_ndimension);
 
-      int block = 1;
-      int nblock = 1;
-      for (int d = 0; d < _ndimension; d++)
-        nblock *= _rdimensions[d];
+    int block = 1;
+    int nblock = 1;
+    for (int d = 0; d < _ndimension; d++)
+      nblock *= _rdimensions[d];
 
-      for (int d = 0; d < _ndimension; d++)
+    for (int d = 0; d < _ndimension; d++)
       {
         nblock /= _rdimensions[d];
         _slice_block[d] = block;
@@ -167,8 +166,9 @@ public:
         _slice_nblock[d] = nblock;
         block = block * _rdimensions[d];
       }
-    };
+  };
 
 };
-}
+
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/cartesian/Cartesian_red_black.h b/Grid/cartesian/Cartesian_red_black.h
index 5a041f65..34f763d2 100644
--- a/Grid/cartesian/Cartesian_red_black.h
+++ b/Grid/cartesian/Cartesian_red_black.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,178 +24,147 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_CARTESIAN_RED_BLACK_H
 #define GRID_CARTESIAN_RED_BLACK_H
 
+NAMESPACE_BEGIN(Grid);
 
-namespace Grid {
-
-  static const int CbRed  =0;
-  static const int CbBlack=1;
-  static const int Even   =CbRed;
-  static const int Odd    =CbBlack;
+static const int CbRed  =0;
+static const int CbBlack=1;
+static const int Even   =CbRed;
+static const int Odd    =CbBlack;
     
 // Specialise this for red black grids storing half the data like a chess board.
 class GridRedBlackCartesian : public GridBase
 {
 public:
-    std::vector<int> _checker_dim_mask;
-    int              _checker_dim;
-    std::vector<int> _checker_board;
+  Coordinate _checker_dim_mask;
+  int              _checker_dim;
+  std::vector<int> _checker_board;
 
-    virtual int CheckerBoarded(int dim){
-      if( dim==_checker_dim) return 1;
-      else return 0;
-    }
-    virtual int CheckerBoard(const std::vector<int> &site){
-      int linear=0;
-      assert(site.size()==_ndimension);
-      for(int d=0;d<_ndimension;d++){ 
-	if(_checker_dim_mask[d])
-	  linear=linear+site[d];
-      }
-      return (linear&0x1);
+  virtual int CheckerBoarded(int dim){
+    if( dim==_checker_dim) return 1;
+    else return 0;
+  }
+  virtual int CheckerBoard(const Coordinate &site){
+    int linear=0;
+    assert(site.size()==_ndimension);
+    for(int d=0;d<_ndimension;d++){ 
+      if(_checker_dim_mask[d])
+	linear=linear+site[d];
     }
+    return (linear&0x1);
+  }
 
+  // Depending on the cb of site, we toggle source cb.
+  // for block #b, element #e = (b, e)
+  // we need 
+  virtual int CheckerBoardShiftForCB(int source_cb,int dim,int shift,int ocb){
+    if(dim != _checker_dim) return shift;
 
-    // Depending on the cb of site, we toggle source cb.
-    // for block #b, element #e = (b, e)
-    // we need 
-    virtual int CheckerBoardShiftForCB(int source_cb,int dim,int shift,int ocb){
-      if(dim != _checker_dim) return shift;
+    int fulldim =_fdimensions[dim];
+    shift = (shift+fulldim)%fulldim;
 
-      int fulldim =_fdimensions[dim];
-      shift = (shift+fulldim)%fulldim;
-
-      // Probably faster with table lookup;
-      // or by looping over x,y,z and multiply rather than computing checkerboard.
+    // Probably faster with table lookup;
+    // or by looping over x,y,z and multiply rather than computing checkerboard.
 	  
-      if ( (source_cb+ocb)&1 ) {
-	return (shift)/2;
-      } else {
-	return (shift+1)/2;
-      }
+    if ( (source_cb+ocb)&1 ) {
+      return (shift)/2;
+    } else {
+      return (shift+1)/2;
     }
-    virtual int  CheckerBoardFromOindexTable (int Oindex) {
-      return _checker_board[Oindex];
-    }
-    virtual int  CheckerBoardFromOindex (int Oindex)
-    {
-      std::vector<int> ocoor;
-      oCoorFromOindex(ocoor,Oindex);
-      return CheckerBoard(ocoor);
-    }
-    virtual int CheckerBoardShift(int source_cb,int dim,int shift,int osite){
+  }
+  virtual int  CheckerBoardFromOindexTable (int Oindex) {
+    return _checker_board[Oindex];
+  }
+  virtual int  CheckerBoardFromOindex (int Oindex)
+  {
+    Coordinate ocoor;
+    oCoorFromOindex(ocoor,Oindex);
+    return CheckerBoard(ocoor);
+  }
+  virtual int CheckerBoardShift(int source_cb,int dim,int shift,int osite){
 
-      if(dim != _checker_dim) return shift;
+    if(dim != _checker_dim) return shift;
 
-      int ocb=CheckerBoardFromOindex(osite);
+    int ocb=CheckerBoardFromOindex(osite);
       
-      return CheckerBoardShiftForCB(source_cb,dim,shift,ocb);
-    }
+    return CheckerBoardShiftForCB(source_cb,dim,shift,ocb);
+  }
     
-    virtual int CheckerBoardDestination(int source_cb,int shift,int dim){
-      if ( _checker_dim_mask[dim]  ) {
-	// If _fdimensions[checker_dim] is odd, then shifting by 1 in other dims
-	// does NOT cause a parity hop.
-	int add=(dim==_checker_dim) ? 0 : _fdimensions[_checker_dim];
-        if ( (shift+add) &0x1) {
-            return 1-source_cb;
-        } else {
-            return source_cb;
-        }
+  virtual int CheckerBoardDestination(int source_cb,int shift,int dim){
+    if ( _checker_dim_mask[dim]  ) {
+      // If _fdimensions[checker_dim] is odd, then shifting by 1 in other dims
+      // does NOT cause a parity hop.
+      int add=(dim==_checker_dim) ? 0 : _fdimensions[_checker_dim];
+      if ( (shift+add) &0x1) {
+	return 1-source_cb;
       } else {
 	return source_cb;
-
       }
-    };
+    } else {
+      return source_cb;
 
-    ////////////////////////////////////////////////////////////
-    // Create Redblack from original grid; require full grid pointer ?
-    ////////////////////////////////////////////////////////////
-    GridRedBlackCartesian(const GridBase *base) : GridBase(base->_processors,*base)
-    {
-      int dims = base->_ndimension;
-      std::vector<int> checker_dim_mask(dims,1);
-      int checker_dim = 0;
-      Init(base->_fdimensions,base->_simd_layout,base->_processors,checker_dim_mask,checker_dim);
-    };
-
-    ////////////////////////////////////////////////////////////
-    // Create redblack from original grid, with non-trivial checker dim mask
-    ////////////////////////////////////////////////////////////
-    GridRedBlackCartesian(const GridBase *base,
-			  const std::vector<int> &checker_dim_mask,
-			  int checker_dim
-			  ) :  GridBase(base->_processors,*base) 
-    {
-      Init(base->_fdimensions,base->_simd_layout,base->_processors,checker_dim_mask,checker_dim)  ;
     }
+  };
 
-    virtual ~GridRedBlackCartesian() = default;
-#if 0
-    ////////////////////////////////////////////////////////////
-    // Create redblack grid ;; deprecate these. Should not
-    // need direct creation of redblack without a full grid to base on
-    ////////////////////////////////////////////////////////////
-    GridRedBlackCartesian(const GridBase *base,
-			  const std::vector<int> &dimensions,
-			  const std::vector<int> &simd_layout,
-			  const std::vector<int> &processor_grid,
-			  const std::vector<int> &checker_dim_mask,
-			  int checker_dim
-			  ) :  GridBase(processor_grid,*base) 
-    {
-      Init(dimensions,simd_layout,processor_grid,checker_dim_mask,checker_dim);
-    }
+  ////////////////////////////////////////////////////////////
+  // Create Redblack from original grid; require full grid pointer ?
+  ////////////////////////////////////////////////////////////
+  GridRedBlackCartesian(const GridBase *base) : GridBase(base->_processors,*base)
+  {
+    int dims = base->_ndimension;
+    Coordinate checker_dim_mask(dims,1);
+    int checker_dim = 0;
+    Init(base->_fdimensions,base->_simd_layout,base->_processors,checker_dim_mask,checker_dim);
+  };
 
-    ////////////////////////////////////////////////////////////
-    // Create redblack grid
-    ////////////////////////////////////////////////////////////
-    GridRedBlackCartesian(const GridBase *base,
-			  const std::vector<int> &dimensions,
-			  const std::vector<int> &simd_layout,
-			  const std::vector<int> &processor_grid) : GridBase(processor_grid,*base) 
-    {
-      std::vector<int> checker_dim_mask(dimensions.size(),1);
-      int checker_dim = 0;
-      Init(dimensions,simd_layout,processor_grid,checker_dim_mask,checker_dim);
-    }
-#endif
+  ////////////////////////////////////////////////////////////
+  // Create redblack from original grid, with non-trivial checker dim mask
+  ////////////////////////////////////////////////////////////
+  GridRedBlackCartesian(const GridBase *base,
+			const Coordinate &checker_dim_mask,
+			int checker_dim
+			) :  GridBase(base->_processors,*base) 
+  {
+    Init(base->_fdimensions,base->_simd_layout,base->_processors,checker_dim_mask,checker_dim)  ;
+  }
 
-    void Init(const std::vector<int> &dimensions,
-              const std::vector<int> &simd_layout,
-              const std::vector<int> &processor_grid,
-              const std::vector<int> &checker_dim_mask,
-              int checker_dim)
-    {
+  virtual ~GridRedBlackCartesian() = default;
+
+  void Init(const Coordinate &dimensions,
+	    const Coordinate &simd_layout,
+	    const Coordinate &processor_grid,
+	    const Coordinate &checker_dim_mask,
+	    int checker_dim)
+  {
 
       _isCheckerBoarded = true;
-      _checker_dim = checker_dim;
-      assert(checker_dim_mask[checker_dim] == 1);
-      _ndimension = dimensions.size();
-      assert(checker_dim_mask.size() == _ndimension);
-      assert(processor_grid.size() == _ndimension);
-      assert(simd_layout.size() == _ndimension);
+    _checker_dim = checker_dim;
+    assert(checker_dim_mask[checker_dim] == 1);
+    _ndimension = dimensions.size();
+    assert(checker_dim_mask.size() == _ndimension);
+    assert(processor_grid.size() == _ndimension);
+    assert(simd_layout.size() == _ndimension);
 
-      _fdimensions.resize(_ndimension);
-      _gdimensions.resize(_ndimension);
-      _ldimensions.resize(_ndimension);
-      _rdimensions.resize(_ndimension);
-      _simd_layout.resize(_ndimension);
-      _lstart.resize(_ndimension);
-      _lend.resize(_ndimension);
+    _fdimensions.resize(_ndimension);
+    _gdimensions.resize(_ndimension);
+    _ldimensions.resize(_ndimension);
+    _rdimensions.resize(_ndimension);
+    _simd_layout.resize(_ndimension);
+    _lstart.resize(_ndimension);
+    _lend.resize(_ndimension);
 
-      _ostride.resize(_ndimension);
-      _istride.resize(_ndimension);
+    _ostride.resize(_ndimension);
+    _istride.resize(_ndimension);
 
-      _fsites = _gsites = _osites = _isites = 1;
+    _fsites = _gsites = _osites = _isites = 1;
 
-      _checker_dim_mask = checker_dim_mask;
+    _checker_dim_mask = checker_dim_mask;
 
-      for (int d = 0; d < _ndimension; d++)
+    for (int d = 0; d < _ndimension; d++)
       {
         _fdimensions[d] = dimensions[d];
         _gdimensions[d] = _fdimensions[d];
@@ -203,11 +172,11 @@ public:
         _gsites = _gsites * _gdimensions[d];
 
         if (d == _checker_dim)
-        {
-          assert((_gdimensions[d] & 0x1) == 0);
-          _gdimensions[d] = _gdimensions[d] / 2; // Remove a checkerboard
-	  _gsites /= 2;
-        }
+	  {
+	    assert((_gdimensions[d] & 0x1) == 0);
+	    _gdimensions[d] = _gdimensions[d] / 2; // Remove a checkerboard
+	    _gsites /= 2;
+	  }
         _ldimensions[d] = _gdimensions[d] / _processors[d];
         assert(_ldimensions[d] * _processors[d] == _gdimensions[d]);
         _lstart[d] = _processor_coor[d] * _ldimensions[d];
@@ -222,42 +191,42 @@ public:
         // all elements of a simd vector must have same checkerboard.
         // If Ls vectorised, this must still be the case; e.g. dwf rb5d
         if (_simd_layout[d] > 1)
-        {
-          if (checker_dim_mask[d])
-          {
-            assert((_rdimensions[d] & 0x1) == 0);
-          }
-        }
+	  {
+	    if (checker_dim_mask[d])
+	      {
+		assert((_rdimensions[d] & 0x1) == 0);
+	      }
+	  }
 
         _osites *= _rdimensions[d];
         _isites *= _simd_layout[d];
 
         // Addressing support
         if (d == 0)
-        {
-          _ostride[d] = 1;
-          _istride[d] = 1;
-        }
+	  {
+	    _ostride[d] = 1;
+	    _istride[d] = 1;
+	  }
         else
-        {
-          _ostride[d] = _ostride[d - 1] * _rdimensions[d - 1];
-          _istride[d] = _istride[d - 1] * _simd_layout[d - 1];
-        }
+	  {
+	    _ostride[d] = _ostride[d - 1] * _rdimensions[d - 1];
+	    _istride[d] = _istride[d - 1] * _simd_layout[d - 1];
+	  }
       }
 
-      ////////////////////////////////////////////////////////////////////////////////////////////
-      // subplane information
-      ////////////////////////////////////////////////////////////////////////////////////////////
-      _slice_block.resize(_ndimension);
-      _slice_stride.resize(_ndimension);
-      _slice_nblock.resize(_ndimension);
+    ////////////////////////////////////////////////////////////////////////////////////////////
+    // subplane information
+    ////////////////////////////////////////////////////////////////////////////////////////////
+    _slice_block.resize(_ndimension);
+    _slice_stride.resize(_ndimension);
+    _slice_nblock.resize(_ndimension);
 
-      int block = 1;
-      int nblock = 1;
-      for (int d = 0; d < _ndimension; d++)
-        nblock *= _rdimensions[d];
+    int block = 1;
+    int nblock = 1;
+    for (int d = 0; d < _ndimension; d++)
+      nblock *= _rdimensions[d];
 
-      for (int d = 0; d < _ndimension; d++)
+    for (int d = 0; d < _ndimension; d++)
       {
         nblock /= _rdimensions[d];
         _slice_block[d] = block;
@@ -266,55 +235,55 @@ public:
         block = block * _rdimensions[d];
       }
 
-      ////////////////////////////////////////////////
-      // Create a checkerboard lookup table
-      ////////////////////////////////////////////////
-      int rvol = 1;
-      for (int d = 0; d < _ndimension; d++)
+    ////////////////////////////////////////////////
+    // Create a checkerboard lookup table
+    ////////////////////////////////////////////////
+    int rvol = 1;
+    for (int d = 0; d < _ndimension; d++)
       {
         rvol = rvol * _rdimensions[d];
       }
-      _checker_board.resize(rvol);
-      for (int osite = 0; osite < _osites; osite++)
+    _checker_board.resize(rvol);
+    for (int osite = 0; osite < _osites; osite++)
       {
         _checker_board[osite] = CheckerBoardFromOindex(osite);
       }
-    };
+  };
 
-  protected:
-    virtual int oIndex(std::vector<int> &coor)
-    {
-      int idx = 0;
-      for (int d = 0; d < _ndimension; d++)
+protected:
+  virtual int oIndex(Coordinate &coor)
+  {
+    int idx = 0;
+    for (int d = 0; d < _ndimension; d++)
       {
         if (d == _checker_dim)
-        {
-          idx += _ostride[d] * ((coor[d] / 2) % _rdimensions[d]);
-        }
+	  {
+	    idx += _ostride[d] * ((coor[d] / 2) % _rdimensions[d]);
+	  }
         else
-        {
-          idx += _ostride[d] * (coor[d] % _rdimensions[d]);
-        }
+	  {
+	    idx += _ostride[d] * (coor[d] % _rdimensions[d]);
+	  }
       }
-      return idx;
-    };
+    return idx;
+  };
 
-    virtual int iIndex(std::vector<int> &lcoor)
-    {
-      int idx = 0;
-      for (int d = 0; d < _ndimension; d++)
+  virtual int iIndex(Coordinate &lcoor)
+  {
+    int idx = 0;
+    for (int d = 0; d < _ndimension; d++)
       {
         if (d == _checker_dim)
-        {
-          idx += _istride[d] * (lcoor[d] / (2 * _rdimensions[d]));
-        }
+	  {
+	    idx += _istride[d] * (lcoor[d] / (2 * _rdimensions[d]));
+	  }
         else
-        {
-          idx += _istride[d] * (lcoor[d] / _rdimensions[d]);
-        }
+	  {
+	    idx += _istride[d] * (lcoor[d] / _rdimensions[d]);
+	  }
       }
-      return idx;
-    }
+    return idx;
+  }
 };
-}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/communicator/Communicator.h b/Grid/communicator/Communicator.h
index d4ec5a13..76d6f37a 100644
--- a/Grid/communicator/Communicator.h
+++ b/Grid/communicator/Communicator.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,11 +23,12 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_COMMUNICATOR_H
 #define GRID_COMMUNICATOR_H
 
+#include <Grid/util/Coordinate.h>
 #include <Grid/communicator/SharedMemory.h>
 #include <Grid/communicator/Communicator_base.h>
 
diff --git a/Grid/communicator/Communicator_base.cc b/Grid/communicator/Communicator_base.cc
index edbf26af..dfa4846b 100644
--- a/Grid/communicator/Communicator_base.cc
+++ b/Grid/communicator/Communicator_base.cc
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,15 +23,15 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #include <Grid/GridCore.h>
 #include <fcntl.h>
 #include <unistd.h>
 #include <limits.h>
 #include <sys/mman.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 ///////////////////////////////////////////////////////////////
 // Info that is setup once and indept of cartesian layout
@@ -47,8 +47,8 @@ int                      CartesianCommunicator::Dimensions(void)        { return
 int                      CartesianCommunicator::IsBoss(void)            { return _processor==0; };
 int                      CartesianCommunicator::BossRank(void)          { return 0; };
 int                      CartesianCommunicator::ThisRank(void)          { return _processor; };
-const std::vector<int> & CartesianCommunicator::ThisProcessorCoor(void) { return _processor_coor; };
-const std::vector<int> & CartesianCommunicator::ProcessorGrid(void)     { return _processors; };
+const Coordinate & CartesianCommunicator::ThisProcessorCoor(void) { return _processor_coor; };
+const Coordinate & CartesianCommunicator::ProcessorGrid(void)     { return _processors; };
 int                      CartesianCommunicator::ProcessorCount(void)    { return _Nprocessors; };
 
 ////////////////////////////////////////////////////////////////////////////////
@@ -72,5 +72,6 @@ void CartesianCommunicator::GlobalSumVector(ComplexD *c,int N)
   GlobalSumVector((double *)c,2*N);
 }
   
-}
+NAMESPACE_END(Grid);
+
 
diff --git a/Grid/communicator/Communicator_base.h b/Grid/communicator/Communicator_base.h
index 7d6911d3..11dbfcbb 100644
--- a/Grid/communicator/Communicator_base.h
+++ b/Grid/communicator/Communicator_base.h
@@ -1,5 +1,5 @@
 
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,8 +24,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_COMMUNICATOR_BASE_H
 #define GRID_COMMUNICATOR_BASE_H
 
@@ -34,7 +34,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 ///////////////////////////////////
 #include <Grid/communicator/SharedMemory.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 class CartesianCommunicator : public SharedMemory {
 
@@ -52,9 +52,9 @@ public:
   // Communicator should know nothing of the physics grid, only processor grid.
   ////////////////////////////////////////////
   int              _Nprocessors;     // How many in all
-  std::vector<int> _processors;      // Which dimensions get relayed out over processors lanes.
+  Coordinate _processors;      // Which dimensions get relayed out over processors lanes.
   int              _processor;       // linear processor rank
-  std::vector<int> _processor_coor;  // linear processor coordinate
+  Coordinate _processor_coor;  // linear processor coordinate
   unsigned long    _ndimension;
   static Grid_MPI_Comm      communicator_world;
   Grid_MPI_Comm             communicator;
@@ -69,34 +69,34 @@ public:
   // Constructors to sub-divide a parent communicator
   // and default to comm world
   ////////////////////////////////////////////////
-  CartesianCommunicator(const std::vector<int> &processors,const CartesianCommunicator &parent,int &srank);
-  CartesianCommunicator(const std::vector<int> &pdimensions_in);
+  CartesianCommunicator(const Coordinate &processors,const CartesianCommunicator &parent,int &srank);
+  CartesianCommunicator(const Coordinate &pdimensions_in);
   virtual ~CartesianCommunicator();
 
- private:
+private:
 
   ////////////////////////////////////////////////
   // Private initialise from an MPI communicator
   // Can use after an MPI_Comm_split, but hidden from user so private
   ////////////////////////////////////////////////
-  void InitFromMPICommunicator(const std::vector<int> &processors, Grid_MPI_Comm communicator_base);
-
- public:
+  void InitFromMPICommunicator(const Coordinate &processors, Grid_MPI_Comm communicator_base);
 
+public:
+  
   
   ////////////////////////////////////////////////////////////////////////////////////////
   // Wraps MPI_Cart routines, or implements equivalent on other impls
   ////////////////////////////////////////////////////////////////////////////////////////
   void ShiftedRanks(int dim,int shift,int & source, int & dest);
-  int  RankFromProcessorCoor(std::vector<int> &coor);
-  void ProcessorCoorFromRank(int rank,std::vector<int> &coor);
+  int  RankFromProcessorCoor(Coordinate &coor);
+  void ProcessorCoorFromRank(int rank,Coordinate &coor);
   
   int                      Dimensions(void)        ;
   int                      IsBoss(void)            ;
   int                      BossRank(void)          ;
   int                      ThisRank(void)          ;
-  const std::vector<int> & ThisProcessorCoor(void) ;
-  const std::vector<int> & ProcessorGrid(void)     ;
+  const Coordinate & ThisProcessorCoor(void) ;
+  const Coordinate & ProcessorGrid(void)     ;
   int                      ProcessorCount(void)    ;
 
   ////////////////////////////////////////////////////////////////////////////////
@@ -197,11 +197,12 @@ public:
   void AllToAll(void  *in,void *out,uint64_t words         ,uint64_t bytes);
   
   template<class obj> void Broadcast(int root,obj &data)
-    {
-      Broadcast(root,(void *)&data,sizeof(data));
-    };
+  {
+    Broadcast(root,(void *)&data,sizeof(data));
+  }
 
 }; 
-}
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/communicator/Communicator_mpi3.cc b/Grid/communicator/Communicator_mpi3.cc
index 5f46cf18..2576b1fa 100644
--- a/Grid/communicator/Communicator_mpi3.cc
+++ b/Grid/communicator/Communicator_mpi3.cc
@@ -23,12 +23,12 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #include <Grid/GridCore.h>
 #include <Grid/communicator/SharedMemory.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 Grid_MPI_Comm       CartesianCommunicator::communicator_world;
 
@@ -44,10 +44,15 @@ void CartesianCommunicator::Init(int *argc, char ***argv)
   MPI_Initialized(&flag); // needed to coexist with other libs apparently
   if ( !flag ) {
     MPI_Init_thread(argc,argv,MPI_THREAD_MULTIPLE,&provided);
+
     //If only 1 comms thread we require any threading mode other than SINGLE, but for multiple comms threads we need MULTIPLE
-    if( (nCommThreads == 1 && provided == MPI_THREAD_SINGLE) ||
-        (nCommThreads > 1 && provided != MPI_THREAD_MULTIPLE) )
+    if( (nCommThreads == 1) && (provided == MPI_THREAD_SINGLE) ) {
       assert(0);
+    }
+
+    if( (nCommThreads > 1) && (provided != MPI_THREAD_MULTIPLE) ) {
+      assert(0);
+    }
   }
 
   // Never clean up as done once.
@@ -69,14 +74,14 @@ void CartesianCommunicator::ShiftedRanks(int dim,int shift,int &source,int &dest
   int ierr=MPI_Cart_shift(communicator,dim,shift,&source,&dest);
   assert(ierr==0);
 }
-int CartesianCommunicator::RankFromProcessorCoor(std::vector<int> &coor)
+int CartesianCommunicator::RankFromProcessorCoor(Coordinate &coor)
 {
   int rank;
   int ierr=MPI_Cart_rank  (communicator, &coor[0], &rank);
   assert(ierr==0);
   return rank;
 }
-void  CartesianCommunicator::ProcessorCoorFromRank(int rank, std::vector<int> &coor)
+void  CartesianCommunicator::ProcessorCoorFromRank(int rank, Coordinate &coor)
 {
   coor.resize(_ndimension);
   int ierr=MPI_Cart_coords  (communicator, rank, _ndimension,&coor[0]);
@@ -86,7 +91,7 @@ void  CartesianCommunicator::ProcessorCoorFromRank(int rank, std::vector<int> &c
 ////////////////////////////////////////////////////////////////////////////////////////////////////////
 // Initialises from communicator_world
 ////////////////////////////////////////////////////////////////////////////////////////////////////////
-CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors) 
+CartesianCommunicator::CartesianCommunicator(const Coordinate &processors) 
 {
   MPI_Comm optimal_comm;
   ////////////////////////////////////////////////////
@@ -105,12 +110,12 @@ CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors)
 //////////////////////////////////
 // Try to subdivide communicator
 //////////////////////////////////
-CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors,const CartesianCommunicator &parent,int &srank)    
+CartesianCommunicator::CartesianCommunicator(const Coordinate &processors,const CartesianCommunicator &parent,int &srank)    
 {
   _ndimension = processors.size();  assert(_ndimension>=1);
   int parent_ndimension = parent._ndimension; assert(_ndimension >= parent._ndimension);
-  std::vector<int> parent_processor_coor(_ndimension,0);
-  std::vector<int> parent_processors    (_ndimension,1);
+  Coordinate parent_processor_coor(_ndimension,0);
+  Coordinate parent_processors    (_ndimension,1);
 
   // Can make 5d grid from 4d etc...
   int pad = _ndimension-parent_ndimension;
@@ -133,9 +138,9 @@ CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors,
   int Nchild = Nparent/childsize;
   assert (childsize * Nchild == Nparent);
 
-  std::vector<int> ccoor(_ndimension); // coor within subcommunicator
-  std::vector<int> scoor(_ndimension); // coor of split within parent
-  std::vector<int> ssize(_ndimension); // coor of split within parent
+  Coordinate ccoor(_ndimension); // coor within subcommunicator
+  Coordinate scoor(_ndimension); // coor of split within parent
+  Coordinate ssize(_ndimension); // coor of split within parent
 
   for(int d=0;d<_ndimension;d++){
     ccoor[d] = parent_processor_coor[d] % processors[d];
@@ -152,36 +157,6 @@ CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors,
   MPI_Comm comm_split;
   if ( Nchild > 1 ) { 
 
-    if(0){
-      std::cout << GridLogMessage<<"Child communicator of "<< std::hex << parent.communicator << std::dec<<std::endl;
-      std::cout << GridLogMessage<<" parent grid["<< parent._ndimension<<"]    ";
-      for(int d=0;d<parent._ndimension;d++)  std::cout << parent._processors[d] << " ";
-      std::cout<<std::endl;
-      
-      std::cout << GridLogMessage<<" child grid["<< _ndimension <<"]    ";
-      for(int d=0;d<processors.size();d++)  std::cout << processors[d] << " ";
-      std::cout<<std::endl;
-      
-      std::cout << GridLogMessage<<" old rank "<< parent._processor<<" coor ["<< parent._ndimension <<"]    ";
-      for(int d=0;d<parent._ndimension;d++)  std::cout << parent._processor_coor[d] << " ";
-      std::cout<<std::endl;
-      
-      std::cout << GridLogMessage<<" new split "<< srank<<" scoor ["<< _ndimension <<"]    ";
-      for(int d=0;d<processors.size();d++)  std::cout << scoor[d] << " ";
-      std::cout<<std::endl;
-      
-      std::cout << GridLogMessage<<" new rank "<< crank<<" coor ["<< _ndimension <<"]    ";
-      for(int d=0;d<processors.size();d++)  std::cout << ccoor[d] << " ";
-      std::cout<<std::endl;
-
-      //////////////////////////////////////////////////////////////////////////////////////////////////////
-      // Declare victory
-      //////////////////////////////////////////////////////////////////////////////////////////////////////
-      std::cout << GridLogMessage<<"Divided communicator "<< parent._Nprocessors<<" into "
-		<< Nchild <<" communicators with " << childsize << " ranks"<<std::endl;
-      std::cout << " Split communicator " <<comm_split <<std::endl;
-    }
-
     ////////////////////////////////////////////////////////////////
     // Split the communicator
     ////////////////////////////////////////////////////////////////
@@ -203,7 +178,7 @@ CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors,
   // Take the right SHM buffers
   //////////////////////////////////////////////////////////////////////////////////////////////////////
   SetCommunicator(comm_split);
-  
+
   ///////////////////////////////////////////////
   // Free the temp communicator 
   ///////////////////////////////////////////////
@@ -220,7 +195,7 @@ CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors,
   }
 }
 
-void CartesianCommunicator::InitFromMPICommunicator(const std::vector<int> &processors, MPI_Comm communicator_base)
+void CartesianCommunicator::InitFromMPICommunicator(const Coordinate &processors, MPI_Comm communicator_base)
 {
   ////////////////////////////////////////////////////
   // Creates communicator, and the communicator_halo
@@ -237,7 +212,7 @@ void CartesianCommunicator::InitFromMPICommunicator(const std::vector<int> &proc
     _Nprocessors*=_processors[i];
   }
 
-  std::vector<int> periodic(_ndimension,1);
+  Coordinate periodic(_ndimension,1);
   MPI_Cart_create(communicator_base, _ndimension,&_processors[0],&periodic[0],0,&communicator);
   MPI_Comm_rank(communicator,&_processor);
   MPI_Cart_coords(communicator,_processor,_ndimension,&_processor_coor[0]);
@@ -474,7 +449,7 @@ void CartesianCommunicator::BroadcastWorld(int root,void* data, int bytes)
 
 void CartesianCommunicator::AllToAll(int dim,void  *in,void *out,uint64_t words,uint64_t bytes)
 {
-  std::vector<int> row(_ndimension,1);
+  Coordinate row(_ndimension,1);
   assert(dim>=0 && dim<_ndimension);
 
   //  Split the communicator
@@ -503,7 +478,6 @@ void CartesianCommunicator::AllToAll(void  *in,void *out,uint64_t words,uint64_t
   MPI_Type_free(&object);
 }
 
+NAMESPACE_END(Grid);
 
 
-}
-
diff --git a/Grid/communicator/Communicator_none.cc b/Grid/communicator/Communicator_none.cc
index 2369b9b5..b8a15a0e 100644
--- a/Grid/communicator/Communicator_none.cc
+++ b/Grid/communicator/Communicator_none.cc
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,11 +23,11 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #include <Grid/GridCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 // Info that is setup once and indept of cartesian layout
@@ -38,18 +38,18 @@ void CartesianCommunicator::Init(int *argc, char *** arv)
 {
   GlobalSharedMemory::Init(communicator_world);
   GlobalSharedMemory::SharedMemoryAllocate(
-		   GlobalSharedMemory::MAX_MPI_SHM_BYTES,
-		   GlobalSharedMemory::Hugepages);
+					   GlobalSharedMemory::MAX_MPI_SHM_BYTES,
+					   GlobalSharedMemory::Hugepages);
 }
 
-CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors,const CartesianCommunicator &parent,int &srank) 
+CartesianCommunicator::CartesianCommunicator(const Coordinate &processors,const CartesianCommunicator &parent,int &srank) 
   : CartesianCommunicator(processors) 
 {
   srank=0;
   SetCommunicator(communicator_world);
 }
 
-CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors)
+CartesianCommunicator::CartesianCommunicator(const Coordinate &processors)
 {
   _processors = processors;
   _ndimension = processors.size();  assert(_ndimension>=1);
@@ -122,8 +122,8 @@ int  CartesianCommunicator::RankWorld(void){return 0;}
 void CartesianCommunicator::Barrier(void){}
 void CartesianCommunicator::Broadcast(int root,void* data, int bytes) {}
 void CartesianCommunicator::BroadcastWorld(int root,void* data, int bytes) { }
-int  CartesianCommunicator::RankFromProcessorCoor(std::vector<int> &coor) {  return 0;}
-void CartesianCommunicator::ProcessorCoorFromRank(int rank, std::vector<int> &coor){  coor = _processor_coor; }
+int  CartesianCommunicator::RankFromProcessorCoor(Coordinate &coor) {  return 0;}
+void CartesianCommunicator::ProcessorCoorFromRank(int rank, Coordinate &coor){  coor = _processor_coor; }
 void CartesianCommunicator::ShiftedRanks(int dim,int shift,int &source,int &dest)
 {
   source =0;
@@ -160,6 +160,6 @@ void CartesianCommunicator::StencilSendToRecvFromComplete(std::vector<CommsReque
 
 void CartesianCommunicator::StencilBarrier(void){};
 
+NAMESPACE_END(Grid);
 
-}
 
diff --git a/Grid/communicator/SharedMemory.cc b/Grid/communicator/SharedMemory.cc
index 4682d420..5bca9764 100644
--- a/Grid/communicator/SharedMemory.cc
+++ b/Grid/communicator/SharedMemory.cc
@@ -28,10 +28,11 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 #include <Grid/GridCore.h>
 
-namespace Grid { 
+NAMESPACE_BEGIN(Grid); 
 
 // static data
 
+int                 GlobalSharedMemory::HPEhypercube = 1;
 uint64_t            GlobalSharedMemory::MAX_MPI_SHM_BYTES   = 1024LL*1024LL*1024LL; 
 int                 GlobalSharedMemory::Hugepages = 0;
 int                 GlobalSharedMemory::_ShmSetup;
@@ -76,6 +77,7 @@ void *SharedMemory::ShmBufferMalloc(size_t bytes){
     std::cout<< " Current value is " << (heap_size/(1024*1024)) <<std::endl;
     assert(heap_bytes<heap_size);
   }
+  //std::cerr << "ShmBufferMalloc "<<std::hex<< ptr<<" - "<<((uint64_t)ptr+bytes)<<std::dec<<std::endl;
   return ptr;
 }
 void SharedMemory::ShmBufferFreeAll(void) { 
@@ -84,9 +86,9 @@ void SharedMemory::ShmBufferFreeAll(void) {
 }
 void *SharedMemory::ShmBufferSelf(void)
 {
+  //std::cerr << "ShmBufferSelf "<<ShmRank<<" "<<std::hex<< ShmCommBufs[ShmRank] <<std::dec<<std::endl;
   return ShmCommBufs[ShmRank];
 }
 
+NAMESPACE_END(Grid); 
 
-
-}
diff --git a/Grid/communicator/SharedMemory.h b/Grid/communicator/SharedMemory.h
index 63cfc7a6..2f89f571 100644
--- a/Grid/communicator/SharedMemory.h
+++ b/Grid/communicator/SharedMemory.h
@@ -25,18 +25,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     See the full license in the file "LICENSE" in the top level distribution directory
 *************************************************************************************/
 /*  END LEGAL */
-
-
-// TODO
-// 1) move includes into SharedMemory.cc
-//
-// 2) split shared memory into a) optimal communicator creation from comm world
-// 
-//                             b) shared memory buffers container
-//                                -- static globally shared; init once
-//                                -- per instance set of buffers.
-//                                   
-
 #pragma once 
 
 #include <Grid/GridCore.h>
@@ -57,26 +45,32 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <numaif.h>
 #endif
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 #if defined (GRID_COMMS_MPI3) 
-  typedef MPI_Comm    Grid_MPI_Comm;
-  typedef MPI_Request CommsRequest_t;
+typedef MPI_Comm    Grid_MPI_Comm;
+typedef MPI_Request CommsRequest_t;
 #else 
-  typedef int CommsRequest_t;
-  typedef int Grid_MPI_Comm;
+typedef int CommsRequest_t;
+typedef int Grid_MPI_Comm;
 #endif
 
 class GlobalSharedMemory {
- private:
+private:
   static const int     MAXLOG2RANKSPERNODE = 16;            
 
+
   // Init once lock on the buffer allocation
   static int      _ShmSetup;
   static int      _ShmAlloc;
   static uint64_t _ShmAllocBytes;
 
- public:
+public:
+  ///////////////////////////////////////
+  // HPE 8600 hypercube optimisation
+  ///////////////////////////////////////
+  static int HPEhypercube;
+
   static int      ShmSetup(void)      { return _ShmSetup; }
   static int      ShmAlloc(void)      { return _ShmAlloc; }
   static uint64_t ShmAllocBytes(void) { return _ShmAllocBytes; }
@@ -102,14 +96,16 @@ class GlobalSharedMemory {
   // Create an optimal reordered communicator that makes MPI_Cart_create get it right
   //////////////////////////////////////////////////////////////////////////////////////
   static void Init(Grid_MPI_Comm comm); // Typically MPI_COMM_WORLD
-  static void OptimalCommunicator(const std::vector<int> &processors,Grid_MPI_Comm & optimal_comm);  // Turns MPI_COMM_WORLD into right layout for Cartesian
-  static void OptimalCommunicatorHypercube(const std::vector<int> &processors,Grid_MPI_Comm & optimal_comm);  // Turns MPI_COMM_WORLD into right layout for Cartesian
-  static void OptimalCommunicatorSharedMemory(const std::vector<int> &processors,Grid_MPI_Comm & optimal_comm);  // Turns MPI_COMM_WORLD into right layout for Cartesian
+  static void OptimalCommunicator            (const Coordinate &processors,Grid_MPI_Comm & optimal_comm);  // Turns MPI_COMM_WORLD into right layout for Cartesian
+  static void OptimalCommunicatorHypercube   (const Coordinate &processors,Grid_MPI_Comm & optimal_comm);  // Turns MPI_COMM_WORLD into right layout for Cartesian
+  static void OptimalCommunicatorSharedMemory(const Coordinate &processors,Grid_MPI_Comm & optimal_comm);  // Turns MPI_COMM_WORLD into right layout for Cartesian
   ///////////////////////////////////////////////////
   // Provide shared memory facilities off comm world
   ///////////////////////////////////////////////////
   static void SharedMemoryAllocate(uint64_t bytes, int flags);
   static void SharedMemoryFree(void);
+  static void SharedMemoryCopy(void *dest,const void *src,size_t bytes);
+  static void SharedMemoryZero(void *dest,size_t bytes);
 
 };
 
@@ -118,14 +114,14 @@ class GlobalSharedMemory {
 //////////////////////////////
 class SharedMemory 
 {
- private:
+private:
   static const int     MAXLOG2RANKSPERNODE = 16;            
 
   size_t heap_top;
   size_t heap_bytes;
   size_t heap_size;
 
- protected:
+protected:
 
   Grid_MPI_Comm    ShmComm; // for barriers
   int    ShmRank; 
@@ -133,7 +129,7 @@ class SharedMemory
   std::vector<void *> ShmCommBufs;
   std::vector<int>    ShmRanks;// Mapping comm ranks to Shm ranks
 
- public:
+public:
   SharedMemory() {};
   ~SharedMemory();
   ///////////////////////////////////////////////////////////////////////////////////////
@@ -150,6 +146,7 @@ class SharedMemory
   // Call on any instance
   ///////////////////////////////////////////////////
   void SharedMemoryTest(void);
+  
   void *ShmBufferSelf(void);
   void *ShmBuffer    (int rank);
   void *ShmBufferTranslate(int rank,void * local_p);
@@ -164,4 +161,5 @@ class SharedMemory
 
 };
 
-}
+NAMESPACE_END(Grid);
+
diff --git a/Grid/communicator/SharedMemoryMPI.cc b/Grid/communicator/SharedMemoryMPI.cc
index b2a6ab53..b2896bda 100644
--- a/Grid/communicator/SharedMemoryMPI.cc
+++ b/Grid/communicator/SharedMemoryMPI.cc
@@ -29,8 +29,12 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/GridCore.h>
 #include <pwd.h>
 
-namespace Grid { 
+#ifdef GRID_NVCC
+#include <cuda_runtime_api.h>
+#endif
 
+NAMESPACE_BEGIN(Grid); 
+#define header "SharedMemoryMpi: "
 /*Construct from an MPI communicator*/
 void GlobalSharedMemory::Init(Grid_MPI_Comm comm)
 {
@@ -46,6 +50,11 @@ void GlobalSharedMemory::Init(Grid_MPI_Comm comm)
   MPI_Comm_split_type(comm, MPI_COMM_TYPE_SHARED, 0, MPI_INFO_NULL,&WorldShmComm);
   MPI_Comm_rank(WorldShmComm     ,&WorldShmRank);
   MPI_Comm_size(WorldShmComm     ,&WorldShmSize);
+
+  if ( WorldRank == 0) {
+    std::cout << header " World communicator of size " <<WorldSize << std::endl;  
+    std::cout << header " Node  communicator of size " <<WorldShmSize << std::endl;
+  }
   // WorldShmComm, WorldShmSize, WorldShmRank
 
   // WorldNodes
@@ -130,7 +139,7 @@ int Log2Size(int TwoToPower,int MAXLOG2)
   }
   return log2size;
 }
-void GlobalSharedMemory::OptimalCommunicator(const std::vector<int> &processors,Grid_MPI_Comm & optimal_comm)
+void GlobalSharedMemory::OptimalCommunicator(const Coordinate &processors,Grid_MPI_Comm & optimal_comm)
 {
   //////////////////////////////////////////////////////////////////////////////
   // Look and see if it looks like an HPE 8600 based on hostname conventions
@@ -143,10 +152,10 @@ void GlobalSharedMemory::OptimalCommunicator(const std::vector<int> &processors,
   gethostname(name,namelen);
   int nscan = sscanf(name,"r%di%dn%d",&R,&I,&N) ;
 
-  if(nscan==3) OptimalCommunicatorHypercube(processors,optimal_comm);
-  else         OptimalCommunicatorSharedMemory(processors,optimal_comm);
+  if(nscan==3 && HPEhypercube ) OptimalCommunicatorHypercube(processors,optimal_comm);
+  else                          OptimalCommunicatorSharedMemory(processors,optimal_comm);
 }
-void GlobalSharedMemory::OptimalCommunicatorHypercube(const std::vector<int> &processors,Grid_MPI_Comm & optimal_comm)
+void GlobalSharedMemory::OptimalCommunicatorHypercube(const Coordinate &processors,Grid_MPI_Comm & optimal_comm)
 {
   ////////////////////////////////////////////////////////////////
   // Assert power of two shm_size.
@@ -188,9 +197,9 @@ void GlobalSharedMemory::OptimalCommunicatorHypercube(const std::vector<int> &pr
   }
 
   std::string hname(name);
-  std::cout << "hostname "<<hname<<std::endl;
-  std::cout << "R " << R << " I " << I << " N "<< N
-            << " hypercoor 0x"<<std::hex<<hypercoor<<std::dec<<std::endl;
+  //  std::cout << "hostname "<<hname<<std::endl;
+  //  std::cout << "R " << R << " I " << I << " N "<< N
+  //            << " hypercoor 0x"<<std::hex<<hypercoor<<std::dec<<std::endl;
 
   //////////////////////////////////////////////////////////////////
   // broadcast node 0's base coordinate for this partition.
@@ -213,7 +222,8 @@ void GlobalSharedMemory::OptimalCommunicatorHypercube(const std::vector<int> &pr
   ////////////////////////////////////////////////////////////////
   int ndimension              = processors.size();
   std::vector<int> processor_coor(ndimension);
-  std::vector<int> WorldDims = processors;   std::vector<int> ShmDims  (ndimension,1);  std::vector<int> NodeDims (ndimension);
+  std::vector<int> WorldDims = processors.toVector();
+  std::vector<int> ShmDims  (ndimension,1);  std::vector<int> NodeDims (ndimension);
   std::vector<int> ShmCoor  (ndimension);    std::vector<int> NodeCoor (ndimension);    std::vector<int> WorldCoor(ndimension);
   std::vector<int> HyperCoor(ndimension);
   int dim = 0;
@@ -221,7 +231,7 @@ void GlobalSharedMemory::OptimalCommunicatorHypercube(const std::vector<int> &pr
     while ( (WorldDims[dim] / ShmDims[dim]) <= 1 ) dim=(dim+1)%ndimension;
     ShmDims[dim]*=2;
     dim=(dim+1)%ndimension;
-  }
+    }
 
   ////////////////////////////////////////////////////////////////
   // Establish torus of processes and nodes with sub-blockings
@@ -240,7 +250,7 @@ void GlobalSharedMemory::OptimalCommunicatorHypercube(const std::vector<int> &pr
      HyperCoor[d]=hcoor & msk;  
      HyperCoor[d]=BinaryToGray(HyperCoor[d]); // Space filling curve magic
      hcoor = hcoor >> bits;
-  } 
+  }
   ////////////////////////////////////////////////////////////////
   // Check processor counts match
   ////////////////////////////////////////////////////////////////
@@ -269,7 +279,7 @@ void GlobalSharedMemory::OptimalCommunicatorHypercube(const std::vector<int> &pr
   int ierr= MPI_Comm_split(WorldComm,0,rank,&optimal_comm);
   assert(ierr==0);
 }
-void GlobalSharedMemory::OptimalCommunicatorSharedMemory(const std::vector<int> &processors,Grid_MPI_Comm & optimal_comm)
+void GlobalSharedMemory::OptimalCommunicatorSharedMemory(const Coordinate &processors,Grid_MPI_Comm & optimal_comm)
 {
   ////////////////////////////////////////////////////////////////
   // Assert power of two shm_size.
@@ -282,9 +292,9 @@ void GlobalSharedMemory::OptimalCommunicatorSharedMemory(const std::vector<int>
   // in a maximally symmetrical way
   ////////////////////////////////////////////////////////////////
   int ndimension              = processors.size();
-  std::vector<int> processor_coor(ndimension);
-  std::vector<int> WorldDims = processors;   std::vector<int> ShmDims  (ndimension,1);  std::vector<int> NodeDims (ndimension);
-  std::vector<int> ShmCoor  (ndimension);    std::vector<int> NodeCoor (ndimension);    std::vector<int> WorldCoor(ndimension);
+  Coordinate processor_coor(ndimension);
+  Coordinate WorldDims = processors; Coordinate ShmDims(ndimension,1);  Coordinate NodeDims (ndimension);
+  Coordinate ShmCoor(ndimension);    Coordinate NodeCoor(ndimension);   Coordinate WorldCoor(ndimension);
   int dim = 0;
   for(int l2=0;l2<log2size;l2++){
     while ( (WorldDims[dim] / ShmDims[dim]) <= 1 ) dim=(dim+1)%ndimension;
@@ -330,7 +340,7 @@ void GlobalSharedMemory::OptimalCommunicatorSharedMemory(const std::vector<int>
 #ifdef GRID_MPI3_SHMGET
 void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 {
-  std::cout << "SharedMemoryAllocate "<< bytes<< " shmget implementation "<<std::endl;
+  std::cout << header "SharedMemoryAllocate "<< bytes<< " shmget implementation "<<std::endl;
   assert(_ShmSetup==1);
   assert(_ShmAlloc==0);
 
@@ -385,14 +395,101 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
   _ShmAllocBytes  = bytes;
 }
 #endif
- 
+
 ////////////////////////////////////////////////////////////////////////////////////////////
 // Hugetlbfs mapping intended
 ////////////////////////////////////////////////////////////////////////////////////////////
+#ifdef GRID_NVCC
+void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
+{
+  void * ShmCommBuf ; 
+  assert(_ShmSetup==1);
+  assert(_ShmAlloc==0);
+
+  //////////////////////////////////////////////////////////////////////////////////////////////////////////
+  // allocate the pointer array for shared windows for our group
+  //////////////////////////////////////////////////////////////////////////////////////////////////////////
+  MPI_Barrier(WorldShmComm);
+  WorldShmCommBufs.resize(WorldShmSize);
+
+  //////////////////////////////////////////////////////////////////////////////////////////////////////////
+  // TODO/FIXME : NOT ALL NVLINK BOARDS have full Peer to peer connectivity.
+  // The annoyance is that they have partial peer 2 peer. This occurs on the 8 GPU blades.
+  // e.g. DGX1, supermicro board, 
+  //////////////////////////////////////////////////////////////////////////////////////////////////////////
+  //  cudaDeviceGetP2PAttribute(&perfRank, cudaDevP2PAttrPerformanceRank, device1, device2);
+  cudaSetDevice(WorldShmRank);
+  ///////////////////////////////////////////////////////////////////////////////////////////////////////////
+  // Each MPI rank should allocate our own buffer
+  ///////////////////////////////////////////////////////////////////////////////////////////////////////////
+  auto err =  cudaMalloc(&ShmCommBuf, bytes);
+  if ( err !=  cudaSuccess) {
+    std::cerr << " SharedMemoryMPI.cc cudaMallocManaged failed for " << bytes<<" bytes " <<cudaGetErrorString(err)<< std::endl;
+    exit(EXIT_FAILURE);  
+  }
+  if (ShmCommBuf == (void *)NULL ) {
+    std::cerr << " SharedMemoryMPI.cc cudaMallocManaged failed NULL pointer for " << bytes<<" bytes " << std::endl;
+    exit(EXIT_FAILURE);  
+  }
+  if ( WorldRank == 0 ){
+    std::cout << header " SharedMemoryMPI.cc cudaMalloc "<< bytes << "bytes at "<< std::hex<< ShmCommBuf <<std::dec<<" for comms buffers " <<std::endl;
+  }
+  SharedMemoryZero(ShmCommBuf,bytes);
+
+  ///////////////////////////////////////////////////////////////////////////////////////////////////////////
+  // Loop over ranks/gpu's on our node
+  ///////////////////////////////////////////////////////////////////////////////////////////////////////////
+  for(int r=0;r<WorldShmSize;r++){
+    
+    //////////////////////////////////////////////////
+    // If it is me, pass around the IPC access key
+    //////////////////////////////////////////////////
+    cudaIpcMemHandle_t handle;
+
+    if ( r==WorldShmRank ) { 
+      err = cudaIpcGetMemHandle(&handle,ShmCommBuf);
+      if ( err !=  cudaSuccess) {
+	std::cerr << " SharedMemoryMPI.cc cudaIpcGetMemHandle failed for rank" << r <<" "<<cudaGetErrorString(err)<< std::endl;
+	exit(EXIT_FAILURE);
+      }
+    }
+    //////////////////////////////////////////////////
+    // Share this IPC handle across the Shm Comm
+    //////////////////////////////////////////////////
+    { 
+      int ierr=MPI_Bcast(&handle,
+			 sizeof(handle),
+			 MPI_BYTE,
+			 r,
+			 WorldShmComm);
+      assert(ierr==0);
+    }
+    
+    ///////////////////////////////////////////////////////////////
+    // If I am not the source, overwrite thisBuf with remote buffer
+    ///////////////////////////////////////////////////////////////
+    void * thisBuf = ShmCommBuf;
+    if ( r!=WorldShmRank ) { 
+      err = cudaIpcOpenMemHandle(&thisBuf,handle,cudaIpcMemLazyEnablePeerAccess);
+      if ( err !=  cudaSuccess) {
+	std::cerr << " SharedMemoryMPI.cc cudaIpcOpenMemHandle failed for rank" << r <<" "<<cudaGetErrorString(err)<< std::endl;
+	exit(EXIT_FAILURE);
+      }
+    }
+    ///////////////////////////////////////////////////////////////
+    // Save a copy of the device buffers
+    ///////////////////////////////////////////////////////////////
+    WorldShmCommBufs[r] = thisBuf;
+  }
+
+  _ShmAllocBytes=bytes;
+  _ShmAlloc=1;
+}
+#else 
 #ifdef GRID_MPI3_SHMMMAP
 void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 {
-  std::cout << "SharedMemoryAllocate "<< bytes<< " MMAP implementation "<< GRID_SHM_PATH <<std::endl;
+  std::cout << header "SharedMemoryAllocate "<< bytes<< " MMAP implementation "<< GRID_SHM_PATH <<std::endl;
   assert(_ShmSetup==1);
   assert(_ShmAlloc==0);
   //////////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -429,7 +526,7 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
     assert(((uint64_t)ptr&0x3F)==0);
     close(fd);
     WorldShmCommBufs[r] =ptr;
-    //    std::cout << "Set WorldShmCommBufs["<<r<<"]="<<ptr<< "("<< bytes<< "bytes)"<<std::endl;
+    //    std::cout << header "Set WorldShmCommBufs["<<r<<"]="<<ptr<< "("<< bytes<< "bytes)"<<std::endl;
   }
   _ShmAlloc=1;
   _ShmAllocBytes  = bytes;
@@ -439,7 +536,7 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 #ifdef GRID_MPI3_SHM_NONE
 void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 {
-  std::cout << "SharedMemoryAllocate "<< bytes<< " MMAP anonymous implementation "<<std::endl;
+  std::cout << header "SharedMemoryAllocate "<< bytes<< " MMAP anonymous implementation "<<std::endl;
   assert(_ShmSetup==1);
   assert(_ShmAlloc==0);
   //////////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -486,7 +583,7 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 ////////////////////////////////////////////////////////////////////////////////////////////
 void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
 { 
-  std::cout << "SharedMemoryAllocate "<< bytes<< " SHMOPEN implementation "<<std::endl;
+  std::cout << header "SharedMemoryAllocate "<< bytes<< " SHMOPEN implementation "<<std::endl;
   assert(_ShmSetup==1);
   assert(_ShmAlloc==0); 
   MPI_Barrier(WorldShmComm);
@@ -552,14 +649,31 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
   _ShmAllocBytes = bytes;
 }
 #endif
+#endif // End NVCC case for GPU device buffers
 
-
-
-
-  ////////////////////////////////////////////////////////
-  // Global shared functionality finished
-  // Now move to per communicator functionality
-  ////////////////////////////////////////////////////////
+/////////////////////////////////////////////////////////////////////////
+// Routines accessing shared memory should route through for GPU safety
+/////////////////////////////////////////////////////////////////////////
+void GlobalSharedMemory::SharedMemoryZero(void *dest,size_t bytes)
+{
+#ifdef GRID_NVCC
+  cudaMemset(dest,0,bytes);
+#else
+  bzero(dest,bytes);
+#endif
+}
+void GlobalSharedMemory::SharedMemoryCopy(void *dest,const void *src,size_t bytes)
+{
+#ifdef GRID_NVCC
+  cudaMemcpy(dest,src,bytes,cudaMemcpyDefault);
+#else   
+  bcopy(src,dest,bytes);
+#endif
+}
+////////////////////////////////////////////////////////
+// Global shared functionality finished
+// Now move to per communicator functionality
+////////////////////////////////////////////////////////
 void SharedMemory::SetCommunicator(Grid_MPI_Comm comm)
 {
   int rank, size;
@@ -587,7 +701,6 @@ void SharedMemory::SetCommunicator(Grid_MPI_Comm comm)
     MPI_Allreduce(MPI_IN_PLACE,&wsr,1,MPI_UINT32_T,MPI_SUM,ShmComm);
 
     ShmCommBufs[r] = GlobalSharedMemory::WorldShmCommBufs[wsr];
-    //    std::cout << "SetCommunicator ShmCommBufs ["<< r<< "] = "<< ShmCommBufs[r]<< "  wsr = "<<wsr<<std::endl;
   }
   ShmBufferFreeAll();
 
@@ -600,6 +713,8 @@ void SharedMemory::SetCommunicator(Grid_MPI_Comm comm)
 
   std::vector<int> ranks(size);   for(int r=0;r<size;r++) ranks[r]=r;
   MPI_Group_translate_ranks (FullGroup,size,&ranks[0],ShmGroup, &ShmRanks[0]); 
+
+  SharedMemoryTest();
 }
 //////////////////////////////////////////////////////////////////
 // On node barrier
@@ -614,24 +729,26 @@ void SharedMemory::ShmBarrier(void)
 void SharedMemory::SharedMemoryTest(void)
 {
   ShmBarrier();
+  uint64_t check[3];
+  uint64_t magic = 0x5A5A5A;
   if ( ShmRank == 0 ) {
-    for(int r=0;r<ShmSize;r++){
-      uint64_t * check = (uint64_t *) ShmCommBufs[r];
-      check[0] = GlobalSharedMemory::WorldNode;
-      check[1] = r;
-      check[2] = 0x5A5A5A;
+    for(uint64_t r=0;r<ShmSize;r++){
+       check[0]=GlobalSharedMemory::WorldNode;
+       check[1]=r;
+       check[2]=magic;
+       GlobalSharedMemory::SharedMemoryCopy( ShmCommBufs[r], check, 3*sizeof(uint64_t));
     }
   }
   ShmBarrier();
-  for(int r=0;r<ShmSize;r++){
-    uint64_t * check = (uint64_t *) ShmCommBufs[r];
-    
+  for(uint64_t r=0;r<ShmSize;r++){
+    ShmBarrier();
+    GlobalSharedMemory::SharedMemoryCopy(check,ShmCommBufs[r], 3*sizeof(uint64_t));
+    ShmBarrier();
     assert(check[0]==GlobalSharedMemory::WorldNode);
     assert(check[1]==r);
-    assert(check[2]==0x5A5A5A);
-    
+    assert(check[2]==magic);
+    ShmBarrier();
   }
-  ShmBarrier();
 }
 
 void *SharedMemory::ShmBuffer(int rank)
@@ -645,7 +762,6 @@ void *SharedMemory::ShmBuffer(int rank)
 }
 void *SharedMemory::ShmBufferTranslate(int rank,void * local_p)
 {
-  static int count =0;
   int gpeer = ShmRanks[rank];
   assert(gpeer!=ShmRank); // never send to self
   if (gpeer == MPI_UNDEFINED){
@@ -664,4 +780,5 @@ SharedMemory::~SharedMemory()
   }
 };
 
-}
+NAMESPACE_END(Grid); 
+
diff --git a/Grid/communicator/SharedMemoryNone.cc b/Grid/communicator/SharedMemoryNone.cc
index a23e3c1c..ed37ab47 100644
--- a/Grid/communicator/SharedMemoryNone.cc
+++ b/Grid/communicator/SharedMemoryNone.cc
@@ -28,7 +28,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 #include <Grid/GridCore.h>
 
-namespace Grid { 
+NAMESPACE_BEGIN(Grid); 
 
 /*Construct from an MPI communicator*/
 void GlobalSharedMemory::Init(Grid_MPI_Comm comm)
@@ -47,7 +47,7 @@ void GlobalSharedMemory::Init(Grid_MPI_Comm comm)
   _ShmSetup=1;
 }
 
-void GlobalSharedMemory::OptimalCommunicator(const std::vector<int> &processors,Grid_MPI_Comm & optimal_comm)
+void GlobalSharedMemory::OptimalCommunicator(const Coordinate &processors,Grid_MPI_Comm & optimal_comm)
 {
   optimal_comm = WorldComm;
 }
@@ -84,10 +84,10 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
   _ShmAlloc=1;
 };
 
-  ////////////////////////////////////////////////////////
-  // Global shared functionality finished
-  // Now move to per communicator functionality
-  ////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////
+// Global shared functionality finished
+// Now move to per communicator functionality
+////////////////////////////////////////////////////////
 void SharedMemory::SetCommunicator(Grid_MPI_Comm comm)
 {
   assert(GlobalSharedMemory::ShmAlloc()==1);
@@ -125,4 +125,5 @@ void *SharedMemory::ShmBufferTranslate(int rank,void * local_p)
 SharedMemory::~SharedMemory()
 {};
 
-}
+NAMESPACE_END(Grid); 
+
diff --git a/Grid/cshift/Cshift.h b/Grid/cshift/Cshift.h
index 7d0caeee..07ec8412 100644
--- a/Grid/cshift/Cshift.h
+++ b/Grid/cshift/Cshift.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,8 +23,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef _GRID_CSHIFT_H_
 #define _GRID_CSHIFT_H_
 
diff --git a/Grid/cshift/Cshift_common.h b/Grid/cshift/Cshift_common.h
index 24e0d2ab..954342cb 100644
--- a/Grid/cshift/Cshift_common.h
+++ b/Grid/cshift/Cshift_common.h
@@ -25,10 +25,9 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     See the full license in the file "LICENSE" in the top level distribution directory
     *************************************************************************************/
     /*  END LEGAL */
-#ifndef _GRID_CSHIFT_COMMON_H_
-#define _GRID_CSHIFT_COMMON_H_
+#pragma once
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 ///////////////////////////////////////////////////////////////////
 // Gather for when there is no need to SIMD split 
@@ -36,20 +35,21 @@ namespace Grid {
 template<class vobj> void 
 Gather_plane_simple (const Lattice<vobj> &rhs,commVector<vobj> &buffer,int dimension,int plane,int cbmask, int off=0)
 {
-  int rd = rhs._grid->_rdimensions[dimension];
+  int rd = rhs.Grid()->_rdimensions[dimension];
 
-  if ( !rhs._grid->CheckerBoarded(dimension) ) {
+  if ( !rhs.Grid()->CheckerBoarded(dimension) ) {
     cbmask = 0x3;
   }
   
-  int so=plane*rhs._grid->_ostride[dimension]; // base offset for start of plane 
-  int e1=rhs._grid->_slice_nblock[dimension];
-  int e2=rhs._grid->_slice_block[dimension];
+  int so=plane*rhs.Grid()->_ostride[dimension]; // base offset for start of plane 
+  int e1=rhs.Grid()->_slice_nblock[dimension];
+  int e2=rhs.Grid()->_slice_block[dimension];
   int ent = 0;
 
-  static std::vector<std::pair<int,int> > table; table.resize(e1*e2);
+  static Vector<std::pair<int,int> > table; table.resize(e1*e2);
+  int stride=rhs.Grid()->_slice_stride[dimension];
 
-  int stride=rhs._grid->_slice_stride[dimension];
+  auto rhs_v = rhs.View();
   if ( cbmask == 0x3 ) { 
     for(int n=0;n<e1;n++){
       for(int b=0;b<e2;b++){
@@ -63,66 +63,68 @@ Gather_plane_simple (const Lattice<vobj> &rhs,commVector<vobj> &buffer,int dimen
      for(int n=0;n<e1;n++){
        for(int b=0;b<e2;b++){
 	 int o  = n*stride;
-	 int ocb=1<<rhs._grid->CheckerBoardFromOindex(o+b);
+	 int ocb=1<<rhs.Grid()->CheckerBoardFromOindex(o+b);
 	 if ( ocb &cbmask ) {
 	   table[ent++]=std::pair<int,int> (off+bo++,so+o+b);
 	 }
        }
      }
   }
-  parallel_for(int i=0;i<ent;i++){
-    buffer[table[i].first]=rhs._odata[table[i].second];
-  }
+  thread_for(i,ent,{
+    buffer[table[i].first]=rhs_v[table[i].second];
+  });
 }
 
 ///////////////////////////////////////////////////////////////////
 // Gather for when there *is* need to SIMD split 
 ///////////////////////////////////////////////////////////////////
 template<class vobj> void 
-Gather_plane_extract(const Lattice<vobj> &rhs,std::vector<typename vobj::scalar_object *> pointers,int dimension,int plane,int cbmask)
+Gather_plane_extract(const Lattice<vobj> &rhs,
+		     ExtractPointerArray<typename vobj::scalar_object> pointers,
+		     int dimension,int plane,int cbmask)
 {
-  int rd = rhs._grid->_rdimensions[dimension];
+  int rd = rhs.Grid()->_rdimensions[dimension];
 
-  if ( !rhs._grid->CheckerBoarded(dimension) ) {
+  if ( !rhs.Grid()->CheckerBoarded(dimension) ) {
     cbmask = 0x3;
   }
 
-  int so  = plane*rhs._grid->_ostride[dimension]; // base offset for start of plane 
+  int so  = plane*rhs.Grid()->_ostride[dimension]; // base offset for start of plane 
 
-  int e1=rhs._grid->_slice_nblock[dimension];
-  int e2=rhs._grid->_slice_block[dimension];
-  int n1=rhs._grid->_slice_stride[dimension];
+  int e1=rhs.Grid()->_slice_nblock[dimension];
+  int e2=rhs.Grid()->_slice_block[dimension];
+  int n1=rhs.Grid()->_slice_stride[dimension];
 
+  auto rhs_v = rhs.View();
   if ( cbmask ==0x3){
-    parallel_for_nest2(int n=0;n<e1;n++){
+    thread_for_collapse(2,n,e1,{
       for(int b=0;b<e2;b++){
 
 	int o      =   n*n1;
 	int offset = b+n*e2;
 	
-	vobj temp =rhs._odata[so+o+b];
+	vobj temp =rhs_v[so+o+b];
 	extract<vobj>(temp,pointers,offset);
-
       }
-    }
+    });
   } else { 
 
     // Case of SIMD split AND checker dim cannot currently be hit, except in 
     // Test_cshift_red_black code.
     std::cout << " Dense packed buffer WARNING " <<std::endl;
-    parallel_for_nest2(int n=0;n<e1;n++){
+    thread_for_collapse(2,n,e1,{
       for(int b=0;b<e2;b++){
 
 	int o=n*n1;
-	int ocb=1<<rhs._grid->CheckerBoardFromOindex(o+b);
+	int ocb=1<<rhs.Grid()->CheckerBoardFromOindex(o+b);
 	int offset = b+n*e2;
 
 	if ( ocb & cbmask ) {
-	  vobj temp =rhs._odata[so+o+b];
+	  vobj temp =rhs_v[so+o+b];
 	  extract<vobj>(temp,pointers,offset);
 	}
       }
-    }
+    });
   }
 }
 
@@ -131,17 +133,17 @@ Gather_plane_extract(const Lattice<vobj> &rhs,std::vector<typename vobj::scalar_
 //////////////////////////////////////////////////////
 template<class vobj> void Scatter_plane_simple (Lattice<vobj> &rhs,commVector<vobj> &buffer, int dimension,int plane,int cbmask)
 {
-  int rd = rhs._grid->_rdimensions[dimension];
+  int rd = rhs.Grid()->_rdimensions[dimension];
 
-  if ( !rhs._grid->CheckerBoarded(dimension) ) {
+  if ( !rhs.Grid()->CheckerBoarded(dimension) ) {
     cbmask=0x3;
   }
 
-  int so  = plane*rhs._grid->_ostride[dimension]; // base offset for start of plane 
+  int so  = plane*rhs.Grid()->_ostride[dimension]; // base offset for start of plane 
     
-  int e1=rhs._grid->_slice_nblock[dimension];
-  int e2=rhs._grid->_slice_block[dimension];
-  int stride=rhs._grid->_slice_stride[dimension];
+  int e1=rhs.Grid()->_slice_nblock[dimension];
+  int e2=rhs.Grid()->_slice_block[dimension];
+  int stride=rhs.Grid()->_slice_stride[dimension];
 
   static std::vector<std::pair<int,int> > table; table.resize(e1*e2);
   int ent    =0;
@@ -150,8 +152,8 @@ template<class vobj> void Scatter_plane_simple (Lattice<vobj> &rhs,commVector<vo
 
     for(int n=0;n<e1;n++){
       for(int b=0;b<e2;b++){
-	int o   =n*rhs._grid->_slice_stride[dimension];
-	int bo  =n*rhs._grid->_slice_block[dimension];
+	int o   =n*rhs.Grid()->_slice_stride[dimension];
+	int bo  =n*rhs.Grid()->_slice_block[dimension];
 	table[ent++] = std::pair<int,int>(so+o+b,bo+b);
       }
     }
@@ -160,57 +162,60 @@ template<class vobj> void Scatter_plane_simple (Lattice<vobj> &rhs,commVector<vo
     int bo=0;
     for(int n=0;n<e1;n++){
       for(int b=0;b<e2;b++){
-	int o   =n*rhs._grid->_slice_stride[dimension];
-	int ocb=1<<rhs._grid->CheckerBoardFromOindex(o+b);// Could easily be a table lookup
+	int o   =n*rhs.Grid()->_slice_stride[dimension];
+	int ocb=1<<rhs.Grid()->CheckerBoardFromOindex(o+b);// Could easily be a table lookup
 	if ( ocb & cbmask ) {
 	  table[ent++]=std::pair<int,int> (so+o+b,bo++);
 	}
       }
     }
   }
-
-  parallel_for(int i=0;i<ent;i++){
-    rhs._odata[table[i].first]=buffer[table[i].second];
-  }
+  
+  auto rhs_v = rhs.View();
+  thread_for(i,ent,{
+    rhs_v[table[i].first]=buffer[table[i].second];
+  });
 }
 
 //////////////////////////////////////////////////////
 // Scatter for when there *is* need to SIMD split
 //////////////////////////////////////////////////////
-template<class vobj> void Scatter_plane_merge(Lattice<vobj> &rhs,std::vector<typename vobj::scalar_object *> pointers,int dimension,int plane,int cbmask)
+template<class vobj> void Scatter_plane_merge(Lattice<vobj> &rhs,ExtractPointerArray<typename vobj::scalar_object> pointers,int dimension,int plane,int cbmask)
 {
-  int rd = rhs._grid->_rdimensions[dimension];
+  int rd = rhs.Grid()->_rdimensions[dimension];
 
-  if ( !rhs._grid->CheckerBoarded(dimension) ) {
+  if ( !rhs.Grid()->CheckerBoarded(dimension) ) {
     cbmask=0x3;
   }
 
-  int so  = plane*rhs._grid->_ostride[dimension]; // base offset for start of plane 
+  int so  = plane*rhs.Grid()->_ostride[dimension]; // base offset for start of plane 
     
-  int e1=rhs._grid->_slice_nblock[dimension];
-  int e2=rhs._grid->_slice_block[dimension];
+  int e1=rhs.Grid()->_slice_nblock[dimension];
+  int e2=rhs.Grid()->_slice_block[dimension];
 
   if(cbmask ==0x3 ) {
-    parallel_for_nest2(int n=0;n<e1;n++){
+    auto rhs_v = rhs.View();
+    thread_for_collapse(2,n,e1,{
       for(int b=0;b<e2;b++){
-	int o      = n*rhs._grid->_slice_stride[dimension];
-	int offset = b+n*rhs._grid->_slice_block[dimension];
-	merge(rhs._odata[so+o+b],pointers,offset);
+	int o      = n*rhs.Grid()->_slice_stride[dimension];
+	int offset = b+n*rhs.Grid()->_slice_block[dimension];
+	merge(rhs_v[so+o+b],pointers,offset);
       }
-    }
+    });
   } else { 
 
     // Case of SIMD split AND checker dim cannot currently be hit, except in 
     // Test_cshift_red_black code.
     //    std::cout << "Scatter_plane merge assert(0); think this is buggy FIXME "<< std::endl;// think this is buggy FIXME
     std::cout<<" Unthreaded warning -- buffer is not densely packed ??"<<std::endl;
+    auto rhs_v = rhs.View();
     for(int n=0;n<e1;n++){
       for(int b=0;b<e2;b++){
-	int o      = n*rhs._grid->_slice_stride[dimension];
-	int offset = b+n*rhs._grid->_slice_block[dimension];
-	int ocb=1<<rhs._grid->CheckerBoardFromOindex(o+b);
+	int o      = n*rhs.Grid()->_slice_stride[dimension];
+	int offset = b+n*rhs.Grid()->_slice_block[dimension];
+	int ocb=1<<rhs.Grid()->CheckerBoardFromOindex(o+b);
 	if ( ocb&cbmask ) {
-	  merge(rhs._odata[so+o+b],pointers,offset);
+	  merge(rhs_v[so+o+b],pointers,offset);
 	}
       }
     }
@@ -222,18 +227,18 @@ template<class vobj> void Scatter_plane_merge(Lattice<vobj> &rhs,std::vector<typ
 //////////////////////////////////////////////////////
 template<class vobj> void Copy_plane(Lattice<vobj>& lhs,const Lattice<vobj> &rhs, int dimension,int lplane,int rplane,int cbmask)
 {
-  int rd = rhs._grid->_rdimensions[dimension];
+  int rd = rhs.Grid()->_rdimensions[dimension];
 
-  if ( !rhs._grid->CheckerBoarded(dimension) ) {
+  if ( !rhs.Grid()->CheckerBoarded(dimension) ) {
     cbmask=0x3;
   }
 
-  int ro  = rplane*rhs._grid->_ostride[dimension]; // base offset for start of plane 
-  int lo  = lplane*lhs._grid->_ostride[dimension]; // base offset for start of plane 
+  int ro  = rplane*rhs.Grid()->_ostride[dimension]; // base offset for start of plane 
+  int lo  = lplane*lhs.Grid()->_ostride[dimension]; // base offset for start of plane 
 
-  int e1=rhs._grid->_slice_nblock[dimension]; // clearly loop invariant for icpc
-  int e2=rhs._grid->_slice_block[dimension];
-  int stride = rhs._grid->_slice_stride[dimension];
+  int e1=rhs.Grid()->_slice_nblock[dimension]; // clearly loop invariant for icpc
+  int e2=rhs.Grid()->_slice_block[dimension];
+  int stride = rhs.Grid()->_slice_stride[dimension];
   static std::vector<std::pair<int,int> > table; table.resize(e1*e2);
   int ent=0;
 
@@ -248,7 +253,7 @@ template<class vobj> void Copy_plane(Lattice<vobj>& lhs,const Lattice<vobj> &rhs
     for(int n=0;n<e1;n++){
       for(int b=0;b<e2;b++){
         int o =n*stride+b;
-        int ocb=1<<lhs._grid->CheckerBoardFromOindex(o);
+        int ocb=1<<lhs.Grid()->CheckerBoardFromOindex(o);
         if ( ocb&cbmask ) {
 	  table[ent++] = std::pair<int,int>(lo+o,ro+o);
 	}
@@ -256,32 +261,33 @@ template<class vobj> void Copy_plane(Lattice<vobj>& lhs,const Lattice<vobj> &rhs
     }
   }
 
-  parallel_for(int i=0;i<ent;i++){
-    lhs._odata[table[i].first]=rhs._odata[table[i].second];
-  }
+  auto rhs_v = rhs.View();
+  auto lhs_v = lhs.View();
+  thread_for(i,ent,{
+    lhs_v[table[i].first]=rhs_v[table[i].second];
+  });
 
 }
 
 template<class vobj> void Copy_plane_permute(Lattice<vobj>& lhs,const Lattice<vobj> &rhs, int dimension,int lplane,int rplane,int cbmask,int permute_type)
 {
  
-  int rd = rhs._grid->_rdimensions[dimension];
+  int rd = rhs.Grid()->_rdimensions[dimension];
 
-  if ( !rhs._grid->CheckerBoarded(dimension) ) {
+  if ( !rhs.Grid()->CheckerBoarded(dimension) ) {
     cbmask=0x3;
   }
 
-  int ro  = rplane*rhs._grid->_ostride[dimension]; // base offset for start of plane 
-  int lo  = lplane*lhs._grid->_ostride[dimension]; // base offset for start of plane 
+  int ro  = rplane*rhs.Grid()->_ostride[dimension]; // base offset for start of plane 
+  int lo  = lplane*lhs.Grid()->_ostride[dimension]; // base offset for start of plane 
 
-  int e1=rhs._grid->_slice_nblock[dimension];
-  int e2=rhs._grid->_slice_block [dimension];
-  int stride = rhs._grid->_slice_stride[dimension];
+  int e1=rhs.Grid()->_slice_nblock[dimension];
+  int e2=rhs.Grid()->_slice_block [dimension];
+  int stride = rhs.Grid()->_slice_stride[dimension];
 
   static std::vector<std::pair<int,int> > table;  table.resize(e1*e2);
   int ent=0;
 
-  double t_tab,t_perm;
   if ( cbmask == 0x3 ) {
     for(int n=0;n<e1;n++){
     for(int b=0;b<e2;b++){
@@ -292,14 +298,16 @@ template<class vobj> void Copy_plane_permute(Lattice<vobj>& lhs,const Lattice<vo
     for(int n=0;n<e1;n++){
     for(int b=0;b<e2;b++){
       int o  =n*stride;
-      int ocb=1<<lhs._grid->CheckerBoardFromOindex(o+b);
+      int ocb=1<<lhs.Grid()->CheckerBoardFromOindex(o+b);
       if ( ocb&cbmask ) table[ent++] = std::pair<int,int>(lo+o+b,ro+o+b);
     }}
   }
 
-  parallel_for(int i=0;i<ent;i++){
-    permute(lhs._odata[table[i].first],rhs._odata[table[i].second],permute_type);
-  }
+  auto rhs_v = rhs.View();
+  auto lhs_v = lhs.View();
+  thread_for(i,ent,{
+    permute(lhs_v[table[i].first],rhs_v[table[i].second],permute_type);
+  });
 }
 
 //////////////////////////////////////////////////////
@@ -309,11 +317,9 @@ template<class vobj> void Cshift_local(Lattice<vobj>& ret,const Lattice<vobj> &r
 {
   int sshift[2];
 
-  sshift[0] = rhs._grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,Even);
-  sshift[1] = rhs._grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,Odd);
+  sshift[0] = rhs.Grid()->CheckerBoardShiftForCB(rhs.Checkerboard(),dimension,shift,Even);
+  sshift[1] = rhs.Grid()->CheckerBoardShiftForCB(rhs.Checkerboard(),dimension,shift,Odd);
 
-  double t_local;
-  
   if ( sshift[0] == sshift[1] ) {
     Cshift_local(ret,rhs,dimension,shift,0x3);
   } else {
@@ -324,7 +330,7 @@ template<class vobj> void Cshift_local(Lattice<vobj>& ret,const Lattice<vobj> &r
 
 template<class vobj> void Cshift_local(Lattice<vobj> &ret,const Lattice<vobj> &rhs,int dimension,int shift,int cbmask)
 {
-  GridBase *grid = rhs._grid;
+  GridBase *grid = rhs.Grid();
   int fd = grid->_fdimensions[dimension];
   int rd = grid->_rdimensions[dimension];
   int ld = grid->_ldimensions[dimension];
@@ -335,18 +341,18 @@ template<class vobj> void Cshift_local(Lattice<vobj> &ret,const Lattice<vobj> &r
   shift = (shift+fd)%fd;
 
   // the permute type
-  ret.checkerboard = grid->CheckerBoardDestination(rhs.checkerboard,shift,dimension);
+  ret.Checkerboard() = grid->CheckerBoardDestination(rhs.Checkerboard(),shift,dimension);
   int permute_dim =grid->PermuteDim(dimension);
   int permute_type=grid->PermuteType(dimension);
   int permute_type_dist;
 
   for(int x=0;x<rd;x++){       
 
-    int o   = 0;
+    //    int o   = 0;
     int bo  = x * grid->_ostride[dimension];
     int cb= (cbmask==0x2)? Odd : Even;
 
-    int sshift = grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,cb);
+    int sshift = grid->CheckerBoardShiftForCB(rhs.Checkerboard(),dimension,shift,cb);
     int sx     = (x+sshift)%rd;
     
     // wrap is whether sshift > rd.
@@ -387,5 +393,5 @@ template<class vobj> void Cshift_local(Lattice<vobj> &ret,const Lattice<vobj> &r
   
   }
 }
-}
-#endif
+NAMESPACE_END(Grid);
+
diff --git a/Grid/cshift/Cshift_mpi.h b/Grid/cshift/Cshift_mpi.h
index 98972135..0f0e80b1 100644
--- a/Grid/cshift/Cshift_mpi.h
+++ b/Grid/cshift/Cshift_mpi.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,33 +24,33 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef _GRID_CSHIFT_MPI_H_
 #define _GRID_CSHIFT_MPI_H_
 
 
-namespace Grid { 
+NAMESPACE_BEGIN(Grid); 
 
 template<class vobj> Lattice<vobj> Cshift(const Lattice<vobj> &rhs,int dimension,int shift)
 {
   typedef typename vobj::vector_type vector_type;
   typedef typename vobj::scalar_type scalar_type;
 
-  Lattice<vobj> ret(rhs._grid); 
+  Lattice<vobj> ret(rhs.Grid()); 
   
-  int fd = rhs._grid->_fdimensions[dimension];
-  int rd = rhs._grid->_rdimensions[dimension];
+  int fd = rhs.Grid()->_fdimensions[dimension];
+  int rd = rhs.Grid()->_rdimensions[dimension];
 
   // Map to always positive shift modulo global full dimension.
   shift = (shift+fd)%fd;
 
-  ret.checkerboard = rhs._grid->CheckerBoardDestination(rhs.checkerboard,shift,dimension);
+  ret.Checkerboard() = rhs.Grid()->CheckerBoardDestination(rhs.Checkerboard(),shift,dimension);
         
   // the permute type
-  int simd_layout     = rhs._grid->_simd_layout[dimension];
-  int comm_dim        = rhs._grid->_processors[dimension] >1 ;
-  int splice_dim      = rhs._grid->_simd_layout[dimension]>1 && (comm_dim);
+  int simd_layout     = rhs.Grid()->_simd_layout[dimension];
+  int comm_dim        = rhs.Grid()->_processors[dimension] >1 ;
+  int splice_dim      = rhs.Grid()->_simd_layout[dimension]>1 && (comm_dim);
 
 
   if ( !comm_dim ) {
@@ -70,10 +70,10 @@ template<class vobj> void Cshift_comms(Lattice<vobj>& ret,const Lattice<vobj> &r
 {
   int sshift[2];
 
-  sshift[0] = rhs._grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,Even);
-  sshift[1] = rhs._grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,Odd);
+  sshift[0] = rhs.Grid()->CheckerBoardShiftForCB(rhs.Checkerboard(),dimension,shift,Even);
+  sshift[1] = rhs.Grid()->CheckerBoardShiftForCB(rhs.Checkerboard(),dimension,shift,Odd);
 
-  //  std::cout << "Cshift_comms dim "<<dimension<<"cb "<<rhs.checkerboard<<"shift "<<shift<<" sshift " << sshift[0]<<" "<<sshift[1]<<std::endl;
+  //  std::cout << "Cshift_comms dim "<<dimension<<"cb "<<rhs.Checkerboard()<<"shift "<<shift<<" sshift " << sshift[0]<<" "<<sshift[1]<<std::endl;
   if ( sshift[0] == sshift[1] ) {
     //    std::cout << "Single pass Cshift_comms" <<std::endl;
     Cshift_comms(ret,rhs,dimension,shift,0x3);
@@ -88,8 +88,8 @@ template<class vobj> void Cshift_comms_simd(Lattice<vobj>& ret,const Lattice<vob
 {
   int sshift[2];
 
-  sshift[0] = rhs._grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,Even);
-  sshift[1] = rhs._grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,Odd);
+  sshift[0] = rhs.Grid()->CheckerBoardShiftForCB(rhs.Checkerboard(),dimension,shift,Even);
+  sshift[1] = rhs.Grid()->CheckerBoardShiftForCB(rhs.Checkerboard(),dimension,shift,Odd);
 
   //std::cout << "Cshift_comms_simd dim "<<dimension<<"cb "<<rhs.checkerboard<<"shift "<<shift<<" sshift " << sshift[0]<<" "<<sshift[1]<<std::endl;
   if ( sshift[0] == sshift[1] ) {
@@ -107,25 +107,25 @@ template<class vobj> void Cshift_comms(Lattice<vobj> &ret,const Lattice<vobj> &r
   typedef typename vobj::vector_type vector_type;
   typedef typename vobj::scalar_type scalar_type;
 
-  GridBase *grid=rhs._grid;
-  Lattice<vobj> temp(rhs._grid);
+  GridBase *grid=rhs.Grid();
+  Lattice<vobj> temp(rhs.Grid());
 
-  int fd              = rhs._grid->_fdimensions[dimension];
-  int rd              = rhs._grid->_rdimensions[dimension];
-  int pd              = rhs._grid->_processors[dimension];
-  int simd_layout     = rhs._grid->_simd_layout[dimension];
-  int comm_dim        = rhs._grid->_processors[dimension] >1 ;
+  int fd              = rhs.Grid()->_fdimensions[dimension];
+  int rd              = rhs.Grid()->_rdimensions[dimension];
+  int pd              = rhs.Grid()->_processors[dimension];
+  int simd_layout     = rhs.Grid()->_simd_layout[dimension];
+  int comm_dim        = rhs.Grid()->_processors[dimension] >1 ;
   assert(simd_layout==1);
   assert(comm_dim==1);
   assert(shift>=0);
   assert(shift<fd);
   
-  int buffer_size = rhs._grid->_slice_nblock[dimension]*rhs._grid->_slice_block[dimension];
+  int buffer_size = rhs.Grid()->_slice_nblock[dimension]*rhs.Grid()->_slice_block[dimension];
   commVector<vobj> send_buf(buffer_size);
   commVector<vobj> recv_buf(buffer_size);
 
   int cb= (cbmask==0x2)? Odd : Even;
-  int sshift= rhs._grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,cb);
+  int sshift= rhs.Grid()->CheckerBoardShiftForCB(rhs.Checkerboard(),dimension,shift,cb);
 
   for(int x=0;x<rd;x++){       
 
@@ -145,7 +145,7 @@ template<class vobj> void Cshift_comms(Lattice<vobj> &ret,const Lattice<vobj> &r
 
       Gather_plane_simple (rhs,send_buf,dimension,sx,cbmask);
 
-      int rank           = grid->_processor;
+      //      int rank           = grid->_processor;
       int recv_from_rank;
       int xmit_to_rank;
       grid->ShiftedRanks(dimension,comm_proc,xmit_to_rank,recv_from_rank);
@@ -165,7 +165,7 @@ template<class vobj> void Cshift_comms(Lattice<vobj> &ret,const Lattice<vobj> &r
 
 template<class vobj> void  Cshift_comms_simd(Lattice<vobj> &ret,const Lattice<vobj> &rhs,int dimension,int shift,int cbmask)
 {
-  GridBase *grid=rhs._grid;
+  GridBase *grid=rhs.Grid();
   const int Nsimd = grid->Nsimd();
   typedef typename vobj::vector_type vector_type;
   typedef typename vobj::scalar_object scalar_object;
@@ -193,21 +193,21 @@ template<class vobj> void  Cshift_comms_simd(Lattice<vobj> &ret,const Lattice<vo
   // Simd direction uses an extract/merge pair
   ///////////////////////////////////////////////
   int buffer_size = grid->_slice_nblock[dimension]*grid->_slice_block[dimension];
-  int words = sizeof(vobj)/sizeof(vector_type);
+  //  int words = sizeof(vobj)/sizeof(vector_type);
 
   std::vector<commVector<scalar_object> >   send_buf_extract(Nsimd,commVector<scalar_object>(buffer_size) );
   std::vector<commVector<scalar_object> >   recv_buf_extract(Nsimd,commVector<scalar_object>(buffer_size) );
 
   int bytes = buffer_size*sizeof(scalar_object);
 
-  std::vector<scalar_object *>  pointers(Nsimd); // 
-  std::vector<scalar_object *> rpointers(Nsimd); // received pointers
+  ExtractPointerArray<scalar_object>  pointers(Nsimd); // 
+  ExtractPointerArray<scalar_object> rpointers(Nsimd); // received pointers
 
   ///////////////////////////////////////////
   // Work out what to send where
   ///////////////////////////////////////////
   int cb    = (cbmask==0x2)? Odd : Even;
-  int sshift= grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,cb);
+  int sshift= grid->CheckerBoardShiftForCB(rhs.Checkerboard(),dimension,shift,cb);
 
   // loop over outer coord planes orthog to dim
   for(int x=0;x<rd;x++){       
@@ -257,6 +257,8 @@ template<class vobj> void  Cshift_comms_simd(Lattice<vobj> &ret,const Lattice<vo
     Scatter_plane_merge(ret,rpointers,dimension,x,cbmask);
   }
 
- }
 }
+
+NAMESPACE_END(Grid); 
+
 #endif
diff --git a/Grid/cshift/Cshift_none.h b/Grid/cshift/Cshift_none.h
index cdc2563e..d3c0bfa0 100644
--- a/Grid/cshift/Cshift_none.h
+++ b/Grid/cshift/Cshift_none.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,17 +23,18 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef _GRID_CSHIFT_NONE_H_
 #define _GRID_CSHIFT_NONE_H_
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 template<class vobj> Lattice<vobj> Cshift(const Lattice<vobj> &rhs,int dimension,int shift)
 {
-  Lattice<vobj> ret(rhs._grid);
-  ret.checkerboard = rhs._grid->CheckerBoardDestination(rhs.checkerboard,shift,dimension);
+  Lattice<vobj> ret(rhs.Grid());
+  ret.Checkerboard() = rhs.Grid()->CheckerBoardDestination(rhs.Checkerboard(),shift,dimension);
   Cshift_local(ret,rhs,dimension,shift);
   return ret;
 }
-}
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/json/json.hpp b/Grid/json/json.hpp
index c8b0cc9e..618aa7a1 100644
--- a/Grid/json/json.hpp
+++ b/Grid/json/json.hpp
@@ -1,3 +1,4 @@
+#ifndef __NVCC__
 /*
     __ _____ _____ _____
  __|  |   __|     |   | |  JSON for Modern C++
@@ -18918,3 +18919,4 @@ inline nlohmann::json::json_pointer operator "" _json_pointer(const char* s, std
 
 
 #endif
+#endif
diff --git a/Grid/lattice/Lattice.h b/Grid/lattice/Lattice.h
index e2bb2a82..1eea98ed 100644
--- a/Grid/lattice/Lattice.h
+++ b/Grid/lattice/Lattice.h
@@ -25,9 +25,22 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     See the full license in the file "LICENSE" in the top level distribution directory
     *************************************************************************************/
     /*  END LEGAL */
-#ifndef GRID_LATTICE_H
-#define GRID_LATTICE_H
-
+#pragma once
 #include <Grid/lattice/Lattice_base.h>
+#include <Grid/lattice/Lattice_conformable.h>
+#include <Grid/lattice/Lattice_ET.h>
+#include <Grid/lattice/Lattice_arith.h>
+#include <Grid/lattice/Lattice_trace.h>
+#include <Grid/lattice/Lattice_transpose.h>
+#include <Grid/lattice/Lattice_local.h>
+#include <Grid/lattice/Lattice_reduction.h>
+#include <Grid/lattice/Lattice_peekpoke.h>
+#include <Grid/lattice/Lattice_reality.h>
+#include <Grid/lattice/Lattice_comparison_utils.h>
+#include <Grid/lattice/Lattice_comparison.h>
+#include <Grid/lattice/Lattice_coordinate.h>
+//#include <Grid/lattice/Lattice_where.h>
+#include <Grid/lattice/Lattice_rng.h>
+#include <Grid/lattice/Lattice_unary.h>
+#include <Grid/lattice/Lattice_transfer.h>
 
-#endif
diff --git a/Grid/lattice/Lattice_ET.h b/Grid/lattice/Lattice_ET.h
index 1bb83901..cf7147b9 100644
--- a/Grid/lattice/Lattice_ET.h
+++ b/Grid/lattice/Lattice_ET.h
@@ -27,7 +27,7 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef GRID_LATTICE_ET_H
 #define GRID_LATTICE_ET_H
 
@@ -36,13 +36,13 @@ directory
 #include <typeinfo>
 #include <vector>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 ////////////////////////////////////////////////////
 // Predicated where support
 ////////////////////////////////////////////////////
 template <class iobj, class vobj, class robj>
-inline vobj predicatedWhere(const iobj &predicate, const vobj &iftrue,
+accelerator_inline vobj predicatedWhere(const iobj &predicate, const vobj &iftrue,
                             const robj &iffalse) {
   typename std::remove_const<vobj>::type ret;
 
@@ -51,11 +51,10 @@ inline vobj predicatedWhere(const iobj &predicate, const vobj &iftrue,
   typedef typename vobj::vector_type vector_type;
 
   const int Nsimd = vobj::vector_type::Nsimd();
-  const int words = sizeof(vobj) / sizeof(vector_type);
 
-  std::vector<Integer> mask(Nsimd);
-  std::vector<scalar_object> truevals(Nsimd);
-  std::vector<scalar_object> falsevals(Nsimd);
+  ExtractBuffer<Integer> mask(Nsimd);
+  ExtractBuffer<scalar_object> truevals(Nsimd);
+  ExtractBuffer<scalar_object> falsevals(Nsimd);
 
   extract(iftrue, truevals);
   extract(iffalse, falsevals);
@@ -69,158 +68,148 @@ inline vobj predicatedWhere(const iobj &predicate, const vobj &iftrue,
   return ret;
 }
 
-////////////////////////////////////////////
-// recursive evaluation of expressions; Could
-// switch to generic approach with variadics, a la
-// Antonin's Lat Sim but the repack to variadic with popped
-// from tuple is hideous; C++14 introduces std::make_index_sequence for this
-////////////////////////////////////////////
-
-// leaf eval of lattice ; should enable if protect using traits
-
-template <typename T>
-using is_lattice = std::is_base_of<LatticeBase, T>;
-
-template <typename T>
-using is_lattice_expr = std::is_base_of<LatticeExpressionBase, T>;
-
-template <typename T> using is_lattice_expr = std::is_base_of<LatticeExpressionBase,T >;
-
+/////////////////////////////////////////////////////
 //Specialization of getVectorType for lattices
+/////////////////////////////////////////////////////
 template<typename T>
 struct getVectorType<Lattice<T> >{
   typedef typename Lattice<T>::vector_object type;
 };
- 
-template<class sobj>
-inline sobj eval(const unsigned int ss, const sobj &arg)
+
+////////////////////////////////////////////
+//--  recursive evaluation of expressions; --
+// handle leaves of syntax tree
+///////////////////////////////////////////////////
+template<class sobj> accelerator_inline 
+sobj eval(const uint64_t ss, const sobj &arg)
 {
   return arg;
 }
-template <class lobj>
-inline const lobj &eval(const unsigned int ss, const Lattice<lobj> &arg) {
-  return arg._odata[ss];
+
+template <class lobj> accelerator_inline 
+const lobj & eval(const uint64_t ss, const LatticeView<lobj> &arg) 
+{
+  return arg[ss];
+}
+template <class lobj> accelerator_inline 
+const lobj & eval(const uint64_t ss, const Lattice<lobj> &arg) 
+{
+  auto view = arg.View();
+  return view[ss];
 }
 
-// handle nodes in syntax tree
-template <typename Op, typename T1>
-auto inline eval(
-    const unsigned int ss,
-    const LatticeUnaryExpression<Op, T1> &expr)  // eval one operand
-    -> decltype(expr.first.func(eval(ss, std::get<0>(expr.second)))) {
-  return expr.first.func(eval(ss, std::get<0>(expr.second)));
+///////////////////////////////////////////////////
+// handle nodes in syntax tree- eval one operand
+///////////////////////////////////////////////////
+template <typename Op, typename T1> accelerator_inline 
+auto eval(const uint64_t ss, const LatticeUnaryExpression<Op, T1> &expr)  
+  -> decltype(expr.op.func( eval(ss, expr.arg1)))
+{
+  return expr.op.func( eval(ss, expr.arg1) );
 }
-
-template <typename Op, typename T1, typename T2>
-auto inline eval(
-    const unsigned int ss,
-    const LatticeBinaryExpression<Op, T1, T2> &expr)  // eval two operands
-    -> decltype(expr.first.func(eval(ss, std::get<0>(expr.second)),
-                                eval(ss, std::get<1>(expr.second)))) {
-  return expr.first.func(eval(ss, std::get<0>(expr.second)),
-                         eval(ss, std::get<1>(expr.second)));
+///////////////////////
+// eval two operands
+///////////////////////
+template <typename Op, typename T1, typename T2> accelerator_inline
+auto eval(const uint64_t ss, const LatticeBinaryExpression<Op, T1, T2> &expr)  
+  -> decltype(expr.op.func( eval(ss,expr.arg1),eval(ss,expr.arg2)))
+{
+  return expr.op.func( eval(ss,expr.arg1), eval(ss,expr.arg2) );
 }
-
-template <typename Op, typename T1, typename T2, typename T3>
-auto inline eval(const unsigned int ss,
-                 const LatticeTrinaryExpression<Op, T1, T2, T3>
-                     &expr)  // eval three operands
-    -> decltype(expr.first.func(eval(ss, std::get<0>(expr.second)),
-                                eval(ss, std::get<1>(expr.second)),
-                                eval(ss, std::get<2>(expr.second)))) {
-  return expr.first.func(eval(ss, std::get<0>(expr.second)),
-                         eval(ss, std::get<1>(expr.second)),
-                         eval(ss, std::get<2>(expr.second)));
+///////////////////////
+// eval three operands
+///////////////////////
+template <typename Op, typename T1, typename T2, typename T3> accelerator_inline
+auto eval(const uint64_t ss, const LatticeTrinaryExpression<Op, T1, T2, T3> &expr)  
+  -> decltype(expr.op.func(eval(ss, expr.arg1), eval(ss, expr.arg2), eval(ss, expr.arg3)))
+{
+  return expr.op.func(eval(ss, expr.arg1), eval(ss, expr.arg2), eval(ss, expr.arg3));
 }
 
 //////////////////////////////////////////////////////////////////////////
 // Obtain the grid from an expression, ensuring conformable. This must follow a
-// tree recursion
+// tree recursion; must retain grid pointer in the LatticeView class which sucks
+// Use a different method, and make it void *.
+// Perhaps a conformable method.
 //////////////////////////////////////////////////////////////////////////
-template <class T1,
-          typename std::enable_if<is_lattice<T1>::value, T1>::type * = nullptr>
-inline void GridFromExpression(GridBase *&grid, const T1 &lat)  // Lattice leaf
+template <class T1,typename std::enable_if<is_lattice<T1>::value, T1>::type * = nullptr>
+accelerator_inline void GridFromExpression(GridBase *&grid, const T1 &lat)  // Lattice leaf
 {
-  if (grid) {
-    conformable(grid, lat._grid);
-  }
-  grid = lat._grid;
+  lat.Conformable(grid);
 }
-template <class T1,
-          typename std::enable_if<!is_lattice<T1>::value, T1>::type * = nullptr>
-inline void GridFromExpression(GridBase *&grid,
-                               const T1 &notlat)  // non-lattice leaf
+
+template <class T1,typename std::enable_if<!is_lattice<T1>::value, T1>::type * = nullptr>
+accelerator_inline 
+void GridFromExpression(GridBase *&grid,const T1 &notlat)  // non-lattice leaf
 {}
+
 template <typename Op, typename T1>
-inline void GridFromExpression(GridBase *&grid,
-                               const LatticeUnaryExpression<Op, T1> &expr) {
-  GridFromExpression(grid, std::get<0>(expr.second));  // recurse
+accelerator_inline 
+void GridFromExpression(GridBase *&grid,const LatticeUnaryExpression<Op, T1> &expr) 
+{
+  GridFromExpression(grid, expr.arg1);  // recurse
 }
 
 template <typename Op, typename T1, typename T2>
-inline void GridFromExpression(
-    GridBase *&grid, const LatticeBinaryExpression<Op, T1, T2> &expr) {
-  GridFromExpression(grid, std::get<0>(expr.second));  // recurse
-  GridFromExpression(grid, std::get<1>(expr.second));
+accelerator_inline 
+void GridFromExpression(GridBase *&grid, const LatticeBinaryExpression<Op, T1, T2> &expr) 
+{
+  GridFromExpression(grid, expr.arg1);  // recurse
+  GridFromExpression(grid, expr.arg2);
 }
 template <typename Op, typename T1, typename T2, typename T3>
-inline void GridFromExpression(
-    GridBase *&grid, const LatticeTrinaryExpression<Op, T1, T2, T3> &expr) {
-  GridFromExpression(grid, std::get<0>(expr.second));  // recurse
-  GridFromExpression(grid, std::get<1>(expr.second));
-  GridFromExpression(grid, std::get<2>(expr.second));
+accelerator_inline 
+void GridFromExpression(GridBase *&grid, const LatticeTrinaryExpression<Op, T1, T2, T3> &expr) 
+{
+  GridFromExpression(grid, expr.arg1);  // recurse
+  GridFromExpression(grid, expr.arg2);  // recurse
+  GridFromExpression(grid, expr.arg3);  // recurse
 }
 
 //////////////////////////////////////////////////////////////////////////
 // Obtain the CB from an expression, ensuring conformable. This must follow a
 // tree recursion
 //////////////////////////////////////////////////////////////////////////
-template <class T1,
-          typename std::enable_if<is_lattice<T1>::value, T1>::type * = nullptr>
+template <class T1,typename std::enable_if<is_lattice<T1>::value, T1>::type * = nullptr>
 inline void CBFromExpression(int &cb, const T1 &lat)  // Lattice leaf
 {
   if ((cb == Odd) || (cb == Even)) {
-    assert(cb == lat.checkerboard);
+    assert(cb == lat.Checkerboard());
   }
-  cb = lat.checkerboard;
-  //  std::cout<<GridLogMessage<<"Lattice leaf cb "<<cb<<std::endl;
+  cb = lat.Checkerboard();
 }
-template <class T1,
-          typename std::enable_if<!is_lattice<T1>::value, T1>::type * = nullptr>
+template <class T1,typename std::enable_if<!is_lattice<T1>::value, T1>::type * = nullptr>
 inline void CBFromExpression(int &cb, const T1 &notlat)  // non-lattice leaf
 {
-  //  std::cout<<GridLogMessage<<"Non lattice leaf cb"<<cb<<std::endl;
-}
-template <typename Op, typename T1>
-inline void CBFromExpression(int &cb,
-                             const LatticeUnaryExpression<Op, T1> &expr) {
-  CBFromExpression(cb, std::get<0>(expr.second));  // recurse
-  //  std::cout<<GridLogMessage<<"Unary node cb "<<cb<<std::endl;
 }
 
-template <typename Op, typename T1, typename T2>
-inline void CBFromExpression(int &cb,
-                             const LatticeBinaryExpression<Op, T1, T2> &expr) {
-  CBFromExpression(cb, std::get<0>(expr.second));  // recurse
-  CBFromExpression(cb, std::get<1>(expr.second));
-  //  std::cout<<GridLogMessage<<"Binary node cb "<<cb<<std::endl;
+template <typename Op, typename T1> inline 
+void CBFromExpression(int &cb,const LatticeUnaryExpression<Op, T1> &expr) 
+{
+  CBFromExpression(cb, expr.arg1);  // recurse AST
+}
+
+template <typename Op, typename T1, typename T2> inline 
+void CBFromExpression(int &cb,const LatticeBinaryExpression<Op, T1, T2> &expr) 
+{
+  CBFromExpression(cb, expr.arg1);  // recurse AST
+  CBFromExpression(cb, expr.arg2);  // recurse AST
 }
 template <typename Op, typename T1, typename T2, typename T3>
-inline void CBFromExpression(
-    int &cb, const LatticeTrinaryExpression<Op, T1, T2, T3> &expr) {
-  CBFromExpression(cb, std::get<0>(expr.second));  // recurse
-  CBFromExpression(cb, std::get<1>(expr.second));
-  CBFromExpression(cb, std::get<2>(expr.second));
-  //  std::cout<<GridLogMessage<<"Trinary node cb "<<cb<<std::endl;
+inline void CBFromExpression(int &cb, const LatticeTrinaryExpression<Op, T1, T2, T3> &expr) 
+{
+  CBFromExpression(cb, expr.arg1);  // recurse AST
+  CBFromExpression(cb, expr.arg2);  // recurse AST
+  CBFromExpression(cb, expr.arg3);  // recurse AST
 }
 
 ////////////////////////////////////////////
 // Unary operators and funcs
 ////////////////////////////////////////////
-#define GridUnopClass(name, ret)                                          \
-  template <class arg>                                                    \
-  struct name {                                                           \
-    static auto inline func(const arg a) -> decltype(ret) { return ret; } \
+#define GridUnopClass(name, ret)					\
+  template <class arg>							\
+  struct name {								\
+    static auto accelerator_inline func(const arg a) -> decltype(ret) { return ret; } \
   };
 
 GridUnopClass(UnarySub, -a);
@@ -250,19 +239,21 @@ GridUnopClass(UnaryExp, exp(a));
 ////////////////////////////////////////////
 // Binary operators
 ////////////////////////////////////////////
-#define GridBinOpClass(name, combination)                      \
-  template <class left, class right>                           \
-  struct name {                                                \
-    static auto inline func(const left &lhs, const right &rhs) \
-        -> decltype(combination) const {                       \
-      return combination;                                      \
-    }                                                          \
-  }
+#define GridBinOpClass(name, combination)			\
+  template <class left, class right>				\
+  struct name {							\
+    static auto accelerator_inline				\
+    func(const left &lhs, const right &rhs)			\
+      -> decltype(combination) const				\
+    {								\
+      return combination;					\
+    }								\
+  };
+
 GridBinOpClass(BinaryAdd, lhs + rhs);
 GridBinOpClass(BinarySub, lhs - rhs);
 GridBinOpClass(BinaryMul, lhs *rhs);
 GridBinOpClass(BinaryDiv, lhs /rhs);
-
 GridBinOpClass(BinaryAnd, lhs &rhs);
 GridBinOpClass(BinaryOr, lhs | rhs);
 GridBinOpClass(BinaryAndAnd, lhs &&rhs);
@@ -271,92 +262,71 @@ GridBinOpClass(BinaryOrOr, lhs || rhs);
 ////////////////////////////////////////////////////
 // Trinary conditional op
 ////////////////////////////////////////////////////
-#define GridTrinOpClass(name, combination)                                     \
-  template <class predicate, class left, class right>                          \
-  struct name {                                                                \
-    static auto inline func(const predicate &pred, const left &lhs,            \
-                            const right &rhs) -> decltype(combination) const { \
-      return combination;                                                      \
-    }                                                                          \
-  }
+#define GridTrinOpClass(name, combination)				\
+  template <class predicate, class left, class right>			\
+  struct name {								\
+    static auto accelerator_inline					\
+    func(const predicate &pred, const left &lhs, const right &rhs)	\
+      -> decltype(combination) const					\
+    {									\
+      return combination;						\
+    }									\
+  };
 
-GridTrinOpClass(
-    TrinaryWhere,
-    (predicatedWhere<predicate, typename std::remove_reference<left>::type,
-                     typename std::remove_reference<right>::type>(pred, lhs,
-                                                                  rhs)));
+GridTrinOpClass(TrinaryWhere,
+		(predicatedWhere<predicate, 
+		 typename std::remove_reference<left>::type,
+		 typename std::remove_reference<right>::type>(pred, lhs,rhs)));
 
 ////////////////////////////////////////////
 // Operator syntactical glue
 ////////////////////////////////////////////
 
-#define GRID_UNOP(name) name<decltype(eval(0, arg))>
-#define GRID_BINOP(name) name<decltype(eval(0, lhs)), decltype(eval(0, rhs))>
-#define GRID_TRINOP(name) \
-  name<decltype(eval(0, pred)), decltype(eval(0, lhs)), decltype(eval(0, rhs))>
+#define GRID_UNOP(name)   name<decltype(eval(0, arg))>
+#define GRID_BINOP(name)  name<decltype(eval(0, lhs)), decltype(eval(0, rhs))>
+#define GRID_TRINOP(name) name<decltype(eval(0, pred)), decltype(eval(0, lhs)), decltype(eval(0, rhs))>
 
-#define GRID_DEF_UNOP(op, name)                                             \
-  template <typename T1,                                                    \
-            typename std::enable_if<is_lattice<T1>::value ||                \
-                                        is_lattice_expr<T1>::value,         \
-                                    T1>::type * = nullptr>                  \
-  inline auto op(const T1 &arg)                                             \
-      ->decltype(LatticeUnaryExpression<GRID_UNOP(name), const T1 &>(       \
-          std::make_pair(GRID_UNOP(name)(), std::forward_as_tuple(arg)))) { \
-    return LatticeUnaryExpression<GRID_UNOP(name), const T1 &>(             \
-        std::make_pair(GRID_UNOP(name)(), std::forward_as_tuple(arg)));     \
+#define GRID_DEF_UNOP(op, name)						\
+  template <typename T1, typename std::enable_if<is_lattice<T1>::value||is_lattice_expr<T1>::value,T1>::type * = nullptr> \
+  inline auto op(const T1 &arg) ->decltype(LatticeUnaryExpression<GRID_UNOP(name),T1>(GRID_UNOP(name)(), arg)) \
+  {									\
+    return     LatticeUnaryExpression<GRID_UNOP(name),T1>(GRID_UNOP(name)(), arg); \
   }
 
-#define GRID_BINOP_LEFT(op, name)                                             \
-  template <typename T1, typename T2,                                         \
-            typename std::enable_if<is_lattice<T1>::value ||                  \
-                                        is_lattice_expr<T1>::value,           \
-                                    T1>::type * = nullptr>                    \
-  inline auto op(const T1 &lhs, const T2 &rhs)                                \
-      ->decltype(                                                             \
-          LatticeBinaryExpression<GRID_BINOP(name), const T1 &, const T2 &>(  \
-              std::make_pair(GRID_BINOP(name)(),                              \
-                             std::forward_as_tuple(lhs, rhs)))) {             \
-    return LatticeBinaryExpression<GRID_BINOP(name), const T1 &, const T2 &>( \
-        std::make_pair(GRID_BINOP(name)(), std::forward_as_tuple(lhs, rhs))); \
+#define GRID_BINOP_LEFT(op, name)					\
+  template <typename T1, typename T2,					\
+            typename std::enable_if<is_lattice<T1>::value||is_lattice_expr<T1>::value,T1>::type * = nullptr> \
+  inline auto op(const T1 &lhs, const T2 &rhs)				\
+    ->decltype(LatticeBinaryExpression<GRID_BINOP(name),T1,T2>(GRID_BINOP(name)(),lhs,rhs)) \
+  {									\
+    return     LatticeBinaryExpression<GRID_BINOP(name),T1,T2>(GRID_BINOP(name)(),lhs,rhs);\
   }
 
-#define GRID_BINOP_RIGHT(op, name)                                            \
-  template <typename T1, typename T2,                                         \
-            typename std::enable_if<!is_lattice<T1>::value &&                 \
-                                        !is_lattice_expr<T1>::value,          \
-                                    T1>::type * = nullptr,                    \
-            typename std::enable_if<is_lattice<T2>::value ||                  \
-                                        is_lattice_expr<T2>::value,           \
-                                    T2>::type * = nullptr>                    \
-  inline auto op(const T1 &lhs, const T2 &rhs)                                \
-      ->decltype(                                                             \
-          LatticeBinaryExpression<GRID_BINOP(name), const T1 &, const T2 &>(  \
-              std::make_pair(GRID_BINOP(name)(),                              \
-                             std::forward_as_tuple(lhs, rhs)))) {             \
-    return LatticeBinaryExpression<GRID_BINOP(name), const T1 &, const T2 &>( \
-        std::make_pair(GRID_BINOP(name)(), std::forward_as_tuple(lhs, rhs))); \
+#define GRID_BINOP_RIGHT(op, name)					\
+  template <typename T1, typename T2,					\
+            typename std::enable_if<!is_lattice<T1>::value&&!is_lattice_expr<T1>::value,T1>::type * = nullptr, \
+            typename std::enable_if< is_lattice<T2>::value|| is_lattice_expr<T2>::value,T2>::type * = nullptr> \
+  inline auto op(const T1 &lhs, const T2 &rhs)				\
+    ->decltype(LatticeBinaryExpression<GRID_BINOP(name),T1,T2>(GRID_BINOP(name)(),lhs, rhs)) \
+  {									\
+    return     LatticeBinaryExpression<GRID_BINOP(name),T1,T2>(GRID_BINOP(name)(),lhs, rhs); \
   }
 
-#define GRID_DEF_BINOP(op, name) \
-  GRID_BINOP_LEFT(op, name);     \
+#define GRID_DEF_BINOP(op, name)		\
+  GRID_BINOP_LEFT(op, name);			\
   GRID_BINOP_RIGHT(op, name);
 
-#define GRID_DEF_TRINOP(op, name)                                              \
-  template <typename T1, typename T2, typename T3>                             \
-  inline auto op(const T1 &pred, const T2 &lhs, const T3 &rhs)                 \
-      ->decltype(                                                              \
-          LatticeTrinaryExpression<GRID_TRINOP(name), const T1 &, const T2 &,  \
-                                   const T3 &>(std::make_pair(                 \
-              GRID_TRINOP(name)(), std::forward_as_tuple(pred, lhs, rhs)))) {  \
-    return LatticeTrinaryExpression<GRID_TRINOP(name), const T1 &, const T2 &, \
-                                    const T3 &>(std::make_pair(                \
-        GRID_TRINOP(name)(), std::forward_as_tuple(pred, lhs, rhs)));          \
+#define GRID_DEF_TRINOP(op, name)					\
+  template <typename T1, typename T2, typename T3>			\
+  inline auto op(const T1 &pred, const T2 &lhs, const T3 &rhs)		\
+    ->decltype(LatticeTrinaryExpression<GRID_TRINOP(name),T1,T2,T3>(GRID_TRINOP(name)(),pred, lhs, rhs)) \
+  {									\
+    return LatticeTrinaryExpression<GRID_TRINOP(name),T1,T2,T3>(GRID_TRINOP(name)(),pred, lhs, rhs); \
   }
+
 ////////////////////////
 // Operator definitions
 ////////////////////////
-
 GRID_DEF_UNOP(operator-, UnarySub);
 GRID_DEF_UNOP(Not, UnaryNot);
 GRID_DEF_UNOP(operator!, UnaryNot);
@@ -400,29 +370,27 @@ GRID_DEF_TRINOP(where, TrinaryWhere);
 /////////////////////////////////////////////////////////////
 template <class Op, class T1>
 auto closure(const LatticeUnaryExpression<Op, T1> &expr)
-    -> Lattice<decltype(expr.first.func(eval(0, std::get<0>(expr.second))))> {
-  Lattice<decltype(expr.first.func(eval(0, std::get<0>(expr.second))))> ret(
-      expr);
+  -> Lattice<decltype(expr.op.func(eval(0, expr.arg1)))> 
+{
+  Lattice<decltype(expr.op.func(eval(0, expr.arg1)))> ret(expr);
   return ret;
 }
 template <class Op, class T1, class T2>
 auto closure(const LatticeBinaryExpression<Op, T1, T2> &expr)
-    -> Lattice<decltype(expr.first.func(eval(0, std::get<0>(expr.second)),
-                                        eval(0, std::get<1>(expr.second))))> {
-  Lattice<decltype(expr.first.func(eval(0, std::get<0>(expr.second)),
-                                   eval(0, std::get<1>(expr.second))))>
-      ret(expr);
+  -> Lattice<decltype(expr.op.func(eval(0, expr.arg1),eval(0, expr.arg2)))> 
+{
+  Lattice<decltype(expr.op.func(eval(0, expr.arg1),eval(0, expr.arg2)))> ret(expr);
   return ret;
 }
 template <class Op, class T1, class T2, class T3>
 auto closure(const LatticeTrinaryExpression<Op, T1, T2, T3> &expr)
-    -> Lattice<decltype(expr.first.func(eval(0, std::get<0>(expr.second)),
-                                        eval(0, std::get<1>(expr.second)),
-                                        eval(0, std::get<2>(expr.second))))> {
-  Lattice<decltype(expr.first.func(eval(0, std::get<0>(expr.second)),
-                                   eval(0, std::get<1>(expr.second)),
-                                   eval(0, std::get<2>(expr.second))))>
-      ret(expr);
+  -> Lattice<decltype(expr.op.func(eval(0, expr.arg1),
+				   eval(0, expr.arg2),
+				   eval(0, expr.arg3)))> 
+{
+  Lattice<decltype(expr.op.func(eval(0, expr.arg1),
+				eval(0, expr.arg2),
+				eval(0, expr.arg3)))>  ret(expr);
   return ret;
 }
 
@@ -433,34 +401,7 @@ auto closure(const LatticeTrinaryExpression<Op, T1, T2, T3> &expr)
 #undef GRID_DEF_UNOP
 #undef GRID_DEF_BINOP
 #undef GRID_DEF_TRINOP
-}
 
-#if 0
-using namespace Grid;
-        
- int main(int argc,char **argv){
-   
-   Lattice<double> v1(16);
-   Lattice<double> v2(16);
-   Lattice<double> v3(16);
-
-   BinaryAdd<double,double> tmp;
-   LatticeBinaryExpression<BinaryAdd<double,double>,Lattice<double> &,Lattice<double> &> 
-     expr(std::make_pair(tmp,
-    std::forward_as_tuple(v1,v2)));
-   tmp.func(eval(0,v1),eval(0,v2));
-
-   auto var = v1+v2;
-   std::cout<<GridLogMessage<<typeid(var).name()<<std::endl;
-
-   v3=v1+v2;
-   v3=v1+v2+v1*v2;
- };
-
-void testit(Lattice<double> &v1,Lattice<double> &v2,Lattice<double> &v3)
-{
-   v3=v1+v2+v1*v2;
-}
-#endif
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/lattice/Lattice_arith.h b/Grid/lattice/Lattice_arith.h
index d1cbc84a..3543d6aa 100644
--- a/Grid/lattice/Lattice_arith.h
+++ b/Grid/lattice/Lattice_arith.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,233 +23,235 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_LATTICE_ARITH_H
 #define GRID_LATTICE_ARITH_H
 
-namespace Grid {
-
-
-  //////////////////////////////////////////////////////////////////////////////////////////////////////
-  //  avoid copy back routines for mult, mac, sub, add
-  //////////////////////////////////////////////////////////////////////////////////////////////////////
-  template<class obj1,class obj2,class obj3> strong_inline
-    void mult(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const Lattice<obj3> &rhs){
-    ret.checkerboard = lhs.checkerboard;
-    conformable(ret,rhs);
-    conformable(lhs,rhs);
-    parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      obj1 tmp;
-      mult(&tmp,&lhs._odata[ss],&rhs._odata[ss]);
-      vstream(ret._odata[ss],tmp);
-#else
-      mult(&ret._odata[ss],&lhs._odata[ss],&rhs._odata[ss]);
-#endif
-    }
-  }
-  
-  template<class obj1,class obj2,class obj3> strong_inline
-    void mac(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const Lattice<obj3> &rhs){
-    ret.checkerboard = lhs.checkerboard;
-    conformable(ret,rhs);
-    conformable(lhs,rhs);
-    parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      obj1 tmp;
-      mac(&tmp,&lhs._odata[ss],&rhs._odata[ss]);
-      vstream(ret._odata[ss],tmp);
-#else
-      mac(&ret._odata[ss],&lhs._odata[ss],&rhs._odata[ss]);
-#endif
-    }
-  }
-  
-  template<class obj1,class obj2,class obj3> strong_inline
-    void sub(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const Lattice<obj3> &rhs){
-    ret.checkerboard = lhs.checkerboard;
-    conformable(ret,rhs);
-    conformable(lhs,rhs);
-    parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      obj1 tmp;
-      sub(&tmp,&lhs._odata[ss],&rhs._odata[ss]);
-      vstream(ret._odata[ss],tmp);
-#else
-      sub(&ret._odata[ss],&lhs._odata[ss],&rhs._odata[ss]);
-#endif
-    }
-  }
-  template<class obj1,class obj2,class obj3> strong_inline
-    void add(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const Lattice<obj3> &rhs){
-    ret.checkerboard = lhs.checkerboard;
-    conformable(ret,rhs);
-    conformable(lhs,rhs);
-    parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      obj1 tmp;
-      add(&tmp,&lhs._odata[ss],&rhs._odata[ss]);
-      vstream(ret._odata[ss],tmp);
-#else
-      add(&ret._odata[ss],&lhs._odata[ss],&rhs._odata[ss]);
-#endif
-    }
-  }
-  
-  //////////////////////////////////////////////////////////////////////////////////////////////////////
-  //  avoid copy back routines for mult, mac, sub, add
-  //////////////////////////////////////////////////////////////////////////////////////////////////////
-  template<class obj1,class obj2,class obj3> strong_inline
-    void mult(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const obj3 &rhs){
-    ret.checkerboard = lhs.checkerboard;
-    conformable(lhs,ret);
-    parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-      obj1 tmp;
-      mult(&tmp,&lhs._odata[ss],&rhs);
-      vstream(ret._odata[ss],tmp);
-    }
-  }
-  
-  template<class obj1,class obj2,class obj3> strong_inline
-    void mac(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const obj3 &rhs){
-    ret.checkerboard = lhs.checkerboard;
-    conformable(ret,lhs);
-    parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-      obj1 tmp;
-      mac(&tmp,&lhs._odata[ss],&rhs);
-      vstream(ret._odata[ss],tmp);
-    }
-  }
-  
-  template<class obj1,class obj2,class obj3> strong_inline
-    void sub(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const obj3 &rhs){
-    ret.checkerboard = lhs.checkerboard;
-    conformable(ret,lhs);
-    parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      obj1 tmp;
-      sub(&tmp,&lhs._odata[ss],&rhs);
-      vstream(ret._odata[ss],tmp);
-#else 
-      sub(&ret._odata[ss],&lhs._odata[ss],&rhs);
-#endif
-    }
-  }
-  template<class obj1,class obj2,class obj3> strong_inline
-    void add(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const obj3 &rhs){
-    ret.checkerboard = lhs.checkerboard;
-    conformable(lhs,ret);
-    parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      obj1 tmp;
-      add(&tmp,&lhs._odata[ss],&rhs);
-      vstream(ret._odata[ss],tmp);
-#else 
-      add(&ret._odata[ss],&lhs._odata[ss],&rhs);
-#endif
-    }
-  }
-
-  //////////////////////////////////////////////////////////////////////////////////////////////////////
-  //  avoid copy back routines for mult, mac, sub, add
-  //////////////////////////////////////////////////////////////////////////////////////////////////////
-    template<class obj1,class obj2,class obj3> strong_inline
-    void mult(Lattice<obj1> &ret,const obj2 &lhs,const Lattice<obj3> &rhs){
-    ret.checkerboard = rhs.checkerboard;
-    conformable(ret,rhs);
-    parallel_for(int ss=0;ss<rhs._grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      obj1 tmp;
-      mult(&tmp,&lhs,&rhs._odata[ss]);
-      vstream(ret._odata[ss],tmp);
-#else 
-      mult(&ret._odata[ss],&lhs,&rhs._odata[ss]);
-#endif
-    }
-  }
-  
-  template<class obj1,class obj2,class obj3> strong_inline
-    void mac(Lattice<obj1> &ret,const obj2 &lhs,const Lattice<obj3> &rhs){
-    ret.checkerboard = rhs.checkerboard;
-    conformable(ret,rhs);
-    parallel_for(int ss=0;ss<rhs._grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      obj1 tmp;
-      mac(&tmp,&lhs,&rhs._odata[ss]);
-      vstream(ret._odata[ss],tmp);
-#else 
-      mac(&ret._odata[ss],&lhs,&rhs._odata[ss]);
-#endif
-    }
-  }
-  
-  template<class obj1,class obj2,class obj3> strong_inline
-    void sub(Lattice<obj1> &ret,const obj2 &lhs,const Lattice<obj3> &rhs){
-    ret.checkerboard = rhs.checkerboard;
-    conformable(ret,rhs);
-    parallel_for(int ss=0;ss<rhs._grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      obj1 tmp;
-      sub(&tmp,&lhs,&rhs._odata[ss]);
-      vstream(ret._odata[ss],tmp);
-#else 
-      sub(&ret._odata[ss],&lhs,&rhs._odata[ss]);
-#endif
-    }
-  }
-  template<class obj1,class obj2,class obj3> strong_inline
-    void add(Lattice<obj1> &ret,const obj2 &lhs,const Lattice<obj3> &rhs){
-    ret.checkerboard = rhs.checkerboard;
-    conformable(ret,rhs);
-    parallel_for(int ss=0;ss<rhs._grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      obj1 tmp;
-      add(&tmp,&lhs,&rhs._odata[ss]);
-      vstream(ret._odata[ss],tmp);
-#else 
-      add(&ret._odata[ss],&lhs,&rhs._odata[ss]);
-#endif
-    }
-  }
-  
-  template<class sobj,class vobj> strong_inline
-  void axpy(Lattice<vobj> &ret,sobj a,const Lattice<vobj> &x,const Lattice<vobj> &y){
-    ret.checkerboard = x.checkerboard;
-    conformable(ret,x);
-    conformable(x,y);
-    parallel_for(int ss=0;ss<x._grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      vobj tmp = a*x._odata[ss]+y._odata[ss];
-      vstream(ret._odata[ss],tmp);
-#else
-      ret._odata[ss]=a*x._odata[ss]+y._odata[ss];
-#endif
-    }
-  }
-  template<class sobj,class vobj> strong_inline
-  void axpby(Lattice<vobj> &ret,sobj a,sobj b,const Lattice<vobj> &x,const Lattice<vobj> &y){
-    ret.checkerboard = x.checkerboard;
-    conformable(ret,x);
-    conformable(x,y);
-    parallel_for(int ss=0;ss<x._grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      vobj tmp = a*x._odata[ss]+b*y._odata[ss];
-      vstream(ret._odata[ss],tmp);
-#else
-      ret._odata[ss]=a*x._odata[ss]+b*y._odata[ss];
-#endif
-    }
-  }
-
-  template<class sobj,class vobj> strong_inline
-  RealD axpy_norm(Lattice<vobj> &ret,sobj a,const Lattice<vobj> &x,const Lattice<vobj> &y){
-    return axpy_norm_fast(ret,a,x,y);
-  }
-  template<class sobj,class vobj> strong_inline
-  RealD axpby_norm(Lattice<vobj> &ret,sobj a,sobj b,const Lattice<vobj> &x,const Lattice<vobj> &y){
-    return axpby_norm_fast(ret,a,b,x,y);
-  }
+NAMESPACE_BEGIN(Grid);
 
+//////////////////////////////////////////////////////////////////////////////////////////////////////
+//  avoid copy back routines for mult, mac, sub, add
+//////////////////////////////////////////////////////////////////////////////////////////////////////
+template<class obj1,class obj2,class obj3> inline
+void mult(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const Lattice<obj3> &rhs){
+  ret.Checkerboard() = lhs.Checkerboard();
+  auto ret_v = ret.View();
+  auto lhs_v = lhs.View();
+  auto rhs_v = rhs.View();
+  conformable(ret,rhs);
+  conformable(lhs,rhs);
+  accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
+    decltype(coalescedRead(obj1())) tmp;
+    auto lhs_t = lhs_v(ss);
+    auto rhs_t = rhs_v(ss);
+    mult(&tmp,&lhs_t,&rhs_t);
+    coalescedWrite(ret_v[ss],tmp);
+  });
 }
+  
+template<class obj1,class obj2,class obj3> inline
+void mac(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const Lattice<obj3> &rhs){
+  ret.Checkerboard() = lhs.Checkerboard();
+  conformable(ret,rhs);
+  conformable(lhs,rhs);
+  auto ret_v = ret.View();
+  auto lhs_v = lhs.View();
+  auto rhs_v = rhs.View();
+  accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
+    decltype(coalescedRead(obj1())) tmp;
+    auto lhs_t=lhs_v(ss);
+    auto rhs_t=rhs_v(ss);
+    mac(&tmp,&lhs_t,&rhs_t);
+    coalescedWrite(ret_v[ss],tmp);
+  });
+}
+  
+template<class obj1,class obj2,class obj3> inline
+void sub(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const Lattice<obj3> &rhs){
+  ret.Checkerboard() = lhs.Checkerboard();
+  conformable(ret,rhs);
+  conformable(lhs,rhs);
+  auto ret_v = ret.View();
+  auto lhs_v = lhs.View();
+  auto rhs_v = rhs.View();
+  accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
+    decltype(coalescedRead(obj1())) tmp;
+    auto lhs_t=lhs_v(ss);
+    auto rhs_t=rhs_v(ss);
+    sub(&tmp,&lhs_t,&rhs_t);
+    coalescedWrite(ret_v[ss],tmp);
+  });
+}
+template<class obj1,class obj2,class obj3> inline
+void add(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const Lattice<obj3> &rhs){
+  ret.Checkerboard() = lhs.Checkerboard();
+  conformable(ret,rhs);
+  conformable(lhs,rhs);
+  auto ret_v = ret.View();
+  auto lhs_v = lhs.View();
+  auto rhs_v = rhs.View();
+  accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
+    decltype(coalescedRead(obj1())) tmp;
+    auto lhs_t=lhs_v(ss);
+    auto rhs_t=rhs_v(ss);
+    add(&tmp,&lhs_t,&rhs_t);
+    coalescedWrite(ret_v[ss],tmp);
+  });
+}
+  
+//////////////////////////////////////////////////////////////////////////////////////////////////////
+//  avoid copy back routines for mult, mac, sub, add
+//////////////////////////////////////////////////////////////////////////////////////////////////////
+template<class obj1,class obj2,class obj3> inline
+void mult(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const obj3 &rhs){
+  ret.Checkerboard() = lhs.Checkerboard();
+  conformable(lhs,ret);
+  auto ret_v = ret.View();
+  auto lhs_v = lhs.View();
+  accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
+    decltype(coalescedRead(obj1())) tmp;
+    mult(&tmp,&lhs_v(ss),&rhs);
+    coalescedWrite(ret_v[ss],tmp);
+  });
+}
+  
+template<class obj1,class obj2,class obj3> inline
+void mac(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const obj3 &rhs){
+  ret.Checkerboard() = lhs.Checkerboard();
+  conformable(ret,lhs);
+  auto ret_v = ret.View();
+  auto lhs_v = lhs.View();
+  accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
+    decltype(coalescedRead(obj1())) tmp;
+    auto lhs_t=lhs_v(ss);
+    mac(&tmp,&lhs_t,&rhs);
+    coalescedWrite(ret_v[ss],tmp);
+  });
+}
+  
+template<class obj1,class obj2,class obj3> inline
+void sub(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const obj3 &rhs){
+  ret.Checkerboard() = lhs.Checkerboard();
+  conformable(ret,lhs);
+  auto ret_v = ret.View();
+  auto lhs_v = lhs.View();
+  accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
+    decltype(coalescedRead(obj1())) tmp;
+    auto lhs_t=lhs_v(ss);
+    sub(&tmp,&lhs_t,&rhs);
+    coalescedWrite(ret_v[ss],tmp);
+  });
+}
+template<class obj1,class obj2,class obj3> inline
+void add(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const obj3 &rhs){
+  ret.Checkerboard() = lhs.Checkerboard();
+  conformable(lhs,ret);
+  auto ret_v = ret.View();
+  auto lhs_v = lhs.View();
+  accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
+    decltype(coalescedRead(obj1())) tmp;
+    auto lhs_t=lhs_v(ss);
+    add(&tmp,&lhs_t,&rhs);
+    coalescedWrite(ret_v[ss],tmp);
+  });
+}
+
+//////////////////////////////////////////////////////////////////////////////////////////////////////
+//  avoid copy back routines for mult, mac, sub, add
+//////////////////////////////////////////////////////////////////////////////////////////////////////
+template<class obj1,class obj2,class obj3> inline
+void mult(Lattice<obj1> &ret,const obj2 &lhs,const Lattice<obj3> &rhs){
+  ret.Checkerboard() = rhs.Checkerboard();
+  conformable(ret,rhs);
+  auto ret_v = ret.View();
+  auto rhs_v = lhs.View();
+  accelerator_for(ss,rhs_v.size(),obj1::Nsimd(),{
+    decltype(coalescedRead(obj1())) tmp;
+    auto rhs_t=rhs_v(ss);
+    mult(&tmp,&lhs,&rhs_t);
+    coalescedWrite(ret_v[ss],tmp);
+  });
+}
+  
+template<class obj1,class obj2,class obj3> inline
+void mac(Lattice<obj1> &ret,const obj2 &lhs,const Lattice<obj3> &rhs){
+  ret.Checkerboard() = rhs.Checkerboard();
+  conformable(ret,rhs);
+  auto ret_v = ret.View();
+  auto rhs_v = lhs.View();
+  accelerator_for(ss,rhs_v.size(),obj1::Nsimd(),{
+    decltype(coalescedRead(obj1())) tmp;
+    auto rhs_t=rhs_v(ss);
+    mac(&tmp,&lhs,&rhs_t);
+    coalescedWrite(ret_v[ss],tmp);
+  });
+}
+  
+template<class obj1,class obj2,class obj3> inline
+void sub(Lattice<obj1> &ret,const obj2 &lhs,const Lattice<obj3> &rhs){
+  ret.Checkerboard() = rhs.Checkerboard();
+  conformable(ret,rhs);
+  auto ret_v = ret.View();
+  auto rhs_v = lhs.View();
+  accelerator_for(ss,rhs_v.size(),obj1::Nsimd(),{
+    decltype(coalescedRead(obj1())) tmp;
+    auto rhs_t=rhs_v(ss);
+    sub(&tmp,&lhs,&rhs_t);
+    coalescedWrite(ret_v[ss],tmp);
+  });
+}
+template<class obj1,class obj2,class obj3> inline
+void add(Lattice<obj1> &ret,const obj2 &lhs,const Lattice<obj3> &rhs){
+  ret.Checkerboard() = rhs.Checkerboard();
+  conformable(ret,rhs);
+  auto ret_v = ret.View();
+  auto rhs_v = lhs.View();
+  accelerator_for(ss,rhs_v.size(),obj1::Nsimd(),{
+    decltype(coalescedRead(obj1())) tmp;
+    auto rhs_t=rhs_v(ss);
+    add(&tmp,&lhs,&rhs_t);
+    coalescedWrite(ret_v[ss],tmp);
+  });
+}
+  
+template<class sobj,class vobj> inline
+void axpy(Lattice<vobj> &ret,sobj a,const Lattice<vobj> &x,const Lattice<vobj> &y){
+  ret.Checkerboard() = x.Checkerboard();
+  conformable(ret,x);
+  conformable(x,y);
+  auto ret_v = ret.View();
+  auto x_v = x.View();
+  auto y_v = y.View();
+  accelerator_for(ss,x_v.size(),vobj::Nsimd(),{
+    auto tmp = a*x_v(ss)+y_v(ss);
+    coalescedWrite(ret_v[ss],tmp);
+  });
+}
+template<class sobj,class vobj> inline
+void axpby(Lattice<vobj> &ret,sobj a,sobj b,const Lattice<vobj> &x,const Lattice<vobj> &y){
+  ret.Checkerboard() = x.Checkerboard();
+  conformable(ret,x);
+  conformable(x,y);
+  auto ret_v = ret.View();
+  auto x_v = x.View();
+  auto y_v = y.View();
+  accelerator_for(ss,x_v.size(),vobj::Nsimd(),{
+    auto tmp = a*x_v(ss)+b*y_v(ss);
+    coalescedWrite(ret_v[ss],tmp);
+  });
+}
+
+template<class sobj,class vobj> inline
+RealD axpy_norm(Lattice<vobj> &ret,sobj a,const Lattice<vobj> &x,const Lattice<vobj> &y)
+{
+    return axpy_norm_fast(ret,a,x,y);
+}
+template<class sobj,class vobj> inline
+RealD axpby_norm(Lattice<vobj> &ret,sobj a,sobj b,const Lattice<vobj> &x,const Lattice<vobj> &y)
+{
+    return axpby_norm_fast(ret,a,b,x,y);
+}
+
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/lattice/Lattice_base.h b/Grid/lattice/Lattice_base.h
index d0ef0f55..d48e5090 100644
--- a/Grid/lattice/Lattice_base.h
+++ b/Grid/lattice/Lattice_base.h
@@ -27,349 +27,440 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
-#ifndef GRID_LATTICE_BASE_H
-#define GRID_LATTICE_BASE_H
+			   /*  END LEGAL */
+#pragma once 
 
 #define STREAMING_STORES
 
-namespace Grid {
-
-// TODO: 
-//       mac,real,imag
-
-// Functionality:
-//     -=,+=,*=,()
-//     add,+,sub,-,mult,mac,*
-//     adj,conjugate
-//     real,imag
-//     transpose,transposeIndex  
-//     trace,traceIndex
-//     peekIndex
-//     innerProduct,outerProduct,
-//     localNorm2
-//     localInnerProduct
+NAMESPACE_BEGIN(Grid);
 
 extern int GridCshiftPermuteMap[4][16];
 
-////////////////////////////////////////////////
-// Basic expressions used in Expression Template
-////////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////
+// Base class which can be used by traits to pick up behaviour
+///////////////////////////////////////////////////////////////////
+class LatticeBase {};
 
-class LatticeBase
+/////////////////////////////////////////////////////////////////////////////////////////
+// Conformable checks; same instance of Grid required
+/////////////////////////////////////////////////////////////////////////////////////////
+void accelerator_inline conformable(GridBase *lhs,GridBase *rhs)
 {
-public:
-    virtual ~LatticeBase(void) = default;
-    GridBase *_grid;
-};
-    
-class LatticeExpressionBase {};
-
-template <typename Op, typename T1>                           
-class LatticeUnaryExpression  : public std::pair<Op,std::tuple<T1> > , public LatticeExpressionBase {
- public:
- LatticeUnaryExpression(const std::pair<Op,std::tuple<T1> > &arg): std::pair<Op,std::tuple<T1> >(arg) {};
-};
-
-template <typename Op, typename T1, typename T2>              
-class LatticeBinaryExpression : public std::pair<Op,std::tuple<T1,T2> > , public LatticeExpressionBase {
- public:
- LatticeBinaryExpression(const std::pair<Op,std::tuple<T1,T2> > &arg): std::pair<Op,std::tuple<T1,T2> >(arg) {};
-};
-
-template <typename Op, typename T1, typename T2, typename T3> 
-class LatticeTrinaryExpression :public std::pair<Op,std::tuple<T1,T2,T3> >, public LatticeExpressionBase {
- public:
- LatticeTrinaryExpression(const std::pair<Op,std::tuple<T1,T2,T3> > &arg): std::pair<Op,std::tuple<T1,T2,T3> >(arg) {};
-};
-
-void inline conformable(GridBase *lhs,GridBase *rhs)
-{
-  assert((lhs == rhs) && " conformable check pointers mismatch ");
+  assert(lhs == rhs);
 }
 
-template<class vobj>
-class Lattice : public LatticeBase
+////////////////////////////////////////////////////////////////////////////
+// Minimal base class containing only data valid to access from accelerator
+// _odata will be a managed pointer in CUDA
+////////////////////////////////////////////////////////////////////////////
+// Force access to lattice through a view object.
+// prevents writing of code that will not offload to GPU, but perhaps annoyingly
+// strict since host could could in principle direct access through the lattice object
+// Need to decide programming model.
+#define LATTICE_VIEW_STRICT
+template<class vobj> class LatticeAccelerator : public LatticeBase
+{
+protected:
+  GridBase *_grid;
+  int checkerboard;
+  vobj     *_odata;    // A managed pointer
+  uint64_t _odata_size;    
+public:
+  accelerator_inline LatticeAccelerator() : checkerboard(0), _odata(nullptr), _odata_size(0), _grid(nullptr) { }; 
+  accelerator_inline uint64_t oSites(void) const { return _odata_size; };
+  accelerator_inline int  Checkerboard(void) const { return checkerboard; };
+  accelerator_inline int &Checkerboard(void) { return this->checkerboard; }; // can assign checkerboard on a container, not a view
+  accelerator_inline void Conformable(GridBase * &grid) const
+  { 
+    if (grid) conformable(grid, _grid);
+    else      grid = _grid;
+  };
+};
+
+/////////////////////////////////////////////////////////////////////////////////////////
+// A View class which provides accessor to the data.
+// This will be safe to call from accelerator_for and is trivially copy constructible
+// The copy constructor for this will need to be used by device lambda functions
+/////////////////////////////////////////////////////////////////////////////////////////
+template<class vobj> 
+class LatticeView : public LatticeAccelerator<vobj>
 {
 public:
-    int checkerboard;
-    Vector<vobj> _odata;
-    
-    // to pthread need a computable loop where loop induction is not required
-    int begin(void) { return 0;};
-    int end(void)   { return _odata.size(); }
-    vobj & operator[](int i) { return _odata[i]; };
-    const vobj & operator[](int i) const { return _odata[i]; };
 
+
+  // Rvalue
+#ifdef __CUDA_ARCH__
+  accelerator_inline const typename vobj::scalar_object operator()(size_t i) const { return coalescedRead(this->_odata[i]); }
+#else 
+  accelerator_inline const vobj & operator()(size_t i) const { return this->_odata[i]; }
+#endif
+
+  accelerator_inline const vobj & operator[](size_t i) const { return this->_odata[i]; };
+  accelerator_inline vobj       & operator[](size_t i)       { return this->_odata[i]; };
+
+  accelerator_inline uint64_t begin(void) const { return 0;};
+  accelerator_inline uint64_t end(void)   const { return this->_odata_size; };
+  accelerator_inline uint64_t size(void)  const { return this->_odata_size; };
+
+  LatticeView(const LatticeAccelerator<vobj> &refer_to_me) : LatticeAccelerator<vobj> (refer_to_me)
+  {
+  }
+};
+
+/////////////////////////////////////////////////////////////////////////////////////////
+// Lattice expression types used by ET to assemble the AST
+// 
+// Need to be able to detect code paths according to the whether a lattice object or not
+// so introduce some trait type things
+/////////////////////////////////////////////////////////////////////////////////////////
+
+class LatticeExpressionBase {};
+
+template <typename T> using is_lattice = std::is_base_of<LatticeBase, T>;
+template <typename T> using is_lattice_expr = std::is_base_of<LatticeExpressionBase,T >;
+
+template<class T, bool isLattice> struct ViewMapBase { typedef T Type; };
+template<class T>                 struct ViewMapBase<T,true> { typedef LatticeView<typename T::vector_object> Type; };
+template<class T> using ViewMap = ViewMapBase<T,std::is_base_of<LatticeBase, T>::value >;
+
+template <typename Op, typename _T1>                           
+class LatticeUnaryExpression : public  LatticeExpressionBase 
+{
 public:
-    typedef typename vobj::scalar_type scalar_type;
-    typedef typename vobj::vector_type vector_type;
-    typedef vobj vector_object;
-   
+  typedef typename ViewMap<_T1>::Type T1;
+  Op op;
+  T1 arg1;
+  LatticeUnaryExpression(Op _op,const _T1 &_arg1) : op(_op), arg1(_arg1) {};
+};
+
+template <typename Op, typename _T1, typename _T2>              
+class LatticeBinaryExpression : public LatticeExpressionBase 
+{
+public:
+  typedef typename ViewMap<_T1>::Type T1;
+  typedef typename ViewMap<_T2>::Type T2;
+  Op op;
+  T1 arg1;
+  T2 arg2;
+  LatticeBinaryExpression(Op _op,const _T1 &_arg1,const _T2 &_arg2) : op(_op), arg1(_arg1), arg2(_arg2) {};
+};
+
+template <typename Op, typename _T1, typename _T2, typename _T3> 
+class LatticeTrinaryExpression : public LatticeExpressionBase 
+{
+public:
+  typedef typename ViewMap<_T1>::Type T1;
+  typedef typename ViewMap<_T2>::Type T2;
+  typedef typename ViewMap<_T3>::Type T3;
+  Op op;
+  T1 arg1;
+  T2 arg2;
+  T3 arg3;
+  LatticeTrinaryExpression(Op _op,const _T1 &_arg1,const _T2 &_arg2,const _T3 &_arg3) : op(_op), arg1(_arg1), arg2(_arg2), arg3(_arg3) {};
+};
+
+/////////////////////////////////////////////////////////////////////////////////////////
+// The real lattice class, with normal copy and assignment semantics.
+// This contains extra (host resident) grid pointer data that may be accessed by host code
+/////////////////////////////////////////////////////////////////////////////////////////
+template<class vobj>
+class Lattice : public LatticeAccelerator<vobj>
+{
+public:
+  GridBase *Grid(void) const { return this->_grid; }
+  ///////////////////////////////////////////////////
+  // Member types
+  ///////////////////////////////////////////////////
+  typedef typename vobj::scalar_type scalar_type;
+  typedef typename vobj::vector_type vector_type;
+  typedef vobj vector_object;
+
+private:
+  void dealloc(void)
+  {
+    alignedAllocator<vobj> alloc;
+    if( this->_odata_size ) {
+      alloc.deallocate(this->_odata,this->_odata_size);
+      this->_odata=nullptr;
+      this->_odata_size=0;
+    }
+  }
+  void resize(uint64_t size)
+  {
+    alignedAllocator<vobj> alloc;
+    if ( this->_odata_size != size ) {
+      dealloc();
+    }
+    this->_odata_size = size;
+    if ( size ) 
+      this->_odata      = alloc.allocate(this->_odata_size);
+    else 
+      this->_odata      = nullptr;
+  }
+public:
+  /////////////////////////////////////////////////////////////////////////////////
+  // Return a view object that may be dereferenced in site loops.
+  // The view is trivially copy constructible and may be copied to an accelerator device
+  // in device lambdas
+  /////////////////////////////////////////////////////////////////////////////////
+  LatticeView<vobj> View (void) const 
+  {
+    LatticeView<vobj> accessor(*( (LatticeAccelerator<vobj> *) this));
+    return accessor;
+  }
+  
+  ~Lattice() { 
+    if ( this->_odata_size ) {
+      dealloc();
+    }
+   }
   ////////////////////////////////////////////////////////////////////////////////
   // Expression Template closure support
   ////////////////////////////////////////////////////////////////////////////////
-  template <typename Op, typename T1>                         strong_inline Lattice<vobj> & operator=(const LatticeUnaryExpression<Op,T1> &expr)
+  template <typename Op, typename T1> inline Lattice<vobj> & operator=(const LatticeUnaryExpression<Op,T1> &expr)
   {
     GridBase *egrid(nullptr);
     GridFromExpression(egrid,expr);
     assert(egrid!=nullptr);
-    conformable(_grid,egrid);
+    conformable(this->_grid,egrid);
 
     int cb=-1;
     CBFromExpression(cb,expr);
     assert( (cb==Odd) || (cb==Even));
-    checkerboard=cb;
+    this->checkerboard=cb;
 
-    parallel_for(int ss=0;ss<_grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      vobj tmp = eval(ss,expr);
-      vstream(_odata[ss] ,tmp);
-#else
-      _odata[ss]=eval(ss,expr);
-#endif
-    }
+    auto me  = View();
+    accelerator_for(ss,me.size(),1,{
+      auto tmp = eval(ss,expr);
+      vstream(me[ss],tmp);
+    });
     return *this;
   }
-  template <typename Op, typename T1,typename T2> strong_inline Lattice<vobj> & operator=(const LatticeBinaryExpression<Op,T1,T2> &expr)
+  template <typename Op, typename T1,typename T2> inline Lattice<vobj> & operator=(const LatticeBinaryExpression<Op,T1,T2> &expr)
   {
     GridBase *egrid(nullptr);
     GridFromExpression(egrid,expr);
     assert(egrid!=nullptr);
-    conformable(_grid,egrid);
+    conformable(this->_grid,egrid);
 
     int cb=-1;
     CBFromExpression(cb,expr);
     assert( (cb==Odd) || (cb==Even));
-    checkerboard=cb;
+    this->checkerboard=cb;
 
-    parallel_for(int ss=0;ss<_grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      vobj tmp = eval(ss,expr);
-      vstream(_odata[ss] ,tmp);
-#else
-      _odata[ss]=eval(ss,expr);
-#endif
-    }
+    auto me  = View();
+    accelerator_for(ss,me.size(),1,{
+      auto tmp = eval(ss,expr);
+      vstream(me[ss],tmp);
+    });
     return *this;
   }
-  template <typename Op, typename T1,typename T2,typename T3> strong_inline Lattice<vobj> & operator=(const LatticeTrinaryExpression<Op,T1,T2,T3> &expr)
+  template <typename Op, typename T1,typename T2,typename T3> inline Lattice<vobj> & operator=(const LatticeTrinaryExpression<Op,T1,T2,T3> &expr)
   {
     GridBase *egrid(nullptr);
     GridFromExpression(egrid,expr);
     assert(egrid!=nullptr);
-    conformable(_grid,egrid);
+    conformable(this->_grid,egrid);
 
     int cb=-1;
     CBFromExpression(cb,expr);
     assert( (cb==Odd) || (cb==Even));
-    checkerboard=cb;
-
-    parallel_for(int ss=0;ss<_grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      //vobj tmp = eval(ss,expr);
-      vstream(_odata[ss] ,eval(ss,expr));
-#else
-      _odata[ss] = eval(ss,expr);
-#endif
-    }
+    this->checkerboard=cb;
+    auto me  = View();
+    accelerator_for(ss,me.size(),1,{
+      auto tmp = eval(ss,expr);
+      vstream(me[ss],tmp);
+    });
     return *this;
   }
   //GridFromExpression is tricky to do
   template<class Op,class T1>
-    Lattice(const LatticeUnaryExpression<Op,T1> & expr) {
-    _grid = nullptr;
-    GridFromExpression(_grid,expr);
-    assert(_grid!=nullptr);
+  Lattice(const LatticeUnaryExpression<Op,T1> & expr) {
+    this->_grid = nullptr;
+    GridFromExpression(this->_grid,expr);
+    assert(this->_grid!=nullptr);
 
     int cb=-1;
     CBFromExpression(cb,expr);
     assert( (cb==Odd) || (cb==Even));
-    checkerboard=cb;
+    this->checkerboard=cb;
 
-    _odata.resize(_grid->oSites());
-    parallel_for(int ss=0;ss<_grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      vobj tmp = eval(ss,expr);
-      vstream(_odata[ss] ,tmp);
-#else
-      _odata[ss]=eval(ss,expr);
-#endif
-    }
-  };
+    resize(this->_grid->oSites());
+
+    *this = expr;
+  }
   template<class Op,class T1, class T2>
   Lattice(const LatticeBinaryExpression<Op,T1,T2> & expr) {
-    _grid = nullptr;
-    GridFromExpression(_grid,expr);
-    assert(_grid!=nullptr);
+    this->_grid = nullptr;
+    GridFromExpression(this->_grid,expr);
+    assert(this->_grid!=nullptr);
 
     int cb=-1;
     CBFromExpression(cb,expr);
     assert( (cb==Odd) || (cb==Even));
-    checkerboard=cb;
+    this->checkerboard=cb;
 
-    _odata.resize(_grid->oSites());
-    parallel_for(int ss=0;ss<_grid->oSites();ss++){
-#ifdef STREAMING_STORES
-      vobj tmp = eval(ss,expr);
-      vstream(_odata[ss] ,tmp);
-#else
-      _odata[ss]=eval(ss,expr);
-#endif
-    }
-  };
+    resize(this->_grid->oSites());
+
+    *this = expr;
+  }
   template<class Op,class T1, class T2, class T3>
   Lattice(const LatticeTrinaryExpression<Op,T1,T2,T3> & expr) {
-    _grid = nullptr;
-    GridFromExpression(_grid,expr);
-    assert(_grid!=nullptr);
+    this->_grid = nullptr;
+    GridFromExpression(this->_grid,expr);
+    assert(this->_grid!=nullptr);
 
     int cb=-1;
     CBFromExpression(cb,expr);
     assert( (cb==Odd) || (cb==Even));
-    checkerboard=cb;
+    this->checkerboard=cb;
 
-    _odata.resize(_grid->oSites());
-    parallel_for(int ss=0;ss<_grid->oSites();ss++){
-      vstream(_odata[ss] ,eval(ss,expr));
-    }
-  };
+    resize(this->_grid->oSites());
+
+    *this = expr;
+  }
+
+  template<class sobj> inline Lattice<vobj> & operator = (const sobj & r){
+    auto me  = View();
+    thread_for(ss,me.size(),{
+      me[ss] = r;
+    });
+    return *this;
+  }
 
   //////////////////////////////////////////////////////////////////
-  // Constructor requires "grid" passed.
-  // what about a default grid?
-  //////////////////////////////////////////////////////////////////
-  Lattice(GridBase *grid) : _odata(grid->oSites()) {
-    _grid = grid;
-    //        _odata.reserve(_grid->oSites());
-    //        _odata.resize(_grid->oSites());
-    //      std::cout << "Constructing lattice object with Grid pointer "<<_grid<<std::endl;
-    assert((((uint64_t)&_odata[0])&0xF) ==0);
-    checkerboard=0;
+  // Follow rule of five, with Constructor requires "grid" passed
+  // to user defined constructor
+  ///////////////////////////////////////////
+  // user defined constructor
+  ///////////////////////////////////////////
+  Lattice(GridBase *grid) { 
+    this->_grid = grid;
+    resize(this->_grid->oSites());
+    assert((((uint64_t)&this->_odata[0])&0xF) ==0);
+    this->checkerboard=0;
   }
   
-  Lattice(const Lattice& r){ // copy constructor
-    _grid = r._grid;
-    checkerboard = r.checkerboard;
-    _odata.resize(_grid->oSites());// essential
-    parallel_for(int ss=0;ss<_grid->oSites();ss++){
-      _odata[ss]=r._odata[ss];
-    }  	
-  }
-
-  Lattice(Lattice&& r){ // move constructor
-    _grid = r._grid;
-    checkerboard = r.checkerboard;
-    _odata=std::move(r._odata);
-  }
-  
-  inline Lattice<vobj> & operator = (Lattice<vobj> && r)
-  {
-    _grid        = r._grid;
-    checkerboard = r.checkerboard;
-    _odata       =std::move(r._odata);
-    return *this;
-  }
-
-  inline Lattice<vobj> & operator = (const Lattice<vobj> & r){
-    _grid        = r._grid;
-    checkerboard = r.checkerboard;
-    _odata.resize(_grid->oSites());// essential
-    
-    parallel_for(int ss=0;ss<_grid->oSites();ss++){
-      _odata[ss]=r._odata[ss];
-    }  	
-    return *this;
-  }
-
-  template<class robj> strong_inline Lattice<vobj> & operator = (const Lattice<robj> & r){
-    this->checkerboard = r.checkerboard;
-    conformable(*this,r);
-    
-    parallel_for(int ss=0;ss<_grid->oSites();ss++){
-      this->_odata[ss]=r._odata[ss];
-    }
-    return *this;
-  }
-
-  virtual ~Lattice(void) = default;
+  //  virtual ~Lattice(void) = default;
     
   void reset(GridBase* grid) {
-    if (_grid != grid) {
-      _grid = grid;
-      _odata.resize(grid->oSites());
-      checkerboard = 0;
+    if (this->_grid != grid) {
+      this->_grid = grid;
+      this->_odata.resize(grid->oSites());
+      this->checkerboard = 0;
     }
   }
-  
-
-  template<class sobj> strong_inline Lattice<vobj> & operator = (const sobj & r){
-    parallel_for(int ss=0;ss<_grid->oSites();ss++){
-      this->_odata[ss]=r;
-    }
+  ///////////////////////////////////////////
+  // copy constructor
+  ///////////////////////////////////////////
+  Lattice(const Lattice& r){ 
+    //    std::cout << "Lattice constructor(const Lattice &) "<<this<<std::endl; 
+    this->_grid = r.Grid();
+    resize(this->_grid->oSites());
+    *this = r;
+  }
+  ///////////////////////////////////////////
+  // move constructor
+  ///////////////////////////////////////////
+  Lattice(Lattice && r){ 
+    this->_grid = r.Grid();
+    this->_odata      = r._odata;
+    this->_odata_size = r._odata_size;
+    this->checkerboard= r.Checkerboard();
+    r._odata      = nullptr;
+    r._odata_size = 0;
+  }
+  ///////////////////////////////////////////
+  // assignment template
+  ///////////////////////////////////////////
+  template<class robj> inline Lattice<vobj> & operator = (const Lattice<robj> & r){
+    typename std::enable_if<!std::is_same<robj,vobj>::value,int>::type i=0;
+    conformable(*this,r);
+    this->checkerboard = r.Checkerboard();
+    auto me =   View();
+    auto him= r.View();
+    accelerator_for(ss,me.size(),vobj::Nsimd(),{
+      coalescedWrite(me[ss],him(ss));
+    });
     return *this;
   }
-  
-  
+
+  ///////////////////////////////////////////
+  // Copy assignment 
+  ///////////////////////////////////////////
+  inline Lattice<vobj> & operator = (const Lattice<vobj> & r){
+    this->checkerboard = r.Checkerboard();
+    conformable(*this,r);
+    auto me =   View();
+    auto him= r.View();
+    accelerator_for(ss,me.size(),vobj::Nsimd(),{
+      coalescedWrite(me[ss],him(ss));
+    });
+    return *this;
+  }
+  ///////////////////////////////////////////
+  // Move assignment possible if same type
+  ///////////////////////////////////////////
+  inline Lattice<vobj> & operator = (Lattice<vobj> && r){
+
+    resize(0); // deletes if appropriate
+    this->_grid       = r.Grid();
+    this->_odata      = r._odata;
+    this->_odata_size = r._odata_size;
+    this->checkerboard= r.Checkerboard();
+
+    r._odata      = nullptr;
+    r._odata_size = 0;
+    
+    return *this;
+  }
+
+  /////////////////////////////////////////////////////////////////////////////
   // *=,+=,-= operators inherit behvour from correspond */+/- operation
-  template<class T> strong_inline Lattice<vobj> &operator *=(const T &r) {
+  /////////////////////////////////////////////////////////////////////////////
+  template<class T> inline Lattice<vobj> &operator *=(const T &r) {
     *this = (*this)*r;
     return *this;
   }
   
-  template<class T> strong_inline Lattice<vobj> &operator -=(const T &r) {
+  template<class T> inline Lattice<vobj> &operator -=(const T &r) {
     *this = (*this)-r;
     return *this;
   }
-  template<class T> strong_inline Lattice<vobj> &operator +=(const T &r) {
+  template<class T> inline Lattice<vobj> &operator +=(const T &r) {
     *this = (*this)+r;
     return *this;
   }
-}; // class Lattice
-  
-  template<class vobj> std::ostream& operator<< (std::ostream& stream, const Lattice<vobj> &o){
-    std::vector<int> gcoor;
-    typedef typename vobj::scalar_object sobj;
-    sobj ss;
-    for(int g=0;g<o._grid->_gsites;g++){
-      o._grid->GlobalIndexToGlobalCoor(g,gcoor);
-      peekSite(ss,o,gcoor);
-      stream<<"[";
-      for(int d=0;d<gcoor.size();d++){
-	stream<<gcoor[d];
-	if(d!=gcoor.size()-1) stream<<",";
-      }
-      stream<<"]\t";
-      stream<<ss<<std::endl;
-    }
-    return stream;
+
+  friend inline void swap(Lattice &l, Lattice &r) { 
+    conformable(l,r);
+    LatticeAccelerator<vobj> tmp;
+    LatticeAccelerator<vobj> *lp = (LatticeAccelerator<vobj> *)&l;
+    LatticeAccelerator<vobj> *rp = (LatticeAccelerator<vobj> *)&r;
+    tmp = *lp;    *lp=*rp;    *rp=tmp;
   }
-  
+
+}; // class Lattice
+
+template<class vobj> std::ostream& operator<< (std::ostream& stream, const Lattice<vobj> &o){
+  typedef typename vobj::scalar_object sobj;
+  for(int g=0;g<o.Grid()->_gsites;g++){
+
+    Coordinate gcoor;
+    o.Grid()->GlobalIndexToGlobalCoor(g,gcoor);
+
+    sobj ss;
+    peekSite(ss,o,gcoor);
+    stream<<"[";
+    for(int d=0;d<gcoor.size();d++){
+      stream<<gcoor[d];
+      if(d!=gcoor.size()-1) stream<<",";
+    }
+    stream<<"]\t";
+    stream<<ss<<std::endl;
+  }
+  return stream;
 }
+  
+NAMESPACE_END(Grid);
 
-
-
-#include "Lattice_conformable.h"
-#define GRID_LATTICE_EXPRESSION_TEMPLATES
-#ifdef  GRID_LATTICE_EXPRESSION_TEMPLATES
-#include "Lattice_ET.h"
-#else 
-#include "Lattice_overload.h"
-#endif
-#include "Lattice_arith.h"
-#include "Lattice_trace.h"
-#include "Lattice_transpose.h"
-#include "Lattice_local.h"
-#include "Lattice_reduction.h"
-#include "Lattice_peekpoke.h"
-#include "Lattice_reality.h"
-#include "Lattice_comparison_utils.h"
-#include "Lattice_comparison.h"
-#include "Lattice_coordinate.h"
-#include "Lattice_where.h"
-#include "Lattice_rng.h"
-#include "Lattice_unary.h"
-#include "Lattice_transfer.h"
-
-
-#endif
diff --git a/Grid/lattice/Lattice_comparison.h b/Grid/lattice/Lattice_comparison.h
index 9bf1fb2d..bbed2ef5 100644
--- a/Grid/lattice/Lattice_comparison.h
+++ b/Grid/lattice/Lattice_comparison.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,146 +24,184 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_LATTICE_COMPARISON_H
 #define GRID_LATTICE_COMPARISON_H
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-    //////////////////////////////////////////////////////////////////////////
-    // relational operators
-    // 
-    // Support <,>,<=,>=,==,!=
-    //
-    //Query supporting bitwise &, |, ^, !
-    //Query supporting logical &&, ||, 
-    //////////////////////////////////////////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////
+// relational operators
+// 
+// Support <,>,<=,>=,==,!=
+//
+//Query supporting bitwise &, |, ^, !
+//Query supporting logical &&, ||, 
+//////////////////////////////////////////////////////////////////////////
 
-  //////////////////////////////////////////////////////////////////////////
-  // compare lattice to lattice
-  //////////////////////////////////////////////////////////////////////////
-  template<class vfunctor,class lobj,class robj>  
-    inline Lattice<vInteger> LLComparison(vfunctor op,const Lattice<lobj> &lhs,const Lattice<robj> &rhs)
-  {
-    Lattice<vInteger> ret(rhs._grid);
-    parallel_for(int ss=0;ss<rhs._grid->oSites(); ss++){
-      ret._odata[ss]=op(lhs._odata[ss],rhs._odata[ss]);
-    }
-    return ret;
+typedef iScalar<vInteger> vPredicate ;
+
+/*
+template <class iobj, class vobj, class robj> accelerator_inline 
+vobj predicatedWhere(const iobj &predicate, const vobj &iftrue, const robj &iffalse) 
+{
+  typename std::remove_const<vobj>::type ret;
+
+  typedef typename vobj::scalar_object scalar_object;
+  typedef typename vobj::scalar_type scalar_type;
+  typedef typename vobj::vector_type vector_type;
+
+  const int Nsimd = vobj::vector_type::Nsimd();
+
+  ExtractBuffer<Integer> mask(Nsimd);
+  ExtractBuffer<scalar_object> truevals(Nsimd);
+  ExtractBuffer<scalar_object> falsevals(Nsimd);
+
+  extract(iftrue, truevals);
+  extract(iffalse, falsevals);
+  extract<vInteger, Integer>(TensorRemove(predicate), mask);
+
+  for (int s = 0; s < Nsimd; s++) {
+    if (mask[s]) falsevals[s] = truevals[s];
   }
-  //////////////////////////////////////////////////////////////////////////
-  // compare lattice to scalar
-  //////////////////////////////////////////////////////////////////////////
-  template<class vfunctor,class lobj,class robj> 
-    inline Lattice<vInteger> LSComparison(vfunctor op,const Lattice<lobj> &lhs,const robj &rhs)
-  {
-    Lattice<vInteger> ret(lhs._grid);
-    parallel_for(int ss=0;ss<lhs._grid->oSites(); ss++){
-      ret._odata[ss]=op(lhs._odata[ss],rhs);
-    }
-    return ret;
-  }
-  //////////////////////////////////////////////////////////////////////////
-  // compare scalar to lattice
-  //////////////////////////////////////////////////////////////////////////
-  template<class vfunctor,class lobj,class robj> 
-    inline Lattice<vInteger> SLComparison(vfunctor op,const lobj &lhs,const Lattice<robj> &rhs)
-  {
-    Lattice<vInteger> ret(rhs._grid);
-    parallel_for(int ss=0;ss<rhs._grid->oSites(); ss++){
-      ret._odata[ss]=op(lhs._odata[ss],rhs);
-    }
-    return ret;
-  }
-  
-  //////////////////////////////////////////////////////////////////////////
-  // Map to functors
-  //////////////////////////////////////////////////////////////////////////
-  // Less than
-  template<class lobj,class robj>
-    inline Lattice<vInteger> operator < (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
-    return LLComparison(vlt<lobj,robj>(),lhs,rhs);
-  }
-  template<class lobj,class robj>
-    inline Lattice<vInteger> operator < (const Lattice<lobj> & lhs, const robj & rhs) {
-    return LSComparison(vlt<lobj,robj>(),lhs,rhs);
-  }
-  template<class lobj,class robj>
-    inline Lattice<vInteger> operator < (const lobj & lhs, const Lattice<robj> & rhs) {
-    return SLComparison(vlt<lobj,robj>(),lhs,rhs);
-  }
-  
-  // Less than equal
-  template<class lobj,class robj>
-    inline Lattice<vInteger> operator <= (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
-    return LLComparison(vle<lobj,robj>(),lhs,rhs);
-  }
-  template<class lobj,class robj>
-    inline Lattice<vInteger> operator <= (const Lattice<lobj> & lhs, const robj & rhs) {
-    return LSComparison(vle<lobj,robj>(),lhs,rhs);
-  }
-  template<class lobj,class robj>
-    inline Lattice<vInteger> operator <= (const lobj & lhs, const Lattice<robj> & rhs) {
-    return SLComparison(vle<lobj,robj>(),lhs,rhs);
-  }
-  
-  // Greater than 
-  template<class lobj,class robj>
-    inline Lattice<vInteger> operator > (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
-    return LLComparison(vgt<lobj,robj>(),lhs,rhs);
-  }
-  template<class lobj,class robj>
-    inline Lattice<vInteger> operator > (const Lattice<lobj> & lhs, const robj & rhs) {
-    return LSComparison(vgt<lobj,robj>(),lhs,rhs);
-  }
-  template<class lobj,class robj>
-    inline Lattice<vInteger> operator > (const lobj & lhs, const Lattice<robj> & rhs) {
-     return SLComparison(vgt<lobj,robj>(),lhs,rhs);
-  }
-  
-  
-  // Greater than equal
-   template<class lobj,class robj>
-     inline Lattice<vInteger> operator >= (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
-     return LLComparison(vge<lobj,robj>(),lhs,rhs);
-   }
-   template<class lobj,class robj>
-   inline Lattice<vInteger> operator >= (const Lattice<lobj> & lhs, const robj & rhs) {
-     return LSComparison(vge<lobj,robj>(),lhs,rhs);
-   }
-   template<class lobj,class robj>
-     inline Lattice<vInteger> operator >= (const lobj & lhs, const Lattice<robj> & rhs) {
-     return SLComparison(vge<lobj,robj>(),lhs,rhs);
-   }
-   
-   // equal
-   template<class lobj,class robj>
-     inline Lattice<vInteger> operator == (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
-     return LLComparison(veq<lobj,robj>(),lhs,rhs);
-   }
-   template<class lobj,class robj>
-     inline Lattice<vInteger> operator == (const Lattice<lobj> & lhs, const robj & rhs) {
-     return LSComparison(veq<lobj,robj>(),lhs,rhs);
-   }
-   template<class lobj,class robj>
-     inline Lattice<vInteger> operator == (const lobj & lhs, const Lattice<robj> & rhs) {
-     return SLComparison(veq<lobj,robj>(),lhs,rhs);
-   }
-   
-   
-   // not equal
-   template<class lobj,class robj>
-     inline Lattice<vInteger> operator != (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
-     return LLComparison(vne<lobj,robj>(),lhs,rhs);
-   }
-   template<class lobj,class robj>
-     inline Lattice<vInteger> operator != (const Lattice<lobj> & lhs, const robj & rhs) {
-     return LSComparison(vne<lobj,robj>(),lhs,rhs);
-   }
-   template<class lobj,class robj>
-     inline Lattice<vInteger> operator != (const lobj & lhs, const Lattice<robj> & rhs) {
-     return SLComparison(vne<lobj,robj>(),lhs,rhs);
-   }
+
+  merge(ret, falsevals);
+  return ret;
 }
+*/
+//////////////////////////////////////////////////////////////////////////
+// compare lattice to lattice
+//////////////////////////////////////////////////////////////////////////
+
+template<class vfunctor,class lobj,class robj>  
+inline Lattice<vPredicate> LLComparison(vfunctor op,const Lattice<lobj> &lhs,const Lattice<robj> &rhs)
+{
+  Lattice<vPredicate> ret(rhs.Grid());
+  auto lhs_v = lhs.View();
+  auto rhs_v = rhs.View();
+  auto ret_v = ret.View();
+  thread_for( ss, rhs_v.size(), {
+      ret_v[ss]=op(lhs_v[ss],rhs_v[ss]);
+  });
+  return ret;
+}
+//////////////////////////////////////////////////////////////////////////
+// compare lattice to scalar
+//////////////////////////////////////////////////////////////////////////
+template<class vfunctor,class lobj,class robj> 
+inline Lattice<vPredicate> LSComparison(vfunctor op,const Lattice<lobj> &lhs,const robj &rhs)
+{
+  Lattice<vPredicate> ret(lhs.Grid());
+  auto lhs_v = lhs.View();
+  auto ret_v = ret.View();
+  thread_for( ss, lhs_v.size(), {
+    ret_v[ss]=op(lhs_v[ss],rhs);
+  });
+  return ret;
+}
+//////////////////////////////////////////////////////////////////////////
+// compare scalar to lattice
+//////////////////////////////////////////////////////////////////////////
+template<class vfunctor,class lobj,class robj> 
+inline Lattice<vPredicate> SLComparison(vfunctor op,const lobj &lhs,const Lattice<robj> &rhs)
+{
+  Lattice<vPredicate> ret(rhs.Grid());
+  auto rhs_v = rhs.View();
+  auto ret_v = ret.View();
+  thread_for( ss, rhs_v.size(), {
+    ret_v[ss]=op(lhs,rhs_v[ss]);
+  });
+  return ret;
+}
+  
+//////////////////////////////////////////////////////////////////////////
+// Map to functors
+//////////////////////////////////////////////////////////////////////////
+// Less than
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator < (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
+  return LLComparison(vlt<lobj,robj>(),lhs,rhs);
+}
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator < (const Lattice<lobj> & lhs, const robj & rhs) {
+  return LSComparison(vlt<lobj,robj>(),lhs,rhs);
+}
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator < (const lobj & lhs, const Lattice<robj> & rhs) {
+  return SLComparison(vlt<lobj,robj>(),lhs,rhs);
+}
+  
+// Less than equal
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator <= (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
+  return LLComparison(vle<lobj,robj>(),lhs,rhs);
+}
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator <= (const Lattice<lobj> & lhs, const robj & rhs) {
+  return LSComparison(vle<lobj,robj>(),lhs,rhs);
+}
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator <= (const lobj & lhs, const Lattice<robj> & rhs) {
+  return SLComparison(vle<lobj,robj>(),lhs,rhs);
+}
+  
+// Greater than 
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator > (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
+  return LLComparison(vgt<lobj,robj>(),lhs,rhs);
+}
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator > (const Lattice<lobj> & lhs, const robj & rhs) {
+  return LSComparison(vgt<lobj,robj>(),lhs,rhs);
+}
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator > (const lobj & lhs, const Lattice<robj> & rhs) {
+  return SLComparison(vgt<lobj,robj>(),lhs,rhs);
+}
+  
+  
+// Greater than equal
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator >= (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
+  return LLComparison(vge<lobj,robj>(),lhs,rhs);
+}
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator >= (const Lattice<lobj> & lhs, const robj & rhs) {
+  return LSComparison(vge<lobj,robj>(),lhs,rhs);
+}
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator >= (const lobj & lhs, const Lattice<robj> & rhs) {
+  return SLComparison(vge<lobj,robj>(),lhs,rhs);
+}
+   
+// equal
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator == (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
+  return LLComparison(veq<lobj,robj>(),lhs,rhs);
+}
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator == (const Lattice<lobj> & lhs, const robj & rhs) {
+  return LSComparison(veq<lobj,robj>(),lhs,rhs);
+}
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator == (const lobj & lhs, const Lattice<robj> & rhs) {
+  return SLComparison(veq<lobj,robj>(),lhs,rhs);
+}
+   
+   
+// not equal
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator != (const Lattice<lobj> & lhs, const Lattice<robj> & rhs) {
+  return LLComparison(vne<lobj,robj>(),lhs,rhs);
+}
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator != (const Lattice<lobj> & lhs, const robj & rhs) {
+  return LSComparison(vne<lobj,robj>(),lhs,rhs);
+}
+template<class lobj,class robj>
+inline Lattice<vPredicate> operator != (const lobj & lhs, const Lattice<robj> & rhs) {
+  return SLComparison(vne<lobj,robj>(),lhs,rhs);
+}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/lattice/Lattice_comparison_utils.h b/Grid/lattice/Lattice_comparison_utils.h
index cbac20ec..431aa9e1 100644
--- a/Grid/lattice/Lattice_comparison_utils.h
+++ b/Grid/lattice/Lattice_comparison_utils.h
@@ -26,10 +26,10 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     See the full license in the file "LICENSE" in the top level distribution directory
     *************************************************************************************/
     /*  END LEGAL */
-#ifndef GRID_COMPARISON_H
-#define GRID_COMPARISON_H
 
-namespace Grid {
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
 
   /////////////////////////////////////////
   // This implementation is a bit poor.
@@ -44,42 +44,42 @@ namespace Grid {
   //
   template<class lobj,class robj> class veq {
   public:
-    vInteger operator()(const lobj &lhs, const robj &rhs)
+    accelerator vInteger operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) == (rhs);
     }
   };
   template<class lobj,class robj> class vne {
   public:
-    vInteger operator()(const lobj &lhs, const robj &rhs)
+    accelerator vInteger operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) != (rhs);
     }
   };
   template<class lobj,class robj> class vlt {
   public:
-    vInteger operator()(const lobj &lhs, const robj &rhs)
+    accelerator vInteger operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) < (rhs);
     }
   };
   template<class lobj,class robj> class vle {
   public:
-    vInteger operator()(const lobj &lhs, const robj &rhs)
+    accelerator vInteger operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) <= (rhs);
     }
   };
   template<class lobj,class robj> class vgt {
   public:
-    vInteger operator()(const lobj &lhs, const robj &rhs)
+    accelerator vInteger operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) > (rhs);
     }
   };
   template<class lobj,class robj> class vge {
     public:
-    vInteger operator()(const lobj &lhs, const robj &rhs)
+    accelerator vInteger operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) >= (rhs);
     }
@@ -88,42 +88,42 @@ namespace Grid {
   // Generic list of functors
   template<class lobj,class robj> class seq {
   public:
-    Integer operator()(const lobj &lhs, const robj &rhs)
+    accelerator Integer operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) == (rhs);
     }
   };
   template<class lobj,class robj> class sne {
   public:
-    Integer operator()(const lobj &lhs, const robj &rhs)
+    accelerator Integer operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) != (rhs);
     }
   };
   template<class lobj,class robj> class slt {
   public:
-    Integer operator()(const lobj &lhs, const robj &rhs)
+    accelerator Integer operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) < (rhs);
     }
   };
   template<class lobj,class robj> class sle {
   public:
-    Integer operator()(const lobj &lhs, const robj &rhs)
+    accelerator Integer operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) <= (rhs);
     }
   };
   template<class lobj,class robj> class sgt {
   public:
-    Integer operator()(const lobj &lhs, const robj &rhs)
+    accelerator Integer operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) > (rhs);
     }
   };
   template<class lobj,class robj> class sge {
   public:
-    Integer operator()(const lobj &lhs, const robj &rhs)
+    accelerator Integer operator()(const lobj &lhs, const robj &rhs)
     { 
       return (lhs) >= (rhs);
     }
@@ -133,12 +133,12 @@ namespace Grid {
   // Integer and real get extra relational functions.
   //////////////////////////////////////////////////////////////////////////////////////////////////////
   template<class sfunctor, class vsimd,IfNotComplex<vsimd> = 0> 
-    inline vInteger Comparison(sfunctor sop,const vsimd & lhs, const vsimd & rhs)
+    accelerator_inline vInteger Comparison(sfunctor sop,const vsimd & lhs, const vsimd & rhs)
     {
       typedef typename vsimd::scalar_type scalar;
-      std::vector<scalar> vlhs(vsimd::Nsimd());   // Use functors to reduce this to single implementation
-      std::vector<scalar> vrhs(vsimd::Nsimd());
-      std::vector<Integer> vpred(vsimd::Nsimd());
+      ExtractBuffer<scalar> vlhs(vsimd::Nsimd());   // Use functors to reduce this to single implementation
+      ExtractBuffer<scalar> vrhs(vsimd::Nsimd());
+      ExtractBuffer<Integer> vpred(vsimd::Nsimd());
       vInteger ret;
       extract<vsimd,scalar>(lhs,vlhs);
       extract<vsimd,scalar>(rhs,vrhs);
@@ -150,11 +150,11 @@ namespace Grid {
     }
 
   template<class sfunctor, class vsimd,IfNotComplex<vsimd> = 0> 
-    inline vInteger Comparison(sfunctor sop,const vsimd & lhs, const typename vsimd::scalar_type & rhs)
+    accelerator_inline vInteger Comparison(sfunctor sop,const vsimd & lhs, const typename vsimd::scalar_type & rhs)
     {
       typedef typename vsimd::scalar_type scalar;
-      std::vector<scalar> vlhs(vsimd::Nsimd());   // Use functors to reduce this to single implementation
-      std::vector<Integer> vpred(vsimd::Nsimd());
+      ExtractBuffer<scalar> vlhs(vsimd::Nsimd());   // Use functors to reduce this to single implementation
+      ExtractBuffer<Integer> vpred(vsimd::Nsimd());
       vInteger ret;
       extract<vsimd,scalar>(lhs,vlhs);
       for(int s=0;s<vsimd::Nsimd();s++){
@@ -165,11 +165,11 @@ namespace Grid {
     }
 
   template<class sfunctor, class vsimd,IfNotComplex<vsimd> = 0> 
-    inline vInteger Comparison(sfunctor sop,const typename vsimd::scalar_type & lhs, const vsimd & rhs)
+    accelerator_inline vInteger Comparison(sfunctor sop,const typename vsimd::scalar_type & lhs, const vsimd & rhs)
     {
       typedef typename vsimd::scalar_type scalar;
-      std::vector<scalar> vrhs(vsimd::Nsimd());   // Use functors to reduce this to single implementation
-      std::vector<Integer> vpred(vsimd::Nsimd());
+      ExtractBuffer<scalar> vrhs(vsimd::Nsimd());   // Use functors to reduce this to single implementation
+      ExtractBuffer<Integer> vpred(vsimd::Nsimd());
       vInteger ret;
       extract<vsimd,scalar>(rhs,vrhs);
       for(int s=0;s<vsimd::Nsimd();s++){
@@ -181,30 +181,30 @@ namespace Grid {
 
 #define DECLARE_RELATIONAL_EQ(op,functor) \
   template<class vsimd,IfSimd<vsimd> = 0>\
-    inline vInteger operator op (const vsimd & lhs, const vsimd & rhs)\
+    accelerator_inline vInteger operator op (const vsimd & lhs, const vsimd & rhs)\
     {\
       typedef typename vsimd::scalar_type scalar;\
       return Comparison(functor<scalar,scalar>(),lhs,rhs);\
     }\
   template<class vsimd,IfSimd<vsimd> = 0>\
-    inline vInteger operator op (const vsimd & lhs, const typename vsimd::scalar_type & rhs) \
+    accelerator_inline vInteger operator op (const vsimd & lhs, const typename vsimd::scalar_type & rhs) \
     {\
       typedef typename vsimd::scalar_type scalar;\
       return Comparison(functor<scalar,scalar>(),lhs,rhs);\
     }\
   template<class vsimd,IfSimd<vsimd> = 0>\
-    inline vInteger operator op (const typename vsimd::scalar_type & lhs, const vsimd & rhs) \
+    accelerator_inline vInteger operator op (const typename vsimd::scalar_type & lhs, const vsimd & rhs) \
     {\
       typedef typename vsimd::scalar_type scalar;\
       return Comparison(functor<scalar,scalar>(),lhs,rhs);\
     }\
   template<class vsimd>\
-    inline vInteger operator op(const iScalar<vsimd> &lhs,const typename vsimd::scalar_type &rhs) \
+    accelerator_inline vInteger operator op(const iScalar<vsimd> &lhs,const typename vsimd::scalar_type &rhs) \
     {									\
       return lhs._internal op rhs;					\
     }									\
   template<class vsimd>\
-    inline vInteger operator op(const typename vsimd::scalar_type &lhs,const iScalar<vsimd> &rhs) \
+    accelerator_inline vInteger operator op(const typename vsimd::scalar_type &lhs,const iScalar<vsimd> &rhs) \
     {									\
       return lhs op rhs._internal;					\
     }									\
@@ -212,7 +212,7 @@ namespace Grid {
 #define DECLARE_RELATIONAL(op,functor) \
   DECLARE_RELATIONAL_EQ(op,functor)    \
   template<class vsimd>\
-    inline vInteger operator op(const iScalar<vsimd> &lhs,const iScalar<vsimd> &rhs)\
+    accelerator_inline vInteger operator op(const iScalar<vsimd> &lhs,const iScalar<vsimd> &rhs)\
     {									\
       return lhs._internal op rhs._internal;				\
     }									
@@ -226,7 +226,7 @@ DECLARE_RELATIONAL(!=,sne);
 
 #undef DECLARE_RELATIONAL
 
-}
+NAMESPACE_END(Grid);
+
 
 
-#endif
diff --git a/Grid/lattice/Lattice_conformable.h b/Grid/lattice/Lattice_conformable.h
index 398b3c90..ce22685e 100644
--- a/Grid/lattice/Lattice_conformable.h
+++ b/Grid/lattice/Lattice_conformable.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,18 +23,18 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_LATTICE_CONFORMABLE_H
 #define GRID_LATTICE_CONFORMABLE_H
 
-namespace Grid {
-
-    template<class obj1,class obj2> void conformable(const Lattice<obj1> &lhs,const Lattice<obj2> &rhs)
-    {
-        assert(lhs._grid == rhs._grid);
-        assert(lhs.checkerboard == rhs.checkerboard);
-    }
+NAMESPACE_BEGIN(Grid);
 
+template<class obj1,class obj2> void conformable(const Lattice<obj1> &lhs,const Lattice<obj2> &rhs)
+{
+  assert(lhs.Grid() == rhs.Grid());
+  assert(lhs.Checkerboard() == rhs.Checkerboard());
 }
+
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/lattice/Lattice_coordinate.h b/Grid/lattice/Lattice_coordinate.h
index 19eceba8..16f3641b 100644
--- a/Grid/lattice/Lattice_coordinate.h
+++ b/Grid/lattice/Lattice_coordinate.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,34 +23,52 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef GRID_LATTICE_COORDINATE_H
-#define GRID_LATTICE_COORDINATE_H
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once 
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-    template<class iobj> inline void LatticeCoordinate(Lattice<iobj> &l,int mu)
-    {
-      typedef typename iobj::scalar_type scalar_type;
-      typedef typename iobj::vector_type vector_type;
+template<class iobj> inline void LatticeCoordinate(Lattice<iobj> &l,int mu)
+{
+  typedef typename iobj::scalar_type scalar_type;
+  typedef typename iobj::vector_type vector_type;
 
-      GridBase *grid = l._grid;
-      int Nsimd = grid->iSites();
+  GridBase *grid = l.Grid();
+  int Nsimd = grid->iSites();
 
-      std::vector<int> gcoor;
-      std::vector<scalar_type> mergebuf(Nsimd);
+  Coordinate gcoor;
+  ExtractBuffer<scalar_type> mergebuf(Nsimd);
 
-      vector_type vI;
-      for(int o=0;o<grid->oSites();o++){
-	for(int i=0;i<grid->iSites();i++){
-	  grid->RankIndexToGlobalCoor(grid->ThisRank(),o,i,gcoor);
-	  mergebuf[i]=(Integer)gcoor[mu];
-	}
-	merge<vector_type,scalar_type>(vI,mergebuf);
-	l._odata[o]=vI;
+  vector_type vI;
+  auto l_v = l.View();
+  for(int o=0;o<grid->oSites();o++){
+    for(int i=0;i<grid->iSites();i++){
+      grid->RankIndexToGlobalCoor(grid->ThisRank(),o,i,gcoor);
+      mergebuf[i]=(Integer)gcoor[mu];
+    }
+    merge<vector_type,scalar_type>(vI,mergebuf);
+    l_v[o]=vI;
+  }
+};
+
+// LatticeCoordinate();
+// FIXME for debug; deprecate this; made obscelete by 
+template<class vobj> void lex_sites(Lattice<vobj> &l){
+  auto l_v = l.View();
+  Real *v_ptr = (Real *)&l_v[0];
+  size_t o_len = l.Grid()->oSites();
+  size_t v_len = sizeof(vobj)/sizeof(vRealF);
+  size_t vec_len = vRealF::Nsimd();
+
+  for(int i=0;i<o_len;i++){
+    for(int j=0;j<v_len;j++){
+      for(int vv=0;vv<vec_len;vv+=2){
+	v_ptr[i*v_len*vec_len+j*vec_len+vv  ]= i+vv*500;
+	v_ptr[i*v_len*vec_len+j*vec_len+vv+1]= i+vv*500;
       }
-    };
-
+    }}
 }
-#endif
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/lattice/Lattice_local.h b/Grid/lattice/Lattice_local.h
index 9dae1cd9..99455e24 100644
--- a/Grid/lattice/Lattice_local.h
+++ b/Grid/lattice/Lattice_local.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,8 +23,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_LATTICE_LOCALREDUCTION_H
 #define GRID_LATTICE_LOCALREDUCTION_H
 
@@ -32,44 +32,56 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 // localInner, localNorm, outerProduct
 ///////////////////////////////////////////////
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  /////////////////////////////////////////////////////
-  // Non site, reduced locally reduced routines
-  /////////////////////////////////////////////////////
+/////////////////////////////////////////////////////
+// Non site, reduced locally reduced routines
+/////////////////////////////////////////////////////
   
-  // localNorm2,
-  template<class vobj>
-    inline auto localNorm2 (const Lattice<vobj> &rhs)-> Lattice<typename vobj::tensor_reduced>
-    {
-      Lattice<typename vobj::tensor_reduced> ret(rhs._grid);
-      parallel_for(int ss=0;ss<rhs._grid->oSites(); ss++){
-	ret._odata[ss]=innerProduct(rhs._odata[ss],rhs._odata[ss]);
-      }
-      return ret;
-    }
-  
-  // localInnerProduct
-  template<class vobj>
-    inline auto localInnerProduct (const Lattice<vobj> &lhs,const Lattice<vobj> &rhs) -> Lattice<typename vobj::tensor_reduced>
-    {
-      Lattice<typename vobj::tensor_reduced> ret(rhs._grid);
-      parallel_for(int ss=0;ss<rhs._grid->oSites(); ss++){
-	ret._odata[ss]=innerProduct(lhs._odata[ss],rhs._odata[ss]);
-      }
-      return ret;
-    }
-  
-  // outerProduct Scalar x Scalar -> Scalar
-  //              Vector x Vector -> Matrix
-  template<class ll,class rr>
-    inline auto outerProduct (const Lattice<ll> &lhs,const Lattice<rr> &rhs) -> Lattice<decltype(outerProduct(lhs._odata[0],rhs._odata[0]))>
-  {
-    Lattice<decltype(outerProduct(lhs._odata[0],rhs._odata[0]))> ret(rhs._grid);
-    parallel_for(int ss=0;ss<rhs._grid->oSites(); ss++){
-      ret._odata[ss]=outerProduct(lhs._odata[ss],rhs._odata[ss]);
-    }
-    return ret;
-  }
+// localNorm2,
+template<class vobj>
+inline auto localNorm2 (const Lattice<vobj> &rhs)-> Lattice<typename vobj::tensor_reduced>
+{
+  Lattice<typename vobj::tensor_reduced> ret(rhs.Grid());
+  auto rhs_v = rhs.View();
+  auto ret_v = ret.View();
+  accelerator_for(ss,rhs_v.size(),vobj::Nsimd(),{
+    coalescedWrite(ret_v[ss],innerProduct(rhs_v(ss),rhs_v(ss)));
+  });
+  return ret;
 }
+  
+// localInnerProduct
+template<class vobj>
+inline auto localInnerProduct (const Lattice<vobj> &lhs,const Lattice<vobj> &rhs) -> Lattice<typename vobj::tensor_reduced>
+{
+  Lattice<typename vobj::tensor_reduced> ret(rhs.Grid());
+  auto lhs_v = lhs.View();
+  auto rhs_v = rhs.View();
+  auto ret_v = ret.View();
+  accelerator_for(ss,rhs_v.size(),vobj::Nsimd(),{
+    coalescedWrite(ret_v[ss],innerProduct(lhs_v(ss),rhs_v(ss)));
+  });
+  return ret;
+}
+  
+// outerProduct Scalar x Scalar -> Scalar
+//              Vector x Vector -> Matrix
+template<class ll,class rr>
+inline auto outerProduct (const Lattice<ll> &lhs,const Lattice<rr> &rhs) -> Lattice<decltype(outerProduct(ll(),rr()))>
+{
+  typedef decltype(coalescedRead(ll())) sll;
+  typedef decltype(coalescedRead(rr())) srr;
+  Lattice<decltype(outerProduct(ll(),rr()))> ret(rhs.Grid());
+  auto lhs_v = lhs.View();
+  auto rhs_v = rhs.View();
+  auto ret_v = ret.View();
+  accelerator_for(ss,rhs_v.size(),1,{
+    // FIXME had issues with scalar version of outer 
+    // Use vector [] operator and don't read coalesce this loop
+    ret_v[ss]=outerProduct(lhs_v[ss],rhs_v[ss]);
+  });
+  return ret;
+}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/lattice/Lattice_matrix_reduction.h b/Grid/lattice/Lattice_matrix_reduction.h
new file mode 100644
index 00000000..0980ad8a
--- /dev/null
+++ b/Grid/lattice/Lattice_matrix_reduction.h
@@ -0,0 +1,202 @@
+/*************************************************************************************
+    Grid physics library, www.github.com/paboyle/Grid 
+    Source file: ./lib/lattice/Lattice_reduction.h
+    Copyright (C) 2015
+Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once 
+#include <Grid/Grid_Eigen_Dense.h>
+
+#ifdef GRID_WARN_SUBOPTIMAL
+#warning "Optimisation alert all these reduction loops are NOT threaded "
+#endif     
+
+NAMESPACE_BEGIN(Grid);
+
+template<class vobj>
+static void sliceMaddMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice<vobj> &X,const Lattice<vobj> &Y,int Orthog,RealD scale=1.0) 
+{    
+  typedef typename vobj::scalar_object sobj;
+  typedef typename vobj::scalar_type scalar_type;
+  typedef typename vobj::vector_type vector_type;
+
+  int Nblock = X.Grid()->GlobalDimensions()[Orthog];
+
+  GridBase *FullGrid  = X.Grid();
+  //  GridBase *SliceGrid = makeSubSliceGrid(FullGrid,Orthog);
+
+  //  Lattice<vobj> Xslice(SliceGrid);
+  //  Lattice<vobj> Rslice(SliceGrid);
+
+  assert( FullGrid->_simd_layout[Orthog]==1);
+
+  //FIXME package in a convenient iterator
+  //Should loop over a plane orthogonal to direction "Orthog"
+  int stride=FullGrid->_slice_stride[Orthog];
+  int block =FullGrid->_slice_block [Orthog];
+  int nblock=FullGrid->_slice_nblock[Orthog];
+  int ostride=FullGrid->_ostride[Orthog];
+  auto X_v = X.View();
+  auto Y_v = Y.View();
+  auto R_v = R.View();
+  thread_region
+  {
+    std::vector<vobj> s_x(Nblock);
+
+    thread_loop_collapse2( (int n=0;n<nblock;n++),{
+      for(int b=0;b<block;b++){
+	int o  = n*stride + b;
+
+	for(int i=0;i<Nblock;i++){
+	  s_x[i] = X_v[o+i*ostride];
+	}
+
+	vobj dot;
+	for(int i=0;i<Nblock;i++){
+	  dot = Y_v[o+i*ostride];
+	  for(int j=0;j<Nblock;j++){
+	    dot = dot + s_x[j]*(scale*aa(j,i));
+	  }
+	  R_v[o+i*ostride]=dot;
+	}
+      }});
+  }
+};
+
+template<class vobj>
+static void sliceMulMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice<vobj> &X,int Orthog,RealD scale=1.0) 
+{    
+  typedef typename vobj::scalar_object sobj;
+  typedef typename vobj::scalar_type scalar_type;
+  typedef typename vobj::vector_type vector_type;
+
+  int Nblock = X.Grid()->GlobalDimensions()[Orthog];
+
+  GridBase *FullGrid  = X.Grid();
+  assert( FullGrid->_simd_layout[Orthog]==1);
+
+  //FIXME package in a convenient iterator
+  //Should loop over a plane orthogonal to direction "Orthog"
+  int stride=FullGrid->_slice_stride[Orthog];
+  int block =FullGrid->_slice_block [Orthog];
+  int nblock=FullGrid->_slice_nblock[Orthog];
+  int ostride=FullGrid->_ostride[Orthog];
+
+  auto X_v = X.View();
+  auto R_v = R.View();
+
+  thread_region
+  {
+    std::vector<vobj> s_x(Nblock);
+    
+    thread_loop_collapse2( (int n=0;n<nblock;n++),{
+      for(int b=0;b<block;b++){
+	int o  = n*stride + b;
+
+	for(int i=0;i<Nblock;i++){
+	  s_x[i] = X_v[o+i*ostride];
+	}
+
+	vobj dot;
+	for(int i=0;i<Nblock;i++){
+	  dot = s_x[0]*(scale*aa(0,i));
+	  for(int j=1;j<Nblock;j++){
+	    dot = dot + s_x[j]*(scale*aa(j,i));
+	  }
+	  R_v[o+i*ostride]=dot;
+	}
+    }});
+  }
+
+};
+
+
+template<class vobj>
+static void sliceInnerProductMatrix(  Eigen::MatrixXcd &mat, const Lattice<vobj> &lhs,const Lattice<vobj> &rhs,int Orthog) 
+{
+  typedef typename vobj::scalar_object sobj;
+  typedef typename vobj::scalar_type scalar_type;
+  typedef typename vobj::vector_type vector_type;
+  
+  GridBase *FullGrid  = lhs.Grid();
+  //  GridBase *SliceGrid = makeSubSliceGrid(FullGrid,Orthog);
+  
+  int Nblock = FullGrid->GlobalDimensions()[Orthog];
+  
+  //  Lattice<vobj> Lslice(SliceGrid);
+  //  Lattice<vobj> Rslice(SliceGrid);
+  
+  mat = Eigen::MatrixXcd::Zero(Nblock,Nblock);
+
+  assert( FullGrid->_simd_layout[Orthog]==1);
+  //  int nh =  FullGrid->_ndimension;
+  //  int nl = SliceGrid->_ndimension;
+  //  int nl = nh-1;
+
+  //FIXME package in a convenient iterator
+  //Should loop over a plane orthogonal to direction "Orthog"
+  int stride=FullGrid->_slice_stride[Orthog];
+  int block =FullGrid->_slice_block [Orthog];
+  int nblock=FullGrid->_slice_nblock[Orthog];
+  int ostride=FullGrid->_ostride[Orthog];
+
+  typedef typename vobj::vector_typeD vector_typeD;
+  auto lhs_v = lhs.View();
+  auto rhs_v = rhs.View();
+  thread_region {
+    std::vector<vobj> Left(Nblock);
+    std::vector<vobj> Right(Nblock);
+    Eigen::MatrixXcd  mat_thread = Eigen::MatrixXcd::Zero(Nblock,Nblock);
+
+    thread_loop_collapse2((int n=0;n<nblock;n++),{
+      for(int b=0;b<block;b++){
+
+	int o  = n*stride + b;
+
+	for(int i=0;i<Nblock;i++){
+	  Left [i] = lhs_v[o+i*ostride];
+	  Right[i] = rhs_v[o+i*ostride];
+	}
+
+	for(int i=0;i<Nblock;i++){
+	  for(int j=0;j<Nblock;j++){
+	    auto tmp = innerProduct(Left[i],Right[j]);
+	    auto rtmp = TensorRemove(tmp);
+	    ComplexD z = Reduce(rtmp);
+	    mat_thread(i,j) += std::complex<double>(real(z),imag(z));
+	  }}
+    }});
+    thread_critical {
+      mat += mat_thread;
+    }  
+  }
+
+  for(int i=0;i<Nblock;i++){
+    for(int j=0;j<Nblock;j++){
+      ComplexD sum = mat(i,j);
+      FullGrid->GlobalSum(sum);
+      mat(i,j)=sum;
+    }}
+
+  return;
+}
+
+NAMESPACE_END(Grid);
+
+
+
diff --git a/Grid/lattice/Lattice_overload.h b/Grid/lattice/Lattice_overload.h
deleted file mode 100644
index 0906b610..00000000
--- a/Grid/lattice/Lattice_overload.h
+++ /dev/null
@@ -1,138 +0,0 @@
-    /*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid 
-
-    Source file: ./lib/lattice/Lattice_overload.h
-
-    Copyright (C) 2015
-
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef GRID_LATTICE_OVERLOAD_H
-#define GRID_LATTICE_OVERLOAD_H
-
-namespace Grid {
-
-  //////////////////////////////////////////////////////////////////////////////////////////////////////
-  // unary negation
-  //////////////////////////////////////////////////////////////////////////////////////////////////////
-  template<class vobj>
-  inline Lattice<vobj> operator -(const Lattice<vobj> &r)
-  {
-    Lattice<vobj> ret(r._grid);
-    parallel_for(int ss=0;ss<r._grid->oSites();ss++){
-      vstream(ret._odata[ss], -r._odata[ss]);
-    }
-    return ret;
-  } 
-  /////////////////////////////////////////////////////////////////////////////////////
-  // Lattice BinOp Lattice,
-  //NB mult performs conformable check. Do not reapply here for performance.
-  /////////////////////////////////////////////////////////////////////////////////////
-  template<class left,class right>
-    inline auto operator * (const Lattice<left> &lhs,const Lattice<right> &rhs)-> Lattice<decltype(lhs._odata[0]*rhs._odata[0])>
-  {
-    Lattice<decltype(lhs._odata[0]*rhs._odata[0])> ret(rhs._grid);
-    mult(ret,lhs,rhs);
-    return ret;
-  }
-  template<class left,class right>
-    inline auto operator + (const Lattice<left> &lhs,const Lattice<right> &rhs)-> Lattice<decltype(lhs._odata[0]+rhs._odata[0])>
-  {
-    Lattice<decltype(lhs._odata[0]+rhs._odata[0])> ret(rhs._grid);
-    add(ret,lhs,rhs);
-    return ret;
-  }
-  template<class left,class right>
-    inline auto operator - (const Lattice<left> &lhs,const Lattice<right> &rhs)-> Lattice<decltype(lhs._odata[0]-rhs._odata[0])>
-  {
-    Lattice<decltype(lhs._odata[0]-rhs._odata[0])> ret(rhs._grid);
-    sub(ret,lhs,rhs);
-    return ret;
-  }
-  
-  // Scalar BinOp Lattice ;generate return type
-  template<class left,class right>
-  inline auto operator * (const left &lhs,const Lattice<right> &rhs) -> Lattice<decltype(lhs*rhs._odata[0])>
-  {
-    Lattice<decltype(lhs*rhs._odata[0])> ret(rhs._grid);
-    parallel_for(int ss=0;ss<rhs._grid->oSites(); ss++){
-      decltype(lhs*rhs._odata[0]) tmp=lhs*rhs._odata[ss]; 
-      vstream(ret._odata[ss],tmp);
-	   //      ret._odata[ss]=lhs*rhs._odata[ss];
-    }
-    return ret;
-  }
-  template<class left,class right>
-    inline auto operator + (const left &lhs,const Lattice<right> &rhs) -> Lattice<decltype(lhs+rhs._odata[0])>
-    {
-      Lattice<decltype(lhs+rhs._odata[0])> ret(rhs._grid);
-      parallel_for(int ss=0;ss<rhs._grid->oSites(); ss++){
-	decltype(lhs+rhs._odata[0]) tmp =lhs-rhs._odata[ss];  
-	vstream(ret._odata[ss],tmp);
-	//	ret._odata[ss]=lhs+rhs._odata[ss];
-      }
-        return ret;
-    }
-  template<class left,class right>
-    inline auto operator - (const left &lhs,const Lattice<right> &rhs) -> Lattice<decltype(lhs-rhs._odata[0])>
-  {
-    Lattice<decltype(lhs-rhs._odata[0])> ret(rhs._grid);
-    parallel_for(int ss=0;ss<rhs._grid->oSites(); ss++){
-      decltype(lhs-rhs._odata[0]) tmp=lhs-rhs._odata[ss];  
-      vstream(ret._odata[ss],tmp);
-    }
-    return ret;
-  }
-    template<class left,class right>
-      inline auto operator * (const Lattice<left> &lhs,const right &rhs) -> Lattice<decltype(lhs._odata[0]*rhs)>
-    {
-      Lattice<decltype(lhs._odata[0]*rhs)> ret(lhs._grid);
-      parallel_for(int ss=0;ss<lhs._grid->oSites(); ss++){
-	decltype(lhs._odata[0]*rhs) tmp =lhs._odata[ss]*rhs;
-	vstream(ret._odata[ss],tmp);
-	//            ret._odata[ss]=lhs._odata[ss]*rhs;
-      }
-      return ret;
-    }
-    template<class left,class right>
-      inline auto operator + (const Lattice<left> &lhs,const right &rhs) -> Lattice<decltype(lhs._odata[0]+rhs)>
-    {
-        Lattice<decltype(lhs._odata[0]+rhs)> ret(lhs._grid);
-	parallel_for(int ss=0;ss<rhs._grid->oSites(); ss++){
-	  decltype(lhs._odata[0]+rhs) tmp=lhs._odata[ss]+rhs; 
-	  vstream(ret._odata[ss],tmp);
-	  //	  ret._odata[ss]=lhs._odata[ss]+rhs;
-        }
-        return ret;
-    }
-    template<class left,class right>
-      inline auto operator - (const Lattice<left> &lhs,const right &rhs) -> Lattice<decltype(lhs._odata[0]-rhs)>
-    {
-      Lattice<decltype(lhs._odata[0]-rhs)> ret(lhs._grid);
-      parallel_for(int ss=0;ss<rhs._grid->oSites(); ss++){
-	  decltype(lhs._odata[0]-rhs) tmp=lhs._odata[ss]-rhs;
-	  vstream(ret._odata[ss],tmp);
-	  //	ret._odata[ss]=lhs._odata[ss]-rhs;
-      }
-      return ret;
-    }
-}
-#endif
diff --git a/Grid/lattice/Lattice_peekpoke.h b/Grid/lattice/Lattice_peekpoke.h
index 3d6268d2..c3e7e931 100644
--- a/Grid/lattice/Lattice_peekpoke.h
+++ b/Grid/lattice/Lattice_peekpoke.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,8 +25,8 @@ Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_LATTICE_PEEK_H
 #define GRID_LATTICE_PEEK_H
 
@@ -34,172 +34,184 @@ Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
 // Peeking and poking around
 ///////////////////////////////////////////////
 
-namespace Grid {
-
-    ////////////////////////////////////////////////////////////////////////////////////////////////////
-    // Peek internal indices of a Lattice object
-    ////////////////////////////////////////////////////////////////////////////////////////////////////
-    template<int Index,class vobj>
-       auto PeekIndex(const Lattice<vobj> &lhs,int i) -> Lattice<decltype(peekIndex<Index>(lhs._odata[0],i))>
-    {
-      Lattice<decltype(peekIndex<Index>(lhs._odata[0],i))> ret(lhs._grid);
-      ret.checkerboard=lhs.checkerboard;
-      parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-	ret._odata[ss] = peekIndex<Index>(lhs._odata[ss],i);
-      }
-      return ret;
-    };
-    template<int Index,class vobj>
-      auto PeekIndex(const Lattice<vobj> &lhs,int i,int j) -> Lattice<decltype(peekIndex<Index>(lhs._odata[0],i,j))>
-    {
-      Lattice<decltype(peekIndex<Index>(lhs._odata[0],i,j))> ret(lhs._grid);
-      ret.checkerboard=lhs.checkerboard;
-      parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-	ret._odata[ss] = peekIndex<Index>(lhs._odata[ss],i,j);
-      }
-      return ret;
-    };
-
-    ////////////////////////////////////////////////////////////////////////////////////////////////////
-    // Poke internal indices of a Lattice object
-    ////////////////////////////////////////////////////////////////////////////////////////////////////
-    template<int Index,class vobj> 
-    void PokeIndex(Lattice<vobj> &lhs,const Lattice<decltype(peekIndex<Index>(lhs._odata[0],0))> & rhs,int i)
-    {
-      parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-	pokeIndex<Index>(lhs._odata[ss],rhs._odata[ss],i);
-      }      
-    }
-    template<int Index,class vobj>
-      void PokeIndex(Lattice<vobj> &lhs,const Lattice<decltype(peekIndex<Index>(lhs._odata[0],0,0))> & rhs,int i,int j)
-    {
-      parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-	pokeIndex<Index>(lhs._odata[ss],rhs._odata[ss],i,j);
-      }      
-    }
-
-    //////////////////////////////////////////////////////
-    // Poke a scalar object into the SIMD array
-    //////////////////////////////////////////////////////
-    template<class vobj,class sobj>
-    void pokeSite(const sobj &s,Lattice<vobj> &l,const std::vector<int> &site){
-
-      GridBase *grid=l._grid;
-
-      typedef typename vobj::scalar_type scalar_type;
-      typedef typename vobj::vector_type vector_type;
-
-      int Nsimd = grid->Nsimd();
-
-      assert( l.checkerboard== l._grid->CheckerBoard(site));
-      assert( sizeof(sobj)*Nsimd == sizeof(vobj));
-
-      int rank,odx,idx;
-      // Optional to broadcast from node 0.
-      grid->GlobalCoorToRankIndex(rank,odx,idx,site);
-      grid->Broadcast(grid->BossRank(),s);
-
-      std::vector<sobj> buf(Nsimd);
-
-      // extract-modify-merge cycle is easiest way and this is not perf critical
-      if ( rank == grid->ThisRank() ) {
-	extract(l._odata[odx],buf);
-	buf[idx] = s;
-	merge(l._odata[odx],buf);
-      }
-
-      return;
-    };
+NAMESPACE_BEGIN(Grid);
 
 
-    //////////////////////////////////////////////////////////
-    // Peek a scalar object from the SIMD array
-    //////////////////////////////////////////////////////////
-    template<class vobj,class sobj>
-      void peekSite(sobj &s,const Lattice<vobj> &l,const std::vector<int> &site){
-        
-      GridBase *grid=l._grid;
-
-      typedef typename vobj::scalar_type scalar_type;
-      typedef typename vobj::vector_type vector_type;
-
-      int Nsimd = grid->Nsimd();
-
-      assert( l.checkerboard == l._grid->CheckerBoard(site));
-
-      int rank,odx,idx;
-      grid->GlobalCoorToRankIndex(rank,odx,idx,site);
-
-      std::vector<sobj> buf(Nsimd);
-      extract(l._odata[odx],buf);
-
-      s = buf[idx];
-
-      grid->Broadcast(rank,s);
-
-      return;
-    };
-
-
-    //////////////////////////////////////////////////////////
-    // Peek a scalar object from the SIMD array
-    //////////////////////////////////////////////////////////
-    template<class vobj,class sobj>
-    void peekLocalSite(sobj &s,const Lattice<vobj> &l,std::vector<int> &site){
-        
-      GridBase *grid = l._grid;
-
-      typedef typename vobj::scalar_type scalar_type;
-      typedef typename vobj::vector_type vector_type;
-
-      int Nsimd = grid->Nsimd();
-
-      assert( l.checkerboard== l._grid->CheckerBoard(site));
-      assert( sizeof(sobj)*Nsimd == sizeof(vobj));
-
-      static const int words=sizeof(vobj)/sizeof(vector_type);
-      int odx,idx;
-      idx= grid->iIndex(site);
-      odx= grid->oIndex(site);
-
-      scalar_type * vp = (scalar_type *)&l._odata[odx];
-      scalar_type * pt = (scalar_type *)&s;
-      
-      for(int w=0;w<words;w++){
-        pt[w] = vp[idx+w*Nsimd];
-      }
-      
-      return;
-    };
-
-    template<class vobj,class sobj>
-    void pokeLocalSite(const sobj &s,Lattice<vobj> &l,std::vector<int> &site){
-
-      GridBase *grid=l._grid;
-
-      typedef typename vobj::scalar_type scalar_type;
-      typedef typename vobj::vector_type vector_type;
-
-      int Nsimd = grid->Nsimd();
-
-      assert( l.checkerboard== l._grid->CheckerBoard(site));
-      assert( sizeof(sobj)*Nsimd == sizeof(vobj));
-
-      static const int words=sizeof(vobj)/sizeof(vector_type);
-      int odx,idx;
-      idx= grid->iIndex(site);
-      odx= grid->oIndex(site);
-
-      scalar_type * vp = (scalar_type *)&l._odata[odx];
-      scalar_type * pt = (scalar_type *)&s;
-      
-      for(int w=0;w<words;w++){
-        vp[idx+w*Nsimd] = pt[w];
-      }
-
-      return;
-    };
+// FIXME accelerator_loop and accelerator_inline these
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Peek internal indices of a Lattice object
+////////////////////////////////////////////////////////////////////////////////////////////////////
+template<int Index,class vobj> 
+auto PeekIndex(const Lattice<vobj> &lhs,int i) -> Lattice<decltype(peekIndex<Index>(vobj(),i))>
+{
+  Lattice<decltype(peekIndex<Index>(vobj(),i))> ret(lhs.Grid());
+  ret.Checkerboard()=lhs.Checkerboard();
+  auto ret_v = ret.View();
+  auto lhs_v = lhs.View();
+  thread_for( ss, lhs_v.size(), {
+    ret_v[ss] = peekIndex<Index>(lhs_v[ss],i);
+  });
+  return ret;
+};
+template<int Index,class vobj> 
+auto PeekIndex(const Lattice<vobj> &lhs,int i,int j) -> Lattice<decltype(peekIndex<Index>(vobj(),i,j))>
+{
+  Lattice<decltype(peekIndex<Index>(vobj(),i,j))> ret(lhs.Grid());
+  ret.Checkerboard()=lhs.Checkerboard();
+  auto ret_v = ret.View();
+  auto lhs_v = lhs.View();
+  thread_for( ss, lhs_v.size(), {
+    ret_v[ss] = peekIndex<Index>(lhs_v[ss],i,j);
+  });
+  return ret;
+};
 
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Poke internal indices of a Lattice object
+////////////////////////////////////////////////////////////////////////////////////////////////////
+template<int Index,class vobj>  
+void PokeIndex(Lattice<vobj> &lhs,const Lattice<decltype(peekIndex<Index>(vobj(),0))> & rhs,int i)
+{
+  auto rhs_v = rhs.View();
+  auto lhs_v = lhs.View();
+  thread_for( ss, lhs_v.size(), {
+    pokeIndex<Index>(lhs_v[ss],rhs_v[ss],i);
+  });
 }
+template<int Index,class vobj> 
+void PokeIndex(Lattice<vobj> &lhs,const Lattice<decltype(peekIndex<Index>(vobj(),0,0))> & rhs,int i,int j)
+{
+  auto rhs_v = rhs.View();
+  auto lhs_v = lhs.View();
+  thread_for( ss, lhs_v.size(), {
+    pokeIndex<Index>(lhs_v[ss],rhs_v[ss],i,j);
+  });
+}
+
+//////////////////////////////////////////////////////
+// Poke a scalar object into the SIMD array
+//////////////////////////////////////////////////////
+template<class vobj,class sobj> 
+void pokeSite(const sobj &s,Lattice<vobj> &l,const Coordinate &site){
+
+  GridBase *grid=l.Grid();
+
+  typedef typename vobj::scalar_type scalar_type;
+  typedef typename vobj::vector_type vector_type;
+
+  int Nsimd = grid->Nsimd();
+
+  assert( l.Checkerboard()== l.Grid()->CheckerBoard(site));
+  assert( sizeof(sobj)*Nsimd == sizeof(vobj));
+
+  int rank,odx,idx;
+  // Optional to broadcast from node 0.
+  grid->GlobalCoorToRankIndex(rank,odx,idx,site);
+  grid->Broadcast(grid->BossRank(),s);
+
+  // extract-modify-merge cycle is easiest way and this is not perf critical
+  ExtractBuffer<sobj> buf(Nsimd);
+  auto l_v = l.View();
+  if ( rank == grid->ThisRank() ) {
+    extract(l_v[odx],buf);
+    buf[idx] = s;
+    merge(l_v[odx],buf);
+  }
+
+  return;
+};
+
+
+//////////////////////////////////////////////////////////
+// Peek a scalar object from the SIMD array
+//////////////////////////////////////////////////////////
+template<class vobj,class sobj>
+void peekSite(sobj &s,const Lattice<vobj> &l,const Coordinate &site){
+        
+  GridBase *grid=l.Grid();
+
+  typedef typename vobj::scalar_type scalar_type;
+  typedef typename vobj::vector_type vector_type;
+
+  int Nsimd = grid->Nsimd();
+
+  assert( l.Checkerboard() == l.Grid()->CheckerBoard(site));
+
+  int rank,odx,idx;
+  grid->GlobalCoorToRankIndex(rank,odx,idx,site);
+
+  ExtractBuffer<sobj> buf(Nsimd);
+  auto l_v = l.View();
+  extract(l_v[odx],buf);
+
+  s = buf[idx];
+
+  grid->Broadcast(rank,s);
+
+  return;
+};
+
+
+//////////////////////////////////////////////////////////
+// Peek a scalar object from the SIMD array
+//////////////////////////////////////////////////////////
+template<class vobj,class sobj>
+void peekLocalSite(sobj &s,const Lattice<vobj> &l,Coordinate &site){
+        
+  GridBase *grid = l.Grid();
+
+  typedef typename vobj::scalar_type scalar_type;
+  typedef typename vobj::vector_type vector_type;
+
+  int Nsimd = grid->Nsimd();
+
+  assert( l.Checkerboard()== l.Grid()->CheckerBoard(site));
+  assert( sizeof(sobj)*Nsimd == sizeof(vobj));
+
+  static const int words=sizeof(vobj)/sizeof(vector_type);
+  int odx,idx;
+  idx= grid->iIndex(site);
+  odx= grid->oIndex(site);
+  
+  auto l_v = l.View();
+  scalar_type * vp = (scalar_type *)&l_v[odx];
+  scalar_type * pt = (scalar_type *)&s;
+      
+  for(int w=0;w<words;w++){
+    pt[w] = vp[idx+w*Nsimd];
+  }
+      
+  return;
+};
+
+template<class vobj,class sobj>
+void pokeLocalSite(const sobj &s,Lattice<vobj> &l,Coordinate &site){
+
+  GridBase *grid=l.Grid();
+
+  typedef typename vobj::scalar_type scalar_type;
+  typedef typename vobj::vector_type vector_type;
+
+  int Nsimd = grid->Nsimd();
+
+  assert( l.Checkerboard()== l.Grid()->CheckerBoard(site));
+  assert( sizeof(sobj)*Nsimd == sizeof(vobj));
+
+  static const int words=sizeof(vobj)/sizeof(vector_type);
+  int odx,idx;
+  idx= grid->iIndex(site);
+  odx= grid->oIndex(site);
+
+  auto l_v = l.View();
+  scalar_type * vp = (scalar_type *)&l_v[odx];
+  scalar_type * pt = (scalar_type *)&s;
+  for(int w=0;w<words;w++){
+    vp[idx+w*Nsimd] = pt[w];
+  }
+
+  return;
+};
+
+NAMESPACE_END(Grid);
 #endif
 
diff --git a/Grid/lattice/Lattice_reality.h b/Grid/lattice/Lattice_reality.h
index 7e7b2631..7373b2f9 100644
--- a/Grid/lattice/Lattice_reality.h
+++ b/Grid/lattice/Lattice_reality.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,8 +25,8 @@ Author: neo <cossu@post.kek.jp>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_LATTICE_REALITY_H
 #define GRID_LATTICE_REALITY_H
 
@@ -36,22 +36,28 @@ Author: neo <cossu@post.kek.jp>
 // The choice of burying complex in the SIMD
 // is making the use of "real" and "imag" very cumbersome
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-    template<class vobj> inline Lattice<vobj> adj(const Lattice<vobj> &lhs){
-        Lattice<vobj> ret(lhs._grid);
-	parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-            ret._odata[ss] = adj(lhs._odata[ss]);
-        }
-        return ret;
-    };
+template<class vobj> inline Lattice<vobj> adj(const Lattice<vobj> &lhs){
+  Lattice<vobj> ret(lhs.Grid());
+  auto lhs_v = lhs.View();
+  auto ret_v = ret.View();
+  accelerator_for( ss, lhs_v.size(), vobj::Nsimd(), {
+    coalescedWrite(ret_v[ss], adj(lhs_v(ss)));
+  });
+  return ret;
+};
+
+template<class vobj> inline Lattice<vobj> conjugate(const Lattice<vobj> &lhs){
+  Lattice<vobj> ret(lhs.Grid());
+  auto lhs_v = lhs.View();
+  auto ret_v = ret.View();
+  accelerator_for( ss, lhs_v.size(), vobj::Nsimd(), {
+    coalescedWrite( ret_v[ss] , conjugate(lhs_v(ss)));
+  });
+  return ret;
+};
+
+NAMESPACE_END(Grid);
 
-    template<class vobj> inline Lattice<vobj> conjugate(const Lattice<vobj> &lhs){
-        Lattice<vobj> ret(lhs._grid);
-	parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-	  ret._odata[ss] = conjugate(lhs._odata[ss]);
-        }
-        return ret;
-    };
-}
 #endif
diff --git a/Grid/lattice/Lattice_reduction.h b/Grid/lattice/Lattice_reduction.h
index 41351c98..3a076038 100644
--- a/Grid/lattice/Lattice_reduction.h
+++ b/Grid/lattice/Lattice_reduction.h
@@ -19,22 +19,76 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     See the full license in the file "LICENSE" in the top level distribution directory
     *************************************************************************************/
     /*  END LEGAL */
-#ifndef GRID_LATTICE_REDUCTION_H
-#define GRID_LATTICE_REDUCTION_H
+#pragma once
 
 #include <Grid/Grid_Eigen_Dense.h>
 
-namespace Grid {
-#ifdef GRID_WARN_SUBOPTIMAL
-#warning "Optimisation alert all these reduction loops are NOT threaded "
-#endif     
 
-  ////////////////////////////////////////////////////////////////////////////////////////////////////
-  // Deterministic Reduction operations
-  ////////////////////////////////////////////////////////////////////////////////////////////////////
+#ifdef GRID_NVCC
+#include <Grid/lattice/Lattice_reduction_gpu.h>
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+//////////////////////////////////////////////////////
+// FIXME this should promote to double and accumulate
+//////////////////////////////////////////////////////
+template<class vobj>
+inline typename vobj::scalar_object sum_cpu(const vobj *arg, Integer osites)
+{
+  typedef typename vobj::scalar_object  sobj;
+
+  const int Nsimd = vobj::Nsimd();
+  const int nthread = GridThread::GetThreads();
+
+  Vector<sobj> sumarray(nthread);
+  for(int i=0;i<nthread;i++){
+    sumarray[i]=Zero();
+  }
+  
+  thread_for(thr,nthread, {
+    int nwork, mywork, myoff;
+    nwork = osites;
+    GridThread::GetWork(nwork,thr,mywork,myoff);
+    vobj vvsum=Zero();
+    for(int ss=myoff;ss<mywork+myoff; ss++){
+      vvsum = vvsum + arg[ss];
+    }
+    sumarray[thr]=Reduce(vvsum);
+  });
+  
+  sobj ssum=Zero();  // sum across threads
+  for(int i=0;i<nthread;i++){
+    ssum = ssum+sumarray[i];
+  } 
+  
+  return ssum;
+}
+template<class vobj>
+inline typename vobj::scalar_object sum(const vobj *arg, Integer osites)
+{
+#ifdef GRID_NVCC
+  return sum_gpu(arg,osites);
+#else
+  return sum_cpu(arg,osites);
+#endif  
+}
+template<class vobj>
+inline typename vobj::scalar_object sum(const Lattice<vobj> &arg)
+{
+  auto arg_v = arg.View();
+  Integer osites = arg.Grid()->oSites();
+  auto ssum= sum(&arg_v[0],osites);
+  arg.Grid()->GlobalSum(ssum);
+  return ssum;
+}
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Deterministic Reduction operations
+////////////////////////////////////////////////////////////////////////////////////////////////////
 template<class vobj> inline RealD norm2(const Lattice<vobj> &arg){
-  auto nrm = innerProduct(arg,arg);
-  return std::real(nrm); 
+  ComplexD nrm = innerProduct(arg,arg);
+  return real(nrm); 
 }
 
 // Double inner product
@@ -43,32 +97,49 @@ inline ComplexD innerProduct(const Lattice<vobj> &left,const Lattice<vobj> &righ
 {
   typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_typeD vector_type;
-  GridBase *grid = left._grid;
-  const int pad = 8;
-
-  ComplexD  inner;
-  Vector<ComplexD> sumarray(grid->SumArraySize()*pad);
-
-  parallel_for(int thr=0;thr<grid->SumArraySize();thr++){
-    int nwork, mywork, myoff;
-    GridThread::GetWork(left._grid->oSites(),thr,mywork,myoff);
-    
-    decltype(innerProductD(left._odata[0],right._odata[0])) vinner=zero; // private to thread; sub summation
-    for(int ss=myoff;ss<mywork+myoff; ss++){
-      vinner = vinner + innerProductD(left._odata[ss],right._odata[ss]);
-    }
-    // All threads sum across SIMD; reduce serial work at end
-    // one write per cacheline with streaming store
-    ComplexD tmp = Reduce(TensorRemove(vinner)) ;
-    vstream(sumarray[thr*pad],tmp);
-  }
+  ComplexD  nrm;
   
-  inner=0.0;
-  for(int i=0;i<grid->SumArraySize();i++){
-    inner = inner+sumarray[i*pad];
-  } 
-  right._grid->GlobalSum(inner);
-  return inner;
+  GridBase *grid = left.Grid();
+  
+  // Might make all code paths go this way.
+  auto left_v = left.View();
+  auto right_v=right.View();
+
+  const uint64_t nsimd = grid->Nsimd();
+  const uint64_t sites = grid->oSites();
+  
+#ifdef GRID_NVCC
+  // GPU - SIMT lane compliance...
+  typedef decltype(innerProduct(left_v[0],right_v[0])) inner_t;
+  Vector<inner_t> inner_tmp(sites);
+  auto inner_tmp_v = &inner_tmp[0];
+  
+
+  accelerator_for( ss, sites, nsimd,{
+      auto x_l = left_v(ss);
+      auto y_l = right_v(ss);
+      coalescedWrite(inner_tmp_v[ss],innerProduct(x_l,y_l));
+  })
+
+  // This is in single precision and fails some tests
+  // Need a sumD that sums in double
+  nrm = TensorRemove(sumD_gpu(inner_tmp_v,sites));  
+#else
+  // CPU 
+  typedef decltype(innerProductD(left_v[0],right_v[0])) inner_t;
+  Vector<inner_t> inner_tmp(sites);
+  auto inner_tmp_v = &inner_tmp[0];
+  
+  accelerator_for( ss, sites, nsimd,{
+      auto x_l = left_v[ss];
+      auto y_l = right_v[ss];
+      inner_tmp_v[ss]=innerProductD(x_l,y_l);
+  })
+  nrm = TensorRemove(sum(inner_tmp_v,sites));
+#endif
+  grid->GlobalSum(nrm);
+
+  return nrm;
 }
 
 /////////////////////////
@@ -86,8 +157,7 @@ axpy_norm_fast(Lattice<vobj> &z,sobj a,const Lattice<vobj> &x,const Lattice<vobj
 template<class sobj,class vobj> strong_inline RealD 
 axpby_norm_fast(Lattice<vobj> &z,sobj a,sobj b,const Lattice<vobj> &x,const Lattice<vobj> &y) 
 {
-  const int pad = 8;
-  z.checkerboard = x.checkerboard;
+  z.Checkerboard() = x.Checkerboard();
   conformable(z,x);
   conformable(x,y);
 
@@ -95,43 +165,57 @@ axpby_norm_fast(Lattice<vobj> &z,sobj a,sobj b,const Lattice<vobj> &x,const Latt
   typedef typename vobj::vector_typeD vector_type;
   RealD  nrm;
   
-  GridBase *grid = x._grid;
+  GridBase *grid = x.Grid();
+
+  auto x_v=x.View();
+  auto y_v=y.View();
+  auto z_v=z.View();
+
+  const uint64_t nsimd = grid->Nsimd();
+  const uint64_t sites = grid->oSites();
   
-  Vector<RealD> sumarray(grid->SumArraySize()*pad);
+#ifdef GRID_NVCC
+  // GPU
+  typedef decltype(innerProduct(x_v[0],y_v[0])) inner_t;
+  Vector<inner_t> inner_tmp(sites);
+  auto inner_tmp_v = &inner_tmp[0];
+
+  accelerator_for( ss, sites, nsimd,{
+      auto tmp = a*x_v(ss)+b*y_v(ss);
+      coalescedWrite(inner_tmp_v[ss],innerProduct(tmp,tmp));
+      coalescedWrite(z_v[ss],tmp);
+  });
+
+  nrm = real(TensorRemove(sumD_gpu(inner_tmp_v,sites)));
+#else
+  // CPU 
+  typedef decltype(innerProductD(x_v[0],y_v[0])) inner_t;
+  Vector<inner_t> inner_tmp(sites);
+  auto inner_tmp_v = &inner_tmp[0];
   
-  parallel_for(int thr=0;thr<grid->SumArraySize();thr++){
-    int nwork, mywork, myoff;
-    GridThread::GetWork(x._grid->oSites(),thr,mywork,myoff);
-    
-    // private to thread; sub summation
-    decltype(innerProductD(z._odata[0],z._odata[0])) vnrm=zero; 
-    for(int ss=myoff;ss<mywork+myoff; ss++){
-      vobj tmp = a*x._odata[ss]+b*y._odata[ss];
-      vnrm = vnrm + innerProductD(tmp,tmp);
-      vstream(z._odata[ss],tmp);
-    }
-    vstream(sumarray[thr*pad],real(Reduce(TensorRemove(vnrm)))) ;
-  }
-  
-  nrm = 0.0; // sum across threads; linear in thread count but fast
-  for(int i=0;i<grid->SumArraySize();i++){
-    nrm = nrm+sumarray[i*pad];
-  } 
-  z._grid->GlobalSum(nrm);
+  accelerator_for( ss, sites, nsimd,{
+      auto tmp = a*x_v(ss)+b*y_v(ss);
+      inner_tmp_v[ss]=innerProductD(tmp,tmp);
+      z_v[ss]=tmp;
+  });
+  // Already promoted to double
+  nrm = real(TensorRemove(sum(inner_tmp_v,sites)));
+#endif
+  grid->GlobalSum(nrm);
   return nrm; 
 }
 
  
 template<class Op,class T1>
 inline auto sum(const LatticeUnaryExpression<Op,T1> & expr)
-  ->typename decltype(expr.first.func(eval(0,std::get<0>(expr.second))))::scalar_object
+  ->typename decltype(expr.op.func(eval(0,expr.arg1)))::scalar_object
 {
   return sum(closure(expr));
 }
 
 template<class Op,class T1,class T2>
 inline auto sum(const LatticeBinaryExpression<Op,T1,T2> & expr)
-      ->typename decltype(expr.first.func(eval(0,std::get<0>(expr.second)),eval(0,std::get<1>(expr.second))))::scalar_object
+      ->typename decltype(expr.op.func(eval(0,expr.arg1),eval(0,expr.arg2)))::scalar_object
 {
   return sum(closure(expr));
 }
@@ -139,54 +223,14 @@ inline auto sum(const LatticeBinaryExpression<Op,T1,T2> & expr)
 
 template<class Op,class T1,class T2,class T3>
 inline auto sum(const LatticeTrinaryExpression<Op,T1,T2,T3> & expr)
-  ->typename decltype(expr.first.func(eval(0,std::get<0>(expr.second)),
-				      eval(0,std::get<1>(expr.second)),
-				      eval(0,std::get<2>(expr.second))
+  ->typename decltype(expr.op.func(eval(0,expr.arg1),
+				      eval(0,expr.arg2),
+				      eval(0,expr.arg3)
 				      ))::scalar_object
 {
   return sum(closure(expr));
 }
 
-template<class vobj>
-inline typename vobj::scalar_object sum(const Lattice<vobj> &arg)
-{
-  GridBase *grid=arg._grid;
-  int Nsimd = grid->Nsimd();
-  
-  std::vector<vobj,alignedAllocator<vobj> > sumarray(grid->SumArraySize());
-  for(int i=0;i<grid->SumArraySize();i++){
-    sumarray[i]=zero;
-  }
-  
-  parallel_for(int thr=0;thr<grid->SumArraySize();thr++){
-    int nwork, mywork, myoff;
-    GridThread::GetWork(grid->oSites(),thr,mywork,myoff);
-    
-    vobj vvsum=zero;
-    for(int ss=myoff;ss<mywork+myoff; ss++){
-      vvsum = vvsum + arg._odata[ss];
-    }
-    sumarray[thr]=vvsum;
-  }
-  
-  vobj vsum=zero;  // sum across threads
-  for(int i=0;i<grid->SumArraySize();i++){
-    vsum = vsum+sumarray[i];
-  } 
-  
-  typedef typename vobj::scalar_object sobj;
-  sobj ssum=zero;
-  
-  std::vector<sobj>               buf(Nsimd);
-  extract(vsum,buf);
-  
-  for(int i=0;i<Nsimd;i++) ssum = ssum + buf[i];
-  arg._grid->GlobalSum(ssum);
-  
-  return ssum;
-}
-
-
 //////////////////////////////////////////////////////////////////////////////////////////////////////////////
 // sliceSum, sliceInnerProduct, sliceAxpy, sliceNorm etc...
 //////////////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -199,7 +243,7 @@ template<class vobj> inline void sliceSum(const Lattice<vobj> &Data,std::vector<
   // But easily avoided by using double precision fields
   ///////////////////////////////////////////////////////
   typedef typename vobj::scalar_object sobj;
-  GridBase  *grid = Data._grid;
+  GridBase  *grid = Data.Grid();
   assert(grid!=NULL);
 
   const int    Nd = grid->_ndimension;
@@ -212,13 +256,13 @@ template<class vobj> inline void sliceSum(const Lattice<vobj> &Data,std::vector<
   int ld=grid->_ldimensions[orthogdim];
   int rd=grid->_rdimensions[orthogdim];
 
-  std::vector<vobj,alignedAllocator<vobj> > lvSum(rd); // will locally sum vectors first
-  std::vector<sobj> lsSum(ld,zero);                    // sum across these down to scalars
-  std::vector<sobj> extracted(Nsimd);                  // splitting the SIMD
+  Vector<vobj> lvSum(rd); // will locally sum vectors first
+  Vector<sobj> lsSum(ld,Zero());                    // sum across these down to scalars
+  ExtractBuffer<sobj> extracted(Nsimd);                  // splitting the SIMD
 
   result.resize(fd); // And then global sum to return the same vector to every node 
   for(int r=0;r<rd;r++){
-    lvSum[r]=zero;
+    lvSum[r]=Zero();
   }
 
   int e1=    grid->_slice_nblock[orthogdim];
@@ -227,20 +271,19 @@ template<class vobj> inline void sliceSum(const Lattice<vobj> &Data,std::vector<
 
   // sum over reduced dimension planes, breaking out orthog dir
   // Parallel over orthog direction
-  parallel_for(int r=0;r<rd;r++){
-
+  auto Data_v=Data.View();
+  thread_for( r,rd, {
     int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
-
     for(int n=0;n<e1;n++){
       for(int b=0;b<e2;b++){
 	int ss= so+n*stride+b;
-	lvSum[r]=lvSum[r]+Data._odata[ss];
+	lvSum[r]=lvSum[r]+Data_v[ss];
       }
     }
-  }
+  });
 
   // Sum across simd lanes in the plane, breaking out orthog dir.
-  std::vector<int> icoor(Nd);
+  Coordinate icoor(Nd);
 
   for(int rt=0;rt<rd;rt++){
 
@@ -265,7 +308,7 @@ template<class vobj> inline void sliceSum(const Lattice<vobj> &Data,std::vector<
     if ( pt == grid->_processor_coor[orthogdim] ) {
       gsum=lsSum[lt];
     } else {
-      gsum=zero;
+      gsum=Zero();
     }
 
     grid->GlobalSum(gsum);
@@ -292,9 +335,9 @@ static void localSliceInnerProductVector(std::vector<ComplexD> &result, const La
   // std::cout << GridLogMessage << "Start prep" << std::endl;
   typedef typename vobj::vector_type   vector_type;
   typedef typename vobj::scalar_type   scalar_type;
-  GridBase  *grid = lhs._grid;
+  GridBase  *grid = lhs.Grid();
   assert(grid!=NULL);
-  conformable(grid,rhs._grid);
+  conformable(grid,rhs.Grid());
 
   const int    Nd = grid->_ndimension;
   const int Nsimd = grid->Nsimd();
@@ -307,14 +350,14 @@ static void localSliceInnerProductVector(std::vector<ComplexD> &result, const La
   int rd=grid->_rdimensions[orthogdim];
   // std::cout << GridLogMessage << "Start alloc" << std::endl;
 
-  std::vector<vector_type,alignedAllocator<vector_type> > lvSum(rd); // will locally sum vectors first
+  Vector<vector_type> lvSum(rd); // will locally sum vectors first
   lsSum.resize(ld,scalar_type(0.0));                    // sum across these down to scalars
-  std::vector<iScalar<scalar_type>> extracted(Nsimd);   // splitting the SIMD  
+  ExtractBuffer<iScalar<scalar_type> > extracted(Nsimd);   // splitting the SIMD  
   // std::cout << GridLogMessage << "End alloc" << std::endl;
 
   result.resize(fd); // And then global sum to return the same vector to every node for IO to file
   for(int r=0;r<rd;r++){
-    lvSum[r]=zero;
+    lvSum[r]=Zero();
   }
 
   int e1=    grid->_slice_nblock[orthogdim];
@@ -323,23 +366,24 @@ static void localSliceInnerProductVector(std::vector<ComplexD> &result, const La
   // std::cout << GridLogMessage << "End prep" << std::endl;
   // std::cout << GridLogMessage << "Start parallel inner product, _rd = " << rd << std::endl;
   vector_type vv;
-  parallel_for(int r=0;r<rd;r++)
-  {
+  auto l_v=lhs.View();
+  auto r_v=rhs.View();
+  thread_for( r,rd,{
 
     int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
 
     for(int n=0;n<e1;n++){
       for(int b=0;b<e2;b++){
         int ss = so + n * stride + b;
-        vv = TensorRemove(innerProduct(lhs._odata[ss], rhs._odata[ss]));
+        vv = TensorRemove(innerProduct(l_v[ss], r_v[ss]));
         lvSum[r] = lvSum[r] + vv;
       }
     }
-  }
+  });
   // std::cout << GridLogMessage << "End parallel inner product" << std::endl;
 
   // Sum across simd lanes in the plane, breaking out orthog dir.
-  std::vector<int> icoor(Nd);
+  Coordinate icoor(Nd);
   for(int rt=0;rt<rd;rt++){
 
     iScalar<vector_type> temp; 
@@ -362,7 +406,7 @@ template <class vobj>
 static void globalSliceInnerProductVector(std::vector<ComplexD> &result, const Lattice<vobj> &lhs, std::vector<typename vobj::scalar_type> &lsSum, int orthogdim)
 {
   typedef typename vobj::scalar_type scalar_type;
-  GridBase *grid = lhs._grid;
+  GridBase *grid = lhs.Grid();
   int fd = result.size();
   int ld = lsSum.size();
   // sum over nodes.
@@ -388,9 +432,9 @@ static void sliceInnerProductVector( std::vector<ComplexD> & result, const Latti
 {
   typedef typename vobj::vector_type   vector_type;
   typedef typename vobj::scalar_type   scalar_type;
-  GridBase  *grid = lhs._grid;
+  GridBase  *grid = lhs.Grid();
   assert(grid!=NULL);
-  conformable(grid,rhs._grid);
+  conformable(grid,rhs.Grid());
 
   const int    Nd = grid->_ndimension;
   const int Nsimd = grid->Nsimd();
@@ -402,34 +446,36 @@ static void sliceInnerProductVector( std::vector<ComplexD> & result, const Latti
   int ld=grid->_ldimensions[orthogdim];
   int rd=grid->_rdimensions[orthogdim];
 
-  std::vector<vector_type,alignedAllocator<vector_type> > lvSum(rd); // will locally sum vectors first
-  std::vector<scalar_type > lsSum(ld,scalar_type(0.0));                    // sum across these down to scalars
-  std::vector<iScalar<scalar_type> > extracted(Nsimd);                  // splitting the SIMD
+  Vector<vector_type> lvSum(rd); // will locally sum vectors first
+  Vector<scalar_type > lsSum(ld,scalar_type(0.0));                    // sum across these down to scalars
+  ExtractBuffer<iScalar<scalar_type> > extracted(Nsimd);   // splitting the SIMD  
 
   result.resize(fd); // And then global sum to return the same vector to every node for IO to file
   for(int r=0;r<rd;r++){
-    lvSum[r]=zero;
+    lvSum[r]=Zero();
   }
 
   int e1=    grid->_slice_nblock[orthogdim];
   int e2=    grid->_slice_block [orthogdim];
   int stride=grid->_slice_stride[orthogdim];
 
-  parallel_for(int r=0;r<rd;r++){
+  auto lhv=lhs.View();
+  auto rhv=rhs.View();
+  thread_for( r,rd,{
 
     int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
 
     for(int n=0;n<e1;n++){
       for(int b=0;b<e2;b++){
 	int ss= so+n*stride+b;
-	vector_type vv = TensorRemove(innerProduct(lhs._odata[ss],rhs._odata[ss]));
+	vector_type vv = TensorRemove(innerProduct(lhv[ss],rhv[ss]));
 	lvSum[r]=lvSum[r]+vv;
       }
     }
-  }
+  });
 
   // Sum across simd lanes in the plane, breaking out orthog dir.
-  std::vector<int> icoor(Nd);
+  Coordinate icoor(Nd);
   for(int rt=0;rt<rd;rt++){
 
     iScalar<vector_type> temp; 
@@ -470,7 +516,7 @@ static void sliceNorm (std::vector<RealD> &sn,const Lattice<vobj> &rhs,int Ortho
   typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_type vector_type;
   
-  int Nblock = rhs._grid->GlobalDimensions()[Orthog];
+  int Nblock = rhs.Grid()->GlobalDimensions()[Orthog];
   std::vector<ComplexD> ip(Nblock);
   sn.resize(Nblock);
   
@@ -492,7 +538,7 @@ static void sliceMaddVector(Lattice<vobj> &R,std::vector<RealD> &a,const Lattice
   
   scalar_type zscale(scale);
 
-  GridBase *grid  = X._grid;
+  GridBase *grid  = X.Grid();
 
   int Nsimd  =grid->Nsimd();
   int Nblock =grid->GlobalDimensions()[orthogdim];
@@ -505,8 +551,7 @@ static void sliceMaddVector(Lattice<vobj> &R,std::vector<RealD> &a,const Lattice
   int e2     =grid->_slice_block [orthogdim];
   int stride =grid->_slice_stride[orthogdim];
 
-  std::vector<int> icoor;
-
+  Coordinate icoor;
   for(int r=0;r<rd;r++){
 
     int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
@@ -522,12 +567,15 @@ static void sliceMaddVector(Lattice<vobj> &R,std::vector<RealD> &a,const Lattice
 
     tensor_reduced at; at=av;
 
-    parallel_for_nest2(int n=0;n<e1;n++){
+    auto Rv=R.View();
+    auto Xv=X.View();
+    auto Yv=Y.View();
+    thread_for_collapse(2, n, e1, {
       for(int b=0;b<e2;b++){
 	int ss= so+n*stride+b;
-	R._odata[ss] = at*X._odata[ss]+Y._odata[ss];
+	Rv[ss] = at*Xv[ss]+Yv[ss];
       }
-    }
+    });
   }
 };
 
@@ -559,18 +607,18 @@ static void sliceMaddMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice
   typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_type vector_type;
 
-  int Nblock = X._grid->GlobalDimensions()[Orthog];
+  int Nblock = X.Grid()->GlobalDimensions()[Orthog];
 
-  GridBase *FullGrid  = X._grid;
+  GridBase *FullGrid  = X.Grid();
   //  GridBase *SliceGrid = makeSubSliceGrid(FullGrid,Orthog);
 
   //  Lattice<vobj> Xslice(SliceGrid);
   //  Lattice<vobj> Rslice(SliceGrid);
 
   assert( FullGrid->_simd_layout[Orthog]==1);
-  int nh =  FullGrid->_ndimension;
+  //  int nh =  FullGrid->_ndimension;
   //  int nl = SliceGrid->_ndimension;
-  int nl = nh-1;
+  //  int nl = nh-1;
 
   //FIXME package in a convenient iterator
   //Should loop over a plane orthogonal to direction "Orthog"
@@ -578,28 +626,31 @@ static void sliceMaddMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice
   int block =FullGrid->_slice_block [Orthog];
   int nblock=FullGrid->_slice_nblock[Orthog];
   int ostride=FullGrid->_ostride[Orthog];
-#pragma omp parallel 
-  {
-    std::vector<vobj> s_x(Nblock);
 
-#pragma omp for collapse(2)
-    for(int n=0;n<nblock;n++){
-    for(int b=0;b<block;b++){
+  auto X_v=X.View();
+  auto Y_v=Y.View();
+  auto R_v=R.View();
+  thread_region
+  {
+    Vector<vobj> s_x(Nblock);
+
+    thread_for_collapse_in_region(2, n,nblock, {
+     for(int b=0;b<block;b++){
       int o  = n*stride + b;
 
       for(int i=0;i<Nblock;i++){
-	s_x[i] = X[o+i*ostride];
+	s_x[i] = X_v[o+i*ostride];
       }
 
       vobj dot;
       for(int i=0;i<Nblock;i++){
-	dot = Y[o+i*ostride];
+	dot = Y_v[o+i*ostride];
 	for(int j=0;j<Nblock;j++){
 	  dot = dot + s_x[j]*(scale*aa(j,i));
 	}
-	R[o+i*ostride]=dot;
+	R_v[o+i*ostride]=dot;
       }
-    }}
+    }});
   }
 };
 
@@ -610,17 +661,17 @@ static void sliceMulMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice<
   typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_type vector_type;
 
-  int Nblock = X._grid->GlobalDimensions()[Orthog];
+  int Nblock = X.Grid()->GlobalDimensions()[Orthog];
 
-  GridBase *FullGrid  = X._grid;
+  GridBase *FullGrid  = X.Grid();
   //  GridBase *SliceGrid = makeSubSliceGrid(FullGrid,Orthog);
   //  Lattice<vobj> Xslice(SliceGrid);
   //  Lattice<vobj> Rslice(SliceGrid);
 
   assert( FullGrid->_simd_layout[Orthog]==1);
-  int nh =  FullGrid->_ndimension;
+  //  int nh =  FullGrid->_ndimension;
   //  int nl = SliceGrid->_ndimension;
-  int nl=1;
+  //  int nl=1;
 
   //FIXME package in a convenient iterator
   //Should loop over a plane orthogonal to direction "Orthog"
@@ -628,17 +679,19 @@ static void sliceMulMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice<
   int block =FullGrid->_slice_block [Orthog];
   int nblock=FullGrid->_slice_nblock[Orthog];
   int ostride=FullGrid->_ostride[Orthog];
-#pragma omp parallel 
+  auto R_v = R.View();
+  auto X_v = X.View();
+  thread_region
   {
     std::vector<vobj> s_x(Nblock);
 
-#pragma omp for collapse(2)
-    for(int n=0;n<nblock;n++){
+
+    thread_for_collapse_in_region( 2 ,n,nblock,{
     for(int b=0;b<block;b++){
       int o  = n*stride + b;
 
       for(int i=0;i<Nblock;i++){
-	s_x[i] = X[o+i*ostride];
+	s_x[i] = X_v[o+i*ostride];
       }
 
       vobj dot;
@@ -647,11 +700,10 @@ static void sliceMulMatrix (Lattice<vobj> &R,Eigen::MatrixXcd &aa,const Lattice<
 	for(int j=1;j<Nblock;j++){
 	  dot = dot + s_x[j]*(scale*aa(j,i));
 	}
-	R[o+i*ostride]=dot;
+	R_v[o+i*ostride]=dot;
       }
-    }}
+    }});
   }
-
 };
 
 
@@ -662,7 +714,7 @@ static void sliceInnerProductMatrix(  Eigen::MatrixXcd &mat, const Lattice<vobj>
   typedef typename vobj::scalar_type scalar_type;
   typedef typename vobj::vector_type vector_type;
   
-  GridBase *FullGrid  = lhs._grid;
+  GridBase *FullGrid  = lhs.Grid();
   //  GridBase *SliceGrid = makeSubSliceGrid(FullGrid,Orthog);
   
   int Nblock = FullGrid->GlobalDimensions()[Orthog];
@@ -673,9 +725,9 @@ static void sliceInnerProductMatrix(  Eigen::MatrixXcd &mat, const Lattice<vobj>
   mat = Eigen::MatrixXcd::Zero(Nblock,Nblock);
 
   assert( FullGrid->_simd_layout[Orthog]==1);
-  int nh =  FullGrid->_ndimension;
+  //  int nh =  FullGrid->_ndimension;
   //  int nl = SliceGrid->_ndimension;
-  int nl = nh-1;
+  //  int nl = nh-1;
 
   //FIXME package in a convenient iterator
   //Should loop over a plane orthogonal to direction "Orthog"
@@ -686,31 +738,33 @@ static void sliceInnerProductMatrix(  Eigen::MatrixXcd &mat, const Lattice<vobj>
 
   typedef typename vobj::vector_typeD vector_typeD;
 
-#pragma omp parallel 
+  auto lhs_v=lhs.View();
+  auto rhs_v=rhs.View();
+  thread_region
   {
     std::vector<vobj> Left(Nblock);
     std::vector<vobj> Right(Nblock);
     Eigen::MatrixXcd  mat_thread = Eigen::MatrixXcd::Zero(Nblock,Nblock);
 
-#pragma omp for collapse(2)
-    for(int n=0;n<nblock;n++){
+    thread_for_collapse_in_region( 2, n,nblock,{
     for(int b=0;b<block;b++){
 
       int o  = n*stride + b;
 
       for(int i=0;i<Nblock;i++){
-	Left [i] = lhs[o+i*ostride];
-	Right[i] = rhs[o+i*ostride];
+	Left [i] = lhs_v[o+i*ostride];
+	Right[i] = rhs_v[o+i*ostride];
       }
 
       for(int i=0;i<Nblock;i++){
       for(int j=0;j<Nblock;j++){
 	auto tmp = innerProduct(Left[i],Right[j]);
 	auto rtmp = TensorRemove(tmp);
-	mat_thread(i,j) += Reduce(rtmp);
+	auto red  =  Reduce(rtmp);
+	mat_thread(i,j) += std::complex<double>(real(red),imag(red));
       }}
-    }}
-#pragma omp critical
+    }});
+    thread_critical
     {
       mat += mat_thread;
     }  
@@ -726,8 +780,8 @@ static void sliceInnerProductMatrix(  Eigen::MatrixXcd &mat, const Lattice<vobj>
   return;
 }
 
-} /*END NAMESPACE GRID*/
-#endif
+NAMESPACE_END(Grid);
+
 
 
 
diff --git a/Grid/lattice/Lattice_reduction_gpu.h b/Grid/lattice/Lattice_reduction_gpu.h
new file mode 100644
index 00000000..c5d75356
--- /dev/null
+++ b/Grid/lattice/Lattice_reduction_gpu.h
@@ -0,0 +1,226 @@
+NAMESPACE_BEGIN(Grid);
+
+#define WARP_SIZE 32
+extern cudaDeviceProp *gpu_props;
+__device__ unsigned int retirementCount = 0;
+
+template <class Iterator>
+unsigned int nextPow2(Iterator x) {
+  --x;
+  x |= x >> 1;
+  x |= x >> 2;
+  x |= x >> 4;
+  x |= x >> 8;
+  x |= x >> 16;
+  return ++x;
+}
+
+template <class Iterator>
+void getNumBlocksAndThreads(const Iterator n, const size_t sizeofsobj, Iterator &threads, Iterator &blocks) {
+  
+  int device;
+  cudaGetDevice(&device);
+  
+  Iterator warpSize            = gpu_props[device].warpSize;
+  Iterator sharedMemPerBlock   = gpu_props[device].sharedMemPerBlock;
+  Iterator maxThreadsPerBlock  = gpu_props[device].maxThreadsPerBlock;
+  Iterator multiProcessorCount = gpu_props[device].multiProcessorCount;
+  
+  std::cout << GridLogDebug << "GPU has:" << std::endl;
+  std::cout << GridLogDebug << "\twarpSize            = " << warpSize << std::endl;
+  std::cout << GridLogDebug << "\tsharedMemPerBlock   = " << sharedMemPerBlock << std::endl;
+  std::cout << GridLogDebug << "\tmaxThreadsPerBlock  = " << maxThreadsPerBlock << std::endl;
+  std::cout << GridLogDebug << "\tmaxThreadsPerBlock  = " << warpSize << std::endl;
+  std::cout << GridLogDebug << "\tmultiProcessorCount = " << multiProcessorCount << std::endl;
+  
+  if (warpSize != WARP_SIZE) {
+    std::cout << GridLogError << "The warp size of the GPU in use does not match the warp size set when compiling Grid." << std::endl;
+    exit(EXIT_FAILURE);
+  }
+  
+  // let the number of threads in a block be a multiple of 2, starting from warpSize
+  threads = warpSize;
+  while( 2*threads*sizeofsobj < sharedMemPerBlock && 2*threads <= maxThreadsPerBlock ) threads *= 2;
+  // keep all the streaming multiprocessors busy
+  blocks = nextPow2(multiProcessorCount);
+  
+}
+
+template <class sobj, class Iterator>
+__device__ void reduceBlock(volatile sobj *sdata, sobj mySum, const Iterator tid) {
+  
+  Iterator blockSize = blockDim.x;
+  
+  // cannot use overloaded operators for sobj as they are not volatile-qualified
+  memcpy((void *)&sdata[tid], (void *)&mySum, sizeof(sobj));
+  __syncwarp();
+  
+  const Iterator VEC = WARP_SIZE;
+  const Iterator vid = tid & (VEC-1);
+  
+  sobj beta, temp;
+  memcpy((void *)&beta, (void *)&mySum, sizeof(sobj));
+  
+  for (int i = VEC/2; i > 0; i>>=1) {
+    if (vid < i) {
+      memcpy((void *)&temp, (void *)&sdata[tid+i], sizeof(sobj));
+      beta += temp;
+      memcpy((void *)&sdata[tid], (void *)&beta, sizeof(sobj));
+    }
+    __syncwarp();
+  }
+  __syncthreads();
+  
+  if (threadIdx.x == 0) {
+    beta  = Zero();
+    for (Iterator i = 0; i < blockSize; i += VEC) {
+      memcpy((void *)&temp, (void *)&sdata[i], sizeof(sobj));
+      beta  += temp;
+    }
+    memcpy((void *)&sdata[0], (void *)&beta, sizeof(sobj));
+  }
+  __syncthreads();
+}
+
+
+template <class vobj, class sobj, class Iterator>
+__device__ void reduceBlocks(const vobj *g_idata, sobj *g_odata, Iterator n) 
+{
+  constexpr Iterator nsimd = vobj::Nsimd();
+  
+  Iterator blockSize = blockDim.x;
+  
+  // force shared memory alignment
+  extern __shared__ __align__(COALESCE_GRANULARITY) unsigned char shmem_pointer[];
+  // it's not possible to have two extern __shared__ arrays with same name
+  // but different types in different scopes -- need to cast each time
+  sobj *sdata = (sobj *)shmem_pointer;
+  
+  // first level of reduction,
+  // each thread writes result in mySum
+  Iterator tid = threadIdx.x;
+  Iterator i = blockIdx.x*(blockSize*2) + threadIdx.x;
+  Iterator gridSize = blockSize*2*gridDim.x;
+  sobj mySum = Zero();
+  
+  while (i < n) {
+    Iterator lane = i % nsimd;
+    Iterator ss   = i / nsimd;
+    auto tmp = extractLane(lane,g_idata[ss]);
+    sobj tmpD;
+    tmpD=tmp;
+    mySum   +=tmpD;
+    
+    if (i + blockSize < n) {
+      lane = (i+blockSize) % nsimd;
+      ss   = (i+blockSize) / nsimd;
+      tmp = extractLane(lane,g_idata[ss]);
+      tmpD = tmp;
+      mySum += tmpD;
+    }
+    i += gridSize;
+  }
+  
+  // copy mySum to shared memory and perform
+  // reduction for all threads in this block
+  reduceBlock(sdata, mySum, tid);
+  if (tid == 0) g_odata[blockIdx.x] = sdata[0];
+}
+
+template <class vobj, class sobj,class Iterator>
+__global__ void reduceKernel(const vobj *lat, sobj *buffer, Iterator n) {
+  
+  Iterator blockSize = blockDim.x;
+  
+  // perform reduction for this block and
+  // write result to global memory buffer
+  reduceBlocks(lat, buffer, n);
+  
+  if (gridDim.x > 1) {
+    
+    const Iterator tid = threadIdx.x;
+    __shared__ bool amLast;
+    // force shared memory alignment
+    extern __shared__ __align__(COALESCE_GRANULARITY) unsigned char shmem_pointer[];
+    // it's not possible to have two extern __shared__ arrays with same name
+    // but different types in different scopes -- need to cast each time
+    sobj *smem = (sobj *)shmem_pointer;
+    
+    // wait until all outstanding memory instructions in this thread are finished
+    __threadfence();
+    
+    if (tid==0) {
+      unsigned int ticket = atomicInc(&retirementCount, gridDim.x);
+      // true if this block is the last block to be done
+      amLast = (ticket == gridDim.x-1);
+    }
+    
+    // each thread must read the correct value of amLast
+    __syncthreads();
+    
+    if (amLast) {
+      // reduce buffer[0], ..., buffer[gridDim.x-1]
+      Iterator i = tid;
+      sobj mySum = Zero();
+      
+      while (i < gridDim.x) {
+        mySum += buffer[i];
+        i += blockSize;
+      }
+      
+      reduceBlock(smem, mySum, tid);
+      
+      if (tid==0) {
+        buffer[0] = smem[0];
+        // reset count variable
+        retirementCount = 0;
+      }
+    }
+  }
+}
+
+/////////////////////////////////////////////////////////////////////////////////////////////////////////
+// Possibly promote to double and sum
+/////////////////////////////////////////////////////////////////////////////////////////////////////////
+template <class vobj>
+inline typename vobj::scalar_objectD sumD_gpu(const vobj *lat, Integer osites) 
+{
+  typedef typename vobj::scalar_objectD sobj;
+  typedef decltype(lat) Iterator;
+  
+  Integer nsimd= vobj::Nsimd();
+  Integer size = osites*nsimd;
+
+  Integer numThreads, numBlocks;
+  getNumBlocksAndThreads(size, sizeof(sobj), numThreads, numBlocks);
+  Integer smemSize = numThreads * sizeof(sobj);
+
+  Vector<sobj> buffer(numBlocks);
+  sobj *buffer_v = &buffer[0];
+  
+  reduceKernel<<< numBlocks, numThreads, smemSize >>>(lat, buffer_v, size);
+  cudaDeviceSynchronize();
+  
+  cudaError err = cudaGetLastError();
+  if ( cudaSuccess != err ) {
+    printf("Cuda error %s\n",cudaGetErrorString( err ));
+    exit(0);
+  }
+  auto result = buffer_v[0];
+  return result;
+}
+/////////////////////////////////////////////////////////////////////////////////////////////////////////
+// Return as same precision as input performing reduction in double precision though
+/////////////////////////////////////////////////////////////////////////////////////////////////////////
+template <class vobj>
+inline typename vobj::scalar_object sum_gpu(const vobj *lat, Integer osites) 
+{
+  typedef typename vobj::scalar_object sobj;
+  sobj result;
+  result = sumD_gpu(lat,osites);
+  return result;
+}
+
+
+
+NAMESPACE_END(Grid);
diff --git a/Grid/lattice/Lattice_rng.h b/Grid/lattice/Lattice_rng.h
index 5348538c..1bb1f087 100644
--- a/Grid/lattice/Lattice_rng.h
+++ b/Grid/lattice/Lattice_rng.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,8 +24,8 @@
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_LATTICE_RNG_H
 #define GRID_LATTICE_RNG_H
 
@@ -41,282 +41,289 @@
 #undef  RNG_FAST_DISCARD
 #endif
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  //////////////////////////////////////////////////////////////
-  // Allow the RNG state to be less dense than the fine grid
-  //////////////////////////////////////////////////////////////
-  inline int RNGfillable(GridBase *coarse,GridBase *fine)
-  {
+//////////////////////////////////////////////////////////////
+// Allow the RNG state to be less dense than the fine grid
+//////////////////////////////////////////////////////////////
+inline int RNGfillable(GridBase *coarse,GridBase *fine)
+{
 
-    int rngdims = coarse->_ndimension;
+  int rngdims = coarse->_ndimension;
 
-    // trivially extended in higher dims, with locality guaranteeing RNG state is local to node
-    int lowerdims   = fine->_ndimension - coarse->_ndimension;
-    assert(lowerdims >= 0);
-    for(int d=0;d<lowerdims;d++){
-      assert(fine->_simd_layout[d]==1);
-      assert(fine->_processors[d]==1);
-    }
-
-    int multiplicity=1;
-    for(int d=0;d<lowerdims;d++){
-      multiplicity=multiplicity*fine->_rdimensions[d];
-    }
-    // local and global volumes subdivide cleanly after SIMDization
-    for(int d=0;d<rngdims;d++){
-      int fd= d+lowerdims;
-      assert(coarse->_processors[d]  == fine->_processors[fd]);
-      assert(coarse->_simd_layout[d] == fine->_simd_layout[fd]);
-      assert(((fine->_rdimensions[fd] / coarse->_rdimensions[d])* coarse->_rdimensions[d])==fine->_rdimensions[fd]); 
-
-      multiplicity = multiplicity *fine->_rdimensions[fd] / coarse->_rdimensions[d]; 
-    }
-    return multiplicity;
+  // trivially extended in higher dims, with locality guaranteeing RNG state is local to node
+  int lowerdims   = fine->_ndimension - coarse->_ndimension;
+  assert(lowerdims >= 0);
+  for(int d=0;d<lowerdims;d++){
+    assert(fine->_simd_layout[d]==1);
+    assert(fine->_processors[d]==1);
   }
 
+  int multiplicity=1;
+  for(int d=0;d<lowerdims;d++){
+    multiplicity=multiplicity*fine->_rdimensions[d];
+  }
+  // local and global volumes subdivide cleanly after SIMDization
+  for(int d=0;d<rngdims;d++){
+    int fd= d+lowerdims;
+    assert(coarse->_processors[d]  == fine->_processors[fd]);
+    assert(coarse->_simd_layout[d] == fine->_simd_layout[fd]);
+    assert(((fine->_rdimensions[fd] / coarse->_rdimensions[d])* coarse->_rdimensions[d])==fine->_rdimensions[fd]); 
+
+    multiplicity = multiplicity *fine->_rdimensions[fd] / coarse->_rdimensions[d]; 
+  }
+  return multiplicity;
+}
+
   
 // merge of April 11 2017
-  // this function is necessary for the LS vectorised field
-  inline int RNGfillable_general(GridBase *coarse,GridBase *fine)
-  {
-    int rngdims = coarse->_ndimension;
+// this function is necessary for the LS vectorised field
+inline int RNGfillable_general(GridBase *coarse,GridBase *fine)
+{
+  int rngdims = coarse->_ndimension;
     
-    // trivially extended in higher dims, with locality guaranteeing RNG state is local to node
-    int lowerdims   = fine->_ndimension - coarse->_ndimension;  assert(lowerdims >= 0);
-    // assumes that the higher dimensions are not using more processors
-    // all further divisions are local
-    for(int d=0;d<lowerdims;d++) assert(fine->_processors[d]==1);
-    for(int d=0;d<rngdims;d++) assert(coarse->_processors[d] == fine->_processors[d+lowerdims]);
+  // trivially extended in higher dims, with locality guaranteeing RNG state is local to node
+  int lowerdims   = fine->_ndimension - coarse->_ndimension;  assert(lowerdims >= 0);
+  // assumes that the higher dimensions are not using more processors
+  // all further divisions are local
+  for(int d=0;d<lowerdims;d++) assert(fine->_processors[d]==1);
+  for(int d=0;d<rngdims;d++) assert(coarse->_processors[d] == fine->_processors[d+lowerdims]);
 
-    // then divide the number of local sites
-    // check that the total number of sims agree, meanse the iSites are the same
-    assert(fine->Nsimd() == coarse->Nsimd());
+  // then divide the number of local sites
+  // check that the total number of sims agree, meanse the iSites are the same
+  assert(fine->Nsimd() == coarse->Nsimd());
 
-    // check that the two grids divide cleanly
-    assert( (fine->lSites() / coarse->lSites() ) * coarse->lSites() == fine->lSites() );
+  // check that the two grids divide cleanly
+  assert( (fine->lSites() / coarse->lSites() ) * coarse->lSites() == fine->lSites() );
 
-    return fine->lSites() / coarse->lSites();
-  }
+  return fine->lSites() / coarse->lSites();
+}
   
-  // real scalars are one component
-  template<class scalar,class distribution,class generator> 
-  void fillScalar(scalar &s,distribution &dist,generator & gen)
-  {
-    s=dist(gen);
-  }
-  template<class distribution,class generator> 
-  void fillScalar(ComplexF &s,distribution &dist, generator &gen)
-  {
-    s=ComplexF(dist(gen),dist(gen));
-  }
-  template<class distribution,class generator> 
-  void fillScalar(ComplexD &s,distribution &dist,generator &gen)
-  {
-    s=ComplexD(dist(gen),dist(gen));
-  }
+// real scalars are one component
+template<class scalar,class distribution,class generator> 
+void fillScalar(scalar &s,distribution &dist,generator & gen)
+{
+  s=dist(gen);
+}
+template<class distribution,class generator> 
+void fillScalar(ComplexF &s,distribution &dist, generator &gen)
+{
+  //  s=ComplexF(dist(gen),dist(gen));
+  s.real(dist(gen));
+  s.imag(dist(gen));
+}
+template<class distribution,class generator> 
+void fillScalar(ComplexD &s,distribution &dist,generator &gen)
+{
+  //  s=ComplexD(dist(gen),dist(gen));
+  s.real(dist(gen));
+  s.imag(dist(gen));
+}
   
-  class GridRNGbase {
-  public:
-    // One generator per site.
-    // Uniform and Gaussian distributions from these generators.
+class GridRNGbase {
+public:
+  // One generator per site.
+  // Uniform and Gaussian distributions from these generators.
 #ifdef RNG_RANLUX
-    typedef std::ranlux48 RngEngine;
-    typedef uint64_t      RngStateType;
-    static const int RngStateCount = 15;
+  typedef std::ranlux48 RngEngine;
+  typedef uint64_t      RngStateType;
+  static const int RngStateCount = 15;
 #endif 
 #ifdef RNG_MT19937 
-    typedef std::mt19937 RngEngine;
-    typedef uint32_t     RngStateType;
-    static const int     RngStateCount = std::mt19937::state_size;
+  typedef std::mt19937 RngEngine;
+  typedef uint32_t     RngStateType;
+  static const int     RngStateCount = std::mt19937::state_size;
 #endif
 #ifdef RNG_SITMO
-    typedef sitmo::prng_engine 	RngEngine;
-    typedef uint64_t    	RngStateType;
-    static const int    	RngStateCount = 13;
+  typedef sitmo::prng_engine 	RngEngine;
+  typedef uint64_t    	RngStateType;
+  static const int    	RngStateCount = 13;
 #endif
 
-    std::vector<RngEngine>                             _generators;
-    std::vector<std::uniform_real_distribution<RealD> > _uniform;
-    std::vector<std::normal_distribution<RealD> >       _gaussian;
-    std::vector<std::discrete_distribution<int32_t> >   _bernoulli;
-    std::vector<std::uniform_int_distribution<uint32_t> > _uid;
+  std::vector<RngEngine>                             _generators;
+  std::vector<std::uniform_real_distribution<RealD> > _uniform;
+  std::vector<std::normal_distribution<RealD> >       _gaussian;
+  std::vector<std::discrete_distribution<int32_t> >   _bernoulli;
+  std::vector<std::uniform_int_distribution<uint32_t> > _uid;
 
-    ///////////////////////
-    // support for parallel init
-    ///////////////////////
+  ///////////////////////
+  // support for parallel init
+  ///////////////////////
 #ifdef RNG_FAST_DISCARD
-    static void Skip(RngEngine &eng,uint64_t site)
-    {
-      /////////////////////////////////////////////////////////////////////////////////////
-      // Skip by 2^40 elements between successive lattice sites
-      // This goes by 10^12.
-      // Consider quenched updating; likely never exceeding rate of 1000 sweeps
-      // per second on any machine. This gives us of order 10^9 seconds, or 100 years
-      // skip ahead.
-      // For HMC unlikely to go at faster than a solve per second, and 
-      // tens of seconds per trajectory so this is clean in all reasonable cases,
-      // and margin of safety is orders of magnitude.
-      // We could hack Sitmo to skip in the higher order words of state if necessary
+  static void Skip(RngEngine &eng,uint64_t site)
+  {
+    /////////////////////////////////////////////////////////////////////////////////////
+    // Skip by 2^40 elements between successive lattice sites
+    // This goes by 10^12.
+    // Consider quenched updating; likely never exceeding rate of 1000 sweeps
+    // per second on any machine. This gives us of order 10^9 seconds, or 100 years
+    // skip ahead.
+    // For HMC unlikely to go at faster than a solve per second, and 
+    // tens of seconds per trajectory so this is clean in all reasonable cases,
+    // and margin of safety is orders of magnitude.
+    // We could hack Sitmo to skip in the higher order words of state if necessary
       //
       // Replace with 2^30 ; avoid problem on large volumes
       //
-      /////////////////////////////////////////////////////////////////////////////////////
-      //      uint64_t skip = site+1;  //   Old init Skipped then drew.  Checked compat with faster init
-      const int shift = 30;
+    /////////////////////////////////////////////////////////////////////////////////////
+    //      uint64_t skip = site+1;  //   Old init Skipped then drew.  Checked compat with faster init
+    const int shift = 30;
 
-      uint64_t skip = site;
+    ////////////////////////////////////////////////////////////////////
+    // Weird compiler bug in Intel 2018.1 under O3 was generating 32bit and not 64 bit left shift.
+    ////////////////////////////////////////////////////////////////////
+    volatile uint64_t skip = site;
 
-      skip = skip<<shift;
+    skip = skip<<shift;
 
-      assert((skip >> shift)==site); // check for overflow
+    assert((skip >> shift)==site); // check for overflow
 
-      eng.discard(skip);
-      //      std::cout << " Engine  " <<site << " state " <<eng<<std::endl;
-    } 
+    eng.discard(skip);
+    //      std::cout << " Engine  " <<site << " state " <<eng<<std::endl;
+  } 
 #endif
-    static RngEngine Reseed(RngEngine &eng)
-    {
-      std::vector<uint32_t> newseed;
-      std::uniform_int_distribution<uint32_t> uid;
-      return Reseed(eng,newseed,uid);
-    }
-    static RngEngine Reseed(RngEngine &eng,std::vector<uint32_t> & newseed,
-			    std::uniform_int_distribution<uint32_t> &uid)
-    {
-      const int reseeds=4;
+  static RngEngine Reseed(RngEngine &eng)
+  {
+    std::vector<uint32_t> newseed;
+    std::uniform_int_distribution<uint32_t> uid;
+    return Reseed(eng,newseed,uid);
+  }
+  static RngEngine Reseed(RngEngine &eng,std::vector<uint32_t> & newseed,
+			  std::uniform_int_distribution<uint32_t> &uid)
+  {
+    const int reseeds=4;
       
-      newseed.resize(reseeds);
-      for(int i=0;i<reseeds;i++){
-	newseed[i] = uid(eng);
-      }
-      std::seed_seq sseq(newseed.begin(),newseed.end());
-      return RngEngine(sseq);
-    }    
+    newseed.resize(reseeds);
+    for(int i=0;i<reseeds;i++){
+      newseed[i] = uid(eng);
+    }
+    std::seed_seq sseq(newseed.begin(),newseed.end());
+    return RngEngine(sseq);
+  }    
 
-    void GetState(std::vector<RngStateType> & saved,RngEngine &eng) {
-      saved.resize(RngStateCount);
-      std::stringstream ss;
-      ss<<eng;
-      ss.seekg(0,ss.beg);
-      for(int i=0;i<RngStateCount;i++){
-        ss>>saved[i];
-      }
+  void GetState(std::vector<RngStateType> & saved,RngEngine &eng) {
+    saved.resize(RngStateCount);
+    std::stringstream ss;
+    ss<<eng;
+    ss.seekg(0,ss.beg);
+    for(int i=0;i<RngStateCount;i++){
+      ss>>saved[i];
     }
-    void GetState(std::vector<RngStateType> & saved,int gen) {
-      GetState(saved,_generators[gen]);
+  }
+  void GetState(std::vector<RngStateType> & saved,int gen) {
+    GetState(saved,_generators[gen]);
+  }
+  void SetState(std::vector<RngStateType> & saved,RngEngine &eng){
+    assert(saved.size()==RngStateCount);
+    std::stringstream ss;
+    for(int i=0;i<RngStateCount;i++){
+      ss<< saved[i]<<" ";
     }
-    void SetState(std::vector<RngStateType> & saved,RngEngine &eng){
-      assert(saved.size()==RngStateCount);
-      std::stringstream ss;
-      for(int i=0;i<RngStateCount;i++){
-        ss<< saved[i]<<" ";
-      }
-      ss.seekg(0,ss.beg);
-      ss>>eng;
-    }
-    void SetState(std::vector<RngStateType> & saved,int gen){
-      SetState(saved,_generators[gen]);
-    }
-    void SetEngine(RngEngine &Eng, int gen){
-      _generators[gen]=Eng;
-    }
-    void GetEngine(RngEngine &Eng, int gen){
-      Eng=_generators[gen];
-    }
-    template<class source> void Seed(source &src, int gen)
-    {
-      _generators[gen] = RngEngine(src);
-    }    
-  };
+    ss.seekg(0,ss.beg);
+    ss>>eng;
+  }
+  void SetState(std::vector<RngStateType> & saved,int gen){
+    SetState(saved,_generators[gen]);
+  }
+  void SetEngine(RngEngine &Eng, int gen){
+    _generators[gen]=Eng;
+  }
+  void GetEngine(RngEngine &Eng, int gen){
+    Eng=_generators[gen];
+  }
+  template<class source> void Seed(source &src, int gen)
+  {
+    _generators[gen] = RngEngine(src);
+  }    
+};
 
-  class GridSerialRNG : public GridRNGbase {
-  public:
+class GridSerialRNG : public GridRNGbase {
+public:
 
-    GridSerialRNG() : GridRNGbase() {
-      _generators.resize(1);
-      _uniform.resize(1,std::uniform_real_distribution<RealD>{0,1});
-      _gaussian.resize(1,std::normal_distribution<RealD>(0.0,1.0) );
-      _bernoulli.resize(1,std::discrete_distribution<int32_t>{1,1});
-      _uid.resize(1,std::uniform_int_distribution<uint32_t>() );
-    }
+  GridSerialRNG() : GridRNGbase() {
+    _generators.resize(1);
+    _uniform.resize(1,std::uniform_real_distribution<RealD>{0,1});
+    _gaussian.resize(1,std::normal_distribution<RealD>(0.0,1.0) );
+    _bernoulli.resize(1,std::discrete_distribution<int32_t>{1,1});
+    _uid.resize(1,std::uniform_int_distribution<uint32_t>() );
+  }
 
-    template <class sobj,class distribution> inline void fill(sobj &l,std::vector<distribution> &dist){
+  template <class sobj,class distribution> inline void fill(sobj &l,std::vector<distribution> &dist){
 
-      typedef typename sobj::scalar_type scalar_type;
+    typedef typename sobj::scalar_type scalar_type;
  
-      int words = sizeof(sobj)/sizeof(scalar_type);
+    int words = sizeof(sobj)/sizeof(scalar_type);
 
-      scalar_type *buf = (scalar_type *) & l;
+    scalar_type *buf = (scalar_type *) & l;
 
-      dist[0].reset();
-      for(int idx=0;idx<words;idx++){
-	fillScalar(buf[idx],dist[0],_generators[0]);
-      }
+    dist[0].reset();
+    for(int idx=0;idx<words;idx++){
+      fillScalar(buf[idx],dist[0],_generators[0]);
+    }
 
-      CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
+    CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
 
-    };
+  }
 
-    template <class distribution>  inline void fill(ComplexF &l,std::vector<distribution> &dist){
-      dist[0].reset();
-      fillScalar(l,dist[0],_generators[0]);
-      CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
+  template <class distribution>  inline void fill(ComplexF &l,std::vector<distribution> &dist){
+    dist[0].reset();
+    fillScalar(l,dist[0],_generators[0]);
+    CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
+  }
+  template <class distribution>  inline void fill(ComplexD &l,std::vector<distribution> &dist){
+    dist[0].reset();
+    fillScalar(l,dist[0],_generators[0]);
+    CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
+  }
+  template <class distribution>  inline void fill(RealF &l,std::vector<distribution> &dist){
+    dist[0].reset();
+    fillScalar(l,dist[0],_generators[0]);
+    CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
+  }
+  template <class distribution>  inline void fill(RealD &l,std::vector<distribution> &dist){
+    dist[0].reset();
+    fillScalar(l,dist[0],_generators[0]);
+    CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
+  }
+  // vector fill
+  template <class distribution>  inline void fill(vComplexF &l,std::vector<distribution> &dist){
+    RealF *pointer=(RealF *)&l;
+    dist[0].reset();
+    for(int i=0;i<2*vComplexF::Nsimd();i++){
+      fillScalar(pointer[i],dist[0],_generators[0]);
     }
-    template <class distribution>  inline void fill(ComplexD &l,std::vector<distribution> &dist){
-      dist[0].reset();
-      fillScalar(l,dist[0],_generators[0]);
-      CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
+    CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
+  }
+  template <class distribution>  inline void fill(vComplexD &l,std::vector<distribution> &dist){
+    RealD *pointer=(RealD *)&l;
+    dist[0].reset();
+    for(int i=0;i<2*vComplexD::Nsimd();i++){
+      fillScalar(pointer[i],dist[0],_generators[0]);
     }
-    template <class distribution>  inline void fill(RealF &l,std::vector<distribution> &dist){
-      dist[0].reset();
-      fillScalar(l,dist[0],_generators[0]);
-      CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
+    CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
+  }
+  template <class distribution>  inline void fill(vRealF &l,std::vector<distribution> &dist){
+    RealF *pointer=(RealF *)&l;
+    dist[0].reset();
+    for(int i=0;i<vRealF::Nsimd();i++){
+      fillScalar(pointer[i],dist[0],_generators[0]);
     }
-    template <class distribution>  inline void fill(RealD &l,std::vector<distribution> &dist){
-      dist[0].reset();
-      fillScalar(l,dist[0],_generators[0]);
-      CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
-    }
-    // vector fill
-    template <class distribution>  inline void fill(vComplexF &l,std::vector<distribution> &dist){
-      RealF *pointer=(RealF *)&l;
-      dist[0].reset();
-      for(int i=0;i<2*vComplexF::Nsimd();i++){
-	fillScalar(pointer[i],dist[0],_generators[0]);
-      }
-      CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
-    }
-    template <class distribution>  inline void fill(vComplexD &l,std::vector<distribution> &dist){
-      RealD *pointer=(RealD *)&l;
-      dist[0].reset();
-      for(int i=0;i<2*vComplexD::Nsimd();i++){
-	fillScalar(pointer[i],dist[0],_generators[0]);
-      }
-      CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
-    }
-    template <class distribution>  inline void fill(vRealF &l,std::vector<distribution> &dist){
-      RealF *pointer=(RealF *)&l;
-      dist[0].reset();
-      for(int i=0;i<vRealF::Nsimd();i++){
-	fillScalar(pointer[i],dist[0],_generators[0]);
-      }
-      CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
-    }
-    template <class distribution>  inline void fill(vRealD &l,std::vector<distribution> &dist){
-      RealD *pointer=(RealD *)&l;
-      dist[0].reset();
-      for(int i=0;i<vRealD::Nsimd();i++){
-	fillScalar(pointer[i],dist[0],_generators[0]);
-      }
-      CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
+    CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
+  }
+  template <class distribution>  inline void fill(vRealD &l,std::vector<distribution> &dist){
+    RealD *pointer=(RealD *)&l;
+    dist[0].reset();
+    for(int i=0;i<vRealD::Nsimd();i++){
+      fillScalar(pointer[i],dist[0],_generators[0]);
     }
+    CartesianCommunicator::BroadcastWorld(0,(void *)&l,sizeof(l));
+  }
     
-    void SeedFixedIntegers(const std::vector<int> &seeds){
-      CartesianCommunicator::BroadcastWorld(0,(void *)&seeds[0],sizeof(int)*seeds.size());
-      std::seed_seq src(seeds.begin(),seeds.end());
-      Seed(src,0);
-    }
+  void SeedFixedIntegers(const std::vector<int> &seeds){
+    CartesianCommunicator::BroadcastWorld(0,(void *)&seeds[0],sizeof(int)*seeds.size());
+    std::seed_seq src(seeds.begin(),seeds.end());
+    Seed(src,0);
+  }
 
     void SeedUniqueString(const std::string &s){
       std::vector<int> seeds;
@@ -330,65 +337,67 @@ namespace Grid {
       std::cout << GridLogMessage << "Seed SHA256: " << sha.str() << std::endl;
       SeedFixedIntegers(seeds);
     }
-  };
+};
 
-  class GridParallelRNG : public GridRNGbase {
+class GridParallelRNG : public GridRNGbase {
+private:
+  double _time_counter;
+  GridBase *_grid;
+  unsigned int _vol;
 
-    double _time_counter;
+public:
+  GridBase *Grid(void) const { return _grid; }
+  int generator_idx(int os,int is) {
+    return is*_grid->oSites()+os;
+  }
 
-  public:
-    GridBase *_grid;
-    unsigned int _vol;
+  GridParallelRNG(GridBase *grid) : GridRNGbase() {
+    _grid = grid;
+    _vol  =_grid->iSites()*_grid->oSites();
 
-    int generator_idx(int os,int is) {
-      return is*_grid->oSites()+os;
-    }
+    _generators.resize(_vol);
+    _uniform.resize(_vol,std::uniform_real_distribution<RealD>{0,1});
+    _gaussian.resize(_vol,std::normal_distribution<RealD>(0.0,1.0) );
+    _bernoulli.resize(_vol,std::discrete_distribution<int32_t>{1,1});
+    _uid.resize(_vol,std::uniform_int_distribution<uint32_t>() );
+  }
 
-    GridParallelRNG(GridBase *grid) : GridRNGbase() {
-      _grid = grid;
-      _vol  =_grid->iSites()*_grid->oSites();
+  template <class vobj,class distribution> inline void fill(Lattice<vobj> &l,std::vector<distribution> &dist){
 
-      _generators.resize(_vol);
-      _uniform.resize(_vol,std::uniform_real_distribution<RealD>{0,1});
-      _gaussian.resize(_vol,std::normal_distribution<RealD>(0.0,1.0) );
-      _bernoulli.resize(_vol,std::discrete_distribution<int32_t>{1,1});
-      _uid.resize(_vol,std::uniform_int_distribution<uint32_t>() );
-    }
+    typedef typename vobj::scalar_object scalar_object;
+    typedef typename vobj::scalar_type scalar_type;
+    typedef typename vobj::vector_type vector_type;
 
-    template <class vobj,class distribution> inline void fill(Lattice<vobj> &l,std::vector<distribution> &dist){
+    double inner_time_counter = usecond();
 
-      typedef typename vobj::scalar_object scalar_object;
-      typedef typename vobj::scalar_type scalar_type;
-      typedef typename vobj::vector_type vector_type;
+    int multiplicity = RNGfillable_general(_grid, l.Grid()); // l has finer or same grid
+    int Nsimd  = _grid->Nsimd();  // guaranteed to be the same for l.Grid() too
+    int osites = _grid->oSites();  // guaranteed to be <= l.Grid()->oSites() by a factor multiplicity
+    int words  = sizeof(scalar_object) / sizeof(scalar_type);
 
-      double inner_time_counter = usecond();
+    auto l_v = l.View();
+    thread_for( ss, osites, {
+      ExtractBuffer<scalar_object> buf(Nsimd);
+      for (int m = 0; m < multiplicity; m++) {  // Draw from same generator multiplicity times
 
-      int multiplicity = RNGfillable_general(_grid, l._grid); // l has finer or same grid
-      int Nsimd  = _grid->Nsimd();  // guaranteed to be the same for l._grid too
-      int osites = _grid->oSites();  // guaranteed to be <= l._grid->oSites() by a factor multiplicity
-      int words  = sizeof(scalar_object) / sizeof(scalar_type);
+	int sm = multiplicity * ss + m;  // Maps the generator site to the fine site
 
-      parallel_for(int ss=0;ss<osites;ss++){
-        std::vector<scalar_object> buf(Nsimd);
-        for (int m = 0; m < multiplicity; m++) {  // Draw from same generator multiplicity times
-
-          int sm = multiplicity * ss + m;  // Maps the generator site to the fine site
-
-          for (int si = 0; si < Nsimd; si++) {
+	for (int si = 0; si < Nsimd; si++) {
             
-            int gdx = generator_idx(ss, si);  // index of generator state
-            scalar_type *pointer = (scalar_type *)&buf[si];
-            dist[gdx].reset();
-            for (int idx = 0; idx < words; idx++) 
-              fillScalar(pointer[idx], dist[gdx], _generators[gdx]);
-          }
-          // merge into SIMD lanes, FIXME suboptimal implementation
-          merge(l._odata[sm], buf);
-        }
+	  int gdx = generator_idx(ss, si);  // index of generator state
+	  scalar_type *pointer = (scalar_type *)&buf[si];
+	  dist[gdx].reset();
+	  for (int idx = 0; idx < words; idx++) 
+	    fillScalar(pointer[idx], dist[gdx], _generators[gdx]);
+	}
+	// merge into SIMD lanes, FIXME suboptimal implementation
+	merge(l_v[sm], buf);
       }
+      });
+    //    });
 
-      _time_counter += usecond()- inner_time_counter;
-    };
+    _time_counter += usecond()- inner_time_counter;
+  }
 
     void SeedUniqueString(const std::string &s){
       std::vector<int> seeds;
@@ -398,119 +407,119 @@ namespace Grid {
       std::cout << GridLogMessage << "Seed SHA256: " << GridChecksum::sha256_string(seeds) << std::endl;
       SeedFixedIntegers(seeds);
     }
-    void SeedFixedIntegers(const std::vector<int> &seeds){
+  void SeedFixedIntegers(const std::vector<int> &seeds){
 
-      // Everyone generates the same seed_seq based on input seeds
-      CartesianCommunicator::BroadcastWorld(0,(void *)&seeds[0],sizeof(int)*seeds.size());
+    // Everyone generates the same seed_seq based on input seeds
+    CartesianCommunicator::BroadcastWorld(0,(void *)&seeds[0],sizeof(int)*seeds.size());
 
-      std::seed_seq source(seeds.begin(),seeds.end());
+    std::seed_seq source(seeds.begin(),seeds.end());
 
-      RngEngine master_engine(source);
+    RngEngine master_engine(source);
 
 #ifdef RNG_FAST_DISCARD
-      ////////////////////////////////////////////////
-      // Skip ahead through a single stream.
-      // Applicable to SITMO and other has based/crypto RNGs
-      // Should be applicable to Mersenne Twister, but the C++11
-      // MT implementation does not implement fast discard even though
-      // in principle this is possible
-      ////////////////////////////////////////////////
-
-      // Everybody loops over global volume.
-      parallel_for(int gidx=0;gidx<_grid->_gsites;gidx++){
+    ////////////////////////////////////////////////
+    // Skip ahead through a single stream.
+    // Applicable to SITMO and other has based/crypto RNGs
+    // Should be applicable to Mersenne Twister, but the C++11
+    // MT implementation does not implement fast discard even though
+    // in principle this is possible
+    ////////////////////////////////////////////////
 
+    // Everybody loops over global volume.
+    thread_for( gidx, _grid->_gsites, {
 	// Where is it?
-	int rank,o_idx,i_idx;
-	std::vector<int> gcoor;
+	int rank;
+	int o_idx;
+	int i_idx;
 
+	Coordinate gcoor;
 	_grid->GlobalIndexToGlobalCoor(gidx,gcoor);
 	_grid->GlobalCoorToRankIndex(rank,o_idx,i_idx,gcoor);
-
+	
 	// If this is one of mine we take it
 	if( rank == _grid->ThisRank() ){
 	  int l_idx=generator_idx(o_idx,i_idx);
 	  _generators[l_idx] = master_engine;
 	  Skip(_generators[l_idx],gidx); // Skip to next RNG sequence
 	}
-
-      }
+    });
 #else 
-      ////////////////////////////////////////////////////////////////
-      // Machine and thread decomposition dependent seeding is efficient
-      // and maximally parallel; but NOT reproducible from machine to machine. 
-      // Not ideal, but fastest way to reseed all nodes.
-      ////////////////////////////////////////////////////////////////
-      {
-	// Obtain one Reseed per processor
-	int Nproc = _grid->ProcessorCount();
-	std::vector<RngEngine> seeders(Nproc);
-	int me= _grid->ThisRank();
-	for(int p=0;p<Nproc;p++){
-	  seeders[p] = Reseed(master_engine);
-	}
-	master_engine = seeders[me];
+    ////////////////////////////////////////////////////////////////
+    // Machine and thread decomposition dependent seeding is efficient
+    // and maximally parallel; but NOT reproducible from machine to machine. 
+    // Not ideal, but fastest way to reseed all nodes.
+    ////////////////////////////////////////////////////////////////
+    {
+      // Obtain one Reseed per processor
+      int Nproc = _grid->ProcessorCount();
+      std::vector<RngEngine> seeders(Nproc);
+      int me= _grid->ThisRank();
+      for(int p=0;p<Nproc;p++){
+	seeders[p] = Reseed(master_engine);
+      }
+      master_engine = seeders[me];
+    }
+
+    {
+      // Obtain one reseeded generator per thread
+      int Nthread = GridThread::GetThreads();
+      std::vector<RngEngine> seeders(Nthread);
+      for(int t=0;t<Nthread;t++){
+	seeders[t] = Reseed(master_engine);
       }
 
-      {
-	// Obtain one reseeded generator per thread
-	int Nthread = GridThread::GetThreads();
-	std::vector<RngEngine> seeders(Nthread);
-	for(int t=0;t<Nthread;t++){
-	  seeders[t] = Reseed(master_engine);
-	}
-
-	parallel_for(int t=0;t<Nthread;t++) {
-	  // set up one per local site in threaded fashion
-	  std::vector<uint32_t> newseeds;
-	  std::uniform_int_distribution<uint32_t> uid;	
-	  for(int l=0;l<_grid->lSites();l++) {
-	    if ( (l%Nthread)==t ) {
-	      _generators[l] = Reseed(seeders[t],newseeds,uid);
-	    }
+      thread_for( t, Nthread, {
+	// set up one per local site in threaded fashion
+	std::vector<uint32_t> newseeds;
+	std::uniform_int_distribution<uint32_t> uid;	
+	for(int l=0;l<_grid->lSites();l++) {
+	  if ( (l%Nthread)==t ) {
+	    _generators[l] = Reseed(seeders[t],newseeds,uid);
 	  }
 	}
-      }
+      });
+    }
 #endif
+  }
+
+  void Report(){
+    std::cout << GridLogMessage << "Time spent in the fill() routine by GridParallelRNG: "<< _time_counter/1e3 << " ms" << std::endl;
+  }
+
+
+  ////////////////////////////////////////////////////////////////////////
+  // Support for rigorous test of RNG's
+  // Return uniform random uint32_t from requested site generator
+  ////////////////////////////////////////////////////////////////////////
+  uint32_t GlobalU01(int gsite){
+
+    uint32_t the_number;
+    // who
+    int rank,o_idx,i_idx;
+    Coordinate gcoor;
+    _grid->GlobalIndexToGlobalCoor(gsite,gcoor);
+    _grid->GlobalCoorToRankIndex(rank,o_idx,i_idx,gcoor);
+
+    // draw
+    int l_idx=generator_idx(o_idx,i_idx);
+    if( rank == _grid->ThisRank() ){
+      the_number = _uid[l_idx](_generators[l_idx]);
     }
-
-    void Report(){
-      std::cout << GridLogMessage << "Time spent in the fill() routine by GridParallelRNG: "<< _time_counter/1e3 << " ms" << std::endl;
-    }
-
-
-    ////////////////////////////////////////////////////////////////////////
-    // Support for rigorous test of RNG's
-    // Return uniform random uint32_t from requested site generator
-    ////////////////////////////////////////////////////////////////////////
-    uint32_t GlobalU01(int gsite){
-
-      uint32_t the_number;
-      // who
-      std::vector<int> gcoor;
-      int rank,o_idx,i_idx;
-      _grid->GlobalIndexToGlobalCoor(gsite,gcoor);
-      _grid->GlobalCoorToRankIndex(rank,o_idx,i_idx,gcoor);
-
-      // draw
-      int l_idx=generator_idx(o_idx,i_idx);
-      if( rank == _grid->ThisRank() ){
-	the_number = _uid[l_idx](_generators[l_idx]);
-      }
       
-      // share & return
-      _grid->Broadcast(rank,(void *)&the_number,sizeof(the_number));
-      return the_number;
-    }
+    // share & return
+    _grid->Broadcast(rank,(void *)&the_number,sizeof(the_number));
+    return the_number;
+  }
 
-  };
+};
 
-  template <class vobj> inline void random(GridParallelRNG &rng,Lattice<vobj> &l)   { rng.fill(l,rng._uniform);  }
-  template <class vobj> inline void gaussian(GridParallelRNG &rng,Lattice<vobj> &l) { rng.fill(l,rng._gaussian); }
-  template <class vobj> inline void bernoulli(GridParallelRNG &rng,Lattice<vobj> &l){ rng.fill(l,rng._bernoulli);}
+template <class vobj> inline void random(GridParallelRNG &rng,Lattice<vobj> &l)   { rng.fill(l,rng._uniform);  }
+template <class vobj> inline void gaussian(GridParallelRNG &rng,Lattice<vobj> &l) { rng.fill(l,rng._gaussian); }
+template <class vobj> inline void bernoulli(GridParallelRNG &rng,Lattice<vobj> &l){ rng.fill(l,rng._bernoulli);}
 
-  template <class sobj> inline void random(GridSerialRNG &rng,sobj &l)   { rng.fill(l,rng._uniform  ); }
-  template <class sobj> inline void gaussian(GridSerialRNG &rng,sobj &l) { rng.fill(l,rng._gaussian ); }
-  template <class sobj> inline void bernoulli(GridSerialRNG &rng,sobj &l){ rng.fill(l,rng._bernoulli); }
+template <class sobj> inline void random(GridSerialRNG &rng,sobj &l)   { rng.fill(l,rng._uniform  ); }
+template <class sobj> inline void gaussian(GridSerialRNG &rng,sobj &l) { rng.fill(l,rng._gaussian ); }
+template <class sobj> inline void bernoulli(GridSerialRNG &rng,sobj &l){ rng.fill(l,rng._bernoulli); }
 
-}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/lattice/Lattice_trace.h b/Grid/lattice/Lattice_trace.h
index 449c55f8..93444e0c 100644
--- a/Grid/lattice/Lattice_trace.h
+++ b/Grid/lattice/Lattice_trace.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,8 +23,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_LATTICE_TRACE_H
 #define GRID_LATTICE_TRACE_H
 
@@ -32,36 +32,38 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 // Tracing, transposing, peeking, poking
 ///////////////////////////////////////////////
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-    ////////////////////////////////////////////////////////////////////////////////////////////////////
-    // Trace
-    ////////////////////////////////////////////////////////////////////////////////////////////////////
-    template<class vobj>
-    inline auto trace(const Lattice<vobj> &lhs)
-      -> Lattice<decltype(trace(lhs._odata[0]))>
-    {
-      Lattice<decltype(trace(lhs._odata[0]))> ret(lhs._grid);
-      parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-            ret._odata[ss] = trace(lhs._odata[ss]);
-        }
-        return ret;
-    };
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Trace
+////////////////////////////////////////////////////////////////////////////////////////////////////
+template<class vobj>
+inline auto trace(const Lattice<vobj> &lhs)  -> Lattice<decltype(trace(vobj()))>
+{
+  Lattice<decltype(trace(vobj()))> ret(lhs.Grid());
+  auto ret_v = ret.View();
+  auto lhs_v = lhs.View();
+  accelerator_for( ss, lhs_v.size(), vobj::Nsimd(), {
+    coalescedWrite(ret_v[ss], trace(lhs_v(ss)));
+  });
+  return ret;
+};
     
-    ////////////////////////////////////////////////////////////////////////////////////////////////////
-    // Trace Index level dependent operation
-    ////////////////////////////////////////////////////////////////////////////////////////////////////
-    template<int Index,class vobj>
-    inline auto TraceIndex(const Lattice<vobj> &lhs) -> Lattice<decltype(traceIndex<Index>(lhs._odata[0]))>
-    {
-      Lattice<decltype(traceIndex<Index>(lhs._odata[0]))> ret(lhs._grid);
-      parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-	ret._odata[ss] = traceIndex<Index>(lhs._odata[ss]);
-      }
-      return ret;
-    };
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Trace Index level dependent operation
+////////////////////////////////////////////////////////////////////////////////////////////////////
+template<int Index,class vobj>
+inline auto TraceIndex(const Lattice<vobj> &lhs) -> Lattice<decltype(traceIndex<Index>(vobj()))>
+{
+  Lattice<decltype(traceIndex<Index>(vobj()))> ret(lhs.Grid());
+  auto ret_v = ret.View();
+  auto lhs_v = lhs.View();
+  accelerator_for( ss, lhs_v.size(), vobj::Nsimd(), {
+    coalescedWrite(ret_v[ss], traceIndex<Index>(lhs_v(ss)));
+  });
+  return ret;
+};
 
-
-}
+NAMESPACE_END(Grid);
 #endif
 
diff --git a/Grid/lattice/Lattice_transfer.h b/Grid/lattice/Lattice_transfer.h
index 69c132ed..865a4b14 100644
--- a/Grid/lattice/Lattice_transfer.h
+++ b/Grid/lattice/Lattice_transfer.h
@@ -23,12 +23,11 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef GRID_LATTICE_TRANSFER_H
-#define GRID_LATTICE_TRANSFER_H
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 inline void subdivides(GridBase *coarse,GridBase *fine)
 {
@@ -45,39 +44,44 @@ inline void subdivides(GridBase *coarse,GridBase *fine)
 }
 
  
-  ////////////////////////////////////////////////////////////////////////////////////////////
-  // remove and insert a half checkerboard
-  ////////////////////////////////////////////////////////////////////////////////////////////
-  template<class vobj> inline void pickCheckerboard(int cb,Lattice<vobj> &half,const Lattice<vobj> &full){
-    half.checkerboard = cb;
+////////////////////////////////////////////////////////////////////////////////////////////
+// remove and insert a half checkerboard
+////////////////////////////////////////////////////////////////////////////////////////////
+template<class vobj> inline void pickCheckerboard(int cb,Lattice<vobj> &half,const Lattice<vobj> &full){
+  half.Checkerboard() = cb;
 
-    parallel_for(int ss=0;ss<full._grid->oSites();ss++){
-      int cbos;
-      std::vector<int> coor;
-      full._grid->oCoorFromOindex(coor,ss);
-      cbos=half._grid->CheckerBoard(coor);
-      
-      if (cbos==cb) {
-	int ssh=half._grid->oIndex(coor);
-	half._odata[ssh] = full._odata[ss];
-      }
-    }
-  }
-  template<class vobj> inline void setCheckerboard(Lattice<vobj> &full,const Lattice<vobj> &half){
-    int cb = half.checkerboard;
-    parallel_for(int ss=0;ss<full._grid->oSites();ss++){
-      std::vector<int> coor;
-      int cbos;
+  auto half_v = half.View();
+  auto full_v = full.View();
+  thread_for(ss, full.Grid()->oSites(),{
+    int cbos;
+    Coordinate coor;
+    full.Grid()->oCoorFromOindex(coor,ss);
+    cbos=half.Grid()->CheckerBoard(coor);
 
-      full._grid->oCoorFromOindex(coor,ss);
-      cbos=half._grid->CheckerBoard(coor);
-      
-      if (cbos==cb) {
-	int ssh=half._grid->oIndex(coor);
-	full._odata[ss]=half._odata[ssh];
-      }
+    if (cbos==cb) {
+      int ssh=half.Grid()->oIndex(coor);
+      half_v[ssh] = full_v[ss];
     }
-  }
+  });
+}
+template<class vobj> inline void setCheckerboard(Lattice<vobj> &full,const Lattice<vobj> &half){
+  int cb = half.Checkerboard();
+  auto half_v = half.View();
+  auto full_v = full.View();
+  thread_for(ss,full.Grid()->oSites(),{
+
+    Coordinate coor;
+    int cbos;
+
+    full.Grid()->oCoorFromOindex(coor,ss);
+    cbos=half.Grid()->CheckerBoard(coor);
+      
+    if (cbos==cb) {
+      int ssh=half.Grid()->oIndex(coor);
+      full_v[ss]=half_v[ssh];
+    }
+  });
+}
   
 
 template<class vobj,class CComplex,int nbasis>
@@ -85,8 +89,8 @@ inline void blockProject(Lattice<iVector<CComplex,nbasis > > &coarseData,
 			 const             Lattice<vobj>   &fineData,
 			 const std::vector<Lattice<vobj> > &Basis)
 {
-  GridBase * fine  = fineData._grid;
-  GridBase * coarse= coarseData._grid;
+  GridBase * fine  = fineData.Grid();
+  GridBase * coarse= coarseData.Grid();
   int  _ndimension = coarse->_ndimension;
 
   // checks
@@ -96,33 +100,33 @@ inline void blockProject(Lattice<iVector<CComplex,nbasis > > &coarseData,
     conformable(Basis[i],fineData);
   }
 
-  std::vector<int>  block_r      (_ndimension);
+  Coordinate block_r      (_ndimension);
   
   for(int d=0 ; d<_ndimension;d++){
     block_r[d] = fine->_rdimensions[d] / coarse->_rdimensions[d];
     assert(block_r[d]*coarse->_rdimensions[d] == fine->_rdimensions[d]);
   }
 
-  coarseData=zero;
+  coarseData=Zero();
 
+  auto fineData_   = fineData.View();
+  auto coarseData_ = coarseData.View();
   // Loop over coars parallel, and then loop over fine associated with coarse.
-  parallel_for(int sf=0;sf<fine->oSites();sf++){
-
+  thread_for( sf, fine->oSites(), {
     int sc;
-    std::vector<int> coor_c(_ndimension);
-    std::vector<int> coor_f(_ndimension);
+    Coordinate coor_c(_ndimension);
+    Coordinate coor_f(_ndimension);
     Lexicographic::CoorFromIndex(coor_f,sf,fine->_rdimensions);
     for(int d=0;d<_ndimension;d++) coor_c[d]=coor_f[d]/block_r[d];
     Lexicographic::IndexFromCoor(coor_c,sc,coarse->_rdimensions);
 
-PARALLEL_CRITICAL
-    for(int i=0;i<nbasis;i++) {
-
-      coarseData._odata[sc](i)=coarseData._odata[sc](i)
-	+ innerProduct(Basis[i]._odata[sf],fineData._odata[sf]);
-
+    thread_critical {
+      for(int i=0;i<nbasis;i++) {
+	auto Basis_      = Basis[i].View();
+	coarseData_[sc](i)=coarseData_[sc](i) + innerProduct(Basis_[sf],fineData_[sf]);
+      }
     }
-  }
+  });
   return;
 }
 
@@ -132,18 +136,18 @@ inline void blockZAXPY(Lattice<vobj> &fineZ,
 		       const Lattice<vobj> &fineX,
 		       const Lattice<vobj> &fineY)
 {
-  GridBase * fine  = fineZ._grid;
-  GridBase * coarse= coarseA._grid;
+  GridBase * fine  = fineZ.Grid();
+  GridBase * coarse= coarseA.Grid();
 
-  fineZ.checkerboard=fineX.checkerboard;
-  assert(fineX.checkerboard==fineY.checkerboard);
+  fineZ.Checkerboard()=fineX.Checkerboard();
+  assert(fineX.Checkerboard()==fineY.Checkerboard());
   subdivides(coarse,fine); // require they map
   conformable(fineX,fineY);
   conformable(fineX,fineZ);
 
   int _ndimension = coarse->_ndimension;
   
-  std::vector<int>  block_r      (_ndimension);
+  Coordinate  block_r      (_ndimension);
 
   // FIXME merge with subdivide checking routine as this is redundant
   for(int d=0 ; d<_ndimension;d++){
@@ -151,48 +155,56 @@ inline void blockZAXPY(Lattice<vobj> &fineZ,
     assert(block_r[d]*coarse->_rdimensions[d]==fine->_rdimensions[d]);
   }
 
-  parallel_for(int sf=0;sf<fine->oSites();sf++){
+  auto fineZ_  = fineZ.View();
+  auto fineX_  = fineX.View();
+  auto fineY_  = fineY.View();
+  auto coarseA_= coarseA.View();
+
+  thread_for(sf, fine->oSites(), {
     
     int sc;
-    std::vector<int> coor_c(_ndimension);
-    std::vector<int> coor_f(_ndimension);
+    Coordinate coor_c(_ndimension);
+    Coordinate coor_f(_ndimension);
 
     Lexicographic::CoorFromIndex(coor_f,sf,fine->_rdimensions);
     for(int d=0;d<_ndimension;d++) coor_c[d]=coor_f[d]/block_r[d];
     Lexicographic::IndexFromCoor(coor_c,sc,coarse->_rdimensions);
 
     // z = A x + y
-    fineZ._odata[sf]=coarseA._odata[sc]*fineX._odata[sf]+fineY._odata[sf];
+    fineZ_[sf]=coarseA_[sc]*fineX_[sf]+fineY_[sf];
 
-  }
+  });
 
   return;
 }
 template<class vobj,class CComplex>
-  inline void blockInnerProduct(Lattice<CComplex> &CoarseInner,
-				const Lattice<vobj> &fineX,
-				const Lattice<vobj> &fineY)
+inline void blockInnerProduct(Lattice<CComplex> &CoarseInner,
+			      const Lattice<vobj> &fineX,
+			      const Lattice<vobj> &fineY)
 {
-  typedef decltype(innerProduct(fineX._odata[0],fineY._odata[0])) dotp;
+  typedef decltype(innerProduct(vobj(),vobj())) dotp;
 
-  GridBase *coarse(CoarseInner._grid);
-  GridBase *fine  (fineX._grid);
+  GridBase *coarse(CoarseInner.Grid());
+  GridBase *fine  (fineX.Grid());
 
-  Lattice<dotp> fine_inner(fine); fine_inner.checkerboard = fineX.checkerboard;
+  Lattice<dotp> fine_inner(fine); fine_inner.Checkerboard() = fineX.Checkerboard();
   Lattice<dotp> coarse_inner(coarse);
 
   // Precision promotion?
+  auto CoarseInner_  = CoarseInner.View();
+  auto coarse_inner_ = coarse_inner.View();
+
   fine_inner = localInnerProduct(fineX,fineY);
   blockSum(coarse_inner,fine_inner);
-  parallel_for(int ss=0;ss<coarse->oSites();ss++){
-    CoarseInner._odata[ss] = coarse_inner._odata[ss];
-  }
+  thread_for(ss, coarse->oSites(),{
+    CoarseInner_[ss] = coarse_inner_[ss];
+  });
 }
 template<class vobj,class CComplex>
 inline void blockNormalise(Lattice<CComplex> &ip,Lattice<vobj> &fineX)
 {
-  GridBase *coarse = ip._grid;
-  Lattice<vobj> zz(fineX._grid); zz=zero; zz.checkerboard=fineX.checkerboard;
+  GridBase *coarse = ip.Grid();
+  Lattice<vobj> zz(fineX.Grid()); zz=Zero(); zz.Checkerboard()=fineX.Checkerboard();
   blockInnerProduct(ip,fineX,fineX);
   ip = pow(ip,-0.5);
   blockZAXPY(fineX,ip,fineX,zz);
@@ -202,14 +214,14 @@ inline void blockNormalise(Lattice<CComplex> &ip,Lattice<vobj> &fineX)
 template<class vobj>
 inline void blockSum(Lattice<vobj> &coarseData,const Lattice<vobj> &fineData)
 {
-  GridBase * fine  = fineData._grid;
-  GridBase * coarse= coarseData._grid;
+  GridBase * fine  = fineData.Grid();
+  GridBase * coarse= coarseData.Grid();
 
   subdivides(coarse,fine); // require they map
 
   int _ndimension = coarse->_ndimension;
   
-  std::vector<int>  block_r      (_ndimension);
+  Coordinate  block_r      (_ndimension);
   
   for(int d=0 ; d<_ndimension;d++){
     block_r[d] = fine->_rdimensions[d] / coarse->_rdimensions[d];
@@ -217,36 +229,36 @@ inline void blockSum(Lattice<vobj> &coarseData,const Lattice<vobj> &fineData)
 
   // Turn this around to loop threaded over sc and interior loop 
   // over sf would thread better
-  coarseData=zero;
-  parallel_region {
+  coarseData=Zero();
+  auto coarseData_ = coarseData.View();
+  auto fineData_   = fineData.View();
 
+  thread_for(sf,fine->oSites(),{
     int sc;
-    std::vector<int> coor_c(_ndimension);
-    std::vector<int> coor_f(_ndimension);
-
-    parallel_for_internal(int sf=0;sf<fine->oSites();sf++){
+    Coordinate coor_c(_ndimension);
+    Coordinate coor_f(_ndimension);
     
-      Lexicographic::CoorFromIndex(coor_f,sf,fine->_rdimensions);
-      for(int d=0;d<_ndimension;d++) coor_c[d]=coor_f[d]/block_r[d];
-      Lexicographic::IndexFromCoor(coor_c,sc,coarse->_rdimensions);
-      
-PARALLEL_CRITICAL
-      coarseData._odata[sc]=coarseData._odata[sc]+fineData._odata[sf];
-
+    Lexicographic::CoorFromIndex(coor_f,sf,fine->_rdimensions);
+    for(int d=0;d<_ndimension;d++) coor_c[d]=coor_f[d]/block_r[d];
+    Lexicographic::IndexFromCoor(coor_c,sc,coarse->_rdimensions);
+    
+    thread_critical { 
+      coarseData_[sc]=coarseData_[sc]+fineData_[sf];
     }
-  }
+
+  });
   return;
 }
 
 template<class vobj>
-inline void blockPick(GridBase *coarse,const Lattice<vobj> &unpicked,Lattice<vobj> &picked,std::vector<int> coor)
+inline void blockPick(GridBase *coarse,const Lattice<vobj> &unpicked,Lattice<vobj> &picked,Coordinate coor)
 {
-  GridBase * fine = unpicked._grid;
+  GridBase * fine = unpicked.Grid();
 
-  Lattice<vobj> zz(fine); zz.checkerboard = unpicked.checkerboard;
+  Lattice<vobj> zz(fine); zz.Checkerboard() = unpicked.Checkerboard();
   Lattice<iScalar<vInteger> > fcoor(fine);
 
-  zz = zero;
+  zz = Zero();
 
   picked = unpicked;
   for(int d=0;d<fine->_ndimension;d++){
@@ -262,16 +274,15 @@ inline void blockPick(GridBase *coarse,const Lattice<vobj> &unpicked,Lattice<vob
 template<class vobj,class CComplex>
 inline void blockOrthogonalise(Lattice<CComplex> &ip,std::vector<Lattice<vobj> > &Basis)
 {
-  GridBase *coarse = ip._grid;
-  GridBase *fine   = Basis[0]._grid;
+  GridBase *coarse = ip.Grid();
+  GridBase *fine   = Basis[0].Grid();
 
   int       nbasis = Basis.size() ;
-  int  _ndimension = coarse->_ndimension;
 
   // checks
   subdivides(coarse,fine); 
   for(int i=0;i<nbasis;i++){
-    conformable(Basis[i]._grid,fine);
+    conformable(Basis[i].Grid(),fine);
   }
 
   for(int v=0;v<nbasis;v++) {
@@ -290,41 +301,41 @@ inline void blockPromote(const Lattice<iVector<CComplex,nbasis > > &coarseData,
 			 Lattice<vobj>   &fineData,
 			 const std::vector<Lattice<vobj> > &Basis)
 {
-  GridBase * fine  = fineData._grid;
-  GridBase * coarse= coarseData._grid;
+  GridBase * fine  = fineData.Grid();
+  GridBase * coarse= coarseData.Grid();
   int  _ndimension = coarse->_ndimension;
 
   // checks
   assert( nbasis == Basis.size() );
   subdivides(coarse,fine); 
   for(int i=0;i<nbasis;i++){
-    conformable(Basis[i]._grid,fine);
+    conformable(Basis[i].Grid(),fine);
   }
 
-  std::vector<int>  block_r      (_ndimension);
+  Coordinate  block_r      (_ndimension);
   
   for(int d=0 ; d<_ndimension;d++){
     block_r[d] = fine->_rdimensions[d] / coarse->_rdimensions[d];
   }
+  auto fineData_   = fineData.View();
+  auto coarseData_ = coarseData.View();
 
   // Loop with a cache friendly loop ordering
-  parallel_region {
+  thread_for(sf,fine->oSites(),{
     int sc;
-    std::vector<int> coor_c(_ndimension);
-    std::vector<int> coor_f(_ndimension);
+    Coordinate coor_c(_ndimension);
+    Coordinate coor_f(_ndimension);
 
-    parallel_for_internal(int sf=0;sf<fine->oSites();sf++){
+    Lexicographic::CoorFromIndex(coor_f,sf,fine->_rdimensions);
+    for(int d=0;d<_ndimension;d++) coor_c[d]=coor_f[d]/block_r[d];
+    Lexicographic::IndexFromCoor(coor_c,sc,coarse->_rdimensions);
 
-      Lexicographic::CoorFromIndex(coor_f,sf,fine->_rdimensions);
-      for(int d=0;d<_ndimension;d++) coor_c[d]=coor_f[d]/block_r[d];
-      Lexicographic::IndexFromCoor(coor_c,sc,coarse->_rdimensions);
-      
-      for(int i=0;i<nbasis;i++) {
-	if(i==0) fineData._odata[sf]=coarseData._odata[sc](i) * Basis[i]._odata[sf];
-	else     fineData._odata[sf]=fineData._odata[sf]+coarseData._odata[sc](i)*Basis[i]._odata[sf];
-      }
+    for(int i=0;i<nbasis;i++) {
+      auto basis_ = Basis[i].View();
+      if(i==0) fineData_[sf]=coarseData_[sc](i) *basis_[sf];
+      else     fineData_[sf]=fineData_[sf]+coarseData_[sc](i)*basis_[sf];
     }
-  }
+  });
   return;
   
 }
@@ -337,8 +348,8 @@ void localConvert(const Lattice<vobj> &in,Lattice<vvobj> &out)
   typedef typename vobj::scalar_object sobj;
   typedef typename vvobj::scalar_object ssobj;
 
-  GridBase *ig = in._grid;
-  GridBase *og = out._grid;
+  GridBase *ig = in.Grid();
+  GridBase *og = out.Grid();
 
   int ni = ig->_ndimension;
   int no = og->_ndimension;
@@ -351,16 +362,16 @@ void localConvert(const Lattice<vobj> &in,Lattice<vvobj> &out)
     assert(ig->lSites() == og->lSites());
   }
 
-  parallel_for(int idx=0;idx<ig->lSites();idx++){
+  thread_for(idx, ig->lSites(),{
     sobj s;
     ssobj ss;
 
-    std::vector<int> lcoor(ni);
+    Coordinate lcoor(ni);
     ig->LocalIndexToLocalCoor(idx,lcoor);
     peekLocalSite(s,in,lcoor);
     ss=s;
     pokeLocalSite(ss,out,lcoor);
-  }
+  });
 }
 
 
@@ -369,8 +380,8 @@ void InsertSlice(const Lattice<vobj> &lowDim,Lattice<vobj> & higherDim,int slice
 {
   typedef typename vobj::scalar_object sobj;
 
-  GridBase *lg = lowDim._grid;
-  GridBase *hg = higherDim._grid;
+  GridBase *lg = lowDim.Grid();
+  GridBase *hg = higherDim.Grid();
   int nl = lg->_ndimension;
   int nh = hg->_ndimension;
 
@@ -389,10 +400,10 @@ void InsertSlice(const Lattice<vobj> &lowDim,Lattice<vobj> & higherDim,int slice
   }
 
   // the above should guarantee that the operations are local
-  parallel_for(int idx=0;idx<lg->lSites();idx++){
+  thread_for(idx,lg->lSites(),{
     sobj s;
-    std::vector<int> lcoor(nl);
-    std::vector<int> hcoor(nh);
+    Coordinate lcoor(nl);
+    Coordinate hcoor(nh);
     lg->LocalIndexToLocalCoor(idx,lcoor);
     int ddl=0;
     hcoor[orthog] = slice;
@@ -403,7 +414,7 @@ void InsertSlice(const Lattice<vobj> &lowDim,Lattice<vobj> & higherDim,int slice
     }
     peekLocalSite(s,lowDim,lcoor);
     pokeLocalSite(s,higherDim,hcoor);
-  }
+  });
 }
 
 template<class vobj>
@@ -411,8 +422,8 @@ void ExtractSlice(Lattice<vobj> &lowDim,const Lattice<vobj> & higherDim,int slic
 {
   typedef typename vobj::scalar_object sobj;
 
-  GridBase *lg = lowDim._grid;
-  GridBase *hg = higherDim._grid;
+  GridBase *lg = lowDim.Grid();
+  GridBase *hg = higherDim.Grid();
   int nl = lg->_ndimension;
   int nh = hg->_ndimension;
 
@@ -422,18 +433,18 @@ void ExtractSlice(Lattice<vobj> &lowDim,const Lattice<vobj> & higherDim,int slic
   assert(hg->_processors[orthog]==1);
 
   int dl; dl = 0;
-    for(int d=0;d<nh;d++){
-      if ( d != orthog) {
-	assert(lg->_processors[dl]  == hg->_processors[d]);
-	assert(lg->_ldimensions[dl] == hg->_ldimensions[d]);
-	dl++;
+  for(int d=0;d<nh;d++){
+    if ( d != orthog) {
+      assert(lg->_processors[dl]  == hg->_processors[d]);
+      assert(lg->_ldimensions[dl] == hg->_ldimensions[d]);
+      dl++;
     }
   }
   // the above should guarantee that the operations are local
-  parallel_for(int idx=0;idx<lg->lSites();idx++){
+  thread_for(idx,lg->lSites(),{
     sobj s;
-    std::vector<int> lcoor(nl);
-    std::vector<int> hcoor(nh);
+    Coordinate lcoor(nl);
+    Coordinate hcoor(nh);
     lg->LocalIndexToLocalCoor(idx,lcoor);
     int ddl=0;
     hcoor[orthog] = slice;
@@ -444,7 +455,7 @@ void ExtractSlice(Lattice<vobj> &lowDim,const Lattice<vobj> & higherDim,int slic
     }
     peekLocalSite(s,higherDim,hcoor);
     pokeLocalSite(s,lowDim,lcoor);
-  }
+  });
 
 }
 
@@ -454,8 +465,8 @@ void InsertSliceLocal(const Lattice<vobj> &lowDim, Lattice<vobj> & higherDim,int
 {
   typedef typename vobj::scalar_object sobj;
 
-  GridBase *lg = lowDim._grid;
-  GridBase *hg = higherDim._grid;
+  GridBase *lg = lowDim.Grid();
+  GridBase *hg = higherDim.Grid();
   int nl = lg->_ndimension;
   int nh = hg->_ndimension;
 
@@ -465,16 +476,16 @@ void InsertSliceLocal(const Lattice<vobj> &lowDim, Lattice<vobj> & higherDim,int
 
   for(int d=0;d<nh;d++){
     if ( d!=orthog ) {
-      assert(lg->_processors[d]  == hg->_processors[d]);
-      assert(lg->_ldimensions[d] == hg->_ldimensions[d]);
-    }
+    assert(lg->_processors[d]  == hg->_processors[d]);
+    assert(lg->_ldimensions[d] == hg->_ldimensions[d]);
+  }
   }
 
   // the above should guarantee that the operations are local
-  parallel_for(int idx=0;idx<lg->lSites();idx++){
+  thread_for(idx,lg->lSites(),{
     sobj s;
-    std::vector<int> lcoor(nl);
-    std::vector<int> hcoor(nh);
+    Coordinate lcoor(nl);
+    Coordinate hcoor(nh);
     lg->LocalIndexToLocalCoor(idx,lcoor);
     if( lcoor[orthog] == slice_lo ) { 
       hcoor=lcoor;
@@ -482,7 +493,7 @@ void InsertSliceLocal(const Lattice<vobj> &lowDim, Lattice<vobj> & higherDim,int
       peekLocalSite(s,lowDim,lcoor);
       pokeLocalSite(s,higherDim,hcoor);
     }
-  }
+  });
 }
 
 
@@ -491,8 +502,8 @@ void ExtractSliceLocal(Lattice<vobj> &lowDim,const Lattice<vobj> & higherDim,int
 {
   typedef typename vobj::scalar_object sobj;
 
-  GridBase *lg = lowDim._grid;
-  GridBase *hg = higherDim._grid;
+  GridBase *lg = lowDim.Grid();
+  GridBase *hg = higherDim.Grid();
   int nl = lg->_ndimension;
   int nh = hg->_ndimension;
 
@@ -502,16 +513,16 @@ void ExtractSliceLocal(Lattice<vobj> &lowDim,const Lattice<vobj> & higherDim,int
 
   for(int d=0;d<nh;d++){
     if ( d!=orthog ) {
-      assert(lg->_processors[d]  == hg->_processors[d]);
-      assert(lg->_ldimensions[d] == hg->_ldimensions[d]);
-    }
+    assert(lg->_processors[d]  == hg->_processors[d]);
+    assert(lg->_ldimensions[d] == hg->_ldimensions[d]);
+  }
   }
 
   // the above should guarantee that the operations are local
-  parallel_for(int idx=0;idx<lg->lSites();idx++){
+  thread_for(idx,lg->lSites(),{
     sobj s;
-    std::vector<int> lcoor(nl);
-    std::vector<int> hcoor(nh);
+    Coordinate lcoor(nl);
+    Coordinate hcoor(nh);
     lg->LocalIndexToLocalCoor(idx,lcoor);
     if( lcoor[orthog] == slice_lo ) { 
       hcoor=lcoor;
@@ -519,7 +530,7 @@ void ExtractSliceLocal(Lattice<vobj> &lowDim,const Lattice<vobj> & higherDim,int
       peekLocalSite(s,higherDim,hcoor);
       pokeLocalSite(s,lowDim,lcoor);
     }
-  }
+  });
 }
 
 
@@ -528,8 +539,8 @@ void Replicate(Lattice<vobj> &coarse,Lattice<vobj> & fine)
 {
   typedef typename vobj::scalar_object sobj;
 
-  GridBase *cg = coarse._grid;
-  GridBase *fg =   fine._grid;
+  GridBase *cg = coarse.Grid();
+  GridBase *fg =   fine.Grid();
 
   int nd = cg->_ndimension;
 
@@ -537,14 +548,14 @@ void Replicate(Lattice<vobj> &coarse,Lattice<vobj> & fine)
 
   assert(cg->_ndimension==fg->_ndimension);
 
-  std::vector<int> ratio(cg->_ndimension);
+  Coordinate ratio(cg->_ndimension);
 
   for(int d=0;d<cg->_ndimension;d++){
     ratio[d] = fg->_fdimensions[d]/cg->_fdimensions[d];
   }
 
-  std::vector<int> fcoor(nd);
-  std::vector<int> ccoor(nd);
+  Coordinate fcoor(nd);
+  Coordinate ccoor(nd);
   for(int g=0;g<fg->gSites();g++){
 
     fg->GlobalIndexToGlobalCoor(g,fcoor);
@@ -567,41 +578,46 @@ unvectorizeToLexOrdArray(std::vector<sobj> &out, const Lattice<vobj> &in)
 
   typedef typename vobj::vector_type vtype;
   
-  GridBase* in_grid = in._grid;
+  GridBase* in_grid = in.Grid();
   out.resize(in_grid->lSites());
   
   int ndim = in_grid->Nd();
   int in_nsimd = vtype::Nsimd();
 
-  std::vector<std::vector<int> > in_icoor(in_nsimd);
+  std::vector<Coordinate > in_icoor(in_nsimd);
       
   for(int lane=0; lane < in_nsimd; lane++){
     in_icoor[lane].resize(ndim);
     in_grid->iCoorFromIindex(in_icoor[lane], lane);
   }
-  
-  parallel_for(int in_oidx = 0; in_oidx < in_grid->oSites(); in_oidx++){ //loop over outer index
-    //Assemble vector of pointers to output elements
-    std::vector<sobj*> out_ptrs(in_nsimd);
 
-    std::vector<int> in_ocoor(ndim);
+  //loop over outer index
+  auto in_v  = in.View();
+  thread_for(in_oidx,in_grid->oSites(),{
+    //Assemble vector of pointers to output elements
+    ExtractPointerArray<sobj> out_ptrs(in_nsimd);
+
+    Coordinate in_ocoor(ndim);
     in_grid->oCoorFromOindex(in_ocoor, in_oidx);
 
-    std::vector<int> lcoor(in_grid->Nd());
+    Coordinate lcoor(in_grid->Nd());
       
     for(int lane=0; lane < in_nsimd; lane++){
-      for(int mu=0;mu<ndim;mu++)
+
+      for(int mu=0;mu<ndim;mu++){
 	lcoor[mu] = in_ocoor[mu] + in_grid->_rdimensions[mu]*in_icoor[lane][mu];
+      }
 
       int lex;
       Lexicographic::IndexFromCoor(lcoor, lex, in_grid->_ldimensions);
+      assert(lex < out.size());
       out_ptrs[lane] = &out[lex];
     }
     
     //Unpack into those ptrs
-    const vobj & in_vobj = in._odata[in_oidx];
-    extract1(in_vobj, out_ptrs, 0);
-  }
+    const vobj & in_vobj = in_v[in_oidx];
+    extract(in_vobj, out_ptrs, 0);
+  });
 }
 
 template<typename vobj, typename sobj>
@@ -617,21 +633,21 @@ unvectorizeToRevLexOrdArray(std::vector<sobj> &out, const Lattice<vobj> &in)
   int ndim = in_grid->Nd();
   int in_nsimd = vtype::Nsimd();
 
-  std::vector<std::vector<int> > in_icoor(in_nsimd);
+  std::vector<Coordinate > in_icoor(in_nsimd);
       
   for(int lane=0; lane < in_nsimd; lane++){
     in_icoor[lane].resize(ndim);
     in_grid->iCoorFromIindex(in_icoor[lane], lane);
   }
   
-  parallel_for(int in_oidx = 0; in_oidx < in_grid->oSites(); in_oidx++){ //loop over outer index
+  thread_for(in_oidx, in_grid->oSites(),{
     //Assemble vector of pointers to output elements
     std::vector<sobj*> out_ptrs(in_nsimd);
 
-    std::vector<int> in_ocoor(ndim);
+    Coordinate in_ocoor(ndim);
     in_grid->oCoorFromOindex(in_ocoor, in_oidx);
 
-    std::vector<int> lcoor(in_grid->Nd());
+    Coordinate lcoor(in_grid->Nd());
       
     for(int lane=0; lane < in_nsimd; lane++){
       for(int mu=0;mu<ndim;mu++)
@@ -645,39 +661,38 @@ unvectorizeToRevLexOrdArray(std::vector<sobj> &out, const Lattice<vobj> &in)
     //Unpack into those ptrs
     const vobj & in_vobj = in._odata[in_oidx];
     extract1(in_vobj, out_ptrs, 0);
-  }
+  });
 }
 
 //Copy SIMD-vectorized lattice to array of scalar objects in lexicographic order
 template<typename vobj, typename sobj>
 typename std::enable_if<isSIMDvectorized<vobj>::value 
-                    && !isSIMDvectorized<sobj>::value, void>::type 
+			&& !isSIMDvectorized<sobj>::value, void>::type 
 vectorizeFromLexOrdArray( std::vector<sobj> &in, Lattice<vobj> &out)
 {
 
   typedef typename vobj::vector_type vtype;
   
-  GridBase* grid = out._grid;
+  GridBase* grid = out.Grid();
   assert(in.size()==grid->lSites());
   
-  int ndim     = grid->Nd();
-  int nsimd    = vtype::Nsimd();
+  const int ndim     = grid->Nd();
+  constexpr int nsimd    = vtype::Nsimd();
 
-  std::vector<std::vector<int> > icoor(nsimd);
+  std::vector<Coordinate > icoor(nsimd);
       
   for(int lane=0; lane < nsimd; lane++){
     icoor[lane].resize(ndim);
     grid->iCoorFromIindex(icoor[lane],lane);
   }
-  
-  parallel_for(uint64_t oidx = 0; oidx < grid->oSites(); oidx++){ //loop over outer index
+  auto out_v = out.View();
+  thread_for(oidx, grid->oSites(),{
     //Assemble vector of pointers to output elements
-    std::vector<sobj*> ptrs(nsimd);
+    ExtractPointerArray<sobj> ptrs(nsimd);
 
-    std::vector<int> ocoor(ndim);
+    Coordinate ocoor(ndim);
+    Coordinate lcoor(ndim);
     grid->oCoorFromOindex(ocoor, oidx);
-
-    std::vector<int> lcoor(grid->Nd());
       
     for(int lane=0; lane < nsimd; lane++){
 
@@ -692,9 +707,9 @@ vectorizeFromLexOrdArray( std::vector<sobj> &in, Lattice<vobj> &out)
     
     //pack from those ptrs
     vobj vecobj;
-    merge1(vecobj, ptrs, 0);
-    out._odata[oidx] = vecobj; 
-  }
+    merge(vecobj, ptrs, 0);
+    out_v[oidx] = vecobj; 
+  });
 }
 
 template<typename vobj, typename sobj>
@@ -711,21 +726,21 @@ vectorizeFromRevLexOrdArray( std::vector<sobj> &in, Lattice<vobj> &out)
   int ndim     = grid->Nd();
   int nsimd    = vtype::Nsimd();
 
-  std::vector<std::vector<int> > icoor(nsimd);
+  std::vector<Coordinate > icoor(nsimd);
       
   for(int lane=0; lane < nsimd; lane++){
     icoor[lane].resize(ndim);
     grid->iCoorFromIindex(icoor[lane],lane);
   }
   
-  parallel_for(uint64_t oidx = 0; oidx < grid->oSites(); oidx++){ //loop over outer index
+  thread_for(oidx, grid->oSites(), {
     //Assemble vector of pointers to output elements
     std::vector<sobj*> ptrs(nsimd);
 
-    std::vector<int> ocoor(ndim);
+    Coordinate ocoor(ndim);
     grid->oCoorFromOindex(ocoor, oidx);
 
-    std::vector<int> lcoor(grid->Nd());
+    Coordinate lcoor(grid->Nd());
       
     for(int lane=0; lane < nsimd; lane++){
 
@@ -742,25 +757,28 @@ vectorizeFromRevLexOrdArray( std::vector<sobj> &in, Lattice<vobj> &out)
     vobj vecobj;
     merge1(vecobj, ptrs, 0);
     out._odata[oidx] = vecobj; 
-  }
+  });
 }
 
 //Convert a Lattice from one precision to another
 template<class VobjOut, class VobjIn>
-void precisionChange(Lattice<VobjOut> &out, const Lattice<VobjIn> &in){
-  assert(out._grid->Nd() == in._grid->Nd());
-  assert(out._grid->FullDimensions() == in._grid->FullDimensions());
-  out.checkerboard = in.checkerboard;
-  GridBase *in_grid=in._grid;
-  GridBase *out_grid = out._grid;
+void precisionChange(Lattice<VobjOut> &out, const Lattice<VobjIn> &in)
+{
+  assert(out.Grid()->Nd() == in.Grid()->Nd());
+  for(int d=0;d<out.Grid()->Nd();d++){
+    assert(out.Grid()->FullDimensions()[d] == in.Grid()->FullDimensions()[d]);
+  }
+  out.Checkerboard() = in.Checkerboard();
+  GridBase *in_grid=in.Grid();
+  GridBase *out_grid = out.Grid();
 
   typedef typename VobjOut::scalar_object SobjOut;
   typedef typename VobjIn::scalar_object SobjIn;
 
-  int ndim = out._grid->Nd();
+  int ndim = out.Grid()->Nd();
   int out_nsimd = out_grid->Nsimd();
     
-  std::vector<std::vector<int> > out_icoor(out_nsimd);
+  std::vector<Coordinate > out_icoor(out_nsimd);
       
   for(int lane=0; lane < out_nsimd; lane++){
     out_icoor[lane].resize(ndim);
@@ -770,13 +788,14 @@ void precisionChange(Lattice<VobjOut> &out, const Lattice<VobjIn> &in){
   std::vector<SobjOut> in_slex_conv(in_grid->lSites());
   unvectorizeToLexOrdArray(in_slex_conv, in);
     
-  parallel_for(uint64_t out_oidx=0;out_oidx<out_grid->oSites();out_oidx++){
-    std::vector<int> out_ocoor(ndim);
+  auto out_v = out.View();
+  thread_for(out_oidx,out_grid->oSites(),{
+    Coordinate out_ocoor(ndim);
     out_grid->oCoorFromOindex(out_ocoor, out_oidx);
 
-    std::vector<SobjOut*> ptrs(out_nsimd);      
+    ExtractPointerArray<SobjOut> ptrs(out_nsimd);      
 
-    std::vector<int> lcoor(out_grid->Nd());
+    Coordinate lcoor(out_grid->Nd());
       
     for(int lane=0; lane < out_nsimd; lane++){
       for(int mu=0;mu<ndim;mu++)
@@ -785,8 +804,8 @@ void precisionChange(Lattice<VobjOut> &out, const Lattice<VobjIn> &in){
       int llex; Lexicographic::IndexFromCoor(lcoor, llex, out_grid->_ldimensions);
       ptrs[lane] = &in_slex_conv[llex];
     }
-    merge(out._odata[out_oidx], ptrs, 0);
-  }
+    merge(out_v[out_oidx], ptrs, 0);
+  });
 }
 
 ////////////////////////////////////////////////////////////////////////////////
@@ -845,8 +864,8 @@ void Grid_split(std::vector<Lattice<Vobj> > & full,Lattice<Vobj>   & split)
 
   assert(full_vecs>=1);
 
-  GridBase * full_grid = full[0]._grid;
-  GridBase *split_grid = split._grid;
+  GridBase * full_grid = full[0].Grid();
+  GridBase *split_grid = split.Grid();
 
   int       ndim  = full_grid->_ndimension;
   int  full_nproc = full_grid->_Nprocessors;
@@ -855,18 +874,18 @@ void Grid_split(std::vector<Lattice<Vobj> > & full,Lattice<Vobj>   & split)
   ////////////////////////////////
   // Checkerboard management
   ////////////////////////////////
-  int cb = full[0].checkerboard;
-  split.checkerboard = cb;
+  int cb = full[0].Checkerboard();
+  split.Checkerboard() = cb;
 
   //////////////////////////////
   // Checks
   //////////////////////////////
   assert(full_grid->_ndimension==split_grid->_ndimension);
   for(int n=0;n<full_vecs;n++){
-    assert(full[n].checkerboard == cb);
+    assert(full[n].Checkerboard() == cb);
     for(int d=0;d<ndim;d++){
-      assert(full[n]._grid->_gdimensions[d]==split._grid->_gdimensions[d]);
-      assert(full[n]._grid->_fdimensions[d]==split._grid->_fdimensions[d]);
+      assert(full[n].Grid()->_gdimensions[d]==split.Grid()->_gdimensions[d]);
+      assert(full[n].Grid()->_fdimensions[d]==split.Grid()->_fdimensions[d]);
     }
   }
 
@@ -874,7 +893,7 @@ void Grid_split(std::vector<Lattice<Vobj> > & full,Lattice<Vobj>   & split)
   assert(nvector*split_nproc==full_nproc);
   assert(nvector == full_vecs);
 
-  std::vector<int> ratio(ndim);
+  Coordinate ratio(ndim);
   for(int d=0;d<ndim;d++){
     ratio[d] = full_grid->_processors[d]/ split_grid->_processors[d];
   }
@@ -887,13 +906,13 @@ void Grid_split(std::vector<Lattice<Vobj> > & full,Lattice<Vobj>   & split)
 
   for(int v=0;v<nvector;v++){
     unvectorizeToLexOrdArray(scalardata,full[v]);    
-    parallel_for(int site=0;site<lsites;site++){
+    thread_for(site,lsites,{
       alldata[v*lsites+site] = scalardata[site];
-    }
+    });
   }
 
   int nvec = nvector; // Counts down to 1 as we collapse dims
-  std::vector<int> ldims = full_grid->_ldimensions;
+  Coordinate ldims = full_grid->_ldimensions;
 
   for(int d=ndim-1;d>=0;d--){
 
@@ -919,8 +938,8 @@ void Grid_split(std::vector<Lattice<Vobj> > & full,Lattice<Vobj>   & split)
       int chunk  = (nvec*fvol)/sP;          assert(chunk*sP == nvec*fvol);
 
       // Loop over reordered data post A2A
-      parallel_for(int c=0;c<chunk;c++){
-	std::vector<int> coor(ndim);
+      thread_for(c, chunk, {
+	Coordinate coor(ndim);
 	for(int m=0;m<M;m++){
 	  for(int s=0;s<sP;s++){
 	    
@@ -942,7 +961,7 @@ void Grid_split(std::vector<Lattice<Vobj> > & full,Lattice<Vobj>   & split)
 
 	  }
 	}
-      }
+      });
       ldims[d]*= ratio[d];
       lsites  *= ratio[d];
 
@@ -954,8 +973,8 @@ void Grid_split(std::vector<Lattice<Vobj> > & full,Lattice<Vobj>   & split)
 template<class Vobj>
 void Grid_split(Lattice<Vobj> &full,Lattice<Vobj>   & split)
 {
-  int nvector = full._grid->_Nprocessors / split._grid->_Nprocessors;
-  std::vector<Lattice<Vobj> > full_v(nvector,full._grid);
+  int nvector = full.Grid()->_Nprocessors / split.Grid()->_Nprocessors;
+  std::vector<Lattice<Vobj> > full_v(nvector,full.Grid());
   for(int n=0;n<nvector;n++){
     full_v[n] = full;
   }
@@ -971,8 +990,8 @@ void Grid_unsplit(std::vector<Lattice<Vobj> > & full,Lattice<Vobj>   & split)
 
   assert(full_vecs>=1);
 
-  GridBase * full_grid = full[0]._grid;
-  GridBase *split_grid = split._grid;
+  GridBase * full_grid = full[0].Grid();
+  GridBase *split_grid = split.Grid();
 
   int       ndim  = full_grid->_ndimension;
   int  full_nproc = full_grid->_Nprocessors;
@@ -981,18 +1000,18 @@ void Grid_unsplit(std::vector<Lattice<Vobj> > & full,Lattice<Vobj>   & split)
   ////////////////////////////////
   // Checkerboard management
   ////////////////////////////////
-  int cb = full[0].checkerboard;
-  split.checkerboard = cb;
+  int cb = full[0].Checkerboard();
+  split.Checkerboard() = cb;
 
   //////////////////////////////
   // Checks
   //////////////////////////////
   assert(full_grid->_ndimension==split_grid->_ndimension);
   for(int n=0;n<full_vecs;n++){
-    assert(full[n].checkerboard == cb);
+    assert(full[n].Checkerboard() == cb);
     for(int d=0;d<ndim;d++){
-      assert(full[n]._grid->_gdimensions[d]==split._grid->_gdimensions[d]);
-      assert(full[n]._grid->_fdimensions[d]==split._grid->_fdimensions[d]);
+      assert(full[n].Grid()->_gdimensions[d]==split.Grid()->_gdimensions[d]);
+      assert(full[n].Grid()->_fdimensions[d]==split.Grid()->_fdimensions[d]);
     }
   }
 
@@ -1000,7 +1019,7 @@ void Grid_unsplit(std::vector<Lattice<Vobj> > & full,Lattice<Vobj>   & split)
   assert(nvector*split_nproc==full_nproc);
   assert(nvector == full_vecs);
 
-  std::vector<int> ratio(ndim);
+  Coordinate ratio(ndim);
   for(int d=0;d<ndim;d++){
     ratio[d] = full_grid->_processors[d]/ split_grid->_processors[d];
   }
@@ -1019,7 +1038,7 @@ void Grid_unsplit(std::vector<Lattice<Vobj> > & full,Lattice<Vobj>   & split)
 
   int nvec = 1;
   uint64_t rsites        = split_grid->lSites();
-  std::vector<int> rdims = split_grid->_ldimensions;
+  Coordinate rdims = split_grid->_ldimensions;
 
   for(int d=0;d<ndim;d++){
 
@@ -1038,8 +1057,8 @@ void Grid_unsplit(std::vector<Lattice<Vobj> > & full,Lattice<Vobj>   & split)
 	
       {
 	// Loop over reordered data post A2A
-	parallel_for(int c=0;c<chunk;c++){
-	  std::vector<int> coor(ndim);
+	thread_for(c, chunk,{
+	  Coordinate coor(ndim);
 	  for(int m=0;m<M;m++){
 	    for(int s=0;s<sP;s++){
 
@@ -1060,7 +1079,7 @@ void Grid_unsplit(std::vector<Lattice<Vobj> > & full,Lattice<Vobj>   & split)
 	      tmpdata[lex_c] = alldata[lex_r];
 	    }
 	  }
-	}
+        });
       }
 
       if ( split_grid->_processors[d] > 1 ) {
@@ -1076,14 +1095,12 @@ void Grid_unsplit(std::vector<Lattice<Vobj> > & full,Lattice<Vobj>   & split)
 
   lsites = full_grid->lSites();
   for(int v=0;v<nvector;v++){
-    //    assert(v<full.size());
-    parallel_for(int site=0;site<lsites;site++){
-      //      assert(v*lsites+site < alldata.size());
+    thread_for(site, lsites,{
       scalardata[site] = alldata[v*lsites+site];
-    }
+    });
     vectorizeFromLexOrdArray(scalardata,full[v]);    
   }
 }
 
-}
-#endif
+NAMESPACE_END(Grid);
+
diff --git a/Grid/lattice/Lattice_transpose.h b/Grid/lattice/Lattice_transpose.h
index 0ae7c6b3..9b0b3483 100644
--- a/Grid/lattice/Lattice_transpose.h
+++ b/Grid/lattice/Lattice_transpose.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,8 +24,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_LATTICE_TRANSPOSE_H
 #define GRID_LATTICE_TRANSPOSE_H
 
@@ -33,31 +33,36 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 // Transpose
 ///////////////////////////////////////////////
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-    ////////////////////////////////////////////////////////////////////////////////////////////////////
-    // Transpose
-    ////////////////////////////////////////////////////////////////////////////////////////////////////
-  template<class vobj>
-    inline Lattice<vobj> transpose(const Lattice<vobj> &lhs){
-    Lattice<vobj> ret(lhs._grid);
-    parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-      ret._odata[ss] = transpose(lhs._odata[ss]);
-    }
-    return ret;
-  };
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Transpose
+////////////////////////////////////////////////////////////////////////////////////////////////////
+template<class vobj>
+inline Lattice<vobj> transpose(const Lattice<vobj> &lhs){
+  Lattice<vobj> ret(lhs.Grid());
+  auto ret_v = ret.View();
+  auto lhs_v = lhs.View();
+  accelerator_for(ss,lhs_v.size(),vobj::Nsimd(),{
+    coalescedWrite(ret_v[ss], transpose(lhs_v(ss)));
+  });
+  return ret;
+};
     
-  ////////////////////////////////////////////////////////////////////////////////////////////////////
-  // Index level dependent transpose
-  ////////////////////////////////////////////////////////////////////////////////////////////////////
-  template<int Index,class vobj>
-    inline auto TransposeIndex(const Lattice<vobj> &lhs) -> Lattice<decltype(transposeIndex<Index>(lhs._odata[0]))>
-  {
-    Lattice<decltype(transposeIndex<Index>(lhs._odata[0]))> ret(lhs._grid);
-    parallel_for(int ss=0;ss<lhs._grid->oSites();ss++){
-      ret._odata[ss] = transposeIndex<Index>(lhs._odata[ss]);
-    }
-    return ret;
-  };
-}
+////////////////////////////////////////////////////////////////////////////////////////////////////
+// Index level dependent transpose
+////////////////////////////////////////////////////////////////////////////////////////////////////
+template<int Index,class vobj>
+inline auto TransposeIndex(const Lattice<vobj> &lhs) -> Lattice<decltype(transposeIndex<Index>(vobj()))>
+{
+  Lattice<decltype(transposeIndex<Index>(vobj()))> ret(lhs.Grid());
+  auto ret_v = ret.View();
+  auto lhs_v = lhs.View();
+  accelerator_for(ss,lhs_v.size(),vobj::Nsimd(),{
+    coalescedWrite(ret_v[ss] , transposeIndex<Index>(lhs_v(ss)));
+  });
+  return ret;
+};
+
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/lattice/Lattice_unary.h b/Grid/lattice/Lattice_unary.h
index 44b7b4f1..591afe72 100644
--- a/Grid/lattice/Lattice_unary.h
+++ b/Grid/lattice/Lattice_unary.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -26,59 +26,55 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_LATTICE_UNARY_H
 #define GRID_LATTICE_UNARY_H
 
-namespace Grid {
-
-  template<class obj> Lattice<obj> pow(const Lattice<obj> &rhs,RealD y){
-    Lattice<obj> ret(rhs._grid);
-    ret.checkerboard = rhs.checkerboard;
-    conformable(ret,rhs);
-    parallel_for(int ss=0;ss<rhs._grid->oSites();ss++){
-      ret._odata[ss]=pow(rhs._odata[ss],y);
-    }
-    return ret;
-  }
-  template<class obj> Lattice<obj> mod(const Lattice<obj> &rhs,Integer y){
-    Lattice<obj> ret(rhs._grid);
-    ret.checkerboard = rhs.checkerboard;
-    conformable(ret,rhs);
-    parallel_for(int ss=0;ss<rhs._grid->oSites();ss++){
-      ret._odata[ss]=mod(rhs._odata[ss],y);
-    }
-    return ret;
-  }
-
-  template<class obj> Lattice<obj> div(const Lattice<obj> &rhs,Integer y){
-    Lattice<obj> ret(rhs._grid);
-    ret.checkerboard = rhs.checkerboard;
-    conformable(ret,rhs);
-    parallel_for(int ss=0;ss<rhs._grid->oSites();ss++){
-      ret._odata[ss]=div(rhs._odata[ss],y);
-    }
-    return ret;
-  }
-
-  template<class obj> Lattice<obj> expMat(const Lattice<obj> &rhs, RealD alpha, Integer Nexp = DEFAULT_MAT_EXP){
-    Lattice<obj> ret(rhs._grid);
-    ret.checkerboard = rhs.checkerboard;
-    conformable(ret,rhs);
-    parallel_for(int ss=0;ss<rhs._grid->oSites();ss++){
-      ret._odata[ss]=Exponentiate(rhs._odata[ss],alpha, Nexp);
-    }
-
-    return ret;
-
-    
-    
-
-    
-  }
-
-
+NAMESPACE_BEGIN(Grid);
 
+template<class obj> Lattice<obj> pow(const Lattice<obj> &rhs_i,RealD y){
+  Lattice<obj> ret_i(rhs_i.Grid());
+  auto rhs = rhs_i.View();
+  auto ret = ret_i.View();
+  ret.Checkerboard() = rhs.Checkerboard();
+  accelerator_for(ss,rhs.size(),1,{
+      ret[ss]=pow(rhs[ss],y);
+  });
+  return ret_i;
 }
+template<class obj> Lattice<obj> mod(const Lattice<obj> &rhs_i,Integer y){
+  Lattice<obj> ret_i(rhs_i.Grid());
+  auto rhs = rhs_i.View();
+  auto ret = ret_i.View();
+  ret.Checkerboard() = rhs.Checkerboard();
+  accelerator_for(ss,rhs.size(),obj::Nsimd(),{
+    coalescedWrite(ret[ss],mod(rhs(ss),y));
+  });
+  return ret_i;
+}
+
+template<class obj> Lattice<obj> div(const Lattice<obj> &rhs_i,Integer y){
+  Lattice<obj> ret_i(rhs_i.Grid());
+  auto ret = ret_i.View();
+  auto rhs = rhs_i.View();
+  ret.Checkerboard() = rhs_i.Checkerboard();
+  accelerator_for(ss,rhs.size(),obj::Nsimd(),{
+    coalescedWrite(ret[ss],div(rhs(ss),y));
+  });
+  return ret_i;
+}
+
+template<class obj> Lattice<obj> expMat(const Lattice<obj> &rhs_i, RealD alpha, Integer Nexp = DEFAULT_MAT_EXP){
+  Lattice<obj> ret_i(rhs_i.Grid());
+  auto rhs = rhs_i.View();
+  auto ret = ret_i.View();
+  ret.Checkerboard() = rhs.Checkerboard();
+  accelerator_for(ss,rhs.size(),obj::Nsimd(),{
+    coalescedWrite(ret[ss],Exponentiate(rhs(ss),alpha, Nexp));
+  });
+  return ret_i;
+}
+
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/log/Log.cc b/Grid/log/Log.cc
index cb4a8521..9302b4cc 100644
--- a/Grid/log/Log.cc
+++ b/Grid/log/Log.cc
@@ -28,27 +28,27 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #include <Grid/GridCore.h>
 #include <Grid/util/CompilerCompatible.h>
 
 #include <cxxabi.h>
 #include <memory>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  std::string demangle(const char* name) {
+std::string demangle(const char* name) {
     
-    int status = -4; // some arbitrary value to eliminate the compiler warning
+  int status = -4; // some arbitrary value to eliminate the compiler warning
     
-    // enable c++11 by passing the flag -std=c++11 to g++
-    std::unique_ptr<char, void(*)(void*)> res {
-      abi::__cxa_demangle(name, NULL, NULL, &status),
-	std::free
-	};
+  // enable c++11 by passing the flag -std=c++11 to g++
+  std::unique_ptr<char, void(*)(void*)> res {
+    abi::__cxa_demangle(name, NULL, NULL, &status),
+      std::free
+      };
     
-    return (status==0) ? res.get() : name ;
-  }
+  return (status==0) ? res.get() : name ;
+}
   
 GridStopWatch Logger::GlobalStopWatch;
 int Logger::timestamp;
@@ -109,8 +109,9 @@ void Grid_quiesce_nodes(void) {
 }
 
 void Grid_unquiesce_nodes(void) {
-#ifdef GRID_COMMS_MPI
+#if defined(GRID_COMMS_MPI) || defined(GRID_COMMS_MPI3) || defined(GRID_COMMS_MPIT)
   std::cout.clear();
 #endif
 }
-}
+NAMESPACE_END(Grid);
+
diff --git a/Grid/log/Log.h b/Grid/log/Log.h
index 322ab32e..d459a4a9 100644
--- a/Grid/log/Log.h
+++ b/Grid/log/Log.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,8 +25,8 @@
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 
 #include <map>
 
@@ -37,13 +37,12 @@
 #include <execinfo.h>
 #endif
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 //////////////////////////////////////////////////////////////////////////////////////////////////
 // Dress the output; use std::chrono for time stamping via the StopWatch class
 //////////////////////////////////////////////////////////////////////////////////////////////////
 
-
 class Colours{
 protected:
   bool is_active;
@@ -57,15 +56,15 @@ public:
   void Active(bool activate){
     is_active=activate;
     if (is_active){
-     colour["BLACK"]  ="\033[30m";
-     colour["RED"]    ="\033[31m";
-     colour["GREEN"]  ="\033[32m";
-     colour["YELLOW"] ="\033[33m";
-     colour["BLUE"]   ="\033[34m";
-     colour["PURPLE"] ="\033[35m";
-     colour["CYAN"]   ="\033[36m";
-     colour["WHITE"]  ="\033[37m";
-     colour["NORMAL"] ="\033[0;39m";
+      colour["BLACK"]  ="\033[30m";
+      colour["RED"]    ="\033[31m";
+      colour["GREEN"]  ="\033[32m";
+      colour["YELLOW"] ="\033[33m";
+      colour["BLUE"]   ="\033[34m";
+      colour["PURPLE"] ="\033[35m";
+      colour["CYAN"]   ="\033[36m";
+      colour["WHITE"]  ="\033[37m";
+      colour["NORMAL"] ="\033[0;39m";
     } else {
       colour["BLACK"] ="";
       colour["RED"]   ="";
@@ -102,14 +101,14 @@ public:
   std::string colour() {return Painter.colour[COLOUR];}
 
   Logger(std::string topNm, int on, std::string nm, Colours& col_class, std::string col)  : active(on),
-    name(nm),
-    topName(topNm),
-    Painter(col_class),
-    timing_mode(0),
-    COLOUR(col) 
-    {
-      StopWatch = & GlobalStopWatch;
-    };
+											    name(nm),
+											    topName(topNm),
+											    Painter(col_class),
+											    timing_mode(0),
+											    COLOUR(col) 
+  {
+    StopWatch = & GlobalStopWatch;
+  };
   
   void Active(int on) {active = on;};
   int  isActive(void) {return active;};
@@ -164,7 +163,7 @@ public:
 class GridLogger: public Logger {
 public:
   GridLogger(int on, std::string nm, Colours&col_class, std::string col_key = "NORMAL"):
-  Logger("Grid", on, nm, col_class, col_key){};
+    Logger("Grid", on, nm, col_class, col_key){};
 };
 
 void GridLogConfigure(std::vector<std::string> &logstreams);
@@ -181,39 +180,39 @@ extern GridLogger GridLogIterative  ;
 extern GridLogger GridLogIntegrator  ;
 extern Colours    GridLogColours;
 
- std::string demangle(const char* name) ;
+std::string demangle(const char* name) ;
 
 #define _NBACKTRACE (256)
 extern void * Grid_backtrace_buffer[_NBACKTRACE];
 
-#define BACKTRACEFILE() {\
-char string[20];					\
-std::sprintf(string,"backtrace.%d",CartesianCommunicator::RankWorld()); \
-std::FILE * fp = std::fopen(string,"w");				\
-BACKTRACEFP(fp)\
-std::fclose(fp);	    \
-}
+#define BACKTRACEFILE() {						\
+    char string[20];							\
+    std::sprintf(string,"backtrace.%d",CartesianCommunicator::RankWorld()); \
+    std::FILE * fp = std::fopen(string,"w");				\
+    BACKTRACEFP(fp)							\
+      std::fclose(fp);							\
+  }
 
 
 #ifdef HAVE_EXECINFO_H
-#define BACKTRACEFP(fp) { \
-int symbols    = backtrace        (Grid_backtrace_buffer,_NBACKTRACE);\
-char **strings = backtrace_symbols(Grid_backtrace_buffer,symbols);\
-for (int i = 0; i < symbols; i++){\
-  std::fprintf (fp,"BackTrace Strings: %d %s\n",i, demangle(strings[i]).c_str()); std::fflush(fp); \
-}\
-}
+#define BACKTRACEFP(fp) {						\
+    int symbols    = backtrace        (Grid_backtrace_buffer,_NBACKTRACE); \
+    char **strings = backtrace_symbols(Grid_backtrace_buffer,symbols);	\
+    for (int i = 0; i < symbols; i++){					\
+      std::fprintf (fp,"BackTrace Strings: %d %s\n",i, demangle(strings[i]).c_str()); std::fflush(fp); \
+    }									\
+  }
 #else 
-#define BACKTRACEFP(fp) { \
-std::fprintf (fp,"BT %d %lx\n",0, __builtin_return_address(0)); std::fflush(fp); \
-std::fprintf (fp,"BT %d %lx\n",1, __builtin_return_address(1)); std::fflush(fp); \
-std::fprintf (fp,"BT %d %lx\n",2, __builtin_return_address(2)); std::fflush(fp); \
-std::fprintf (fp,"BT %d %lx\n",3, __builtin_return_address(3)); std::fflush(fp); \
-}
+#define BACKTRACEFP(fp) {						\
+    std::fprintf (fp,"BT %d %lx\n",0, __builtin_return_address(0)); std::fflush(fp); \
+    std::fprintf (fp,"BT %d %lx\n",1, __builtin_return_address(1)); std::fflush(fp); \
+    std::fprintf (fp,"BT %d %lx\n",2, __builtin_return_address(2)); std::fflush(fp); \
+    std::fprintf (fp,"BT %d %lx\n",3, __builtin_return_address(3)); std::fflush(fp); \
+  }
 #endif
 
 #define BACKTRACE() BACKTRACEFP(stdout) 
 
+NAMESPACE_END(Grid);
 
-}
 #endif
diff --git a/Grid/parallelIO/BinaryIO.h b/Grid/parallelIO/BinaryIO.h
index 144ff29f..f90c34a9 100644
--- a/Grid/parallelIO/BinaryIO.h
+++ b/Grid/parallelIO/BinaryIO.h
@@ -26,8 +26,7 @@
     See the full license in the file "LICENSE" in the top level distribution directory
     *************************************************************************************/
     /*  END LEGAL */
-#ifndef GRID_BINARY_IO_H
-#define GRID_BINARY_IO_H
+#pragma once
 
 #if defined(GRID_COMMS_MPI) || defined(GRID_COMMS_MPI3) || defined(GRID_COMMS_MPIT) 
 #define USE_MPI_IO
@@ -42,8 +41,7 @@
 #include <arpa/inet.h>
 #include <algorithm>
 
-namespace Grid { 
-
+NAMESPACE_BEGIN(Grid);
 
 /////////////////////////////////////////////////////////////////////////////////
 // Byte reversal garbage
@@ -91,7 +89,7 @@ class BinaryIO {
   {
     typedef typename vobj::scalar_object sobj;
 
-    GridBase *grid = lat._grid;
+    GridBase *grid = lat.Grid();
     uint64_t lsites = grid->lSites();
 
     std::vector<sobj> scalardata(lsites); 
@@ -111,21 +109,20 @@ class BinaryIO {
       lsites = 1;
     }
 
-PARALLEL_REGION
+    thread_region
     {
       uint32_t nersc_csum_thr = 0;
 
-PARALLEL_FOR_LOOP_INTERN
-      for (uint64_t local_site = 0; local_site < lsites; local_site++)
+      thread_for_in_region( local_site, lsites, 
       {
         uint32_t *site_buf = (uint32_t *)&fbuf[local_site];
         for (uint64_t j = 0; j < size32; j++)
         {
           nersc_csum_thr = nersc_csum_thr + site_buf[j];
         }
-      }
+      });
 
-PARALLEL_CRITICAL
+      thread_critical
       {
         nersc_csum += nersc_csum_thr;
       }
@@ -134,28 +131,25 @@ PARALLEL_CRITICAL
 
   template<class fobj> static inline void ScidacChecksum(GridBase *grid,std::vector<fobj> &fbuf,uint32_t &scidac_csuma,uint32_t &scidac_csumb)
   {
-    const uint64_t size32 = sizeof(fobj)/sizeof(uint32_t);
-
-
     int nd = grid->_ndimension;
 
     uint64_t lsites              =grid->lSites();
     if (fbuf.size()==1) {
       lsites=1;
     }
-    std::vector<int> local_vol   =grid->LocalDimensions();
-    std::vector<int> local_start =grid->LocalStarts();
-    std::vector<int> global_vol  =grid->FullDimensions();
+    Coordinate local_vol   =grid->LocalDimensions();
+    Coordinate local_start =grid->LocalStarts();
+    Coordinate global_vol  =grid->FullDimensions();
 
-PARALLEL_REGION
+    thread_region
     { 
-      std::vector<int> coor(nd);
+      Coordinate coor(nd);
       uint32_t scidac_csuma_thr=0;
       uint32_t scidac_csumb_thr=0;
       uint32_t site_crc=0;
 
-PARALLEL_FOR_LOOP_INTERN
-      for(uint64_t local_site=0;local_site<lsites;local_site++){
+      thread_for_in_region( local_site, lsites, 
+      {
 
 	uint32_t * site_buf = (uint32_t *)&fbuf[local_site];
 
@@ -182,9 +176,9 @@ PARALLEL_FOR_LOOP_INTERN
 	//	std::cout << "Site "<<local_site << std::hex<<site_buf[0] <<site_buf[1]<<std::dec <<std::endl;
 	scidac_csuma_thr ^= site_crc<<gsite29 | site_crc>>(32-gsite29);
 	scidac_csumb_thr ^= site_crc<<gsite31 | site_crc>>(32-gsite31);
-      }
+      });
 
-PARALLEL_CRITICAL
+      thread_critical
       {
 	scidac_csuma^= scidac_csuma_thr;
 	scidac_csumb^= scidac_csumb_thr;
@@ -202,9 +196,9 @@ PARALLEL_CRITICAL
   {
     uint32_t * f = (uint32_t *)file_object;
     uint64_t count = bytes/sizeof(uint32_t);
-    parallel_for(uint64_t i=0;i<count;i++){  
+    thread_for( i, count, {  
       f[i] = ntohl(f[i]);
-    }
+    });
   }
   // LE must Swap and switch to host
   static inline void le32toh_v(void *file_object,uint64_t bytes)
@@ -212,13 +206,13 @@ PARALLEL_CRITICAL
     uint32_t *fp = (uint32_t *)file_object;
 
     uint64_t count = bytes/sizeof(uint32_t);
-    parallel_for(uint64_t i=0;i<count;i++){  
+    thread_for(i,count,{
       uint32_t f;
       f = fp[i];
       // got network order and the network to host
       f = ((f&0xFF)<<24) | ((f&0xFF00)<<8) | ((f&0xFF0000)>>8) | ((f&0xFF000000UL)>>24) ; 
       fp[i] = ntohl(f);
-    }
+    });
   }
 
   // BE is same as network
@@ -226,9 +220,9 @@ PARALLEL_CRITICAL
   {
     uint64_t * f = (uint64_t *)file_object;
     uint64_t count = bytes/sizeof(uint64_t);
-    parallel_for(uint64_t i=0;i<count;i++){  
+    thread_for( i, count, {
       f[i] = Grid_ntohll(f[i]);
-    }
+    });
   }
   
   // LE must swap and switch;
@@ -236,7 +230,7 @@ PARALLEL_CRITICAL
   {
     uint64_t *fp = (uint64_t *)file_object;
     uint64_t count = bytes/sizeof(uint64_t);
-    parallel_for(uint64_t i=0;i<count;i++){  
+    thread_for( i, count, {
       uint64_t f,g;
       f = fp[i];
       // got network order and the network to host
@@ -245,7 +239,7 @@ PARALLEL_CRITICAL
       f = f >> 32;
       g|= ((f&0xFF)<<24) | ((f&0xFF00)<<8) | ((f&0xFF0000)>>8) | ((f&0xFF000000UL)>>24) ; 
       fp[i] = Grid_ntohll(g);
-    }
+    });
   }
   /////////////////////////////////////////////////////////////////////////////
   // Real action:
@@ -281,13 +275,13 @@ PARALLEL_CRITICAL
     int nrank                = grid->ProcessorCount();
     int myrank               = grid->ThisRank();
 
-    std::vector<int>  psizes = grid->ProcessorGrid(); 
-    std::vector<int>  pcoor  = grid->ThisProcessorCoor();
-    std::vector<int> gLattice= grid->GlobalDimensions();
-    std::vector<int> lLattice= grid->LocalDimensions();
+    Coordinate  psizes = grid->ProcessorGrid(); 
+    Coordinate  pcoor  = grid->ThisProcessorCoor();
+    Coordinate gLattice= grid->GlobalDimensions();
+    Coordinate lLattice= grid->LocalDimensions();
 
-    std::vector<int> lStart(ndim);
-    std::vector<int> gStart(ndim);
+    Coordinate lStart(ndim);
+    Coordinate gStart(ndim);
 
     // Flatten the file
     uint64_t lsites = grid->lSites();
@@ -546,7 +540,7 @@ PARALLEL_CRITICAL
     typedef typename vobj::scalar_object sobj;
     typedef typename vobj::Realified::scalar_type word;    word w=0;
 
-    GridBase *grid = Umu._grid;
+    GridBase *grid = Umu.Grid();
     uint64_t lsites = grid->lSites();
 
     std::vector<sobj> scalardata(lsites); 
@@ -558,7 +552,7 @@ PARALLEL_CRITICAL
     GridStopWatch timer; 
     timer.Start();
 
-    parallel_for(uint64_t x=0;x<lsites;x++) munge(iodata[x], scalardata[x]);
+    thread_for(x,lsites, { munge(iodata[x], scalardata[x]); });
 
     vectorizeFromLexOrdArray(scalardata,Umu);    
     grid->Barrier();
@@ -582,7 +576,7 @@ PARALLEL_CRITICAL
   {
     typedef typename vobj::scalar_object sobj;
     typedef typename vobj::Realified::scalar_type word;    word w=0;
-    GridBase *grid = Umu._grid;
+    GridBase *grid = Umu.Grid();
     uint64_t lsites = grid->lSites(), offsetCopy = offset;
     int attemptsLeft = std::max(0, BinaryIO::latticeWriteMaxRetry);
     bool checkWrite = (BinaryIO::latticeWriteMaxRetry >= 0);
@@ -596,7 +590,7 @@ PARALLEL_CRITICAL
     GridStopWatch timer; timer.Start();
     unvectorizeToLexOrdArray(scalardata,Umu);    
 
-    parallel_for(uint64_t x=0;x<lsites;x++) munge(scalardata[x],iodata[x]);
+    thread_for(x, lsites, { munge(scalardata[x],iodata[x]); });
 
     grid->Barrier();
     timer.Stop();
@@ -619,7 +613,7 @@ PARALLEL_CRITICAL
         {
           std::cout << GridLogMessage << "writeLatticeObject: read test checksum failure, re-writing (" << attemptsLeft << " attempt(s) remaining)" << std::endl;
           offset = offsetCopy;
-          parallel_for(uint64_t x=0;x<lsites;x++) munge(scalardata[x],iodata[x]);
+          thread_for(x,lsites, { munge(scalardata[x],iodata[x]); });
         }
         else
         {
@@ -637,8 +631,8 @@ PARALLEL_CRITICAL
   /////////////////////////////////////////////////////////////////////////////
   // Read a RNG;  use IOobject and lexico map to an array of state 
   //////////////////////////////////////////////////////////////////////////////////////
-  static inline void readRNG(GridSerialRNG &serial,
-			     GridParallelRNG &parallel,
+  static inline void readRNG(GridSerialRNG &serial_rng,
+			     GridParallelRNG &parallel_rng,
 			     std::string file,
 			     uint64_t offset,
 			     uint32_t &nersc_csum,
@@ -652,7 +646,7 @@ PARALLEL_CRITICAL
 
     std::string format = "IEEE32BIG";
 
-    GridBase *grid = parallel._grid;
+    GridBase *grid = parallel_rng.Grid();
     uint64_t gsites = grid->gSites();
     uint64_t lsites = grid->lSites();
 
@@ -669,11 +663,11 @@ PARALLEL_CRITICAL
 	     nersc_csum,scidac_csuma,scidac_csumb);
 
     timer.Start();
-    parallel_for(uint64_t lidx=0;lidx<lsites;lidx++){
+    thread_for(lidx,lsites,{
       std::vector<RngStateType> tmp(RngStateCount);
       std::copy(iodata[lidx].begin(),iodata[lidx].end(),tmp.begin());
-      parallel.SetState(tmp,lidx);
-    }
+      parallel_rng.SetState(tmp,lidx);
+      });
     timer.Stop();
 
     iodata.resize(1);
@@ -683,7 +677,7 @@ PARALLEL_CRITICAL
     {
       std::vector<RngStateType> tmp(RngStateCount);
       std::copy(iodata[0].begin(),iodata[0].end(),tmp.begin());
-      serial.SetState(tmp,0);
+      serial_rng.SetState(tmp,0);
     }
 
     nersc_csum   = nersc_csum   + nersc_csum_tmp;
@@ -699,8 +693,8 @@ PARALLEL_CRITICAL
   /////////////////////////////////////////////////////////////////////////////
   // Write a RNG; lexico map to an array of state and use IOobject
   //////////////////////////////////////////////////////////////////////////////////////
-  static inline void writeRNG(GridSerialRNG &serial,
-			      GridParallelRNG &parallel,
+  static inline void writeRNG(GridSerialRNG &serial_rng,
+			      GridParallelRNG &parallel_rng,
 			      std::string file,
 			      uint64_t offset,
 			      uint32_t &nersc_csum,
@@ -712,7 +706,7 @@ PARALLEL_CRITICAL
     const int RngStateCount = GridSerialRNG::RngStateCount;
     typedef std::array<RngStateType,RngStateCount> RNGstate;
 
-    GridBase *grid = parallel._grid;
+    GridBase *grid = parallel_rng.Grid();
     uint64_t gsites = grid->gSites();
     uint64_t lsites = grid->lSites();
 
@@ -727,11 +721,11 @@ PARALLEL_CRITICAL
 
     timer.Start();
     std::vector<RNGstate> iodata(lsites);
-    parallel_for(uint64_t lidx=0;lidx<lsites;lidx++){
+    thread_for(lidx,lsites,{
       std::vector<RngStateType> tmp(RngStateCount);
-      parallel.GetState(tmp,lidx);
+      parallel_rng.GetState(tmp,lidx);
       std::copy(tmp.begin(),tmp.end(),iodata[lidx].begin());
-    }
+    });
     timer.Stop();
 
     IOobject(w,grid,iodata,file,offset,format,BINARYIO_WRITE|BINARYIO_LEXICOGRAPHIC,
@@ -739,7 +733,7 @@ PARALLEL_CRITICAL
     iodata.resize(1);
     {
       std::vector<RngStateType> tmp(RngStateCount);
-      serial.GetState(tmp,0);
+      serial_rng.GetState(tmp,0);
       std::copy(tmp.begin(),tmp.end(),iodata[0].begin());
     }
     IOobject(w,grid,iodata,file,offset,format,BINARYIO_WRITE|BINARYIO_MASTER_APPEND,
@@ -756,5 +750,4 @@ PARALLEL_CRITICAL
   }
 };
 
-}
-#endif
+NAMESPACE_END(Grid);
diff --git a/Grid/parallelIO/IldgIO.h b/Grid/parallelIO/IldgIO.h
index db07cac0..b564371b 100644
--- a/Grid/parallelIO/IldgIO.h
+++ b/Grid/parallelIO/IldgIO.h
@@ -24,8 +24,7 @@ See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
 /*  END LEGAL */
-#ifndef GRID_ILDG_IO_H
-#define GRID_ILDG_IO_H
+#pragma once
 
 #ifdef HAVE_LIME
 #include <algorithm>
@@ -43,8 +42,7 @@ extern "C" {
 #include "lime.h"
 }
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 #define GRID_FIELD_NORM "FieldNormMetaData"
 #define GRID_FIELD_NORM_CALC(FieldNormMetaData_, n2ck) \
@@ -140,7 +138,7 @@ assert(GRID_FIELD_NORM_CALC(FieldNormMetaData_, n2ck) < 1.0e-5);
    /////////////////////////////////////
    // Scidac Private File structure
    /////////////////////////////////////
-   _scidacFile              = scidacFile(field._grid);
+   _scidacFile              = scidacFile(field.Grid());
 
    /////////////////////////////////////
    // Scidac Private Record structure
@@ -227,10 +225,10 @@ class GridLimeReader : public BinaryIO {
 
 	//	std::cout << GridLogMessage<< " readLimeLatticeBinaryObject matches ! " <<std::endl;
 
-	uint64_t PayloadSize = sizeof(sobj) * field._grid->_gsites;
+	uint64_t PayloadSize = sizeof(sobj) * field.Grid()->_gsites;
 
 	//	std::cout << "R sizeof(sobj)= " <<sizeof(sobj)<<std::endl;
-	//	std::cout << "R Gsites " <<field._grid->_gsites<<std::endl;
+	//	std::cout << "R Gsites " <<field.Grid()->_gsites<<std::endl;
 	//	std::cout << "R Payload expected " <<PayloadSize<<std::endl;
 	//	std::cout << "R file size " <<file_bytes <<std::endl;
 
@@ -406,7 +404,7 @@ class GridLimeWriter : public BinaryIO
   ////////////////////////////////////////////////////
   // Write a generic lattice field and csum
   // This routine is Collectively called by all nodes
-  // in communicator used by the field._grid
+  // in communicator used by the field.Grid()
   ////////////////////////////////////////////////////
   template<class vobj>
   void writeLimeLatticeBinaryObject(Lattice<vobj> &field,std::string record_name)
@@ -425,8 +423,8 @@ class GridLimeWriter : public BinaryIO
     //  v) Continue writing scidac record.
     ////////////////////////////////////////////////////////////////////
     
-    GridBase *grid = field._grid;
-    assert(boss_node == field._grid->IsBoss() );
+    GridBase *grid = field.Grid();
+    assert(boss_node == field.Grid()->IsBoss() );
 
     FieldNormMetaData FNMD; FNMD.norm2 = norm2(field);
 
@@ -443,7 +441,7 @@ class GridLimeWriter : public BinaryIO
     }
     
     //    std::cout << "W sizeof(sobj)"      <<sizeof(sobj)<<std::endl;
-    //    std::cout << "W Gsites "           <<field._grid->_gsites<<std::endl;
+    //    std::cout << "W Gsites "           <<field.Grid()->_gsites<<std::endl;
     //    std::cout << "W Payload expected " <<PayloadSize<<std::endl;
 
     ////////////////////////////////////////////////
@@ -515,7 +513,7 @@ class ScidacWriter : public GridLimeWriter {
   void writeScidacFieldRecord(Lattice<vobj> &field,userRecord _userRecord,
                               const unsigned int recordScientificPrec = 0) 
   {
-    GridBase * grid = field._grid;
+    GridBase * grid = field.Grid();
 
     ////////////////////////////////////////
     // fill the Grid header
@@ -557,7 +555,7 @@ class ScidacReader : public GridLimeReader {
   void readScidacFieldRecord(Lattice<vobj> &field,userRecord &_userRecord) 
   {
     typedef typename vobj::scalar_object sobj;
-    GridBase * grid = field._grid;
+    GridBase * grid = field.Grid();
 
     ////////////////////////////////////////
     // fill the Grid header
@@ -624,7 +622,7 @@ class IldgWriter : public ScidacWriter {
   template <class vsimd>
   void writeConfiguration(Lattice<iLorentzColourMatrix<vsimd> > &Umu,int sequence,std::string LFN,std::string description) 
   {
-    GridBase * grid = Umu._grid;
+    GridBase * grid = Umu.Grid();
     typedef Lattice<iLorentzColourMatrix<vsimd> > GaugeField;
     typedef iLorentzColourMatrix<vsimd> vobj;
     typedef typename vobj::scalar_object sobj;
@@ -717,9 +715,9 @@ class IldgReader : public GridLimeReader {
     typedef LorentzColourMatrixF fobj;
     typedef LorentzColourMatrixD dobj;
 
-    GridBase *grid = Umu._grid;
+    GridBase *grid = Umu.Grid();
 
-    std::vector<int> dims = Umu._grid->FullDimensions();
+    Coordinate dims = Umu.Grid()->FullDimensions();
 
     assert(dims.size()==4);
 
@@ -853,6 +851,7 @@ class IldgReader : public GridLimeReader {
     // Minimally must find binary segment and checksum
     // Since this is an ILDG reader require ILDG format
     //////////////////////////////////////////////////////
+    assert(found_ildgLFN);
     assert(found_ildgBinary);
     assert(found_ildgFormat);
     assert(found_scidacChecksum);
@@ -930,9 +929,9 @@ class IldgReader : public GridLimeReader {
   }
  };
 
-}}
+NAMESPACE_END(Grid);
+
 
 //HAVE_LIME
 #endif
 
-#endif
diff --git a/Grid/parallelIO/IldgIOtypes.h b/Grid/parallelIO/IldgIOtypes.h
index 281b20f2..ddc0969c 100644
--- a/Grid/parallelIO/IldgIOtypes.h
+++ b/Grid/parallelIO/IldgIOtypes.h
@@ -23,7 +23,7 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef GRID_ILDGTYPES_IO_H
 #define GRID_ILDGTYPES_IO_H
 
@@ -32,7 +32,7 @@ extern "C" { // for linkage
 #include "lime.h"
 }
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 /////////////////////////////////////////////////////////////////////////////////
 // Data representation of records that enter ILDG and SciDac formats
@@ -51,12 +51,12 @@ namespace Grid {
 // Unused SCIDAC records names; could move to support this functionality
 #define SCIDAC_SITELIST           "scidac-sitelist"
 
-  ////////////////////////////////////////////////////////////
-  const int GRID_IO_SINGLEFILE = 0; // hardcode lift from QIO compat
-  const int GRID_IO_MULTIFILE  = 1; // hardcode lift from QIO compat
-  const int GRID_IO_FIELD      = 0; // hardcode lift from QIO compat
-  const int GRID_IO_GLOBAL     = 1; // hardcode lift from QIO compat
-  ////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////
+const int GRID_IO_SINGLEFILE = 0; // hardcode lift from QIO compat
+const int GRID_IO_MULTIFILE  = 1; // hardcode lift from QIO compat
+const int GRID_IO_FIELD      = 0; // hardcode lift from QIO compat
+const int GRID_IO_GLOBAL     = 1; // hardcode lift from QIO compat
+////////////////////////////////////////////////////////////
 
 /////////////////////////////////////////////////////////////////////////////////
 // QIO uses mandatory "private" records fixed format
@@ -74,7 +74,7 @@ struct emptyUserRecord : Serializable {
 // <?xml version="1.0" encoding="UTF-8"?><scidacFile><version>1.1</version><spacetime>4</spacetime><dims>16 16 16 32 </dims><volfmt>0</volfmt></scidacFile>
 ////////////////////////
 struct scidacFile : Serializable {
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(scidacFile,
                                   double, version,
                                   int, spacetime,
@@ -91,7 +91,7 @@ struct scidacFile : Serializable {
     return dimensions;
   }
 
-  void setDimensions(std::vector<int> dimensions) { 
+  void setDimensions(Coordinate dimensions) { 
     char delimiter = ' ';
     std::stringstream stream;
     for(int i=0;i<dimensions.size();i++){ 
@@ -124,7 +124,7 @@ struct scidacFile : Serializable {
 ///////////////////////////////////////////////////////////////////////
 
 struct scidacRecord : Serializable {
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(scidacRecord,
                                   double, version,
                                   std::string, date,
@@ -160,7 +160,7 @@ public:
 // USQCD info
 ////////////////////////
 struct usqcdInfo : Serializable { 
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(usqcdInfo,
 				  double, version,
 				  double, plaq,
@@ -174,7 +174,7 @@ struct usqcdInfo : Serializable {
 // Scidac Checksum
 ////////////////////////
 struct scidacChecksum : Serializable { 
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(scidacChecksum,
 				  double, version,
 				  std::string, suma,
@@ -201,7 +201,7 @@ struct scidacChecksum : Serializable {
 // From http://www.physics.utah.edu/~detar/scidac/qio_2p3.pdf
 ////////////////////////////////////////////////////////////////////////////////////////
 struct usqcdPropFile : Serializable { 
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(usqcdPropFile,
 				  double, version,
 				  std::string, type,
@@ -211,7 +211,7 @@ struct usqcdPropFile : Serializable {
   };
 };
 struct usqcdSourceInfo : Serializable { 
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(usqcdSourceInfo,
 				  double, version,
 				  std::string, info);
@@ -220,7 +220,7 @@ struct usqcdSourceInfo : Serializable {
   };
 };
 struct usqcdPropInfo : Serializable { 
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(usqcdPropInfo,
 				  double, version,
 				  int, spin,
@@ -232,6 +232,6 @@ struct usqcdPropInfo : Serializable {
 };
 #endif
 
-}
+NAMESPACE_END(Grid);
 #endif
 #endif
diff --git a/Grid/parallelIO/MetaData.h b/Grid/parallelIO/MetaData.h
index 4cc93d03..2e211838 100644
--- a/Grid/parallelIO/MetaData.h
+++ b/Grid/parallelIO/MetaData.h
@@ -36,23 +36,24 @@
 #include <sys/utsname.h>
 #include <pwd.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  ///////////////////////////////////////////////////////
-  // Precision mapping
-  ///////////////////////////////////////////////////////
-  template<class vobj> static std::string getFormatString (void)
-  {
-    std::string format;
-    typedef typename getPrecision<vobj>::real_scalar_type stype;
-    if ( sizeof(stype) == sizeof(float) ) {
-      format = std::string("IEEE32BIG");
-    }
-    if ( sizeof(stype) == sizeof(double) ) {
-      format = std::string("IEEE64BIG");
-    }
-    return format;
+///////////////////////////////////////////////////////
+// Precision mapping
+///////////////////////////////////////////////////////
+template<class vobj> static std::string getFormatString (void)
+{
+  std::string format;
+  typedef typename getPrecision<vobj>::real_scalar_type stype;
+  if ( sizeof(stype) == sizeof(float) ) {
+    format = std::string("IEEE32BIG");
   }
+  if ( sizeof(stype) == sizeof(double) ) {
+    format = std::string("IEEE64BIG");
+  }
+  return format;
+};
+
   ////////////////////////////////////////////////////////////////////////////////
   // header specification/interpretation
   ////////////////////////////////////////////////////////////////////////////////
@@ -93,146 +94,145 @@ namespace Grid {
       link_trace(0.), plaquette(0.), checksum(0),
       scidac_checksuma(0), scidac_checksumb(0), sequence_number(0)
       {}
-    };
+  };
 
-  namespace QCD {
+// PB disable using namespace - this is a header and forces namesapce visibility for all 
+// including files
+//using namespace Grid;
 
-    using namespace Grid;
+//////////////////////////////////////////////////////////////////////
+// Bit and Physical Checksumming and QA of data
+//////////////////////////////////////////////////////////////////////
+inline void GridMetaData(GridBase *grid,FieldMetaData &header)
+{
+  int nd = grid->_ndimension;
+  header.nd = nd;
+  header.dimension.resize(nd);
+  header.boundary.resize(nd);
+  header.data_start = 0;
+  for(int d=0;d<nd;d++) {
+    header.dimension[d] = grid->_fdimensions[d];
+  }
+  for(int d=0;d<nd;d++) {
+    header.boundary[d] = std::string("PERIODIC");
+  }
+}
 
+inline void MachineCharacteristics(FieldMetaData &header)
+{
+  // Who
+  struct passwd *pw = getpwuid (getuid());
+  if (pw) header.creator = std::string(pw->pw_name); 
 
-    //////////////////////////////////////////////////////////////////////
-    // Bit and Physical Checksumming and QA of data
-    //////////////////////////////////////////////////////////////////////
-    inline void GridMetaData(GridBase *grid,FieldMetaData &header)
-    {
-      int nd = grid->_ndimension;
-      header.nd = nd;
-      header.dimension.resize(nd);
-      header.boundary.resize(nd);
-      header.data_start = 0;
-      for(int d=0;d<nd;d++) {
-	header.dimension[d] = grid->_fdimensions[d];
-      }
-      for(int d=0;d<nd;d++) {
-	header.boundary[d] = std::string("PERIODIC");
-      }
-    }
+  // When
+  std::time_t t = std::time(nullptr);
+  std::tm tm_ = *std::localtime(&t);
+  std::ostringstream oss; 
+  //      oss << std::put_time(&tm_, "%c %Z");
+  header.creation_date = oss.str();
+  header.archive_date  = header.creation_date;
 
-    inline void MachineCharacteristics(FieldMetaData &header)
-    {
-      // Who
-      struct passwd *pw = getpwuid (getuid());
-      if (pw) header.creator = std::string(pw->pw_name); 
-
-      // When
-      std::time_t t = std::time(nullptr);
-      std::tm tm_ = *std::localtime(&t);
-      std::ostringstream oss; 
-      //      oss << std::put_time(&tm_, "%c %Z");
-      header.creation_date = oss.str();
-      header.archive_date  = header.creation_date;
-
-      // What
-      struct utsname name;  uname(&name);
-      header.creator_hardware = std::string(name.nodename)+"-";
-      header.creator_hardware+= std::string(name.machine)+"-";
-      header.creator_hardware+= std::string(name.sysname)+"-";
-      header.creator_hardware+= std::string(name.release);
-    }
+  // What
+  struct utsname name;  uname(&name);
+  header.creator_hardware = std::string(name.nodename)+"-";
+  header.creator_hardware+= std::string(name.machine)+"-";
+  header.creator_hardware+= std::string(name.sysname)+"-";
+  header.creator_hardware+= std::string(name.release);
+}
 
 #define dump_meta_data(field, s)					\
-      s << "BEGIN_HEADER"      << std::endl;				\
-      s << "HDR_VERSION = "    << field.hdr_version    << std::endl;	\
-      s << "DATATYPE = "       << field.data_type      << std::endl;	\
-      s << "STORAGE_FORMAT = " << field.storage_format << std::endl;	\
-      for(int i=0;i<4;i++){						\
-	s << "DIMENSION_" << i+1 << " = " << field.dimension[i] << std::endl ; \
-      }									\
-      s << "LINK_TRACE = " << std::setprecision(10) << field.link_trace << std::endl; \
-      s << "PLAQUETTE  = " << std::setprecision(10) << field.plaquette  << std::endl; \
-      for(int i=0;i<4;i++){						\
-	s << "BOUNDARY_"<<i+1<<" = " << field.boundary[i] << std::endl;	\
-      }									\
+  s << "BEGIN_HEADER"      << std::endl;				\
+  s << "HDR_VERSION = "    << field.hdr_version    << std::endl;	\
+  s << "DATATYPE = "       << field.data_type      << std::endl;	\
+  s << "STORAGE_FORMAT = " << field.storage_format << std::endl;	\
+  for(int i=0;i<4;i++){							\
+    s << "DIMENSION_" << i+1 << " = " << field.dimension[i] << std::endl ; \
+  }									\
+  s << "LINK_TRACE = " << std::setprecision(10) << field.link_trace << std::endl; \
+  s << "PLAQUETTE  = " << std::setprecision(10) << field.plaquette  << std::endl; \
+  for(int i=0;i<4;i++){							\
+    s << "BOUNDARY_"<<i+1<<" = " << field.boundary[i] << std::endl;	\
+  }									\
 									\
-      s << "CHECKSUM = "<< std::hex << std::setw(10) << field.checksum << std::dec<<std::endl; \
-      s << "SCIDAC_CHECKSUMA = "<< std::hex << std::setw(10) << field.scidac_checksuma << std::dec<<std::endl; \
-      s << "SCIDAC_CHECKSUMB = "<< std::hex << std::setw(10) << field.scidac_checksumb << std::dec<<std::endl; \
-      s << "ENSEMBLE_ID = "     << field.ensemble_id      << std::endl;	\
-      s << "ENSEMBLE_LABEL = "  << field.ensemble_label   << std::endl;	\
-      s << "SEQUENCE_NUMBER = " << field.sequence_number  << std::endl;	\
-      s << "CREATOR = "         << field.creator          << std::endl;	\
-      s << "CREATOR_HARDWARE = "<< field.creator_hardware << std::endl;	\
-      s << "CREATION_DATE = "   << field.creation_date    << std::endl;	\
-      s << "ARCHIVE_DATE = "    << field.archive_date     << std::endl;	\
-      s << "FLOATING_POINT = "  << field.floating_point   << std::endl;	\
-      s << "END_HEADER"         << std::endl;
+  s << "CHECKSUM = "<< std::hex << std::setw(10) << field.checksum << std::dec<<std::endl; \
+  s << "SCIDAC_CHECKSUMA = "<< std::hex << std::setw(10) << field.scidac_checksuma << std::dec<<std::endl; \
+  s << "SCIDAC_CHECKSUMB = "<< std::hex << std::setw(10) << field.scidac_checksumb << std::dec<<std::endl; \
+  s << "ENSEMBLE_ID = "     << field.ensemble_id      << std::endl;	\
+  s << "ENSEMBLE_LABEL = "  << field.ensemble_label   << std::endl;	\
+  s << "SEQUENCE_NUMBER = " << field.sequence_number  << std::endl;	\
+  s << "CREATOR = "         << field.creator          << std::endl;	\
+  s << "CREATOR_HARDWARE = "<< field.creator_hardware << std::endl;	\
+  s << "CREATION_DATE = "   << field.creation_date    << std::endl;	\
+  s << "ARCHIVE_DATE = "    << field.archive_date     << std::endl;	\
+  s << "FLOATING_POINT = "  << field.floating_point   << std::endl;	\
+  s << "END_HEADER"         << std::endl;
 
 template<class vobj> inline void PrepareMetaData(Lattice<vobj> & field, FieldMetaData &header)
 {
-  GridBase *grid = field._grid;
+  GridBase *grid = field.Grid();
   std::string format = getFormatString<vobj>();
-   header.floating_point = format;
-   header.checksum = 0x0; // Nersc checksum unused in ILDG, Scidac
-   GridMetaData(grid,header); 
-   MachineCharacteristics(header);
- }
- inline void GaugeStatistics(Lattice<vLorentzColourMatrixF> & data,FieldMetaData &header)
- {
-   // How to convert data precision etc...
-   header.link_trace=Grid::QCD::WilsonLoops<PeriodicGimplF>::linkTrace(data);
-   header.plaquette =Grid::QCD::WilsonLoops<PeriodicGimplF>::avgPlaquette(data);
- }
- inline void GaugeStatistics(Lattice<vLorentzColourMatrixD> & data,FieldMetaData &header)
- {
-   // How to convert data precision etc...
-   header.link_trace=Grid::QCD::WilsonLoops<PeriodicGimplD>::linkTrace(data);
-   header.plaquette =Grid::QCD::WilsonLoops<PeriodicGimplD>::avgPlaquette(data);
- }
- template<> inline void PrepareMetaData<vLorentzColourMatrixF>(Lattice<vLorentzColourMatrixF> & field, FieldMetaData &header)
- {
+  header.floating_point = format;
+  header.checksum = 0x0; // Nersc checksum unused in ILDG, Scidac
+  GridMetaData(grid,header); 
+  MachineCharacteristics(header);
+}
+inline void GaugeStatistics(Lattice<vLorentzColourMatrixF> & data,FieldMetaData &header)
+{
+  // How to convert data precision etc...
+  header.link_trace=WilsonLoops<PeriodicGimplF>::linkTrace(data);
+  header.plaquette =WilsonLoops<PeriodicGimplF>::avgPlaquette(data);
+}
+inline void GaugeStatistics(Lattice<vLorentzColourMatrixD> & data,FieldMetaData &header)
+{
+  // How to convert data precision etc...
+  header.link_trace=WilsonLoops<PeriodicGimplD>::linkTrace(data);
+  header.plaquette =WilsonLoops<PeriodicGimplD>::avgPlaquette(data);
+}
+template<> inline void PrepareMetaData<vLorentzColourMatrixF>(Lattice<vLorentzColourMatrixF> & field, FieldMetaData &header)
+{
    
-   GridBase *grid = field._grid;
-   std::string format = getFormatString<vLorentzColourMatrixF>();
-   header.floating_point = format;
-   header.checksum = 0x0; // Nersc checksum unused in ILDG, Scidac
-   GridMetaData(grid,header); 
-   GaugeStatistics(field,header);
-   MachineCharacteristics(header);
- }
- template<> inline void PrepareMetaData<vLorentzColourMatrixD>(Lattice<vLorentzColourMatrixD> & field, FieldMetaData &header)
- {
-   GridBase *grid = field._grid;
-   std::string format = getFormatString<vLorentzColourMatrixD>();
-   header.floating_point = format;
-   header.checksum = 0x0; // Nersc checksum unused in ILDG, Scidac
-   GridMetaData(grid,header); 
-   GaugeStatistics(field,header);
-   MachineCharacteristics(header);
- }
+  GridBase *grid = field.Grid();
+  std::string format = getFormatString<vLorentzColourMatrixF>();
+  header.floating_point = format;
+  header.checksum = 0x0; // Nersc checksum unused in ILDG, Scidac
+  GridMetaData(grid,header); 
+  GaugeStatistics(field,header);
+  MachineCharacteristics(header);
+}
+template<> inline void PrepareMetaData<vLorentzColourMatrixD>(Lattice<vLorentzColourMatrixD> & field, FieldMetaData &header)
+{
+  GridBase *grid = field.Grid();
+  std::string format = getFormatString<vLorentzColourMatrixD>();
+  header.floating_point = format;
+  header.checksum = 0x0; // Nersc checksum unused in ILDG, Scidac
+  GridMetaData(grid,header); 
+  GaugeStatistics(field,header);
+  MachineCharacteristics(header);
+}
 
-    //////////////////////////////////////////////////////////////////////
-    // Utilities ; these are QCD aware
-    //////////////////////////////////////////////////////////////////////
-    inline void reconstruct3(LorentzColourMatrix & cm)
-    {
-      const int x=0;
-      const int y=1;
-      const int z=2;
-      for(int mu=0;mu<Nd;mu++){
-	cm(mu)()(2,x) = adj(cm(mu)()(0,y)*cm(mu)()(1,z)-cm(mu)()(0,z)*cm(mu)()(1,y)); //x= yz-zy
-	cm(mu)()(2,y) = adj(cm(mu)()(0,z)*cm(mu)()(1,x)-cm(mu)()(0,x)*cm(mu)()(1,z)); //y= zx-xz
-	cm(mu)()(2,z) = adj(cm(mu)()(0,x)*cm(mu)()(1,y)-cm(mu)()(0,y)*cm(mu)()(1,x)); //z= xy-yx
-      }
-    }
+//////////////////////////////////////////////////////////////////////
+// Utilities ; these are QCD aware
+//////////////////////////////////////////////////////////////////////
+inline void reconstruct3(LorentzColourMatrix & cm)
+{
+  const int x=0;
+  const int y=1;
+  const int z=2;
+  for(int mu=0;mu<Nd;mu++){
+    cm(mu)()(2,x) = adj(cm(mu)()(0,y)*cm(mu)()(1,z)-cm(mu)()(0,z)*cm(mu)()(1,y)); //x= yz-zy
+    cm(mu)()(2,y) = adj(cm(mu)()(0,z)*cm(mu)()(1,x)-cm(mu)()(0,x)*cm(mu)()(1,z)); //y= zx-xz
+    cm(mu)()(2,z) = adj(cm(mu)()(0,x)*cm(mu)()(1,y)-cm(mu)()(0,y)*cm(mu)()(1,x)); //z= xy-yx
+  }
+}
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // Some data types for intermediate storage
-    ////////////////////////////////////////////////////////////////////////////////
-    template<typename vtype> using iLorentzColour2x3 = iVector<iVector<iVector<vtype, Nc>, 2>, Nd >;
+////////////////////////////////////////////////////////////////////////////////
+// Some data types for intermediate storage
+////////////////////////////////////////////////////////////////////////////////
+template<typename vtype> using iLorentzColour2x3 = iVector<iVector<iVector<vtype, Nc>, 2>, Nd >;
 
-    typedef iLorentzColour2x3<Complex>  LorentzColour2x3;
-    typedef iLorentzColour2x3<ComplexF> LorentzColour2x3F;
-    typedef iLorentzColour2x3<ComplexD> LorentzColour2x3D;
+typedef iLorentzColour2x3<Complex>  LorentzColour2x3;
+typedef iLorentzColour2x3<ComplexF> LorentzColour2x3F;
+typedef iLorentzColour2x3<ComplexD> LorentzColour2x3D;
 
 /////////////////////////////////////////////////////////////////////////////////
 // Simple classes for precision conversion
@@ -276,56 +276,55 @@ struct BinarySimpleMunger {
 };
 
 
-    template<class fobj,class sobj>
-    struct GaugeSimpleMunger{
-      void operator()(fobj &in, sobj &out) {
-        for (int mu = 0; mu < Nd; mu++) {
-          for (int i = 0; i < Nc; i++) {
-          for (int j = 0; j < Nc; j++) {
-	    out(mu)()(i, j) = in(mu)()(i, j);
-	  }}
-        }
-      };
-    };
+template<class fobj,class sobj>
+struct GaugeSimpleMunger{
+  void operator()(fobj &in, sobj &out) {
+    for (int mu = 0; mu < Nd; mu++) {
+      for (int i = 0; i < Nc; i++) {
+	for (int j = 0; j < Nc; j++) {
+	  out(mu)()(i, j) = in(mu)()(i, j);
+	}}
+    }
+  };
+};
 
-    template <class fobj, class sobj>
-    struct GaugeSimpleUnmunger {
+template <class fobj, class sobj>
+struct GaugeSimpleUnmunger {
 
-      void operator()(sobj &in, fobj &out) {
-        for (int mu = 0; mu < Nd; mu++) {
-          for (int i = 0; i < Nc; i++) {
-          for (int j = 0; j < Nc; j++) {
-	    out(mu)()(i, j) = in(mu)()(i, j);
-	  }}
-        }
-      };
-    };
+  void operator()(sobj &in, fobj &out) {
+    for (int mu = 0; mu < Nd; mu++) {
+      for (int i = 0; i < Nc; i++) {
+	for (int j = 0; j < Nc; j++) {
+	  out(mu)()(i, j) = in(mu)()(i, j);
+	}}
+    }
+  };
+};
 
-    template<class fobj,class sobj>
-    struct Gauge3x2munger{
-      void operator() (fobj &in,sobj &out){
-	for(int mu=0;mu<Nd;mu++){
-	  for(int i=0;i<2;i++){
-	  for(int j=0;j<3;j++){
-	    out(mu)()(i,j) = in(mu)(i)(j);
-	  }}
-	}
-	reconstruct3(out);
-      }
-    };
-
-    template<class fobj,class sobj>
-    struct Gauge3x2unmunger{
-      void operator() (sobj &in,fobj &out){
-	for(int mu=0;mu<Nd;mu++){
-	  for(int i=0;i<2;i++){
-	  for(int j=0;j<3;j++){
-	    out(mu)(i)(j) = in(mu)()(i,j);
-	  }}
-	}
-      }
-    };
+template<class fobj,class sobj>
+struct Gauge3x2munger{
+  void operator() (fobj &in,sobj &out){
+    for(int mu=0;mu<Nd;mu++){
+      for(int i=0;i<2;i++){
+	for(int j=0;j<3;j++){
+	  out(mu)()(i,j) = in(mu)(i)(j);
+	}}
+    }
+    reconstruct3(out);
   }
+};
 
+template<class fobj,class sobj>
+struct Gauge3x2unmunger{
+  void operator() (sobj &in,fobj &out){
+    for(int mu=0;mu<Nd;mu++){
+      for(int i=0;i<2;i++){
+	for(int j=0;j<3;j++){
+	  out(mu)(i)(j) = in(mu)()(i,j);
+	}}
+    }
+  }
+};
+
+NAMESPACE_END(Grid);
 
-}
diff --git a/Grid/parallelIO/NerscIO.h b/Grid/parallelIO/NerscIO.h
index 0eb52071..499295c3 100644
--- a/Grid/parallelIO/NerscIO.h
+++ b/Grid/parallelIO/NerscIO.h
@@ -30,334 +30,330 @@
 #ifndef GRID_NERSC_IO_H
 #define GRID_NERSC_IO_H
 
-namespace Grid {
-  namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
-    using namespace Grid;
+using namespace Grid;
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // Write and read from fstream; comput header offset for payload
-    ////////////////////////////////////////////////////////////////////////////////
-    class NerscIO : public BinaryIO { 
-    public:
+////////////////////////////////////////////////////////////////////////////////
+// Write and read from fstream; comput header offset for payload
+////////////////////////////////////////////////////////////////////////////////
+class NerscIO : public BinaryIO { 
+public:
 
-      static inline void truncate(std::string file){
-	std::ofstream fout(file,std::ios::out);
-      }
+  static inline void truncate(std::string file){
+    std::ofstream fout(file,std::ios::out);
+  }
   
-      static inline unsigned int writeHeader(FieldMetaData &field,std::string file)
-      {
-      std::ofstream fout(file,std::ios::out|std::ios::in);
-      fout.seekp(0,std::ios::beg);
-      dump_meta_data(field, fout);
-      field.data_start = fout.tellp();
-      return field.data_start;
-    }
+  static inline unsigned int writeHeader(FieldMetaData &field,std::string file)
+  {
+    std::ofstream fout(file,std::ios::out|std::ios::in);
+    fout.seekp(0,std::ios::beg);
+    dump_meta_data(field, fout);
+    field.data_start = fout.tellp();
+    return field.data_start;
+  }
 
-      // for the header-reader
-      static inline int readHeader(std::string file,GridBase *grid,  FieldMetaData &field)
-      {
-      uint64_t offset=0;
-      std::map<std::string,std::string> header;
-      std::string line;
+  // for the header-reader
+  static inline int readHeader(std::string file,GridBase *grid,  FieldMetaData &field)
+  {
+    std::map<std::string,std::string> header;
+    std::string line;
 
-      //////////////////////////////////////////////////
-      // read the header
-      //////////////////////////////////////////////////
-      std::ifstream fin(file);
+    //////////////////////////////////////////////////
+    // read the header
+    //////////////////////////////////////////////////
+    std::ifstream fin(file);
 
-      getline(fin,line); // read one line and insist is 
+    getline(fin,line); // read one line and insist is 
 
-      removeWhitespace(line);
-      std::cout << GridLogMessage << "* " << line << std::endl;
+    removeWhitespace(line);
+    std::cout << GridLogMessage << "* " << line << std::endl;
 
-      assert(line==std::string("BEGIN_HEADER"));
+    assert(line==std::string("BEGIN_HEADER"));
 
-      do {
+    do {
       getline(fin,line); // read one line
       std::cout << GridLogMessage << "* "<<line<< std::endl;
       int eq = line.find("=");
       if(eq >0) {
-      std::string key=line.substr(0,eq);
-      std::string val=line.substr(eq+1);
-      removeWhitespace(key);
-      removeWhitespace(val);
+	std::string key=line.substr(0,eq);
+	std::string val=line.substr(eq+1);
+	removeWhitespace(key);
+	removeWhitespace(val);
       
-      header[key] = val;
-    }
+	header[key] = val;
+      }
     } while( line.find("END_HEADER") == std::string::npos );
 
-      field.data_start = fin.tellg();
+    field.data_start = fin.tellg();
 
-      //////////////////////////////////////////////////
-      // chomp the values
-      //////////////////////////////////////////////////
-      field.hdr_version    = header["HDR_VERSION"];
-      field.data_type      = header["DATATYPE"];
-      field.storage_format = header["STORAGE_FORMAT"];
+    //////////////////////////////////////////////////
+    // chomp the values
+    //////////////////////////////////////////////////
+    field.hdr_version    = header["HDR_VERSION"];
+    field.data_type      = header["DATATYPE"];
+    field.storage_format = header["STORAGE_FORMAT"];
   
-      field.dimension[0] = std::stol(header["DIMENSION_1"]);
-      field.dimension[1] = std::stol(header["DIMENSION_2"]);
-      field.dimension[2] = std::stol(header["DIMENSION_3"]);
-      field.dimension[3] = std::stol(header["DIMENSION_4"]);
+    field.dimension[0] = std::stol(header["DIMENSION_1"]);
+    field.dimension[1] = std::stol(header["DIMENSION_2"]);
+    field.dimension[2] = std::stol(header["DIMENSION_3"]);
+    field.dimension[3] = std::stol(header["DIMENSION_4"]);
 
-      assert(grid->_ndimension == 4);
-      for(int d=0;d<4;d++){
+    assert(grid->_ndimension == 4);
+    for(int d=0;d<4;d++){
       assert(grid->_fdimensions[d]==field.dimension[d]);
     }
 
-      field.link_trace = std::stod(header["LINK_TRACE"]);
-      field.plaquette  = std::stod(header["PLAQUETTE"]);
+    field.link_trace = std::stod(header["LINK_TRACE"]);
+    field.plaquette  = std::stod(header["PLAQUETTE"]);
 
-      field.boundary[0] = header["BOUNDARY_1"];
-      field.boundary[1] = header["BOUNDARY_2"];
-      field.boundary[2] = header["BOUNDARY_3"];
-      field.boundary[3] = header["BOUNDARY_4"];
+    field.boundary[0] = header["BOUNDARY_1"];
+    field.boundary[1] = header["BOUNDARY_2"];
+    field.boundary[2] = header["BOUNDARY_3"];
+    field.boundary[3] = header["BOUNDARY_4"];
 
-      field.checksum = std::stoul(header["CHECKSUM"],0,16);
-      field.ensemble_id      = header["ENSEMBLE_ID"];
-      field.ensemble_label   = header["ENSEMBLE_LABEL"];
-      field.sequence_number  = std::stol(header["SEQUENCE_NUMBER"]);
-      field.creator          = header["CREATOR"];
-      field.creator_hardware = header["CREATOR_HARDWARE"];
-      field.creation_date    = header["CREATION_DATE"];
-      field.archive_date     = header["ARCHIVE_DATE"];
-      field.floating_point   = header["FLOATING_POINT"];
+    field.checksum = std::stoul(header["CHECKSUM"],0,16);
+    field.ensemble_id      = header["ENSEMBLE_ID"];
+    field.ensemble_label   = header["ENSEMBLE_LABEL"];
+    field.sequence_number  = std::stol(header["SEQUENCE_NUMBER"]);
+    field.creator          = header["CREATOR"];
+    field.creator_hardware = header["CREATOR_HARDWARE"];
+    field.creation_date    = header["CREATION_DATE"];
+    field.archive_date     = header["ARCHIVE_DATE"];
+    field.floating_point   = header["FLOATING_POINT"];
 
-      return field.data_start;
+    return field.data_start;
+  }
+
+  /////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+  // Now the meat: the object readers
+  /////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+  template<class vsimd>
+  static inline void readConfiguration(Lattice<iLorentzColourMatrix<vsimd> > &Umu,
+				       FieldMetaData& header,
+				       std::string file)
+  {
+    typedef Lattice<iLorentzColourMatrix<vsimd> > GaugeField;
+
+    GridBase *grid = Umu.Grid();
+    uint64_t offset = readHeader(file,Umu.Grid(),header);
+
+    FieldMetaData clone(header);
+
+    std::string format(header.floating_point);
+
+    int ieee32big = (format == std::string("IEEE32BIG"));
+    int ieee32    = (format == std::string("IEEE32"));
+    int ieee64big = (format == std::string("IEEE64BIG"));
+    int ieee64    = (format == std::string("IEEE64"));
+
+    uint32_t nersc_csum,scidac_csuma,scidac_csumb;
+    // depending on datatype, set up munger;
+    // munger is a function of <floating point, Real, data_type>
+    if ( header.data_type == std::string("4D_SU3_GAUGE") ) {
+      if ( ieee32 || ieee32big ) {
+	BinaryIO::readLatticeObject<iLorentzColourMatrix<vsimd>, LorentzColour2x3F> 
+	  (Umu,file,Gauge3x2munger<LorentzColour2x3F,LorentzColourMatrix>(), offset,format,
+	   nersc_csum,scidac_csuma,scidac_csumb);
+      }
+      if ( ieee64 || ieee64big ) {
+	BinaryIO::readLatticeObject<iLorentzColourMatrix<vsimd>, LorentzColour2x3D> 
+	  (Umu,file,Gauge3x2munger<LorentzColour2x3D,LorentzColourMatrix>(),offset,format,
+	   nersc_csum,scidac_csuma,scidac_csumb);
+      }
+    } else if ( header.data_type == std::string("4D_SU3_GAUGE_3x3") ) {
+      if ( ieee32 || ieee32big ) {
+	BinaryIO::readLatticeObject<iLorentzColourMatrix<vsimd>,LorentzColourMatrixF>
+	  (Umu,file,GaugeSimpleMunger<LorentzColourMatrixF,LorentzColourMatrix>(),offset,format,
+	   nersc_csum,scidac_csuma,scidac_csumb);
+      }
+      if ( ieee64 || ieee64big ) {
+	BinaryIO::readLatticeObject<iLorentzColourMatrix<vsimd>,LorentzColourMatrixD>
+	  (Umu,file,GaugeSimpleMunger<LorentzColourMatrixD,LorentzColourMatrix>(),offset,format,
+	   nersc_csum,scidac_csuma,scidac_csumb);
+      }
+    } else {
+      assert(0);
     }
 
-    /////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-    // Now the meat: the object readers
-    /////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+    GaugeStatistics(Umu,clone);
 
-    template<class vsimd>
-    static inline void readConfiguration(Lattice<iLorentzColourMatrix<vsimd> > &Umu,
-					 FieldMetaData& header,
-					 std::string file)
-    {
-      typedef Lattice<iLorentzColourMatrix<vsimd> > GaugeField;
+    std::cout<<GridLogMessage <<"NERSC Configuration "<<file<<" checksum "<<std::hex<<nersc_csum<< std::dec
+	     <<" header   "<<std::hex<<header.checksum<<std::dec <<std::endl;
+    std::cout<<GridLogMessage <<"NERSC Configuration "<<file<<" plaquette "<<clone.plaquette
+	     <<" header    "<<header.plaquette<<std::endl;
+    std::cout<<GridLogMessage <<"NERSC Configuration "<<file<<" link_trace "<<clone.link_trace
+	     <<" header    "<<header.link_trace<<std::endl;
 
-      GridBase *grid = Umu._grid;
-      uint64_t offset = readHeader(file,Umu._grid,header);
-
-      FieldMetaData clone(header);
-
-      std::string format(header.floating_point);
-
-      int ieee32big = (format == std::string("IEEE32BIG"));
-      int ieee32    = (format == std::string("IEEE32"));
-      int ieee64big = (format == std::string("IEEE64BIG"));
-      int ieee64    = (format == std::string("IEEE64"));
-
-      uint32_t nersc_csum,scidac_csuma,scidac_csumb;
-      // depending on datatype, set up munger;
-      // munger is a function of <floating point, Real, data_type>
-      if ( header.data_type == std::string("4D_SU3_GAUGE") ) {
-	if ( ieee32 || ieee32big ) {
-	  BinaryIO::readLatticeObject<iLorentzColourMatrix<vsimd>, LorentzColour2x3F> 
-	    (Umu,file,Gauge3x2munger<LorentzColour2x3F,LorentzColourMatrix>(), offset,format,
-	     nersc_csum,scidac_csuma,scidac_csumb);
-	}
-	if ( ieee64 || ieee64big ) {
-	  BinaryIO::readLatticeObject<iLorentzColourMatrix<vsimd>, LorentzColour2x3D> 
-	    (Umu,file,Gauge3x2munger<LorentzColour2x3D,LorentzColourMatrix>(),offset,format,
-	     nersc_csum,scidac_csuma,scidac_csumb);
-	}
-      } else if ( header.data_type == std::string("4D_SU3_GAUGE_3x3") ) {
-	if ( ieee32 || ieee32big ) {
-	  BinaryIO::readLatticeObject<iLorentzColourMatrix<vsimd>,LorentzColourMatrixF>
-	    (Umu,file,GaugeSimpleMunger<LorentzColourMatrixF,LorentzColourMatrix>(),offset,format,
-	     nersc_csum,scidac_csuma,scidac_csumb);
-	}
-	if ( ieee64 || ieee64big ) {
-	  BinaryIO::readLatticeObject<iLorentzColourMatrix<vsimd>,LorentzColourMatrixD>
-	    (Umu,file,GaugeSimpleMunger<LorentzColourMatrixD,LorentzColourMatrix>(),offset,format,
-	     nersc_csum,scidac_csuma,scidac_csumb);
-	}
-      } else {
-	assert(0);
-      }
-
-      GaugeStatistics(Umu,clone);
-
-      std::cout<<GridLogMessage <<"NERSC Configuration "<<file<<" checksum "<<std::hex<<nersc_csum<< std::dec
-	       <<" header   "<<std::hex<<header.checksum<<std::dec <<std::endl;
-      std::cout<<GridLogMessage <<"NERSC Configuration "<<file<<" plaquette "<<clone.plaquette
-	       <<" header    "<<header.plaquette<<std::endl;
-      std::cout<<GridLogMessage <<"NERSC Configuration "<<file<<" link_trace "<<clone.link_trace
-	       <<" header    "<<header.link_trace<<std::endl;
-
-      if ( fabs(clone.plaquette -header.plaquette ) >=  1.0e-5 ) { 
-	std::cout << " Plaquette mismatch "<<std::endl;
-	std::cout << Umu[0]<<std::endl;
-	std::cout << Umu[1]<<std::endl;
-      }
-      if ( nersc_csum != header.checksum ) { 
-	std::cerr << " checksum mismatch " << std::endl;
-	std::cerr << " plaqs " << clone.plaquette << " " << header.plaquette << std::endl;
-	std::cerr << " trace " << clone.link_trace<< " " << header.link_trace<< std::endl;
-	std::cerr << " nersc_csum  " <<std::hex<< nersc_csum << " " << header.checksum<< std::dec<< std::endl;
-	exit(0);
-      }
-      assert(fabs(clone.plaquette -header.plaquette ) < 1.0e-5 );
-      assert(fabs(clone.link_trace-header.link_trace) < 1.0e-6 );
-      assert(nersc_csum == header.checksum );
+    if ( fabs(clone.plaquette -header.plaquette ) >=  1.0e-5 ) { 
+      std::cout << " Plaquette mismatch "<<std::endl;
+    }
+    if ( nersc_csum != header.checksum ) { 
+      std::cerr << " checksum mismatch " << std::endl;
+      std::cerr << " plaqs " << clone.plaquette << " " << header.plaquette << std::endl;
+      std::cerr << " trace " << clone.link_trace<< " " << header.link_trace<< std::endl;
+      std::cerr << " nersc_csum  " <<std::hex<< nersc_csum << " " << header.checksum<< std::dec<< std::endl;
+      exit(0);
+    }
+    assert(fabs(clone.plaquette -header.plaquette ) < 1.0e-5 );
+    assert(fabs(clone.link_trace-header.link_trace) < 1.0e-6 );
+    assert(nersc_csum == header.checksum );
       
-      std::cout<<GridLogMessage <<"NERSC Configuration "<<file<< " and plaquette, link trace, and checksum agree"<<std::endl;
-    }
+    std::cout<<GridLogMessage <<"NERSC Configuration "<<file<< " and plaquette, link trace, and checksum agree"<<std::endl;
+  }
 
-      template<class vsimd>
-      static inline void writeConfiguration(Lattice<iLorentzColourMatrix<vsimd> > &Umu,
-					    std::string file, 
-					    int two_row,
-					    int bits32)
-      {
-	typedef Lattice<iLorentzColourMatrix<vsimd> > GaugeField;
+  template<class vsimd>
+  static inline void writeConfiguration(Lattice<iLorentzColourMatrix<vsimd> > &Umu,
+					std::string file, 
+					int two_row,
+					int bits32)
+  {
+    typedef Lattice<iLorentzColourMatrix<vsimd> > GaugeField;
 
-	typedef iLorentzColourMatrix<vsimd> vobj;
-	typedef typename vobj::scalar_object sobj;
+    typedef iLorentzColourMatrix<vsimd> vobj;
+    typedef typename vobj::scalar_object sobj;
 
-	FieldMetaData header;
-	///////////////////////////////////////////
-	// Following should become arguments
-	///////////////////////////////////////////
-	header.sequence_number = 1;
-	header.ensemble_id     = "UKQCD";
-	header.ensemble_label  = "DWF";
+    FieldMetaData header;
+    ///////////////////////////////////////////
+    // Following should become arguments
+    ///////////////////////////////////////////
+    header.sequence_number = 1;
+    header.ensemble_id     = "UKQCD";
+    header.ensemble_label  = "DWF";
 
-	typedef LorentzColourMatrixD fobj3D;
-	typedef LorentzColour2x3D    fobj2D;
+    typedef LorentzColourMatrixD fobj3D;
+    typedef LorentzColour2x3D    fobj2D;
   
-	GridBase *grid = Umu._grid;
+    GridBase *grid = Umu.Grid();
 
-	GridMetaData(grid,header);
-	assert(header.nd==4);
-	GaugeStatistics(Umu,header);
-	MachineCharacteristics(header);
+    GridMetaData(grid,header);
+    assert(header.nd==4);
+    GaugeStatistics(Umu,header);
+    MachineCharacteristics(header);
 
 	uint64_t offset;
 
-	// Sod it -- always write 3x3 double
-	header.floating_point = std::string("IEEE64BIG");
-	header.data_type      = std::string("4D_SU3_GAUGE_3x3");
-	GaugeSimpleUnmunger<fobj3D,sobj> munge;
+    // Sod it -- always write 3x3 double
+    header.floating_point = std::string("IEEE64BIG");
+    header.data_type      = std::string("4D_SU3_GAUGE_3x3");
+    GaugeSimpleUnmunger<fobj3D,sobj> munge;
 	if ( grid->IsBoss() ) { 
 	  truncate(file);
-	  offset = writeHeader(header,file);
+    offset = writeHeader(header,file);
 	}
 	grid->Broadcast(0,(void *)&offset,sizeof(offset));
 
-	uint32_t nersc_csum,scidac_csuma,scidac_csumb;
-	BinaryIO::writeLatticeObject<vobj,fobj3D>(Umu,file,munge,offset,header.floating_point,
-								  nersc_csum,scidac_csuma,scidac_csumb);
-	header.checksum = nersc_csum;
+    uint32_t nersc_csum,scidac_csuma,scidac_csumb;
+    BinaryIO::writeLatticeObject<vobj,fobj3D>(Umu,file,munge,offset,header.floating_point,
+					      nersc_csum,scidac_csuma,scidac_csumb);
+    header.checksum = nersc_csum;
 	if ( grid->IsBoss() ) { 
-	  writeHeader(header,file);
+    writeHeader(header,file);
 	}
 
-	std::cout<<GridLogMessage <<"Written NERSC Configuration on "<< file << " checksum "
-		 <<std::hex<<header.checksum
-		 <<std::dec<<" plaq "<< header.plaquette <<std::endl;
+    std::cout<<GridLogMessage <<"Written NERSC Configuration on "<< file << " checksum "
+	     <<std::hex<<header.checksum
+	     <<std::dec<<" plaq "<< header.plaquette <<std::endl;
 
-      }
-      ///////////////////////////////
-      // RNG state
-      ///////////////////////////////
-      static inline void writeRNGState(GridSerialRNG &serial,GridParallelRNG &parallel,std::string file)
-      {
-	typedef typename GridParallelRNG::RngStateType RngStateType;
+  }
+  ///////////////////////////////
+  // RNG state
+  ///////////////////////////////
+  static inline void writeRNGState(GridSerialRNG &serial,GridParallelRNG &parallel,std::string file)
+  {
+    typedef typename GridParallelRNG::RngStateType RngStateType;
 
-	// Following should become arguments
-	FieldMetaData header;
-	header.sequence_number = 1;
-	header.ensemble_id     = "UKQCD";
-	header.ensemble_label  = "DWF";
+    // Following should become arguments
+    FieldMetaData header;
+    header.sequence_number = 1;
+    header.ensemble_id     = "UKQCD";
+    header.ensemble_label  = "DWF";
 
-	GridBase *grid = parallel._grid;
+    GridBase *grid = parallel.Grid();
 
-	GridMetaData(grid,header);
-	assert(header.nd==4);
-	header.link_trace=0.0;
-	header.plaquette=0.0;
-	MachineCharacteristics(header);
+    GridMetaData(grid,header);
+    assert(header.nd==4);
+    header.link_trace=0.0;
+    header.plaquette=0.0;
+    MachineCharacteristics(header);
 
 	uint64_t offset;
   
 #ifdef RNG_RANLUX
-	header.floating_point = std::string("UINT64");
-	header.data_type      = std::string("RANLUX48");
+    header.floating_point = std::string("UINT64");
+    header.data_type      = std::string("RANLUX48");
 #endif
 #ifdef RNG_MT19937
-	header.floating_point = std::string("UINT32");
-	header.data_type      = std::string("MT19937");
+    header.floating_point = std::string("UINT32");
+    header.data_type      = std::string("MT19937");
 #endif
 #ifdef RNG_SITMO
-	header.floating_point = std::string("UINT64");
-	header.data_type      = std::string("SITMO");
+    header.floating_point = std::string("UINT64");
+    header.data_type      = std::string("SITMO");
 #endif
 
 	if ( grid->IsBoss() ) { 
-	  truncate(file);
-	  offset = writeHeader(header,file);
+    truncate(file);
+    offset = writeHeader(header,file);
 	}
 	grid->Broadcast(0,(void *)&offset,sizeof(offset));
 	
-	uint32_t nersc_csum,scidac_csuma,scidac_csumb;
-	BinaryIO::writeRNG(serial,parallel,file,offset,nersc_csum,scidac_csuma,scidac_csumb);
-	header.checksum = nersc_csum;
+    uint32_t nersc_csum,scidac_csuma,scidac_csumb;
+    BinaryIO::writeRNG(serial,parallel,file,offset,nersc_csum,scidac_csuma,scidac_csumb);
+    header.checksum = nersc_csum;
 	if ( grid->IsBoss() ) { 
-	  offset = writeHeader(header,file);
+    offset = writeHeader(header,file);
 	}
 
-	std::cout<<GridLogMessage 
-		 <<"Written NERSC RNG STATE "<<file<< " checksum "
-		 <<std::hex<<header.checksum
-		 <<std::dec<<std::endl;
+    std::cout<<GridLogMessage 
+	     <<"Written NERSC RNG STATE "<<file<< " checksum "
+	     <<std::hex<<header.checksum
+	     <<std::dec<<std::endl;
 
-      }
+  }
     
-      static inline void readRNGState(GridSerialRNG &serial,GridParallelRNG & parallel,FieldMetaData& header,std::string file)
-      {
-	typedef typename GridParallelRNG::RngStateType RngStateType;
+  static inline void readRNGState(GridSerialRNG &serial,GridParallelRNG & parallel,FieldMetaData& header,std::string file)
+  {
+    typedef typename GridParallelRNG::RngStateType RngStateType;
 
-	GridBase *grid = parallel._grid;
+    GridBase *grid = parallel.Grid();
 
 	uint64_t offset = readHeader(file,grid,header);
 
-	FieldMetaData clone(header);
+    FieldMetaData clone(header);
 
-	std::string format(header.floating_point);
-	std::string data_type(header.data_type);
+    std::string format(header.floating_point);
+    std::string data_type(header.data_type);
 
 #ifdef RNG_RANLUX
-	assert(format == std::string("UINT64"));
-	assert(data_type == std::string("RANLUX48"));
+    assert(format == std::string("UINT64"));
+    assert(data_type == std::string("RANLUX48"));
 #endif
 #ifdef RNG_MT19937
-	assert(format == std::string("UINT32"));
-	assert(data_type == std::string("MT19937"));
+    assert(format == std::string("UINT32"));
+    assert(data_type == std::string("MT19937"));
 #endif
 #ifdef RNG_SITMO
-	assert(format == std::string("UINT64"));
-	assert(data_type == std::string("SITMO"));
+    assert(format == std::string("UINT64"));
+    assert(data_type == std::string("SITMO"));
 #endif
 
-	// depending on datatype, set up munger;
-	// munger is a function of <floating point, Real, data_type>
-	uint32_t nersc_csum,scidac_csuma,scidac_csumb;
-	BinaryIO::readRNG(serial,parallel,file,offset,nersc_csum,scidac_csuma,scidac_csumb);
+    // depending on datatype, set up munger;
+    // munger is a function of <floating point, Real, data_type>
+    uint32_t nersc_csum,scidac_csuma,scidac_csumb;
+    BinaryIO::readRNG(serial,parallel,file,offset,nersc_csum,scidac_csuma,scidac_csumb);
 
-	if ( nersc_csum != header.checksum ) { 
-	  std::cerr << "checksum mismatch "<<std::hex<< nersc_csum <<" "<<header.checksum<<std::dec<<std::endl;
-	  exit(0);
-	}
-	assert(nersc_csum == header.checksum );
+    if ( nersc_csum != header.checksum ) { 
+      std::cerr << "checksum mismatch "<<std::hex<< nersc_csum <<" "<<header.checksum<<std::dec<<std::endl;
+      exit(0);
+    }
+    assert(nersc_csum == header.checksum );
 
-	std::cout<<GridLogMessage <<"Read NERSC RNG file "<<file<< " format "<< data_type <<std::endl;
-      }
+    std::cout<<GridLogMessage <<"Read NERSC RNG file "<<file<< " format "<< data_type <<std::endl;
+  }
+};
 
-    };
+NAMESPACE_END(QCD);
 
-  }}
 #endif
diff --git a/Grid/perfmon/PerfCount.cc b/Grid/perfmon/PerfCount.cc
index c6f92b9f..2062bb59 100644
--- a/Grid/perfmon/PerfCount.cc
+++ b/Grid/perfmon/PerfCount.cc
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,13 +23,13 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 
 #include <Grid/GridCore.h>
 #include <Grid/perfmon/PerfCount.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 #define CacheControl(L,O,R) ((PERF_COUNT_HW_CACHE_##L)|(PERF_COUNT_HW_CACHE_OP_##O<<8)| (PERF_COUNT_HW_CACHE_RESULT_##R<<16))
 #define RawConfig(A,B) (A<<8|B)
@@ -39,16 +39,16 @@ const PerformanceCounter::PerformanceCounterConfig PerformanceCounter::Performan
   { PERF_TYPE_HARDWARE, PERF_COUNT_HW_CACHE_MISSES        ,  "CACHE_MISSES......." , CACHE_REFERENCES},
   { PERF_TYPE_HARDWARE, PERF_COUNT_HW_CPU_CYCLES          ,  "CPUCYCLES.........." , INSTRUCTIONS},
   { PERF_TYPE_HARDWARE, PERF_COUNT_HW_INSTRUCTIONS        ,  "INSTRUCTIONS......." , CPUCYCLES   },
-    // 4
+  // 4
 #ifdef KNL
-    { PERF_TYPE_RAW, RawConfig(0x40,0x04), "ALL_LOADS..........", CPUCYCLES    },
-    { PERF_TYPE_RAW, RawConfig(0x01,0x04), "L1_MISS_LOADS......", L1D_READ_ACCESS  },
-    { PERF_TYPE_RAW, RawConfig(0x40,0x04), "ALL_LOADS..........", L1D_READ_ACCESS    },
-    { PERF_TYPE_RAW, RawConfig(0x02,0x04), "L2_HIT_LOADS.......", L1D_READ_ACCESS  },
-    { PERF_TYPE_RAW, RawConfig(0x04,0x04), "L2_MISS_LOADS......", L1D_READ_ACCESS  },
-    { PERF_TYPE_RAW, RawConfig(0x10,0x04), "UTLB_MISS_LOADS....", L1D_READ_ACCESS },
-    { PERF_TYPE_RAW, RawConfig(0x08,0x04), "DTLB_MISS_LOADS....", L1D_READ_ACCESS },
-    // 11
+  { PERF_TYPE_RAW, RawConfig(0x40,0x04), "ALL_LOADS..........", CPUCYCLES    },
+  { PERF_TYPE_RAW, RawConfig(0x01,0x04), "L1_MISS_LOADS......", L1D_READ_ACCESS },
+  { PERF_TYPE_RAW, RawConfig(0x40,0x04), "ALL_LOADS..........", L1D_READ_ACCESS },
+  { PERF_TYPE_RAW, RawConfig(0x02,0x04), "L2_HIT_LOADS.......", L1D_READ_ACCESS },
+  { PERF_TYPE_RAW, RawConfig(0x04,0x04), "L2_MISS_LOADS......", L1D_READ_ACCESS },
+  { PERF_TYPE_RAW, RawConfig(0x10,0x04), "UTLB_MISS_LOADS....", L1D_READ_ACCESS },
+  { PERF_TYPE_RAW, RawConfig(0x08,0x04), "DTLB_MISS_LOADS....", L1D_READ_ACCESS },
+  // 11
 #else
   { PERF_TYPE_HW_CACHE, CacheControl(L1D,READ,ACCESS)     ,  "L1D_READ_ACCESS....",INSTRUCTIONS},
   { PERF_TYPE_HW_CACHE, CacheControl(L1D,READ,MISS)       ,  "L1D_READ_MISS......",L1D_READ_ACCESS},
@@ -57,19 +57,20 @@ const PerformanceCounter::PerformanceCounterConfig PerformanceCounter::Performan
   { PERF_TYPE_HW_CACHE, CacheControl(L1D,PREFETCH,MISS)   ,  "L1D_PREFETCH_MISS..",L1D_READ_ACCESS},
   { PERF_TYPE_HW_CACHE, CacheControl(L1D,PREFETCH,ACCESS) ,  "L1D_PREFETCH_ACCESS",L1D_READ_ACCESS},
   { PERF_TYPE_HW_CACHE, CacheControl(L1D,PREFETCH,ACCESS) ,  "L1D_PREFETCH_ACCESS",L1D_READ_ACCESS},
-    // 11
+   // 11
 #endif
   { PERF_TYPE_HW_CACHE, CacheControl(LL,READ,MISS)        ,  "LL_READ_MISS.......",L1D_READ_ACCESS},
   { PERF_TYPE_HW_CACHE, CacheControl(LL,READ,ACCESS)      ,  "LL_READ_ACCESS.....",L1D_READ_ACCESS},
   { PERF_TYPE_HW_CACHE, CacheControl(LL,WRITE,MISS)       ,  "LL_WRITE_MISS......",L1D_READ_ACCESS},
   { PERF_TYPE_HW_CACHE, CacheControl(LL,WRITE,ACCESS)     ,  "LL_WRITE_ACCESS....",L1D_READ_ACCESS},
-    //15
+  //15
   { PERF_TYPE_HW_CACHE, CacheControl(LL,PREFETCH,MISS)    ,  "LL_PREFETCH_MISS...",L1D_READ_ACCESS},
   { PERF_TYPE_HW_CACHE, CacheControl(LL,PREFETCH,ACCESS)  ,  "LL_PREFETCH_ACCESS.",L1D_READ_ACCESS},
   { PERF_TYPE_HW_CACHE, CacheControl(L1I,READ,MISS)       ,  "L1I_READ_MISS......",INSTRUCTIONS},
   { PERF_TYPE_HW_CACHE, CacheControl(L1I,READ,ACCESS)     ,  "L1I_READ_ACCESS....",INSTRUCTIONS}
-    //19
+  //19
   //  { PERF_TYPE_HARDWARE, PERF_COUNT_HW_STALLED_CYCLES_FRONTEND, "STALL_CYCLES" },
 #endif
 };
-}
+NAMESPACE_END(Grid);
+
diff --git a/Grid/perfmon/PerfCount.h b/Grid/perfmon/PerfCount.h
index 73d2c70f..13d59719 100644
--- a/Grid/perfmon/PerfCount.h
+++ b/Grid/perfmon/PerfCount.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,8 +25,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_PERFCOUNT_H
 #define GRID_PERFCOUNT_H
 
@@ -47,7 +47,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #include <x86intrin.h>
 #endif
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 #ifdef __linux__
 static long perf_event_open(struct perf_event_attr *hw_event, pid_t pid,
@@ -84,11 +84,14 @@ inline uint64_t cyclecount(void){
 
 #ifdef __bgq__
 inline uint64_t cyclecount(void){ 
-   uint64_t tmp;
-   asm volatile ("mfspr %0,0x10C" : "=&r" (tmp)  );
-   return tmp;
+  uint64_t tmp;
+  asm volatile ("mfspr %0,0x10C" : "=&r" (tmp)  );
+  return tmp;
 }
 #elif defined __x86_64__
+#ifdef GRID_NVCC
+accelerator_inline uint64_t __rdtsc(void) {  return 0; }
+#endif
 inline uint64_t cyclecount(void){ 
   return __rdtsc();
   //  unsigned int dummy;
@@ -97,7 +100,7 @@ inline uint64_t cyclecount(void){
 #else
 
 inline uint64_t cyclecount(void){ 
-   return 0;
+  return 0;
 }
 
 #endif
@@ -212,7 +215,7 @@ public:
       ::ioctl(cyclefd, PERF_EVENT_IOC_DISABLE, 0);
       ign=::read(fd, &count, sizeof(long long));
       ign+=::read(cyclefd, &cycles, sizeof(long long));
-      assert(ign=2*sizeof(long long));
+      assert(ign==2*sizeof(long long));
     }
     elapsed = cyclecount() - begin;
 #else
@@ -225,8 +228,8 @@ public:
     int N = PerformanceCounterConfigs[PCT].normalisation;
     const char * sn = PerformanceCounterConfigs[N].name ;
     const char * sc = PerformanceCounterConfigs[PCT].name;
-      std::printf("tsc = %llu %s = %llu  %s = %20llu\n (%s/%s) rate = %lf\n", elapsed,sn ,cycles, 
-		  sc, count, sc,sn, (double)count/(double)cycles);
+    std::printf("tsc = %llu %s = %llu  %s = %20llu\n (%s/%s) rate = %lf\n", elapsed,sn ,cycles, 
+		sc, count, sc,sn, (double)count/(double)cycles);
 #else
     std::printf("%llu cycles \n", elapsed );
 #endif
@@ -241,5 +244,6 @@ public:
 
 };
 
-}
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/perfmon/Stat.cc b/Grid/perfmon/Stat.cc
index 3f47fd83..4c3be254 100644
--- a/Grid/perfmon/Stat.cc
+++ b/Grid/perfmon/Stat.cc
@@ -2,7 +2,7 @@
 #include <Grid/perfmon/PerfCount.h>
 #include <Grid/perfmon/Stat.h>
 
-namespace Grid { 
+NAMESPACE_BEGIN(Grid); 
 
 bool PmuStat::pmu_initialized=false;
 
@@ -175,39 +175,39 @@ void PmuStat::KNLevsetup(const char *ename, int &fd, int event, int umask)
 }
 
 
- void PmuStat::KNLsetup(void){
+void PmuStat::KNLsetup(void){
 
-   int ret;
-   char fname[1024];
+  int ret;
+  char fname[1024];
 
-   // MC RPQ inserts and WPQ inserts (reads & writes)
-   for (int mc = 0; mc < NMC; ++mc)
-     {
-       ::snprintf(fname, sizeof(fname), "/sys/devices/uncore_imc_%d",mc);
-       // RPQ Inserts
-       KNLevsetup(fname, gbl.mc_rd[mc], 0x1, 0x1);
-       // WPQ Inserts
-       KNLevsetup(fname, gbl.mc_wr[mc], 0x2, 0x1);
-     }
-   // EDC RPQ inserts and WPQ inserts
-   for (int edc=0; edc < NEDC; ++edc)
-     {
-       ::snprintf(fname, sizeof(fname), "/sys/devices/uncore_edc_eclk_%d",edc);
-       // RPQ inserts
-       KNLevsetup(fname, gbl.edc_rd[edc], 0x1, 0x1);
-       // WPQ inserts
-       KNLevsetup(fname, gbl.edc_wr[edc], 0x2, 0x1);
-     }
-   // EDC HitE, HitM, MissE, MissM
-   for (int edc=0; edc < NEDC; ++edc)
-     {
-       ::snprintf(fname, sizeof(fname), "/sys/devices/uncore_edc_uclk_%d", edc);
-       KNLevsetup(fname, gbl.edc_hite[edc], 0x2, 0x1);
-       KNLevsetup(fname, gbl.edc_hitm[edc], 0x2, 0x2);
-       KNLevsetup(fname, gbl.edc_misse[edc], 0x2, 0x4);
-       KNLevsetup(fname, gbl.edc_missm[edc], 0x2, 0x8);
-     }
- }
+  // MC RPQ inserts and WPQ inserts (reads & writes)
+  for (int mc = 0; mc < NMC; ++mc)
+    {
+      ::snprintf(fname, sizeof(fname), "/sys/devices/uncore_imc_%d",mc);
+      // RPQ Inserts
+      KNLevsetup(fname, gbl.mc_rd[mc], 0x1, 0x1);
+      // WPQ Inserts
+      KNLevsetup(fname, gbl.mc_wr[mc], 0x2, 0x1);
+    }
+  // EDC RPQ inserts and WPQ inserts
+  for (int edc=0; edc < NEDC; ++edc)
+    {
+      ::snprintf(fname, sizeof(fname), "/sys/devices/uncore_edc_eclk_%d",edc);
+      // RPQ inserts
+      KNLevsetup(fname, gbl.edc_rd[edc], 0x1, 0x1);
+      // WPQ inserts
+      KNLevsetup(fname, gbl.edc_wr[edc], 0x2, 0x1);
+    }
+  // EDC HitE, HitM, MissE, MissM
+  for (int edc=0; edc < NEDC; ++edc)
+    {
+      ::snprintf(fname, sizeof(fname), "/sys/devices/uncore_edc_uclk_%d", edc);
+      KNLevsetup(fname, gbl.edc_hite[edc], 0x2, 0x1);
+      KNLevsetup(fname, gbl.edc_hitm[edc], 0x2, 0x2);
+      KNLevsetup(fname, gbl.edc_misse[edc], 0x2, 0x4);
+      KNLevsetup(fname, gbl.edc_missm[edc], 0x2, 0x8);
+    }
+}
 
 uint64_t PmuStat::KNLreadctr(int fd)
 {
@@ -242,4 +242,5 @@ void PmuStat::KNLreadctrs(ctrs &c)
 }
 
 #endif
-}
+NAMESPACE_END(Grid); 
+
diff --git a/Grid/perfmon/Stat.h b/Grid/perfmon/Stat.h
index 96bd594a..30baec29 100644
--- a/Grid/perfmon/Stat.h
+++ b/Grid/perfmon/Stat.h
@@ -5,7 +5,7 @@
 #define _KNIGHTS_LANDING_ROOTONLY
 #endif
 
-namespace Grid { 
+NAMESPACE_BEGIN(Grid); 
 
 ///////////////////////////////////////////////////////////////////////////////
 // Extra KNL counters from MCDRAM
@@ -15,14 +15,14 @@ namespace Grid {
 #define NEDC 8
 struct ctrs
 {
-    uint64_t mcrd[NMC];
-    uint64_t mcwr[NMC];
-    uint64_t edcrd[NEDC]; 
-    uint64_t edcwr[NEDC];
-    uint64_t edchite[NEDC];
-    uint64_t edchitm[NEDC];
-    uint64_t edcmisse[NEDC];
-    uint64_t edcmissm[NEDC];
+  uint64_t mcrd[NMC];
+  uint64_t mcwr[NMC];
+  uint64_t edcrd[NEDC]; 
+  uint64_t edcwr[NEDC];
+  uint64_t edchite[NEDC];
+  uint64_t edchitm[NEDC];
+  uint64_t edcmisse[NEDC];
+  uint64_t edcmissm[NEDC];
 };
 // Peter/Azusa:
 // Our modification of a code provided by Larry Meadows from Intel
@@ -44,61 +44,62 @@ struct knl_gbl_
 
 class PmuStat
 {
-    uint64_t counters[8][256];
+  uint64_t counters[8][256];
 #ifdef _KNIGHTS_LANDING_
-    static struct knl_gbl_ gbl;
+  static struct knl_gbl_ gbl;
 #endif
-    const char *name;
+  const char *name;
 
-    uint64_t reads;     // memory reads
-    uint64_t writes;    // memory writes
-    uint64_t mrstart;   // memory read counter at start of parallel region
-    uint64_t mrend;     // memory read counter at end of parallel region
-    uint64_t mwstart;   // memory write counter at start of parallel region
-    uint64_t mwend;     // memory write counter at end of parallel region
+  uint64_t reads;     // memory reads
+  uint64_t writes;    // memory writes
+  uint64_t mrstart;   // memory read counter at start of parallel region
+  uint64_t mrend;     // memory read counter at end of parallel region
+  uint64_t mwstart;   // memory write counter at start of parallel region
+  uint64_t mwend;     // memory write counter at end of parallel region
 
-    // cumulative counters
-    uint64_t count;     // number of invocations
-    uint64_t tregion;   // total time in parallel region (from thread 0)
-    uint64_t tcycles;   // total cycles inside parallel region
-    uint64_t inst, ref, cyc;   // fixed counters
-    uint64_t pmc0, pmc1;// pmu
-    // add memory counters here
-    // temp variables
-    uint64_t tstart;    // tsc at start of parallel region
-    uint64_t tend;      // tsc at end of parallel region
-    // map for ctrs values
-    // 0 pmc0 start
-    // 1 pmc0 end
-    // 2 pmc1 start
-    // 3 pmc1 end
-    // 4 tsc start
-    // 5 tsc end
-    static bool pmu_initialized;
+  // cumulative counters
+  uint64_t count;     // number of invocations
+  uint64_t tregion;   // total time in parallel region (from thread 0)
+  uint64_t tcycles;   // total cycles inside parallel region
+  uint64_t inst, ref, cyc;   // fixed counters
+  uint64_t pmc0, pmc1;// pmu
+  // add memory counters here
+  // temp variables
+  uint64_t tstart;    // tsc at start of parallel region
+  uint64_t tend;      // tsc at end of parallel region
+  // map for ctrs values
+  // 0 pmc0 start
+  // 1 pmc0 end
+  // 2 pmc1 start
+  // 3 pmc1 end
+  // 4 tsc start
+  // 5 tsc end
+  static bool pmu_initialized;
 public:
-    static bool is_init(void){ return pmu_initialized;}
-    static void pmu_init(void);
-    static void pmu_fini(void);
-    static void pmu_start(void);
-    static void pmu_stop(void);
-    void accum(int nthreads);
-    static void xmemctrs(uint64_t *mr, uint64_t *mw);
-    void start(void);
-    void enter(int t);
-    void exit(int t);
-    void print(void);
-    void init(const char *regname);
-    void clear(void);
+  static bool is_init(void){ return pmu_initialized;}
+  static void pmu_init(void);
+  static void pmu_fini(void);
+  static void pmu_start(void);
+  static void pmu_stop(void);
+  void accum(int nthreads);
+  static void xmemctrs(uint64_t *mr, uint64_t *mw);
+  void start(void);
+  void enter(int t);
+  void exit(int t);
+  void print(void);
+  void init(const char *regname);
+  void clear(void);
 #ifdef _KNIGHTS_LANDING_
-    static void     KNLsetup(void);
-    static uint64_t KNLreadctr(int fd);
-    static void     KNLreadctrs(ctrs &c);
-    static void     KNLevsetup(const char *ename, int &fd, int event, int umask);
+  static void     KNLsetup(void);
+  static uint64_t KNLreadctr(int fd);
+  static void     KNLreadctrs(ctrs &c);
+  static void     KNLevsetup(const char *ename, int &fd, int event, int umask);
 #endif
     
-  };
+};
+
+NAMESPACE_END(Grid); 
 
-}
 #endif
 
 
diff --git a/Grid/perfmon/Timer.h b/Grid/perfmon/Timer.h
index ce1b5d76..88b4e1cc 100644
--- a/Grid/perfmon/Timer.h
+++ b/Grid/perfmon/Timer.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,8 +24,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_TIME_H
 #define GRID_TIME_H
 
@@ -33,11 +33,9 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <ctime>
 #include <chrono>
 
-namespace Grid {
-
-
-  // Dress the output; use std::chrono
+NAMESPACE_BEGIN(Grid)
 
+// Dress the output; use std::chrono
 // C++11 time facilities better?
 inline double usecond(void) {
   struct timeval tv;
@@ -80,7 +78,7 @@ inline std::ostream& operator<< (std::ostream & stream, const GridUsecs & now)
   stream.fill(fill);
   return stream;
 }
-
+ 
 
 class GridStopWatch {
 private:
@@ -125,5 +123,6 @@ public:
   }
 };
 
-}
+NAMESPACE_END(Grid)
+
 #endif
diff --git a/Grid/pugixml/pugixml.cc b/Grid/pugixml/pugixml.cc
index dd08092c..e7b395ad 100644
--- a/Grid/pugixml/pugixml.cc
+++ b/Grid/pugixml/pugixml.cc
@@ -14,7 +14,12 @@
 #ifndef SOURCE_PUGIXML_CPP
 #define SOURCE_PUGIXML_CPP
 
-#include <Grid/pugixml/pugixml.h>
+#ifdef __NVCC__
+#pragma push
+#pragma diag_suppress declared_but_not_referenced // suppress "function was declared but never referenced warning"
+#endif
+
+#include "pugixml.h"
 
 #include <stdlib.h>
 #include <stdio.h>
@@ -202,7 +207,7 @@ PUGI__NS_BEGIN
 	// Without a template<> we'll get multiple definitions of the same static
 	template <typename T> allocation_function xml_memory_management_function_storage<T>::allocate = default_allocate;
 	template <typename T> deallocation_function xml_memory_management_function_storage<T>::deallocate = default_deallocate;
-
+        template struct xml_memory_management_function_storage<int>;
 	typedef xml_memory_management_function_storage<int> xml_memory;
 PUGI__NS_END
 
@@ -12768,6 +12773,10 @@ namespace pugi
 #undef PUGI__THROW_ERROR
 #undef PUGI__CHECK_ERROR
 
+#ifdef GRID_NVCC
+#pragma pop
+#endif
+
 #endif
 
 /**
diff --git a/Grid/qcd/QCD.h b/Grid/qcd/QCD.h
index 7e9b2da0..2c8e60da 100644
--- a/Grid/qcd/QCD.h
+++ b/Grid/qcd/QCD.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -27,114 +27,112 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef GRID_QCD_BASE_H
-#define GRID_QCD_BASE_H
-namespace Grid{
-namespace QCD {
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
 
-    static const int Xdir = 0;
-    static const int Ydir = 1;
-    static const int Zdir = 2;
-    static const int Tdir = 3;
+NAMESPACE_BEGIN(Grid);
 
-  
-    static const int Xp = 0;
-    static const int Yp = 1;
-    static const int Zp = 2;
-    static const int Tp = 3;
-    static const int Xm = 4;
-    static const int Ym = 5;
-    static const int Zm = 6;
-    static const int Tm = 7;
+static constexpr int Xdir = 0;
+static constexpr int Ydir = 1;
+static constexpr int Zdir = 2;
+static constexpr int Tdir = 3;
 
-    static const int Nc=3;
-    static const int Ns=4;
-    static const int Nd=4;
-    static const int Nhs=2; // half spinor
-    static const int Nds=8; // double stored gauge field
-    static const int Ngp=2; // gparity index range
+static constexpr int Xp = 0;
+static constexpr int Yp = 1;
+static constexpr int Zp = 2;
+static constexpr int Tp = 3;
+static constexpr int Xm = 4;
+static constexpr int Ym = 5;
+static constexpr int Zm = 6;
+static constexpr int Tm = 7;
 
-    //////////////////////////////////////////////////////////////////////////////
-    // QCD iMatrix types
-    // Index conventions:                            Lorentz x Spin x Colour
-    // note: static const int or constexpr will work for type deductions
-    //       with the intel compiler (up to version 17)
-    //////////////////////////////////////////////////////////////////////////////
-    #define ColourIndex  2
-    #define SpinIndex    1
-    #define LorentzIndex 0
+static constexpr int Nc=3;
+static constexpr int Ns=4;
+static constexpr int Nd=4;
+static constexpr int Nhs=2; // half spinor
+static constexpr int Nds=8; // double stored gauge field
+static constexpr int Ngp=2; // gparity index range
 
-    // Also should make these a named enum type
-    static const int DaggerNo=0;
-    static const int DaggerYes=1;
-    static const int InverseNo=0;
-    static const int InverseYes=1;
+//////////////////////////////////////////////////////////////////////////////
+// QCD iMatrix types
+// Index conventions:                            Lorentz x Spin x Colour
+// note: static constexpr int or constexpr will work for type deductions
+//       with the intel compiler (up to version 17)
+//////////////////////////////////////////////////////////////////////////////
+#define ColourIndex  (2)
+#define SpinIndex    (1)
+#define LorentzIndex (0)
 
-    // Useful traits is this a spin index
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+// Also should make these a named enum type
+static constexpr int DaggerNo=0;
+static constexpr int DaggerYes=1;
+static constexpr int InverseNo=0;
+static constexpr int InverseYes=1;
 
-    const int SpinorIndex = 2;
-    template<typename T> struct isSpinor {
-      static const bool value = (SpinorIndex==T::TensorLevel);
-    };
-    template <typename T> using IfSpinor    = Invoke<std::enable_if< isSpinor<T>::value,int> > ;
-    template <typename T> using IfNotSpinor = Invoke<std::enable_if<!isSpinor<T>::value,int> > ;
+// Useful traits is this a spin index
+//typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
 
-    // ChrisK very keen to add extra space for Gparity doubling.
-    //
-    // Also add domain wall index, in a way where Wilson operator 
-    // naturally distributes across the 5th dimensions.
-    //
-    // That probably makes for GridRedBlack4dCartesian grid.
+const int SpinorIndex = 2;
+template<typename T> struct isSpinor {
+  static constexpr bool value = (SpinorIndex==T::TensorLevel);
+};
+template <typename T> using IfSpinor    = Invoke<std::enable_if< isSpinor<T>::value,int> > ;
+template <typename T> using IfNotSpinor = Invoke<std::enable_if<!isSpinor<T>::value,int> > ;
 
-    // s,sp,c,spc,lc
+// ChrisK very keen to add extra space for Gparity doubling.
+//
+// Also add domain wall index, in a way where Wilson operator 
+// naturally distributes across the 5th dimensions.
+//
+// That probably makes for GridRedBlack4dCartesian grid.
 
-    template<typename vtype> using iSinglet                     = iScalar<iScalar<iScalar<vtype> > >;
-    template<typename vtype> using iSpinMatrix                  = iScalar<iMatrix<iScalar<vtype>, Ns> >;
-    template<typename vtype> using iColourMatrix                = iScalar<iScalar<iMatrix<vtype, Nc> > > ;
-    template<typename vtype> using iSpinColourMatrix            = iScalar<iMatrix<iMatrix<vtype, Nc>, Ns> >;
-    template<typename vtype> using iLorentzColourMatrix         = iVector<iScalar<iMatrix<vtype, Nc> >, Nd > ;
-    template<typename vtype> using iDoubleStoredColourMatrix    = iVector<iScalar<iMatrix<vtype, Nc> >, Nds > ;
-    template<typename vtype> using iSpinVector                  = iScalar<iVector<iScalar<vtype>, Ns> >;
-    template<typename vtype> using iColourVector                = iScalar<iScalar<iVector<vtype, Nc> > >;
-    template<typename vtype> using iSpinColourVector            = iScalar<iVector<iVector<vtype, Nc>, Ns> >;
-    template<typename vtype> using iHalfSpinVector              = iScalar<iVector<iScalar<vtype>, Nhs> >;
-    template<typename vtype> using iHalfSpinColourVector        = iScalar<iVector<iVector<vtype, Nc>, Nhs> >;
+// s,sp,c,spc,lc
+
+template<typename vtype> using iSinglet                   = iScalar<iScalar<iScalar<vtype> > >;
+template<typename vtype> using iSpinMatrix                = iScalar<iMatrix<iScalar<vtype>, Ns> >;
+template<typename vtype> using iColourMatrix              = iScalar<iScalar<iMatrix<vtype, Nc> > > ;
+template<typename vtype> using iSpinColourMatrix          = iScalar<iMatrix<iMatrix<vtype, Nc>, Ns> >;
+template<typename vtype> using iLorentzColourMatrix       = iVector<iScalar<iMatrix<vtype, Nc> >, Nd > ;
+template<typename vtype> using iDoubleStoredColourMatrix  = iVector<iScalar<iMatrix<vtype, Nc> >, Nds > ;
+template<typename vtype> using iSpinVector                = iScalar<iVector<iScalar<vtype>, Ns> >;
+template<typename vtype> using iColourVector              = iScalar<iScalar<iVector<vtype, Nc> > >;
+template<typename vtype> using iSpinColourVector          = iScalar<iVector<iVector<vtype, Nc>, Ns> >;
+template<typename vtype> using iHalfSpinVector            = iScalar<iVector<iScalar<vtype>, Nhs> >;
+template<typename vtype> using iHalfSpinColourVector      = iScalar<iVector<iVector<vtype, Nc>, Nhs> >;
     template<typename vtype> using iSpinColourSpinColourMatrix  = iScalar<iMatrix<iMatrix<iMatrix<iMatrix<vtype, Nc>, Ns>, Nc>, Ns> >;
 
 
-    template<typename vtype> using iGparitySpinColourVector       = iVector<iVector<iVector<vtype, Nc>, Ns>, Ngp >;
-    template<typename vtype> using iGparityHalfSpinColourVector   = iVector<iVector<iVector<vtype, Nc>, Nhs>, Ngp >;
+template<typename vtype> using iGparitySpinColourVector       = iVector<iVector<iVector<vtype, Nc>, Ns>, Ngp >;
+template<typename vtype> using iGparityHalfSpinColourVector   = iVector<iVector<iVector<vtype, Nc>, Nhs>, Ngp >;
 
-    // Spin matrix
-    typedef iSpinMatrix<Complex  >          SpinMatrix;
-    typedef iSpinMatrix<ComplexF >          SpinMatrixF;
-    typedef iSpinMatrix<ComplexD >          SpinMatrixD;
+// Spin matrix
+typedef iSpinMatrix<Complex  >          SpinMatrix;
+typedef iSpinMatrix<ComplexF >          SpinMatrixF;
+typedef iSpinMatrix<ComplexD >          SpinMatrixD;
 
-    typedef iSpinMatrix<vComplex >          vSpinMatrix;
-    typedef iSpinMatrix<vComplexF>          vSpinMatrixF;
-    typedef iSpinMatrix<vComplexD>          vSpinMatrixD;
+typedef iSpinMatrix<vComplex >          vSpinMatrix;
+typedef iSpinMatrix<vComplexF>          vSpinMatrixF;
+typedef iSpinMatrix<vComplexD>          vSpinMatrixD;
 
-    // Colour Matrix
-    typedef iColourMatrix<Complex  >        ColourMatrix;
-    typedef iColourMatrix<ComplexF >        ColourMatrixF;
-    typedef iColourMatrix<ComplexD >        ColourMatrixD;
+// Colour Matrix
+typedef iColourMatrix<Complex  >        ColourMatrix;
+typedef iColourMatrix<ComplexF >        ColourMatrixF;
+typedef iColourMatrix<ComplexD >        ColourMatrixD;
 
-    typedef iColourMatrix<vComplex >        vColourMatrix;
-    typedef iColourMatrix<vComplexF>        vColourMatrixF;
-    typedef iColourMatrix<vComplexD>        vColourMatrixD;
+typedef iColourMatrix<vComplex >        vColourMatrix;
+typedef iColourMatrix<vComplexF>        vColourMatrixF;
+typedef iColourMatrix<vComplexD>        vColourMatrixD;
+
+// SpinColour matrix
+typedef iSpinColourMatrix<Complex  >    SpinColourMatrix;
+typedef iSpinColourMatrix<ComplexF >    SpinColourMatrixF;
+typedef iSpinColourMatrix<ComplexD >    SpinColourMatrixD;
+
+typedef iSpinColourMatrix<vComplex >    vSpinColourMatrix;
+typedef iSpinColourMatrix<vComplexF>    vSpinColourMatrixF;
+typedef iSpinColourMatrix<vComplexD>    vSpinColourMatrixD;
 
-    // SpinColour matrix
-    typedef iSpinColourMatrix<Complex  >    SpinColourMatrix;
-    typedef iSpinColourMatrix<ComplexF >    SpinColourMatrixF;
-    typedef iSpinColourMatrix<ComplexD >    SpinColourMatrixD;
-    
-    typedef iSpinColourMatrix<vComplex >    vSpinColourMatrix;
-    typedef iSpinColourMatrix<vComplexF>    vSpinColourMatrixF;
-    typedef iSpinColourMatrix<vComplexD>    vSpinColourMatrixD;
-    
     // SpinColourSpinColour matrix
     typedef iSpinColourSpinColourMatrix<Complex  >    SpinColourSpinColourMatrix;
     typedef iSpinColourSpinColourMatrix<ComplexF >    SpinColourSpinColourMatrixF;
@@ -153,383 +151,379 @@ namespace QCD {
     typedef iSpinColourSpinColourMatrix<vComplexF>    vSpinColourSpinColourMatrixF;
     typedef iSpinColourSpinColourMatrix<vComplexD>    vSpinColourSpinColourMatrixD;
 
-    // LorentzColour
-    typedef iLorentzColourMatrix<Complex  > LorentzColourMatrix;
-    typedef iLorentzColourMatrix<ComplexF > LorentzColourMatrixF;
-    typedef iLorentzColourMatrix<ComplexD > LorentzColourMatrixD;
+// LorentzColour
+typedef iLorentzColourMatrix<Complex  > LorentzColourMatrix;
+typedef iLorentzColourMatrix<ComplexF > LorentzColourMatrixF;
+typedef iLorentzColourMatrix<ComplexD > LorentzColourMatrixD;
 
-    typedef iLorentzColourMatrix<vComplex > vLorentzColourMatrix;
-    typedef iLorentzColourMatrix<vComplexF> vLorentzColourMatrixF;
-    typedef iLorentzColourMatrix<vComplexD> vLorentzColourMatrixD;
+typedef iLorentzColourMatrix<vComplex > vLorentzColourMatrix;
+typedef iLorentzColourMatrix<vComplexF> vLorentzColourMatrixF;
+typedef iLorentzColourMatrix<vComplexD> vLorentzColourMatrixD;
 
-    // DoubleStored gauge field
-    typedef iDoubleStoredColourMatrix<Complex  > DoubleStoredColourMatrix;
-    typedef iDoubleStoredColourMatrix<ComplexF > DoubleStoredColourMatrixF;
-    typedef iDoubleStoredColourMatrix<ComplexD > DoubleStoredColourMatrixD;
+// DoubleStored gauge field
+typedef iDoubleStoredColourMatrix<Complex  > DoubleStoredColourMatrix;
+typedef iDoubleStoredColourMatrix<ComplexF > DoubleStoredColourMatrixF;
+typedef iDoubleStoredColourMatrix<ComplexD > DoubleStoredColourMatrixD;
 
-    typedef iDoubleStoredColourMatrix<vComplex > vDoubleStoredColourMatrix;
-    typedef iDoubleStoredColourMatrix<vComplexF> vDoubleStoredColourMatrixF;
-    typedef iDoubleStoredColourMatrix<vComplexD> vDoubleStoredColourMatrixD;
+typedef iDoubleStoredColourMatrix<vComplex > vDoubleStoredColourMatrix;
+typedef iDoubleStoredColourMatrix<vComplexF> vDoubleStoredColourMatrixF;
+typedef iDoubleStoredColourMatrix<vComplexD> vDoubleStoredColourMatrixD;
 
-    // Spin vector
-    typedef iSpinVector<Complex >           SpinVector;
-    typedef iSpinVector<ComplexF>           SpinVectorF;
-    typedef iSpinVector<ComplexD>           SpinVectorD;
+// Spin vector
+typedef iSpinVector<Complex >           SpinVector;
+typedef iSpinVector<ComplexF>           SpinVectorF;
+typedef iSpinVector<ComplexD>           SpinVectorD;
 
-    typedef iSpinVector<vComplex >           vSpinVector;
-    typedef iSpinVector<vComplexF>           vSpinVectorF;
-    typedef iSpinVector<vComplexD>           vSpinVectorD;
+typedef iSpinVector<vComplex >           vSpinVector;
+typedef iSpinVector<vComplexF>           vSpinVectorF;
+typedef iSpinVector<vComplexD>           vSpinVectorD;
 
-    // Colour vector
-    typedef iColourVector<Complex >         ColourVector;
-    typedef iColourVector<ComplexF>         ColourVectorF;
-    typedef iColourVector<ComplexD>         ColourVectorD;
+// Colour vector
+typedef iColourVector<Complex >         ColourVector;
+typedef iColourVector<ComplexF>         ColourVectorF;
+typedef iColourVector<ComplexD>         ColourVectorD;
 
-    typedef iColourVector<vComplex >         vColourVector;
-    typedef iColourVector<vComplexF>         vColourVectorF;
-    typedef iColourVector<vComplexD>         vColourVectorD;
+typedef iColourVector<vComplex >         vColourVector;
+typedef iColourVector<vComplexF>         vColourVectorF;
+typedef iColourVector<vComplexD>         vColourVectorD;
 
-    // SpinColourVector
-    typedef iSpinColourVector<Complex >     SpinColourVector;
-    typedef iSpinColourVector<ComplexF>     SpinColourVectorF;
-    typedef iSpinColourVector<ComplexD>     SpinColourVectorD;
+// SpinColourVector
+typedef iSpinColourVector<Complex >     SpinColourVector;
+typedef iSpinColourVector<ComplexF>     SpinColourVectorF;
+typedef iSpinColourVector<ComplexD>     SpinColourVectorD;
 
-    typedef iSpinColourVector<vComplex >     vSpinColourVector;
-    typedef iSpinColourVector<vComplexF>     vSpinColourVectorF;
-    typedef iSpinColourVector<vComplexD>     vSpinColourVectorD;
+typedef iSpinColourVector<vComplex >     vSpinColourVector;
+typedef iSpinColourVector<vComplexF>     vSpinColourVectorF;
+typedef iSpinColourVector<vComplexD>     vSpinColourVectorD;
 
-    // HalfSpin vector
-    typedef iHalfSpinVector<Complex >       HalfSpinVector;
-    typedef iHalfSpinVector<ComplexF>       HalfSpinVectorF;
-    typedef iHalfSpinVector<ComplexD>       HalfSpinVectorD;
+// HalfSpin vector
+typedef iHalfSpinVector<Complex >       HalfSpinVector;
+typedef iHalfSpinVector<ComplexF>       HalfSpinVectorF;
+typedef iHalfSpinVector<ComplexD>       HalfSpinVectorD;
 
-    typedef iHalfSpinVector<vComplex >       vHalfSpinVector;
-    typedef iHalfSpinVector<vComplexF>       vHalfSpinVectorF;
-    typedef iHalfSpinVector<vComplexD>       vHalfSpinVectorD;
+typedef iHalfSpinVector<vComplex >       vHalfSpinVector;
+typedef iHalfSpinVector<vComplexF>       vHalfSpinVectorF;
+typedef iHalfSpinVector<vComplexD>       vHalfSpinVectorD;
 
-    // HalfSpinColour vector
-    typedef iHalfSpinColourVector<Complex > HalfSpinColourVector;
-    typedef iHalfSpinColourVector<ComplexF> HalfSpinColourVectorF;
-    typedef iHalfSpinColourVector<ComplexD> HalfSpinColourVectorD;
+// HalfSpinColour vector
+typedef iHalfSpinColourVector<Complex > HalfSpinColourVector;
+typedef iHalfSpinColourVector<ComplexF> HalfSpinColourVectorF;
+typedef iHalfSpinColourVector<ComplexD> HalfSpinColourVectorD;
     
-    typedef iHalfSpinColourVector<vComplex > vHalfSpinColourVector;
-    typedef iHalfSpinColourVector<vComplexF> vHalfSpinColourVectorF;
-    typedef iHalfSpinColourVector<vComplexD> vHalfSpinColourVectorD;
+typedef iHalfSpinColourVector<vComplex > vHalfSpinColourVector;
+typedef iHalfSpinColourVector<vComplexF> vHalfSpinColourVectorF;
+typedef iHalfSpinColourVector<vComplexD> vHalfSpinColourVectorD;
     
-    // singlets
-    typedef iSinglet<Complex >         TComplex;     // FIXME This is painful. Tensor singlet complex type.
-    typedef iSinglet<ComplexF>         TComplexF;    // FIXME This is painful. Tensor singlet complex type.
-    typedef iSinglet<ComplexD>         TComplexD;    // FIXME This is painful. Tensor singlet complex type.
+// singlets
+typedef iSinglet<Complex >         TComplex;     // FIXME This is painful. Tensor singlet complex type.
+typedef iSinglet<ComplexF>         TComplexF;    // FIXME This is painful. Tensor singlet complex type.
+typedef iSinglet<ComplexD>         TComplexD;    // FIXME This is painful. Tensor singlet complex type.
 
-    typedef iSinglet<vComplex >        vTComplex ;   // what if we don't know the tensor structure
-    typedef iSinglet<vComplexF>        vTComplexF;   // what if we don't know the tensor structure
-    typedef iSinglet<vComplexD>        vTComplexD;   // what if we don't know the tensor structure
+typedef iSinglet<vComplex >        vTComplex ;   // what if we don't know the tensor structure
+typedef iSinglet<vComplexF>        vTComplexF;   // what if we don't know the tensor structure
+typedef iSinglet<vComplexD>        vTComplexD;   // what if we don't know the tensor structure
 
-    typedef iSinglet<Real >            TReal;        // Shouldn't need these; can I make it work without?
-    typedef iSinglet<RealF>            TRealF;       // Shouldn't need these; can I make it work without?
-    typedef iSinglet<RealD>            TRealD;       // Shouldn't need these; can I make it work without?
+typedef iSinglet<Real >            TReal;        // Shouldn't need these; can I make it work without?
+typedef iSinglet<RealF>            TRealF;       // Shouldn't need these; can I make it work without?
+typedef iSinglet<RealD>            TRealD;       // Shouldn't need these; can I make it work without?
 
-    typedef iSinglet<vReal >           vTReal;      
-    typedef iSinglet<vRealF>           vTRealF;      
-    typedef iSinglet<vRealD>           vTRealD;      
+typedef iSinglet<vReal >           vTReal;      
+typedef iSinglet<vRealF>           vTRealF;      
+typedef iSinglet<vRealD>           vTRealD;      
 
-    typedef iSinglet<vInteger>         vTInteger;
-    typedef iSinglet<Integer >         TInteger;
+typedef iSinglet<vInteger>         vTInteger;
+typedef iSinglet<Integer >         TInteger;
 
 
-    // Lattices of these
-    typedef Lattice<vColourMatrix>          LatticeColourMatrix;
-    typedef Lattice<vColourMatrixF>         LatticeColourMatrixF;
-    typedef Lattice<vColourMatrixD>         LatticeColourMatrixD;
+// Lattices of these
+typedef Lattice<vColourMatrix>          LatticeColourMatrix;
+typedef Lattice<vColourMatrixF>         LatticeColourMatrixF;
+typedef Lattice<vColourMatrixD>         LatticeColourMatrixD;
 
-    typedef Lattice<vSpinMatrix>            LatticeSpinMatrix;
-    typedef Lattice<vSpinMatrixF>           LatticeSpinMatrixF;
-    typedef Lattice<vSpinMatrixD>           LatticeSpinMatrixD;
+typedef Lattice<vSpinMatrix>            LatticeSpinMatrix;
+typedef Lattice<vSpinMatrixF>           LatticeSpinMatrixF;
+typedef Lattice<vSpinMatrixD>           LatticeSpinMatrixD;
 
-    typedef Lattice<vSpinColourMatrix>      LatticeSpinColourMatrix;
-    typedef Lattice<vSpinColourMatrixF>     LatticeSpinColourMatrixF;
-    typedef Lattice<vSpinColourMatrixD>     LatticeSpinColourMatrixD;
+typedef Lattice<vSpinColourMatrix>      LatticeSpinColourMatrix;
+typedef Lattice<vSpinColourMatrixF>     LatticeSpinColourMatrixF;
+typedef Lattice<vSpinColourMatrixD>     LatticeSpinColourMatrixD;
 
-    typedef Lattice<vSpinColourSpinColourMatrix>      LatticeSpinColourSpinColourMatrix;
-    typedef Lattice<vSpinColourSpinColourMatrixF>     LatticeSpinColourSpinColourMatrixF;
-    typedef Lattice<vSpinColourSpinColourMatrixD>     LatticeSpinColourSpinColourMatrixD;
+typedef Lattice<vSpinColourSpinColourMatrix>      LatticeSpinColourSpinColourMatrix;
+typedef Lattice<vSpinColourSpinColourMatrixF>     LatticeSpinColourSpinColourMatrixF;
+typedef Lattice<vSpinColourSpinColourMatrixD>     LatticeSpinColourSpinColourMatrixD;
 
-    typedef Lattice<vLorentzColourMatrix>  LatticeLorentzColourMatrix;
-    typedef Lattice<vLorentzColourMatrixF> LatticeLorentzColourMatrixF;
-    typedef Lattice<vLorentzColourMatrixD> LatticeLorentzColourMatrixD;
+typedef Lattice<vLorentzColourMatrix>  LatticeLorentzColourMatrix;
+typedef Lattice<vLorentzColourMatrixF> LatticeLorentzColourMatrixF;
+typedef Lattice<vLorentzColourMatrixD> LatticeLorentzColourMatrixD;
 
-    // DoubleStored gauge field
-    typedef Lattice<vDoubleStoredColourMatrix>  LatticeDoubleStoredColourMatrix;
-    typedef Lattice<vDoubleStoredColourMatrixF> LatticeDoubleStoredColourMatrixF;
-    typedef Lattice<vDoubleStoredColourMatrixD> LatticeDoubleStoredColourMatrixD;
+// DoubleStored gauge field
+typedef Lattice<vDoubleStoredColourMatrix>  LatticeDoubleStoredColourMatrix;
+typedef Lattice<vDoubleStoredColourMatrixF> LatticeDoubleStoredColourMatrixF;
+typedef Lattice<vDoubleStoredColourMatrixD> LatticeDoubleStoredColourMatrixD;
 
-    typedef Lattice<vSpinVector>            LatticeSpinVector;
-    typedef Lattice<vSpinVectorF>           LatticeSpinVectorF;
-    typedef Lattice<vSpinVectorD>           LatticeSpinVectorD;
+typedef Lattice<vSpinVector>            LatticeSpinVector;
+typedef Lattice<vSpinVectorF>           LatticeSpinVectorF;
+typedef Lattice<vSpinVectorD>           LatticeSpinVectorD;
 
-    typedef Lattice<vColourVector>          LatticeColourVector;
-    typedef Lattice<vColourVectorF>         LatticeColourVectorF;
-    typedef Lattice<vColourVectorD>         LatticeColourVectorD;
+typedef Lattice<vColourVector>          LatticeColourVector;
+typedef Lattice<vColourVectorF>         LatticeColourVectorF;
+typedef Lattice<vColourVectorD>         LatticeColourVectorD;
 
-    typedef Lattice<vSpinColourVector>      LatticeSpinColourVector;
-    typedef Lattice<vSpinColourVectorF>     LatticeSpinColourVectorF;
-    typedef Lattice<vSpinColourVectorD>     LatticeSpinColourVectorD;
+typedef Lattice<vSpinColourVector>      LatticeSpinColourVector;
+typedef Lattice<vSpinColourVectorF>     LatticeSpinColourVectorF;
+typedef Lattice<vSpinColourVectorD>     LatticeSpinColourVectorD;
 
-    typedef Lattice<vHalfSpinVector>        LatticeHalfSpinVector;
-    typedef Lattice<vHalfSpinVectorF>       LatticeHalfSpinVectorF;
-    typedef Lattice<vHalfSpinVectorD>       LatticeHalfSpinVectorD;
+typedef Lattice<vHalfSpinVector>        LatticeHalfSpinVector;
+typedef Lattice<vHalfSpinVectorF>       LatticeHalfSpinVectorF;
+typedef Lattice<vHalfSpinVectorD>       LatticeHalfSpinVectorD;
 
-    typedef Lattice<vHalfSpinColourVector>  LatticeHalfSpinColourVector;
-    typedef Lattice<vHalfSpinColourVectorF> LatticeHalfSpinColourVectorF;
-    typedef Lattice<vHalfSpinColourVectorD> LatticeHalfSpinColourVectorD;
+typedef Lattice<vHalfSpinColourVector>  LatticeHalfSpinColourVector;
+typedef Lattice<vHalfSpinColourVectorF> LatticeHalfSpinColourVectorF;
+typedef Lattice<vHalfSpinColourVectorD> LatticeHalfSpinColourVectorD;
 
-    typedef Lattice<vTReal>            LatticeReal;
-    typedef Lattice<vTRealF>           LatticeRealF;
-    typedef Lattice<vTRealD>           LatticeRealD;
+typedef Lattice<vTReal>            LatticeReal;
+typedef Lattice<vTRealF>           LatticeRealF;
+typedef Lattice<vTRealD>           LatticeRealD;
 
-    typedef Lattice<vTComplex>         LatticeComplex;
-    typedef Lattice<vTComplexF>        LatticeComplexF;
-    typedef Lattice<vTComplexD>        LatticeComplexD;
+typedef Lattice<vTComplex>         LatticeComplex;
+typedef Lattice<vTComplexF>        LatticeComplexF;
+typedef Lattice<vTComplexD>        LatticeComplexD;
 
-    typedef Lattice<vTInteger>         LatticeInteger; // Predicates for "where"
+typedef Lattice<vTInteger>         LatticeInteger; // Predicates for "where"
 
 
-    ///////////////////////////////////////////
-    // Physical names for things
-    ///////////////////////////////////////////
-    typedef LatticeHalfSpinColourVector  LatticeHalfFermion;
-    typedef LatticeHalfSpinColourVectorF LatticeHalfFermionF;
-    typedef LatticeHalfSpinColourVectorF LatticeHalfFermionD;
+///////////////////////////////////////////
+// Physical names for things
+///////////////////////////////////////////
+typedef LatticeHalfSpinColourVector  LatticeHalfFermion;
+typedef LatticeHalfSpinColourVectorF LatticeHalfFermionF;
+typedef LatticeHalfSpinColourVectorF LatticeHalfFermionD;
 
-    typedef LatticeSpinColourVector      LatticeFermion;
-    typedef LatticeSpinColourVectorF     LatticeFermionF;
-    typedef LatticeSpinColourVectorD     LatticeFermionD;
+typedef LatticeSpinColourVector      LatticeFermion;
+typedef LatticeSpinColourVectorF     LatticeFermionF;
+typedef LatticeSpinColourVectorD     LatticeFermionD;
 
-    typedef LatticeSpinColourMatrix                LatticePropagator;
-    typedef LatticeSpinColourMatrixF               LatticePropagatorF;
-    typedef LatticeSpinColourMatrixD               LatticePropagatorD;
+typedef LatticeSpinColourMatrix                LatticePropagator;
+typedef LatticeSpinColourMatrixF               LatticePropagatorF;
+typedef LatticeSpinColourMatrixD               LatticePropagatorD;
 
-    typedef LatticeLorentzColourMatrix             LatticeGaugeField;
-    typedef LatticeLorentzColourMatrixF            LatticeGaugeFieldF;
-    typedef LatticeLorentzColourMatrixD            LatticeGaugeFieldD;
+typedef LatticeLorentzColourMatrix             LatticeGaugeField;
+typedef LatticeLorentzColourMatrixF            LatticeGaugeFieldF;
+typedef LatticeLorentzColourMatrixD            LatticeGaugeFieldD;
 
-    typedef LatticeDoubleStoredColourMatrix        LatticeDoubledGaugeField;
-    typedef LatticeDoubleStoredColourMatrixF       LatticeDoubledGaugeFieldF;
-    typedef LatticeDoubleStoredColourMatrixD       LatticeDoubledGaugeFieldD;
+typedef LatticeDoubleStoredColourMatrix        LatticeDoubledGaugeField;
+typedef LatticeDoubleStoredColourMatrixF       LatticeDoubledGaugeFieldF;
+typedef LatticeDoubleStoredColourMatrixD       LatticeDoubledGaugeFieldD;
 
-    template<class GF> using LorentzScalar = Lattice<iScalar<typename GF::vector_object::element> >;
+template<class GF> using LorentzScalar = Lattice<iScalar<typename GF::vector_object::element> >;
 
-    // Uhgg... typing this hurt  ;)
-    // (my keyboard got burning hot when I typed this, must be the anti-Fermion)
-    typedef Lattice<vColourVector>          LatticeStaggeredFermion;    
-    typedef Lattice<vColourVectorF>         LatticeStaggeredFermionF;    
-    typedef Lattice<vColourVectorD>         LatticeStaggeredFermionD;    
+// Uhgg... typing this hurt  ;)
+// (my keyboard got burning hot when I typed this, must be the anti-Fermion)
+typedef Lattice<vColourVector>          LatticeStaggeredFermion;    
+typedef Lattice<vColourVectorF>         LatticeStaggeredFermionF;    
+typedef Lattice<vColourVectorD>         LatticeStaggeredFermionD;    
 
-    typedef Lattice<vColourMatrix>          LatticeStaggeredPropagator; 
-    typedef Lattice<vColourMatrixF>         LatticeStaggeredPropagatorF; 
-    typedef Lattice<vColourMatrixD>         LatticeStaggeredPropagatorD; 
+typedef Lattice<vColourMatrix>          LatticeStaggeredPropagator; 
+typedef Lattice<vColourMatrixF>         LatticeStaggeredPropagatorF; 
+typedef Lattice<vColourMatrixD>         LatticeStaggeredPropagatorD; 
 
-    //////////////////////////////////////////////////////////////////////////////
-    // Peek and Poke named after physics attributes
-    //////////////////////////////////////////////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////////
+// Peek and Poke named after physics attributes
+//////////////////////////////////////////////////////////////////////////////
 
-    //spin
-    template<class vobj> auto peekSpin(const vobj &rhs,int i) -> decltype(PeekIndex<SpinIndex>(rhs,0))
-    {
-      return PeekIndex<SpinIndex>(rhs,i);
-    }
-    template<class vobj> auto peekSpin(const vobj &rhs,int i,int j) -> decltype(PeekIndex<SpinIndex>(rhs,0,0))
-    {
-      return PeekIndex<SpinIndex>(rhs,i,j);
-    }
-    template<class vobj> auto peekSpin(const Lattice<vobj> &rhs,int i) -> decltype(PeekIndex<SpinIndex>(rhs,0))
-    {
-      return PeekIndex<SpinIndex>(rhs,i);
-    }
-    template<class vobj> auto peekSpin(const Lattice<vobj> &rhs,int i,int j) -> decltype(PeekIndex<SpinIndex>(rhs,0,0))
-    {
-      return PeekIndex<SpinIndex>(rhs,i,j);
-    }
-    //colour
-    template<class vobj> auto peekColour(const vobj &rhs,int i) -> decltype(PeekIndex<ColourIndex>(rhs,0))
-    {
-      return PeekIndex<ColourIndex>(rhs,i);
-    }
-    template<class vobj> auto peekColour(const vobj &rhs,int i,int j) -> decltype(PeekIndex<ColourIndex>(rhs,0,0))
-    {
-      return PeekIndex<ColourIndex>(rhs,i,j);
-    }
-    template<class vobj> auto peekColour(const Lattice<vobj> &rhs,int i) -> decltype(PeekIndex<ColourIndex>(rhs,0))
-    {
-      return PeekIndex<ColourIndex>(rhs,i);
-    }
-    template<class vobj> auto peekColour(const Lattice<vobj> &rhs,int i,int j) -> decltype(PeekIndex<ColourIndex>(rhs,0,0))
-    {
-      return PeekIndex<ColourIndex>(rhs,i,j);
-    }
-    //lorentz
-    template<class vobj> auto peekLorentz(const vobj &rhs,int i) -> decltype(PeekIndex<LorentzIndex>(rhs,0))
-    {
-      return PeekIndex<LorentzIndex>(rhs,i);
-    }
-    template<class vobj> auto peekLorentz(const Lattice<vobj> &rhs,int i) -> decltype(PeekIndex<LorentzIndex>(rhs,0))
-    {
-      return PeekIndex<LorentzIndex>(rhs,i);
-    }
+//spin
+template<class vobj> auto peekSpin(const vobj &rhs,int i) -> decltype(PeekIndex<SpinIndex>(rhs,0))
+{
+  return PeekIndex<SpinIndex>(rhs,i);
+}
+template<class vobj> auto peekSpin(const vobj &rhs,int i,int j) -> decltype(PeekIndex<SpinIndex>(rhs,0,0))
+{
+  return PeekIndex<SpinIndex>(rhs,i,j);
+}
+template<class vobj> auto peekSpin(const Lattice<vobj> &rhs,int i) -> decltype(PeekIndex<SpinIndex>(rhs,0))
+{
+  return PeekIndex<SpinIndex>(rhs,i);
+}
+template<class vobj> auto peekSpin(const Lattice<vobj> &rhs,int i,int j) -> decltype(PeekIndex<SpinIndex>(rhs,0,0))
+{
+  return PeekIndex<SpinIndex>(rhs,i,j);
+}
+//colour
+template<class vobj> auto peekColour(const vobj &rhs,int i) -> decltype(PeekIndex<ColourIndex>(rhs,0))
+{
+  return PeekIndex<ColourIndex>(rhs,i);
+}
+template<class vobj> auto peekColour(const vobj &rhs,int i,int j) -> decltype(PeekIndex<ColourIndex>(rhs,0,0))
+{
+  return PeekIndex<ColourIndex>(rhs,i,j);
+}
+template<class vobj> auto peekColour(const Lattice<vobj> &rhs,int i) -> decltype(PeekIndex<ColourIndex>(rhs,0))
+{
+  return PeekIndex<ColourIndex>(rhs,i);
+}
+template<class vobj> auto peekColour(const Lattice<vobj> &rhs,int i,int j) -> decltype(PeekIndex<ColourIndex>(rhs,0,0))
+{
+  return PeekIndex<ColourIndex>(rhs,i,j);
+}
+//lorentz
+template<class vobj> auto peekLorentz(const vobj &rhs,int i) -> decltype(PeekIndex<LorentzIndex>(rhs,0))
+{
+  return PeekIndex<LorentzIndex>(rhs,i);
+}
+template<class vobj> auto peekLorentz(const Lattice<vobj> &rhs,int i) -> decltype(PeekIndex<LorentzIndex>(rhs,0))
+{
+  return PeekIndex<LorentzIndex>(rhs,i);
+}
 
-    //////////////////////////////////////////////
-    // Poke lattice
-    //////////////////////////////////////////////
-    template<class vobj> 
-      void pokeColour(Lattice<vobj> &lhs,
-              const Lattice<decltype(peekIndex<ColourIndex>(lhs._odata[0],0))> & rhs,
+//////////////////////////////////////////////
+// Poke lattice
+//////////////////////////////////////////////
+template<class vobj> 
+void pokeColour(Lattice<vobj> &lhs,
+		const Lattice<decltype(peekIndex<ColourIndex>(vobj(),0))> & rhs,
+		int i)
+{
+  PokeIndex<ColourIndex>(lhs,rhs,i);
+}
+template<class vobj> 
+void pokeColour(Lattice<vobj> &lhs,
+		const Lattice<decltype(peekIndex<ColourIndex>(vobj(),0,0))> & rhs,
+		int i,int j)
+{
+  PokeIndex<ColourIndex>(lhs,rhs,i,j);
+}
+template<class vobj> 
+void pokeSpin(Lattice<vobj> &lhs,
+              const Lattice<decltype(peekIndex<SpinIndex>(vobj(),0))> & rhs,
               int i)
-    {
-      PokeIndex<ColourIndex>(lhs,rhs,i);
-    }
-    template<class vobj> 
-      void pokeColour(Lattice<vobj> &lhs,
-              const Lattice<decltype(peekIndex<ColourIndex>(lhs._odata[0],0,0))> & rhs,
+{
+  PokeIndex<SpinIndex>(lhs,rhs,i);
+}
+template<class vobj> 
+void pokeSpin(Lattice<vobj> &lhs,
+              const Lattice<decltype(peekIndex<SpinIndex>(vobj(),0,0))> & rhs,
               int i,int j)
-    {
-      PokeIndex<ColourIndex>(lhs,rhs,i,j);
-    }
-    template<class vobj> 
-      void pokeSpin(Lattice<vobj> &lhs,
-              const Lattice<decltype(peekIndex<SpinIndex>(lhs._odata[0],0))> & rhs,
-              int i)
-    {
-      PokeIndex<SpinIndex>(lhs,rhs,i);
-    }
-    template<class vobj> 
-      void pokeSpin(Lattice<vobj> &lhs,
-              const Lattice<decltype(peekIndex<SpinIndex>(lhs._odata[0],0,0))> & rhs,
-              int i,int j)
-    {
-      PokeIndex<SpinIndex>(lhs,rhs,i,j);
-    }
-    template<class vobj> 
-      void pokeLorentz(Lattice<vobj> &lhs,
-              const Lattice<decltype(peekIndex<LorentzIndex>(lhs._odata[0],0))> & rhs,
-              int i)
-    {
-      PokeIndex<LorentzIndex>(lhs,rhs,i);
-    }
+{
+  PokeIndex<SpinIndex>(lhs,rhs,i,j);
+}
+template<class vobj> 
+void pokeLorentz(Lattice<vobj> &lhs,
+		 const Lattice<decltype(peekIndex<LorentzIndex>(vobj(),0))> & rhs,
+		 int i)
+{
+  PokeIndex<LorentzIndex>(lhs,rhs,i);
+}
 
-    //////////////////////////////////////////////
-    // Poke scalars
-    //////////////////////////////////////////////
-    template<class vobj> void pokeSpin(vobj &lhs,const decltype(peekIndex<SpinIndex>(lhs,0)) & rhs,int i)
-    {
-      pokeIndex<SpinIndex>(lhs,rhs,i);
-    }
-    template<class vobj> void pokeSpin(vobj &lhs,const decltype(peekIndex<SpinIndex>(lhs,0,0)) & rhs,int i,int j)
-    {
-      pokeIndex<SpinIndex>(lhs,rhs,i,j);
-    }
+//////////////////////////////////////////////
+// Poke scalars
+//////////////////////////////////////////////
+template<class vobj> void pokeSpin(vobj &lhs,const decltype(peekIndex<SpinIndex>(lhs,0)) & rhs,int i)
+{
+  pokeIndex<SpinIndex>(lhs,rhs,i);
+}
+template<class vobj> void pokeSpin(vobj &lhs,const decltype(peekIndex<SpinIndex>(lhs,0,0)) & rhs,int i,int j)
+{
+  pokeIndex<SpinIndex>(lhs,rhs,i,j);
+}
 
-    template<class vobj> void pokeColour(vobj &lhs,const decltype(peekIndex<ColourIndex>(lhs,0)) & rhs,int i)
-    {
-      pokeIndex<ColourIndex>(lhs,rhs,i);
-    }
-    template<class vobj> void pokeColour(vobj &lhs,const decltype(peekIndex<ColourIndex>(lhs,0,0)) & rhs,int i,int j)
-    {
-      pokeIndex<ColourIndex>(lhs,rhs,i,j);
-    }
+template<class vobj> void pokeColour(vobj &lhs,const decltype(peekIndex<ColourIndex>(lhs,0)) & rhs,int i)
+{
+  pokeIndex<ColourIndex>(lhs,rhs,i);
+}
+template<class vobj> void pokeColour(vobj &lhs,const decltype(peekIndex<ColourIndex>(lhs,0,0)) & rhs,int i,int j)
+{
+  pokeIndex<ColourIndex>(lhs,rhs,i,j);
+}
 
-    template<class vobj> void pokeLorentz(vobj &lhs,const decltype(peekIndex<LorentzIndex>(lhs,0)) & rhs,int i)
-    {
-      pokeIndex<LorentzIndex>(lhs,rhs,i);
-    }
+template<class vobj> void pokeLorentz(vobj &lhs,const decltype(peekIndex<LorentzIndex>(lhs,0)) & rhs,int i)
+{
+  pokeIndex<LorentzIndex>(lhs,rhs,i);
+}
 
-    //////////////////////////////////////////////
-    // Fermion <-> propagator assignements
-    //////////////////////////////////////////////
+//////////////////////////////////////////////
+// Fermion <-> propagator assignements
+//////////////////////////////////////////////
     //template <class Prop, class Ferm>
     template <class Fimpl>
       void FermToProp(typename Fimpl::PropagatorField &p, const typename Fimpl::FermionField &f, const int s, const int c)
+{
+  for(int j = 0; j < Ns; ++j)
     {
-      for(int j = 0; j < Ns; ++j)
-        {
-            auto pjs = peekSpin(p, j, s);
-            auto fj  = peekSpin(f, j);
+      auto pjs = peekSpin(p, j, s);
+      auto fj  = peekSpin(f, j);
             
             for(int i = 0; i < Fimpl::Dimension; ++i)
-            {
-                pokeColour(pjs, peekColour(fj, i), i, c);
-            }
-            pokeSpin(p, pjs, j, s);
-        }
+	{
+	  pokeColour(pjs, peekColour(fj, i), i, c);
+	}
+      pokeSpin(p, pjs, j, s);
     }
+}
     
     //template <class Prop, class Ferm>
     template <class Fimpl>
       void PropToFerm(typename Fimpl::FermionField &f, const typename Fimpl::PropagatorField &p, const int s, const int c)
+{
+  for(int j = 0; j < Ns; ++j)
     {
-        for(int j = 0; j < Ns; ++j)
-        {
-            auto pjs = peekSpin(p, j, s);
-            auto fj  = peekSpin(f, j);
+      auto pjs = peekSpin(p, j, s);
+      auto fj  = peekSpin(f, j);
             
             for(int i = 0; i < Fimpl::Dimension; ++i)
-            {
-                pokeColour(fj, peekColour(pjs, i, c), i);
-            }
-            pokeSpin(f, fj, j);
-        }
+	{
+	  pokeColour(fj, peekColour(pjs, i, c), i);
+	}
+      pokeSpin(f, fj, j);
     }
+}
     
-    //////////////////////////////////////////////
-    // transpose array and scalar
-    //////////////////////////////////////////////
-    template<int Index,class vobj> inline Lattice<vobj> transposeSpin(const Lattice<vobj> &lhs){
-      return transposeIndex<SpinIndex>(lhs);
-    }
-    template<int Index,class vobj> inline Lattice<vobj> transposeColour(const Lattice<vobj> &lhs){
-      return transposeIndex<ColourIndex>(lhs);
-    }
-    template<int Index,class vobj> inline vobj transposeSpin(const vobj &lhs){
-      return transposeIndex<SpinIndex>(lhs);
-    }
-    template<int Index,class vobj> inline vobj transposeColour(const vobj &lhs){
-      return transposeIndex<ColourIndex>(lhs);
-    }
+//////////////////////////////////////////////
+// transpose array and scalar
+//////////////////////////////////////////////
+template<int Index,class vobj> inline Lattice<vobj> transposeSpin(const Lattice<vobj> &lhs){
+  return transposeIndex<SpinIndex>(lhs);
+}
+template<int Index,class vobj> inline Lattice<vobj> transposeColour(const Lattice<vobj> &lhs){
+  return transposeIndex<ColourIndex>(lhs);
+}
+template<int Index,class vobj> inline vobj transposeSpin(const vobj &lhs){
+  return transposeIndex<SpinIndex>(lhs);
+}
+template<int Index,class vobj> inline vobj transposeColour(const vobj &lhs){
+  return transposeIndex<ColourIndex>(lhs);
+}
 
-    //////////////////////////////////////////
-    // Trace lattice and non-lattice
-    //////////////////////////////////////////
-    template<int Index,class vobj>
-    inline auto traceSpin(const Lattice<vobj> &lhs) -> Lattice<decltype(traceIndex<SpinIndex>(lhs._odata[0]))>
-    {
-      return traceIndex<SpinIndex>(lhs);
-    }
-    template<int Index,class vobj>
-    inline auto traceColour(const Lattice<vobj> &lhs) -> Lattice<decltype(traceIndex<ColourIndex>(lhs._odata[0]))>
-    {
-      return traceIndex<ColourIndex>(lhs);
-    }
-    template<int Index,class vobj>
-    inline auto traceSpin(const vobj &lhs) -> Lattice<decltype(traceIndex<SpinIndex>(lhs))>
-    {
-      return traceIndex<SpinIndex>(lhs);
-    }
-    template<int Index,class vobj>
-    inline auto traceColour(const vobj &lhs) -> Lattice<decltype(traceIndex<ColourIndex>(lhs))>
-    {
-      return traceIndex<ColourIndex>(lhs);
-    }
+//////////////////////////////////////////
+// Trace lattice and non-lattice
+//////////////////////////////////////////
+template<int Index,class vobj>
+inline auto traceSpin(const Lattice<vobj> &lhs) -> Lattice<decltype(traceIndex<SpinIndex>(vobj()))>
+{
+  return traceIndex<SpinIndex>(lhs);
+}
+template<int Index,class vobj>
+inline auto traceColour(const Lattice<vobj> &lhs) -> Lattice<decltype(traceIndex<ColourIndex>(vobj()))>
+{
+  return traceIndex<ColourIndex>(lhs);
+}
+template<int Index,class vobj>
+inline auto traceSpin(const vobj &lhs) -> Lattice<decltype(traceIndex<SpinIndex>(lhs))>
+{
+  return traceIndex<SpinIndex>(lhs);
+}
+template<int Index,class vobj>
+inline auto traceColour(const vobj &lhs) -> Lattice<decltype(traceIndex<ColourIndex>(lhs))>
+{
+  return traceIndex<ColourIndex>(lhs);
+}
 
-    //////////////////////////////////////////
-    // Current types
-    //////////////////////////////////////////
-    GRID_SERIALIZABLE_ENUM(Current, undef,
-                           Vector,  0,
-                           Axial,   1,
-                           Tadpole, 2);
+//////////////////////////////////////////
+// Current types
+//////////////////////////////////////////
+GRID_SERIALIZABLE_ENUM(Current, undef,
+		       Vector,  0,
+		       Axial,   1,
+		       Tadpole, 2);
 
-}   //namespace QCD
-} // Grid
+NAMESPACE_END(Grid);
 
-
-
-#endif
diff --git a/Grid/qcd/action/Action.h b/Grid/qcd/action/Action.h
index 7272c90d..737c1ff0 100644
--- a/Grid/qcd/action/Action.h
+++ b/Grid/qcd/action/Action.h
@@ -37,14 +37,18 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 // Abstract base interface
 ////////////////////////////////////////////
 #include <Grid/qcd/action/ActionCore.h>
+NAMESPACE_CHECK(ActionCore);
 ////////////////////////////////////////////////////////////////////////
 // Fermion actions; prevent coupling fermion.cc files to other headers
 ////////////////////////////////////////////////////////////////////////
 #include <Grid/qcd/action/fermion/FermionCore.h>
+NAMESPACE_CHECK(FermionCore);
 #include <Grid/qcd/action/fermion/Fermion.h>
+NAMESPACE_CHECK(Fermion);
 ////////////////////////////////////////
 // Pseudo fermion combinations for HMC
 ////////////////////////////////////////
 #include <Grid/qcd/action/pseudofermion/PseudoFermion.h>
+NAMESPACE_CHECK(PseudoFermion);
 
 #endif
diff --git a/Grid/qcd/action/ActionBase.h b/Grid/qcd/action/ActionBase.h
index 8d853d45..bff21d1d 100644
--- a/Grid/qcd/action/ActionBase.h
+++ b/Grid/qcd/action/ActionBase.h
@@ -27,19 +27,18 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 
 #ifndef ACTION_BASE_H
 #define ACTION_BASE_H
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 template <class GaugeField >
 class Action 
 {
 
- public:
+public:
   bool is_smeared = false;
   // Heatbath?
   virtual void refresh(const GaugeField& U, GridParallelRNG& pRNG) = 0; // refresh pseudofermions
@@ -50,7 +49,6 @@ class Action
   virtual ~Action(){}
 };
 
-}
-}
+NAMESPACE_END(Grid);
 
 #endif // ACTION_BASE_H
diff --git a/Grid/qcd/action/ActionCore.h b/Grid/qcd/action/ActionCore.h
index 7a5caf15..6544318d 100644
--- a/Grid/qcd/action/ActionCore.h
+++ b/Grid/qcd/action/ActionCore.h
@@ -31,29 +31,37 @@ directory
 #define QCD_ACTION_CORE
 
 #include <Grid/qcd/action/ActionBase.h>
+NAMESPACE_CHECK(ActionBase);
 #include <Grid/qcd/action/ActionSet.h>
+NAMESPACE_CHECK(ActionSet);
 #include <Grid/qcd/action/ActionParams.h>
+NAMESPACE_CHECK(ActionParams);
 
 ////////////////////////////////////////////
 // Gauge Actions
 ////////////////////////////////////////////
 #include <Grid/qcd/action/gauge/Gauge.h>
+NAMESPACE_CHECK(Gauge);
 
 ////////////////////////////////////////////
 // Fermion prereqs
 ////////////////////////////////////////////
 #include <Grid/qcd/action/fermion/FermionCore.h>
+NAMESPACE_CHECK(ActionFermionCore);
 
 ////////////////////////////////////////////
 // Scalar Actions
 ////////////////////////////////////////////
 #include <Grid/qcd/action/scalar/Scalar.h>
+NAMESPACE_CHECK(Scalar);
 
 ////////////////////////////////////////////
 // Utility functions
 ////////////////////////////////////////////
 #include <Grid/qcd/utils/Metric.h>
+NAMESPACE_CHECK(Metric);
 #include <Grid/qcd/utils/CovariantLaplacian.h>
+NAMESPACE_CHECK(CovariantLaplacian);
 
 
 
diff --git a/Grid/qcd/action/ActionParams.h b/Grid/qcd/action/ActionParams.h
index 88de777d..0e6a11c6 100644
--- a/Grid/qcd/action/ActionParams.h
+++ b/Grid/qcd/action/ActionParams.h
@@ -27,37 +27,35 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 
 #ifndef GRID_QCD_ACTION_PARAMS_H
 #define GRID_QCD_ACTION_PARAMS_H
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
-  // These can move into a params header and be given MacroMagic serialisation
-  struct GparityWilsonImplParams {
-    bool overlapCommsCompute;
-    std::vector<int> twists;
-    GparityWilsonImplParams() : twists(Nd, 0), overlapCommsCompute(false){};
-  };
+// These can move into a params header and be given MacroMagic serialisation
+struct GparityWilsonImplParams {
+  Coordinate twists;
+  GparityWilsonImplParams() : twists(Nd, 0) {};
+};
   
-  struct WilsonImplParams {
-    bool overlapCommsCompute;
-    std::vector<Real> twist_n_2pi_L;
-    std::vector<Complex> boundary_phases;
-    WilsonImplParams() : overlapCommsCompute(false) {
-      boundary_phases.resize(Nd, 1.0);
+struct WilsonImplParams {
+  bool overlapCommsCompute;
+  AcceleratorVector<Real,Nd> twist_n_2pi_L;
+  AcceleratorVector<Complex,Nd> boundary_phases;
+  WilsonImplParams()  {
+    boundary_phases.resize(Nd, 1.0);
       twist_n_2pi_L.resize(Nd, 0.0);
-    };
-    WilsonImplParams(const std::vector<Complex> phi) : boundary_phases(phi), overlapCommsCompute(false) {
-      twist_n_2pi_L.resize(Nd, 0.0);
-    }
   };
+  WilsonImplParams(const AcceleratorVector<Complex,Nd> phi) : boundary_phases(phi), overlapCommsCompute(false) {
+    twist_n_2pi_L.resize(Nd, 0.0);
+  }
+};
 
-  struct StaggeredImplParams {
-    StaggeredImplParams()  {};
-  };
+struct StaggeredImplParams {
+  StaggeredImplParams()  {};
+};
   
   struct OneFlavourRationalParams : Serializable {
     GRID_SERIALIZABLE_CLASS_MEMBERS(OneFlavourRationalParams, 
@@ -69,10 +67,10 @@ namespace QCD {
 				    int,   precision,
 				    int,   BoundsCheckFreq);
     
-    // MaxIter and tolerance, vectors??
+  // MaxIter and tolerance, vectors??
     
-    // constructor 
-    OneFlavourRationalParams(	RealD _lo      = 0.0, 
+  // constructor 
+  OneFlavourRationalParams(	RealD _lo      = 0.0, 
 				RealD _hi      = 1.0, 
 				int _maxit     = 1000,
 				RealD tol      = 1.0e-8, 
@@ -88,11 +86,6 @@ namespace QCD {
         BoundsCheckFreq(_BoundsCheckFreq){};
   };
   
-  
-}
-}
-
-
-
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/ActionSet.h b/Grid/qcd/action/ActionSet.h
index 4ed6a582..e6879fe5 100644
--- a/Grid/qcd/action/ActionSet.h
+++ b/Grid/qcd/action/ActionSet.h
@@ -26,14 +26,11 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef ACTION_SET_H
 #define ACTION_SET_H
 
-namespace Grid {
-
-// Should drop this namespace here
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 //////////////////////////////////
 // Indexing of tuple types
@@ -62,7 +59,7 @@ struct Index<T, std::tuple<U, Types...>> {
 
 template <class Field, class Repr = NoHirep >
 struct ActionLevel {
- public:
+public:
   unsigned int multiplier;
 
   // Fundamental repr actions separated because of the smearing
@@ -77,7 +74,7 @@ struct ActionLevel {
   std::vector<ActPtr>& actions;
 
   explicit ActionLevel(unsigned int mul = 1) : 
-  actions(std::get<0>(actions_hirep)), multiplier(mul) {
+    actions(std::get<0>(actions_hirep)), multiplier(mul) {
     // initialize the hirep vectors to zero.
     // apply(this->resize, actions_hirep, 0); //need a working resize
     assert(mul >= 1);
@@ -87,7 +84,7 @@ struct ActionLevel {
   void push_back(Action<GenField>* ptr) {
     // insert only in the correct vector
     std::get< Index < GenField, action_hirep_types>::value >(actions_hirep).push_back(ptr);
-  };
+  }
 
   template <class ActPtr>
   static void resize(ActPtr ap, unsigned int n) {
@@ -110,7 +107,6 @@ struct ActionLevel {
 template <class GaugeField, class R>
 using ActionSet = std::vector<ActionLevel<GaugeField, R> >;
 
-} // QCD
-} // Grid
+NAMESPACE_END(Grid);
 
 #endif  // ACTION_SET_H
diff --git a/Grid/qcd/action/fermion/AbstractEOFAFermion.h b/Grid/qcd/action/fermion/AbstractEOFAFermion.h
index 15faa401..18bcb394 100644
--- a/Grid/qcd/action/fermion/AbstractEOFAFermion.h
+++ b/Grid/qcd/action/fermion/AbstractEOFAFermion.h
@@ -26,75 +26,75 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 
 See the full license in the file "LICENSE" in the top level distribution directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef  GRID_QCD_ABSTRACT_EOFA_FERMION_H
 #define  GRID_QCD_ABSTRACT_EOFA_FERMION_H
 
 #include <Grid/qcd/action/fermion/CayleyFermion5D.h>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
-  // DJM: Abstract base class for EOFA fermion types.
-  // Defines layout of additional EOFA-specific parameters and operators.
-  // Use to construct EOFA pseudofermion actions that are agnostic to
-  // Shamir / Mobius / etc., and ensure that no one can construct EOFA
-  // pseudofermion action with non-EOFA fermion type.
-  template<class Impl>
-  class AbstractEOFAFermion : public CayleyFermion5D<Impl> {
-    public:
-      INHERIT_IMPL_TYPES(Impl);
+// DJM: Abstract base class for EOFA fermion types.
+// Defines layout of additional EOFA-specific parameters and operators.
+// Use to construct EOFA pseudofermion actions that are agnostic to
+// Shamir / Mobius / etc., and ensure that no one can construct EOFA
+// pseudofermion action with non-EOFA fermion type.
+template<class Impl>
+class AbstractEOFAFermion : public CayleyFermion5D<Impl> {
+public:
+  INHERIT_IMPL_TYPES(Impl);
 
-    public:
-      // Fermion operator: D(mq1) + shift*\gamma_{5}*R_{5}*\Delta_{\pm}(mq2,mq3)*P_{\pm}
-      RealD mq1;
-      RealD mq2;
-      RealD mq3;
-      RealD shift;
-      int pm;
+public:
+  // Fermion operator: D(mq1) + shift*\gamma_{5}*R_{5}*\Delta_{\pm}(mq2,mq3)*P_{\pm}
+  RealD mq1;
+  RealD mq2;
+  RealD mq3;
+  RealD shift;
+  int pm;
 
-      RealD alpha; // Mobius scale
-      RealD k;     // EOFA normalization constant
+  RealD alpha; // Mobius scale
+  RealD k;     // EOFA normalization constant
 
-      virtual void Instantiatable(void) = 0;
+  virtual void Instantiatable(void) = 0;
 
-      // EOFA-specific operations
-      // Force user to implement in derived classes
-      virtual void  Omega    (const FermionField& in, FermionField& out, int sign, int dag) = 0;
-      virtual void  Dtilde   (const FermionField& in, FermionField& out) = 0;
-      virtual void  DtildeInv(const FermionField& in, FermionField& out) = 0;
+  // EOFA-specific operations
+  // Force user to implement in derived classes
+  virtual void  Omega    (const FermionField& in, FermionField& out, int sign, int dag) = 0;
+  virtual void  Dtilde   (const FermionField& in, FermionField& out) = 0;
+  virtual void  DtildeInv(const FermionField& in, FermionField& out) = 0;
 
-      // Implement derivatives in base class:
-      // for EOFA both DWF and Mobius just need d(Dw)/dU
-      virtual void MDeriv(GaugeField& mat, const FermionField& U, const FermionField& V, int dag){
-        this->DhopDeriv(mat, U, V, dag);
-      };
-      virtual void MoeDeriv(GaugeField& mat, const FermionField& U, const FermionField& V, int dag){
-        this->DhopDerivOE(mat, U, V, dag);
-      };
-      virtual void MeoDeriv(GaugeField& mat, const FermionField& U, const FermionField& V, int dag){
-        this->DhopDerivEO(mat, U, V, dag);
-      };
-
-      // Recompute 5D coefficients for different value of shift constant
-      // (needed for heatbath loop over poles)
-      virtual void RefreshShiftCoefficients(RealD new_shift) = 0;
-
-      // Constructors
-      AbstractEOFAFermion(GaugeField& _Umu, GridCartesian& FiveDimGrid, GridRedBlackCartesian& FiveDimRedBlackGrid,
-        GridCartesian& FourDimGrid, GridRedBlackCartesian& FourDimRedBlackGrid,
-        RealD _mq1, RealD _mq2, RealD _mq3, RealD _shift, int _pm,
-        RealD _M5, RealD _b, RealD _c, const ImplParams& p=ImplParams())
-        : CayleyFermion5D<Impl>(_Umu, FiveDimGrid, FiveDimRedBlackGrid, FourDimGrid, FourDimRedBlackGrid,
-          _mq1, _M5, p), mq1(_mq1), mq2(_mq2), mq3(_mq3), shift(_shift), pm(_pm)
-      {
-        int Ls = this->Ls;
-        this->alpha = _b + _c;
-        this->k = this->alpha * (_mq3-_mq2) * std::pow(this->alpha+1.0,2*Ls) /
-                    ( std::pow(this->alpha+1.0,Ls) + _mq2*std::pow(this->alpha-1.0,Ls) ) /
-                    ( std::pow(this->alpha+1.0,Ls) + _mq3*std::pow(this->alpha-1.0,Ls) );
-      };
+  // Implement derivatives in base class:
+  // for EOFA both DWF and Mobius just need d(Dw)/dU
+  virtual void MDeriv(GaugeField& mat, const FermionField& U, const FermionField& V, int dag){
+    this->DhopDeriv(mat, U, V, dag);
   };
-}}
+  virtual void MoeDeriv(GaugeField& mat, const FermionField& U, const FermionField& V, int dag){
+    this->DhopDerivOE(mat, U, V, dag);
+  };
+  virtual void MeoDeriv(GaugeField& mat, const FermionField& U, const FermionField& V, int dag){
+    this->DhopDerivEO(mat, U, V, dag);
+  };
+
+  // Recompute 5D coefficients for different value of shift constant
+  // (needed for heatbath loop over poles)
+  virtual void RefreshShiftCoefficients(RealD new_shift) = 0;
+
+  // Constructors
+  AbstractEOFAFermion(GaugeField& _Umu, GridCartesian& FiveDimGrid, GridRedBlackCartesian& FiveDimRedBlackGrid,
+		      GridCartesian& FourDimGrid, GridRedBlackCartesian& FourDimRedBlackGrid,
+		      RealD _mq1, RealD _mq2, RealD _mq3, RealD _shift, int _pm,
+		      RealD _M5, RealD _b, RealD _c, const ImplParams& p=ImplParams())
+    : CayleyFermion5D<Impl>(_Umu, FiveDimGrid, FiveDimRedBlackGrid, FourDimGrid, FourDimRedBlackGrid,
+			    _mq1, _M5, p), mq1(_mq1), mq2(_mq2), mq3(_mq3), shift(_shift), pm(_pm)
+  {
+    int Ls = this->Ls;
+    this->alpha = _b + _c;
+    this->k = this->alpha * (_mq3-_mq2) * std::pow(this->alpha+1.0,2*Ls) /
+      ( std::pow(this->alpha+1.0,Ls) + _mq2*std::pow(this->alpha-1.0,Ls) ) /
+      ( std::pow(this->alpha+1.0,Ls) + _mq3*std::pow(this->alpha-1.0,Ls) );
+  };
+};
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/fermion/CayleyFermion5D.h b/Grid/qcd/action/fermion/CayleyFermion5D.h
index 203a015e..333ba49b 100644
--- a/Grid/qcd/action/fermion/CayleyFermion5D.h
+++ b/Grid/qcd/action/fermion/CayleyFermion5D.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid
 
@@ -24,203 +24,146 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef  GRID_QCD_CAYLEY_FERMION_H
-#define  GRID_QCD_CAYLEY_FERMION_H
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
 
 #include <Grid/qcd/action/fermion/WilsonFermion5D.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class CayleyFermion5D : public WilsonFermion5D<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+public:
 
-     template<typename T> struct switcheroo   {
-       static inline int iscomplex()  { return 0; }
+  // override multiply
+  virtual RealD  M    (const FermionField &in, FermionField &out);
+  virtual RealD  Mdag (const FermionField &in, FermionField &out);
 
-       template<class vec>
-       static inline vec mult(vec a, vec b) {
-	 return real_mult(a,b);
-       }
-     };
-     template<> struct switcheroo<ComplexD> {
-       static inline int iscomplex()  { return 1; }
+  // half checkerboard operations
+  virtual void   Meooe       (const FermionField &in, FermionField &out);
+  virtual void   MeooeDag    (const FermionField &in, FermionField &out);
+  virtual void   Mooee       (const FermionField &in, FermionField &out);
+  virtual void   MooeeDag    (const FermionField &in, FermionField &out);
+  virtual void   MooeeInv    (const FermionField &in, FermionField &out);
+  virtual void   MooeeInvDag (const FermionField &in, FermionField &out);
+  virtual void   Meo5D (const FermionField &psi, FermionField &chi);
 
-       template<class vec>
-       static inline vec mult(vec a, vec b) {
-	 return a*b;
-       }
-     };
-     template<> struct switcheroo<ComplexF> {
-       static inline int iscomplex()  { return 1; }
-       template<class vec>
-       static inline vec mult(vec a, vec b) {
-	 return a*b;
-       }
-     };
+  virtual void   M5D   (const FermionField &psi, FermionField &chi);
+  virtual void   M5Ddag(const FermionField &psi, FermionField &chi);
 
+  ///////////////////////////////////////////////////////////////
+  // Physical surface field utilities
+  ///////////////////////////////////////////////////////////////
+  virtual void Dminus(const FermionField &psi, FermionField &chi);
+  virtual void DminusDag(const FermionField &psi, FermionField &chi);
+  virtual void ExportPhysicalFermionSolution(const FermionField &solution5d,FermionField &exported4d);
+  virtual void ExportPhysicalFermionSource(const FermionField &solution5d, FermionField &exported4d);
+  virtual void ImportPhysicalFermionSource(const FermionField &input4d,FermionField &imported5d);
+  virtual void ImportUnphysicalFermion(const FermionField &solution5d, FermionField &exported4d);
 
-    template<class Impl>
-    class CayleyFermion5D : public WilsonFermion5D<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-    public:
+  ///////////////////////////////////////////////////////////////
+  // Support for MADWF tricks
+  ///////////////////////////////////////////////////////////////
+  RealD Mass(void) { return mass; };
+  void  SetMass(RealD _mass) { 
+    mass=_mass; 
+    SetCoefficientsInternal(_zolo_hi,_gamma,_b,_c);  // Reset coeffs
+  } ;
+  void  P(const FermionField &psi, FermionField &chi);
+  void  Pdag(const FermionField &psi, FermionField &chi);
+  
+  /////////////////////////////////////////////////////
+  // Instantiate different versions depending on Impl
+  /////////////////////////////////////////////////////
+  void M5D(const FermionField &psi,
+	   const FermionField &phi,
+	   FermionField &chi,
+	   Vector<Coeff_t> &lower,
+	   Vector<Coeff_t> &diag,
+	   Vector<Coeff_t> &upper);
 
-      // override multiply
-      virtual RealD  M    (const FermionField &in, FermionField &out);
-      virtual RealD  Mdag (const FermionField &in, FermionField &out);
+  void M5Ddag(const FermionField &psi,
+	      const FermionField &phi,
+	      FermionField &chi,
+	      Vector<Coeff_t> &lower,
+	      Vector<Coeff_t> &diag,
+	      Vector<Coeff_t> &upper);
 
-      // half checkerboard operations
-      virtual void   Meooe       (const FermionField &in, FermionField &out);
-      virtual void   MeooeDag    (const FermionField &in, FermionField &out);
-      virtual void   Mooee       (const FermionField &in, FermionField &out);
-      virtual void   MooeeDag    (const FermionField &in, FermionField &out);
-      virtual void   MooeeInv    (const FermionField &in, FermionField &out);
-      virtual void   MooeeInvDag (const FermionField &in, FermionField &out);
-      virtual void   Meo5D (const FermionField &psi, FermionField &chi);
+  virtual void   Instantiatable(void)=0;
 
-      virtual void   M5D   (const FermionField &psi, FermionField &chi);
-      virtual void   M5Ddag(const FermionField &psi, FermionField &chi);
+  // force terms; five routines; default to Dhop on diagonal
+  virtual void MDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  virtual void MoeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  virtual void MeoDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
 
-      ///////////////////////////////////////////////////////////////
-      // Physical surface field utilities
-      ///////////////////////////////////////////////////////////////
-      virtual void   Dminus(const FermionField &psi, FermionField &chi);
-      virtual void   DminusDag(const FermionField &psi, FermionField &chi);
-      virtual void ExportPhysicalFermionSolution(const FermionField &solution5d,FermionField &exported4d);
-      virtual void ExportPhysicalFermionSource(const FermionField &solution5d, FermionField &exported4d);
-      virtual void ImportPhysicalFermionSource(const FermionField &input4d,FermionField &imported5d);
-      virtual void ImportUnphysicalFermion(const FermionField &solution5d, FermionField &exported4d);
+  // Efficient support for multigrid coarsening
+  virtual void  Mdir (const FermionField &in, FermionField &out,int dir,int disp);
 
-      ///////////////////////////////////////////////////////////////
-      // Support for MADWF tricks
-      ///////////////////////////////////////////////////////////////
-      RealD Mass(void) { return mass; };
-      void  SetMass(RealD _mass) { 
-	mass=_mass; 
-	SetCoefficientsInternal(_zolo_hi,_gamma,_b,_c);  // Reset coeffs
-      } ;
-      void  P(const FermionField &psi, FermionField &chi);
-      void  Pdag(const FermionField &psi, FermionField &chi);
+  void   Meooe5D       (const FermionField &in, FermionField &out);
+  void   MeooeDag5D    (const FermionField &in, FermionField &out);
 
-      /////////////////////////////////////////////////////
-      // Instantiate different versions depending on Impl
-      /////////////////////////////////////////////////////
-      void M5D(const FermionField &psi,
-	       const FermionField &phi,
-	       FermionField &chi,
-	       std::vector<Coeff_t> &lower,
-	       std::vector<Coeff_t> &diag,
-	       std::vector<Coeff_t> &upper);
+  //    protected:
+  RealD mass;
 
-      void M5Ddag(const FermionField &psi,
-		  const FermionField &phi,
-		  FermionField &chi,
-		  std::vector<Coeff_t> &lower,
-		  std::vector<Coeff_t> &diag,
-		  std::vector<Coeff_t> &upper);
+  // Save arguments to SetCoefficientsInternal
+  Vector<Coeff_t> _gamma;
+  RealD                _zolo_hi;
+  RealD                _b;
+  RealD                _c;
 
-      void MooeeInternal(const FermionField &in, FermionField &out,int dag,int inv);
-      void MooeeInternalCompute(int dag, int inv, Vector<iSinglet<Simd> > & Matp, Vector<iSinglet<Simd> > & Matm);
+  // Cayley form Moebius (tanh and zolotarev)
+  Vector<Coeff_t> omega;
+  Vector<Coeff_t> bs;    // S dependent coeffs
+  Vector<Coeff_t> cs;
+  Vector<Coeff_t> as;
+  // For preconditioning Cayley form
+  Vector<Coeff_t> bee;
+  Vector<Coeff_t> cee;
+  Vector<Coeff_t> aee;
+  Vector<Coeff_t> beo;
+  Vector<Coeff_t> ceo;
+  Vector<Coeff_t> aeo;
+  // LDU factorisation of the eeoo matrix
+  Vector<Coeff_t> lee;
+  Vector<Coeff_t> leem;
+  Vector<Coeff_t> uee;
+  Vector<Coeff_t> ueem;
+  Vector<Coeff_t> dee;
 
-      void MooeeInternalAsm(const FermionField &in, FermionField &out,
-			    int LLs, int site,
-			    Vector<iSinglet<Simd> > &Matp,
-			    Vector<iSinglet<Simd> > &Matm);
-      void MooeeInternalZAsm(const FermionField &in, FermionField &out,
-			    int LLs, int site,
-			    Vector<iSinglet<Simd> > &Matp,
-			    Vector<iSinglet<Simd> > &Matm);
+  // Matrices of 5d ee inverse params
+  Vector<iSinglet<Simd> >  MatpInv;
+  Vector<iSinglet<Simd> >  MatmInv;
+  Vector<iSinglet<Simd> >  MatpInvDag;
+  Vector<iSinglet<Simd> >  MatmInvDag;
 
+  // Constructors
+  CayleyFermion5D(GaugeField &_Umu,
+		  GridCartesian         &FiveDimGrid,
+		  GridRedBlackCartesian &FiveDimRedBlackGrid,
+		  GridCartesian         &FourDimGrid,
+		  GridRedBlackCartesian &FourDimRedBlackGrid,
+		  RealD _mass,RealD _M5,const ImplParams &p= ImplParams());
 
-      virtual void   Instantiatable(void)=0;
+  void CayleyReport(void);
+  void CayleyZeroCounters(void);
 
-      // force terms; five routines; default to Dhop on diagonal
-      virtual void MDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
-      virtual void MoeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
-      virtual void MeoDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  double M5Dflops;
+  double M5Dcalls;
+  double M5Dtime;
 
-      // Efficient support for multigrid coarsening
-      virtual void  Mdir (const FermionField &in, FermionField &out,int dir,int disp);
+  double MooeeInvFlops;
+  double MooeeInvCalls;
+  double MooeeInvTime;
 
-      void   Meooe5D       (const FermionField &in, FermionField &out);
-      void   MeooeDag5D    (const FermionField &in, FermionField &out);
+protected:
+  virtual void SetCoefficientsZolotarev(RealD zolohi,Approx::zolotarev_data *zdata,RealD b,RealD c);
+  virtual void SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD b,RealD c);
+  virtual void SetCoefficientsInternal(RealD zolo_hi,Vector<Coeff_t> & gamma,RealD b,RealD c);
+};
 
-      //    protected:
-      RealD mass;
+NAMESPACE_END(Grid);
 
-      // Save arguments to SetCoefficientsInternal
-      std::vector<Coeff_t> _gamma;
-      RealD                _zolo_hi;
-      RealD                _b;
-      RealD                _c;
-
-      // Cayley form Moebius (tanh and zolotarev)
-      std::vector<Coeff_t> omega;
-      std::vector<Coeff_t> bs;    // S dependent coeffs
-      std::vector<Coeff_t> cs;
-      std::vector<Coeff_t> as;
-      // For preconditioning Cayley form
-      std::vector<Coeff_t> bee;
-      std::vector<Coeff_t> cee;
-      std::vector<Coeff_t> aee;
-      std::vector<Coeff_t> beo;
-      std::vector<Coeff_t> ceo;
-      std::vector<Coeff_t> aeo;
-      // LDU factorisation of the eeoo matrix
-      std::vector<Coeff_t> lee;
-      std::vector<Coeff_t> leem;
-      std::vector<Coeff_t> uee;
-      std::vector<Coeff_t> ueem;
-      std::vector<Coeff_t> dee;
-
-      // Matrices of 5d ee inverse params
-      Vector<iSinglet<Simd> >  MatpInv;
-      Vector<iSinglet<Simd> >  MatmInv;
-      Vector<iSinglet<Simd> >  MatpInvDag;
-      Vector<iSinglet<Simd> >  MatmInvDag;
-
-      // Constructors
-      CayleyFermion5D(GaugeField &_Umu,
-		      GridCartesian         &FiveDimGrid,
-		      GridRedBlackCartesian &FiveDimRedBlackGrid,
-		      GridCartesian         &FourDimGrid,
-		      GridRedBlackCartesian &FourDimRedBlackGrid,
-		      RealD _mass,RealD _M5,const ImplParams &p= ImplParams());
-
-
-
-     void CayleyReport(void);
-     void CayleyZeroCounters(void);
-
-     double M5Dflops;
-     double M5Dcalls;
-     double M5Dtime;
-
-     double MooeeInvFlops;
-     double MooeeInvCalls;
-     double MooeeInvTime;
-
-    protected:
-      virtual void SetCoefficientsZolotarev(RealD zolohi,Approx::zolotarev_data *zdata,RealD b,RealD c);
-      virtual void SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD b,RealD c);
-      virtual void SetCoefficientsInternal(RealD zolo_hi,std::vector<Coeff_t> & gamma,RealD b,RealD c);
-    };
-
-  }
-}
-#define INSTANTIATE_DPERP(A)\
-template void CayleyFermion5D< A >::M5D(const FermionField &psi,const FermionField &phi,FermionField &chi,\
-					std::vector<Coeff_t> &lower,std::vector<Coeff_t> &diag,std::vector<Coeff_t> &upper); \
-template void CayleyFermion5D< A >::M5Ddag(const FermionField &psi,const FermionField &phi,FermionField &chi,\
-					   std::vector<Coeff_t> &lower,std::vector<Coeff_t> &diag,std::vector<Coeff_t> &upper); \
-template void CayleyFermion5D< A >::MooeeInv    (const FermionField &psi, FermionField &chi); \
-template void CayleyFermion5D< A >::MooeeInvDag (const FermionField &psi, FermionField &chi);
-
-#undef  CAYLEY_DPERP_DENSE
-#define  CAYLEY_DPERP_CACHE
-#undef  CAYLEY_DPERP_LINALG
-#define CAYLEY_DPERP_VEC
-
-#endif
diff --git a/Grid/qcd/action/fermion/CayleyFermion5Dcache.cc b/Grid/qcd/action/fermion/CayleyFermion5Dcache.cc
deleted file mode 100644
index dd6ec7bf..00000000
--- a/Grid/qcd/action/fermion/CayleyFermion5Dcache.cc
+++ /dev/null
@@ -1,249 +0,0 @@
-    /*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid 
-
-    Source file: ./lib/qcd/action/fermion/CayleyFermion5D.cc
-
-    Copyright (C) 2015
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/CayleyFermion5D.h>
-
-
-namespace Grid {
-namespace QCD {
-
-  // FIXME -- make a version of these routines with site loop outermost for cache reuse.
-
-  // Pminus fowards
-  // Pplus  backwards..
-template<class Impl>  
-void CayleyFermion5D<Impl>::M5D(const FermionField &psi,
-				const FermionField &phi, 
-				FermionField &chi,
-				std::vector<Coeff_t> &lower,
-				std::vector<Coeff_t> &diag,
-				std::vector<Coeff_t> &upper)
-{
-  int Ls =this->Ls;
-  GridBase *grid=psi._grid;
-  assert(phi.checkerboard == psi.checkerboard);
-  chi.checkerboard=psi.checkerboard;
-  // Flops = 6.0*(Nc*Ns) *Ls*vol
-  M5Dcalls++;
-  M5Dtime-=usecond();
-
-  parallel_for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
-    for(int s=0;s<Ls;s++){
-      auto tmp = psi._odata[0];
-      if ( s==0 ) {
- 	                            spProj5m(tmp,psi._odata[ss+s+1]);
-	chi[ss+s]=diag[s]*phi[ss+s]+upper[s]*tmp;
-
-	                    spProj5p(tmp,psi._odata[ss+Ls-1]);
-	chi[ss+s]=chi[ss+s]+lower[s]*tmp;
-      } else if ( s==(Ls-1)) {
-	                            spProj5m(tmp,psi._odata[ss+0]);
-	chi[ss+s]=diag[s]*phi[ss+s]+upper[s]*tmp;
-
- 	                    spProj5p(tmp,psi._odata[ss+s-1]);
-	chi[ss+s]=chi[ss+s]+lower[s]*tmp;
-      } else { 
-	                            spProj5m(tmp,psi._odata[ss+s+1]);
-	chi[ss+s]=diag[s]*phi[ss+s]+upper[s]*tmp;
-
-	                    spProj5p(tmp,psi._odata[ss+s-1]);
-	chi[ss+s]=chi[ss+s]+lower[s]*tmp;
-      }
-    }
-  }
-  M5Dtime+=usecond();
-}
-
-template<class Impl>  
-void CayleyFermion5D<Impl>::M5Ddag(const FermionField &psi,
-				   const FermionField &phi, 
-				   FermionField &chi,
-				   std::vector<Coeff_t> &lower,
-				   std::vector<Coeff_t> &diag,
-				   std::vector<Coeff_t> &upper)
-{
-  int Ls =this->Ls;
-  GridBase *grid=psi._grid;
-  assert(phi.checkerboard == psi.checkerboard);
-  chi.checkerboard=psi.checkerboard;
-
-  // Flops = 6.0*(Nc*Ns) *Ls*vol
-  M5Dcalls++;
-  M5Dtime-=usecond();
-
-  parallel_for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
-    auto tmp = psi._odata[0];
-    for(int s=0;s<Ls;s++){
-      if ( s==0 ) {
-	spProj5p(tmp,psi._odata[ss+s+1]);
-	chi[ss+s]=diag[s]*phi[ss+s]+upper[s]*tmp;
-
-	spProj5m(tmp,psi._odata[ss+Ls-1]);
-	chi[ss+s]=chi[ss+s]+lower[s]*tmp;
-      } else if ( s==(Ls-1)) {
-	spProj5p(tmp,psi._odata[ss+0]);
-	chi[ss+s]=diag[s]*phi[ss+s]+upper[s]*tmp;
-
-	spProj5m(tmp,psi._odata[ss+s-1]);
-	chi[ss+s]=chi[ss+s]+lower[s]*tmp;
-      } else { 
-	spProj5p(tmp,psi._odata[ss+s+1]);
-	chi[ss+s]=diag[s]*phi[ss+s]+upper[s]*tmp;
-
-	spProj5m(tmp,psi._odata[ss+s-1]);
-	chi[ss+s]=chi[ss+s]+lower[s]*tmp;
-      }
-    }
-  }
-  M5Dtime+=usecond();
-}
-
-template<class Impl>
-void CayleyFermion5D<Impl>::MooeeInv    (const FermionField &psi, FermionField &chi)
-{
-  GridBase *grid=psi._grid;
-  int Ls=this->Ls;
-
-  chi.checkerboard=psi.checkerboard;
-
-  MooeeInvCalls++;
-  MooeeInvTime-=usecond();
-
-  parallel_for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
-    auto tmp = psi._odata[0];
-
-    // flops = 12*2*Ls + 12*2*Ls + 3*12*Ls + 12*2*Ls  = 12*Ls * (9) = 108*Ls flops
-    // Apply (L^{\prime})^{-1}
-    chi[ss]=psi[ss]; // chi[0]=psi[0]
-    for(int s=1;s<Ls;s++){
-                            spProj5p(tmp,chi[ss+s-1]);  
-      chi[ss+s] = psi[ss+s]-lee[s-1]*tmp;
-    }
-    // L_m^{-1} 
-    for (int s=0;s<Ls-1;s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
-                                   spProj5m(tmp,chi[ss+s]);    
-      chi[ss+Ls-1] = chi[ss+Ls-1] - leem[s]*tmp;
-    }
-    // U_m^{-1} D^{-1}
-    for (int s=0;s<Ls-1;s++){
-      // Chi[s] + 1/d chi[s] 
-                                                spProj5p(tmp,chi[ss+Ls-1]); 
-      chi[ss+s] = (1.0/dee[s])*chi[ss+s]-(ueem[s]/dee[Ls-1])*tmp;
-    }	
-    chi[ss+Ls-1]= (1.0/dee[Ls-1])*chi[ss+Ls-1];
-      
-    // Apply U^{-1}
-    for (int s=Ls-2;s>=0;s--){
-                            spProj5m(tmp,chi[ss+s+1]);  
-      chi[ss+s] = chi[ss+s] - uee[s]*tmp;
-    }
-  }
-
-  MooeeInvTime+=usecond();
-
-}
-
-template<class Impl>
-void CayleyFermion5D<Impl>::MooeeInvDag (const FermionField &psi, FermionField &chi)
-{
-  GridBase *grid=psi._grid;
-  int Ls=this->Ls;
-
-  assert(psi.checkerboard == psi.checkerboard);
-  chi.checkerboard=psi.checkerboard;
-
-  std::vector<Coeff_t> ueec(Ls);
-  std::vector<Coeff_t> deec(Ls);
-  std::vector<Coeff_t> leec(Ls);
-  std::vector<Coeff_t> ueemc(Ls);
-  std::vector<Coeff_t> leemc(Ls);
-  for(int s=0;s<ueec.size();s++){
-    ueec[s] = conjugate(uee[s]);
-    deec[s] = conjugate(dee[s]);
-    leec[s] = conjugate(lee[s]);
-    ueemc[s]= conjugate(ueem[s]);
-    leemc[s]= conjugate(leem[s]);
-  }
-  MooeeInvCalls++;
-  MooeeInvTime-=usecond();
-
-  parallel_for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
-
-    auto tmp = psi._odata[0];
-
-    // Apply (U^{\prime})^{-dagger}
-    chi[ss]=psi[ss];
-    for (int s=1;s<Ls;s++){
-                            spProj5m(tmp,chi[ss+s-1]);
-      chi[ss+s] = psi[ss+s]-ueec[s-1]*tmp;
-    }
-    // U_m^{-\dagger} 
-    for (int s=0;s<Ls-1;s++){
-                                   spProj5p(tmp,chi[ss+s]);
-      chi[ss+Ls-1] = chi[ss+Ls-1] - ueemc[s]*tmp;
-    }
-
-    // L_m^{-\dagger} D^{-dagger}
-    for (int s=0;s<Ls-1;s++){
-      spProj5m(tmp,chi[ss+Ls-1]);
-      chi[ss+s] = (1.0/deec[s])*chi[ss+s]-(leemc[s]/deec[Ls-1])*tmp;
-    }	
-    chi[ss+Ls-1]= (1.0/deec[Ls-1])*chi[ss+Ls-1];
-  
-    // Apply L^{-dagger}
-    for (int s=Ls-2;s>=0;s--){
-      spProj5p(tmp,chi[ss+s+1]);
-      chi[ss+s] = chi[ss+s] - leec[s]*tmp;
-    }
-  }
-
-  MooeeInvTime+=usecond();
-
-}
-
-#ifdef CAYLEY_DPERP_CACHE
-  INSTANTIATE_DPERP(WilsonImplF);
-  INSTANTIATE_DPERP(WilsonImplD);
-  INSTANTIATE_DPERP(GparityWilsonImplF);
-  INSTANTIATE_DPERP(GparityWilsonImplD);
-  INSTANTIATE_DPERP(ZWilsonImplF);
-  INSTANTIATE_DPERP(ZWilsonImplD);
-
-  INSTANTIATE_DPERP(WilsonImplFH);
-  INSTANTIATE_DPERP(WilsonImplDF);
-  INSTANTIATE_DPERP(GparityWilsonImplFH);
-  INSTANTIATE_DPERP(GparityWilsonImplDF);
-  INSTANTIATE_DPERP(ZWilsonImplFH);
-  INSTANTIATE_DPERP(ZWilsonImplDF);
-#endif
-
-}}
diff --git a/Grid/qcd/action/fermion/CayleyFermion5Dvec.cc b/Grid/qcd/action/fermion/CayleyFermion5Dvec.cc
deleted file mode 100644
index 2b2eace7..00000000
--- a/Grid/qcd/action/fermion/CayleyFermion5Dvec.cc
+++ /dev/null
@@ -1,828 +0,0 @@
-    /*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid 
-
-    Source file: ./lib/qcd/action/fermion/CayleyFermion5D.cc
-
-    Copyright (C) 2015
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-
-
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/CayleyFermion5D.h>
-
-
-namespace Grid {
-namespace QCD {  
-  /*
-   * Dense matrix versions of routines
-   */
-template<class Impl>
-void CayleyFermion5D<Impl>::MooeeInvDag (const FermionField &psi, FermionField &chi)
-{
-  this->MooeeInternal(psi,chi,DaggerYes,InverseYes);
-}
-  
-template<class Impl>
-void CayleyFermion5D<Impl>::MooeeInv(const FermionField &psi, FermionField &chi)
-{
-  this->MooeeInternal(psi,chi,DaggerNo,InverseYes);
-}
-template<class Impl>  
-void CayleyFermion5D<Impl>::M5D(const FermionField &psi,
-				const FermionField &phi, 
-				FermionField &chi,
-				std::vector<Coeff_t> &lower,
-				std::vector<Coeff_t> &diag,
-				std::vector<Coeff_t> &upper)
-{
-  GridBase *grid=psi._grid;
-  int Ls   = this->Ls;
-  int LLs  = grid->_rdimensions[0];
-  const int nsimd= Simd::Nsimd();
-
-  Vector<iSinglet<Simd> > u(LLs);
-  Vector<iSinglet<Simd> > l(LLs);
-  Vector<iSinglet<Simd> > d(LLs);
-
-  assert(Ls/LLs==nsimd);
-  assert(phi.checkerboard == psi.checkerboard);
-
-  chi.checkerboard=psi.checkerboard;
-
-  // just directly address via type pun
-  typedef typename Simd::scalar_type scalar_type;
-  scalar_type * u_p = (scalar_type *)&u[0];
-  scalar_type * l_p = (scalar_type *)&l[0];
-  scalar_type * d_p = (scalar_type *)&d[0];
-
-  for(int o=0;o<LLs;o++){ // outer
-  for(int i=0;i<nsimd;i++){ //inner
-    int s  = o+i*LLs;
-    int ss = o*nsimd+i;
-    u_p[ss] = upper[s];
-    l_p[ss] = lower[s];
-    d_p[ss] = diag[s];
-  }}
-
-
-  M5Dcalls++;
-  M5Dtime-=usecond();
-
-  assert(Nc==3);
-
-  parallel_for(int ss=0;ss<grid->oSites();ss+=LLs){ // adds LLs
-#if 0
-      alignas(64) SiteHalfSpinor hp;
-      alignas(64) SiteHalfSpinor hm;
-      alignas(64) SiteSpinor fp;
-      alignas(64) SiteSpinor fm;
-
-      for(int v=0;v<LLs;v++){
-
-	int vp=(v+1)%LLs;
-	int vm=(v+LLs-1)%LLs;
-
-	spProj5m(hp,psi[ss+vp]);
-	spProj5p(hm,psi[ss+vm]);
-
-	if ( vp<=v ) rotate(hp,hp,1);
-	if ( vm>=v ) rotate(hm,hm,nsimd-1);
-	
-	hp=0.5*hp;
-        hm=0.5*hm;
-
-	spRecon5m(fp,hp);
-	spRecon5p(fm,hm);
-
-	chi[ss+v] = d[v]*phi[ss+v];
-	chi[ss+v] = chi[ss+v]     +u[v]*fp;
-	chi[ss+v] = chi[ss+v]     +l[v]*fm;
-
-      }
-#else
-      for(int v=0;v<LLs;v++){
-
-	vprefetch(psi[ss+v+LLs]);
-
-	int vp= (v==LLs-1) ? 0     : v+1;
-	int vm= (v==0    ) ? LLs-1 : v-1;
-	
-	Simd hp_00 = psi[ss+vp]()(2)(0); 
-	Simd hp_01 = psi[ss+vp]()(2)(1); 
-	Simd hp_02 = psi[ss+vp]()(2)(2); 
-	Simd hp_10 = psi[ss+vp]()(3)(0); 
-	Simd hp_11 = psi[ss+vp]()(3)(1); 
-	Simd hp_12 = psi[ss+vp]()(3)(2); 
-	
-	Simd hm_00 = psi[ss+vm]()(0)(0); 
-	Simd hm_01 = psi[ss+vm]()(0)(1); 
-	Simd hm_02 = psi[ss+vm]()(0)(2); 
-	Simd hm_10 = psi[ss+vm]()(1)(0); 
-	Simd hm_11 = psi[ss+vm]()(1)(1); 
-	Simd hm_12 = psi[ss+vm]()(1)(2); 
-
-	if ( vp<=v ) {
-	  hp_00.v = Optimization::Rotate::tRotate<2>(hp_00.v);
-	  hp_01.v = Optimization::Rotate::tRotate<2>(hp_01.v);
-	  hp_02.v = Optimization::Rotate::tRotate<2>(hp_02.v);
-	  hp_10.v = Optimization::Rotate::tRotate<2>(hp_10.v);
-	  hp_11.v = Optimization::Rotate::tRotate<2>(hp_11.v);
-	  hp_12.v = Optimization::Rotate::tRotate<2>(hp_12.v);
-	}
-	if ( vm>=v ) {
-	  hm_00.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_00.v);
-	  hm_01.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_01.v);
-	  hm_02.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_02.v);
-	  hm_10.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_10.v);
-	  hm_11.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_11.v);
-	  hm_12.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_12.v);
-	}
-
-	// Can force these to real arithmetic and save 2x.
-	Simd p_00  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(0))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_00); 
-	Simd p_01  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(1))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_01); 
-	Simd p_02  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(2))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_02); 
-	Simd p_10  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(0))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_10); 
-	Simd p_11  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(1))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_11); 
-	Simd p_12  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(2))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_12); 
-	Simd p_20  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(0))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_00); 
-	Simd p_21  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(1))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_01); 
-	Simd p_22  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(2))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_02);  
-	Simd p_30  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(0))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_10); 
-	Simd p_31  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(1))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_11); 
-	Simd p_32  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(2))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_12); 
-
-	vstream(chi[ss+v]()(0)(0),p_00);
-	vstream(chi[ss+v]()(0)(1),p_01);
-	vstream(chi[ss+v]()(0)(2),p_02);
-	vstream(chi[ss+v]()(1)(0),p_10);
-	vstream(chi[ss+v]()(1)(1),p_11);
-	vstream(chi[ss+v]()(1)(2),p_12);
-	vstream(chi[ss+v]()(2)(0),p_20);
-	vstream(chi[ss+v]()(2)(1),p_21);
-	vstream(chi[ss+v]()(2)(2),p_22);
-	vstream(chi[ss+v]()(3)(0),p_30);
-	vstream(chi[ss+v]()(3)(1),p_31);
-	vstream(chi[ss+v]()(3)(2),p_32);
-
-      }
-#endif
-  }
-  M5Dtime+=usecond();
-}
-
-template<class Impl>  
-void CayleyFermion5D<Impl>::M5Ddag(const FermionField &psi,
-				   const FermionField &phi, 
-				   FermionField &chi,
-				   std::vector<Coeff_t> &lower,
-				   std::vector<Coeff_t> &diag,
-				   std::vector<Coeff_t> &upper)
-{
-  GridBase *grid=psi._grid;
-  int Ls   = this->Ls;
-  int LLs  = grid->_rdimensions[0];
-  int nsimd= Simd::Nsimd();
-
-  Vector<iSinglet<Simd> > u(LLs);
-  Vector<iSinglet<Simd> > l(LLs);
-  Vector<iSinglet<Simd> > d(LLs);
-
-  assert(Ls/LLs==nsimd);
-  assert(phi.checkerboard == psi.checkerboard);
-
-  chi.checkerboard=psi.checkerboard;
-
-  // just directly address via type pun
-  typedef typename Simd::scalar_type scalar_type;
-  scalar_type * u_p = (scalar_type *)&u[0];
-  scalar_type * l_p = (scalar_type *)&l[0];
-  scalar_type * d_p = (scalar_type *)&d[0];
-
-  for(int o=0;o<LLs;o++){ // outer
-  for(int i=0;i<nsimd;i++){ //inner
-    int s  = o+i*LLs;
-    int ss = o*nsimd+i;
-    u_p[ss] = upper[s];
-    l_p[ss] = lower[s];
-    d_p[ss] = diag[s];
-  }}
-
-  M5Dcalls++;
-  M5Dtime-=usecond();
-  parallel_for(int ss=0;ss<grid->oSites();ss+=LLs){ // adds LLs
-#if 0
-    alignas(64) SiteHalfSpinor hp;
-    alignas(64) SiteHalfSpinor hm;
-    alignas(64) SiteSpinor fp;
-    alignas(64) SiteSpinor fm;
-
-    for(int v=0;v<LLs;v++){
-
-      int vp=(v+1)%LLs;
-      int vm=(v+LLs-1)%LLs;
-
-      spProj5p(hp,psi[ss+vp]);
-      spProj5m(hm,psi[ss+vm]);
-
-      if ( vp<=v ) rotate(hp,hp,1);
-      if ( vm>=v ) rotate(hm,hm,nsimd-1);
-      
-      hp=hp*0.5;
-      hm=hm*0.5;
-      spRecon5p(fp,hp);
-      spRecon5m(fm,hm);
-
-      chi[ss+v] = d[v]*phi[ss+v]+u[v]*fp;
-      chi[ss+v] = chi[ss+v]     +l[v]*fm;
-
-    }
-#else
-      for(int v=0;v<LLs;v++){
-
-	vprefetch(psi[ss+v+LLs]);
-
-	int vp= (v==LLs-1) ? 0     : v+1;
-	int vm= (v==0    ) ? LLs-1 : v-1;
-	
-	Simd hp_00 = psi[ss+vp]()(0)(0); 
-	Simd hp_01 = psi[ss+vp]()(0)(1); 
-	Simd hp_02 = psi[ss+vp]()(0)(2); 
-	Simd hp_10 = psi[ss+vp]()(1)(0); 
-	Simd hp_11 = psi[ss+vp]()(1)(1); 
-	Simd hp_12 = psi[ss+vp]()(1)(2); 
-	
-	Simd hm_00 = psi[ss+vm]()(2)(0); 
-	Simd hm_01 = psi[ss+vm]()(2)(1); 
-	Simd hm_02 = psi[ss+vm]()(2)(2); 
-	Simd hm_10 = psi[ss+vm]()(3)(0); 
-	Simd hm_11 = psi[ss+vm]()(3)(1); 
-	Simd hm_12 = psi[ss+vm]()(3)(2); 
-
-	if ( vp<=v ) {
-	  hp_00.v = Optimization::Rotate::tRotate<2>(hp_00.v);
-	  hp_01.v = Optimization::Rotate::tRotate<2>(hp_01.v);
-	  hp_02.v = Optimization::Rotate::tRotate<2>(hp_02.v);
-	  hp_10.v = Optimization::Rotate::tRotate<2>(hp_10.v);
-	  hp_11.v = Optimization::Rotate::tRotate<2>(hp_11.v);
-	  hp_12.v = Optimization::Rotate::tRotate<2>(hp_12.v);
-	}
-	if ( vm>=v ) {
-	  hm_00.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_00.v);
-	  hm_01.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_01.v);
-	  hm_02.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_02.v);
-	  hm_10.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_10.v);
-	  hm_11.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_11.v);
-	  hm_12.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_12.v);
-	}
-
-	Simd p_00  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(0))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_00); 
-	Simd p_01  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(1))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_01); 
-	Simd p_02  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(2))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_02); 
-	Simd p_10  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(0))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_10); 
-	Simd p_11  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(1))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_11); 
-	Simd p_12  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(2))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_12); 
-
-	Simd p_20  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(0))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_00); 
-	Simd p_21  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(1))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_01); 
-	Simd p_22  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(2))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_02);  
-	Simd p_30  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(0))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_10); 
-	Simd p_31  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(1))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_11); 
-	Simd p_32  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(2))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_12); 
-
-	vstream(chi[ss+v]()(0)(0),p_00);
-	vstream(chi[ss+v]()(0)(1),p_01);
-	vstream(chi[ss+v]()(0)(2),p_02);
-	vstream(chi[ss+v]()(1)(0),p_10);
-	vstream(chi[ss+v]()(1)(1),p_11);
-	vstream(chi[ss+v]()(1)(2),p_12);
-	vstream(chi[ss+v]()(2)(0),p_20);
-	vstream(chi[ss+v]()(2)(1),p_21);
-	vstream(chi[ss+v]()(2)(2),p_22);
-	vstream(chi[ss+v]()(3)(0),p_30);
-	vstream(chi[ss+v]()(3)(1),p_31);
-	vstream(chi[ss+v]()(3)(2),p_32);
-      }
-#endif
-  }
-  M5Dtime+=usecond();
-}
-
-
-#ifdef AVX512 
-#include <simd/Intel512common.h>
-#include <simd/Intel512avx.h>
-#include <simd/Intel512single.h>
-#endif 
-
-template<class Impl>
-void CayleyFermion5D<Impl>::MooeeInternalAsm(const FermionField &psi, FermionField &chi,
-					     int LLs, int site,
-					     Vector<iSinglet<Simd> > &Matp,
-					     Vector<iSinglet<Simd> > &Matm)
-{
-#ifndef AVX512
-  {
-  SiteHalfSpinor BcastP;
-  SiteHalfSpinor BcastM;
-  SiteHalfSpinor SiteChiP;
-  SiteHalfSpinor SiteChiM;
-
-  // Ls*Ls * 2 * 12 * vol flops
-  for(int s1=0;s1<LLs;s1++){ 
-    for(int s2=0;s2<LLs;s2++){ 
-      for(int  l=0; l<Simd::Nsimd();l++){ // simd lane
-
-        int s=s2+l*LLs;
-	int lex=s2+LLs*site;
-	
-	if ( s2==0 && l==0) {
-	  SiteChiP=zero;
-	  SiteChiM=zero;
-	}
-	
-	for(int sp=0;sp<2;sp++){
-        for(int co=0;co<Nc;co++){
-	  vbroadcast(BcastP()(sp  )(co),psi[lex]()(sp)(co),l);
-	}}
-	for(int sp=0;sp<2;sp++){
-        for(int co=0;co<Nc;co++){
-	  vbroadcast(BcastM()(sp  )(co),psi[lex]()(sp+2)(co),l);
-	}}
-
-	for(int sp=0;sp<2;sp++){
-        for(int co=0;co<Nc;co++){
-	  SiteChiP()(sp)(co)=real_madd(Matp[LLs*s+s1]()()(),BcastP()(sp)(co),SiteChiP()(sp)(co)); // 1100 us.
-	  SiteChiM()(sp)(co)=real_madd(Matm[LLs*s+s1]()()(),BcastM()(sp)(co),SiteChiM()(sp)(co)); // each found by commenting out
-	}}
-
-    }}
-    {
-      int lex = s1+LLs*site;
-      for(int sp=0;sp<2;sp++){
-      for(int co=0;co<Nc;co++){
-	vstream(chi[lex]()(sp)(co), SiteChiP()(sp)(co));
-	vstream(chi[lex]()(sp+2)(co), SiteChiM()(sp)(co));
-      }}
-    }
-  }
-
-  }
-#else
-  {
-  // pointers
-    //  MASK_REGS;
-#define Chi_00 %%zmm1
-#define Chi_01 %%zmm2
-#define Chi_02 %%zmm3
-#define Chi_10 %%zmm4
-#define Chi_11 %%zmm5
-#define Chi_12 %%zmm6
-#define Chi_20 %%zmm7
-#define Chi_21 %%zmm8
-#define Chi_22 %%zmm9
-#define Chi_30 %%zmm10
-#define Chi_31 %%zmm11
-#define Chi_32 %%zmm12
-
-#define BCAST0   %%zmm13
-#define BCAST1   %%zmm14
-#define BCAST2   %%zmm15
-#define BCAST3   %%zmm16
-#define BCAST4   %%zmm17
-#define BCAST5   %%zmm18
-#define BCAST6   %%zmm19
-#define BCAST7   %%zmm20
-#define BCAST8   %%zmm21
-#define BCAST9   %%zmm22
-#define BCAST10  %%zmm23
-#define BCAST11  %%zmm24
-
-  int incr=LLs*LLs*sizeof(iSinglet<Simd>);
-  for(int s1=0;s1<LLs;s1++){ 
-    for(int s2=0;s2<LLs;s2++){ 
-      int lex=s2+LLs*site;
-      uint64_t a0 = (uint64_t)&Matp[LLs*s2+s1]; // should be cacheable
-      uint64_t a1 = (uint64_t)&Matm[LLs*s2+s1];
-      uint64_t a2 = (uint64_t)&psi[lex];
-      for(int  l=0; l<Simd::Nsimd();l++){ // simd lane
-	if ( (s2+l)==0 ) {
-	  asm (
-  	           VPREFETCH1(0,%2)  	     VPREFETCH1(0,%1)
-  	           VPREFETCH1(12,%2)  	     VPREFETCH1(13,%2)
-  	           VPREFETCH1(14,%2)  	     VPREFETCH1(15,%2)         
-		   VBCASTCDUP(0,%2,BCAST0)   
-		   VBCASTCDUP(1,%2,BCAST1)   
-		   VBCASTCDUP(2,%2,BCAST2)   
-		   VBCASTCDUP(3,%2,BCAST3)   
-		   VBCASTCDUP(4,%2,BCAST4)     VMULMEM (0,%0,BCAST0,Chi_00)
-		   VBCASTCDUP(5,%2,BCAST5)     VMULMEM (0,%0,BCAST1,Chi_01)
-		   VBCASTCDUP(6,%2,BCAST6)     VMULMEM (0,%0,BCAST2,Chi_02)
-		   VBCASTCDUP(7,%2,BCAST7)     VMULMEM (0,%0,BCAST3,Chi_10)
-		   VBCASTCDUP(8,%2,BCAST8)     VMULMEM (0,%0,BCAST4,Chi_11)
-		   VBCASTCDUP(9,%2,BCAST9)     VMULMEM (0,%0,BCAST5,Chi_12)
-		   VBCASTCDUP(10,%2,BCAST10)   VMULMEM (0,%1,BCAST6,Chi_20)
-		   VBCASTCDUP(11,%2,BCAST11)   VMULMEM (0,%1,BCAST7,Chi_21)
-		   VMULMEM (0,%1,BCAST8,Chi_22)         
-		   VMULMEM (0,%1,BCAST9,Chi_30)
-		   VMULMEM (0,%1,BCAST10,Chi_31)       
-		   VMULMEM (0,%1,BCAST11,Chi_32)
-		   : : "r" (a0), "r" (a1), "r" (a2)  );
-	} else { 
-	  asm (
-		   VBCASTCDUP(0,%2,BCAST0)   VMADDMEM (0,%0,BCAST0,Chi_00)
-		   VBCASTCDUP(1,%2,BCAST1)   VMADDMEM (0,%0,BCAST1,Chi_01)
-		   VBCASTCDUP(2,%2,BCAST2)   VMADDMEM (0,%0,BCAST2,Chi_02)
-		   VBCASTCDUP(3,%2,BCAST3)   VMADDMEM (0,%0,BCAST3,Chi_10)
-		   VBCASTCDUP(4,%2,BCAST4)   VMADDMEM (0,%0,BCAST4,Chi_11)
-		   VBCASTCDUP(5,%2,BCAST5)   VMADDMEM (0,%0,BCAST5,Chi_12)
-		   VBCASTCDUP(6,%2,BCAST6)   VMADDMEM (0,%1,BCAST6,Chi_20)
-		   VBCASTCDUP(7,%2,BCAST7)   VMADDMEM (0,%1,BCAST7,Chi_21)
-		   VBCASTCDUP(8,%2,BCAST8)   VMADDMEM (0,%1,BCAST8,Chi_22)
-		   VBCASTCDUP(9,%2,BCAST9)   VMADDMEM (0,%1,BCAST9,Chi_30)
-		   VBCASTCDUP(10,%2,BCAST10)  VMADDMEM (0,%1,BCAST10,Chi_31)
-		   VBCASTCDUP(11,%2,BCAST11)  VMADDMEM (0,%1,BCAST11,Chi_32) 
-		   : : "r" (a0), "r" (a1), "r" (a2)  );
-	}
-	a0 = a0+incr;
-	a1 = a1+incr;
-	a2 = a2+sizeof(typename Simd::scalar_type);
-      }}
-    {
-      int lexa = s1+LLs*site;
-      asm (
-	       VSTORE(0,%0,Chi_00) VSTORE(1 ,%0,Chi_01)  VSTORE(2 ,%0,Chi_02)		
-	       VSTORE(3,%0,Chi_10) VSTORE(4 ,%0,Chi_11)  VSTORE(5 ,%0,Chi_12)		
-	       VSTORE(6,%0,Chi_20) VSTORE(7 ,%0,Chi_21)  VSTORE(8 ,%0,Chi_22)		
-	       VSTORE(9,%0,Chi_30) VSTORE(10,%0,Chi_31)  VSTORE(11,%0,Chi_32)		
-	       : : "r" ((uint64_t)&chi[lexa]) : "memory" );
-
-    }
-  }
-  }
-#undef Chi_00
-#undef Chi_01
-#undef Chi_02
-#undef Chi_10
-#undef Chi_11
-#undef Chi_12
-#undef Chi_20
-#undef Chi_21
-#undef Chi_22
-#undef Chi_30
-#undef Chi_31
-#undef Chi_32
-
-#undef BCAST0
-#undef BCAST1
-#undef BCAST2
-#undef BCAST3
-#undef BCAST4
-#undef BCAST5
-#undef BCAST6
-#undef BCAST7
-#undef BCAST8
-#undef BCAST9
-#undef BCAST10
-#undef BCAST11
-#endif
-};
-
-  // Z-mobius version
-template<class Impl>
-void CayleyFermion5D<Impl>::MooeeInternalZAsm(const FermionField &psi, FermionField &chi,
-					     int LLs, int site, Vector<iSinglet<Simd> > &Matp, Vector<iSinglet<Simd> > &Matm)
-{
-#ifndef AVX512
-  {
-  SiteHalfSpinor BcastP;
-  SiteHalfSpinor BcastM;
-  SiteHalfSpinor SiteChiP;
-  SiteHalfSpinor SiteChiM;
-
-  // Ls*Ls * 2 * 12 * vol flops
-  for(int s1=0;s1<LLs;s1++){ 
-    for(int s2=0;s2<LLs;s2++){ 
-      for(int  l=0; l<Simd::Nsimd();l++){ // simd lane
-
-        int s=s2+l*LLs;
-	int lex=s2+LLs*site;
-	
-	if ( s2==0 && l==0) {
-	  SiteChiP=zero;
-	  SiteChiM=zero;
-	}
-	
-	for(int sp=0;sp<2;sp++){
-        for(int co=0;co<Nc;co++){
-	  vbroadcast(BcastP()(sp  )(co),psi[lex]()(sp)(co),l);
-	}}
-	for(int sp=0;sp<2;sp++){
-        for(int co=0;co<Nc;co++){
-	  vbroadcast(BcastM()(sp  )(co),psi[lex]()(sp+2)(co),l);
-	}}
-
-	for(int sp=0;sp<2;sp++){
-        for(int co=0;co<Nc;co++){
-	  SiteChiP()(sp)(co)=SiteChiP()(sp)(co)+ Matp[LLs*s+s1]()()()*BcastP()(sp)(co); 
-	  SiteChiM()(sp)(co)=SiteChiM()(sp)(co)+ Matm[LLs*s+s1]()()()*BcastM()(sp)(co); 
-	}}
-
-
-    }}
-    {
-      int lex = s1+LLs*site;
-      for(int sp=0;sp<2;sp++){
-      for(int co=0;co<Nc;co++){
-	vstream(chi[lex]()(sp)(co), SiteChiP()(sp)(co));
-	vstream(chi[lex]()(sp+2)(co), SiteChiM()(sp)(co));
-      }}
-    }
-  }
-
-  }
-#else
-  {
-  // pointers
-  //  MASK_REGS;
-#define Chi_00 %zmm0
-#define Chi_01 %zmm1
-#define Chi_02 %zmm2
-#define Chi_10 %zmm3
-#define Chi_11 %zmm4
-#define Chi_12 %zmm5
-#define Chi_20 %zmm6
-#define Chi_21 %zmm7
-#define Chi_22 %zmm8
-#define Chi_30 %zmm9
-#define Chi_31 %zmm10
-#define Chi_32 %zmm11
-#define pChi_00 %%zmm0
-#define pChi_01 %%zmm1
-#define pChi_02 %%zmm2
-#define pChi_10 %%zmm3
-#define pChi_11 %%zmm4
-#define pChi_12 %%zmm5
-#define pChi_20 %%zmm6
-#define pChi_21 %%zmm7
-#define pChi_22 %%zmm8
-#define pChi_30 %%zmm9
-#define pChi_31 %%zmm10
-#define pChi_32 %%zmm11
-
-#define BCAST_00   %zmm12
-#define  SHUF_00   %zmm13
-#define BCAST_01   %zmm14
-#define  SHUF_01   %zmm15
-#define BCAST_02   %zmm16
-#define  SHUF_02   %zmm17
-#define BCAST_10   %zmm18
-#define  SHUF_10   %zmm19
-#define BCAST_11   %zmm20
-#define  SHUF_11   %zmm21
-#define BCAST_12   %zmm22
-#define  SHUF_12   %zmm23
-
-#define Mp  %zmm24
-#define Mps %zmm25
-#define Mm  %zmm26
-#define Mms %zmm27
-#define N 8
-  int incr=LLs*LLs*sizeof(iSinglet<Simd>);
-  for(int s1=0;s1<LLs;s1++){ 
-    for(int s2=0;s2<LLs;s2++){ 
-      int lex=s2+LLs*site;
-      uint64_t a0 = (uint64_t)&Matp[LLs*s2+s1]; // should be cacheable
-      uint64_t a1 = (uint64_t)&Matm[LLs*s2+s1];
-      uint64_t a2 = (uint64_t)&psi[lex];
-      for(int  l=0; l<Simd::Nsimd();l++){ // simd lane
-	if ( (s2+l)==0 ) {
-	  LOAD64(%r8,a0);
-	  LOAD64(%r9,a1);
-	  LOAD64(%r10,a2);
-	  asm (
-	       VLOAD(0,%r8,Mp)// i r
-	       VLOAD(0,%r9,Mm)
-	       VSHUF(Mp,Mps)  // r i 
-	       VSHUF(Mm,Mms)
-	       VPREFETCH1(12,%r10)  	     VPREFETCH1(13,%r10)
-	       VPREFETCH1(14,%r10)  	     VPREFETCH1(15,%r10)         
-
-	       VMULIDUP(0*N,%r10,Mps,Chi_00)
-	       VMULIDUP(1*N,%r10,Mps,Chi_01)
-	       VMULIDUP(2*N,%r10,Mps,Chi_02)
-	       VMULIDUP(3*N,%r10,Mps,Chi_10)
-	       VMULIDUP(4*N,%r10,Mps,Chi_11)
-	       VMULIDUP(5*N,%r10,Mps,Chi_12)
-
-	       VMULIDUP(6*N ,%r10,Mms,Chi_20)
-	       VMULIDUP(7*N ,%r10,Mms,Chi_21)
-	       VMULIDUP(8*N ,%r10,Mms,Chi_22)
-	       VMULIDUP(9*N ,%r10,Mms,Chi_30)
-	       VMULIDUP(10*N,%r10,Mms,Chi_31)
-	       VMULIDUP(11*N,%r10,Mms,Chi_32)
-
-	       VMADDSUBRDUP(0*N,%r10,Mp,Chi_00)
-	       VMADDSUBRDUP(1*N,%r10,Mp,Chi_01)
-	       VMADDSUBRDUP(2*N,%r10,Mp,Chi_02)
-	       VMADDSUBRDUP(3*N,%r10,Mp,Chi_10)
-	       VMADDSUBRDUP(4*N,%r10,Mp,Chi_11)
-	       VMADDSUBRDUP(5*N,%r10,Mp,Chi_12)
-
-	       VMADDSUBRDUP(6*N ,%r10,Mm,Chi_20)
-	       VMADDSUBRDUP(7*N ,%r10,Mm,Chi_21)
-	       VMADDSUBRDUP(8*N ,%r10,Mm,Chi_22)
-	       VMADDSUBRDUP(9*N ,%r10,Mm,Chi_30)
-	       VMADDSUBRDUP(10*N,%r10,Mm,Chi_31)
-	       VMADDSUBRDUP(11*N,%r10,Mm,Chi_32)
-	       );
-	} else { 
-	  LOAD64(%r8,a0);
-	  LOAD64(%r9,a1);
-	  LOAD64(%r10,a2);
-	  asm (
-	       VLOAD(0,%r8,Mp)
-	       VSHUF(Mp,Mps)
-
-	       VLOAD(0,%r9,Mm)
-	       VSHUF(Mm,Mms)
-
-	       VMADDSUBIDUP(0*N,%r10,Mps,Chi_00) //  Mri * Pii +- Cir
-	       VMADDSUBIDUP(1*N,%r10,Mps,Chi_01)
-	       VMADDSUBIDUP(2*N,%r10,Mps,Chi_02)
-	       VMADDSUBIDUP(3*N,%r10,Mps,Chi_10)
-	       VMADDSUBIDUP(4*N,%r10,Mps,Chi_11)
-	       VMADDSUBIDUP(5*N,%r10,Mps,Chi_12)
-
-	       VMADDSUBIDUP(6 *N,%r10,Mms,Chi_20)
-	       VMADDSUBIDUP(7 *N,%r10,Mms,Chi_21)
-	       VMADDSUBIDUP(8 *N,%r10,Mms,Chi_22)
-	       VMADDSUBIDUP(9 *N,%r10,Mms,Chi_30)
-	       VMADDSUBIDUP(10*N,%r10,Mms,Chi_31)
-	       VMADDSUBIDUP(11*N,%r10,Mms,Chi_32)
-
-	       VMADDSUBRDUP(0*N,%r10,Mp,Chi_00) //  Cir = Mir * Prr +- ( Mri * Pii +- Cir) 
-	       VMADDSUBRDUP(1*N,%r10,Mp,Chi_01) //  Ci = MiPr + Ci + MrPi ;    Cr = MrPr - ( MiPi - Cr)
-	       VMADDSUBRDUP(2*N,%r10,Mp,Chi_02)
-	       VMADDSUBRDUP(3*N,%r10,Mp,Chi_10)
-	       VMADDSUBRDUP(4*N,%r10,Mp,Chi_11)
-	       VMADDSUBRDUP(5*N,%r10,Mp,Chi_12)
-
-	       VMADDSUBRDUP(6 *N,%r10,Mm,Chi_20)
-	       VMADDSUBRDUP(7 *N,%r10,Mm,Chi_21)
-	       VMADDSUBRDUP(8 *N,%r10,Mm,Chi_22)
-	       VMADDSUBRDUP(9 *N,%r10,Mm,Chi_30)
-	       VMADDSUBRDUP(10*N,%r10,Mm,Chi_31)
-	       VMADDSUBRDUP(11*N,%r10,Mm,Chi_32)
-	       );
-	}
-	a0 = a0+incr;
-	a1 = a1+incr;
-	a2 = a2+sizeof(typename Simd::scalar_type);
-      }}
-    {
-      int lexa = s1+LLs*site;
-      /*
-      SiteSpinor tmp;
-      asm (
-	       VSTORE(0,%0,pChi_00) VSTORE(1 ,%0,pChi_01)  VSTORE(2 ,%0,pChi_02)		
-	       VSTORE(3,%0,pChi_10) VSTORE(4 ,%0,pChi_11)  VSTORE(5 ,%0,pChi_12)		
-	       VSTORE(6,%0,pChi_20) VSTORE(7 ,%0,pChi_21)  VSTORE(8 ,%0,pChi_22)		
-	       VSTORE(9,%0,pChi_30) VSTORE(10,%0,pChi_31)  VSTORE(11,%0,pChi_32)		
-	       : : "r" ((uint64_t)&tmp) : "memory" );
-      */
-
-      asm (
-	       VSTORE(0,%0,pChi_00) VSTORE(1 ,%0,pChi_01)  VSTORE(2 ,%0,pChi_02)		
-	       VSTORE(3,%0,pChi_10) VSTORE(4 ,%0,pChi_11)  VSTORE(5 ,%0,pChi_12)		
-	       VSTORE(6,%0,pChi_20) VSTORE(7 ,%0,pChi_21)  VSTORE(8 ,%0,pChi_22)		
-	       VSTORE(9,%0,pChi_30) VSTORE(10,%0,pChi_31)  VSTORE(11,%0,pChi_32)		
-	       : : "r" ((uint64_t)&chi[lexa]) : "memory" );
-
-      //      if ( 1 || (site==0) ) { 
-      //	std::cout<<site << " s1 "<<s1<<"\n\t"<<tmp << "\n't" << chi[lexa] <<"\n\t"<<tmp-chi[lexa]<<std::endl;
-      //      }
-    }
-  }
-  }
-#undef Chi_00
-#undef Chi_01
-#undef Chi_02
-#undef Chi_10
-#undef Chi_11
-#undef Chi_12
-#undef Chi_20
-#undef Chi_21
-#undef Chi_22
-#undef Chi_30
-#undef Chi_31
-#undef Chi_32
-
-#undef BCAST0
-#undef BCAST1
-#undef BCAST2
-#undef BCAST3
-#undef BCAST4
-#undef BCAST5
-#undef BCAST6
-#undef BCAST7
-#undef BCAST8
-#undef BCAST9
-#undef BCAST10
-#undef BCAST11
-
-#endif
-};
-
-
-template<class Impl>
-void CayleyFermion5D<Impl>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv)
-{
-  int Ls=this->Ls;
-  int LLs = psi._grid->_rdimensions[0];
-  int vol = psi._grid->oSites()/LLs;
-
-  chi.checkerboard=psi.checkerboard;
-  
-  Vector<iSinglet<Simd> >  Matp;
-  Vector<iSinglet<Simd> >  Matm;
-  Vector<iSinglet<Simd> >  *_Matp;
-  Vector<iSinglet<Simd> >  *_Matm;
-  
-  //  MooeeInternalCompute(dag,inv,Matp,Matm);
-  if ( inv && dag ) { 
-    _Matp = &MatpInvDag;
-    _Matm = &MatmInvDag;
-  }
-  if ( inv && (!dag) ) { 
-    _Matp = &MatpInv;
-    _Matm = &MatmInv;
-  } 
-  if ( !inv ) {
-    MooeeInternalCompute(dag,inv,Matp,Matm);
-    _Matp = &Matp;
-    _Matm = &Matm;
-  }
-  assert(_Matp->size()==Ls*LLs);
-
-  MooeeInvCalls++;
-  MooeeInvTime-=usecond();
-
-  if ( switcheroo<Coeff_t>::iscomplex() ) {
-    parallel_for(auto site=0;site<vol;site++){
-      MooeeInternalZAsm(psi,chi,LLs,site,*_Matp,*_Matm);
-    }
-  } else { 
-    parallel_for(auto site=0;site<vol;site++){
-      MooeeInternalAsm(psi,chi,LLs,site,*_Matp,*_Matm);
-    }
-  }
-  MooeeInvTime+=usecond();
-}
-
-INSTANTIATE_DPERP(DomainWallVec5dImplD);
-INSTANTIATE_DPERP(DomainWallVec5dImplF);
-INSTANTIATE_DPERP(ZDomainWallVec5dImplD);
-INSTANTIATE_DPERP(ZDomainWallVec5dImplF);
-
-INSTANTIATE_DPERP(DomainWallVec5dImplDF);
-INSTANTIATE_DPERP(DomainWallVec5dImplFH);
-INSTANTIATE_DPERP(ZDomainWallVec5dImplDF);
-INSTANTIATE_DPERP(ZDomainWallVec5dImplFH);
-
-template void CayleyFermion5D<DomainWallVec5dImplF>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-template void CayleyFermion5D<DomainWallVec5dImplD>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-template void CayleyFermion5D<ZDomainWallVec5dImplF>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-template void CayleyFermion5D<ZDomainWallVec5dImplD>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-
-template void CayleyFermion5D<DomainWallVec5dImplFH>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-template void CayleyFermion5D<DomainWallVec5dImplDF>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-template void CayleyFermion5D<ZDomainWallVec5dImplFH>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-template void CayleyFermion5D<ZDomainWallVec5dImplDF>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
-
-
-
-}}
diff --git a/Grid/qcd/action/fermion/ContinuedFractionFermion5D.cc b/Grid/qcd/action/fermion/ContinuedFractionFermion5D.cc
deleted file mode 100644
index f6857115..00000000
--- a/Grid/qcd/action/fermion/ContinuedFractionFermion5D.cc
+++ /dev/null
@@ -1,323 +0,0 @@
-    /*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid 
-
-    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
-
-    Copyright (C) 2015
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
-
-namespace Grid {
-  namespace QCD {
-
-    template<class Impl>
-    void ContinuedFractionFermion5D<Impl>::SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD scale)
-    {
-      SetCoefficientsZolotarev(1.0/scale,zdata);
-    }
-    template<class Impl>
-    void ContinuedFractionFermion5D<Impl>::SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata)
-    {
-      // How to check Ls matches??
-      //      std::cout<<GridLogMessage << Ls << " Ls"<<std::endl;
-      //      std::cout<<GridLogMessage << zdata->n  << " - n"<<std::endl;
-      //      std::cout<<GridLogMessage << zdata->da << " -da "<<std::endl;
-      //      std::cout<<GridLogMessage << zdata->db << " -db"<<std::endl;
-      //      std::cout<<GridLogMessage << zdata->dn << " -dn"<<std::endl;
-      //      std::cout<<GridLogMessage << zdata->dd << " -dd"<<std::endl;
-      int Ls = this->Ls;
-      assert(zdata->db==Ls);// Beta has Ls coeffs
-
-      R=(1+this->mass)/(1-this->mass);
-
-      Beta.resize(Ls);
-      cc.resize(Ls);
-      cc_d.resize(Ls);
-      sqrt_cc.resize(Ls);
-      for(int i=0; i < Ls ; i++){
-	Beta[i] = zdata -> beta[i];
-	cc[i] = 1.0/Beta[i];
-	cc_d[i]=sqrt(cc[i]);
-      }
-    
-      cc_d[Ls-1]=1.0;
-      for(int i=0; i < Ls-1 ; i++){
-	sqrt_cc[i]= sqrt(cc[i]*cc[i+1]);
-      }    
-      sqrt_cc[Ls-2]=sqrt(cc[Ls-2]);
-
-
-      ZoloHiInv =1.0/zolo_hi;
-      dw_diag = (4.0-this->M5)*ZoloHiInv;
-    
-      See.resize(Ls);
-      Aee.resize(Ls);
-      int sign=1;
-      for(int s=0;s<Ls;s++){
-	Aee[s] = sign * Beta[s] * dw_diag;
-	sign   = - sign;
-      }
-      Aee[Ls-1] += R;
-    
-      See[0] = Aee[0];
-      for(int s=1;s<Ls;s++){
-	See[s] = Aee[s] - 1.0/See[s-1];
-      }
-      for(int s=0;s<Ls;s++){
-	std::cout<<GridLogMessage <<"s = "<<s<<" Beta "<<Beta[s]<<" Aee "<<Aee[s] <<" See "<<See[s] <<std::endl;
-      }
-    }
-
-
-
-    template<class Impl>
-    RealD  ContinuedFractionFermion5D<Impl>::M           (const FermionField &psi, FermionField &chi)
-    {
-      int Ls = this->Ls;
-
-      FermionField D(psi._grid);
-
-      this->DW(psi,D,DaggerNo); 
-
-      int sign=1;
-      for(int s=0;s<Ls;s++){
-	if ( s==0 ) {
-	  ag5xpby_ssp(chi,cc[0]*Beta[0]*sign*ZoloHiInv,D,sqrt_cc[0],psi,s,s+1); // Multiplies Dw by G5 so Hw
-	} else if ( s==(Ls-1) ){
-	  RealD R=(1.0+mass)/(1.0-mass);
-	  ag5xpby_ssp(chi,Beta[s]*ZoloHiInv,D,sqrt_cc[s-1],psi,s,s-1);
-	  ag5xpby_ssp(chi,R,psi,1.0,chi,s,s);
-	} else {
-	  ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*ZoloHiInv,D,sqrt_cc[s],psi,s,s+1);
-  	  axpby_ssp(chi,1.0,chi,sqrt_cc[s-1],psi,s,s-1);
-	}
-	sign=-sign; 
-      }
-      return norm2(chi);
-    }
-    template<class Impl>
-    RealD  ContinuedFractionFermion5D<Impl>::Mdag        (const FermionField &psi, FermionField &chi)
-    {
-      // This matrix is already hermitian. (g5 Dw) = Dw dag g5 = (g5 Dw)dag
-      // The rest of matrix is symmetric.
-      // Can ignore "dag"
-      return M(psi,chi);
-    }
-    template<class Impl>
-    void  ContinuedFractionFermion5D<Impl>::Mdir (const FermionField &psi, FermionField &chi,int dir,int disp){
-      int Ls = this->Ls;
-
-      this->DhopDir(psi,chi,dir,disp); // Dslash on diagonal. g5 Dslash is hermitian
-
-      int sign=1;
-      for(int s=0;s<Ls;s++){
-	if ( s==(Ls-1) ){
-	  ag5xpby_ssp(chi,Beta[s]*ZoloHiInv,chi,0.0,chi,s,s);
-	} else {
-	  ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*ZoloHiInv,chi,0.0,chi,s,s);
-	}
-	sign=-sign; 
-      }
-    }
-    template<class Impl>
-    void   ContinuedFractionFermion5D<Impl>::Meooe       (const FermionField &psi, FermionField &chi)
-    {
-      int Ls = this->Ls;
-
-      // Apply 4d dslash
-      if ( psi.checkerboard == Odd ) {
-	this->DhopEO(psi,chi,DaggerNo); // Dslash on diagonal. g5 Dslash is hermitian
-      } else {
-	this->DhopOE(psi,chi,DaggerNo); // Dslash on diagonal. g5 Dslash is hermitian
-      }
-      
-      int sign=1;
-      for(int s=0;s<Ls;s++){
-	if ( s==(Ls-1) ){
-	  ag5xpby_ssp(chi,Beta[s]*ZoloHiInv,chi,0.0,chi,s,s);
-	} else {
-	  ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*ZoloHiInv,chi,0.0,chi,s,s);
-	}
-	sign=-sign; 
-      }
-    }
-    template<class Impl>
-    void   ContinuedFractionFermion5D<Impl>::MeooeDag    (const FermionField &psi, FermionField &chi)
-    {
-      this->Meooe(psi,chi);
-    }
-    template<class Impl>
-    void   ContinuedFractionFermion5D<Impl>::Mooee       (const FermionField &psi, FermionField &chi)
-    {
-      int Ls = this->Ls;
-
-      int sign=1;
-      for(int s=0;s<Ls;s++){
-	if ( s==0 ) {
-	  ag5xpby_ssp(chi,cc[0]*Beta[0]*sign*dw_diag,psi,sqrt_cc[0],psi,s,s+1); // Multiplies Dw by G5 so Hw
-	} else if ( s==(Ls-1) ){
-	  // Drop the CC here.
-	  double R=(1+mass)/(1-mass);
-	  ag5xpby_ssp(chi,Beta[s]*dw_diag,psi,sqrt_cc[s-1],psi,s,s-1);
-	  ag5xpby_ssp(chi,R,psi,1.0,chi,s,s);
-	} else {
-	  ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*dw_diag,psi,sqrt_cc[s],psi,s,s+1);
-	  axpby_ssp(chi,1.0,chi,sqrt_cc[s-1],psi,s,s-1);
-	}
-	sign=-sign; 
-      }
-    }
-
-    template<class Impl>
-    void   ContinuedFractionFermion5D<Impl>::MooeeDag    (const FermionField &psi, FermionField &chi)
-    {
-      this->Mooee(psi,chi);
-    }
-    template<class Impl>
-    void   ContinuedFractionFermion5D<Impl>::MooeeInv    (const FermionField &psi, FermionField &chi)
-    {
-      int Ls = this->Ls;
-
-      // Apply Linv
-      axpby_ssp(chi,1.0/cc_d[0],psi,0.0,psi,0,0); 
-      for(int s=1;s<Ls;s++){
-	axpbg5y_ssp(chi,1.0/cc_d[s],psi,-1.0/See[s-1],chi,s,s-1);
-      }
-      // Apply Dinv
-      for(int s=0;s<Ls;s++){
-	ag5xpby_ssp(chi,1.0/See[s],chi,0.0,chi,s,s); //only appearance of See[0]
-      }
-      // Apply Uinv = (Linv)^T
-      axpby_ssp(chi,1.0/cc_d[Ls-1],chi,0.0,chi,Ls-1,Ls-1);
-      for(int s=Ls-2;s>=0;s--){
-	axpbg5y_ssp(chi,1.0/cc_d[s],chi,-1.0*cc_d[s+1]/See[s]/cc_d[s],chi,s,s+1);
-      }
-    }
-    template<class Impl>
-    void   ContinuedFractionFermion5D<Impl>::MooeeInvDag (const FermionField &psi, FermionField &chi)
-    {
-      this->MooeeInv(psi,chi);
-    }
-
-  // force terms; five routines; default to Dhop on diagonal
-    template<class Impl>
-   void ContinuedFractionFermion5D<Impl>::MDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
-  {
-    int Ls = this->Ls;
-
-    FermionField D(V._grid);
-
-    int sign=1;
-    for(int s=0;s<Ls;s++){
-      if ( s==(Ls-1) ){
-	ag5xpby_ssp(D,Beta[s]*ZoloHiInv,U,0.0,U,s,s);
-      } else {
-	ag5xpby_ssp(D,cc[s]*Beta[s]*sign*ZoloHiInv,U,0.0,U,s,s);
-      }
-      sign=-sign; 
-    }
-    this->DhopDeriv(mat,D,V,DaggerNo); 
-  };
-    template<class Impl>
-   void ContinuedFractionFermion5D<Impl>::MoeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
-  {
-    int Ls = this->Ls;
-
-    FermionField D(V._grid);
-
-    int sign=1;
-    for(int s=0;s<Ls;s++){
-      if ( s==(Ls-1) ){
-	ag5xpby_ssp(D,Beta[s]*ZoloHiInv,U,0.0,U,s,s);
-      } else {
-	ag5xpby_ssp(D,cc[s]*Beta[s]*sign*ZoloHiInv,U,0.0,U,s,s);
-      }
-      sign=-sign; 
-    }
-    this->DhopDerivOE(mat,D,V,DaggerNo); 
-  };
-  template<class Impl>
-  void ContinuedFractionFermion5D<Impl>::MeoDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
-  {
-    int Ls = this->Ls;
-
-    FermionField D(V._grid);
-
-    int sign=1;
-    for(int s=0;s<Ls;s++){
-      if ( s==(Ls-1) ){
-	ag5xpby_ssp(D,Beta[s]*ZoloHiInv,U,0.0,U,s,s);
-      } else {
-	ag5xpby_ssp(D,cc[s]*Beta[s]*sign*ZoloHiInv,U,0.0,U,s,s);
-      }
-      sign=-sign; 
-    }
-    this->DhopDerivEO(mat,D,V,DaggerNo); 
-  };
-    
-    // Constructors
-    template<class Impl>
-    ContinuedFractionFermion5D<Impl>::ContinuedFractionFermion5D(
-							   GaugeField &_Umu,
-							   GridCartesian         &FiveDimGrid,
-							   GridRedBlackCartesian &FiveDimRedBlackGrid,
-							   GridCartesian         &FourDimGrid,
-							   GridRedBlackCartesian &FourDimRedBlackGrid,
-							   RealD _mass,RealD M5,const ImplParams &p) :
-      WilsonFermion5D<Impl>(_Umu,
-			    FiveDimGrid, FiveDimRedBlackGrid,
-			    FourDimGrid, FourDimRedBlackGrid,M5,p),
-      mass(_mass)
-    {
-      int Ls = this->Ls;
-      assert((Ls&0x1)==1); // Odd Ls required
-    }
-
-    template<class Impl>
-    void ContinuedFractionFermion5D<Impl>::ExportPhysicalFermionSolution(const FermionField &solution5d,FermionField &exported4d)
-    {
-      int Ls = this->Ls;
-      conformable(solution5d._grid,this->FermionGrid());
-      conformable(exported4d._grid,this->GaugeGrid());
-      ExtractSlice(exported4d, solution5d, Ls-1, Ls-1);
-    }
-    template<class Impl>
-    void ContinuedFractionFermion5D<Impl>::ImportPhysicalFermionSource(const FermionField &input4d,FermionField &imported5d)
-    {
-      int Ls = this->Ls;
-      conformable(imported5d._grid,this->FermionGrid());
-      conformable(input4d._grid   ,this->GaugeGrid());
-      FermionField tmp(this->FermionGrid());
-      tmp=zero;
-      InsertSlice(input4d, tmp, Ls-1, Ls-1);
-      tmp=Gamma(Gamma::Algebra::Gamma5)*tmp;
-      this->Dminus(tmp,imported5d);
-    }
-
-    FermOpTemplateInstantiate(ContinuedFractionFermion5D);
-
-  }
-}
-
diff --git a/Grid/qcd/action/fermion/ContinuedFractionFermion5D.h b/Grid/qcd/action/fermion/ContinuedFractionFermion5D.h
index b551fc28..379c5f8f 100644
--- a/Grid/qcd/action/fermion/ContinuedFractionFermion5D.h
+++ b/Grid/qcd/action/fermion/ContinuedFractionFermion5D.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,46 +24,44 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef  GRID_QCD_CONTINUED_FRACTION_H
 #define  GRID_QCD_CONTINUED_FRACTION_H
 
 #include <Grid/qcd/action/fermion/WilsonFermion5D.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class ContinuedFractionFermion5D : public WilsonFermion5D<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+public:
 
-    template<class Impl>
-    class ContinuedFractionFermion5D : public WilsonFermion5D<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-    public:
+  // override multiply
+  virtual RealD  M    (const FermionField &in, FermionField &out);
+  virtual RealD  Mdag (const FermionField &in, FermionField &out);
 
-      // override multiply
-      virtual RealD  M    (const FermionField &in, FermionField &out);
-      virtual RealD  Mdag (const FermionField &in, FermionField &out);
+  // half checkerboard operaions
+  virtual void   Meooe       (const FermionField &in, FermionField &out);
+  virtual void   MeooeDag    (const FermionField &in, FermionField &out);
+  virtual void   Mooee       (const FermionField &in, FermionField &out);
+  virtual void   MooeeDag    (const FermionField &in, FermionField &out);
+  virtual void   MooeeInv    (const FermionField &in, FermionField &out);
+  virtual void   MooeeInvDag (const FermionField &in, FermionField &out);
 
-      // half checkerboard operaions
-      virtual void   Meooe       (const FermionField &in, FermionField &out);
-      virtual void   MeooeDag    (const FermionField &in, FermionField &out);
-      virtual void   Mooee       (const FermionField &in, FermionField &out);
-      virtual void   MooeeDag    (const FermionField &in, FermionField &out);
-      virtual void   MooeeInv    (const FermionField &in, FermionField &out);
-      virtual void   MooeeInvDag (const FermionField &in, FermionField &out);
+  // force terms; five routines; default to Dhop on diagonal
+  virtual void MDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  virtual void MoeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  virtual void MeoDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
 
-      // force terms; five routines; default to Dhop on diagonal
-      virtual void MDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
-      virtual void MoeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
-      virtual void MeoDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  //      virtual void   Instantiatable(void)=0;
+  virtual void   Instantiatable(void) =0;
 
-      //      virtual void   Instantiatable(void)=0;
-      virtual void   Instantiatable(void) =0;
-
-      // Efficient support for multigrid coarsening
-      virtual void  Mdir (const FermionField &in, FermionField &out,int dir,int disp);
+  // Efficient support for multigrid coarsening
+  virtual void  Mdir (const FermionField &in, FermionField &out,int dir,int disp);
 
       ///////////////////////////////////////////////////////////////
       // Physical surface field utilities
@@ -73,35 +71,34 @@ namespace Grid {
       virtual void ExportPhysicalFermionSolution(const FermionField &solution5d,FermionField &exported4d);
       virtual void ImportPhysicalFermionSource  (const FermionField &input4d,FermionField &imported5d);
 
-      // Constructors
-      ContinuedFractionFermion5D(GaugeField &_Umu,
-				 GridCartesian         &FiveDimGrid,
-				 GridRedBlackCartesian &FiveDimRedBlackGrid,
-				 GridCartesian         &FourDimGrid,
-				 GridRedBlackCartesian &FourDimRedBlackGrid,
-				 RealD _mass,RealD M5,const ImplParams &p= ImplParams());
+  // Constructors
+  ContinuedFractionFermion5D(GaugeField &_Umu,
+			     GridCartesian         &FiveDimGrid,
+			     GridRedBlackCartesian &FiveDimRedBlackGrid,
+			     GridCartesian         &FourDimGrid,
+			     GridRedBlackCartesian &FourDimRedBlackGrid,
+			     RealD _mass,RealD M5,const ImplParams &p= ImplParams());
 
-    protected:
+protected:
 
-      void SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD scale);
-      void SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata);;
+  void SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD scale);
+  void SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata);;
 
-      // Cont frac
-      RealD dw_diag;
-      RealD mass;
-      RealD R;
-      RealD ZoloHiInv;
-      std::vector<double> Beta;
-      std::vector<double> cc;;
-      std::vector<double> cc_d;;
-      std::vector<double> sqrt_cc;
-      std::vector<double> See;
-      std::vector<double> Aee;
+  // Cont frac
+  RealD dw_diag;
+  RealD mass;
+  RealD R;
+  RealD ZoloHiInv;
+  Vector<double> Beta;
+  Vector<double> cc;;
+  Vector<double> cc_d;;
+  Vector<double> sqrt_cc;
+  Vector<double> See;
+  Vector<double> Aee;
 
-    };
+};
 
 
-  }
-}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/fermion/DomainWallEOFAFermion.cc b/Grid/qcd/action/fermion/DomainWallEOFAFermion.cc
deleted file mode 100644
index 37ab5fa6..00000000
--- a/Grid/qcd/action/fermion/DomainWallEOFAFermion.cc
+++ /dev/null
@@ -1,438 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
-
-Copyright (C) 2017
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-Author: David Murphy <dmurphy@phys.columbia.edu>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-#include <Grid/Grid_Eigen_Dense.h>
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
-
-namespace Grid {
-namespace QCD {
-
-    template<class Impl>
-    DomainWallEOFAFermion<Impl>::DomainWallEOFAFermion(
-      GaugeField            &_Umu,
-      GridCartesian         &FiveDimGrid,
-      GridRedBlackCartesian &FiveDimRedBlackGrid,
-      GridCartesian         &FourDimGrid,
-      GridRedBlackCartesian &FourDimRedBlackGrid,
-      RealD _mq1, RealD _mq2, RealD _mq3,
-      RealD _shift, int _pm, RealD _M5, const ImplParams &p) :
-    AbstractEOFAFermion<Impl>(_Umu, FiveDimGrid, FiveDimRedBlackGrid,
-        FourDimGrid, FourDimRedBlackGrid, _mq1, _mq2, _mq3,
-        _shift, _pm, _M5, 1.0, 0.0, p)
-    {
-        RealD eps = 1.0;
-        Approx::zolotarev_data *zdata = Approx::higham(eps,this->Ls);
-        assert(zdata->n == this->Ls);
-
-        std::cout << GridLogMessage << "DomainWallEOFAFermion with Ls=" << this->Ls << std::endl;
-        this->SetCoefficientsTanh(zdata, 1.0, 0.0);
-
-        Approx::zolotarev_free(zdata);
-    }
-
-    /***************************************************************
-     * Additional EOFA operators only called outside the inverter.
-     * Since speed is not essential, simple axpby-style
-     * implementations should be fine.
-     ***************************************************************/
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::Omega(const FermionField& psi, FermionField& Din, int sign, int dag)
-    {
-        int Ls = this->Ls;
-
-        Din = zero;
-        if((sign == 1) && (dag == 0)){ axpby_ssp(Din, 0.0, psi, 1.0, psi, Ls-1, 0); }
-        else if((sign == -1) && (dag == 0)){ axpby_ssp(Din, 0.0, psi, 1.0, psi, 0, 0); }
-        else if((sign == 1 ) && (dag == 1)){ axpby_ssp(Din, 0.0, psi, 1.0, psi, 0, Ls-1); }
-        else if((sign == -1) && (dag == 1)){ axpby_ssp(Din, 0.0, psi, 1.0, psi, 0, 0); }
-    }
-
-    // This is just the identity for DWF
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::Dtilde(const FermionField& psi, FermionField& chi){ chi = psi; }
-
-    // This is just the identity for DWF
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::DtildeInv(const FermionField& psi, FermionField& chi){ chi = psi; }
-
-    /*****************************************************************************************************/
-
-    template<class Impl>
-    RealD DomainWallEOFAFermion<Impl>::M(const FermionField& psi, FermionField& chi)
-    {
-        int Ls = this->Ls;
-
-        FermionField Din(psi._grid);
-
-        this->Meooe5D(psi, Din);
-        this->DW(Din, chi, DaggerNo);
-        axpby(chi, 1.0, 1.0, chi, psi);
-        this->M5D(psi, chi);
-        return(norm2(chi));
-    }
-
-    template<class Impl>
-    RealD DomainWallEOFAFermion<Impl>::Mdag(const FermionField& psi, FermionField& chi)
-    {
-        int Ls = this->Ls;
-
-        FermionField Din(psi._grid);
-
-        this->DW(psi, Din, DaggerYes);
-        this->MeooeDag5D(Din, chi);
-        this->M5Ddag(psi, chi);
-        axpby(chi, 1.0, 1.0, chi, psi);
-        return(norm2(chi));
-    }
-
-    /********************************************************************
-     * Performance critical fermion operators called inside the inverter
-     ********************************************************************/
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::M5D(const FermionField& psi, FermionField& chi)
-    {
-        int   Ls    = this->Ls;
-        int   pm    = this->pm;
-        RealD shift = this->shift;
-        RealD mq1   = this->mq1;
-        RealD mq2   = this->mq2;
-        RealD mq3   = this->mq3;
-
-        // coefficients for shift operator ( = shift*\gamma_{5}*R_{5}*\Delta_{\pm}(mq2,mq3)*P_{\pm} )
-        Coeff_t shiftp(0.0), shiftm(0.0);
-        if(shift != 0.0){
-          if(pm == 1){ shiftp = shift*(mq3-mq2); }
-          else{ shiftm = -shift*(mq3-mq2); }
-        }
-
-        std::vector<Coeff_t> diag(Ls,1.0);
-        std::vector<Coeff_t> upper(Ls,-1.0); upper[Ls-1] = mq1 + shiftm;
-        std::vector<Coeff_t> lower(Ls,-1.0); lower[0]    = mq1 + shiftp;
-
-        #if(0)
-            std::cout << GridLogMessage << "DomainWallEOFAFermion::M5D(FF&,FF&):" << std::endl;
-            for(int i=0; i<diag.size(); ++i){
-                std::cout << GridLogMessage << "diag[" << i << "] =" << diag[i] << std::endl;
-            }
-            for(int i=0; i<upper.size(); ++i){
-                std::cout << GridLogMessage << "upper[" << i << "] =" << upper[i] << std::endl;
-            }
-            for(int i=0; i<lower.size(); ++i){
-                std::cout << GridLogMessage << "lower[" << i << "] =" << lower[i] << std::endl;
-            }
-        #endif
-
-        this->M5D(psi, chi, chi, lower, diag, upper);
-    }
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::M5Ddag(const FermionField& psi, FermionField& chi)
-    {
-        int   Ls    = this->Ls;
-        int   pm    = this->pm;
-        RealD shift = this->shift;
-        RealD mq1   = this->mq1;
-        RealD mq2   = this->mq2;
-        RealD mq3   = this->mq3;
-
-        // coefficients for shift operator ( = shift*\gamma_{5}*R_{5}*\Delta_{\pm}(mq2,mq3)*P_{\pm} )
-        Coeff_t shiftp(0.0), shiftm(0.0);
-        if(shift != 0.0){
-          if(pm == 1){ shiftp = shift*(mq3-mq2); }
-          else{ shiftm = -shift*(mq3-mq2); }
-        }
-
-        std::vector<Coeff_t> diag(Ls,1.0);
-        std::vector<Coeff_t> upper(Ls,-1.0); upper[Ls-1] = mq1 + shiftp;
-        std::vector<Coeff_t> lower(Ls,-1.0); lower[0]    = mq1 + shiftm;
-
-        #if(0)
-            std::cout << GridLogMessage << "DomainWallEOFAFermion::M5Ddag(FF&,FF&):" << std::endl;
-            for(int i=0; i<diag.size(); ++i){
-                std::cout << GridLogMessage << "diag[" << i << "] =" << diag[i] << std::endl;
-            }
-            for(int i=0; i<upper.size(); ++i){
-                std::cout << GridLogMessage << "upper[" << i << "] =" << upper[i] << std::endl;
-            }
-            for(int i=0; i<lower.size(); ++i){
-                std::cout << GridLogMessage << "lower[" << i << "] =" << lower[i] << std::endl;
-            }
-        #endif
-
-        this->M5Ddag(psi, chi, chi, lower, diag, upper);
-    }
-
-    // half checkerboard operations
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::Mooee(const FermionField& psi, FermionField& chi)
-    {
-        int Ls = this->Ls;
-
-        std::vector<Coeff_t> diag = this->bee;
-        std::vector<Coeff_t> upper(Ls);
-        std::vector<Coeff_t> lower(Ls);
-
-        for(int s=0; s<Ls; s++){
-          upper[s] = -this->cee[s];
-          lower[s] = -this->cee[s];
-        }
-        upper[Ls-1] = this->dm;
-        lower[0]    = this->dp;
-
-        this->M5D(psi, psi, chi, lower, diag, upper);
-    }
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::MooeeDag(const FermionField& psi, FermionField& chi)
-    {
-        int Ls = this->Ls;
-
-        std::vector<Coeff_t> diag = this->bee;
-        std::vector<Coeff_t> upper(Ls);
-        std::vector<Coeff_t> lower(Ls);
-
-        for(int s=0; s<Ls; s++){
-          upper[s] = -this->cee[s];
-          lower[s] = -this->cee[s];
-        }
-        upper[Ls-1] = this->dp;
-        lower[0]    = this->dm;
-
-        this->M5Ddag(psi, psi, chi, lower, diag, upper);
-    }
-
-    /****************************************************************************************/
-
-    //Zolo
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::SetCoefficientsInternal(RealD zolo_hi, std::vector<Coeff_t>& gamma, RealD b, RealD c)
-    {
-        int   Ls    = this->Ls;
-        int   pm    = this->pm;
-        RealD mq1   = this->mq1;
-        RealD mq2   = this->mq2;
-        RealD mq3   = this->mq3;
-        RealD shift = this->shift;
-
-        ////////////////////////////////////////////////////////
-        // Constants for the preconditioned matrix Cayley form
-        ////////////////////////////////////////////////////////
-        this->bs.resize(Ls);
-        this->cs.resize(Ls);
-        this->aee.resize(Ls);
-        this->aeo.resize(Ls);
-        this->bee.resize(Ls);
-        this->beo.resize(Ls);
-        this->cee.resize(Ls);
-        this->ceo.resize(Ls);
-
-        for(int i=0; i<Ls; ++i){
-          this->bee[i] = 4.0 - this->M5 + 1.0;
-          this->cee[i] = 1.0;
-        }
-
-        for(int i=0; i<Ls; ++i){
-          this->aee[i] = this->cee[i];
-          this->bs[i] = this->beo[i] = 1.0;
-          this->cs[i] = this->ceo[i] = 0.0;
-        }
-
-        //////////////////////////////////////////
-        // EOFA shift terms
-        //////////////////////////////////////////
-        if(pm == 1){
-          this->dp = mq1*this->cee[0] + shift*(mq3-mq2);
-          this->dm = mq1*this->cee[Ls-1];
-        } else if(this->pm == -1) {
-          this->dp = mq1*this->cee[0];
-          this->dm = mq1*this->cee[Ls-1] - shift*(mq3-mq2);
-        } else {
-          this->dp = mq1*this->cee[0];
-          this->dm = mq1*this->cee[Ls-1];
-        }
-
-        //////////////////////////////////////////
-        // LDU decomposition of eeoo
-        //////////////////////////////////////////
-        this->dee.resize(Ls+1);
-        this->lee.resize(Ls);
-        this->leem.resize(Ls);
-        this->uee.resize(Ls);
-        this->ueem.resize(Ls);
-
-        for(int i=0; i<Ls; ++i){
-
-          if(i < Ls-1){
-
-            this->lee[i] = -this->cee[i+1]/this->bee[i]; // sub-diag entry on the ith column
-
-            this->leem[i] = this->dm/this->bee[i];
-            for(int j=0; j<i; j++){ this->leem[i] *= this->aee[j]/this->bee[j]; }
-
-            this->dee[i] = this->bee[i];
-
-            this->uee[i] = -this->aee[i]/this->bee[i];   // up-diag entry on the ith row
-
-            this->ueem[i] = this->dp / this->bee[0];
-            for(int j=1; j<=i; j++){ this->ueem[i] *= this->cee[j]/this->bee[j]; }
-
-          } else {
-
-            this->lee[i]  = 0.0;
-            this->leem[i] = 0.0;
-            this->uee[i]  = 0.0;
-            this->ueem[i] = 0.0;
-
-          }
-        }
-
-        {
-          Coeff_t delta_d = 1.0 / this->bee[0];
-          for(int j=1; j<Ls-1; j++){ delta_d *= this->cee[j] / this->bee[j]; }
-          this->dee[Ls-1] = this->bee[Ls-1] + this->cee[0] * this->dm * delta_d;
-          this->dee[Ls] = this->bee[Ls-1] + this->cee[Ls-1] * this->dp * delta_d;
-        }
-
-        int inv = 1;
-        this->MooeeInternalCompute(0, inv, this->MatpInv, this->MatmInv);
-        this->MooeeInternalCompute(1, inv, this->MatpInvDag, this->MatmInvDag);
-    }
-
-    // Recompute Cayley-form coefficients for different shift
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::RefreshShiftCoefficients(RealD new_shift)
-    {
-        this->shift = new_shift;
-        Approx::zolotarev_data *zdata = Approx::higham(1.0, this->Ls);
-        this->SetCoefficientsTanh(zdata, 1.0, 0.0);
-    }
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::MooeeInternalCompute(int dag, int inv,
-        Vector<iSinglet<Simd> >& Matp, Vector<iSinglet<Simd> >& Matm)
-    {
-        int Ls = this->Ls;
-
-        GridBase* grid = this->FermionRedBlackGrid();
-        int LLs = grid->_rdimensions[0];
-
-        if(LLs == Ls){ return; } // Not vectorised in 5th direction
-
-        Eigen::MatrixXcd Pplus  = Eigen::MatrixXcd::Zero(Ls,Ls);
-        Eigen::MatrixXcd Pminus = Eigen::MatrixXcd::Zero(Ls,Ls);
-
-        for(int s=0; s<Ls; s++){
-            Pplus(s,s)  = this->bee[s];
-            Pminus(s,s) = this->bee[s];
-        }
-
-        for(int s=0; s<Ls-1; s++){
-            Pminus(s,s+1) = -this->cee[s];
-        }
-
-        for(int s=0; s<Ls-1; s++){
-            Pplus(s+1,s) = -this->cee[s+1];
-        }
-
-        Pplus (0,Ls-1) = this->dp;
-        Pminus(Ls-1,0) = this->dm;
-
-        Eigen::MatrixXcd PplusMat ;
-        Eigen::MatrixXcd PminusMat;
-
-        #if(0)
-            std::cout << GridLogMessage << "Pplus:" << std::endl;
-            for(int s=0; s<Ls; ++s){
-                for(int ss=0; ss<Ls; ++ss){
-                    std::cout << Pplus(s,ss) << "\t";
-                }
-                std::cout << std::endl;
-            }
-            std::cout << GridLogMessage << "Pminus:" << std::endl;
-            for(int s=0; s<Ls; ++s){
-                for(int ss=0; ss<Ls; ++ss){
-                    std::cout << Pminus(s,ss) << "\t";
-                }
-                std::cout << std::endl;
-            }
-        #endif
-
-        if(inv) {
-            PplusMat  = Pplus.inverse();
-            PminusMat = Pminus.inverse();
-        } else {
-            PplusMat  = Pplus;
-            PminusMat = Pminus;
-        }
-
-        if(dag){
-            PplusMat.adjointInPlace();
-            PminusMat.adjointInPlace();
-        }
-
-        typedef typename SiteHalfSpinor::scalar_type scalar_type;
-        const int Nsimd = Simd::Nsimd();
-        Matp.resize(Ls*LLs);
-        Matm.resize(Ls*LLs);
-
-        for(int s2=0; s2<Ls; s2++){
-        for(int s1=0; s1<LLs; s1++){
-            int istride = LLs;
-            int ostride = 1;
-            Simd Vp;
-            Simd Vm;
-            scalar_type *sp = (scalar_type*) &Vp;
-            scalar_type *sm = (scalar_type*) &Vm;
-            for(int l=0; l<Nsimd; l++){
-                if(switcheroo<Coeff_t>::iscomplex()) {
-                    sp[l] = PplusMat (l*istride+s1*ostride,s2);
-                    sm[l] = PminusMat(l*istride+s1*ostride,s2);
-                } else {
-                    // if real
-                    scalar_type tmp;
-                    tmp = PplusMat (l*istride+s1*ostride,s2);
-                    sp[l] = scalar_type(tmp.real(),tmp.real());
-                    tmp = PminusMat(l*istride+s1*ostride,s2);
-                    sm[l] = scalar_type(tmp.real(),tmp.real());
-                }
-            }
-            Matp[LLs*s2+s1] = Vp;
-            Matm[LLs*s2+s1] = Vm;
-        }}
-    }
-
-    FermOpTemplateInstantiate(DomainWallEOFAFermion);
-    GparityFermOpTemplateInstantiate(DomainWallEOFAFermion);
-
-}}
diff --git a/Grid/qcd/action/fermion/DomainWallEOFAFermion.h b/Grid/qcd/action/fermion/DomainWallEOFAFermion.h
index 5362cda8..a2d0e733 100644
--- a/Grid/qcd/action/fermion/DomainWallEOFAFermion.h
+++ b/Grid/qcd/action/fermion/DomainWallEOFAFermion.h
@@ -26,90 +26,65 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 
 See the full license in the file "LICENSE" in the top level distribution directory
 *************************************************************************************/
-/*  END LEGAL */
-#ifndef  GRID_QCD_DOMAIN_WALL_EOFA_FERMION_H
-#define  GRID_QCD_DOMAIN_WALL_EOFA_FERMION_H
+			   /*  END LEGAL */
+#pragma once
 
 #include <Grid/qcd/action/fermion/AbstractEOFAFermion.h>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
-  template<class Impl>
-  class DomainWallEOFAFermion : public AbstractEOFAFermion<Impl>
-  {
-    public:
-      INHERIT_IMPL_TYPES(Impl);
+template<class Impl>
+class DomainWallEOFAFermion : public AbstractEOFAFermion<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
 
-    public:
-      // Modified (0,Ls-1) and (Ls-1,0) elements of Mooee
-      // for red-black preconditioned Shamir EOFA
-      Coeff_t dm;
-      Coeff_t dp;
+public:
+  // Modified (0,Ls-1) and (Ls-1,0) elements of Mooee
+  // for red-black preconditioned Shamir EOFA
+  Coeff_t dm;
+  Coeff_t dp;
 
-      virtual void Instantiatable(void) {};
+  virtual void Instantiatable(void) {};
 
-      // EOFA-specific operations
-      virtual void  Omega      (const FermionField& in, FermionField& out, int sign, int dag);
-      virtual void  Dtilde     (const FermionField& in, FermionField& out);
-      virtual void  DtildeInv  (const FermionField& in, FermionField& out);
+  // EOFA-specific operations
+  virtual void  Omega      (const FermionField& in, FermionField& out, int sign, int dag);
+  virtual void  Dtilde     (const FermionField& in, FermionField& out);
+  virtual void  DtildeInv  (const FermionField& in, FermionField& out);
 
-      // override multiply
-      virtual RealD M          (const FermionField& in, FermionField& out);
-      virtual RealD Mdag       (const FermionField& in, FermionField& out);
+  // override multiply
+  virtual RealD M          (const FermionField& in, FermionField& out);
+  virtual RealD Mdag       (const FermionField& in, FermionField& out);
 
-      // half checkerboard operations
-      virtual void  Mooee      (const FermionField& in, FermionField& out);
-      virtual void  MooeeDag   (const FermionField& in, FermionField& out);
-      virtual void  MooeeInv   (const FermionField& in, FermionField& out);
-      virtual void  MooeeInvDag(const FermionField& in, FermionField& out);
+  // half checkerboard operations
+  virtual void  Mooee      (const FermionField& in, FermionField& out);
+  virtual void  MooeeDag   (const FermionField& in, FermionField& out);
+  virtual void  MooeeInv   (const FermionField& in, FermionField& out);
+  virtual void  MooeeInvDag(const FermionField& in, FermionField& out);
 
-      virtual void   M5D       (const FermionField& psi, FermionField& chi);
-      virtual void   M5Ddag    (const FermionField& psi, FermionField& chi);
+  virtual void   M5D       (const FermionField& psi, FermionField& chi);
+  virtual void   M5Ddag    (const FermionField& psi, FermionField& chi);
 
-      /////////////////////////////////////////////////////
-      // Instantiate different versions depending on Impl
-      /////////////////////////////////////////////////////
-      void M5D(const FermionField& psi, const FermionField& phi, FermionField& chi,
-        std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper);
+  /////////////////////////////////////////////////////
+  // Instantiate different versions depending on Impl
+  /////////////////////////////////////////////////////
+  void M5D(const FermionField& psi, const FermionField& phi, FermionField& chi,
+	   Vector<Coeff_t>& lower, Vector<Coeff_t>& diag, Vector<Coeff_t>& upper);
 
-      void M5Ddag(const FermionField& psi, const FermionField& phi, FermionField& chi,
-        std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper);
+  void M5Ddag(const FermionField& psi, const FermionField& phi, FermionField& chi,
+	      Vector<Coeff_t>& lower, Vector<Coeff_t>& diag, Vector<Coeff_t>& upper);
 
-      void MooeeInternal(const FermionField& in, FermionField& out, int dag, int inv);
+  virtual void RefreshShiftCoefficients(RealD new_shift);
 
-      void MooeeInternalCompute(int dag, int inv, Vector<iSinglet<Simd>>& Matp, Vector<iSinglet<Simd>>& Matm);
+  // Constructors
+  DomainWallEOFAFermion(GaugeField& _Umu, GridCartesian& FiveDimGrid, GridRedBlackCartesian& FiveDimRedBlackGrid,
+			GridCartesian& FourDimGrid, GridRedBlackCartesian& FourDimRedBlackGrid,
+			RealD _mq1, RealD _mq2, RealD _mq3, RealD _shift, int pm,
+			RealD _M5, const ImplParams& p=ImplParams());
 
-      void MooeeInternalAsm(const FermionField& in, FermionField& out, int LLs, int site,
-        Vector<iSinglet<Simd>>& Matp, Vector<iSinglet<Simd>>& Matm);
+protected:
+  void SetCoefficientsInternal(RealD zolo_hi, Vector<Coeff_t>& gamma, RealD b, RealD c);
+};
 
-      void MooeeInternalZAsm(const FermionField& in, FermionField& out, int LLs, int site,
-        Vector<iSinglet<Simd>>& Matp, Vector<iSinglet<Simd>>& Matm);
+NAMESPACE_END(Grid);
 
-      virtual void RefreshShiftCoefficients(RealD new_shift);
-
-      // Constructors
-      DomainWallEOFAFermion(GaugeField& _Umu, GridCartesian& FiveDimGrid, GridRedBlackCartesian& FiveDimRedBlackGrid,
-        GridCartesian& FourDimGrid, GridRedBlackCartesian& FourDimRedBlackGrid,
-        RealD _mq1, RealD _mq2, RealD _mq3, RealD _shift, int pm,
-        RealD _M5, const ImplParams& p=ImplParams());
-
-    protected:
-      void SetCoefficientsInternal(RealD zolo_hi, std::vector<Coeff_t>& gamma, RealD b, RealD c);
-  };
-}}
-
-#define INSTANTIATE_DPERP_DWF_EOFA(A)\
-template void DomainWallEOFAFermion<A>::M5D(const FermionField& psi, const FermionField& phi, FermionField& chi, \
-  std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper); \
-template void DomainWallEOFAFermion<A>::M5Ddag(const FermionField& psi, const FermionField& phi, FermionField& chi, \
-  std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper); \
-template void DomainWallEOFAFermion<A>::MooeeInv(const FermionField& psi, FermionField& chi); \
-template void DomainWallEOFAFermion<A>::MooeeInvDag(const FermionField& psi, FermionField& chi);
-
-#undef  DOMAIN_WALL_EOFA_DPERP_DENSE
-#define DOMAIN_WALL_EOFA_DPERP_CACHE
-#undef  DOMAIN_WALL_EOFA_DPERP_LINALG
-#define DOMAIN_WALL_EOFA_DPERP_VEC
-
-#endif
diff --git a/Grid/qcd/action/fermion/DomainWallEOFAFermioncache.cc b/Grid/qcd/action/fermion/DomainWallEOFAFermioncache.cc
deleted file mode 100644
index 0b214d31..00000000
--- a/Grid/qcd/action/fermion/DomainWallEOFAFermioncache.cc
+++ /dev/null
@@ -1,248 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermioncache.cc
-
-Copyright (C) 2017
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-Author: David Murphy <dmurphy@phys.columbia.edu>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
-
-namespace Grid {
-namespace QCD {
-
-    // FIXME -- make a version of these routines with site loop outermost for cache reuse.
-
-    // Pminus fowards
-    // Pplus  backwards..
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::M5D(const FermionField& psi, const FermionField& phi,
-        FermionField& chi, std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper)
-    {
-        int Ls = this->Ls;
-        GridBase* grid = psi._grid;
-
-        assert(phi.checkerboard == psi.checkerboard);
-        chi.checkerboard = psi.checkerboard;
-        // Flops = 6.0*(Nc*Ns) *Ls*vol
-        this->M5Dcalls++;
-        this->M5Dtime -= usecond();
-
-        parallel_for(int ss=0; ss<grid->oSites(); ss+=Ls){ // adds Ls
-            for(int s=0; s<Ls; s++){
-                auto tmp = psi._odata[0];
-                if(s==0) {
-                    spProj5m(tmp, psi._odata[ss+s+1]);
-                    chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-                    spProj5p(tmp, psi._odata[ss+Ls-1]);
-                    chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-                } else if(s==(Ls-1)) {
-                    spProj5m(tmp, psi._odata[ss+0]);
-                    chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-                    spProj5p(tmp, psi._odata[ss+s-1]);
-                    chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-                } else {
-                    spProj5m(tmp, psi._odata[ss+s+1]);
-                    chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-                    spProj5p(tmp, psi._odata[ss+s-1]);
-                    chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-                }
-            }
-        }
-
-        this->M5Dtime += usecond();
-    }
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::M5Ddag(const FermionField& psi, const FermionField& phi,
-        FermionField& chi, std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper)
-    {
-        int Ls = this->Ls;
-        GridBase* grid = psi._grid;
-        assert(phi.checkerboard == psi.checkerboard);
-        chi.checkerboard=psi.checkerboard;
-
-        // Flops = 6.0*(Nc*Ns) *Ls*vol
-        this->M5Dcalls++;
-        this->M5Dtime -= usecond();
-
-        parallel_for(int ss=0; ss<grid->oSites(); ss+=Ls){ // adds Ls
-            auto tmp = psi._odata[0];
-            for(int s=0; s<Ls; s++){
-                if(s==0) {
-                    spProj5p(tmp, psi._odata[ss+s+1]);
-                    chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-                    spProj5m(tmp, psi._odata[ss+Ls-1]);
-                    chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-                } else if(s==(Ls-1)) {
-                    spProj5p(tmp, psi._odata[ss+0]);
-                    chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-                    spProj5m(tmp, psi._odata[ss+s-1]);
-                    chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-                } else {
-                    spProj5p(tmp, psi._odata[ss+s+1]);
-                    chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-                    spProj5m(tmp, psi._odata[ss+s-1]);
-                    chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-                }
-            }
-        }
-
-        this->M5Dtime += usecond();
-    }
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::MooeeInv(const FermionField& psi, FermionField& chi)
-    {
-        GridBase* grid = psi._grid;
-        int Ls = this->Ls;
-
-        chi.checkerboard = psi.checkerboard;
-
-        this->MooeeInvCalls++;
-        this->MooeeInvTime -= usecond();
-
-        parallel_for(int ss=0; ss<grid->oSites(); ss+=Ls){ // adds Ls
-
-            auto tmp1 = psi._odata[0];
-            auto tmp2 = psi._odata[0];
-
-            // flops = 12*2*Ls + 12*2*Ls + 3*12*Ls + 12*2*Ls  = 12*Ls * (9) = 108*Ls flops
-            // Apply (L^{\prime})^{-1}
-            chi[ss] = psi[ss]; // chi[0]=psi[0]
-            for(int s=1; s<Ls; s++){
-                spProj5p(tmp1, chi[ss+s-1]);
-                chi[ss+s] = psi[ss+s] - this->lee[s-1]*tmp1;
-            }
-
-            // L_m^{-1}
-            for(int s=0; s<Ls-1; s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
-                spProj5m(tmp1, chi[ss+s]);
-                chi[ss+Ls-1] = chi[ss+Ls-1] - this->leem[s]*tmp1;
-            }
-
-            // U_m^{-1} D^{-1}
-            for(int s=0; s<Ls-1; s++){ // Chi[s] + 1/d chi[s]
-                spProj5p(tmp1, chi[ss+Ls-1]);
-                chi[ss+s] = (1.0/this->dee[s])*chi[ss+s] - (this->ueem[s]/this->dee[Ls])*tmp1;
-            }
-            spProj5m(tmp2, chi[ss+Ls-1]);
-            chi[ss+Ls-1] = (1.0/this->dee[Ls])*tmp1 + (1.0/this->dee[Ls-1])*tmp2;
-
-            // Apply U^{-1}
-            for(int s=Ls-2; s>=0; s--){
-                spProj5m(tmp1, chi[ss+s+1]);
-                chi[ss+s] = chi[ss+s] - this->uee[s]*tmp1;
-            }
-        }
-
-        this->MooeeInvTime += usecond();
-    }
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::MooeeInvDag(const FermionField& psi, FermionField& chi)
-    {
-        GridBase* grid = psi._grid;
-        int Ls = this->Ls;
-
-        assert(psi.checkerboard == psi.checkerboard);
-        chi.checkerboard = psi.checkerboard;
-
-        std::vector<Coeff_t> ueec(Ls);
-        std::vector<Coeff_t> deec(Ls+1);
-        std::vector<Coeff_t> leec(Ls);
-        std::vector<Coeff_t> ueemc(Ls);
-        std::vector<Coeff_t> leemc(Ls);
-
-        for(int s=0; s<ueec.size(); s++){
-            ueec[s]  = conjugate(this->uee[s]);
-            deec[s]  = conjugate(this->dee[s]);
-            leec[s]  = conjugate(this->lee[s]);
-            ueemc[s] = conjugate(this->ueem[s]);
-            leemc[s] = conjugate(this->leem[s]);
-        }
-        deec[Ls] = conjugate(this->dee[Ls]);
-
-        this->MooeeInvCalls++;
-        this->MooeeInvTime -= usecond();
-
-        parallel_for(int ss=0; ss<grid->oSites(); ss+=Ls){ // adds Ls
-
-            auto tmp1 = psi._odata[0];
-            auto tmp2 = psi._odata[0];
-
-            // Apply (U^{\prime})^{-dagger}
-            chi[ss] = psi[ss];
-            for(int s=1; s<Ls; s++){
-                spProj5m(tmp1, chi[ss+s-1]);
-                chi[ss+s] = psi[ss+s] - ueec[s-1]*tmp1;
-            }
-
-            // U_m^{-\dagger}
-            for(int s=0; s<Ls-1; s++){
-                spProj5p(tmp1, chi[ss+s]);
-                chi[ss+Ls-1] = chi[ss+Ls-1] - ueemc[s]*tmp1;
-            }
-
-            // L_m^{-\dagger} D^{-dagger}
-            for(int s=0; s<Ls-1; s++){
-                spProj5m(tmp1, chi[ss+Ls-1]);
-                chi[ss+s] = (1.0/deec[s])*chi[ss+s] - (leemc[s]/deec[Ls-1])*tmp1;
-            }
-            spProj5p(tmp2, chi[ss+Ls-1]);
-            chi[ss+Ls-1] = (1.0/deec[Ls-1])*tmp1 + (1.0/deec[Ls])*tmp2;
-
-            // Apply L^{-dagger}
-            for(int s=Ls-2; s>=0; s--){
-                spProj5p(tmp1, chi[ss+s+1]);
-                chi[ss+s] = chi[ss+s] - leec[s]*tmp1;
-            }
-        }
-
-        this->MooeeInvTime += usecond();
-    }
-
-    #ifdef DOMAIN_WALL_EOFA_DPERP_CACHE
-
-        INSTANTIATE_DPERP_DWF_EOFA(WilsonImplF);
-        INSTANTIATE_DPERP_DWF_EOFA(WilsonImplD);
-        INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplF);
-        INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplD);
-        INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplF);
-        INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplD);
-
-        INSTANTIATE_DPERP_DWF_EOFA(WilsonImplFH);
-        INSTANTIATE_DPERP_DWF_EOFA(WilsonImplDF);
-        INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplFH);
-        INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplDF);
-        INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplFH);
-        INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplDF);
-
-    #endif
-
-}}
diff --git a/Grid/qcd/action/fermion/DomainWallEOFAFermiondense.cc b/Grid/qcd/action/fermion/DomainWallEOFAFermiondense.cc
deleted file mode 100644
index c27074d9..00000000
--- a/Grid/qcd/action/fermion/DomainWallEOFAFermiondense.cc
+++ /dev/null
@@ -1,159 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermiondense.cc
-
-Copyright (C) 2017
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-Author: David Murphy <dmurphy@phys.columbia.edu>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-#include <Grid/Grid_Eigen_Dense.h>
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
-
-namespace Grid {
-namespace QCD {
-
-    /*
-    * Dense matrix versions of routines
-    */
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::MooeeInvDag(const FermionField& psi, FermionField& chi)
-    {
-        this->MooeeInternal(psi, chi, DaggerYes, InverseYes);
-    }
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::MooeeInv(const FermionField& psi, FermionField& chi)
-    {
-        this->MooeeInternal(psi, chi, DaggerNo, InverseYes);
-    }
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv)
-    {
-        int Ls = this->Ls;
-        int LLs = psi._grid->_rdimensions[0];
-        int vol = psi._grid->oSites()/LLs;
-
-        chi.checkerboard = psi.checkerboard;
-
-        assert(Ls==LLs);
-
-        Eigen::MatrixXd Pplus  = Eigen::MatrixXd::Zero(Ls,Ls);
-        Eigen::MatrixXd Pminus = Eigen::MatrixXd::Zero(Ls,Ls);
-
-        for(int s=0;s<Ls;s++){
-            Pplus(s,s)  = this->bee[s];
-            Pminus(s,s) = this->bee[s];
-        }
-
-        for(int s=0; s<Ls-1; s++){
-            Pminus(s,s+1) = -this->cee[s];
-        }
-
-        for(int s=0; s<Ls-1; s++){
-            Pplus(s+1,s) = -this->cee[s+1];
-        }
-
-        Pplus (0,Ls-1) = this->dp;
-        Pminus(Ls-1,0) = this->dm;
-
-        Eigen::MatrixXd PplusMat ;
-        Eigen::MatrixXd PminusMat;
-
-        if(inv) {
-            PplusMat  = Pplus.inverse();
-            PminusMat = Pminus.inverse();
-        } else {
-            PplusMat  = Pplus;
-            PminusMat = Pminus;
-        }
-
-        if(dag){
-            PplusMat.adjointInPlace();
-            PminusMat.adjointInPlace();
-        }
-
-        // For the non-vectorised s-direction this is simple
-
-        for(auto site=0; site<vol; site++){
-
-            SiteSpinor     SiteChi;
-            SiteHalfSpinor SitePplus;
-            SiteHalfSpinor SitePminus;
-
-            for(int s1=0; s1<Ls; s1++){
-                SiteChi = zero;
-                for(int s2=0; s2<Ls; s2++){
-                    int lex2 = s2 + Ls*site;
-                    if(PplusMat(s1,s2) != 0.0){
-                        spProj5p(SitePplus,psi[lex2]);
-                        accumRecon5p(SiteChi, PplusMat(s1,s2)*SitePplus);
-                    }
-                    if(PminusMat(s1,s2) != 0.0){
-                        spProj5m(SitePminus, psi[lex2]);
-                        accumRecon5m(SiteChi, PminusMat(s1,s2)*SitePminus);
-                    }
-                }
-                chi[s1+Ls*site] = SiteChi*0.5;
-            }
-        }
-    }
-
-    #ifdef DOMAIN_WALL_EOFA_DPERP_DENSE
-
-        INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplF);
-        INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplD);
-        INSTANTIATE_DPERP_DWF_EOFA(WilsonImplF);
-        INSTANTIATE_DPERP_DWF_EOFA(WilsonImplD);
-        INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplF);
-        INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplD);
-
-        template void DomainWallEOFAFermion<GparityWilsonImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<GparityWilsonImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<WilsonImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<WilsonImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<ZWilsonImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<ZWilsonImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-
-        INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplFH);
-        INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplDF);
-        INSTANTIATE_DPERP_DWF_EOFA(WilsonImplFH);
-        INSTANTIATE_DPERP_DWF_EOFA(WilsonImplDF);
-        INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplFH);
-        INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplDF);
-
-        template void DomainWallEOFAFermion<GparityWilsonImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<GparityWilsonImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<WilsonImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<WilsonImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<ZWilsonImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<ZWilsonImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-
-    #endif
-
-}}
diff --git a/Grid/qcd/action/fermion/DomainWallEOFAFermionssp.cc b/Grid/qcd/action/fermion/DomainWallEOFAFermionssp.cc
deleted file mode 100644
index 80a4bf09..00000000
--- a/Grid/qcd/action/fermion/DomainWallEOFAFermionssp.cc
+++ /dev/null
@@ -1,168 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermionssp.cc
-
-Copyright (C) 2017
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-Author: David Murphy <dmurphy@phys.columbia.edu>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
-
-namespace Grid {
-namespace QCD {
-
-    // FIXME -- make a version of these routines with site loop outermost for cache reuse.
-    // Pminus fowards
-    // Pplus  backwards
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::M5D(const FermionField& psi, const FermionField& phi,
-        FermionField& chi, std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper)
-    {
-        Coeff_t one(1.0);
-        int Ls = this->Ls;
-        for(int s=0; s<Ls; s++){
-            if(s==0) {
-              axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, s+1);
-              axpby_ssp_pplus (chi, one, chi, lower[s], psi, s, Ls-1);
-            } else if (s==(Ls-1)) {
-              axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, 0);
-              axpby_ssp_pplus (chi, one, chi, lower[s], psi, s, s-1);
-            } else {
-              axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, s+1);
-              axpby_ssp_pplus(chi, one, chi, lower[s], psi, s, s-1);
-            }
-        }
-    }
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::M5Ddag(const FermionField& psi, const FermionField& phi,
-        FermionField& chi, std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper)
-    {
-        Coeff_t one(1.0);
-        int Ls = this->Ls;
-        for(int s=0; s<Ls; s++){
-            if(s==0) {
-              axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, s+1);
-              axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, Ls-1);
-            } else if (s==(Ls-1)) {
-              axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, 0);
-              axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, s-1);
-            } else {
-              axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, s+1);
-              axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, s-1);
-            }
-        }
-    }
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::MooeeInv(const FermionField& psi, FermionField& chi)
-    {
-        Coeff_t one(1.0);
-        Coeff_t czero(0.0);
-        chi.checkerboard = psi.checkerboard;
-        int Ls = this->Ls;
-
-        FermionField tmp(psi._grid);
-
-        // Apply (L^{\prime})^{-1}
-        axpby_ssp(chi, one, psi, czero, psi, 0, 0);      // chi[0]=psi[0]
-        for(int s=1; s<Ls; s++){
-            axpby_ssp_pplus(chi, one, psi, -this->lee[s-1], chi, s, s-1);// recursion Psi[s] -lee P_+ chi[s-1]
-        }
-
-        // L_m^{-1}
-        for(int s=0; s<Ls-1; s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
-            axpby_ssp_pminus(chi, one, chi, -this->leem[s], chi, Ls-1, s);
-        }
-
-        // U_m^{-1} D^{-1}
-        for(int s=0; s<Ls-1; s++){
-            axpby_ssp_pplus(chi, one/this->dee[s], chi, -this->ueem[s]/this->dee[Ls], chi, s, Ls-1);
-        }
-        axpby_ssp_pminus(tmp, czero, chi, one/this->dee[Ls-1], chi, Ls-1, Ls-1);
-        axpby_ssp_pplus(chi, one, tmp, one/this->dee[Ls], chi, Ls-1, Ls-1);
-
-        // Apply U^{-1}
-        for(int s=Ls-2; s>=0; s--){
-            axpby_ssp_pminus(chi, one, chi, -this->uee[s], chi, s, s+1);  // chi[Ls]
-        }
-    }
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::MooeeInvDag(const FermionField& psi, FermionField& chi)
-    {
-        Coeff_t one(1.0);
-        Coeff_t czero(0.0);
-        chi.checkerboard = psi.checkerboard;
-        int Ls = this->Ls;
-
-        FermionField tmp(psi._grid);
-
-        // Apply (U^{\prime})^{-dagger}
-        axpby_ssp(chi, one, psi, czero, psi, 0, 0);      // chi[0]=psi[0]
-        for(int s=1; s<Ls; s++){
-            axpby_ssp_pminus(chi, one, psi, -conjugate(this->uee[s-1]), chi, s, s-1);
-        }
-
-        // U_m^{-\dagger}
-        for(int s=0; s<Ls-1; s++){
-            axpby_ssp_pplus(chi, one, chi, -conjugate(this->ueem[s]), chi, Ls-1, s);
-        }
-
-        // L_m^{-\dagger} D^{-dagger}
-        for(int s=0; s<Ls-1; s++){
-            axpby_ssp_pminus(chi, one/conjugate(this->dee[s]), chi, -conjugate(this->leem[s]/this->dee[Ls-1]), chi, s, Ls-1);
-        }
-        axpby_ssp_pminus(tmp, czero, chi, one/conjugate(this->dee[Ls-1]), chi, Ls-1, Ls-1);
-        axpby_ssp_pplus(chi, one, tmp, one/conjugate(this->dee[Ls]), chi, Ls-1, Ls-1);
-
-        // Apply L^{-dagger}
-        for(int s=Ls-2; s>=0; s--){
-            axpby_ssp_pplus(chi, one, chi, -conjugate(this->lee[s]), chi, s, s+1);  // chi[Ls]
-        }
-    }
-
-    #ifdef DOMAIN_WALL_EOFA_DPERP_LINALG
-
-        INSTANTIATE_DPERP_DWF_EOFA(WilsonImplF);
-        INSTANTIATE_DPERP_DWF_EOFA(WilsonImplD);
-        INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplF);
-        INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplD);
-        INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplF);
-        INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplD);
-
-        INSTANTIATE_DPERP_DWF_EOFA(WilsonImplFH);
-        INSTANTIATE_DPERP_DWF_EOFA(WilsonImplDF);
-        INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplFH);
-        INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplDF);
-        INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplFH);
-        INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplDF);
-
-    #endif
-
-}}
diff --git a/Grid/qcd/action/fermion/DomainWallEOFAFermionvec.cc b/Grid/qcd/action/fermion/DomainWallEOFAFermionvec.cc
deleted file mode 100644
index c95172a5..00000000
--- a/Grid/qcd/action/fermion/DomainWallEOFAFermionvec.cc
+++ /dev/null
@@ -1,605 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermionvec.cc
-
-Copyright (C) 2017
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-Author: David Murphy <dmurphy@phys.columbia.edu>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
-
-namespace Grid {
-namespace QCD {
-
-    /*
-    * Dense matrix versions of routines
-    */
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::MooeeInvDag(const FermionField& psi, FermionField& chi)
-    {
-        this->MooeeInternal(psi, chi, DaggerYes, InverseYes);
-    }
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::MooeeInv(const FermionField& psi, FermionField& chi)
-    {
-        this->MooeeInternal(psi, chi, DaggerNo, InverseYes);
-    }
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::M5D(const FermionField& psi, const FermionField& phi,
-        FermionField& chi, std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper)
-    {
-        GridBase* grid = psi._grid;
-        int Ls  = this->Ls;
-        int LLs = grid->_rdimensions[0];
-        const int nsimd = Simd::Nsimd();
-
-        Vector<iSinglet<Simd> > u(LLs);
-        Vector<iSinglet<Simd> > l(LLs);
-        Vector<iSinglet<Simd> > d(LLs);
-
-        assert(Ls/LLs == nsimd);
-        assert(phi.checkerboard == psi.checkerboard);
-
-        chi.checkerboard = psi.checkerboard;
-
-        // just directly address via type pun
-        typedef typename Simd::scalar_type scalar_type;
-        scalar_type* u_p = (scalar_type*) &u[0];
-        scalar_type* l_p = (scalar_type*) &l[0];
-        scalar_type* d_p = (scalar_type*) &d[0];
-
-        for(int o=0;o<LLs;o++){ // outer
-        for(int i=0;i<nsimd;i++){ //inner
-            int s  = o + i*LLs;
-            int ss = o*nsimd + i;
-            u_p[ss] = upper[s];
-            l_p[ss] = lower[s];
-            d_p[ss] = diag[s];
-        }}
-
-        this->M5Dcalls++;
-        this->M5Dtime -= usecond();
-
-        assert(Nc == 3);
-
-        parallel_for(int ss=0; ss<grid->oSites(); ss+=LLs){ // adds LLs
-
-            #if 0
-
-                alignas(64) SiteHalfSpinor hp;
-                alignas(64) SiteHalfSpinor hm;
-                alignas(64) SiteSpinor fp;
-                alignas(64) SiteSpinor fm;
-
-                for(int v=0; v<LLs; v++){
-
-                    int vp = (v+1)%LLs;
-                    int vm = (v+LLs-1)%LLs;
-
-                    spProj5m(hp, psi[ss+vp]);
-                    spProj5p(hm, psi[ss+vm]);
-
-                    if (vp <= v){ rotate(hp, hp, 1); }
-                    if (vm >= v){ rotate(hm, hm, nsimd-1); }
-
-                    hp = 0.5*hp;
-                    hm = 0.5*hm;
-
-                    spRecon5m(fp, hp);
-                    spRecon5p(fm, hm);
-
-                    chi[ss+v] = d[v]*phi[ss+v];
-                    chi[ss+v] = chi[ss+v] + u[v]*fp;
-                    chi[ss+v] = chi[ss+v] + l[v]*fm;
-
-                }
-
-            #else
-
-                for(int v=0; v<LLs; v++){
-
-                    vprefetch(psi[ss+v+LLs]);
-
-                    int vp = (v==LLs-1) ? 0     : v+1;
-                    int vm = (v==0)     ? LLs-1 : v-1;
-
-                    Simd hp_00 = psi[ss+vp]()(2)(0);
-                    Simd hp_01 = psi[ss+vp]()(2)(1);
-                    Simd hp_02 = psi[ss+vp]()(2)(2);
-                    Simd hp_10 = psi[ss+vp]()(3)(0);
-                    Simd hp_11 = psi[ss+vp]()(3)(1);
-                    Simd hp_12 = psi[ss+vp]()(3)(2);
-
-                    Simd hm_00 = psi[ss+vm]()(0)(0);
-                    Simd hm_01 = psi[ss+vm]()(0)(1);
-                    Simd hm_02 = psi[ss+vm]()(0)(2);
-                    Simd hm_10 = psi[ss+vm]()(1)(0);
-                    Simd hm_11 = psi[ss+vm]()(1)(1);
-                    Simd hm_12 = psi[ss+vm]()(1)(2);
-
-                    if(vp <= v){
-                        hp_00.v = Optimization::Rotate::tRotate<2>(hp_00.v);
-                        hp_01.v = Optimization::Rotate::tRotate<2>(hp_01.v);
-                        hp_02.v = Optimization::Rotate::tRotate<2>(hp_02.v);
-                        hp_10.v = Optimization::Rotate::tRotate<2>(hp_10.v);
-                        hp_11.v = Optimization::Rotate::tRotate<2>(hp_11.v);
-                        hp_12.v = Optimization::Rotate::tRotate<2>(hp_12.v);
-                    }
-
-                    if(vm >= v){
-                        hm_00.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_00.v);
-                        hm_01.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_01.v);
-                        hm_02.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_02.v);
-                        hm_10.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_10.v);
-                        hm_11.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_11.v);
-                        hm_12.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_12.v);
-                    }
-
-                    // Can force these to real arithmetic and save 2x.
-                    Simd p_00 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_00);
-                    Simd p_01 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_01);
-                    Simd p_02 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_02);
-                    Simd p_10 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_10);
-                    Simd p_11 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_11);
-                    Simd p_12 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_12);
-                    Simd p_20 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_00);
-                    Simd p_21 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_01);
-                    Simd p_22 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_02);
-                    Simd p_30 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_10);
-                    Simd p_31 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_11);
-                    Simd p_32 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_12);
-
-                    vstream(chi[ss+v]()(0)(0), p_00);
-                    vstream(chi[ss+v]()(0)(1), p_01);
-                    vstream(chi[ss+v]()(0)(2), p_02);
-                    vstream(chi[ss+v]()(1)(0), p_10);
-                    vstream(chi[ss+v]()(1)(1), p_11);
-                    vstream(chi[ss+v]()(1)(2), p_12);
-                    vstream(chi[ss+v]()(2)(0), p_20);
-                    vstream(chi[ss+v]()(2)(1), p_21);
-                    vstream(chi[ss+v]()(2)(2), p_22);
-                    vstream(chi[ss+v]()(3)(0), p_30);
-                    vstream(chi[ss+v]()(3)(1), p_31);
-                    vstream(chi[ss+v]()(3)(2), p_32);
-                }
-
-            #endif
-        }
-
-        this->M5Dtime += usecond();
-    }
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::M5Ddag(const FermionField& psi, const FermionField& phi,
-        FermionField& chi, std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper)
-    {
-        GridBase* grid = psi._grid;
-        int Ls  = this->Ls;
-        int LLs = grid->_rdimensions[0];
-        int nsimd = Simd::Nsimd();
-
-        Vector<iSinglet<Simd> > u(LLs);
-        Vector<iSinglet<Simd> > l(LLs);
-        Vector<iSinglet<Simd> > d(LLs);
-
-        assert(Ls/LLs == nsimd);
-        assert(phi.checkerboard == psi.checkerboard);
-
-        chi.checkerboard = psi.checkerboard;
-
-        // just directly address via type pun
-        typedef typename Simd::scalar_type scalar_type;
-        scalar_type* u_p = (scalar_type*) &u[0];
-        scalar_type* l_p = (scalar_type*) &l[0];
-        scalar_type* d_p = (scalar_type*) &d[0];
-
-        for(int o=0; o<LLs; o++){ // outer
-        for(int i=0; i<nsimd; i++){ //inner
-            int s  = o + i*LLs;
-            int ss = o*nsimd + i;
-            u_p[ss] = upper[s];
-            l_p[ss] = lower[s];
-            d_p[ss] = diag[s];
-        }}
-
-        this->M5Dcalls++;
-        this->M5Dtime -= usecond();
-
-        parallel_for(int ss=0; ss<grid->oSites(); ss+=LLs){ // adds LLs
-
-        #if 0
-
-            alignas(64) SiteHalfSpinor hp;
-            alignas(64) SiteHalfSpinor hm;
-            alignas(64) SiteSpinor fp;
-            alignas(64) SiteSpinor fm;
-
-            for(int v=0; v<LLs; v++){
-
-                int vp = (v+1)%LLs;
-                int vm = (v+LLs-1)%LLs;
-
-                spProj5p(hp, psi[ss+vp]);
-                spProj5m(hm, psi[ss+vm]);
-
-                if(vp <= v){ rotate(hp, hp, 1); }
-                if(vm >= v){ rotate(hm, hm, nsimd-1); }
-
-                hp = hp*0.5;
-                hm = hm*0.5;
-                spRecon5p(fp, hp);
-                spRecon5m(fm, hm);
-
-                chi[ss+v] = d[v]*phi[ss+v]+u[v]*fp;
-                chi[ss+v] = chi[ss+v]     +l[v]*fm;
-            }
-
-        #else
-
-            for(int v=0; v<LLs; v++){
-
-                vprefetch(psi[ss+v+LLs]);
-
-                int vp = (v == LLs-1) ? 0     : v+1;
-                int vm = (v == 0    ) ? LLs-1 : v-1;
-
-                Simd hp_00 = psi[ss+vp]()(0)(0);
-                Simd hp_01 = psi[ss+vp]()(0)(1);
-                Simd hp_02 = psi[ss+vp]()(0)(2);
-                Simd hp_10 = psi[ss+vp]()(1)(0);
-                Simd hp_11 = psi[ss+vp]()(1)(1);
-                Simd hp_12 = psi[ss+vp]()(1)(2);
-
-                Simd hm_00 = psi[ss+vm]()(2)(0);
-                Simd hm_01 = psi[ss+vm]()(2)(1);
-                Simd hm_02 = psi[ss+vm]()(2)(2);
-                Simd hm_10 = psi[ss+vm]()(3)(0);
-                Simd hm_11 = psi[ss+vm]()(3)(1);
-                Simd hm_12 = psi[ss+vm]()(3)(2);
-
-                if (vp <= v){
-                    hp_00.v = Optimization::Rotate::tRotate<2>(hp_00.v);
-                    hp_01.v = Optimization::Rotate::tRotate<2>(hp_01.v);
-                    hp_02.v = Optimization::Rotate::tRotate<2>(hp_02.v);
-                    hp_10.v = Optimization::Rotate::tRotate<2>(hp_10.v);
-                    hp_11.v = Optimization::Rotate::tRotate<2>(hp_11.v);
-                    hp_12.v = Optimization::Rotate::tRotate<2>(hp_12.v);
-                }
-
-                if(vm >= v){
-                    hm_00.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_00.v);
-                    hm_01.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_01.v);
-                    hm_02.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_02.v);
-                    hm_10.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_10.v);
-                    hm_11.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_11.v);
-                    hm_12.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_12.v);
-                }
-
-                Simd p_00 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_00);
-                Simd p_01 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_01);
-                Simd p_02 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_02);
-                Simd p_10 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_10);
-                Simd p_11 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_11);
-                Simd p_12 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_12);
-                Simd p_20 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_00);
-                Simd p_21 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_01);
-                Simd p_22 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_02);
-                Simd p_30 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_10);
-                Simd p_31 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_11);
-                Simd p_32 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_12);
-
-                vstream(chi[ss+v]()(0)(0), p_00);
-                vstream(chi[ss+v]()(0)(1), p_01);
-                vstream(chi[ss+v]()(0)(2), p_02);
-                vstream(chi[ss+v]()(1)(0), p_10);
-                vstream(chi[ss+v]()(1)(1), p_11);
-                vstream(chi[ss+v]()(1)(2), p_12);
-                vstream(chi[ss+v]()(2)(0), p_20);
-                vstream(chi[ss+v]()(2)(1), p_21);
-                vstream(chi[ss+v]()(2)(2), p_22);
-                vstream(chi[ss+v]()(3)(0), p_30);
-                vstream(chi[ss+v]()(3)(1), p_31);
-                vstream(chi[ss+v]()(3)(2), p_32);
-            }
-        #endif
-
-        }
-
-        this->M5Dtime += usecond();
-    }
-
-    #ifdef AVX512
-        #include<simd/Intel512common.h>
-        #include<simd/Intel512avx.h>
-        #include<simd/Intel512single.h>
-    #endif
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::MooeeInternalAsm(const FermionField& psi, FermionField& chi,
-        int LLs, int site, Vector<iSinglet<Simd> >& Matp, Vector<iSinglet<Simd> >& Matm)
-    {
-        #ifndef AVX512
-        {
-            SiteHalfSpinor BcastP;
-            SiteHalfSpinor BcastM;
-            SiteHalfSpinor SiteChiP;
-            SiteHalfSpinor SiteChiM;
-
-            // Ls*Ls * 2 * 12 * vol flops
-            for(int s1=0; s1<LLs; s1++){
-
-                for(int s2=0; s2<LLs; s2++){
-                for(int l=0; l < Simd::Nsimd(); l++){ // simd lane
-
-                    int s = s2 + l*LLs;
-                    int lex = s2 + LLs*site;
-
-                    if( s2==0 && l==0 ){
-                        SiteChiP=zero;
-                        SiteChiM=zero;
-                    }
-
-                    for(int sp=0; sp<2;  sp++){
-                    for(int co=0; co<Nc; co++){
-                        vbroadcast(BcastP()(sp)(co), psi[lex]()(sp)(co), l);
-                    }}
-
-                    for(int sp=0; sp<2;  sp++){
-                    for(int co=0; co<Nc; co++){
-                        vbroadcast(BcastM()(sp)(co), psi[lex]()(sp+2)(co), l);
-                    }}
-
-                    for(int sp=0; sp<2;  sp++){
-                    for(int co=0; co<Nc; co++){
-                        SiteChiP()(sp)(co) = real_madd(Matp[LLs*s+s1]()()(), BcastP()(sp)(co), SiteChiP()(sp)(co)); // 1100 us.
-                        SiteChiM()(sp)(co) = real_madd(Matm[LLs*s+s1]()()(), BcastM()(sp)(co), SiteChiM()(sp)(co)); // each found by commenting out
-                    }}
-                }}
-
-                {
-                    int lex = s1 + LLs*site;
-                    for(int sp=0; sp<2;  sp++){
-                    for(int co=0; co<Nc; co++){
-                        vstream(chi[lex]()(sp)(co),   SiteChiP()(sp)(co));
-                        vstream(chi[lex]()(sp+2)(co), SiteChiM()(sp)(co));
-                    }}
-                }
-            }
-
-        }
-        #else
-        {
-            // pointers
-            //  MASK_REGS;
-            #define Chi_00 %%zmm1
-            #define Chi_01 %%zmm2
-            #define Chi_02 %%zmm3
-            #define Chi_10 %%zmm4
-            #define Chi_11 %%zmm5
-            #define Chi_12 %%zmm6
-            #define Chi_20 %%zmm7
-            #define Chi_21 %%zmm8
-            #define Chi_22 %%zmm9
-            #define Chi_30 %%zmm10
-            #define Chi_31 %%zmm11
-            #define Chi_32 %%zmm12
-
-            #define BCAST0  %%zmm13
-            #define BCAST1  %%zmm14
-            #define BCAST2  %%zmm15
-            #define BCAST3  %%zmm16
-            #define BCAST4  %%zmm17
-            #define BCAST5  %%zmm18
-            #define BCAST6  %%zmm19
-            #define BCAST7  %%zmm20
-            #define BCAST8  %%zmm21
-            #define BCAST9  %%zmm22
-            #define BCAST10 %%zmm23
-            #define BCAST11 %%zmm24
-
-            int incr = LLs*LLs*sizeof(iSinglet<Simd>);
-            for(int s1=0; s1<LLs; s1++){
-
-                for(int s2=0; s2<LLs; s2++){
-
-                    int lex = s2 + LLs*site;
-                    uint64_t a0 = (uint64_t) &Matp[LLs*s2+s1]; // should be cacheable
-                    uint64_t a1 = (uint64_t) &Matm[LLs*s2+s1];
-                    uint64_t a2 = (uint64_t) &psi[lex];
-
-                    for(int l=0; l<Simd::Nsimd(); l++){ // simd lane
-                        if((s2+l)==0) {
-                            asm(
-                                    VPREFETCH1(0,%2)              VPREFETCH1(0,%1)
-                                    VPREFETCH1(12,%2)  	          VPREFETCH1(13,%2)
-                                    VPREFETCH1(14,%2)  	          VPREFETCH1(15,%2)
-                                    VBCASTCDUP(0,%2,BCAST0)
-                                    VBCASTCDUP(1,%2,BCAST1)
-                                    VBCASTCDUP(2,%2,BCAST2)
-                                    VBCASTCDUP(3,%2,BCAST3)
-                                    VBCASTCDUP(4,%2,BCAST4)       VMULMEM(0,%0,BCAST0,Chi_00)
-                                    VBCASTCDUP(5,%2,BCAST5)       VMULMEM(0,%0,BCAST1,Chi_01)
-                                    VBCASTCDUP(6,%2,BCAST6)       VMULMEM(0,%0,BCAST2,Chi_02)
-                                    VBCASTCDUP(7,%2,BCAST7)       VMULMEM(0,%0,BCAST3,Chi_10)
-                                    VBCASTCDUP(8,%2,BCAST8)       VMULMEM(0,%0,BCAST4,Chi_11)
-                                    VBCASTCDUP(9,%2,BCAST9)       VMULMEM(0,%0,BCAST5,Chi_12)
-                                    VBCASTCDUP(10,%2,BCAST10)     VMULMEM(0,%1,BCAST6,Chi_20)
-                                    VBCASTCDUP(11,%2,BCAST11)     VMULMEM(0,%1,BCAST7,Chi_21)
-                                    VMULMEM(0,%1,BCAST8,Chi_22)
-                                    VMULMEM(0,%1,BCAST9,Chi_30)
-                                    VMULMEM(0,%1,BCAST10,Chi_31)
-                                    VMULMEM(0,%1,BCAST11,Chi_32)
-                                    : : "r" (a0), "r" (a1), "r" (a2)                            );
-                        } else {
-                            asm(
-                                    VBCASTCDUP(0,%2,BCAST0)   VMADDMEM(0,%0,BCAST0,Chi_00)
-                                    VBCASTCDUP(1,%2,BCAST1)   VMADDMEM(0,%0,BCAST1,Chi_01)
-                                    VBCASTCDUP(2,%2,BCAST2)   VMADDMEM(0,%0,BCAST2,Chi_02)
-                                    VBCASTCDUP(3,%2,BCAST3)   VMADDMEM(0,%0,BCAST3,Chi_10)
-                                    VBCASTCDUP(4,%2,BCAST4)   VMADDMEM(0,%0,BCAST4,Chi_11)
-                                    VBCASTCDUP(5,%2,BCAST5)   VMADDMEM(0,%0,BCAST5,Chi_12)
-                                    VBCASTCDUP(6,%2,BCAST6)   VMADDMEM(0,%1,BCAST6,Chi_20)
-                                    VBCASTCDUP(7,%2,BCAST7)   VMADDMEM(0,%1,BCAST7,Chi_21)
-                                    VBCASTCDUP(8,%2,BCAST8)   VMADDMEM(0,%1,BCAST8,Chi_22)
-                                    VBCASTCDUP(9,%2,BCAST9)   VMADDMEM(0,%1,BCAST9,Chi_30)
-                                    VBCASTCDUP(10,%2,BCAST10) VMADDMEM(0,%1,BCAST10,Chi_31)
-                                    VBCASTCDUP(11,%2,BCAST11) VMADDMEM(0,%1,BCAST11,Chi_32)
-                                    : : "r" (a0), "r" (a1), "r" (a2)                            );
-                        }
-                        a0 = a0 + incr;
-                        a1 = a1 + incr;
-                        a2 = a2 + sizeof(typename Simd::scalar_type);
-                    }
-                }
-
-                {
-                  int lexa = s1+LLs*site;
-                  asm (
-                     VSTORE(0,%0,Chi_00) VSTORE(1 ,%0,Chi_01)  VSTORE(2 ,%0,Chi_02)
-                     VSTORE(3,%0,Chi_10) VSTORE(4 ,%0,Chi_11)  VSTORE(5 ,%0,Chi_12)
-                     VSTORE(6,%0,Chi_20) VSTORE(7 ,%0,Chi_21)  VSTORE(8 ,%0,Chi_22)
-                     VSTORE(9,%0,Chi_30) VSTORE(10,%0,Chi_31)  VSTORE(11,%0,Chi_32)
-                     : : "r" ((uint64_t)&chi[lexa]) : "memory" );
-
-                }
-            }
-        }
-
-        #undef Chi_00
-        #undef Chi_01
-        #undef Chi_02
-        #undef Chi_10
-        #undef Chi_11
-        #undef Chi_12
-        #undef Chi_20
-        #undef Chi_21
-        #undef Chi_22
-        #undef Chi_30
-        #undef Chi_31
-        #undef Chi_32
-
-        #undef BCAST0
-        #undef BCAST1
-        #undef BCAST2
-        #undef BCAST3
-        #undef BCAST4
-        #undef BCAST5
-        #undef BCAST6
-        #undef BCAST7
-        #undef BCAST8
-        #undef BCAST9
-        #undef BCAST10
-        #undef BCAST11
-        #endif
-    };
-
-    // Z-mobius version
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::MooeeInternalZAsm(const FermionField& psi, FermionField& chi,
-        int LLs, int site, Vector<iSinglet<Simd> >& Matp, Vector<iSinglet<Simd> >& Matm)
-    {
-        std::cout << "Error: zMobius not implemented for EOFA" << std::endl;
-        exit(-1);
-    };
-
-    template<class Impl>
-    void DomainWallEOFAFermion<Impl>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv)
-    {
-        int Ls  = this->Ls;
-        int LLs = psi._grid->_rdimensions[0];
-        int vol = psi._grid->oSites()/LLs;
-
-        chi.checkerboard = psi.checkerboard;
-
-        Vector<iSinglet<Simd> > Matp;
-        Vector<iSinglet<Simd> > Matm;
-        Vector<iSinglet<Simd> > *_Matp;
-        Vector<iSinglet<Simd> > *_Matm;
-
-        //  MooeeInternalCompute(dag,inv,Matp,Matm);
-        if(inv && dag){
-            _Matp = &this->MatpInvDag;
-            _Matm = &this->MatmInvDag;
-        }
-
-        if(inv && (!dag)){
-            _Matp = &this->MatpInv;
-            _Matm = &this->MatmInv;
-        }
-
-        if(!inv){
-            MooeeInternalCompute(dag, inv, Matp, Matm);
-            _Matp = &Matp;
-            _Matm = &Matm;
-        }
-
-        assert(_Matp->size() == Ls*LLs);
-
-        this->MooeeInvCalls++;
-        this->MooeeInvTime -= usecond();
-
-        if(switcheroo<Coeff_t>::iscomplex()){
-            parallel_for(auto site=0; site<vol; site++){
-                MooeeInternalZAsm(psi, chi, LLs, site, *_Matp, *_Matm);
-            }
-        } else {
-            parallel_for(auto site=0; site<vol; site++){
-                MooeeInternalAsm(psi, chi, LLs, site, *_Matp, *_Matm);
-            }
-        }
-
-        this->MooeeInvTime += usecond();
-    }
-
-    #ifdef DOMAIN_WALL_EOFA_DPERP_VEC
-
-        INSTANTIATE_DPERP_DWF_EOFA(DomainWallVec5dImplD);
-        INSTANTIATE_DPERP_DWF_EOFA(DomainWallVec5dImplF);
-        INSTANTIATE_DPERP_DWF_EOFA(ZDomainWallVec5dImplD);
-        INSTANTIATE_DPERP_DWF_EOFA(ZDomainWallVec5dImplF);
-
-        INSTANTIATE_DPERP_DWF_EOFA(DomainWallVec5dImplDF);
-        INSTANTIATE_DPERP_DWF_EOFA(DomainWallVec5dImplFH);
-        INSTANTIATE_DPERP_DWF_EOFA(ZDomainWallVec5dImplDF);
-        INSTANTIATE_DPERP_DWF_EOFA(ZDomainWallVec5dImplFH);
-
-        template void DomainWallEOFAFermion<DomainWallVec5dImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<DomainWallVec5dImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<ZDomainWallVec5dImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<ZDomainWallVec5dImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-
-        template void DomainWallEOFAFermion<DomainWallVec5dImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<DomainWallVec5dImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<ZDomainWallVec5dImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-        template void DomainWallEOFAFermion<ZDomainWallVec5dImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-
-    #endif
-
-}}
diff --git a/Grid/qcd/action/fermion/DomainWallFermion.h b/Grid/qcd/action/fermion/DomainWallFermion.h
index b6824dc4..e32e5917 100644
--- a/Grid/qcd/action/fermion/DomainWallFermion.h
+++ b/Grid/qcd/action/fermion/DomainWallFermion.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,34 +25,33 @@ Author: Vera Guelpers <V.M.Guelpers@soton.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef  GRID_QCD_DOMAIN_WALL_FERMION_H
 #define  GRID_QCD_DOMAIN_WALL_FERMION_H
 
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class DomainWallFermion : public CayleyFermion5D<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+public:
 
-    template<class Impl>
-    class DomainWallFermion : public CayleyFermion5D<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-    public:
+  void FreePropagator(const FermionField &in,FermionField &out,RealD mass,std::vector<Complex> boundary, std::vector<double> twist, bool fiveD) {
+	FermionField in_k(in.Grid());
+	FermionField prop_k(in.Grid());
 
-      void FreePropagator(const FermionField &in,FermionField &out,RealD mass,std::vector<Complex> boundary, std::vector<double> twist, bool fiveD) {
-	FermionField in_k(in._grid);
-	FermionField prop_k(in._grid);
-
-	FFT theFFT((GridCartesian *) in._grid);
+	FFT theFFT((GridCartesian *) in.Grid());
 
 	//phase for boundary condition
-	ComplexField coor(in._grid);
-	ComplexField ph(in._grid);  ph = zero;
-	FermionField in_buf(in._grid); in_buf = zero;
+	ComplexField coor(in.Grid());
+	ComplexField ph(in.Grid());  ph = Zero();
+	FermionField in_buf(in.Grid()); in_buf = Zero();
+	typedef typename Simd::scalar_type Scalar;
 	Scalar ci(0.0,1.0);
 	assert(twist.size() == Nd);//check that twist is Nd
 	assert(boundary.size() == Nd);//check that boundary conditions is Nd
@@ -63,13 +62,12 @@ namespace Grid {
 	  // Shift coordinate lattice index by 1 to account for 5th dimension.
           LatticeCoordinate(coor, nu + shift);
 	  double boundary_phase = ::acos(real(boundary[nu]));
-	  ph = ph + boundary_phase*coor*((1./(in._grid->_fdimensions[nu+shift])));
+	  ph = ph + boundary_phase*coor*((1./(in.Grid()->_fdimensions[nu+shift])));
 	  //momenta for propagator shifted by twist+boundary
 	  twist[nu] = twist[nu] + boundary_phase/((2.0*M_PI));
 	}
 	in_buf = exp(ci*ph*(-1.0))*in;
 
-
 	if(fiveD){//FFT only on temporal and spatial dimensions
           std::vector<int> mask(Nd+1,1); mask[0] = 0;
 	  theFFT.FFT_dim_mask(in_k,in_buf,mask,FFT::forward);
@@ -82,7 +80,7 @@ namespace Grid {
 	  theFFT.FFT_all_dim(out,prop_k,FFT::backward);
         }
 	//phase for boundary condition
-	out = out * exp(ci*ph);
+	out = out * exp(Scalar(2.0*M_PI)*ci*ph);
       };
 
       virtual void FreePropagator(const FermionField &in,FermionField &out,RealD mass,std::vector<Complex> boundary,std::vector<double> twist) {
@@ -105,38 +103,37 @@ namespace Grid {
 	FreePropagator(in,out,mass,boundary,twist,fiveD);
       };
 
-      virtual void   Instantiatable(void) {};
-      // Constructors
-      DomainWallFermion(GaugeField &_Umu,
-			GridCartesian         &FiveDimGrid,
-			GridRedBlackCartesian &FiveDimRedBlackGrid,
-			GridCartesian         &FourDimGrid,
-			GridRedBlackCartesian &FourDimRedBlackGrid,
-			RealD _mass,RealD _M5,const ImplParams &p= ImplParams()) : 
+  virtual void   Instantiatable(void) {};
+  // Constructors
+  DomainWallFermion(GaugeField &_Umu,
+		    GridCartesian         &FiveDimGrid,
+		    GridRedBlackCartesian &FiveDimRedBlackGrid,
+		    GridCartesian         &FourDimGrid,
+		    GridRedBlackCartesian &FourDimRedBlackGrid,
+		    RealD _mass,RealD _M5,const ImplParams &p= ImplParams()) : 
 
 
-      CayleyFermion5D<Impl>(_Umu,
-			    FiveDimGrid,
-			    FiveDimRedBlackGrid,
-			    FourDimGrid,
-			    FourDimRedBlackGrid,_mass,_M5,p)
+    CayleyFermion5D<Impl>(_Umu,
+			  FiveDimGrid,
+			  FiveDimRedBlackGrid,
+			  FourDimGrid,
+			  FourDimRedBlackGrid,_mass,_M5,p)
 
-      {
-	RealD eps = 1.0;
+  {
+    RealD eps = 1.0;
 
-	Approx::zolotarev_data *zdata = Approx::higham(eps,this->Ls);// eps is ignored for higham
-	assert(zdata->n==this->Ls);
+    Approx::zolotarev_data *zdata = Approx::higham(eps,this->Ls);// eps is ignored for higham
+    assert(zdata->n==this->Ls);
 	
-	std::cout<<GridLogMessage << "DomainWallFermion with Ls="<<this->Ls<<std::endl;
-	// Call base setter
-	this->SetCoefficientsTanh(zdata,1.0,0.0);
-
-	Approx::zolotarev_free(zdata);
-      }
-
-    };
+    //    std::cout<<GridLogMessage << "DomainWallFermion with Ls="<<this->Ls<<std::endl;
+    // Call base setter
+    this->SetCoefficientsTanh(zdata,1.0,0.0);
 
+    Approx::zolotarev_free(zdata);
   }
-}
+
+};
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/fermion/DomainWallVec5dImpl.h b/Grid/qcd/action/fermion/DomainWallVec5dImpl.h
new file mode 100644
index 00000000..890c680b
--- /dev/null
+++ b/Grid/qcd/action/fermion/DomainWallVec5dImpl.h
@@ -0,0 +1,213 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/FermionOperatorImpl.h
+
+Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+			   /*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+template<class S,class Representation = FundamentalRepresentation, class Options=CoeffReal>
+class DomainWallVec5dImpl :  public PeriodicGaugeImpl< GaugeImplTypes< S,Representation::Dimension> > { 
+public:
+
+  typedef PeriodicGaugeImpl<GaugeImplTypes<S, Representation::Dimension> > Gimpl;
+  INHERIT_GIMPL_TYPES(Gimpl);
+
+  static const int Dimension = Representation::Dimension;
+  static const bool isFundamental = Representation::isFundamental;
+  static const bool LsVectorised=true;
+  static const int Nhcs = Options::Nhcs;
+      
+  typedef typename Options::_Coeff_t Coeff_t;      
+  typedef typename Options::template PrecisionMapper<Simd>::LowerPrecVector SimdL;
+  
+  template <typename vtype> using iImplSpinor            = iScalar<iVector<iVector<vtype, Dimension>, Ns> >;
+  template <typename vtype> using iImplPropagator        = iScalar<iMatrix<iMatrix<vtype, Dimension>, Ns> >;
+  template <typename vtype> using iImplHalfSpinor        = iScalar<iVector<iVector<vtype, Dimension>, Nhs> >;
+  template <typename vtype> using iImplHalfCommSpinor    = iScalar<iVector<iVector<vtype, Dimension>, Nhcs> >;
+  template <typename vtype> using iImplDoubledGaugeField = iVector<iScalar<iMatrix<vtype, Dimension> >, Nds>;
+  template <typename vtype> using iImplGaugeField        = iVector<iScalar<iMatrix<vtype, Dimension> >, Nd>;
+  template <typename vtype> using iImplGaugeLink         = iScalar<iScalar<iMatrix<vtype, Dimension> > >;
+  
+  typedef iImplSpinor<Simd>            SiteSpinor;
+  typedef iImplPropagator<Simd>        SitePropagator;
+  typedef iImplHalfSpinor<Simd>        SiteHalfSpinor;
+  typedef iImplHalfCommSpinor<SimdL>   SiteHalfCommSpinor;
+  typedef Lattice<SiteSpinor>          FermionField;
+  typedef Lattice<SitePropagator>      PropagatorField;
+
+  /////////////////////////////////////////////////
+  // Make the doubled gauge field a *scalar*
+  /////////////////////////////////////////////////
+  typedef iImplDoubledGaugeField<typename Simd::scalar_type>  SiteDoubledGaugeField;  // This is a scalar
+  typedef iImplGaugeField<typename Simd::scalar_type>         SiteScalarGaugeField;  // scalar
+  typedef iImplGaugeLink<typename Simd::scalar_type>          SiteScalarGaugeLink;  // scalar
+  typedef Lattice<SiteDoubledGaugeField>                      DoubledGaugeField;
+      
+  typedef WilsonCompressor<SiteHalfCommSpinor,SiteHalfSpinor, SiteSpinor> Compressor;
+  typedef WilsonImplParams ImplParams;
+  typedef WilsonStencil<SiteSpinor, SiteHalfSpinor,ImplParams> StencilImpl;
+  typedef typename StencilImpl::View_type StencilView;
+  
+  ImplParams Params;
+
+  DomainWallVec5dImpl(const ImplParams &p = ImplParams()) : Params(p){};
+      
+  template <class ref>
+  static accelerator_inline void loadLinkElement(Simd &reg, ref &memory) 
+  {
+    vsplat(reg, memory);
+  }
+
+  template<class _Spinor>
+  static accelerator_inline void multLink(_Spinor &phi, const SiteDoubledGaugeField &U,
+					  const _Spinor &chi, int mu, StencilEntry *SE,
+					  StencilView &St) 
+  {
+#ifdef GPU_VEC
+    // Gauge link is scalarised
+    mult(&phi(), &U(mu), &chi());
+#else
+    SiteGaugeLink UU;
+    for (int i = 0; i < Dimension; i++) {
+      for (int j = 0; j < Dimension; j++) {
+        vsplat(UU()()(i, j), U(mu)()(i, j));
+      }
+    }
+    mult(&phi(), &UU(), &chi());
+#endif
+  }
+
+  inline void DoubleStore(GridBase *GaugeGrid, DoubledGaugeField &Uds,const GaugeField &Umu) 
+  {
+    SiteScalarGaugeField  ScalarUmu;
+    SiteDoubledGaugeField ScalarUds;
+    
+    GaugeLinkField U(Umu.Grid());
+    GaugeField  Uadj(Umu.Grid());
+    for (int mu = 0; mu < Nd; mu++) {
+      U = PeekIndex<LorentzIndex>(Umu, mu);
+      U = adj(Cshift(U, mu, -1));
+      PokeIndex<LorentzIndex>(Uadj, U, mu);
+    }
+    
+    for (int lidx = 0; lidx < GaugeGrid->lSites(); lidx++) {
+      Coordinate lcoor;
+      GaugeGrid->LocalIndexToLocalCoor(lidx, lcoor);
+      
+      peekLocalSite(ScalarUmu, Umu, lcoor);
+      for (int mu = 0; mu < 4; mu++) ScalarUds(mu) = ScalarUmu(mu);
+      
+      peekLocalSite(ScalarUmu, Uadj, lcoor);
+      for (int mu = 0; mu < 4; mu++) ScalarUds(mu + 4) = ScalarUmu(mu);
+      
+      pokeLocalSite(ScalarUds, Uds, lcoor);
+    }
+  }
+      
+  inline void InsertForce4D(GaugeField &mat, FermionField &Btilde,FermionField &A, int mu) 
+  {
+    assert(0);
+  }
+
+  inline void outerProductImpl(PropagatorField &mat, const FermionField &Btilde, const FermionField &A){
+    assert(0);
+  } 
+
+  inline void TraceSpinImpl(GaugeLinkField &mat, PropagatorField&P) {
+    assert(0);
+  }
+
+  inline void extractLinkField(std::vector<GaugeLinkField> &mat, DoubledGaugeField &Uds){
+    assert(0);
+  }
+
+
+  inline void InsertForce5D(GaugeField &mat, FermionField &Btilde, FermionField &Atilde, int mu) {
+
+    assert(0);
+    // Following lines to be revised after Peter's addition of half prec
+    // missing put lane...
+    /*
+      typedef decltype(traceIndex<SpinIndex>(outerProduct(Btilde[0], Atilde[0]))) result_type;
+      unsigned int LLs = Btilde.Grid()->_rdimensions[0];
+      conformable(Atilde.Grid(),Btilde.Grid());
+      GridBase* grid = mat.Grid();
+      GridBase* Bgrid = Btilde.Grid();
+      unsigned int dimU = grid->Nd();
+      unsigned int dimF = Bgrid->Nd();
+      GaugeLinkField tmp(grid); 
+      tmp = Zero();
+    
+      // FIXME 
+      // Current implementation works, thread safe, probably suboptimal
+      // Passing through the local coordinate for grid transformation
+      // the force grid is in general very different from the Ls vectorized grid
+
+      for (int so = 0; so < grid->oSites(); so++) {
+      std::vector<typename result_type::scalar_object> vres(Bgrid->Nsimd());
+      std::vector<int> ocoor;  grid->oCoorFromOindex(ocoor,so); 
+      for (int si = 0; si < tmp.Grid()->iSites(); si++){
+      typename result_type::scalar_object scalar_object; scalar_object = Zero();
+      std::vector<int> local_coor;      
+      std::vector<int> icoor; grid->iCoorFromIindex(icoor,si);
+      grid->InOutCoorToLocalCoor(ocoor, icoor, local_coor);
+      for (int s = 0; s < LLs; s++) {
+      std::vector<int> slocal_coor(dimF);
+      slocal_coor[0] = s;
+      for (int s4d = 1; s4d< dimF; s4d++) slocal_coor[s4d] = local_coor[s4d-1];
+      int sF = Bgrid->oIndexReduced(slocal_coor);  
+      assert(sF < Bgrid->oSites());
+
+      extract(traceIndex<SpinIndex>(outerProduct(Btilde[sF], Atilde[sF])), vres); 
+      // sum across the 5d dimension
+      for (auto v : vres) scalar_object += v;  
+      }
+      tmp[so].putlane(scalar_object, si);
+      }
+      }
+      PokeIndex<LorentzIndex>(mat, tmp, mu);
+    */
+  }
+};
+typedef DomainWallVec5dImpl<vComplex ,FundamentalRepresentation, CoeffReal> DomainWallVec5dImplR; // Real.. whichever prec
+typedef DomainWallVec5dImpl<vComplexF,FundamentalRepresentation, CoeffReal> DomainWallVec5dImplF; // Float
+typedef DomainWallVec5dImpl<vComplexD,FundamentalRepresentation, CoeffReal> DomainWallVec5dImplD; // Double
+ 
+typedef DomainWallVec5dImpl<vComplex ,FundamentalRepresentation, CoeffRealHalfComms> DomainWallVec5dImplRL; // Real.. whichever prec
+typedef DomainWallVec5dImpl<vComplexF,FundamentalRepresentation, CoeffRealHalfComms> DomainWallVec5dImplFH; // Float
+typedef DomainWallVec5dImpl<vComplexD,FundamentalRepresentation, CoeffRealHalfComms> DomainWallVec5dImplDF; // Double
+ 
+typedef DomainWallVec5dImpl<vComplex ,FundamentalRepresentation,CoeffComplex> ZDomainWallVec5dImplR; // Real.. whichever prec
+typedef DomainWallVec5dImpl<vComplexF,FundamentalRepresentation,CoeffComplex> ZDomainWallVec5dImplF; // Float
+typedef DomainWallVec5dImpl<vComplexD,FundamentalRepresentation,CoeffComplex> ZDomainWallVec5dImplD; // Double
+ 
+typedef DomainWallVec5dImpl<vComplex ,FundamentalRepresentation,CoeffComplexHalfComms> ZDomainWallVec5dImplRL; // Real.. whichever prec
+typedef DomainWallVec5dImpl<vComplexF,FundamentalRepresentation,CoeffComplexHalfComms> ZDomainWallVec5dImplFH; // Float
+typedef DomainWallVec5dImpl<vComplexD,FundamentalRepresentation,CoeffComplexHalfComms> ZDomainWallVec5dImplDF; // Double
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/Fermion.h b/Grid/qcd/action/fermion/Fermion.h
index 77a4681f..fb6f18bb 100644
--- a/Grid/qcd/action/fermion/Fermion.h
+++ b/Grid/qcd/action/fermion/Fermion.h
@@ -23,10 +23,9 @@ Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef  GRID_QCD_FERMION_H
-#define  GRID_QCD_FERMION_H
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 // Explicit explicit template instantiation is still required in the .cc files
@@ -50,12 +49,17 @@ Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 ////////////////////////////////////////////
 
 #include <Grid/qcd/action/fermion/WilsonFermion.h>       // 4d wilson like
-#include <Grid/qcd/action/fermion/WilsonTMFermion.h>     // 4d wilson like
+NAMESPACE_CHECK(Wilson);
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>       // 4d wilson like
+NAMESPACE_CHECK(WilsonTM);
 #include <Grid/qcd/action/fermion/WilsonCloverFermion.h> // 4d wilson clover fermions
+NAMESPACE_CHECK(WilsonClover);
 #include <Grid/qcd/action/fermion/WilsonFermion5D.h>     // 5d base used by all 5d overlap types
+NAMESPACE_CHECK(Wilson5D);
 
 #include <Grid/qcd/action/fermion/ImprovedStaggeredFermion.h>
 #include <Grid/qcd/action/fermion/ImprovedStaggeredFermion5D.h>
+NAMESPACE_CHECK(Staggered);
 
 #include <Grid/qcd/action/fermion/CayleyFermion5D.h>     // Cayley types
 #include <Grid/qcd/action/fermion/DomainWallFermion.h>
@@ -63,7 +67,8 @@ Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 #include <Grid/qcd/action/fermion/MobiusFermion.h>
 #include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
 #include <Grid/qcd/action/fermion/ZMobiusFermion.h>
-#include <Grid/qcd/action/fermion/SchurDiagTwoKappa.h>
+NAMESPACE_CHECK(DomainWall);
+
 #include <Grid/qcd/action/fermion/ScaledShamirFermion.h>
 #include <Grid/qcd/action/fermion/MobiusZolotarevFermion.h>
 #include <Grid/qcd/action/fermion/ShamirZolotarevFermion.h>
@@ -75,6 +80,7 @@ Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 #include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>                 // Partial fraction
 #include <Grid/qcd/action/fermion/OverlapWilsonPartialFractionTanhFermion.h>
 #include <Grid/qcd/action/fermion/OverlapWilsonPartialFractionZolotarevFermion.h>
+NAMESPACE_CHECK(Overlap);
 ///////////////////////////////////////////////////////////////////////////////
 // G5 herm -- this has to live in QCD since dirac matrix is not in the broader sector of code
 ///////////////////////////////////////////////////////////////////////////////
@@ -84,14 +90,17 @@ Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 // Fourier accelerated Pauli Villars inverse support
 ///////////////////////////////////////////////////////////////////////////////
 #include <Grid/qcd/action/fermion/WilsonTMFermion5D.h>   
+NAMESPACE_CHECK(WilsonTM5);
 
 ////////////////////////////////////////////////////////////////////////////////
 // Move this group to a DWF specific tools/algorithms subdir? 
 ////////////////////////////////////////////////////////////////////////////////
+#include <Grid/qcd/action/fermion/SchurDiagTwoKappa.h>
 #include <Grid/qcd/action/fermion/FourierAcceleratedPV.h>
 #include <Grid/qcd/action/fermion/PauliVillarsInverters.h>
 #include <Grid/qcd/action/fermion/Reconstruct5Dprop.h>
 #include <Grid/qcd/action/fermion/MADWF.h>
+NAMESPACE_CHECK(DWFutils);
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 // More maintainable to maintain the following typedef list centrally, as more "impl" targets
@@ -99,8 +108,7 @@ Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 
 // Cayley 5d
-namespace Grid {
-  namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 typedef WilsonFermion<WilsonImplR> WilsonFermionR;
 typedef WilsonFermion<WilsonImplF> WilsonFermionF;
@@ -186,46 +194,6 @@ typedef ZMobiusFermion<ZWilsonImplFH> ZMobiusFermionFH;
 typedef ZMobiusFermion<ZWilsonImplDF> ZMobiusFermionDF;
 
 // Ls vectorised
-typedef DomainWallFermion<DomainWallVec5dImplR> DomainWallFermionVec5dR;
-typedef DomainWallFermion<DomainWallVec5dImplF> DomainWallFermionVec5dF;
-typedef DomainWallFermion<DomainWallVec5dImplD> DomainWallFermionVec5dD;
-
-typedef DomainWallFermion<DomainWallVec5dImplRL> DomainWallFermionVec5dRL;
-typedef DomainWallFermion<DomainWallVec5dImplFH> DomainWallFermionVec5dFH;
-typedef DomainWallFermion<DomainWallVec5dImplDF> DomainWallFermionVec5dDF;
-
-typedef DomainWallEOFAFermion<DomainWallVec5dImplR> DomainWallEOFAFermionVec5dR;
-typedef DomainWallEOFAFermion<DomainWallVec5dImplF> DomainWallEOFAFermionVec5dF;
-typedef DomainWallEOFAFermion<DomainWallVec5dImplD> DomainWallEOFAFermionVec5dD;
-
-typedef DomainWallEOFAFermion<DomainWallVec5dImplRL> DomainWallEOFAFermionVec5dRL;
-typedef DomainWallEOFAFermion<DomainWallVec5dImplFH> DomainWallEOFAFermionVec5dFH;
-typedef DomainWallEOFAFermion<DomainWallVec5dImplDF> DomainWallEOFAFermionVec5dDF;
-
-typedef MobiusFermion<DomainWallVec5dImplR> MobiusFermionVec5dR;
-typedef MobiusFermion<DomainWallVec5dImplF> MobiusFermionVec5dF;
-typedef MobiusFermion<DomainWallVec5dImplD> MobiusFermionVec5dD;
-
-typedef MobiusFermion<DomainWallVec5dImplRL> MobiusFermionVec5dRL;
-typedef MobiusFermion<DomainWallVec5dImplFH> MobiusFermionVec5dFH;
-typedef MobiusFermion<DomainWallVec5dImplDF> MobiusFermionVec5dDF;
-
-typedef MobiusEOFAFermion<DomainWallVec5dImplR> MobiusEOFAFermionVec5dR;
-typedef MobiusEOFAFermion<DomainWallVec5dImplF> MobiusEOFAFermionVec5dF;
-typedef MobiusEOFAFermion<DomainWallVec5dImplD> MobiusEOFAFermionVec5dD;
-
-typedef MobiusEOFAFermion<DomainWallVec5dImplRL> MobiusEOFAFermionVec5dRL;
-typedef MobiusEOFAFermion<DomainWallVec5dImplFH> MobiusEOFAFermionVec5dFH;
-typedef MobiusEOFAFermion<DomainWallVec5dImplDF> MobiusEOFAFermionVec5dDF;
-
-typedef ZMobiusFermion<ZDomainWallVec5dImplR> ZMobiusFermionVec5dR;
-typedef ZMobiusFermion<ZDomainWallVec5dImplF> ZMobiusFermionVec5dF;
-typedef ZMobiusFermion<ZDomainWallVec5dImplD> ZMobiusFermionVec5dD;
-
-typedef ZMobiusFermion<ZDomainWallVec5dImplRL> ZMobiusFermionVec5dRL;
-typedef ZMobiusFermion<ZDomainWallVec5dImplFH> ZMobiusFermionVec5dFH;
-typedef ZMobiusFermion<ZDomainWallVec5dImplDF> ZMobiusFermionVec5dDF;
-
 typedef ScaledShamirFermion<WilsonImplR> ScaledShamirFermionR;
 typedef ScaledShamirFermion<WilsonImplF> ScaledShamirFermionF;
 typedef ScaledShamirFermion<WilsonImplD> ScaledShamirFermionD;
@@ -318,12 +286,13 @@ typedef ImprovedStaggeredFermion5D<StaggeredImplR> ImprovedStaggeredFermion5DR;
 typedef ImprovedStaggeredFermion5D<StaggeredImplF> ImprovedStaggeredFermion5DF;
 typedef ImprovedStaggeredFermion5D<StaggeredImplD> ImprovedStaggeredFermion5DD;
 
+#ifndef GRID_NVCC
 typedef ImprovedStaggeredFermion5D<StaggeredVec5dImplR> ImprovedStaggeredFermionVec5dR;
 typedef ImprovedStaggeredFermion5D<StaggeredVec5dImplF> ImprovedStaggeredFermionVec5dF;
 typedef ImprovedStaggeredFermion5D<StaggeredVec5dImplD> ImprovedStaggeredFermionVec5dD;
+#endif
 
-
-  }}
+NAMESPACE_END(Grid);
 
 ////////////////////
 // Scalar QED actions
@@ -332,4 +301,4 @@ typedef ImprovedStaggeredFermion5D<StaggeredVec5dImplD> ImprovedStaggeredFermion
 #include <Grid/qcd/action/scalar/Scalar.h>
 #include <Grid/qcd/action/gauge/Photon.h>
 
-#endif
+
diff --git a/Grid/qcd/action/fermion/FermionCore.h b/Grid/qcd/action/fermion/FermionCore.h
index 60632c3a..6745032e 100644
--- a/Grid/qcd/action/fermion/FermionCore.h
+++ b/Grid/qcd/action/fermion/FermionCore.h
@@ -36,58 +36,13 @@ Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 // Fermion prereqs
 ////////////////////////////////////////////
 #include <Grid/qcd/action/fermion/WilsonCompressor.h>     //used by all wilson type fermions
+NAMESPACE_CHECK(Compressor);
 #include <Grid/qcd/action/fermion/FermionOperatorImpl.h>
+NAMESPACE_CHECK(FermionOperatorImpl);
 #include <Grid/qcd/action/fermion/FermionOperator.h>
+NAMESPACE_CHECK(FermionOperator);
 #include <Grid/qcd/action/fermion/WilsonKernels.h>        //used by all wilson type fermions
 #include <Grid/qcd/action/fermion/StaggeredKernels.h>        //used by all wilson type fermions
-
-#define FermOpStaggeredTemplateInstantiate(A) \
-  template class A<StaggeredImplF>; \
-  template class A<StaggeredImplD>; 
-
-#define FermOpStaggeredVec5dTemplateInstantiate(A) \
-  template class A<StaggeredVec5dImplF>; \
-  template class A<StaggeredVec5dImplD>; 
-
-#define FermOp4dVecTemplateInstantiate(A) \
-  template class A<WilsonImplF>;		\
-  template class A<WilsonImplD>;		\
-  template class A<ZWilsonImplF>;		\
-  template class A<ZWilsonImplD>;		\
-  template class A<GparityWilsonImplF>;		\
-  template class A<GparityWilsonImplD>;		\
-  template class A<WilsonImplFH>;		\
-  template class A<WilsonImplDF>;		\
-  template class A<ZWilsonImplFH>;		\
-  template class A<ZWilsonImplDF>;		\
-  template class A<GparityWilsonImplFH>;		\
-  template class A<GparityWilsonImplDF>;		
-
-
-#define AdjointFermOpTemplateInstantiate(A) \
-  template class A<WilsonAdjImplF>; \
-  template class A<WilsonAdjImplD>; 
-
-#define TwoIndexFermOpTemplateInstantiate(A) \
-  template class A<WilsonTwoIndexSymmetricImplF>; \
-  template class A<WilsonTwoIndexSymmetricImplD>; \
-  template class A<WilsonTwoIndexAntiSymmetricImplF>; \
-  template class A<WilsonTwoIndexAntiSymmetricImplD>;
-
-#define FermOp5dVecTemplateInstantiate(A) \
-  template class A<DomainWallVec5dImplF>;	\
-  template class A<DomainWallVec5dImplD>;	\
-  template class A<ZDomainWallVec5dImplF>;	\
-  template class A<ZDomainWallVec5dImplD>;	\
-  template class A<DomainWallVec5dImplFH>;	\
-  template class A<DomainWallVec5dImplDF>;	\
-  template class A<ZDomainWallVec5dImplFH>;	\
-  template class A<ZDomainWallVec5dImplDF>;	
-
-#define FermOpTemplateInstantiate(A) \
- FermOp4dVecTemplateInstantiate(A) \
- FermOp5dVecTemplateInstantiate(A) 
-
-#define GparityFermOpTemplateInstantiate(A) 
+NAMESPACE_CHECK(Kernels);
 
 #endif
diff --git a/Grid/qcd/action/fermion/FermionOperator.h b/Grid/qcd/action/fermion/FermionOperator.h
index 221f2bfd..c60a2e84 100644
--- a/Grid/qcd/action/fermion/FermionOperator.h
+++ b/Grid/qcd/action/fermion/FermionOperator.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -26,86 +26,87 @@ Author: Vera Guelpers <V.M.Guelpers@soton.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef  GRID_QCD_FERMION_OPERATOR_H
-#define  GRID_QCD_FERMION_OPERATOR_H
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
-
-    ////////////////////////////////////////////////////////////////
-    // Allow to select  between gauge representation rank bc's, flavours etc.
-    // and single/double precision.
-    ////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////
+// Allow to select  between gauge representation rank bc's, flavours etc.
+// and single/double precision.
+////////////////////////////////////////////////////////////////
     
-    template<class Impl>
-    class FermionOperator : public CheckerBoardedSparseMatrixBase<typename Impl::FermionField>, public Impl
-    {
-    public:
+template<class Impl>
+class FermionOperator : public CheckerBoardedSparseMatrixBase<typename Impl::FermionField>, public Impl
+{
+public:
 
-      INHERIT_IMPL_TYPES(Impl);
+  INHERIT_IMPL_TYPES(Impl);
 
-      FermionOperator(const ImplParams &p= ImplParams()) : Impl(p) {};
-      virtual ~FermionOperator(void) = default;
+  FermionOperator(const ImplParams &p= ImplParams()) : Impl(p) {};
+  virtual ~FermionOperator(void) = default;
 
-      virtual FermionField &tmp(void) = 0;
+  virtual FermionField &tmp(void) = 0;
 
-      GridBase * Grid(void)   { return FermionGrid(); };   // this is all the linalg routines need to know
-      GridBase * RedBlackGrid(void) { return FermionRedBlackGrid(); };
+  GridBase * Grid(void)   { return FermionGrid(); };   // this is all the linalg routines need to know
+  GridBase * RedBlackGrid(void) { return FermionRedBlackGrid(); };
 
-      virtual GridBase *FermionGrid(void)         =0;
-      virtual GridBase *FermionRedBlackGrid(void) =0;
-      virtual GridBase *GaugeGrid(void)           =0;
-      virtual GridBase *GaugeRedBlackGrid(void)   =0;
+  virtual GridBase *FermionGrid(void)         =0;
+  virtual GridBase *FermionRedBlackGrid(void) =0;
+  virtual GridBase *GaugeGrid(void)           =0;
+  virtual GridBase *GaugeRedBlackGrid(void)   =0;
 
-      // override multiply
-      virtual RealD  M    (const FermionField &in, FermionField &out)=0;
-      virtual RealD  Mdag (const FermionField &in, FermionField &out)=0;
+  // override multiply
+  virtual RealD  M    (const FermionField &in, FermionField &out)=0;
+  virtual RealD  Mdag (const FermionField &in, FermionField &out)=0;
 
-      // half checkerboard operaions
-      virtual void   Meooe       (const FermionField &in, FermionField &out)=0;
-      virtual void   MeooeDag    (const FermionField &in, FermionField &out)=0;
-      virtual void   Mooee       (const FermionField &in, FermionField &out)=0;
-      virtual void   MooeeDag    (const FermionField &in, FermionField &out)=0;
-      virtual void   MooeeInv    (const FermionField &in, FermionField &out)=0;
-      virtual void   MooeeInvDag (const FermionField &in, FermionField &out)=0;
+  // half checkerboard operaions
+  virtual void   Meooe       (const FermionField &in, FermionField &out)=0;
+  virtual void   MeooeDag    (const FermionField &in, FermionField &out)=0;
+  virtual void   Mooee       (const FermionField &in, FermionField &out)=0;
+  virtual void   MooeeDag    (const FermionField &in, FermionField &out)=0;
+  virtual void   MooeeInv    (const FermionField &in, FermionField &out)=0;
+  virtual void   MooeeInvDag (const FermionField &in, FermionField &out)=0;
 
-      // non-hermitian hopping term; half cb or both
-      virtual void Dhop  (const FermionField &in, FermionField &out,int dag)=0;
-      virtual void DhopOE(const FermionField &in, FermionField &out,int dag)=0;
-      virtual void DhopEO(const FermionField &in, FermionField &out,int dag)=0;
-      virtual void DhopDir(const FermionField &in, FermionField &out,int dir,int disp)=0; // implemented by WilsonFermion and WilsonFermion5D
+  // non-hermitian hopping term; half cb or both
+  virtual void Dhop  (const FermionField &in, FermionField &out,int dag)=0;
+  virtual void DhopOE(const FermionField &in, FermionField &out,int dag)=0;
+  virtual void DhopEO(const FermionField &in, FermionField &out,int dag)=0;
+  virtual void DhopDir(const FermionField &in, FermionField &out,int dir,int disp)=0; // implemented by WilsonFermion and WilsonFermion5D
 
-      // force terms; five routines; default to Dhop on diagonal
-      virtual void MDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag){DhopDeriv(mat,U,V,dag);};
-      virtual void MoeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag){DhopDerivOE(mat,U,V,dag);};
-      virtual void MeoDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag){DhopDerivEO(mat,U,V,dag);};
-      virtual void MooDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag){mat=zero;}; // Clover can override these
-      virtual void MeeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag){mat=zero;};
+  // force terms; five routines; default to Dhop on diagonal
+  virtual void MDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag){DhopDeriv(mat,U,V,dag);};
+  virtual void MoeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag){DhopDerivOE(mat,U,V,dag);};
+  virtual void MeoDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag){DhopDerivEO(mat,U,V,dag);};
+  virtual void MooDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag){mat=Zero();}; // Clover can override these
+  virtual void MeeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag){mat=Zero();};
 
-      virtual void DhopDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag)=0;
-      virtual void DhopDerivEO(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)=0;
-      virtual void DhopDerivOE(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)=0;
+  virtual void DhopDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag)=0;
+  virtual void DhopDerivEO(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)=0;
+  virtual void DhopDerivOE(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)=0;
 
 
-      virtual void  Mdiag  (const FermionField &in, FermionField &out) { Mooee(in,out);};   // Same as Mooee applied to both CB's
-      virtual void  Mdir   (const FermionField &in, FermionField &out,int dir,int disp)=0;   // case by case Wilson, Clover, Cayley, ContFrac, PartFrac
+  virtual void  Mdiag  (const FermionField &in, FermionField &out) { Mooee(in,out);};   // Same as Mooee applied to both CB's
+  virtual void  Mdir   (const FermionField &in, FermionField &out,int dir,int disp)=0;   // case by case Wilson, Clover, Cayley, ContFrac, PartFrac
 
 
       virtual void  MomentumSpacePropagator(FermionField &out,const FermionField &in,RealD _m,std::vector<double> twist) { assert(0);};
 
-      virtual void  FreePropagator(const FermionField &in,FermionField &out,RealD mass,std::vector<Complex> boundary,std::vector<double> twist) {
-	FFT theFFT((GridCartesian *) in._grid);
+      virtual void  FreePropagator(const FermionField &in,FermionField &out,RealD mass,std::vector<Complex> boundary,std::vector<double> twist) 
+      {
+	FFT theFFT((GridCartesian *) in.Grid());
 
-	FermionField in_k(in._grid);
-	FermionField prop_k(in._grid);
+	typedef typename Simd::scalar_type Scalar;
+
+	FermionField in_k(in.Grid());
+	FermionField prop_k(in.Grid());
 
 	//phase for boundary condition
-	ComplexField coor(in._grid);
-	ComplexField ph(in._grid);  ph = zero;
-	FermionField in_buf(in._grid); in_buf = zero;
+	ComplexField coor(in.Grid());
+	ComplexField ph(in.Grid());  ph = Zero();
+	FermionField in_buf(in.Grid()); in_buf = Zero();
+
 	Scalar ci(0.0,1.0);
 	assert(twist.size() == Nd);//check that twist is Nd
 	assert(boundary.size() == Nd);//check that boundary conditions is Nd
@@ -113,7 +114,7 @@ namespace Grid {
 	{
           LatticeCoordinate(coor, nu);
 	  double boundary_phase = ::acos(real(boundary[nu]));
-	  ph = ph + boundary_phase*coor*((1./(in._grid->_fdimensions[nu])));
+	  ph = ph + boundary_phase*coor*((1./(in.Grid()->_fdimensions[nu])));
 	  //momenta for propagator shifted by twist+boundary
 	  twist[nu] = twist[nu] + boundary_phase/((2.0*M_PI));
 	}
@@ -124,43 +125,42 @@ namespace Grid {
 	theFFT.FFT_all_dim(out,prop_k,FFT::backward);
 
 	//phase for boundary condition
-	out = out * exp(ci*ph);
+        out = out * exp(Scalar(2.0*M_PI)*ci*ph);
 
       };
 
       virtual void FreePropagator(const FermionField &in,FermionField &out,RealD mass) {
-		std::vector<Complex> boundary;
-		for(int i=0;i<Nd;i++) boundary.push_back(1);//default: periodic boundary conditions
-		std::vector<double> twist(Nd,0.0); //default: periodic boundarys in all directions
-	        FreePropagator(in,out,mass,boundary,twist);
+	std::vector<Complex> boundary;
+	for(int i=0;i<Nd;i++) boundary.push_back(1);//default: periodic boundary conditions
+	std::vector<double> twist(Nd,0.0); //default: periodic boundarys in all directions
+	FreePropagator(in,out,mass,boundary,twist);
       };
 
-      ///////////////////////////////////////////////
-      // Updates gauge field during HMC
-      ///////////////////////////////////////////////
-      virtual void ImportGauge(const GaugeField & _U)=0;
+  ///////////////////////////////////////////////
+  // Updates gauge field during HMC
+  ///////////////////////////////////////////////
+  virtual void ImportGauge(const GaugeField & _U)=0;
 
-      //////////////////////////////////////////////////////////////////////
-      // Conserved currents, either contract at sink or insert sequentially.
-      //////////////////////////////////////////////////////////////////////
-
-      virtual void ContractConservedCurrent(PropagatorField &q_in_1,
-                                            PropagatorField &q_in_2,
-                                            PropagatorField &q_out,
-                                            Current curr_type,
-                                            unsigned int mu)=0;
-      virtual void SeqConservedCurrent(PropagatorField &q_in, 
-                                       PropagatorField &q_out,
-                                       Current curr_type,
-                                       unsigned int mu,
-                                       unsigned int tmin, 
-                                       unsigned int tmax,
-                                       ComplexField &lattice_cmplx)=0;
+  //////////////////////////////////////////////////////////////////////
+  // Conserved currents, either contract at sink or insert sequentially.
+  //////////////////////////////////////////////////////////////////////
+  virtual void ContractConservedCurrent(PropagatorField &q_in_1,
+					PropagatorField &q_in_2,
+					PropagatorField &q_out,
+					Current curr_type,
+					unsigned int mu)=0;
+  virtual void SeqConservedCurrent(PropagatorField &q_in, 
+				   PropagatorField &q_out,
+				   Current curr_type,
+				   unsigned int mu,
+				   unsigned int tmin, 
+				   unsigned int tmax,
+				   ComplexField &lattice_cmplx)=0;
 
       // Only reimplemented in Wilson5D 
       // Default to just a zero correlation function
-      virtual void ContractJ5q(FermionField &q_in   ,ComplexField &J5q) { J5q=zero; };
-      virtual void ContractJ5q(PropagatorField &q_in,ComplexField &J5q) { J5q=zero; };
+  virtual void ContractJ5q(FermionField &q_in   ,ComplexField &J5q) { J5q=Zero(); };
+  virtual void ContractJ5q(PropagatorField &q_in,ComplexField &J5q) { J5q=Zero(); };
 
       ///////////////////////////////////////////////
       // Physical field import/export
@@ -183,9 +183,7 @@ namespace Grid {
       {
 	exported=solution;
       };
-    };
+};
 
-  }
-}
+NAMESPACE_END(Grid);
 
-#endif
diff --git a/Grid/qcd/action/fermion/FermionOperatorImpl.h b/Grid/qcd/action/fermion/FermionOperatorImpl.h
index 721004e1..b444f6dc 100644
--- a/Grid/qcd/action/fermion/FermionOperatorImpl.h
+++ b/Grid/qcd/action/fermion/FermionOperatorImpl.h
@@ -28,1056 +28,162 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
-#ifndef GRID_QCD_FERMION_OPERATOR_IMPL_H
-#define GRID_QCD_FERMION_OPERATOR_IMPL_H
+			   /*  END LEGAL */
 
-namespace Grid {
-namespace QCD {
+#pragma once
 
-  //////////////////////////////////////////////
-  // Template parameter class constructs to package
-  // externally control Fermion implementations
-  // in orthogonal directions
-  //
-  // Ultimately need Impl to always define types where XXX is opaque
-  //
-  //    typedef typename XXX               Simd;
-  //    typedef typename XXX     GaugeLinkField;        
-  //    typedef typename XXX         GaugeField;
-  //    typedef typename XXX      GaugeActField;
-  //    typedef typename XXX       FermionField;
-  //    typedef typename XXX    PropagatorField;
-  //    typedef typename XXX  DoubledGaugeField;
-  //    typedef typename XXX         SiteSpinor;
-  //    typedef typename XXX     SitePropagator;
-  //    typedef typename XXX     SiteHalfSpinor;	
-  //    typedef typename XXX         Compressor;	
-  //
-  // and Methods:
-  //    void ImportGauge(GridBase *GaugeGrid,DoubledGaugeField &Uds,const GaugeField &Umu)
-  //    void DoubleStore(GridBase *GaugeGrid,DoubledGaugeField &Uds,const GaugeField &Umu)
-  //    void multLink(SiteHalfSpinor &phi,const SiteDoubledGaugeField &U,const SiteHalfSpinor &chi,int mu,StencilEntry *SE,StencilImpl &St)
-  //    void InsertForce4D(GaugeField &mat,const FermionField &Btilde,const FermionField &A,int mu)
-  //    void InsertForce5D(GaugeField &mat,const FermionField &Btilde,const FermionField &A,int mu)
-  //
-  //
-  // To acquire the typedefs from "Base" (either a base class or template param) use:
-  //
-  // INHERIT_GIMPL_TYPES(Base)
-  // INHERIT_FIMPL_TYPES(Base)
-  // INHERIT_IMPL_TYPES(Base)
-  //
-  // The Fermion operators will do the following:
-  //
-  // struct MyOpParams { 
-  //   RealD mass;
-  // };
-  //
-  //
-  // template<class Impl>
-  // class MyOp : public<Impl> { 
-  // public:
-  //
-  //    INHERIT_ALL_IMPL_TYPES(Impl);
-  //
-  //    MyOp(MyOpParams Myparm, ImplParams &ImplParam) :  Impl(ImplParam)
-  //    {
-  //
-  //    };
-  //    
-  //  }
-  //////////////////////////////////////////////
+NAMESPACE_BEGIN(Grid);
 
-  template <class T> struct SamePrecisionMapper {
-    typedef T HigherPrecVector ;
-    typedef T LowerPrecVector ;
-  };
-  template <class T> struct LowerPrecisionMapper {  };
-  template <> struct LowerPrecisionMapper<vRealF> {
-    typedef vRealF HigherPrecVector ;
-    typedef vRealH LowerPrecVector ;
-  };
-  template <> struct LowerPrecisionMapper<vRealD> {
-    typedef vRealD HigherPrecVector ;
-    typedef vRealF LowerPrecVector ;
-  };
-  template <> struct LowerPrecisionMapper<vComplexF> {
-    typedef vComplexF HigherPrecVector ;
-    typedef vComplexH LowerPrecVector ;
-  };
-  template <> struct LowerPrecisionMapper<vComplexD> {
-    typedef vComplexD HigherPrecVector ;
-    typedef vComplexF LowerPrecVector ;
-  };
+//////////////////////////////////////////////
+// Template parameter class constructs to package
+// externally control Fermion implementations
+// in orthogonal directions
+//
+// Ultimately need Impl to always define types where XXX is opaque
+//
+//    typedef typename XXX               Simd;
+//    typedef typename XXX     GaugeLinkField;        
+//    typedef typename XXX         GaugeField;
+//    typedef typename XXX      GaugeActField;
+//    typedef typename XXX       FermionField;
+//    typedef typename XXX    PropagatorField;
+//    typedef typename XXX  DoubledGaugeField;
+//    typedef typename XXX         SiteSpinor;
+//    typedef typename XXX     SitePropagator;
+//    typedef typename XXX     SiteHalfSpinor;	
+//    typedef typename XXX         Compressor;	
+//
+// and Methods:
+//    void ImportGauge(GridBase *GaugeGrid,DoubledGaugeField &Uds,const GaugeField &Umu)
+//    void DoubleStore(GridBase *GaugeGrid,DoubledGaugeField &Uds,const GaugeField &Umu)
+//    void multLink(SiteHalfSpinor &phi,const SiteDoubledGaugeField &U,const SiteHalfSpinor &chi,int mu,StencilEntry *SE,StencilImpl::View_type &St)
+//    void InsertForce4D(GaugeField &mat,const FermionField &Btilde,const FermionField &A,int mu)
+//    void InsertForce5D(GaugeField &mat,const FermionField &Btilde,const FermionField &A,int mu)
+//
+//
+// To acquire the typedefs from "Base" (either a base class or template param) use:
+//
+// INHERIT_GIMPL_TYPES(Base)
+// INHERIT_FIMPL_TYPES(Base)
+// INHERIT_IMPL_TYPES(Base)
+//
+// The Fermion operators will do the following:
+//
+// struct MyOpParams { 
+//   RealD mass;
+// };
+//
+//
+// template<class Impl>
+// class MyOp : public<Impl> { 
+// public:
+//
+//    INHERIT_ALL_IMPL_TYPES(Impl);
+//
+//    MyOp(MyOpParams Myparm, ImplParams &ImplParam) :  Impl(ImplParam)
+//    {
+//
+//    };
+//    
+//  }
+//////////////////////////////////////////////
 
-  struct CoeffReal {
-  public:
-    typedef RealD _Coeff_t;
-    static const int Nhcs = 2;
-    template<class Simd> using PrecisionMapper = SamePrecisionMapper<Simd>;
-  };
-  struct CoeffRealHalfComms {
-  public:
-    typedef RealD _Coeff_t;
-    static const int Nhcs = 1;
-    template<class Simd> using PrecisionMapper = LowerPrecisionMapper<Simd>;
-  };
-  struct CoeffComplex {
-  public:
-    typedef ComplexD _Coeff_t;
-    static const int Nhcs = 2;
-    template<class Simd> using PrecisionMapper = SamePrecisionMapper<Simd>;
-  };
-  struct CoeffComplexHalfComms {
-  public:
-    typedef ComplexD _Coeff_t;
-    static const int Nhcs = 1;
-    template<class Simd> using PrecisionMapper = LowerPrecisionMapper<Simd>;
-  };
+template <class T> struct SamePrecisionMapper {
+  typedef T HigherPrecVector ;
+  typedef T LowerPrecVector ;
+};
+template <class T> struct LowerPrecisionMapper {  };
+template <> struct LowerPrecisionMapper<vRealF> {
+  typedef vRealF HigherPrecVector ;
+  typedef vRealH LowerPrecVector ;
+};
+template <> struct LowerPrecisionMapper<vRealD> {
+  typedef vRealD HigherPrecVector ;
+  typedef vRealF LowerPrecVector ;
+};
+template <> struct LowerPrecisionMapper<vComplexF> {
+  typedef vComplexF HigherPrecVector ;
+  typedef vComplexH LowerPrecVector ;
+};
+template <> struct LowerPrecisionMapper<vComplexD> {
+  typedef vComplexD HigherPrecVector ;
+  typedef vComplexF LowerPrecVector ;
+};
 
-  ////////////////////////////////////////////////////////////////////////
-  // Implementation dependent fermion types
-  ////////////////////////////////////////////////////////////////////////
+struct CoeffReal {
+public:
+  typedef RealD _Coeff_t;
+  static const int Nhcs = 2;
+  template<class Simd> using PrecisionMapper = SamePrecisionMapper<Simd>;
+};
+struct CoeffRealHalfComms {
+public:
+  typedef RealD _Coeff_t;
+  static const int Nhcs = 1;
+  template<class Simd> using PrecisionMapper = LowerPrecisionMapper<Simd>;
+};
+struct CoeffComplex {
+public:
+  typedef ComplexD _Coeff_t;
+  static const int Nhcs = 2;
+  template<class Simd> using PrecisionMapper = SamePrecisionMapper<Simd>;
+};
+struct CoeffComplexHalfComms {
+public:
+  typedef ComplexD _Coeff_t;
+  static const int Nhcs = 1;
+  template<class Simd> using PrecisionMapper = LowerPrecisionMapper<Simd>;
+};
+
+////////////////////////////////////////////////////////////////////////
+// Implementation dependent fermion types
+////////////////////////////////////////////////////////////////////////
   
 #define INHERIT_FIMPL_TYPES(Impl)\
+  typedef typename Impl::Coeff_t                     Coeff_t;           \
   typedef Impl Impl_t;							\
   typedef typename Impl::FermionField           FermionField;		\
   typedef typename Impl::PropagatorField     PropagatorField;		\
   typedef typename Impl::DoubledGaugeField DoubledGaugeField;		\
+  typedef typename Impl::SiteDoubledGaugeField SiteDoubledGaugeField;	\
   typedef typename Impl::SiteSpinor               SiteSpinor;		\
   typedef typename Impl::SitePropagator       SitePropagator;		\
   typedef typename Impl::SiteHalfSpinor       SiteHalfSpinor;		\
   typedef typename Impl::Compressor               Compressor;		\
   typedef typename Impl::StencilImpl             StencilImpl;		\
   typedef typename Impl::ImplParams               ImplParams;	        \
-  typedef typename Impl::Coeff_t                     Coeff_t;           \
-  
-#define INHERIT_IMPL_TYPES(Base) \
-  INHERIT_GIMPL_TYPES(Base)      \
+  typedef typename Impl::StencilImpl::View_type  StencilView;		\
+  typedef typename ViewMap<FermionField>::Type      FermionFieldView;	\
+  typedef typename ViewMap<DoubledGaugeField>::Type DoubledGaugeFieldView;
+
+#define INHERIT_IMPL_TYPES(Base)		\
+  INHERIT_GIMPL_TYPES(Base)			\
   INHERIT_FIMPL_TYPES(Base)
-  
-  /////////////////////////////////////////////////////////////////////////////
-  // Single flavour four spinors with colour index
-  /////////////////////////////////////////////////////////////////////////////
-  template <class S, class Representation = FundamentalRepresentation,class Options = CoeffReal >
-  class WilsonImpl : public PeriodicGaugeImpl<GaugeImplTypes<S, Representation::Dimension > > {
-    public:
 
-    static const int Dimension = Representation::Dimension;
-    static const bool isFundamental = Representation::isFundamental;
-    static const bool LsVectorised=false;
-    static const int Nhcs = Options::Nhcs;
-
-    typedef PeriodicGaugeImpl<GaugeImplTypes<S, Dimension > > Gimpl;
-    INHERIT_GIMPL_TYPES(Gimpl);
-      
-    //Necessary?
-    constexpr bool is_fundamental() const{return Dimension == Nc ? 1 : 0;}
-    
-    typedef typename Options::_Coeff_t Coeff_t;
-    typedef typename Options::template PrecisionMapper<Simd>::LowerPrecVector SimdL;
-      
-    template <typename vtype> using iImplSpinor            = iScalar<iVector<iVector<vtype, Dimension>, Ns> >;
-    template <typename vtype> using iImplPropagator        = iScalar<iMatrix<iMatrix<vtype, Dimension>, Ns> >;
-    template <typename vtype> using iImplHalfSpinor        = iScalar<iVector<iVector<vtype, Dimension>, Nhs> >;
-    template <typename vtype> using iImplHalfCommSpinor    = iScalar<iVector<iVector<vtype, Dimension>, Nhcs> >;
-    template <typename vtype> using iImplDoubledGaugeField = iVector<iScalar<iMatrix<vtype, Dimension> >, Nds>;
-    
-    typedef iImplSpinor<Simd>            SiteSpinor;
-    typedef iImplPropagator<Simd>        SitePropagator;
-    typedef iImplHalfSpinor<Simd>        SiteHalfSpinor;
-    typedef iImplHalfCommSpinor<SimdL>   SiteHalfCommSpinor;
-    typedef iImplDoubledGaugeField<Simd> SiteDoubledGaugeField;
-    
-    typedef Lattice<SiteSpinor>            FermionField;
-    typedef Lattice<SitePropagator>        PropagatorField;
-    typedef Lattice<SiteDoubledGaugeField> DoubledGaugeField;
-    
-    typedef WilsonCompressor<SiteHalfCommSpinor,SiteHalfSpinor, SiteSpinor> Compressor;
-    typedef WilsonImplParams ImplParams;
-    typedef WilsonStencil<SiteSpinor, SiteHalfSpinor> StencilImpl;
-    
-    ImplParams Params;
-    
-    WilsonImpl(const ImplParams &p = ImplParams()) : Params(p){
-      assert(Params.boundary_phases.size() == Nd);
-    };
-      
-    bool overlapCommsCompute(void) { return Params.overlapCommsCompute; };
-      
-    inline void multLink(SiteHalfSpinor &phi,
-                         const SiteDoubledGaugeField &U,
-                         const SiteHalfSpinor &chi,
-                         int mu,
-                         StencilEntry *SE,
-                         StencilImpl &St) {
-      mult(&phi(), &U(mu), &chi());
-    }
-    
-    inline void multLinkProp(SitePropagator &phi,
-                             const SiteDoubledGaugeField &U,
-                             const SitePropagator &chi,
-                             int mu) {
-       mult(&phi(), &U(mu), &chi());
-    }
-      
-    template <class ref>
-    inline void loadLinkElement(Simd &reg, ref &memory) {
-      reg = memory;
-    }
-      
-    inline void DoubleStore(GridBase *GaugeGrid,
-                            DoubledGaugeField &Uds,
-                            const GaugeField &Umu) 
-    {
-      typedef typename Simd::scalar_type scalar_type;
-
-      conformable(Uds._grid, GaugeGrid);
-      conformable(Umu._grid, GaugeGrid);
-
-      GaugeLinkField U(GaugeGrid);
-      GaugeLinkField tmp(GaugeGrid);
-
-      Lattice<iScalar<vInteger> > coor(GaugeGrid);
-      ////////////////////////////////////////////////////
-      // apply any boundary phase or twists
-      ////////////////////////////////////////////////////
-      for (int mu = 0; mu < Nd; mu++) {
-
-	////////// boundary phase /////////////
-	auto pha = Params.boundary_phases[mu];
-	scalar_type phase( real(pha),imag(pha) );
-
-	int L   = GaugeGrid->GlobalDimensions()[mu];
-        int Lmu = L - 1;
-
-        LatticeCoordinate(coor, mu);
-
-        U = PeekIndex<LorentzIndex>(Umu, mu);
-
-	// apply any twists
-	RealD theta = Params.twist_n_2pi_L[mu] * 2*M_PI / L;
-	if ( theta != 0.0) { 
-	  scalar_type twphase(::cos(theta),::sin(theta));
-	  U = twphase*U;
-	  std::cout << GridLogMessage << " Twist ["<<mu<<"] "<< Params.twist_n_2pi_L[mu]<< " phase"<<phase <<std::endl;
-	}
-
-        tmp = where(coor == Lmu, phase * U, U);
-        PokeIndex<LorentzIndex>(Uds, tmp, mu);
-
-        U = adj(Cshift(U, mu, -1));
-        U = where(coor == 0, conjugate(phase) * U, U); 
-        PokeIndex<LorentzIndex>(Uds, U, mu + 4);
-      }
-    }
-
-    inline void InsertForce4D(GaugeField &mat, FermionField &Btilde, FermionField &A,int mu){
-      GaugeLinkField link(mat._grid);
-      link = TraceIndex<SpinIndex>(outerProduct(Btilde,A)); 
-      PokeIndex<LorentzIndex>(mat,link,mu);
-    }  
-    
-    inline void outerProductImpl(PropagatorField &mat, const FermionField &B, const FermionField &A){
-      mat = outerProduct(B,A); 
-    }  
-
-    inline void TraceSpinImpl(GaugeLinkField &mat, PropagatorField&P) {
-      mat = TraceIndex<SpinIndex>(P); 
-    }
-      
-    inline void extractLinkField(std::vector<GaugeLinkField> &mat, DoubledGaugeField &Uds){
-      for (int mu = 0; mu < Nd; mu++)
-      mat[mu] = PeekIndex<LorentzIndex>(Uds, mu);
-    }
-
-
-    inline void InsertForce5D(GaugeField &mat, FermionField &Btilde, FermionField &Atilde,int mu){
-      
-      int Ls=Btilde._grid->_fdimensions[0];
-      GaugeLinkField tmp(mat._grid);
-      tmp = zero;
-      
-      parallel_for(int sss=0;sss<tmp._grid->oSites();sss++){
-	    int sU=sss;
-	    for(int s=0;s<Ls;s++){
-	        int sF = s+Ls*sU;
-	        tmp[sU] = tmp[sU]+ traceIndex<SpinIndex>(outerProduct(Btilde[sF],Atilde[sF])); // ordering here
-	    }
-      }
-      PokeIndex<LorentzIndex>(mat,tmp,mu);
-      
-    }
-  };
-
-  ////////////////////////////////////////////////////////////////////////////////////
-  // Single flavour four spinors with colour index, 5d redblack
-  ////////////////////////////////////////////////////////////////////////////////////
-template<class S,class Representation = FundamentalRepresentation, class Options=CoeffReal>
-class DomainWallVec5dImpl :  public PeriodicGaugeImpl< GaugeImplTypes< S,Representation::Dimension> > { 
-  public:
-
-  typedef PeriodicGaugeImpl<GaugeImplTypes<S, Representation::Dimension> > Gimpl;
-  INHERIT_GIMPL_TYPES(Gimpl);
-
-  static const int Dimension = Representation::Dimension;
-  static const bool isFundamental = Representation::isFundamental;
-  static const bool LsVectorised=true;
-  static const int Nhcs = Options::Nhcs;
-      
-  typedef typename Options::_Coeff_t Coeff_t;      
-  typedef typename Options::template PrecisionMapper<Simd>::LowerPrecVector SimdL;
-  
-  template <typename vtype> using iImplSpinor            = iScalar<iVector<iVector<vtype, Dimension>, Ns> >;
-  template <typename vtype> using iImplPropagator        = iScalar<iMatrix<iMatrix<vtype, Dimension>, Ns> >;
-  template <typename vtype> using iImplHalfSpinor        = iScalar<iVector<iVector<vtype, Dimension>, Nhs> >;
-  template <typename vtype> using iImplHalfCommSpinor    = iScalar<iVector<iVector<vtype, Dimension>, Nhcs> >;
-  template <typename vtype> using iImplDoubledGaugeField = iVector<iScalar<iMatrix<vtype, Dimension> >, Nds>;
-  template <typename vtype> using iImplGaugeField        = iVector<iScalar<iMatrix<vtype, Dimension> >, Nd>;
-  template <typename vtype> using iImplGaugeLink         = iScalar<iScalar<iMatrix<vtype, Dimension> > >;
-  
-  typedef iImplSpinor<Simd>            SiteSpinor;
-  typedef iImplPropagator<Simd>        SitePropagator;
-  typedef iImplHalfSpinor<Simd>        SiteHalfSpinor;
-  typedef iImplHalfCommSpinor<SimdL>   SiteHalfCommSpinor;
-  typedef Lattice<SiteSpinor>          FermionField;
-  typedef Lattice<SitePropagator>      PropagatorField;
-
-  /////////////////////////////////////////////////
-  // Make the doubled gauge field a *scalar*
-  /////////////////////////////////////////////////
-  typedef iImplDoubledGaugeField<typename Simd::scalar_type>  SiteDoubledGaugeField;  // This is a scalar
-  typedef iImplGaugeField<typename Simd::scalar_type>         SiteScalarGaugeField;  // scalar
-  typedef iImplGaugeLink<typename Simd::scalar_type>          SiteScalarGaugeLink;  // scalar
-  typedef Lattice<SiteDoubledGaugeField>                      DoubledGaugeField;
-      
-  typedef WilsonCompressor<SiteHalfCommSpinor,SiteHalfSpinor, SiteSpinor> Compressor;
-  typedef WilsonImplParams ImplParams;
-  typedef WilsonStencil<SiteSpinor, SiteHalfSpinor> StencilImpl;
-  
-  ImplParams Params;
-  
-  DomainWallVec5dImpl(const ImplParams &p = ImplParams()) : Params(p){};
-      
-  bool overlapCommsCompute(void) { return false; };
-      
-  template <class ref>
-  inline void loadLinkElement(Simd &reg, ref &memory) {
-    vsplat(reg, memory);
-  }
-
-  inline void multLink(SiteHalfSpinor &phi, const SiteDoubledGaugeField &U,
-                       const SiteHalfSpinor &chi, int mu, StencilEntry *SE,
-                       StencilImpl &St) {
-    SiteGaugeLink UU;
-    for (int i = 0; i < Dimension; i++) {
-      for (int j = 0; j < Dimension; j++) {
-        vsplat(UU()()(i, j), U(mu)()(i, j));
-      }
-    }
-    mult(&phi(), &UU(), &chi());
-  }
-
-  inline void multLinkProp(SitePropagator &phi,
-                           const SiteDoubledGaugeField &U,
-                           const SitePropagator &chi,
-                           int mu) {
-    SiteGaugeLink UU;
-    for (int i = 0; i < Dimension; i++) {
-      for (int j = 0; j < Dimension; j++) {
-        vsplat(UU()()(i, j), U(mu)()(i, j));
-      }
-    }
-    mult(&phi(), &UU(), &chi());
-  }
-
-  inline void DoubleStore(GridBase *GaugeGrid, DoubledGaugeField &Uds,const GaugeField &Umu) 
-  {
-    SiteScalarGaugeField  ScalarUmu;
-    SiteDoubledGaugeField ScalarUds;
-    
-    GaugeLinkField U(Umu._grid);
-    GaugeField  Uadj(Umu._grid);
-    for (int mu = 0; mu < Nd; mu++) {
-      U = PeekIndex<LorentzIndex>(Umu, mu);
-      U = adj(Cshift(U, mu, -1));
-      PokeIndex<LorentzIndex>(Uadj, U, mu);
-    }
-    
-    for (int lidx = 0; lidx < GaugeGrid->lSites(); lidx++) {
-      std::vector<int> lcoor;
-      GaugeGrid->LocalIndexToLocalCoor(lidx, lcoor);
-      
-      peekLocalSite(ScalarUmu, Umu, lcoor);
-      for (int mu = 0; mu < 4; mu++) ScalarUds(mu) = ScalarUmu(mu);
-      
-      peekLocalSite(ScalarUmu, Uadj, lcoor);
-      for (int mu = 0; mu < 4; mu++) ScalarUds(mu + 4) = ScalarUmu(mu);
-      
-      pokeLocalSite(ScalarUds, Uds, lcoor);
-    }
-  }
-      
-  inline void InsertForce4D(GaugeField &mat, FermionField &Btilde,FermionField &A, int mu) 
-  {
-    assert(0);
-  }
-
-  inline void outerProductImpl(PropagatorField &mat, const FermionField &Btilde, const FermionField &A){
-    assert(0);
-  } 
-
-  inline void TraceSpinImpl(GaugeLinkField &mat, PropagatorField&P) {
-    assert(0);
-  }
-
-  inline void extractLinkField(std::vector<GaugeLinkField> &mat, DoubledGaugeField &Uds){
-    assert(0);
-  }
-
-
-  inline void InsertForce5D(GaugeField &mat, FermionField &Btilde, FermionField &Atilde, int mu) {
-
-    assert(0);
-    // Following lines to be revised after Peter's addition of half prec
-    // missing put lane...
-    /*
-    typedef decltype(traceIndex<SpinIndex>(outerProduct(Btilde[0], Atilde[0]))) result_type;
-    unsigned int LLs = Btilde._grid->_rdimensions[0];
-    conformable(Atilde._grid,Btilde._grid);
-    GridBase* grid = mat._grid;
-    GridBase* Bgrid = Btilde._grid;
-    unsigned int dimU = grid->Nd();
-    unsigned int dimF = Bgrid->Nd();
-    GaugeLinkField tmp(grid); 
-    tmp = zero;
-    
-    // FIXME 
-    // Current implementation works, thread safe, probably suboptimal
-    // Passing through the local coordinate for grid transformation
-    // the force grid is in general very different from the Ls vectorized grid
-
-    PARALLEL_FOR_LOOP
-    for (int so = 0; so < grid->oSites(); so++) {
-      std::vector<typename result_type::scalar_object> vres(Bgrid->Nsimd());
-      std::vector<int> ocoor;  grid->oCoorFromOindex(ocoor,so); 
-      for (int si = 0; si < tmp._grid->iSites(); si++){
-        typename result_type::scalar_object scalar_object; scalar_object = zero;
-        std::vector<int> local_coor;      
-        std::vector<int> icoor; grid->iCoorFromIindex(icoor,si);
-        grid->InOutCoorToLocalCoor(ocoor, icoor, local_coor);
-        for (int s = 0; s < LLs; s++) {
-          std::vector<int> slocal_coor(dimF);
-          slocal_coor[0] = s;
-          for (int s4d = 1; s4d< dimF; s4d++) slocal_coor[s4d] = local_coor[s4d-1];
-          int sF = Bgrid->oIndexReduced(slocal_coor);  
-          assert(sF < Bgrid->oSites());
-
-          extract(traceIndex<SpinIndex>(outerProduct(Btilde[sF], Atilde[sF])), vres); 
-          // sum across the 5d dimension
-          for (auto v : vres) scalar_object += v;  
-        }
-        tmp._odata[so].putlane(scalar_object, si);
-      }
-    }
-    PokeIndex<LorentzIndex>(mat, tmp, mu);
-    */
-  }
-};
-    
-    ////////////////////////////////////////////////////////////////////////////////////////
-    // Flavour doubled spinors; is Gparity the only? what about C*?
-    ////////////////////////////////////////////////////////////////////////////////////////
-template <class S, class Representation = FundamentalRepresentation, class Options=CoeffReal>
-class GparityWilsonImpl : public ConjugateGaugeImpl<GaugeImplTypes<S, Representation::Dimension> > {
- public:
-
- static const int Dimension = Representation::Dimension;
- static const bool isFundamental = Representation::isFundamental;
- static const int Nhcs = Options::Nhcs;
- static const bool LsVectorised=false;
-
- typedef ConjugateGaugeImpl< GaugeImplTypes<S,Dimension> > Gimpl;
- INHERIT_GIMPL_TYPES(Gimpl);
-
- typedef typename Options::_Coeff_t Coeff_t;
- typedef typename Options::template PrecisionMapper<Simd>::LowerPrecVector SimdL;
-      
- template <typename vtype> using iImplSpinor            = iVector<iVector<iVector<vtype, Dimension>, Ns>,   Ngp>;
- template <typename vtype> using iImplPropagator        = iVector<iMatrix<iMatrix<vtype, Dimension>, Ns>,   Ngp>;
- template <typename vtype> using iImplHalfSpinor        = iVector<iVector<iVector<vtype, Dimension>, Nhs>,  Ngp>;
- template <typename vtype> using iImplHalfCommSpinor    = iVector<iVector<iVector<vtype, Dimension>, Nhcs>, Ngp>;
- template <typename vtype> using iImplDoubledGaugeField = iVector<iVector<iScalar<iMatrix<vtype, Dimension> >, Nds>, Ngp>;
-
- typedef iImplSpinor<Simd>            SiteSpinor;
- typedef iImplPropagator<Simd>        SitePropagator;
- typedef iImplHalfSpinor<Simd>        SiteHalfSpinor;
- typedef iImplHalfCommSpinor<SimdL>   SiteHalfCommSpinor;
- typedef iImplDoubledGaugeField<Simd> SiteDoubledGaugeField;
-
- typedef Lattice<SiteSpinor> FermionField;
- typedef Lattice<SitePropagator> PropagatorField;
- typedef Lattice<SiteDoubledGaugeField> DoubledGaugeField;
- 
- typedef WilsonCompressor<SiteHalfCommSpinor,SiteHalfSpinor, SiteSpinor> Compressor;
- typedef WilsonStencil<SiteSpinor, SiteHalfSpinor> StencilImpl;
- 
- typedef GparityWilsonImplParams ImplParams;
-      
- ImplParams Params;
-
- GparityWilsonImpl(const ImplParams &p = ImplParams()) : Params(p){};
-
- bool overlapCommsCompute(void) { return Params.overlapCommsCompute; };
-
- // provide the multiply by link that is differentiated between Gparity (with
- // flavour index) and non-Gparity
- inline void multLink(SiteHalfSpinor &phi, const SiteDoubledGaugeField &U,
-                      const SiteHalfSpinor &chi, int mu, StencilEntry *SE,
-                      StencilImpl &St) {
-
-   typedef SiteHalfSpinor vobj;
-   typedef typename SiteHalfSpinor::scalar_object sobj;
-	
-   vobj vtmp;
-   sobj stmp;
-        
-   GridBase *grid = St._grid;
-        
-   const int Nsimd = grid->Nsimd();
-        
-   int direction = St._directions[mu];
-   int distance = St._distances[mu];
-   int ptype = St._permute_type[mu];
-   int sl = St._grid->_simd_layout[direction];
+NAMESPACE_END(Grid);
+NAMESPACE_CHECK(ImplBase);  
+/////////////////////////////////////////////////////////////////////////////
+// Single flavour four spinors with colour index
+/////////////////////////////////////////////////////////////////////////////
+#include <Grid/qcd/action/fermion/WilsonImpl.h> 
+NAMESPACE_CHECK(ImplWilson);  
    
-   // Fixme X.Y.Z.T hardcode in stencil
-   int mmu = mu % Nd;
-        
-   // assert our assumptions
-   assert((distance == 1) || (distance == -1));  // nearest neighbour stencil hard code
-   assert((sl == 1) || (sl == 2));
-   
-   std::vector<int> icoor;
-        
-   if ( SE->_around_the_world && Params.twists[mmu] ) {
-
-     if ( sl == 2 ) {
-       
-       std::vector<sobj> vals(Nsimd);
-
-       extract(chi,vals);
-       for(int s=0;s<Nsimd;s++){
-
-         grid->iCoorFromIindex(icoor,s);
-              
-         assert((icoor[direction]==0)||(icoor[direction]==1));
-              
-         int permute_lane;
-         if ( distance == 1) {
-           permute_lane = icoor[direction]?1:0;
-         } else {
-           permute_lane = icoor[direction]?0:1;
-         }
-              
-         if ( permute_lane ) { 
-           stmp(0) = vals[s](1);
-           stmp(1) = vals[s](0);
-           vals[s] = stmp;
-              }
-       }
-       merge(vtmp,vals);
-            
-     } else { 
-       vtmp(0) = chi(1);
-       vtmp(1) = chi(0);
-     }
-     mult(&phi(0),&U(0)(mu),&vtmp(0));
-     mult(&phi(1),&U(1)(mu),&vtmp(1));
-     
-   } else { 
-     mult(&phi(0),&U(0)(mu),&chi(0));
-     mult(&phi(1),&U(1)(mu),&chi(1));
-   }
-   
- }
-    // Fixme: Gparity prop * link
-    inline void multLinkProp(SitePropagator &phi, const SiteDoubledGaugeField &U,
-                             const SitePropagator &chi, int mu)
-    {
-        assert(0);
-    }
-
- template <class ref>
- inline void loadLinkElement(Simd &reg, ref &memory) {
-   reg = memory;
- }
-
- inline void DoubleStore(GridBase *GaugeGrid,DoubledGaugeField &Uds,const GaugeField &Umu)
- {
-   conformable(Uds._grid,GaugeGrid);
-   conformable(Umu._grid,GaugeGrid);
-   
-   GaugeLinkField Utmp (GaugeGrid);
-   GaugeLinkField U    (GaugeGrid);
-   GaugeLinkField Uconj(GaugeGrid);
-   
-   Lattice<iScalar<vInteger> > coor(GaugeGrid);
-        
-   for(int mu=0;mu<Nd;mu++){
-          
-     LatticeCoordinate(coor,mu);
-          
-     U     = PeekIndex<LorentzIndex>(Umu,mu);
-     Uconj = conjugate(U);
-     
-     // This phase could come from a simple bc 1,1,-1,1 ..
-     int neglink = GaugeGrid->GlobalDimensions()[mu]-1;
-     if ( Params.twists[mu] ) { 
-       Uconj = where(coor==neglink,-Uconj,Uconj);
-     }
-	  
-     parallel_for(auto ss=U.begin();ss<U.end();ss++){
-       Uds[ss](0)(mu) = U[ss]();
-       Uds[ss](1)(mu) = Uconj[ss]();
-     }
-          
-     U     = adj(Cshift(U    ,mu,-1));      // correct except for spanning the boundary
-     Uconj = adj(Cshift(Uconj,mu,-1));
- 
-     Utmp = U;
-     if ( Params.twists[mu] ) { 
-       Utmp = where(coor==0,Uconj,Utmp);
-     }
-
-	  
-     parallel_for(auto ss=U.begin();ss<U.end();ss++){
-       Uds[ss](0)(mu+4) = Utmp[ss]();
-     }
-          
-     Utmp = Uconj;
-     if ( Params.twists[mu] ) { 
-       Utmp = where(coor==0,U,Utmp);
-     }
-	  
-     parallel_for(auto ss=U.begin();ss<U.end();ss++){
-       Uds[ss](1)(mu+4) = Utmp[ss]();
-     }
-          
-   }
- }
-      
- inline void InsertForce4D(GaugeField &mat, FermionField &Btilde, FermionField &A, int mu) {
-
-   // DhopDir provides U or Uconj depending on coor/flavour.
-   GaugeLinkField link(mat._grid);
-   // use lorentz for flavour as hack.
-   auto tmp = TraceIndex<SpinIndex>(outerProduct(Btilde, A));
-   parallel_for(auto ss = tmp.begin(); ss < tmp.end(); ss++) {
-     link[ss]() = tmp[ss](0, 0) + conjugate(tmp[ss](1, 1));
-   }
-   PokeIndex<LorentzIndex>(mat, link, mu);
-   return;
- }
-      
- inline void outerProductImpl(PropagatorField &mat, const FermionField &Btilde, const FermionField &A){
-   //mat = outerProduct(Btilde, A);
-   assert(0);
-  }
-
-  inline void TraceSpinImpl(GaugeLinkField &mat, PropagatorField&P) {
-    assert(0);
-    /*
-    auto tmp = TraceIndex<SpinIndex>(P);
-    parallel_for(auto ss = tmp.begin(); ss < tmp.end(); ss++) {
-      mat[ss]() = tmp[ss](0, 0) + conjugate(tmp[ss](1, 1));
-    }
-    */
-  }
-
-  inline void extractLinkField(std::vector<GaugeLinkField> &mat, DoubledGaugeField &Uds){
-    assert(0);
-  }
-  
- inline void InsertForce5D(GaugeField &mat, FermionField &Btilde, FermionField &Atilde, int mu) {
-
-   int Ls = Btilde._grid->_fdimensions[0];
-        
-   GaugeLinkField tmp(mat._grid);
-   tmp = zero;
-   parallel_for(int ss = 0; ss < tmp._grid->oSites(); ss++) {
-     for (int s = 0; s < Ls; s++) {
-       int sF = s + Ls * ss;
-       auto ttmp = traceIndex<SpinIndex>(outerProduct(Btilde[sF], Atilde[sF]));
-       tmp[ss]() = tmp[ss]() + ttmp(0, 0) + conjugate(ttmp(1, 1));
-     }
-   }
-   PokeIndex<LorentzIndex>(mat, tmp, mu);
-   return;
- }
-
-};
+////////////////////////////////////////////////////////////////////////////////////////
+// Flavour doubled spinors; is Gparity the only? what about C*?
+////////////////////////////////////////////////////////////////////////////////////////
+#include <Grid/qcd/action/fermion/GparityWilsonImpl.h> 
+NAMESPACE_CHECK(ImplGparityWilson);  
 
 /////////////////////////////////////////////////////////////////////////////
 // Single flavour one component spinors with colour index
 /////////////////////////////////////////////////////////////////////////////
-template <class S, class Representation = FundamentalRepresentation >
-class StaggeredImpl : public PeriodicGaugeImpl<GaugeImplTypes<S, Representation::Dimension > > {
+#include <Grid/qcd/action/fermion/StaggeredImpl.h> 
+NAMESPACE_CHECK(ImplStaggered);  
 
-    public:
-
-    typedef RealD  _Coeff_t ;
-    static const int Dimension = Representation::Dimension;
-    static const bool isFundamental = Representation::isFundamental;
-    static const bool LsVectorised=false;
-    typedef PeriodicGaugeImpl<GaugeImplTypes<S, Dimension > > Gimpl;
-      
-    //Necessary?
-    constexpr bool is_fundamental() const{return Dimension == Nc ? 1 : 0;}
-    
-    typedef _Coeff_t Coeff_t;
-
-    INHERIT_GIMPL_TYPES(Gimpl);
-      
-    template <typename vtype> using iImplSpinor            = iScalar<iScalar<iVector<vtype, Dimension> > >;
-    template <typename vtype> using iImplHalfSpinor        = iScalar<iScalar<iVector<vtype, Dimension> > >;
-    template <typename vtype> using iImplDoubledGaugeField = iVector<iScalar<iMatrix<vtype, Dimension> >, Nds>;
-    template <typename vtype> using iImplPropagator        = iScalar<iScalar<iMatrix<vtype, Dimension> > >;
-    
-    typedef iImplSpinor<Simd>            SiteSpinor;
-    typedef iImplHalfSpinor<Simd>        SiteHalfSpinor;
-    typedef iImplDoubledGaugeField<Simd> SiteDoubledGaugeField;
-    typedef iImplPropagator<Simd>        SitePropagator;
-    
-    typedef Lattice<SiteSpinor>            FermionField;
-    typedef Lattice<SiteDoubledGaugeField> DoubledGaugeField;
-    typedef Lattice<SitePropagator> PropagatorField;
-    
-    typedef SimpleCompressor<SiteSpinor> Compressor;
-    typedef StaggeredImplParams ImplParams;
-    typedef CartesianStencil<SiteSpinor, SiteSpinor> StencilImpl;
-    
-    ImplParams Params;
-    
-    StaggeredImpl(const ImplParams &p = ImplParams()) : Params(p){};
-      
-    inline void multLink(SiteSpinor &phi,
-			 const SiteDoubledGaugeField &U,
-			 const SiteSpinor &chi,
-			 int mu){
-      mult(&phi(), &U(mu), &chi());
-    }
-    inline void multLinkAdd(SiteSpinor &phi,
-			    const SiteDoubledGaugeField &U,
-			    const SiteSpinor &chi,
-			    int mu){
-      mac(&phi(), &U(mu), &chi());
-    }
-      
-    template <class ref>
-    inline void loadLinkElement(Simd &reg, ref &memory) {
-      reg = memory;
-    }
-
-    inline void InsertGaugeField(DoubledGaugeField &U_ds,
-				 const GaugeLinkField &U,int mu)
-    {
-      PokeIndex<LorentzIndex>(U_ds, U, mu);
-    }
-    inline void DoubleStore(GridBase *GaugeGrid,
-			    DoubledGaugeField &UUUds, // for Naik term
-			    DoubledGaugeField &Uds,
-			    const GaugeField &Uthin,
-			    const GaugeField &Ufat) {
-      conformable(Uds._grid, GaugeGrid);
-      conformable(Uthin._grid, GaugeGrid);
-      conformable(Ufat._grid, GaugeGrid);
-      GaugeLinkField U(GaugeGrid);
-      GaugeLinkField UU(GaugeGrid);
-      GaugeLinkField UUU(GaugeGrid);
-      GaugeLinkField Udag(GaugeGrid);
-      GaugeLinkField UUUdag(GaugeGrid);
-      for (int mu = 0; mu < Nd; mu++) {
-
-	// Staggered Phase.
-	Lattice<iScalar<vInteger> > coor(GaugeGrid);
-	Lattice<iScalar<vInteger> > x(GaugeGrid); LatticeCoordinate(x,0);
-	Lattice<iScalar<vInteger> > y(GaugeGrid); LatticeCoordinate(y,1);
-	Lattice<iScalar<vInteger> > z(GaugeGrid); LatticeCoordinate(z,2);
-	Lattice<iScalar<vInteger> > t(GaugeGrid); LatticeCoordinate(t,3);
-
-	Lattice<iScalar<vInteger> > lin_z(GaugeGrid); lin_z=x+y;
-	Lattice<iScalar<vInteger> > lin_t(GaugeGrid); lin_t=x+y+z;
-
-	ComplexField phases(GaugeGrid);	phases=1.0;
-
-	if ( mu == 1 ) phases = where( mod(x    ,2)==(Integer)0, phases,-phases);
-	if ( mu == 2 ) phases = where( mod(lin_z,2)==(Integer)0, phases,-phases);
-	if ( mu == 3 ) phases = where( mod(lin_t,2)==(Integer)0, phases,-phases);
-
-	// 1 hop based on fat links
-	U      = PeekIndex<LorentzIndex>(Ufat, mu);
-	Udag   = adj( Cshift(U, mu, -1));
-
-	U    = U    *phases;
-	Udag = Udag *phases;
-
-	InsertGaugeField(Uds,U,mu);
-	InsertGaugeField(Uds,Udag,mu+4);
-	//	PokeIndex<LorentzIndex>(Uds, U, mu);
-	//	PokeIndex<LorentzIndex>(Uds, Udag, mu + 4);
-
-	// 3 hop based on thin links. Crazy huh ?
-	U  = PeekIndex<LorentzIndex>(Uthin, mu);
-	UU = Gimpl::CovShiftForward(U,mu,U);
-	UUU= Gimpl::CovShiftForward(U,mu,UU);
-	
-	UUUdag = adj( Cshift(UUU, mu, -3));
-
-	UUU    = UUU    *phases;
-	UUUdag = UUUdag *phases;
-
-	InsertGaugeField(UUUds,UUU,mu);
-	InsertGaugeField(UUUds,UUUdag,mu+4);
-
-      }
-    }
-
-    inline void InsertForce4D(GaugeField &mat, FermionField &Btilde, FermionField &A,int mu){
-      GaugeLinkField link(mat._grid);
-      link = TraceIndex<SpinIndex>(outerProduct(Btilde,A)); 
-      PokeIndex<LorentzIndex>(mat,link,mu);
-    } 
-          
-    inline void InsertForce5D(GaugeField &mat, FermionField &Btilde, FermionField &Atilde,int mu){
-      assert (0); 
-      // Must never hit
-    }
-  };
-
-  /////////////////////////////////////////////////////////////////////////////
-  // Single flavour one component spinors with colour index. 5d vec
-  /////////////////////////////////////////////////////////////////////////////
-  template <class S, class Representation = FundamentalRepresentation >
-  class StaggeredVec5dImpl : public PeriodicGaugeImpl<GaugeImplTypes<S, Representation::Dimension > > {
-
-    public:
-
-    static const int Dimension = Representation::Dimension;
-    static const bool isFundamental = Representation::isFundamental;
-    static const bool LsVectorised=true;
-    typedef RealD   Coeff_t ;
-    typedef PeriodicGaugeImpl<GaugeImplTypes<S, Dimension > > Gimpl;
-      
-    //Necessary?
-    constexpr bool is_fundamental() const{return Dimension == Nc ? 1 : 0;}
+/////////////////////////////////////////////////////////////////////////////
+// Single flavour one component spinors with colour index. 5d vec
+/////////////////////////////////////////////////////////////////////////////
+#include <Grid/qcd/action/fermion/StaggeredVec5dImpl.h> 
+NAMESPACE_CHECK(ImplStaggered5dVec);  
 
 
-    INHERIT_GIMPL_TYPES(Gimpl);
-
-    template <typename vtype> using iImplSpinor            = iScalar<iScalar<iVector<vtype, Dimension> > >;
-    template <typename vtype> using iImplHalfSpinor        = iScalar<iScalar<iVector<vtype, Dimension> > >;
-    template <typename vtype> using iImplDoubledGaugeField = iVector<iScalar<iMatrix<vtype, Dimension> >, Nds>;
-    template <typename vtype> using iImplGaugeField        = iVector<iScalar<iMatrix<vtype, Dimension> >, Nd>;
-    template <typename vtype> using iImplGaugeLink         = iScalar<iScalar<iMatrix<vtype, Dimension> > >;
-    template <typename vtype> using iImplPropagator        = iScalar<iScalar<iMatrix<vtype, Dimension> > >;
-
-    // Make the doubled gauge field a *scalar*
-    typedef iImplDoubledGaugeField<typename Simd::scalar_type>  SiteDoubledGaugeField;  // This is a scalar
-    typedef iImplGaugeField<typename Simd::scalar_type>         SiteScalarGaugeField;  // scalar
-    typedef iImplGaugeLink<typename Simd::scalar_type>          SiteScalarGaugeLink;  // scalar
-    typedef iImplPropagator<Simd>        SitePropagator;
-
-    typedef Lattice<SiteDoubledGaugeField> DoubledGaugeField;
-    typedef Lattice<SitePropagator> PropagatorField;
-    
-    typedef iImplSpinor<Simd>            SiteSpinor;
-    typedef iImplHalfSpinor<Simd>        SiteHalfSpinor;
-
-    
-    typedef Lattice<SiteSpinor>            FermionField;
-    
-    typedef SimpleCompressor<SiteSpinor> Compressor;
-    typedef StaggeredImplParams ImplParams;
-    typedef CartesianStencil<SiteSpinor, SiteSpinor> StencilImpl;
-    
-    ImplParams Params;
-    
-    StaggeredVec5dImpl(const ImplParams &p = ImplParams()) : Params(p){};
-
-    template <class ref>
-    inline void loadLinkElement(Simd &reg, ref &memory) {
-      vsplat(reg, memory);
-    }
-
-    inline void multLink(SiteHalfSpinor &phi, const SiteDoubledGaugeField &U,
-			 const SiteHalfSpinor &chi, int mu) {
-      SiteGaugeLink UU;
-      for (int i = 0; i < Dimension; i++) {
-	for (int j = 0; j < Dimension; j++) {
-	  vsplat(UU()()(i, j), U(mu)()(i, j));
-	}
-      }
-      mult(&phi(), &UU(), &chi());
-    }
-    inline void multLinkAdd(SiteHalfSpinor &phi, const SiteDoubledGaugeField &U,
-			    const SiteHalfSpinor &chi, int mu) {
-      SiteGaugeLink UU;
-      for (int i = 0; i < Dimension; i++) {
-	for (int j = 0; j < Dimension; j++) {
-	  vsplat(UU()()(i, j), U(mu)()(i, j));
-	}
-      }
-      mac(&phi(), &UU(), &chi());
-    }
-      
-    inline void InsertGaugeField(DoubledGaugeField &U_ds,const GaugeLinkField &U,int mu)
-    {
-      GridBase *GaugeGrid = U_ds._grid;
-      parallel_for (int lidx = 0; lidx < GaugeGrid->lSites(); lidx++) {
-
-	SiteScalarGaugeLink   ScalarU;
-	SiteDoubledGaugeField ScalarUds;
-	
-	std::vector<int> lcoor;
-	GaugeGrid->LocalIndexToLocalCoor(lidx, lcoor);
-	peekLocalSite(ScalarUds, U_ds, lcoor);
-	
-	peekLocalSite(ScalarU, U, lcoor);
-	ScalarUds(mu) = ScalarU();
-	
-      }
-    }
-    inline void DoubleStore(GridBase *GaugeGrid,
-			    DoubledGaugeField &UUUds, // for Naik term
-			    DoubledGaugeField &Uds,
-			    const GaugeField &Uthin,
-			    const GaugeField &Ufat) 
-    {
-
-      GridBase * InputGrid = Uthin._grid;
-      conformable(InputGrid,Ufat._grid);
-
-      GaugeLinkField U(InputGrid);
-      GaugeLinkField UU(InputGrid);
-      GaugeLinkField UUU(InputGrid);
-      GaugeLinkField Udag(InputGrid);
-      GaugeLinkField UUUdag(InputGrid);
-
-      for (int mu = 0; mu < Nd; mu++) {
-
-	// Staggered Phase.
-	Lattice<iScalar<vInteger> > coor(InputGrid);
-	Lattice<iScalar<vInteger> > x(InputGrid); LatticeCoordinate(x,0);
-	Lattice<iScalar<vInteger> > y(InputGrid); LatticeCoordinate(y,1);
-	Lattice<iScalar<vInteger> > z(InputGrid); LatticeCoordinate(z,2);
-	Lattice<iScalar<vInteger> > t(InputGrid); LatticeCoordinate(t,3);
-
-	Lattice<iScalar<vInteger> > lin_z(InputGrid); lin_z=x+y;
-	Lattice<iScalar<vInteger> > lin_t(InputGrid); lin_t=x+y+z;
-
-	ComplexField phases(InputGrid);	phases=1.0;
-
-	if ( mu == 1 ) phases = where( mod(x    ,2)==(Integer)0, phases,-phases);
-	if ( mu == 2 ) phases = where( mod(lin_z,2)==(Integer)0, phases,-phases);
-	if ( mu == 3 ) phases = where( mod(lin_t,2)==(Integer)0, phases,-phases);
-
-	// 1 hop based on fat links
-	U      = PeekIndex<LorentzIndex>(Ufat, mu);
-	Udag   = adj( Cshift(U, mu, -1));
-
-	U    = U    *phases;
-	Udag = Udag *phases;
-
-	InsertGaugeField(Uds,U,mu);
-	InsertGaugeField(Uds,Udag,mu+4);
-
-	// 3 hop based on thin links. Crazy huh ?
-	U  = PeekIndex<LorentzIndex>(Uthin, mu);
-	UU = Gimpl::CovShiftForward(U,mu,U);
-	UUU= Gimpl::CovShiftForward(U,mu,UU);
-	
-	UUUdag = adj( Cshift(UUU, mu, -3));
-
-	UUU    = UUU    *phases;
-	UUUdag = UUUdag *phases;
-
-	InsertGaugeField(UUUds,UUU,mu);
-	InsertGaugeField(UUUds,UUUdag,mu+4);
-
-      }
-    }
-
-    inline void InsertForce4D(GaugeField &mat, FermionField &Btilde, FermionField &A,int mu){
-      assert(0);
-    }   
-      
-    inline void InsertForce5D(GaugeField &mat, FermionField &Btilde, FermionField &Atilde,int mu){
-      assert (0); 
-    }
-  };
-
-typedef WilsonImpl<vComplex,  FundamentalRepresentation, CoeffReal > WilsonImplR;  // Real.. whichever prec
-typedef WilsonImpl<vComplexF, FundamentalRepresentation, CoeffReal > WilsonImplF;  // Float
-typedef WilsonImpl<vComplexD, FundamentalRepresentation, CoeffReal > WilsonImplD;  // Double
-
-typedef WilsonImpl<vComplex,  FundamentalRepresentation, CoeffRealHalfComms > WilsonImplRL;  // Real.. whichever prec
-typedef WilsonImpl<vComplexF, FundamentalRepresentation, CoeffRealHalfComms > WilsonImplFH;  // Float
-typedef WilsonImpl<vComplexD, FundamentalRepresentation, CoeffRealHalfComms > WilsonImplDF;  // Double
-
-typedef WilsonImpl<vComplex,  FundamentalRepresentation, CoeffComplex > ZWilsonImplR; // Real.. whichever prec
-typedef WilsonImpl<vComplexF, FundamentalRepresentation, CoeffComplex > ZWilsonImplF; // Float
-typedef WilsonImpl<vComplexD, FundamentalRepresentation, CoeffComplex > ZWilsonImplD; // Double
-
-typedef WilsonImpl<vComplex,  FundamentalRepresentation, CoeffComplexHalfComms > ZWilsonImplRL; // Real.. whichever prec
-typedef WilsonImpl<vComplexF, FundamentalRepresentation, CoeffComplexHalfComms > ZWilsonImplFH; // Float
-typedef WilsonImpl<vComplexD, FundamentalRepresentation, CoeffComplexHalfComms > ZWilsonImplDF; // Double
- 
-typedef WilsonImpl<vComplex,  AdjointRepresentation, CoeffReal > WilsonAdjImplR;   // Real.. whichever prec
-typedef WilsonImpl<vComplexF, AdjointRepresentation, CoeffReal > WilsonAdjImplF;  // Float
-typedef WilsonImpl<vComplexD, AdjointRepresentation, CoeffReal > WilsonAdjImplD;  // Double
- 
-typedef WilsonImpl<vComplex,  TwoIndexSymmetricRepresentation, CoeffReal > WilsonTwoIndexSymmetricImplR;   // Real.. whichever prec
-typedef WilsonImpl<vComplexF, TwoIndexSymmetricRepresentation, CoeffReal > WilsonTwoIndexSymmetricImplF;  // Float
-typedef WilsonImpl<vComplexD, TwoIndexSymmetricRepresentation, CoeffReal > WilsonTwoIndexSymmetricImplD;  // Double
- 
-typedef WilsonImpl<vComplex,  TwoIndexAntiSymmetricRepresentation, CoeffReal > WilsonTwoIndexAntiSymmetricImplR;   // Real.. whichever prec
-typedef WilsonImpl<vComplexF, TwoIndexAntiSymmetricRepresentation, CoeffReal > WilsonTwoIndexAntiSymmetricImplF;  // Float
-typedef WilsonImpl<vComplexD, TwoIndexAntiSymmetricRepresentation, CoeffReal > WilsonTwoIndexAntiSymmetricImplD;  // Double
-
-typedef DomainWallVec5dImpl<vComplex ,FundamentalRepresentation, CoeffReal> DomainWallVec5dImplR; // Real.. whichever prec
-typedef DomainWallVec5dImpl<vComplexF,FundamentalRepresentation, CoeffReal> DomainWallVec5dImplF; // Float
-typedef DomainWallVec5dImpl<vComplexD,FundamentalRepresentation, CoeffReal> DomainWallVec5dImplD; // Double
- 
-typedef DomainWallVec5dImpl<vComplex ,FundamentalRepresentation, CoeffRealHalfComms> DomainWallVec5dImplRL; // Real.. whichever prec
-typedef DomainWallVec5dImpl<vComplexF,FundamentalRepresentation, CoeffRealHalfComms> DomainWallVec5dImplFH; // Float
-typedef DomainWallVec5dImpl<vComplexD,FundamentalRepresentation, CoeffRealHalfComms> DomainWallVec5dImplDF; // Double
- 
-typedef DomainWallVec5dImpl<vComplex ,FundamentalRepresentation,CoeffComplex> ZDomainWallVec5dImplR; // Real.. whichever prec
-typedef DomainWallVec5dImpl<vComplexF,FundamentalRepresentation,CoeffComplex> ZDomainWallVec5dImplF; // Float
-typedef DomainWallVec5dImpl<vComplexD,FundamentalRepresentation,CoeffComplex> ZDomainWallVec5dImplD; // Double
- 
-typedef DomainWallVec5dImpl<vComplex ,FundamentalRepresentation,CoeffComplexHalfComms> ZDomainWallVec5dImplRL; // Real.. whichever prec
-typedef DomainWallVec5dImpl<vComplexF,FundamentalRepresentation,CoeffComplexHalfComms> ZDomainWallVec5dImplFH; // Float
-typedef DomainWallVec5dImpl<vComplexD,FundamentalRepresentation,CoeffComplexHalfComms> ZDomainWallVec5dImplDF; // Double
- 
-typedef GparityWilsonImpl<vComplex , FundamentalRepresentation,CoeffReal> GparityWilsonImplR;  // Real.. whichever prec
-typedef GparityWilsonImpl<vComplexF, FundamentalRepresentation,CoeffReal> GparityWilsonImplF;  // Float
-typedef GparityWilsonImpl<vComplexD, FundamentalRepresentation,CoeffReal> GparityWilsonImplD;  // Double
- 
-typedef GparityWilsonImpl<vComplex , FundamentalRepresentation,CoeffRealHalfComms> GparityWilsonImplRL;  // Real.. whichever prec
-typedef GparityWilsonImpl<vComplexF, FundamentalRepresentation,CoeffRealHalfComms> GparityWilsonImplFH;  // Float
-typedef GparityWilsonImpl<vComplexD, FundamentalRepresentation,CoeffRealHalfComms> GparityWilsonImplDF;  // Double
-
-typedef StaggeredImpl<vComplex,  FundamentalRepresentation > StaggeredImplR;   // Real.. whichever prec
-typedef StaggeredImpl<vComplexF, FundamentalRepresentation > StaggeredImplF;  // Float
-typedef StaggeredImpl<vComplexD, FundamentalRepresentation > StaggeredImplD;  // Double
-
-typedef StaggeredVec5dImpl<vComplex,  FundamentalRepresentation > StaggeredVec5dImplR;   // Real.. whichever prec
-typedef StaggeredVec5dImpl<vComplexF, FundamentalRepresentation > StaggeredVec5dImplF;  // Float
-typedef StaggeredVec5dImpl<vComplexD, FundamentalRepresentation > StaggeredVec5dImplD;  // Double
-
-}}
-
-#endif
diff --git a/Grid/qcd/action/fermion/FourierAcceleratedPV.h b/Grid/qcd/action/fermion/FourierAcceleratedPV.h
index d6196eee..bf23ff75 100644
--- a/Grid/qcd/action/fermion/FourierAcceleratedPV.h
+++ b/Grid/qcd/action/fermion/FourierAcceleratedPV.h
@@ -28,8 +28,8 @@ Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
     *************************************************************************************/
     /*  END LEGAL */
 #pragma once
-namespace Grid {
-namespace QCD {
+
+NAMESPACE_BEGIN(Grid);
 
   template<typename M>
     void get_real_const_bc(M& m, RealD& _b, RealD& _c) {
@@ -63,8 +63,8 @@ class FourierAcceleratedPV {
    : dwfPV(_dwfPV), Umu(_Umu), cg(_cg), group_in_s(_group_in_s) 
   {
     assert( dwfPV.FermionGrid()->_fdimensions[0] % (2*group_in_s) == 0);
-    grid5D = QCD::SpaceTimeGrid::makeFiveDimGrid(2*group_in_s, (GridCartesian*)Umu._grid);
-    gridRB5D = QCD::SpaceTimeGrid::makeFiveDimRedBlackGrid(2*group_in_s, (GridCartesian*)Umu._grid);
+    grid5D   = SpaceTimeGrid::makeFiveDimGrid(2*group_in_s, (GridCartesian*)Umu.Grid());
+    gridRB5D = SpaceTimeGrid::makeFiveDimRedBlackGrid(2*group_in_s, (GridCartesian*)Umu.Grid());
   }
 
   void rotatePV(const Vi& _src, Vi& dst, bool forward) const {
@@ -72,13 +72,13 @@ class FourierAcceleratedPV {
     GridStopWatch gsw1, gsw2;
 
     typedef typename Vi::scalar_type Coeff_t;
-    int Ls = dst._grid->_fdimensions[0];
+    int Ls = dst.Grid()->_fdimensions[0];
 
-    Vi _tmp(dst._grid);
+    Vi _tmp(dst.Grid());
     double phase = M_PI / (double)Ls;
     Coeff_t bzero(0.0,0.0);
 
-    FFT theFFT((GridCartesian*)dst._grid);
+    FFT theFFT((GridCartesian*)dst.Grid());
 
     if (!forward) {
       gsw1.Start();
@@ -115,7 +115,7 @@ class FourierAcceleratedPV {
     std::cout << GridLogMessage << "Fourier-Accelerated Outer Pauli Villars"<<std::endl;
 
     typedef typename Vi::scalar_type Coeff_t;
-    int Ls = _dst._grid->_fdimensions[0];
+    int Ls = _dst.Grid()->_fdimensions[0];
 
     GridStopWatch gswT;
     gswT.Start();
@@ -126,12 +126,12 @@ class FourierAcceleratedPV {
     
     // U(true) Rightinv TMinv U(false) = Minv
 
-    Vi _src_diag(_dst._grid);
+    Vi _src_diag(_dst.Grid());
     Vi _src_diag_slice(dwfPV.GaugeGrid());
     Vi _dst_diag_slice(dwfPV.GaugeGrid());
     Vi _src_diag_slices(grid5D);
     Vi _dst_diag_slices(grid5D);
-    Vi _dst_diag(_dst._grid);
+    Vi _dst_diag(_dst.Grid());
 
     rotatePV(_src,_src_diag,false);
 
@@ -163,7 +163,7 @@ class FourierAcceleratedPV {
       for (int sidx=0;sidx<group_in_s;sidx++) {
 
 	int s = sgroup*group_in_s + sidx;
-	int sprime = Ls-s-1;
+	//	int sprime = Ls-s-1;
 
 	RealD phase = M_PI / (RealD)Ls * (2.0 * s + 1.0);
 	RealD cosp = ::cos(phase);
@@ -196,7 +196,7 @@ class FourierAcceleratedPV {
 
       GridStopWatch gsw;
       gsw.Start();
-      _dst_diag_slices = zero; // zero guess
+      _dst_diag_slices = Zero(); // zero guess
       sol(tm,_src_diag_slices,_dst_diag_slices);
       gsw.Stop();
       std::cout << GridLogMessage << "Solve[sgroup=" << sgroup << "] completed in " << gsw.Elapsed() << ", " << gswA.Elapsed() << std::endl;
@@ -212,7 +212,7 @@ class FourierAcceleratedPV {
 
 	// now rotate with inverse of
 	Coeff_t pA = b + c*cosp;
-	Coeff_t pB = - Coeff_t(0.0,1.0)*c*sinp;
+	Coeff_t pB = - Coeff_t(0.0,1.0)*Coeff_t(c*sinp);
 	Coeff_t pABden = pA*pA - pB*pB;
 	// (pA + pB * G5) * (pA - pB*G5) = (pA^2 - pB^2)
       
@@ -234,4 +234,5 @@ class FourierAcceleratedPV {
   }
 
 };
-}}
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/GparityWilsonImpl.h b/Grid/qcd/action/fermion/GparityWilsonImpl.h
new file mode 100644
index 00000000..f87d2260
--- /dev/null
+++ b/Grid/qcd/action/fermion/GparityWilsonImpl.h
@@ -0,0 +1,321 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/FermionOperatorImpl.h
+
+Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+			   /*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+template <class S, class Representation = FundamentalRepresentation, class Options=CoeffReal>
+class GparityWilsonImpl : public ConjugateGaugeImpl<GaugeImplTypes<S, Representation::Dimension> > {
+public:
+
+ static const int Dimension = Representation::Dimension;
+ static const bool isFundamental = Representation::isFundamental;
+ static const int Nhcs = Options::Nhcs;
+ static const bool LsVectorised=false;
+
+ typedef ConjugateGaugeImpl< GaugeImplTypes<S,Dimension> > Gimpl;
+ INHERIT_GIMPL_TYPES(Gimpl);
+ 
+ typedef typename Options::_Coeff_t Coeff_t;
+ typedef typename Options::template PrecisionMapper<Simd>::LowerPrecVector SimdL;
+      
+ template <typename vtype> using iImplSpinor            = iVector<iVector<iVector<vtype, Dimension>, Ns>,   Ngp>;
+ template <typename vtype> using iImplPropagator        = iVector<iMatrix<iMatrix<vtype, Dimension>, Ns>,   Ngp>;
+ template <typename vtype> using iImplHalfSpinor        = iVector<iVector<iVector<vtype, Dimension>, Nhs>,  Ngp>;
+ template <typename vtype> using iImplHalfCommSpinor    = iVector<iVector<iVector<vtype, Dimension>, Nhcs>, Ngp>;
+ template <typename vtype> using iImplDoubledGaugeField = iVector<iVector<iScalar<iMatrix<vtype, Dimension> >, Nds>, Ngp>;
+
+  typedef iImplSpinor<Simd>            SiteSpinor;
+  typedef iImplPropagator<Simd>        SitePropagator;
+  typedef iImplHalfSpinor<Simd>        SiteHalfSpinor;
+  typedef iImplHalfCommSpinor<SimdL>   SiteHalfCommSpinor;
+  typedef iImplDoubledGaugeField<Simd> SiteDoubledGaugeField;
+
+  typedef Lattice<SiteSpinor> FermionField;
+  typedef Lattice<SitePropagator> PropagatorField;
+  typedef Lattice<SiteDoubledGaugeField> DoubledGaugeField;
+ 
+  typedef GparityWilsonImplParams ImplParams;
+  typedef WilsonCompressor<SiteHalfCommSpinor,SiteHalfSpinor, SiteSpinor> Compressor;
+  typedef WilsonStencil<SiteSpinor, SiteHalfSpinor, ImplParams> StencilImpl;
+  typedef typename StencilImpl::View_type StencilView;
+      
+  ImplParams Params;
+
+  GparityWilsonImpl(const ImplParams &p = ImplParams()) : Params(p){};
+
+  // provide the multiply by link that is differentiated between Gparity (with
+  // flavour index) and non-Gparity
+  template<class _Spinor>
+  static accelerator_inline void multLink(_Spinor &phi, 
+					  const SiteDoubledGaugeField &U,
+					  const _Spinor &chi, 
+					  int mu) 
+  {
+    assert(0);
+  } 
+  template<class _Spinor>
+  static accelerator_inline void multLink(_Spinor &phi, 
+					  const SiteDoubledGaugeField &U,
+					  const _Spinor &chi, 
+					  int mu, 
+					  StencilEntry *SE,
+					  StencilView &St) 
+  {
+    int direction = St._directions[mu];
+    int distance  = St._distances[mu];
+    int ptype     = St._permute_type[mu];
+    int sl        = St._simd_layout[direction];
+    Coordinate icoor;
+
+#ifdef __CUDA_ARCH__
+    _Spinor tmp;
+
+    const int Nsimd =SiteDoubledGaugeField::Nsimd();
+    int s = SIMTlane(Nsimd);
+    St.iCoorFromIindex(icoor,s);
+
+    int mmu = mu % Nd;
+    if ( SE->_around_the_world && St.parameters.twists[mmu] ) {
+      
+      int permute_lane = (sl==1) 
+    	|| ((distance== 1)&&(icoor[direction]==1))
+	|| ((distance==-1)&&(icoor[direction]==0));
+
+      if ( permute_lane ) { 
+	tmp(0) = chi(1);
+	tmp(1) = chi(0);
+      } else {
+	tmp(0) = chi(0);
+	tmp(1) = chi(1);
+      }
+
+      auto UU0=coalescedRead(U(0)(mu));
+      auto UU1=coalescedRead(U(1)(mu));
+
+      mult(&phi(0),&UU0,&tmp(0));
+      mult(&phi(1),&UU1,&tmp(1));
+
+    } else {
+
+      auto UU0=coalescedRead(U(0)(mu));
+      auto UU1=coalescedRead(U(1)(mu));
+
+      mult(&phi(0),&UU0,&chi(0));
+      mult(&phi(1),&UU1,&chi(1));
+
+    }
+
+#else
+    typedef _Spinor vobj;
+    typedef typename SiteHalfSpinor::scalar_object sobj;
+    typedef typename SiteHalfSpinor::vector_type   vector_type;
+	
+    vobj vtmp;
+    sobj stmp;
+        
+    const int Nsimd =vector_type::Nsimd();
+    
+    // Fixme X.Y.Z.T hardcode in stencil
+    int mmu = mu % Nd;
+        
+    // assert our assumptions
+    assert((distance == 1) || (distance == -1));  // nearest neighbour stencil hard code
+    assert((sl == 1) || (sl == 2));
+
+    if ( SE->_around_the_world && St.parameters.twists[mmu] ) {
+
+      if ( sl == 2 ) {
+       
+	ExtractBuffer<sobj> vals(Nsimd);
+
+	extract(chi,vals);
+	for(int s=0;s<Nsimd;s++){
+
+	  St.iCoorFromIindex(icoor,s);
+              
+	  assert((icoor[direction]==0)||(icoor[direction]==1));
+              
+	  int permute_lane;
+	  if ( distance == 1) {
+	    permute_lane = icoor[direction]?1:0;
+	  } else {
+	    permute_lane = icoor[direction]?0:1;
+	  }
+              
+	  if ( permute_lane ) { 
+	    stmp(0) = vals[s](1);
+	    stmp(1) = vals[s](0);
+	    vals[s] = stmp;
+	  }
+	}
+	merge(vtmp,vals);
+            
+      } else { 
+	vtmp(0) = chi(1);
+	vtmp(1) = chi(0);
+      }
+      mult(&phi(0),&U(0)(mu),&vtmp(0));
+      mult(&phi(1),&U(1)(mu),&vtmp(1));
+     
+    } else { 
+      mult(&phi(0),&U(0)(mu),&chi(0));
+      mult(&phi(1),&U(1)(mu),&chi(1));
+    }
+#endif   
+  }
+
+  template <class ref>
+  static accelerator_inline void loadLinkElement(Simd &reg, ref &memory) 
+  {
+    reg = memory;
+  }
+
+  inline void DoubleStore(GridBase *GaugeGrid,DoubledGaugeField &Uds,const GaugeField &Umu)
+  {
+    conformable(Uds.Grid(),GaugeGrid);
+    conformable(Umu.Grid(),GaugeGrid);
+   
+    GaugeLinkField Utmp (GaugeGrid);
+    GaugeLinkField U    (GaugeGrid);
+    GaugeLinkField Uconj(GaugeGrid);
+   
+    Lattice<iScalar<vInteger> > coor(GaugeGrid);
+        
+    for(int mu=0;mu<Nd;mu++){
+          
+      LatticeCoordinate(coor,mu);
+          
+      U     = PeekIndex<LorentzIndex>(Umu,mu);
+      Uconj = conjugate(U);
+     
+      // This phase could come from a simple bc 1,1,-1,1 ..
+      int neglink = GaugeGrid->GlobalDimensions()[mu]-1;
+      if ( Params.twists[mu] ) { 
+	Uconj = where(coor==neglink,-Uconj,Uconj);
+      }
+	  
+      auto U_v = U.View();
+      auto Uds_v = Uds.View();
+      auto Uconj_v = Uconj.View();
+      auto Utmp_v= Utmp.View();
+      thread_foreach(ss,U_v,{
+	Uds_v[ss](0)(mu) = U_v[ss]();
+	Uds_v[ss](1)(mu) = Uconj_v[ss]();
+      });
+          
+      U     = adj(Cshift(U    ,mu,-1));      // correct except for spanning the boundary
+      Uconj = adj(Cshift(Uconj,mu,-1));
+ 
+      Utmp = U;
+      if ( Params.twists[mu] ) { 
+	Utmp = where(coor==0,Uconj,Utmp);
+      }
+
+      thread_foreach(ss,Utmp_v,{
+	Uds_v[ss](0)(mu+4) = Utmp_v[ss]();
+      });
+          
+      Utmp = Uconj;
+      if ( Params.twists[mu] ) { 
+	Utmp = where(coor==0,U,Utmp);
+      }
+	  
+      thread_foreach(ss,Utmp_v,{
+        Uds_v[ss](1)(mu+4) = Utmp_v[ss]();
+      });
+          
+    }
+  }
+      
+  inline void InsertForce4D(GaugeField &mat, FermionField &Btilde, FermionField &A, int mu) {
+
+    // DhopDir provides U or Uconj depending on coor/flavour.
+    GaugeLinkField link(mat.Grid());
+    // use lorentz for flavour as hack.
+    auto tmp = TraceIndex<SpinIndex>(outerProduct(Btilde, A));
+    auto link_v = link.View();
+    auto tmp_v = tmp.View();
+    thread_foreach(ss,tmp_v,{
+      link_v[ss]() = tmp_v[ss](0, 0) + conjugate(tmp_v[ss](1, 1));
+    });
+    PokeIndex<LorentzIndex>(mat, link, mu);
+    return;
+  }
+      
+ inline void outerProductImpl(PropagatorField &mat, const FermionField &Btilde, const FermionField &A){
+   //mat = outerProduct(Btilde, A);
+   assert(0);
+  }
+
+  inline void TraceSpinImpl(GaugeLinkField &mat, PropagatorField&P) {
+    assert(0);
+    /*
+    auto tmp = TraceIndex<SpinIndex>(P);
+    parallel_for(auto ss = tmp.begin(); ss < tmp.end(); ss++) {
+      mat[ss]() = tmp[ss](0, 0) + conjugate(tmp[ss](1, 1));
+    }
+    */
+  }
+
+  inline void extractLinkField(std::vector<GaugeLinkField> &mat, DoubledGaugeField &Uds){
+    assert(0);
+  }
+  
+  inline void InsertForce5D(GaugeField &mat, FermionField &Btilde, FermionField &Atilde, int mu) {
+
+    int Ls = Btilde.Grid()->_fdimensions[0];
+        
+    GaugeLinkField tmp(mat.Grid());
+    tmp = Zero();
+    auto tmp_v = tmp.View();
+    auto Atilde_v = Atilde.View();
+    auto Btilde_v = Btilde.View();
+    thread_for(ss,tmp.Grid()->oSites(),{
+      for (int s = 0; s < Ls; s++) {
+	int sF = s + Ls * ss;
+	auto ttmp = traceIndex<SpinIndex>(outerProduct(Btilde_v[sF], Atilde_v[sF]));
+	tmp_v[ss]() = tmp_v[ss]() + ttmp(0, 0) + conjugate(ttmp(1, 1));
+      }
+    });
+    PokeIndex<LorentzIndex>(mat, tmp, mu);
+    return;
+  }
+  
+};
+
+typedef GparityWilsonImpl<vComplex , FundamentalRepresentation,CoeffReal> GparityWilsonImplR;  // Real.. whichever prec
+typedef GparityWilsonImpl<vComplexF, FundamentalRepresentation,CoeffReal> GparityWilsonImplF;  // Float
+typedef GparityWilsonImpl<vComplexD, FundamentalRepresentation,CoeffReal> GparityWilsonImplD;  // Double
+ 
+typedef GparityWilsonImpl<vComplex , FundamentalRepresentation,CoeffRealHalfComms> GparityWilsonImplRL;  // Real.. whichever prec
+typedef GparityWilsonImpl<vComplexF, FundamentalRepresentation,CoeffRealHalfComms> GparityWilsonImplFH;  // Float
+typedef GparityWilsonImpl<vComplexD, FundamentalRepresentation,CoeffRealHalfComms> GparityWilsonImplDF;  // Double
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/ImprovedStaggeredFermion.h b/Grid/qcd/action/fermion/ImprovedStaggeredFermion.h
index 3aa78d1b..b4d8d60b 100644
--- a/Grid/qcd/action/fermion/ImprovedStaggeredFermion.h
+++ b/Grid/qcd/action/fermion/ImprovedStaggeredFermion.h
@@ -25,16 +25,14 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef GRID_QCD_IMPR_STAG_FERMION_H
 #define GRID_QCD_IMPR_STAG_FERMION_H
 
-namespace Grid {
-
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 class ImprovedStaggeredFermionStatic {
- public:
+public:
   static const std::vector<int> directions;
   static const std::vector<int> displacements;
   static const int npoint = 16;
@@ -42,7 +40,7 @@ class ImprovedStaggeredFermionStatic {
 
 template <class Impl>
 class ImprovedStaggeredFermion : public StaggeredKernels<Impl>, public ImprovedStaggeredFermionStatic {
- public:
+public:
   INHERIT_IMPL_TYPES(Impl);
   typedef StaggeredKernels<Impl> Kernels;
 
@@ -139,7 +137,7 @@ class ImprovedStaggeredFermion : public StaggeredKernels<Impl>, public ImprovedS
 
   // DoubleStore impl dependent
   void ImportGauge      (const GaugeField &_Uthin ) { assert(0); }
-  void ImportGauge      (const GaugeField &_Uthin  ,const GaugeField &_Ufat);
+  void ImportGauge(const GaugeField &_Uthin, const GaugeField &_Ufat);
   void ImportGaugeSimple(const GaugeField &_UUU    ,const GaugeField &_U);
   void ImportGaugeSimple(const DoubledGaugeField &_UUU,const DoubledGaugeField &_U);
   DoubledGaugeField &GetU(void)   { return Umu ; } ;
@@ -151,7 +149,7 @@ class ImprovedStaggeredFermion : public StaggeredKernels<Impl>, public ImprovedS
   ///////////////////////////////////////////////////////////////
 
   //    protected:
- public:
+public:
   // any other parameters of action ???
   virtual int   isTrivialEE(void) { return 1; };
   virtual RealD Mass(void) { return mass; }
@@ -188,11 +186,11 @@ class ImprovedStaggeredFermion : public StaggeredKernels<Impl>, public ImprovedS
                                 PropagatorField &q_out,
                                 Current curr_type,
                                 unsigned int mu);
-  void SeqConservedCurrent(PropagatorField &q_in, 
+  void SeqConservedCurrent(PropagatorField &q_in,
                            PropagatorField &q_out,
-                           Current curr_type, 
-                           unsigned int mu,
-                           unsigned int tmin, 
+                           Current curr_type,
+                           unsigned int mu, 
+                           unsigned int tmin,
                            unsigned int tmax,
 			   ComplexField &lattice_cmplx);
 };
@@ -200,6 +198,6 @@ class ImprovedStaggeredFermion : public StaggeredKernels<Impl>, public ImprovedS
 typedef ImprovedStaggeredFermion<StaggeredImplF> ImprovedStaggeredFermionF;
 typedef ImprovedStaggeredFermion<StaggeredImplD> ImprovedStaggeredFermionD;
 
-}
-}
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/qcd/action/fermion/ImprovedStaggeredFermion5D.h b/Grid/qcd/action/fermion/ImprovedStaggeredFermion5D.h
index 0d729e60..b10c0356 100644
--- a/Grid/qcd/action/fermion/ImprovedStaggeredFermion5D.h
+++ b/Grid/qcd/action/fermion/ImprovedStaggeredFermion5D.h
@@ -1,5 +1,5 @@
 
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,101 +25,99 @@ Author: AzusaYamaguchi <ayamaguc@staffmail.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef  GRID_QCD_IMPROVED_STAGGERED_FERMION_5D_H
-#define  GRID_QCD_IMPROVED_STAGGERED_FERMION_5D_H
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
-  ////////////////////////////////////////////////////////////////////////////////
-  // This is the 4d red black case appropriate to support
-  ////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////
+// This is the 4d red black case appropriate to support
+////////////////////////////////////////////////////////////////////////////////
 
-    class ImprovedStaggeredFermion5DStatic { 
-    public:
-      // S-direction is INNERMOST and takes no part in the parity.
-      static const std::vector<int> directions;
-      static const std::vector<int> displacements;
-      const int npoint = 16;
-    };
+class ImprovedStaggeredFermion5DStatic { 
+public:
+  // S-direction is INNERMOST and takes no part in the parity.
+  static const std::vector<int> directions;
+  static const std::vector<int> displacements;
+  const int npoint = 16;
+};
 
-    template<class Impl>
-    class ImprovedStaggeredFermion5D :  public StaggeredKernels<Impl>, public ImprovedStaggeredFermion5DStatic 
-    {
-    public:
-      INHERIT_IMPL_TYPES(Impl);
-      typedef StaggeredKernels<Impl> Kernels;
+template<class Impl>
+class ImprovedStaggeredFermion5D :  public StaggeredKernels<Impl>, public ImprovedStaggeredFermion5DStatic 
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+  typedef StaggeredKernels<Impl> Kernels;
 
-      FermionField _tmp;
-      FermionField &tmp(void) { return _tmp; }
+  FermionField _tmp;
+  FermionField &tmp(void) { return _tmp; }
 
-      ////////////////////////////////////////
-      // Performance monitoring
-      ////////////////////////////////////////
-      void Report(void);
-      void ZeroCounters(void);
-      double DhopTotalTime;
-      double DhopCalls;
-      double DhopCommTime;
-      double DhopComputeTime;
+  ////////////////////////////////////////
+  // Performance monitoring
+  ////////////////////////////////////////
+  void Report(void);
+  void ZeroCounters(void);
+  double DhopTotalTime;
+  double DhopCalls;
+  double DhopCommTime;
+  double DhopComputeTime;
       double DhopComputeTime2;
       double DhopFaceTime;
 
-      ///////////////////////////////////////////////////////////////
-      // Implement the abstract base
-      ///////////////////////////////////////////////////////////////
-      GridBase *GaugeGrid(void)              { return _FourDimGrid ;}
-      GridBase *GaugeRedBlackGrid(void)      { return _FourDimRedBlackGrid ;}
-      GridBase *FermionGrid(void)            { return _FiveDimGrid;}
-      GridBase *FermionRedBlackGrid(void)    { return _FiveDimRedBlackGrid;}
+  ///////////////////////////////////////////////////////////////
+  // Implement the abstract base
+  ///////////////////////////////////////////////////////////////
+  GridBase *GaugeGrid(void)              { return _FourDimGrid ;}
+  GridBase *GaugeRedBlackGrid(void)      { return _FourDimRedBlackGrid ;}
+  GridBase *FermionGrid(void)            { return _FiveDimGrid;}
+  GridBase *FermionRedBlackGrid(void)    { return _FiveDimRedBlackGrid;}
 
-      // full checkerboard operations; leave unimplemented as abstract for now
-      RealD  M    (const FermionField &in, FermionField &out);
-      RealD  Mdag (const FermionField &in, FermionField &out);
+  // full checkerboard operations; leave unimplemented as abstract for now
+  RealD  M    (const FermionField &in, FermionField &out);
+  RealD  Mdag (const FermionField &in, FermionField &out);
 
-      // half checkerboard operations
-      void   Meooe       (const FermionField &in, FermionField &out);
-      void   Mooee       (const FermionField &in, FermionField &out);
-      void   MooeeInv    (const FermionField &in, FermionField &out);
+  // half checkerboard operations
+  void   Meooe       (const FermionField &in, FermionField &out);
+  void   Mooee       (const FermionField &in, FermionField &out);
+  void   MooeeInv    (const FermionField &in, FermionField &out);
 
-      void   MeooeDag    (const FermionField &in, FermionField &out);
-      void   MooeeDag    (const FermionField &in, FermionField &out);
-      void   MooeeInvDag (const FermionField &in, FermionField &out);
+  void   MeooeDag    (const FermionField &in, FermionField &out);
+  void   MooeeDag    (const FermionField &in, FermionField &out);
+  void   MooeeInvDag (const FermionField &in, FermionField &out);
 
-      void   Mdir   (const FermionField &in, FermionField &out,int dir,int disp);
-      void DhopDir(const FermionField &in, FermionField &out,int dir,int disp);
+  void   Mdir   (const FermionField &in, FermionField &out,int dir,int disp);
+  void DhopDir(const FermionField &in, FermionField &out,int dir,int disp);
 
-      // These can be overridden by fancy 5d chiral action
-      void DhopDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
-      void DhopDerivEO(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
-      void DhopDerivOE(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  // These can be overridden by fancy 5d chiral action
+  void DhopDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  void DhopDerivEO(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  void DhopDerivOE(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
 
-      // Implement hopping term non-hermitian hopping term; half cb or both
-      void Dhop  (const FermionField &in, FermionField &out,int dag);
-      void DhopOE(const FermionField &in, FermionField &out,int dag);
-      void DhopEO(const FermionField &in, FermionField &out,int dag);
+  // Implement hopping term non-hermitian hopping term; half cb or both
+  void Dhop  (const FermionField &in, FermionField &out,int dag);
+  void DhopOE(const FermionField &in, FermionField &out,int dag);
+  void DhopEO(const FermionField &in, FermionField &out,int dag);
 
     
-    ///////////////////////////////////////////////////////////////
-    // New methods added 
-    ///////////////////////////////////////////////////////////////
-    void DerivInternal(StencilImpl & st,
-		       DoubledGaugeField & U,
-		       DoubledGaugeField & UUU,
-		       GaugeField &mat,
-		       const FermionField &A,
-		       const FermionField &B,
-		       int dag);
+  ///////////////////////////////////////////////////////////////
+  // New methods added 
+  ///////////////////////////////////////////////////////////////
+  void DerivInternal(StencilImpl & st,
+		     DoubledGaugeField & U,
+		     DoubledGaugeField & UUU,
+		     GaugeField &mat,
+		     const FermionField &A,
+		     const FermionField &B,
+		     int dag);
     
-    void DhopInternal(StencilImpl & st,
-		      LebesgueOrder &lo,
-		      DoubledGaugeField &U,
-		      DoubledGaugeField &UUU,
-		      const FermionField &in, 
-		      FermionField &out,
-		      int dag);
+  void DhopInternal(StencilImpl & st,
+		    LebesgueOrder &lo,
+		    DoubledGaugeField &U,
+		    DoubledGaugeField &UUU,
+		    const FermionField &in, 
+		    FermionField &out,
+		    int dag);
     
     void DhopInternalOverlappedComms(StencilImpl & st,
 		      LebesgueOrder &lo,
@@ -138,17 +136,17 @@ namespace QCD {
 		      int dag);
     
     
-    // Constructors
+  // Constructors
     ////////////////////////////////////////////////////////////////////////////////////////////////
     // Grid internal interface -- Thin link and fat link, with coefficients
     ////////////////////////////////////////////////////////////////////////////////////////////////
-    ImprovedStaggeredFermion5D(GaugeField &_Uthin,
-			       GaugeField &_Ufat,
-			       GridCartesian         &FiveDimGrid,
-			       GridRedBlackCartesian &FiveDimRedBlackGrid,
-			       GridCartesian         &FourDimGrid,
-			       GridRedBlackCartesian &FourDimRedBlackGrid,
-			       double _mass,
+  ImprovedStaggeredFermion5D(GaugeField &_Uthin,
+			     GaugeField &_Ufat,
+			     GridCartesian         &FiveDimGrid,
+			     GridRedBlackCartesian &FiveDimRedBlackGrid,
+			     GridCartesian         &FourDimGrid,
+			     GridRedBlackCartesian &FourDimRedBlackGrid,
+			     double _mass,
 			       RealD _c1, RealD _c2,RealD _u0,
 			       const ImplParams &p= ImplParams());
     ////////////////////////////////////////////////////////////////////////////////////////////////
@@ -160,11 +158,11 @@ namespace QCD {
 			       GridRedBlackCartesian &FourDimRedBlackGrid,
 			       double _mass,
 			       RealD _c1=1.0, RealD _c2=1.0,RealD _u0=1.0,
-			       const ImplParams &p= ImplParams());
-
+			     const ImplParams &p= ImplParams());
+    
     // DoubleStore gauge field in operator
     void ImportGauge      (const GaugeField &_Uthin ) { assert(0); }
-    void ImportGauge      (const GaugeField &_Uthin  ,const GaugeField &_Ufat);
+  void ImportGauge(const GaugeField &_Uthin,const GaugeField &_Ufat);
     void ImportGaugeSimple(const GaugeField &_UUU,const GaugeField &_U);
     void ImportGaugeSimple(const DoubledGaugeField &_UUU,const DoubledGaugeField &_U);
     // Give a reference; can be used to do an assignment or copy back out after import
@@ -173,62 +171,61 @@ namespace QCD {
     DoubledGaugeField &GetUUU(void) { return UUUmu; };
     void CopyGaugeCheckerboards(void);
     
-    ///////////////////////////////////////////////////////////////
-    // Data members require to support the functionality
-    ///////////////////////////////////////////////////////////////
-  public:
-
+  ///////////////////////////////////////////////////////////////
+  // Data members require to support the functionality
+  ///////////////////////////////////////////////////////////////
+public:
+    
     virtual int   isTrivialEE(void) { return 1; };
     virtual RealD Mass(void) { return mass; }
     
-    GridBase *_FourDimGrid;
-    GridBase *_FourDimRedBlackGrid;
-    GridBase *_FiveDimGrid;
-    GridBase *_FiveDimRedBlackGrid;
+  GridBase *_FourDimGrid;
+  GridBase *_FourDimRedBlackGrid;
+  GridBase *_FiveDimGrid;
+  GridBase *_FiveDimRedBlackGrid;
     
-    RealD mass;
-    RealD c1;
-    RealD c2;
-    RealD u0;
-    int Ls;
+  RealD mass;
+  RealD c1;
+  RealD c2;
+  RealD u0;
+  int Ls;
     
-    //Defines the stencils for even and odd
-    StencilImpl Stencil; 
-    StencilImpl StencilEven; 
-    StencilImpl StencilOdd; 
+  //Defines the stencils for even and odd
+  StencilImpl Stencil; 
+  StencilImpl StencilEven; 
+  StencilImpl StencilOdd; 
     
-    // Copy of the gauge field , with even and odd subsets
-    DoubledGaugeField Umu;
-    DoubledGaugeField UmuEven;
-    DoubledGaugeField UmuOdd;
+  // Copy of the gauge field , with even and odd subsets
+  DoubledGaugeField Umu;
+  DoubledGaugeField UmuEven;
+  DoubledGaugeField UmuOdd;
 
-    DoubledGaugeField UUUmu;
-    DoubledGaugeField UUUmuEven;
-    DoubledGaugeField UUUmuOdd;
+  DoubledGaugeField UUUmu;
+  DoubledGaugeField UUUmuEven;
+  DoubledGaugeField UUUmuOdd;
     
-    LebesgueOrder Lebesgue;
-    LebesgueOrder LebesgueEvenOdd;
+  LebesgueOrder Lebesgue;
+  LebesgueOrder LebesgueEvenOdd;
     
-    // Comms buffer
-    std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  comm_buf;
+  // Comms buffer
+  std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  comm_buf;
     
-    ///////////////////////////////////////////////////////////////
-    // Conserved current utilities
-    ///////////////////////////////////////////////////////////////
-    void ContractConservedCurrent(PropagatorField &q_in_1,
-                                  PropagatorField &q_in_2,
-                                  PropagatorField &q_out,
-                                  Current curr_type,
-                                  unsigned int mu);
-    void SeqConservedCurrent(PropagatorField &q_in, 
-                             PropagatorField &q_out,
-                             Current curr_type, 
-                             unsigned int mu,
-                             unsigned int tmin, 
+  ///////////////////////////////////////////////////////////////
+  // Conserved current utilities
+  ///////////////////////////////////////////////////////////////
+  void ContractConservedCurrent(PropagatorField &q_in_1,
+				PropagatorField &q_in_2,
+				PropagatorField &q_out,
+				Current curr_type,
+				unsigned int mu);
+  void SeqConservedCurrent(PropagatorField &q_in,
+			   PropagatorField &q_out,
+			   Current curr_type,
+			   unsigned int mu, 
+			   unsigned int tmin,
                              unsigned int tmax,
                  	     ComplexField &lattice_cmplx);
-  };
+};
 
-}}
+NAMESPACE_END(Grid);
 
-#endif
diff --git a/Grid/qcd/action/fermion/MADWF.h b/Grid/qcd/action/fermion/MADWF.h
index 064b13a8..f7f0ee1b 100644
--- a/Grid/qcd/action/fermion/MADWF.h
+++ b/Grid/qcd/action/fermion/MADWF.h
@@ -27,8 +27,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     /*  END LEGAL */
 #pragma once
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 template <class Fieldi, class Fieldo,IfNotSame<Fieldi,Fieldo> X=0>
 inline void convert(const Fieldi &from,Fieldo &to) 
@@ -109,7 +108,7 @@ class MADWF
     std::cout << GridLogMessage << " b    " <<norm2(b)<<std::endl;
 
     defect = b;
-    sol5=zero;
+    sol5=Zero();
     for (int i=0;i<maxiter;i++) {
 
       ///////////////////////////////////////
@@ -122,7 +121,7 @@ class MADWF
       ////////////////////////////////////////////////
       // Solve the inner system with surface term c0
       ////////////////////////////////////////////////
-      ci = zero;  
+      ci = Zero();  
       convert(c0,c0i); // Possible precison change
       InsertSlice(c0i,ci,0, 0);
 
@@ -190,4 +189,4 @@ class MADWF
 
 };
 
-}}
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/MobiusEOFAFermion.cc b/Grid/qcd/action/fermion/MobiusEOFAFermion.cc
deleted file mode 100644
index 0344afbf..00000000
--- a/Grid/qcd/action/fermion/MobiusEOFAFermion.cc
+++ /dev/null
@@ -1,502 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
-
-Copyright (C) 2017
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-Author: David Murphy <dmurphy@phys.columbia.edu>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-#include <Grid/Grid_Eigen_Dense.h>
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
-
-namespace Grid {
-namespace QCD {
-
-  template<class Impl>
-    MobiusEOFAFermion<Impl>::MobiusEOFAFermion(
-      GaugeField            &_Umu,
-      GridCartesian         &FiveDimGrid,
-      GridRedBlackCartesian &FiveDimRedBlackGrid,
-      GridCartesian         &FourDimGrid,
-      GridRedBlackCartesian &FourDimRedBlackGrid,
-      RealD _mq1, RealD _mq2, RealD _mq3,
-      RealD _shift, int _pm, RealD _M5,
-      RealD _b, RealD _c, const ImplParams &p) :
-    AbstractEOFAFermion<Impl>(_Umu, FiveDimGrid, FiveDimRedBlackGrid,
-        FourDimGrid, FourDimRedBlackGrid, _mq1, _mq2, _mq3,
-        _shift, _pm, _M5, _b, _c, p)
-    {
-      int Ls = this->Ls;
-
-      RealD eps = 1.0;
-      Approx::zolotarev_data *zdata = Approx::higham(eps, this->Ls);
-      assert(zdata->n == this->Ls);
-
-      std::cout << GridLogMessage << "MobiusEOFAFermion (b=" << _b <<
-        ",c=" << _c << ") with Ls=" << Ls << std::endl;
-      this->SetCoefficientsTanh(zdata, _b, _c);
-      std::cout << GridLogMessage << "EOFA parameters: (mq1=" << _mq1 <<
-        ",mq2=" << _mq2 << ",mq3=" << _mq3 << ",shift=" << _shift <<
-        ",pm=" << _pm << ")" << std::endl;
-
-      Approx::zolotarev_free(zdata);
-
-      if(_shift != 0.0){
-        SetCoefficientsPrecondShiftOps();
-      } else {
-        Mooee_shift.resize(Ls, 0.0);
-        MooeeInv_shift_lc.resize(Ls, 0.0);
-        MooeeInv_shift_norm.resize(Ls, 0.0);
-        MooeeInvDag_shift_lc.resize(Ls, 0.0);
-        MooeeInvDag_shift_norm.resize(Ls, 0.0);
-      }
-    }
-
-    /****************************************************************
-     * Additional EOFA operators only called outside the inverter.  
-     * Since speed is not essential, simple axpby-style
-     * implementations should be fine.
-     ***************************************************************/
-    template<class Impl>
-    void MobiusEOFAFermion<Impl>::Omega(const FermionField& psi, FermionField& Din, int sign, int dag)
-    {
-      int Ls = this->Ls;
-      RealD alpha = this->alpha;
-
-      Din = zero;
-      if((sign == 1) && (dag == 0)) { // \Omega_{+}
-        for(int s=0; s<Ls; ++s){
-          axpby_ssp(Din, 0.0, psi, 2.0*std::pow(1.0-alpha,Ls-s-1)/std::pow(1.0+alpha,Ls-s), psi, s, 0);
-        }
-      } else if((sign == -1) && (dag == 0)) { // \Omega_{-}
-        for(int s=0; s<Ls; ++s){
-          axpby_ssp(Din, 0.0, psi, 2.0*std::pow(1.0-alpha,s)/std::pow(1.0+alpha,s+1), psi, s, 0);
-        }
-      } else if((sign == 1 ) && (dag == 1)) { // \Omega_{+}^{\dagger}
-        for(int sp=0; sp<Ls; ++sp){
-          axpby_ssp(Din, 1.0, Din, 2.0*std::pow(1.0-alpha,Ls-sp-1)/std::pow(1.0+alpha,Ls-sp), psi, 0, sp);
-        }
-      } else if((sign == -1) && (dag == 1)) { // \Omega_{-}^{\dagger}
-        for(int sp=0; sp<Ls; ++sp){
-          axpby_ssp(Din, 1.0, Din, 2.0*std::pow(1.0-alpha,sp)/std::pow(1.0+alpha,sp+1), psi, 0, sp);
-        }
-      }
-    }
-
-    // This is the operator relating the usual Ddwf to TWQCD's EOFA Dirac operator (arXiv:1706.05843, Eqn. 6).
-    // It also relates the preconditioned and unpreconditioned systems described in Appendix B.2.
-    template<class Impl>
-    void MobiusEOFAFermion<Impl>::Dtilde(const FermionField& psi, FermionField& chi)
-    {
-      int Ls    = this->Ls;
-      RealD b   = 0.5 * ( 1.0 + this->alpha );
-      RealD c   = 0.5 * ( 1.0 - this->alpha );
-      RealD mq1 = this->mq1;
-
-      for(int s=0; s<Ls; ++s){
-        if(s == 0) {
-          axpby_ssp_pminus(chi, b, psi, -c, psi, s, s+1);
-          axpby_ssp_pplus (chi, 1.0, chi, mq1*c, psi, s, Ls-1);
-        } else if(s == (Ls-1)) {
-          axpby_ssp_pminus(chi, b, psi, mq1*c, psi, s, 0);
-          axpby_ssp_pplus (chi, 1.0, chi, -c, psi, s, s-1);
-        } else {
-          axpby_ssp_pminus(chi, b, psi, -c, psi, s, s+1);
-          axpby_ssp_pplus (chi, 1.0, chi, -c, psi, s, s-1);
-        }
-      }
-    }
-
-    template<class Impl>
-    void MobiusEOFAFermion<Impl>::DtildeInv(const FermionField& psi, FermionField& chi)
-    {
-      int Ls = this->Ls;
-      RealD m = this->mq1;
-      RealD c = 0.5 * this->alpha;
-      RealD d = 0.5;
-
-      RealD DtInv_p(0.0), DtInv_m(0.0);
-      RealD N = std::pow(c+d,Ls) + m*std::pow(c-d,Ls);
-      FermionField tmp(this->FermionGrid());
-
-      for(int s=0; s<Ls; ++s){
-      for(int sp=0; sp<Ls; ++sp){
-
-        DtInv_p = m * std::pow(-1.0,s-sp+1) * std::pow(c-d,Ls+s-sp) / std::pow(c+d,s-sp+1) / N;
-        DtInv_p += (s < sp) ? 0.0 : std::pow(-1.0,s-sp) * std::pow(c-d,s-sp) / std::pow(c+d,s-sp+1);
-        DtInv_m = m * std::pow(-1.0,sp-s+1) * std::pow(c-d,Ls+sp-s) / std::pow(c+d,sp-s+1) / N;
-        DtInv_m += (s > sp) ? 0.0 : std::pow(-1.0,sp-s) * std::pow(c-d,sp-s) / std::pow(c+d,sp-s+1);
-
-        if(sp == 0){
-          axpby_ssp_pplus (tmp, 0.0, tmp, DtInv_p, psi, s, sp);
-          axpby_ssp_pminus(tmp, 0.0, tmp, DtInv_m, psi, s, sp);
-        } else {
-          axpby_ssp_pplus (tmp, 1.0, tmp, DtInv_p, psi, s, sp);
-          axpby_ssp_pminus(tmp, 1.0, tmp, DtInv_m, psi, s, sp);
-        }
-
-      }}
-    }
-
-    /*****************************************************************************************************/
-
-    template<class Impl>
-    RealD MobiusEOFAFermion<Impl>::M(const FermionField& psi, FermionField& chi)
-    {
-      int Ls = this->Ls;
-
-      FermionField Din(psi._grid);
-
-      this->Meooe5D(psi, Din);
-      this->DW(Din, chi, DaggerNo);
-      axpby(chi, 1.0, 1.0, chi, psi);
-      this->M5D(psi, chi);
-      return(norm2(chi));
-    }
-
-    template<class Impl>
-    RealD MobiusEOFAFermion<Impl>::Mdag(const FermionField& psi, FermionField& chi)
-    {
-      int Ls = this->Ls;
-
-      FermionField Din(psi._grid);
-
-      this->DW(psi, Din, DaggerYes);
-      this->MeooeDag5D(Din, chi);
-      this->M5Ddag(psi, chi);
-      axpby(chi, 1.0, 1.0, chi, psi);
-      return(norm2(chi));
-    }
-
-    /********************************************************************
-     * Performance critical fermion operators called inside the inverter
-     ********************************************************************/
-
-    template<class Impl>
-    void MobiusEOFAFermion<Impl>::M5D(const FermionField& psi, FermionField& chi)
-    {
-      int Ls = this->Ls;
-
-      std::vector<Coeff_t> diag(Ls,1.0);
-      std::vector<Coeff_t> upper(Ls,-1.0);  upper[Ls-1] = this->mq1;
-      std::vector<Coeff_t> lower(Ls,-1.0);  lower[0]    = this->mq1;
-
-      // no shift term
-      if(this->shift == 0.0){ this->M5D(psi, chi, chi, lower, diag, upper); }
-
-      // fused M + shift operation
-      else{ this->M5D_shift(psi, chi, chi, lower, diag, upper, Mooee_shift); }
-    }
-
-    template<class Impl>
-    void MobiusEOFAFermion<Impl>::M5Ddag(const FermionField& psi, FermionField& chi)
-    {
-      int Ls = this->Ls;
-
-      std::vector<Coeff_t> diag(Ls,1.0);
-      std::vector<Coeff_t> upper(Ls,-1.0);  upper[Ls-1] = this->mq1;
-      std::vector<Coeff_t> lower(Ls,-1.0);  lower[0]    = this->mq1;
-
-      // no shift term
-      if(this->shift == 0.0){ this->M5Ddag(psi, chi, chi, lower, diag, upper); }
-
-      // fused M + shift operation
-      else{ this->M5Ddag_shift(psi, chi, chi, lower, diag, upper, Mooee_shift); }
-    }
-
-    // half checkerboard operations
-    template<class Impl>
-    void MobiusEOFAFermion<Impl>::Mooee(const FermionField& psi, FermionField& chi)
-    {
-      int Ls = this->Ls;
-
-      // coefficients of Mooee
-      std::vector<Coeff_t> diag = this->bee;
-      std::vector<Coeff_t> upper(Ls);
-      std::vector<Coeff_t> lower(Ls);
-      for(int s=0; s<Ls; s++){
-        upper[s] = -this->cee[s];
-        lower[s] = -this->cee[s];
-      }
-      upper[Ls-1] *= -this->mq1;
-      lower[0]    *= -this->mq1;
-
-      // no shift term
-      if(this->shift == 0.0){ this->M5D(psi, psi, chi, lower, diag, upper); }
-
-      // fused M + shift operation
-      else { this->M5D_shift(psi, psi, chi, lower, diag, upper, Mooee_shift); }
-    }
-
-    template<class Impl>
-    void MobiusEOFAFermion<Impl>::MooeeDag(const FermionField& psi, FermionField& chi)
-    {
-      int Ls = this->Ls;
-
-      // coefficients of MooeeDag
-      std::vector<Coeff_t> diag = this->bee;
-      std::vector<Coeff_t> upper(Ls);
-      std::vector<Coeff_t> lower(Ls);
-      for(int s=0; s<Ls; s++){
-        if(s==0) {
-          upper[s] = -this->cee[s+1];
-          lower[s] = this->mq1*this->cee[Ls-1];
-        } else if(s==(Ls-1)) {
-          upper[s] = this->mq1*this->cee[0];
-          lower[s] = -this->cee[s-1];
-        } else {
-          upper[s] = -this->cee[s+1];
-          lower[s] = -this->cee[s-1];
-        }
-      }
-
-      // no shift term
-      if(this->shift == 0.0){ this->M5Ddag(psi, psi, chi, lower, diag, upper); }
-
-      // fused M + shift operation
-      else{ this->M5Ddag_shift(psi, psi, chi, lower, diag, upper, Mooee_shift); }
-    }
-
-    /****************************************************************************************/
-
-    // Computes coefficients for applying Cayley preconditioned shift operators
-    //  (Mooee + \Delta) --> Mooee_shift
-    //  (Mooee + \Delta)^{-1} --> MooeeInv_shift_lc, MooeeInv_shift_norm
-    //  (Mooee + \Delta)^{-dag} --> MooeeInvDag_shift_lc, MooeeInvDag_shift_norm
-    // For the latter two cases, the operation takes the form
-    //  [ (Mooee + \Delta)^{-1} \psi ]_{i} = Mooee_{ij} \psi_{j} +
-    //      ( MooeeInv_shift_norm )_{i} ( \sum_{j} [ MooeeInv_shift_lc ]_{j} P_{pm} \psi_{j} )
-    template<class Impl>
-    void MobiusEOFAFermion<Impl>::SetCoefficientsPrecondShiftOps()
-    {
-      int   Ls    = this->Ls;
-      int   pm    = this->pm;
-      RealD alpha = this->alpha;
-      RealD k     = this->k;
-      RealD mq1   = this->mq1;
-      RealD shift = this->shift;
-
-      // Initialize
-      Mooee_shift.resize(Ls);
-      MooeeInv_shift_lc.resize(Ls);
-      MooeeInv_shift_norm.resize(Ls);
-      MooeeInvDag_shift_lc.resize(Ls);
-      MooeeInvDag_shift_norm.resize(Ls);
-
-      // Construct Mooee_shift
-      int idx(0);
-      Coeff_t N = ( (pm == 1) ? 1.0 : -1.0 ) * (2.0*shift*k) *
-                  ( std::pow(alpha+1.0,Ls) + mq1*std::pow(alpha-1.0,Ls) );
-      for(int s=0; s<Ls; ++s){
-        idx = (pm == 1) ? (s) : (Ls-1-s);
-        Mooee_shift[idx] = N * std::pow(-1.0,s) * std::pow(alpha-1.0,s) / std::pow(alpha+1.0,Ls+s+1);
-      }
-
-      // Tridiagonal solve for MooeeInvDag_shift_lc
-      {
-        Coeff_t m(0.0);
-        std::vector<Coeff_t> d = Mooee_shift;
-        std::vector<Coeff_t> u(Ls,0.0);
-        std::vector<Coeff_t> y(Ls,0.0);
-        std::vector<Coeff_t> q(Ls,0.0);
-        if(pm == 1){ u[0] = 1.0; }
-        else{ u[Ls-1] = 1.0; }
-
-        // Tridiagonal matrix algorithm + Sherman-Morrison formula
-        //
-        // We solve
-        //  ( Mooee' + u \otimes v ) MooeeInvDag_shift_lc = Mooee_shift
-        // where Mooee' is the tridiagonal part of Mooee_{+}, and
-        // u = (1,0,...,0) and v = (0,...,0,mq1*cee[0]) are chosen
-        // so that the outer-product u \otimes v gives the (0,Ls-1)
-        // entry of Mooee_{+}.
-        //
-        // We do this as two solves: Mooee'*y = d and Mooee'*q = u,
-        // and then construct the solution to the original system
-        //  MooeeInvDag_shift_lc = y - <v,y> / ( 1 + <v,q> ) q
-        if(pm == 1){
-          for(int s=1; s<Ls; ++s){
-            m = -this->cee[s] / this->bee[s-1];
-            d[s] -= m*d[s-1];
-            u[s] -= m*u[s-1];
-          }
-        }
-        y[Ls-1] = d[Ls-1] / this->bee[Ls-1];
-        q[Ls-1] = u[Ls-1] / this->bee[Ls-1];
-        for(int s=Ls-2; s>=0; --s){
-          if(pm == 1){
-            y[s] = d[s] / this->bee[s];
-            q[s] = u[s] / this->bee[s];
-          } else {
-            y[s] = ( d[s] + this->cee[s]*y[s+1] ) / this->bee[s];
-            q[s] = ( u[s] + this->cee[s]*q[s+1] ) / this->bee[s];
-          }
-        }
-
-        // Construct MooeeInvDag_shift_lc
-        for(int s=0; s<Ls; ++s){
-          if(pm == 1){
-            MooeeInvDag_shift_lc[s] = y[s] - mq1*this->cee[0]*y[Ls-1] /
-              (1.0+mq1*this->cee[0]*q[Ls-1]) * q[s];
-          } else {
-            MooeeInvDag_shift_lc[s] = y[s] - mq1*this->cee[Ls-1]*y[0] /
-              (1.0+mq1*this->cee[Ls-1]*q[0]) * q[s];
-          }
-        }
-
-        // Compute remaining coefficients
-        N = (pm == 1) ? (1.0 + MooeeInvDag_shift_lc[Ls-1]) : (1.0 + MooeeInvDag_shift_lc[0]);
-        for(int s=0; s<Ls; ++s){
-
-          // MooeeInv_shift_lc
-          if(pm == 1){ MooeeInv_shift_lc[s] = std::pow(this->bee[s],s) * std::pow(this->cee[s],Ls-1-s); }
-          else{ MooeeInv_shift_lc[s] = std::pow(this->bee[s],Ls-1-s) * std::pow(this->cee[s],s); }
-
-          // MooeeInv_shift_norm
-          MooeeInv_shift_norm[s] = -MooeeInvDag_shift_lc[s] /
-            ( std::pow(this->bee[s],Ls) + mq1*std::pow(this->cee[s],Ls) ) / N;
-
-          // MooeeInvDag_shift_norm
-          if(pm == 1){ MooeeInvDag_shift_norm[s] = -std::pow(this->bee[s],s) * std::pow(this->cee[s],Ls-1-s) /
-            ( std::pow(this->bee[s],Ls) + mq1*std::pow(this->cee[s],Ls) ) / N; }
-          else{ MooeeInvDag_shift_norm[s] = -std::pow(this->bee[s],Ls-1-s) * std::pow(this->cee[s],s) /
-            ( std::pow(this->bee[s],Ls) + mq1*std::pow(this->cee[s],Ls) ) / N; }
-        }
-      }
-    }
-
-    // Recompute coefficients for a different value of shift constant
-    template<class Impl>
-    void MobiusEOFAFermion<Impl>::RefreshShiftCoefficients(RealD new_shift)
-    {
-      this->shift = new_shift;
-      if(new_shift != 0.0){
-        SetCoefficientsPrecondShiftOps();
-      } else {
-        int Ls = this->Ls;
-        Mooee_shift.resize(Ls,0.0);
-        MooeeInv_shift_lc.resize(Ls,0.0);
-        MooeeInv_shift_norm.resize(Ls,0.0);
-        MooeeInvDag_shift_lc.resize(Ls,0.0);
-        MooeeInvDag_shift_norm.resize(Ls,0.0);
-      }
-    }
-
-    template<class Impl>
-    void MobiusEOFAFermion<Impl>::MooeeInternalCompute(int dag, int inv,
-      Vector<iSinglet<Simd> >& Matp, Vector<iSinglet<Simd> >& Matm)
-    {
-      int Ls = this->Ls;
-
-      GridBase* grid = this->FermionRedBlackGrid();
-      int LLs = grid->_rdimensions[0];
-
-      if(LLs == Ls){ return; } // Not vectorised in 5th direction
-
-      Eigen::MatrixXcd Pplus  = Eigen::MatrixXcd::Zero(Ls,Ls);
-      Eigen::MatrixXcd Pminus = Eigen::MatrixXcd::Zero(Ls,Ls);
-
-      for(int s=0; s<Ls; s++){
-        Pplus(s,s)  = this->bee[s];
-        Pminus(s,s) = this->bee[s];
-      }
-
-      for(int s=0; s<Ls-1; s++){
-        Pminus(s,s+1) = -this->cee[s];
-        Pplus(s+1,s) = -this->cee[s+1];
-      }
-
-      Pplus (0,Ls-1) = this->mq1*this->cee[0];
-      Pminus(Ls-1,0) = this->mq1*this->cee[Ls-1];
-
-      if(this->shift != 0.0){
-        RealD c = 0.5 * this->alpha;
-        RealD d = 0.5;
-        RealD N = this->shift * this->k * ( std::pow(c+d,Ls) + this->mq1*std::pow(c-d,Ls) );
-        if(this->pm == 1) {
-          for(int s=0; s<Ls; ++s){
-            Pplus(s,Ls-1) += N * std::pow(-1.0,s) * std::pow(c-d,s) / std::pow(c+d,Ls+s+1);
-          }
-        } else {
-          for(int s=0; s<Ls; ++s){
-            Pminus(s,0) += N * std::pow(-1.0,s+1) * std::pow(c-d,Ls-1-s) / std::pow(c+d,2*Ls-s);
-          }
-        }
-      }
-
-      Eigen::MatrixXcd PplusMat ;
-      Eigen::MatrixXcd PminusMat;
-
-      if(inv) {
-        PplusMat  = Pplus.inverse();
-        PminusMat = Pminus.inverse();
-      } else {
-        PplusMat  = Pplus;
-        PminusMat = Pminus;
-      }
-
-      if(dag){
-        PplusMat.adjointInPlace();
-        PminusMat.adjointInPlace();
-      }
-
-      typedef typename SiteHalfSpinor::scalar_type scalar_type;
-      const int Nsimd = Simd::Nsimd();
-      Matp.resize(Ls*LLs);
-      Matm.resize(Ls*LLs);
-
-      for(int s2=0; s2<Ls; s2++){
-      for(int s1=0; s1<LLs; s1++){
-        int istride = LLs;
-        int ostride = 1;
-        Simd Vp;
-        Simd Vm;
-        scalar_type *sp = (scalar_type*) &Vp;
-        scalar_type *sm = (scalar_type*) &Vm;
-        for(int l=0; l<Nsimd; l++){
-          if(switcheroo<Coeff_t>::iscomplex()) {
-            sp[l] = PplusMat (l*istride+s1*ostride,s2);
-            sm[l] = PminusMat(l*istride+s1*ostride,s2);
-          } else {
-            // if real
-            scalar_type tmp;
-            tmp = PplusMat (l*istride+s1*ostride,s2);
-            sp[l] = scalar_type(tmp.real(),tmp.real());
-            tmp = PminusMat(l*istride+s1*ostride,s2);
-            sm[l] = scalar_type(tmp.real(),tmp.real());
-          }
-        }
-        Matp[LLs*s2+s1] = Vp;
-        Matm[LLs*s2+s1] = Vm;
-      }}
-  }
-
-  FermOpTemplateInstantiate(MobiusEOFAFermion);
-  GparityFermOpTemplateInstantiate(MobiusEOFAFermion);
-
-}}
diff --git a/Grid/qcd/action/fermion/MobiusEOFAFermion.h b/Grid/qcd/action/fermion/MobiusEOFAFermion.h
index 519b49e7..6b214233 100644
--- a/Grid/qcd/action/fermion/MobiusEOFAFermion.h
+++ b/Grid/qcd/action/fermion/MobiusEOFAFermion.h
@@ -26,108 +26,79 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 
 See the full license in the file "LICENSE" in the top level distribution directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef  GRID_QCD_MOBIUS_EOFA_FERMION_H
 #define  GRID_QCD_MOBIUS_EOFA_FERMION_H
 
 #include <Grid/qcd/action/fermion/AbstractEOFAFermion.h>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
-  template<class Impl>
-  class MobiusEOFAFermion : public AbstractEOFAFermion<Impl>
-  {
-    public:
-      INHERIT_IMPL_TYPES(Impl);
+template<class Impl>
+class MobiusEOFAFermion : public AbstractEOFAFermion<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
 
-    public:
-      // Shift operator coefficients for red-black preconditioned Mobius EOFA
-      std::vector<Coeff_t> Mooee_shift;
-      std::vector<Coeff_t> MooeeInv_shift_lc;
-      std::vector<Coeff_t> MooeeInv_shift_norm;
-      std::vector<Coeff_t> MooeeInvDag_shift_lc;
-      std::vector<Coeff_t> MooeeInvDag_shift_norm;
+public:
+  // Shift operator coefficients for red-black preconditioned Mobius EOFA
+  Vector<Coeff_t> Mooee_shift;
+  Vector<Coeff_t> MooeeInv_shift_lc;
+  Vector<Coeff_t> MooeeInv_shift_norm;
+  Vector<Coeff_t> MooeeInvDag_shift_lc;
+  Vector<Coeff_t> MooeeInvDag_shift_norm;
 
-      virtual void Instantiatable(void) {};
+  virtual void Instantiatable(void) {};
 
-      // EOFA-specific operations
-      virtual void  Omega            (const FermionField& in, FermionField& out, int sign, int dag);
-      virtual void  Dtilde           (const FermionField& in, FermionField& out);
-      virtual void  DtildeInv        (const FermionField& in, FermionField& out);
+  // EOFA-specific operations
+  virtual void  Omega            (const FermionField& in, FermionField& out, int sign, int dag);
+  virtual void  Dtilde           (const FermionField& in, FermionField& out);
+  virtual void  DtildeInv        (const FermionField& in, FermionField& out);
 
-      // override multiply
-      virtual RealD M                (const FermionField& in, FermionField& out);
-      virtual RealD Mdag             (const FermionField& in, FermionField& out);
+  // override multiply
+  virtual RealD M                (const FermionField& in, FermionField& out);
+  virtual RealD Mdag             (const FermionField& in, FermionField& out);
 
-      // half checkerboard operations
-      virtual void  Mooee            (const FermionField& in, FermionField& out);
-      virtual void  MooeeDag         (const FermionField& in, FermionField& out);
-      virtual void  MooeeInv         (const FermionField& in, FermionField& out);
-      virtual void  MooeeInv_shift   (const FermionField& in, FermionField& out);
-      virtual void  MooeeInvDag      (const FermionField& in, FermionField& out);
-      virtual void  MooeeInvDag_shift(const FermionField& in, FermionField& out);
+  // half checkerboard operations
+  virtual void  Mooee            (const FermionField& in, FermionField& out);
+  virtual void  MooeeDag         (const FermionField& in, FermionField& out);
+  virtual void  MooeeInv         (const FermionField& in, FermionField& out);
+  virtual void  MooeeInv_shift   (const FermionField& in, FermionField& out);
+  virtual void  MooeeInvDag      (const FermionField& in, FermionField& out);
+  virtual void  MooeeInvDag_shift(const FermionField& in, FermionField& out);
 
-      virtual void   M5D             (const FermionField& psi, FermionField& chi);
-      virtual void   M5Ddag          (const FermionField& psi, FermionField& chi);
+  virtual void   M5D             (const FermionField& psi, FermionField& chi);
+  virtual void   M5Ddag          (const FermionField& psi, FermionField& chi);
 
-      /////////////////////////////////////////////////////
-      // Instantiate different versions depending on Impl
-      /////////////////////////////////////////////////////
-      void M5D(const FermionField& psi, const FermionField& phi, FermionField& chi,
-        std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper);
+  /////////////////////////////////////////////////////
+  // Instantiate different versions depending on Impl
+  /////////////////////////////////////////////////////
+  void M5D(const FermionField& psi, const FermionField& phi, FermionField& chi,
+	   Vector<Coeff_t>& lower, Vector<Coeff_t>& diag, Vector<Coeff_t>& upper);
 
-      void M5D_shift(const FermionField& psi, const FermionField& phi, FermionField& chi,
-        std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper,
-        std::vector<Coeff_t>& shift_coeffs);
+  void M5D_shift(const FermionField& psi, const FermionField& phi, FermionField& chi,
+		 Vector<Coeff_t>& lower, Vector<Coeff_t>& diag, Vector<Coeff_t>& upper,
+		 Vector<Coeff_t>& shift_coeffs);
 
-      void M5Ddag(const FermionField& psi, const FermionField& phi, FermionField& chi,
-        std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper);
+  void M5Ddag(const FermionField& psi, const FermionField& phi, FermionField& chi,
+	      Vector<Coeff_t>& lower, Vector<Coeff_t>& diag, Vector<Coeff_t>& upper);
 
-      void M5Ddag_shift(const FermionField& psi, const FermionField& phi, FermionField& chi,
-        std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper,
-        std::vector<Coeff_t>& shift_coeffs);
+  void M5Ddag_shift(const FermionField& psi, const FermionField& phi, FermionField& chi,
+		    Vector<Coeff_t>& lower, Vector<Coeff_t>& diag, Vector<Coeff_t>& upper,
+		    Vector<Coeff_t>& shift_coeffs);
 
-      void MooeeInternal(const FermionField& in, FermionField& out, int dag, int inv);
+  virtual void RefreshShiftCoefficients(RealD new_shift);
 
-      void MooeeInternalCompute(int dag, int inv, Vector<iSinglet<Simd>>& Matp, Vector<iSinglet<Simd>>& Matm);
+  // Constructors
+  MobiusEOFAFermion(GaugeField& _Umu, GridCartesian& FiveDimGrid, GridRedBlackCartesian& FiveDimRedBlackGrid,
+		    GridCartesian& FourDimGrid, GridRedBlackCartesian& FourDimRedBlackGrid,
+		    RealD _mq1, RealD _mq2, RealD _mq3, RealD _shift, int pm,
+		    RealD _M5, RealD _b, RealD _c, const ImplParams& p=ImplParams());
 
-      void MooeeInternalAsm(const FermionField& in, FermionField& out, int LLs, int site,
-        Vector<iSinglet<Simd>>& Matp, Vector<iSinglet<Simd>>& Matm);
+protected:
+  void SetCoefficientsPrecondShiftOps(void);
+};
 
-      void MooeeInternalZAsm(const FermionField& in, FermionField& out, int LLs, int site,
-        Vector<iSinglet<Simd>>& Matp, Vector<iSinglet<Simd>>& Matm);
-
-      virtual void RefreshShiftCoefficients(RealD new_shift);
-
-      // Constructors
-      MobiusEOFAFermion(GaugeField& _Umu, GridCartesian& FiveDimGrid, GridRedBlackCartesian& FiveDimRedBlackGrid,
-        GridCartesian& FourDimGrid, GridRedBlackCartesian& FourDimRedBlackGrid,
-        RealD _mq1, RealD _mq2, RealD _mq3, RealD _shift, int pm,
-        RealD _M5, RealD _b, RealD _c, const ImplParams& p=ImplParams());
-
-    protected:
-      void SetCoefficientsPrecondShiftOps(void);
-  };
-}}
-
-#define INSTANTIATE_DPERP_MOBIUS_EOFA(A)\
-template void MobiusEOFAFermion<A>::M5D(const FermionField& psi, const FermionField& phi, FermionField& chi, \
-  std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper); \
-template void MobiusEOFAFermion<A>::M5D_shift(const FermionField& psi, const FermionField& phi, FermionField& chi, \
-  std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper, std::vector<Coeff_t>& shift_coeffs); \
-template void MobiusEOFAFermion<A>::M5Ddag(const FermionField& psi, const FermionField& phi, FermionField& chi, \
-  std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper); \
-template void MobiusEOFAFermion<A>::M5Ddag_shift(const FermionField& psi, const FermionField& phi, FermionField& chi, \
-  std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper, std::vector<Coeff_t>& shift_coeffs); \
-template void MobiusEOFAFermion<A>::MooeeInv(const FermionField& psi, FermionField& chi); \
-template void MobiusEOFAFermion<A>::MooeeInv_shift(const FermionField& psi, FermionField& chi); \
-template void MobiusEOFAFermion<A>::MooeeInvDag(const FermionField& psi, FermionField& chi); \
-template void MobiusEOFAFermion<A>::MooeeInvDag_shift(const FermionField& psi, FermionField& chi);
-
-#undef  MOBIUS_EOFA_DPERP_DENSE
-#define MOBIUS_EOFA_DPERP_CACHE
-#undef  MOBIUS_EOFA_DPERP_LINALG
-#define MOBIUS_EOFA_DPERP_VEC
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/fermion/MobiusEOFAFermioncache.cc b/Grid/qcd/action/fermion/MobiusEOFAFermioncache.cc
deleted file mode 100644
index 420f6390..00000000
--- a/Grid/qcd/action/fermion/MobiusEOFAFermioncache.cc
+++ /dev/null
@@ -1,429 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/qcd/action/fermion/MobiusEOFAFermioncache.cc
-
-Copyright (C) 2017
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-Author: David Murphy <dmurphy@phys.columbia.edu>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
-
-namespace Grid {
-namespace QCD {
-
-  // FIXME -- make a version of these routines with site loop outermost for cache reuse.
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::M5D(const FermionField &psi, const FermionField &phi, FermionField &chi,
-    std::vector<Coeff_t> &lower, std::vector<Coeff_t> &diag, std::vector<Coeff_t> &upper)
-  {
-    int Ls = this->Ls;
-    GridBase *grid = psi._grid;
-
-    assert(phi.checkerboard == psi.checkerboard);
-    chi.checkerboard = psi.checkerboard;
-
-    // Flops = 6.0*(Nc*Ns) *Ls*vol
-    this->M5Dcalls++;
-    this->M5Dtime -= usecond();
-
-    parallel_for(int ss=0; ss<grid->oSites(); ss+=Ls){
-      for(int s=0; s<Ls; s++){
-        auto tmp = psi._odata[0];
-        if(s==0){
-          spProj5m(tmp, psi._odata[ss+s+1]);
-          chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-          spProj5p(tmp, psi._odata[ss+Ls-1]);
-          chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-        } else if(s==(Ls-1)) {
-          spProj5m(tmp, psi._odata[ss+0]);
-          chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-          spProj5p(tmp, psi._odata[ss+s-1]);
-          chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-        } else {
-          spProj5m(tmp, psi._odata[ss+s+1]);
-          chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-          spProj5p(tmp, psi._odata[ss+s-1]);
-          chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-        }
-      }
-    }
-
-    this->M5Dtime += usecond();
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::M5D_shift(const FermionField &psi, const FermionField &phi, FermionField &chi,
-    std::vector<Coeff_t> &lower, std::vector<Coeff_t> &diag, std::vector<Coeff_t> &upper,
-    std::vector<Coeff_t> &shift_coeffs)
-  {
-    int Ls = this->Ls;
-    int shift_s = (this->pm == 1) ? (Ls-1) : 0; // s-component modified by shift operator
-    GridBase *grid = psi._grid;
-
-    assert(phi.checkerboard == psi.checkerboard);
-    chi.checkerboard = psi.checkerboard;
-
-    // Flops = 6.0*(Nc*Ns) *Ls*vol
-    this->M5Dcalls++;
-    this->M5Dtime -= usecond();
-
-    parallel_for(int ss=0; ss<grid->oSites(); ss+=Ls){
-      for(int s=0; s<Ls; s++){
-        auto tmp = psi._odata[0];
-        if(s==0){
-          spProj5m(tmp, psi._odata[ss+s+1]);
-          chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-          spProj5p(tmp, psi._odata[ss+Ls-1]);
-          chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-        } else if(s==(Ls-1)) {
-          spProj5m(tmp, psi._odata[ss+0]);
-          chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-          spProj5p(tmp, psi._odata[ss+s-1]);
-          chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-        } else {
-          spProj5m(tmp, psi._odata[ss+s+1]);
-          chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-          spProj5p(tmp, psi._odata[ss+s-1]);
-          chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-        }
-        if(this->pm == 1){ spProj5p(tmp, psi._odata[ss+shift_s]); }
-        else{ spProj5m(tmp, psi._odata[ss+shift_s]); }
-        chi[ss+s] = chi[ss+s] + shift_coeffs[s]*tmp;
-      }
-    }
-
-    this->M5Dtime += usecond();
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::M5Ddag(const FermionField &psi, const FermionField &phi, FermionField &chi,
-    std::vector<Coeff_t> &lower, std::vector<Coeff_t> &diag, std::vector<Coeff_t> &upper)
-  {
-    int Ls = this->Ls;
-    GridBase *grid = psi._grid;
-
-    assert(phi.checkerboard == psi.checkerboard);
-    chi.checkerboard = psi.checkerboard;
-
-    // Flops = 6.0*(Nc*Ns) *Ls*vol
-    this->M5Dcalls++;
-    this->M5Dtime -= usecond();
-
-    parallel_for(int ss=0; ss<grid->oSites(); ss+=Ls){
-      auto tmp = psi._odata[0];
-      for(int s=0; s<Ls; s++){
-        if(s==0) {
-          spProj5p(tmp, psi._odata[ss+s+1]);
-          chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-          spProj5m(tmp, psi._odata[ss+Ls-1]);
-          chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-        } else if(s==(Ls-1)) {
-          spProj5p(tmp, psi._odata[ss+0]);
-          chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-          spProj5m(tmp, psi._odata[ss+s-1]);
-          chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-        } else {
-          spProj5p(tmp, psi._odata[ss+s+1]);
-          chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-          spProj5m(tmp, psi._odata[ss+s-1]);
-          chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-        }
-      }
-    }
-
-    this->M5Dtime += usecond();
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::M5Ddag_shift(const FermionField &psi, const FermionField &phi, FermionField &chi,
-    std::vector<Coeff_t> &lower, std::vector<Coeff_t> &diag, std::vector<Coeff_t> &upper,
-    std::vector<Coeff_t> &shift_coeffs)
-  {
-    int Ls = this->Ls;
-    int shift_s = (this->pm == 1) ? (Ls-1) : 0; // s-component modified by shift operator
-    GridBase *grid = psi._grid;
-
-    assert(phi.checkerboard == psi.checkerboard);
-    chi.checkerboard = psi.checkerboard;
-
-    // Flops = 6.0*(Nc*Ns) *Ls*vol
-    this->M5Dcalls++;
-    this->M5Dtime -= usecond();
-
-    parallel_for(int ss=0; ss<grid->oSites(); ss+=Ls){
-      chi[ss+Ls-1] = zero;
-      auto tmp = psi._odata[0];
-      for(int s=0; s<Ls; s++){
-        if(s==0) {
-          spProj5p(tmp, psi._odata[ss+s+1]);
-          chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-          spProj5m(tmp, psi._odata[ss+Ls-1]);
-          chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-        } else if(s==(Ls-1)) {
-          spProj5p(tmp, psi._odata[ss+0]);
-          chi[ss+s] = chi[ss+s] + diag[s]*phi[ss+s] + upper[s]*tmp;
-          spProj5m(tmp, psi._odata[ss+s-1]);
-          chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-        } else {
-          spProj5p(tmp, psi._odata[ss+s+1]);
-          chi[ss+s] = diag[s]*phi[ss+s] + upper[s]*tmp;
-          spProj5m(tmp, psi._odata[ss+s-1]);
-          chi[ss+s] = chi[ss+s] + lower[s]*tmp;
-        }
-        if(this->pm == 1){ spProj5p(tmp, psi._odata[ss+s]); }
-        else{ spProj5m(tmp, psi._odata[ss+s]); }
-        chi[ss+shift_s] = chi[ss+shift_s] + shift_coeffs[s]*tmp;
-      }
-    }
-
-    this->M5Dtime += usecond();
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInv(const FermionField &psi, FermionField &chi)
-  {
-    if(this->shift != 0.0){ MooeeInv_shift(psi,chi); return; }
-
-    GridBase *grid = psi._grid;
-    int Ls = this->Ls;
-
-    chi.checkerboard = psi.checkerboard;
-
-    this->MooeeInvCalls++;
-    this->MooeeInvTime -= usecond();
-
-    parallel_for(int ss=0; ss<grid->oSites(); ss+=Ls){
-
-      auto tmp = psi._odata[0];
-
-      // Apply (L^{\prime})^{-1}
-      chi[ss] = psi[ss]; // chi[0]=psi[0]
-      for(int s=1; s<Ls; s++){
-        spProj5p(tmp, chi[ss+s-1]);
-        chi[ss+s] = psi[ss+s] - this->lee[s-1]*tmp;
-      }
-
-      // L_m^{-1}
-      for(int s=0; s<Ls-1; s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
-        spProj5m(tmp, chi[ss+s]);
-        chi[ss+Ls-1] = chi[ss+Ls-1] - this->leem[s]*tmp;
-      }
-
-      // U_m^{-1} D^{-1}
-      for(int s=0; s<Ls-1; s++){ // Chi[s] + 1/d chi[s]
-        spProj5p(tmp, chi[ss+Ls-1]);
-        chi[ss+s] = (1.0/this->dee[s])*chi[ss+s] - (this->ueem[s]/this->dee[Ls-1])*tmp;
-      }
-      chi[ss+Ls-1] = (1.0/this->dee[Ls-1])*chi[ss+Ls-1];
-
-      // Apply U^{-1}
-      for(int s=Ls-2; s>=0; s--){
-        spProj5m(tmp, chi[ss+s+1]);
-        chi[ss+s] = chi[ss+s] - this->uee[s]*tmp;
-      }
-    }
-
-    this->MooeeInvTime += usecond();
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInv_shift(const FermionField &psi, FermionField &chi)
-  {
-    GridBase *grid = psi._grid;
-    int Ls = this->Ls;
-
-    chi.checkerboard = psi.checkerboard;
-
-    this->MooeeInvCalls++;
-    this->MooeeInvTime -= usecond();
-
-    parallel_for(int ss=0; ss<grid->oSites(); ss+=Ls){
-
-      auto tmp1        = psi._odata[0];
-      auto tmp2        = psi._odata[0];
-      auto tmp2_spProj = psi._odata[0];
-
-      // Apply (L^{\prime})^{-1} and accumulate MooeeInv_shift_lc[j]*psi[j] in tmp2
-      chi[ss] = psi[ss]; // chi[0]=psi[0]
-      tmp2 = MooeeInv_shift_lc[0]*psi[ss];
-      for(int s=1; s<Ls; s++){
-        spProj5p(tmp1, chi[ss+s-1]);
-        chi[ss+s] = psi[ss+s] - this->lee[s-1]*tmp1;
-        tmp2 = tmp2 + MooeeInv_shift_lc[s]*psi[ss+s];
-      }
-      if(this->pm == 1){ spProj5p(tmp2_spProj, tmp2);}
-      else{ spProj5m(tmp2_spProj, tmp2); }
-
-      // L_m^{-1}
-      for(int s=0; s<Ls-1; s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
-        spProj5m(tmp1, chi[ss+s]);
-        chi[ss+Ls-1] = chi[ss+Ls-1] - this->leem[s]*tmp1;
-      }
-
-      // U_m^{-1} D^{-1}
-      for(int s=0; s<Ls-1; s++){ // Chi[s] + 1/d chi[s]
-        spProj5p(tmp1, chi[ss+Ls-1]);
-        chi[ss+s] = (1.0/this->dee[s])*chi[ss+s] - (this->ueem[s]/this->dee[Ls-1])*tmp1;
-      }
-      // chi[ss+Ls-1] = (1.0/this->dee[Ls-1])*chi[ss+Ls-1] + MooeeInv_shift_norm[Ls-1]*tmp2_spProj;
-      chi[ss+Ls-1] = (1.0/this->dee[Ls-1])*chi[ss+Ls-1];
-      spProj5m(tmp1, chi[ss+Ls-1]);
-      chi[ss+Ls-1] = chi[ss+Ls-1] + MooeeInv_shift_norm[Ls-1]*tmp2_spProj;
-
-      // Apply U^{-1} and add shift term
-      for(int s=Ls-2; s>=0; s--){
-        chi[ss+s] = chi[ss+s] - this->uee[s]*tmp1;
-        spProj5m(tmp1, chi[ss+s]);
-        chi[ss+s] = chi[ss+s] + MooeeInv_shift_norm[s]*tmp2_spProj;
-      }
-    }
-
-    this->MooeeInvTime += usecond();
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInvDag(const FermionField &psi, FermionField &chi)
-  {
-    if(this->shift != 0.0){ MooeeInvDag_shift(psi,chi); return; }
-
-    GridBase *grid = psi._grid;
-    int Ls = this->Ls;
-
-    chi.checkerboard = psi.checkerboard;
-
-    this->MooeeInvCalls++;
-    this->MooeeInvTime -= usecond();
-
-    parallel_for(int ss=0; ss<grid->oSites(); ss+=Ls){
-
-      auto tmp = psi._odata[0];
-
-      // Apply (U^{\prime})^{-dag}
-      chi[ss] = psi[ss];
-      for(int s=1; s<Ls; s++){
-        spProj5m(tmp, chi[ss+s-1]);
-        chi[ss+s] = psi[ss+s] - this->uee[s-1]*tmp;
-      }
-
-      // U_m^{-\dag}
-      for(int s=0; s<Ls-1; s++){
-        spProj5p(tmp, chi[ss+s]);
-        chi[ss+Ls-1] = chi[ss+Ls-1] - this->ueem[s]*tmp;
-      }
-
-      // L_m^{-\dag} D^{-dag}
-      for(int s=0; s<Ls-1; s++){
-        spProj5m(tmp, chi[ss+Ls-1]);
-        chi[ss+s] = (1.0/this->dee[s])*chi[ss+s] - (this->leem[s]/this->dee[Ls-1])*tmp;
-      }
-      chi[ss+Ls-1] = (1.0/this->dee[Ls-1])*chi[ss+Ls-1];
-
-      // Apply L^{-dag}
-      for(int s=Ls-2; s>=0; s--){
-        spProj5p(tmp, chi[ss+s+1]);
-        chi[ss+s] = chi[ss+s] - this->lee[s]*tmp;
-      }
-    }
-
-    this->MooeeInvTime += usecond();
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInvDag_shift(const FermionField &psi, FermionField &chi)
-  {
-    GridBase *grid = psi._grid;
-    int Ls = this->Ls;
-
-    chi.checkerboard = psi.checkerboard;
-
-    this->MooeeInvCalls++;
-    this->MooeeInvTime -= usecond();
-
-    parallel_for(int ss=0; ss<grid->oSites(); ss+=Ls){
-
-      auto tmp1        = psi._odata[0];
-      auto tmp2        = psi._odata[0];
-      auto tmp2_spProj = psi._odata[0];
-
-      // Apply (U^{\prime})^{-dag} and accumulate MooeeInvDag_shift_lc[j]*psi[j] in tmp2
-      chi[ss] = psi[ss];
-      tmp2 = MooeeInvDag_shift_lc[0]*psi[ss];
-      for(int s=1; s<Ls; s++){
-        spProj5m(tmp1, chi[ss+s-1]);
-        chi[ss+s] = psi[ss+s] - this->uee[s-1]*tmp1;
-        tmp2 = tmp2 + MooeeInvDag_shift_lc[s]*psi[ss+s];
-      }
-      if(this->pm == 1){ spProj5p(tmp2_spProj, tmp2);}
-      else{ spProj5m(tmp2_spProj, tmp2); }
-
-      // U_m^{-\dag}
-      for(int s=0; s<Ls-1; s++){
-        spProj5p(tmp1, chi[ss+s]);
-        chi[ss+Ls-1] = chi[ss+Ls-1] - this->ueem[s]*tmp1;
-      }
-
-      // L_m^{-\dag} D^{-dag}
-      for(int s=0; s<Ls-1; s++){
-        spProj5m(tmp1, chi[ss+Ls-1]);
-        chi[ss+s] = (1.0/this->dee[s])*chi[ss+s] - (this->leem[s]/this->dee[Ls-1])*tmp1;
-      }
-      chi[ss+Ls-1] = (1.0/this->dee[Ls-1])*chi[ss+Ls-1];
-      spProj5p(tmp1, chi[ss+Ls-1]);
-      chi[ss+Ls-1] = chi[ss+Ls-1] + MooeeInvDag_shift_norm[Ls-1]*tmp2_spProj;
-
-      // Apply L^{-dag}
-      for(int s=Ls-2; s>=0; s--){
-        chi[ss+s] = chi[ss+s] - this->lee[s]*tmp1;
-        spProj5p(tmp1, chi[ss+s]);
-        chi[ss+s] = chi[ss+s] + MooeeInvDag_shift_norm[s]*tmp2_spProj;
-      }
-    }
-
-    this->MooeeInvTime += usecond();
-  }
-
-  #ifdef MOBIUS_EOFA_DPERP_CACHE
-
-    INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplD);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplD);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplD);
-
-    INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplFH);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplDF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplFH);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplDF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplFH);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplDF);
-
-  #endif
-
-}}
diff --git a/Grid/qcd/action/fermion/MobiusEOFAFermiondense.cc b/Grid/qcd/action/fermion/MobiusEOFAFermiondense.cc
deleted file mode 100644
index d66b8cd9..00000000
--- a/Grid/qcd/action/fermion/MobiusEOFAFermiondense.cc
+++ /dev/null
@@ -1,184 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/qcd/action/fermion/MobiusEOFAFermiondense.cc
-
-Copyright (C) 2017
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-Author: David Murphy <dmurphy@phys.columbia.edu>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-#include <Grid/Grid_Eigen_Dense.h>
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
-
-namespace Grid {
-namespace QCD {
-
-  /*
-  * Dense matrix versions of routines
-  */
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInv(const FermionField& psi, FermionField& chi)
-  {
-    this->MooeeInternal(psi, chi, DaggerNo, InverseYes);
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInv_shift(const FermionField& psi, FermionField& chi)
-  {
-    this->MooeeInternal(psi, chi, DaggerNo, InverseYes);
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInvDag(const FermionField& psi, FermionField& chi)
-  {
-    this->MooeeInternal(psi, chi, DaggerYes, InverseYes);
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInvDag_shift(const FermionField& psi, FermionField& chi)
-  {
-    this->MooeeInternal(psi, chi, DaggerYes, InverseYes);
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv)
-  {
-    int Ls = this->Ls;
-    int LLs = psi._grid->_rdimensions[0];
-    int vol = psi._grid->oSites()/LLs;
-
-    int pm      = this->pm;
-    RealD shift = this->shift;
-    RealD alpha = this->alpha;
-    RealD k     = this->k;
-    RealD mq1   = this->mq1;
-
-    chi.checkerboard = psi.checkerboard;
-
-    assert(Ls==LLs);
-
-    Eigen::MatrixXd Pplus  = Eigen::MatrixXd::Zero(Ls,Ls);
-    Eigen::MatrixXd Pminus = Eigen::MatrixXd::Zero(Ls,Ls);
-
-    for(int s=0;s<Ls;s++){
-        Pplus(s,s)  = this->bee[s];
-        Pminus(s,s) = this->bee[s];
-    }
-
-    for(int s=0; s<Ls-1; s++){
-        Pminus(s,s+1) = -this->cee[s];
-    }
-
-    for(int s=0; s<Ls-1; s++){
-        Pplus(s+1,s) = -this->cee[s+1];
-    }
-    Pplus (0,Ls-1) = mq1*this->cee[0];
-    Pminus(Ls-1,0) = mq1*this->cee[Ls-1];
-
-    if(shift != 0.0){
-      Coeff_t N = 2.0 * ( std::pow(alpha+1.0,Ls) + mq1*std::pow(alpha-1.0,Ls) );
-      for(int s=0; s<Ls; ++s){
-        if(pm == 1){ Pplus(s,Ls-1) += shift * k * N * std::pow(-1.0,s) * std::pow(alpha-1.0,s) / std::pow(alpha+1.0,Ls+s+1); }
-        else{ Pminus(Ls-1-s,Ls-1) -= shift * k * N * std::pow(-1.0,s) * std::pow(alpha-1.0,s) / std::pow(alpha+1.0,Ls+s+1); }
-      }
-    }
-
-    Eigen::MatrixXd PplusMat ;
-    Eigen::MatrixXd PminusMat;
-
-    if(inv){
-      PplusMat  = Pplus.inverse();
-      PminusMat = Pminus.inverse();
-    } else {
-      PplusMat  = Pplus;
-      PminusMat = Pminus;
-    }
-
-    if(dag){
-      PplusMat.adjointInPlace();
-      PminusMat.adjointInPlace();
-    }
-
-    // For the non-vectorised s-direction this is simple
-
-    for(auto site=0; site<vol; site++){
-
-        SiteSpinor     SiteChi;
-        SiteHalfSpinor SitePplus;
-        SiteHalfSpinor SitePminus;
-
-        for(int s1=0; s1<Ls; s1++){
-            SiteChi = zero;
-            for(int s2=0; s2<Ls; s2++){
-                int lex2 = s2 + Ls*site;
-                if(PplusMat(s1,s2) != 0.0){
-                    spProj5p(SitePplus,psi[lex2]);
-                    accumRecon5p(SiteChi, PplusMat(s1,s2)*SitePplus);
-                }
-                if(PminusMat(s1,s2) != 0.0){
-                    spProj5m(SitePminus, psi[lex2]);
-                    accumRecon5m(SiteChi, PminusMat(s1,s2)*SitePminus);
-                }
-            }
-            chi[s1+Ls*site] = SiteChi*0.5;
-        }
-    }
-  }
-
-  #ifdef MOBIUS_EOFA_DPERP_DENSE
-
-    INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplD);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplD);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplD);
-
-    template void MobiusEOFAFermion<GparityWilsonImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<GparityWilsonImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<WilsonImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<WilsonImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<ZWilsonImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<ZWilsonImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-
-    INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplFH);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplDF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplFH);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplDF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplFH);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplDF);
-
-    template void MobiusEOFAFermion<GparityWilsonImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<GparityWilsonImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<WilsonImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<WilsonImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<ZWilsonImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<ZWilsonImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-
-  #endif
-
-}}
diff --git a/Grid/qcd/action/fermion/MobiusEOFAFermionssp.cc b/Grid/qcd/action/fermion/MobiusEOFAFermionssp.cc
deleted file mode 100644
index c86bb995..00000000
--- a/Grid/qcd/action/fermion/MobiusEOFAFermionssp.cc
+++ /dev/null
@@ -1,290 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/qcd/action/fermion/MobiusEOFAFermionssp.cc
-
-Copyright (C) 2017
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-Author: David Murphy <dmurphy@phys.columbia.edu>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
-
-namespace Grid {
-namespace QCD {
-
-  // FIXME -- make a version of these routines with site loop outermost for cache reuse.
-  // Pminus fowards
-  // Pplus  backwards
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::M5D(const FermionField& psi, const FermionField& phi,
-    FermionField& chi, std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper)
-  {
-    Coeff_t one(1.0);
-    int Ls = this->Ls;
-    for(int s=0; s<Ls; s++){
-      if(s==0) {
-        axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, s+1);
-        axpby_ssp_pplus (chi, one, chi, lower[s], psi, s, Ls-1);
-      } else if (s==(Ls-1)) {
-        axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, 0);
-        axpby_ssp_pplus (chi, one, chi, lower[s], psi, s, s-1);
-      } else {
-        axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, s+1);
-        axpby_ssp_pplus(chi, one, chi, lower[s], psi, s, s-1);
-      }
-    }
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::M5D_shift(const FermionField& psi, const FermionField& phi,
-    FermionField& chi, std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper,
-    std::vector<Coeff_t>& shift_coeffs)
-  {
-    Coeff_t one(1.0);
-    int Ls = this->Ls;
-    for(int s=0; s<Ls; s++){
-      if(s==0) {
-        axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, s+1);
-        axpby_ssp_pplus (chi, one, chi, lower[s], psi, s, Ls-1);
-      } else if (s==(Ls-1)) {
-        axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, 0);
-        axpby_ssp_pplus (chi, one, chi, lower[s], psi, s, s-1);
-      } else {
-        axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, s+1);
-        axpby_ssp_pplus(chi, one, chi, lower[s], psi, s, s-1);
-      }
-      if(this->pm == 1){ axpby_ssp_pplus(chi, one, chi, shift_coeffs[s], psi, s, Ls-1); }
-      else{ axpby_ssp_pminus(chi, one, chi, shift_coeffs[s], psi, s, 0); }
-    }
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::M5Ddag(const FermionField& psi, const FermionField& phi,
-    FermionField& chi, std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper)
-  {
-    Coeff_t one(1.0);
-    int Ls = this->Ls;
-    for(int s=0; s<Ls; s++){
-      if(s==0) {
-        axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, s+1);
-        axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, Ls-1);
-      } else if (s==(Ls-1)) {
-        axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, 0);
-        axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, s-1);
-      } else {
-        axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, s+1);
-        axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, s-1);
-      }
-    }
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::M5Ddag_shift(const FermionField& psi, const FermionField& phi,
-    FermionField& chi, std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper,
-    std::vector<Coeff_t>& shift_coeffs)
-  {
-    Coeff_t one(1.0);
-    int Ls = this->Ls;
-    for(int s=0; s<Ls; s++){
-      if(s==0) {
-        axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, s+1);
-        axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, Ls-1);
-      } else if (s==(Ls-1)) {
-        axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, 0);
-        axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, s-1);
-      } else {
-        axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, s+1);
-        axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, s-1);
-      }
-      if(this->pm == 1){ axpby_ssp_pplus(chi, one, chi, shift_coeffs[s], psi, Ls-1, s); }
-      else{ axpby_ssp_pminus(chi, one, chi, shift_coeffs[s], psi, 0, s); }
-    }
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInv(const FermionField& psi, FermionField& chi)
-  {
-    if(this->shift != 0.0){ MooeeInv_shift(psi,chi); return; }
-
-    Coeff_t one(1.0);
-    Coeff_t czero(0.0);
-    chi.checkerboard = psi.checkerboard;
-    int Ls = this->Ls;
-
-    // Apply (L^{\prime})^{-1}
-    axpby_ssp(chi, one, psi, czero, psi, 0, 0);      // chi[0]=psi[0]
-    for(int s=1; s<Ls; s++){
-      axpby_ssp_pplus(chi, one, psi, -this->lee[s-1], chi, s, s-1);// recursion Psi[s] -lee P_+ chi[s-1]
-    }
-
-    // L_m^{-1}
-    for(int s=0; s<Ls-1; s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
-      axpby_ssp_pminus(chi, one, chi, -this->leem[s], chi, Ls-1, s);
-    }
-
-    // U_m^{-1} D^{-1}
-    for(int s=0; s<Ls-1; s++){
-      axpby_ssp_pplus(chi, one/this->dee[s], chi, -this->ueem[s]/this->dee[Ls-1], chi, s, Ls-1);
-    }
-    axpby_ssp(chi, one/this->dee[Ls-1], chi, czero, chi, Ls-1, Ls-1);
-
-    // Apply U^{-1}
-    for(int s=Ls-2; s>=0; s--){
-      axpby_ssp_pminus(chi, one, chi, -this->uee[s], chi, s, s+1);  // chi[Ls]
-    }
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInv_shift(const FermionField& psi, FermionField& chi)
-  {
-    Coeff_t one(1.0);
-    Coeff_t czero(0.0);
-    chi.checkerboard = psi.checkerboard;
-    int Ls = this->Ls;
-
-    FermionField tmp(psi._grid);
-
-    // Apply (L^{\prime})^{-1}
-    axpby_ssp(chi, one, psi, czero, psi, 0, 0);      // chi[0]=psi[0]
-    axpby_ssp(tmp, czero, tmp, this->MooeeInv_shift_lc[0], psi, 0, 0);
-    for(int s=1; s<Ls; s++){
-      axpby_ssp_pplus(chi, one, psi, -this->lee[s-1], chi, s, s-1);// recursion Psi[s] -lee P_+ chi[s-1]
-      axpby_ssp(tmp, one, tmp, this->MooeeInv_shift_lc[s], psi, 0, s);
-    }
-
-    // L_m^{-1}
-    for(int s=0; s<Ls-1; s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
-      axpby_ssp_pminus(chi, one, chi, -this->leem[s], chi, Ls-1, s);
-    }
-
-    // U_m^{-1} D^{-1}
-    for(int s=0; s<Ls-1; s++){
-      axpby_ssp_pplus(chi, one/this->dee[s], chi, -this->ueem[s]/this->dee[Ls-1], chi, s, Ls-1);
-    }
-    axpby_ssp(chi, one/this->dee[Ls-1], chi, czero, chi, Ls-1, Ls-1);
-
-    // Apply U^{-1} and add shift term
-    if(this->pm == 1){ axpby_ssp_pplus(chi, one, chi, this->MooeeInv_shift_norm[Ls-1], tmp, Ls-1, 0); }
-    else{ axpby_ssp_pminus(chi, one, chi, this->MooeeInv_shift_norm[Ls-1], tmp, Ls-1, 0); }
-    for(int s=Ls-2; s>=0; s--){
-      axpby_ssp_pminus(chi, one, chi, -this->uee[s], chi, s, s+1);  // chi[Ls]
-      if(this->pm == 1){ axpby_ssp_pplus(chi, one, chi, this->MooeeInv_shift_norm[s], tmp, s, 0); }
-      else{ axpby_ssp_pminus(chi, one, chi, this->MooeeInv_shift_norm[s], tmp, s, 0); }
-    }
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInvDag(const FermionField& psi, FermionField& chi)
-  {
-    if(this->shift != 0.0){ MooeeInvDag_shift(psi,chi); return; }
-
-    Coeff_t one(1.0);
-    Coeff_t czero(0.0);
-    chi.checkerboard = psi.checkerboard;
-    int Ls = this->Ls;
-
-    // Apply (U^{\prime})^{-dagger}
-    axpby_ssp(chi, one, psi, czero, psi, 0, 0);      // chi[0]=psi[0]
-    for(int s=1; s<Ls; s++){
-      axpby_ssp_pminus(chi, one, psi, -conjugate(this->uee[s-1]), chi, s, s-1);
-    }
-
-    // U_m^{-\dagger}
-    for(int s=0; s<Ls-1; s++){
-      axpby_ssp_pplus(chi, one, chi, -conjugate(this->ueem[s]), chi, Ls-1, s);
-    }
-
-    // L_m^{-\dagger} D^{-dagger}
-    for(int s=0; s<Ls-1; s++){
-      axpby_ssp_pminus(chi, one/conjugate(this->dee[s]), chi, -conjugate(this->leem[s]/this->dee[Ls-1]), chi, s, Ls-1);
-    }
-    axpby_ssp(chi, one/conjugate(this->dee[Ls-1]), chi, czero, chi, Ls-1, Ls-1);
-
-    // Apply L^{-dagger}
-    for(int s=Ls-2; s>=0; s--){
-      axpby_ssp_pplus(chi, one, chi, -conjugate(this->lee[s]), chi, s, s+1);  // chi[Ls]
-    }
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInvDag_shift(const FermionField& psi, FermionField& chi)
-  {
-    Coeff_t one(1.0);
-    Coeff_t czero(0.0);
-    chi.checkerboard = psi.checkerboard;
-    int Ls = this->Ls;
-
-    FermionField tmp(psi._grid);
-
-    // Apply (U^{\prime})^{-dagger} and accumulate (MooeeInvDag_shift_lc)_{j} \psi_{j} in tmp[0]
-    axpby_ssp(chi, one, psi, czero, psi, 0, 0);      // chi[0]=psi[0]
-    axpby_ssp(tmp, czero, tmp, this->MooeeInvDag_shift_lc[0], psi, 0, 0);
-    for(int s=1; s<Ls; s++){
-      axpby_ssp_pminus(chi, one, psi, -conjugate(this->uee[s-1]), chi, s, s-1);
-      axpby_ssp(tmp, one, tmp, this->MooeeInvDag_shift_lc[s], psi, 0, s);
-    }
-
-    // U_m^{-\dagger}
-    for(int s=0; s<Ls-1; s++){
-      axpby_ssp_pplus(chi, one, chi, -conjugate(this->ueem[s]), chi, Ls-1, s);
-    }
-
-    // L_m^{-\dagger} D^{-dagger}
-    for(int s=0; s<Ls-1; s++){
-      axpby_ssp_pminus(chi, one/conjugate(this->dee[s]), chi, -conjugate(this->leem[s]/this->dee[Ls-1]), chi, s, Ls-1);
-    }
-    axpby_ssp(chi, one/conjugate(this->dee[Ls-1]), chi, czero, chi, Ls-1, Ls-1);
-
-    // Apply L^{-dagger} and add shift
-    if(this->pm == 1){ axpby_ssp_pplus(chi, one, chi, this->MooeeInvDag_shift_norm[Ls-1], tmp, Ls-1, 0); }
-    else{ axpby_ssp_pminus(chi, one, chi, this->MooeeInvDag_shift_norm[Ls-1], tmp, Ls-1, 0); }
-    for(int s=Ls-2; s>=0; s--){
-      axpby_ssp_pplus(chi, one, chi, -conjugate(this->lee[s]), chi, s, s+1);  // chi[Ls]
-      if(this->pm == 1){ axpby_ssp_pplus(chi, one, chi, this->MooeeInvDag_shift_norm[s], tmp, s, 0); }
-      else{ axpby_ssp_pminus(chi, one, chi, this->MooeeInvDag_shift_norm[s], tmp, s, 0); }
-    }
-  }
-
-  #ifdef MOBIUS_EOFA_DPERP_LINALG
-
-    INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplD);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplD);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplD);
-
-    INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplFH);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplDF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplFH);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplDF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplFH);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplDF);
-
-  #endif
-
-}}
diff --git a/Grid/qcd/action/fermion/MobiusEOFAFermionvec.cc b/Grid/qcd/action/fermion/MobiusEOFAFermionvec.cc
deleted file mode 100644
index 290ba158..00000000
--- a/Grid/qcd/action/fermion/MobiusEOFAFermionvec.cc
+++ /dev/null
@@ -1,983 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/qcd/action/fermion/MobiusEOFAFermionvec.cc
-
-Copyright (C) 2017
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-Author: David Murphy <dmurphy@phys.columbia.edu>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
-
-namespace Grid {
-namespace QCD {
-
-  /*
-  * Dense matrix versions of routines
-  */
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInv(const FermionField& psi, FermionField& chi)
-  {
-    this->MooeeInternal(psi, chi, DaggerNo, InverseYes);
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInv_shift(const FermionField& psi, FermionField& chi)
-  {
-    this->MooeeInternal(psi, chi, DaggerNo, InverseYes);
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInvDag(const FermionField& psi, FermionField& chi)
-  {
-    this->MooeeInternal(psi, chi, DaggerYes, InverseYes);
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInvDag_shift(const FermionField& psi, FermionField& chi)
-  {
-    this->MooeeInternal(psi, chi, DaggerYes, InverseYes);
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::M5D(const FermionField& psi, const FermionField& phi,
-    FermionField& chi, std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper)
-  {
-    GridBase* grid  = psi._grid;
-    int Ls          = this->Ls;
-    int LLs         = grid->_rdimensions[0];
-    const int nsimd = Simd::Nsimd();
-
-    Vector<iSinglet<Simd>> u(LLs);
-    Vector<iSinglet<Simd>> l(LLs);
-    Vector<iSinglet<Simd>> d(LLs);
-
-    assert(Ls/LLs == nsimd);
-    assert(phi.checkerboard == psi.checkerboard);
-
-    chi.checkerboard = psi.checkerboard;
-
-    // just directly address via type pun
-    typedef typename Simd::scalar_type scalar_type;
-    scalar_type* u_p = (scalar_type*) &u[0];
-    scalar_type* l_p = (scalar_type*) &l[0];
-    scalar_type* d_p = (scalar_type*) &d[0];
-
-    for(int o=0; o<LLs; o++){ // outer
-    for(int i=0; i<nsimd; i++){ //inner
-      int s   = o + i*LLs;
-      int ss  = o*nsimd + i;
-      u_p[ss] = upper[s];
-      l_p[ss] = lower[s];
-      d_p[ss] = diag[s];
-    }}
-
-    this->M5Dcalls++;
-    this->M5Dtime -= usecond();
-
-    assert(Nc == 3);
-
-    parallel_for(int ss=0; ss<grid->oSites(); ss+=LLs){ // adds LLs
-
-      #if 0
-
-        alignas(64) SiteHalfSpinor hp;
-        alignas(64) SiteHalfSpinor hm;
-        alignas(64) SiteSpinor fp;
-        alignas(64) SiteSpinor fm;
-
-        for(int v=0; v<LLs; v++){
-
-          int vp = (v+1)%LLs;
-          int vm = (v+LLs-1)%LLs;
-
-          spProj5m(hp, psi[ss+vp]);
-          spProj5p(hm, psi[ss+vm]);
-
-          if (vp <= v){ rotate(hp, hp, 1); }
-          if (vm >= v){ rotate(hm, hm, nsimd-1); }
-
-          hp = 0.5*hp;
-          hm = 0.5*hm;
-
-          spRecon5m(fp, hp);
-          spRecon5p(fm, hm);
-
-          chi[ss+v] = d[v]*phi[ss+v];
-          chi[ss+v] = chi[ss+v] + u[v]*fp;
-          chi[ss+v] = chi[ss+v] + l[v]*fm;
-
-        }
-
-      #else
-
-        for(int v=0; v<LLs; v++){
-
-          vprefetch(psi[ss+v+LLs]);
-
-          int vp = (v == LLs-1) ? 0     : v+1;
-          int vm = (v == 0)     ? LLs-1 : v-1;
-
-          Simd hp_00 = psi[ss+vp]()(2)(0);
-          Simd hp_01 = psi[ss+vp]()(2)(1);
-          Simd hp_02 = psi[ss+vp]()(2)(2);
-          Simd hp_10 = psi[ss+vp]()(3)(0);
-          Simd hp_11 = psi[ss+vp]()(3)(1);
-          Simd hp_12 = psi[ss+vp]()(3)(2);
-
-          Simd hm_00 = psi[ss+vm]()(0)(0);
-          Simd hm_01 = psi[ss+vm]()(0)(1);
-          Simd hm_02 = psi[ss+vm]()(0)(2);
-          Simd hm_10 = psi[ss+vm]()(1)(0);
-          Simd hm_11 = psi[ss+vm]()(1)(1);
-          Simd hm_12 = psi[ss+vm]()(1)(2);
-
-          if(vp <= v){
-            hp_00.v = Optimization::Rotate::tRotate<2>(hp_00.v);
-            hp_01.v = Optimization::Rotate::tRotate<2>(hp_01.v);
-            hp_02.v = Optimization::Rotate::tRotate<2>(hp_02.v);
-            hp_10.v = Optimization::Rotate::tRotate<2>(hp_10.v);
-            hp_11.v = Optimization::Rotate::tRotate<2>(hp_11.v);
-            hp_12.v = Optimization::Rotate::tRotate<2>(hp_12.v);
-          }
-
-          if(vm >= v){
-            hm_00.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_00.v);
-            hm_01.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_01.v);
-            hm_02.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_02.v);
-            hm_10.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_10.v);
-            hm_11.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_11.v);
-            hm_12.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_12.v);
-          }
-
-          // Can force these to real arithmetic and save 2x.
-          Simd p_00 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_00);
-          Simd p_01 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_01);
-          Simd p_02 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_02);
-          Simd p_10 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_10);
-          Simd p_11 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_11);
-          Simd p_12 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_12);
-          Simd p_20 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_00);
-          Simd p_21 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_01);
-          Simd p_22 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_02);
-          Simd p_30 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_10);
-          Simd p_31 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_11);
-          Simd p_32 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_12);
-
-          vstream(chi[ss+v]()(0)(0), p_00);
-          vstream(chi[ss+v]()(0)(1), p_01);
-          vstream(chi[ss+v]()(0)(2), p_02);
-          vstream(chi[ss+v]()(1)(0), p_10);
-          vstream(chi[ss+v]()(1)(1), p_11);
-          vstream(chi[ss+v]()(1)(2), p_12);
-          vstream(chi[ss+v]()(2)(0), p_20);
-          vstream(chi[ss+v]()(2)(1), p_21);
-          vstream(chi[ss+v]()(2)(2), p_22);
-          vstream(chi[ss+v]()(3)(0), p_30);
-          vstream(chi[ss+v]()(3)(1), p_31);
-          vstream(chi[ss+v]()(3)(2), p_32);
-        }
-
-      #endif
-    }
-
-    this->M5Dtime += usecond();
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::M5D_shift(const FermionField& psi, const FermionField& phi,
-    FermionField& chi, std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper,
-    std::vector<Coeff_t>& shift_coeffs)
-  {
-    #if 0
-
-      this->M5D(psi, phi, chi, lower, diag, upper);
-
-      // FIXME: possible gain from vectorizing shift operation as well?
-      Coeff_t one(1.0);
-      int Ls = this->Ls;
-      for(int s=0; s<Ls; s++){
-        if(this->pm == 1){ axpby_ssp_pplus(chi, one, chi, shift_coeffs[s], psi, s, Ls-1); }
-        else{ axpby_ssp_pminus(chi, one, chi, shift_coeffs[s], psi, s, 0); }
-      }
-
-    #else
-
-      GridBase* grid  = psi._grid;
-      int Ls          = this->Ls;
-      int LLs         = grid->_rdimensions[0];
-      const int nsimd = Simd::Nsimd();
-
-      Vector<iSinglet<Simd>> u(LLs);
-      Vector<iSinglet<Simd>> l(LLs);
-      Vector<iSinglet<Simd>> d(LLs);
-      Vector<iSinglet<Simd>> s(LLs);
-
-      assert(Ls/LLs == nsimd);
-      assert(phi.checkerboard == psi.checkerboard);
-
-      chi.checkerboard = psi.checkerboard;
-
-      // just directly address via type pun
-      typedef typename Simd::scalar_type scalar_type;
-      scalar_type* u_p = (scalar_type*) &u[0];
-      scalar_type* l_p = (scalar_type*) &l[0];
-      scalar_type* d_p = (scalar_type*) &d[0];
-      scalar_type* s_p = (scalar_type*) &s[0];
-
-      for(int o=0; o<LLs; o++){ // outer
-      for(int i=0; i<nsimd; i++){ //inner
-        int s   = o + i*LLs;
-        int ss  = o*nsimd + i;
-        u_p[ss] = upper[s];
-        l_p[ss] = lower[s];
-        d_p[ss] = diag[s];
-        s_p[ss] = shift_coeffs[s];
-      }}
-
-      this->M5Dcalls++;
-      this->M5Dtime -= usecond();
-
-      assert(Nc == 3);
-
-      parallel_for(int ss=0; ss<grid->oSites(); ss+=LLs){ // adds LLs
-
-        int vs     = (this->pm == 1) ? LLs-1 : 0;
-        Simd hs_00 = (this->pm == 1) ? psi[ss+vs]()(2)(0) : psi[ss+vs]()(0)(0);
-        Simd hs_01 = (this->pm == 1) ? psi[ss+vs]()(2)(1) : psi[ss+vs]()(0)(1);
-        Simd hs_02 = (this->pm == 1) ? psi[ss+vs]()(2)(2) : psi[ss+vs]()(0)(2);
-        Simd hs_10 = (this->pm == 1) ? psi[ss+vs]()(3)(0) : psi[ss+vs]()(1)(0);
-        Simd hs_11 = (this->pm == 1) ? psi[ss+vs]()(3)(1) : psi[ss+vs]()(1)(1);
-        Simd hs_12 = (this->pm == 1) ? psi[ss+vs]()(3)(2) : psi[ss+vs]()(1)(2);
-
-        for(int v=0; v<LLs; v++){
-
-          vprefetch(psi[ss+v+LLs]);
-
-          int vp = (v == LLs-1) ? 0     : v+1;
-          int vm = (v == 0)     ? LLs-1 : v-1;
-
-          Simd hp_00 = psi[ss+vp]()(2)(0);
-          Simd hp_01 = psi[ss+vp]()(2)(1);
-          Simd hp_02 = psi[ss+vp]()(2)(2);
-          Simd hp_10 = psi[ss+vp]()(3)(0);
-          Simd hp_11 = psi[ss+vp]()(3)(1);
-          Simd hp_12 = psi[ss+vp]()(3)(2);
-
-          Simd hm_00 = psi[ss+vm]()(0)(0);
-          Simd hm_01 = psi[ss+vm]()(0)(1);
-          Simd hm_02 = psi[ss+vm]()(0)(2);
-          Simd hm_10 = psi[ss+vm]()(1)(0);
-          Simd hm_11 = psi[ss+vm]()(1)(1);
-          Simd hm_12 = psi[ss+vm]()(1)(2);
-
-          if(vp <= v){
-            hp_00.v = Optimization::Rotate::tRotate<2>(hp_00.v);
-            hp_01.v = Optimization::Rotate::tRotate<2>(hp_01.v);
-            hp_02.v = Optimization::Rotate::tRotate<2>(hp_02.v);
-            hp_10.v = Optimization::Rotate::tRotate<2>(hp_10.v);
-            hp_11.v = Optimization::Rotate::tRotate<2>(hp_11.v);
-            hp_12.v = Optimization::Rotate::tRotate<2>(hp_12.v);
-          }
-
-          if(this->pm == 1 && vs <= v){
-            hs_00.v = Optimization::Rotate::tRotate<2>(hs_00.v);
-            hs_01.v = Optimization::Rotate::tRotate<2>(hs_01.v);
-            hs_02.v = Optimization::Rotate::tRotate<2>(hs_02.v);
-            hs_10.v = Optimization::Rotate::tRotate<2>(hs_10.v);
-            hs_11.v = Optimization::Rotate::tRotate<2>(hs_11.v);
-            hs_12.v = Optimization::Rotate::tRotate<2>(hs_12.v);
-          }
-
-          if(vm >= v){
-            hm_00.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_00.v);
-            hm_01.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_01.v);
-            hm_02.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_02.v);
-            hm_10.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_10.v);
-            hm_11.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_11.v);
-            hm_12.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_12.v);
-          }
-
-          if(this->pm == -1 && vs >= v){
-            hs_00.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hs_00.v);
-            hs_01.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hs_01.v);
-            hs_02.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hs_02.v);
-            hs_10.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hs_10.v);
-            hs_11.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hs_11.v);
-            hs_12.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hs_12.v);
-          }
-
-          // Can force these to real arithmetic and save 2x.
-          Simd p_00 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_00)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_00)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_00);
-          Simd p_01 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_01)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_01)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_01);
-          Simd p_02 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_02)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_02)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_02);
-          Simd p_10 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_10)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_10)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_10);
-          Simd p_11 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_11)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_11)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_11);
-          Simd p_12 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_12)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_12)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_12);
-          Simd p_20 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_00)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_00)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_00);
-          Simd p_21 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_01)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_01)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_01);
-          Simd p_22 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_02)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_02)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_02);
-          Simd p_30 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_10)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_10)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_10);
-          Simd p_31 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_11)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_11)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_11);
-          Simd p_32 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_12)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_12)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_12);
-
-          vstream(chi[ss+v]()(0)(0), p_00);
-          vstream(chi[ss+v]()(0)(1), p_01);
-          vstream(chi[ss+v]()(0)(2), p_02);
-          vstream(chi[ss+v]()(1)(0), p_10);
-          vstream(chi[ss+v]()(1)(1), p_11);
-          vstream(chi[ss+v]()(1)(2), p_12);
-          vstream(chi[ss+v]()(2)(0), p_20);
-          vstream(chi[ss+v]()(2)(1), p_21);
-          vstream(chi[ss+v]()(2)(2), p_22);
-          vstream(chi[ss+v]()(3)(0), p_30);
-          vstream(chi[ss+v]()(3)(1), p_31);
-          vstream(chi[ss+v]()(3)(2), p_32);
-        }
-      }
-
-      this->M5Dtime += usecond();
-
-    #endif
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::M5Ddag(const FermionField& psi, const FermionField& phi,
-    FermionField& chi, std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper)
-  {
-    GridBase* grid = psi._grid;
-    int Ls  = this->Ls;
-    int LLs = grid->_rdimensions[0];
-    int nsimd = Simd::Nsimd();
-
-    Vector<iSinglet<Simd>> u(LLs);
-    Vector<iSinglet<Simd>> l(LLs);
-    Vector<iSinglet<Simd>> d(LLs);
-
-    assert(Ls/LLs == nsimd);
-    assert(phi.checkerboard == psi.checkerboard);
-
-    chi.checkerboard = psi.checkerboard;
-
-    // just directly address via type pun
-    typedef typename Simd::scalar_type scalar_type;
-    scalar_type* u_p = (scalar_type*) &u[0];
-    scalar_type* l_p = (scalar_type*) &l[0];
-    scalar_type* d_p = (scalar_type*) &d[0];
-
-    for(int o=0; o<LLs; o++){ // outer
-    for(int i=0; i<nsimd; i++){ //inner
-      int s  = o + i*LLs;
-      int ss = o*nsimd + i;
-      u_p[ss] = upper[s];
-      l_p[ss] = lower[s];
-      d_p[ss] = diag[s];
-    }}
-
-    this->M5Dcalls++;
-    this->M5Dtime -= usecond();
-
-    parallel_for(int ss=0; ss<grid->oSites(); ss+=LLs){ // adds LLs
-
-      #if 0
-
-        alignas(64) SiteHalfSpinor hp;
-        alignas(64) SiteHalfSpinor hm;
-        alignas(64) SiteSpinor fp;
-        alignas(64) SiteSpinor fm;
-
-        for(int v=0; v<LLs; v++){
-
-          int vp = (v+1)%LLs;
-          int vm = (v+LLs-1)%LLs;
-
-          spProj5p(hp, psi[ss+vp]);
-          spProj5m(hm, psi[ss+vm]);
-
-          if(vp <= v){ rotate(hp, hp, 1); }
-          if(vm >= v){ rotate(hm, hm, nsimd-1); }
-
-          hp = hp*0.5;
-          hm = hm*0.5;
-          spRecon5p(fp, hp);
-          spRecon5m(fm, hm);
-
-          chi[ss+v] = d[v]*phi[ss+v]+u[v]*fp;
-          chi[ss+v] = chi[ss+v]     +l[v]*fm;
-
-        }
-
-      #else
-
-        for(int v=0; v<LLs; v++){
-
-          vprefetch(psi[ss+v+LLs]);
-
-          int vp = (v == LLs-1) ? 0     : v+1;
-          int vm = (v == 0    ) ? LLs-1 : v-1;
-
-          Simd hp_00 = psi[ss+vp]()(0)(0);
-          Simd hp_01 = psi[ss+vp]()(0)(1);
-          Simd hp_02 = psi[ss+vp]()(0)(2);
-          Simd hp_10 = psi[ss+vp]()(1)(0);
-          Simd hp_11 = psi[ss+vp]()(1)(1);
-          Simd hp_12 = psi[ss+vp]()(1)(2);
-
-          Simd hm_00 = psi[ss+vm]()(2)(0);
-          Simd hm_01 = psi[ss+vm]()(2)(1);
-          Simd hm_02 = psi[ss+vm]()(2)(2);
-          Simd hm_10 = psi[ss+vm]()(3)(0);
-          Simd hm_11 = psi[ss+vm]()(3)(1);
-          Simd hm_12 = psi[ss+vm]()(3)(2);
-
-          if (vp <= v){
-            hp_00.v = Optimization::Rotate::tRotate<2>(hp_00.v);
-            hp_01.v = Optimization::Rotate::tRotate<2>(hp_01.v);
-            hp_02.v = Optimization::Rotate::tRotate<2>(hp_02.v);
-            hp_10.v = Optimization::Rotate::tRotate<2>(hp_10.v);
-            hp_11.v = Optimization::Rotate::tRotate<2>(hp_11.v);
-            hp_12.v = Optimization::Rotate::tRotate<2>(hp_12.v);
-          }
-
-          if(vm >= v){
-            hm_00.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_00.v);
-            hm_01.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_01.v);
-            hm_02.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_02.v);
-            hm_10.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_10.v);
-            hm_11.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_11.v);
-            hm_12.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_12.v);
-          }
-
-          Simd p_00 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_00);
-          Simd p_01 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_01);
-          Simd p_02 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_02);
-          Simd p_10 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_10);
-          Simd p_11 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_11);
-          Simd p_12 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_12);
-          Simd p_20 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_00);
-          Simd p_21 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_01);
-          Simd p_22 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_02);
-          Simd p_30 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_10);
-          Simd p_31 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_11);
-          Simd p_32 = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_12);
-
-          vstream(chi[ss+v]()(0)(0), p_00);
-          vstream(chi[ss+v]()(0)(1), p_01);
-          vstream(chi[ss+v]()(0)(2), p_02);
-          vstream(chi[ss+v]()(1)(0), p_10);
-          vstream(chi[ss+v]()(1)(1), p_11);
-          vstream(chi[ss+v]()(1)(2), p_12);
-          vstream(chi[ss+v]()(2)(0), p_20);
-          vstream(chi[ss+v]()(2)(1), p_21);
-          vstream(chi[ss+v]()(2)(2), p_22);
-          vstream(chi[ss+v]()(3)(0), p_30);
-          vstream(chi[ss+v]()(3)(1), p_31);
-          vstream(chi[ss+v]()(3)(2), p_32);
-
-        }
-
-      #endif
-
-    }
-
-    this->M5Dtime += usecond();
-  }
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::M5Ddag_shift(const FermionField& psi, const FermionField& phi,
-    FermionField& chi, std::vector<Coeff_t>& lower, std::vector<Coeff_t>& diag, std::vector<Coeff_t>& upper,
-    std::vector<Coeff_t>& shift_coeffs)
-  {
-    #if 0
-
-      this->M5Ddag(psi, phi, chi, lower, diag, upper);
-
-      // FIXME: possible gain from vectorizing shift operation as well?
-      Coeff_t one(1.0);
-      int Ls = this->Ls;
-      for(int s=0; s<Ls; s++){
-        if(this->pm == 1){ axpby_ssp_pplus(chi, one, chi, shift_coeffs[s], psi, Ls-1, s); }
-        else{ axpby_ssp_pminus(chi, one, chi, shift_coeffs[s], psi, 0, s); }
-      }
-
-    #else
-
-      GridBase* grid = psi._grid;
-      int Ls  = this->Ls;
-      int LLs = grid->_rdimensions[0];
-      int nsimd = Simd::Nsimd();
-
-      Vector<iSinglet<Simd>> u(LLs);
-      Vector<iSinglet<Simd>> l(LLs);
-      Vector<iSinglet<Simd>> d(LLs);
-      Vector<iSinglet<Simd>> s(LLs);
-
-      assert(Ls/LLs == nsimd);
-      assert(phi.checkerboard == psi.checkerboard);
-
-      chi.checkerboard = psi.checkerboard;
-
-      // just directly address via type pun
-      typedef typename Simd::scalar_type scalar_type;
-      scalar_type* u_p = (scalar_type*) &u[0];
-      scalar_type* l_p = (scalar_type*) &l[0];
-      scalar_type* d_p = (scalar_type*) &d[0];
-      scalar_type* s_p = (scalar_type*) &s[0];
-
-      for(int o=0; o<LLs; o++){ // outer
-      for(int i=0; i<nsimd; i++){ //inner
-        int s  = o + i*LLs;
-        int ss = o*nsimd + i;
-        u_p[ss] = upper[s];
-        l_p[ss] = lower[s];
-        d_p[ss] = diag[s];
-        s_p[ss] = shift_coeffs[s];
-      }}
-
-      this->M5Dcalls++;
-      this->M5Dtime -= usecond();
-
-      parallel_for(int ss=0; ss<grid->oSites(); ss+=LLs){ // adds LLs
-
-        int vs     = (this->pm == 1) ? LLs-1 : 0;
-        Simd hs_00 = (this->pm == 1) ? psi[ss+vs]()(0)(0) : psi[ss+vs]()(2)(0);
-        Simd hs_01 = (this->pm == 1) ? psi[ss+vs]()(0)(1) : psi[ss+vs]()(2)(1);
-        Simd hs_02 = (this->pm == 1) ? psi[ss+vs]()(0)(2) : psi[ss+vs]()(2)(2);
-        Simd hs_10 = (this->pm == 1) ? psi[ss+vs]()(1)(0) : psi[ss+vs]()(3)(0);
-        Simd hs_11 = (this->pm == 1) ? psi[ss+vs]()(1)(1) : psi[ss+vs]()(3)(1);
-        Simd hs_12 = (this->pm == 1) ? psi[ss+vs]()(1)(2) : psi[ss+vs]()(3)(2);
-
-        for(int v=0; v<LLs; v++){
-
-          vprefetch(psi[ss+v+LLs]);
-
-          int vp = (v == LLs-1) ? 0     : v+1;
-          int vm = (v == 0    ) ? LLs-1 : v-1;
-
-          Simd hp_00 = psi[ss+vp]()(0)(0);
-          Simd hp_01 = psi[ss+vp]()(0)(1);
-          Simd hp_02 = psi[ss+vp]()(0)(2);
-          Simd hp_10 = psi[ss+vp]()(1)(0);
-          Simd hp_11 = psi[ss+vp]()(1)(1);
-          Simd hp_12 = psi[ss+vp]()(1)(2);
-
-          Simd hm_00 = psi[ss+vm]()(2)(0);
-          Simd hm_01 = psi[ss+vm]()(2)(1);
-          Simd hm_02 = psi[ss+vm]()(2)(2);
-          Simd hm_10 = psi[ss+vm]()(3)(0);
-          Simd hm_11 = psi[ss+vm]()(3)(1);
-          Simd hm_12 = psi[ss+vm]()(3)(2);
-
-          if (vp <= v){
-            hp_00.v = Optimization::Rotate::tRotate<2>(hp_00.v);
-            hp_01.v = Optimization::Rotate::tRotate<2>(hp_01.v);
-            hp_02.v = Optimization::Rotate::tRotate<2>(hp_02.v);
-            hp_10.v = Optimization::Rotate::tRotate<2>(hp_10.v);
-            hp_11.v = Optimization::Rotate::tRotate<2>(hp_11.v);
-            hp_12.v = Optimization::Rotate::tRotate<2>(hp_12.v);
-          }
-
-          if(this->pm == 1 && vs <= v){
-            hs_00.v = Optimization::Rotate::tRotate<2>(hs_00.v);
-            hs_01.v = Optimization::Rotate::tRotate<2>(hs_01.v);
-            hs_02.v = Optimization::Rotate::tRotate<2>(hs_02.v);
-            hs_10.v = Optimization::Rotate::tRotate<2>(hs_10.v);
-            hs_11.v = Optimization::Rotate::tRotate<2>(hs_11.v);
-            hs_12.v = Optimization::Rotate::tRotate<2>(hs_12.v);
-          }
-
-          if(vm >= v){
-            hm_00.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_00.v);
-            hm_01.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_01.v);
-            hm_02.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_02.v);
-            hm_10.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_10.v);
-            hm_11.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_11.v);
-            hm_12.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_12.v);
-          }
-
-          if(this->pm == -1 && vs >= v){
-            hs_00.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hs_00.v);
-            hs_01.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hs_01.v);
-            hs_02.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hs_02.v);
-            hs_10.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hs_10.v);
-            hs_11.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hs_11.v);
-            hs_12.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hs_12.v);
-          }
-
-          Simd p_00 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_00)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_00)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_00);
-          Simd p_01 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_01)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_01)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_01);
-          Simd p_02 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_02)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_02)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_02);
-          Simd p_10 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_10)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_10)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(0)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_10);
-          Simd p_11 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_11)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_11)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(1)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_11);
-          Simd p_12 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_12)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_12)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(2)) + switcheroo<Coeff_t>::mult(u[v]()()(), hp_12);
-          Simd p_20 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_00)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_00)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_00);
-          Simd p_21 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_01)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_01)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_01);
-          Simd p_22 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_02)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_02)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_02);
-          Simd p_30 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_10)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(0)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_10)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_10);
-          Simd p_31 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_11)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(1)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_11)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_11);
-          Simd p_32 = (this->pm == 1) ? switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_12)
-                                      : switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(2)) + switcheroo<Coeff_t>::mult(l[v]()()(), hm_12)
-                                                                                                 + switcheroo<Coeff_t>::mult(s[v]()()(), hs_12);
-
-          vstream(chi[ss+v]()(0)(0), p_00);
-          vstream(chi[ss+v]()(0)(1), p_01);
-          vstream(chi[ss+v]()(0)(2), p_02);
-          vstream(chi[ss+v]()(1)(0), p_10);
-          vstream(chi[ss+v]()(1)(1), p_11);
-          vstream(chi[ss+v]()(1)(2), p_12);
-          vstream(chi[ss+v]()(2)(0), p_20);
-          vstream(chi[ss+v]()(2)(1), p_21);
-          vstream(chi[ss+v]()(2)(2), p_22);
-          vstream(chi[ss+v]()(3)(0), p_30);
-          vstream(chi[ss+v]()(3)(1), p_31);
-          vstream(chi[ss+v]()(3)(2), p_32);
-
-        }
-
-      }
-
-      this->M5Dtime += usecond();
-
-    #endif
-  }
-
-  #ifdef AVX512
-    #include<simd/Intel512common.h>
-    #include<simd/Intel512avx.h>
-    #include<simd/Intel512single.h>
-  #endif
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInternalAsm(const FermionField& psi, FermionField& chi,
-    int LLs, int site, Vector<iSinglet<Simd> >& Matp, Vector<iSinglet<Simd> >& Matm)
-  {
-    #ifndef AVX512
-      {
-        SiteHalfSpinor BcastP;
-        SiteHalfSpinor BcastM;
-        SiteHalfSpinor SiteChiP;
-        SiteHalfSpinor SiteChiM;
-
-        // Ls*Ls * 2 * 12 * vol flops
-        for(int s1=0; s1<LLs; s1++){
-
-          for(int s2=0; s2<LLs; s2++){
-          for(int l=0; l < Simd::Nsimd(); l++){ // simd lane
-
-            int s = s2 + l*LLs;
-            int lex = s2 + LLs*site;
-
-            if( s2==0 && l==0 ){
-              SiteChiP=zero;
-              SiteChiM=zero;
-            }
-
-            for(int sp=0; sp<2;  sp++){
-            for(int co=0; co<Nc; co++){
-              vbroadcast(BcastP()(sp)(co), psi[lex]()(sp)(co), l);
-            }}
-
-            for(int sp=0; sp<2;  sp++){
-            for(int co=0; co<Nc; co++){
-              vbroadcast(BcastM()(sp)(co), psi[lex]()(sp+2)(co), l);
-            }}
-
-            for(int sp=0; sp<2;  sp++){
-            for(int co=0; co<Nc; co++){
-              SiteChiP()(sp)(co) = real_madd(Matp[LLs*s+s1]()()(), BcastP()(sp)(co), SiteChiP()(sp)(co)); // 1100 us.
-              SiteChiM()(sp)(co) = real_madd(Matm[LLs*s+s1]()()(), BcastM()(sp)(co), SiteChiM()(sp)(co)); // each found by commenting out
-            }}
-          }}
-
-          {
-            int lex = s1 + LLs*site;
-            for(int sp=0; sp<2;  sp++){
-            for(int co=0; co<Nc; co++){
-              vstream(chi[lex]()(sp)(co),   SiteChiP()(sp)(co));
-              vstream(chi[lex]()(sp+2)(co), SiteChiM()(sp)(co));
-            }}
-          }
-        }
-      }
-    #else
-      {
-        // pointers
-        //  MASK_REGS;
-        #define Chi_00 %%zmm1
-        #define Chi_01 %%zmm2
-        #define Chi_02 %%zmm3
-        #define Chi_10 %%zmm4
-        #define Chi_11 %%zmm5
-        #define Chi_12 %%zmm6
-        #define Chi_20 %%zmm7
-        #define Chi_21 %%zmm8
-        #define Chi_22 %%zmm9
-        #define Chi_30 %%zmm10
-        #define Chi_31 %%zmm11
-        #define Chi_32 %%zmm12
-
-        #define BCAST0  %%zmm13
-        #define BCAST1  %%zmm14
-        #define BCAST2  %%zmm15
-        #define BCAST3  %%zmm16
-        #define BCAST4  %%zmm17
-        #define BCAST5  %%zmm18
-        #define BCAST6  %%zmm19
-        #define BCAST7  %%zmm20
-        #define BCAST8  %%zmm21
-        #define BCAST9  %%zmm22
-        #define BCAST10 %%zmm23
-        #define BCAST11 %%zmm24
-
-        int incr = LLs*LLs*sizeof(iSinglet<Simd>);
-
-        for(int s1=0; s1<LLs; s1++){
-
-          for(int s2=0; s2<LLs; s2++){
-
-            int lex = s2 + LLs*site;
-            uint64_t a0 = (uint64_t) &Matp[LLs*s2+s1]; // should be cacheable
-            uint64_t a1 = (uint64_t) &Matm[LLs*s2+s1];
-            uint64_t a2 = (uint64_t) &psi[lex];
-
-            for(int l=0; l<Simd::Nsimd(); l++){ // simd lane
-
-              if((s2+l)==0) {
-                asm(
-                      VPREFETCH1(0,%2)              VPREFETCH1(0,%1)
-                      VPREFETCH1(12,%2)  	          VPREFETCH1(13,%2)
-                      VPREFETCH1(14,%2)  	          VPREFETCH1(15,%2)
-                      VBCASTCDUP(0,%2,BCAST0)
-                      VBCASTCDUP(1,%2,BCAST1)
-                      VBCASTCDUP(2,%2,BCAST2)
-                      VBCASTCDUP(3,%2,BCAST3)
-                      VBCASTCDUP(4,%2,BCAST4)       VMULMEM(0,%0,BCAST0,Chi_00)
-                      VBCASTCDUP(5,%2,BCAST5)       VMULMEM(0,%0,BCAST1,Chi_01)
-                      VBCASTCDUP(6,%2,BCAST6)       VMULMEM(0,%0,BCAST2,Chi_02)
-                      VBCASTCDUP(7,%2,BCAST7)       VMULMEM(0,%0,BCAST3,Chi_10)
-                      VBCASTCDUP(8,%2,BCAST8)       VMULMEM(0,%0,BCAST4,Chi_11)
-                      VBCASTCDUP(9,%2,BCAST9)       VMULMEM(0,%0,BCAST5,Chi_12)
-                      VBCASTCDUP(10,%2,BCAST10)     VMULMEM(0,%1,BCAST6,Chi_20)
-                      VBCASTCDUP(11,%2,BCAST11)     VMULMEM(0,%1,BCAST7,Chi_21)
-                      VMULMEM(0,%1,BCAST8,Chi_22)
-                      VMULMEM(0,%1,BCAST9,Chi_30)
-                      VMULMEM(0,%1,BCAST10,Chi_31)
-                      VMULMEM(0,%1,BCAST11,Chi_32)
-                      : : "r" (a0), "r" (a1), "r" (a2)                            );
-              } else {
-                asm(
-                      VBCASTCDUP(0,%2,BCAST0)   VMADDMEM(0,%0,BCAST0,Chi_00)
-                      VBCASTCDUP(1,%2,BCAST1)   VMADDMEM(0,%0,BCAST1,Chi_01)
-                      VBCASTCDUP(2,%2,BCAST2)   VMADDMEM(0,%0,BCAST2,Chi_02)
-                      VBCASTCDUP(3,%2,BCAST3)   VMADDMEM(0,%0,BCAST3,Chi_10)
-                      VBCASTCDUP(4,%2,BCAST4)   VMADDMEM(0,%0,BCAST4,Chi_11)
-                      VBCASTCDUP(5,%2,BCAST5)   VMADDMEM(0,%0,BCAST5,Chi_12)
-                      VBCASTCDUP(6,%2,BCAST6)   VMADDMEM(0,%1,BCAST6,Chi_20)
-                      VBCASTCDUP(7,%2,BCAST7)   VMADDMEM(0,%1,BCAST7,Chi_21)
-                      VBCASTCDUP(8,%2,BCAST8)   VMADDMEM(0,%1,BCAST8,Chi_22)
-                      VBCASTCDUP(9,%2,BCAST9)   VMADDMEM(0,%1,BCAST9,Chi_30)
-                      VBCASTCDUP(10,%2,BCAST10) VMADDMEM(0,%1,BCAST10,Chi_31)
-                      VBCASTCDUP(11,%2,BCAST11) VMADDMEM(0,%1,BCAST11,Chi_32)
-                      : : "r" (a0), "r" (a1), "r" (a2)                            );
-              }
-
-              a0 = a0 + incr;
-              a1 = a1 + incr;
-              a2 = a2 + sizeof(typename Simd::scalar_type);
-            }
-          }
-
-          {
-            int lexa = s1+LLs*site;
-            asm (
-               VSTORE(0,%0,Chi_00) VSTORE(1 ,%0,Chi_01)  VSTORE(2 ,%0,Chi_02)
-               VSTORE(3,%0,Chi_10) VSTORE(4 ,%0,Chi_11)  VSTORE(5 ,%0,Chi_12)
-               VSTORE(6,%0,Chi_20) VSTORE(7 ,%0,Chi_21)  VSTORE(8 ,%0,Chi_22)
-               VSTORE(9,%0,Chi_30) VSTORE(10,%0,Chi_31)  VSTORE(11,%0,Chi_32)
-               : : "r" ((uint64_t)&chi[lexa]) : "memory" );
-          }
-        }
-      }
-
-      #undef Chi_00
-      #undef Chi_01
-      #undef Chi_02
-      #undef Chi_10
-      #undef Chi_11
-      #undef Chi_12
-      #undef Chi_20
-      #undef Chi_21
-      #undef Chi_22
-      #undef Chi_30
-      #undef Chi_31
-      #undef Chi_32
-
-      #undef BCAST0
-      #undef BCAST1
-      #undef BCAST2
-      #undef BCAST3
-      #undef BCAST4
-      #undef BCAST5
-      #undef BCAST6
-      #undef BCAST7
-      #undef BCAST8
-      #undef BCAST9
-      #undef BCAST10
-      #undef BCAST11
-
-    #endif
-  };
-
-  // Z-mobius version
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInternalZAsm(const FermionField& psi, FermionField& chi,
-    int LLs, int site, Vector<iSinglet<Simd> >& Matp, Vector<iSinglet<Simd> >& Matm)
-  {
-    std::cout << "Error: zMobius not implemented for EOFA" << std::endl;
-    exit(-1);
-  };
-
-  template<class Impl>
-  void MobiusEOFAFermion<Impl>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv)
-  {
-    int Ls  = this->Ls;
-    int LLs = psi._grid->_rdimensions[0];
-    int vol = psi._grid->oSites()/LLs;
-
-    chi.checkerboard = psi.checkerboard;
-
-    Vector<iSinglet<Simd>>   Matp;
-    Vector<iSinglet<Simd>>   Matm;
-    Vector<iSinglet<Simd>>* _Matp;
-    Vector<iSinglet<Simd>>* _Matm;
-
-    //  MooeeInternalCompute(dag,inv,Matp,Matm);
-    if(inv && dag){
-      _Matp = &this->MatpInvDag;
-      _Matm = &this->MatmInvDag;
-    }
-
-    if(inv && (!dag)){
-      _Matp = &this->MatpInv;
-      _Matm = &this->MatmInv;
-    }
-
-    if(!inv){
-      MooeeInternalCompute(dag, inv, Matp, Matm);
-      _Matp = &Matp;
-      _Matm = &Matm;
-    }
-
-    assert(_Matp->size() == Ls*LLs);
-
-    this->MooeeInvCalls++;
-    this->MooeeInvTime -= usecond();
-
-    if(switcheroo<Coeff_t>::iscomplex()){
-      parallel_for(auto site=0; site<vol; site++){
-        MooeeInternalZAsm(psi, chi, LLs, site, *_Matp, *_Matm);
-      }
-    } else {
-      parallel_for(auto site=0; site<vol; site++){
-        MooeeInternalAsm(psi, chi, LLs, site, *_Matp, *_Matm);
-      }
-    }
-
-    this->MooeeInvTime += usecond();
-  }
-
-  #ifdef MOBIUS_EOFA_DPERP_VEC
-
-    INSTANTIATE_DPERP_MOBIUS_EOFA(DomainWallVec5dImplD);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(DomainWallVec5dImplF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZDomainWallVec5dImplD);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZDomainWallVec5dImplF);
-
-    INSTANTIATE_DPERP_MOBIUS_EOFA(DomainWallVec5dImplDF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(DomainWallVec5dImplFH);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZDomainWallVec5dImplDF);
-    INSTANTIATE_DPERP_MOBIUS_EOFA(ZDomainWallVec5dImplFH);
-
-    template void MobiusEOFAFermion<DomainWallVec5dImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<DomainWallVec5dImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<ZDomainWallVec5dImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<ZDomainWallVec5dImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-
-    template void MobiusEOFAFermion<DomainWallVec5dImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<DomainWallVec5dImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<ZDomainWallVec5dImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-    template void MobiusEOFAFermion<ZDomainWallVec5dImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
-
-  #endif
-
-}}
diff --git a/Grid/qcd/action/fermion/MobiusFermion.h b/Grid/qcd/action/fermion/MobiusFermion.h
index b61c26d5..1cbb6609 100644
--- a/Grid/qcd/action/fermion/MobiusFermion.h
+++ b/Grid/qcd/action/fermion/MobiusFermion.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,57 +24,54 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef  GRID_QCD_MOBIUS_FERMION_H
 #define  GRID_QCD_MOBIUS_FERMION_H
 
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class MobiusFermion : public CayleyFermion5D<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+public:
 
-    template<class Impl>
-    class MobiusFermion : public CayleyFermion5D<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-    public:
-
-      virtual void   Instantiatable(void) {};
-      // Constructors
-      MobiusFermion(GaugeField &_Umu,
-		    GridCartesian         &FiveDimGrid,
-		    GridRedBlackCartesian &FiveDimRedBlackGrid,
-		    GridCartesian         &FourDimGrid,
-		    GridRedBlackCartesian &FourDimRedBlackGrid,
-		    RealD _mass,RealD _M5,
-		    RealD b, RealD c,const ImplParams &p= ImplParams()) : 
+  virtual void   Instantiatable(void) {};
+  // Constructors
+  MobiusFermion(GaugeField &_Umu,
+		GridCartesian         &FiveDimGrid,
+		GridRedBlackCartesian &FiveDimRedBlackGrid,
+		GridCartesian         &FourDimGrid,
+		GridRedBlackCartesian &FourDimRedBlackGrid,
+		RealD _mass,RealD _M5,
+		RealD b, RealD c,const ImplParams &p= ImplParams()) : 
       
-      CayleyFermion5D<Impl>(_Umu,
-			    FiveDimGrid,
-			    FiveDimRedBlackGrid,
-			    FourDimGrid,
-			    FourDimRedBlackGrid,_mass,_M5,p)
+    CayleyFermion5D<Impl>(_Umu,
+			  FiveDimGrid,
+			  FiveDimRedBlackGrid,
+			  FourDimGrid,
+			  FourDimRedBlackGrid,_mass,_M5,p)
 
-      {
-	RealD eps = 1.0;
+  {
+    RealD eps = 1.0;
 
-	std::cout<<GridLogMessage << "MobiusFermion (b="<<b<<",c="<<c<<") with Ls= "<<this->Ls<<" Tanh approx"<<std::endl;
-	Approx::zolotarev_data *zdata = Approx::higham(eps,this->Ls);// eps is ignored for higham
-	assert(zdata->n==this->Ls);
+    std::cout<<GridLogMessage << "MobiusFermion (b="<<b<<",c="<<c<<") with Ls= "<<this->Ls<<" Tanh approx"<<std::endl;
+    Approx::zolotarev_data *zdata = Approx::higham(eps,this->Ls);// eps is ignored for higham
+    assert(zdata->n==this->Ls);
 	
-	// Call base setter
-	this->SetCoefficientsTanh(zdata,b,c);
+    // Call base setter
+    this->SetCoefficientsTanh(zdata,b,c);
 
-	Approx::zolotarev_free(zdata);
+    Approx::zolotarev_free(zdata);
  
-      }
-
-    };
-
   }
-}
+
+};
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/fermion/MobiusZolotarevFermion.h b/Grid/qcd/action/fermion/MobiusZolotarevFermion.h
index 078d4f3e..48496773 100644
--- a/Grid/qcd/action/fermion/MobiusZolotarevFermion.h
+++ b/Grid/qcd/action/fermion/MobiusZolotarevFermion.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,58 +24,55 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef  GRID_QCD_MOBIUS_ZOLOTAREV_FERMION_H
 #define  GRID_QCD_MOBIUS_ZOLOTAREV_FERMION_H
 
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class MobiusZolotarevFermion : public CayleyFermion5D<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+public:
 
-    template<class Impl>
-    class MobiusZolotarevFermion : public CayleyFermion5D<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-    public:
-
-      virtual void   Instantiatable(void) {};
-      // Constructors
-       MobiusZolotarevFermion(GaugeField &_Umu,
-			      GridCartesian         &FiveDimGrid,
-			      GridRedBlackCartesian &FiveDimRedBlackGrid,
-			      GridCartesian         &FourDimGrid,
-			      GridRedBlackCartesian &FourDimRedBlackGrid,
-			      RealD _mass,RealD _M5,
-			      RealD b, RealD c,
-			      RealD lo, RealD hi,const ImplParams &p= ImplParams()) : 
+  virtual void   Instantiatable(void) {};
+  // Constructors
+  MobiusZolotarevFermion(GaugeField &_Umu,
+			 GridCartesian         &FiveDimGrid,
+			 GridRedBlackCartesian &FiveDimRedBlackGrid,
+			 GridCartesian         &FourDimGrid,
+			 GridRedBlackCartesian &FourDimRedBlackGrid,
+			 RealD _mass,RealD _M5,
+			 RealD b, RealD c,
+			 RealD lo, RealD hi,const ImplParams &p= ImplParams()) : 
       
-      CayleyFermion5D<Impl>(_Umu,
-			    FiveDimGrid,
-			    FiveDimRedBlackGrid,
-			    FourDimGrid,
-			    FourDimRedBlackGrid,_mass,_M5,p)
+    CayleyFermion5D<Impl>(_Umu,
+			  FiveDimGrid,
+			  FiveDimRedBlackGrid,
+			  FourDimGrid,
+			  FourDimRedBlackGrid,_mass,_M5,p)
 
-      {
-	RealD eps = lo/hi;
+  {
+    RealD eps = lo/hi;
 
-	Approx::zolotarev_data *zdata = Approx::zolotarev(eps,this->Ls,0);
-	assert(zdata->n==this->Ls);
+    Approx::zolotarev_data *zdata = Approx::zolotarev(eps,this->Ls,0);
+    assert(zdata->n==this->Ls);
 
-	std::cout<<GridLogMessage << "MobiusZolotarevFermion (b="<<b<<",c="<<c<<") with Ls= "<<this->Ls<<" Zolotarev range ["<<lo<<","<<hi<<"]"<<std::endl;
+    std::cout<<GridLogMessage << "MobiusZolotarevFermion (b="<<b<<",c="<<c<<") with Ls= "<<this->Ls<<" Zolotarev range ["<<lo<<","<<hi<<"]"<<std::endl;
 	
-	// Call base setter
-	this->SetCoefficientsZolotarev(hi,zdata,b,c);
+    // Call base setter
+    this->SetCoefficientsZolotarev(hi,zdata,b,c);
  
-	Approx::zolotarev_free(zdata);
-      }
-
-    };
-
+    Approx::zolotarev_free(zdata);
   }
-}
+
+};
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/fermion/OverlapWilsonCayleyTanhFermion.h b/Grid/qcd/action/fermion/OverlapWilsonCayleyTanhFermion.h
index fd7d74df..350e89e2 100644
--- a/Grid/qcd/action/fermion/OverlapWilsonCayleyTanhFermion.h
+++ b/Grid/qcd/action/fermion/OverlapWilsonCayleyTanhFermion.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,46 +24,44 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef OVERLAP_WILSON_CAYLEY_TANH_FERMION_H
 #define OVERLAP_WILSON_CAYLEY_TANH_FERMION_H
 
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
-
-    template<class Impl>
-    class OverlapWilsonCayleyTanhFermion : public MobiusFermion<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-    public:
+template<class Impl>
+class OverlapWilsonCayleyTanhFermion : public MobiusFermion<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+public:
 
      void  MomentumSpacePropagator(FermionField &out,const FermionField &in,RealD _m,std::vector<double> twist) {
        this->MomentumSpacePropagatorHw(out,in,_m,twist);
-     };
+  };
 
-     // Constructors
-    OverlapWilsonCayleyTanhFermion(GaugeField &_Umu,
-				   GridCartesian         &FiveDimGrid,
-				   GridRedBlackCartesian &FiveDimRedBlackGrid,
-				   GridCartesian         &FourDimGrid,
-				   GridRedBlackCartesian &FourDimRedBlackGrid,
-				   RealD _mass,RealD _M5,
-				   RealD scale,const ImplParams &p= ImplParams()) :
+  // Constructors
+  OverlapWilsonCayleyTanhFermion(GaugeField &_Umu,
+				 GridCartesian         &FiveDimGrid,
+				 GridRedBlackCartesian &FiveDimRedBlackGrid,
+				 GridCartesian         &FourDimGrid,
+				 GridRedBlackCartesian &FourDimRedBlackGrid,
+				 RealD _mass,RealD _M5,
+				 RealD scale,const ImplParams &p= ImplParams()) :
       
-      // b+c=scale, b-c = 0 <=> b =c = scale/2
-      MobiusFermion<Impl>(_Umu,
-			  FiveDimGrid,
-			  FiveDimRedBlackGrid,
-			  FourDimGrid,
-			  FourDimRedBlackGrid,_mass,_M5,0.5*scale,0.5*scale,p)
-	{
-	}
-    };
+    // b+c=scale, b-c = 0 <=> b =c = scale/2
+    MobiusFermion<Impl>(_Umu,
+			FiveDimGrid,
+			FiveDimRedBlackGrid,
+			FourDimGrid,
+			FourDimRedBlackGrid,_mass,_M5,0.5*scale,0.5*scale,p)
+  {
   }
-}
+};
+
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/qcd/action/fermion/OverlapWilsonCayleyZolotarevFermion.h b/Grid/qcd/action/fermion/OverlapWilsonCayleyZolotarevFermion.h
index 4f1adbbf..d15690fa 100644
--- a/Grid/qcd/action/fermion/OverlapWilsonCayleyZolotarevFermion.h
+++ b/Grid/qcd/action/fermion/OverlapWilsonCayleyZolotarevFermion.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,45 +24,42 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef  OVERLAP_WILSON_CAYLEY_ZOLOTAREV_FERMION_H
 #define  OVERLAP_WILSON_CAYLEY_ZOLOTAREV_FERMION_H
 
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class OverlapWilsonCayleyZolotarevFermion : public MobiusZolotarevFermion<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+public:
 
-    template<class Impl>
-    class OverlapWilsonCayleyZolotarevFermion : public MobiusZolotarevFermion<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-    public:
+  // Constructors
 
-      // Constructors
+  OverlapWilsonCayleyZolotarevFermion(GaugeField &_Umu,
+				      GridCartesian         &FiveDimGrid,
+				      GridRedBlackCartesian &FiveDimRedBlackGrid,
+				      GridCartesian         &FourDimGrid,
+				      GridRedBlackCartesian &FourDimRedBlackGrid,
+				      RealD _mass,RealD _M5,
+				      RealD lo, RealD hi,const ImplParams &p= ImplParams()) : 
+    // b+c=1.0, b-c = 0 <=> b =c = 1/2
+    MobiusZolotarevFermion<Impl>(_Umu,
+				 FiveDimGrid,
+				 FiveDimRedBlackGrid,
+				 FourDimGrid,
+				 FourDimRedBlackGrid,_mass,_M5,0.5,0.5,lo,hi,p)
 
-    OverlapWilsonCayleyZolotarevFermion(GaugeField &_Umu,
-					GridCartesian         &FiveDimGrid,
-					GridRedBlackCartesian &FiveDimRedBlackGrid,
-					GridCartesian         &FourDimGrid,
-					GridRedBlackCartesian &FourDimRedBlackGrid,
-					RealD _mass,RealD _M5,
-					RealD lo, RealD hi,const ImplParams &p= ImplParams()) : 
-      // b+c=1.0, b-c = 0 <=> b =c = 1/2
-      MobiusZolotarevFermion<Impl>(_Umu,
-				   FiveDimGrid,
-				   FiveDimRedBlackGrid,
-				   FourDimGrid,
-				   FourDimRedBlackGrid,_mass,_M5,0.5,0.5,lo,hi,p)
+  {}
 
-      {}
+};
 
-    };
-
-  }
-}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/fermion/OverlapWilsonContfracTanhFermion.h b/Grid/qcd/action/fermion/OverlapWilsonContfracTanhFermion.h
index 38d0fda2..9d1a9a86 100644
--- a/Grid/qcd/action/fermion/OverlapWilsonContfracTanhFermion.h
+++ b/Grid/qcd/action/fermion/OverlapWilsonContfracTanhFermion.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,48 +24,47 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef OVERLAP_WILSON_CONTFRAC_TANH_FERMION_H
 #define OVERLAP_WILSON_CONTFRAC_TANH_FERMION_H
 
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class OverlapWilsonContFracTanhFermion : public ContinuedFractionFermion5D<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+public:
 
-    template<class Impl>
-    class OverlapWilsonContFracTanhFermion : public ContinuedFractionFermion5D<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-    public:
-
-      virtual void   Instantiatable(void){};
-      // Constructors
-    OverlapWilsonContFracTanhFermion(GaugeField &_Umu,
-				     GridCartesian         &FiveDimGrid,
-				     GridRedBlackCartesian &FiveDimRedBlackGrid,
-				     GridCartesian         &FourDimGrid,
-				     GridRedBlackCartesian &FourDimRedBlackGrid,
-				     RealD _mass,RealD _M5,
-				     RealD scale,const ImplParams &p= ImplParams()) :
+  virtual void   Instantiatable(void){};
+  // Constructors
+  OverlapWilsonContFracTanhFermion(GaugeField &_Umu,
+				   GridCartesian         &FiveDimGrid,
+				   GridRedBlackCartesian &FiveDimRedBlackGrid,
+				   GridCartesian         &FourDimGrid,
+				   GridRedBlackCartesian &FourDimRedBlackGrid,
+				   RealD _mass,RealD _M5,
+				   RealD scale,const ImplParams &p= ImplParams()) :
       
-      // b+c=scale, b-c = 0 <=> b =c = scale/2
-      ContinuedFractionFermion5D<Impl>(_Umu,
-				       FiveDimGrid,
-				       FiveDimRedBlackGrid,
-				       FourDimGrid,
-				       FourDimRedBlackGrid,_mass,_M5,p)
-	{
-	  assert((this->Ls&0x1)==1); // Odd Ls required
-	  int nrational=this->Ls-1;// Even rational order
-	  Approx::zolotarev_data *zdata = Approx::higham(1.0,nrational);// eps is ignored for higham
-	  this->SetCoefficientsTanh(zdata,scale);
-	  Approx::zolotarev_free(zdata);
-	}
-    };
+    // b+c=scale, b-c = 0 <=> b =c = scale/2
+    ContinuedFractionFermion5D<Impl>(_Umu,
+				     FiveDimGrid,
+				     FiveDimRedBlackGrid,
+				     FourDimGrid,
+				     FourDimRedBlackGrid,_mass,_M5,p)
+  {
+    assert((this->Ls&0x1)==1); // Odd Ls required
+    int nrational=this->Ls-1;// Even rational order
+    Approx::zolotarev_data *zdata = Approx::higham(1.0,nrational);// eps is ignored for higham
+    this->SetCoefficientsTanh(zdata,scale);
+    Approx::zolotarev_free(zdata);
   }
-}
+};
+
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/qcd/action/fermion/OverlapWilsonContfracZolotarevFermion.h b/Grid/qcd/action/fermion/OverlapWilsonContfracZolotarevFermion.h
index 6773b4d2..ce796d4a 100644
--- a/Grid/qcd/action/fermion/OverlapWilsonContfracZolotarevFermion.h
+++ b/Grid/qcd/action/fermion/OverlapWilsonContfracZolotarevFermion.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,51 +24,49 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef OVERLAP_WILSON_CONTFRAC_ZOLOTAREV_FERMION_H
 #define OVERLAP_WILSON_CONTFRAC_ZOLOTAREV_FERMION_H
 
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class OverlapWilsonContFracZolotarevFermion : public ContinuedFractionFermion5D<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
 
-    template<class Impl>
-    class OverlapWilsonContFracZolotarevFermion : public ContinuedFractionFermion5D<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-
-      virtual void   Instantiatable(void){};
-      // Constructors
-    OverlapWilsonContFracZolotarevFermion(GaugeField &_Umu,
-					  GridCartesian         &FiveDimGrid,
-					  GridRedBlackCartesian &FiveDimRedBlackGrid,
-					  GridCartesian         &FourDimGrid,
-					  GridRedBlackCartesian &FourDimRedBlackGrid,
-					  RealD _mass,RealD _M5,
-					  RealD lo,RealD hi,const ImplParams &p= ImplParams()):
+  virtual void   Instantiatable(void){};
+  // Constructors
+  OverlapWilsonContFracZolotarevFermion(GaugeField &_Umu,
+					GridCartesian         &FiveDimGrid,
+					GridRedBlackCartesian &FiveDimRedBlackGrid,
+					GridCartesian         &FourDimGrid,
+					GridRedBlackCartesian &FourDimRedBlackGrid,
+					RealD _mass,RealD _M5,
+					RealD lo,RealD hi,const ImplParams &p= ImplParams()):
       
-      // b+c=scale, b-c = 0 <=> b =c = scale/2
-      ContinuedFractionFermion5D<Impl>(_Umu,
-				       FiveDimGrid,
-				       FiveDimRedBlackGrid,
-				       FourDimGrid,
-				       FourDimRedBlackGrid,_mass,_M5,p)
-	{
-	  assert((this->Ls&0x1)==1); // Odd Ls required
+    // b+c=scale, b-c = 0 <=> b =c = scale/2
+    ContinuedFractionFermion5D<Impl>(_Umu,
+				     FiveDimGrid,
+				     FiveDimRedBlackGrid,
+				     FourDimGrid,
+				     FourDimRedBlackGrid,_mass,_M5,p)
+  {
+    assert((this->Ls&0x1)==1); // Odd Ls required
 
-	  int nrational=this->Ls;// Odd rational order
-	  RealD eps = lo/hi;
+    int nrational=this->Ls;// Odd rational order
+    RealD eps = lo/hi;
 
-	  Approx::zolotarev_data *zdata = Approx::zolotarev(eps,nrational,0);
-	  this->SetCoefficientsZolotarev(hi,zdata);
-	  Approx::zolotarev_free(zdata);
+    Approx::zolotarev_data *zdata = Approx::zolotarev(eps,nrational,0);
+    this->SetCoefficientsZolotarev(hi,zdata);
+    Approx::zolotarev_free(zdata);
 
-	}
-    };
   }
-}
+};
+
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/qcd/action/fermion/OverlapWilsonPartialFractionTanhFermion.h b/Grid/qcd/action/fermion/OverlapWilsonPartialFractionTanhFermion.h
index 84c4f597..f2fb46cd 100644
--- a/Grid/qcd/action/fermion/OverlapWilsonPartialFractionTanhFermion.h
+++ b/Grid/qcd/action/fermion/OverlapWilsonPartialFractionTanhFermion.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,48 +24,46 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef OVERLAP_WILSON_PARTFRAC_TANH_FERMION_H
 #define OVERLAP_WILSON_PARTFRAC_TANH_FERMION_H
 
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class OverlapWilsonPartialFractionTanhFermion : public PartialFractionFermion5D<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+public:
 
-    template<class Impl>
-    class OverlapWilsonPartialFractionTanhFermion : public PartialFractionFermion5D<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-    public:
-
-      virtual void   Instantiatable(void){};
-      // Constructors
-    OverlapWilsonPartialFractionTanhFermion(GaugeField &_Umu,
-					    GridCartesian         &FiveDimGrid,
-					    GridRedBlackCartesian &FiveDimRedBlackGrid,
-					    GridCartesian         &FourDimGrid,
-					    GridRedBlackCartesian &FourDimRedBlackGrid,
-					    RealD _mass,RealD _M5,
-					    RealD scale,const ImplParams &p= ImplParams()) :
+  virtual void   Instantiatable(void){};
+  // Constructors
+  OverlapWilsonPartialFractionTanhFermion(GaugeField &_Umu,
+					  GridCartesian         &FiveDimGrid,
+					  GridRedBlackCartesian &FiveDimRedBlackGrid,
+					  GridCartesian         &FourDimGrid,
+					  GridRedBlackCartesian &FourDimRedBlackGrid,
+					  RealD _mass,RealD _M5,
+					  RealD scale,const ImplParams &p= ImplParams()) :
       
-      // b+c=scale, b-c = 0 <=> b =c = scale/2
-      PartialFractionFermion5D<Impl>(_Umu,
-				     FiveDimGrid,
-				     FiveDimRedBlackGrid,
-				     FourDimGrid,
-				     FourDimRedBlackGrid,_mass,_M5,p)
-	{
-	  assert((this->Ls&0x1)==1); // Odd Ls required
-	  int nrational=this->Ls-1;// Even rational order
-	  Approx::zolotarev_data *zdata = Approx::higham(1.0,nrational);// eps is ignored for higham
-	  this->SetCoefficientsTanh(zdata,scale);
-	  Approx::zolotarev_free(zdata);
-	}
-    };
+    // b+c=scale, b-c = 0 <=> b =c = scale/2
+    PartialFractionFermion5D<Impl>(_Umu,
+				   FiveDimGrid,
+				   FiveDimRedBlackGrid,
+				   FourDimGrid,
+				   FourDimRedBlackGrid,_mass,_M5,p)
+  {
+    assert((this->Ls&0x1)==1); // Odd Ls required
+    int nrational=this->Ls-1;// Even rational order
+    Approx::zolotarev_data *zdata = Approx::higham(1.0,nrational);// eps is ignored for higham
+    this->SetCoefficientsTanh(zdata,scale);
+    Approx::zolotarev_free(zdata);
   }
-}
+};
+
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/qcd/action/fermion/OverlapWilsonPartialFractionZolotarevFermion.h b/Grid/qcd/action/fermion/OverlapWilsonPartialFractionZolotarevFermion.h
index dc275852..f98b64a9 100644
--- a/Grid/qcd/action/fermion/OverlapWilsonPartialFractionZolotarevFermion.h
+++ b/Grid/qcd/action/fermion/OverlapWilsonPartialFractionZolotarevFermion.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,51 +24,50 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef OVERLAP_WILSON_PARTFRAC_ZOLOTAREV_FERMION_H
 #define OVERLAP_WILSON_PARTFRAC_ZOLOTAREV_FERMION_H
 
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class OverlapWilsonPartialFractionZolotarevFermion : public PartialFractionFermion5D<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
 
-    template<class Impl>
-    class OverlapWilsonPartialFractionZolotarevFermion : public PartialFractionFermion5D<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-
-      virtual void   Instantiatable(void){};
-      // Constructors
-    OverlapWilsonPartialFractionZolotarevFermion(GaugeField &_Umu,
-						 GridCartesian         &FiveDimGrid,
-						 GridRedBlackCartesian &FiveDimRedBlackGrid,
-						 GridCartesian         &FourDimGrid,
-						 GridRedBlackCartesian &FourDimRedBlackGrid,
-						 RealD _mass,RealD _M5,
-						 RealD lo,RealD hi,const ImplParams &p= ImplParams()):
+  virtual void   Instantiatable(void){};
+  // Constructors
+  OverlapWilsonPartialFractionZolotarevFermion(GaugeField &_Umu,
+					       GridCartesian         &FiveDimGrid,
+					       GridRedBlackCartesian &FiveDimRedBlackGrid,
+					       GridCartesian         &FourDimGrid,
+					       GridRedBlackCartesian &FourDimRedBlackGrid,
+					       RealD _mass,RealD _M5,
+					       RealD lo,RealD hi,const ImplParams &p= ImplParams()):
       
-      // b+c=scale, b-c = 0 <=> b =c = scale/2
-      PartialFractionFermion5D<Impl>(_Umu,
-				     FiveDimGrid,
-				     FiveDimRedBlackGrid,
-				     FourDimGrid,
-				     FourDimRedBlackGrid,_mass,_M5,p)
-	{
-	  assert((this->Ls&0x1)==1); // Odd Ls required
+    // b+c=scale, b-c = 0 <=> b =c = scale/2
+    PartialFractionFermion5D<Impl>(_Umu,
+				   FiveDimGrid,
+				   FiveDimRedBlackGrid,
+				   FourDimGrid,
+				   FourDimRedBlackGrid,_mass,_M5,p)
+  {
+    assert((this->Ls&0x1)==1); // Odd Ls required
 
-	  int nrational=this->Ls;// Odd rational order
-	  RealD eps = lo/hi;
+    int nrational=this->Ls;// Odd rational order
+    RealD eps = lo/hi;
 
-	  Approx::zolotarev_data *zdata = Approx::zolotarev(eps,nrational,0);
-	  this->SetCoefficientsZolotarev(hi,zdata);
-	  Approx::zolotarev_free(zdata);
+    Approx::zolotarev_data *zdata = Approx::zolotarev(eps,nrational,0);
+    this->SetCoefficientsZolotarev(hi,zdata);
+    Approx::zolotarev_free(zdata);
 
-	}
-    };
   }
-}
+};
+
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/qcd/action/fermion/PartialFractionFermion5D.cc b/Grid/qcd/action/fermion/PartialFractionFermion5D.cc
deleted file mode 100644
index 11840027..00000000
--- a/Grid/qcd/action/fermion/PartialFractionFermion5D.cc
+++ /dev/null
@@ -1,459 +0,0 @@
-    /*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid 
-
-    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
-
-    Copyright (C) 2015
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
-
-namespace Grid {
-  namespace QCD {
-
-
-    template<class Impl>
-    void  PartialFractionFermion5D<Impl>::Mdir (const FermionField &psi, FermionField &chi,int dir,int disp){
-      // this does both dag and undag but is trivial; make a common helper routing
-
-      int sign = 1;
-      int Ls = this->Ls;
-
-      this->DhopDir(psi,chi,dir,disp);
-
-      int nblock=(Ls-1)/2;
-      for(int b=0;b<nblock;b++){
-	int s = 2*b;
-	ag5xpby_ssp(chi,-scale,chi,0.0,chi,s,s); 
-	ag5xpby_ssp(chi, scale,chi,0.0,chi,s+1,s+1); 
-      }
-      ag5xpby_ssp(chi,p[nblock]*scale/amax,chi,0.0,chi,Ls-1,Ls-1);
-
-    }
-    template<class Impl>
-    void   PartialFractionFermion5D<Impl>::Meooe_internal(const FermionField &psi, FermionField &chi,int dag)
-    {
-      int Ls = this->Ls;
-      int sign = dag ? (-1) : 1;
-
-      if ( psi.checkerboard == Odd ) {
-	this->DhopEO(psi,chi,DaggerNo);
-      } else {
-	this->DhopOE(psi,chi,DaggerNo);
-      }
-
-      int nblock=(Ls-1)/2;
-      for(int b=0;b<nblock;b++){
-	int s = 2*b;
-	ag5xpby_ssp(chi,-scale,chi,0.0,chi,s,s); 
-	ag5xpby_ssp(chi, scale,chi,0.0,chi,s+1,s+1); 
-      }
-      ag5xpby_ssp(chi,p[nblock]*scale/amax,chi,0.0,chi,Ls-1,Ls-1);
-    }
-
-    template<class Impl>
-    void   PartialFractionFermion5D<Impl>::Mooee_internal(const FermionField &psi, FermionField &chi,int dag)
-    {
-      // again dag and undag are trivially related
-      int sign = dag ? (-1) : 1;
-      int Ls = this->Ls;
-      
-      int nblock=(Ls-1)/2;
-      for(int b=0;b<nblock;b++){
-	
-	int s = 2*b;
-	RealD pp = p[nblock-1-b];
-	RealD qq = q[nblock-1-b];
-	
-	// Do each 2x2 block aligned at s and multiplies Dw site diagonal by G5 so Hw
-	ag5xpby_ssp(chi,-dw_diag*scale,psi,amax*sqrt(qq)*scale,psi, s  ,s+1); 
-	ag5xpby_ssp(chi, dw_diag*scale,psi,amax*sqrt(qq)*scale,psi, s+1,s);
-	axpby_ssp  (chi, 1.0, chi,sqrt(amax*pp)*scale*sign,psi,s+1,Ls-1);
-      }
-      
-      {
-	RealD R=(1+mass)/(1-mass);
-	//R g5 psi[Ls-1] + p[0] H
-	ag5xpbg5y_ssp(chi,R*scale,psi,p[nblock]*scale*dw_diag/amax,psi,Ls-1,Ls-1);
-	
-	for(int b=0;b<nblock;b++){
-	  int s = 2*b+1;
-	  RealD pp = p[nblock-1-b];
-	  axpby_ssp(chi,1.0,chi,-sqrt(amax*pp)*scale*sign,psi,Ls-1,s);
-	}
-      }
-    }
-
-    template<class Impl>
-    void   PartialFractionFermion5D<Impl>::MooeeInv_internal(const FermionField &psi, FermionField &chi,int dag)
-    {
-      int sign = dag ? (-1) : 1;
-      int Ls = this->Ls;
-
-      FermionField tmp(psi._grid);
-      
-      ///////////////////////////////////////////////////////////////////////////////////////
-      //Linv
-      ///////////////////////////////////////////////////////////////////////////////////////
-      int nblock=(Ls-1)/2;
-
-      axpy(chi,0.0,psi,psi); // Identity piece
-      
-      for(int b=0;b<nblock;b++){
-	int s = 2*b;
-	RealD pp = p[nblock-1-b];
-	RealD qq = q[nblock-1-b];
-	RealD coeff1=sign*sqrt(amax*amax*amax*pp*qq) / ( dw_diag*dw_diag + amax*amax* qq);
-	RealD coeff2=sign*sqrt(amax*pp)*dw_diag / ( dw_diag*dw_diag + amax*amax* qq); // Implicit g5 here
-	axpby_ssp  (chi,1.0,chi,coeff1,psi,Ls-1,s);
-	axpbg5y_ssp(chi,1.0,chi,coeff2,psi,Ls-1,s+1);
-      }
-      
-      ///////////////////////////////////////////////////////////////////////////////////////
-      //Dinv (note D isn't really diagonal -- just diagonal enough that we can still invert)
-      // Compute Seeinv (coeff of gamma5)
-      ///////////////////////////////////////////////////////////////////////////////////////
-      RealD R=(1+mass)/(1-mass);
-      RealD Seeinv = R + p[nblock]*dw_diag/amax;
-      for(int b=0;b<nblock;b++){
-	Seeinv += p[nblock-1-b]*dw_diag/amax / ( dw_diag*dw_diag/amax/amax + q[nblock-1-b]);
-      }    
-      Seeinv = 1.0/Seeinv;
-      
-      for(int b=0;b<nblock;b++){
-	int s = 2*b;
-	RealD pp = p[nblock-1-b];
-	RealD qq = q[nblock-1-b];
-	RealD coeff1=dw_diag / ( dw_diag*dw_diag + amax*amax* qq); // Implicit g5 here
-	RealD coeff2=amax*sqrt(qq) / ( dw_diag*dw_diag + amax*amax* qq);
-	ag5xpby_ssp  (tmp,-coeff1,chi,coeff2,chi,s,s+1);
-	ag5xpby_ssp  (tmp, coeff1,chi,coeff2,chi,s+1,s);
-      }
-      ag5xpby_ssp  (tmp, Seeinv,chi,0.0,chi,Ls-1,Ls-1);
-      
-      ///////////////////////////////////////////////////////////////////////////////////////
-      // Uinv
-      ///////////////////////////////////////////////////////////////////////////////////////
-      for(int b=0;b<nblock;b++){
-	int s = 2*b;
-	RealD pp = p[nblock-1-b];
-	RealD qq = q[nblock-1-b];
-	RealD coeff1=-sign*sqrt(amax*amax*amax*pp*qq) / ( dw_diag*dw_diag + amax*amax* qq);
-	RealD coeff2=-sign*sqrt(amax*pp)*dw_diag / ( dw_diag*dw_diag + amax*amax* qq); // Implicit g5 here
-	axpby_ssp  (chi,1.0/scale,tmp,coeff1/scale,tmp,s,Ls-1);
-	axpbg5y_ssp(chi,1.0/scale,tmp,coeff2/scale,tmp,s+1,Ls-1);
-      }
-      axpby_ssp  (chi, 1.0/scale,tmp,0.0,tmp,Ls-1,Ls-1);
-    }
-
-    template<class Impl>
-    void   PartialFractionFermion5D<Impl>::M_internal(const FermionField &psi, FermionField &chi,int dag)
-    {
-      FermionField D(psi._grid);
-  
-      int Ls = this->Ls;
-      int sign = dag ? (-1) : 1;
-
-      // For partial frac Hw case (b5=c5=1) chroma quirkily computes
-      //
-      // Conventions for partfrac appear to be a mess.
-      // Tony's Nara lectures have
-      //
-      // BlockDiag(  H/p_i  1             | 1       )    
-      //          (  1      p_i H / q_i^2 | 0       )  
-      //           ---------------------------------
-      //           ( -1      0                | R  +p0 H  )
-      //
-      //Chroma     ( -2H    2sqrt(q_i)    |   0         )
-      //           (2 sqrt(q_i)   2H      |  2 sqrt(p_i) )
-      //           ---------------------------------
-      //           ( 0     -2 sqrt(p_i)   |  2 R gamma_5 + p0 2H
-      //
-      // Edwards/Joo/Kennedy/Wenger
-      //
-      // Here, the "beta's" selected by chroma to scale the unphysical bulk constraint fields
-      // incorporate the approx scale factor. This is obtained by propagating the
-      // scale on "H" out to the off diagonal elements as follows:
-      //
-      // BlockDiag(  H/p_i  1             | 1       ) 
-      //          (  1      p_i H / q_i^2 | 0       )  
-      //           ---------------------------------
-      //          ( -1      0                | R  + p_0 H  )
-      //
-      // becomes:
-      // BlockDiag(  H/ sp_i  1               | 1             ) 
-      //          (  1      sp_i H / s^2q_i^2 | 0             )  
-      //           ---------------------------------
-      //           ( -1      0                | R + p_0/s H   )
-      //
-      //
-      // This is implemented in Chroma by
-      //           p0' = p0/approxMax
-      //           p_i' = p_i*approxMax
-      //           q_i' = q_i*approxMax*approxMax
-      //
-      // After the equivalence transform is applied the matrix becomes
-      // 
-      //Chroma     ( -2H    sqrt(q'_i)    |   0         )
-      //           (sqrt(q'_i)   2H       |   sqrt(p'_i) )
-      //           ---------------------------------
-      //           ( 0     -sqrt(p'_i)    |  2 R gamma_5 + p'0 2H
-      //
-      //     =     ( -2H    sqrt(q_i)amax    |   0              )
-      //           (sqrt(q_i)amax   2H       |   sqrt(p_i*amax) )
-      //           ---------------------------------
-      //           ( 0     -sqrt(p_i)*amax   |  2 R gamma_5 + p0/amax 2H
-      //
-
-      this->DW(psi,D,DaggerNo); 
-
-      int nblock=(Ls-1)/2;
-      for(int b=0;b<nblock;b++){
-	
-	int s = 2*b;
-	double pp = p[nblock-1-b];
-	double qq = q[nblock-1-b];
-	
-	// Do each 2x2 block aligned at s and
-	ag5xpby_ssp(chi,-1.0*scale,D,amax*sqrt(qq)*scale,psi, s  ,s+1); // Multiplies Dw by G5 so Hw
-	ag5xpby_ssp(chi, 1.0*scale,D,amax*sqrt(qq)*scale,psi, s+1,s);
-	
-	// Pick up last column
-	axpby_ssp  (chi, 1.0, chi,sqrt(amax*pp)*scale*sign,psi,s+1,Ls-1);
-      }
-	
-      {
-	double R=(1+this->mass)/(1-this->mass);
-	//R g5 psi[Ls] + p[0] H
-	ag5xpbg5y_ssp(chi,R*scale,psi,p[nblock]*scale/amax,D,Ls-1,Ls-1);
-	
-	for(int b=0;b<nblock;b++){
-	  int s = 2*b+1;
-	  double pp = p[nblock-1-b];
-	  axpby_ssp(chi,1.0,chi,-sqrt(amax*pp)*scale*sign,psi,Ls-1,s);
-	}
-      }
-
-    }
-
-    template<class Impl>
-    RealD  PartialFractionFermion5D<Impl>::M    (const FermionField &in, FermionField &out)
-    {
-      M_internal(in,out,DaggerNo);
-      return norm2(out);
-    }
-    template<class Impl>
-    RealD  PartialFractionFermion5D<Impl>::Mdag (const FermionField &in, FermionField &out)
-    {
-      M_internal(in,out,DaggerYes);
-      return norm2(out);
-    }
-
-    template<class Impl>
-    void PartialFractionFermion5D<Impl>::Meooe       (const FermionField &in, FermionField &out)
-    {
-      Meooe_internal(in,out,DaggerNo);
-    }
-    template<class Impl>
-    void PartialFractionFermion5D<Impl>::MeooeDag    (const FermionField &in, FermionField &out)
-    {
-      Meooe_internal(in,out,DaggerYes);
-    }
-    template<class Impl>
-    void PartialFractionFermion5D<Impl>::Mooee       (const FermionField &in, FermionField &out)
-    {
-      Mooee_internal(in,out,DaggerNo);
-    }
-    template<class Impl>
-    void PartialFractionFermion5D<Impl>::MooeeDag    (const FermionField &in, FermionField &out)
-    {
-      Mooee_internal(in,out,DaggerYes);
-    }
-
-    template<class Impl>
-    void PartialFractionFermion5D<Impl>::MooeeInv    (const FermionField &in, FermionField &out)
-    {
-      MooeeInv_internal(in,out,DaggerNo);
-    }
-    template<class Impl>
-    void PartialFractionFermion5D<Impl>::MooeeInvDag (const FermionField &in, FermionField &out)
-    {
-      MooeeInv_internal(in,out,DaggerYes);
-    }
-
-
-  // force terms; five routines; default to Dhop on diagonal
-    template<class Impl>
-   void PartialFractionFermion5D<Impl>::MDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
-  {
-    int Ls = this->Ls;
-
-    FermionField D(V._grid);
-
-    int nblock=(Ls-1)/2;
-    for(int b=0;b<nblock;b++){
-      int s = 2*b;
-      ag5xpby_ssp(D,-scale,U,0.0,U,s,s); 
-      ag5xpby_ssp(D, scale,U,0.0,U,s+1,s+1); 
-    }
-    ag5xpby_ssp(D,p[nblock]*scale/amax,U,0.0,U,Ls-1,Ls-1);
-
-    this->DhopDeriv(mat,D,V,DaggerNo); 
-  };
-    template<class Impl>
-   void PartialFractionFermion5D<Impl>::MoeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
-  {
-    int Ls = this->Ls;
-
-    FermionField D(V._grid);
-
-    int nblock=(Ls-1)/2;
-    for(int b=0;b<nblock;b++){
-      int s = 2*b;
-      ag5xpby_ssp(D,-scale,U,0.0,U,s,s); 
-      ag5xpby_ssp(D, scale,U,0.0,U,s+1,s+1); 
-    }
-    ag5xpby_ssp(D,p[nblock]*scale/amax,U,0.0,U,Ls-1,Ls-1);
-
-    this->DhopDerivOE(mat,D,V,DaggerNo); 
-  };
-    template<class Impl>
-   void PartialFractionFermion5D<Impl>::MeoDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
-  {
-    int Ls = this->Ls;
-
-    FermionField D(V._grid);
-
-    int nblock=(Ls-1)/2;
-    for(int b=0;b<nblock;b++){
-      int s = 2*b;
-      ag5xpby_ssp(D,-scale,U,0.0,U,s,s); 
-      ag5xpby_ssp(D, scale,U,0.0,U,s+1,s+1); 
-    }
-    ag5xpby_ssp(D,p[nblock]*scale/amax,U,0.0,U,Ls-1,Ls-1);
-
-    this->DhopDerivEO(mat,D,V,DaggerNo); 
-  };
-
-    template<class Impl>
-    void  PartialFractionFermion5D<Impl>::SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD scale){
-      SetCoefficientsZolotarev(1.0/scale,zdata);
-    }
-    template<class Impl>
-    void  PartialFractionFermion5D<Impl>::SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata){
-
-      // check on degree matching
-      //      std::cout<<GridLogMessage << Ls << " Ls"<<std::endl;
-      //      std::cout<<GridLogMessage << zdata->n  << " - n"<<std::endl;
-      //      std::cout<<GridLogMessage << zdata->da << " -da "<<std::endl;
-      //      std::cout<<GridLogMessage << zdata->db << " -db"<<std::endl;
-      //      std::cout<<GridLogMessage << zdata->dn << " -dn"<<std::endl;
-      //      std::cout<<GridLogMessage << zdata->dd << " -dd"<<std::endl;
-      int Ls = this->Ls;
-
-      assert(Ls == (2*zdata->da -1) );
-
-      // Part frac
-      //      RealD R;
-      R=(1+mass)/(1-mass);
-      dw_diag = (4.0-this->M5);
-
-      //      std::vector<RealD> p; 
-      //      std::vector<RealD> q;
-      p.resize(zdata->da);
-      q.resize(zdata->dd);
-	
-      for(int n=0;n<zdata->da;n++){
-	p[n] = zdata -> alpha[n];
-      }
-      for(int n=0;n<zdata->dd;n++){
-	q[n] = -zdata -> ap[n];
-      }
-      
-      scale= part_frac_chroma_convention ? 2.0 : 1.0; // Chroma conventions annoy me
-
-      amax=zolo_hi;
-    }
-
-    template<class Impl>
-    void PartialFractionFermion5D<Impl>::ExportPhysicalFermionSolution(const FermionField &solution5d,FermionField &exported4d)
-    {
-      int Ls = this->Ls;
-      conformable(solution5d._grid,this->FermionGrid());
-      conformable(exported4d._grid,this->GaugeGrid());
-      ExtractSlice(exported4d, solution5d, Ls-1, Ls-1);
-    }
-    template<class Impl>
-    void PartialFractionFermion5D<Impl>::ImportPhysicalFermionSource(const FermionField &input4d,FermionField &imported5d)
-    {
-      int Ls = this->Ls;
-      conformable(imported5d._grid,this->FermionGrid());
-      conformable(input4d._grid   ,this->GaugeGrid());
-      FermionField tmp(this->FermionGrid());
-      tmp=zero;
-      InsertSlice(input4d, tmp, Ls-1, Ls-1);
-      tmp=Gamma(Gamma::Algebra::Gamma5)*tmp;
-      this->Dminus(tmp,imported5d);
-    }
-
-      // Constructors
-    template<class Impl>
-    PartialFractionFermion5D<Impl>::PartialFractionFermion5D(GaugeField &_Umu,
-							     GridCartesian         &FiveDimGrid,
-							     GridRedBlackCartesian &FiveDimRedBlackGrid,
-							     GridCartesian         &FourDimGrid,
-							     GridRedBlackCartesian &FourDimRedBlackGrid,
-							     RealD _mass,RealD M5,
-							     const ImplParams &p) :
-      WilsonFermion5D<Impl>(_Umu,
-			    FiveDimGrid, FiveDimRedBlackGrid,
-			    FourDimGrid, FourDimRedBlackGrid,M5,p),
-      mass(_mass)
-
-    {
-      int Ls = this->Ls;
-
-      assert((Ls&0x1)==1); // Odd Ls required
-      int nrational=Ls-1;
-
-
-      Approx::zolotarev_data *zdata = Approx::higham(1.0,nrational);
-
-      // NB: chroma uses a cast to "float" for the zolotarev range(!?).
-      // this creates a real difference in the operator which I do not like but we can replicate here
-      // to demonstrate compatibility
-      //      RealD eps = (zolo_lo / zolo_hi);
-      //      zdata = bfm_zolotarev(eps,nrational,0);
-      
-      SetCoefficientsTanh(zdata,1.0);
-
-      Approx::zolotarev_free(zdata);
-
-    }
- 
-    FermOpTemplateInstantiate(PartialFractionFermion5D);
-
- }
-}
-
diff --git a/Grid/qcd/action/fermion/PartialFractionFermion5D.h b/Grid/qcd/action/fermion/PartialFractionFermion5D.h
index 91f1bd3c..d61515f0 100644
--- a/Grid/qcd/action/fermion/PartialFractionFermion5D.h
+++ b/Grid/qcd/action/fermion/PartialFractionFermion5D.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,51 +24,49 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef  GRID_QCD_PARTIAL_FRACTION_H
 #define  GRID_QCD_PARTIAL_FRACTION_H
 
 #include <Grid/qcd/action/fermion/WilsonFermion5D.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class PartialFractionFermion5D : public WilsonFermion5D<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
 
-    template<class Impl>
-    class PartialFractionFermion5D : public WilsonFermion5D<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
+  const int part_frac_chroma_convention=1;
 
-      const int part_frac_chroma_convention=1;
+  void   Meooe_internal(const FermionField &in, FermionField &out,int dag);
+  void   Mooee_internal(const FermionField &in, FermionField &out,int dag);
+  void   MooeeInv_internal(const FermionField &in, FermionField &out,int dag);
+  void   M_internal(const FermionField &in, FermionField &out,int dag);
 
-      void   Meooe_internal(const FermionField &in, FermionField &out,int dag);
-      void   Mooee_internal(const FermionField &in, FermionField &out,int dag);
-      void   MooeeInv_internal(const FermionField &in, FermionField &out,int dag);
-      void   M_internal(const FermionField &in, FermionField &out,int dag);
+  // override multiply
+  virtual RealD  M    (const FermionField &in, FermionField &out);
+  virtual RealD  Mdag (const FermionField &in, FermionField &out);
 
-      // override multiply
-      virtual RealD  M    (const FermionField &in, FermionField &out);
-      virtual RealD  Mdag (const FermionField &in, FermionField &out);
+  // half checkerboard operaions
+  virtual void   Meooe       (const FermionField &in, FermionField &out);
+  virtual void   MeooeDag    (const FermionField &in, FermionField &out);
+  virtual void   Mooee       (const FermionField &in, FermionField &out);
+  virtual void   MooeeDag    (const FermionField &in, FermionField &out);
+  virtual void   MooeeInv    (const FermionField &in, FermionField &out);
+  virtual void   MooeeInvDag (const FermionField &in, FermionField &out);
 
-      // half checkerboard operaions
-      virtual void   Meooe       (const FermionField &in, FermionField &out);
-      virtual void   MeooeDag    (const FermionField &in, FermionField &out);
-      virtual void   Mooee       (const FermionField &in, FermionField &out);
-      virtual void   MooeeDag    (const FermionField &in, FermionField &out);
-      virtual void   MooeeInv    (const FermionField &in, FermionField &out);
-      virtual void   MooeeInvDag (const FermionField &in, FermionField &out);
+  // force terms; five routines; default to Dhop on diagonal
+  virtual void MDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  virtual void MoeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  virtual void MeoDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
 
-      // force terms; five routines; default to Dhop on diagonal
-      virtual void MDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
-      virtual void MoeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
-      virtual void MeoDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  virtual void   Instantiatable(void) =0; // ensure no make-eee
 
-      virtual void   Instantiatable(void) =0; // ensure no make-eee
-
-      // Efficient support for multigrid coarsening
-      virtual void  Mdir (const FermionField &in, FermionField &out,int dir,int disp);
+  // Efficient support for multigrid coarsening
+  virtual void  Mdir (const FermionField &in, FermionField &out,int dir,int disp);
 
       ///////////////////////////////////////////////////////////////
       // Physical surface field utilities
@@ -76,32 +74,30 @@ namespace Grid {
       virtual void ExportPhysicalFermionSolution(const FermionField &solution5d,FermionField &exported4d);
       virtual void ImportPhysicalFermionSource  (const FermionField &input4d,FermionField &imported5d);
 
-      // Constructors
-      PartialFractionFermion5D(GaugeField &_Umu,
-			       GridCartesian         &FiveDimGrid,
-			       GridRedBlackCartesian &FiveDimRedBlackGrid,
-			       GridCartesian         &FourDimGrid,
-			       GridRedBlackCartesian &FourDimRedBlackGrid,
-			       RealD _mass,RealD M5,const ImplParams &p= ImplParams());
+  // Constructors
+  PartialFractionFermion5D(GaugeField &_Umu,
+			   GridCartesian         &FiveDimGrid,
+			   GridRedBlackCartesian &FiveDimRedBlackGrid,
+			   GridCartesian         &FourDimGrid,
+			   GridRedBlackCartesian &FourDimRedBlackGrid,
+			   RealD _mass,RealD M5,const ImplParams &p= ImplParams());
 
-    protected:
+protected:
 
-      virtual void SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD scale);
-      virtual void SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata);
+  virtual void SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD scale);
+  virtual void SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata);
 
-      // Part frac
-      RealD mass;
-      RealD dw_diag;
-      RealD R;
-      RealD amax;
-      RealD scale;
-      std::vector<double> p; 
-      std::vector<double> q;
+  // Part frac
+  RealD mass;
+  RealD dw_diag;
+  RealD R;
+  RealD amax;
+  RealD scale;
+  Vector<double> p; 
+  Vector<double> q;
 
-    };
+};
 
-
-  }
-}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/fermion/PauliVillarsInverters.h b/Grid/qcd/action/fermion/PauliVillarsInverters.h
index b10640e3..7d003087 100644
--- a/Grid/qcd/action/fermion/PauliVillarsInverters.h
+++ b/Grid/qcd/action/fermion/PauliVillarsInverters.h
@@ -27,8 +27,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     /*  END LEGAL */
 #pragma once
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 template<class Field>
 class PauliVillarsSolverUnprec
@@ -90,6 +89,4 @@ class PauliVillarsSolverFourierAccel
   };
 };
 
-
-}
-}
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/Reconstruct5Dprop.h b/Grid/qcd/action/fermion/Reconstruct5Dprop.h
index 6862c5ee..93af1ab8 100644
--- a/Grid/qcd/action/fermion/Reconstruct5Dprop.h
+++ b/Grid/qcd/action/fermion/Reconstruct5Dprop.h
@@ -27,8 +27,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     /*  END LEGAL */
 #pragma once
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 template<class Field,class PVinverter> class Reconstruct5DfromPhysical {
  private:
@@ -131,5 +130,5 @@ template<class Field,class PVinverter> class Reconstruct5DfromPhysical {
   }
 };
 
-}
-}
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/ScaledShamirFermion.h b/Grid/qcd/action/fermion/ScaledShamirFermion.h
index b779b9c0..67ae8012 100644
--- a/Grid/qcd/action/fermion/ScaledShamirFermion.h
+++ b/Grid/qcd/action/fermion/ScaledShamirFermion.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,46 +24,43 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef  GRID_QCD_SCALED_SHAMIR_FERMION_H
 #define  GRID_QCD_SCALED_SHAMIR_FERMION_H
 
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class ScaledShamirFermion : public MobiusFermion<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
 
-    template<class Impl>
-    class ScaledShamirFermion : public MobiusFermion<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-
-      // Constructors
-    ScaledShamirFermion(GaugeField &_Umu,
-			GridCartesian         &FiveDimGrid,
-			GridRedBlackCartesian &FiveDimRedBlackGrid,
-			GridCartesian         &FourDimGrid,
-			GridRedBlackCartesian &FourDimRedBlackGrid,
-			RealD _mass,RealD _M5,
-//			RealD scale):
-			RealD scale,const ImplParams &p= ImplParams()) :
+  // Constructors
+  ScaledShamirFermion(GaugeField &_Umu,
+		      GridCartesian         &FiveDimGrid,
+		      GridRedBlackCartesian &FiveDimRedBlackGrid,
+		      GridCartesian         &FourDimGrid,
+		      GridRedBlackCartesian &FourDimRedBlackGrid,
+		      RealD _mass,RealD _M5,
+		      //			RealD scale):
+		      RealD scale,const ImplParams &p= ImplParams()) :
       
-      // b+c=scale, b-c = 1 <=> 2b = scale+1; 2c = scale-1
-      MobiusFermion<Impl>(_Umu,
-		    FiveDimGrid,
-		    FiveDimRedBlackGrid,
-		    FourDimGrid,
-	FourDimRedBlackGrid,_mass,_M5,0.5*(scale+1.0),0.5*(scale-1.0),p)
-//		    FourDimRedBlackGrid,_mass,_M5,0.5*(scale+1.0),0.5*(scale-1.0))
-      {
-      }
-
-    };
-
+    // b+c=scale, b-c = 1 <=> 2b = scale+1; 2c = scale-1
+    MobiusFermion<Impl>(_Umu,
+			FiveDimGrid,
+			FiveDimRedBlackGrid,
+			FourDimGrid,
+			FourDimRedBlackGrid,_mass,_M5,0.5*(scale+1.0),0.5*(scale-1.0),p)
+    //		    FourDimRedBlackGrid,_mass,_M5,0.5*(scale+1.0),0.5*(scale-1.0))
+  {
   }
-}
+
+};
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/fermion/SchurDiagTwoKappa.h b/Grid/qcd/action/fermion/SchurDiagTwoKappa.h
index 8305f98a..1545c245 100644
--- a/Grid/qcd/action/fermion/SchurDiagTwoKappa.h
+++ b/Grid/qcd/action/fermion/SchurDiagTwoKappa.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,40 +24,40 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef  _SCHUR_DIAG_TWO_KAPPA_H
-#define  _SCHUR_DIAG_TWO_KAPPA_H
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  // This is specific to (Z)mobius fermions
-  template<class Matrix, class Field>
-    class KappaSimilarityTransform {
-  public:
-    INHERIT_IMPL_TYPES(Matrix);
-    std::vector<Coeff_t> kappa, kappaDag, kappaInv, kappaInvDag;
+// This is specific to (Z)mobius fermions
+template<class Matrix, class Field>
+class KappaSimilarityTransform {
+public:
+  INHERIT_IMPL_TYPES(Matrix);
+  Vector<Coeff_t> kappa, kappaDag, kappaInv, kappaInvDag;
 
-    KappaSimilarityTransform (Matrix &zmob) {
-      for (int i=0;i<(int)zmob.bs.size();i++) {
-	Coeff_t k = 1.0 / ( 2.0 * (zmob.bs[i] *(4 - zmob.M5) + 1.0) );
-	kappa.push_back( k );
-	kappaDag.push_back( conj(k) );
-	kappaInv.push_back( 1.0 / k );
-	kappaInvDag.push_back( 1.0 / conj(k) );
-      }
+  KappaSimilarityTransform (Matrix &zmob) {
+    for (int i=0;i<(int)zmob.bs.size();i++) {
+      Coeff_t k = 1.0 / ( 2.0 * (zmob.bs[i] *(4 - zmob.M5) + 1.0) );
+      kappa.push_back( k );
+      kappaDag.push_back( conj(k) );
+      kappaInv.push_back( 1.0 / k );
+      kappaInvDag.push_back( 1.0 / conj(k) );
     }
+  }
 
   template<typename vobj>
-    void sscale(const Lattice<vobj>& in, Lattice<vobj>& out, Coeff_t* s) {
-    GridBase *grid=out._grid;
-    out.checkerboard = in.checkerboard;
+  void sscale(const Lattice<vobj>& in, Lattice<vobj>& out, Coeff_t* s) {
+    GridBase *grid=out.Grid();
+    out.Checkerboard() = in.Checkerboard();
     assert(grid->_simd_layout[0] == 1); // should be fine for ZMobius for now
     int Ls = grid->_rdimensions[0];
-    parallel_for(int ss=0;ss<grid->oSites();ss++){
-      vobj tmp = s[ss % Ls]*in._odata[ss];
-      vstream(out._odata[ss],tmp);
-    }
+    thread_for(ss, grid->oSites(),
+    {
+      vobj tmp = s[ss % Ls]*in[ss];
+      vstream(out[ss],tmp);
+    });
   }
 
   RealD sscale_norm(const Field& in, Field& out, Coeff_t* s) {
@@ -70,33 +70,33 @@ namespace Grid {
   virtual RealD MInv    (const Field& in, Field& out) { return sscale_norm(in,out,&kappaInv[0]);}
   virtual RealD MInvDag (const Field& in, Field& out) { return sscale_norm(in,out,&kappaInvDag[0]);}
 
-  };
+};
 
-  template<class Matrix,class Field>
-    class SchurDiagTwoKappaOperator :  public SchurOperatorBase<Field> {
-  public:
-    KappaSimilarityTransform<Matrix, Field> _S;
-    SchurDiagTwoOperator<Matrix, Field> _Mat;
+template<class Matrix,class Field>
+class SchurDiagTwoKappaOperator :  public SchurOperatorBase<Field> {
+public:
+  KappaSimilarityTransform<Matrix, Field> _S;
+  SchurDiagTwoOperator<Matrix, Field> _Mat;
 
-    SchurDiagTwoKappaOperator (Matrix &Mat): _S(Mat), _Mat(Mat) {};
+  SchurDiagTwoKappaOperator (Matrix &Mat): _S(Mat), _Mat(Mat) {};
 
-    virtual  RealD Mpc      (const Field &in, Field &out) {
-      Field tmp(in._grid);
+  virtual  RealD Mpc      (const Field &in, Field &out) {
+    Field tmp(in.Grid());
 
-      _S.MInv(in,out);
-      _Mat.Mpc(out,tmp);
-      return _S.M(tmp,out);
+    _S.MInv(in,out);
+    _Mat.Mpc(out,tmp);
+    return _S.M(tmp,out);
 
-    }
-    virtual  RealD MpcDag   (const Field &in, Field &out){
-      Field tmp(in._grid);
+  }
+  virtual  RealD MpcDag   (const Field &in, Field &out){
+    Field tmp(in.Grid());
 
-      _S.MDag(in,out);
-      _Mat.MpcDag(out,tmp);
-      return _S.MInvDag(tmp,out);
-    }
-  };
+    _S.MDag(in,out);
+    _Mat.MpcDag(out,tmp);
+    return _S.MInvDag(tmp,out);
+  }
+};
+
+NAMESPACE_END(Grid);
 
-}
 
-#endif
diff --git a/Grid/qcd/action/fermion/ShamirZolotarevFermion.h b/Grid/qcd/action/fermion/ShamirZolotarevFermion.h
index f9397911..54463a9f 100644
--- a/Grid/qcd/action/fermion/ShamirZolotarevFermion.h
+++ b/Grid/qcd/action/fermion/ShamirZolotarevFermion.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,46 +24,43 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef  GRID_QCD_SHAMIR_ZOLOTAREV_FERMION_H
 #define  GRID_QCD_SHAMIR_ZOLOTAREV_FERMION_H
 
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class ShamirZolotarevFermion : public MobiusZolotarevFermion<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
 
-    template<class Impl>
-    class ShamirZolotarevFermion : public MobiusZolotarevFermion<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-
-      // Constructors
+  // Constructors
 
 
-    ShamirZolotarevFermion(GaugeField &_Umu,
-			   GridCartesian         &FiveDimGrid,
-			   GridRedBlackCartesian &FiveDimRedBlackGrid,
-			   GridCartesian         &FourDimGrid,
-			   GridRedBlackCartesian &FourDimRedBlackGrid,
-			   RealD _mass,RealD _M5,
-			   RealD lo, RealD hi,const ImplParams &p= ImplParams()) : 
+  ShamirZolotarevFermion(GaugeField &_Umu,
+			 GridCartesian         &FiveDimGrid,
+			 GridRedBlackCartesian &FiveDimRedBlackGrid,
+			 GridCartesian         &FourDimGrid,
+			 GridRedBlackCartesian &FourDimRedBlackGrid,
+			 RealD _mass,RealD _M5,
+			 RealD lo, RealD hi,const ImplParams &p= ImplParams()) : 
       
-      // b+c = 1; b-c = 1 => b=1, c=0
-      MobiusZolotarevFermion<Impl>(_Umu,
-				   FiveDimGrid,
-				   FiveDimRedBlackGrid,
-				   FourDimGrid,
-				   FourDimRedBlackGrid,_mass,_M5,1.0,0.0,lo,hi,p)
+    // b+c = 1; b-c = 1 => b=1, c=0
+    MobiusZolotarevFermion<Impl>(_Umu,
+				 FiveDimGrid,
+				 FiveDimRedBlackGrid,
+				 FourDimGrid,
+				 FourDimRedBlackGrid,_mass,_M5,1.0,0.0,lo,hi,p)
       
-      {}
+  {}
 
-    };
+};
 
-  }
-}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/fermion/StaggeredImpl.h b/Grid/qcd/action/fermion/StaggeredImpl.h
new file mode 100644
index 00000000..8adf45a4
--- /dev/null
+++ b/Grid/qcd/action/fermion/StaggeredImpl.h
@@ -0,0 +1,175 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/FermionOperatorImpl.h
+
+Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+			   /*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+template <class S, class Representation = FundamentalRepresentation >
+class StaggeredImpl : public PeriodicGaugeImpl<GaugeImplTypes<S, Representation::Dimension > > 
+{
+
+public:
+
+  typedef RealD  _Coeff_t ;
+  static const int Dimension = Representation::Dimension;
+  static const bool isFundamental = Representation::isFundamental;
+  static const bool LsVectorised=false;
+  typedef PeriodicGaugeImpl<GaugeImplTypes<S, Dimension > > Gimpl;
+      
+  //Necessary?
+  constexpr bool is_fundamental() const{return Dimension == Nc ? 1 : 0;}
+    
+  typedef _Coeff_t Coeff_t;
+
+  INHERIT_GIMPL_TYPES(Gimpl);
+      
+  template <typename vtype> using iImplSpinor            = iScalar<iScalar<iVector<vtype, Dimension> > >;
+  template <typename vtype> using iImplHalfSpinor        = iScalar<iScalar<iVector<vtype, Dimension> > >;
+  template <typename vtype> using iImplDoubledGaugeField = iVector<iScalar<iMatrix<vtype, Dimension> >, Nds>;
+  template <typename vtype> using iImplPropagator        = iScalar<iScalar<iMatrix<vtype, Dimension> > >;
+    
+  typedef iImplSpinor<Simd>            SiteSpinor;
+  typedef iImplHalfSpinor<Simd>        SiteHalfSpinor;
+  typedef iImplDoubledGaugeField<Simd> SiteDoubledGaugeField;
+  typedef iImplPropagator<Simd>        SitePropagator;
+    
+  typedef Lattice<SiteSpinor>            FermionField;
+  typedef Lattice<SiteDoubledGaugeField> DoubledGaugeField;
+  typedef Lattice<SitePropagator> PropagatorField;
+    
+  typedef StaggeredImplParams ImplParams;
+  typedef SimpleCompressor<SiteSpinor> Compressor;
+  typedef CartesianStencil<SiteSpinor, SiteSpinor, ImplParams> StencilImpl;
+  typedef typename StencilImpl::View_type StencilView;
+
+  ImplParams Params;
+    
+  StaggeredImpl(const ImplParams &p = ImplParams()) : Params(p){};
+      
+  static accelerator_inline void multLink(SiteSpinor &phi,
+		       const SiteDoubledGaugeField &U,
+		       const SiteSpinor &chi,
+		       int mu)
+  {
+    mult(&phi(), &U(mu), &chi());
+  }
+  static accelerator_inline void multLinkAdd(SiteSpinor &phi,
+			  const SiteDoubledGaugeField &U,
+			  const SiteSpinor &chi,
+			  int mu)
+  {
+    mac(&phi(), &U(mu), &chi());
+  }
+      
+  template <class ref>
+  static accelerator_inline void loadLinkElement(Simd &reg, ref &memory) 
+  {
+    reg = memory;
+  }
+      
+    inline void InsertGaugeField(DoubledGaugeField &U_ds,
+				 const GaugeLinkField &U,int mu)
+    {
+      PokeIndex<LorentzIndex>(U_ds, U, mu);
+    }
+  inline void DoubleStore(GridBase *GaugeGrid,
+			  DoubledGaugeField &UUUds, // for Naik term
+			  DoubledGaugeField &Uds,
+			  const GaugeField &Uthin,
+			  const GaugeField &Ufat) {
+    conformable(Uds.Grid(), GaugeGrid);
+    conformable(Uthin.Grid(), GaugeGrid);
+    conformable(Ufat.Grid(), GaugeGrid);
+    GaugeLinkField U(GaugeGrid);
+    GaugeLinkField UU(GaugeGrid);
+    GaugeLinkField UUU(GaugeGrid);
+    GaugeLinkField Udag(GaugeGrid);
+    GaugeLinkField UUUdag(GaugeGrid);
+    for (int mu = 0; mu < Nd; mu++) {
+
+      // Staggered Phase.
+      Lattice<iScalar<vInteger> > coor(GaugeGrid);
+      Lattice<iScalar<vInteger> > x(GaugeGrid); LatticeCoordinate(x,0);
+      Lattice<iScalar<vInteger> > y(GaugeGrid); LatticeCoordinate(y,1);
+      Lattice<iScalar<vInteger> > z(GaugeGrid); LatticeCoordinate(z,2);
+      Lattice<iScalar<vInteger> > t(GaugeGrid); LatticeCoordinate(t,3);
+
+      Lattice<iScalar<vInteger> > lin_z(GaugeGrid); lin_z=x+y;
+      Lattice<iScalar<vInteger> > lin_t(GaugeGrid); lin_t=x+y+z;
+
+      ComplexField phases(GaugeGrid);	phases=1.0;
+
+      if ( mu == 1 ) phases = where( mod(x    ,2)==(Integer)0, phases,-phases);
+      if ( mu == 2 ) phases = where( mod(lin_z,2)==(Integer)0, phases,-phases);
+      if ( mu == 3 ) phases = where( mod(lin_t,2)==(Integer)0, phases,-phases);
+
+      // 1 hop based on fat links
+      U      = PeekIndex<LorentzIndex>(Ufat, mu);
+      Udag   = adj( Cshift(U, mu, -1));
+
+      U    = U    *phases;
+      Udag = Udag *phases;
+
+	InsertGaugeField(Uds,U,mu);
+	InsertGaugeField(Uds,Udag,mu+4);
+	//	PokeIndex<LorentzIndex>(Uds, U, mu);
+	//	PokeIndex<LorentzIndex>(Uds, Udag, mu + 4);
+
+      // 3 hop based on thin links. Crazy huh ?
+      U  = PeekIndex<LorentzIndex>(Uthin, mu);
+      UU = Gimpl::CovShiftForward(U,mu,U);
+      UUU= Gimpl::CovShiftForward(U,mu,UU);
+	
+      UUUdag = adj( Cshift(UUU, mu, -3));
+
+      UUU    = UUU    *phases;
+      UUUdag = UUUdag *phases;
+
+	InsertGaugeField(UUUds,UUU,mu);
+	InsertGaugeField(UUUds,UUUdag,mu+4);
+
+    }
+  }
+
+  inline void InsertForce4D(GaugeField &mat, FermionField &Btilde, FermionField &A,int mu){
+    GaugeLinkField link(mat.Grid());
+    link = TraceIndex<SpinIndex>(outerProduct(Btilde,A)); 
+    PokeIndex<LorentzIndex>(mat,link,mu);
+  }   
+      
+  inline void InsertForce5D(GaugeField &mat, FermionField &Btilde, FermionField &Atilde,int mu){
+    assert (0); 
+    // Must never hit
+  }
+};
+typedef StaggeredImpl<vComplex,  FundamentalRepresentation > StaggeredImplR;   // Real.. whichever prec
+typedef StaggeredImpl<vComplexF, FundamentalRepresentation > StaggeredImplF;  // Float
+typedef StaggeredImpl<vComplexD, FundamentalRepresentation > StaggeredImplD;  // Double
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/StaggeredKernels.h b/Grid/qcd/action/fermion/StaggeredKernels.h
index 79de1a68..6ef0ab9d 100644
--- a/Grid/qcd/action/fermion/StaggeredKernels.h
+++ b/Grid/qcd/action/fermion/StaggeredKernels.h
@@ -26,11 +26,9 @@ See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
 /*  END LEGAL */
-#ifndef GRID_QCD_STAGGERED_KERNELS_H
-#define GRID_QCD_STAGGERED_KERNELS_H
+#pragma once
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid)
 
   ////////////////////////////////////////////////////////////////////////////////////////////////////////////////
   // Helper routines that implement Staggered stencil for a single site.
@@ -51,72 +49,69 @@ template<class Impl> class StaggeredKernels : public FermionOperator<Impl> , pub
    
 public:
     
-   void DhopDir(StencilImpl &st, DoubledGaugeField &U, DoubledGaugeField &UUU, SiteSpinor * buf,
-		      int sF, int sU, const FermionField &in, FermionField &out, int dir,int disp);
+   void DhopDirKernel(StencilImpl &st, DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU, SiteSpinor * buf,
+		      int sF, int sU, const FermionFieldView &in, FermionFieldView &out, int dir,int disp);
 
    ///////////////////////////////////////////////////////////////////////////////////////
    // Generic Nc kernels
    ///////////////////////////////////////////////////////////////////////////////////////
    void DhopSiteGeneric(StencilImpl &st, LebesgueOrder &lo, 
-			DoubledGaugeField &U, DoubledGaugeField &UUU, 
+			DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU, 
 			SiteSpinor * buf, int LLs, int sU, 
-			const FermionField &in, FermionField &out,int dag);
+			const FermionFieldView &in, FermionFieldView &out,int dag);
    void DhopSiteGenericInt(StencilImpl &st, LebesgueOrder &lo, 
-			   DoubledGaugeField &U, DoubledGaugeField &UUU, 
+			   DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU, 
 			   SiteSpinor * buf, int LLs, int sU, 
-			   const FermionField &in, FermionField &out,int dag);
+			   const FermionFieldView &in, FermionFieldView &out,int dag);
    void DhopSiteGenericExt(StencilImpl &st, LebesgueOrder &lo, 
-			   DoubledGaugeField &U, DoubledGaugeField &UUU,
+			   DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU,
 			   SiteSpinor * buf, int LLs, int sU, 
-			   const FermionField &in, FermionField &out,int dag);
+			   const FermionFieldView &in, FermionFieldView &out,int dag);
 
    ///////////////////////////////////////////////////////////////////////////////////////
    // Nc=3 specific kernels
    ///////////////////////////////////////////////////////////////////////////////////////
    void DhopSiteHand(StencilImpl &st, LebesgueOrder &lo, 
-		     DoubledGaugeField &U,DoubledGaugeField &UUU, 
+		     DoubledGaugeFieldView &U,DoubledGaugeFieldView &UUU, 
 		     SiteSpinor * buf, int LLs, int sU, 
-		     const FermionField &in, FermionField &out,int dag);
+		     const FermionFieldView &in, FermionFieldView &out,int dag);
    void DhopSiteHandInt(StencilImpl &st, LebesgueOrder &lo, 
-			DoubledGaugeField &U,DoubledGaugeField &UUU, 
+			DoubledGaugeFieldView &U,DoubledGaugeFieldView &UUU, 
 			SiteSpinor * buf, int LLs, int sU, 
-			const FermionField &in, FermionField &out,int dag);
+			const FermionFieldView &in, FermionFieldView &out,int dag);
    void DhopSiteHandExt(StencilImpl &st, LebesgueOrder &lo, 
-			DoubledGaugeField &U,DoubledGaugeField &UUU, 
+			DoubledGaugeFieldView &U,DoubledGaugeFieldView &UUU, 
 			SiteSpinor * buf, int LLs, int sU, 
-			const FermionField &in, FermionField &out,int dag);
+			const FermionFieldView &in, FermionFieldView &out,int dag);
 
    ///////////////////////////////////////////////////////////////////////////////////////
    // Asm Nc=3 specific kernels
    ///////////////////////////////////////////////////////////////////////////////////////
    void DhopSiteAsm(StencilImpl &st, LebesgueOrder &lo, 
-		    DoubledGaugeField &U,DoubledGaugeField &UUU, 
+		    DoubledGaugeFieldView &U,DoubledGaugeFieldView &UUU, 
 		    SiteSpinor * buf, int LLs, int sU, 
-		    const FermionField &in, FermionField &out,int dag);
+		    const FermionFieldView &in, FermionFieldView &out,int dag);
    ///////////////////////////////////////////////////////////////////////////////////////////////////
    // Generic interface; fan out to right routine
    ///////////////////////////////////////////////////////////////////////////////////////////////////
    void DhopSite(StencilImpl &st, LebesgueOrder &lo, 
-		 DoubledGaugeField &U, DoubledGaugeField &UUU, 
+		 DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU, 
 		 SiteSpinor * buf, int LLs, int sU,
-		 const FermionField &in, FermionField &out, int interior=1,int exterior=1);
+		 const FermionFieldView &in, FermionFieldView &out, int interior=1,int exterior=1);
 
    void DhopSiteDag(StencilImpl &st, LebesgueOrder &lo, 
-		    DoubledGaugeField &U, DoubledGaugeField &UUU, 
+		    DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU, 
 		    SiteSpinor * buf, int LLs, int sU,
-		    const FermionField &in, FermionField &out, int interior=1,int exterior=1);
+		    const FermionFieldView &in, FermionFieldView &out, int interior=1,int exterior=1);
 
    void DhopSite(StencilImpl &st, LebesgueOrder &lo, 
-		 DoubledGaugeField &U, DoubledGaugeField &UUU, 
+		 DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU, 
 		 SiteSpinor * buf, int LLs, int sU,
-		 const FermionField &in, FermionField &out, int dag, int interior,int exterior);
+		 const FermionFieldView &in, FermionFieldView &out, int dag, int interior,int exterior);
   
 public:
 
   StaggeredKernels(const ImplParams &p = ImplParams());
 
 };
-    
-}}
-
-#endif
+NAMESPACE_END(Grid);    
diff --git a/Grid/qcd/action/fermion/StaggeredVec5dImpl.h b/Grid/qcd/action/fermion/StaggeredVec5dImpl.h
new file mode 100644
index 00000000..2d4de18e
--- /dev/null
+++ b/Grid/qcd/action/fermion/StaggeredVec5dImpl.h
@@ -0,0 +1,203 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/FermionOperatorImpl.h
+
+Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+			   /*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+template <class S, class Representation = FundamentalRepresentation >
+class StaggeredVec5dImpl : public PeriodicGaugeImpl<GaugeImplTypes<S, Representation::Dimension > > {
+
+public:
+
+  static const int Dimension = Representation::Dimension;
+    static const bool isFundamental = Representation::isFundamental;
+  static const bool LsVectorised=true;
+  typedef RealD   Coeff_t ;
+  typedef PeriodicGaugeImpl<GaugeImplTypes<S, Dimension > > Gimpl;
+      
+  //Necessary?
+  constexpr bool is_fundamental() const{return Dimension == Nc ? 1 : 0;}
+
+
+  INHERIT_GIMPL_TYPES(Gimpl);
+
+  template <typename vtype> using iImplSpinor            = iScalar<iScalar<iVector<vtype, Dimension> > >;
+  template <typename vtype> using iImplHalfSpinor        = iScalar<iScalar<iVector<vtype, Dimension> > >;
+  template <typename vtype> using iImplDoubledGaugeField = iVector<iScalar<iMatrix<vtype, Dimension> >, Nds>;
+  template <typename vtype> using iImplGaugeField        = iVector<iScalar<iMatrix<vtype, Dimension> >, Nd>;
+  template <typename vtype> using iImplGaugeLink         = iScalar<iScalar<iMatrix<vtype, Dimension> > >;
+  template <typename vtype> using iImplPropagator        = iScalar<iScalar<iMatrix<vtype, Dimension> > >;
+
+  // Make the doubled gauge field a *scalar*
+  typedef iImplDoubledGaugeField<typename Simd::scalar_type>  SiteDoubledGaugeField;  // This is a scalar
+  typedef iImplGaugeField<typename Simd::scalar_type>         SiteScalarGaugeField;  // scalar
+  typedef iImplGaugeLink<typename Simd::scalar_type>          SiteScalarGaugeLink;  // scalar
+  typedef iImplPropagator<Simd>        SitePropagator;
+
+  typedef Lattice<SiteDoubledGaugeField> DoubledGaugeField;
+  typedef Lattice<SitePropagator> PropagatorField;
+    
+  typedef iImplSpinor<Simd>            SiteSpinor;
+  typedef iImplHalfSpinor<Simd>        SiteHalfSpinor;
+
+    
+  typedef Lattice<SiteSpinor>            FermionField;
+    
+  typedef StaggeredImplParams ImplParams;
+  typedef SimpleCompressor<SiteSpinor> Compressor;
+  typedef CartesianStencil<SiteSpinor, SiteSpinor, ImplParams> StencilImpl;
+  typedef typename StencilImpl::View_type StencilView;
+    
+  ImplParams Params;
+    
+  StaggeredVec5dImpl(const ImplParams &p = ImplParams()) : Params(p){};
+
+  template <class ref>
+  static accelerator_inline void loadLinkElement(Simd &reg, ref &memory) 
+  {
+    vsplat(reg, memory);
+  }
+
+  static accelerator_inline void multLink(SiteHalfSpinor &phi, 
+					  const SiteDoubledGaugeField &U,
+					  const SiteHalfSpinor &chi, 
+					  int mu) 
+  {
+    SiteGaugeLink UU;
+    for (int i = 0; i < Dimension; i++) {
+      for (int j = 0; j < Dimension; j++) {
+	vsplat(UU()()(i, j), U(mu)()(i, j));
+      }
+    }
+    mult(&phi(), &UU(), &chi());
+  }
+  static accelerator_inline void multLinkAdd(SiteHalfSpinor &phi, 
+					     const SiteDoubledGaugeField &U,
+					     const SiteHalfSpinor &chi, 
+					     int mu) 
+  {
+    SiteGaugeLink UU;
+    for (int i = 0; i < Dimension; i++) {
+      for (int j = 0; j < Dimension; j++) {
+	vsplat(UU()()(i, j), U(mu)()(i, j));
+      }
+    }
+    mac(&phi(), &UU(), &chi());
+  }
+      
+  inline void InsertGaugeField(DoubledGaugeField &U_ds,const GaugeLinkField &U,int mu)
+  {
+    GridBase *GaugeGrid = U_ds.Grid();
+    thread_for(lidx, GaugeGrid->lSites(),{
+
+	SiteScalarGaugeLink   ScalarU;
+	SiteDoubledGaugeField ScalarUds;
+	
+	Coordinate lcoor;
+	GaugeGrid->LocalIndexToLocalCoor(lidx, lcoor);
+	peekLocalSite(ScalarUds, U_ds, lcoor);
+	
+	peekLocalSite(ScalarU, U, lcoor);
+	ScalarUds(mu) = ScalarU();
+	
+    });
+  }
+  inline void DoubleStore(GridBase *GaugeGrid,
+			  DoubledGaugeField &UUUds, // for Naik term
+			  DoubledGaugeField &Uds,
+			  const GaugeField &Uthin,
+			  const GaugeField &Ufat) 
+  {
+
+    GridBase * InputGrid = Uthin.Grid();
+    conformable(InputGrid,Ufat.Grid());
+
+    GaugeLinkField U(InputGrid);
+    GaugeLinkField UU(InputGrid);
+    GaugeLinkField UUU(InputGrid);
+    GaugeLinkField Udag(InputGrid);
+    GaugeLinkField UUUdag(InputGrid);
+
+    for (int mu = 0; mu < Nd; mu++) {
+
+      // Staggered Phase.
+      Lattice<iScalar<vInteger> > coor(InputGrid);
+      Lattice<iScalar<vInteger> > x(InputGrid); LatticeCoordinate(x,0);
+      Lattice<iScalar<vInteger> > y(InputGrid); LatticeCoordinate(y,1);
+      Lattice<iScalar<vInteger> > z(InputGrid); LatticeCoordinate(z,2);
+      Lattice<iScalar<vInteger> > t(InputGrid); LatticeCoordinate(t,3);
+
+      Lattice<iScalar<vInteger> > lin_z(InputGrid); lin_z=x+y;
+      Lattice<iScalar<vInteger> > lin_t(InputGrid); lin_t=x+y+z;
+
+      ComplexField phases(InputGrid);	phases=1.0;
+
+      if ( mu == 1 ) phases = where( mod(x    ,2)==(Integer)0, phases,-phases);
+      if ( mu == 2 ) phases = where( mod(lin_z,2)==(Integer)0, phases,-phases);
+      if ( mu == 3 ) phases = where( mod(lin_t,2)==(Integer)0, phases,-phases);
+
+      // 1 hop based on fat links
+      U      = PeekIndex<LorentzIndex>(Ufat, mu);
+      Udag   = adj( Cshift(U, mu, -1));
+
+      U    = U    *phases;
+      Udag = Udag *phases;
+
+      InsertGaugeField(Uds,U,mu);
+      InsertGaugeField(Uds,Udag,mu+4);
+
+      // 3 hop based on thin links. Crazy huh ?
+      U  = PeekIndex<LorentzIndex>(Uthin, mu);
+      UU = Gimpl::CovShiftForward(U,mu,U);
+      UUU= Gimpl::CovShiftForward(U,mu,UU);
+	
+      UUUdag = adj( Cshift(UUU, mu, -3));
+
+      UUU    = UUU    *phases;
+      UUUdag = UUUdag *phases;
+
+      InsertGaugeField(UUUds,UUU,mu);
+      InsertGaugeField(UUUds,UUUdag,mu+4);
+
+    }
+  }
+
+  inline void InsertForce4D(GaugeField &mat, FermionField &Btilde, FermionField &A,int mu){
+    assert(0);
+  }   
+      
+  inline void InsertForce5D(GaugeField &mat, FermionField &Btilde, FermionField &Atilde,int mu){
+    assert (0); 
+  }
+};
+typedef StaggeredVec5dImpl<vComplex,  FundamentalRepresentation > StaggeredVec5dImplR;   // Real.. whichever prec
+typedef StaggeredVec5dImpl<vComplexF, FundamentalRepresentation > StaggeredVec5dImplF;  // Float
+typedef StaggeredVec5dImpl<vComplexD, FundamentalRepresentation > StaggeredVec5dImplD;  // Double
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/WilsonCloverFermion.h b/Grid/qcd/action/fermion/WilsonCloverFermion.h
index 40d08a76..3847b0d9 100644
--- a/Grid/qcd/action/fermion/WilsonCloverFermion.h
+++ b/Grid/qcd/action/fermion/WilsonCloverFermion.h
@@ -27,15 +27,11 @@
     *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef GRID_QCD_WILSON_CLOVER_FERMION_H
-#define GRID_QCD_WILSON_CLOVER_FERMION_H
+#pragma once
 
 #include <Grid/Grid.h>
 
-namespace Grid
-{
-namespace QCD
-{
+NAMESPACE_BEGIN(Grid);
 
 ///////////////////////////////////////////////////////////////////
 // Wilson Clover
@@ -131,22 +127,22 @@ public:
   // Derivative parts unpreconditioned pseudofermions
   void MDeriv(GaugeField &force, const FermionField &X, const FermionField &Y, int dag)
   {
-    conformable(X._grid, Y._grid);
-    conformable(X._grid, force._grid);
-    GaugeLinkField force_mu(force._grid), lambda(force._grid);
-    GaugeField clover_force(force._grid);
-    PropagatorField Lambda(force._grid);
+    conformable(X.Grid(), Y.Grid());
+    conformable(X.Grid(), force.Grid());
+    GaugeLinkField force_mu(force.Grid()), lambda(force.Grid());
+    GaugeField clover_force(force.Grid());
+    PropagatorField Lambda(force.Grid());
 
     // Guido: Here we are hitting some performance issues:
     // need to extract the components of the DoubledGaugeField
     // for each call
     // Possible solution
     // Create a vector object to store them? (cons: wasting space)
-    std::vector<GaugeLinkField> U(Nd, this->Umu._grid);
+    std::vector<GaugeLinkField> U(Nd, this->Umu.Grid());
 
     Impl::extractLinkField(U, this->Umu);
 
-    force = zero;
+    force = Zero();
     // Derivative of the Wilson hopping term
     this->DhopDeriv(force, X, Y, dag);
 
@@ -179,10 +175,10 @@ public:
     */
 
     int count = 0;
-    clover_force = zero;
+    clover_force = Zero();
     for (int mu = 0; mu < 4; mu++)
     {
-      force_mu = zero;
+      force_mu = Zero();
       for (int nu = 0; nu < 4; nu++)
       {
         if (mu == nu)
@@ -212,8 +208,8 @@ public:
   // Computing C_{\mu \nu}(x) as in Eq.(B.39) in Zbigniew Sroczynski's PhD thesis
   GaugeLinkField Cmunu(std::vector<GaugeLinkField> &U, GaugeLinkField &lambda, int mu, int nu)
   {
-    conformable(lambda._grid, U[0]._grid);
-    GaugeLinkField out(lambda._grid), tmp(lambda._grid);
+    conformable(lambda.Grid(), U[0].Grid());
+    GaugeLinkField out(lambda.Grid()), tmp(lambda.Grid());
     // insertion in upper staple
     // please check redundancy of shift operations
 
@@ -266,102 +262,113 @@ private:
   // using the DeGrand-Rossi basis for the gamma matrices
   CloverFieldType fillCloverYZ(const GaugeLinkField &F)
   {
-    CloverFieldType T(F._grid);
-    T = zero;
-    PARALLEL_FOR_LOOP
-    for (int i = 0; i < CloverTerm._grid->oSites(); i++)
+    CloverFieldType T(F.Grid());
+    T = Zero();
+    auto T_v = T.View();
+    auto F_v = F.View();
+    thread_for(i, CloverTerm.Grid()->oSites(),
     {
-      T._odata[i]()(0, 1) = timesMinusI(F._odata[i]()());
-      T._odata[i]()(1, 0) = timesMinusI(F._odata[i]()());
-      T._odata[i]()(2, 3) = timesMinusI(F._odata[i]()());
-      T._odata[i]()(3, 2) = timesMinusI(F._odata[i]()());
-    }
+      T_v[i]()(0, 1) = timesMinusI(F_v[i]()());
+      T_v[i]()(1, 0) = timesMinusI(F_v[i]()());
+      T_v[i]()(2, 3) = timesMinusI(F_v[i]()());
+      T_v[i]()(3, 2) = timesMinusI(F_v[i]()());
+    });
 
     return T;
   }
 
   CloverFieldType fillCloverXZ(const GaugeLinkField &F)
   {
-    CloverFieldType T(F._grid);
-    T = zero;
-    PARALLEL_FOR_LOOP
-    for (int i = 0; i < CloverTerm._grid->oSites(); i++)
+    CloverFieldType T(F.Grid());
+    T = Zero();
+    
+    auto T_v = T.View();
+    auto F_v = F.View();
+    thread_for(i, CloverTerm.Grid()->oSites(),
     {
-      T._odata[i]()(0, 1) = -F._odata[i]()();
-      T._odata[i]()(1, 0) = F._odata[i]()();
-      T._odata[i]()(2, 3) = -F._odata[i]()();
-      T._odata[i]()(3, 2) = F._odata[i]()();
-    }
+      T_v[i]()(0, 1) = -F_v[i]()();
+      T_v[i]()(1, 0) = F_v[i]()();
+      T_v[i]()(2, 3) = -F_v[i]()();
+      T_v[i]()(3, 2) = F_v[i]()();
+    });
 
     return T;
   }
 
   CloverFieldType fillCloverXY(const GaugeLinkField &F)
   {
-    CloverFieldType T(F._grid);
-    T = zero;
-    PARALLEL_FOR_LOOP
-    for (int i = 0; i < CloverTerm._grid->oSites(); i++)
-    {
+    CloverFieldType T(F.Grid());
+    T = Zero();
 
-      T._odata[i]()(0, 0) = timesMinusI(F._odata[i]()());
-      T._odata[i]()(1, 1) = timesI(F._odata[i]()());
-      T._odata[i]()(2, 2) = timesMinusI(F._odata[i]()());
-      T._odata[i]()(3, 3) = timesI(F._odata[i]()());
-    }
+    auto T_v = T.View();
+    auto F_v = F.View();
+    thread_for(i, CloverTerm.Grid()->oSites(),
+    {
+      T_v[i]()(0, 0) = timesMinusI(F_v[i]()());
+      T_v[i]()(1, 1) = timesI(F_v[i]()());
+      T_v[i]()(2, 2) = timesMinusI(F_v[i]()());
+      T_v[i]()(3, 3) = timesI(F_v[i]()());
+    });
 
     return T;
   }
 
   CloverFieldType fillCloverXT(const GaugeLinkField &F)
   {
-    CloverFieldType T(F._grid);
-    T = zero;
-    PARALLEL_FOR_LOOP
-    for (int i = 0; i < CloverTerm._grid->oSites(); i++)
+    CloverFieldType T(F.Grid());
+    T = Zero();
+
+    auto T_v = T.View();
+    auto F_v = F.View();
+    thread_for(i, CloverTerm.Grid()->oSites(),
     {
-      T._odata[i]()(0, 1) = timesI(F._odata[i]()());
-      T._odata[i]()(1, 0) = timesI(F._odata[i]()());
-      T._odata[i]()(2, 3) = timesMinusI(F._odata[i]()());
-      T._odata[i]()(3, 2) = timesMinusI(F._odata[i]()());
-    }
+      T_v[i]()(0, 1) = timesI(F_v[i]()());
+      T_v[i]()(1, 0) = timesI(F_v[i]()());
+      T_v[i]()(2, 3) = timesMinusI(F_v[i]()());
+      T_v[i]()(3, 2) = timesMinusI(F_v[i]()());
+    });
 
     return T;
   }
 
   CloverFieldType fillCloverYT(const GaugeLinkField &F)
   {
-    CloverFieldType T(F._grid);
-    T = zero;
-    PARALLEL_FOR_LOOP
-    for (int i = 0; i < CloverTerm._grid->oSites(); i++)
+    CloverFieldType T(F.Grid());
+    T = Zero();
+    
+    auto T_v = T.View();
+    auto F_v = F.View();
+    thread_for(i, CloverTerm.Grid()->oSites(),
     {
-      T._odata[i]()(0, 1) = -(F._odata[i]()());
-      T._odata[i]()(1, 0) = (F._odata[i]()());
-      T._odata[i]()(2, 3) = (F._odata[i]()());
-      T._odata[i]()(3, 2) = -(F._odata[i]()());
-    }
+      T_v[i]()(0, 1) = -(F_v[i]()());
+      T_v[i]()(1, 0) = (F_v[i]()());
+      T_v[i]()(2, 3) = (F_v[i]()());
+      T_v[i]()(3, 2) = -(F_v[i]()());
+    });
 
     return T;
   }
 
   CloverFieldType fillCloverZT(const GaugeLinkField &F)
   {
-    CloverFieldType T(F._grid);
-    T = zero;
-    PARALLEL_FOR_LOOP
-    for (int i = 0; i < CloverTerm._grid->oSites(); i++)
+    CloverFieldType T(F.Grid());
+
+    T = Zero();
+
+    auto T_v = T.View();
+    auto F_v = F.View();
+    thread_for(i, CloverTerm.Grid()->oSites(),
     {
-      T._odata[i]()(0, 0) = timesI(F._odata[i]()());
-      T._odata[i]()(1, 1) = timesMinusI(F._odata[i]()());
-      T._odata[i]()(2, 2) = timesMinusI(F._odata[i]()());
-      T._odata[i]()(3, 3) = timesI(F._odata[i]()());
-    }
+      T_v[i]()(0, 0) = timesI(F_v[i]()());
+      T_v[i]()(1, 1) = timesMinusI(F_v[i]()());
+      T_v[i]()(2, 2) = timesMinusI(F_v[i]()());
+      T_v[i]()(3, 3) = timesI(F_v[i]()());
+    });
 
     return T;
   }
 };
-}
-}
+NAMESPACE_END(Grid);
+
+
 
-#endif // GRID_QCD_WILSON_CLOVER_FERMION_H
diff --git a/Grid/qcd/action/fermion/WilsonCompressor.h b/Grid/qcd/action/fermion/WilsonCompressor.h
index 7553743e..10e98f33 100644
--- a/Grid/qcd/action/fermion/WilsonCompressor.h
+++ b/Grid/qcd/action/fermion/WilsonCompressor.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,13 +25,12 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef  GRID_QCD_WILSON_COMPRESSOR_H
 #define  GRID_QCD_WILSON_COMPRESSOR_H
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 /////////////////////////////////////////////////////////////////////////////////////////////
 // optimised versions supporting half precision too
@@ -43,9 +42,9 @@ class WilsonCompressorTemplate;
 
 template<class _HCspinor,class _Hspinor,class _Spinor, class projector>
 class WilsonCompressorTemplate< _HCspinor, _Hspinor, _Spinor, projector,
-  typename std::enable_if<std::is_same<_HCspinor,_Hspinor>::value>::type >
+				typename std::enable_if<std::is_same<_HCspinor,_Hspinor>::value>::type >
 {
- public:
+public:
   
   int mu,dag;  
 
@@ -62,15 +61,16 @@ class WilsonCompressorTemplate< _HCspinor, _Hspinor, _Spinor, projector,
   typedef typename SiteHalfSpinor::vector_type     vComplexHigh;
   constexpr static int Nw=sizeof(SiteHalfSpinor)/sizeof(vComplexHigh);
 
-  inline int CommDatumSize(void) {
+  accelerator_inline int CommDatumSize(void) {
     return sizeof(SiteHalfCommSpinor);
   }
 
   /*****************************************************/
   /* Compress includes precision change if mpi data is not same */
   /*****************************************************/
-  inline void Compress(SiteHalfSpinor * __restrict__ buf,Integer o,const SiteSpinor &in) {
-    SiteHalfSpinor tmp;
+  template<class _SiteHalfSpinor, class _SiteSpinor>
+  accelerator_inline void Compress(_SiteHalfSpinor *buf,Integer o,const _SiteSpinor &in) {
+    _SiteHalfSpinor tmp;
     projector::Proj(tmp,in,mu,dag);
     vstream(buf[o],tmp);
   }
@@ -78,10 +78,10 @@ class WilsonCompressorTemplate< _HCspinor, _Hspinor, _Spinor, projector,
   /*****************************************************/
   /* Exchange includes precision change if mpi data is not same */
   /*****************************************************/
-  inline void Exchange(SiteHalfSpinor * __restrict__ mp,
-                       const SiteHalfSpinor * __restrict__ vp0,
-                       const SiteHalfSpinor * __restrict__ vp1,
-		       Integer type,Integer o){
+  accelerator_inline void Exchange(SiteHalfSpinor *mp,
+				   const SiteHalfSpinor * __restrict__ vp0,
+				   const SiteHalfSpinor * __restrict__ vp1,
+				   Integer type,Integer o){
     SiteHalfSpinor tmp1;
     SiteHalfSpinor tmp2;
     exchange(tmp1,tmp2,vp0[o],vp1[o],type);
@@ -92,19 +92,21 @@ class WilsonCompressorTemplate< _HCspinor, _Hspinor, _Spinor, projector,
   /*****************************************************/
   /* Have a decompression step if mpi data is not same */
   /*****************************************************/
-  inline void Decompress(SiteHalfSpinor * __restrict__ out,
-			 SiteHalfSpinor * __restrict__ in, Integer o) {    
+  accelerator_inline void Decompress(SiteHalfSpinor * __restrict__ out,
+				     SiteHalfSpinor * __restrict__ in, Integer o) {    
     assert(0);
   }
 
   /*****************************************************/
   /* Compress Exchange                                 */
   /*****************************************************/
-  inline void CompressExchange(SiteHalfSpinor * __restrict__ out0,
-			       SiteHalfSpinor * __restrict__ out1,
-			       const SiteSpinor * __restrict__ in,
-			       Integer j,Integer k, Integer m,Integer type){
-    SiteHalfSpinor temp1, temp2,temp3,temp4;
+  accelerator_inline void CompressExchange(SiteHalfSpinor * __restrict__ out0,
+					   SiteHalfSpinor * __restrict__ out1,
+					   const SiteSpinor * __restrict__ in,
+					   Integer j,Integer k, Integer m,Integer type)
+  {
+    SiteHalfSpinor temp1, temp2;
+    SiteHalfSpinor temp3, temp4;
     projector::Proj(temp1,in[k],mu,dag);
     projector::Proj(temp2,in[m],mu,dag);
     exchange(temp3,temp4,temp1,temp2,type);
@@ -115,15 +117,15 @@ class WilsonCompressorTemplate< _HCspinor, _Hspinor, _Spinor, projector,
   /*****************************************************/
   /* Pass the info to the stencil */
   /*****************************************************/
-  inline bool DecompressionStep(void) { return false; }
+  accelerator_inline bool DecompressionStep(void) { return false; }
 
 };
 
 template<class _HCspinor,class _Hspinor,class _Spinor, class projector>
 class WilsonCompressorTemplate< _HCspinor, _Hspinor, _Spinor, projector,
-  typename std::enable_if<!std::is_same<_HCspinor,_Hspinor>::value>::type >
+				typename std::enable_if<!std::is_same<_HCspinor,_Hspinor>::value>::type >
 {
- public:
+public:
   
   int mu,dag;  
 
@@ -140,15 +142,16 @@ class WilsonCompressorTemplate< _HCspinor, _Hspinor, _Spinor, projector,
   typedef typename SiteHalfSpinor::vector_type     vComplexHigh;
   constexpr static int Nw=sizeof(SiteHalfSpinor)/sizeof(vComplexHigh);
 
-  inline int CommDatumSize(void) {
+  accelerator_inline int CommDatumSize(void) {
     return sizeof(SiteHalfCommSpinor);
   }
 
   /*****************************************************/
   /* Compress includes precision change if mpi data is not same */
   /*****************************************************/
-  inline void Compress(SiteHalfSpinor *buf,Integer o,const SiteSpinor &in) {
-    SiteHalfSpinor hsp;
+  template<class _SiteHalfSpinor, class _SiteSpinor>
+  accelerator_inline void Compress(_SiteHalfSpinor *buf,Integer o,const _SiteSpinor &in) {
+    _SiteHalfSpinor hsp;
     SiteHalfCommSpinor *hbuf = (SiteHalfCommSpinor *)buf;
     projector::Proj(hsp,in,mu,dag);
     precisionChange((vComplexLow *)&hbuf[o],(vComplexHigh *)&hsp,Nw);
@@ -157,7 +160,7 @@ class WilsonCompressorTemplate< _HCspinor, _Hspinor, _Spinor, projector,
   /*****************************************************/
   /* Exchange includes precision change if mpi data is not same */
   /*****************************************************/
-  inline void Exchange(SiteHalfSpinor *mp,
+  accelerator_inline void Exchange(SiteHalfSpinor *mp,
                        SiteHalfSpinor *vp0,
                        SiteHalfSpinor *vp1,
 		       Integer type,Integer o){
@@ -172,8 +175,7 @@ class WilsonCompressorTemplate< _HCspinor, _Hspinor, _Spinor, projector,
   /*****************************************************/
   /* Have a decompression step if mpi data is not same */
   /*****************************************************/
-  inline void Decompress(SiteHalfSpinor *out,
-			 SiteHalfSpinor *in, Integer o){
+  accelerator_inline void Decompress(SiteHalfSpinor *out, SiteHalfSpinor *in, Integer o){
     SiteHalfCommSpinor *hin=(SiteHalfCommSpinor *)in;
     precisionChange((vComplexHigh *)&out[o],(vComplexLow *)&hin[o],Nw);
   }
@@ -181,7 +183,7 @@ class WilsonCompressorTemplate< _HCspinor, _Hspinor, _Spinor, projector,
   /*****************************************************/
   /* Compress Exchange                                 */
   /*****************************************************/
-  inline void CompressExchange(SiteHalfSpinor *out0,
+  accelerator_inline void CompressExchange(SiteHalfSpinor *out0,
 			       SiteHalfSpinor *out1,
 			       const SiteSpinor *in,
 			       Integer j,Integer k, Integer m,Integer type){
@@ -198,19 +200,19 @@ class WilsonCompressorTemplate< _HCspinor, _Hspinor, _Spinor, projector,
   /*****************************************************/
   /* Pass the info to the stencil */
   /*****************************************************/
-  inline bool DecompressionStep(void) { return true; }
+  accelerator_inline bool DecompressionStep(void) { return true; }
 
 };
 
 #define DECLARE_PROJ(Projector,Compressor,spProj)			\
   class Projector {							\
   public:								\
-    template<class hsp,class fsp>					\
-    static void Proj(hsp &result,const fsp &in,int mu,int dag){			\
-      spProj(result,in);						\
-    }									\
+  template<class hsp,class fsp>						\
+  static accelerator void Proj(hsp &result,const fsp &in,int mu,int dag){ \
+    spProj(result,in);							\
+  }									\
   };									\
-template<typename HCS,typename HS,typename S> using Compressor = WilsonCompressorTemplate<HCS,HS,S,Projector>;
+  template<typename HCS,typename HS,typename S> using Compressor = WilsonCompressorTemplate<HCS,HS,S,Projector>;
 
 DECLARE_PROJ(WilsonXpProjector,WilsonXpCompressor,spProjXp);
 DECLARE_PROJ(WilsonYpProjector,WilsonYpCompressor,spProjYp);
@@ -222,9 +224,9 @@ DECLARE_PROJ(WilsonZmProjector,WilsonZmCompressor,spProjZm);
 DECLARE_PROJ(WilsonTmProjector,WilsonTmCompressor,spProjTm);
 
 class WilsonProjector {
- public:
+public:
   template<class hsp,class fsp>
-  static void Proj(hsp &result,const fsp &in,int mu,int dag){
+  static accelerator void Proj(hsp &result,const fsp &in,int mu,int dag){
     int mudag=dag? mu : (mu+Nd)%(2*Nd);
     switch(mudag) {
     case Xp:	spProjXp(result,in);	break;
@@ -243,9 +245,14 @@ template<typename HCS,typename HS,typename S> using WilsonCompressor = WilsonCom
 
 // Fast comms buffer manipulation which should inline right through (avoid direction
 // dependent logic that prevents inlining
-template<class vobj,class cobj>
-class WilsonStencil : public CartesianStencil<vobj,cobj> {
+template<class vobj,class cobj,class Parameters>
+class WilsonStencil : public CartesianStencil<vobj,cobj,Parameters> {
 public:
+
+  typedef CartesianStencil<vobj,cobj,Parameters> Base;
+  typedef typename Base::View_type View_type;
+  typedef typename Base::StencilVector StencilVector;
+
   double timer0;
   double timer1;
   double timer2;
@@ -274,16 +281,40 @@ public:
     if ( timer4 ) std::cout << GridLogMessage << " timer4 " <<timer4 <<std::endl;
   }
 
+  std::vector<int> surface_list;
+
   WilsonStencil(GridBase *grid,
 		int npoints,
 		int checkerboard,
 		const std::vector<int> &directions,
-		const std::vector<int> &distances)  
-    : CartesianStencil<vobj,cobj> (grid,npoints,checkerboard,directions,distances) 
+		const std::vector<int> &distances,Parameters p)  
+    : CartesianStencil<vobj,cobj,Parameters> (grid,npoints,checkerboard,directions,distances,p) 
   { 
     ZeroCountersi();
+    surface_list.resize(0);
+    this->same_node.resize(npoints);
   };
 
+  void BuildSurfaceList(int Ls,int vol4){
+
+    // find same node for SHM
+    // Here we know the distance is 1 for WilsonStencil
+    for(int point=0;point<this->_npoints;point++){
+      this->same_node[point] = this->SameNode(point);
+    }
+    
+    for(int site = 0 ;site< vol4;site++){
+      int local = 1;
+      for(int point=0;point<this->_npoints;point++){
+	if( (!this->GetNodeLocal(site*Ls,point)) && (!this->same_node[point]) ){ 
+	  local = 0;
+	}
+      }
+      if(local == 0) { 
+	surface_list.push_back(site);
+      }
+    }
+  }
 
   template < class compressor>
   void HaloExchangeOpt(const Lattice<vobj> &source,compressor &compress) 
@@ -292,8 +323,6 @@ public:
     this->HaloExchangeOptGather(source,compress);
     double t1=usecond();
     // Asynchronous MPI calls multidirectional, Isend etc...
-    //    this->CommunicateBegin(reqs);
-    //    this->CommunicateComplete(reqs);
     // Non-overlapped directions within a thread. Asynchronous calls except MPI3, threaded up to comm threads ways.
     this->Communicate();
     double t2=usecond(); timer1 += t2-t1;
@@ -327,7 +356,7 @@ public:
     this->_grid->StencilBarrier();
     this->mpi3synctime_g+=usecond();
 
-    assert(source._grid==this->_grid);
+    assert(source.Grid()==this->_grid);
     this->halogtime-=usecond();
     
     this->u_comm_offset=0;
@@ -365,9 +394,10 @@ public:
     this->face_table_computed=1;
     assert(this->u_comm_offset==this->_unified_buffer_size);
     this->halogtime+=usecond();
+    accelerator_barrier();
   }
 
- };
+};
 
-}} // namespace close
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/qcd/action/fermion/WilsonFermion.h b/Grid/qcd/action/fermion/WilsonFermion.h
index 8bfe6c1a..3a712435 100644
--- a/Grid/qcd/action/fermion/WilsonFermion.h
+++ b/Grid/qcd/action/fermion/WilsonFermion.h
@@ -27,16 +27,13 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
-#ifndef GRID_QCD_WILSON_FERMION_H
-#define GRID_QCD_WILSON_FERMION_H
+			   /*  END LEGAL */
+#pragma once
 
-namespace Grid {
-
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 class WilsonFermionStatic {
- public:
+public:
   static int HandOptDslash;  // these are a temporary hack
   static int MortonOrder;
   static const std::vector<int> directions;
@@ -60,8 +57,9 @@ class WilsonFermionStatic {
 };
 
 template <class Impl>
-class WilsonFermion : public WilsonKernels<Impl>, public WilsonFermionStatic {
- public:
+class WilsonFermion : public WilsonKernels<Impl>, public WilsonFermionStatic 
+{
+public:
   INHERIT_IMPL_TYPES(Impl);
   typedef WilsonKernels<Impl> Kernels;
 
@@ -138,10 +136,10 @@ class WilsonFermion : public WilsonKernels<Impl>, public WilsonFermionStatic {
 
   // Constructor
   WilsonFermion(GaugeField &_Umu, GridCartesian &Fgrid,
-                GridRedBlackCartesian &Hgrid, RealD _mass, 
+                GridRedBlackCartesian &Hgrid, RealD _mass,
                 const ImplParams &p = ImplParams(), 
                 const WilsonAnisotropyCoefficients &anis = WilsonAnisotropyCoefficients() );
-  
+
   // DoubleStore impl dependent
   void ImportGauge(const GaugeField &_Umu);
 
@@ -150,7 +148,7 @@ class WilsonFermion : public WilsonKernels<Impl>, public WilsonFermionStatic {
   ///////////////////////////////////////////////////////////////
 
   //    protected:
- public:
+public:
   virtual RealD Mass(void) { return mass; }
   virtual int   isTrivialEE(void) { return 1; };
   RealD mass;
@@ -171,7 +169,7 @@ class WilsonFermion : public WilsonKernels<Impl>, public WilsonFermionStatic {
 
   LebesgueOrder Lebesgue;
   LebesgueOrder LebesgueEvenOdd;
-
+  
   WilsonAnisotropyCoefficients anisotropyCoeff;
   
   ///////////////////////////////////////////////////////////////
@@ -182,11 +180,11 @@ class WilsonFermion : public WilsonKernels<Impl>, public WilsonFermionStatic {
                                 PropagatorField &q_out,
                                 Current curr_type,
                                 unsigned int mu);
-  void SeqConservedCurrent(PropagatorField &q_in, 
-                             PropagatorField &q_out,
-                             Current curr_type, 
-                             unsigned int mu,
-                             unsigned int tmin, 
+  void SeqConservedCurrent(PropagatorField &q_in,
+                           PropagatorField &q_out,
+                           Current curr_type,
+                           unsigned int mu, 
+                           unsigned int tmin,
                              unsigned int tmax,
 			     ComplexField &lattice_cmplx);
 };
@@ -194,7 +192,6 @@ class WilsonFermion : public WilsonKernels<Impl>, public WilsonFermionStatic {
 typedef WilsonFermion<WilsonImplF> WilsonFermionF;
 typedef WilsonFermion<WilsonImplD> WilsonFermionD;
 
+NAMESPACE_END(Grid);
+
 
-}
-}
-#endif
diff --git a/Grid/qcd/action/fermion/WilsonFermion5D.h b/Grid/qcd/action/fermion/WilsonFermion5D.h
index 4a31bb43..8f1073db 100644
--- a/Grid/qcd/action/fermion/WilsonFermion5D.h
+++ b/Grid/qcd/action/fermion/WilsonFermion5D.h
@@ -1,5 +1,5 @@
 
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -26,216 +26,215 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef  GRID_QCD_WILSON_FERMION_5D_H
 #define  GRID_QCD_WILSON_FERMION_5D_H
 
 #include <Grid/perfmon/Stat.h>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
-  ////////////////////////////////////////////////////////////////////////////////
-  // This is the 4d red black case appropriate to support
-  //
-  // parity = (x+y+z+t)|2;
-  // generalised five dim fermions like mobius, zolotarev etc..	
-  //
-  // i.e. even even contains fifth dim hopping term.
-  //
-  // [DIFFERS from original CPS red black implementation parity = (x+y+z+t+s)|2 ]
-  ////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////
+// This is the 4d red black case appropriate to support
+//
+// parity = (x+y+z+t)|2;
+// generalised five dim fermions like mobius, zolotarev etc..	
+//
+// i.e. even even contains fifth dim hopping term.
+//
+// [DIFFERS from original CPS red black implementation parity = (x+y+z+t+s)|2 ]
+////////////////////////////////////////////////////////////////////////////////
 
-    ////////////////////////////////////////////////////////////////////////////////
-    // This is the 4d red black case appropriate to support
-    //
-    // parity = (x+y+z+t)|2;
-    // generalised five dim fermions like mobius, zolotarev etc..	
-    //
-    // i.e. even even contains fifth dim hopping term.
-    //
-    // [DIFFERS from original CPS red black implementation parity = (x+y+z+t+s)|2 ]
-    ////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////
+// This is the 4d red black case appropriate to support
+//
+// parity = (x+y+z+t)|2;
+// generalised five dim fermions like mobius, zolotarev etc..	
+//
+// i.e. even even contains fifth dim hopping term.
+//
+// [DIFFERS from original CPS red black implementation parity = (x+y+z+t+s)|2 ]
+////////////////////////////////////////////////////////////////////////////////
 
-    class WilsonFermion5DStatic { 
-    public:
-      // S-direction is INNERMOST and takes no part in the parity.
-      static const std::vector<int> directions;
-      static const std::vector<int> displacements;
-      const int npoint = 8;
-    };
+class WilsonFermion5DStatic { 
+public:
+  // S-direction is INNERMOST and takes no part in the parity.
+  static const std::vector<int> directions;
+  static const std::vector<int> displacements;
+  static constexpr int npoint = 8;
+};
 
-    template<class Impl>
-    class WilsonFermion5D : public WilsonKernels<Impl>, public WilsonFermion5DStatic
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-     typedef WilsonKernels<Impl> Kernels;
-     PmuStat stat;
+template<class Impl>
+class WilsonFermion5D : public WilsonKernels<Impl>, public WilsonFermion5DStatic
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+  typedef WilsonKernels<Impl> Kernels;
+  PmuStat stat;
 
-     FermionField _tmp;
-     FermionField &tmp(void) { return _tmp; }
+  FermionField _tmp;
+  FermionField &tmp(void) { return _tmp; }
 
-     void Report(void);
-     void ZeroCounters(void);
-     double DhopCalls;
-     double DhopCommTime;
-     double DhopComputeTime;
-     double DhopComputeTime2;
-     double DhopFaceTime;
-     double DhopTotalTime;
+  void Report(void);
+  void ZeroCounters(void);
+  double DhopCalls;
+  double DhopCommTime;
+  double DhopComputeTime;
+  double DhopComputeTime2;
+  double DhopFaceTime;
+  double DhopTotalTime;
 
-     double DerivCalls;
-     double DerivCommTime;
-     double DerivComputeTime;
-     double DerivDhopComputeTime;
+  double DerivCalls;
+  double DerivCommTime;
+  double DerivComputeTime;
+  double DerivDhopComputeTime;
 
-      ///////////////////////////////////////////////////////////////
-      // Implement the abstract base
-      ///////////////////////////////////////////////////////////////
-      GridBase *GaugeGrid(void)              { return _FourDimGrid ;}
-      GridBase *GaugeRedBlackGrid(void)      { return _FourDimRedBlackGrid ;}
-      GridBase *FermionGrid(void)            { return _FiveDimGrid;}
-      GridBase *FermionRedBlackGrid(void)    { return _FiveDimRedBlackGrid;}
+  ///////////////////////////////////////////////////////////////
+  // Implement the abstract base
+  ///////////////////////////////////////////////////////////////
+  GridBase *GaugeGrid(void)              { return _FourDimGrid ;}
+  GridBase *GaugeRedBlackGrid(void)      { return _FourDimRedBlackGrid ;}
+  GridBase *FermionGrid(void)            { return _FiveDimGrid;}
+  GridBase *FermionRedBlackGrid(void)    { return _FiveDimRedBlackGrid;}
 
-      // full checkerboard operations; leave unimplemented as abstract for now
-      virtual RealD  M    (const FermionField &in, FermionField &out){assert(0); return 0.0;};
-      virtual RealD  Mdag (const FermionField &in, FermionField &out){assert(0); return 0.0;};
+  // full checkerboard operations; leave unimplemented as abstract for now
+  virtual RealD  M    (const FermionField &in, FermionField &out){assert(0); return 0.0;};
+  virtual RealD  Mdag (const FermionField &in, FermionField &out){assert(0); return 0.0;};
 
-      // half checkerboard operations; leave unimplemented as abstract for now
-      virtual void   Meooe       (const FermionField &in, FermionField &out){assert(0);};
-      virtual void   Mooee       (const FermionField &in, FermionField &out){assert(0);};
-      virtual void   MooeeInv    (const FermionField &in, FermionField &out){assert(0);};
+  // half checkerboard operations; leave unimplemented as abstract for now
+  virtual void   Meooe       (const FermionField &in, FermionField &out){assert(0);};
+  virtual void   Mooee       (const FermionField &in, FermionField &out){assert(0);};
+  virtual void   MooeeInv    (const FermionField &in, FermionField &out){assert(0);};
 
-      virtual void   MeooeDag    (const FermionField &in, FermionField &out){assert(0);};
-      virtual void   MooeeDag    (const FermionField &in, FermionField &out){assert(0);};
-      virtual void   MooeeInvDag (const FermionField &in, FermionField &out){assert(0);};
-      virtual void   Mdir   (const FermionField &in, FermionField &out,int dir,int disp){assert(0);};   // case by case Wilson, Clover, Cayley, ContFrac, PartFrac
+  virtual void   MeooeDag    (const FermionField &in, FermionField &out){assert(0);};
+  virtual void   MooeeDag    (const FermionField &in, FermionField &out){assert(0);};
+  virtual void   MooeeInvDag (const FermionField &in, FermionField &out){assert(0);};
+  virtual void   Mdir   (const FermionField &in, FermionField &out,int dir,int disp){assert(0);};   // case by case Wilson, Clover, Cayley, ContFrac, PartFrac
 
-      // These can be overridden by fancy 5d chiral action
-      virtual void DhopDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
-      virtual void DhopDerivEO(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
-      virtual void DhopDerivOE(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  // These can be overridden by fancy 5d chiral action
+  virtual void DhopDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  virtual void DhopDerivEO(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  virtual void DhopDerivOE(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
 
       void MomentumSpacePropagatorHt_5d(FermionField &out,const FermionField &in,RealD mass,std::vector<double> twist) ;
       void MomentumSpacePropagatorHt(FermionField &out,const FermionField &in,RealD mass,std::vector<double> twist) ;
       void MomentumSpacePropagatorHw(FermionField &out,const FermionField &in,RealD mass,std::vector<double> twist) ;
 
-      // Implement hopping term non-hermitian hopping term; half cb or both
-      // Implement s-diagonal DW
-      void DW    (const FermionField &in, FermionField &out,int dag);
-      void Dhop  (const FermionField &in, FermionField &out,int dag);
-      void DhopOE(const FermionField &in, FermionField &out,int dag);
-      void DhopEO(const FermionField &in, FermionField &out,int dag);
+  // Implement hopping term non-hermitian hopping term; half cb or both
+  // Implement s-diagonal DW
+  void DW    (const FermionField &in, FermionField &out,int dag);
+  void Dhop  (const FermionField &in, FermionField &out,int dag);
+  void DhopOE(const FermionField &in, FermionField &out,int dag);
+  void DhopEO(const FermionField &in, FermionField &out,int dag);
 
-      // add a DhopComm
-      // -- suboptimal interface will presently trigger multiple comms.
-    void DhopDir(const FermionField &in, FermionField &out,int dir,int disp);
+  // add a DhopComm
+  // -- suboptimal interface will presently trigger multiple comms.
+  void DhopDir(const FermionField &in, FermionField &out,int dir,int disp);
     
-    ///////////////////////////////////////////////////////////////
-    // New methods added 
-    ///////////////////////////////////////////////////////////////
-    void DerivInternal(StencilImpl & st,
-		       DoubledGaugeField & U,
-		       GaugeField &mat,
-		       const FermionField &A,
-		       const FermionField &B,
-		       int dag);
+  ///////////////////////////////////////////////////////////////
+  // New methods added 
+  ///////////////////////////////////////////////////////////////
+  void DerivInternal(StencilImpl & st,
+		     DoubledGaugeField & U,
+		     GaugeField &mat,
+		     const FermionField &A,
+		     const FermionField &B,
+		     int dag);
     
-    void DhopInternal(StencilImpl & st,
-		      LebesgueOrder &lo,
-		      DoubledGaugeField &U,
-		      const FermionField &in, 
-		      FermionField &out,
-		      int dag);
+  void DhopInternal(StencilImpl & st,
+		    LebesgueOrder &lo,
+		    DoubledGaugeField &U,
+		    const FermionField &in, 
+		    FermionField &out,
+		    int dag);
 
-    void DhopInternalOverlappedComms(StencilImpl & st,
-				     LebesgueOrder &lo,
-				     DoubledGaugeField &U,
-				     const FermionField &in, 
-				     FermionField &out,
-				     int dag);
+  void DhopInternalOverlappedComms(StencilImpl & st,
+				   LebesgueOrder &lo,
+				   DoubledGaugeField &U,
+				   const FermionField &in, 
+				   FermionField &out,
+				   int dag);
 
-    void DhopInternalSerialComms(StencilImpl & st,
-				 LebesgueOrder &lo,
-				 DoubledGaugeField &U,
-				 const FermionField &in, 
-				 FermionField &out,
-				 int dag);
+  void DhopInternalSerialComms(StencilImpl & st,
+			       LebesgueOrder &lo,
+			       DoubledGaugeField &U,
+			       const FermionField &in, 
+			       FermionField &out,
+			       int dag);
     
-    // Constructors
-    WilsonFermion5D(GaugeField &_Umu,
-		    GridCartesian         &FiveDimGrid,
-		    GridRedBlackCartesian &FiveDimRedBlackGrid,
-		    GridCartesian         &FourDimGrid,
-		    GridRedBlackCartesian &FourDimRedBlackGrid,
-		    double _M5,const ImplParams &p= ImplParams());
+  // Constructors
+  WilsonFermion5D(GaugeField &_Umu,
+		  GridCartesian         &FiveDimGrid,
+		  GridRedBlackCartesian &FiveDimRedBlackGrid,
+		  GridCartesian         &FourDimGrid,
+		  GridRedBlackCartesian &FourDimRedBlackGrid,
+		  double _M5,const ImplParams &p= ImplParams());
     
-    // Constructors
-    /*
-      WilsonFermion5D(int simd, 
-      GaugeField &_Umu,
-      GridCartesian         &FiveDimGrid,
-      GridRedBlackCartesian &FiveDimRedBlackGrid,
-      GridCartesian         &FourDimGrid,
-      double _M5,const ImplParams &p= ImplParams());
-    */
+  // Constructors
+  /*
+    WilsonFermion5D(int simd, 
+    GaugeField &_Umu,
+    GridCartesian         &FiveDimGrid,
+    GridRedBlackCartesian &FiveDimRedBlackGrid,
+    GridCartesian         &FourDimGrid,
+    double _M5,const ImplParams &p= ImplParams());
+  */
     
-    // DoubleStore
-    void ImportGauge(const GaugeField &_Umu);
+  // DoubleStore
+  void ImportGauge(const GaugeField &_Umu);
     
-    ///////////////////////////////////////////////////////////////
-    // Data members require to support the functionality
-    ///////////////////////////////////////////////////////////////
-  public:
+  ///////////////////////////////////////////////////////////////
+  // Data members require to support the functionality
+  ///////////////////////////////////////////////////////////////
+public:
     
-    // Add these to the support from Wilson
-    GridBase *_FourDimGrid;
-    GridBase *_FourDimRedBlackGrid;
-    GridBase *_FiveDimGrid;
-    GridBase *_FiveDimRedBlackGrid;
+  // Add these to the support from Wilson
+  GridBase *_FourDimGrid;
+  GridBase *_FourDimRedBlackGrid;
+  GridBase *_FiveDimGrid;
+  GridBase *_FiveDimRedBlackGrid;
     
-    double                        M5;
-    int Ls;
+  double                        M5;
+  int Ls;
     
-    //Defines the stencils for even and odd
-    StencilImpl Stencil; 
-    StencilImpl StencilEven; 
-    StencilImpl StencilOdd; 
+  //Defines the stencils for even and odd
+  StencilImpl Stencil; 
+  StencilImpl StencilEven; 
+  StencilImpl StencilOdd; 
     
-    // Copy of the gauge field , with even and odd subsets
-    DoubledGaugeField Umu;
-    DoubledGaugeField UmuEven;
-    DoubledGaugeField UmuOdd;
+  // Copy of the gauge field , with even and odd subsets
+  DoubledGaugeField Umu;
+  DoubledGaugeField UmuEven;
+  DoubledGaugeField UmuOdd;
     
-    LebesgueOrder Lebesgue;
-    LebesgueOrder LebesgueEvenOdd;
+  LebesgueOrder Lebesgue;
+  LebesgueOrder LebesgueEvenOdd;
     
-    // Comms buffer
-    std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  comm_buf;
+  // Comms buffer
+  std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  comm_buf;
     
-    ///////////////////////////////////////////////////////////////
-    // Conserved current utilities
-    ///////////////////////////////////////////////////////////////
-    void ContractConservedCurrent(PropagatorField &q_in_1,
-                                  PropagatorField &q_in_2,
-                                  PropagatorField &q_out,
-                                  Current curr_type, 
-                                  unsigned int mu);
-    void SeqConservedCurrent(PropagatorField &q_in, 
-                             PropagatorField &q_out,
-                             Current curr_type, 
-                             unsigned int mu,
-                             unsigned int tmin, 
-                             unsigned int tmax,
-			     ComplexField &lattice_cmplx);
+  ///////////////////////////////////////////////////////////////
+  // Conserved current utilities
+  ///////////////////////////////////////////////////////////////
+  void ContractConservedCurrent(PropagatorField &q_in_1,
+				PropagatorField &q_in_2,
+				PropagatorField &q_out,
+				Current curr_type, 
+				unsigned int mu);
+  void SeqConservedCurrent(PropagatorField &q_in,
+			   PropagatorField &q_out,
+			   Current curr_type,
+			   unsigned int mu,
+			   unsigned int tmin,
+			   unsigned int tmax,
+			   ComplexField &lattice_cmplx);
 
-    void ContractJ5q(PropagatorField &q_in,ComplexField &J5q);
-    void ContractJ5q(FermionField &q_in,ComplexField &J5q);
+  void ContractJ5q(PropagatorField &q_in,ComplexField &J5q);
+  void ContractJ5q(FermionField &q_in,ComplexField &J5q);
 
-  };
+};
 
-}}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/fermion/WilsonImpl.h b/Grid/qcd/action/fermion/WilsonImpl.h
new file mode 100644
index 00000000..47160730
--- /dev/null
+++ b/Grid/qcd/action/fermion/WilsonImpl.h
@@ -0,0 +1,226 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/FermionOperatorImpl.h
+
+Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+			   /*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+  
+/////////////////////////////////////////////////////////////////////////////
+// Single flavour four spinors with colour index
+/////////////////////////////////////////////////////////////////////////////
+template <class S, class Representation = FundamentalRepresentation,class Options = CoeffReal >
+class WilsonImpl : public PeriodicGaugeImpl<GaugeImplTypes<S, Representation::Dimension > > {
+public:
+
+  static const int Dimension = Representation::Dimension;
+  static const bool isFundamental = Representation::isFundamental;
+  static const bool LsVectorised=false;
+  static const int Nhcs = Options::Nhcs;
+
+  typedef PeriodicGaugeImpl<GaugeImplTypes<S, Dimension > > Gimpl;
+  INHERIT_GIMPL_TYPES(Gimpl);
+      
+  //Necessary?
+  constexpr bool is_fundamental() const{return Dimension == Nc ? 1 : 0;}
+    
+  typedef typename Options::_Coeff_t Coeff_t;
+  typedef typename Options::template PrecisionMapper<Simd>::LowerPrecVector SimdL;
+      
+  template <typename vtype> using iImplSpinor            = iScalar<iVector<iVector<vtype, Dimension>, Ns> >;
+  template <typename vtype> using iImplPropagator        = iScalar<iMatrix<iMatrix<vtype, Dimension>, Ns> >;
+  template <typename vtype> using iImplHalfSpinor        = iScalar<iVector<iVector<vtype, Dimension>, Nhs> >;
+  template <typename vtype> using iImplHalfCommSpinor    = iScalar<iVector<iVector<vtype, Dimension>, Nhcs> >;
+  template <typename vtype> using iImplDoubledGaugeField = iVector<iScalar<iMatrix<vtype, Dimension> >, Nds>;
+    
+  typedef iImplSpinor<Simd>            SiteSpinor;
+  typedef iImplPropagator<Simd>        SitePropagator;
+  typedef iImplHalfSpinor<Simd>        SiteHalfSpinor;
+  typedef iImplHalfCommSpinor<SimdL>   SiteHalfCommSpinor;
+  typedef iImplDoubledGaugeField<Simd> SiteDoubledGaugeField;
+    
+  typedef Lattice<SiteSpinor>            FermionField;
+  typedef Lattice<SitePropagator>        PropagatorField;
+  typedef Lattice<SiteDoubledGaugeField> DoubledGaugeField;
+    
+  typedef WilsonCompressor<SiteHalfCommSpinor,SiteHalfSpinor, SiteSpinor> Compressor;
+  typedef WilsonImplParams ImplParams;
+  typedef WilsonStencil<SiteSpinor, SiteHalfSpinor,ImplParams> StencilImpl;
+  typedef typename StencilImpl::View_type StencilView;
+    
+  ImplParams Params;
+
+  WilsonImpl(const ImplParams &p = ImplParams()) : Params(p){
+    assert(Params.boundary_phases.size() == Nd);
+  };
+
+  template<class _Spinor>
+  static accelerator_inline void multLink(_Spinor &phi,
+					  const SiteDoubledGaugeField &U,
+					  const _Spinor &chi,
+					  int mu) 
+  {
+    auto UU = coalescedRead(U(mu));
+    mult(&phi(), &UU, &chi());
+  }
+  template<class _Spinor>
+  static accelerator_inline void multLink(_Spinor &phi,
+					  const SiteDoubledGaugeField &U,
+					  const _Spinor &chi,
+					  int mu,
+					  StencilEntry *SE,
+					  StencilView &St) 
+  {
+    multLink(phi,U,chi,mu);
+  }
+    
+      
+  template <class ref>
+  static accelerator_inline void loadLinkElement(Simd &reg, ref &memory) 
+  {
+    reg = memory;
+  }
+      
+  inline void DoubleStore(GridBase *GaugeGrid,
+			  DoubledGaugeField &Uds,
+			  const GaugeField &Umu) 
+  {
+    typedef typename Simd::scalar_type scalar_type;
+
+    conformable(Uds.Grid(), GaugeGrid);
+    conformable(Umu.Grid(), GaugeGrid);
+
+    GaugeLinkField U(GaugeGrid);
+    GaugeLinkField tmp(GaugeGrid);
+
+    Lattice<iScalar<vInteger> > coor(GaugeGrid);
+      ////////////////////////////////////////////////////
+      // apply any boundary phase or twists
+      ////////////////////////////////////////////////////
+    for (int mu = 0; mu < Nd; mu++) {
+
+	////////// boundary phase /////////////
+      auto pha = Params.boundary_phases[mu];
+      scalar_type phase( real(pha),imag(pha) );
+
+	int L   = GaugeGrid->GlobalDimensions()[mu];
+        int Lmu = L - 1;
+
+      LatticeCoordinate(coor, mu);
+
+      U = PeekIndex<LorentzIndex>(Umu, mu);
+
+	// apply any twists
+	RealD theta = Params.twist_n_2pi_L[mu] * 2*M_PI / L;
+	if ( theta != 0.0) { 
+	  scalar_type twphase(::cos(theta),::sin(theta));
+	  U = twphase*U;
+	  std::cout << GridLogMessage << " Twist ["<<mu<<"] "<< Params.twist_n_2pi_L[mu]<< " phase"<<phase <<std::endl;
+	}
+
+      tmp = where(coor == Lmu, phase * U, U);
+      PokeIndex<LorentzIndex>(Uds, tmp, mu);
+
+      U = adj(Cshift(U, mu, -1));
+      U = where(coor == 0, conjugate(phase) * U, U); 
+      PokeIndex<LorentzIndex>(Uds, U, mu + 4);
+    }
+  }
+
+  inline void InsertForce4D(GaugeField &mat, FermionField &Btilde, FermionField &A,int mu){
+    GaugeLinkField link(mat.Grid());
+    link = TraceIndex<SpinIndex>(outerProduct(Btilde,A)); 
+    PokeIndex<LorentzIndex>(mat,link,mu);
+  }   
+      
+    inline void outerProductImpl(PropagatorField &mat, const FermionField &B, const FermionField &A){
+      mat = outerProduct(B,A); 
+    }  
+
+    inline void TraceSpinImpl(GaugeLinkField &mat, PropagatorField&P) {
+      mat = TraceIndex<SpinIndex>(P); 
+    }
+      
+    inline void extractLinkField(std::vector<GaugeLinkField> &mat, DoubledGaugeField &Uds){
+      for (int mu = 0; mu < Nd; mu++)
+      mat[mu] = PeekIndex<LorentzIndex>(Uds, mu);
+    }
+
+
+  inline void InsertForce5D(GaugeField &mat, FermionField &Btilde, FermionField &Atilde,int mu){
+      
+    int Ls=Btilde.Grid()->_fdimensions[0];
+    GaugeLinkField tmp(mat.Grid());
+    tmp = Zero();
+    auto tmp_v = tmp.View();
+    auto Btilde_v = Btilde.View();
+    auto Atilde_v = Atilde.View();
+    thread_for(sss,tmp.Grid()->oSites(),{
+      int sU=sss;
+      for(int s=0;s<Ls;s++){
+	int sF = s+Ls*sU;
+	tmp_v[sU] = tmp_v[sU]+ traceIndex<SpinIndex>(outerProduct(Btilde_v[sF],Atilde_v[sF])); // ordering here
+      }
+    });
+    PokeIndex<LorentzIndex>(mat,tmp,mu);
+      
+  }
+};
+
+
+typedef WilsonImpl<vComplex,  FundamentalRepresentation, CoeffReal > WilsonImplR;  // Real.. whichever prec
+typedef WilsonImpl<vComplexF, FundamentalRepresentation, CoeffReal > WilsonImplF;  // Float
+typedef WilsonImpl<vComplexD, FundamentalRepresentation, CoeffReal > WilsonImplD;  // Double
+
+typedef WilsonImpl<vComplex,  FundamentalRepresentation, CoeffRealHalfComms > WilsonImplRL;  // Real.. whichever prec
+typedef WilsonImpl<vComplexF, FundamentalRepresentation, CoeffRealHalfComms > WilsonImplFH;  // Float
+typedef WilsonImpl<vComplexD, FundamentalRepresentation, CoeffRealHalfComms > WilsonImplDF;  // Double
+
+typedef WilsonImpl<vComplex,  FundamentalRepresentation, CoeffComplex > ZWilsonImplR; // Real.. whichever prec
+typedef WilsonImpl<vComplexF, FundamentalRepresentation, CoeffComplex > ZWilsonImplF; // Float
+typedef WilsonImpl<vComplexD, FundamentalRepresentation, CoeffComplex > ZWilsonImplD; // Double
+
+typedef WilsonImpl<vComplex,  FundamentalRepresentation, CoeffComplexHalfComms > ZWilsonImplRL; // Real.. whichever prec
+typedef WilsonImpl<vComplexF, FundamentalRepresentation, CoeffComplexHalfComms > ZWilsonImplFH; // Float
+typedef WilsonImpl<vComplexD, FundamentalRepresentation, CoeffComplexHalfComms > ZWilsonImplDF; // Double
+ 
+typedef WilsonImpl<vComplex,  AdjointRepresentation, CoeffReal > WilsonAdjImplR;   // Real.. whichever prec
+typedef WilsonImpl<vComplexF, AdjointRepresentation, CoeffReal > WilsonAdjImplF;  // Float
+typedef WilsonImpl<vComplexD, AdjointRepresentation, CoeffReal > WilsonAdjImplD;  // Double
+ 
+typedef WilsonImpl<vComplex,  TwoIndexSymmetricRepresentation, CoeffReal > WilsonTwoIndexSymmetricImplR;   // Real.. whichever prec
+typedef WilsonImpl<vComplexF, TwoIndexSymmetricRepresentation, CoeffReal > WilsonTwoIndexSymmetricImplF;  // Float
+typedef WilsonImpl<vComplexD, TwoIndexSymmetricRepresentation, CoeffReal > WilsonTwoIndexSymmetricImplD;  // Double
+ 
+typedef WilsonImpl<vComplex,  TwoIndexAntiSymmetricRepresentation, CoeffReal > WilsonTwoIndexAntiSymmetricImplR;   // Real.. whichever prec
+typedef WilsonImpl<vComplexF, TwoIndexAntiSymmetricRepresentation, CoeffReal > WilsonTwoIndexAntiSymmetricImplF;  // Float
+typedef WilsonImpl<vComplexD, TwoIndexAntiSymmetricRepresentation, CoeffReal > WilsonTwoIndexAntiSymmetricImplD;  // Double
+
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/WilsonKernels.cc b/Grid/qcd/action/fermion/WilsonKernels.cc
deleted file mode 100644
index dc66db23..00000000
--- a/Grid/qcd/action/fermion/WilsonKernels.cc
+++ /dev/null
@@ -1,455 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
-
-Copyright (C) 2015
-
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution
-directory
-*************************************************************************************/
-/*  END LEGAL */
-#include <Grid/qcd/action/fermion/FermionCore.h>
-
-namespace Grid {
-namespace QCD {
-
-int WilsonKernelsStatic::Opt   = WilsonKernelsStatic::OptGeneric;
-int WilsonKernelsStatic::Comms = WilsonKernelsStatic::CommsAndCompute;
-
-template <class Impl>
-WilsonKernels<Impl>::WilsonKernels(const ImplParams &p) : Base(p){};
-
-////////////////////////////////////////////
-// Generic implementation; move to different file?
-////////////////////////////////////////////
-  
-#define GENERIC_STENCIL_LEG(Dir,spProj,Recon)			\
-  SE = st.GetEntry(ptype, Dir, sF);				\
-  if (SE->_is_local) {						\
-    chi_p = &chi;						\
-    if (SE->_permute) {						\
-      spProj(tmp, in._odata[SE->_offset]);			\
-      permute(chi, tmp, ptype);					\
-    } else {							\
-      spProj(chi, in._odata[SE->_offset]);			\
-    }								\
-  } else {							\
-    chi_p = &buf[SE->_offset];					\
-  }								\
-  Impl::multLink(Uchi, U._odata[sU], *chi_p, Dir, SE, st);	\
-  Recon(result, Uchi);
-  
-#define GENERIC_STENCIL_LEG_INT(Dir,spProj,Recon)		\
-  SE = st.GetEntry(ptype, Dir, sF);				\
-  if (SE->_is_local) {						\
-    chi_p = &chi;						\
-    if (SE->_permute) {						\
-      spProj(tmp, in._odata[SE->_offset]);			\
-      permute(chi, tmp, ptype);					\
-    } else {							\
-      spProj(chi, in._odata[SE->_offset]);			\
-    }								\
-  } else if ( st.same_node[Dir] ) {				\
-      chi_p = &buf[SE->_offset];				\
-  }								\
-  if (SE->_is_local || st.same_node[Dir] ) {			\
-    Impl::multLink(Uchi, U._odata[sU], *chi_p, Dir, SE, st);	\
-    Recon(result, Uchi);					\
-  }
-
-#define GENERIC_STENCIL_LEG_EXT(Dir,spProj,Recon)		\
-  SE = st.GetEntry(ptype, Dir, sF);				\
-  if ((!SE->_is_local) && (!st.same_node[Dir]) ) {		\
-    chi_p = &buf[SE->_offset];					\
-    Impl::multLink(Uchi, U._odata[sU], *chi_p, Dir, SE, st);	\
-    Recon(result, Uchi);					\
-    nmu++;							\
-  }
-
-#define GENERIC_DHOPDIR_LEG(Dir,spProj,Recon)			\
-  if (gamma == Dir) {						\
-    if (SE->_is_local && SE->_permute) {			\
-      spProj(tmp, in._odata[SE->_offset]);			\
-      permute(chi, tmp, ptype);					\
-    } else if (SE->_is_local) {					\
-      spProj(chi, in._odata[SE->_offset]);			\
-    } else {							\
-      chi = buf[SE->_offset];					\
-    }								\
-    Impl::multLink(Uchi, U._odata[sU], chi, dir, SE, st);	\
-    Recon(result, Uchi);					\
-  }
-
-  ////////////////////////////////////////////////////////////////////
-  // All legs kernels ; comms then compute
-  ////////////////////////////////////////////////////////////////////
-template <class Impl>
-void WilsonKernels<Impl>::GenericDhopSiteDag(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
-					     SiteHalfSpinor *buf, int sF,
-					     int sU, const FermionField &in, FermionField &out)
-{
-  SiteHalfSpinor tmp;
-  SiteHalfSpinor chi;
-  SiteHalfSpinor *chi_p;
-  SiteHalfSpinor Uchi;
-  SiteSpinor result;
-  StencilEntry *SE;
-  int ptype;
-
-  GENERIC_STENCIL_LEG(Xp,spProjXp,spReconXp);
-  GENERIC_STENCIL_LEG(Yp,spProjYp,accumReconYp);
-  GENERIC_STENCIL_LEG(Zp,spProjZp,accumReconZp);
-  GENERIC_STENCIL_LEG(Tp,spProjTp,accumReconTp);
-  GENERIC_STENCIL_LEG(Xm,spProjXm,accumReconXm);
-  GENERIC_STENCIL_LEG(Ym,spProjYm,accumReconYm);
-  GENERIC_STENCIL_LEG(Zm,spProjZm,accumReconZm);
-  GENERIC_STENCIL_LEG(Tm,spProjTm,accumReconTm);
-  vstream(out._odata[sF], result);
-};
-
-template <class Impl>
-void WilsonKernels<Impl>::GenericDhopSite(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
-					  SiteHalfSpinor *buf, int sF,
-					  int sU, const FermionField &in, FermionField &out) 
-{
-  SiteHalfSpinor tmp;
-  SiteHalfSpinor chi;
-  SiteHalfSpinor *chi_p;
-  SiteHalfSpinor Uchi;
-  SiteSpinor result;
-  StencilEntry *SE;
-  int ptype;
-
-  GENERIC_STENCIL_LEG(Xm,spProjXp,spReconXp);
-  GENERIC_STENCIL_LEG(Ym,spProjYp,accumReconYp);
-  GENERIC_STENCIL_LEG(Zm,spProjZp,accumReconZp);
-  GENERIC_STENCIL_LEG(Tm,spProjTp,accumReconTp);
-  GENERIC_STENCIL_LEG(Xp,spProjXm,accumReconXm);
-  GENERIC_STENCIL_LEG(Yp,spProjYm,accumReconYm);
-  GENERIC_STENCIL_LEG(Zp,spProjZm,accumReconZm);
-  GENERIC_STENCIL_LEG(Tp,spProjTm,accumReconTm);
-  vstream(out._odata[sF], result);
-};
-  ////////////////////////////////////////////////////////////////////
-  // Interior kernels
-  ////////////////////////////////////////////////////////////////////
-template <class Impl>
-void WilsonKernels<Impl>::GenericDhopSiteDagInt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
-						SiteHalfSpinor *buf, int sF,
-						int sU, const FermionField &in, FermionField &out)
-{
-  SiteHalfSpinor tmp;
-  SiteHalfSpinor chi;
-  SiteHalfSpinor *chi_p;
-  SiteHalfSpinor Uchi;
-  SiteSpinor result;
-  StencilEntry *SE;
-  int ptype;
-
-  result=zero;
-  GENERIC_STENCIL_LEG_INT(Xp,spProjXp,accumReconXp);
-  GENERIC_STENCIL_LEG_INT(Yp,spProjYp,accumReconYp);
-  GENERIC_STENCIL_LEG_INT(Zp,spProjZp,accumReconZp);
-  GENERIC_STENCIL_LEG_INT(Tp,spProjTp,accumReconTp);
-  GENERIC_STENCIL_LEG_INT(Xm,spProjXm,accumReconXm);
-  GENERIC_STENCIL_LEG_INT(Ym,spProjYm,accumReconYm);
-  GENERIC_STENCIL_LEG_INT(Zm,spProjZm,accumReconZm);
-  GENERIC_STENCIL_LEG_INT(Tm,spProjTm,accumReconTm);
-  vstream(out._odata[sF], result);
-};
-
-template <class Impl>
-void WilsonKernels<Impl>::GenericDhopSiteInt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
-					     SiteHalfSpinor *buf, int sF,
-					     int sU, const FermionField &in, FermionField &out) 
-{
-  SiteHalfSpinor tmp;
-  SiteHalfSpinor chi;
-  SiteHalfSpinor *chi_p;
-  SiteHalfSpinor Uchi;
-  SiteSpinor result;
-  StencilEntry *SE;
-  int ptype;
-  result=zero;
-  GENERIC_STENCIL_LEG_INT(Xm,spProjXp,accumReconXp);
-  GENERIC_STENCIL_LEG_INT(Ym,spProjYp,accumReconYp);
-  GENERIC_STENCIL_LEG_INT(Zm,spProjZp,accumReconZp);
-  GENERIC_STENCIL_LEG_INT(Tm,spProjTp,accumReconTp);
-  GENERIC_STENCIL_LEG_INT(Xp,spProjXm,accumReconXm);
-  GENERIC_STENCIL_LEG_INT(Yp,spProjYm,accumReconYm);
-  GENERIC_STENCIL_LEG_INT(Zp,spProjZm,accumReconZm);
-  GENERIC_STENCIL_LEG_INT(Tp,spProjTm,accumReconTm);
-  vstream(out._odata[sF], result);
-};
-////////////////////////////////////////////////////////////////////
-// Exterior kernels
-////////////////////////////////////////////////////////////////////
-template <class Impl>
-void WilsonKernels<Impl>::GenericDhopSiteDagExt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
-						SiteHalfSpinor *buf, int sF,
-						int sU, const FermionField &in, FermionField &out)
-{
-  SiteHalfSpinor tmp;
-  SiteHalfSpinor chi;
-  SiteHalfSpinor *chi_p;
-  SiteHalfSpinor Uchi;
-  SiteSpinor result;
-  StencilEntry *SE;
-  int ptype;
-  int nmu=0;
-  result=zero;
-  GENERIC_STENCIL_LEG_EXT(Xp,spProjXp,accumReconXp);
-  GENERIC_STENCIL_LEG_EXT(Yp,spProjYp,accumReconYp);
-  GENERIC_STENCIL_LEG_EXT(Zp,spProjZp,accumReconZp);
-  GENERIC_STENCIL_LEG_EXT(Tp,spProjTp,accumReconTp);
-  GENERIC_STENCIL_LEG_EXT(Xm,spProjXm,accumReconXm);
-  GENERIC_STENCIL_LEG_EXT(Ym,spProjYm,accumReconYm);
-  GENERIC_STENCIL_LEG_EXT(Zm,spProjZm,accumReconZm);
-  GENERIC_STENCIL_LEG_EXT(Tm,spProjTm,accumReconTm);
-  if ( nmu ) { 
-    out._odata[sF] = out._odata[sF] + result; 
-  }
-};
-
-template <class Impl>
-void WilsonKernels<Impl>::GenericDhopSiteExt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
-					     SiteHalfSpinor *buf, int sF,
-					     int sU, const FermionField &in, FermionField &out) 
-{
-  SiteHalfSpinor tmp;
-  SiteHalfSpinor chi;
-  SiteHalfSpinor *chi_p;
-  SiteHalfSpinor Uchi;
-  SiteSpinor result;
-  StencilEntry *SE;
-  int ptype;
-  int nmu=0;
-  result=zero;
-  GENERIC_STENCIL_LEG_EXT(Xm,spProjXp,accumReconXp);
-  GENERIC_STENCIL_LEG_EXT(Ym,spProjYp,accumReconYp);
-  GENERIC_STENCIL_LEG_EXT(Zm,spProjZp,accumReconZp);
-  GENERIC_STENCIL_LEG_EXT(Tm,spProjTp,accumReconTp);
-  GENERIC_STENCIL_LEG_EXT(Xp,spProjXm,accumReconXm);
-  GENERIC_STENCIL_LEG_EXT(Yp,spProjYm,accumReconYm);
-  GENERIC_STENCIL_LEG_EXT(Zp,spProjZm,accumReconZm);
-  GENERIC_STENCIL_LEG_EXT(Tp,spProjTm,accumReconTm);
-  if ( nmu ) { 
-    out._odata[sF] = out._odata[sF] + result; 
-  }
-};
-
-template <class Impl>
-void WilsonKernels<Impl>::DhopDir( StencilImpl &st, DoubledGaugeField &U,SiteHalfSpinor *buf, int sF,
-					   int sU, const FermionField &in, FermionField &out, int dir, int gamma) {
-
-  SiteHalfSpinor tmp;
-  SiteHalfSpinor chi;
-  SiteSpinor result;
-  SiteHalfSpinor Uchi;
-  StencilEntry *SE;
-  int ptype;
-
-  SE = st.GetEntry(ptype, dir, sF);
-  GENERIC_DHOPDIR_LEG(Xp,spProjXp,spReconXp);
-  GENERIC_DHOPDIR_LEG(Yp,spProjYp,spReconYp);
-  GENERIC_DHOPDIR_LEG(Zp,spProjZp,spReconZp);
-  GENERIC_DHOPDIR_LEG(Tp,spProjTp,spReconTp);
-  GENERIC_DHOPDIR_LEG(Xm,spProjXm,spReconXm);
-  GENERIC_DHOPDIR_LEG(Ym,spProjYm,spReconYm);
-  GENERIC_DHOPDIR_LEG(Zm,spProjZm,spReconZm);
-  GENERIC_DHOPDIR_LEG(Tm,spProjTm,spReconTm);
-  vstream(out._odata[sF], result);
-}
-
-/*******************************************************************************
- * Conserved current utilities for Wilson fermions, for contracting propagators
- * to make a conserved current sink or inserting the conserved current 
- * sequentially. Common to both 4D and 5D.
- ******************************************************************************/
-// N.B. Functions below assume a -1/2 factor within U.
-#define WilsonCurrentFwd(expr, mu) ((expr - Gamma::gmu[mu]*expr))
-#define WilsonCurrentBwd(expr, mu) ((expr + Gamma::gmu[mu]*expr))
-
-/*******************************************************************************
- * Name: ContractConservedCurrentSiteFwd
- * Operation: (1/2) * q2[x] * U(x) * (g[mu] - 1) * q1[x + mu]
- * Notes: - DoubledGaugeField U assumed to contain -1/2 factor.
- *        - Pass in q_in_1 shifted in +ve mu direction.
- ******************************************************************************/
-template<class Impl>
-void WilsonKernels<Impl>::ContractConservedCurrentSiteFwd(
-                                                  const SitePropagator &q_in_1,
-                                                  const SitePropagator &q_in_2,
-                                                  SitePropagator &q_out,
-                                                  DoubledGaugeField &U,
-                                                  unsigned int sU,
-                                                  unsigned int mu,
-                                                  bool switch_sign)
-{
-    SitePropagator result, tmp;
-    Gamma g5(Gamma::Algebra::Gamma5);
-    Impl::multLinkProp(tmp, U._odata[sU], q_in_1, mu);
-    result = g5 * adj(q_in_2) * g5 * WilsonCurrentFwd(tmp, mu);
-    if (switch_sign)
-    {
-        q_out -= result;
-    }
-    else
-    {
-        q_out += result;
-    }
-}
-
-/*******************************************************************************
- * Name: ContractConservedCurrentSiteBwd
- * Operation: (1/2) * q2[x + mu] * U^dag(x) * (g[mu] + 1) * q1[x]
- * Notes: - DoubledGaugeField U assumed to contain -1/2 factor.
- *        - Pass in q_in_2 shifted in +ve mu direction.
- ******************************************************************************/
-template<class Impl>
-void WilsonKernels<Impl>::ContractConservedCurrentSiteBwd(
-                                                  const SitePropagator &q_in_1,
-                                                  const SitePropagator &q_in_2,
-                                                  SitePropagator &q_out,
-                                                  DoubledGaugeField &U,
-                                                  unsigned int sU,
-                                                  unsigned int mu,
-                                                  bool switch_sign)
-{
-    SitePropagator result, tmp;
-    Gamma g5(Gamma::Algebra::Gamma5);
-    Impl::multLinkProp(tmp, U._odata[sU], q_in_1, mu + Nd);
-    result = g5 * adj(q_in_2) * g5 * WilsonCurrentBwd(tmp, mu);
-    if (switch_sign)
-    {
-        q_out += result;
-    }
-    else
-    {
-        q_out -= result;
-    }
-}
-
-// G-parity requires more specialised implementation.
-#define NO_CURR_SITE(Impl) \
-template <> \
-void WilsonKernels<Impl>::ContractConservedCurrentSiteFwd( \
-                                                  const SitePropagator &q_in_1, \
-                                                  const SitePropagator &q_in_2, \
-                                                  SitePropagator &q_out,        \
-                                                  DoubledGaugeField &U,         \
-                                                  unsigned int sU,              \
-                                                  unsigned int mu,              \
-                                                  bool switch_sign)             \
-{ \
-    assert(0); \
-} \
-template <> \
-void WilsonKernels<Impl>::ContractConservedCurrentSiteBwd( \
-                                                  const SitePropagator &q_in_1, \
-                                                  const SitePropagator &q_in_2, \
-                                                  SitePropagator &q_out,        \
-                                                  DoubledGaugeField &U,         \
-                                                  unsigned int mu,              \
-                                                  unsigned int sU,              \
-                                                  bool switch_sign)             \
-{ \
-    assert(0); \
-}
-
-NO_CURR_SITE(GparityWilsonImplF);
-NO_CURR_SITE(GparityWilsonImplD);
-NO_CURR_SITE(GparityWilsonImplFH);
-NO_CURR_SITE(GparityWilsonImplDF);
-
-
-/*******************************************************************************
- * Name: SeqConservedCurrentSiteFwd
- * Operation: (1/2) * U(x) * (g[mu] - 1) * q[x + mu]
- * Notes: - DoubledGaugeField U assumed to contain -1/2 factor.
- *        - Pass in q_in shifted in +ve mu direction.
- ******************************************************************************/
-template<class Impl>
-void WilsonKernels<Impl>::SeqConservedCurrentSiteFwd(const SitePropagator &q_in,
-                                                     SitePropagator &q_out,
-                                                     DoubledGaugeField &U,
-                                                     unsigned int sU,
-                                                     unsigned int mu,
-                                                     vInteger t_mask,
-                                                     bool switch_sign)
-{
-    SitePropagator result;
-    Impl::multLinkProp(result, U._odata[sU], q_in, mu);
-    result = WilsonCurrentFwd(result, mu);
-
-    // Zero any unwanted timeslice entries.
-    result = predicatedWhere(t_mask, result, 0.*result);
-
-    if (switch_sign)
-    {
-        q_out -= result;
-    }
-    else
-    {
-        q_out += result;
-    }
-}
-
-/*******************************************************************************
- * Name: SeqConservedCurrentSiteFwd
- * Operation: (1/2) * U^dag(x) * (g[mu] + 1) * q[x - mu]
- * Notes: - DoubledGaugeField U assumed to contain -1/2 factor.
- *        - Pass in q_in shifted in -ve mu direction.
- ******************************************************************************/
-template<class Impl>
-void WilsonKernels<Impl>::SeqConservedCurrentSiteBwd(const SitePropagator &q_in, 
-                                                     SitePropagator &q_out,
-                                                     DoubledGaugeField &U,
-                                                     unsigned int sU,
-                                                     unsigned int mu,
-                                                     vInteger t_mask,
-                                                     bool switch_sign)
-{
-    SitePropagator result;
-    Impl::multLinkProp(result, U._odata[sU], q_in, mu + Nd);
-    result = WilsonCurrentBwd(result, mu);
-
-    // Zero any unwanted timeslice entries.
-    result = predicatedWhere(t_mask, result, 0.*result);
-
-    if (switch_sign)
-    {
-        q_out += result;
-    }
-    else
-    {
-        q_out -= result;
-    }
-}
-
-FermOpTemplateInstantiate(WilsonKernels);
-AdjointFermOpTemplateInstantiate(WilsonKernels);
-TwoIndexFermOpTemplateInstantiate(WilsonKernels);
-
-}}
-
diff --git a/Grid/qcd/action/fermion/WilsonKernels.h b/Grid/qcd/action/fermion/WilsonKernels.h
index 2b150896..35715097 100644
--- a/Grid/qcd/action/fermion/WilsonKernels.h
+++ b/Grid/qcd/action/fermion/WilsonKernels.h
@@ -27,19 +27,17 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
-#ifndef GRID_QCD_DHOP_H
-#define GRID_QCD_DHOP_H
+			   /*  END LEGAL */
+#pragma once
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
-  ////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-  // Helper routines that implement Wilson stencil for a single site.
-  // Common to both the WilsonFermion and WilsonFermion5D
-  ////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// Helper routines that implement Wilson stencil for a single site.
+// Common to both the WilsonFermion and WilsonFermion5D
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////
 class WilsonKernelsStatic { 
- public:
+public:
   enum { OptGeneric, OptHandUnroll, OptInlineAsm };
   enum { CommsAndCompute, CommsThenCompute };
   static int Opt;  
@@ -47,235 +45,123 @@ class WilsonKernelsStatic {
 };
  
 template<class Impl> class WilsonKernels : public FermionOperator<Impl> , public WilsonKernelsStatic { 
- public:
-   
+public:
+
   INHERIT_IMPL_TYPES(Impl);
   typedef FermionOperator<Impl> Base;
    
 public:
 
-  template <bool EnableBool = true>
-  typename std::enable_if<Impl::isFundamental==true && Nc == 3 &&EnableBool, void>::type
-  DhopSite(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-		   int sF, int sU, int Ls, int Ns, const FermionField &in, FermionField &out,int interior=1,int exterior=1) 
-  {
-    bgq_l1p_optimisation(1);
-    switch(Opt) {
-#if defined(AVX512) || defined (QPX)
-    case OptInlineAsm:
-      if(interior&&exterior) WilsonKernels<Impl>::AsmDhopSite   (st,lo,U,buf,sF,sU,Ls,Ns,in,out);
-      else if (interior)     WilsonKernels<Impl>::AsmDhopSiteInt(st,lo,U,buf,sF,sU,Ls,Ns,in,out);
-      else if (exterior)     WilsonKernels<Impl>::AsmDhopSiteExt(st,lo,U,buf,sF,sU,Ls,Ns,in,out);
-      else assert(0);
-      break;
-#endif
-    case OptHandUnroll:
-         for (int site = 0; site < Ns; site++) {
-	   for (int s = 0; s < Ls; s++) {
-	     if(interior&&exterior) WilsonKernels<Impl>::HandDhopSite(st,lo,U,buf,sF,sU,in,out);
-	     else if (interior)     WilsonKernels<Impl>::HandDhopSiteInt(st,lo,U,buf,sF,sU,in,out);
-	     else if (exterior)     WilsonKernels<Impl>::HandDhopSiteExt(st,lo,U,buf,sF,sU,in,out);
-	     sF++;
-	   }
-	   sU++;
-         }
-      break;
-    case OptGeneric:
-         for (int site = 0; site < Ns; site++) {
-	   for (int s = 0; s < Ls; s++) {
-	     if(interior&&exterior) WilsonKernels<Impl>::GenericDhopSite(st,lo,U,buf,sF,sU,in,out);
-	     else if (interior)     WilsonKernels<Impl>::GenericDhopSiteInt(st,lo,U,buf,sF,sU,in,out);
-	     else if (exterior)     WilsonKernels<Impl>::GenericDhopSiteExt(st,lo,U,buf,sF,sU,in,out);
-	     else assert(0);
-	     sF++;
-	   }
-	   sU++;
-       } 
-      break;
-    default:
-      assert(0);
-    }
-    bgq_l1p_optimisation(0);
-  }
-     
-  template <bool EnableBool = true>
-  typename std::enable_if<(Impl::isFundamental==false || (Impl::isFundamental==true && Nc != 3)) && EnableBool, void>::type
-  DhopSite(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-	   int sF, int sU, int Ls, int Ns, const FermionField &in, FermionField &out,int interior=1,int exterior=1 ) {
-    // no kernel choice  
-    for (int site = 0; site < Ns; site++) {
-      for (int s = 0; s < Ls; s++) {
-	if(interior&&exterior) WilsonKernels<Impl>::GenericDhopSite(st,lo,U,buf,sF,sU,in,out);
-	else if (interior)     WilsonKernels<Impl>::GenericDhopSiteInt(st,lo,U,buf,sF,sU,in,out);
-	else if (exterior)     WilsonKernels<Impl>::GenericDhopSiteExt(st,lo,U,buf,sF,sU,in,out);
-	else assert(0);
-	sF++;
-      }
-      sU++;
-    }
-  }
-     
-  template <bool EnableBool = true>
-  typename std::enable_if<Impl::isFundamental==true && Nc == 3 && EnableBool,void>::type
-  DhopSiteDag(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-	      int sF, int sU, int Ls, int Ns, const FermionField &in, FermionField &out,int interior=1,int exterior=1) 
-{
-    bgq_l1p_optimisation(1);
-    switch(Opt) {
-#if defined(AVX512) || defined (QPX)
-    case OptInlineAsm:
-      if(interior&&exterior) WilsonKernels<Impl>::AsmDhopSiteDag   (st,lo,U,buf,sF,sU,Ls,Ns,in,out);
-      else if (interior)     WilsonKernels<Impl>::AsmDhopSiteDagInt(st,lo,U,buf,sF,sU,Ls,Ns,in,out);
-      else if (exterior)     WilsonKernels<Impl>::AsmDhopSiteDagExt(st,lo,U,buf,sF,sU,Ls,Ns,in,out);
-      else assert(0);
-      break;
-#endif
-    case OptHandUnroll:
-      for (int site = 0; site < Ns; site++) {
-	for (int s = 0; s < Ls; s++) {
-	  if(interior&&exterior) WilsonKernels<Impl>::HandDhopSiteDag(st,lo,U,buf,sF,sU,in,out);
-	  else if (interior)     WilsonKernels<Impl>::HandDhopSiteDagInt(st,lo,U,buf,sF,sU,in,out);
-	  else if (exterior)     WilsonKernels<Impl>::HandDhopSiteDagExt(st,lo,U,buf,sF,sU,in,out);
-	  else assert(0);
-	  sF++;
-	}
-	sU++;
-      }
-      break;
-    case OptGeneric:
-      for (int site = 0; site < Ns; site++) {
-	for (int s = 0; s < Ls; s++) {
-	  if(interior&&exterior) WilsonKernels<Impl>::GenericDhopSiteDag(st,lo,U,buf,sF,sU,in,out);
-	  else if (interior)     WilsonKernels<Impl>::GenericDhopSiteDagInt(st,lo,U,buf,sF,sU,in,out);
-	  else if (exterior)     WilsonKernels<Impl>::GenericDhopSiteDagExt(st,lo,U,buf,sF,sU,in,out);
-	  else assert(0);
-	  sF++;
-	}
-	sU++;
-      }
-      break;
-    default:
-      assert(0);
-    }
-    bgq_l1p_optimisation(0);
-  }
+  static void DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField &U, SiteHalfSpinor * buf,
+			 int Ls, int Nsite, const FermionField &in, FermionField &out,
+			 int interior=1,int exterior=1) ;
 
-  template <bool EnableBool = true>
-  typename std::enable_if<(Impl::isFundamental==false || (Impl::isFundamental==true && Nc != 3)) && EnableBool,void>::type
-  DhopSiteDag(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,SiteHalfSpinor * buf,
-		      int sF, int sU, int Ls, int Ns, const FermionField &in, FermionField &out,int interior=1,int exterior=1) {
+  static void DhopDagKernel(int Opt,StencilImpl &st,  DoubledGaugeField &U, SiteHalfSpinor * buf,
+			    int Ls, int Nsite, const FermionField &in, FermionField &out,
+			    int interior=1,int exterior=1) ;
 
-    for (int site = 0; site < Ns; site++) {
-      for (int s = 0; s < Ls; s++) {
-	if(interior&&exterior) WilsonKernels<Impl>::GenericDhopSiteDag(st,lo,U,buf,sF,sU,in,out);
-	else if (interior)     WilsonKernels<Impl>::GenericDhopSiteDagInt(st,lo,U,buf,sF,sU,in,out);
-	else if (exterior)     WilsonKernels<Impl>::GenericDhopSiteDagExt(st,lo,U,buf,sF,sU,in,out);
-	else assert(0);
-	sF++;
-      }
-      sU++;
-    }
-  }
+  static void DhopDirKernel(StencilImpl &st, DoubledGaugeField &U,SiteHalfSpinor * buf,
+			    int Ls, int Nsite, const FermionField &in, FermionField &out, int dirdisp, int gamma);
 
-  void DhopDir(StencilImpl &st, DoubledGaugeField &U,SiteHalfSpinor * buf,
-		       int sF, int sU, const FermionField &in, FermionField &out, int dirdisp, int gamma);
-      
   //////////////////////////////////////////////////////////////////////////////
   // Utilities for inserting Wilson conserved current.
   //////////////////////////////////////////////////////////////////////////////
-  void ContractConservedCurrentSiteFwd(const SitePropagator &q_in_1,
+  static void ContractConservedCurrentSiteFwd(const SitePropagator &q_in_1,
                                        const SitePropagator &q_in_2,
                                        SitePropagator &q_out,
-                                       DoubledGaugeField &U,
+                                       DoubledGaugeFieldView &U,
                                        unsigned int sU,
                                        unsigned int mu,
                                        bool switch_sign = false);
-  void ContractConservedCurrentSiteBwd(const SitePropagator &q_in_1,
+
+  static void ContractConservedCurrentSiteBwd(const SitePropagator &q_in_1,
                                        const SitePropagator &q_in_2,
                                        SitePropagator &q_out,
-                                       DoubledGaugeField &U,
+                                       DoubledGaugeFieldView &U,
                                        unsigned int sU,
                                        unsigned int mu,
                                        bool switch_sign = false);
-  void SeqConservedCurrentSiteFwd(const SitePropagator &q_in, 
+
+  static void SeqConservedCurrentSiteFwd(const SitePropagator &q_in, 
                                   SitePropagator &q_out,
-                                  DoubledGaugeField &U,
+                                  DoubledGaugeFieldView &U,
                                   unsigned int sU,
                                   unsigned int mu,
-                                  vInteger t_mask,
+                                  vPredicate t_mask,
                                   bool switch_sign = false);
-  void SeqConservedCurrentSiteBwd(const SitePropagator &q_in,
+
+  static void SeqConservedCurrentSiteBwd(const SitePropagator &q_in,
                                   SitePropagator &q_out,
-                                  DoubledGaugeField &U,
+                                  DoubledGaugeFieldView &U,
                                   unsigned int sU,
                                   unsigned int mu,
-                                  vInteger t_mask,
+                                  vPredicate t_mask,
                                   bool switch_sign = false);
 
 private:
-     // Specialised variants
-  void GenericDhopSite(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-		       int sF, int sU, const FermionField &in, FermionField &out);
+
+  static accelerator void DhopDirK(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor * buf,
+				   int sF, int sU, const FermionFieldView &in, FermionFieldView &out, int dirdisp, int gamma);
       
-  void GenericDhopSiteDag(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-			  int sF, int sU, const FermionField &in, FermionField &out);
-
-  void GenericDhopSiteInt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-			  int sF, int sU, const FermionField &in, FermionField &out);
+  // Specialised variants
+  static accelerator void GenericDhopSite(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+					  int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
       
-  void GenericDhopSiteDagInt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-			     int sF, int sU, const FermionField &in, FermionField &out);
-
-  void GenericDhopSiteExt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-			  int sF, int sU, const FermionField &in, FermionField &out);
+  static accelerator void GenericDhopSiteDag(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+						    int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+  
+  static accelerator void GenericDhopSiteInt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+						    int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
       
-  void GenericDhopSiteDagExt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-			     int sF, int sU, const FermionField &in, FermionField &out);
-
-
-  void AsmDhopSite(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-		   int sF, int sU, int Ls, int Ns, const FermionField &in,FermionField &out);
-
-  void AsmDhopSiteDag(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-		      int sF, int sU, int Ls, int Ns, const FermionField &in, FermionField &out);
-
-  void AsmDhopSiteInt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-		      int sF, int sU, int Ls, int Ns, const FermionField &in,FermionField &out);
-
-  void AsmDhopSiteDagInt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-			 int sF, int sU, int Ls, int Ns, const FermionField &in, FermionField &out);
-
-  void AsmDhopSiteExt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-		      int sF, int sU, int Ls, int Ns, const FermionField &in,FermionField &out);
-
-  void AsmDhopSiteDagExt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-			 int sF, int sU, int Ls, int Ns, const FermionField &in, FermionField &out);
-
-
-  void HandDhopSite(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-		    int sF, int sU, const FermionField &in, FermionField &out);
-
-  void HandDhopSiteDag(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-		       int sF, int sU, const FermionField &in, FermionField &out);
+  static accelerator void GenericDhopSiteDagInt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+						int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+  
+  static accelerator void GenericDhopSiteExt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+					     int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
       
-  void HandDhopSiteInt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-		       int sF, int sU, const FermionField &in, FermionField &out);
-  
-  void HandDhopSiteDagInt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-			  int sF, int sU, const FermionField &in, FermionField &out);
-  
-  void HandDhopSiteExt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-		       int sF, int sU, const FermionField &in, FermionField &out);
-  
-  void HandDhopSiteDagExt(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, SiteHalfSpinor * buf,
-			  int sF, int sU, const FermionField &in, FermionField &out);
-  
-public:
+  static accelerator void GenericDhopSiteDagExt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+						       int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
 
-  WilsonKernels(const ImplParams &p = ImplParams());
+  static void AsmDhopSite(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+			  int sF, int sU, int Ls, int Nsite, const FermionFieldView &in,FermionFieldView &out);
+  
+  static void AsmDhopSiteDag(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+			     int sF, int sU, int Ls, int Nsite, const FermionFieldView &in, FermionFieldView &out);
+  
+  static void AsmDhopSiteInt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+			     int sF, int sU, int Ls, int Nsite, const FermionFieldView &in,FermionFieldView &out);
+  
+  static void AsmDhopSiteDagInt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+				int sF, int sU, int Ls, int Nsite, const FermionFieldView &in, FermionFieldView &out);
+  
+  static void AsmDhopSiteExt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+			     int sF, int sU, int Ls, int Nsite, const FermionFieldView &in,FermionFieldView &out);
+  
+  static void AsmDhopSiteDagExt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+				int sF, int sU, int Ls, int Nsite, const FermionFieldView &in, FermionFieldView &out);
 
+// Keep Hand unrolled temporarily  
+  static accelerator void HandDhopSite(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+				       int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+  
+  static accelerator void HandDhopSiteDag(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+					  int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+  
+  static accelerator void HandDhopSiteInt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+					  int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+  
+  static accelerator void HandDhopSiteDagInt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+					     int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+  
+  static accelerator void HandDhopSiteExt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+					  int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+  
+  static accelerator void HandDhopSiteDagExt(StencilView &st,  DoubledGaugeFieldView &U, SiteHalfSpinor * buf,
+					     int sF, int sU, const FermionFieldView &in, FermionFieldView &out);
+ public:
+ WilsonKernels(const ImplParams &p = ImplParams()) : Base(p){};
 };
     
-}}
+NAMESPACE_END(Grid);
+
 
-#endif
diff --git a/Grid/qcd/action/fermion/WilsonKernelsAsm.cc b/Grid/qcd/action/fermion/WilsonKernelsAsm.cc
deleted file mode 100644
index 55911988..00000000
--- a/Grid/qcd/action/fermion/WilsonKernelsAsm.cc
+++ /dev/null
@@ -1,127 +0,0 @@
-/*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid 
-
-
-
-    Source file: ./lib/qcd/action/fermion/WilsonKernelsAsm.cc
-
-    Copyright (C) 2015
-
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-Author: Guido Cossu <guido.cossu@ed.ac.uk>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-#include <Grid/qcd/action/fermion/FermionCore.h>
-
-namespace Grid {
-namespace QCD {
-
-
-///////////////////////////////////////////////////////////
-// Default to no assembler implementation
-///////////////////////////////////////////////////////////
-template<class Impl> void 
-WilsonKernels<Impl >::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-					  int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-{
-  assert(0);
-}
-
-template<class Impl> void 
-WilsonKernels<Impl >::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-					     int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-{
-  assert(0);
-}
-
-template<class Impl> void 
-WilsonKernels<Impl >::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-					  int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-{
-  assert(0);
-}
-
-template<class Impl> void 
-WilsonKernels<Impl >::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-					     int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-{
-  assert(0);
-}
-
-template<class Impl> void 
-WilsonKernels<Impl >::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-					  int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-{
-  assert(0);
-}
-
-template<class Impl> void 
-WilsonKernels<Impl >::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-					     int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-{
-  assert(0);
-}
-
-#include <Grid/qcd/action/fermion/WilsonKernelsAsmAvx512.h>
-#include <Grid/qcd/action/fermion/WilsonKernelsAsmQPX.h>
-
-#define INSTANTIATE_ASM(A)\
-template void WilsonKernels<A>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,\
-                                  int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out);\
- \
-template void WilsonKernels<A>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,\
-                                  int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out);\
-template void WilsonKernels<A>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,\
-                                  int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out);\
- \
-template void WilsonKernels<A>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,\
-                                  int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out);\
-template void WilsonKernels<A>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,\
-                                  int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out);\
- \
-template void WilsonKernels<A>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,\
-                                  int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out);\
-
-INSTANTIATE_ASM(WilsonImplF);
-INSTANTIATE_ASM(WilsonImplD);
-INSTANTIATE_ASM(ZWilsonImplF);
-INSTANTIATE_ASM(ZWilsonImplD);
-INSTANTIATE_ASM(GparityWilsonImplF);
-INSTANTIATE_ASM(GparityWilsonImplD);
-INSTANTIATE_ASM(DomainWallVec5dImplF);
-INSTANTIATE_ASM(DomainWallVec5dImplD);
-INSTANTIATE_ASM(ZDomainWallVec5dImplF);
-INSTANTIATE_ASM(ZDomainWallVec5dImplD);
-
-INSTANTIATE_ASM(WilsonImplFH);
-INSTANTIATE_ASM(WilsonImplDF);
-INSTANTIATE_ASM(ZWilsonImplFH);
-INSTANTIATE_ASM(ZWilsonImplDF);
-INSTANTIATE_ASM(GparityWilsonImplFH);
-INSTANTIATE_ASM(GparityWilsonImplDF);
-INSTANTIATE_ASM(DomainWallVec5dImplFH);
-INSTANTIATE_ASM(DomainWallVec5dImplDF);
-INSTANTIATE_ASM(ZDomainWallVec5dImplFH);
-INSTANTIATE_ASM(ZDomainWallVec5dImplDF);
-
-}}
-
diff --git a/Grid/qcd/action/fermion/WilsonKernelsAsmAvx512.h b/Grid/qcd/action/fermion/WilsonKernelsAsmAvx512.h
deleted file mode 100644
index 948c16a2..00000000
--- a/Grid/qcd/action/fermion/WilsonKernelsAsmAvx512.h
+++ /dev/null
@@ -1,650 +0,0 @@
-/*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid 
-
-
-
-    Source file: ./lib/qcd/action/fermion/WilsonKernelsAsmAvx512.h
-
-    Copyright (C) 2015
-
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-
-#if defined(AVX512) 
-    ///////////////////////////////////////////////////////////
-    // If we are AVX512 specialise the single precision routine
-    ///////////////////////////////////////////////////////////
-#include <simd/Intel512wilson.h>
-#include <simd/Intel512single.h>
-    
-static Vector<vComplexF> signsF;
-
-  template<typename vtype>    
-  int setupSigns(Vector<vtype>& signs ){
-    Vector<vtype> bother(2);
-    signs = bother;
-    vrsign(signs[0]);
-    visign(signs[1]);
-    return 1;
-  }
-
-  static int signInitF = setupSigns(signsF);
-
-#define MAYBEPERM(A,perm) if (perm) { A ; }
-#define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN(ptr,pf)
-#define COMPLEX_SIGNS(isigns) vComplexF *isigns = &signsF[0];  
-  
-/////////////////////////////////////////////////////////////////
-// XYZT vectorised, undag Kernel, single
-/////////////////////////////////////////////////////////////////
-#undef KERNEL_DAG
-#define INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<WilsonImplF>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<ZWilsonImplF>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<WilsonImplFH>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<ZWilsonImplFH>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#define INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<WilsonImplF>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<ZWilsonImplF>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<WilsonImplFH>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<ZWilsonImplFH>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-
-#undef INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#define EXTERIOR
-template<> void 
-WilsonKernels<WilsonImplF>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<ZWilsonImplF>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<WilsonImplFH>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<ZWilsonImplFH>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-      
-/////////////////////////////////////////////////////////////////
-// XYZT vectorised, dag Kernel, single
-/////////////////////////////////////////////////////////////////
-#define KERNEL_DAG
-#define INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<WilsonImplF>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<ZWilsonImplF>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<WilsonImplFH>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<ZWilsonImplFH>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#define INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<WilsonImplF>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<ZWilsonImplF>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<WilsonImplFH>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<ZWilsonImplFH>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#define EXTERIOR
-template<> void 
-WilsonKernels<WilsonImplF>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-				    
-template<> void 
-WilsonKernels<ZWilsonImplF>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-				    
-template<> void 
-WilsonKernels<WilsonImplFH>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-				    
-template<> void 
-WilsonKernels<ZWilsonImplFH>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-				    
-#undef MAYBEPERM
-#undef MULT_2SPIN
-#define MAYBEPERM(A,B) 
-#define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN_LS(ptr,pf)
-				    
-/////////////////////////////////////////////////////////////////
-// Ls vectorised, undag Kernel, single
-/////////////////////////////////////////////////////////////////
-#undef KERNEL_DAG
-#define INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<DomainWallVec5dImplF>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<DomainWallVec5dImplFH>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplFH>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#define INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<DomainWallVec5dImplF>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<DomainWallVec5dImplFH>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplFH>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#define EXTERIOR
-#undef  MULT_2SPIN
-#define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN_LSNOPF(ptr,pf)
-template<> void 
-WilsonKernels<DomainWallVec5dImplF>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-				    
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-				    
-template<> void 
-WilsonKernels<DomainWallVec5dImplFH>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-				    
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplFH>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-				    
-/////////////////////////////////////////////////////////////////
-// Ls vectorised, dag Kernel, single
-/////////////////////////////////////////////////////////////////
-#define KERNEL_DAG
-#define INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<DomainWallVec5dImplF>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<DomainWallVec5dImplFH>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplFH>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#define INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<DomainWallVec5dImplF>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<DomainWallVec5dImplFH>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplFH>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#define EXTERIOR
-template<> void 
-WilsonKernels<DomainWallVec5dImplF>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<DomainWallVec5dImplFH>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplFH>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef COMPLEX_SIGNS
-#undef MAYBEPERM
-#undef MULT_2SPIN
-	
-
-
-///////////////////////////////////////////////////////////
-// If we are AVX512 specialise the double precision routine
-///////////////////////////////////////////////////////////
-
-#include <simd/Intel512double.h>
-    
-static Vector<vComplexD> signsD;
-static int signInitD = setupSigns(signsD);
-    
-#define MAYBEPERM(A,perm) if (perm) { A ; }
-#define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN(ptr,pf)
-#define COMPLEX_SIGNS(isigns) vComplexD *isigns = &signsD[0];  
-
-
-#define INTERIOR_AND_EXTERIOR    
-#undef  INTERIOR
-#undef  EXTERIOR
-  
-/////////////////////////////////////////////////////////////////
-// XYZT vectorised, undag Kernel, single
-/////////////////////////////////////////////////////////////////
-#undef KERNEL_DAG
-#define INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<WilsonImplD>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZWilsonImplD>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<WilsonImplDF>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZWilsonImplDF>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#define INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<WilsonImplD>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZWilsonImplD>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<WilsonImplDF>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZWilsonImplDF>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#define EXTERIOR
-template<> void 
-WilsonKernels<WilsonImplD>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZWilsonImplD>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-      
-template<> void 
-WilsonKernels<WilsonImplDF>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZWilsonImplDF>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-      
-/////////////////////////////////////////////////////////////////
-// XYZT vectorised, dag Kernel, single
-/////////////////////////////////////////////////////////////////
-#define KERNEL_DAG
-#define INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<WilsonImplD>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZWilsonImplD>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<WilsonImplDF>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZWilsonImplDF>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#define INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<WilsonImplD>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZWilsonImplD>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<WilsonImplDF>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZWilsonImplDF>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#define EXTERIOR
-template<> void 
-WilsonKernels<WilsonImplD>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZWilsonImplD>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-				    
-template<> void 
-WilsonKernels<WilsonImplDF>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZWilsonImplDF>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-				    
-#undef MAYBEPERM
-#undef MULT_2SPIN
-#define MAYBEPERM(A,B) 
-#define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN_LS(ptr,pf)
-				    
-/////////////////////////////////////////////////////////////////
-// Ls vectorised, undag Kernel, single
-/////////////////////////////////////////////////////////////////
-#undef KERNEL_DAG
-#define INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<DomainWallVec5dImplD>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<DomainWallVec5dImplDF>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplDF>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#define INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<DomainWallVec5dImplD>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<DomainWallVec5dImplDF>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplDF>::AsmDhopSiteInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#define EXTERIOR
-#undef  MULT_2SPIN
-#define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN_LSNOPF(ptr,pf)
-template<> void 
-WilsonKernels<DomainWallVec5dImplD>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-				    
-template<> void 
-WilsonKernels<DomainWallVec5dImplDF>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplDF>::AsmDhopSiteExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
-							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-				    
-/////////////////////////////////////////////////////////////////
-// Ls vectorised, dag Kernel, single
-/////////////////////////////////////////////////////////////////
-#define KERNEL_DAG
-#define INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<DomainWallVec5dImplD>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<DomainWallVec5dImplDF>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplDF>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#define INTERIOR
-#undef EXTERIOR
-template<> void 
-WilsonKernels<DomainWallVec5dImplD>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<DomainWallVec5dImplDF>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplDF>::AsmDhopSiteDagInt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef INTERIOR_AND_EXTERIOR
-#undef INTERIOR
-#define EXTERIOR
-template<> void 
-WilsonKernels<DomainWallVec5dImplD>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-template<> void 
-WilsonKernels<DomainWallVec5dImplDF>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-template<> void 
-WilsonKernels<ZDomainWallVec5dImplDF>::AsmDhopSiteDagExt(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
-#undef COMPLEX_SIGNS
-#undef MAYBEPERM
-#undef MULT_2SPIN
-
-#endif //AVX512
diff --git a/Grid/qcd/action/fermion/WilsonTMFermion.cc b/Grid/qcd/action/fermion/WilsonTMFermion.cc
deleted file mode 100644
index d4604b10..00000000
--- a/Grid/qcd/action/fermion/WilsonTMFermion.cc
+++ /dev/null
@@ -1,99 +0,0 @@
-    /*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid 
-
-    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
-
-    Copyright (C) 2015
-
-Author: paboyle <paboyle@ph.ed.ac.uk>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#include <Grid/qcd/action/fermion/FermionCore.h>
-#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
-
-namespace Grid {
-namespace QCD {
-
-    /*
-     * BF sequence
-     *
-      void bfmbase<Float>::MooeeInv(Fermion_t psi, 
-			       Fermion_t chi, 
-			      int dag, int cb)
-
-    double m    = this->mass;
-    double tm   = this->twistedmass;
-    double mtil = 4.0+this->mass;
-
-    double sq = mtil*mtil + tm*tm;
-
-    double a = mtil/sq;
-    double b = -tm /sq;
-    if(dag) b=-b;
-    axpibg5x(chi,psi,a,b);
-
-      void bfmbase<Float>::Mooee(Fermion_t psi, 
-			   Fermion_t chi, 
-			   int dag,int cb)
-    double a = 4.0+this->mass;
-    double b = this->twistedmass;
-    if(dag) b=-b;
-    axpibg5x(chi,psi,a,b);
-    */
-
-  template<class Impl>
-  void WilsonTMFermion<Impl>::Mooee(const FermionField &in, FermionField &out) {
-    RealD a = 4.0+this->mass;
-    RealD b = this->mu;
-    out.checkerboard = in.checkerboard;
-    axpibg5x(out,in,a,b);
-  }
-  template<class Impl>
-  void WilsonTMFermion<Impl>::MooeeDag(const FermionField &in, FermionField &out) {
-    RealD a = 4.0+this->mass;
-    RealD b = -this->mu;
-    out.checkerboard = in.checkerboard;
-    axpibg5x(out,in,a,b);
-  }
-  template<class Impl>
-  void WilsonTMFermion<Impl>::MooeeInv(const FermionField &in, FermionField &out) {
-    RealD m    = this->mass;
-    RealD tm   = this->mu;
-    RealD mtil = 4.0+this->mass;
-    RealD sq   = mtil*mtil+tm*tm;
-    RealD a    = mtil/sq;
-    RealD b    = -tm /sq;
-    axpibg5x(out,in,a,b);
-  }
-  template<class Impl>
-  void WilsonTMFermion<Impl>::MooeeInvDag(const FermionField &in, FermionField &out) {
-    RealD m    = this->mass;
-    RealD tm   = this->mu;
-    RealD mtil = 4.0+this->mass;
-    RealD sq   = mtil*mtil+tm*tm;
-    RealD a    = mtil/sq;
-    RealD b    = tm /sq;
-    axpibg5x(out,in,a,b);
-  }
-
-  FermOpTemplateInstantiate(WilsonTMFermion);
-
-}
-}
diff --git a/Grid/qcd/action/fermion/WilsonTMFermion.h b/Grid/qcd/action/fermion/WilsonTMFermion.h
index f75c287b..12c4b71a 100644
--- a/Grid/qcd/action/fermion/WilsonTMFermion.h
+++ b/Grid/qcd/action/fermion/WilsonTMFermion.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,55 +23,52 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef  GRID_QCD_WILSON_TM_FERMION_H
-#define  GRID_QCD_WILSON_TM_FERMION_H
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once 
 
 #include <Grid/qcd/action/fermion/FermionCore.h>
 #include <Grid/qcd/action/fermion/WilsonFermion.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class WilsonTMFermion : public WilsonFermion<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+public:
 
-    template<class Impl>
-    class WilsonTMFermion : public WilsonFermion<Impl>
-    {
-    public:
-      INHERIT_IMPL_TYPES(Impl);
-    public:
+  virtual void   Instantiatable(void) {};
+  // Constructors
+  WilsonTMFermion(GaugeField &_Umu,
+		  GridCartesian         &Fgrid,
+		  GridRedBlackCartesian &Hgrid, 
+		  RealD _mass,
+		  RealD _mu,
+		  const ImplParams &p= ImplParams()
+		  ) :
+    WilsonFermion<Impl>(_Umu,
+			Fgrid,
+			Hgrid,
+			_mass,p)
 
-      virtual void   Instantiatable(void) {};
-      // Constructors
-      WilsonTMFermion(GaugeField &_Umu,
-		    GridCartesian         &Fgrid,
-		    GridRedBlackCartesian &Hgrid, 
-		    RealD _mass,
-		    RealD _mu,
-		    const ImplParams &p= ImplParams()
-		      ) :
-	WilsonFermion<Impl>(_Umu,
-			    Fgrid,
-			    Hgrid,
-			    _mass,p)
-
-      {
-	mu = _mu;
-      }
+  {
+    mu = _mu;
+  }
 
 
-    // allow override for twisted mass and clover
-    virtual void Mooee(const FermionField &in, FermionField &out) ;
-    virtual void MooeeDag(const FermionField &in, FermionField &out) ;
-    virtual void MooeeInv(const FermionField &in, FermionField &out) ;
-    virtual void MooeeInvDag(const FermionField &in, FermionField &out) ;
+  // allow override for twisted mass and clover
+  virtual void Mooee(const FermionField &in, FermionField &out) ;
+  virtual void MooeeDag(const FermionField &in, FermionField &out) ;
+  virtual void MooeeInv(const FermionField &in, FermionField &out) ;
+  virtual void MooeeInvDag(const FermionField &in, FermionField &out) ;
 
-  private:
-     RealD mu; // TwistedMass parameter
+private:
+  RealD mu; // TwistedMass parameter
 
-  };
+};
+
+NAMESPACE_END(Grid);
 
-}}
 
-#endif
diff --git a/Grid/qcd/action/fermion/WilsonTMFermion5D.h b/Grid/qcd/action/fermion/WilsonTMFermion5D.h
index 6c75d908..71acf763 100644
--- a/Grid/qcd/action/fermion/WilsonTMFermion5D.h
+++ b/Grid/qcd/action/fermion/WilsonTMFermion5D.h
@@ -30,126 +30,123 @@ Author: paboyle <paboyle@ph.ed.ac.uk> ; NB Christoph did similar in GPT
 #include <Grid/qcd/action/fermion/FermionCore.h>
 #include <Grid/qcd/action/fermion/WilsonFermion.h>
 
-
-namespace Grid {
-
-  namespace QCD {
+NAMESPACE_BEGIN(Grid);
     
-    template<class Impl>
-      class WilsonTMFermion5D : public WilsonFermion5D<Impl>
-      {
-      public:
-	INHERIT_IMPL_TYPES(Impl);
-      public:
-
-	virtual void   Instantiatable(void) {};
-
-	// Constructors
-        WilsonTMFermion5D(GaugeField &_Umu,
-			  GridCartesian         &Fgrid,
-			  GridRedBlackCartesian &Frbgrid, 
-			  GridCartesian         &Ugrid,
-			  GridRedBlackCartesian &Urbgrid, 
-			  const std::vector<RealD> _mass,
-			  const std::vector<RealD> _mu,
-			  const ImplParams &p= ImplParams()
-			  ) :
-	WilsonFermion5D<Impl>(_Umu,
-			      Fgrid,
-			      Frbgrid,
-			      Ugrid,
-			      Urbgrid,
-			      4.0,p)
-	
-	  {
-	    update(_mass,_mu);
-	  }
-
-	virtual void Meooe(const FermionField &in, FermionField &out) {
-	  if (in.checkerboard == Odd) {
-	    this->DhopEO(in, out, DaggerNo);
-	  } else {
-	    this->DhopOE(in, out, DaggerNo);
-	  }
-	}
-
-	virtual void MeooeDag(const FermionField &in, FermionField &out) {
-	  if (in.checkerboard == Odd) {
-	    this->DhopEO(in, out, DaggerYes);
-	  } else {
-	    this->DhopOE(in, out, DaggerYes);
-	  }
-	}	
-	
-	// allow override for twisted mass and clover
-	virtual void Mooee(const FermionField &in, FermionField &out) {
-	  out.checkerboard = in.checkerboard;
-	  //axpibg5x(out,in,a,b); // out = a*in + b*i*G5*in
-	  for (int s=0;s<(int)this->mass.size();s++) {
-	    ComplexD a = 4.0+this->mass[s];
-	    ComplexD b(0.0,this->mu[s]);
-	    axpbg5y_ssp(out,a,in,b,in,s,s);
-	  }
-	}
-
-	virtual void MooeeDag(const FermionField &in, FermionField &out) {
-	  out.checkerboard = in.checkerboard;
-	  for (int s=0;s<(int)this->mass.size();s++) {
-	    ComplexD a = 4.0+this->mass[s];
-	    ComplexD b(0.0,-this->mu[s]);
-	    axpbg5y_ssp(out,a,in,b,in,s,s);
-	  }
-	}
-	virtual void MooeeInv(const FermionField &in, FermionField &out) {
-	  for (int s=0;s<(int)this->mass.size();s++) {
-	    RealD m    = this->mass[s];
-	    RealD tm   = this->mu[s];
-	    RealD mtil = 4.0+this->mass[s];
-	    RealD sq   = mtil*mtil+tm*tm;
-	    ComplexD a    = mtil/sq;
-	    ComplexD b(0.0, -tm /sq);
-	    axpbg5y_ssp(out,a,in,b,in,s,s);
-	  }
-	}
-	virtual void MooeeInvDag(const FermionField &in, FermionField &out) {
-	  for (int s=0;s<(int)this->mass.size();s++) {
-	    RealD m    = this->mass[s];
-	    RealD tm   = this->mu[s];
-	    RealD mtil = 4.0+this->mass[s];
-	    RealD sq   = mtil*mtil+tm*tm;
-	    ComplexD a    = mtil/sq;
-	    ComplexD b(0.0,tm /sq);
-	    axpbg5y_ssp(out,a,in,b,in,s,s);
-	  }
-	}
-
-	virtual RealD M(const FermionField &in, FermionField &out) {
-	  out.checkerboard = in.checkerboard;
-	  this->Dhop(in, out, DaggerNo);
-	  FermionField tmp(out._grid);
-	  for (int s=0;s<(int)this->mass.size();s++) {
-	    ComplexD a = 4.0+this->mass[s];
-	    ComplexD b(0.0,this->mu[s]);
-	    axpbg5y_ssp(tmp,a,in,b,in,s,s);
-	  }
-	  return axpy_norm(out, 1.0, tmp, out);
-	}
-	
-	// needed for fast PV
-	void update(const std::vector<RealD>& _mass, const std::vector<RealD>& _mu) {
-	  assert(_mass.size() == _mu.size());
-	  assert(_mass.size() == this->FermionGrid()->_fdimensions[0]);
-	  this->mass = _mass;
-	  this->mu = _mu;
-	}
-	
-      private:
-	std::vector<RealD> mu;
-	std::vector<RealD> mass;
-	
-      };
+template<class Impl>
+class WilsonTMFermion5D : public WilsonFermion5D<Impl>
+{
+ public:
+  INHERIT_IMPL_TYPES(Impl);
+ public:
+  
+  virtual void   Instantiatable(void) {};
+  
+  // Constructors
+ WilsonTMFermion5D(GaugeField &_Umu,
+		   GridCartesian         &Fgrid,
+		   GridRedBlackCartesian &Frbgrid, 
+		   GridCartesian         &Ugrid,
+		   GridRedBlackCartesian &Urbgrid, 
+		   const std::vector<RealD> _mass,
+		   const std::vector<RealD> _mu,
+		   const ImplParams &p= ImplParams()
+		   ) :
+  WilsonFermion5D<Impl>(_Umu,
+			Fgrid,
+			Frbgrid,
+			Ugrid,
+			Urbgrid,
+			4.0,p)
    
-    typedef WilsonTMFermion5D<WilsonImplF> WilsonTMFermion5DF; 
-    typedef WilsonTMFermion5D<WilsonImplD> WilsonTMFermion5DD; 
+    {
+      update(_mass,_mu);
+    }
+  
+  virtual void Meooe(const FermionField &in, FermionField &out) {
+    if (in.Checkerboard() == Odd) {
+      this->DhopEO(in, out, DaggerNo);
+    } else {
+      this->DhopOE(in, out, DaggerNo);
+    }
+  }
+  
+  virtual void MeooeDag(const FermionField &in, FermionField &out) {
+    if (in.Checkerboard() == Odd) {
+      this->DhopEO(in, out, DaggerYes);
+    } else {
+      this->DhopOE(in, out, DaggerYes);
+    }
+  }	
+  
+  // allow override for twisted mass and clover
+  virtual void Mooee(const FermionField &in, FermionField &out) {
+    out.Checkerboard() = in.Checkerboard();
+    //axpibg5x(out,in,a,b); // out = a*in + b*i*G5*in
+    for (int s=0;s<(int)this->mass.size();s++) {
+      ComplexD a = 4.0+this->mass[s];
+      ComplexD b(0.0,this->mu[s]);
+      axpbg5y_ssp(out,a,in,b,in,s,s);
+    }
+  }
+  
+  virtual void MooeeDag(const FermionField &in, FermionField &out) {
+    out.Checkerboard() = in.Checkerboard();
+    for (int s=0;s<(int)this->mass.size();s++) {
+      ComplexD a = 4.0+this->mass[s];
+      ComplexD b(0.0,-this->mu[s]);
+      axpbg5y_ssp(out,a,in,b,in,s,s);
+    }
+  }
+  virtual void MooeeInv(const FermionField &in, FermionField &out) {
+    for (int s=0;s<(int)this->mass.size();s++) {
+      RealD m    = this->mass[s];
+      RealD tm   = this->mu[s];
+      RealD mtil = 4.0+this->mass[s];
+      RealD sq   = mtil*mtil+tm*tm;
+      ComplexD a    = mtil/sq;
+      ComplexD b(0.0, -tm /sq);
+      axpbg5y_ssp(out,a,in,b,in,s,s);
+    }
+  }
+  virtual void MooeeInvDag(const FermionField &in, FermionField &out) {
+    for (int s=0;s<(int)this->mass.size();s++) {
+      RealD m    = this->mass[s];
+      RealD tm   = this->mu[s];
+      RealD mtil = 4.0+this->mass[s];
+      RealD sq   = mtil*mtil+tm*tm;
+      ComplexD a    = mtil/sq;
+      ComplexD b(0.0,tm /sq);
+      axpbg5y_ssp(out,a,in,b,in,s,s);
+    }
+  }
+  
+  virtual RealD M(const FermionField &in, FermionField &out) {
+    out.Checkerboard() = in.Checkerboard();
+    this->Dhop(in, out, DaggerNo);
+    FermionField tmp(out.Grid());
+    for (int s=0;s<(int)this->mass.size();s++) {
+      ComplexD a = 4.0+this->mass[s];
+      ComplexD b(0.0,this->mu[s]);
+      axpbg5y_ssp(tmp,a,in,b,in,s,s);
+    }
+    return axpy_norm(out, 1.0, tmp, out);
+  }
+  
+  // needed for fast PV
+  void update(const std::vector<RealD>& _mass, const std::vector<RealD>& _mu) {
+    assert(_mass.size() == _mu.size());
+    assert(_mass.size() == this->FermionGrid()->_fdimensions[0]);
+    this->mass = _mass;
+    this->mu = _mu;
+  }
+  
+ private:
+  std::vector<RealD> mu;
+  std::vector<RealD> mass;
+  
+};
+   
+typedef WilsonTMFermion5D<WilsonImplF> WilsonTMFermion5DF; 
+typedef WilsonTMFermion5D<WilsonImplD> WilsonTMFermion5DD; 
 
-}}
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/ZMobiusFermion.h b/Grid/qcd/action/fermion/ZMobiusFermion.h
index 32ff7670..fc8a7439 100644
--- a/Grid/qcd/action/fermion/ZMobiusFermion.h
+++ b/Grid/qcd/action/fermion/ZMobiusFermion.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,56 +24,50 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef  GRID_QCD_ZMOBIUS_FERMION_H
-#define  GRID_QCD_ZMOBIUS_FERMION_H
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
 
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  namespace QCD {
+template<class Impl>
+class ZMobiusFermion : public CayleyFermion5D<Impl>
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
+public:
 
-    template<class Impl>
-    class ZMobiusFermion : public CayleyFermion5D<Impl>
-    {
-    public:
-     INHERIT_IMPL_TYPES(Impl);
-    public:
-
-      virtual void   Instantiatable(void) {};
-      // Constructors
-      ZMobiusFermion(GaugeField &_Umu,
-		     GridCartesian         &FiveDimGrid,
-		     GridRedBlackCartesian &FiveDimRedBlackGrid,
-		     GridCartesian         &FourDimGrid,
-		     GridRedBlackCartesian &FourDimRedBlackGrid,
-		     RealD _mass,RealD _M5,
-		     std::vector<ComplexD> &gamma, RealD b,RealD c,const ImplParams &p= ImplParams()) : 
+  virtual void   Instantiatable(void) {};
+  // Constructors
+  ZMobiusFermion(GaugeField &_Umu,
+		 GridCartesian         &FiveDimGrid,
+		 GridRedBlackCartesian &FiveDimRedBlackGrid,
+		 GridCartesian         &FourDimGrid,
+		 GridRedBlackCartesian &FourDimRedBlackGrid,
+		 RealD _mass,RealD _M5,
+		 std::vector<ComplexD> &gamma, RealD b,RealD c,const ImplParams &p= ImplParams()) : 
       
-      CayleyFermion5D<Impl>(_Umu,
-			    FiveDimGrid,
-			    FiveDimRedBlackGrid,
-			    FourDimGrid,
-			    FourDimRedBlackGrid,_mass,_M5,p)
+    CayleyFermion5D<Impl>(_Umu,
+			  FiveDimGrid,
+			  FiveDimRedBlackGrid,
+			  FourDimGrid,
+			  FourDimRedBlackGrid,_mass,_M5,p)
 
-      {
-	RealD eps = 1.0;
-	
-	std::cout<<GridLogMessage << "ZMobiusFermion (b="<<b<<",c="<<c<<") with Ls= "<<this->Ls<<" gamma passed in"<<std::endl;
-	std::vector<Coeff_t> zgamma(this->Ls);
-	for(int s=0;s<this->Ls;s++){
-	  zgamma[s] = gamma[s];
-	}
-
-	// Call base setter
-	this->SetCoefficientsInternal(1.0,zgamma,b,c);
-      }
-
-    };
+  {
+    //    RealD eps = 1.0;
+    std::cout<<GridLogMessage << "ZMobiusFermion (b="<<b<<",c="<<c<<") with Ls= "<<this->Ls<<" gamma passed in"<<std::endl;
+    Vector<Coeff_t> zgamma(this->Ls);
+    for(int s=0;s<this->Ls;s++){
+      zgamma[s] = gamma[s];
+    }
 
+    // Call base setter
+    this->SetCoefficientsInternal(1.0,zgamma,b,c);
   }
-}
 
-#endif
+};
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/CayleyFermion5Ddense.cc b/Grid/qcd/action/fermion/deprecated/CayleyFermion5Ddense.h
similarity index 91%
rename from Grid/qcd/action/fermion/CayleyFermion5Ddense.cc
rename to Grid/qcd/action/fermion/deprecated/CayleyFermion5Ddense.h
index 4014675a..f1acb50c 100644
--- a/Grid/qcd/action/fermion/CayleyFermion5Ddense.cc
+++ b/Grid/qcd/action/fermion/deprecated/CayleyFermion5Ddense.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -26,19 +26,19 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 
 #include <Grid/Grid_Eigen_Dense.h>
 #include <Grid/qcd/action/fermion/FermionCore.h>
 #include <Grid/qcd/action/fermion/CayleyFermion5D.h>
 
 
-namespace Grid {
-namespace QCD {
-  /*
-   * Dense matrix versions of routines
-   */
+NAMESPACE_BEGIN(Grid);
+
+/*
+ * Dense matrix versions of routines
+ */
 template<class Impl>
 void CayleyFermion5D<Impl>::MooeeInvDag (const FermionField &psi, FermionField &chi)
 {
@@ -54,10 +54,10 @@ template<class Impl>
 void CayleyFermion5D<Impl>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv)
 {
   int Ls=this->Ls;
-  int LLs = psi._grid->_rdimensions[0];
-  int vol = psi._grid->oSites()/LLs;
+  int LLs = psi.Grid()->_rdimensions[0];
+  int vol = psi.Grid()->oSites()/LLs;
   
-  chi.checkerboard=psi.checkerboard;
+  chi.Checkerboard()=psi.Checkerboard();
   
   assert(Ls==LLs);
   
@@ -96,15 +96,14 @@ void CayleyFermion5D<Impl>::MooeeInternal(const FermionField &psi, FermionField
   }
 
   // For the non-vectorised s-direction this is simple
-  
-  for(auto site=0;site<vol;site++){
+  thread_loop( (auto site=0;site<vol;site++), {
     
     SiteSpinor     SiteChi;
     SiteHalfSpinor SitePplus;
     SiteHalfSpinor SitePminus;
     
     for(int s1=0;s1<Ls;s1++){
-      SiteChi =zero;
+      SiteChi =Zero();
       for(int s2=0;s2<Ls;s2++){
 	int lex2 = s2+Ls*site;
 	
@@ -120,7 +119,7 @@ void CayleyFermion5D<Impl>::MooeeInternal(const FermionField &psi, FermionField
       }
       chi[s1+Ls*site] = SiteChi*0.5;
     }
-  }
+  });
 }
 
 #ifdef CAYLEY_DPERP_DENSE
@@ -153,4 +152,4 @@ template void CayleyFermion5D<ZWilsonImplFH>::MooeeInternal(const FermionField &
 template void CayleyFermion5D<ZWilsonImplDF>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv);
 #endif
 
-}}
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/CayleyFermion5Dssp.cc b/Grid/qcd/action/fermion/deprecated/CayleyFermion5Dssp.h
similarity index 79%
rename from Grid/qcd/action/fermion/CayleyFermion5Dssp.cc
rename to Grid/qcd/action/fermion/deprecated/CayleyFermion5Dssp.h
index cb9b2957..650c391c 100644
--- a/Grid/qcd/action/fermion/CayleyFermion5Dssp.cc
+++ b/Grid/qcd/action/fermion/deprecated/CayleyFermion5Dssp.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -26,26 +26,24 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 
 #include <Grid/qcd/action/fermion/FermionCore.h>
 #include <Grid/qcd/action/fermion/CayleyFermion5D.h>
 
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
-  // FIXME -- make a version of these routines with site loop outermost for cache reuse.
-  // Pminus fowards
-  // Pplus  backwards
+// Pminus fowards
+// Pplus  backwards
 template<class Impl>  
 void CayleyFermion5D<Impl>::M5D(const FermionField &psi,
 				const FermionField &phi, 
 				FermionField &chi,
-				std::vector<Coeff_t> &lower,
-				std::vector<Coeff_t> &diag,
-				std::vector<Coeff_t> &upper)
+				Vector<Coeff_t> &lower,
+				Vector<Coeff_t> &diag,
+				Vector<Coeff_t> &upper)
 {
   Coeff_t one(1.0);
   int Ls=this->Ls;
@@ -66,9 +64,9 @@ template<class Impl>
 void CayleyFermion5D<Impl>::M5Ddag(const FermionField &psi,
 				   const FermionField &phi, 
 				   FermionField &chi,
-				   std::vector<Coeff_t> &lower,
-				   std::vector<Coeff_t> &diag,
-				   std::vector<Coeff_t> &upper)
+				   Vector<Coeff_t> &lower,
+				   Vector<Coeff_t> &diag,
+				   Vector<Coeff_t> &upper)
 {
   Coeff_t one(1.0);
   int Ls=this->Ls;
@@ -91,7 +89,7 @@ void CayleyFermion5D<Impl>::MooeeInv    (const FermionField &psi, FermionField &
 {
   Coeff_t one(1.0);
   Coeff_t czero(0.0);
-  chi.checkerboard=psi.checkerboard;
+  chi.Checkerboard()=psi.Checkerboard();
   int Ls=this->Ls;
   // Apply (L^{\prime})^{-1}
   axpby_ssp (chi,one,psi,     czero,psi,0,0);      // chi[0]=psi[0]
@@ -120,7 +118,7 @@ void CayleyFermion5D<Impl>::MooeeInvDag (const FermionField &psi, FermionField &
 {
   Coeff_t one(1.0);
   Coeff_t czero(0.0);
-  chi.checkerboard=psi.checkerboard;
+  chi.Checkerboard()=psi.Checkerboard();
   int Ls=this->Ls;
   // Apply (U^{\prime})^{-dagger}
   axpby_ssp (chi,one,psi,     czero,psi,0,0);      // chi[0]=psi[0]
@@ -145,20 +143,19 @@ void CayleyFermion5D<Impl>::MooeeInvDag (const FermionField &psi, FermionField &
 
 
 #ifdef CAYLEY_DPERP_LINALG
-  INSTANTIATE_DPERP(WilsonImplF);
-  INSTANTIATE_DPERP(WilsonImplD);
-  INSTANTIATE_DPERP(GparityWilsonImplF);
-  INSTANTIATE_DPERP(GparityWilsonImplD);
-  INSTANTIATE_DPERP(ZWilsonImplF);
-  INSTANTIATE_DPERP(ZWilsonImplD);
+INSTANTIATE_DPERP(WilsonImplF);
+INSTANTIATE_DPERP(WilsonImplD);
+INSTANTIATE_DPERP(GparityWilsonImplF);
+INSTANTIATE_DPERP(GparityWilsonImplD);
+INSTANTIATE_DPERP(ZWilsonImplF);
+INSTANTIATE_DPERP(ZWilsonImplD);
 
-  INSTANTIATE_DPERP(WilsonImplFH);
-  INSTANTIATE_DPERP(WilsonImplDF);
-  INSTANTIATE_DPERP(GparityWilsonImplFH);
-  INSTANTIATE_DPERP(GparityWilsonImplDF);
-  INSTANTIATE_DPERP(ZWilsonImplFH);
-  INSTANTIATE_DPERP(ZWilsonImplDF);
+INSTANTIATE_DPERP(WilsonImplFH);
+INSTANTIATE_DPERP(WilsonImplDF);
+INSTANTIATE_DPERP(GparityWilsonImplFH);
+INSTANTIATE_DPERP(GparityWilsonImplDF);
+INSTANTIATE_DPERP(ZWilsonImplFH);
+INSTANTIATE_DPERP(ZWilsonImplDF);
 #endif
 
-}
-}
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/deprecated/DomainWallEOFAFermiondense.h b/Grid/qcd/action/fermion/deprecated/DomainWallEOFAFermiondense.h
new file mode 100644
index 00000000..3ec4fbac
--- /dev/null
+++ b/Grid/qcd/action/fermion/deprecated/DomainWallEOFAFermiondense.h
@@ -0,0 +1,158 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermiondense.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+
+NAMESPACE_BEGIN(Grid);
+
+/*
+ * Dense matrix versions of routines
+ */
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::MooeeInvDag(const FermionField& psi, FermionField& chi)
+{
+  this->MooeeInternal(psi, chi, DaggerYes, InverseYes);
+}
+
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::MooeeInv(const FermionField& psi, FermionField& chi)
+{
+  this->MooeeInternal(psi, chi, DaggerNo, InverseYes);
+}
+
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv)
+{
+  int Ls = this->Ls;
+  int LLs = psi.Grid()->_rdimensions[0];
+  int vol = psi.Grid()->oSites()/LLs;
+
+  chi.Checkerboard() = psi.Checkerboard();
+
+  assert(Ls==LLs);
+
+  Eigen::MatrixXd Pplus  = Eigen::MatrixXd::Zero(Ls,Ls);
+  Eigen::MatrixXd Pminus = Eigen::MatrixXd::Zero(Ls,Ls);
+
+  for(int s=0;s<Ls;s++){
+    Pplus(s,s)  = this->bee[s];
+    Pminus(s,s) = this->bee[s];
+  }
+
+  for(int s=0; s<Ls-1; s++){
+    Pminus(s,s+1) = -this->cee[s];
+  }
+
+  for(int s=0; s<Ls-1; s++){
+    Pplus(s+1,s) = -this->cee[s+1];
+  }
+
+  Pplus (0,Ls-1) = this->dp;
+  Pminus(Ls-1,0) = this->dm;
+
+  Eigen::MatrixXd PplusMat ;
+  Eigen::MatrixXd PminusMat;
+
+  if(inv) {
+    PplusMat  = Pplus.inverse();
+    PminusMat = Pminus.inverse();
+  } else {
+    PplusMat  = Pplus;
+    PminusMat = Pminus;
+  }
+
+  if(dag){
+    PplusMat.adjointInPlace();
+    PminusMat.adjointInPlace();
+  }
+
+  // For the non-vectorised s-direction this is simple
+
+  for(auto site=0; site<vol; site++){
+
+    SiteSpinor     SiteChi;
+    SiteHalfSpinor SitePplus;
+    SiteHalfSpinor SitePminus;
+
+    for(int s1=0; s1<Ls; s1++){
+      SiteChi = Zero();
+      for(int s2=0; s2<Ls; s2++){
+	int lex2 = s2 + Ls*site;
+	if(PplusMat(s1,s2) != 0.0){
+	  spProj5p(SitePplus,psi[lex2]);
+	  accumRecon5p(SiteChi, PplusMat(s1,s2)*SitePplus);
+	}
+	if(PminusMat(s1,s2) != 0.0){
+	  spProj5m(SitePminus, psi[lex2]);
+	  accumRecon5m(SiteChi, PminusMat(s1,s2)*SitePminus);
+	}
+      }
+      chi[s1+Ls*site] = SiteChi*0.5;
+    }
+  }
+}
+
+#ifdef DOMAIN_WALL_EOFA_DPERP_DENSE
+
+INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplF);
+INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplD);
+INSTANTIATE_DPERP_DWF_EOFA(WilsonImplF);
+INSTANTIATE_DPERP_DWF_EOFA(WilsonImplD);
+INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplF);
+INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplD);
+
+template void DomainWallEOFAFermion<GparityWilsonImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void DomainWallEOFAFermion<GparityWilsonImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void DomainWallEOFAFermion<WilsonImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void DomainWallEOFAFermion<WilsonImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void DomainWallEOFAFermion<ZWilsonImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void DomainWallEOFAFermion<ZWilsonImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+
+INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplFH);
+INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplDF);
+INSTANTIATE_DPERP_DWF_EOFA(WilsonImplFH);
+INSTANTIATE_DPERP_DWF_EOFA(WilsonImplDF);
+INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplFH);
+INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplDF);
+
+template void DomainWallEOFAFermion<GparityWilsonImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void DomainWallEOFAFermion<GparityWilsonImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void DomainWallEOFAFermion<WilsonImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void DomainWallEOFAFermion<WilsonImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void DomainWallEOFAFermion<ZWilsonImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void DomainWallEOFAFermion<ZWilsonImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+
+#endif
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/deprecated/DomainWallEOFAFermionssp.h b/Grid/qcd/action/fermion/deprecated/DomainWallEOFAFermionssp.h
new file mode 100644
index 00000000..c9e638e5
--- /dev/null
+++ b/Grid/qcd/action/fermion/deprecated/DomainWallEOFAFermionssp.h
@@ -0,0 +1,167 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermionssp.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+
+NAMESPACE_BEGIN(Grid);
+
+// FIXME -- make a version of these routines with site loop outermost for cache reuse.
+// Pminus fowards
+// Pplus  backwards
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::M5D(const FermionField& psi, const FermionField& phi,
+				      FermionField& chi, Vector<Coeff_t>& lower, Vector<Coeff_t>& diag, Vector<Coeff_t>& upper)
+{
+  Coeff_t one(1.0);
+  int Ls = this->Ls;
+  for(int s=0; s<Ls; s++){
+    if(s==0) {
+      axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, s+1);
+      axpby_ssp_pplus (chi, one, chi, lower[s], psi, s, Ls-1);
+    } else if (s==(Ls-1)) {
+      axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, 0);
+      axpby_ssp_pplus (chi, one, chi, lower[s], psi, s, s-1);
+    } else {
+      axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, s+1);
+      axpby_ssp_pplus(chi, one, chi, lower[s], psi, s, s-1);
+    }
+  }
+}
+
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::M5Ddag(const FermionField& psi, const FermionField& phi,
+					 FermionField& chi, Vector<Coeff_t>& lower, Vector<Coeff_t>& diag, Vector<Coeff_t>& upper)
+{
+  Coeff_t one(1.0);
+  int Ls = this->Ls;
+  for(int s=0; s<Ls; s++){
+    if(s==0) {
+      axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, s+1);
+      axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, Ls-1);
+    } else if (s==(Ls-1)) {
+      axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, 0);
+      axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, s-1);
+    } else {
+      axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, s+1);
+      axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, s-1);
+    }
+  }
+}
+
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::MooeeInv(const FermionField& psi, FermionField& chi)
+{
+  Coeff_t one(1.0);
+  Coeff_t czero(0.0);
+  chi.Checkerboard() = psi.Checkerboard();
+  int Ls = this->Ls;
+
+  FermionField tmp(psi.Grid());
+
+  // Apply (L^{\prime})^{-1}
+  axpby_ssp(chi, one, psi, czero, psi, 0, 0);      // chi[0]=psi[0]
+  for(int s=1; s<Ls; s++){
+    axpby_ssp_pplus(chi, one, psi, -this->lee[s-1], chi, s, s-1);// recursion Psi[s] -lee P_+ chi[s-1]
+  }
+
+  // L_m^{-1}
+  for(int s=0; s<Ls-1; s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
+    axpby_ssp_pminus(chi, one, chi, -this->leem[s], chi, Ls-1, s);
+  }
+
+  // U_m^{-1} D^{-1}
+  for(int s=0; s<Ls-1; s++){
+    axpby_ssp_pplus(chi, one/this->dee[s], chi, -this->ueem[s]/this->dee[Ls], chi, s, Ls-1);
+  }
+  axpby_ssp_pminus(tmp, czero, chi, one/this->dee[Ls-1], chi, Ls-1, Ls-1);
+  axpby_ssp_pplus(chi, one, tmp, one/this->dee[Ls], chi, Ls-1, Ls-1);
+
+  // Apply U^{-1}
+  for(int s=Ls-2; s>=0; s--){
+    axpby_ssp_pminus(chi, one, chi, -this->uee[s], chi, s, s+1);  // chi[Ls]
+  }
+}
+
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::MooeeInvDag(const FermionField& psi, FermionField& chi)
+{
+  Coeff_t one(1.0);
+  Coeff_t czero(0.0);
+  chi.Checkerboard() = psi.Checkerboard();
+  int Ls = this->Ls;
+
+  FermionField tmp(psi.Grid());
+
+  // Apply (U^{\prime})^{-dagger}
+  axpby_ssp(chi, one, psi, czero, psi, 0, 0);      // chi[0]=psi[0]
+  for(int s=1; s<Ls; s++){
+    axpby_ssp_pminus(chi, one, psi, -conjugate(this->uee[s-1]), chi, s, s-1);
+  }
+
+  // U_m^{-\dagger}
+  for(int s=0; s<Ls-1; s++){
+    axpby_ssp_pplus(chi, one, chi, -conjugate(this->ueem[s]), chi, Ls-1, s);
+  }
+
+  // L_m^{-\dagger} D^{-dagger}
+  for(int s=0; s<Ls-1; s++){
+    axpby_ssp_pminus(chi, one/conjugate(this->dee[s]), chi, -conjugate(this->leem[s]/this->dee[Ls-1]), chi, s, Ls-1);
+  }
+  axpby_ssp_pminus(tmp, czero, chi, one/conjugate(this->dee[Ls-1]), chi, Ls-1, Ls-1);
+  axpby_ssp_pplus(chi, one, tmp, one/conjugate(this->dee[Ls]), chi, Ls-1, Ls-1);
+
+  // Apply L^{-dagger}
+  for(int s=Ls-2; s>=0; s--){
+    axpby_ssp_pplus(chi, one, chi, -conjugate(this->lee[s]), chi, s, s+1);  // chi[Ls]
+  }
+}
+
+#ifdef DOMAIN_WALL_EOFA_DPERP_LINALG
+
+INSTANTIATE_DPERP_DWF_EOFA(WilsonImplF);
+INSTANTIATE_DPERP_DWF_EOFA(WilsonImplD);
+INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplF);
+INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplD);
+INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplF);
+INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplD);
+
+INSTANTIATE_DPERP_DWF_EOFA(WilsonImplFH);
+INSTANTIATE_DPERP_DWF_EOFA(WilsonImplDF);
+INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplFH);
+INSTANTIATE_DPERP_DWF_EOFA(GparityWilsonImplDF);
+INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplFH);
+INSTANTIATE_DPERP_DWF_EOFA(ZWilsonImplDF);
+
+#endif
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/deprecated/MobiusEOFAFermiondense.h b/Grid/qcd/action/fermion/deprecated/MobiusEOFAFermiondense.h
new file mode 100644
index 00000000..8091f344
--- /dev/null
+++ b/Grid/qcd/action/fermion/deprecated/MobiusEOFAFermiondense.h
@@ -0,0 +1,183 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermiondense.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+
+NAMESPACE_BEGIN(Grid);
+
+/*
+ * Dense matrix versions of routines
+ */
+template<class Impl>
+void MobiusEOFAFermion<Impl>::MooeeInv(const FermionField& psi, FermionField& chi)
+{
+  this->MooeeInternal(psi, chi, DaggerNo, InverseYes);
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::MooeeInv_shift(const FermionField& psi, FermionField& chi)
+{
+  this->MooeeInternal(psi, chi, DaggerNo, InverseYes);
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::MooeeInvDag(const FermionField& psi, FermionField& chi)
+{
+  this->MooeeInternal(psi, chi, DaggerYes, InverseYes);
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::MooeeInvDag_shift(const FermionField& psi, FermionField& chi)
+{
+  this->MooeeInternal(psi, chi, DaggerYes, InverseYes);
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv)
+{
+  int Ls = this->Ls;
+  int LLs = psi.Grid()->_rdimensions[0];
+  int vol = psi.Grid()->oSites()/LLs;
+
+  int pm      = this->pm;
+  RealD shift = this->shift;
+  RealD alpha = this->alpha;
+  RealD k     = this->k;
+  RealD mq1   = this->mq1;
+
+  chi.Checkerboard() = psi.Checkerboard();
+
+  assert(Ls==LLs);
+
+  Eigen::MatrixXd Pplus  = Eigen::MatrixXd::Zero(Ls,Ls);
+  Eigen::MatrixXd Pminus = Eigen::MatrixXd::Zero(Ls,Ls);
+
+  for(int s=0;s<Ls;s++){
+    Pplus(s,s)  = this->bee[s];
+    Pminus(s,s) = this->bee[s];
+  }
+
+  for(int s=0; s<Ls-1; s++){
+    Pminus(s,s+1) = -this->cee[s];
+  }
+
+  for(int s=0; s<Ls-1; s++){
+    Pplus(s+1,s) = -this->cee[s+1];
+  }
+  Pplus (0,Ls-1) = mq1*this->cee[0];
+  Pminus(Ls-1,0) = mq1*this->cee[Ls-1];
+
+  if(shift != 0.0){
+    Coeff_t N = 2.0 * ( std::pow(alpha+1.0,Ls) + mq1*std::pow(alpha-1.0,Ls) );
+    for(int s=0; s<Ls; ++s){
+      if(pm == 1){ Pplus(s,Ls-1) += shift * k * N * std::pow(-1.0,s) * std::pow(alpha-1.0,s) / std::pow(alpha+1.0,Ls+s+1); }
+      else{ Pminus(Ls-1-s,Ls-1) -= shift * k * N * std::pow(-1.0,s) * std::pow(alpha-1.0,s) / std::pow(alpha+1.0,Ls+s+1); }
+    }
+  }
+
+  Eigen::MatrixXd PplusMat ;
+  Eigen::MatrixXd PminusMat;
+
+  if(inv){
+    PplusMat  = Pplus.inverse();
+    PminusMat = Pminus.inverse();
+  } else {
+    PplusMat  = Pplus;
+    PminusMat = Pminus;
+  }
+
+  if(dag){
+    PplusMat.adjointInPlace();
+    PminusMat.adjointInPlace();
+  }
+
+  // For the non-vectorised s-direction this is simple
+
+  for(auto site=0; site<vol; site++){
+
+    SiteSpinor     SiteChi;
+    SiteHalfSpinor SitePplus;
+    SiteHalfSpinor SitePminus;
+
+    for(int s1=0; s1<Ls; s1++){
+      SiteChi = Zero();
+      for(int s2=0; s2<Ls; s2++){
+	int lex2 = s2 + Ls*site;
+	if(PplusMat(s1,s2) != 0.0){
+	  spProj5p(SitePplus,psi[lex2]);
+	  accumRecon5p(SiteChi, PplusMat(s1,s2)*SitePplus);
+	}
+	if(PminusMat(s1,s2) != 0.0){
+	  spProj5m(SitePminus, psi[lex2]);
+	  accumRecon5m(SiteChi, PminusMat(s1,s2)*SitePminus);
+	}
+      }
+      chi[s1+Ls*site] = SiteChi*0.5;
+    }
+  }
+}
+
+#ifdef MOBIUS_EOFA_DPERP_DENSE
+
+INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplF);
+INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplD);
+INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplF);
+INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplD);
+INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplF);
+INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplD);
+
+template void MobiusEOFAFermion<GparityWilsonImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void MobiusEOFAFermion<GparityWilsonImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void MobiusEOFAFermion<WilsonImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void MobiusEOFAFermion<WilsonImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void MobiusEOFAFermion<ZWilsonImplF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void MobiusEOFAFermion<ZWilsonImplD>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+
+INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplFH);
+INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplDF);
+INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplFH);
+INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplDF);
+INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplFH);
+INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplDF);
+
+template void MobiusEOFAFermion<GparityWilsonImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void MobiusEOFAFermion<GparityWilsonImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void MobiusEOFAFermion<WilsonImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void MobiusEOFAFermion<WilsonImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void MobiusEOFAFermion<ZWilsonImplFH>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+template void MobiusEOFAFermion<ZWilsonImplDF>::MooeeInternal(const FermionField& psi, FermionField& chi, int dag, int inv);
+
+#endif
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/deprecated/MobiusEOFAFermionssp.h b/Grid/qcd/action/fermion/deprecated/MobiusEOFAFermionssp.h
new file mode 100644
index 00000000..254cdb54
--- /dev/null
+++ b/Grid/qcd/action/fermion/deprecated/MobiusEOFAFermionssp.h
@@ -0,0 +1,289 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermionssp.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+
+NAMESPACE_BEGIN(Grid);
+
+// FIXME -- make a version of these routines with site loop outermost for cache reuse.
+// Pminus fowards
+// Pplus  backwards
+template<class Impl>
+void MobiusEOFAFermion<Impl>::M5D(const FermionField& psi, const FermionField& phi,
+				  FermionField& chi, Vector<Coeff_t>& lower, Vector<Coeff_t>& diag, Vector<Coeff_t>& upper)
+{
+  Coeff_t one(1.0);
+  int Ls = this->Ls;
+  for(int s=0; s<Ls; s++){
+    if(s==0) {
+      axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, s+1);
+      axpby_ssp_pplus (chi, one, chi, lower[s], psi, s, Ls-1);
+    } else if (s==(Ls-1)) {
+      axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, 0);
+      axpby_ssp_pplus (chi, one, chi, lower[s], psi, s, s-1);
+    } else {
+      axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, s+1);
+      axpby_ssp_pplus(chi, one, chi, lower[s], psi, s, s-1);
+    }
+  }
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::M5D_shift(const FermionField& psi, const FermionField& phi,
+					FermionField& chi, Vector<Coeff_t>& lower, Vector<Coeff_t>& diag, Vector<Coeff_t>& upper,
+					Vector<Coeff_t>& shift_coeffs)
+{
+  Coeff_t one(1.0);
+  int Ls = this->Ls;
+  for(int s=0; s<Ls; s++){
+    if(s==0) {
+      axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, s+1);
+      axpby_ssp_pplus (chi, one, chi, lower[s], psi, s, Ls-1);
+    } else if (s==(Ls-1)) {
+      axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, 0);
+      axpby_ssp_pplus (chi, one, chi, lower[s], psi, s, s-1);
+    } else {
+      axpby_ssp_pminus(chi, diag[s], phi, upper[s], psi, s, s+1);
+      axpby_ssp_pplus(chi, one, chi, lower[s], psi, s, s-1);
+    }
+    if(this->pm == 1){ axpby_ssp_pplus(chi, one, chi, shift_coeffs[s], psi, s, Ls-1); }
+    else{ axpby_ssp_pminus(chi, one, chi, shift_coeffs[s], psi, s, 0); }
+  }
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::M5Ddag(const FermionField& psi, const FermionField& phi,
+				     FermionField& chi, Vector<Coeff_t>& lower, Vector<Coeff_t>& diag, Vector<Coeff_t>& upper)
+{
+  Coeff_t one(1.0);
+  int Ls = this->Ls;
+  for(int s=0; s<Ls; s++){
+    if(s==0) {
+      axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, s+1);
+      axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, Ls-1);
+    } else if (s==(Ls-1)) {
+      axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, 0);
+      axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, s-1);
+    } else {
+      axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, s+1);
+      axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, s-1);
+    }
+  }
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::M5Ddag_shift(const FermionField& psi, const FermionField& phi,
+					   FermionField& chi, Vector<Coeff_t>& lower, Vector<Coeff_t>& diag, Vector<Coeff_t>& upper,
+					   Vector<Coeff_t>& shift_coeffs)
+{
+  Coeff_t one(1.0);
+  int Ls = this->Ls;
+  for(int s=0; s<Ls; s++){
+    if(s==0) {
+      axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, s+1);
+      axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, Ls-1);
+    } else if (s==(Ls-1)) {
+      axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, 0);
+      axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, s-1);
+    } else {
+      axpby_ssp_pplus (chi, diag[s], phi, upper[s], psi, s, s+1);
+      axpby_ssp_pminus(chi, one, chi, lower[s], psi, s, s-1);
+    }
+    if(this->pm == 1){ axpby_ssp_pplus(chi, one, chi, shift_coeffs[s], psi, Ls-1, s); }
+    else{ axpby_ssp_pminus(chi, one, chi, shift_coeffs[s], psi, 0, s); }
+  }
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::MooeeInv(const FermionField& psi, FermionField& chi)
+{
+  if(this->shift != 0.0){ MooeeInv_shift(psi,chi); return; }
+
+  Coeff_t one(1.0);
+  Coeff_t czero(0.0);
+  chi.Checkerboard() = psi.Checkerboard();
+  int Ls = this->Ls;
+
+  // Apply (L^{\prime})^{-1}
+  axpby_ssp(chi, one, psi, czero, psi, 0, 0);      // chi[0]=psi[0]
+  for(int s=1; s<Ls; s++){
+    axpby_ssp_pplus(chi, one, psi, -this->lee[s-1], chi, s, s-1);// recursion Psi[s] -lee P_+ chi[s-1]
+  }
+
+  // L_m^{-1}
+  for(int s=0; s<Ls-1; s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
+    axpby_ssp_pminus(chi, one, chi, -this->leem[s], chi, Ls-1, s);
+  }
+
+  // U_m^{-1} D^{-1}
+  for(int s=0; s<Ls-1; s++){
+    axpby_ssp_pplus(chi, one/this->dee[s], chi, -this->ueem[s]/this->dee[Ls-1], chi, s, Ls-1);
+  }
+  axpby_ssp(chi, one/this->dee[Ls-1], chi, czero, chi, Ls-1, Ls-1);
+
+  // Apply U^{-1}
+  for(int s=Ls-2; s>=0; s--){
+    axpby_ssp_pminus(chi, one, chi, -this->uee[s], chi, s, s+1);  // chi[Ls]
+  }
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::MooeeInv_shift(const FermionField& psi, FermionField& chi)
+{
+  Coeff_t one(1.0);
+  Coeff_t czero(0.0);
+  chi.Checkerboard() = psi.Checkerboard();
+  int Ls = this->Ls;
+
+  FermionField tmp(psi.Grid());
+
+  // Apply (L^{\prime})^{-1}
+  axpby_ssp(chi, one, psi, czero, psi, 0, 0);      // chi[0]=psi[0]
+  axpby_ssp(tmp, czero, tmp, this->MooeeInv_shift_lc[0], psi, 0, 0);
+  for(int s=1; s<Ls; s++){
+    axpby_ssp_pplus(chi, one, psi, -this->lee[s-1], chi, s, s-1);// recursion Psi[s] -lee P_+ chi[s-1]
+    axpby_ssp(tmp, one, tmp, this->MooeeInv_shift_lc[s], psi, 0, s);
+  }
+
+  // L_m^{-1}
+  for(int s=0; s<Ls-1; s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
+    axpby_ssp_pminus(chi, one, chi, -this->leem[s], chi, Ls-1, s);
+  }
+
+  // U_m^{-1} D^{-1}
+  for(int s=0; s<Ls-1; s++){
+    axpby_ssp_pplus(chi, one/this->dee[s], chi, -this->ueem[s]/this->dee[Ls-1], chi, s, Ls-1);
+  }
+  axpby_ssp(chi, one/this->dee[Ls-1], chi, czero, chi, Ls-1, Ls-1);
+
+  // Apply U^{-1} and add shift term
+  if(this->pm == 1){ axpby_ssp_pplus(chi, one, chi, this->MooeeInv_shift_norm[Ls-1], tmp, Ls-1, 0); }
+  else{ axpby_ssp_pminus(chi, one, chi, this->MooeeInv_shift_norm[Ls-1], tmp, Ls-1, 0); }
+  for(int s=Ls-2; s>=0; s--){
+    axpby_ssp_pminus(chi, one, chi, -this->uee[s], chi, s, s+1);  // chi[Ls]
+    if(this->pm == 1){ axpby_ssp_pplus(chi, one, chi, this->MooeeInv_shift_norm[s], tmp, s, 0); }
+    else{ axpby_ssp_pminus(chi, one, chi, this->MooeeInv_shift_norm[s], tmp, s, 0); }
+  }
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::MooeeInvDag(const FermionField& psi, FermionField& chi)
+{
+  if(this->shift != 0.0){ MooeeInvDag_shift(psi,chi); return; }
+
+  Coeff_t one(1.0);
+  Coeff_t czero(0.0);
+  chi.Checkerboard() = psi.Checkerboard();
+  int Ls = this->Ls;
+
+  // Apply (U^{\prime})^{-dagger}
+  axpby_ssp(chi, one, psi, czero, psi, 0, 0);      // chi[0]=psi[0]
+  for(int s=1; s<Ls; s++){
+    axpby_ssp_pminus(chi, one, psi, -conjugate(this->uee[s-1]), chi, s, s-1);
+  }
+
+  // U_m^{-\dagger}
+  for(int s=0; s<Ls-1; s++){
+    axpby_ssp_pplus(chi, one, chi, -conjugate(this->ueem[s]), chi, Ls-1, s);
+  }
+
+  // L_m^{-\dagger} D^{-dagger}
+  for(int s=0; s<Ls-1; s++){
+    axpby_ssp_pminus(chi, one/conjugate(this->dee[s]), chi, -conjugate(this->leem[s]/this->dee[Ls-1]), chi, s, Ls-1);
+  }
+  axpby_ssp(chi, one/conjugate(this->dee[Ls-1]), chi, czero, chi, Ls-1, Ls-1);
+
+  // Apply L^{-dagger}
+  for(int s=Ls-2; s>=0; s--){
+    axpby_ssp_pplus(chi, one, chi, -conjugate(this->lee[s]), chi, s, s+1);  // chi[Ls]
+  }
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::MooeeInvDag_shift(const FermionField& psi, FermionField& chi)
+{
+  Coeff_t one(1.0);
+  Coeff_t czero(0.0);
+  chi.Checkerboard() = psi.Checkerboard();
+  int Ls = this->Ls;
+
+  FermionField tmp(psi.Grid());
+
+  // Apply (U^{\prime})^{-dagger} and accumulate (MooeeInvDag_shift_lc)_{j} \psi_{j} in tmp[0]
+  axpby_ssp(chi, one, psi, czero, psi, 0, 0);      // chi[0]=psi[0]
+  axpby_ssp(tmp, czero, tmp, this->MooeeInvDag_shift_lc[0], psi, 0, 0);
+  for(int s=1; s<Ls; s++){
+    axpby_ssp_pminus(chi, one, psi, -conjugate(this->uee[s-1]), chi, s, s-1);
+    axpby_ssp(tmp, one, tmp, this->MooeeInvDag_shift_lc[s], psi, 0, s);
+  }
+
+  // U_m^{-\dagger}
+  for(int s=0; s<Ls-1; s++){
+    axpby_ssp_pplus(chi, one, chi, -conjugate(this->ueem[s]), chi, Ls-1, s);
+  }
+
+  // L_m^{-\dagger} D^{-dagger}
+  for(int s=0; s<Ls-1; s++){
+    axpby_ssp_pminus(chi, one/conjugate(this->dee[s]), chi, -conjugate(this->leem[s]/this->dee[Ls-1]), chi, s, Ls-1);
+  }
+  axpby_ssp(chi, one/conjugate(this->dee[Ls-1]), chi, czero, chi, Ls-1, Ls-1);
+
+  // Apply L^{-dagger} and add shift
+  if(this->pm == 1){ axpby_ssp_pplus(chi, one, chi, this->MooeeInvDag_shift_norm[Ls-1], tmp, Ls-1, 0); }
+  else{ axpby_ssp_pminus(chi, one, chi, this->MooeeInvDag_shift_norm[Ls-1], tmp, Ls-1, 0); }
+  for(int s=Ls-2; s>=0; s--){
+    axpby_ssp_pplus(chi, one, chi, -conjugate(this->lee[s]), chi, s, s+1);  // chi[Ls]
+    if(this->pm == 1){ axpby_ssp_pplus(chi, one, chi, this->MooeeInvDag_shift_norm[s], tmp, s, 0); }
+    else{ axpby_ssp_pminus(chi, one, chi, this->MooeeInvDag_shift_norm[s], tmp, s, 0); }
+  }
+}
+
+#ifdef MOBIUS_EOFA_DPERP_LINALG
+
+INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplF);
+INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplD);
+INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplF);
+INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplD);
+INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplF);
+INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplD);
+
+INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplFH);
+INSTANTIATE_DPERP_MOBIUS_EOFA(WilsonImplDF);
+INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplFH);
+INSTANTIATE_DPERP_MOBIUS_EOFA(GparityWilsonImplDF);
+INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplFH);
+INSTANTIATE_DPERP_MOBIUS_EOFA(ZWilsonImplDF);
+
+#endif
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/g5HermitianLinop.h b/Grid/qcd/action/fermion/g5HermitianLinop.h
index cca7a113..2e417ceb 100644
--- a/Grid/qcd/action/fermion/g5HermitianLinop.h
+++ b/Grid/qcd/action/fermion/g5HermitianLinop.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,13 +23,12 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef G5_HERMITIAN_LINOP
 #define G5_HERMITIAN_LINOP
 
-namespace Grid {
-  namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 ////////////////////////////////////////////////////////////////////
 // Wrap an already herm matrix
@@ -46,12 +45,12 @@ public:
     HermOp(in,out);
   }
   void OpDiag (const Field &in, Field &out) {
-    Field tmp(in._grid);
+    Field tmp(in.Grid());
     _Mat.Mdiag(in,tmp);
     G5R5(out,tmp);
   }
   void OpDir  (const Field &in, Field &out,int dir,int disp) {
-    Field tmp(in._grid);
+    Field tmp(in.Grid());
     _Mat.Mdir(in,tmp,dir,disp);
     G5R5(out,tmp);
   }
@@ -68,7 +67,7 @@ public:
     n2=real(dot);
   }
   void HermOp(const Field &in, Field &out){
-    Field tmp(in._grid);
+    Field tmp(in.Grid());
     _Mat.M(in,tmp);
     G5R5(out,tmp);
   }
@@ -80,7 +79,7 @@ class Gamma5HermitianLinearOperator : public LinearOperatorBase<Field> {
   Matrix &_Mat;
   Gamma g5;
 public:
-    Gamma5HermitianLinearOperator(Matrix &Mat): _Mat(Mat), g5(Gamma::Algebra::Gamma5) {};
+  Gamma5HermitianLinearOperator(Matrix &Mat): _Mat(Mat), g5(Gamma::Algebra::Gamma5) {};
   void Op     (const Field &in, Field &out){
     HermOp(in,out);
   }
@@ -88,12 +87,12 @@ public:
     HermOp(in,out);
   }
   void OpDiag (const Field &in, Field &out) {
-    Field tmp(in._grid);
+    Field tmp(in.Grid());
     _Mat.Mdiag(in,tmp);
     out=g5*tmp;
   }
   void OpDir  (const Field &in, Field &out,int dir,int disp) {
-    Field tmp(in._grid);
+    Field tmp(in.Grid());
     _Mat.Mdir(in,tmp,dir,disp);
     out=g5*tmp;
   }
@@ -110,12 +109,11 @@ public:
     n2=real(dot);
   }
   void HermOp(const Field &in, Field &out){
-    Field tmp(in._grid);
+    Field tmp(in.Grid());
     _Mat.M(in,tmp);
     out=g5*tmp;
   }
 };
 
-
-}}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/qcd/action/fermion/CayleyFermion5D.cc b/Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h
similarity index 80%
rename from Grid/qcd/action/fermion/CayleyFermion5D.cc
rename to Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h
index a95ea4a0..bab21aea 100644
--- a/Grid/qcd/action/fermion/CayleyFermion5D.cc
+++ b/Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -26,31 +26,30 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 
 #include <Grid/Grid_Eigen_Dense.h>
 #include <Grid/qcd/action/fermion/FermionCore.h>
 #include <Grid/qcd/action/fermion/CayleyFermion5D.h>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
- template<class Impl>
- CayleyFermion5D<Impl>::CayleyFermion5D(GaugeField &_Umu,
-					GridCartesian         &FiveDimGrid,
-					GridRedBlackCartesian &FiveDimRedBlackGrid,
-					GridCartesian         &FourDimGrid,
-					GridRedBlackCartesian &FourDimRedBlackGrid,
-					RealD _mass,RealD _M5,const ImplParams &p) :
-   WilsonFermion5D<Impl>(_Umu,
-		   FiveDimGrid,
-		   FiveDimRedBlackGrid,
-		   FourDimGrid,
- 	 	   FourDimRedBlackGrid,_M5,p),
-   mass(_mass)
- { 
- }
+template<class Impl>
+CayleyFermion5D<Impl>::CayleyFermion5D(GaugeField &_Umu,
+				       GridCartesian         &FiveDimGrid,
+				       GridRedBlackCartesian &FiveDimRedBlackGrid,
+				       GridCartesian         &FourDimGrid,
+				       GridRedBlackCartesian &FourDimRedBlackGrid,
+				       RealD _mass,RealD _M5,const ImplParams &p) :
+  WilsonFermion5D<Impl>(_Umu,
+			FiveDimGrid,
+			FiveDimRedBlackGrid,
+			FourDimGrid,
+			FourDimRedBlackGrid,_M5,p),
+  mass(_mass)
+{ 
+}
 
 ///////////////////////////////////////////////////////////////
 // Physical surface field utilities
@@ -61,8 +60,8 @@ void CayleyFermion5D<Impl>::ExportPhysicalFermionSolution(const FermionField &so
   int Ls = this->Ls;
   FermionField tmp(this->FermionGrid());
   tmp = solution5d;
-  conformable(solution5d._grid,this->FermionGrid());
-  conformable(exported4d._grid,this->GaugeGrid());
+  conformable(solution5d.Grid(),this->FermionGrid());
+  conformable(exported4d.Grid(),this->GaugeGrid());
   axpby_ssp_pminus(tmp, 0., solution5d, 1., solution5d, 0, 0);
   axpby_ssp_pplus (tmp, 1., tmp       , 1., solution5d, 0, Ls-1);
   ExtractSlice(exported4d, tmp, 0, 0);
@@ -71,7 +70,7 @@ template<class Impl>
 void CayleyFermion5D<Impl>::P(const FermionField &psi, FermionField &chi)
 {
   int Ls= this->Ls;
-  chi=zero;
+  chi=Zero();
   for(int s=0;s<Ls;s++){
     axpby_ssp_pminus(chi,1.0,chi,1.0,psi,s,s);
     axpby_ssp_pplus (chi,1.0,chi,1.0,psi,s,(s+1)%Ls);
@@ -81,7 +80,7 @@ template<class Impl>
 void CayleyFermion5D<Impl>::Pdag(const FermionField &psi, FermionField &chi)
 {
   int Ls= this->Ls;
-  chi=zero;
+  chi=Zero();
   for(int s=0;s<Ls;s++){
     axpby_ssp_pminus(chi,1.0,chi,1.0,psi,s,s);
     axpby_ssp_pplus (chi,1.0,chi,1.0,psi,s,(s-1+Ls)%Ls);
@@ -93,8 +92,8 @@ void CayleyFermion5D<Impl>::ExportPhysicalFermionSource(const FermionField &solu
   int Ls = this->Ls;
   FermionField tmp(this->FermionGrid());
   tmp = solution5d;
-  conformable(solution5d._grid,this->FermionGrid());
-  conformable(exported4d._grid,this->GaugeGrid());
+  conformable(solution5d.Grid(),this->FermionGrid());
+  conformable(exported4d.Grid(),this->GaugeGrid());
   axpby_ssp_pplus (tmp, 0., solution5d, 1., solution5d, 0, 0);
   axpby_ssp_pminus(tmp, 1., tmp       , 1., solution5d, 0, Ls-1);
   ExtractSlice(exported4d, tmp, 0, 0);
@@ -104,9 +103,9 @@ void CayleyFermion5D<Impl>::ImportUnphysicalFermion(const FermionField &input4d,
 {
   int Ls = this->Ls;
   FermionField tmp(this->FermionGrid());
-  conformable(imported5d._grid,this->FermionGrid());
-  conformable(input4d._grid   ,this->GaugeGrid());
-  tmp = zero;
+  conformable(imported5d.Grid(),this->FermionGrid());
+  conformable(input4d.Grid()   ,this->GaugeGrid());
+  tmp = Zero();
   InsertSlice(input4d, tmp, 0   , 0);
   InsertSlice(input4d, tmp, Ls-1, 0);
   axpby_ssp_pplus (tmp, 0., tmp, 1., tmp, 0, 0);
@@ -119,9 +118,9 @@ void CayleyFermion5D<Impl>::ImportPhysicalFermionSource(const FermionField &inpu
 {
   int Ls = this->Ls;
   FermionField tmp(this->FermionGrid());
-  conformable(imported5d._grid,this->FermionGrid());
-  conformable(input4d._grid   ,this->GaugeGrid());
-  tmp = zero;
+  conformable(imported5d.Grid(),this->FermionGrid());
+  conformable(input4d.Grid()   ,this->GaugeGrid());
+  tmp = Zero();
   InsertSlice(input4d, tmp, 0   , 0);
   InsertSlice(input4d, tmp, Ls-1, 0);
   axpby_ssp_pplus (tmp, 0., tmp, 1., tmp, 0, 0);
@@ -156,7 +155,7 @@ void CayleyFermion5D<Impl>::DminusDag(const FermionField &psi, FermionField &chi
 template<class Impl> void CayleyFermion5D<Impl>::CayleyReport(void)
 {
   this->Report();
-  std::vector<int> latt = GridDefaultLatt();          
+  Coordinate latt = GridDefaultLatt();          
   RealD volume = this->Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt[mu];
   RealD NP     = this->_FourDimGrid->_Nprocessors;
   if ( M5Dcalls > 0 ) {
@@ -164,10 +163,16 @@ template<class Impl> void CayleyFermion5D<Impl>::CayleyReport(void)
     std::cout << GridLogMessage << "CayleyFermion5D Number of M5D Calls     : " << M5Dcalls   << std::endl;
     std::cout << GridLogMessage << "CayleyFermion5D ComputeTime/Calls       : " << M5Dtime / M5Dcalls << " us" << std::endl;
 
-    // Flops = 6.0*(Nc*Ns) *Ls*vol
-    RealD mflops = 6.0*12*volume*M5Dcalls/M5Dtime/2; // 2 for red black counting
+    // Flops = 10.0*(Nc*Ns) *Ls*vol
+    RealD mflops = 10.0*(Nc*Ns)*volume*M5Dcalls/M5Dtime/2; // 2 for red black counting
     std::cout << GridLogMessage << "Average mflops/s per call                : " << mflops << std::endl;
     std::cout << GridLogMessage << "Average mflops/s per call per rank       : " << mflops/NP << std::endl;
+
+    // Bytes = sizeof(Real) * (Nc*Ns*Nreim) * Ls * vol * (read+write) (/2 for red black counting)
+    // read = 2 ( psi[ss+s+1] and psi[ss+s-1] count as 1 )
+    // write = 1
+    RealD Gbytes = sizeof(Real) * (Nc*Ns*2) * volume * 3 /2. * 1.e-9;
+    std::cout << GridLogMessage << "Average bandwidth (GB/s)                 : " << Gbytes/M5Dtime*M5Dcalls*1.e6 << std::endl;
   }
 
   if ( MooeeInvCalls > 0 ) {
@@ -175,11 +180,16 @@ template<class Impl> void CayleyFermion5D<Impl>::CayleyReport(void)
     std::cout << GridLogMessage << "#### MooeeInv calls report " << std::endl;
     std::cout << GridLogMessage << "CayleyFermion5D Number of MooeeInv Calls     : " << MooeeInvCalls   << std::endl;
     std::cout << GridLogMessage << "CayleyFermion5D ComputeTime/Calls            : " << MooeeInvTime / MooeeInvCalls << " us" << std::endl;
-
+#ifdef GRID_NVCC
+    RealD mflops = ( -16.*Nc*Ns+this->Ls*(1.+18.*Nc*Ns) )*volume*MooeeInvCalls/MooeeInvTime/2; // 2 for red black counting
+    std::cout << GridLogMessage << "Average mflops/s per call                : " << mflops << std::endl;
+    std::cout << GridLogMessage << "Average mflops/s per call per rank       : " << mflops/NP << std::endl;
+#else
     // Flops = MADD * Ls *Ls *4dvol * spin/colour/complex
     RealD mflops = 2.0*24*this->Ls*volume*MooeeInvCalls/MooeeInvTime/2; // 2 for red black counting
     std::cout << GridLogMessage << "Average mflops/s per call                : " << mflops << std::endl;
     std::cout << GridLogMessage << "Average mflops/s per call per rank       : " << mflops/NP << std::endl;
+#endif
   }
 
 }
@@ -198,18 +208,18 @@ template<class Impl>
 void CayleyFermion5D<Impl>::M5D   (const FermionField &psi, FermionField &chi)
 {
   int Ls=this->Ls;
-  std::vector<Coeff_t> diag (Ls,1.0);
-  std::vector<Coeff_t> upper(Ls,-1.0); upper[Ls-1]=mass;
-  std::vector<Coeff_t> lower(Ls,-1.0); lower[0]   =mass;
+  Vector<Coeff_t> diag (Ls,1.0);
+  Vector<Coeff_t> upper(Ls,-1.0); upper[Ls-1]=mass;
+  Vector<Coeff_t> lower(Ls,-1.0); lower[0]   =mass;
   M5D(psi,chi,chi,lower,diag,upper);
 }
 template<class Impl>
 void CayleyFermion5D<Impl>::Meooe5D    (const FermionField &psi, FermionField &Din)
 {
   int Ls=this->Ls;
-  std::vector<Coeff_t> diag = bs;
-  std::vector<Coeff_t> upper= cs;
-  std::vector<Coeff_t> lower= cs; 
+  Vector<Coeff_t> diag = bs;
+  Vector<Coeff_t> upper= cs;
+  Vector<Coeff_t> lower= cs; 
   upper[Ls-1]=-mass*upper[Ls-1];
   lower[0]   =-mass*lower[0];
   M5D(psi,psi,Din,lower,diag,upper);
@@ -218,9 +228,9 @@ void CayleyFermion5D<Impl>::Meooe5D    (const FermionField &psi, FermionField &D
 template<class Impl> void CayleyFermion5D<Impl>::Meo5D     (const FermionField &psi, FermionField &chi)
 {
   int Ls=this->Ls;
-  std::vector<Coeff_t> diag = beo;
-  std::vector<Coeff_t> upper(Ls);
-  std::vector<Coeff_t> lower(Ls);
+  Vector<Coeff_t> diag = beo;
+  Vector<Coeff_t> upper(Ls);
+  Vector<Coeff_t> lower(Ls);
   for(int i=0;i<Ls;i++) {
     upper[i]=-ceo[i];
     lower[i]=-ceo[i];
@@ -233,9 +243,9 @@ template<class Impl>
 void CayleyFermion5D<Impl>::Mooee       (const FermionField &psi, FermionField &chi)
 {
   int Ls=this->Ls;
-  std::vector<Coeff_t> diag = bee;
-  std::vector<Coeff_t> upper(Ls);
-  std::vector<Coeff_t> lower(Ls);
+  Vector<Coeff_t> diag = bee;
+  Vector<Coeff_t> upper(Ls);
+  Vector<Coeff_t> lower(Ls);
   for(int i=0;i<Ls;i++) {
     upper[i]=-cee[i];
     lower[i]=-cee[i];
@@ -248,9 +258,9 @@ template<class Impl>
 void CayleyFermion5D<Impl>::MooeeDag    (const FermionField &psi, FermionField &chi)
 {
   int Ls=this->Ls;
-  std::vector<Coeff_t> diag = bee;
-  std::vector<Coeff_t> upper(Ls);
-  std::vector<Coeff_t> lower(Ls);
+  Vector<Coeff_t> diag = bee;
+  Vector<Coeff_t> upper(Ls);
+  Vector<Coeff_t> lower(Ls);
 
   for (int s=0;s<Ls;s++){
     // Assemble the 5d matrix
@@ -278,9 +288,9 @@ template<class Impl>
 void CayleyFermion5D<Impl>::M5Ddag (const FermionField &psi, FermionField &chi)
 {
   int Ls=this->Ls;
-  std::vector<Coeff_t> diag(Ls,1.0);
-  std::vector<Coeff_t> upper(Ls,-1.0);
-  std::vector<Coeff_t> lower(Ls,-1.0);
+  Vector<Coeff_t> diag(Ls,1.0);
+  Vector<Coeff_t> upper(Ls,-1.0);
+  Vector<Coeff_t> lower(Ls,-1.0);
   upper[Ls-1]=-mass*upper[Ls-1];
   lower[0]   =-mass*lower[0];
   M5Ddag(psi,chi,chi,lower,diag,upper);
@@ -290,9 +300,9 @@ template<class Impl>
 void CayleyFermion5D<Impl>::MeooeDag5D    (const FermionField &psi, FermionField &Din)
 {
   int Ls=this->Ls;
-  std::vector<Coeff_t> diag =bs;
-  std::vector<Coeff_t> upper=cs;
-  std::vector<Coeff_t> lower=cs; 
+  Vector<Coeff_t> diag =bs;
+  Vector<Coeff_t> upper=cs;
+  Vector<Coeff_t> lower=cs; 
 
   for (int s=0;s<Ls;s++){
     if ( s== 0 ) {
@@ -315,9 +325,7 @@ void CayleyFermion5D<Impl>::MeooeDag5D    (const FermionField &psi, FermionField
 template<class Impl>
 RealD CayleyFermion5D<Impl>::M    (const FermionField &psi, FermionField &chi)
 {
-  int Ls=this->Ls;
-  
-  FermionField Din(psi._grid);
+  FermionField Din(psi.Grid());
   
   // Assemble Din
   Meooe5D(psi,Din);
@@ -337,7 +345,7 @@ RealD CayleyFermion5D<Impl>::Mdag (const FermionField &psi, FermionField &chi)
   //D1+        D1- P-    ->   D1+^dag   P+ D2-^dag
   //D2- P+     D2+            P-D1-^dag D2+dag
   
-  FermionField Din(psi._grid);
+  FermionField Din(psi.Grid());
   // Apply Dw
   this->DW(psi,Din,DaggerYes); 
   
@@ -353,11 +361,9 @@ RealD CayleyFermion5D<Impl>::Mdag (const FermionField &psi, FermionField &chi)
 template<class Impl>
 void CayleyFermion5D<Impl>::Meooe       (const FermionField &psi, FermionField &chi)
 {
-  int Ls=this->Ls;
-
   Meooe5D(psi,this->tmp()); 
 
-  if ( psi.checkerboard == Odd ) {
+  if ( psi.Checkerboard() == Odd ) {
     this->DhopEO(this->tmp(),chi,DaggerNo);
   } else {
     this->DhopOE(this->tmp(),chi,DaggerNo);
@@ -368,7 +374,7 @@ template<class Impl>
 void CayleyFermion5D<Impl>::MeooeDag    (const FermionField &psi, FermionField &chi)
 {
   // Apply 4d dslash
-  if ( psi.checkerboard == Odd ) {
+  if ( psi.Checkerboard() == Odd ) {
     this->DhopEO(psi,this->tmp(),DaggerYes);
   } else {
     this->DhopOE(psi,this->tmp(),DaggerYes);
@@ -386,7 +392,7 @@ void  CayleyFermion5D<Impl>::Mdir (const FermionField &psi, FermionField &chi,in
 template<class Impl>
 void CayleyFermion5D<Impl>::MDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
 {
-  FermionField Din(V._grid);
+  FermionField Din(V.Grid());
   
   if ( dag == DaggerNo ) {
     //      U d/du [D_w D5] V = U d/du DW D5 V
@@ -401,7 +407,7 @@ void CayleyFermion5D<Impl>::MDeriv  (GaugeField &mat,const FermionField &U,const
 template<class Impl>
 void CayleyFermion5D<Impl>::MoeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
 {
-  FermionField Din(V._grid);
+  FermionField Din(V.Grid());
   
   if ( dag == DaggerNo ) {
     //      U d/du [D_w D5] V = U d/du DW D5 V
@@ -416,7 +422,7 @@ void CayleyFermion5D<Impl>::MoeDeriv(GaugeField &mat,const FermionField &U,const
 template<class Impl>
 void CayleyFermion5D<Impl>::MeoDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
 {
-  FermionField Din(V._grid);
+  FermionField Din(V.Grid());
   
   if ( dag == DaggerNo ) {
     //      U d/du [D_w D5] V = U d/du DW D5 V
@@ -433,7 +439,7 @@ void CayleyFermion5D<Impl>::MeoDeriv(GaugeField &mat,const FermionField &U,const
 template<class Impl>
 void CayleyFermion5D<Impl>::SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD b,RealD c)
 {
-  std::vector<Coeff_t> gamma(this->Ls);
+  Vector<Coeff_t> gamma(this->Ls);
   for(int s=0;s<this->Ls;s++) gamma[s] = zdata->gamma[s];
   SetCoefficientsInternal(1.0,gamma,b,c);
 }
@@ -441,13 +447,13 @@ void CayleyFermion5D<Impl>::SetCoefficientsTanh(Approx::zolotarev_data *zdata,Re
 template<class Impl>
 void CayleyFermion5D<Impl>::SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata,RealD b,RealD c)
 {
-  std::vector<Coeff_t> gamma(this->Ls);
+  Vector<Coeff_t> gamma(this->Ls);
   for(int s=0;s<this->Ls;s++) gamma[s] = zdata->gamma[s];
   SetCoefficientsInternal(zolo_hi,gamma,b,c);
 }
 //Zolo
 template<class Impl>
-void CayleyFermion5D<Impl>::SetCoefficientsInternal(RealD zolo_hi,std::vector<Coeff_t> & gamma,RealD b,RealD c)
+void CayleyFermion5D<Impl>::SetCoefficientsInternal(RealD zolo_hi,Vector<Coeff_t> & gamma,RealD b,RealD c)
 {
   int Ls=this->Ls;
 
@@ -568,12 +574,12 @@ void CayleyFermion5D<Impl>::SetCoefficientsInternal(RealD zolo_hi,std::vector<Co
     dee[Ls-1] += delta_d;
   }  
 
-  int inv=1;
-  this->MooeeInternalCompute(0,inv,MatpInv,MatmInv);
-  this->MooeeInternalCompute(1,inv,MatpInvDag,MatmInvDag);
+  //  int inv=1;
+  //  this->MooeeInternalCompute(0,inv,MatpInv,MatmInv);
+  //  this->MooeeInternalCompute(1,inv,MatpInvDag,MatmInvDag);
 }
 
-
+#if 0
 template<class Impl>
 void CayleyFermion5D<Impl>::MooeeInternalCompute(int dag, int inv,
 						 Vector<iSinglet<Simd> > & Matp,
@@ -628,35 +634,32 @@ void CayleyFermion5D<Impl>::MooeeInternalCompute(int dag, int inv,
   Matm.resize(Ls*LLs);
 
   for(int s2=0;s2<Ls;s2++){
-  for(int s1=0;s1<LLs;s1++){
-    int istride = LLs;
-    int ostride = 1;
-    Simd Vp;
-    Simd Vm;
-    scalar_type *sp = (scalar_type *)&Vp;
-    scalar_type *sm = (scalar_type *)&Vm;
-    for(int l=0;l<Nsimd;l++){
-      if ( switcheroo<Coeff_t>::iscomplex() ) {
-	sp[l] = PplusMat (l*istride+s1*ostride,s2);
-	sm[l] = PminusMat(l*istride+s1*ostride,s2);
-      } else { 
-      // if real
-	scalar_type tmp;
-	tmp = PplusMat (l*istride+s1*ostride,s2);
-	sp[l] = scalar_type(tmp.real(),tmp.real());
-	tmp = PminusMat(l*istride+s1*ostride,s2);
-	sm[l] = scalar_type(tmp.real(),tmp.real());
+    for(int s1=0;s1<LLs;s1++){
+      int istride = LLs;
+      int ostride = 1;
+      Simd Vp;
+      Simd Vm;
+      scalar_type *sp = (scalar_type *)&Vp;
+      scalar_type *sm = (scalar_type *)&Vm;
+      for(int l=0;l<Nsimd;l++){
+	if ( switcheroo<Coeff_t>::iscomplex() ) {
+	  sp[l] = PplusMat (l*istride+s1*ostride,s2);
+	  sm[l] = PminusMat(l*istride+s1*ostride,s2);
+	} else { 
+	  // if real
+	  scalar_type tmp;
+	  tmp = PplusMat (l*istride+s1*ostride,s2);
+	  sp[l] = scalar_type(tmp.real(),tmp.real());
+	  tmp = PminusMat(l*istride+s1*ostride,s2);
+	  sm[l] = scalar_type(tmp.real(),tmp.real());
+	}
       }
-    }
-    Matp[LLs*s2+s1] = Vp;
-    Matm[LLs*s2+s1] = Vm;
-  }}
+      Matp[LLs*s2+s1] = Vp;
+      Matm[LLs*s2+s1] = Vm;
+    }}
 }
-
-
-  FermOpTemplateInstantiate(CayleyFermion5D);
-  GparityFermOpTemplateInstantiate(CayleyFermion5D);
-
-}}
+#endif
+
+NAMESPACE_END(Grid);
 
 
diff --git a/Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h b/Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h
new file mode 100644
index 00000000..8af3e7c0
--- /dev/null
+++ b/Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h
@@ -0,0 +1,235 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/CayleyFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/CayleyFermion5D.h>
+
+
+NAMESPACE_BEGIN(Grid);
+
+// Pminus fowards
+// Pplus  backwards..
+template<class Impl>  
+void
+CayleyFermion5D<Impl>::M5D(const FermionField &psi_i,
+			       const FermionField &phi_i, 
+			       FermionField &chi_i,
+			       Vector<Coeff_t> &lower,
+			       Vector<Coeff_t> &diag,
+			       Vector<Coeff_t> &upper)
+{
+  
+  chi_i.Checkerboard()=psi_i.Checkerboard();
+  GridBase *grid=psi_i.Grid();
+  auto psi = psi_i.View();
+  auto phi = phi_i.View();
+  auto chi = chi_i.View();
+  assert(phi.Checkerboard() == psi.Checkerboard());
+
+  int Ls =this->Ls;
+
+  // 10 = 3 complex mult + 2 complex add
+  // Flops = 10.0*(Nc*Ns) *Ls*vol (/2 for red black counting)
+  M5Dcalls++;
+  M5Dtime-=usecond();
+
+  uint64_t nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+    uint64_t ss= sss*Ls;
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    spinor tmp1, tmp2;
+    for(int s=0;s<Ls;s++){
+      uint64_t idx_u = ss+((s+1)%Ls);
+      uint64_t idx_l = ss+((s+Ls-1)%Ls);
+      spProj5m(tmp1,psi(idx_u));
+      spProj5p(tmp2,psi(idx_l));
+      coalescedWrite(chi[ss+s],diag[s]*phi(ss+s)+upper[s]*tmp1+lower[s]*tmp2);
+    }
+  });
+  M5Dtime+=usecond();
+}
+
+template<class Impl>  
+void
+CayleyFermion5D<Impl>::M5Ddag(const FermionField &psi_i,
+			      const FermionField &phi_i, 
+			      FermionField &chi_i,
+			      Vector<Coeff_t> &lower,
+			      Vector<Coeff_t> &diag,
+			      Vector<Coeff_t> &upper)
+{
+  chi_i.Checkerboard()=psi_i.Checkerboard();
+  GridBase *grid=psi_i.Grid();
+  auto psi = psi_i.View();
+  auto phi = phi_i.View();
+  auto chi = chi_i.View();
+  assert(phi.Checkerboard() == psi.Checkerboard());
+
+  int Ls=this->Ls;
+
+  // Flops = 6.0*(Nc*Ns) *Ls*vol
+  M5Dcalls++;
+  M5Dtime-=usecond();
+
+  uint64_t nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+    uint64_t ss=sss*Ls;
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    spinor tmp1,tmp2;
+    for(int s=0;s<Ls;s++){
+      uint64_t idx_u = ss+((s+1)%Ls);
+      uint64_t idx_l = ss+((s+Ls-1)%Ls);
+      spProj5p(tmp1,psi(idx_u));
+      spProj5m(tmp2,psi(idx_l));
+      coalescedWrite(chi[ss+s],diag[s]*phi(ss+s)+upper[s]*tmp1+lower[s]*tmp2);
+    }
+  });
+  M5Dtime+=usecond();
+}
+
+template<class Impl>
+void
+CayleyFermion5D<Impl>::MooeeInv    (const FermionField &psi_i, FermionField &chi_i)
+{
+  chi_i.Checkerboard()=psi_i.Checkerboard();
+  GridBase *grid=psi_i.Grid();
+
+  auto psi = psi_i.View();
+  auto chi = chi_i.View();
+
+  int Ls=this->Ls;
+
+  auto plee  = & lee [0];
+  auto pdee  = & dee [0];
+  auto puee  = & uee [0];
+  auto pleem = & leem[0];
+  auto pueem = & ueem[0];
+
+  MooeeInvCalls++;
+  MooeeInvTime-=usecond();
+  uint64_t nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+    uint64_t ss=sss*Ls;
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    spinor tmp;
+
+    // flops = 12*2*Ls + 12*2*Ls + 3*12*Ls + 12*2*Ls  = 12*Ls * (9) = 108*Ls flops
+    // Apply (L^{\prime})^{-1}
+    coalescedWrite(chi[ss],psi(ss)); // chi[0]=psi[0]
+    for(int s=1;s<Ls;s++){
+      spProj5p(tmp,chi(ss+s-1));  
+      coalescedWrite(chi[ss+s] , psi(ss+s)-plee[s-1]*tmp);
+    }
+
+    // L_m^{-1} 
+    for (int s=0;s<Ls-1;s++){ // Chi[ee] = 1 - sum[s<Ls-1] -pleem[s]P_- chi
+      spProj5m(tmp,chi(ss+s));    
+      coalescedWrite(chi[ss+Ls-1], chi(ss+Ls-1) - pleem[s]*tmp);
+    }
+
+    // U_m^{-1} D^{-1}
+    for (int s=0;s<Ls-1;s++){
+      // Chi[s] + 1/d chi[s] 
+      spProj5p(tmp,chi(ss+Ls-1)); 
+      coalescedWrite(chi[ss+s], (1.0/pdee[s])*chi(ss+s)-(pueem[s]/pdee[Ls-1])*tmp);
+    }	
+    coalescedWrite(chi[ss+Ls-1], (1.0/pdee[Ls-1])*chi(ss+Ls-1));
+      
+    // Apply U^{-1}
+    for (int s=Ls-2;s>=0;s--){
+      spProj5m(tmp,chi(ss+s+1));  
+      coalescedWrite(chi[ss+s], chi(ss+s) - puee[s]*tmp);
+    }
+  });
+
+  MooeeInvTime+=usecond();
+
+}
+
+template<class Impl>
+void
+CayleyFermion5D<Impl>::MooeeInvDag (const FermionField &psi_i, FermionField &chi_i)
+{
+  chi_i.Checkerboard()=psi_i.Checkerboard();
+  GridBase *grid=psi_i.Grid();
+  int Ls=this->Ls;
+
+  auto psi = psi_i.View();
+  auto chi = chi_i.View();
+
+  auto plee  = & lee [0];
+  auto pdee  = & dee [0];
+  auto puee  = & uee [0];
+  auto pleem = & leem[0];
+  auto pueem = & ueem[0];
+
+  assert(psi.Checkerboard() == psi.Checkerboard());
+
+  MooeeInvCalls++;
+  MooeeInvTime-=usecond();
+
+
+  uint64_t nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+    uint64_t ss=sss*Ls;
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    spinor tmp;
+
+    // Apply (U^{\prime})^{-dagger}
+    coalescedWrite(chi[ss],psi(ss));
+    for (int s=1;s<Ls;s++){
+      spProj5m(tmp,chi(ss+s-1));
+      coalescedWrite(chi[ss+s], psi(ss+s)-conjugate(puee[s-1])*tmp);
+    }
+    // U_m^{-\dagger} 
+    for (int s=0;s<Ls-1;s++){
+      spProj5p(tmp,chi(ss+s));
+      coalescedWrite(chi[ss+Ls-1], chi(ss+Ls-1) - conjugate(pueem[s])*tmp);
+    }
+
+    // L_m^{-\dagger} D^{-dagger}
+    for (int s=0;s<Ls-1;s++){
+      spProj5m(tmp,chi(ss+Ls-1));
+      coalescedWrite(chi[ss+s], conjugate(1.0/pdee[s])*chi(ss+s)-conjugate(pleem[s]/pdee[Ls-1])*tmp);
+    }	
+    coalescedWrite(chi[ss+Ls-1], conjugate(1.0/pdee[Ls-1])*chi(ss+Ls-1));
+  
+    // Apply L^{-dagger}
+    for (int s=Ls-2;s>=0;s--){
+      spProj5p(tmp,chi(ss+s+1));
+      coalescedWrite(chi[ss+s], chi(ss+s) - conjugate(plee[s])*tmp);
+    }
+  });
+  MooeeInvTime+=usecond();
+
+}
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h b/Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h
new file mode 100644
index 00000000..034ce642
--- /dev/null
+++ b/Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h
@@ -0,0 +1,831 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/CayleyFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/CayleyFermion5D.h>
+
+NAMESPACE_BEGIN(Grid);
+
+/*
+ * Dense matrix versions of routines
+ */
+template<class Impl>
+void
+CayleyFermion5D<Impl>::MooeeInvDag(const FermionField &psi, FermionField &chi)
+{
+  EnableIf<Impl::LsVectorised&&EnableBool,int> sfinae=0;
+  this->MooeeInternal(psi,chi,DaggerYes,InverseYes);
+}
+  
+template<class Impl>
+void
+CayleyFermion5D<Impl>::MooeeInv(const FermionField &psi, FermionField &chi)
+{
+  EnableIf<Impl::LsVectorised&&EnableBool,int> sfinae=0;
+  this->MooeeInternal(psi,chi,DaggerNo,InverseYes);
+}
+template<class Impl>  
+void
+CayleyFermion5D<Impl>::M5D(const FermionField &psi_i,
+			   const FermionField &phi_i, 
+			   FermionField &chi_i,
+			   Vector<Coeff_t> &lower,
+			   Vector<Coeff_t> &diag,
+			   Vector<Coeff_t> &upper)
+{
+  EnableIf<Impl::LsVectorised&&EnableBool,int> sfinae=0;
+  chi_i.Checkerboard()=psi_i.Checkerboard();
+  GridBase *grid=psi_i.Grid();
+  auto psi = psi_i.View();
+  auto phi = phi_i.View();
+  auto chi = chi_i.View();
+  int Ls   = this->Ls;
+  int LLs  = grid->_rdimensions[0];
+  const int nsimd= Simd::Nsimd();
+
+  Vector<iSinglet<Simd> > u(LLs);
+  Vector<iSinglet<Simd> > l(LLs);
+  Vector<iSinglet<Simd> > d(LLs);
+
+  assert(Ls/LLs==nsimd);
+  assert(phi.Checkerboard() == psi.Checkerboard());
+
+  // just directly address via type pun
+  typedef typename Simd::scalar_type scalar_type;
+  scalar_type * u_p = (scalar_type *)&u[0];
+  scalar_type * l_p = (scalar_type *)&l[0];
+  scalar_type * d_p = (scalar_type *)&d[0];
+
+  for(int o=0;o<LLs;o++){ // outer
+    for(int i=0;i<nsimd;i++){ //inner
+      int s  = o+i*LLs;
+      int ss = o*nsimd+i;
+      u_p[ss] = upper[s];
+      l_p[ss] = lower[s];
+      d_p[ss] = diag[s];
+    }}
+
+
+  M5Dcalls++;
+  M5Dtime-=usecond();
+
+  assert(Nc==3);
+
+  thread_loop( (int ss=0;ss<grid->oSites();ss+=LLs),{ // adds LLs
+#if 0
+    alignas(64) SiteHalfSpinor hp;
+    alignas(64) SiteHalfSpinor hm;
+    alignas(64) SiteSpinor fp;
+    alignas(64) SiteSpinor fm;
+
+    for(int v=0;v<LLs;v++){
+
+      int vp=(v+1)%LLs;
+      int vm=(v+LLs-1)%LLs;
+
+      spProj5m(hp,psi[ss+vp]);
+      spProj5p(hm,psi[ss+vm]);
+
+      if ( vp<=v ) rotate(hp,hp,1);
+      if ( vm>=v ) rotate(hm,hm,nsimd-1);
+	
+      hp=0.5*hp;
+      hm=0.5*hm;
+
+      spRecon5m(fp,hp);
+      spRecon5p(fm,hm);
+
+      chi[ss+v] = d[v]*phi[ss+v];
+      chi[ss+v] = chi[ss+v]     +u[v]*fp;
+      chi[ss+v] = chi[ss+v]     +l[v]*fm;
+
+    }
+#else
+    for(int v=0;v<LLs;v++){
+      
+      vprefetch(psi[ss+v+LLs]);
+
+      int vp= (v==LLs-1) ? 0     : v+1;
+      int vm= (v==0    ) ? LLs-1 : v-1;
+	
+      Simd hp_00 = psi[ss+vp]()(2)(0); 
+      Simd hp_01 = psi[ss+vp]()(2)(1); 
+      Simd hp_02 = psi[ss+vp]()(2)(2); 
+      Simd hp_10 = psi[ss+vp]()(3)(0); 
+      Simd hp_11 = psi[ss+vp]()(3)(1); 
+      Simd hp_12 = psi[ss+vp]()(3)(2); 
+	
+      Simd hm_00 = psi[ss+vm]()(0)(0); 
+      Simd hm_01 = psi[ss+vm]()(0)(1); 
+      Simd hm_02 = psi[ss+vm]()(0)(2); 
+      Simd hm_10 = psi[ss+vm]()(1)(0); 
+      Simd hm_11 = psi[ss+vm]()(1)(1); 
+      Simd hm_12 = psi[ss+vm]()(1)(2); 
+
+      if ( vp<=v ) {
+	hp_00.v = Optimization::Rotate::tRotate<2>(hp_00.v);
+	hp_01.v = Optimization::Rotate::tRotate<2>(hp_01.v);
+	hp_02.v = Optimization::Rotate::tRotate<2>(hp_02.v);
+	hp_10.v = Optimization::Rotate::tRotate<2>(hp_10.v);
+	hp_11.v = Optimization::Rotate::tRotate<2>(hp_11.v);
+	hp_12.v = Optimization::Rotate::tRotate<2>(hp_12.v);
+      }
+      if ( vm>=v ) {
+	hm_00.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_00.v);
+	hm_01.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_01.v);
+	hm_02.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_02.v);
+	hm_10.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_10.v);
+	hm_11.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_11.v);
+	hm_12.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_12.v);
+      }
+
+      // Can force these to real arithmetic and save 2x.
+      Simd p_00  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(0))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_00); 
+      Simd p_01  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(1))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_01); 
+      Simd p_02  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(2))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_02); 
+      Simd p_10  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(0))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_10); 
+      Simd p_11  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(1))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_11); 
+      Simd p_12  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(2))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_12); 
+      Simd p_20  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(0))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_00); 
+      Simd p_21  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(1))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_01); 
+      Simd p_22  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(2))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_02);  
+      Simd p_30  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(0))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_10); 
+      Simd p_31  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(1))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_11); 
+      Simd p_32  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(2))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_12); 
+
+      vstream(chi[ss+v]()(0)(0),p_00);
+      vstream(chi[ss+v]()(0)(1),p_01);
+      vstream(chi[ss+v]()(0)(2),p_02);
+      vstream(chi[ss+v]()(1)(0),p_10);
+      vstream(chi[ss+v]()(1)(1),p_11);
+      vstream(chi[ss+v]()(1)(2),p_12);
+      vstream(chi[ss+v]()(2)(0),p_20);
+      vstream(chi[ss+v]()(2)(1),p_21);
+      vstream(chi[ss+v]()(2)(2),p_22);
+      vstream(chi[ss+v]()(3)(0),p_30);
+      vstream(chi[ss+v]()(3)(1),p_31);
+      vstream(chi[ss+v]()(3)(2),p_32);
+
+    }
+#endif
+  });
+  M5Dtime+=usecond();
+}
+
+template<class Impl>  
+void
+CayleyFermion5D<Impl>::M5Ddag(const FermionField &psi_i,
+			      const FermionField &phi_i, 
+			      FermionField &chi_i,
+			      Vector<Coeff_t> &lower,
+			      Vector<Coeff_t> &diag,
+			      Vector<Coeff_t> &upper)
+{
+  EnableIf<Impl::LsVectorised&&EnableBool,int> sfinae=0;
+  chi_i.Checkerboard()=psi_i.Checkerboard();
+  GridBase *grid=psi_i.Grid();
+  auto psi=psi_i.View();
+  auto phi=phi_i.View();
+  auto chi=chi_i.View();
+  int Ls   = this->Ls;
+  int LLs  = grid->_rdimensions[0];
+  int nsimd= Simd::Nsimd();
+
+  Vector<iSinglet<Simd> > u(LLs);
+  Vector<iSinglet<Simd> > l(LLs);
+  Vector<iSinglet<Simd> > d(LLs);
+
+  assert(Ls/LLs==nsimd);
+  assert(phi.Checkerboard() == psi.Checkerboard());
+
+  // just directly address via type pun
+  typedef typename Simd::scalar_type scalar_type;
+  scalar_type * u_p = (scalar_type *)&u[0];
+  scalar_type * l_p = (scalar_type *)&l[0];
+  scalar_type * d_p = (scalar_type *)&d[0];
+
+  for(int o=0;o<LLs;o++){ // outer
+    for(int i=0;i<nsimd;i++){ //inner
+      int s  = o+i*LLs;
+      int ss = o*nsimd+i;
+      u_p[ss] = upper[s];
+      l_p[ss] = lower[s];
+      d_p[ss] = diag[s];
+    }}
+
+  M5Dcalls++;
+  M5Dtime-=usecond();
+  thread_loop( (int ss=0;ss<grid->oSites();ss+=LLs),{ // adds LLs
+#if 0
+    alignas(64) SiteHalfSpinor hp;
+    alignas(64) SiteHalfSpinor hm;
+    alignas(64) SiteSpinor fp;
+    alignas(64) SiteSpinor fm;
+
+    for(int v=0;v<LLs;v++){
+
+      int vp=(v+1)%LLs;
+      int vm=(v+LLs-1)%LLs;
+
+      spProj5p(hp,psi[ss+vp]);
+      spProj5m(hm,psi[ss+vm]);
+
+      if ( vp<=v ) rotate(hp,hp,1);
+      if ( vm>=v ) rotate(hm,hm,nsimd-1);
+      
+      hp=hp*0.5;
+      hm=hm*0.5;
+      spRecon5p(fp,hp);
+      spRecon5m(fm,hm);
+
+      chi[ss+v] = d[v]*phi[ss+v]+u[v]*fp;
+      chi[ss+v] = chi[ss+v]     +l[v]*fm;
+
+    }
+#else
+    for(int v=0;v<LLs;v++){
+
+      vprefetch(psi[ss+v+LLs]);
+
+      int vp= (v==LLs-1) ? 0     : v+1;
+      int vm= (v==0    ) ? LLs-1 : v-1;
+	
+      Simd hp_00 = psi[ss+vp]()(0)(0); 
+      Simd hp_01 = psi[ss+vp]()(0)(1); 
+      Simd hp_02 = psi[ss+vp]()(0)(2); 
+      Simd hp_10 = psi[ss+vp]()(1)(0); 
+      Simd hp_11 = psi[ss+vp]()(1)(1); 
+      Simd hp_12 = psi[ss+vp]()(1)(2); 
+	
+      Simd hm_00 = psi[ss+vm]()(2)(0); 
+      Simd hm_01 = psi[ss+vm]()(2)(1); 
+      Simd hm_02 = psi[ss+vm]()(2)(2); 
+      Simd hm_10 = psi[ss+vm]()(3)(0); 
+      Simd hm_11 = psi[ss+vm]()(3)(1); 
+      Simd hm_12 = psi[ss+vm]()(3)(2); 
+
+      if ( vp<=v ) {
+	hp_00.v = Optimization::Rotate::tRotate<2>(hp_00.v);
+	hp_01.v = Optimization::Rotate::tRotate<2>(hp_01.v);
+	hp_02.v = Optimization::Rotate::tRotate<2>(hp_02.v);
+	hp_10.v = Optimization::Rotate::tRotate<2>(hp_10.v);
+	hp_11.v = Optimization::Rotate::tRotate<2>(hp_11.v);
+	hp_12.v = Optimization::Rotate::tRotate<2>(hp_12.v);
+      }
+      if ( vm>=v ) {
+	hm_00.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_00.v);
+	hm_01.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_01.v);
+	hm_02.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_02.v);
+	hm_10.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_10.v);
+	hm_11.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_11.v);
+	hm_12.v = Optimization::Rotate::tRotate<2*Simd::Nsimd()-2>(hm_12.v);
+      }
+
+      Simd p_00  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(0))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_00); 
+      Simd p_01  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(1))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_01); 
+      Simd p_02  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(0)(2))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_02); 
+      Simd p_10  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(0))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_10); 
+      Simd p_11  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(1))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_11); 
+      Simd p_12  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(1)(2))  + switcheroo<Coeff_t>::mult(u[v]()()(),hp_12); 
+
+      Simd p_20  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(0))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_00); 
+      Simd p_21  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(1))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_01); 
+      Simd p_22  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(2)(2))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_02);  
+      Simd p_30  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(0))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_10); 
+      Simd p_31  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(1))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_11); 
+      Simd p_32  = switcheroo<Coeff_t>::mult(d[v]()()(), phi[ss+v]()(3)(2))  + switcheroo<Coeff_t>::mult(l[v]()()(),hm_12); 
+
+      vstream(chi[ss+v]()(0)(0),p_00);
+      vstream(chi[ss+v]()(0)(1),p_01);
+      vstream(chi[ss+v]()(0)(2),p_02);
+      vstream(chi[ss+v]()(1)(0),p_10);
+      vstream(chi[ss+v]()(1)(1),p_11);
+      vstream(chi[ss+v]()(1)(2),p_12);
+      vstream(chi[ss+v]()(2)(0),p_20);
+      vstream(chi[ss+v]()(2)(1),p_21);
+      vstream(chi[ss+v]()(2)(2),p_22);
+      vstream(chi[ss+v]()(3)(0),p_30);
+      vstream(chi[ss+v]()(3)(1),p_31);
+      vstream(chi[ss+v]()(3)(2),p_32);
+    }
+#endif
+  });
+  M5Dtime+=usecond();
+}
+
+
+#ifdef AVX512 
+#include <simd/Intel512common.h>
+#include <simd/Intel512avx.h>
+#include <simd/Intel512single.h>
+#endif 
+
+template<class Impl>
+void
+CayleyFermion5D<Impl>::MooeeInternalAsm(const FermionField &psi_i, FermionField &chi_i,
+					int LLs, int site,
+					Vector<iSinglet<Simd> > &Matp,
+					Vector<iSinglet<Simd> > &Matm)
+{
+  EnableIf<Impl::LsVectorised&&EnableBool,int> sfinae=0;
+  auto psi = psi_i.View();
+  auto chi = chi_i.View();
+#ifndef AVX512
+  {
+    SiteHalfSpinor BcastP;
+    SiteHalfSpinor BcastM;
+    SiteHalfSpinor SiteChiP;
+    SiteHalfSpinor SiteChiM;
+
+    // Ls*Ls * 2 * 12 * vol flops
+    for(int s1=0;s1<LLs;s1++){ 
+      for(int s2=0;s2<LLs;s2++){ 
+	for(int  l=0; l<Simd::Nsimd();l++){ // simd lane
+
+	  int s=s2+l*LLs;
+	  int lex=s2+LLs*site;
+	
+	  if ( s2==0 && l==0) {
+	    SiteChiP=Zero();
+	    SiteChiM=Zero();
+	  }
+	
+	  for(int sp=0;sp<2;sp++){
+	    for(int co=0;co<Nc;co++){
+	      vbroadcast(BcastP()(sp  )(co),psi[lex]()(sp)(co),l);
+	    }}
+	  for(int sp=0;sp<2;sp++){
+	    for(int co=0;co<Nc;co++){
+	      vbroadcast(BcastM()(sp  )(co),psi[lex]()(sp+2)(co),l);
+	    }}
+
+	  for(int sp=0;sp<2;sp++){
+	    for(int co=0;co<Nc;co++){
+	      SiteChiP()(sp)(co)=real_madd(Matp[LLs*s+s1]()()(),BcastP()(sp)(co),SiteChiP()(sp)(co)); // 1100 us.
+	      SiteChiM()(sp)(co)=real_madd(Matm[LLs*s+s1]()()(),BcastM()(sp)(co),SiteChiM()(sp)(co)); // each found by commenting out
+	    }}
+
+	}}
+      {
+	int lex = s1+LLs*site;
+	for(int sp=0;sp<2;sp++){
+	  for(int co=0;co<Nc;co++){
+	    vstream(chi[lex]()(sp)(co), SiteChiP()(sp)(co));
+	    vstream(chi[lex]()(sp+2)(co), SiteChiM()(sp)(co));
+	  }}
+      }
+    }
+
+  }
+#else
+  {
+    // pointers
+    //  MASK_REGS;
+#define Chi_00 %%zmm1
+#define Chi_01 %%zmm2
+#define Chi_02 %%zmm3
+#define Chi_10 %%zmm4
+#define Chi_11 %%zmm5
+#define Chi_12 %%zmm6
+#define Chi_20 %%zmm7
+#define Chi_21 %%zmm8
+#define Chi_22 %%zmm9
+#define Chi_30 %%zmm10
+#define Chi_31 %%zmm11
+#define Chi_32 %%zmm12
+
+#define BCAST0   %%zmm13
+#define BCAST1   %%zmm14
+#define BCAST2   %%zmm15
+#define BCAST3   %%zmm16
+#define BCAST4   %%zmm17
+#define BCAST5   %%zmm18
+#define BCAST6   %%zmm19
+#define BCAST7   %%zmm20
+#define BCAST8   %%zmm21
+#define BCAST9   %%zmm22
+#define BCAST10  %%zmm23
+#define BCAST11  %%zmm24
+
+    int incr=LLs*LLs*sizeof(iSinglet<Simd>);
+    for(int s1=0;s1<LLs;s1++){ 
+      for(int s2=0;s2<LLs;s2++){ 
+	int lex=s2+LLs*site;
+	uint64_t a0 = (uint64_t)&Matp[LLs*s2+s1]; // should be cacheable
+	uint64_t a1 = (uint64_t)&Matm[LLs*s2+s1];
+	uint64_t a2 = (uint64_t)&psi[lex];
+	for(int  l=0; l<Simd::Nsimd();l++){ // simd lane
+	  if ( (s2+l)==0 ) {
+	    asm (
+		 VPREFETCH1(0,%2)  	     VPREFETCH1(0,%1)
+		 VPREFETCH1(12,%2)  	     VPREFETCH1(13,%2)
+		 VPREFETCH1(14,%2)  	     VPREFETCH1(15,%2)         
+		 VBCASTCDUP(0,%2,BCAST0)   
+		 VBCASTCDUP(1,%2,BCAST1)   
+		 VBCASTCDUP(2,%2,BCAST2)   
+		 VBCASTCDUP(3,%2,BCAST3)   
+		 VBCASTCDUP(4,%2,BCAST4)     VMULMEM (0,%0,BCAST0,Chi_00)
+		 VBCASTCDUP(5,%2,BCAST5)     VMULMEM (0,%0,BCAST1,Chi_01)
+		 VBCASTCDUP(6,%2,BCAST6)     VMULMEM (0,%0,BCAST2,Chi_02)
+		 VBCASTCDUP(7,%2,BCAST7)     VMULMEM (0,%0,BCAST3,Chi_10)
+		 VBCASTCDUP(8,%2,BCAST8)     VMULMEM (0,%0,BCAST4,Chi_11)
+		 VBCASTCDUP(9,%2,BCAST9)     VMULMEM (0,%0,BCAST5,Chi_12)
+		 VBCASTCDUP(10,%2,BCAST10)   VMULMEM (0,%1,BCAST6,Chi_20)
+		 VBCASTCDUP(11,%2,BCAST11)   VMULMEM (0,%1,BCAST7,Chi_21)
+		 VMULMEM (0,%1,BCAST8,Chi_22)         
+		 VMULMEM (0,%1,BCAST9,Chi_30)
+		 VMULMEM (0,%1,BCAST10,Chi_31)       
+		 VMULMEM (0,%1,BCAST11,Chi_32)
+		 : : "r" (a0), "r" (a1), "r" (a2)  );
+	  } else { 
+	    asm (
+		 VBCASTCDUP(0,%2,BCAST0)   VMADDMEM (0,%0,BCAST0,Chi_00)
+		 VBCASTCDUP(1,%2,BCAST1)   VMADDMEM (0,%0,BCAST1,Chi_01)
+		 VBCASTCDUP(2,%2,BCAST2)   VMADDMEM (0,%0,BCAST2,Chi_02)
+		 VBCASTCDUP(3,%2,BCAST3)   VMADDMEM (0,%0,BCAST3,Chi_10)
+		 VBCASTCDUP(4,%2,BCAST4)   VMADDMEM (0,%0,BCAST4,Chi_11)
+		 VBCASTCDUP(5,%2,BCAST5)   VMADDMEM (0,%0,BCAST5,Chi_12)
+		 VBCASTCDUP(6,%2,BCAST6)   VMADDMEM (0,%1,BCAST6,Chi_20)
+		 VBCASTCDUP(7,%2,BCAST7)   VMADDMEM (0,%1,BCAST7,Chi_21)
+		 VBCASTCDUP(8,%2,BCAST8)   VMADDMEM (0,%1,BCAST8,Chi_22)
+		 VBCASTCDUP(9,%2,BCAST9)   VMADDMEM (0,%1,BCAST9,Chi_30)
+		 VBCASTCDUP(10,%2,BCAST10)  VMADDMEM (0,%1,BCAST10,Chi_31)
+		 VBCASTCDUP(11,%2,BCAST11)  VMADDMEM (0,%1,BCAST11,Chi_32) 
+		 : : "r" (a0), "r" (a1), "r" (a2)  );
+	  }
+	  a0 = a0+incr;
+	  a1 = a1+incr;
+	a2 = a2+sizeof(typename Simd::scalar_type);
+	}}
+      {
+	int lexa = s1+LLs*site;
+	asm (
+	     VSTORE(0,%0,Chi_00) VSTORE(1 ,%0,Chi_01)  VSTORE(2 ,%0,Chi_02)		
+	     VSTORE(3,%0,Chi_10) VSTORE(4 ,%0,Chi_11)  VSTORE(5 ,%0,Chi_12)		
+	     VSTORE(6,%0,Chi_20) VSTORE(7 ,%0,Chi_21)  VSTORE(8 ,%0,Chi_22)		
+	     VSTORE(9,%0,Chi_30) VSTORE(10,%0,Chi_31)  VSTORE(11,%0,Chi_32)		
+	     : : "r" ((uint64_t)&chi[lexa]) : "memory" );
+
+      }
+    }
+  }
+#undef Chi_00
+#undef Chi_01
+#undef Chi_02
+#undef Chi_10
+#undef Chi_11
+#undef Chi_12
+#undef Chi_20
+#undef Chi_21
+#undef Chi_22
+#undef Chi_30
+#undef Chi_31
+#undef Chi_32
+
+#undef BCAST0
+#undef BCAST1
+#undef BCAST2
+#undef BCAST3
+#undef BCAST4
+#undef BCAST5
+#undef BCAST6
+#undef BCAST7
+#undef BCAST8
+#undef BCAST9
+#undef BCAST10
+#undef BCAST11
+#endif
+};
+
+// Z-mobius version
+template<class Impl>
+void
+CayleyFermion5D<Impl>::MooeeInternalZAsm(const FermionField &psi_i, FermionField &chi_i,
+					 int LLs, int site, Vector<iSinglet<Simd> > &Matp, Vector<iSinglet<Simd> > &Matm)
+{
+  EnableIf<Impl::LsVectorised,int> sfinae=0;
+#ifndef AVX512
+  {
+    auto psi = psi_i.View();
+    auto chi = chi_i.View();
+
+    SiteHalfSpinor BcastP;
+    SiteHalfSpinor BcastM;
+    SiteHalfSpinor SiteChiP;
+    SiteHalfSpinor SiteChiM;
+
+    // Ls*Ls * 2 * 12 * vol flops
+    for(int s1=0;s1<LLs;s1++){ 
+      for(int s2=0;s2<LLs;s2++){ 
+	for(int  l=0; l<Simd::Nsimd();l++){ // simd lane
+
+	  int s=s2+l*LLs;
+	  int lex=s2+LLs*site;
+	
+	  if ( s2==0 && l==0) {
+	    SiteChiP=Zero();
+	    SiteChiM=Zero();
+	  }
+	
+	  for(int sp=0;sp<2;sp++){
+	    for(int co=0;co<Nc;co++){
+	      vbroadcast(BcastP()(sp  )(co),psi[lex]()(sp)(co),l);
+	    }}
+	  for(int sp=0;sp<2;sp++){
+	    for(int co=0;co<Nc;co++){
+	      vbroadcast(BcastM()(sp  )(co),psi[lex]()(sp+2)(co),l);
+	    }}
+
+	  for(int sp=0;sp<2;sp++){
+	    for(int co=0;co<Nc;co++){
+	      SiteChiP()(sp)(co)=SiteChiP()(sp)(co)+ Matp[LLs*s+s1]()()()*BcastP()(sp)(co); 
+	      SiteChiM()(sp)(co)=SiteChiM()(sp)(co)+ Matm[LLs*s+s1]()()()*BcastM()(sp)(co); 
+	    }}
+
+
+	}}
+      {
+	int lex = s1+LLs*site;
+	for(int sp=0;sp<2;sp++){
+	  for(int co=0;co<Nc;co++){
+	    vstream(chi[lex]()(sp)(co), SiteChiP()(sp)(co));
+	    vstream(chi[lex]()(sp+2)(co), SiteChiM()(sp)(co));
+	  }}
+      }
+    }
+
+  }
+#else
+  {
+    auto psi = psi_i.View();
+    auto chi = chi_i.View();
+    // pointers
+    //  MASK_REGS;
+#define Chi_00 %zmm0
+#define Chi_01 %zmm1
+#define Chi_02 %zmm2
+#define Chi_10 %zmm3
+#define Chi_11 %zmm4
+#define Chi_12 %zmm5
+#define Chi_20 %zmm6
+#define Chi_21 %zmm7
+#define Chi_22 %zmm8
+#define Chi_30 %zmm9
+#define Chi_31 %zmm10
+#define Chi_32 %zmm11
+#define pChi_00 %%zmm0
+#define pChi_01 %%zmm1
+#define pChi_02 %%zmm2
+#define pChi_10 %%zmm3
+#define pChi_11 %%zmm4
+#define pChi_12 %%zmm5
+#define pChi_20 %%zmm6
+#define pChi_21 %%zmm7
+#define pChi_22 %%zmm8
+#define pChi_30 %%zmm9
+#define pChi_31 %%zmm10
+#define pChi_32 %%zmm11
+
+#define BCAST_00   %zmm12
+#define  SHUF_00   %zmm13
+#define BCAST_01   %zmm14
+#define  SHUF_01   %zmm15
+#define BCAST_02   %zmm16
+#define  SHUF_02   %zmm17
+#define BCAST_10   %zmm18
+#define  SHUF_10   %zmm19
+#define BCAST_11   %zmm20
+#define  SHUF_11   %zmm21
+#define BCAST_12   %zmm22
+#define  SHUF_12   %zmm23
+
+#define Mp  %zmm24
+#define Mps %zmm25
+#define Mm  %zmm26
+#define Mms %zmm27
+#define N 8
+    int incr=LLs*LLs*sizeof(iSinglet<Simd>);
+    for(int s1=0;s1<LLs;s1++){ 
+      for(int s2=0;s2<LLs;s2++){ 
+	int lex=s2+LLs*site;
+	uint64_t a0 = (uint64_t)&Matp[LLs*s2+s1]; // should be cacheable
+	uint64_t a1 = (uint64_t)&Matm[LLs*s2+s1];
+	uint64_t a2 = (uint64_t)&psi[lex];
+	for(int  l=0; l<Simd::Nsimd();l++){ // simd lane
+	  if ( (s2+l)==0 ) {
+	    LOAD64(%r8,a0);
+	    LOAD64(%r9,a1);
+	    LOAD64(%r10,a2);
+	    asm (
+		 VLOAD(0,%r8,Mp)// i r
+		 VLOAD(0,%r9,Mm)
+		 VSHUF(Mp,Mps)  // r i 
+		 VSHUF(Mm,Mms)
+		 VPREFETCH1(12,%r10)  	     VPREFETCH1(13,%r10)
+		 VPREFETCH1(14,%r10)  	     VPREFETCH1(15,%r10)         
+
+		 VMULIDUP(0*N,%r10,Mps,Chi_00)
+		 VMULIDUP(1*N,%r10,Mps,Chi_01)
+		 VMULIDUP(2*N,%r10,Mps,Chi_02)
+		 VMULIDUP(3*N,%r10,Mps,Chi_10)
+		 VMULIDUP(4*N,%r10,Mps,Chi_11)
+		 VMULIDUP(5*N,%r10,Mps,Chi_12)
+
+		 VMULIDUP(6*N ,%r10,Mms,Chi_20)
+		 VMULIDUP(7*N ,%r10,Mms,Chi_21)
+		 VMULIDUP(8*N ,%r10,Mms,Chi_22)
+		 VMULIDUP(9*N ,%r10,Mms,Chi_30)
+		 VMULIDUP(10*N,%r10,Mms,Chi_31)
+		 VMULIDUP(11*N,%r10,Mms,Chi_32)
+
+		 VMADDSUBRDUP(0*N,%r10,Mp,Chi_00)
+		 VMADDSUBRDUP(1*N,%r10,Mp,Chi_01)
+		 VMADDSUBRDUP(2*N,%r10,Mp,Chi_02)
+		 VMADDSUBRDUP(3*N,%r10,Mp,Chi_10)
+		 VMADDSUBRDUP(4*N,%r10,Mp,Chi_11)
+		 VMADDSUBRDUP(5*N,%r10,Mp,Chi_12)
+
+		 VMADDSUBRDUP(6*N ,%r10,Mm,Chi_20)
+		 VMADDSUBRDUP(7*N ,%r10,Mm,Chi_21)
+		 VMADDSUBRDUP(8*N ,%r10,Mm,Chi_22)
+		 VMADDSUBRDUP(9*N ,%r10,Mm,Chi_30)
+		 VMADDSUBRDUP(10*N,%r10,Mm,Chi_31)
+		 VMADDSUBRDUP(11*N,%r10,Mm,Chi_32)
+		 );
+	  } else { 
+	    LOAD64(%r8,a0);
+	    LOAD64(%r9,a1);
+	    LOAD64(%r10,a2);
+	    asm (
+		 VLOAD(0,%r8,Mp)
+		 VSHUF(Mp,Mps)
+
+		 VLOAD(0,%r9,Mm)
+		 VSHUF(Mm,Mms)
+
+		 VMADDSUBIDUP(0*N,%r10,Mps,Chi_00) //  Mri * Pii +- Cir
+		 VMADDSUBIDUP(1*N,%r10,Mps,Chi_01)
+		 VMADDSUBIDUP(2*N,%r10,Mps,Chi_02)
+		 VMADDSUBIDUP(3*N,%r10,Mps,Chi_10)
+		 VMADDSUBIDUP(4*N,%r10,Mps,Chi_11)
+		 VMADDSUBIDUP(5*N,%r10,Mps,Chi_12)
+
+		 VMADDSUBIDUP(6 *N,%r10,Mms,Chi_20)
+		 VMADDSUBIDUP(7 *N,%r10,Mms,Chi_21)
+		 VMADDSUBIDUP(8 *N,%r10,Mms,Chi_22)
+		 VMADDSUBIDUP(9 *N,%r10,Mms,Chi_30)
+		 VMADDSUBIDUP(10*N,%r10,Mms,Chi_31)
+		 VMADDSUBIDUP(11*N,%r10,Mms,Chi_32)
+
+		 VMADDSUBRDUP(0*N,%r10,Mp,Chi_00) //  Cir = Mir * Prr +- ( Mri * Pii +- Cir) 
+		 VMADDSUBRDUP(1*N,%r10,Mp,Chi_01) //  Ci = MiPr + Ci + MrPi ;    Cr = MrPr - ( MiPi - Cr)
+		 VMADDSUBRDUP(2*N,%r10,Mp,Chi_02)
+		 VMADDSUBRDUP(3*N,%r10,Mp,Chi_10)
+		 VMADDSUBRDUP(4*N,%r10,Mp,Chi_11)
+		 VMADDSUBRDUP(5*N,%r10,Mp,Chi_12)
+
+		 VMADDSUBRDUP(6 *N,%r10,Mm,Chi_20)
+		 VMADDSUBRDUP(7 *N,%r10,Mm,Chi_21)
+		 VMADDSUBRDUP(8 *N,%r10,Mm,Chi_22)
+		 VMADDSUBRDUP(9 *N,%r10,Mm,Chi_30)
+		 VMADDSUBRDUP(10*N,%r10,Mm,Chi_31)
+		 VMADDSUBRDUP(11*N,%r10,Mm,Chi_32)
+		 );
+	  }
+	  a0 = a0+incr;
+	  a1 = a1+incr;
+	a2 = a2+sizeof(typename Simd::scalar_type);
+	}}
+      {
+	int lexa = s1+LLs*site;
+	/*
+	  SiteSpinor tmp;
+	  asm (
+	  VSTORE(0,%0,pChi_00) VSTORE(1 ,%0,pChi_01)  VSTORE(2 ,%0,pChi_02)		
+	  VSTORE(3,%0,pChi_10) VSTORE(4 ,%0,pChi_11)  VSTORE(5 ,%0,pChi_12)		
+	  VSTORE(6,%0,pChi_20) VSTORE(7 ,%0,pChi_21)  VSTORE(8 ,%0,pChi_22)		
+	  VSTORE(9,%0,pChi_30) VSTORE(10,%0,pChi_31)  VSTORE(11,%0,pChi_32)		
+	  : : "r" ((uint64_t)&tmp) : "memory" );
+	*/
+
+	asm (
+	     VSTORE(0,%0,pChi_00) VSTORE(1 ,%0,pChi_01)  VSTORE(2 ,%0,pChi_02)		
+	     VSTORE(3,%0,pChi_10) VSTORE(4 ,%0,pChi_11)  VSTORE(5 ,%0,pChi_12)		
+	     VSTORE(6,%0,pChi_20) VSTORE(7 ,%0,pChi_21)  VSTORE(8 ,%0,pChi_22)		
+	     VSTORE(9,%0,pChi_30) VSTORE(10,%0,pChi_31)  VSTORE(11,%0,pChi_32)		
+	     : : "r" ((uint64_t)&chi[lexa]) : "memory" );
+
+	//      if ( 1 || (site==0) ) { 
+	//	std::cout<<site << " s1 "<<s1<<"\n\t"<<tmp << "\n't" << chi[lexa] <<"\n\t"<<tmp-chi[lexa]<<std::endl;
+	//      }
+      }
+    }
+  }
+#undef Chi_00
+#undef Chi_01
+#undef Chi_02
+#undef Chi_10
+#undef Chi_11
+#undef Chi_12
+#undef Chi_20
+#undef Chi_21
+#undef Chi_22
+#undef Chi_30
+#undef Chi_31
+#undef Chi_32
+
+#undef BCAST0
+#undef BCAST1
+#undef BCAST2
+#undef BCAST3
+#undef BCAST4
+#undef BCAST5
+#undef BCAST6
+#undef BCAST7
+#undef BCAST8
+#undef BCAST9
+#undef BCAST10
+#undef BCAST11
+
+#endif
+};
+
+
+template<class Impl>
+void
+CayleyFermion5D<Impl>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv)
+{
+  EnableIf<Impl::LsVectorised,int> sfinae=0;
+  chi.Checkerboard()=psi.Checkerboard();
+
+  int Ls=this->Ls;
+  int LLs = psi.Grid()->_rdimensions[0];
+  int vol = psi.Grid()->oSites()/LLs;
+
+  
+  Vector<iSinglet<Simd> >  Matp;
+  Vector<iSinglet<Simd> >  Matm;
+  Vector<iSinglet<Simd> >  *_Matp;
+  Vector<iSinglet<Simd> >  *_Matm;
+  
+  //  MooeeInternalCompute(dag,inv,Matp,Matm);
+  if ( inv && dag ) { 
+    _Matp = &MatpInvDag;
+    _Matm = &MatmInvDag;
+  }
+  if ( inv && (!dag) ) { 
+    _Matp = &MatpInv;
+    _Matm = &MatmInv;
+  } 
+  if ( !inv ) {
+    MooeeInternalCompute(dag,inv,Matp,Matm);
+    _Matp = &Matp;
+    _Matm = &Matm;
+  }
+  assert(_Matp->size()==Ls*LLs);
+
+  MooeeInvCalls++;
+  MooeeInvTime-=usecond();
+
+  if ( switcheroo<Coeff_t>::iscomplex() ) {
+    thread_loop( (auto site=0;site<vol;site++),{
+      MooeeInternalZAsm(psi,chi,LLs,site,*_Matp,*_Matm);
+    });
+  } else { 
+    thread_loop( (auto site=0;site<vol;site++),{
+      MooeeInternalAsm(psi,chi,LLs,site,*_Matp,*_Matm);
+    });
+  }
+  MooeeInvTime+=usecond();
+}
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h b/Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h
new file mode 100644
index 00000000..7af02263
--- /dev/null
+++ b/Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h
@@ -0,0 +1,321 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
+
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+template<class Impl>
+void ContinuedFractionFermion5D<Impl>::SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD scale)
+{
+  SetCoefficientsZolotarev(1.0/scale,zdata);
+}
+template<class Impl>
+void ContinuedFractionFermion5D<Impl>::SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata)
+{
+  // How to check Ls matches??
+  //      std::cout<<GridLogMessage << Ls << " Ls"<<std::endl;
+  //      std::cout<<GridLogMessage << zdata->n  << " - n"<<std::endl;
+  //      std::cout<<GridLogMessage << zdata->da << " -da "<<std::endl;
+  //      std::cout<<GridLogMessage << zdata->db << " -db"<<std::endl;
+  //      std::cout<<GridLogMessage << zdata->dn << " -dn"<<std::endl;
+  //      std::cout<<GridLogMessage << zdata->dd << " -dd"<<std::endl;
+  int Ls = this->Ls;
+  assert(zdata->db==Ls);// Beta has Ls coeffs
+
+  R=(1+this->mass)/(1-this->mass);
+
+  Beta.resize(Ls);
+  cc.resize(Ls);
+  cc_d.resize(Ls);
+  sqrt_cc.resize(Ls);
+  for(int i=0; i < Ls ; i++){
+    Beta[i] = zdata -> beta[i];
+    cc[i] = 1.0/Beta[i];
+    cc_d[i]=std::sqrt(cc[i]);
+  }
+    
+  cc_d[Ls-1]=1.0;
+  for(int i=0; i < Ls-1 ; i++){
+    sqrt_cc[i]= std::sqrt(cc[i]*cc[i+1]);
+  }    
+  sqrt_cc[Ls-2]=std::sqrt(cc[Ls-2]);
+
+
+  ZoloHiInv =1.0/zolo_hi;
+  dw_diag = (4.0-this->M5)*ZoloHiInv;
+    
+  See.resize(Ls);
+  Aee.resize(Ls);
+  int sign=1;
+  for(int s=0;s<Ls;s++){
+    Aee[s] = sign * Beta[s] * dw_diag;
+    sign   = - sign;
+  }
+  Aee[Ls-1] += R;
+    
+  See[0] = Aee[0];
+  for(int s=1;s<Ls;s++){
+    See[s] = Aee[s] - 1.0/See[s-1];
+  }
+  for(int s=0;s<Ls;s++){
+    std::cout<<GridLogMessage <<"s = "<<s<<" Beta "<<Beta[s]<<" Aee "<<Aee[s] <<" See "<<See[s] <<std::endl;
+  }
+}
+
+
+
+template<class Impl>
+RealD  ContinuedFractionFermion5D<Impl>::M           (const FermionField &psi, FermionField &chi)
+{
+  int Ls = this->Ls;
+
+  FermionField D(psi.Grid());
+
+  this->DW(psi,D,DaggerNo); 
+
+  int sign=1;
+  for(int s=0;s<Ls;s++){
+    if ( s==0 ) {
+      ag5xpby_ssp(chi,cc[0]*Beta[0]*sign*ZoloHiInv,D,sqrt_cc[0],psi,s,s+1); // Multiplies Dw by G5 so Hw
+    } else if ( s==(Ls-1) ){
+      RealD R=(1.0+mass)/(1.0-mass);
+      ag5xpby_ssp(chi,Beta[s]*ZoloHiInv,D,sqrt_cc[s-1],psi,s,s-1);
+      ag5xpby_ssp(chi,R,psi,1.0,chi,s,s);
+    } else {
+      ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*ZoloHiInv,D,sqrt_cc[s],psi,s,s+1);
+      axpby_ssp(chi,1.0,chi,sqrt_cc[s-1],psi,s,s-1);
+    }
+    sign=-sign; 
+  }
+  return norm2(chi);
+}
+template<class Impl>
+RealD  ContinuedFractionFermion5D<Impl>::Mdag        (const FermionField &psi, FermionField &chi)
+{
+  // This matrix is already hermitian. (g5 Dw) = Dw dag g5 = (g5 Dw)dag
+  // The rest of matrix is symmetric.
+  // Can ignore "dag"
+  return M(psi,chi);
+}
+template<class Impl>
+void  ContinuedFractionFermion5D<Impl>::Mdir (const FermionField &psi, FermionField &chi,int dir,int disp){
+  int Ls = this->Ls;
+
+  this->DhopDir(psi,chi,dir,disp); // Dslash on diagonal. g5 Dslash is hermitian
+
+  int sign=1;
+  for(int s=0;s<Ls;s++){
+    if ( s==(Ls-1) ){
+      ag5xpby_ssp(chi,Beta[s]*ZoloHiInv,chi,0.0,chi,s,s);
+    } else {
+      ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*ZoloHiInv,chi,0.0,chi,s,s);
+    }
+    sign=-sign; 
+  }
+}
+template<class Impl>
+void   ContinuedFractionFermion5D<Impl>::Meooe       (const FermionField &psi, FermionField &chi)
+{
+  int Ls = this->Ls;
+
+  // Apply 4d dslash
+  if ( psi.Checkerboard() == Odd ) {
+    this->DhopEO(psi,chi,DaggerNo); // Dslash on diagonal. g5 Dslash is hermitian
+  } else {
+    this->DhopOE(psi,chi,DaggerNo); // Dslash on diagonal. g5 Dslash is hermitian
+  }
+      
+  int sign=1;
+  for(int s=0;s<Ls;s++){
+    if ( s==(Ls-1) ){
+      ag5xpby_ssp(chi,Beta[s]*ZoloHiInv,chi,0.0,chi,s,s);
+    } else {
+      ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*ZoloHiInv,chi,0.0,chi,s,s);
+    }
+    sign=-sign; 
+  }
+}
+template<class Impl>
+void   ContinuedFractionFermion5D<Impl>::MeooeDag    (const FermionField &psi, FermionField &chi)
+{
+  this->Meooe(psi,chi);
+}
+template<class Impl>
+void   ContinuedFractionFermion5D<Impl>::Mooee       (const FermionField &psi, FermionField &chi)
+{
+  int Ls = this->Ls;
+
+  int sign=1;
+  for(int s=0;s<Ls;s++){
+    if ( s==0 ) {
+      ag5xpby_ssp(chi,cc[0]*Beta[0]*sign*dw_diag,psi,sqrt_cc[0],psi,s,s+1); // Multiplies Dw by G5 so Hw
+    } else if ( s==(Ls-1) ){
+      // Drop the CC here.
+      double R=(1+mass)/(1-mass);
+      ag5xpby_ssp(chi,Beta[s]*dw_diag,psi,sqrt_cc[s-1],psi,s,s-1);
+      ag5xpby_ssp(chi,R,psi,1.0,chi,s,s);
+    } else {
+      ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*dw_diag,psi,sqrt_cc[s],psi,s,s+1);
+      axpby_ssp(chi,1.0,chi,sqrt_cc[s-1],psi,s,s-1);
+    }
+    sign=-sign; 
+  }
+}
+
+template<class Impl>
+void   ContinuedFractionFermion5D<Impl>::MooeeDag    (const FermionField &psi, FermionField &chi)
+{
+  this->Mooee(psi,chi);
+}
+template<class Impl>
+void   ContinuedFractionFermion5D<Impl>::MooeeInv    (const FermionField &psi, FermionField &chi)
+{
+  int Ls = this->Ls;
+
+  // Apply Linv
+  axpby_ssp(chi,1.0/cc_d[0],psi,0.0,psi,0,0); 
+  for(int s=1;s<Ls;s++){
+    axpbg5y_ssp(chi,1.0/cc_d[s],psi,-1.0/See[s-1],chi,s,s-1);
+  }
+  // Apply Dinv
+  for(int s=0;s<Ls;s++){
+    ag5xpby_ssp(chi,1.0/See[s],chi,0.0,chi,s,s); //only appearance of See[0]
+  }
+  // Apply Uinv = (Linv)^T
+  axpby_ssp(chi,1.0/cc_d[Ls-1],chi,0.0,chi,Ls-1,Ls-1);
+  for(int s=Ls-2;s>=0;s--){
+    axpbg5y_ssp(chi,1.0/cc_d[s],chi,-1.0*cc_d[s+1]/See[s]/cc_d[s],chi,s,s+1);
+  }
+}
+template<class Impl>
+void   ContinuedFractionFermion5D<Impl>::MooeeInvDag (const FermionField &psi, FermionField &chi)
+{
+  this->MooeeInv(psi,chi);
+}
+
+// force terms; five routines; default to Dhop on diagonal
+template<class Impl>
+void ContinuedFractionFermion5D<Impl>::MDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
+{
+  int Ls = this->Ls;
+
+  FermionField D(V.Grid());
+
+  int sign=1;
+  for(int s=0;s<Ls;s++){
+    if ( s==(Ls-1) ){
+      ag5xpby_ssp(D,Beta[s]*ZoloHiInv,U,0.0,U,s,s);
+    } else {
+      ag5xpby_ssp(D,cc[s]*Beta[s]*sign*ZoloHiInv,U,0.0,U,s,s);
+    }
+    sign=-sign; 
+  }
+  this->DhopDeriv(mat,D,V,DaggerNo); 
+};
+template<class Impl>
+void ContinuedFractionFermion5D<Impl>::MoeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
+{
+  int Ls = this->Ls;
+
+  FermionField D(V.Grid());
+
+  int sign=1;
+  for(int s=0;s<Ls;s++){
+    if ( s==(Ls-1) ){
+      ag5xpby_ssp(D,Beta[s]*ZoloHiInv,U,0.0,U,s,s);
+    } else {
+      ag5xpby_ssp(D,cc[s]*Beta[s]*sign*ZoloHiInv,U,0.0,U,s,s);
+    }
+    sign=-sign; 
+  }
+  this->DhopDerivOE(mat,D,V,DaggerNo); 
+};
+template<class Impl>
+void ContinuedFractionFermion5D<Impl>::MeoDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
+{
+  int Ls = this->Ls;
+
+  FermionField D(V.Grid());
+
+  int sign=1;
+  for(int s=0;s<Ls;s++){
+    if ( s==(Ls-1) ){
+      ag5xpby_ssp(D,Beta[s]*ZoloHiInv,U,0.0,U,s,s);
+    } else {
+      ag5xpby_ssp(D,cc[s]*Beta[s]*sign*ZoloHiInv,U,0.0,U,s,s);
+    }
+    sign=-sign; 
+  }
+  this->DhopDerivEO(mat,D,V,DaggerNo); 
+};
+    
+// Constructors
+template<class Impl>
+ContinuedFractionFermion5D<Impl>::ContinuedFractionFermion5D(
+							     GaugeField &_Umu,
+							     GridCartesian         &FiveDimGrid,
+							     GridRedBlackCartesian &FiveDimRedBlackGrid,
+							     GridCartesian         &FourDimGrid,
+							     GridRedBlackCartesian &FourDimRedBlackGrid,
+							     RealD _mass,RealD M5,const ImplParams &p) :
+  WilsonFermion5D<Impl>(_Umu,
+			FiveDimGrid, FiveDimRedBlackGrid,
+			FourDimGrid, FourDimRedBlackGrid,M5,p),
+  mass(_mass)
+{
+  int Ls = this->Ls;
+  assert((Ls&0x1)==1); // Odd Ls required
+}
+
+    template<class Impl>
+    void ContinuedFractionFermion5D<Impl>::ExportPhysicalFermionSolution(const FermionField &solution5d,FermionField &exported4d)
+    {
+      int Ls = this->Ls;
+      conformable(solution5d.Grid(),this->FermionGrid());
+      conformable(exported4d.Grid(),this->GaugeGrid());
+      ExtractSlice(exported4d, solution5d, Ls-1, Ls-1);
+    }
+    template<class Impl>
+    void ContinuedFractionFermion5D<Impl>::ImportPhysicalFermionSource(const FermionField &input4d,FermionField &imported5d)
+    {
+      int Ls = this->Ls;
+      conformable(imported5d.Grid(),this->FermionGrid());
+      conformable(input4d.Grid()   ,this->GaugeGrid());
+      FermionField tmp(this->FermionGrid());
+      tmp=Zero();
+      InsertSlice(input4d, tmp, Ls-1, Ls-1);
+      tmp=Gamma(Gamma::Algebra::Gamma5)*tmp;
+      this->Dminus(tmp,imported5d);
+    }
+
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h b/Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h
new file mode 100644
index 00000000..a3eca650
--- /dev/null
+++ b/Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h
@@ -0,0 +1,227 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermioncache.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+
+NAMESPACE_BEGIN(Grid);
+
+// FIXME -- make a version of these routines with site loop outermost for cache reuse.
+// Pminus fowards
+// Pplus  backwards..
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::M5D(const FermionField& psi_i, const FermionField& phi_i,FermionField& chi_i, 
+				      Vector<Coeff_t>& lower, Vector<Coeff_t>& diag, Vector<Coeff_t>& upper)
+{
+  chi_i.Checkerboard() = psi_i.Checkerboard();
+  int Ls = this->Ls;
+  GridBase* grid = psi_i.Grid();
+  auto phi = phi_i.View();
+  auto psi = psi_i.View();
+  auto chi = chi_i.View();
+  assert(phi.Checkerboard() == psi.Checkerboard());
+  // Flops = 6.0*(Nc*Ns) *Ls*vol
+  this->M5Dcalls++;
+  this->M5Dtime -= usecond();
+  
+  auto nloop=grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+    auto ss=sss*Ls;
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    for(int s=0; s<Ls; s++){
+      spinor tmp1, tmp2;
+      uint64_t idx_u = ss+((s+1)%Ls);
+      uint64_t idx_l = ss+((s+Ls-1)%Ls);
+      spProj5m(tmp1, psi(idx_u));
+      spProj5p(tmp2, psi(idx_l));
+      coalescedWrite(chi[ss+s], diag[s]*phi(ss+s) + upper[s]*tmp1 + lower[s]*tmp2);
+    }
+  });
+
+  this->M5Dtime += usecond();
+}
+
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::M5Ddag(const FermionField& psi_i, const FermionField& phi_i, FermionField& chi_i, 
+					 Vector<Coeff_t>& lower, Vector<Coeff_t>& diag, Vector<Coeff_t>& upper)
+{
+  chi_i.Checkerboard() = psi_i.Checkerboard();
+  GridBase* grid = psi_i.Grid();
+  int Ls = this->Ls;
+
+  auto psi = psi_i.View();
+  auto phi = phi_i.View();
+  auto chi = chi_i.View();
+  assert(phi.Checkerboard() == psi.Checkerboard());
+
+  // Flops = 6.0*(Nc*Ns) *Ls*vol
+  this->M5Dcalls++;
+  this->M5Dtime -= usecond();
+
+  auto nloop=grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    auto ss=sss*Ls;
+    for(int s=0; s<Ls; s++){
+      spinor tmp1, tmp2;
+      uint64_t idx_u = ss+((s+1)%Ls);
+      uint64_t idx_l = ss+((s+Ls-1)%Ls);
+      spProj5p(tmp1, psi(idx_u));
+      spProj5m(tmp2, psi(idx_l));
+      coalescedWrite(chi[ss+s], diag[s]*phi(ss+s) + upper[s]*tmp1 + lower[s]*tmp2);
+    }
+  });
+
+  this->M5Dtime += usecond();
+}
+
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::MooeeInv(const FermionField& psi_i, FermionField& chi_i)
+{
+  chi_i.Checkerboard() = psi_i.Checkerboard();
+  GridBase* grid = psi_i.Grid();
+  auto psi=psi_i.View();
+  auto chi=chi_i.View();
+  int Ls = this->Ls;
+
+  auto plee  = & this->lee[0];
+  auto pdee  = & this->dee[0];
+  auto puee  = & this->uee[0];
+
+  auto pleem = & this->leem[0];
+  auto pueem = & this->ueem[0];
+
+  this->MooeeInvCalls++;
+  this->MooeeInvTime -= usecond();
+  uint64_t nloop=grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+    auto ss=sss*Ls;
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    spinor tmp1,tmp2;
+
+    // flops = 12*2*Ls + 12*2*Ls + 3*12*Ls + 12*2*Ls  = 12*Ls * (9) = 108*Ls flops
+    // Apply (L^{\prime})^{-1}
+    coalescedWrite(chi[ss],psi(ss)); // chi[0]=psi[0]
+    for(int s=1; s<Ls; s++){
+      spProj5p(tmp1, chi(ss+s-1));
+      coalescedWrite(chi[ss+s], psi(ss+s) - plee[s-1]*tmp1);
+    }
+
+    // L_m^{-1}
+    for(int s=0; s<Ls-1; s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
+      spProj5m(tmp1, chi(ss+s));
+      coalescedWrite(chi[ss+Ls-1], chi(ss+Ls-1) - pleem[s]*tmp1);
+    }
+
+    // U_m^{-1} D^{-1}
+    for(int s=0; s<Ls-1; s++){ // Chi[s] + 1/d chi[s]
+      spProj5p(tmp1, chi(ss+Ls-1));
+      coalescedWrite(chi[ss+s], (1.0/pdee[s])*chi(ss+s) - (pueem[s]/pdee[Ls])*tmp1);
+    }
+    spProj5m(tmp2, chi(ss+Ls-1));
+    coalescedWrite(chi[ss+Ls-1],(1.0/pdee[Ls])*tmp1 + (1.0/pdee[Ls-1])*tmp2);
+
+    // Apply U^{-1}
+    for(int s=Ls-2; s>=0; s--){
+      spProj5m(tmp1, chi(ss+s+1));
+      coalescedWrite(chi[ss+s], chi(ss+s) - puee[s]*tmp1);
+    }
+  });
+  this->MooeeInvTime += usecond();
+}
+
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::MooeeInvDag(const FermionField& psi_i, FermionField& chi_i)
+{
+  chi_i.Checkerboard() = psi_i.Checkerboard();
+  GridBase* grid = psi_i.Grid();
+  auto psi = psi_i.View();
+  auto chi = chi_i.View();
+  int Ls = this->Ls;
+
+  assert(psi.Checkerboard() == psi.Checkerboard());
+
+  Vector<Coeff_t> ueec(Ls);
+  Vector<Coeff_t> deec(Ls+1);
+  Vector<Coeff_t> leec(Ls);
+  Vector<Coeff_t> ueemc(Ls);
+  Vector<Coeff_t> leemc(Ls);
+
+  for(int s=0; s<ueec.size(); s++){
+    ueec[s]  = conjugate(this->uee[s]);
+    deec[s]  = conjugate(this->dee[s]);
+    leec[s]  = conjugate(this->lee[s]);
+    ueemc[s] = conjugate(this->ueem[s]);
+    leemc[s] = conjugate(this->leem[s]);
+  }
+  deec[Ls] = conjugate(this->dee[Ls]);
+
+  this->MooeeInvCalls++;
+  this->MooeeInvTime -= usecond();
+  auto nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    spinor tmp1,tmp2;
+    auto ss=sss*Ls;
+
+    // Apply (U^{\prime})^{-dagger}
+    coalescedWrite(chi[ss], psi(ss));
+    for(int s=1; s<Ls; s++){
+      spProj5m(tmp1, chi(ss+s-1));
+      coalescedWrite(chi[ss+s], psi(ss+s) - ueec[s-1]*tmp1);
+    }
+
+    // U_m^{-\dagger}
+    for(int s=0; s<Ls-1; s++){
+      spProj5p(tmp1, chi(ss+s));
+      coalescedWrite(chi[ss+Ls-1], chi(ss+Ls-1) - ueemc[s]*tmp1);
+    }
+
+    // L_m^{-\dagger} D^{-dagger}
+    for(int s=0; s<Ls-1; s++){
+      spProj5m(tmp1, chi(ss+Ls-1));
+      coalescedWrite(chi[ss+s] ,(1.0/deec[s])*chi(ss+s) - (leemc[s]/deec[Ls-1])*tmp1);
+    }
+    spProj5p(tmp2, chi(ss+Ls-1));
+    coalescedWrite(chi[ss+Ls-1], (1.0/deec[Ls-1])*tmp1 + (1.0/deec[Ls])*tmp2);
+
+    // Apply L^{-dagger}
+    for(int s=Ls-2; s>=0; s--){
+      spProj5p(tmp1, chi(ss+s+1));
+      coalescedWrite(chi[ss+s],chi(ss+s) - leec[s]*tmp1);
+    }
+  });
+
+  this->MooeeInvTime += usecond();
+}
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h b/Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h
new file mode 100644
index 00000000..3684fd6c
--- /dev/null
+++ b/Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h
@@ -0,0 +1,321 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#pragma once
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+
+NAMESPACE_BEGIN(Grid);
+
+template<class Impl>
+DomainWallEOFAFermion<Impl>::DomainWallEOFAFermion(
+						   GaugeField            &_Umu,
+						   GridCartesian         &FiveDimGrid,
+						   GridRedBlackCartesian &FiveDimRedBlackGrid,
+						   GridCartesian         &FourDimGrid,
+						   GridRedBlackCartesian &FourDimRedBlackGrid,
+						   RealD _mq1, RealD _mq2, RealD _mq3,
+						   RealD _shift, int _pm, RealD _M5, const ImplParams &p) :
+  AbstractEOFAFermion<Impl>(_Umu, FiveDimGrid, FiveDimRedBlackGrid,
+			    FourDimGrid, FourDimRedBlackGrid, _mq1, _mq2, _mq3,
+			    _shift, _pm, _M5, 1.0, 0.0, p)
+{
+  RealD eps = 1.0;
+  Approx::zolotarev_data *zdata = Approx::higham(eps,this->Ls);
+  assert(zdata->n == this->Ls);
+
+  std::cout << GridLogMessage << "DomainWallEOFAFermion with Ls=" << this->Ls << std::endl;
+  this->SetCoefficientsTanh(zdata, 1.0, 0.0);
+
+  Approx::zolotarev_free(zdata);
+}
+
+/***************************************************************
+ * Additional EOFA operators only called outside the inverter.
+ * Since speed is not essential, simple axpby-style
+ * implementations should be fine.
+ ***************************************************************/
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::Omega(const FermionField& psi, FermionField& Din, int sign, int dag)
+{
+  int Ls = this->Ls;
+
+  Din = Zero();
+  if((sign == 1) && (dag == 0)){ axpby_ssp(Din, 0.0, psi, 1.0, psi, Ls-1, 0); }
+  else if((sign == -1) && (dag == 0)){ axpby_ssp(Din, 0.0, psi, 1.0, psi, 0, 0); }
+  else if((sign == 1 ) && (dag == 1)){ axpby_ssp(Din, 0.0, psi, 1.0, psi, 0, Ls-1); }
+  else if((sign == -1) && (dag == 1)){ axpby_ssp(Din, 0.0, psi, 1.0, psi, 0, 0); }
+}
+
+// This is just the identity for DWF
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::Dtilde(const FermionField& psi, FermionField& chi){ chi = psi; }
+
+// This is just the identity for DWF
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::DtildeInv(const FermionField& psi, FermionField& chi){ chi = psi; }
+
+/*****************************************************************************************************/
+
+template<class Impl>
+RealD DomainWallEOFAFermion<Impl>::M(const FermionField& psi, FermionField& chi)
+{
+  FermionField Din(psi.Grid());
+
+  this->Meooe5D(psi, Din);
+  this->DW(Din, chi, DaggerNo);
+  axpby(chi, 1.0, 1.0, chi, psi);
+  this->M5D(psi, chi);
+  return(norm2(chi));
+}
+
+template<class Impl>
+RealD DomainWallEOFAFermion<Impl>::Mdag(const FermionField& psi, FermionField& chi)
+{
+  FermionField Din(psi.Grid());
+
+  this->DW(psi, Din, DaggerYes);
+  this->MeooeDag5D(Din, chi);
+  this->M5Ddag(psi, chi);
+  axpby(chi, 1.0, 1.0, chi, psi);
+  return(norm2(chi));
+}
+
+/********************************************************************
+ * Performance critical fermion operators called inside the inverter
+ ********************************************************************/
+
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::M5D(const FermionField& psi, FermionField& chi)
+{
+  int   Ls    = this->Ls;
+  int   pm    = this->pm;
+  RealD shift = this->shift;
+  RealD mq1   = this->mq1;
+  RealD mq2   = this->mq2;
+  RealD mq3   = this->mq3;
+
+  // coefficients for shift operator ( = shift*\gamma_{5}*R_{5}*\Delta_{\pm}(mq2,mq3)*P_{\pm} )
+  Coeff_t shiftp(0.0), shiftm(0.0);
+  if(shift != 0.0){
+    if(pm == 1){ shiftp = shift*(mq3-mq2); }
+    else{ shiftm = -shift*(mq3-mq2); }
+  }
+
+  Vector<Coeff_t> diag(Ls,1.0);
+  Vector<Coeff_t> upper(Ls,-1.0); upper[Ls-1] = mq1 + shiftm;
+  Vector<Coeff_t> lower(Ls,-1.0); lower[0]    = mq1 + shiftp;
+
+#if(0)
+  std::cout << GridLogMessage << "DomainWallEOFAFermion::M5D(FF&,FF&):" << std::endl;
+  for(int i=0; i<diag.size(); ++i){
+    std::cout << GridLogMessage << "diag[" << i << "] =" << diag[i] << std::endl;
+  }
+  for(int i=0; i<upper.size(); ++i){
+    std::cout << GridLogMessage << "upper[" << i << "] =" << upper[i] << std::endl;
+  }
+  for(int i=0; i<lower.size(); ++i){
+    std::cout << GridLogMessage << "lower[" << i << "] =" << lower[i] << std::endl;
+  }
+#endif
+
+  this->M5D(psi, chi, chi, lower, diag, upper);
+}
+
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::M5Ddag(const FermionField& psi, FermionField& chi)
+{
+  int   Ls    = this->Ls;
+  int   pm    = this->pm;
+  RealD shift = this->shift;
+  RealD mq1   = this->mq1;
+  RealD mq2   = this->mq2;
+  RealD mq3   = this->mq3;
+
+  // coefficients for shift operator ( = shift*\gamma_{5}*R_{5}*\Delta_{\pm}(mq2,mq3)*P_{\pm} )
+  Coeff_t shiftp(0.0), shiftm(0.0);
+  if(shift != 0.0){
+    if(pm == 1){ shiftp = shift*(mq3-mq2); }
+    else{ shiftm = -shift*(mq3-mq2); }
+  }
+
+  Vector<Coeff_t> diag(Ls,1.0);
+  Vector<Coeff_t> upper(Ls,-1.0); upper[Ls-1] = mq1 + shiftp;
+  Vector<Coeff_t> lower(Ls,-1.0); lower[0]    = mq1 + shiftm;
+
+  this->M5Ddag(psi, chi, chi, lower, diag, upper);
+}
+
+// half checkerboard operations
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::Mooee(const FermionField& psi, FermionField& chi)
+{
+  int Ls = this->Ls;
+
+  Vector<Coeff_t> diag = this->bee;
+  Vector<Coeff_t> upper(Ls);
+  Vector<Coeff_t> lower(Ls);
+
+  for(int s=0; s<Ls; s++){
+    upper[s] = -this->cee[s];
+    lower[s] = -this->cee[s];
+  }
+  upper[Ls-1] = this->dm;
+  lower[0]    = this->dp;
+
+  this->M5D(psi, psi, chi, lower, diag, upper);
+}
+
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::MooeeDag(const FermionField& psi, FermionField& chi)
+{
+  int Ls = this->Ls;
+
+  Vector<Coeff_t> diag = this->bee;
+  Vector<Coeff_t> upper(Ls);
+  Vector<Coeff_t> lower(Ls);
+
+  for(int s=0; s<Ls; s++){
+    upper[s] = -this->cee[s];
+    lower[s] = -this->cee[s];
+  }
+  upper[Ls-1] = this->dp;
+  lower[0]    = this->dm;
+
+  this->M5Ddag(psi, psi, chi, lower, diag, upper);
+}
+
+/****************************************************************************************/
+
+//Zolo
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::SetCoefficientsInternal(RealD zolo_hi, Vector<Coeff_t>& gamma, RealD b, RealD c)
+{
+  int   Ls    = this->Ls;
+  int   pm    = this->pm;
+  RealD mq1   = this->mq1;
+  RealD mq2   = this->mq2;
+  RealD mq3   = this->mq3;
+  RealD shift = this->shift;
+
+  ////////////////////////////////////////////////////////
+  // Constants for the preconditioned matrix Cayley form
+  ////////////////////////////////////////////////////////
+  this->bs.resize(Ls);
+  this->cs.resize(Ls);
+  this->aee.resize(Ls);
+  this->aeo.resize(Ls);
+  this->bee.resize(Ls);
+  this->beo.resize(Ls);
+  this->cee.resize(Ls);
+  this->ceo.resize(Ls);
+
+  for(int i=0; i<Ls; ++i){
+    this->bee[i] = 4.0 - this->M5 + 1.0;
+    this->cee[i] = 1.0;
+  }
+
+  for(int i=0; i<Ls; ++i){
+    this->aee[i] = this->cee[i];
+    this->bs[i] = this->beo[i] = 1.0;
+    this->cs[i] = this->ceo[i] = 0.0;
+  }
+
+  //////////////////////////////////////////
+  // EOFA shift terms
+  //////////////////////////////////////////
+  if(pm == 1){
+    this->dp = mq1*this->cee[0] + shift*(mq3-mq2);
+    this->dm = mq1*this->cee[Ls-1];
+  } else if(this->pm == -1) {
+    this->dp = mq1*this->cee[0];
+    this->dm = mq1*this->cee[Ls-1] - shift*(mq3-mq2);
+  } else {
+    this->dp = mq1*this->cee[0];
+    this->dm = mq1*this->cee[Ls-1];
+  }
+
+  //////////////////////////////////////////
+  // LDU decomposition of eeoo
+  //////////////////////////////////////////
+  this->dee.resize(Ls+1);
+  this->lee.resize(Ls);
+  this->leem.resize(Ls);
+  this->uee.resize(Ls);
+  this->ueem.resize(Ls);
+
+  for(int i=0; i<Ls; ++i){
+
+    if(i < Ls-1){
+
+      this->lee[i] = -this->cee[i+1]/this->bee[i]; // sub-diag entry on the ith column
+
+      this->leem[i] = this->dm/this->bee[i];
+      for(int j=0; j<i; j++){ this->leem[i] *= this->aee[j]/this->bee[j]; }
+
+      this->dee[i] = this->bee[i];
+
+      this->uee[i] = -this->aee[i]/this->bee[i];   // up-diag entry on the ith row
+
+      this->ueem[i] = this->dp / this->bee[0];
+      for(int j=1; j<=i; j++){ this->ueem[i] *= this->cee[j]/this->bee[j]; }
+
+    } else {
+
+      this->lee[i]  = 0.0;
+      this->leem[i] = 0.0;
+      this->uee[i]  = 0.0;
+      this->ueem[i] = 0.0;
+
+    }
+  }
+
+  {
+    Coeff_t delta_d = 1.0 / this->bee[0];
+    for(int j=1; j<Ls-1; j++){ delta_d *= this->cee[j] / this->bee[j]; }
+    this->dee[Ls-1] = this->bee[Ls-1] + this->cee[0] * this->dm * delta_d;
+    this->dee[Ls] = this->bee[Ls-1] + this->cee[Ls-1] * this->dp * delta_d;
+  }
+}
+
+// Recompute Cayley-form coefficients for different shift
+template<class Impl>
+void DomainWallEOFAFermion<Impl>::RefreshShiftCoefficients(RealD new_shift)
+{
+  this->shift = new_shift;
+  Approx::zolotarev_data *zdata = Approx::higham(1.0, this->Ls);
+  this->SetCoefficientsTanh(zdata, 1.0, 0.0);
+}
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/ImprovedStaggeredFermion5D.cc b/Grid/qcd/action/fermion/implementation/ImprovedStaggeredFermion5DImplementation.h
similarity index 79%
rename from Grid/qcd/action/fermion/ImprovedStaggeredFermion5D.cc
rename to Grid/qcd/action/fermion/implementation/ImprovedStaggeredFermion5DImplementation.h
index 0713a115..c42e896f 100644
--- a/Grid/qcd/action/fermion/ImprovedStaggeredFermion5D.cc
+++ b/Grid/qcd/action/fermion/implementation/ImprovedStaggeredFermion5DImplementation.h
@@ -24,22 +24,17 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #include <Grid/qcd/action/fermion/FermionCore.h>
 #include <Grid/qcd/action/fermion/ImprovedStaggeredFermion5D.h>
 #include <Grid/perfmon/PerfCount.h>
 
-namespace Grid {
-namespace QCD {
-  
-// S-direction is INNERMOST and takes no part in the parity.
-const std::vector<int> 
-ImprovedStaggeredFermion5DStatic::directions({1,2,3,4,1,2,3,4,1,2,3,4,1,2,3,4});
-const std::vector<int> 
-ImprovedStaggeredFermion5DStatic::displacements({1, 1, 1, 1, -1, -1, -1, -1, 3, 3, 3, 3, -3, -3, -3, -3});
+#pragma once
 
-  // 5d lattice for DWF.
+NAMESPACE_BEGIN(Grid);
+
+// 5d lattice for DWF.
 template<class Impl>
 ImprovedStaggeredFermion5D<Impl>::ImprovedStaggeredFermion5D(GridCartesian         &FiveDimGrid,
 							     GridRedBlackCartesian &FiveDimRedBlackGrid,
@@ -53,9 +48,9 @@ ImprovedStaggeredFermion5D<Impl>::ImprovedStaggeredFermion5D(GridCartesian
   _FiveDimRedBlackGrid(&FiveDimRedBlackGrid),
   _FourDimGrid        (&FourDimGrid),
   _FourDimRedBlackGrid(&FourDimRedBlackGrid),
-  Stencil    (&FiveDimGrid,npoint,Even,directions,displacements),
-  StencilEven(&FiveDimRedBlackGrid,npoint,Even,directions,displacements), // source is Even
-  StencilOdd (&FiveDimRedBlackGrid,npoint,Odd ,directions,displacements), // source is Odd
+  Stencil    (&FiveDimGrid,npoint,Even,directions,displacements,p),
+  StencilEven(&FiveDimRedBlackGrid,npoint,Even,directions,displacements,p), // source is Even
+  StencilOdd (&FiveDimRedBlackGrid,npoint,Odd ,directions,displacements,p), // source is Odd
   mass(_mass),
   c1(_c1),
   c2(_c2),
@@ -108,8 +103,8 @@ ImprovedStaggeredFermion5D<Impl>::ImprovedStaggeredFermion5D(GridCartesian
     assert(FiveDimRedBlackGrid._simd_layout[0]==nsimd);
 
     for(int d=0;d<4;d++){
-      assert(FourDimGrid._simd_layout[d]=1);
-      assert(FourDimRedBlackGrid._simd_layout[d]=1);
+      assert(FourDimGrid._simd_layout[d]==1);
+      assert(FourDimRedBlackGrid._simd_layout[d]==1);
       assert(FiveDimRedBlackGrid._simd_layout[d+1]==1);
     }
 
@@ -226,24 +221,27 @@ void ImprovedStaggeredFermion5D<Impl>::DhopDir(const FermionField &in, FermionFi
 
   Compressor compressor;
   Stencil.HaloExchange(in,compressor);
-
-  parallel_for(int ss=0;ss<Umu._grid->oSites();ss++){
+  auto Umu_v   = Umu.View();
+  auto UUUmu_v = UUUmu.View();
+  auto in_v    = in.View();
+  auto out_v   = out.View();
+  thread_for( ss,Umu.Grid()->oSites(),{
     for(int s=0;s<Ls;s++){
       int sU=ss;
       int sF = s+Ls*sU; 
-      Kernels::DhopDir(Stencil, Umu, UUUmu, Stencil.CommBuf(), sF, sU, in, out, dir, disp);
+      Kernels::DhopDirKernel(Stencil, Umu_v, UUUmu_v, Stencil.CommBuf(), sF, sU, in_v, out_v, dir, disp);
     }
-  }
+  });
 };
 
 template<class Impl>
 void ImprovedStaggeredFermion5D<Impl>::DerivInternal(StencilImpl & st,
-            DoubledGaugeField & U,
-            DoubledGaugeField & UUU,
-            GaugeField &mat,
-            const FermionField &A,
-            const FermionField &B,
-            int dag)
+						     DoubledGaugeField & U,
+						     DoubledGaugeField & UUU,
+						     GaugeField &mat,
+						     const FermionField &A,
+						     const FermionField &B,
+						     int dag)
 {
   // No force terms in multi-rhs solver staggered
   assert(0);
@@ -251,18 +249,18 @@ void ImprovedStaggeredFermion5D<Impl>::DerivInternal(StencilImpl & st,
 
 template<class Impl>
 void ImprovedStaggeredFermion5D<Impl>::DhopDeriv(GaugeField &mat,
-				      const FermionField &A,
-				      const FermionField &B,
-				      int dag)
+						 const FermionField &A,
+						 const FermionField &B,
+						 int dag)
 {
   assert(0);
 }
 
 template<class Impl>
 void ImprovedStaggeredFermion5D<Impl>::DhopDerivEO(GaugeField &mat,
-					const FermionField &A,
-					const FermionField &B,
-					int dag)
+						   const FermionField &A,
+						   const FermionField &B,
+						   int dag)
 {
   assert(0);
 }
@@ -270,9 +268,9 @@ void ImprovedStaggeredFermion5D<Impl>::DhopDerivEO(GaugeField &mat,
 
 template<class Impl>
 void ImprovedStaggeredFermion5D<Impl>::DhopDerivOE(GaugeField &mat,
-					const FermionField &A,
-					const FermionField &B,
-					int dag)
+						   const FermionField &A,
+						   const FermionField &B,
+						   int dag)
 {
   assert(0);
 }
@@ -301,8 +299,8 @@ void ImprovedStaggeredFermion5D<Impl>::DhopInternalOverlappedComms(StencilImpl &
 
   Compressor compressor; 
 
-  int LLs = in._grid->_rdimensions[0];
-  int len =  U._grid->oSites();
+  int LLs = in.Grid()->_rdimensions[0];
+  int len =  U.Grid()->oSites();
 
   DhopFaceTime-=usecond();
   st.Prepare();
@@ -328,7 +326,7 @@ void ImprovedStaggeredFermion5D<Impl>::DhopInternalOverlappedComms(StencilImpl &
       double start = usecond();
       nthreads -= ncomms;
       int ttid  = tid - ncomms;
-      int n     = U._grid->oSites(); // 4d vol
+      int n     = U.Grid()->oSites(); // 4d vol
       int chunk = n / nthreads;
       int rem   = n % nthreads;
       int myblock, myn;
@@ -341,17 +339,22 @@ void ImprovedStaggeredFermion5D<Impl>::DhopInternalOverlappedComms(StencilImpl &
       }
 
       // do the compute
+      auto   U_v  =   U.View();
+      auto UUU_v  = UUU.View();
+      auto  in_v  =  in.View();
+      auto out_v  = out.View();
+
       if (dag == DaggerYes) {
         for (int ss = myblock; ss < myblock+myn; ++ss) {
           int sU = ss;
 	  // Interior = 1; Exterior = 0; must implement for staggered
-          Kernels::DhopSiteDag(st,lo,U,UUU,st.CommBuf(),LLs,sU,in,out,1,0); //<---------
+          Kernels::DhopSiteDag(st,lo,U_v,UUU_v,st.CommBuf(),LLs,sU,in_v,out_v,1,0); //<---------
         }
       } else {
         for (int ss = myblock; ss < myblock+myn; ++ss) {
 	  // Interior = 1; Exterior = 0;
           int sU = ss;
-          Kernels::DhopSite(st,lo,U,UUU,st.CommBuf(),LLs,sU,in,out,1,0); //<------------
+          Kernels::DhopSite(st,lo,U_v,UUU_v,st.CommBuf(),LLs,sU,in_v,out_v,1,0); //<------------
         }
       }
         ptime = usecond() - start;
@@ -372,18 +375,23 @@ void ImprovedStaggeredFermion5D<Impl>::DhopInternalOverlappedComms(StencilImpl &
   DhopFaceTime+=usecond();
 
   DhopComputeTime2-=usecond();
+
+  auto   U_v  =   U.View();
+  auto UUU_v  = UUU.View();
+  auto  in_v  =  in.View();
+  auto out_v  = out.View();
   if (dag == DaggerYes) {
     int sz=st.surface_list.size();
-    parallel_for (int ss = 0; ss < sz; ss++) {
+    thread_for( ss,sz,{
       int sU = st.surface_list[ss];
-      Kernels::DhopSiteDag(st,lo,U,UUU,st.CommBuf(),LLs,sU,in,out,0,1); //<----------
-    }
+      Kernels::DhopSiteDag(st,lo,U_v,UUU_v,st.CommBuf(),LLs,sU,in_v,out_v,0,1); //<----------
+    });
   } else {
     int sz=st.surface_list.size();
-    parallel_for (int ss = 0; ss < sz; ss++) {
+    thread_for( ss,sz,{
       int sU = st.surface_list[ss];
-      Kernels::DhopSite(st,lo,U,UUU,st.CommBuf(),LLs,sU,in,out,0,1);//<----------
-    }
+      Kernels::DhopSite(st,lo,U_v,UUU_v,st.CommBuf(),LLs,sU,in_v,out_v,0,1);//<----------
+    });
   }
   DhopComputeTime2+=usecond();
 #else
@@ -398,7 +406,7 @@ void ImprovedStaggeredFermion5D<Impl>::DhopInternalSerialComms(StencilImpl & st,
 						    const FermionField &in, FermionField &out,int dag)
 {
   Compressor compressor;
-  int LLs = in._grid->_rdimensions[0];
+  int LLs = in.Grid()->_rdimensions[0];
 
 
 
@@ -410,16 +418,20 @@ void ImprovedStaggeredFermion5D<Impl>::DhopInternalSerialComms(StencilImpl & st,
   
   DhopComputeTime -= usecond();
   // Dhop takes the 4d grid from U, and makes a 5d index for fermion
+  auto   U_v  =   U.View();
+  auto UUU_v  = UUU.View();
+  auto  in_v  =  in.View();
+  auto out_v  = out.View();
   if (dag == DaggerYes) {
-    parallel_for (int ss = 0; ss < U._grid->oSites(); ss++) {
+    thread_for( ss,U.Grid()->oSites(),{
       int sU=ss;
-      Kernels::DhopSiteDag(st, lo, U, UUU, st.CommBuf(), LLs, sU,in, out);
-    }
+      Kernels::DhopSiteDag(st, lo, U_v, UUU_v, st.CommBuf(), LLs, sU,in_v, out_v);
+    });
   } else {
-    parallel_for (int ss = 0; ss < U._grid->oSites(); ss++) {
+    thread_for( ss,U.Grid()->oSites(),{
       int sU=ss;
-      Kernels::DhopSite(st,lo,U,UUU,st.CommBuf(),LLs,sU,in,out);
-    }
+      Kernels::DhopSite(st,lo,U_v,UUU_v,st.CommBuf(),LLs,sU,in_v,out_v);
+    });
   }
   DhopComputeTime += usecond();
   DhopTotalTime   += usecond();
@@ -432,50 +444,17 @@ void ImprovedStaggeredFermion5D<Impl>::DhopInternalSerialComms(StencilImpl & st,
 }
 /*CHANGE END*/
 
-/* ORG
-template<class Impl>
-void ImprovedStaggeredFermion5D<Impl>::DhopInternal(StencilImpl & st, LebesgueOrder &lo,
-						    DoubledGaugeField & U,DoubledGaugeField & UUU,
-						    const FermionField &in, FermionField &out,int dag)
-{
-  Compressor compressor;
-  int LLs = in._grid->_rdimensions[0];
-
-
-
-  DhopTotalTime -= usecond();
-  DhopCommTime -= usecond();
-  st.HaloExchange(in,compressor);
-  DhopCommTime += usecond();
-  
-  DhopComputeTime -= usecond();
-  // Dhop takes the 4d grid from U, and makes a 5d index for fermion
-  if (dag == DaggerYes) {
-    parallel_for (int ss = 0; ss < U._grid->oSites(); ss++) {
-      int sU=ss;
-      Kernels::DhopSiteDag(st, lo, U, UUU, st.CommBuf(), LLs, sU,in, out);
-    }
-  } else {
-    parallel_for (int ss = 0; ss < U._grid->oSites(); ss++) {
-      int sU=ss;
-	Kernels::DhopSite(st,lo,U,UUU,st.CommBuf(),LLs,sU,in,out);
-    }
-  }
-  DhopComputeTime += usecond();
-  DhopTotalTime   += usecond();
-}
-*/
 
 
 template<class Impl>
 void ImprovedStaggeredFermion5D<Impl>::DhopOE(const FermionField &in, FermionField &out,int dag)
 {
   DhopCalls+=1;
-  conformable(in._grid,FermionRedBlackGrid());    // verifies half grid
-  conformable(in._grid,out._grid); // drops the cb check
+  conformable(in.Grid(),FermionRedBlackGrid());    // verifies half grid
+  conformable(in.Grid(),out.Grid()); // drops the cb check
 
-  assert(in.checkerboard==Even);
-  out.checkerboard = Odd;
+  assert(in.Checkerboard()==Even);
+  out.Checkerboard() = Odd;
 
   DhopInternal(StencilEven,LebesgueEvenOdd,UmuOdd,UUUmuOdd,in,out,dag);
 }
@@ -483,11 +462,11 @@ template<class Impl>
 void ImprovedStaggeredFermion5D<Impl>::DhopEO(const FermionField &in, FermionField &out,int dag)
 {
   DhopCalls+=1;
-  conformable(in._grid,FermionRedBlackGrid());    // verifies half grid
-  conformable(in._grid,out._grid); // drops the cb check
+  conformable(in.Grid(),FermionRedBlackGrid());    // verifies half grid
+  conformable(in.Grid(),out.Grid()); // drops the cb check
 
-  assert(in.checkerboard==Odd);
-  out.checkerboard = Even;
+  assert(in.Checkerboard()==Odd);
+  out.Checkerboard() = Even;
 
   DhopInternal(StencilOdd,LebesgueEvenOdd,UmuEven,UUUmuEven,in,out,dag);
 }
@@ -495,10 +474,10 @@ template<class Impl>
 void ImprovedStaggeredFermion5D<Impl>::Dhop(const FermionField &in, FermionField &out,int dag)
 {
   DhopCalls+=2;
-  conformable(in._grid,FermionGrid()); // verifies full grid
-  conformable(in._grid,out._grid);
+  conformable(in.Grid(),FermionGrid()); // verifies full grid
+  conformable(in.Grid(),out.Grid());
 
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
 
   DhopInternal(Stencil,Lebesgue,Umu,UUUmu,in,out,dag);
 }
@@ -506,7 +485,7 @@ void ImprovedStaggeredFermion5D<Impl>::Dhop(const FermionField &in, FermionField
 template<class Impl>
 void ImprovedStaggeredFermion5D<Impl>::Report(void) 
 {
-  std::vector<int> latt = GridDefaultLatt();          
+  Coordinate latt = GridDefaultLatt();          
   RealD volume = Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt[mu];
   RealD NP = _FourDimGrid->_Nprocessors;
   RealD NN = _FourDimGrid->NodeCount();
@@ -564,21 +543,21 @@ void ImprovedStaggeredFermion5D<Impl>::Mdir(const FermionField &in, FermionField
 }
 template <class Impl>
 RealD ImprovedStaggeredFermion5D<Impl>::M(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   Dhop(in, out, DaggerNo);
   return axpy_norm(out, mass, in, out);
 }
 
 template <class Impl>
 RealD ImprovedStaggeredFermion5D<Impl>::Mdag(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   Dhop(in, out, DaggerYes);
   return axpy_norm(out, mass, in, out);
 }
 
 template <class Impl>
 void ImprovedStaggeredFermion5D<Impl>::Meooe(const FermionField &in, FermionField &out) {
-  if (in.checkerboard == Odd) {
+  if (in.Checkerboard() == Odd) {
     DhopEO(in, out, DaggerNo);
   } else {
     DhopOE(in, out, DaggerNo);
@@ -586,7 +565,7 @@ void ImprovedStaggeredFermion5D<Impl>::Meooe(const FermionField &in, FermionFiel
 }
 template <class Impl>
 void ImprovedStaggeredFermion5D<Impl>::MeooeDag(const FermionField &in, FermionField &out) {
-  if (in.checkerboard == Odd) {
+  if (in.Checkerboard() == Odd) {
     DhopEO(in, out, DaggerYes);
   } else {
     DhopOE(in, out, DaggerYes);
@@ -595,27 +574,27 @@ void ImprovedStaggeredFermion5D<Impl>::MeooeDag(const FermionField &in, FermionF
 
 template <class Impl>
 void ImprovedStaggeredFermion5D<Impl>::Mooee(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   typename FermionField::scalar_type scal(mass);
   out = scal * in;
 }
 
 template <class Impl>
 void ImprovedStaggeredFermion5D<Impl>::MooeeDag(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   Mooee(in, out);
 }
 
 template <class Impl>
 void ImprovedStaggeredFermion5D<Impl>::MooeeInv(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   out = (1.0 / (mass)) * in;
 }
 
 template <class Impl>
 void ImprovedStaggeredFermion5D<Impl>::MooeeInvDag(const FermionField &in,
-                                      FermionField &out) {
-  out.checkerboard = in.checkerboard;
+						   FermionField &out) {
+  out.Checkerboard() = in.Checkerboard();
   MooeeInv(in, out);
 }
 
@@ -624,31 +603,28 @@ void ImprovedStaggeredFermion5D<Impl>::MooeeInvDag(const FermionField &in,
 ////////////////////////////////////////////////////////
 template <class Impl>
 void ImprovedStaggeredFermion5D<Impl>::ContractConservedCurrent(PropagatorField &q_in_1,
-                                                         PropagatorField &q_in_2,
-                                                         PropagatorField &q_out,
-                                                         Current curr_type,
-                                                         unsigned int mu)
+								PropagatorField &q_in_2,
+								PropagatorField &q_out,
+								Current curr_type,
+								unsigned int mu)
 {
-    assert(0);
+  assert(0);
 }
 
 template <class Impl>
-void ImprovedStaggeredFermion5D<Impl>::SeqConservedCurrent(PropagatorField &q_in, 
-                                              PropagatorField &q_out,
-                                              Current curr_type,
-                                              unsigned int mu,
-                                              unsigned int tmin, 
+void ImprovedStaggeredFermion5D<Impl>::SeqConservedCurrent(PropagatorField &q_in,
+							   PropagatorField &q_out,
+							   Current curr_type,
+							   unsigned int mu, 
+							   unsigned int tmin,
                                               unsigned int tmax,
 					      ComplexField &lattice_cmplx)
 {
-    assert(0);
+  assert(0);
 
 }
-
-FermOpStaggeredTemplateInstantiate(ImprovedStaggeredFermion5D);
-FermOpStaggeredVec5dTemplateInstantiate(ImprovedStaggeredFermion5D);
   
-}}
+NAMESPACE_END(Grid);
 
 
 
diff --git a/Grid/qcd/action/fermion/ImprovedStaggeredFermion.cc b/Grid/qcd/action/fermion/implementation/ImprovedStaggeredFermionImplementation.h
similarity index 75%
rename from Grid/qcd/action/fermion/ImprovedStaggeredFermion.cc
rename to Grid/qcd/action/fermion/implementation/ImprovedStaggeredFermionImplementation.h
index 883db902..e2605d81 100644
--- a/Grid/qcd/action/fermion/ImprovedStaggeredFermion.cc
+++ b/Grid/qcd/action/fermion/implementation/ImprovedStaggeredFermionImplementation.h
@@ -28,40 +28,35 @@ directory
 /*  END LEGAL */
 #include <Grid/Grid.h>
 
-namespace Grid {
-namespace QCD {
+#pragma once 
 
-const std::vector<int> 
-ImprovedStaggeredFermionStatic::directions({0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3});
-const std::vector<int> 
-ImprovedStaggeredFermionStatic::displacements({1, 1, 1, 1, -1, -1, -1, -1, 3, 3, 3, 3, -3, -3, -3, -3});
+NAMESPACE_BEGIN(Grid);
 
 /////////////////////////////////
 // Constructor and gauge import
 /////////////////////////////////
 
-
 template <class Impl>
 ImprovedStaggeredFermion<Impl>::ImprovedStaggeredFermion(GridCartesian &Fgrid, GridRedBlackCartesian &Hgrid, 
 							 RealD _mass,
 							 RealD _c1, RealD _c2,RealD _u0,
 							 const ImplParams &p)
-    : Kernels(p),
-      _grid(&Fgrid),
-      _cbgrid(&Hgrid),
-      Stencil(&Fgrid, npoint, Even, directions, displacements),
-      StencilEven(&Hgrid, npoint, Even, directions, displacements),  // source is Even
-      StencilOdd(&Hgrid, npoint, Odd, directions, displacements),  // source is Odd
-      mass(_mass),
-      Lebesgue(_grid),
-      LebesgueEvenOdd(_cbgrid),
-      Umu(&Fgrid),
-      UmuEven(&Hgrid),
-      UmuOdd(&Hgrid),
-      UUUmu(&Fgrid),
-      UUUmuEven(&Hgrid),
-      UUUmuOdd(&Hgrid) ,
-      _tmp(&Hgrid)
+  : Kernels(p),
+    _grid(&Fgrid),
+    _cbgrid(&Hgrid),
+    Stencil(&Fgrid, npoint, Even, directions, displacements,p),
+    StencilEven(&Hgrid, npoint, Even, directions, displacements,p),  // source is Even
+    StencilOdd(&Hgrid, npoint, Odd, directions, displacements,p),  // source is Odd
+    mass(_mass),
+    Lebesgue(_grid),
+    LebesgueEvenOdd(_cbgrid),
+    Umu(&Fgrid),
+    UmuEven(&Hgrid),
+    UmuOdd(&Hgrid),
+    UUUmu(&Fgrid),
+    UUUmuEven(&Hgrid),
+    UUUmuOdd(&Hgrid) ,
+    _tmp(&Hgrid)
 {
   int vol4;
   int LLs=1;
@@ -85,17 +80,17 @@ ImprovedStaggeredFermion<Impl>::ImprovedStaggeredFermion(GaugeField &_Uthin, Gau
   ImportGauge(_Uthin,_Ufat);
 }
 
-  ////////////////////////////////////////////////////////////
-  // Momentum space propagator should be 
-  // https://arxiv.org/pdf/hep-lat/9712010.pdf
-  //
-  // mom space action.
-  //   gamma_mu i ( c1 sin pmu + c2 sin 3 pmu ) + m
-  //
-  // must track through staggered flavour/spin reduction in literature to 
-  // turn to free propagator for the one component chi field, a la page 4/5
-  // of above link to implmement fourier based solver.
-  ////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////
+// Momentum space propagator should be 
+// https://arxiv.org/pdf/hep-lat/9712010.pdf
+//
+// mom space action.
+//   gamma_mu i ( c1 sin pmu + c2 sin 3 pmu ) + m
+//
+// must track through staggered flavour/spin reduction in literature to 
+// turn to free propagator for the one component chi field, a la page 4/5
+// of above link to implmement fourier based solver.
+////////////////////////////////////////////////////////////
 template <class Impl>
 void ImprovedStaggeredFermion<Impl>::ImportGaugeSimple(const GaugeField &_Utriple,const GaugeField &_Ufat) 
 {
@@ -177,21 +172,21 @@ void ImprovedStaggeredFermion<Impl>::ImportGauge(const GaugeField &_Uthin,const
 
 template <class Impl>
 RealD ImprovedStaggeredFermion<Impl>::M(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   Dhop(in, out, DaggerNo);
   return axpy_norm(out, mass, in, out);
 }
 
 template <class Impl>
 RealD ImprovedStaggeredFermion<Impl>::Mdag(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   Dhop(in, out, DaggerYes);
   return axpy_norm(out, mass, in, out);
 }
 
 template <class Impl>
 void ImprovedStaggeredFermion<Impl>::Meooe(const FermionField &in, FermionField &out) {
-  if (in.checkerboard == Odd) {
+  if (in.Checkerboard() == Odd) {
     DhopEO(in, out, DaggerNo);
   } else {
     DhopOE(in, out, DaggerNo);
@@ -199,7 +194,7 @@ void ImprovedStaggeredFermion<Impl>::Meooe(const FermionField &in, FermionField
 }
 template <class Impl>
 void ImprovedStaggeredFermion<Impl>::MeooeDag(const FermionField &in, FermionField &out) {
-  if (in.checkerboard == Odd) {
+  if (in.Checkerboard() == Odd) {
     DhopEO(in, out, DaggerYes);
   } else {
     DhopOE(in, out, DaggerYes);
@@ -208,27 +203,27 @@ void ImprovedStaggeredFermion<Impl>::MeooeDag(const FermionField &in, FermionFie
 
 template <class Impl>
 void ImprovedStaggeredFermion<Impl>::Mooee(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   typename FermionField::scalar_type scal(mass);
   out = scal * in;
 }
 
 template <class Impl>
 void ImprovedStaggeredFermion<Impl>::MooeeDag(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   Mooee(in, out);
 }
 
 template <class Impl>
 void ImprovedStaggeredFermion<Impl>::MooeeInv(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   out = (1.0 / (mass)) * in;
 }
 
 template <class Impl>
 void ImprovedStaggeredFermion<Impl>::MooeeInvDag(const FermionField &in,
-                                      FermionField &out) {
-  out.checkerboard = in.checkerboard;
+						 FermionField &out) {
+  out.Checkerboard() = in.Checkerboard();
   MooeeInv(in, out);
 }
 
@@ -244,8 +239,8 @@ void ImprovedStaggeredFermion<Impl>::DerivInternal(StencilImpl &st, DoubledGauge
 
   Compressor compressor;
 
-  FermionField Btilde(B._grid);
-  FermionField Atilde(B._grid);
+  FermionField Btilde(B.Grid());
+  FermionField Atilde(B.Grid());
   Atilde = A;
 
   st.HaloExchange(B, compressor);
@@ -255,10 +250,13 @@ void ImprovedStaggeredFermion<Impl>::DerivInternal(StencilImpl &st, DoubledGauge
     ////////////////////////
     // Call the single hop
     ////////////////////////
-    PARALLEL_FOR_LOOP
-    for (int sss = 0; sss < B._grid->oSites(); sss++) {
-      Kernels::DhopDir(st, U, UUU, st.CommBuf(), sss, sss, B, Btilde, mu,1);
-    }
+    auto U_v   = U.View();
+    auto UUU_v = UUU.View();
+    auto B_v   = B.View();
+    auto Btilde_v   = Btilde.View();
+    thread_for(sss,B.Grid()->oSites(),{
+      Kernels::DhopDirKernel(st, U_v, UUU_v, st.CommBuf(), sss, sss, B_v, Btilde_v, mu,1);
+    });
 
     // Force in three link terms
     //
@@ -288,11 +286,11 @@ void ImprovedStaggeredFermion<Impl>::DerivInternal(StencilImpl &st, DoubledGauge
 template <class Impl>
 void ImprovedStaggeredFermion<Impl>::DhopDeriv(GaugeField &mat, const FermionField &U, const FermionField &V, int dag) {
 
-  conformable(U._grid, _grid);
-  conformable(U._grid, V._grid);
-  conformable(U._grid, mat._grid);
+  conformable(U.Grid(), _grid);
+  conformable(U.Grid(), V.Grid());
+  conformable(U.Grid(), mat.Grid());
 
-  mat.checkerboard = U.checkerboard;
+  mat.Checkerboard() = U.Checkerboard();
 
   DerivInternal(Stencil, Umu, UUUmu, mat, U, V, dag);
 }
@@ -300,13 +298,13 @@ void ImprovedStaggeredFermion<Impl>::DhopDeriv(GaugeField &mat, const FermionFie
 template <class Impl>
 void ImprovedStaggeredFermion<Impl>::DhopDerivOE(GaugeField &mat, const FermionField &U, const FermionField &V, int dag) {
 
-  conformable(U._grid, _cbgrid);
-  conformable(U._grid, V._grid);
-  conformable(U._grid, mat._grid);
+  conformable(U.Grid(), _cbgrid);
+  conformable(U.Grid(), V.Grid());
+  conformable(U.Grid(), mat.Grid());
 
-  assert(V.checkerboard == Even);
-  assert(U.checkerboard == Odd);
-  mat.checkerboard = Odd;
+  assert(V.Checkerboard() == Even);
+  assert(U.Checkerboard() == Odd);
+  mat.Checkerboard() = Odd;
 
   DerivInternal(StencilEven, UmuOdd, UUUmuOdd, mat, U, V, dag);
 }
@@ -314,48 +312,51 @@ void ImprovedStaggeredFermion<Impl>::DhopDerivOE(GaugeField &mat, const FermionF
 template <class Impl>
 void ImprovedStaggeredFermion<Impl>::DhopDerivEO(GaugeField &mat, const FermionField &U, const FermionField &V, int dag) {
 
-  conformable(U._grid, _cbgrid);
-  conformable(U._grid, V._grid);
-  conformable(U._grid, mat._grid);
+  conformable(U.Grid(), _cbgrid);
+  conformable(U.Grid(), V.Grid());
+  conformable(U.Grid(), mat.Grid());
 
-  assert(V.checkerboard == Odd);
-  assert(U.checkerboard == Even);
-  mat.checkerboard = Even;
+  assert(V.Checkerboard() == Odd);
+  assert(U.Checkerboard() == Even);
+  mat.Checkerboard() = Even;
 
   DerivInternal(StencilOdd, UmuEven, UUUmuEven, mat, U, V, dag);
 }
 
 template <class Impl>
-void ImprovedStaggeredFermion<Impl>::Dhop(const FermionField &in, FermionField &out, int dag) {
+void ImprovedStaggeredFermion<Impl>::Dhop(const FermionField &in, FermionField &out, int dag) 
+{
   DhopCalls+=2;
-  conformable(in._grid, _grid);  // verifies full grid
-  conformable(in._grid, out._grid);
+  conformable(in.Grid(), _grid);  // verifies full grid
+  conformable(in.Grid(), out.Grid());
 
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
 
   DhopInternal(Stencil, Lebesgue, Umu, UUUmu, in, out, dag);
 }
 
 template <class Impl>
-void ImprovedStaggeredFermion<Impl>::DhopOE(const FermionField &in, FermionField &out, int dag) {
+void ImprovedStaggeredFermion<Impl>::DhopOE(const FermionField &in, FermionField &out, int dag) 
+{
   DhopCalls+=1;
-  conformable(in._grid, _cbgrid);    // verifies half grid
-  conformable(in._grid, out._grid);  // drops the cb check
+  conformable(in.Grid(), _cbgrid);    // verifies half grid
+  conformable(in.Grid(), out.Grid());  // drops the cb check
 
-  assert(in.checkerboard == Even);
-  out.checkerboard = Odd;
+  assert(in.Checkerboard() == Even);
+  out.Checkerboard() = Odd;
 
   DhopInternal(StencilEven, LebesgueEvenOdd, UmuOdd, UUUmuOdd, in, out, dag);
 }
 
 template <class Impl>
-void ImprovedStaggeredFermion<Impl>::DhopEO(const FermionField &in, FermionField &out, int dag) {
+void ImprovedStaggeredFermion<Impl>::DhopEO(const FermionField &in, FermionField &out, int dag) 
+{
   DhopCalls+=1;
-  conformable(in._grid, _cbgrid);    // verifies half grid
-  conformable(in._grid, out._grid);  // drops the cb check
+  conformable(in.Grid(), _cbgrid);    // verifies half grid
+  conformable(in.Grid(), out.Grid());  // drops the cb check
 
-  assert(in.checkerboard == Odd);
-  out.checkerboard = Even;
+  assert(in.Checkerboard() == Odd);
+  out.Checkerboard() = Even;
 
   DhopInternal(StencilOdd, LebesgueEvenOdd, UmuEven, UUUmuEven, in, out, dag);
 }
@@ -370,11 +371,13 @@ void ImprovedStaggeredFermion<Impl>::DhopDir(const FermionField &in, FermionFiel
 
   Compressor compressor;
   Stencil.HaloExchange(in, compressor);
-
-  PARALLEL_FOR_LOOP
-  for (int sss = 0; sss < in._grid->oSites(); sss++) {
-    Kernels::DhopDir(Stencil, Umu, UUUmu, Stencil.CommBuf(), sss, sss, in, out, dir, disp);
-  }
+  auto Umu_v   =   Umu.View();
+  auto UUUmu_v = UUUmu.View();
+  auto in_v    =  in.View();
+  auto out_v   = out.View();
+  thread_for( sss, in.Grid()->oSites(),{
+    Kernels::DhopDirKernel(Stencil, Umu_v, UUUmu_v, Stencil.CommBuf(), sss, sss, in_v, out_v, dir, disp);
+  });
 };
 
 template <class Impl>
@@ -400,7 +403,7 @@ void ImprovedStaggeredFermion<Impl>::DhopInternalOverlappedComms(StencilImpl &st
 {
 #ifdef GRID_OMP
   Compressor compressor; 
-  int len =  U._grid->oSites();
+  int len =  U.Grid()->oSites();
   const int LLs =  1;
 
   DhopTotalTime   -= usecond();
@@ -439,17 +442,21 @@ void ImprovedStaggeredFermion<Impl>::DhopInternalOverlappedComms(StencilImpl &st
       }
 
       // do the compute
+      auto U_v   = U.View();
+      auto UUU_v = UUU.View();
+      auto in_v  = in.View();
+      auto out_v = out.View();
       if (dag == DaggerYes) {
         for (int ss = myblock; ss < myblock+myn; ++ss) {
           int sU = ss;
 	  // Interior = 1; Exterior = 0; must implement for staggered
-          Kernels::DhopSiteDag(st,lo,U,UUU,st.CommBuf(),1,sU,in,out,1,0); 
+          Kernels::DhopSiteDag(st,lo,U_v,UUU_v,st.CommBuf(),1,sU,in_v,out_v,1,0); 
         }
       } else {
         for (int ss = myblock; ss < myblock+myn; ++ss) {
 	  // Interior = 1; Exterior = 0;
           int sU = ss;
-          Kernels::DhopSite(st,lo,U,UUU,st.CommBuf(),1,sU,in,out,1,0);
+          Kernels::DhopSite(st,lo,U_v,UUU_v,st.CommBuf(),1,sU,in_v,out_v,1,0);
         }
       }
     } else {
@@ -464,17 +471,23 @@ void ImprovedStaggeredFermion<Impl>::DhopInternalOverlappedComms(StencilImpl &st
   DhopFaceTime    -= usecond();
 
   DhopComputeTime2    -= usecond();
-  if (dag == DaggerYes) {
-    int sz=st.surface_list.size();
-    parallel_for (int ss = 0; ss < sz; ss++) {
-      int sU = st.surface_list[ss];
-      Kernels::DhopSiteDag(st,lo,U,UUU,st.CommBuf(),1,sU,in,out,0,1);
-    }
-  } else {
-    int sz=st.surface_list.size();
-    parallel_for (int ss = 0; ss < sz; ss++) {
-      int sU = st.surface_list[ss];
-      Kernels::DhopSite(st,lo,U,UUU,st.CommBuf(),1,sU,in,out,0,1);
+  {
+    auto U_v   = U.View();
+    auto UUU_v = UUU.View();
+    auto in_v  = in.View();
+    auto out_v = out.View();
+    if (dag == DaggerYes) {
+      int sz=st.surface_list.size();
+      thread_for(ss,sz,{
+	int sU = st.surface_list[ss];
+	Kernels::DhopSiteDag(st,lo,U_v,UUU_v,st.CommBuf(),1,sU,in_v,out_v,0,1);
+      });
+    } else {
+      int sz=st.surface_list.size();
+      thread_for(ss,sz,{
+	int sU = st.surface_list[ss];
+	Kernels::DhopSite(st,lo,U_v,UUU_v,st.CommBuf(),1,sU,in_v,out_v,0,1);
+      });
     }
   }
   DhopComputeTime2    += usecond();
@@ -500,15 +513,19 @@ void ImprovedStaggeredFermion<Impl>::DhopInternalSerialComms(StencilImpl &st, Le
   st.HaloExchange(in, compressor);
   DhopCommTime    += usecond();
 
+  auto U_v   =   U.View();
+  auto UUU_v = UUU.View();
+  auto in_v  =  in.View();
+  auto out_v = out.View();
   DhopComputeTime -= usecond();
   if (dag == DaggerYes) {
-    parallel_for (int sss = 0; sss < in._grid->oSites(); sss++) {
-      Kernels::DhopSiteDag(st, lo, U, UUU, st.CommBuf(), 1, sss, in, out);
-    }
+    thread_for(sss, in.Grid()->oSites(),{
+      Kernels::DhopSiteDag(st, lo, U_v, UUU_v, st.CommBuf(), 1, sss, in_v, out_v);
+    });
   } else {
-    parallel_for (int sss = 0; sss < in._grid->oSites(); sss++) {
-      Kernels::DhopSite(st, lo, U, UUU, st.CommBuf(), 1, sss, in, out);
-    }
+    thread_for(sss, in.Grid()->oSites(),{
+      Kernels::DhopSite(st, lo, U_v, UUU_v, st.CommBuf(), 1, sss, in_v, out_v);
+    });
   }
   DhopComputeTime += usecond();
   DhopTotalTime   += usecond();
@@ -520,7 +537,7 @@ void ImprovedStaggeredFermion<Impl>::DhopInternalSerialComms(StencilImpl &st, Le
 template<class Impl>
 void ImprovedStaggeredFermion<Impl>::Report(void) 
 {
-  std::vector<int> latt = GridDefaultLatt();          
+  Coordinate latt = _grid->GlobalDimensions();
   RealD volume = 1;  for(int mu=0;mu<Nd;mu++) volume=volume*latt[mu];
   RealD NP = _grid->_Nprocessors;
   RealD NN = _grid->NodeCount();
@@ -574,31 +591,25 @@ void ImprovedStaggeredFermion<Impl>::ZeroCounters(void)
 ////////////////////////////////////////////////////////
 template <class Impl>
 void ImprovedStaggeredFermion<Impl>::ContractConservedCurrent(PropagatorField &q_in_1,
-                                                        PropagatorField &q_in_2,
-                                                        PropagatorField &q_out,
-                                                        Current curr_type,
-                                                        unsigned int mu)
+							      PropagatorField &q_in_2,
+							      PropagatorField &q_out,
+							      Current curr_type,
+							      unsigned int mu)
 {
-    assert(0);
+  assert(0);
 }
 
 template <class Impl>
-void ImprovedStaggeredFermion<Impl>::SeqConservedCurrent(PropagatorField &q_in, 
-                                              PropagatorField &q_out,
-                                              Current curr_type,
-                                              unsigned int mu,
-                                              unsigned int tmin, 
+void ImprovedStaggeredFermion<Impl>::SeqConservedCurrent(PropagatorField &q_in,
+                                                         PropagatorField &q_out,
+                                                         Current curr_type,
+                                                         unsigned int mu, 
+                                                         unsigned int tmin,
                                               unsigned int tmax,
 					      ComplexField &lattice_cmplx)
 {
-    assert(0);
+  assert(0);
 
 }
 
-
-FermOpStaggeredTemplateInstantiate(ImprovedStaggeredFermion);
-
-  //AdjointFermOpTemplateInstantiate(ImprovedStaggeredFermion);
-  //TwoIndexFermOpTemplateInstantiate(ImprovedStaggeredFermion);
-
-}}
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h b/Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h
new file mode 100644
index 00000000..650435fc
--- /dev/null
+++ b/Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h
@@ -0,0 +1,453 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermioncache.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+
+NAMESPACE_BEGIN(Grid);
+
+ 
+template<class Impl>
+void MobiusEOFAFermion<Impl>::M5D(const FermionField &psi_i, const FermionField &phi_i, FermionField &chi_i,
+				  Vector<Coeff_t> &lower, Vector<Coeff_t> &diag, Vector<Coeff_t> &upper)
+{
+  chi_i.Checkerboard() = psi_i.Checkerboard();
+  GridBase *grid = psi_i.Grid();
+  int Ls = this->Ls;
+  auto psi = psi_i.View();
+  auto phi = phi_i.View();
+  auto chi = chi_i.View();
+
+  assert(phi.Checkerboard() == psi.Checkerboard());
+
+  // Flops = 6.0*(Nc*Ns) *Ls*vol
+  this->M5Dcalls++;
+  this->M5Dtime -= usecond();
+
+  int nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+    uint64_t ss = sss*Ls;
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    spinor tmp1;
+    spinor tmp2;
+    for(int s=0; s<Ls; s++){
+      uint64_t idx_u = ss+((s+1)%Ls);
+      uint64_t idx_l = ss+((s+Ls-1)%Ls);
+      spProj5m(tmp1, psi(idx_u));
+      spProj5p(tmp2, psi(idx_l));
+      coalescedWrite(chi[ss+s], diag[s]*phi(ss+s) + upper[s]*tmp1 + lower[s]*tmp2);
+    }
+  });
+
+  this->M5Dtime += usecond();
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::M5D_shift(const FermionField &psi_i, const FermionField &phi_i, FermionField &chi_i,
+					Vector<Coeff_t> &lower, Vector<Coeff_t> &diag, Vector<Coeff_t> &upper,
+					Vector<Coeff_t> &shift_coeffs)
+{
+  chi_i.Checkerboard() = psi_i.Checkerboard();
+  GridBase *grid = psi_i.Grid();
+  int Ls = this->Ls;
+  auto psi = psi_i.View();
+  auto phi = phi_i.View();
+  auto chi = chi_i.View();
+
+  auto pm  = this->pm;
+  int shift_s = (pm == 1) ? (Ls-1) : 0; // s-component modified by shift operator
+
+  assert(phi.Checkerboard() == psi.Checkerboard());
+
+  // Flops = 6.0*(Nc*Ns) *Ls*vol
+  this->M5Dcalls++;
+  this->M5Dtime -= usecond();
+
+  int nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+    uint64_t ss = sss*Ls;
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    spinor tmp1;
+    spinor tmp2;
+    spinor tmp;
+    for(int s=0; s<Ls; s++){
+      uint64_t idx_u = ss+((s+1)%Ls);
+      uint64_t idx_l = ss+((s+Ls-1)%Ls);
+      spProj5m(tmp1, psi(idx_u));
+      spProj5p(tmp2, psi(idx_l));
+
+      if(pm == 1){ spProj5p(tmp, psi(ss+shift_s)); }
+      else       { spProj5m(tmp, psi(ss+shift_s)); }
+
+      coalescedWrite(chi[ss+s], diag[s]*phi(ss+s) + upper[s]*tmp1 +lower[s]*tmp2 + shift_coeffs[s]*tmp);
+    }
+  });
+
+  this->M5Dtime += usecond();
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::M5Ddag(const FermionField &psi_i, const FermionField &phi_i, FermionField &chi_i,
+				     Vector<Coeff_t> &lower, Vector<Coeff_t> &diag, Vector<Coeff_t> &upper)
+{
+  chi_i.Checkerboard() = psi_i.Checkerboard();
+  GridBase *grid = psi_i.Grid();
+  int Ls = this->Ls;
+  auto psi = psi_i.View();
+  auto phi = phi_i.View();
+  auto chi = chi_i.View();
+
+  assert(phi.Checkerboard() == psi.Checkerboard());
+
+  // Flops = 6.0*(Nc*Ns) *Ls*vol
+  this->M5Dcalls++;
+  this->M5Dtime -= usecond();
+
+  int nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(), {
+    uint64_t ss = sss*Ls;
+
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    spinor tmp1, tmp2;
+
+    for(int s=0; s<Ls; s++){
+      uint64_t idx_u = ss+((s+1)%Ls);
+      uint64_t idx_l = ss+((s+Ls-1)%Ls);
+      spProj5p(tmp1, psi(idx_u));
+      spProj5m(tmp2, psi(idx_l));
+      coalescedWrite(chi[ss+s], diag[s]*phi(ss+s) + upper[s]*tmp1 + lower[s]*tmp2);
+    }
+  });
+
+  this->M5Dtime += usecond();
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::M5Ddag_shift(const FermionField &psi_i, const FermionField &phi_i, FermionField &chi_i,
+					   Vector<Coeff_t> &lower, Vector<Coeff_t> &diag, Vector<Coeff_t> &upper,
+					   Vector<Coeff_t> &shift_coeffs)
+{
+  chi_i.Checkerboard() = psi_i.Checkerboard();
+  GridBase *grid = psi_i.Grid();
+  int Ls = this->Ls;
+  int shift_s = (this->pm == 1) ? (Ls-1) : 0; // s-component modified by shift operator
+  auto psi = psi_i.View();
+  auto phi = phi_i.View();
+  auto chi = chi_i.View();
+
+  assert(phi.Checkerboard() == psi.Checkerboard());
+
+  // Flops = 6.0*(Nc*Ns) *Ls*vol
+  this->M5Dcalls++;
+  this->M5Dtime -= usecond();
+
+  auto pm = this->pm;
+
+  int nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+    uint64_t ss = sss*Ls;
+
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    spinor tmp1, tmp2, tmp;
+    tmp1=Zero();
+    coalescedWrite(chi[ss+Ls-1],tmp1);
+
+    for(int s=0; s<Ls; s++){
+
+      uint64_t idx_u = ss+((s+1)%Ls);
+      uint64_t idx_l = ss+((s+Ls-1)%Ls);
+
+      spProj5p(tmp1, psi(idx_u));
+      spProj5m(tmp2, psi(idx_l));
+
+      if(s==(Ls-1)) coalescedWrite(chi[ss+s], chi(ss+s)+ diag[s]*phi(ss+s) + upper[s]*tmp1 + lower[s]*tmp2);
+      else          coalescedWrite(chi[ss+s], diag[s]*phi(ss+s) + upper[s]*tmp1 + lower[s]*tmp2);
+      if(pm == 1){ spProj5p(tmp, psi(ss+s)); }
+      else       { spProj5m(tmp, psi(ss+s)); }
+
+      coalescedWrite(chi[ss+shift_s],chi(ss+shift_s)+shift_coeffs[s]*tmp);
+    }
+  });
+
+  this->M5Dtime += usecond();
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::MooeeInv(const FermionField &psi_i, FermionField &chi_i)
+{
+  chi_i.Checkerboard() = psi_i.Checkerboard();
+  GridBase *grid = psi_i.Grid();
+  int Ls = this->Ls;
+  auto psi = psi_i.View();
+  auto chi = chi_i.View();
+
+  auto plee = & this->lee [0];
+  auto pdee = & this->dee [0];
+  auto puee = & this->uee [0];
+  auto pleem= & this->leem[0];
+  auto pueem= & this->ueem[0];
+
+  if(this->shift != 0.0){ MooeeInv_shift(psi_i,chi_i); return; }
+
+  this->MooeeInvCalls++;
+  this->MooeeInvTime -= usecond();
+
+  int nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+
+    uint64_t ss = sss*Ls;
+
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    spinor tmp;
+
+    // Apply (L^{\prime})^{-1}
+    coalescedWrite(chi[ss], psi(ss)); // chi[0]=psi[0]
+    for(int s=1; s<Ls; s++){
+      spProj5p(tmp, chi(ss+s-1));
+      coalescedWrite(chi[ss+s], psi(ss+s) - plee[s-1]*tmp);
+    }
+
+    // L_m^{-1}
+    for(int s=0; s<Ls-1; s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
+      spProj5m(tmp, chi(ss+s));
+      coalescedWrite(chi[ss+Ls-1], chi(ss+Ls-1) - pleem[s]*tmp);
+    }
+
+    // U_m^{-1} D^{-1}
+    for(int s=0; s<Ls-1; s++){ // Chi[s] + 1/d chi[s]
+      spProj5p(tmp, chi(ss+Ls-1));
+      coalescedWrite(chi[ss+s], (1.0/pdee[s])*chi(ss+s) - (pueem[s]/pdee[Ls-1])*tmp);
+    }
+    coalescedWrite(chi[ss+Ls-1], (1.0/pdee[Ls-1])*chi(ss+Ls-1));
+
+    // Apply U^{-1}
+    for(int s=Ls-2; s>=0; s--){
+      spProj5m(tmp, chi(ss+s+1));
+      coalescedWrite(chi[ss+s], chi(ss+s) - puee[s]*tmp);
+    }
+  });
+   
+  this->MooeeInvTime += usecond();
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::MooeeInv_shift(const FermionField &psi_i, FermionField &chi_i)
+{
+  chi_i.Checkerboard() = psi_i.Checkerboard();
+  GridBase *grid = psi_i.Grid();
+  int Ls = this->Ls;
+  auto psi = psi_i.View();
+  auto chi = chi_i.View();
+
+  auto pm = this->pm;
+  auto plee = & this->lee [0];
+  auto pdee = & this->dee [0];
+  auto puee = & this->uee [0];
+  auto pleem= & this->leem[0];
+  auto pueem= & this->ueem[0];
+  auto pMooeeInv_shift_lc   = &MooeeInv_shift_lc[0];
+  auto pMooeeInv_shift_norm = &MooeeInv_shift_norm[0];
+  this->MooeeInvCalls++;
+  this->MooeeInvTime -= usecond();
+
+  int nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+
+    uint64_t ss = sss*Ls;
+
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    spinor tmp1,tmp2,tmp2_spProj;
+
+    // Apply (L^{\prime})^{-1} and accumulate MooeeInv_shift_lc[j]*psi[j] in tmp2
+    coalescedWrite(chi[ss], psi(ss)); // chi[0]=psi[0]
+    tmp2 = pMooeeInv_shift_lc[0]*psi(ss);
+    for(int s=1; s<Ls; s++){
+      spProj5p(tmp1, chi(ss+s-1));
+      coalescedWrite(chi[ss+s], psi(ss+s) - plee[s-1]*tmp1);
+      tmp2 = tmp2 + pMooeeInv_shift_lc[s]*psi(ss+s);
+    }
+    if(pm == 1){ spProj5p(tmp2_spProj, tmp2);}
+    else       { spProj5m(tmp2_spProj, tmp2); }
+
+    // L_m^{-1}
+    for(int s=0; s<Ls-1; s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
+      spProj5m(tmp1, chi(ss+s));
+      coalescedWrite(chi[ss+Ls-1], chi(ss+Ls-1) - pleem[s]*tmp1);
+    }
+
+    // U_m^{-1} D^{-1}
+    for(int s=0; s<Ls-1; s++){ // Chi[s] + 1/d chi[s]
+      spProj5p(tmp1, chi(ss+Ls-1));
+      coalescedWrite(chi[ss+s], (1.0/pdee[s])*chi(ss+s) - (pueem[s]/pdee[Ls-1])*tmp1);
+    }
+    // chi[ss+Ls-1] = (1.0/pdee[Ls-1])*chi[ss+Ls-1] + MooeeInv_shift_norm[Ls-1]*tmp2_spProj;
+    coalescedWrite(chi[ss+Ls-1], (1.0/pdee[Ls-1])*chi(ss+Ls-1));
+    spProj5m(tmp1, chi(ss+Ls-1));
+    coalescedWrite(chi[ss+Ls-1], chi(ss+Ls-1) + pMooeeInv_shift_norm[Ls-1]*tmp2_spProj);
+
+    // Apply U^{-1} and add shift term
+    for(int s=Ls-2; s>=0; s--){
+      coalescedWrite(chi[ss+s] , chi(ss+s) - puee[s]*tmp1);
+      spProj5m(tmp1, chi(ss+s));
+      coalescedWrite(chi[ss+s], chi(ss+s) + pMooeeInv_shift_norm[s]*tmp2_spProj);
+    }
+  });
+
+  this->MooeeInvTime += usecond();
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::MooeeInvDag(const FermionField &psi_i, FermionField &chi_i)
+{
+  if(this->shift != 0.0){ MooeeInvDag_shift(psi_i,chi_i); return; }
+
+  chi_i.Checkerboard() = psi_i.Checkerboard();
+  GridBase *grid = psi_i.Grid();
+  int Ls = this->Ls;
+  auto psi = psi_i.View();
+  auto chi = chi_i.View();
+
+  auto plee = & this->lee [0];
+  auto pdee = & this->dee [0];
+  auto puee = & this->uee [0];
+  auto pleem= & this->leem[0];
+  auto pueem= & this->ueem[0];
+
+  this->MooeeInvCalls++;
+  this->MooeeInvTime -= usecond();
+
+  int nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+
+    uint64_t ss = sss*Ls;
+
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    spinor tmp;
+
+    // Apply (U^{\prime})^{-dag}
+    coalescedWrite(chi[ss], psi(ss));
+    for(int s=1; s<Ls; s++){
+      spProj5m(tmp, chi(ss+s-1));
+      coalescedWrite(chi[ss+s], psi(ss+s) - puee[s-1]*tmp);
+    }
+    
+    // U_m^{-\dag}
+    for(int s=0; s<Ls-1; s++){
+      spProj5p(tmp, chi(ss+s));
+      coalescedWrite(chi[ss+Ls-1], chi(ss+Ls-1) - pueem[s]*tmp);
+    }
+
+    // L_m^{-\dag} D^{-dag}
+    for(int s=0; s<Ls-1; s++){
+      spProj5m(tmp, chi(ss+Ls-1));
+      coalescedWrite(chi[ss+s], (1.0/pdee[s])*chi(ss+s) - (pleem[s]/pdee[Ls-1])*tmp);
+    }
+    coalescedWrite(chi[ss+Ls-1], (1.0/pdee[Ls-1])*chi(ss+Ls-1));
+
+    // Apply L^{-dag}
+    for(int s=Ls-2; s>=0; s--){
+      spProj5p(tmp, chi(ss+s+1));
+      coalescedWrite(chi[ss+s], chi(ss+s) - plee[s]*tmp);
+    }
+  });
+
+  this->MooeeInvTime += usecond();
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::MooeeInvDag_shift(const FermionField &psi_i, FermionField &chi_i)
+{
+  chi_i.Checkerboard() = psi_i.Checkerboard();
+  GridBase *grid = psi_i.Grid();
+  auto psi = psi_i.View();
+  auto chi = chi_i.View();
+  int Ls = this->Ls;
+
+  auto pm = this->pm;
+  auto plee = & this->lee [0];
+  auto pdee = & this->dee [0];
+  auto puee = & this->uee [0];
+  auto pleem= & this->leem[0];
+  auto pueem= & this->ueem[0];
+  auto pMooeeInvDag_shift_lc   = &MooeeInvDag_shift_lc[0];
+  auto pMooeeInvDag_shift_norm = &MooeeInvDag_shift_norm[0];
+
+  this->MooeeInvCalls++;
+  this->MooeeInvTime -= usecond();
+
+  int nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,Simd::Nsimd(),{
+
+    uint64_t ss = sss*Ls;
+
+    typedef decltype(coalescedRead(psi[0])) spinor;
+    spinor tmp1,tmp2,tmp2_spProj;
+
+    // Apply (U^{\prime})^{-dag} and accumulate MooeeInvDag_shift_lc[j]*psi[j] in tmp2
+    coalescedWrite(chi[ss], psi(ss));
+    tmp2 = pMooeeInvDag_shift_lc[0]*psi(ss);
+    for(int s=1; s<Ls; s++){
+      spProj5m(tmp1, chi(ss+s-1));
+      coalescedWrite(chi[ss+s],psi(ss+s) - puee[s-1]*tmp1);
+      tmp2 = tmp2 + pMooeeInvDag_shift_lc[s]*psi(ss+s);
+    }
+
+    if(pm == 1){ spProj5p(tmp2_spProj, tmp2);}
+    else       { spProj5m(tmp2_spProj, tmp2);}
+
+    // U_m^{-\dag}
+    for(int s=0; s<Ls-1; s++){
+      spProj5p(tmp1, chi(ss+s));
+      coalescedWrite(chi[ss+Ls-1], chi(ss+Ls-1) - pueem[s]*tmp1);
+    }
+
+    // L_m^{-\dag} D^{-dag}
+    for(int s=0; s<Ls-1; s++){
+      spProj5m(tmp1, chi(ss+Ls-1));
+      coalescedWrite(chi[ss+s], (1.0/pdee[s])*chi(ss+s) - (pleem[s]/pdee[Ls-1])*tmp1);
+    }
+    coalescedWrite(chi[ss+Ls-1], (1.0/pdee[Ls-1])*chi(ss+Ls-1));
+    spProj5p(tmp1, chi(ss+Ls-1));
+    coalescedWrite(chi[ss+Ls-1], chi(ss+Ls-1) + pMooeeInvDag_shift_norm[Ls-1]*tmp2_spProj);
+
+    // Apply L^{-dag}
+    for(int s=Ls-2; s>=0; s--){
+      coalescedWrite(chi[ss+s], chi(ss+s) - plee[s]*tmp1);
+      spProj5p(tmp1, chi(ss+s));
+      coalescedWrite(chi[ss+s], chi(ss+s) + pMooeeInvDag_shift_norm[s]*tmp2_spProj);
+    }
+  });
+
+  this->MooeeInvTime += usecond();
+}
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h b/Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h
new file mode 100644
index 00000000..256423e6
--- /dev/null
+++ b/Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h
@@ -0,0 +1,407 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#pragma once
+
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+
+NAMESPACE_BEGIN(Grid);
+
+template<class Impl>
+MobiusEOFAFermion<Impl>::MobiusEOFAFermion(
+					   GaugeField            &_Umu,
+					   GridCartesian         &FiveDimGrid,
+					   GridRedBlackCartesian &FiveDimRedBlackGrid,
+					   GridCartesian         &FourDimGrid,
+					   GridRedBlackCartesian &FourDimRedBlackGrid,
+					   RealD _mq1, RealD _mq2, RealD _mq3,
+					   RealD _shift, int _pm, RealD _M5,
+					   RealD _b, RealD _c, const ImplParams &p) :
+  AbstractEOFAFermion<Impl>(_Umu, FiveDimGrid, FiveDimRedBlackGrid,
+			    FourDimGrid, FourDimRedBlackGrid, _mq1, _mq2, _mq3,
+			    _shift, _pm, _M5, _b, _c, p)
+{
+  int Ls = this->Ls;
+
+  RealD eps = 1.0;
+  Approx::zolotarev_data *zdata = Approx::higham(eps, this->Ls);
+  assert(zdata->n == this->Ls);
+
+  std::cout << GridLogMessage << "MobiusEOFAFermion (b=" << _b <<
+    ",c=" << _c << ") with Ls=" << Ls << std::endl;
+  this->SetCoefficientsTanh(zdata, _b, _c);
+  std::cout << GridLogMessage << "EOFA parameters: (mq1=" << _mq1 <<
+    ",mq2=" << _mq2 << ",mq3=" << _mq3 << ",shift=" << _shift <<
+    ",pm=" << _pm << ")" << std::endl;
+
+  Approx::zolotarev_free(zdata);
+
+  if(_shift != 0.0){
+    SetCoefficientsPrecondShiftOps();
+  } else {
+    Mooee_shift.resize(Ls, 0.0);
+    MooeeInv_shift_lc.resize(Ls, 0.0);
+    MooeeInv_shift_norm.resize(Ls, 0.0);
+    MooeeInvDag_shift_lc.resize(Ls, 0.0);
+    MooeeInvDag_shift_norm.resize(Ls, 0.0);
+  }
+}
+
+/****************************************************************
+ * Additional EOFA operators only called outside the inverter.  
+ * Since speed is not essential, simple axpby-style
+ * implementations should be fine.
+ ***************************************************************/
+template<class Impl>
+void MobiusEOFAFermion<Impl>::Omega(const FermionField& psi, FermionField& Din, int sign, int dag)
+{
+  int Ls = this->Ls;
+  RealD alpha = this->alpha;
+
+  Din = Zero();
+  if((sign == 1) && (dag == 0)) { // \Omega_{+}
+    for(int s=0; s<Ls; ++s){
+      axpby_ssp(Din, 0.0, psi, 2.0*std::pow(1.0-alpha,Ls-s-1)/std::pow(1.0+alpha,Ls-s), psi, s, 0);
+    }
+  } else if((sign == -1) && (dag == 0)) { // \Omega_{-}
+    for(int s=0; s<Ls; ++s){
+      axpby_ssp(Din, 0.0, psi, 2.0*std::pow(1.0-alpha,s)/std::pow(1.0+alpha,s+1), psi, s, 0);
+    }
+  } else if((sign == 1 ) && (dag == 1)) { // \Omega_{+}^{\dagger}
+    for(int sp=0; sp<Ls; ++sp){
+      axpby_ssp(Din, 1.0, Din, 2.0*std::pow(1.0-alpha,Ls-sp-1)/std::pow(1.0+alpha,Ls-sp), psi, 0, sp);
+    }
+  } else if((sign == -1) && (dag == 1)) { // \Omega_{-}^{\dagger}
+    for(int sp=0; sp<Ls; ++sp){
+      axpby_ssp(Din, 1.0, Din, 2.0*std::pow(1.0-alpha,sp)/std::pow(1.0+alpha,sp+1), psi, 0, sp);
+    }
+  }
+}
+
+// This is the operator relating the usual Ddwf to TWQCD's EOFA Dirac operator (arXiv:1706.05843, Eqn. 6).
+// It also relates the preconditioned and unpreconditioned systems described in Appendix B.2.
+template<class Impl>
+void MobiusEOFAFermion<Impl>::Dtilde(const FermionField& psi, FermionField& chi)
+{
+  int Ls    = this->Ls;
+  RealD b   = 0.5 * ( 1.0 + this->alpha );
+  RealD c   = 0.5 * ( 1.0 - this->alpha );
+  RealD mq1 = this->mq1;
+
+  for(int s=0; s<Ls; ++s){
+    if(s == 0) {
+      axpby_ssp_pminus(chi, b, psi, -c, psi, s, s+1);
+      axpby_ssp_pplus (chi, 1.0, chi, mq1*c, psi, s, Ls-1);
+    } else if(s == (Ls-1)) {
+      axpby_ssp_pminus(chi, b, psi, mq1*c, psi, s, 0);
+      axpby_ssp_pplus (chi, 1.0, chi, -c, psi, s, s-1);
+    } else {
+      axpby_ssp_pminus(chi, b, psi, -c, psi, s, s+1);
+      axpby_ssp_pplus (chi, 1.0, chi, -c, psi, s, s-1);
+    }
+  }
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::DtildeInv(const FermionField& psi, FermionField& chi)
+{
+  int Ls = this->Ls;
+  RealD m = this->mq1;
+  RealD c = 0.5 * this->alpha;
+  RealD d = 0.5;
+
+  RealD DtInv_p(0.0), DtInv_m(0.0);
+  RealD N = std::pow(c+d,Ls) + m*std::pow(c-d,Ls);
+  FermionField tmp(this->FermionGrid());
+
+  for(int s=0; s<Ls; ++s){
+    for(int sp=0; sp<Ls; ++sp){
+
+      DtInv_p = m * std::pow(-1.0,s-sp+1) * std::pow(c-d,Ls+s-sp) / std::pow(c+d,s-sp+1) / N;
+      DtInv_p += (s < sp) ? 0.0 : std::pow(-1.0,s-sp) * std::pow(c-d,s-sp) / std::pow(c+d,s-sp+1);
+      DtInv_m = m * std::pow(-1.0,sp-s+1) * std::pow(c-d,Ls+sp-s) / std::pow(c+d,sp-s+1) / N;
+      DtInv_m += (s > sp) ? 0.0 : std::pow(-1.0,sp-s) * std::pow(c-d,sp-s) / std::pow(c+d,sp-s+1);
+
+      if(sp == 0){
+	axpby_ssp_pplus (tmp, 0.0, tmp, DtInv_p, psi, s, sp);
+	axpby_ssp_pminus(tmp, 0.0, tmp, DtInv_m, psi, s, sp);
+      } else {
+	axpby_ssp_pplus (tmp, 1.0, tmp, DtInv_p, psi, s, sp);
+	axpby_ssp_pminus(tmp, 1.0, tmp, DtInv_m, psi, s, sp);
+      }
+
+    }}
+}
+
+/*****************************************************************************************************/
+
+template<class Impl>
+RealD MobiusEOFAFermion<Impl>::M(const FermionField& psi, FermionField& chi)
+{
+  FermionField Din(psi.Grid());
+
+  this->Meooe5D(psi, Din);
+  this->DW(Din, chi, DaggerNo);
+  axpby(chi, 1.0, 1.0, chi, psi);
+  this->M5D(psi, chi);
+  return(norm2(chi));
+}
+
+template<class Impl>
+RealD MobiusEOFAFermion<Impl>::Mdag(const FermionField& psi, FermionField& chi)
+{
+  FermionField Din(psi.Grid());
+
+  this->DW(psi, Din, DaggerYes);
+  this->MeooeDag5D(Din, chi);
+  this->M5Ddag(psi, chi);
+  axpby(chi, 1.0, 1.0, chi, psi);
+  return(norm2(chi));
+}
+
+/********************************************************************
+ * Performance critical fermion operators called inside the inverter
+ ********************************************************************/
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::M5D(const FermionField& psi, FermionField& chi)
+{
+  int Ls = this->Ls;
+
+  Vector<Coeff_t> diag(Ls,1.0);
+  Vector<Coeff_t> upper(Ls,-1.0);  upper[Ls-1] = this->mq1;
+  Vector<Coeff_t> lower(Ls,-1.0);  lower[0]    = this->mq1;
+
+  // no shift term
+  if(this->shift == 0.0){ this->M5D(psi, chi, chi, lower, diag, upper); }
+
+  // fused M + shift operation
+  else{ this->M5D_shift(psi, chi, chi, lower, diag, upper, Mooee_shift); }
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::M5Ddag(const FermionField& psi, FermionField& chi)
+{
+  int Ls = this->Ls;
+
+  Vector<Coeff_t> diag(Ls,1.0);
+  Vector<Coeff_t> upper(Ls,-1.0);  upper[Ls-1] = this->mq1;
+  Vector<Coeff_t> lower(Ls,-1.0);  lower[0]    = this->mq1;
+
+  // no shift term
+  if(this->shift == 0.0){ this->M5Ddag(psi, chi, chi, lower, diag, upper); }
+
+  // fused M + shift operation
+  else{ this->M5Ddag_shift(psi, chi, chi, lower, diag, upper, Mooee_shift); }
+}
+
+// half checkerboard operations
+template<class Impl>
+void MobiusEOFAFermion<Impl>::Mooee(const FermionField& psi, FermionField& chi)
+{
+  int Ls = this->Ls;
+
+  // coefficients of Mooee
+  Vector<Coeff_t> diag = this->bee;
+  Vector<Coeff_t> upper(Ls);
+  Vector<Coeff_t> lower(Ls);
+  for(int s=0; s<Ls; s++){
+    upper[s] = -this->cee[s];
+    lower[s] = -this->cee[s];
+  }
+  upper[Ls-1] *= -this->mq1;
+  lower[0]    *= -this->mq1;
+
+  // no shift term
+  if(this->shift == 0.0){ this->M5D(psi, psi, chi, lower, diag, upper); }
+
+  // fused M + shift operation
+  else { this->M5D_shift(psi, psi, chi, lower, diag, upper, Mooee_shift); }
+}
+
+template<class Impl>
+void MobiusEOFAFermion<Impl>::MooeeDag(const FermionField& psi, FermionField& chi)
+{
+  int Ls = this->Ls;
+
+  // coefficients of MooeeDag
+  Vector<Coeff_t> diag = this->bee;
+  Vector<Coeff_t> upper(Ls);
+  Vector<Coeff_t> lower(Ls);
+  for(int s=0; s<Ls; s++){
+    if(s==0) {
+      upper[s] = -this->cee[s+1];
+      lower[s] = this->mq1*this->cee[Ls-1];
+    } else if(s==(Ls-1)) {
+      upper[s] = this->mq1*this->cee[0];
+      lower[s] = -this->cee[s-1];
+    } else {
+      upper[s] = -this->cee[s+1];
+      lower[s] = -this->cee[s-1];
+    }
+  }
+
+  // no shift term
+  if(this->shift == 0.0){ this->M5Ddag(psi, psi, chi, lower, diag, upper); }
+
+  // fused M + shift operation
+  else{ this->M5Ddag_shift(psi, psi, chi, lower, diag, upper, Mooee_shift); }
+}
+
+/****************************************************************************************/
+
+// Computes coefficients for applying Cayley preconditioned shift operators
+//  (Mooee + \Delta) --> Mooee_shift
+//  (Mooee + \Delta)^{-1} --> MooeeInv_shift_lc, MooeeInv_shift_norm
+//  (Mooee + \Delta)^{-dag} --> MooeeInvDag_shift_lc, MooeeInvDag_shift_norm
+// For the latter two cases, the operation takes the form
+//  [ (Mooee + \Delta)^{-1} \psi ]_{i} = Mooee_{ij} \psi_{j} +
+//      ( MooeeInv_shift_norm )_{i} ( \sum_{j} [ MooeeInv_shift_lc ]_{j} P_{pm} \psi_{j} )
+template<class Impl>
+void MobiusEOFAFermion<Impl>::SetCoefficientsPrecondShiftOps()
+{
+  int   Ls    = this->Ls;
+  int   pm    = this->pm;
+  RealD alpha = this->alpha;
+  RealD k     = this->k;
+  RealD mq1   = this->mq1;
+  RealD shift = this->shift;
+
+  // Initialize
+  Mooee_shift.resize(Ls);
+  MooeeInv_shift_lc.resize(Ls);
+  MooeeInv_shift_norm.resize(Ls);
+  MooeeInvDag_shift_lc.resize(Ls);
+  MooeeInvDag_shift_norm.resize(Ls);
+
+  // Construct Mooee_shift
+  int idx(0);
+  Coeff_t N = ( (pm == 1) ? 1.0 : -1.0 ) * (2.0*shift*k) *
+    ( std::pow(alpha+1.0,Ls) + mq1*std::pow(alpha-1.0,Ls) );
+  for(int s=0; s<Ls; ++s){
+    idx = (pm == 1) ? (s) : (Ls-1-s);
+    Mooee_shift[idx] = N * std::pow(-1.0,s) * std::pow(alpha-1.0,s) / std::pow(alpha+1.0,Ls+s+1);
+  }
+
+  // Tridiagonal solve for MooeeInvDag_shift_lc
+  {
+    Coeff_t m(0.0);
+    Vector<Coeff_t> d = Mooee_shift;
+    Vector<Coeff_t> u(Ls,0.0);
+    Vector<Coeff_t> y(Ls,0.0);
+    Vector<Coeff_t> q(Ls,0.0);
+    if(pm == 1){ u[0] = 1.0; }
+    else{ u[Ls-1] = 1.0; }
+
+    // Tridiagonal matrix algorithm + Sherman-Morrison formula
+    //
+    // We solve
+    //  ( Mooee' + u \otimes v ) MooeeInvDag_shift_lc = Mooee_shift
+    // where Mooee' is the tridiagonal part of Mooee_{+}, and
+    // u = (1,0,...,0) and v = (0,...,0,mq1*cee[0]) are chosen
+    // so that the outer-product u \otimes v gives the (0,Ls-1)
+    // entry of Mooee_{+}.
+    //
+    // We do this as two solves: Mooee'*y = d and Mooee'*q = u,
+    // and then construct the solution to the original system
+    //  MooeeInvDag_shift_lc = y - <v,y> / ( 1 + <v,q> ) q
+    if(pm == 1){
+      for(int s=1; s<Ls; ++s){
+	m = -this->cee[s] / this->bee[s-1];
+	d[s] -= m*d[s-1];
+	u[s] -= m*u[s-1];
+      }
+    }
+    y[Ls-1] = d[Ls-1] / this->bee[Ls-1];
+    q[Ls-1] = u[Ls-1] / this->bee[Ls-1];
+    for(int s=Ls-2; s>=0; --s){
+      if(pm == 1){
+	y[s] = d[s] / this->bee[s];
+	q[s] = u[s] / this->bee[s];
+      } else {
+	y[s] = ( d[s] + this->cee[s]*y[s+1] ) / this->bee[s];
+	q[s] = ( u[s] + this->cee[s]*q[s+1] ) / this->bee[s];
+      }
+    }
+
+    // Construct MooeeInvDag_shift_lc
+    for(int s=0; s<Ls; ++s){
+      if(pm == 1){
+	MooeeInvDag_shift_lc[s] = y[s] - mq1*this->cee[0]*y[Ls-1] /
+	  (1.0+mq1*this->cee[0]*q[Ls-1]) * q[s];
+      } else {
+	MooeeInvDag_shift_lc[s] = y[s] - mq1*this->cee[Ls-1]*y[0] /
+	  (1.0+mq1*this->cee[Ls-1]*q[0]) * q[s];
+      }
+    }
+
+    // Compute remaining coefficients
+    N = (pm == 1) ? (1.0 + MooeeInvDag_shift_lc[Ls-1]) : (1.0 + MooeeInvDag_shift_lc[0]);
+    for(int s=0; s<Ls; ++s){
+
+      // MooeeInv_shift_lc
+      if(pm == 1){ MooeeInv_shift_lc[s] = pow(this->bee[s],s)      * pow(this->cee[s],Ls-1-s); }
+      else       { MooeeInv_shift_lc[s] = pow(this->bee[s],Ls-1-s) * pow(this->cee[s],s); }
+
+      // MooeeInv_shift_norm
+      MooeeInv_shift_norm[s] = -MooeeInvDag_shift_lc[s] /
+	( pow(this->bee[s],Ls) + mq1*pow(this->cee[s],Ls) ) / N;
+
+      // MooeeInvDag_shift_norm
+      if(pm == 1){ MooeeInvDag_shift_norm[s] = -pow(this->bee[s],s) * pow(this->cee[s],(Ls-1-s)) /
+     	  ( pow(this->bee[s],Ls) + mq1*pow(this->cee[s],Ls) ) / N; }
+      else{ MooeeInvDag_shift_norm[s] = -pow(this->bee[s],(Ls-1-s)) * pow(this->cee[s],s) /
+	  ( pow(this->bee[s],Ls) + mq1*pow(this->cee[s],Ls) ) / N; }
+    }
+  }
+}
+
+// Recompute coefficients for a different value of shift constant
+template<class Impl>
+void MobiusEOFAFermion<Impl>::RefreshShiftCoefficients(RealD new_shift)
+{
+  this->shift = new_shift;
+  if(new_shift != 0.0){
+    SetCoefficientsPrecondShiftOps();
+  } else {
+    int Ls = this->Ls;
+    Mooee_shift.resize(Ls,0.0);
+    MooeeInv_shift_lc.resize(Ls,0.0);
+    MooeeInv_shift_norm.resize(Ls,0.0);
+    MooeeInvDag_shift_lc.resize(Ls,0.0);
+    MooeeInvDag_shift_norm.resize(Ls,0.0);
+  }
+}
+
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h b/Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h
new file mode 100644
index 00000000..9f8f91ad
--- /dev/null
+++ b/Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h
@@ -0,0 +1,450 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
+
+NAMESPACE_BEGIN(Grid);
+
+template<class Impl>
+void  PartialFractionFermion5D<Impl>::Mdir (const FermionField &psi, FermionField &chi,int dir,int disp){
+  // this does both dag and undag but is trivial; make a common helper routing
+  int Ls = this->Ls;
+
+  this->DhopDir(psi,chi,dir,disp);
+
+  int nblock=(Ls-1)/2;
+  for(int b=0;b<nblock;b++){
+    int s = 2*b;
+    ag5xpby_ssp(chi,-scale,chi,0.0,chi,s,s); 
+    ag5xpby_ssp(chi, scale,chi,0.0,chi,s+1,s+1); 
+  }
+  ag5xpby_ssp(chi,p[nblock]*scale/amax,chi,0.0,chi,Ls-1,Ls-1);
+
+}
+template<class Impl>
+void   PartialFractionFermion5D<Impl>::Meooe_internal(const FermionField &psi, FermionField &chi,int dag)
+{
+  int Ls = this->Ls;
+  if ( psi.Checkerboard() == Odd ) {
+    this->DhopEO(psi,chi,DaggerNo);
+  } else {
+    this->DhopOE(psi,chi,DaggerNo);
+  }
+
+  int nblock=(Ls-1)/2;
+  for(int b=0;b<nblock;b++){
+    int s = 2*b;
+    ag5xpby_ssp(chi,-scale,chi,0.0,chi,s,s); 
+    ag5xpby_ssp(chi, scale,chi,0.0,chi,s+1,s+1); 
+  }
+  ag5xpby_ssp(chi,p[nblock]*scale/amax,chi,0.0,chi,Ls-1,Ls-1);
+}
+
+template<class Impl>
+void   PartialFractionFermion5D<Impl>::Mooee_internal(const FermionField &psi, FermionField &chi,int dag)
+{
+  // again dag and undag are trivially related
+  int sign = dag ? (-1) : 1;
+  int Ls = this->Ls;
+      
+  int nblock=(Ls-1)/2;
+  for(int b=0;b<nblock;b++){
+	
+    int s = 2*b;
+    RealD pp = p[nblock-1-b];
+    RealD qq = q[nblock-1-b];
+	
+    // Do each 2x2 block aligned at s and multiplies Dw site diagonal by G5 so Hw
+    ag5xpby_ssp(chi,-dw_diag*scale,psi,amax*sqrt(qq)*scale,psi, s  ,s+1); 
+    ag5xpby_ssp(chi, dw_diag*scale,psi,amax*sqrt(qq)*scale,psi, s+1,s);
+    axpby_ssp  (chi, 1.0, chi,sqrt(amax*pp)*scale*sign,psi,s+1,Ls-1);
+  }
+      
+  {
+    RealD R=(1+mass)/(1-mass);
+    //R g5 psi[Ls-1] + p[0] H
+    ag5xpbg5y_ssp(chi,R*scale,psi,p[nblock]*scale*dw_diag/amax,psi,Ls-1,Ls-1);
+	
+    for(int b=0;b<nblock;b++){
+      int s = 2*b+1;
+      RealD pp = p[nblock-1-b];
+      axpby_ssp(chi,1.0,chi,-sqrt(amax*pp)*scale*sign,psi,Ls-1,s);
+    }
+  }
+}
+
+template<class Impl>
+void   PartialFractionFermion5D<Impl>::MooeeInv_internal(const FermionField &psi, FermionField &chi,int dag)
+{
+  int sign = dag ? (-1) : 1;
+  int Ls = this->Ls;
+
+  FermionField tmp(psi.Grid());
+      
+  ///////////////////////////////////////////////////////////////////////////////////////
+  //Linv
+  ///////////////////////////////////////////////////////////////////////////////////////
+  int nblock=(Ls-1)/2;
+
+  axpy(chi,0.0,psi,psi); // Identity piece
+      
+  for(int b=0;b<nblock;b++){
+    int s = 2*b;
+    RealD pp = p[nblock-1-b];
+    RealD qq = q[nblock-1-b];
+    RealD coeff1=sign*sqrt(amax*amax*amax*pp*qq) / ( dw_diag*dw_diag + amax*amax* qq);
+    RealD coeff2=sign*sqrt(amax*pp)*dw_diag / ( dw_diag*dw_diag + amax*amax* qq); // Implicit g5 here
+    axpby_ssp  (chi,1.0,chi,coeff1,psi,Ls-1,s);
+    axpbg5y_ssp(chi,1.0,chi,coeff2,psi,Ls-1,s+1);
+  }
+      
+  ///////////////////////////////////////////////////////////////////////////////////////
+  //Dinv (note D isn't really diagonal -- just diagonal enough that we can still invert)
+  // Compute Seeinv (coeff of gamma5)
+  ///////////////////////////////////////////////////////////////////////////////////////
+  RealD R=(1+mass)/(1-mass);
+  RealD Seeinv = R + p[nblock]*dw_diag/amax;
+  for(int b=0;b<nblock;b++){
+    Seeinv += p[nblock-1-b]*dw_diag/amax / ( dw_diag*dw_diag/amax/amax + q[nblock-1-b]);
+  }    
+  Seeinv = 1.0/Seeinv;
+      
+  for(int b=0;b<nblock;b++){
+    int s = 2*b;
+    RealD pp = p[nblock-1-b];
+    RealD qq = q[nblock-1-b];
+    RealD coeff1=dw_diag / ( dw_diag*dw_diag + amax*amax* qq); // Implicit g5 here
+    RealD coeff2=amax*sqrt(qq) / ( dw_diag*dw_diag + amax*amax* qq);
+    ag5xpby_ssp  (tmp,-coeff1,chi,coeff2,chi,s,s+1);
+    ag5xpby_ssp  (tmp, coeff1,chi,coeff2,chi,s+1,s);
+  }
+  ag5xpby_ssp  (tmp, Seeinv,chi,0.0,chi,Ls-1,Ls-1);
+      
+  ///////////////////////////////////////////////////////////////////////////////////////
+  // Uinv
+  ///////////////////////////////////////////////////////////////////////////////////////
+  for(int b=0;b<nblock;b++){
+    int s = 2*b;
+    RealD pp = p[nblock-1-b];
+    RealD qq = q[nblock-1-b];
+    RealD coeff1=-sign*sqrt(amax*amax*amax*pp*qq) / ( dw_diag*dw_diag + amax*amax* qq);
+    RealD coeff2=-sign*sqrt(amax*pp)*dw_diag / ( dw_diag*dw_diag + amax*amax* qq); // Implicit g5 here
+    axpby_ssp  (chi,1.0/scale,tmp,coeff1/scale,tmp,s,Ls-1);
+    axpbg5y_ssp(chi,1.0/scale,tmp,coeff2/scale,tmp,s+1,Ls-1);
+  }
+  axpby_ssp  (chi, 1.0/scale,tmp,0.0,tmp,Ls-1,Ls-1);
+}
+
+template<class Impl>
+void   PartialFractionFermion5D<Impl>::M_internal(const FermionField &psi, FermionField &chi,int dag)
+{
+  FermionField D(psi.Grid());
+  
+  int Ls = this->Ls;
+  int sign = dag ? (-1) : 1;
+
+  // For partial frac Hw case (b5=c5=1) chroma quirkily computes
+  //
+  // Conventions for partfrac appear to be a mess.
+  // Tony's Nara lectures have
+  //
+  // BlockDiag(  H/p_i  1             | 1       )    
+  //          (  1      p_i H / q_i^2 | 0       )  
+  //           ---------------------------------
+  //           ( -1      0                | R  +p0 H  )
+  //
+  //Chroma     ( -2H    2sqrt(q_i)    |   0         )
+  //           (2 sqrt(q_i)   2H      |  2 sqrt(p_i) )
+  //           ---------------------------------
+  //           ( 0     -2 sqrt(p_i)   |  2 R gamma_5 + p0 2H
+  //
+  // Edwards/Joo/Kennedy/Wenger
+  //
+  // Here, the "beta's" selected by chroma to scale the unphysical bulk constraint fields
+  // incorporate the approx scale factor. This is obtained by propagating the
+  // scale on "H" out to the off diagonal elements as follows:
+  //
+  // BlockDiag(  H/p_i  1             | 1       ) 
+  //          (  1      p_i H / q_i^2 | 0       )  
+  //           ---------------------------------
+  //          ( -1      0                | R  + p_0 H  )
+  //
+  // becomes:
+  // BlockDiag(  H/ sp_i  1               | 1             ) 
+  //          (  1      sp_i H / s^2q_i^2 | 0             )  
+  //           ---------------------------------
+  //           ( -1      0                | R + p_0/s H   )
+  //
+  //
+  // This is implemented in Chroma by
+  //           p0' = p0/approxMax
+  //           p_i' = p_i*approxMax
+  //           q_i' = q_i*approxMax*approxMax
+  //
+  // After the equivalence transform is applied the matrix becomes
+  // 
+  //Chroma     ( -2H    sqrt(q'_i)    |   0         )
+  //           (sqrt(q'_i)   2H       |   sqrt(p'_i) )
+  //           ---------------------------------
+  //           ( 0     -sqrt(p'_i)    |  2 R gamma_5 + p'0 2H
+  //
+  //     =     ( -2H    sqrt(q_i)amax    |   0              )
+  //           (sqrt(q_i)amax   2H       |   sqrt(p_i*amax) )
+  //           ---------------------------------
+  //           ( 0     -sqrt(p_i)*amax   |  2 R gamma_5 + p0/amax 2H
+  //
+
+  this->DW(psi,D,DaggerNo); 
+
+  int nblock=(Ls-1)/2;
+  for(int b=0;b<nblock;b++){
+	
+    int s = 2*b;
+    double pp = p[nblock-1-b];
+    double qq = q[nblock-1-b];
+	
+    // Do each 2x2 block aligned at s and
+    ag5xpby_ssp(chi,-1.0*scale,D,amax*sqrt(qq)*scale,psi, s  ,s+1); // Multiplies Dw by G5 so Hw
+    ag5xpby_ssp(chi, 1.0*scale,D,amax*sqrt(qq)*scale,psi, s+1,s);
+	
+    // Pick up last column
+    axpby_ssp  (chi, 1.0, chi,sqrt(amax*pp)*scale*sign,psi,s+1,Ls-1);
+  }
+	
+  {
+    double R=(1+this->mass)/(1-this->mass);
+    //R g5 psi[Ls] + p[0] H
+    ag5xpbg5y_ssp(chi,R*scale,psi,p[nblock]*scale/amax,D,Ls-1,Ls-1);
+	
+    for(int b=0;b<nblock;b++){
+      int s = 2*b+1;
+      double pp = p[nblock-1-b];
+      axpby_ssp(chi,1.0,chi,-sqrt(amax*pp)*scale*sign,psi,Ls-1,s);
+    }
+  }
+
+}
+
+template<class Impl>
+RealD  PartialFractionFermion5D<Impl>::M    (const FermionField &in, FermionField &out)
+{
+  M_internal(in,out,DaggerNo);
+  return norm2(out);
+}
+template<class Impl>
+RealD  PartialFractionFermion5D<Impl>::Mdag (const FermionField &in, FermionField &out)
+{
+  M_internal(in,out,DaggerYes);
+  return norm2(out);
+}
+
+template<class Impl>
+void PartialFractionFermion5D<Impl>::Meooe       (const FermionField &in, FermionField &out)
+{
+  Meooe_internal(in,out,DaggerNo);
+}
+template<class Impl>
+void PartialFractionFermion5D<Impl>::MeooeDag    (const FermionField &in, FermionField &out)
+{
+  Meooe_internal(in,out,DaggerYes);
+}
+template<class Impl>
+void PartialFractionFermion5D<Impl>::Mooee       (const FermionField &in, FermionField &out)
+{
+  Mooee_internal(in,out,DaggerNo);
+}
+template<class Impl>
+void PartialFractionFermion5D<Impl>::MooeeDag    (const FermionField &in, FermionField &out)
+{
+  Mooee_internal(in,out,DaggerYes);
+}
+
+template<class Impl>
+void PartialFractionFermion5D<Impl>::MooeeInv    (const FermionField &in, FermionField &out)
+{
+  MooeeInv_internal(in,out,DaggerNo);
+}
+template<class Impl>
+void PartialFractionFermion5D<Impl>::MooeeInvDag (const FermionField &in, FermionField &out)
+{
+  MooeeInv_internal(in,out,DaggerYes);
+}
+
+
+// force terms; five routines; default to Dhop on diagonal
+template<class Impl>
+void PartialFractionFermion5D<Impl>::MDeriv  (GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
+{
+  int Ls = this->Ls;
+
+  FermionField D(V.Grid());
+
+  int nblock=(Ls-1)/2;
+  for(int b=0;b<nblock;b++){
+    int s = 2*b;
+    ag5xpby_ssp(D,-scale,U,0.0,U,s,s); 
+    ag5xpby_ssp(D, scale,U,0.0,U,s+1,s+1); 
+  }
+  ag5xpby_ssp(D,p[nblock]*scale/amax,U,0.0,U,Ls-1,Ls-1);
+
+  this->DhopDeriv(mat,D,V,DaggerNo); 
+};
+template<class Impl>
+void PartialFractionFermion5D<Impl>::MoeDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
+{
+  int Ls = this->Ls;
+
+  FermionField D(V.Grid());
+
+  int nblock=(Ls-1)/2;
+  for(int b=0;b<nblock;b++){
+    int s = 2*b;
+    ag5xpby_ssp(D,-scale,U,0.0,U,s,s); 
+    ag5xpby_ssp(D, scale,U,0.0,U,s+1,s+1); 
+  }
+  ag5xpby_ssp(D,p[nblock]*scale/amax,U,0.0,U,Ls-1,Ls-1);
+
+  this->DhopDerivOE(mat,D,V,DaggerNo); 
+};
+template<class Impl>
+void PartialFractionFermion5D<Impl>::MeoDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
+{
+  int Ls = this->Ls;
+
+  FermionField D(V.Grid());
+
+  int nblock=(Ls-1)/2;
+  for(int b=0;b<nblock;b++){
+    int s = 2*b;
+    ag5xpby_ssp(D,-scale,U,0.0,U,s,s); 
+    ag5xpby_ssp(D, scale,U,0.0,U,s+1,s+1); 
+  }
+  ag5xpby_ssp(D,p[nblock]*scale/amax,U,0.0,U,Ls-1,Ls-1);
+
+  this->DhopDerivEO(mat,D,V,DaggerNo); 
+};
+
+template<class Impl>
+void  PartialFractionFermion5D<Impl>::SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD scale){
+  SetCoefficientsZolotarev(1.0/scale,zdata);
+}
+template<class Impl>
+void  PartialFractionFermion5D<Impl>::SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata){
+
+  // check on degree matching
+  //      std::cout<<GridLogMessage << Ls << " Ls"<<std::endl;
+  //      std::cout<<GridLogMessage << zdata->n  << " - n"<<std::endl;
+  //      std::cout<<GridLogMessage << zdata->da << " -da "<<std::endl;
+  //      std::cout<<GridLogMessage << zdata->db << " -db"<<std::endl;
+  //      std::cout<<GridLogMessage << zdata->dn << " -dn"<<std::endl;
+  //      std::cout<<GridLogMessage << zdata->dd << " -dd"<<std::endl;
+  int Ls = this->Ls;
+
+  assert(Ls == (2*zdata->da -1) );
+
+  // Part frac
+  //      RealD R;
+  R=(1+mass)/(1-mass);
+  dw_diag = (4.0-this->M5);
+
+  //      std::vector<RealD> p; 
+  //      std::vector<RealD> q;
+  p.resize(zdata->da);
+  q.resize(zdata->dd);
+	
+  for(int n=0;n<zdata->da;n++){
+    p[n] = zdata -> alpha[n];
+  }
+  for(int n=0;n<zdata->dd;n++){
+    q[n] = -zdata -> ap[n];
+  }
+      
+  scale= part_frac_chroma_convention ? 2.0 : 1.0; // Chroma conventions annoy me
+
+  amax=zolo_hi;
+}
+
+    template<class Impl>
+    void PartialFractionFermion5D<Impl>::ExportPhysicalFermionSolution(const FermionField &solution5d,FermionField &exported4d)
+    {
+      int Ls = this->Ls;
+      conformable(solution5d.Grid(),this->FermionGrid());
+      conformable(exported4d.Grid(),this->GaugeGrid());
+      ExtractSlice(exported4d, solution5d, Ls-1, Ls-1);
+    }
+    template<class Impl>
+    void PartialFractionFermion5D<Impl>::ImportPhysicalFermionSource(const FermionField &input4d,FermionField &imported5d)
+    {
+      int Ls = this->Ls;
+      conformable(imported5d.Grid(),this->FermionGrid());
+      conformable(input4d.Grid()   ,this->GaugeGrid());
+      FermionField tmp(this->FermionGrid());
+      tmp=Zero();
+      InsertSlice(input4d, tmp, Ls-1, Ls-1);
+      tmp=Gamma(Gamma::Algebra::Gamma5)*tmp;
+      this->Dminus(tmp,imported5d);
+    }
+
+// Constructors
+template<class Impl>
+PartialFractionFermion5D<Impl>::PartialFractionFermion5D(GaugeField &_Umu,
+							 GridCartesian         &FiveDimGrid,
+							 GridRedBlackCartesian &FiveDimRedBlackGrid,
+							 GridCartesian         &FourDimGrid,
+							 GridRedBlackCartesian &FourDimRedBlackGrid,
+							 RealD _mass,RealD M5,
+							 const ImplParams &p) :
+  WilsonFermion5D<Impl>(_Umu,
+			FiveDimGrid, FiveDimRedBlackGrid,
+			FourDimGrid, FourDimRedBlackGrid,M5,p),
+  mass(_mass)
+
+{
+  int Ls = this->Ls;
+
+  assert((Ls&0x1)==1); // Odd Ls required
+  int nrational=Ls-1;
+
+
+  Approx::zolotarev_data *zdata = Approx::higham(1.0,nrational);
+
+  // NB: chroma uses a cast to "float" for the zolotarev range(!?).
+  // this creates a real difference in the operator which I do not like but we can replicate here
+  // to demonstrate compatibility
+  //      RealD eps = (zolo_lo / zolo_hi);
+  //      zdata = bfm_zolotarev(eps,nrational,0);
+      
+  SetCoefficientsTanh(zdata,1.0);
+
+  Approx::zolotarev_free(zdata);
+
+}
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/StaggeredKernelsAsm.cc b/Grid/qcd/action/fermion/implementation/StaggeredKernelsAsm.h
similarity index 93%
rename from Grid/qcd/action/fermion/StaggeredKernelsAsm.cc
rename to Grid/qcd/action/fermion/implementation/StaggeredKernelsAsm.h
index 9711c487..1a13e73a 100644
--- a/Grid/qcd/action/fermion/StaggeredKernelsAsm.cc
+++ b/Grid/qcd/action/fermion/implementation/StaggeredKernelsAsm.h
@@ -26,6 +26,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     See the full license in the file "LICENSE" in the top level distribution directory
     *************************************************************************************/
     /*  END LEGAL */
+#pragma once
+
 #include <Grid/Grid.h>
 
 #ifdef AVX512
@@ -586,11 +588,11 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
   VADD(UChi_00,UChi_10,UChi_00)				\
   VADD(UChi_01,UChi_11,UChi_01)				\
   VADD(UChi_02,UChi_12,UChi_02)	);			\
-  asm (									\
-       VSTORE(0,%0,pUChi_00)						\
-       VSTORE(1,%0,pUChi_01)						\
-       VSTORE(2,%0,pUChi_02)						\
-       : : "r" (out) : "memory" );
+  asm (							\
+  VSTORE(0,%0,pUChi_00)					\
+  VSTORE(1,%0,pUChi_01)					\
+  VSTORE(2,%0,pUChi_02)					\
+  : : "r" (out) : "memory" );
 
 // FIXME is sign right in the VSUB ?
 #define nREDUCEa(out)					\
@@ -613,20 +615,20 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
       permute##dir(Chi_1,Chi_1);\
       permute##dir(Chi_2,Chi_2);
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 template <class Impl>
 void StaggeredKernels<Impl>::DhopSiteAsm(StencilImpl &st, LebesgueOrder &lo, 
-					 DoubledGaugeField &U, DoubledGaugeField &UUU,
-					 SiteSpinor *buf, int LLs, int sU, 
-					 const FermionField &in, FermionField &out,int dag) 
+					 DoubledGaugeFieldView &U,
+					 DoubledGaugeFieldView &UUU,
+					 SiteSpinor *buf, int LLs,
+					 int sU, const FermionFieldView &in, FermionFieldView &out,int dag) 
 {
   assert(0);
 };
 
 
-//#define CONDITIONAL_MOVE(l,o,out) if ( l ) { out = (uint64_t) &in._odata[o] ; } else { out =(uint64_t) &buf[o]; }
+//#define CONDITIONAL_MOVE(l,o,out) if ( l ) { out = (uint64_t) &in[o] ; } else { out =(uint64_t) &buf[o]; }
 
 #define CONDITIONAL_MOVE(l,o,out) { const SiteSpinor *ptr = l? in_p : buf; out = (uint64_t) &ptr[o]; }
 
@@ -673,22 +675,23 @@ void StaggeredKernels<Impl>::DhopSiteAsm(StencilImpl &st, LebesgueOrder &lo,
   CONDITIONAL_MOVE(l3,o3,addr3);					\
   PF_CHI(addr3);							\
   									\
-  gauge0 =(uint64_t)&UU._odata[sU]( X );				\
-  gauge1 =(uint64_t)&UU._odata[sU]( Y );				\
-  gauge2 =(uint64_t)&UU._odata[sU]( Z );				\
-  gauge3 =(uint64_t)&UU._odata[sU]( T ); 
+  gauge0 =(uint64_t)&UU[sU]( X );				\
+  gauge1 =(uint64_t)&UU[sU]( Y );				\
+  gauge2 =(uint64_t)&UU[sU]( Z );				\
+  gauge3 =(uint64_t)&UU[sU]( T ); 
   
   // This is the single precision 5th direction vectorised kernel
 #include <Grid/simd/Intel512single.h>
 template <> void StaggeredKernels<StaggeredVec5dImplF>::DhopSiteAsm(StencilImpl &st, LebesgueOrder &lo, 
-								    DoubledGaugeField &U, DoubledGaugeField &UUU,
-								    SiteSpinor *buf, int LLs, int sU, 
-								    const FermionField &in, FermionField &out,int dag) 
+								    DoubledGaugeFieldView &U,
+								    DoubledGaugeFieldView &UUU,
+								    SiteSpinor *buf, int LLs,
+								    int sU, const FermionFieldView &in, FermionFieldView &out,int dag) 
 {
 #ifdef AVX512
   uint64_t gauge0,gauge1,gauge2,gauge3;
   uint64_t addr0,addr1,addr2,addr3;
-  const SiteSpinor *in_p; in_p = &in._odata[0];
+  const SiteSpinor *in_p; in_p = &in[0];
 
   int o0,o1,o2,o3; // offsets
   int l0,l1,l2,l3; // local 
@@ -719,7 +722,7 @@ template <> void StaggeredKernels<StaggeredVec5dImplF>::DhopSiteAsm(StencilImpl
     LOAD_CHI(addr0,addr1,addr2,addr3);
     MULT_ADD_LS(gauge0,gauge1,gauge2,gauge3);
 
-    addr0 = (uint64_t) &out._odata[sF];
+    addr0 = (uint64_t) &out[sF];
     if ( dag ) {
       nREDUCE(addr0);
     } else { 
@@ -734,14 +737,15 @@ template <> void StaggeredKernels<StaggeredVec5dImplF>::DhopSiteAsm(StencilImpl
 
 #include <Grid/simd/Intel512double.h>
 template <> void StaggeredKernels<StaggeredVec5dImplD>::DhopSiteAsm(StencilImpl &st, LebesgueOrder &lo, 
-								    DoubledGaugeField &U, DoubledGaugeField &UUU,
-								    SiteSpinor *buf, int LLs, int sU, 
-								    const FermionField &in, FermionField &out,int dag) 
+								    DoubledGaugeFieldView &U,
+								    DoubledGaugeFieldView &UUU,
+								    SiteSpinor *buf, int LLs,
+								    int sU, const FermionFieldView &in, FermionFieldView &out, int dag) 
 {
 #ifdef AVX512
   uint64_t gauge0,gauge1,gauge2,gauge3;
   uint64_t addr0,addr1,addr2,addr3;
-  const SiteSpinor *in_p; in_p = &in._odata[0];
+  const SiteSpinor *in_p; in_p = &in[0];
 
   int o0,o1,o2,o3; // offsets
   int l0,l1,l2,l3; // local 
@@ -771,7 +775,7 @@ template <> void StaggeredKernels<StaggeredVec5dImplD>::DhopSiteAsm(StencilImpl
     LOAD_CHI(addr0,addr1,addr2,addr3);
     MULT_ADD_LS(gauge0,gauge1,gauge2,gauge3);
 
-    addr0 = (uint64_t) &out._odata[sF];
+    addr0 = (uint64_t) &out[sF];
     if ( dag ) {
       nREDUCE(addr0);
     } else { 
@@ -818,14 +822,15 @@ template <> void StaggeredKernels<StaggeredVec5dImplD>::DhopSiteAsm(StencilImpl
 
 #include <Grid/simd/Intel512single.h>
 template <> void StaggeredKernels<StaggeredImplF>::DhopSiteAsm(StencilImpl &st, LebesgueOrder &lo, 
-							       DoubledGaugeField &U, DoubledGaugeField &UUU,
-							       SiteSpinor *buf, int LLs, int sU, 
-							       const FermionField &in, FermionField &out,int dag) 
+							       DoubledGaugeFieldView &U,
+							       DoubledGaugeFieldView &UUU,
+							       SiteSpinor *buf, int LLs,
+							       int sU, const FermionFieldView &in, FermionFieldView &out,int dag) 
 {
 #ifdef AVX512
   uint64_t gauge0,gauge1,gauge2,gauge3;
   uint64_t addr0,addr1,addr2,addr3;
-  const SiteSpinor *in_p; in_p = &in._odata[0];
+  const SiteSpinor *in_p; in_p = &in[0];
 
   int o0,o1,o2,o3; // offsets
   int l0,l1,l2,l3; // local 
@@ -872,7 +877,7 @@ template <> void StaggeredKernels<StaggeredImplF>::DhopSiteAsm(StencilImpl &st,
     PERMUTE23;
     MULT_ADD_XYZT(gauge2,gauge3);  
 
-    addr0 = (uint64_t) &out._odata[sF];
+    addr0 = (uint64_t) &out[sF];
     if ( dag ) { 
       nREDUCEa(addr0);
     } else { 
@@ -886,14 +891,15 @@ template <> void StaggeredKernels<StaggeredImplF>::DhopSiteAsm(StencilImpl &st,
 
 #include <Grid/simd/Intel512double.h>
 template <> void StaggeredKernels<StaggeredImplD>::DhopSiteAsm(StencilImpl &st, LebesgueOrder &lo, 
-							       DoubledGaugeField &U, DoubledGaugeField &UUU,
-							       SiteSpinor *buf, int LLs, int sU, 
-							       const FermionField &in, FermionField &out,int dag) 
+							       DoubledGaugeFieldView &U,
+							       DoubledGaugeFieldView &UUU,
+							       SiteSpinor *buf, int LLs,
+							       int sU, const FermionFieldView &in, FermionFieldView &out,int dag) 
 {
 #ifdef AVX512
   uint64_t gauge0,gauge1,gauge2,gauge3;
   uint64_t addr0,addr1,addr2,addr3;
-  const SiteSpinor *in_p; in_p = &in._odata[0];
+  const SiteSpinor *in_p; in_p = &in[0];
 
   int o0,o1,o2,o3; // offsets
   int l0,l1,l2,l3; // local 
@@ -940,7 +946,7 @@ template <> void StaggeredKernels<StaggeredImplD>::DhopSiteAsm(StencilImpl &st,
     PERMUTE23;
     MULT_ADD_XYZT(gauge2,gauge3);  
     
-    addr0 = (uint64_t) &out._odata[sF];
+    addr0 = (uint64_t) &out[sF];
     if ( dag ) {
       nREDUCEa(addr0);
     } else { 
@@ -952,17 +958,5 @@ template <> void StaggeredKernels<StaggeredImplD>::DhopSiteAsm(StencilImpl &st,
 #endif
 }
 
-#define KERNEL_INSTANTIATE(CLASS,FUNC,IMPL)			    \
-  template void CLASS<IMPL>::FUNC(StencilImpl &st, LebesgueOrder &lo,	\
-				  DoubledGaugeField &U,			\
-				  DoubledGaugeField &UUU,		\
-				  SiteSpinor *buf, int LLs,		\
-				  int sU, const FermionField &in, FermionField &out,int dag);
-
-KERNEL_INSTANTIATE(StaggeredKernels,DhopSiteAsm,StaggeredImplD);
-KERNEL_INSTANTIATE(StaggeredKernels,DhopSiteAsm,StaggeredImplF);
-KERNEL_INSTANTIATE(StaggeredKernels,DhopSiteAsm,StaggeredVec5dImplD);
-KERNEL_INSTANTIATE(StaggeredKernels,DhopSiteAsm,StaggeredVec5dImplF);
-
-}}
+NAMESPACE_END(Grid);
 
diff --git a/Grid/qcd/action/fermion/StaggeredKernelsHand.cc b/Grid/qcd/action/fermion/implementation/StaggeredKernelsHand.h
similarity index 86%
rename from Grid/qcd/action/fermion/StaggeredKernelsHand.cc
rename to Grid/qcd/action/fermion/implementation/StaggeredKernelsHand.h
index f304b00f..62370d75 100644
--- a/Grid/qcd/action/fermion/StaggeredKernelsHand.cc
+++ b/Grid/qcd/action/fermion/implementation/StaggeredKernelsHand.h
@@ -28,6 +28,9 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     /*  END LEGAL */
 #include <Grid/Grid.h>
 
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
 
 #define LOAD_CHI(b)		\
   const SiteSpinor & ref (b[offset]);	\
@@ -38,7 +41,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 // To splat or not to splat depends on the implementation
 #define MULT(A,UChi)				\
-  auto & ref(U._odata[sU](A));			\
+  auto & ref(U[sU](A));			\
    Impl::loadLinkElement(U_00,ref()(0,0));      \
    Impl::loadLinkElement(U_10,ref()(1,0));      \
    Impl::loadLinkElement(U_20,ref()(2,0));      \
@@ -59,7 +62,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     UChi ## _2 += U_22*Chi_2;
 
 #define MULT_ADD(U,A,UChi)			\
-  auto & ref(U._odata[sU](A));			\
+  auto & ref(U[sU](A));			\
    Impl::loadLinkElement(U_00,ref()(0,0));      \
    Impl::loadLinkElement(U_10,ref()(1,0));      \
    Impl::loadLinkElement(U_20,ref()(2,0));      \
@@ -92,7 +95,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
   local  = SE->_is_local;		\
   perm   = SE->_permute;		\
   if ( local ) {						\
-    LOAD_CHI(in._odata);					\
+    LOAD_CHI(in);					\
     if ( perm) {						\
       PERMUTE_DIR(Perm);					\
     }								\
@@ -120,14 +123,14 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
   local  = SE->_is_local;				\
   perm   = SE->_permute;				\
   if ( local ) {					\
-    LOAD_CHI(in._odata);				\
+    LOAD_CHI(in);				\
     if ( perm) {					\
       PERMUTE_DIR(Perm);				\
     }							\
   } else if ( st.same_node[Dir] ) {			\
     LOAD_CHI(buf);					\
   }							\
-  if (SE->_is_local || st.same_node[Dir] ) {		\
+  if (local || st.same_node[Dir] ) {		\
     MULT_ADD(U,Dir,even);				\
   }
 
@@ -135,22 +138,18 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
   SE=st.GetEntry(ptype,Dir+skew,sF);			\
   offset = SE->_offset;					\
   local  = SE->_is_local;				\
-  perm   = SE->_permute;				\
-  if ((!SE->_is_local) && (!st.same_node[Dir]) ) {		\
+  if ((!local) && (!st.same_node[Dir]) ) {		\
     nmu++;							\
     { LOAD_CHI(buf);	  }					\
     { MULT_ADD(U,Dir,even); }					\
   }								
 
-namespace Grid {
-namespace QCD {
-
 
 template <class Impl>
 void StaggeredKernels<Impl>::DhopSiteHand(StencilImpl &st, LebesgueOrder &lo, 
-					  DoubledGaugeField &U,DoubledGaugeField &UUU,
+					  DoubledGaugeFieldView &U,DoubledGaugeFieldView &UUU,
 					  SiteSpinor *buf, int LLs, int sU, 
-					  const FermionField &in, FermionField &out,int dag) 
+					  const FermionFieldView &in, FermionFieldView &out,int dag) 
 {
   typedef typename Simd::scalar_type S;
   typedef typename Simd::vector_type V;
@@ -213,16 +212,16 @@ void StaggeredKernels<Impl>::DhopSiteHand(StencilImpl &st, LebesgueOrder &lo,
       result()()(1) = even_1 + odd_1;
       result()()(2) = even_2 + odd_2;
     }
-    vstream(out._odata[sF],result);
+    vstream(out[sF],result);
   }
 }
 
 
 template <class Impl>
 void StaggeredKernels<Impl>::DhopSiteHandInt(StencilImpl &st, LebesgueOrder &lo, 
-					     DoubledGaugeField &U, DoubledGaugeField &UUU,
+					     DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU,
 					     SiteSpinor *buf, int LLs, int sU, 
-					     const FermionField &in, FermionField &out,int dag) 
+					     const FermionFieldView &in, FermionFieldView &out,int dag) 
 {
   typedef typename Simd::scalar_type S;
   typedef typename Simd::vector_type V;
@@ -249,7 +248,7 @@ void StaggeredKernels<Impl>::DhopSiteHandInt(StencilImpl &st, LebesgueOrder &lo,
   Simd U_22; 
 
   SiteSpinor result;
-  int offset,local,perm, ptype;
+  int offset, ptype, local, perm;
 
   StencilEntry *SE;
   int skew;
@@ -257,8 +256,8 @@ void StaggeredKernels<Impl>::DhopSiteHandInt(StencilImpl &st, LebesgueOrder &lo,
   for(int s=0;s<LLs;s++){
     int sF=s+LLs*sU;
 
-    even_0 = zero;    even_1 = zero;    even_2 = zero;
-     odd_0 = zero;     odd_1 = zero;     odd_2 = zero;
+    even_0 = Zero();    even_1 = Zero();    even_2 = Zero();
+     odd_0 = Zero();     odd_1 = Zero();     odd_2 = Zero();
 
     skew = 0;
     HAND_STENCIL_LEG_INT(U,Xp,3,skew,even);  
@@ -289,16 +288,16 @@ void StaggeredKernels<Impl>::DhopSiteHandInt(StencilImpl &st, LebesgueOrder &lo,
       result()()(1) = even_1 + odd_1;
       result()()(2) = even_2 + odd_2;
     }
-    vstream(out._odata[sF],result);
+    vstream(out[sF],result);
   }
 }
 
 
 template <class Impl>
 void StaggeredKernels<Impl>::DhopSiteHandExt(StencilImpl &st, LebesgueOrder &lo, 
-					     DoubledGaugeField &U, DoubledGaugeField &UUU,
+					     DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU,
 					     SiteSpinor *buf, int LLs, int sU, 
-					     const FermionField &in, FermionField &out,int dag) 
+					     const FermionFieldView &in, FermionFieldView &out,int dag) 
 {
   typedef typename Simd::scalar_type S;
   typedef typename Simd::vector_type V;
@@ -325,7 +324,7 @@ void StaggeredKernels<Impl>::DhopSiteHandExt(StencilImpl &st, LebesgueOrder &lo,
   Simd U_22; 
 
   SiteSpinor result;
-  int offset,local,perm, ptype;
+  int offset, ptype, local;
 
   StencilEntry *SE;
   int skew;
@@ -333,8 +332,8 @@ void StaggeredKernels<Impl>::DhopSiteHandExt(StencilImpl &st, LebesgueOrder &lo,
   for(int s=0;s<LLs;s++){
     int sF=s+LLs*sU;
 
-    even_0 = zero;    even_1 = zero;    even_2 = zero;
-     odd_0 = zero;     odd_1 = zero;     odd_2 = zero;
+    even_0 = Zero();    even_1 = Zero();    even_2 = Zero();
+     odd_0 = Zero();     odd_1 = Zero();     odd_2 = Zero();
     int nmu=0;
     skew = 0;
     HAND_STENCIL_LEG_EXT(U,Xp,3,skew,even);  
@@ -366,34 +365,29 @@ void StaggeredKernels<Impl>::DhopSiteHandExt(StencilImpl &st, LebesgueOrder &lo,
 	result()()(1) = even_1 + odd_1;
 	result()()(2) = even_2 + odd_2;
       }
-      out._odata[sF] = out._odata[sF] + result;
+      out[sF] = out[sF] + result;
     }
   }
 }
 
-
 #define DHOP_SITE_HAND_INSTANTIATE(IMPL)				\
   template void StaggeredKernels<IMPL>::DhopSiteHand(StencilImpl &st, LebesgueOrder &lo, \
-						     DoubledGaugeField &U,DoubledGaugeField &UUU, \
+						     DoubledGaugeFieldView &U,DoubledGaugeFieldView &UUU, \
 						     SiteSpinor *buf, int LLs, int sU, \
-						     const FermionField &in, FermionField &out, int dag); \
+						     const FermionFieldView &in, FermionFieldView &out, int dag); \
 									\
   template void StaggeredKernels<IMPL>::DhopSiteHandInt(StencilImpl &st, LebesgueOrder &lo, \
-						     DoubledGaugeField &U,DoubledGaugeField &UUU, \
+						     DoubledGaugeFieldView &U,DoubledGaugeFieldView &UUU, \
 						     SiteSpinor *buf, int LLs, int sU, \
-						     const FermionField &in, FermionField &out, int dag); \
+						     const FermionFieldView &in, FermionFieldView &out, int dag); \
 									\
   template void StaggeredKernels<IMPL>::DhopSiteHandExt(StencilImpl &st, LebesgueOrder &lo, \
-						     DoubledGaugeField &U,DoubledGaugeField &UUU, \
+						     DoubledGaugeFieldView &U,DoubledGaugeFieldView &UUU, \
 						     SiteSpinor *buf, int LLs, int sU, \
-						     const FermionField &in, FermionField &out, int dag); \
+						     const FermionFieldView &in, FermionFieldView &out, int dag); \
 
-DHOP_SITE_HAND_INSTANTIATE(StaggeredImplD);
-DHOP_SITE_HAND_INSTANTIATE(StaggeredImplF);
-DHOP_SITE_HAND_INSTANTIATE(StaggeredVec5dImplD);
-DHOP_SITE_HAND_INSTANTIATE(StaggeredVec5dImplF);
+#undef LOAD_CHI
+
+NAMESPACE_END(Grid);
 
 
-}
-}
-
diff --git a/Grid/qcd/action/fermion/StaggeredKernels.cc b/Grid/qcd/action/fermion/implementation/StaggeredKernelsImplementation.h
similarity index 83%
rename from Grid/qcd/action/fermion/StaggeredKernels.cc
rename to Grid/qcd/action/fermion/implementation/StaggeredKernelsImplementation.h
index b7e568c2..d301556c 100644
--- a/Grid/qcd/action/fermion/StaggeredKernels.cc
+++ b/Grid/qcd/action/fermion/implementation/StaggeredKernelsImplementation.h
@@ -28,40 +28,38 @@ directory
 /*  END LEGAL */
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
-namespace Grid {
-namespace QCD {
+#pragma once
 
-int StaggeredKernelsStatic::Opt= StaggeredKernelsStatic::OptGeneric;
-int StaggeredKernelsStatic::Comms = StaggeredKernelsStatic::CommsAndCompute;
+NAMESPACE_BEGIN(Grid);
 
 #define GENERIC_STENCIL_LEG(U,Dir,skew,multLink)		\
   SE = st.GetEntry(ptype, Dir+skew, sF);			\
   if (SE->_is_local ) {						\
     if (SE->_permute) {						\
       chi_p = &chi;						\
-      permute(chi,  in._odata[SE->_offset], ptype);		\
+      permute(chi,  in[SE->_offset], ptype);		\
     } else {							\
-      chi_p = &in._odata[SE->_offset];				\
+      chi_p = &in[SE->_offset];				\
     }								\
   } else {							\
     chi_p = &buf[SE->_offset];					\
   }								\
-  multLink(Uchi, U._odata[sU], *chi_p, Dir);			
+  multLink(Uchi, U[sU], *chi_p, Dir);			
 
 #define GENERIC_STENCIL_LEG_INT(U,Dir,skew,multLink)		\
   SE = st.GetEntry(ptype, Dir+skew, sF);			\
   if (SE->_is_local ) {						\
     if (SE->_permute) {						\
       chi_p = &chi;						\
-      permute(chi,  in._odata[SE->_offset], ptype);		\
+      permute(chi,  in[SE->_offset], ptype);		\
     } else {							\
-      chi_p = &in._odata[SE->_offset];				\
+      chi_p = &in[SE->_offset];				\
     }								\
   } else if ( st.same_node[Dir] ) {				\
     chi_p = &buf[SE->_offset];					\
   }								\
   if (SE->_is_local || st.same_node[Dir] ) {			\
-    multLink(Uchi, U._odata[sU], *chi_p, Dir);			\
+    multLink(Uchi, U[sU], *chi_p, Dir);			\
   }
 
 #define GENERIC_STENCIL_LEG_EXT(U,Dir,skew,multLink)		\
@@ -69,7 +67,7 @@ int StaggeredKernelsStatic::Comms = StaggeredKernelsStatic::CommsAndCompute;
   if ((!SE->_is_local) && (!st.same_node[Dir]) ) {		\
     nmu++;							\
     chi_p = &buf[SE->_offset];					\
-    multLink(Uchi, U._odata[sU], *chi_p, Dir);			\
+    multLink(Uchi, U[sU], *chi_p, Dir);			\
   }
 
 template <class Impl>
@@ -81,9 +79,9 @@ StaggeredKernels<Impl>::StaggeredKernels(const ImplParams &p) : Base(p){};
 ////////////////////////////////////////////////////////////////////////////////////
 template <class Impl>
 void StaggeredKernels<Impl>::DhopSiteGeneric(StencilImpl &st, LebesgueOrder &lo, 
-					     DoubledGaugeField &U, DoubledGaugeField &UUU,
+					     DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU,
 					     SiteSpinor *buf, int LLs, int sU, 
-					     const FermionField &in, FermionField &out, int dag) {
+					     const FermionFieldView &in, FermionFieldView &out, int dag) {
   const SiteSpinor *chi_p;
   SiteSpinor chi;
   SiteSpinor Uchi;
@@ -114,7 +112,7 @@ void StaggeredKernels<Impl>::DhopSiteGeneric(StencilImpl &st, LebesgueOrder &lo,
     if ( dag ) { 
       Uchi = - Uchi;
     } 
-    vstream(out._odata[sF], Uchi);
+    vstream(out[sF], Uchi);
   }
 };
 
@@ -123,9 +121,9 @@ void StaggeredKernels<Impl>::DhopSiteGeneric(StencilImpl &st, LebesgueOrder &lo,
   ///////////////////////////////////////////////////
 template <class Impl>
 void StaggeredKernels<Impl>::DhopSiteGenericInt(StencilImpl &st, LebesgueOrder &lo, 
-						DoubledGaugeField &U, DoubledGaugeField &UUU,
+						DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU,
 						SiteSpinor *buf, int LLs, int sU, 
-						const FermionField &in, FermionField &out,int dag) {
+						const FermionFieldView &in, FermionFieldView &out,int dag) {
   const SiteSpinor *chi_p;
   SiteSpinor chi;
   SiteSpinor Uchi;
@@ -136,7 +134,7 @@ void StaggeredKernels<Impl>::DhopSiteGenericInt(StencilImpl &st, LebesgueOrder &
   for(int s=0;s<LLs;s++){
     int sF=LLs*sU+s;
     skew = 0;
-    Uchi=zero;
+    Uchi=Zero();
     GENERIC_STENCIL_LEG_INT(U,Xp,skew,Impl::multLinkAdd);
     GENERIC_STENCIL_LEG_INT(U,Yp,skew,Impl::multLinkAdd);
     GENERIC_STENCIL_LEG_INT(U,Zp,skew,Impl::multLinkAdd);
@@ -157,7 +155,7 @@ void StaggeredKernels<Impl>::DhopSiteGenericInt(StencilImpl &st, LebesgueOrder &
     if ( dag ) {
       Uchi = - Uchi;
     }
-    vstream(out._odata[sF], Uchi);
+    vstream(out[sF], Uchi);
   }
 };
 
@@ -167,11 +165,11 @@ void StaggeredKernels<Impl>::DhopSiteGenericInt(StencilImpl &st, LebesgueOrder &
   ///////////////////////////////////////////////////
 template <class Impl>
 void StaggeredKernels<Impl>::DhopSiteGenericExt(StencilImpl &st, LebesgueOrder &lo, 
-						DoubledGaugeField &U, DoubledGaugeField &UUU,
+						DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU,
 						SiteSpinor *buf, int LLs, int sU,
-						const FermionField &in, FermionField &out,int dag) {
+						const FermionFieldView &in, FermionFieldView &out,int dag) {
   const SiteSpinor *chi_p;
-  SiteSpinor chi;
+  //  SiteSpinor chi;
   SiteSpinor Uchi;
   StencilEntry *SE;
   int ptype;
@@ -181,7 +179,7 @@ void StaggeredKernels<Impl>::DhopSiteGenericExt(StencilImpl &st, LebesgueOrder &
   for(int s=0;s<LLs;s++){
     int sF=LLs*sU+s;
     skew = 0;
-    Uchi=zero;
+    Uchi=Zero();
     GENERIC_STENCIL_LEG_EXT(U,Xp,skew,Impl::multLinkAdd);
     GENERIC_STENCIL_LEG_EXT(U,Yp,skew,Impl::multLinkAdd);
     GENERIC_STENCIL_LEG_EXT(U,Zp,skew,Impl::multLinkAdd);
@@ -202,9 +200,9 @@ void StaggeredKernels<Impl>::DhopSiteGenericExt(StencilImpl &st, LebesgueOrder &
 
     if ( nmu ) { 
       if ( dag ) { 
-	out._odata[sF] = out._odata[sF] - Uchi;
+	out[sF] = out[sF] - Uchi;
       } else { 
-	out._odata[sF] = out._odata[sF] + Uchi;
+	out[sF] = out[sF] + Uchi;
       }
     }
   }
@@ -215,9 +213,9 @@ void StaggeredKernels<Impl>::DhopSiteGenericExt(StencilImpl &st, LebesgueOrder &
 ////////////////////////////////////////////////////////////////////////////////////
 
 template <class Impl>
-void StaggeredKernels<Impl>::DhopSiteDag(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, DoubledGaugeField &UUU,
+void StaggeredKernels<Impl>::DhopSiteDag(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU,
 					 SiteSpinor *buf, int LLs, int sU,
-					 const FermionField &in, FermionField &out,
+					 const FermionFieldView &in, FermionFieldView &out,
 					 int interior,int exterior)
 {
   int dag=1;
@@ -225,9 +223,9 @@ void StaggeredKernels<Impl>::DhopSiteDag(StencilImpl &st, LebesgueOrder &lo, Dou
 };
 
 template <class Impl>
-void StaggeredKernels<Impl>::DhopSite(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, DoubledGaugeField &UUU,
+void StaggeredKernels<Impl>::DhopSite(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU,
 				      SiteSpinor *buf, int LLs, int sU,
-				      const FermionField &in, FermionField &out,
+				      const FermionFieldView &in, FermionFieldView &out,
 				      int interior,int exterior)
 {
   int dag=0;
@@ -235,9 +233,9 @@ void StaggeredKernels<Impl>::DhopSite(StencilImpl &st, LebesgueOrder &lo, Double
 };
 
 template <class Impl>
-void StaggeredKernels<Impl>::DhopSite(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U, DoubledGaugeField &UUU,
+void StaggeredKernels<Impl>::DhopSite(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeFieldView &U, DoubledGaugeFieldView &UUU,
 				      SiteSpinor *buf, int LLs,
-				      int sU, const FermionField &in, FermionField &out,
+				      int sU, const FermionFieldView &in, FermionFieldView &out,
 				      int dag,int interior,int exterior) 
 {
   switch(Opt) {
@@ -277,8 +275,8 @@ void StaggeredKernels<Impl>::DhopSite(StencilImpl &st, LebesgueOrder &lo, Double
 };
 
 template <class Impl>
-void StaggeredKernels<Impl>::DhopDir( StencilImpl &st, DoubledGaugeField &U,  DoubledGaugeField &UUU, SiteSpinor *buf, int sF,
-				      int sU, const FermionField &in, FermionField &out, int dir, int disp) 
+void StaggeredKernels<Impl>::DhopDirKernel( StencilImpl &st, DoubledGaugeFieldView &U,  DoubledGaugeFieldView &UUU, SiteSpinor *buf, int sF,
+					    int sU, const FermionFieldView &in, FermionFieldView &out, int dir, int disp) 
 {
   // Disp should be either +1,-1,+3,-3
   // What about "dag" ?
@@ -287,8 +285,6 @@ void StaggeredKernels<Impl>::DhopDir( StencilImpl &st, DoubledGaugeField &U,  Do
   assert(0);
 }
 
-FermOpStaggeredTemplateInstantiate(StaggeredKernels);
-FermOpStaggeredVec5dTemplateInstantiate(StaggeredKernels);
+NAMESPACE_END(Grid);
 
-}}
 
diff --git a/Grid/qcd/action/fermion/WilsonCloverFermion.cc b/Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h
similarity index 85%
rename from Grid/qcd/action/fermion/WilsonCloverFermion.cc
rename to Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h
index ed06c673..5744d3bb 100644
--- a/Grid/qcd/action/fermion/WilsonCloverFermion.cc
+++ b/Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h
@@ -26,23 +26,21 @@
     See the full license in the file "LICENSE" in the top level distribution directory
     *************************************************************************************/
 /*  END LEGAL */
-#include <Grid/Grid.h>
-//#include <Grid/Eigen/Dense>
-#include <Grid/qcd/spin/Dirac.h>
 
-namespace Grid
-{
-namespace QCD
-{
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+
+NAMESPACE_BEGIN(Grid);
 
 // *NOT* EO
 template <class Impl>
 RealD WilsonCloverFermion<Impl>::M(const FermionField &in, FermionField &out)
 {
-  FermionField temp(out._grid);
+  FermionField temp(out.Grid());
 
   // Wilson term
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   this->Dhop(in, out, DaggerNo);
 
   // Clover term
@@ -55,10 +53,10 @@ RealD WilsonCloverFermion<Impl>::M(const FermionField &in, FermionField &out)
 template <class Impl>
 RealD WilsonCloverFermion<Impl>::Mdag(const FermionField &in, FermionField &out)
 {
-  FermionField temp(out._grid);
+  FermionField temp(out.Grid());
 
   // Wilson term
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   this->Dhop(in, out, DaggerYes);
 
   // Clover term
@@ -72,7 +70,7 @@ template <class Impl>
 void WilsonCloverFermion<Impl>::ImportGauge(const GaugeField &_Umu)
 {
   WilsonFermion<Impl>::ImportGauge(_Umu);
-  GridBase *grid = _Umu._grid;
+  GridBase *grid = _Umu.Grid();
   typename Impl::GaugeLinkField Bx(grid), By(grid), Bz(grid), Ex(grid), Ey(grid), Ez(grid);
 
   // Compute the field strength terms mu>nu
@@ -93,27 +91,29 @@ void WilsonCloverFermion<Impl>::ImportGauge(const GaugeField &_Umu)
   CloverTerm += fillCloverZT(Ez) * csw_t;
   CloverTerm += diag_mass;
 
-  int lvol = _Umu._grid->lSites();
+  int lvol = _Umu.Grid()->lSites();
   int DimRep = Impl::Dimension;
 
   Eigen::MatrixXcd EigenCloverOp = Eigen::MatrixXcd::Zero(Ns * DimRep, Ns * DimRep);
   Eigen::MatrixXcd EigenInvCloverOp = Eigen::MatrixXcd::Zero(Ns * DimRep, Ns * DimRep);
 
-  std::vector<int> lcoor;
-  typename SiteCloverType::scalar_object Qx = zero, Qxinv = zero;
+  Coordinate lcoor;
+  typename SiteCloverType::scalar_object Qx = Zero(), Qxinv = Zero();
 
   for (int site = 0; site < lvol; site++)
   {
     grid->LocalIndexToLocalCoor(site, lcoor);
     EigenCloverOp = Eigen::MatrixXcd::Zero(Ns * DimRep, Ns * DimRep);
     peekLocalSite(Qx, CloverTerm, lcoor);
-    Qxinv = zero;
+    Qxinv = Zero();
     //if (csw!=0){
     for (int j = 0; j < Ns; j++)
       for (int k = 0; k < Ns; k++)
         for (int a = 0; a < DimRep; a++)
-          for (int b = 0; b < DimRep; b++)
-            EigenCloverOp(a + j * DimRep, b + k * DimRep) = Qx()(j, k)(a, b);
+          for (int b = 0; b < DimRep; b++){
+	    auto zz =  Qx()(j, k)(a, b);
+            EigenCloverOp(a + j * DimRep, b + k * DimRep) = std::complex<double>(zz);
+	  }
     //   if (site==0) std::cout << "site =" << site << "\n" << EigenCloverOp << std::endl;
 
     EigenInvCloverOp = EigenCloverOp.inverse();
@@ -169,15 +169,15 @@ void WilsonCloverFermion<Impl>::MooeeInvDag(const FermionField &in, FermionField
 template <class Impl>
 void WilsonCloverFermion<Impl>::MooeeInternal(const FermionField &in, FermionField &out, int dag, int inv)
 {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   CloverFieldType *Clover;
-  assert(in.checkerboard == Odd || in.checkerboard == Even);
+  assert(in.Checkerboard() == Odd || in.Checkerboard() == Even);
 
   if (dag)
   {
-    if (in._grid->_isCheckerBoarded)
+    if (in.Grid()->_isCheckerBoarded)
     {
-      if (in.checkerboard == Odd)
+      if (in.Checkerboard() == Odd)
       {
         Clover = (inv) ? &CloverTermInvDagOdd : &CloverTermDagOdd;
       }
@@ -195,10 +195,10 @@ void WilsonCloverFermion<Impl>::MooeeInternal(const FermionField &in, FermionFie
   }
   else
   {
-    if (in._grid->_isCheckerBoarded)
+    if (in.Grid()->_isCheckerBoarded)
     {
 
-      if (in.checkerboard == Odd)
+      if (in.Checkerboard() == Odd)
       {
         //  std::cout << "Calling clover term Odd" << std::endl;
         Clover = (inv) ? &CloverTermInvOdd : &CloverTermOdd;
@@ -209,7 +209,7 @@ void WilsonCloverFermion<Impl>::MooeeInternal(const FermionField &in, FermionFie
         Clover = (inv) ? &CloverTermInvEven : &CloverTermEven;
       }
       out = *Clover * in;
-      //  std::cout << GridLogMessage << "*Clover.checkerboard "  << (*Clover).checkerboard << std::endl;
+      //  std::cout << GridLogMessage << "*Clover.Checkerboard() "  << (*Clover).Checkerboard() << std::endl;
     }
     else
     {
@@ -235,9 +235,4 @@ void WilsonCloverFermion<Impl>::MeeDeriv(GaugeField &mat, const FermionField &U,
   assert(0); // not implemented yet
 }
 
-FermOpTemplateInstantiate(WilsonCloverFermion);
-AdjointFermOpTemplateInstantiate(WilsonCloverFermion);
-TwoIndexFermOpTemplateInstantiate(WilsonCloverFermion);
-//GparityFermOpTemplateInstantiate(WilsonCloverFermion);
-}
-}
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/WilsonFermion5D.cc b/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
similarity index 74%
rename from Grid/qcd/action/fermion/WilsonFermion5D.cc
rename to Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
index fb3fd861..1bdc9a64 100644
--- a/Grid/qcd/action/fermion/WilsonFermion5D.cc
+++ b/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
@@ -36,13 +36,8 @@ Author: Vera Guelpers <V.M.Guelpers@soton.ac.uk>
 #include <Grid/qcd/action/fermion/WilsonFermion5D.h>
 #include <Grid/perfmon/PerfCount.h>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
   
-// S-direction is INNERMOST and takes no part in the parity.
-const std::vector<int> WilsonFermion5DStatic::directions   ({1,2,3,4, 1, 2, 3, 4});
-const std::vector<int> WilsonFermion5DStatic::displacements({1,1,1,1,-1,-1,-1,-1});
-
   // 5d lattice for DWF.
 template<class Impl>
 WilsonFermion5D<Impl>::WilsonFermion5D(GaugeField &_Umu,
@@ -56,9 +51,9 @@ WilsonFermion5D<Impl>::WilsonFermion5D(GaugeField &_Umu,
   _FiveDimRedBlackGrid(&FiveDimRedBlackGrid),
   _FourDimGrid        (&FourDimGrid),
   _FourDimRedBlackGrid(&FourDimRedBlackGrid),
-  Stencil    (_FiveDimGrid,npoint,Even,directions,displacements),
-  StencilEven(_FiveDimRedBlackGrid,npoint,Even,directions,displacements), // source is Even
-  StencilOdd (_FiveDimRedBlackGrid,npoint,Odd ,directions,displacements), // source is Odd
+  Stencil    (_FiveDimGrid,npoint,Even,directions,displacements,p),
+  StencilEven(_FiveDimRedBlackGrid,npoint,Even,directions,displacements,p), // source is Even
+  StencilOdd (_FiveDimRedBlackGrid,npoint,Odd ,directions,displacements,p), // source is Odd
   M5(_M5),
   Umu(_FourDimGrid),
   UmuEven(_FourDimRedBlackGrid),
@@ -105,8 +100,8 @@ WilsonFermion5D<Impl>::WilsonFermion5D(GaugeField &_Umu,
     assert(FiveDimRedBlackGrid._simd_layout[0]==nsimd);
 
     for(int d=0;d<4;d++){
-      assert(FourDimGrid._simd_layout[d]=1);
-      assert(FourDimRedBlackGrid._simd_layout[d]=1);
+      assert(FourDimGrid._simd_layout[d]==1);
+      assert(FourDimRedBlackGrid._simd_layout[d]==1);
       assert(FiveDimRedBlackGrid._simd_layout[d+1]==1);
     }
 
@@ -141,7 +136,7 @@ void WilsonFermion5D<Impl>::Report(void)
   RealD NP     = _FourDimGrid->_Nprocessors;
   RealD NN     = _FourDimGrid->NodeCount();
   RealD volume = Ls;  
-  std::vector<int> latt = _FourDimGrid->GlobalDimensions();
+  Coordinate latt = _FourDimGrid->GlobalDimensions();
   for(int mu=0;mu<Nd;mu++) volume=volume*latt[mu];
 
   if ( DhopCalls > 0 ) {
@@ -221,7 +216,7 @@ void WilsonFermion5D<Impl>::ZeroCounters(void) {
 template<class Impl>
 void WilsonFermion5D<Impl>::ImportGauge(const GaugeField &_Umu)
 {
-  GaugeField HUmu(_Umu._grid);
+  GaugeField HUmu(_Umu.Grid());
   HUmu = _Umu*(-0.5);
   Impl::DoubleStore(GaugeGrid(),Umu,HUmu);
   pickCheckerboard(Even,UmuEven,Umu);
@@ -235,51 +230,43 @@ void WilsonFermion5D<Impl>::DhopDir(const FermionField &in, FermionField &out,in
   //  assert( (disp==1)||(disp==-1) );
   //  assert( (dir>=0)&&(dir<4) ); //must do x,y,z or t;
 
-  Compressor compressor(DaggerNo);
-  Stencil.HaloExchange(in,compressor);
-  
   int skip = (disp==1) ? 0 : 1;
-
   int dirdisp = dir+skip*4;
   int gamma   = dir+(1-skip)*4;
 
-  assert(dirdisp<=7);
-  assert(dirdisp>=0);
+  Compressor compressor(DaggerNo);
+  Stencil.HaloExchange(in,compressor);
+  
+  uint64_t Nsite = Umu.Grid()->oSites();
+  Kernels::DhopDirKernel(Stencil,Umu,Stencil.CommBuf(),Ls,Nsite,in,out,dirdisp,gamma);
 
-  parallel_for(int ss=0;ss<Umu._grid->oSites();ss++){
-    for(int s=0;s<Ls;s++){
-      int sU=ss;
-      int sF = s+Ls*sU; 
-      Kernels::DhopDir(Stencil,Umu,Stencil.CommBuf(),sF,sU,in,out,dirdisp,gamma);
-    }
-  }
 };
 
 template<class Impl>
 void WilsonFermion5D<Impl>::DerivInternal(StencilImpl & st,
-            DoubledGaugeField & U,
-            GaugeField &mat,
-            const FermionField &A,
-            const FermionField &B,
-            int dag)
+					  DoubledGaugeField & U,
+					  GaugeField &mat,
+					  const FermionField &A,
+					  const FermionField &B,
+					  int dag)
 {
   DerivCalls++;
   assert((dag==DaggerNo) ||(dag==DaggerYes));
 
-  conformable(st._grid,A._grid);
-  conformable(st._grid,B._grid);
+  conformable(st.Grid(),A.Grid());
+  conformable(st.Grid(),B.Grid());
 
   Compressor compressor(dag);
   
-  FermionField Btilde(B._grid);
-  FermionField Atilde(B._grid);
+  FermionField Btilde(B.Grid());
+  FermionField Atilde(B.Grid());
 
   DerivCommTime-=usecond();
   st.HaloExchange(B,compressor);
   DerivCommTime+=usecond();
 
   Atilde=A;
-  int LLs = B._grid->_rdimensions[0];
+  int LLs = B.Grid()->_rdimensions[0];
 
 
   DerivComputeTime-=usecond();
@@ -295,21 +282,11 @@ void WilsonFermion5D<Impl>::DerivInternal(StencilImpl & st,
     ////////////////////////
 
     DerivDhopComputeTime -= usecond();
-    parallel_for (int sss = 0; sss < U._grid->oSites(); sss++) {
-      for (int s = 0; s < Ls; s++) {
-        int sU = sss;
-        int sF = s + Ls * sU;
 
-        assert(sF < B._grid->oSites());
-        assert(sU < U._grid->oSites());
+    int Usites = U.Grid()->oSites();
 
-        Kernels::DhopDir(st, U, st.CommBuf(), sF, sU, B, Btilde, mu, gamma);
+    Kernels::DhopDirKernel(st, U, st.CommBuf(), Ls, Usites, B, Btilde, mu,gamma);
 
-        ////////////////////////////
-        // spin trace outer product
-        ////////////////////////////
-      }
-    }
     ////////////////////////////
     // spin trace outer product
     ////////////////////////////
@@ -325,12 +302,13 @@ void WilsonFermion5D<Impl>::DhopDeriv(GaugeField &mat,
                                       const FermionField &B,
                                       int dag)
 {
-  conformable(A._grid,FermionGrid());  
-  conformable(A._grid,B._grid);
+  conformable(A.Grid(),FermionGrid());  
+  conformable(A.Grid(),B.Grid());
 
-  //conformable(GaugeGrid(),mat._grid);// this is not general! leaving as a comment
+  //conformable(GaugeGrid(),mat.Grid());// this is not general! leaving as a comment
 
-  mat.checkerboard = A.checkerboard;
+  mat.Checkerboard() = A.Checkerboard();
+  //  mat.checkerboard = A.checkerboard;
 
   DerivInternal(Stencil,Umu,mat,A,B,dag);
 }
@@ -341,12 +319,12 @@ void WilsonFermion5D<Impl>::DhopDerivEO(GaugeField &mat,
                                         const FermionField &B,
                                         int dag)
 {
-  conformable(A._grid,FermionRedBlackGrid());
-  conformable(A._grid,B._grid);
+  conformable(A.Grid(),FermionRedBlackGrid());
+  conformable(A.Grid(),B.Grid());
 
-  assert(B.checkerboard==Odd);
-  assert(A.checkerboard==Even);
-  mat.checkerboard = Even;
+  assert(B.Checkerboard()==Odd);
+  assert(A.Checkerboard()==Even);
+  mat.Checkerboard() = Even;
 
   DerivInternal(StencilOdd,UmuEven,mat,A,B,dag);
 }
@@ -358,12 +336,12 @@ void WilsonFermion5D<Impl>::DhopDerivOE(GaugeField &mat,
                                         const FermionField &B,
                                         int dag)
 {
-  conformable(A._grid,FermionRedBlackGrid());
-  conformable(A._grid,B._grid);
+  conformable(A.Grid(),FermionRedBlackGrid());
+  conformable(A.Grid(),B.Grid());
 
-  assert(B.checkerboard==Even);
-  assert(A.checkerboard==Odd);
-  mat.checkerboard = Odd;
+  assert(B.Checkerboard()==Even);
+  assert(A.Checkerboard()==Odd);
+  mat.Checkerboard() = Odd;
 
   DerivInternal(StencilEven,UmuOdd,mat,A,B,dag);
 }
@@ -374,11 +352,9 @@ void WilsonFermion5D<Impl>::DhopInternal(StencilImpl & st, LebesgueOrder &lo,
                                          const FermionField &in, FermionField &out,int dag)
 {
   DhopTotalTime-=usecond();
-#ifdef GRID_OMP
   if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute )
     DhopInternalOverlappedComms(st,lo,U,in,out,dag);
   else 
-#endif
     DhopInternalSerialComms(st,lo,U,in,out,dag);
   DhopTotalTime+=usecond();
 }
@@ -389,131 +365,84 @@ void WilsonFermion5D<Impl>::DhopInternalOverlappedComms(StencilImpl & st, Lebesg
 							DoubledGaugeField & U,
 							const FermionField &in, FermionField &out,int dag)
 {
-#ifdef GRID_OMP
-  //  assert((dag==DaggerNo) ||(dag==DaggerYes));
-
   Compressor compressor(dag);
 
-  int LLs = in._grid->_rdimensions[0];
-  int len =  U._grid->oSites();
+  int LLs = in.Grid()->_rdimensions[0];
+  int len =  U.Grid()->oSites();
 
+  /////////////////////////////
+  // Start comms  // Gather intranode and extra node differentiated??
+  /////////////////////////////
   DhopFaceTime-=usecond();
   st.HaloExchangeOptGather(in,compressor);
-  st.CommsMergeSHM(compressor);// Could do this inside parallel region overlapped with comms
   DhopFaceTime+=usecond();
 
-  double ctime=0;
-  double ptime=0;
+  DhopCommTime -=usecond();
+  std::vector<std::vector<CommsRequest_t> > requests;
+  st.CommunicateBegin(requests);
 
-  //////////////////////////////////////////////////////////////////////////////////////////////////////
-  // Ugly explicit thread mapping introduced for OPA reasons.
-  //////////////////////////////////////////////////////////////////////////////////////////////////////
-#pragma omp parallel reduction(max:ctime) reduction(max:ptime)
-  { 
-    int tid = omp_get_thread_num();
-    int nthreads = omp_get_num_threads();
-    int ncomms = CartesianCommunicator::nCommThreads;
-    if (ncomms == -1) ncomms = 1;
-    assert(nthreads > ncomms);
-    if (tid >= ncomms) {
-      double start = usecond();
-      nthreads -= ncomms;
-      int ttid = tid - ncomms;
-      int n = U._grid->oSites();
-      int chunk = n / nthreads;
-      int rem = n % nthreads;
-      int myblock, myn;
-      if (ttid < rem) {
-	myblock = ttid * chunk + ttid;
-	myn = chunk+1;
-      } else {
-	myblock = ttid*chunk + rem;
-	myn = chunk;
-      }
+  /////////////////////////////
+  // Overlap with comms
+  /////////////////////////////
+  DhopFaceTime-=usecond();
+  st.CommsMergeSHM(compressor);// Could do this inside parallel region overlapped with comms
+  DhopFaceTime+=usecond();
       
-      // do the compute
-      if (dag == DaggerYes) {
-	for (int ss = myblock; ss < myblock+myn; ++ss) {
-	  int sU = ss;
-	  int sF = LLs * sU;
-	  Kernels::DhopSiteDag(st,lo,U,st.CommBuf(),sF,sU,LLs,1,in,out,1,0);
-	}
-      } else {
-	for (int ss = myblock; ss < myblock+myn; ++ss) {
-	  int sU = ss;
-	  int sF = LLs * sU;
-	  Kernels::DhopSite(st,lo,U,st.CommBuf(),sF,sU,LLs,1,in,out,1,0);
-	}
-      }
-	ptime = usecond() - start;
-    } else {
-      double start = usecond();
-      st.CommunicateThreaded();
-      ctime = usecond() - start;
-    }
+  /////////////////////////////
+  // do the compute interior
+  /////////////////////////////
+  int Opt = WilsonKernelsStatic::Opt; // Why pass this. Kernels should know
+  DhopComputeTime-=usecond();
+  if (dag == DaggerYes) {
+    Kernels::DhopDagKernel(Opt,st,U,st.CommBuf(),LLs,U.oSites(),in,out,1,0);
+  } else {
+    Kernels::DhopKernel   (Opt,st,U,st.CommBuf(),LLs,U.oSites(),in,out,1,0);
   }
-  DhopCommTime += ctime;
-  DhopComputeTime+=ptime;
+  DhopComputeTime+=usecond();
 
-  // First to enter, last to leave timing
-  st.CollateThreads();
+  /////////////////////////////
+  // Complete comms
+  /////////////////////////////
+  st.CommunicateComplete(requests);
+  DhopCommTime   +=usecond();
 
+  /////////////////////////////
+  // do the compute exterior
+  /////////////////////////////
   DhopFaceTime-=usecond();
   st.CommsMerge(compressor);
   DhopFaceTime+=usecond();
 
   DhopComputeTime2-=usecond();
   if (dag == DaggerYes) {
-    int sz=st.surface_list.size();
-    parallel_for (int ss = 0; ss < sz; ss++) {
-      int sU = st.surface_list[ss];
-      int sF = LLs * sU;
-      Kernels::DhopSiteDag(st,lo,U,st.CommBuf(),sF,sU,LLs,1,in,out,0,1);
-    }
+    Kernels::DhopDagKernel(Opt,st,U,st.CommBuf(),LLs,U.oSites(),in,out,0,1);
   } else {
-    int sz=st.surface_list.size();
-    parallel_for (int ss = 0; ss < sz; ss++) {
-      int sU = st.surface_list[ss];
-      int sF = LLs * sU;
-      Kernels::DhopSite(st,lo,U,st.CommBuf(),sF,sU,LLs,1,in,out,0,1);
-    }
+    Kernels::DhopKernel   (Opt,st,U,st.CommBuf(),LLs,U.oSites(),in,out,0,1);
   }
   DhopComputeTime2+=usecond();
-#else 
-  assert(0);
-#endif
 }
 
 
 template<class Impl>
 void WilsonFermion5D<Impl>::DhopInternalSerialComms(StencilImpl & st, LebesgueOrder &lo,
-					 DoubledGaugeField & U,
-					 const FermionField &in, FermionField &out,int dag)
+						    DoubledGaugeField & U,
+						    const FermionField &in, 
+						    FermionField &out,int dag)
 {
-  //  assert((dag==DaggerNo) ||(dag==DaggerYes));
   Compressor compressor(dag);
 
-  int LLs = in._grid->_rdimensions[0];
+  int LLs = in.Grid()->_rdimensions[0];
   
   DhopCommTime-=usecond();
   st.HaloExchangeOpt(in,compressor);
   DhopCommTime+=usecond();
   
   DhopComputeTime-=usecond();
-  // Dhop takes the 4d grid from U, and makes a 5d index for fermion
-
+  int Opt = WilsonKernelsStatic::Opt;
   if (dag == DaggerYes) {
-    parallel_for (int ss = 0; ss < U._grid->oSites(); ss++) {
-      int sU = ss;
-      int sF = LLs * sU;
-      Kernels::DhopSiteDag(st,lo,U,st.CommBuf(),sF,sU,LLs,1,in,out);
-    }
+    Kernels::DhopDagKernel(Opt,st,U,st.CommBuf(),LLs,U.oSites(),in,out);
   } else {
-    parallel_for (int ss = 0; ss < U._grid->oSites(); ss++) {
-      int sU = ss;
-      int sF = LLs * sU;
-      Kernels::DhopSite(st,lo,U,st.CommBuf(),sF,sU,LLs,1,in,out);
-    }
+    Kernels::DhopKernel(Opt,st,U,st.CommBuf(),LLs,U.oSites(),in,out);
   }
   DhopComputeTime+=usecond();
 }
@@ -523,11 +452,11 @@ template<class Impl>
 void WilsonFermion5D<Impl>::DhopOE(const FermionField &in, FermionField &out,int dag)
 {
   DhopCalls++;
-  conformable(in._grid,FermionRedBlackGrid());    // verifies half grid
-  conformable(in._grid,out._grid); // drops the cb check
+  conformable(in.Grid(),FermionRedBlackGrid());    // verifies half grid
+  conformable(in.Grid(),out.Grid()); // drops the cb check
 
-  assert(in.checkerboard==Even);
-  out.checkerboard = Odd;
+  assert(in.Checkerboard()==Even);
+  out.Checkerboard() = Odd;
 
   DhopInternal(StencilEven,LebesgueEvenOdd,UmuOdd,in,out,dag);
 }
@@ -535,11 +464,11 @@ template<class Impl>
 void WilsonFermion5D<Impl>::DhopEO(const FermionField &in, FermionField &out,int dag)
 {
   DhopCalls++;
-  conformable(in._grid,FermionRedBlackGrid());    // verifies half grid
-  conformable(in._grid,out._grid); // drops the cb check
+  conformable(in.Grid(),FermionRedBlackGrid());    // verifies half grid
+  conformable(in.Grid(),out.Grid()); // drops the cb check
 
-  assert(in.checkerboard==Odd);
-  out.checkerboard = Even;
+  assert(in.Checkerboard()==Odd);
+  out.Checkerboard() = Even;
 
   DhopInternal(StencilOdd,LebesgueEvenOdd,UmuEven,in,out,dag);
 }
@@ -547,17 +476,17 @@ template<class Impl>
 void WilsonFermion5D<Impl>::Dhop(const FermionField &in, FermionField &out,int dag)
 {
   DhopCalls+=2;
-  conformable(in._grid,FermionGrid()); // verifies full grid
-  conformable(in._grid,out._grid);
+  conformable(in.Grid(),FermionGrid()); // verifies full grid
+  conformable(in.Grid(),out.Grid());
 
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
 
   DhopInternal(Stencil,Lebesgue,Umu,in,out,dag);
 }
 template<class Impl>
 void WilsonFermion5D<Impl>::DW(const FermionField &in, FermionField &out,int dag)
 {
-  out.checkerboard=in.checkerboard;
+  out.Checkerboard()=in.Checkerboard();
   Dhop(in,out,dag); // -0.5 is included
   axpy(out,4.0-M5,in,out);
 }
@@ -569,7 +498,7 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHt_5d(FermionField &out,const
   GridBase *_grid = _FourDimGrid;
   GridBase *_5dgrid = _FiveDimGrid;
 
-  conformable(_5dgrid,out._grid);
+  conformable(_5dgrid,out.Grid());
 
   FermionField   PRsource(_5dgrid);
   FermionField   PLsource(_5dgrid);
@@ -580,7 +509,7 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHt_5d(FermionField &out,const
   FermionField   bufL_4d(_grid);
   FermionField   bufR_4d(_grid);
 
-  unsigned int Ls = in._grid->_rdimensions[0];
+  unsigned int Ls = in.Grid()->_rdimensions[0];
   
   typedef typename FermionField::vector_type vector_type;
   typedef typename FermionField::scalar_type ScalComplex;
@@ -596,12 +525,12 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHt_5d(FermionField &out,const
 
   Gamma g5(Gamma::Algebra::Gamma5);
 
-  std::vector<int> latt_size   = _grid->_fdimensions;
+  Coordinate latt_size   = _grid->_fdimensions;
 
-  LatComplex    sk(_grid);  sk = zero;
-  LatComplex    sk2(_grid); sk2= zero;
-  LatComplex    W(_grid); W= zero;
-  LatComplex    a(_grid); a= zero;
+  LatComplex    sk(_grid);  sk = Zero();
+  LatComplex    sk2(_grid); sk2= Zero();
+  LatComplex    W(_grid); W= Zero();
+  LatComplex    a(_grid); a= Zero();
   LatComplex    one  (_grid); one = ScalComplex(1.0,0.0);
   LatComplex 	cosha(_grid);
   LatComplex 	kmu(_grid);
@@ -643,9 +572,9 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHt_5d(FermionField &out,const
 
   // FIXME Need a Lattice acosh
   for(int idx=0;idx<_grid->lSites();idx++){
-    std::vector<int> lcoor(Nd);
+    Coordinate lcoor(Nd);
     Tcomplex cc;
-    RealD sgn;
+    //    RealD sgn;
     _grid->LocalIndexToLocalCoor(idx,lcoor);
     peekLocalSite(cc,cosha,lcoor);
     assert((double)real(cc)>=1.0);
@@ -678,8 +607,8 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHt_5d(FermionField &out,const
   //calculate GR, GL
   for(unsigned int ss=1;ss<=Ls;ss++)
   {
-    bufR_4d = zero;
-    bufL_4d = zero;
+    bufR_4d = Zero();
+    bufL_4d = Zero();
     for(unsigned int tt=1;tt<=Ls;tt++)
     {
       //possible sign if W<0
@@ -688,7 +617,7 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHt_5d(FermionField &out,const
 
       unsigned int f = (ss > tt) ? ss-tt : tt-ss; //f = abs(ss-tt)
       //GR
-      buf1_4d = zero;
+      buf1_4d = Zero();
       ExtractSlice(buf1_4d, PRsource, (tt-1), 0);
       //G(s,t)
       bufR_4d = bufR_4d + A * exp(a*Ls) * exp(-a*f) * signW * buf1_4d + A * exp(-a*Ls) * exp(a*f) * signW * buf1_4d;
@@ -702,7 +631,7 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHt_5d(FermionField &out,const
       bufR_4d = bufR_4d + Amm * exp(-a*ss) * exp(-a*tt) * signW * buf1_4d ;
 
       //GL
-      buf2_4d = zero;
+      buf2_4d = Zero();
       ExtractSlice(buf2_4d, PLsource, (tt-1), 0);
       //G(s,t)
       bufL_4d = bufL_4d + A * exp(a*Ls) * exp(-a*f) * signW * buf2_4d + A * exp(-a*Ls) * exp(a*f) * signW * buf2_4d;
@@ -722,13 +651,13 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHt_5d(FermionField &out,const
 //calculate propagator
   for(unsigned int ss=1;ss<=Ls;ss++)
   {
-    bufR_4d = zero;
-    bufL_4d = zero;
+    bufR_4d = Zero();
+    bufL_4d = Zero();
 
     //(i*gamma_mu*sin(p_mu) - W)*(GL*P- source)
-    buf1_4d = zero;
+    buf1_4d = Zero();
     ExtractSlice(buf1_4d, GL, (ss-1), 0);
-    buf2_4d = zero;
+    buf2_4d = Zero();
     for(int mu=0;mu<Nd;mu++) {
       LatticeCoordinate(kmu,mu);
       RealD TwoPiL =  M_PI * 2.0/ latt_size[mu];
@@ -738,9 +667,9 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHt_5d(FermionField &out,const
     bufL_4d = buf2_4d - W * buf1_4d;
 
     //(i*gamma_mu*sin(p_mu) - W)*(GR*P+ source)
-    buf1_4d = zero;
+    buf1_4d = Zero();
     ExtractSlice(buf1_4d, GR, (ss-1), 0);
-    buf2_4d = zero;
+    buf2_4d = Zero();
     for(int mu=0;mu<Nd;mu++) {
       LatticeCoordinate(kmu,mu);
       RealD TwoPiL =  M_PI * 2.0/ latt_size[mu];
@@ -781,7 +710,7 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHt(FermionField &out,const Fe
 {
   // what type LatticeComplex 
   GridBase *_grid = _FourDimGrid;
-  conformable(_grid,out._grid);
+  conformable(_grid,out.Grid());
   
   typedef typename FermionField::vector_type vector_type;
   typedef typename FermionField::scalar_type ScalComplex;
@@ -795,17 +724,17 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHt(FermionField &out,const Fe
     Gamma::Algebra::GammaT
   };
 
-  std::vector<int> latt_size   = _grid->_fdimensions;
+  Coordinate latt_size   = _grid->_fdimensions;
 
   
-  FermionField   num  (_grid); num  = zero;
+  FermionField   num  (_grid); num  = Zero();
 
-  LatComplex    sk(_grid);  sk = zero;
-  LatComplex    sk2(_grid); sk2= zero;
-  LatComplex    W(_grid); W= zero;
-  LatComplex    a(_grid); a= zero;
+  LatComplex    sk(_grid);  sk = Zero();
+  LatComplex    sk2(_grid); sk2= Zero();
+  LatComplex    W(_grid); W= Zero();
+  LatComplex    a(_grid); a= Zero();
   LatComplex    one  (_grid); one = ScalComplex(1.0,0.0);
-  LatComplex denom(_grid); denom= zero;
+  LatComplex denom(_grid); denom= Zero();
   LatComplex cosha(_grid); 
   LatComplex kmu(_grid); 
   LatComplex Wea(_grid); 
@@ -838,9 +767,9 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHt(FermionField &out,const Fe
 
   // FIXME Need a Lattice acosh
   for(int idx=0;idx<_grid->lSites();idx++){
-    std::vector<int> lcoor(Nd);
+    Coordinate lcoor(Nd);
     Tcomplex cc;
-    RealD sgn;
+    //    RealD sgn;
     _grid->LocalIndexToLocalCoor(idx,lcoor);
     peekLocalSite(cc,cosha,lcoor);
     assert((double)real(cc)>=1.0);
@@ -868,7 +797,7 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHw(FermionField &out,const Fe
     };
 
     GridBase *_grid = _FourDimGrid;
-    conformable(_grid,out._grid);
+    conformable(_grid,out.Grid());
 
     typedef typename FermionField::vector_type vector_type;
     typedef typename FermionField::scalar_type ScalComplex;
@@ -876,18 +805,18 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHw(FermionField &out,const Fe
     typedef Lattice<iSinglet<vector_type> > LatComplex;
 
 
-    std::vector<int> latt_size   = _grid->_fdimensions;
+    Coordinate latt_size   = _grid->_fdimensions;
 
-    LatComplex    sk(_grid);  sk = zero;
-    LatComplex    sk2(_grid); sk2= zero;
+    LatComplex    sk(_grid);  sk = Zero();
+    LatComplex    sk2(_grid); sk2= Zero();
 
-    LatComplex    w_k(_grid); w_k= zero;
-    LatComplex    b_k(_grid); b_k= zero;
+    LatComplex    w_k(_grid); w_k= Zero();
+    LatComplex    b_k(_grid); b_k= Zero();
 
     LatComplex     one  (_grid); one = ScalComplex(1.0,0.0);
 
-    FermionField   num  (_grid); num  = zero;
-    LatComplex denom(_grid); denom= zero;
+    FermionField   num  (_grid); num  = Zero();
+    LatComplex denom(_grid); denom= Zero();
     LatComplex kmu(_grid); 
     ScalComplex ci(0.0,1.0);
 
@@ -928,7 +857,7 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHw(FermionField &out,const Fe
 // Helper macro to reverse Simd vector. Fixme: slow, generic implementation.
 #define REVERSE_LS(qSite, qSiteRev, Nsimd) \
 { \
-    std::vector<typename SitePropagator::scalar_object> qSiteVec(Nsimd); \
+    ExtractBuffer<typename SitePropagator::scalar_object> qSiteVec(Nsimd);	\
     extract(qSite, qSiteVec); \
     for (int i = 0; i < Nsimd / 2; ++i) \
     { \
@@ -946,31 +875,35 @@ void WilsonFermion5D<Impl>::MomentumSpacePropagatorHw(FermionField &out,const Fe
 template<class vobj> 
 Lattice<vobj> spProj5p(const Lattice<vobj> & in)
 {
-  GridBase *grid=in._grid;
+  GridBase *grid=in.Grid();
   Gamma G5(Gamma::Algebra::Gamma5);
   Lattice<vobj> ret(grid);
-  parallel_for(int ss=0;ss<grid->oSites();ss++){
-    ret._odata[ss] = in._odata[ss] + G5*in._odata[ss];
-  }
+  auto ret_v = ret.View();
+  auto in_v  =  in.View();
+  thread_for(ss,grid->oSites(),{
+    ret_v[ss] = in_v[ss] + G5*in_v[ss];
+  });
   return ret;
 }
 template<class vobj> 
 Lattice<vobj> spProj5m(const Lattice<vobj> & in)
 {
   Gamma G5(Gamma::Algebra::Gamma5);
-  GridBase *grid=in._grid;
+  GridBase *grid=in.Grid();
   Lattice<vobj> ret(grid);
-  parallel_for(int ss=0;ss<grid->oSites();ss++){
-    ret._odata[ss] = in._odata[ss] - G5*in._odata[ss];
-  }
+  auto ret_v = ret.View();
+  auto in_v  =  in.View();
+  thread_for(ss,grid->oSites(),{
+    ret_v[ss] = in_v[ss] - G5*in_v[ss];
+  });
   return ret;
 }
 
 template <class Impl>
 void WilsonFermion5D<Impl>::ContractJ5q(FermionField &q_in,ComplexField &J5q)
 {
-  conformable(GaugeGrid(), J5q._grid);
-  conformable(q_in._grid, FermionGrid());
+  conformable(GaugeGrid(), J5q.Grid());
+  conformable(q_in.Grid(), FermionGrid());
 
   // 4d field
   int Ls = this->Ls;
@@ -990,8 +923,8 @@ void WilsonFermion5D<Impl>::ContractJ5q(FermionField &q_in,ComplexField &J5q)
 template <class Impl>
 void WilsonFermion5D<Impl>::ContractJ5q(PropagatorField &q_in,ComplexField &J5q)
 {
-  conformable(GaugeGrid(), J5q._grid);
-  conformable(q_in._grid, FermionGrid());
+  conformable(GaugeGrid(), J5q.Grid());
+  conformable(q_in.Grid(), FermionGrid());
 
   // 4d field
   int Ls = this->Ls;
@@ -1015,20 +948,26 @@ void WilsonFermion5D<Impl>::ContractConservedCurrent(PropagatorField &q_in_1,
                                                      Current curr_type,
                                                      unsigned int mu)
 {
-    conformable(q_in_1._grid, FermionGrid());
-    conformable(q_in_1._grid, q_in_2._grid);
-    conformable(_FourDimGrid, q_out._grid);
+    conformable(q_in_1.Grid(), FermionGrid());
+    conformable(q_in_1.Grid(), q_in_2.Grid());
+    conformable(_FourDimGrid, q_out.Grid());
 
     PropagatorField tmp1(FermionGrid()), tmp2(FermionGrid());
-    unsigned int LLs = q_in_1._grid->_rdimensions[0];
-    q_out = zero;
+    unsigned int LLs = q_in_1.Grid()->_rdimensions[0];
+    q_out = Zero();
 
     // Forward, need q1(x + mu, s), q2(x, Ls - 1 - s). Backward, need q1(x, s), 
     // q2(x + mu, Ls - 1 - s). 5D lattice so shift 4D coordinate mu by one.
     tmp1 = Cshift(q_in_1, mu + 1, 1);
     tmp2 = Cshift(q_in_2, mu + 1, 1);
-    parallel_for (unsigned int sU = 0; sU < Umu._grid->oSites(); ++sU)
-    {
+    auto q_in_1_v = q_in_1.View();
+    auto q_in_2_v = q_in_2.View();
+    auto tmp1_v   = tmp1.View();
+    auto tmp2_v   = tmp2.View();
+    auto q_out_v  = q_out.View();
+    auto Umu_v    = Umu.View();
+    thread_for(sU, Umu.Grid()->oSites(),{
+
         unsigned int sF1 = sU * LLs;
         unsigned int sF2 = (sU + 1) * LLs - 1;
 
@@ -1042,26 +981,26 @@ void WilsonFermion5D<Impl>::ContractConservedCurrent(PropagatorField &q_in_1,
             // sites correctly.
             if (Impl::LsVectorised)
             {
-                REVERSE_LS(q_in_2._odata[sF2], qSite2, Ls / LLs);
-                REVERSE_LS(tmp2._odata[sF2], qmuSite2, Ls / LLs);
+                REVERSE_LS(q_in_2_v[sF2], qSite2, Ls / LLs);
+                REVERSE_LS(tmp2_v[sF2], qmuSite2, Ls / LLs);
             }
             else
             {
-                qSite2   = q_in_2._odata[sF2];
-                qmuSite2 = tmp2._odata[sF2];
+                qSite2   = q_in_2_v[sF2];
+                qmuSite2 = tmp2_v[sF2];
             }
-            Kernels::ContractConservedCurrentSiteFwd(tmp1._odata[sF1], 
+            Kernels::ContractConservedCurrentSiteFwd(tmp1_v[sF1], 
                                                      qSite2, 
-                                                     q_out._odata[sU],
-                                                     Umu, sU, mu, axial_sign);
-            Kernels::ContractConservedCurrentSiteBwd(q_in_1._odata[sF1],
+                                                     q_out_v[sU],
+                                                     Umu_v, sU, mu, axial_sign);
+            Kernels::ContractConservedCurrentSiteBwd(q_in_1_v[sF1],
                                                      qmuSite2,
-                                                     q_out._odata[sU],
-                                                     Umu, sU, mu, axial_sign);
+                                                     q_out_v[sU],
+                                                     Umu_v, sU, mu, axial_sign);
             sF1++;
             sF2--;
         }
-    }
+    });
 }
 
 
@@ -1074,18 +1013,21 @@ void WilsonFermion5D<Impl>::SeqConservedCurrent(PropagatorField &q_in,
                                                 unsigned int tmax,
 						ComplexField &lattice_cmplx)
 {
-    conformable(q_in._grid, FermionGrid());
-    conformable(q_in._grid, q_out._grid);
+    conformable(q_in.Grid(), FermionGrid());
+    conformable(q_in.Grid(), q_out.Grid());
     PropagatorField tmp(GaugeGrid()),tmp2(GaugeGrid());
     unsigned int tshift = (mu == Tp) ? 1 : 0;
-    unsigned int LLs = q_in._grid->_rdimensions[0];
+    unsigned int LLs = q_in.Grid()->_rdimensions[0];
     unsigned int LLt    = GridDefaultLatt()[Tp];
 
-    q_out = zero;
+    q_out = Zero();
     LatticeInteger coords(_FourDimGrid);
     LatticeCoordinate(coords, Tp);
-
-
+    
+    auto q_out_v = q_out.View();
+    auto tmp2_v  = tmp2.View();
+    auto coords_v= coords.View();
+    auto Umu_v   = Umu.View();
     for (unsigned int s = 0; s < LLs; ++s)
     {
         bool axial_sign = ((curr_type == Current::Axial) && (s < (LLs / 2)));
@@ -1098,59 +1040,51 @@ void WilsonFermion5D<Impl>::SeqConservedCurrent(PropagatorField &q_in,
         tmp = Cshift(tmp2, mu, 1);	 //q(x+mu,s)
         tmp2 = tmp*lattice_cmplx;	 //q(x+mu,s)*A(x)	
 
-    	parallel_for (unsigned int sU = 0; sU < Umu._grid->oSites(); ++sU)
-    	{
+    	thread_for(sU, Umu.Grid()->oSites(),{
             // Compute the sequential conserved current insertion only if our simd
             // object contains a timeslice we need.
-            vInteger t_mask   = ((coords._odata[sU] >= tmin) &&
-                	         (coords._odata[sU] <= tmax));
-            Integer timeSlices = Reduce(t_mask);
+            vPredicate t_mask;
+	    t_mask() = ((coords_v[sU] >= tmin) && (coords_v[sU] <= tmax));
+            Integer timeSlices = Reduce(t_mask());
 
             if (timeSlices > 0)
             {
 		unsigned int sF = sU * LLs + s;
-                Kernels::SeqConservedCurrentSiteFwd(tmp2._odata[sU], 
-                                              q_out._odata[sF], Umu, sU,
-                                              mu, t_mask, switch_sgn);
+                Kernels::SeqConservedCurrentSiteFwd(tmp2_v[sU], 
+						    q_out_v[sF], Umu_v, sU,
+						    mu, t_mask, switch_sgn);
             }
 
-        }
+        });
 
         //backward direction: Need q(x - mu, s)*A(x-mu)
         ExtractSlice(tmp2, q_in, s, 0);  //q(x,s)
         tmp = lattice_cmplx*tmp2;	 //q(x,s)*A(x)
         tmp2 = Cshift(tmp, mu, -1);	 //q(x-mu,s)*A(x-mu,s)
 
-    	parallel_for (unsigned int sU = 0; sU < Umu._grid->oSites(); ++sU)
+    	thread_for(sU, Umu.Grid()->oSites(),
     	{
-            vInteger  t_mask     = ((coords._odata[sU] >= (tmin + tshift)) && 
-                   	  	    (coords._odata[sU] <= (tmax + tshift)));
+	  vPredicate t_mask;
+	  t_mask()= ((coords_v[sU] >= (tmin + tshift)) && (coords_v[sU] <= (tmax + tshift)));
 
-	    //if tmax = LLt-1 (last timeslice) include timeslice 0 if the time is shifted (mu=3)	
-	    unsigned int t0 = 0;
-	    if((tmax==LLt-1) && (tshift==1)) t_mask = (t_mask || (coords._odata[sU] == t0 ));
-
-            Integer timeSlices = Reduce(t_mask);
-
-            if (timeSlices > 0)
-            {
-		unsigned int sF = sU * LLs + s; 
-        	Kernels::SeqConservedCurrentSiteBwd(tmp2._odata[sU], 
-                                             q_out._odata[sF], Umu, sU,
-                                             mu, t_mask, axial_sign);
-            }
-	}
+	  //if tmax = LLt-1 (last timeslice) include timeslice 0 if the time is shifted (mu=3)	
+	  unsigned int t0 = 0;
+	  if((tmax==LLt-1) && (tshift==1)) t_mask() = (t_mask() || (coords_v[sU] == t0 ));
+	  
+	  Integer timeSlices = Reduce(t_mask());
+	  
+	  if (timeSlices > 0) {
+	    unsigned int sF = sU * LLs + s; 
+	    Kernels::SeqConservedCurrentSiteBwd(tmp2_v[sU], 
+						q_out_v[sF], Umu_v, sU,
+						mu, t_mask, axial_sign);
+	  }
+	});
     }
 }
-
-
-
-
-
-FermOpTemplateInstantiate(WilsonFermion5D);
-GparityFermOpTemplateInstantiate(WilsonFermion5D);
   
-}}
+NAMESPACE_END(Grid);
+
 
 
 
diff --git a/Grid/qcd/action/fermion/WilsonFermion.cc b/Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h
similarity index 57%
rename from Grid/qcd/action/fermion/WilsonFermion.cc
rename to Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h
index ffc41396..756bdbf4 100644
--- a/Grid/qcd/action/fermion/WilsonFermion.cc
+++ b/Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h
@@ -1,4 +1,3 @@
-
 /*************************************************************************************
 
 Grid physics library, www.github.com/paboyle/Grid
@@ -29,16 +28,11 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #include <Grid/qcd/action/fermion/FermionCore.h>
 #include <Grid/qcd/action/fermion/WilsonFermion.h>
 
-namespace Grid {
-namespace QCD {
-
-const std::vector<int> WilsonFermionStatic::directions({0, 1, 2, 3, 0, 1, 2, 3});
-const std::vector<int> WilsonFermionStatic::displacements({1, 1, 1, 1, -1, -1, -1, -1});
-int WilsonFermionStatic::HandOptDslash;
+NAMESPACE_BEGIN(Grid);
 
 /////////////////////////////////
 // Constructor and gauge import
@@ -49,18 +43,19 @@ WilsonFermion<Impl>::WilsonFermion(GaugeField &_Umu, GridCartesian &Fgrid,
                                    GridRedBlackCartesian &Hgrid, RealD _mass,
                                    const ImplParams &p,
                                    const WilsonAnisotropyCoefficients &anis)
-    : Kernels(p),
-      _grid(&Fgrid),
-      _cbgrid(&Hgrid),
-      Stencil(&Fgrid, npoint, Even, directions, displacements),
-      StencilEven(&Hgrid, npoint, Even, directions,displacements),  // source is Even
-      StencilOdd(&Hgrid, npoint, Odd, directions,displacements),  // source is Odd
-      mass(_mass),
-      Lebesgue(_grid),
-      LebesgueEvenOdd(_cbgrid),
-      Umu(&Fgrid),
-      UmuEven(&Hgrid),
-      UmuOdd(&Hgrid),
+  : 
+    Kernels(p),
+    _grid(&Fgrid),
+    _cbgrid(&Hgrid),
+    Stencil(&Fgrid, npoint, Even, directions, displacements,p),
+    StencilEven(&Hgrid, npoint, Even, directions,displacements,p),  // source is Even
+    StencilOdd(&Hgrid, npoint, Odd, directions,displacements,p),  // source is Odd
+    mass(_mass),
+    Lebesgue(_grid),
+    LebesgueEvenOdd(_cbgrid),
+    Umu(&Fgrid),
+    UmuEven(&Hgrid),
+    UmuOdd(&Hgrid),
       _tmp(&Hgrid),
       anisotropyCoeff(anis)
 {
@@ -76,8 +71,9 @@ WilsonFermion<Impl>::WilsonFermion(GaugeField &_Umu, GridCartesian &Fgrid,
 }
 
 template <class Impl>
-void WilsonFermion<Impl>::ImportGauge(const GaugeField &_Umu) {
-  GaugeField HUmu(_Umu._grid);
+void WilsonFermion<Impl>::ImportGauge(const GaugeField &_Umu) 
+{
+  GaugeField HUmu(_Umu.Grid());
 
   //Here multiply the anisotropy coefficients
   if (anisotropyCoeff.isAnisotropic)
@@ -107,21 +103,21 @@ void WilsonFermion<Impl>::ImportGauge(const GaugeField &_Umu) {
 
 template <class Impl>
 RealD WilsonFermion<Impl>::M(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   Dhop(in, out, DaggerNo);
   return axpy_norm(out, diag_mass, in, out);
 }
 
 template <class Impl>
 RealD WilsonFermion<Impl>::Mdag(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   Dhop(in, out, DaggerYes);
   return axpy_norm(out, diag_mass, in, out);
 }
 
 template <class Impl>
 void WilsonFermion<Impl>::Meooe(const FermionField &in, FermionField &out) {
-  if (in.checkerboard == Odd) {
+  if (in.Checkerboard() == Odd) {
     DhopEO(in, out, DaggerNo);
   } else {
     DhopOE(in, out, DaggerNo);
@@ -130,7 +126,7 @@ void WilsonFermion<Impl>::Meooe(const FermionField &in, FermionField &out) {
 
 template <class Impl>
 void WilsonFermion<Impl>::MeooeDag(const FermionField &in, FermionField &out) {
-  if (in.checkerboard == Odd) {
+  if (in.Checkerboard() == Odd) {
     DhopEO(in, out, DaggerYes);
   } else {
     DhopOE(in, out, DaggerYes);
@@ -139,26 +135,26 @@ void WilsonFermion<Impl>::MeooeDag(const FermionField &in, FermionField &out) {
   
 template <class Impl>
 void WilsonFermion<Impl>::Mooee(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   typename FermionField::scalar_type scal(diag_mass);
   out = scal * in;
 }
 
 template <class Impl>
 void WilsonFermion<Impl>::MooeeDag(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   Mooee(in, out);
 }
 
 template<class Impl>
 void WilsonFermion<Impl>::MooeeInv(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   out = (1.0/(diag_mass))*in;
 }
   
 template<class Impl>
 void WilsonFermion<Impl>::MooeeInvDag(const FermionField &in, FermionField &out) {
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
   MooeeInv(in,out);
 }
 template<class Impl>
@@ -169,7 +165,7 @@ void WilsonFermion<Impl>::MomentumSpacePropagator(FermionField &out, const Fermi
   typedef Lattice<iSinglet<vector_type> > LatComplex;
   
   // what type LatticeComplex 
-  conformable(_grid,out._grid);
+  conformable(_grid,out.Grid());
   
   Gamma::Algebra Gmu [] = {
     Gamma::Algebra::GammaX,
@@ -178,13 +174,13 @@ void WilsonFermion<Impl>::MomentumSpacePropagator(FermionField &out, const Fermi
     Gamma::Algebra::GammaT
   };
   
-  std::vector<int> latt_size   = _grid->_fdimensions;
+  Coordinate latt_size   = _grid->_fdimensions;
   
-  FermionField   num  (_grid); num  = zero;
-  LatComplex    wilson(_grid); wilson= zero;
+  FermionField   num  (_grid); num  = Zero();
+  LatComplex    wilson(_grid); wilson= Zero();
   LatComplex     one  (_grid); one = ScalComplex(1.0,0.0);
   
-  LatComplex denom(_grid); denom= zero;
+  LatComplex denom(_grid); denom= Zero();
   LatComplex kmu(_grid); 
   ScalComplex ci(0.0,1.0);
   // momphase = n * 2pi / L
@@ -229,9 +225,9 @@ void WilsonFermion<Impl>::DerivInternal(StencilImpl &st, DoubledGaugeField &U,
 
   Compressor compressor(dag);
 
-  FermionField Btilde(B._grid);
-  FermionField Atilde(B._grid);
-  Atilde = A;//redundant
+  FermionField Btilde(B.Grid());
+  FermionField Atilde(B.Grid());
+  Atilde = A;
 
   st.HaloExchange(B, compressor);
 
@@ -242,12 +238,8 @@ void WilsonFermion<Impl>::DerivInternal(StencilImpl &st, DoubledGaugeField &U,
     int gamma = mu;
     if (!dag) gamma += Nd;
 
-    ////////////////////////
-    // Call the single hop
-    ////////////////////////
-    parallel_for (int sss = 0; sss < B._grid->oSites(); sss++) {
-      Kernels::DhopDir(st, U, st.CommBuf(), sss, sss, B, Btilde, mu, gamma);
-    }
+    int Ls=1;
+    Kernels::DhopDirKernel(st, U, st.CommBuf(), Ls, B.Grid()->oSites(), B, Btilde, mu, gamma);
 
     //////////////////////////////////////////////////
     // spin trace outer product
@@ -258,70 +250,70 @@ void WilsonFermion<Impl>::DerivInternal(StencilImpl &st, DoubledGaugeField &U,
 
 template <class Impl>
 void WilsonFermion<Impl>::DhopDeriv(GaugeField &mat, const FermionField &U, const FermionField &V, int dag) {
-  conformable(U._grid, _grid);
-  conformable(U._grid, V._grid);
-  conformable(U._grid, mat._grid);
+  conformable(U.Grid(), _grid);
+  conformable(U.Grid(), V.Grid());
+  conformable(U.Grid(), mat.Grid());
 
-  mat.checkerboard = U.checkerboard;
+  mat.Checkerboard() = U.Checkerboard();
 
   DerivInternal(Stencil, Umu, mat, U, V, dag);
 }
 
 template <class Impl>
 void WilsonFermion<Impl>::DhopDerivOE(GaugeField &mat, const FermionField &U, const FermionField &V, int dag) {
-  conformable(U._grid, _cbgrid);
-  conformable(U._grid, V._grid);
-  //conformable(U._grid, mat._grid); not general, leaving as a comment (Guido)
+  conformable(U.Grid(), _cbgrid);
+  conformable(U.Grid(), V.Grid());
+  //conformable(U.Grid(), mat.Grid()); not general, leaving as a comment (Guido)
   // Motivation: look at the SchurDiff operator
   
-  assert(V.checkerboard == Even);
-  assert(U.checkerboard == Odd);
-  mat.checkerboard = Odd;
+  assert(V.Checkerboard() == Even);
+  assert(U.Checkerboard() == Odd);
+  mat.Checkerboard() = Odd;
 
   DerivInternal(StencilEven, UmuOdd, mat, U, V, dag);
 }
 
 template <class Impl>
 void WilsonFermion<Impl>::DhopDerivEO(GaugeField &mat, const FermionField &U, const FermionField &V, int dag) {
-  conformable(U._grid, _cbgrid);
-  conformable(U._grid, V._grid);
-  //conformable(U._grid, mat._grid);
+  conformable(U.Grid(), _cbgrid);
+  conformable(U.Grid(), V.Grid());
+  //conformable(U.Grid(), mat.Grid());
 
-  assert(V.checkerboard == Odd);
-  assert(U.checkerboard == Even);
-  mat.checkerboard = Even;
+  assert(V.Checkerboard() == Odd);
+  assert(U.Checkerboard() == Even);
+  mat.Checkerboard() = Even;
 
   DerivInternal(StencilOdd, UmuEven, mat, U, V, dag);
 }
 
 template <class Impl>
 void WilsonFermion<Impl>::Dhop(const FermionField &in, FermionField &out, int dag) {
-  conformable(in._grid, _grid);  // verifies full grid
-  conformable(in._grid, out._grid);
+  conformable(in.Grid(), _grid);  // verifies full grid
+  conformable(in.Grid(), out.Grid());
 
-  out.checkerboard = in.checkerboard;
+  out.Checkerboard() = in.Checkerboard();
 
   DhopInternal(Stencil, Lebesgue, Umu, in, out, dag);
 }
 
 template <class Impl>
 void WilsonFermion<Impl>::DhopOE(const FermionField &in, FermionField &out, int dag) {
-  conformable(in._grid, _cbgrid);    // verifies half grid
-  conformable(in._grid, out._grid);  // drops the cb check
+  conformable(in.Grid(), _cbgrid);    // verifies half grid
+  conformable(in.Grid(), out.Grid());  // drops the cb check
 
-  assert(in.checkerboard == Even);
-  out.checkerboard = Odd;
+  assert(in.Checkerboard() == Even);
+  out.Checkerboard() = Odd;
 
   DhopInternal(StencilEven, LebesgueEvenOdd, UmuOdd, in, out, dag);
 }
 
 template <class Impl>
 void WilsonFermion<Impl>::DhopEO(const FermionField &in, FermionField &out,int dag) {
-  conformable(in._grid, _cbgrid);    // verifies half grid
-  conformable(in._grid, out._grid);  // drops the cb check
+  conformable(in.Grid(), _cbgrid);    // verifies half grid
+  conformable(in.Grid(), out.Grid());  // drops the cb check
 
-  assert(in.checkerboard == Odd);
-  out.checkerboard = Even;
+  assert(in.Checkerboard() == Odd);
+  out.Checkerboard() = Even;
 
   DhopInternal(StencilOdd, LebesgueEvenOdd, UmuEven, in, out, dag);
 }
@@ -332,7 +324,8 @@ void WilsonFermion<Impl>::Mdir(const FermionField &in, FermionField &out, int di
 }
 
 template <class Impl>
-void WilsonFermion<Impl>::DhopDir(const FermionField &in, FermionField &out, int dir, int disp) {
+void WilsonFermion<Impl>::DhopDir(const FermionField &in, FermionField &out, int dir, int disp) 
+{
   int skip = (disp == 1) ? 0 : 1;
   int dirdisp = dir + skip * 4;
   int gamma = dir + (1 - skip) * 4;
@@ -341,16 +334,16 @@ void WilsonFermion<Impl>::DhopDir(const FermionField &in, FermionField &out, int
 };
 
 template <class Impl>
-void WilsonFermion<Impl>::DhopDirDisp(const FermionField &in, FermionField &out,int dirdisp, int gamma, int dag) {
+void WilsonFermion<Impl>::DhopDirDisp(const FermionField &in, FermionField &out,int dirdisp, int gamma, int dag) 
+{
   Compressor compressor(dag);
 
   Stencil.HaloExchange(in, compressor);
+  int Ls=1;
+  int Nsite=in.oSites();
+  Kernels::DhopDirKernel(Stencil, Umu, Stencil.CommBuf(), Ls, Nsite, in, out, dirdisp, gamma);
+};
 
-  parallel_for (int sss = 0; sss < in._grid->oSites(); sss++) {
-    Kernels::DhopDir(Stencil, Umu, Stencil.CommBuf(), sss, sss, in, out, dirdisp, gamma);
-  }
-} 
-/*Change starts*/
 template <class Impl>
 void WilsonFermion<Impl>::DhopInternal(StencilImpl &st, LebesgueOrder &lo,
                                        DoubledGaugeField &U,
@@ -367,71 +360,51 @@ void WilsonFermion<Impl>::DhopInternal(StencilImpl &st, LebesgueOrder &lo,
 
 template <class Impl>
 void WilsonFermion<Impl>::DhopInternalOverlappedComms(StencilImpl &st, LebesgueOrder &lo,
-                                       DoubledGaugeField &U,
-                                       const FermionField &in,
-                                       FermionField &out, int dag) {
+						      DoubledGaugeField &U,
+						      const FermionField &in,
+						      FermionField &out, int dag) {
   assert((dag == DaggerNo) || (dag == DaggerYes));
-#ifdef GRID_OMP
-  Compressor compressor;
-  int len =  U._grid->oSites();
-  const int LLs =  1;
-
-  st.Prepare();
-  st.HaloGather(in,compressor);
-  st.CommsMergeSHM(compressor);
-#pragma omp parallel
-  {
-    int tid = omp_get_thread_num();
-    int nthreads = omp_get_num_threads();
-    int ncomms = CartesianCommunicator::nCommThreads;
-    if (ncomms == -1) ncomms = 1;
-    assert(nthreads > ncomms);
-    if (tid >= ncomms) {
-      nthreads -= ncomms;
-      int ttid  = tid - ncomms;
-      int n     = len;
-      int chunk = n / nthreads;
-      int rem   = n % nthreads;
-      int myblock, myn;
-      if (ttid < rem) {
-        myblock = ttid * chunk + ttid;
-        myn = chunk+1;
-      } else {
-        myblock = ttid*chunk + rem;
-        myn = chunk;
-      }
-      // do the compute
-     if (dag == DaggerYes) {
-
-        for (int sss = myblock; sss < myblock+myn; ++sss) {
-         Kernels::DhopSiteDag(st, lo, U, st.CommBuf(), sss, sss, 1, 1, in, out);
-       }
-     } else {
-        for (int sss = myblock; sss < myblock+myn; ++sss) {
-         Kernels::DhopSite(st, lo, U, st.CommBuf(), sss, sss, 1, 1, in, out);
-       }
-    } //else
-
-    } else {
-      st.CommunicateThreaded();
-    }
 
   Compressor compressor(dag);
+  int len =  U.Grid()->oSites();
 
+  /////////////////////////////
+  // Start comms  // Gather intranode and extra node differentiated??
+  /////////////////////////////
+  std::vector<std::vector<CommsRequest_t> > requests;
+  st.Prepare();
+  st.HaloGather(in,compressor);
+  st.CommunicateBegin(requests);
+
+  /////////////////////////////
+  // Overlap with comms
+  /////////////////////////////
+  st.CommsMergeSHM(compressor);
+
+  /////////////////////////////
+  // do the compute interior
+  /////////////////////////////
+  int Opt = WilsonKernelsStatic::Opt;
   if (dag == DaggerYes) {
-    parallel_for (int sss = 0; sss < in._grid->oSites(); sss++) {
-      Kernels::DhopSiteDag(st, lo, U, st.CommBuf(), sss, sss, 1, 1, in, out);
-    }
+    Kernels::DhopDagKernel(Opt,st,U,st.CommBuf(),1,U.oSites(),in,out,1,0);
   } else {
-    parallel_for (int sss = 0; sss < in._grid->oSites(); sss++) {
-      Kernels::DhopSite(st, lo, U, st.CommBuf(), sss, sss, 1, 1, in, out);
-    }
-  }
+    Kernels::DhopKernel(Opt,st,U,st.CommBuf(),1,U.oSites(),in,out,1,0);
+  } 
 
-  }  //pragma
-#else
-  assert(0);
-#endif
+  /////////////////////////////
+  // Complete comms
+  /////////////////////////////
+  st.CommunicateComplete(requests);
+  st.CommsMerge(compressor);
+
+  /////////////////////////////
+  // do the compute exterior
+  /////////////////////////////
+  if (dag == DaggerYes) {
+    Kernels::DhopDagKernel(Opt,st,U,st.CommBuf(),1,U.oSites(),in,out,0,1);
+  } else {
+    Kernels::DhopKernel(Opt,st,U,st.CommBuf(),1,U.oSites(),in,out,0,1);
+  }
 };
 
 
@@ -444,14 +417,11 @@ void WilsonFermion<Impl>::DhopInternalSerial(StencilImpl &st, LebesgueOrder &lo,
   Compressor compressor(dag);
   st.HaloExchange(in, compressor);
 
+  int Opt = WilsonKernelsStatic::Opt;
   if (dag == DaggerYes) {
-    parallel_for (int sss = 0; sss < in._grid->oSites(); sss++) {
-      Kernels::DhopSiteDag(st, lo, U, st.CommBuf(), sss, sss, 1, 1, in, out);
-    }
+    Kernels::DhopDagKernel(Opt,st,U,st.CommBuf(),1,U.oSites(),in,out);
   } else {
-    parallel_for (int sss = 0; sss < in._grid->oSites(); sss++) {
-      Kernels::DhopSite(st, lo, U, st.CommBuf(), sss, sss, 1, 1, in, out);
-    }
+    Kernels::DhopKernel(Opt,st,U,st.CommBuf(),1,U.oSites(),in,out);
   }
 };
 /*Change ends */
@@ -468,28 +438,33 @@ void WilsonFermion<Impl>::ContractConservedCurrent(PropagatorField &q_in_1,
                                                    Current curr_type,
                                                    unsigned int mu)
 {
-    Gamma g5(Gamma::Algebra::Gamma5);
-    conformable(_grid, q_in_1._grid);
-    conformable(_grid, q_in_2._grid);
-    conformable(_grid, q_out._grid);
-    PropagatorField tmp1(_grid), tmp2(_grid);
-    q_out = zero;
+  Gamma g5(Gamma::Algebra::Gamma5);
+  conformable(_grid, q_in_1.Grid());
+  conformable(_grid, q_in_2.Grid());
+  conformable(_grid, q_out.Grid());
+  PropagatorField tmp1(_grid), tmp2(_grid);
+  q_out = Zero();
 
-    // Forward, need q1(x + mu), q2(x). Backward, need q1(x), q2(x + mu).
-    // Inefficient comms method but not performance critical.
-    tmp1 = Cshift(q_in_1, mu, 1);
-    tmp2 = Cshift(q_in_2, mu, 1);
-    parallel_for (unsigned int sU = 0; sU < Umu._grid->oSites(); ++sU)
-    {
-        Kernels::ContractConservedCurrentSiteFwd(tmp1._odata[sU],
-                                                 q_in_2._odata[sU],
-                                                 q_out._odata[sU],
-                                                 Umu, sU, mu);
-        Kernels::ContractConservedCurrentSiteBwd(q_in_1._odata[sU],
-                                                 tmp2._odata[sU],
-                                                 q_out._odata[sU],
-                                                 Umu, sU, mu);
-    }
+  // Forward, need q1(x + mu), q2(x). Backward, need q1(x), q2(x + mu).
+  // Inefficient comms method but not performance critical.
+  tmp1 = Cshift(q_in_1, mu, 1);
+  tmp2 = Cshift(q_in_2, mu, 1);
+  auto tmp1_v  =  tmp1.View();
+  auto tmp2_v  =  tmp2.View();
+  auto q_in_1_v=q_in_1.View();
+  auto q_in_2_v=q_in_2.View();
+  auto q_out_v = q_out.View();
+  auto Umu_v   =   Umu.View();
+  thread_for(sU, Umu.Grid()->oSites(),{
+      Kernels::ContractConservedCurrentSiteFwd(tmp1_v[sU],
+					       q_in_2_v[sU],
+					       q_out_v[sU],
+					       Umu_v, sU, mu);
+      Kernels::ContractConservedCurrentSiteBwd(q_in_1_v[sU],
+					       tmp2_v[sU],
+					       q_out_v[sU],
+					       Umu_v, sU, mu);
+  });
 }
 
 
@@ -502,61 +477,61 @@ void WilsonFermion<Impl>::SeqConservedCurrent(PropagatorField &q_in,
                                               unsigned int tmax,
 					      ComplexField &lattice_cmplx)
 {
-    conformable(_grid, q_in._grid);
-    conformable(_grid, q_out._grid);
-    PropagatorField tmpFwd(_grid), tmpBwd(_grid), tmp(_grid);
-    unsigned int tshift = (mu == Tp) ? 1 : 0;
-    unsigned int LLt    = GridDefaultLatt()[Tp];
+  conformable(_grid, q_in.Grid());
+  conformable(_grid, q_out.Grid());
 
-    q_out = zero;
-    LatticeInteger coords(_grid);
-    LatticeCoordinate(coords, Tp);
+  //  Lattice<iSinglet<Simd>> ph(_grid), coor(_grid);
+  Complex i(0.0,1.0);
+  PropagatorField tmpFwd(_grid), tmpBwd(_grid), tmp(_grid);
+  unsigned int tshift = (mu == Tp) ? 1 : 0;
+  unsigned int LLt    = GridDefaultLatt()[Tp];
 
-    // Need q(x + mu) and q(x - mu).
-    tmp = Cshift(q_in, mu, 1);
-    tmpFwd = tmp*lattice_cmplx;
-    tmp = lattice_cmplx*q_in;
-    tmpBwd = Cshift(tmp, mu, -1);
+  q_out = Zero();
+  LatticeInteger coords(_grid);
+  LatticeCoordinate(coords, Tp);
 
-    parallel_for (unsigned int sU = 0; sU < Umu._grid->oSites(); ++sU)
-    {
-        // Compute the sequential conserved current insertion only if our simd
-        // object contains a timeslice we need.
-        vInteger t_mask   = ((coords._odata[sU] >= tmin) &&
-                             (coords._odata[sU] <= tmax));
-        Integer timeSlices = Reduce(t_mask);
+  // Need q(x + mu) and q(x - mu).
+  tmp    = Cshift(q_in, mu, 1);
+  tmpFwd = tmp*lattice_cmplx;
+  tmp    = lattice_cmplx*q_in;
+  tmpBwd = Cshift(tmp, mu, -1);
 
-        if (timeSlices > 0)
-        {
-            Kernels::SeqConservedCurrentSiteFwd(tmpFwd._odata[sU], 
-                                                q_out._odata[sU], 
-                                                Umu, sU, mu, t_mask);
-        }
+  auto coords_v = coords.View();
+  auto tmpFwd_v = tmpFwd.View();
+  auto tmpBwd_v = tmpBwd.View();
+  auto Umu_v    = Umu.View();
+  auto q_out_v  = q_out.View();
 
-        // Repeat for backward direction.
-        t_mask     = ((coords._odata[sU] >= (tmin + tshift)) && 
-                      (coords._odata[sU] <= (tmax + tshift)));
+  thread_for(sU, Umu.Grid()->oSites(), {
 
-	//if tmax = LLt-1 (last timeslice) include timeslice 0 if the time is shifted (mu=3)	
-	unsigned int t0 = 0;
-	if((tmax==LLt-1) && (tshift==1)) t_mask = (t_mask || (coords._odata[sU] == t0 ));
+    // Compute the sequential conserved current insertion only if our simd
+    // object contains a timeslice we need.
+    vPredicate t_mask;
+    t_mask() = ((coords_v[sU] >= tmin) && (coords_v[sU] <= tmax));
+    Integer timeSlices = Reduce(t_mask());
 
-        timeSlices = Reduce(t_mask);
-
-        if (timeSlices > 0)
-        {
-            Kernels::SeqConservedCurrentSiteBwd(tmpBwd._odata[sU], 
-                                                q_out._odata[sU], 
-                                                Umu, sU, mu, t_mask);
-        }
+    if (timeSlices > 0) {
+      Kernels::SeqConservedCurrentSiteFwd(tmpFwd_v[sU], 
+					  q_out_v[sU], 
+					  Umu_v, sU, mu, t_mask);
     }
 
+    // Repeat for backward direction.
+    t_mask()     = ((coords_v[sU] >= (tmin + tshift)) && 
+		    (coords_v[sU] <= (tmax + tshift)));
+    
+    //if tmax = LLt-1 (last timeslice) include timeslice 0 if the time is shifted (mu=3)	
+    unsigned int t0 = 0;
+    if((tmax==LLt-1) && (tshift==1)) t_mask() = (t_mask() || (coords_v[sU] == t0 ));
+    
+    timeSlices = Reduce(t_mask());
 
+    if (timeSlices > 0) {
+      Kernels::SeqConservedCurrentSiteBwd(tmpBwd_v[sU], 
+					  q_out_v[sU], 
+					  Umu_v, sU, mu, t_mask);
+    }
+  });
 }
 
-FermOpTemplateInstantiate(WilsonFermion);
-AdjointFermOpTemplateInstantiate(WilsonFermion);
-TwoIndexFermOpTemplateInstantiate(WilsonFermion);
-GparityFermOpTemplateInstantiate(WilsonFermion);
-}
-}
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmAvx512.h b/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmAvx512.h
new file mode 100644
index 00000000..4aed13bf
--- /dev/null
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmAvx512.h
@@ -0,0 +1,716 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+
+
+    Source file: ./lib/qcd/action/fermion/WilsonKernelsAsmAvx512.h
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+#if defined(AVX512) 
+    ///////////////////////////////////////////////////////////
+    // If we are AVX512 specialise the single precision routine
+    ///////////////////////////////////////////////////////////
+#include <simd/Intel512wilson.h>
+#include <simd/Intel512single.h>
+
+/// Switch off the 5d vectorised code optimisations
+#undef DWFVEC5D
+
+static Vector<vComplexF> signsF;
+
+  template<typename vtype>    
+  int setupSigns(Vector<vtype>& signs ){
+    Vector<vtype> bother(2);
+    signs = bother;
+    vrsign(signs[0]);
+    visign(signs[1]);
+    return 1;
+  }
+
+  static int signInitF = setupSigns(signsF);
+
+#define MAYBEPERM(A,perm) if (perm) { A ; }
+#define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN(ptr,pf)
+#define COMPLEX_SIGNS(isigns) vComplexF *isigns = &signsF[0];  
+  
+/////////////////////////////////////////////////////////////////
+// XYZT vectorised, undag Kernel, single
+/////////////////////////////////////////////////////////////////
+#undef KERNEL_DAG
+#define INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<WilsonImplF>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<ZWilsonImplF>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<WilsonImplFH>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<ZWilsonImplFH>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#define INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<WilsonImplF>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<ZWilsonImplF>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<WilsonImplFH>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<ZWilsonImplFH>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+
+#undef INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#define EXTERIOR
+template<> void 
+WilsonKernels<WilsonImplF>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<ZWilsonImplF>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<WilsonImplFH>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<ZWilsonImplFH>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+      
+/////////////////////////////////////////////////////////////////
+// XYZT vectorised, dag Kernel, single
+/////////////////////////////////////////////////////////////////
+#define KERNEL_DAG
+#define INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<WilsonImplF>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<ZWilsonImplF>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<WilsonImplFH>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<ZWilsonImplFH>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#define INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<WilsonImplF>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<ZWilsonImplF>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<WilsonImplFH>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<ZWilsonImplFH>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#define EXTERIOR
+template<> void 
+WilsonKernels<WilsonImplF>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+				    
+template<> void 
+WilsonKernels<ZWilsonImplF>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+				    
+template<> void 
+WilsonKernels<WilsonImplFH>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+				    
+template<> void 
+WilsonKernels<ZWilsonImplFH>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+				    
+#undef MAYBEPERM
+#undef MULT_2SPIN
+#define MAYBEPERM(A,B) 
+#define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN_LS(ptr,pf)
+				    
+/////////////////////////////////////////////////////////////////
+// Ls vectorised, undag Kernel, single
+/////////////////////////////////////////////////////////////////
+
+#ifdef DWFVEC5D
+
+#undef KERNEL_DAG
+#define INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<DomainWallVec5dImplF>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<DomainWallVec5dImplFH>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplFH>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#define INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<DomainWallVec5dImplF>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<DomainWallVec5dImplFH>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplFH>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#define EXTERIOR
+#undef  MULT_2SPIN
+#define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN_LSNOPF(ptr,pf)
+template<> void 
+WilsonKernels<DomainWallVec5dImplF>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+				    
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+				    
+template<> void 
+WilsonKernels<DomainWallVec5dImplFH>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+				    
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplFH>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+				    
+/////////////////////////////////////////////////////////////////
+// Ls vectorised, dag Kernel, single
+/////////////////////////////////////////////////////////////////
+#define KERNEL_DAG
+#define INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<DomainWallVec5dImplF>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<DomainWallVec5dImplFH>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplFH>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#define INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<DomainWallVec5dImplF>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<DomainWallVec5dImplFH>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplFH>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#define EXTERIOR
+template<> void 
+WilsonKernels<DomainWallVec5dImplF>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<DomainWallVec5dImplFH>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplFH>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#endif  // VEC 5D
+
+#undef COMPLEX_SIGNS
+#undef MAYBEPERM
+#undef MULT_2SPIN
+	
+
+
+///////////////////////////////////////////////////////////
+// If we are AVX512 specialise the double precision routine
+///////////////////////////////////////////////////////////
+
+#include <simd/Intel512double.h>
+    
+static Vector<vComplexD> signsD;
+static int signInitD = setupSigns(signsD);
+    
+#define MAYBEPERM(A,perm) if (perm) { A ; }
+#define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN(ptr,pf)
+#define COMPLEX_SIGNS(isigns) vComplexD *isigns = &signsD[0];  
+
+
+#define INTERIOR_AND_EXTERIOR    
+#undef  INTERIOR
+#undef  EXTERIOR
+  
+/////////////////////////////////////////////////////////////////
+// XYZT vectorised, undag Kernel, single
+/////////////////////////////////////////////////////////////////
+#undef KERNEL_DAG
+#define INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<WilsonImplD>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZWilsonImplD>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<WilsonImplDF>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZWilsonImplDF>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#define INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<WilsonImplD>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZWilsonImplD>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<WilsonImplDF>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZWilsonImplDF>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#define EXTERIOR
+template<> void 
+WilsonKernels<WilsonImplD>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZWilsonImplD>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+      
+template<> void 
+WilsonKernels<WilsonImplDF>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZWilsonImplDF>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+      
+/////////////////////////////////////////////////////////////////
+// XYZT vectorised, dag Kernel, single
+/////////////////////////////////////////////////////////////////
+#define KERNEL_DAG
+#define INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<WilsonImplD>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZWilsonImplD>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<WilsonImplDF>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZWilsonImplDF>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#define INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<WilsonImplD>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZWilsonImplD>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<WilsonImplDF>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZWilsonImplDF>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#define EXTERIOR
+template<> void 
+WilsonKernels<WilsonImplD>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZWilsonImplD>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+				    
+template<> void 
+WilsonKernels<WilsonImplDF>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZWilsonImplDF>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+				    
+#undef MAYBEPERM
+#undef MULT_2SPIN
+#define MAYBEPERM(A,B) 
+#define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN_LS(ptr,pf)
+				    
+/////////////////////////////////////////////////////////////////
+// Ls vectorised, undag Kernel, single
+/////////////////////////////////////////////////////////////////
+#ifdef DWFVEC5D
+
+#undef KERNEL_DAG
+#define INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<DomainWallVec5dImplD>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<DomainWallVec5dImplDF>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplDF>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#define INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<DomainWallVec5dImplD>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<DomainWallVec5dImplDF>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplDF>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#define EXTERIOR
+#undef  MULT_2SPIN
+#define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN_LSNOPF(ptr,pf)
+template<> void 
+WilsonKernels<DomainWallVec5dImplD>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+				    
+template<> void 
+WilsonKernels<DomainWallVec5dImplDF>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplDF>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U, SiteHalfSpinor *buf,
+							 int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+				    
+/////////////////////////////////////////////////////////////////
+// Ls vectorised, dag Kernel, single
+/////////////////////////////////////////////////////////////////
+#define KERNEL_DAG
+#define INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<DomainWallVec5dImplD>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<DomainWallVec5dImplDF>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplDF>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#define INTERIOR
+#undef EXTERIOR
+template<> void 
+WilsonKernels<DomainWallVec5dImplD>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<DomainWallVec5dImplDF>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplDF>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#undef INTERIOR_AND_EXTERIOR
+#undef INTERIOR
+#define EXTERIOR
+
+template<> void 
+WilsonKernels<DomainWallVec5dImplD>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+template<> void 
+WilsonKernels<DomainWallVec5dImplDF>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+template<> void 
+WilsonKernels<ZDomainWallVec5dImplDF>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+							    int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
+
+#endif  // VEC 5D
+
+#undef COMPLEX_SIGNS
+#undef MAYBEPERM
+#undef MULT_2SPIN
+
+#undef Chi_00
+#undef Chi_01
+#undef Chi_02
+#undef Chi_10
+#undef Chi_11
+#undef Chi_12
+#undef Chi_20
+#undef Chi_21
+#undef Chi_22
+#undef Chi_30
+#undef Chi_31
+#undef Chi_32
+
+#undef UChi_00
+#undef UChi_01
+#undef UChi_02
+#undef UChi_10
+#undef UChi_11
+#undef UChi_12
+#undef UChi_20
+#undef UChi_21
+#undef UChi_22
+#undef UChi_30
+#undef UChi_31
+#undef UChi_32
+
+#undef Psi_00
+#undef Psi_01
+#undef Psi_02
+#undef Psi_10
+#undef Psi_11
+#undef Psi_12
+#undef Psi_20
+#undef Psi_21
+#undef Psi_22
+#undef Psi_30
+#undef Psi_31
+#undef Psi_32
+
+#undef Phi_00
+#undef Phi_01
+#undef Phi_02
+#undef Phi_10
+#undef Phi_11
+#undef Phi_12
+#undef Phi_20
+#undef Phi_21
+#undef Phi_22
+#undef Phi_30
+#undef Phi_31
+#undef Phi_32
+
+
+#endif //AVX512
diff --git a/Grid/qcd/action/fermion/WilsonKernelsAsmBody.h b/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmBody.h
similarity index 96%
rename from Grid/qcd/action/fermion/WilsonKernelsAsmBody.h
rename to Grid/qcd/action/fermion/implementation/WilsonKernelsAsmBody.h
index db8651ab..9f38bfc8 100644
--- a/Grid/qcd/action/fermion/WilsonKernelsAsmBody.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmBody.h
@@ -130,16 +130,18 @@
   int local,perm, ptype;
   uint64_t base;
   uint64_t basep;
-  const uint64_t plocal =(uint64_t) & in._odata[0];
+  const uint64_t plocal =(uint64_t) & in[0];
 
   COMPLEX_SIGNS(isigns);
   MASK_REGS;
-  int nmax=U._grid->oSites();
+  int nmax=U.oSites();
   for(int site=0;site<Ns;site++) {
 #ifndef EXTERIOR
-    int sU =lo.Reorder(ssU);
+    //    int sU =lo.Reorder(ssU);
+    int sU =ssU;
     int ssn=ssU+1;     if(ssn>=nmax) ssn=0;
-    int sUn=lo.Reorder(ssn);
+    //    int sUn=lo.Reorder(ssn);
+    int sUn=ssn;
     LOCK_GAUGE(0);
 #else
     int sU =ssU;
@@ -166,7 +168,7 @@
       if (nmu==0) break;
       //      if (nmu!=0) std::cout << "EXT "<<sU<<std::endl;
 #endif
-      base = (uint64_t) &out._odata[ss];
+      base = (uint64_t) &out[ss];
       basep= st.GetPFInfo(nent,plocal); nent++;
       RESULT(base,basep);
     }
diff --git a/Grid/qcd/action/fermion/WilsonKernelsAsmBody.h.ab b/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmBody.h.ab
similarity index 100%
rename from Grid/qcd/action/fermion/WilsonKernelsAsmBody.h.ab
rename to Grid/qcd/action/fermion/implementation/WilsonKernelsAsmBody.h.ab
diff --git a/Grid/qcd/action/fermion/WilsonKernelsAsmBody.h.abc b/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmBody.h.abc
similarity index 100%
rename from Grid/qcd/action/fermion/WilsonKernelsAsmBody.h.abc
rename to Grid/qcd/action/fermion/implementation/WilsonKernelsAsmBody.h.abc
diff --git a/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h
new file mode 100644
index 00000000..0c956f7e
--- /dev/null
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h
@@ -0,0 +1,86 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+
+
+    Source file: ./lib/qcd/action/fermion/WilsonKernelsAsm.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+
+#pragma once
+
+#include <Grid/qcd/action/fermion/FermionCore.h>
+
+NAMESPACE_BEGIN(Grid);
+
+///////////////////////////////////////////////////////////
+// Default to no assembler implementation
+// Will specialise to 
+///////////////////////////////////////////////////////////
+template<class Impl> void 
+WilsonKernels<Impl >::AsmDhopSite(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+				  int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+{
+  assert(0);
+}
+
+template<class Impl> void 
+WilsonKernels<Impl >::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+				     int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+{
+  assert(0);
+}
+
+template<class Impl> void 
+WilsonKernels<Impl >::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+				     int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+{
+  assert(0);
+}
+
+template<class Impl> void 
+WilsonKernels<Impl >::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+					int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+{
+  assert(0);
+}
+
+template<class Impl> void 
+WilsonKernels<Impl >::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+				     int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+{
+  assert(0);
+}
+
+template<class Impl> void 
+WilsonKernels<Impl >::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+					int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
+{
+  assert(0);
+}
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/WilsonKernelsAsmQPX.h b/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmQPX.h
similarity index 77%
rename from Grid/qcd/action/fermion/WilsonKernelsAsmQPX.h
rename to Grid/qcd/action/fermion/implementation/WilsonKernelsAsmQPX.h
index 612234d7..1f4a5f36 100644
--- a/Grid/qcd/action/fermion/WilsonKernelsAsmQPX.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmQPX.h
@@ -28,7 +28,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     See the full license in the file "LICENSE" in the top level distribution directory
 *************************************************************************************/
 /*  END LEGAL */
-
+#pragma once
 
 #if defined(QPX) 
 
@@ -52,18 +52,18 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 /////////////////////////////////////////////////////////////////
 #undef KERNEL_DAG
 template<> void 
-WilsonKernels<WilsonImplF>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
+WilsonKernels<WilsonImplF>::AsmDhopSite(StencilView &st, DoubledGaugeField &U, SiteHalfSpinor *buf,
 						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
       
 /////////////////////////////////////////////////////////////////
 // XYZT vectorised, dag Kernel, single
 /////////////////////////////////////////////////////////////////
 #define KERNEL_DAG
 template<> void 
-WilsonKernels<WilsonImplF>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
+WilsonKernels<WilsonImplF>::AsmDhopSiteDag(StencilView &st, DoubledGaugeField &U,SiteHalfSpinor *buf,
 						   int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
 				    
 #undef MAYBEPERM
 #undef MULT_2SPIN
@@ -75,18 +75,18 @@ WilsonKernels<WilsonImplF>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,Do
 /////////////////////////////////////////////////////////////////
 #undef KERNEL_DAG
 template<> void 
-WilsonKernels<DomainWallVec5dImplF>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
+WilsonKernels<DomainWallVec5dImplF>::AsmDhopSite(StencilView &st, DoubledGaugeField &U, SiteHalfSpinor *buf,
 							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
 				    
 /////////////////////////////////////////////////////////////////
 // Ls vectorised, dag Kernel, single
 /////////////////////////////////////////////////////////////////
 #define KERNEL_DAG
 template<> void 
-WilsonKernels<DomainWallVec5dImplF>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
+WilsonKernels<DomainWallVec5dImplF>::AsmDhopSiteDag(StencilView &st, DoubledGaugeField &U,SiteHalfSpinor *buf,
 							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
 #undef MAYBEPERM
 #undef MULT_2SPIN
 	
@@ -104,9 +104,9 @@ WilsonKernels<DomainWallVec5dImplF>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrde
 /////////////////////////////////////////////////////////////////
 #undef KERNEL_DAG
 template<> void 
-WilsonKernels<WilsonImplD>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
+WilsonKernels<WilsonImplD>::AsmDhopSite(StencilView &st, DoubledGaugeField &U, SiteHalfSpinor *buf,
 						int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
 /////////////////////////////////////////////////////////////////
       
 
@@ -115,9 +115,9 @@ WilsonKernels<WilsonImplD>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,Doubl
 /////////////////////////////////////////////////////////////////
 #define KERNEL_DAG
 template<> void 
-WilsonKernels<WilsonImplD>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
+WilsonKernels<WilsonImplD>::AsmDhopSiteDag(StencilView &st, DoubledGaugeField &U,SiteHalfSpinor *buf,
 						   int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
 /////////////////////////////////////////////////////////////////
 
 #undef MAYBEPERM
@@ -129,9 +129,9 @@ WilsonKernels<WilsonImplD>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,Do
 /////////////////////////////////////////////////////////////////
 #undef KERNEL_DAG
 template<> void 
-WilsonKernels<DomainWallVec5dImplD>::AsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, SiteHalfSpinor *buf,
+WilsonKernels<DomainWallVec5dImplD>::AsmDhopSite(StencilView &st, DoubledGaugeField &U, SiteHalfSpinor *buf,
 							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
 /////////////////////////////////////////////////////////////////
 				    
 /////////////////////////////////////////////////////////////////
@@ -139,9 +139,9 @@ WilsonKernels<DomainWallVec5dImplD>::AsmDhopSite(StencilImpl &st,LebesgueOrder &
 /////////////////////////////////////////////////////////////////
 #define KERNEL_DAG
 template<> void 
-WilsonKernels<DomainWallVec5dImplD>::AsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
+WilsonKernels<DomainWallVec5dImplD>::AsmDhopSiteDag(StencilView &st, DoubledGaugeField &U,SiteHalfSpinor *buf,
 							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-#include <qcd/action/fermion/WilsonKernelsAsmBody.h>
+#include <qcd/action/fermion/implementation/WilsonKernelsAsmBody.h>
 /////////////////////////////////////////////////////////////////
 	
 #undef MAYBEPERM
diff --git a/Grid/qcd/action/fermion/WilsonKernelsHandGparity.cc b/Grid/qcd/action/fermion/implementation/WilsonKernelsHandGparityImplementation.h
similarity index 79%
rename from Grid/qcd/action/fermion/WilsonKernelsHandGparity.cc
rename to Grid/qcd/action/fermion/implementation/WilsonKernelsHandGparityImplementation.h
index 3bf2f10a..2150938f 100644
--- a/Grid/qcd/action/fermion/WilsonKernelsHandGparity.cc
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsHandGparityImplementation.h
@@ -26,6 +26,9 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     See the full license in the file "LICENSE" in the top level distribution directory
     *************************************************************************************/
     /*  END LEGAL */
+
+#pragma once
+
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
 #define REGISTER
@@ -45,7 +48,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
   Chimu_32=ref(F)(3)(2)
 
 #define LOAD_CHIMU(DIR,F,PERM)						\
-  { const SiteSpinor & ref (in._odata[offset]); LOAD_CHIMU_BODY(F); }
+  { const SiteSpinor & ref (in[offset]); LOAD_CHIMU_BODY(F); }
 
 #define LOAD_CHI_BODY(F)				\
     Chi_00 = ref(F)(0)(0);\
@@ -92,9 +95,9 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
   g = F;								\
   direction = st._directions[DIR];				\
   distance = st._distances[DIR];				\
-  sl = st._grid->_simd_layout[direction];			\
+  sl = st._simd_layout[direction];			        \
   inplace_twist = 0;						\
-  if(SE->_around_the_world && this->Params.twists[DIR % 4]){		\
+  if(SE->_around_the_world && st.parameters.twists[DIR % 4]){		\
     if(sl == 1){							\
       g = (F+1) % 2;							\
     }else{								\
@@ -103,7 +106,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
   }  
 
 #define LOAD_CHIMU_GPARITY_INPLACE_TWIST(DIR,F,PERM)			\
-  { const SiteSpinor &ref(in._odata[offset]);				\
+  { const SiteSpinor &ref(in[offset]);				\
     LOAD_CHI_SETUP(DIR,F);						\
     if(!inplace_twist){							\
       LOAD_CHIMU_BODY(g);						\
@@ -201,10 +204,10 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 
 #define MULT_2SPIN(A,F)					\
-  {auto & ref(U._odata[sU](A)); MULT_2SPIN_BODY; }
+  {auto & ref(U[sU](A)); MULT_2SPIN_BODY; }
 
 #define MULT_2SPIN_GPARITY(A,F)				\
-  {auto & ref(U._odata[sU](F)(A)); MULT_2SPIN_BODY; }
+  {auto & ref(U[sU](F)(A)); MULT_2SPIN_BODY; }
 
 
 #define PERMUTE_DIR(dir)			\
@@ -468,8 +471,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define HAND_STENCIL_LEG_EXT(PROJ,PERM,DIR,RECON,F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL)	\
   SE=st.GetEntry(ptype,DIR,ss);			\
   offset = SE->_offset;				\
-  local  = SE->_is_local;			\
-  perm   = SE->_permute;			\
+  perm   = SE->_permute;				\
   if((!SE->_is_local)&&(!st.same_node[DIR]) ) {	\
     LOAD_CHI_IMPL(DIR,F,PERM);			\
     MULT_2SPIN_IMPL(DIR,F);			\
@@ -479,7 +481,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 #define HAND_RESULT(ss,F)			\
   {						\
-    SiteSpinor & ref (out._odata[ss]);		\
+    SiteSpinor & ref (out[ss]);		\
     vstream(ref(F)(0)(0),result_00);		\
     vstream(ref(F)(0)(1),result_01);		\
     vstream(ref(F)(0)(2),result_02);		\
@@ -496,7 +498,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 #define HAND_RESULT_EXT(ss,F)			\
   if (nmu){					\
-    SiteSpinor & ref (out._odata[ss]);		\
+    SiteSpinor & ref (out[ss]);		\
     ref(F)(0)(0)+=result_00;		\
     ref(F)(0)(1)+=result_01;		\
     ref(F)(0)(2)+=result_02;		\
@@ -545,18 +547,18 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
   Simd U_21;
 
 #define ZERO_RESULT				\
-  result_00=zero;				\
-  result_01=zero;				\
-  result_02=zero;				\
-  result_10=zero;				\
-  result_11=zero;				\
-  result_12=zero;				\
-  result_20=zero;				\
-  result_21=zero;				\
-  result_22=zero;				\
-  result_30=zero;				\
-  result_31=zero;				\
-  result_32=zero;			
+  result_00=Zero();				\
+  result_01=Zero();				\
+  result_02=Zero();				\
+  result_10=Zero();				\
+  result_11=Zero();				\
+  result_12=Zero();				\
+  result_20=Zero();				\
+  result_21=Zero();				\
+  result_22=Zero();				\
+  result_30=Zero();				\
+  result_31=Zero();				\
+  result_32=Zero();			
 
 #define Chimu_00 Chi_00
 #define Chimu_01 Chi_01
@@ -571,21 +573,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define Chimu_31 UChi_11
 #define Chimu_32 UChi_12
 
-namespace Grid {
-namespace QCD {
-
-template<class Impl> void 
-WilsonKernels<Impl>::HandDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor  *buf,
-					  int ss,int sU,const FermionField &in, FermionField &out)
-{
-// T==0, Z==1, Y==2, Z==3 expect 1,2,2,2 simd layout etc...
-  typedef typename Simd::scalar_type S;
-  typedef typename Simd::vector_type V;
-
-  HAND_DECLARATIONS(ignore);
-
-  int offset,local,perm, ptype;
-  StencilEntry *SE;
+NAMESPACE_BEGIN(Grid);
 
 #define HAND_DOP_SITE(F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL) \
   HAND_STENCIL_LEG(XM_PROJ,3,Xp,XM_RECON,F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL); \
@@ -598,21 +586,6 @@ WilsonKernels<Impl>::HandDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGauge
   HAND_STENCIL_LEG(TP_PROJ,0,Tm,TP_RECON_ACCUM,F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL); \
   HAND_RESULT(ss,F)
 
-  HAND_DOP_SITE(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN);
-}
-
-template<class Impl>
-void WilsonKernels<Impl>::HandDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-						  int ss,int sU,const FermionField &in, FermionField &out)
-{
-  typedef typename Simd::scalar_type S;
-  typedef typename Simd::vector_type V;
-
-  HAND_DECLARATIONS(ignore);
-
-  StencilEntry *SE;
-  int offset,local,perm, ptype;
-
 #define HAND_DOP_SITE_DAG(F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL) \
   HAND_STENCIL_LEG(XP_PROJ,3,Xp,XP_RECON,F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL); \
   HAND_STENCIL_LEG(YP_PROJ,2,Yp,YP_RECON_ACCUM,F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL); \
@@ -624,22 +597,6 @@ void WilsonKernels<Impl>::HandDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,Doub
   HAND_STENCIL_LEG(TM_PROJ,0,Tm,TM_RECON_ACCUM,F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL); \
   HAND_RESULT(ss,F)
 
-  HAND_DOP_SITE_DAG(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN);
-}
-
-template<class Impl> void 
-WilsonKernels<Impl>::HandDhopSiteInt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor  *buf,
-					  int ss,int sU,const FermionField &in, FermionField &out)
-{
-// T==0, Z==1, Y==2, Z==3 expect 1,2,2,2 simd layout etc...
-  typedef typename Simd::scalar_type S;
-  typedef typename Simd::vector_type V;
-
-  HAND_DECLARATIONS(ignore);
-
-  int offset,local,perm, ptype;
-  StencilEntry *SE;
-
 #define HAND_DOP_SITE_INT(F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL) \
   ZERO_RESULT; \
   HAND_STENCIL_LEG_INT(XM_PROJ,3,Xp,XM_RECON_ACCUM,F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL); \
@@ -652,21 +609,6 @@ WilsonKernels<Impl>::HandDhopSiteInt(StencilImpl &st,LebesgueOrder &lo,DoubledGa
   HAND_STENCIL_LEG_INT(TP_PROJ,0,Tm,TP_RECON_ACCUM,F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL); \
   HAND_RESULT(ss,F)
 
-  HAND_DOP_SITE_INT(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN);
-}
-
-template<class Impl>
-void WilsonKernels<Impl>::HandDhopSiteDagInt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-						  int ss,int sU,const FermionField &in, FermionField &out)
-{
-  typedef typename Simd::scalar_type S;
-  typedef typename Simd::vector_type V;
-
-  HAND_DECLARATIONS(ignore);
-
-  StencilEntry *SE;
-  int offset,local,perm, ptype;
-
 #define HAND_DOP_SITE_DAG_INT(F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL)				\
   ZERO_RESULT;							\
   HAND_STENCIL_LEG_INT(XP_PROJ,3,Xp,XP_RECON_ACCUM,F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL);		\
@@ -678,23 +620,6 @@ void WilsonKernels<Impl>::HandDhopSiteDagInt(StencilImpl &st,LebesgueOrder &lo,D
   HAND_STENCIL_LEG_INT(ZM_PROJ,1,Zm,ZM_RECON_ACCUM,F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL);		\
   HAND_STENCIL_LEG_INT(TM_PROJ,0,Tm,TM_RECON_ACCUM,F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL);		\
   HAND_RESULT(ss,F)
-  
-  HAND_DOP_SITE_DAG_INT(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN);
-}
-
-template<class Impl> void 
-WilsonKernels<Impl>::HandDhopSiteExt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor  *buf,
-					  int ss,int sU,const FermionField &in, FermionField &out)
-{
-// T==0, Z==1, Y==2, Z==3 expect 1,2,2,2 simd layout etc...
-  typedef typename Simd::scalar_type S;
-  typedef typename Simd::vector_type V;
-
-  HAND_DECLARATIONS(ignore);
-
-  int offset,local,perm, ptype;
-  StencilEntry *SE;
-  int nmu=0;
 
 #define HAND_DOP_SITE_EXT(F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL) \
   ZERO_RESULT; \
@@ -708,22 +633,6 @@ WilsonKernels<Impl>::HandDhopSiteExt(StencilImpl &st,LebesgueOrder &lo,DoubledGa
   HAND_STENCIL_LEG_EXT(TP_PROJ,0,Tm,TP_RECON_ACCUM,F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL); \
   HAND_RESULT_EXT(ss,F)
 
-  HAND_DOP_SITE_EXT(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN);
-}
-
-template<class Impl>
-void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-						  int ss,int sU,const FermionField &in, FermionField &out)
-{
-  typedef typename Simd::scalar_type S;
-  typedef typename Simd::vector_type V;
-
-  HAND_DECLARATIONS(ignore);
-
-  StencilEntry *SE;
-  int offset,local,perm, ptype;
-  int nmu=0;
-
 #define HAND_DOP_SITE_DAG_EXT(F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL) \
   ZERO_RESULT; \
   HAND_STENCIL_LEG_EXT(XP_PROJ,3,Xp,XP_RECON_ACCUM,F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL); \
@@ -736,13 +645,10 @@ void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilImpl &st,LebesgueOrder &lo,D
   HAND_STENCIL_LEG_EXT(TM_PROJ,0,Tm,TM_RECON_ACCUM,F,LOAD_CHI_IMPL,LOAD_CHIMU_IMPL,MULT_2SPIN_IMPL); \
   HAND_RESULT_EXT(ss,F)
 
-  HAND_DOP_SITE_DAG_EXT(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN);
-}
-
 #define HAND_SPECIALISE_GPARITY(IMPL)					\
-  template<> void							\
-  WilsonKernels<IMPL>::HandDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor  *buf, \
-				    int ss,int sU,const FermionField &in, FermionField &out) \
+  template<> void						\
+  WilsonKernels<IMPL>::HandDhopSite(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor  *buf, \
+				    int ss,int sU,const FermionFieldView &in, FermionFieldView &out) \
   {									\
     typedef IMPL Impl;							\
     typedef typename Simd::scalar_type S;				\
@@ -756,9 +662,9 @@ void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilImpl &st,LebesgueOrder &lo,D
     HAND_DOP_SITE(1, LOAD_CHI_GPARITY,LOAD_CHIMU_GPARITY,MULT_2SPIN_GPARITY); \
   }									\
 									\
-  template<>								\
-  void WilsonKernels<IMPL>::HandDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf, \
-					    int ss,int sU,const FermionField &in, FermionField &out) \
+  template<> void						\
+  WilsonKernels<IMPL>::HandDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf, \
+				       int ss,int sU,const FermionFieldView &in, FermionFieldView &out) \
   {									\
     typedef IMPL Impl;							\
     typedef typename Simd::scalar_type S;				\
@@ -772,9 +678,9 @@ void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilImpl &st,LebesgueOrder &lo,D
     HAND_DOP_SITE_DAG(1, LOAD_CHI_GPARITY,LOAD_CHIMU_GPARITY,MULT_2SPIN_GPARITY); \
   }									\
 									\
-  template<> void							\
-  WilsonKernels<IMPL>::HandDhopSiteInt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor  *buf, \
-						     int ss,int sU,const FermionField &in, FermionField &out) \
+  template<> void						\
+  WilsonKernels<IMPL>::HandDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor  *buf, \
+				       int ss,int sU,const FermionFieldView &in, FermionFieldView &out) \
   {									\
     typedef IMPL Impl;							\
     typedef typename Simd::scalar_type S;				\
@@ -788,9 +694,9 @@ void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilImpl &st,LebesgueOrder &lo,D
     HAND_DOP_SITE_INT(1, LOAD_CHI_GPARITY,LOAD_CHIMU_GPARITY,MULT_2SPIN_GPARITY); \
   }									\
 									\
-  template<>								\
-  void WilsonKernels<IMPL>::HandDhopSiteDagInt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf, \
-							     int ss,int sU,const FermionField &in, FermionField &out) \
+  template<> void						\
+  WilsonKernels<IMPL>::HandDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf, \
+					  int ss,int sU,const FermionFieldView &in, FermionFieldView &out) \
   {									\
     typedef IMPL Impl;							\
     typedef typename Simd::scalar_type S;				\
@@ -805,8 +711,8 @@ void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilImpl &st,LebesgueOrder &lo,D
   }									\
 									\
   template<> void							\
-  WilsonKernels<IMPL>::HandDhopSiteExt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor  *buf, \
-						     int ss,int sU,const FermionField &in, FermionField &out) \
+  WilsonKernels<IMPL>::HandDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor  *buf, \
+				       int ss,int sU,const FermionFieldView &in, FermionFieldView &out) \
   {									\
     typedef IMPL Impl;							\
     typedef typename Simd::scalar_type S;				\
@@ -814,16 +720,16 @@ void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilImpl &st,LebesgueOrder &lo,D
 									\
     HAND_DECLARATIONS(ignore);						\
 									\
-    int offset,local,perm, ptype, g, direction, distance, sl, inplace_twist; \
+    int offset,perm, ptype, g, direction, distance, sl, inplace_twist; \
     StencilEntry *SE;							\
     int nmu=0;								\
     HAND_DOP_SITE_EXT(0, LOAD_CHI_GPARITY,LOAD_CHIMU_GPARITY,MULT_2SPIN_GPARITY); \
     nmu = 0;								\
     HAND_DOP_SITE_EXT(1, LOAD_CHI_GPARITY,LOAD_CHIMU_GPARITY,MULT_2SPIN_GPARITY); \
   }									\
-  template<>								\
-  void WilsonKernels<IMPL>::HandDhopSiteDagExt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf, \
-							     int ss,int sU,const FermionField &in, FermionField &out) \
+  template<> void						\
+  WilsonKernels<IMPL>::HandDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf, \
+					  int ss,int sU,const FermionFieldView &in, FermionFieldView &out) \
   {									\
     typedef IMPL Impl;							\
     typedef typename Simd::scalar_type S;				\
@@ -832,47 +738,11 @@ void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilImpl &st,LebesgueOrder &lo,D
     HAND_DECLARATIONS(ignore);						\
 									\
     StencilEntry *SE;							\
-    int offset,local,perm, ptype, g, direction, distance, sl, inplace_twist; \
+    int offset,perm, ptype, g, direction, distance, sl, inplace_twist; \
     int nmu=0;								\
     HAND_DOP_SITE_DAG_EXT(0, LOAD_CHI_GPARITY,LOAD_CHIMU_GPARITY,MULT_2SPIN_GPARITY); \
     nmu = 0;								\
     HAND_DOP_SITE_DAG_EXT(1, LOAD_CHI_GPARITY,LOAD_CHIMU_GPARITY,MULT_2SPIN_GPARITY); \
   }
 
-
-HAND_SPECIALISE_GPARITY(GparityWilsonImplF);
-HAND_SPECIALISE_GPARITY(GparityWilsonImplD);
-HAND_SPECIALISE_GPARITY(GparityWilsonImplFH);
-HAND_SPECIALISE_GPARITY(GparityWilsonImplDF);
-
-
-
-
-
-
-
-
-
-
-  
-////////////// Wilson ; uses this implementation /////////////////////
-
-#define INSTANTIATE_THEM(A) \
-template void WilsonKernels<A>::HandDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf,\
-					     int ss,int sU,const FermionField &in, FermionField &out); \
-template void WilsonKernels<A>::HandDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf, \
-						int ss,int sU,const FermionField &in, FermionField &out);\
-template void WilsonKernels<A>::HandDhopSiteInt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf,\
-						int ss,int sU,const FermionField &in, FermionField &out); \
-template void WilsonKernels<A>::HandDhopSiteDagInt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf, \
-						   int ss,int sU,const FermionField &in, FermionField &out); \
-template void WilsonKernels<A>::HandDhopSiteExt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf,\
-						int ss,int sU,const FermionField &in, FermionField &out); \
-template void WilsonKernels<A>::HandDhopSiteDagExt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf, \
-						   int ss,int sU,const FermionField &in, FermionField &out); 
-
-INSTANTIATE_THEM(GparityWilsonImplF);
-INSTANTIATE_THEM(GparityWilsonImplD);
-INSTANTIATE_THEM(GparityWilsonImplFH);
-INSTANTIATE_THEM(GparityWilsonImplDF);
-}}
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/WilsonKernelsHand.cc b/Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h
similarity index 81%
rename from Grid/qcd/action/fermion/WilsonKernelsHand.cc
rename to Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h
index 50816495..f7b018fa 100644
--- a/Grid/qcd/action/fermion/WilsonKernelsHand.cc
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h
@@ -26,12 +26,58 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     See the full license in the file "LICENSE" in the top level distribution directory
     *************************************************************************************/
     /*  END LEGAL */
+
+#pragma once
+
 #include <Grid/qcd/action/fermion/FermionCore.h>
 
+
+#undef LOAD_CHIMU  
+#undef LOAD_CHI 
+#undef MULT_2SPIN
+#undef PERMUTE_DIR
+#undef XP_PROJ  
+#undef YP_PROJ  
+#undef ZP_PROJ  
+#undef TP_PROJ  
+#undef XM_PROJ  
+#undef YM_PROJ  
+#undef ZM_PROJ  
+#undef TM_PROJ  
+#undef XP_RECON 
+#undef XP_RECON_ACCUM 
+#undef XM_RECON 
+#undef XM_RECON_ACCUM 
+#undef YP_RECON_ACCUM 
+#undef YM_RECON_ACCUM 
+#undef ZP_RECON_ACCUM 
+#undef ZM_RECON_ACCUM 
+#undef TP_RECON_ACCUM 
+#undef TM_RECON_ACCUM 
+#undef ZERO_RESULT				 
+#undef Chimu_00
+#undef Chimu_01
+#undef Chimu_02
+#undef Chimu_10
+#undef Chimu_11
+#undef Chimu_12
+#undef Chimu_20
+#undef Chimu_21
+#undef Chimu_22
+#undef Chimu_30
+#undef Chimu_31
+#undef Chimu_32
+#undef HAND_STENCIL_LEG
+#undef HAND_STENCIL_LEG_INT
+#undef HAND_STENCIL_LEG_EXT
+#undef HAND_RESULT
+#undef HAND_RESULT_INT
+#undef HAND_RESULT_EXT
+
 #define REGISTER
 
 #define LOAD_CHIMU \
-  {const SiteSpinor & ref (in._odata[offset]);	\
+  {const SiteSpinor & ref (in[offset]);	\
     Chimu_00=ref()(0)(0);\
     Chimu_01=ref()(0)(1);\
     Chimu_02=ref()(0)(2);\
@@ -56,7 +102,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 // To splat or not to splat depends on the implementation
 #define MULT_2SPIN(A)\
-  {auto & ref(U._odata[sU](A));			\
+  {auto & ref(U[sU](A));			\
    Impl::loadLinkElement(U_00,ref()(0,0));	\
    Impl::loadLinkElement(U_10,ref()(1,0));	\
    Impl::loadLinkElement(U_20,ref()(2,0));	\
@@ -355,7 +401,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 #define HAND_RESULT(ss)				\
   {						\
-    SiteSpinor & ref (out._odata[ss]);		\
+    SiteSpinor & ref (out[ss]);		\
     vstream(ref()(0)(0),result_00);		\
     vstream(ref()(0)(1),result_01);		\
     vstream(ref()(0)(2),result_02);		\
@@ -372,7 +418,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 #define HAND_RESULT_EXT(ss)			\
   if (nmu){					\
-    SiteSpinor & ref (out._odata[ss]);		\
+    SiteSpinor & ref (out[ss]);		\
     ref()(0)(0)+=result_00;		\
     ref()(0)(1)+=result_01;		\
     ref()(0)(2)+=result_02;		\
@@ -421,18 +467,18 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
   Simd U_21;
 
 #define ZERO_RESULT				\
-  result_00=zero;				\
-  result_01=zero;				\
-  result_02=zero;				\
-  result_10=zero;				\
-  result_11=zero;				\
-  result_12=zero;				\
-  result_20=zero;				\
-  result_21=zero;				\
-  result_22=zero;				\
-  result_30=zero;				\
-  result_31=zero;				\
-  result_32=zero;			
+  result_00=Zero();				\
+  result_01=Zero();				\
+  result_02=Zero();				\
+  result_10=Zero();				\
+  result_11=Zero();				\
+  result_12=Zero();				\
+  result_20=Zero();				\
+  result_21=Zero();				\
+  result_22=Zero();				\
+  result_30=Zero();				\
+  result_31=Zero();				\
+  result_32=Zero();			
 
 #define Chimu_00 Chi_00
 #define Chimu_01 Chi_01
@@ -447,12 +493,11 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define Chimu_31 UChi_11
 #define Chimu_32 UChi_12
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 template<class Impl> void 
-WilsonKernels<Impl>::HandDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor  *buf,
-					  int ss,int sU,const FermionField &in, FermionField &out)
+WilsonKernels<Impl>::HandDhopSite(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor  *buf,
+				  int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
 {
 // T==0, Z==1, Y==2, Z==3 expect 1,2,2,2 simd layout etc...
   typedef typename Simd::scalar_type S;
@@ -475,8 +520,8 @@ WilsonKernels<Impl>::HandDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGauge
 }
 
 template<class Impl>
-void WilsonKernels<Impl>::HandDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-						  int ss,int sU,const FermionField &in, FermionField &out)
+void WilsonKernels<Impl>::HandDhopSiteDag(StencilView &st,DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+					  int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
 {
   typedef typename Simd::scalar_type S;
   typedef typename Simd::vector_type V;
@@ -498,8 +543,8 @@ void WilsonKernels<Impl>::HandDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,Doub
 }
 
 template<class Impl> void 
-WilsonKernels<Impl>::HandDhopSiteInt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor  *buf,
-					  int ss,int sU,const FermionField &in, FermionField &out)
+WilsonKernels<Impl>::HandDhopSiteInt(StencilView &st,DoubledGaugeFieldView &U,SiteHalfSpinor  *buf,
+					  int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
 {
 // T==0, Z==1, Y==2, Z==3 expect 1,2,2,2 simd layout etc...
   typedef typename Simd::scalar_type S;
@@ -522,8 +567,8 @@ WilsonKernels<Impl>::HandDhopSiteInt(StencilImpl &st,LebesgueOrder &lo,DoubledGa
 }
 
 template<class Impl>
-void WilsonKernels<Impl>::HandDhopSiteDagInt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-						  int ss,int sU,const FermionField &in, FermionField &out)
+void WilsonKernels<Impl>::HandDhopSiteDagInt(StencilView &st,DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+						  int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
 {
   typedef typename Simd::scalar_type S;
   typedef typename Simd::vector_type V;
@@ -545,8 +590,8 @@ void WilsonKernels<Impl>::HandDhopSiteDagInt(StencilImpl &st,LebesgueOrder &lo,D
 }
 
 template<class Impl> void 
-WilsonKernels<Impl>::HandDhopSiteExt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor  *buf,
-					  int ss,int sU,const FermionField &in, FermionField &out)
+WilsonKernels<Impl>::HandDhopSiteExt(StencilView &st,DoubledGaugeFieldView &U,SiteHalfSpinor  *buf,
+					  int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
 {
 // T==0, Z==1, Y==2, Z==3 expect 1,2,2,2 simd layout etc...
   typedef typename Simd::scalar_type S;
@@ -554,7 +599,7 @@ WilsonKernels<Impl>::HandDhopSiteExt(StencilImpl &st,LebesgueOrder &lo,DoubledGa
 
   HAND_DECLARATIONS(ignore);
 
-  int offset,local,perm, ptype;
+  int offset, ptype;
   StencilEntry *SE;
   int nmu=0;
   ZERO_RESULT;
@@ -570,8 +615,8 @@ WilsonKernels<Impl>::HandDhopSiteExt(StencilImpl &st,LebesgueOrder &lo,DoubledGa
 }
 
 template<class Impl>
-void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf,
-						  int ss,int sU,const FermionField &in, FermionField &out)
+void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilView &st,DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
+						  int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
 {
   typedef typename Simd::scalar_type S;
   typedef typename Simd::vector_type V;
@@ -579,7 +624,7 @@ void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilImpl &st,LebesgueOrder &lo,D
   HAND_DECLARATIONS(ignore);
 
   StencilEntry *SE;
-  int offset,local,perm, ptype;
+  int offset, ptype;
   int nmu=0;
   ZERO_RESULT;
   HAND_STENCIL_LEG_EXT(XP_PROJ,3,Xp,XP_RECON_ACCUM);
@@ -595,37 +640,45 @@ void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilImpl &st,LebesgueOrder &lo,D
 
 ////////////// Wilson ; uses this implementation /////////////////////
 
-#define INSTANTIATE_THEM(A) \
-template void WilsonKernels<A>::HandDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf,\
-					     int ss,int sU,const FermionField &in, FermionField &out); \
-template void WilsonKernels<A>::HandDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf, \
-						int ss,int sU,const FermionField &in, FermionField &out);\
-template void WilsonKernels<A>::HandDhopSiteInt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf,\
-						int ss,int sU,const FermionField &in, FermionField &out); \
-template void WilsonKernels<A>::HandDhopSiteDagInt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf, \
-						   int ss,int sU,const FermionField &in, FermionField &out); \
-template void WilsonKernels<A>::HandDhopSiteExt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf,\
-						int ss,int sU,const FermionField &in, FermionField &out); \
-template void WilsonKernels<A>::HandDhopSiteDagExt(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,SiteHalfSpinor *buf, \
-						   int ss,int sU,const FermionField &in, FermionField &out); 
-
-INSTANTIATE_THEM(WilsonImplF);
-INSTANTIATE_THEM(WilsonImplD);
-INSTANTIATE_THEM(ZWilsonImplF);
-INSTANTIATE_THEM(ZWilsonImplD);
-INSTANTIATE_THEM(DomainWallVec5dImplF);
-INSTANTIATE_THEM(DomainWallVec5dImplD);
-INSTANTIATE_THEM(ZDomainWallVec5dImplF);
-INSTANTIATE_THEM(ZDomainWallVec5dImplD);
-INSTANTIATE_THEM(WilsonImplFH);
-INSTANTIATE_THEM(WilsonImplDF);
-INSTANTIATE_THEM(ZWilsonImplFH);
-INSTANTIATE_THEM(ZWilsonImplDF);
-INSTANTIATE_THEM(DomainWallVec5dImplFH);
-INSTANTIATE_THEM(DomainWallVec5dImplDF);
-INSTANTIATE_THEM(ZDomainWallVec5dImplFH);
-INSTANTIATE_THEM(ZDomainWallVec5dImplDF);
-INSTANTIATE_THEM(WilsonTwoIndexAntiSymmetricImplF);
-INSTANTIATE_THEM(WilsonTwoIndexAntiSymmetricImplD);
-
-}}
+NAMESPACE_END(Grid);
+#undef LOAD_CHIMU  
+#undef LOAD_CHI 
+#undef MULT_2SPIN
+#undef PERMUTE_DIR
+#undef XP_PROJ  
+#undef YP_PROJ  
+#undef ZP_PROJ  
+#undef TP_PROJ  
+#undef XM_PROJ  
+#undef YM_PROJ  
+#undef ZM_PROJ  
+#undef TM_PROJ  
+#undef XP_RECON 
+#undef XP_RECON_ACCUM 
+#undef XM_RECON 
+#undef XM_RECON_ACCUM 
+#undef YP_RECON_ACCUM 
+#undef YM_RECON_ACCUM 
+#undef ZP_RECON_ACCUM 
+#undef ZM_RECON_ACCUM 
+#undef TP_RECON_ACCUM 
+#undef TM_RECON_ACCUM 
+#undef ZERO_RESULT				 
+#undef Chimu_00
+#undef Chimu_01
+#undef Chimu_02
+#undef Chimu_10
+#undef Chimu_11
+#undef Chimu_12
+#undef Chimu_20
+#undef Chimu_21
+#undef Chimu_22
+#undef Chimu_30
+#undef Chimu_31
+#undef Chimu_32
+#undef HAND_STENCIL_LEG
+#undef HAND_STENCIL_LEG_INT
+#undef HAND_STENCIL_LEG_EXT
+#undef HAND_RESULT
+#undef HAND_RESULT_INT
+#undef HAND_RESULT_EXT
diff --git a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
new file mode 100644
index 00000000..a787fa79
--- /dev/null
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
@@ -0,0 +1,551 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+#include <Grid/qcd/action/fermion/FermionCore.h>
+
+NAMESPACE_BEGIN(Grid);
+
+
+////////////////////////////////////////////
+// Generic implementation; move to different file?
+////////////////////////////////////////////
+
+accelerator_inline void get_stencil(StencilEntry * mem, StencilEntry &chip)
+{
+#ifdef __CUDA_ARCH__
+  static_assert(sizeof(StencilEntry)==sizeof(uint4),"Unexpected Stencil Entry Size"); 
+  uint4 * mem_pun  = (uint4 *)mem; // force 128 bit loads
+  uint4 * chip_pun = (uint4 *)&chip;
+  * chip_pun = * mem_pun;
+#else 
+  chip = *mem;
+#endif
+  return;
+}
+  
+#define GENERIC_STENCIL_LEG(Dir,spProj,Recon)			\
+  SE = st.GetEntry(ptype, Dir, sF);				\
+  if (SE->_is_local) {						\
+    int perm= SE->_permute;					\
+    auto tmp = coalescedReadPermute(in[SE->_offset],ptype,perm,lane);	\
+    spProj(chi,tmp);						\
+  } else {							\
+    chi = coalescedRead(buf[SE->_offset],lane);			\
+  }								\
+  synchronise();						\
+  Impl::multLink(Uchi, U[sU], chi, Dir, SE, st);		\
+  Recon(result, Uchi);
+  
+#define GENERIC_STENCIL_LEG_INT(Dir,spProj,Recon)		\
+  SE = st.GetEntry(ptype, Dir, sF);				\
+  if (SE->_is_local) {						\
+    int perm= SE->_permute;					\
+    auto tmp = coalescedReadPermute(in[SE->_offset],ptype,perm,lane);	\
+    spProj(chi,tmp);						\
+  } else if ( st.same_node[Dir] ) {				\
+    chi = coalescedRead(buf[SE->_offset],lane);			\
+  }								\
+  synchronise();						\
+  if (SE->_is_local || st.same_node[Dir] ) {			\
+    Impl::multLink(Uchi, U[sU], chi, Dir, SE, st);		\
+    Recon(result, Uchi);					\
+  }								\
+  synchronise();						
+
+#define GENERIC_STENCIL_LEG_EXT(Dir,spProj,Recon)		\
+  SE = st.GetEntry(ptype, Dir, sF);				\
+  if ((!SE->_is_local) && (!st.same_node[Dir]) ) {		\
+    auto chi = coalescedRead(buf[SE->_offset],lane);		\
+    Impl::multLink(Uchi, U[sU], chi, Dir, SE, st);		\
+    Recon(result, Uchi);					\
+    nmu++;							\
+  }								\
+  synchronise();						
+
+#define GENERIC_DHOPDIR_LEG(Dir,spProj,Recon)			\
+  if (gamma == Dir) {						\
+    if (SE->_is_local ) {					\
+      int perm= SE->_permute;					\
+      auto tmp = coalescedReadPermute(in[SE->_offset],ptype,perm,lane);	\
+      spProj(chi,tmp);						\
+    } else {							\
+      chi = coalescedRead(buf[SE->_offset],lane);		\
+    }								\
+    synchronise();						\
+    Impl::multLink(Uchi, U[sU], chi, dir, SE, st);		\
+    Recon(result, Uchi);					\
+    synchronise();						\
+  }
+
+  ////////////////////////////////////////////////////////////////////
+  // All legs kernels ; comms then compute
+  ////////////////////////////////////////////////////////////////////
+template <class Impl>
+void WilsonKernels<Impl>::GenericDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,
+					     SiteHalfSpinor *buf, int sF,
+					     int sU, const FermionFieldView &in, FermionFieldView &out)
+{
+  typedef decltype(coalescedRead(buf[0]))   calcHalfSpinor;
+  typedef decltype(coalescedRead(in[0])) calcSpinor;
+  calcHalfSpinor chi;
+  //  calcHalfSpinor *chi_p;
+  calcHalfSpinor Uchi;
+  calcSpinor result;
+  StencilEntry *SE;
+  int ptype;
+  const int Nsimd = SiteHalfSpinor::Nsimd();
+  const int lane=SIMTlane(Nsimd);
+  GENERIC_STENCIL_LEG(Xp,spProjXp,spReconXp);
+  GENERIC_STENCIL_LEG(Yp,spProjYp,accumReconYp);
+  GENERIC_STENCIL_LEG(Zp,spProjZp,accumReconZp);
+  GENERIC_STENCIL_LEG(Tp,spProjTp,accumReconTp);
+  GENERIC_STENCIL_LEG(Xm,spProjXm,accumReconXm);
+  GENERIC_STENCIL_LEG(Ym,spProjYm,accumReconYm);
+  GENERIC_STENCIL_LEG(Zm,spProjZm,accumReconZm);
+  GENERIC_STENCIL_LEG(Tm,spProjTm,accumReconTm);
+  coalescedWrite(out[sF],result,lane);
+};
+
+template <class Impl>
+void WilsonKernels<Impl>::GenericDhopSite(StencilView &st, DoubledGaugeFieldView &U,
+					  SiteHalfSpinor *buf, int sF,
+					  int sU, const FermionFieldView &in, FermionFieldView &out) 
+{
+  typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
+  typedef decltype(coalescedRead(in[0]))  calcSpinor;
+  calcHalfSpinor chi;
+  //  calcHalfSpinor *chi_p;
+  calcHalfSpinor Uchi;
+  calcSpinor result;
+  StencilEntry *SE;
+  int ptype;
+
+  const int Nsimd = SiteHalfSpinor::Nsimd();
+  const int lane=SIMTlane(Nsimd);
+  GENERIC_STENCIL_LEG(Xm,spProjXp,spReconXp);
+  GENERIC_STENCIL_LEG(Ym,spProjYp,accumReconYp);
+  GENERIC_STENCIL_LEG(Zm,spProjZp,accumReconZp);
+  GENERIC_STENCIL_LEG(Tm,spProjTp,accumReconTp);
+  GENERIC_STENCIL_LEG(Xp,spProjXm,accumReconXm);
+  GENERIC_STENCIL_LEG(Yp,spProjYm,accumReconYm);
+  GENERIC_STENCIL_LEG(Zp,spProjZm,accumReconZm);
+  GENERIC_STENCIL_LEG(Tp,spProjTm,accumReconTm);
+  coalescedWrite(out[sF], result,lane);
+};
+  ////////////////////////////////////////////////////////////////////
+  // Interior kernels
+  ////////////////////////////////////////////////////////////////////
+template <class Impl>
+void WilsonKernels<Impl>::GenericDhopSiteDagInt(StencilView &st,  DoubledGaugeFieldView &U,
+						SiteHalfSpinor *buf, int sF,
+						int sU, const FermionFieldView &in, FermionFieldView &out)
+{
+  typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
+  typedef decltype(coalescedRead(in[0]))  calcSpinor;
+  calcHalfSpinor chi;
+  //  calcHalfSpinor *chi_p;
+  calcHalfSpinor Uchi;
+  calcSpinor result;
+  StencilEntry *SE;
+  int ptype;
+  const int Nsimd = SiteHalfSpinor::Nsimd();
+  const int lane=SIMTlane(Nsimd);
+
+  result=Zero();
+  GENERIC_STENCIL_LEG_INT(Xp,spProjXp,accumReconXp);
+  GENERIC_STENCIL_LEG_INT(Yp,spProjYp,accumReconYp);
+  GENERIC_STENCIL_LEG_INT(Zp,spProjZp,accumReconZp);
+  GENERIC_STENCIL_LEG_INT(Tp,spProjTp,accumReconTp);
+  GENERIC_STENCIL_LEG_INT(Xm,spProjXm,accumReconXm);
+  GENERIC_STENCIL_LEG_INT(Ym,spProjYm,accumReconYm);
+  GENERIC_STENCIL_LEG_INT(Zm,spProjZm,accumReconZm);
+  GENERIC_STENCIL_LEG_INT(Tm,spProjTm,accumReconTm);
+  coalescedWrite(out[sF], result,lane);
+};
+
+template <class Impl>
+void WilsonKernels<Impl>::GenericDhopSiteInt(StencilView &st,  DoubledGaugeFieldView &U,
+							 SiteHalfSpinor *buf, int sF,
+							 int sU, const FermionFieldView &in, FermionFieldView &out) 
+{
+  typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
+  typedef decltype(coalescedRead(in[0]))  calcSpinor;
+  const int Nsimd = SiteHalfSpinor::Nsimd();
+  const int lane=SIMTlane(Nsimd);
+
+  calcHalfSpinor chi;
+  //  calcHalfSpinor *chi_p;
+  calcHalfSpinor Uchi;
+  calcSpinor result;
+  StencilEntry *SE;
+  int ptype;
+  result=Zero();
+  GENERIC_STENCIL_LEG_INT(Xm,spProjXp,accumReconXp);
+  GENERIC_STENCIL_LEG_INT(Ym,spProjYp,accumReconYp);
+  GENERIC_STENCIL_LEG_INT(Zm,spProjZp,accumReconZp);
+  GENERIC_STENCIL_LEG_INT(Tm,spProjTp,accumReconTp);
+  GENERIC_STENCIL_LEG_INT(Xp,spProjXm,accumReconXm);
+  GENERIC_STENCIL_LEG_INT(Yp,spProjYm,accumReconYm);
+  GENERIC_STENCIL_LEG_INT(Zp,spProjZm,accumReconZm);
+  GENERIC_STENCIL_LEG_INT(Tp,spProjTm,accumReconTm);
+  coalescedWrite(out[sF], result,lane);
+};
+////////////////////////////////////////////////////////////////////
+// Exterior kernels
+////////////////////////////////////////////////////////////////////
+template <class Impl>
+void WilsonKernels<Impl>::GenericDhopSiteDagExt(StencilView &st,  DoubledGaugeFieldView &U,
+						SiteHalfSpinor *buf, int sF,
+						int sU, const FermionFieldView &in, FermionFieldView &out)
+{
+  typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
+  typedef decltype(coalescedRead(in[0]))  calcSpinor;
+  //  calcHalfSpinor *chi_p;
+  calcHalfSpinor Uchi;
+  calcSpinor result;
+  StencilEntry *SE;
+  int ptype;
+  int nmu=0;
+  const int Nsimd = SiteHalfSpinor::Nsimd();
+  const int lane=SIMTlane(Nsimd);
+  result=Zero();
+  GENERIC_STENCIL_LEG_EXT(Xp,spProjXp,accumReconXp);
+  GENERIC_STENCIL_LEG_EXT(Yp,spProjYp,accumReconYp);
+  GENERIC_STENCIL_LEG_EXT(Zp,spProjZp,accumReconZp);
+  GENERIC_STENCIL_LEG_EXT(Tp,spProjTp,accumReconTp);
+  GENERIC_STENCIL_LEG_EXT(Xm,spProjXm,accumReconXm);
+  GENERIC_STENCIL_LEG_EXT(Ym,spProjYm,accumReconYm);
+  GENERIC_STENCIL_LEG_EXT(Zm,spProjZm,accumReconZm);
+  GENERIC_STENCIL_LEG_EXT(Tm,spProjTm,accumReconTm);
+  if ( nmu ) { 
+    auto out_t = coalescedRead(out[sF],lane);
+    out_t = out_t + result;
+    coalescedWrite(out[sF],out_t,lane);
+  }
+};
+
+template <class Impl>
+void WilsonKernels<Impl>::GenericDhopSiteExt(StencilView &st,  DoubledGaugeFieldView &U,
+					     SiteHalfSpinor *buf, int sF,
+					     int sU, const FermionFieldView &in, FermionFieldView &out) 
+{
+  typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
+  typedef decltype(coalescedRead(in[0]))  calcSpinor;
+  //  calcHalfSpinor *chi_p;
+  calcHalfSpinor Uchi;
+  calcSpinor result;
+  StencilEntry *SE;
+  int ptype;
+  int nmu=0;
+  const int Nsimd = SiteHalfSpinor::Nsimd();
+  const int lane=SIMTlane(Nsimd);
+  result=Zero();
+  GENERIC_STENCIL_LEG_EXT(Xm,spProjXp,accumReconXp);
+  GENERIC_STENCIL_LEG_EXT(Ym,spProjYp,accumReconYp);
+  GENERIC_STENCIL_LEG_EXT(Zm,spProjZp,accumReconZp);
+  GENERIC_STENCIL_LEG_EXT(Tm,spProjTp,accumReconTp);
+  GENERIC_STENCIL_LEG_EXT(Xp,spProjXm,accumReconXm);
+  GENERIC_STENCIL_LEG_EXT(Yp,spProjYm,accumReconYm);
+  GENERIC_STENCIL_LEG_EXT(Zp,spProjZm,accumReconZm);
+  GENERIC_STENCIL_LEG_EXT(Tp,spProjTm,accumReconTm);
+  if ( nmu ) { 
+    auto out_t = coalescedRead(out[sF],lane);
+    out_t = out_t + result;
+    coalescedWrite(out[sF],out_t,lane);
+  }
+};
+
+template <class Impl>
+void WilsonKernels<Impl>::DhopDirK( StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf, int sF,
+				    int sU, const FermionFieldView &in, FermionFieldView &out, int dir, int gamma) 
+{
+  typedef decltype(coalescedRead(buf[0])) calcHalfSpinor;
+  typedef decltype(coalescedRead(in[0]))  calcSpinor;
+  calcHalfSpinor chi;
+  calcSpinor result;
+  calcHalfSpinor Uchi;
+  StencilEntry *SE;
+  int ptype;
+  const int Nsimd = SiteHalfSpinor::Nsimd();
+  const int lane=SIMTlane(Nsimd);
+
+  SE = st.GetEntry(ptype, dir, sF);
+  if (gamma == Xp) {						
+    if (SE->_is_local ) {					
+      int perm= SE->_permute;					
+      auto tmp = coalescedReadPermute(in[SE->_offset],ptype,perm,lane);	
+      spProjXp(chi,tmp);						
+    } else {							
+      chi = coalescedRead(buf[SE->_offset],lane);			
+    }								
+    Impl::multLink(Uchi, U[sU], chi, dir, SE, st);		
+    spReconXp(result, Uchi);					
+  }
+
+  GENERIC_DHOPDIR_LEG(Yp,spProjYp,spReconYp);
+  GENERIC_DHOPDIR_LEG(Zp,spProjZp,spReconZp);
+  GENERIC_DHOPDIR_LEG(Tp,spProjTp,spReconTp);
+  GENERIC_DHOPDIR_LEG(Xm,spProjXm,spReconXm);
+  GENERIC_DHOPDIR_LEG(Ym,spProjYm,spReconYm);
+  GENERIC_DHOPDIR_LEG(Zm,spProjZm,spReconZm);
+  GENERIC_DHOPDIR_LEG(Tm,spProjTm,spReconTm);
+  coalescedWrite(out[sF], result,lane);
+}
+
+template <class Impl>
+void WilsonKernels<Impl>::DhopDirKernel( StencilImpl &st, DoubledGaugeField &U,SiteHalfSpinor *buf, int Ls,
+					 int Nsite, const FermionField &in, FermionField &out, int dirdisp, int gamma) 
+{
+  assert(dirdisp<=7);
+  assert(dirdisp>=0);
+
+   auto U_v   = U.View();
+   auto in_v  = in.View();
+   auto out_v = out.View();
+   auto st_v  = st.View();
+   accelerator_for(ss,Nsite,Simd::Nsimd(),{
+    for(int s=0;s<Ls;s++){
+      int sU=ss;
+      int sF = s+Ls*sU; 
+      DhopDirK(st_v,U_v,st.CommBuf(),sF,sU,in_v,out_v,dirdisp,gamma);
+    }
+  });
+} 
+
+#define KERNEL_CALLNB(A) \
+  const uint64_t    NN = Nsite*Ls;					\
+  accelerator_forNB( ss, NN, Simd::Nsimd(), {				\
+      int sF = ss;							\
+      int sU = ss/Ls;							\
+      WilsonKernels<Impl>::A(st_v,U_v,buf,sF,sU,in_v,out_v);		\
+  });
+
+#define KERNEL_CALL(A) KERNEL_CALLNB(A); accelerator_barrier(); 
+
+#define ASM_CALL(A)							\
+  thread_for( ss, Nsite, {						\
+    int sU = ss;							\
+    int sF = ss*Ls;							\
+    WilsonKernels<Impl>::A(st_v,U_v,buf,sF,sU,Ls,1,in_v,out_v);		\
+  });
+
+template <class Impl>
+void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField &U, SiteHalfSpinor * buf,
+				     int Ls, int Nsite, const FermionField &in, FermionField &out,
+				     int interior,int exterior) 
+{
+    auto U_v   =   U.View();
+    auto in_v  =  in.View();
+    auto out_v = out.View();
+    auto st_v  =  st.View();
+
+   if( interior && exterior ) { 
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSite); return;}
+#ifndef GRID_NVCC
+     if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSite);    return;}
+     if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSite); printf(".");    return;}
+#endif
+   } else if( interior ) {
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALLNB(GenericDhopSiteInt); return;}
+#ifndef GRID_NVCC
+     if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALLNB(HandDhopSiteInt);    return;}
+     if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteInt); printf("-");    return;}
+#endif
+   } else if( exterior ) { 
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteExt); return;}
+#ifndef GRID_NVCC
+     if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteExt);    return;}
+     if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteExt); printf("+");    return;}
+#endif
+   }
+   assert(0 && " Kernel optimisation case not covered ");
+  }
+  template <class Impl>
+  void WilsonKernels<Impl>::DhopDagKernel(int Opt,StencilImpl &st,  DoubledGaugeField &U, SiteHalfSpinor * buf,
+					  int Ls, int Nsite, const FermionField &in, FermionField &out,
+					  int interior,int exterior) 
+  {
+    auto U_v   = U.View();
+    auto in_v  = in.View();
+    auto out_v = out.View();
+    auto st_v  = st.View();
+
+   if( interior && exterior ) { 
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteDag); return;}
+#ifndef GRID_NVCC
+     if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteDag);    return;}
+     if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDag);     return;}
+#endif
+   } else if( interior ) {
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteDagInt); return;}
+#ifndef GRID_NVCC
+     if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteDagInt);    return;}
+     if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDagInt);     return;}
+#endif
+   } else if( exterior ) { 
+     if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteDagExt); return;}
+#ifndef GRID_NVCC
+     if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteDagExt);    return;}
+     if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDagExt);     return;}
+#endif
+   }
+   assert(0 && " Kernel optimisation case not covered ");
+  }
+
+/*******************************************************************************
+ * Conserved current utilities for Wilson fermions, for contracting propagators
+ * to make a conserved current sink or inserting the conserved current 
+ * sequentially. Common to both 4D and 5D.
+ ******************************************************************************/
+// N.B. Functions below assume a -1/2 factor within U.
+#define WilsonCurrentFwd(expr, mu) ((expr - Gamma::gmu[mu]*expr))
+#define WilsonCurrentBwd(expr, mu) ((expr + Gamma::gmu[mu]*expr))
+
+/*******************************************************************************
+ * Name: ContractConservedCurrentSiteFwd
+ * Operation: (1/2) * q2[x] * U(x) * (g[mu] - 1) * q1[x + mu]
+ * Notes: - DoubledGaugeField U assumed to contain -1/2 factor.
+ *        - Pass in q_in_1 shifted in +ve mu direction.
+ ******************************************************************************/
+template<class Impl>
+void WilsonKernels<Impl>::ContractConservedCurrentSiteFwd(const SitePropagator &q_in_1,
+							  const SitePropagator &q_in_2,
+							  SitePropagator &q_out,
+							  DoubledGaugeFieldView &U,
+							  unsigned int sU,
+							  unsigned int mu,
+							  bool switch_sign)
+{
+  SitePropagator result, tmp;
+  Gamma g5(Gamma::Algebra::Gamma5);
+
+  Impl::multLink(tmp, U[sU], q_in_1, mu);
+
+  result = g5 * adj(q_in_2) * g5 * WilsonCurrentFwd(tmp, mu);
+
+  if (switch_sign) {
+    q_out -= result;
+  } else {
+    q_out += result;
+  }
+}
+
+/*******************************************************************************
+ * Name: ContractConservedCurrentSiteBwd
+ * Operation: (1/2) * q2[x + mu] * U^dag(x) * (g[mu] + 1) * q1[x]
+ * Notes: - DoubledGaugeField U assumed to contain -1/2 factor.
+ *        - Pass in q_in_2 shifted in +ve mu direction.
+ ******************************************************************************/
+template<class Impl>
+void WilsonKernels<Impl>::ContractConservedCurrentSiteBwd(const SitePropagator &q_in_1,
+							  const SitePropagator &q_in_2,
+							  SitePropagator &q_out,
+							  DoubledGaugeFieldView &U,
+							  unsigned int sU,
+							  unsigned int mu,
+							  bool switch_sign)
+{
+  SitePropagator result, tmp;
+  Gamma g5(Gamma::Algebra::Gamma5);
+
+  Impl::multLink(tmp, U[sU], q_in_1, mu + Nd);
+
+  result = g5 * adj(q_in_2) * g5 * WilsonCurrentBwd(tmp, mu);
+  if (switch_sign) {
+    q_out += result;
+  } else {
+    q_out -= result;
+  }
+}
+
+/*******************************************************************************
+ * Name: SeqConservedCurrentSiteFwd
+ * Operation: (1/2) * U(x) * (g[mu] - 1) * q[x + mu]
+ * Notes: - DoubledGaugeField U assumed to contain -1/2 factor.
+ *        - Pass in q_in shifted in +ve mu direction.
+ ******************************************************************************/
+template<class Impl>
+void WilsonKernels<Impl>::SeqConservedCurrentSiteFwd(const SitePropagator &q_in,
+                                                     SitePropagator &q_out,
+                                                     DoubledGaugeFieldView &U,
+                                                     unsigned int sU,
+                                                     unsigned int mu,
+                                                     vPredicate t_mask,
+                                                     bool switch_sign)
+{
+  SitePropagator result;
+  
+  Impl::multLink(result, U[sU], q_in, mu);
+  result = WilsonCurrentFwd(result, mu);
+
+  // Zero any unwanted timeslice entries.
+  result = predicatedWhere(t_mask, result, 0.*result);
+  
+  if (switch_sign) {
+    q_out -= result;
+  } else {
+    q_out += result;
+  }
+}
+
+/*******************************************************************************
+ * Name: SeqConservedCurrentSiteFwd
+ * Operation: (1/2) * U^dag(x) * (g[mu] + 1) * q[x - mu]
+ * Notes: - DoubledGaugeField U assumed to contain -1/2 factor.
+ *        - Pass in q_in shifted in -ve mu direction.
+ ******************************************************************************/
+template<class Impl>
+void WilsonKernels<Impl>::SeqConservedCurrentSiteBwd(const SitePropagator &q_in, 
+                                                     SitePropagator &q_out,
+                                                     DoubledGaugeFieldView &U,
+                                                     unsigned int sU,
+                                                     unsigned int mu,
+                                                     vPredicate t_mask,
+                                                     bool switch_sign)
+{
+  SitePropagator result;
+  Impl::multLink(result, U[sU], q_in, mu + Nd);
+  result = WilsonCurrentBwd(result, mu);
+
+  // Zero any unwanted timeslice entries.
+  result = predicatedWhere(t_mask, result, 0.*result);
+  
+  if (switch_sign) {
+    q_out += result;
+  } else {
+    q_out -= result;
+  }
+}
+
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h
new file mode 100644
index 00000000..9a1a152c
--- /dev/null
+++ b/Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h
@@ -0,0 +1,97 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
+
+/*
+ * BF sequence
+ *
+ void bfmbase<Float>::MooeeInv(Fermion_t psi, 
+ Fermion_t chi, 
+ int dag, int cb)
+
+ double m    = this->mass;
+ double tm   = this->twistedmass;
+ double mtil = 4.0+this->mass;
+
+ double sq = mtil*mtil + tm*tm;
+
+ double a = mtil/sq;
+ double b = -tm /sq;
+ if(dag) b=-b;
+ axpibg5x(chi,psi,a,b);
+
+ void bfmbase<Float>::Mooee(Fermion_t psi, 
+ Fermion_t chi, 
+ int dag,int cb)
+ double a = 4.0+this->mass;
+ double b = this->twistedmass;
+ if(dag) b=-b;
+ axpibg5x(chi,psi,a,b);
+*/
+
+template<class Impl>
+void WilsonTMFermion<Impl>::Mooee(const FermionField &in, FermionField &out) {
+  RealD a = 4.0+this->mass;
+  RealD b = this->mu;
+  out.Checkerboard() = in.Checkerboard();
+  axpibg5x(out,in,a,b);
+}
+template<class Impl>
+void WilsonTMFermion<Impl>::MooeeDag(const FermionField &in, FermionField &out) {
+  RealD a = 4.0+this->mass;
+  RealD b = -this->mu;
+  out.Checkerboard() = in.Checkerboard();
+  axpibg5x(out,in,a,b);
+}
+template<class Impl>
+void WilsonTMFermion<Impl>::MooeeInv(const FermionField &in, FermionField &out) {
+  RealD m    = this->mass;
+  RealD tm   = this->mu;
+  RealD mtil = 4.0+m;
+  RealD sq   = mtil*mtil+tm*tm;
+  RealD a    = mtil/sq;
+  RealD b    = -tm /sq;
+  axpibg5x(out,in,a,b);
+}
+template<class Impl>
+void WilsonTMFermion<Impl>::MooeeInvDag(const FermionField &in, FermionField &out) {
+  RealD m    = this->mass;
+  RealD tm   = this->mu;
+  RealD mtil = 4.0+m;
+  RealD sq   = mtil*mtil+tm*tm;
+  RealD a    = mtil/sq;
+  RealD b    = tm /sq;
+  axpibg5x(out,in,a,b);
+}
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/CayleyFermion5DInstantiation.cc.master b/Grid/qcd/action/fermion/instantiation/CayleyFermion5DInstantiation.cc.master
new file mode 100644
index 00000000..5130db9c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/CayleyFermion5DInstantiation.cc.master
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h>
+
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h>
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dgpu.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class CayleyFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ContinuedFractionFermion5DInstantiation.cc.master b/Grid/qcd/action/fermion/instantiation/ContinuedFractionFermion5DInstantiation.cc.master
new file mode 100644
index 00000000..ca0d6cea
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ContinuedFractionFermion5DInstantiation.cc.master
@@ -0,0 +1,38 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ContinuedFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/DomainWallEOFAFermionInstantiation.cc.master b/Grid/qcd/action/fermion/instantiation/DomainWallEOFAFermionInstantiation.cc.master
new file mode 100644
index 00000000..f7198131
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/DomainWallEOFAFermionInstantiation.cc.master
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class DomainWallEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/CayleyFermion5DInstantiationGparityWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/CayleyFermion5DInstantiationGparityWilsonImplD.cc
new file mode 100644
index 00000000..5130db9c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/CayleyFermion5DInstantiationGparityWilsonImplD.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h>
+
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h>
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dgpu.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class CayleyFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/ContinuedFractionFermion5DInstantiationGparityWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/ContinuedFractionFermion5DInstantiationGparityWilsonImplD.cc
new file mode 100644
index 00000000..ca0d6cea
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/ContinuedFractionFermion5DInstantiationGparityWilsonImplD.cc
@@ -0,0 +1,38 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ContinuedFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/DomainWallEOFAFermionInstantiationGparityWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/DomainWallEOFAFermionInstantiationGparityWilsonImplD.cc
new file mode 100644
index 00000000..f7198131
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/DomainWallEOFAFermionInstantiationGparityWilsonImplD.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class DomainWallEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/MobiusEOFAFermionInstantiationGparityWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/MobiusEOFAFermionInstantiationGparityWilsonImplD.cc
new file mode 100644
index 00000000..ce7eaac9
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/MobiusEOFAFermionInstantiationGparityWilsonImplD.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class MobiusEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/PartialFractionFermion5DInstantiationGparityWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/PartialFractionFermion5DInstantiationGparityWilsonImplD.cc
new file mode 100644
index 00000000..757719b6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/PartialFractionFermion5DInstantiationGparityWilsonImplD.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+ 
+#include "impl.h"
+template class PartialFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonCloverFermionInstantiationGparityWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonCloverFermionInstantiationGparityWilsonImplD.cc
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonCloverFermionInstantiationGparityWilsonImplD.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonFermion5DInstantiationGparityWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonFermion5DInstantiationGparityWilsonImplD.cc
new file mode 100644
index 00000000..0dac989c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonFermion5DInstantiationGparityWilsonImplD.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonFermionInstantiationGparityWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonFermionInstantiationGparityWilsonImplD.cc
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonFermionInstantiationGparityWilsonImplD.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonKernelsInstantiationGparityWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonKernelsInstantiationGparityWilsonImplD.cc
new file mode 100644
index 00000000..75f143cb
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonKernelsInstantiationGparityWilsonImplD.cc
@@ -0,0 +1,74 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandGparityImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+// Move these
+
+#include "impl.h"
+
+// G-parity requires more specialised implementation.
+template <>
+void WilsonKernels<IMPLEMENTATION>::ContractConservedCurrentSiteFwd(const SitePropagator &q_in_1, 
+							  const SitePropagator &q_in_2, 
+							  SitePropagator &q_out, 
+							  DoubledGaugeFieldView &U,
+							  unsigned int sU, 
+							  unsigned int mu, 
+							  bool switch_sign)
+{			
+  assert(0);		
+}			
+template <>		
+void WilsonKernels<IMPLEMENTATION>::ContractConservedCurrentSiteBwd( const SitePropagator &q_in_1,
+							   const SitePropagator &q_in_2,
+							   SitePropagator &q_out,
+							   DoubledGaugeFieldView &U,
+							   unsigned int mu,
+							   unsigned int sU,
+							   bool switch_sign)
+{			
+  assert(0);		
+}
+
+HAND_SPECIALISE_GPARITY(IMPLEMENTATION);
+
+
+template class WilsonKernels<IMPLEMENTATION>; 
+
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonTMFermionInstantiationGparityWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonTMFermionInstantiationGparityWilsonImplD.cc
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/WilsonTMFermionInstantiationGparityWilsonImplD.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/impl.h b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/impl.h
new file mode 100644
index 00000000..930380b1
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplD/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION GparityWilsonImplD
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/CayleyFermion5DInstantiationGparityWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/CayleyFermion5DInstantiationGparityWilsonImplDF.cc
new file mode 100644
index 00000000..5130db9c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/CayleyFermion5DInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h>
+
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h>
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dgpu.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class CayleyFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/ContinuedFractionFermion5DInstantiationGparityWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/ContinuedFractionFermion5DInstantiationGparityWilsonImplDF.cc
new file mode 100644
index 00000000..ca0d6cea
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/ContinuedFractionFermion5DInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1,38 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ContinuedFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/DomainWallEOFAFermionInstantiationGparityWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/DomainWallEOFAFermionInstantiationGparityWilsonImplDF.cc
new file mode 100644
index 00000000..f7198131
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/DomainWallEOFAFermionInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class DomainWallEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/MobiusEOFAFermionInstantiationGparityWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/MobiusEOFAFermionInstantiationGparityWilsonImplDF.cc
new file mode 100644
index 00000000..ce7eaac9
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/MobiusEOFAFermionInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class MobiusEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/PartialFractionFermion5DInstantiationGparityWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/PartialFractionFermion5DInstantiationGparityWilsonImplDF.cc
new file mode 100644
index 00000000..757719b6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/PartialFractionFermion5DInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+ 
+#include "impl.h"
+template class PartialFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonCloverFermionInstantiationGparityWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonCloverFermionInstantiationGparityWilsonImplDF.cc
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonCloverFermionInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonFermion5DInstantiationGparityWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonFermion5DInstantiationGparityWilsonImplDF.cc
new file mode 100644
index 00000000..0dac989c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonFermion5DInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonFermionInstantiationGparityWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonFermionInstantiationGparityWilsonImplDF.cc
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonFermionInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonKernelsInstantiationGparityWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonKernelsInstantiationGparityWilsonImplDF.cc
new file mode 100644
index 00000000..75f143cb
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonKernelsInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1,74 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandGparityImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+// Move these
+
+#include "impl.h"
+
+// G-parity requires more specialised implementation.
+template <>
+void WilsonKernels<IMPLEMENTATION>::ContractConservedCurrentSiteFwd(const SitePropagator &q_in_1, 
+							  const SitePropagator &q_in_2, 
+							  SitePropagator &q_out, 
+							  DoubledGaugeFieldView &U,
+							  unsigned int sU, 
+							  unsigned int mu, 
+							  bool switch_sign)
+{			
+  assert(0);		
+}			
+template <>		
+void WilsonKernels<IMPLEMENTATION>::ContractConservedCurrentSiteBwd( const SitePropagator &q_in_1,
+							   const SitePropagator &q_in_2,
+							   SitePropagator &q_out,
+							   DoubledGaugeFieldView &U,
+							   unsigned int mu,
+							   unsigned int sU,
+							   bool switch_sign)
+{			
+  assert(0);		
+}
+
+HAND_SPECIALISE_GPARITY(IMPLEMENTATION);
+
+
+template class WilsonKernels<IMPLEMENTATION>; 
+
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonTMFermionInstantiationGparityWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonTMFermionInstantiationGparityWilsonImplDF.cc
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonTMFermionInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/impl.h b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/impl.h
new file mode 100644
index 00000000..2f13ce8a
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION GparityWilsonImplDF
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/CayleyFermion5DInstantiationGparityWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/CayleyFermion5DInstantiationGparityWilsonImplF.cc
new file mode 100644
index 00000000..5130db9c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/CayleyFermion5DInstantiationGparityWilsonImplF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h>
+
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h>
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dgpu.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class CayleyFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/ContinuedFractionFermion5DInstantiationGparityWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/ContinuedFractionFermion5DInstantiationGparityWilsonImplF.cc
new file mode 100644
index 00000000..ca0d6cea
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/ContinuedFractionFermion5DInstantiationGparityWilsonImplF.cc
@@ -0,0 +1,38 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ContinuedFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/DomainWallEOFAFermionInstantiationGparityWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/DomainWallEOFAFermionInstantiationGparityWilsonImplF.cc
new file mode 100644
index 00000000..f7198131
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/DomainWallEOFAFermionInstantiationGparityWilsonImplF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class DomainWallEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/MobiusEOFAFermionInstantiationGparityWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/MobiusEOFAFermionInstantiationGparityWilsonImplF.cc
new file mode 100644
index 00000000..ce7eaac9
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/MobiusEOFAFermionInstantiationGparityWilsonImplF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class MobiusEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/PartialFractionFermion5DInstantiationGparityWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/PartialFractionFermion5DInstantiationGparityWilsonImplF.cc
new file mode 100644
index 00000000..757719b6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/PartialFractionFermion5DInstantiationGparityWilsonImplF.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+ 
+#include "impl.h"
+template class PartialFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonCloverFermionInstantiationGparityWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonCloverFermionInstantiationGparityWilsonImplF.cc
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonCloverFermionInstantiationGparityWilsonImplF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonFermion5DInstantiationGparityWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonFermion5DInstantiationGparityWilsonImplF.cc
new file mode 100644
index 00000000..0dac989c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonFermion5DInstantiationGparityWilsonImplF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonFermionInstantiationGparityWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonFermionInstantiationGparityWilsonImplF.cc
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonFermionInstantiationGparityWilsonImplF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonKernelsInstantiationGparityWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonKernelsInstantiationGparityWilsonImplF.cc
new file mode 100644
index 00000000..75f143cb
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonKernelsInstantiationGparityWilsonImplF.cc
@@ -0,0 +1,74 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandGparityImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+// Move these
+
+#include "impl.h"
+
+// G-parity requires more specialised implementation.
+template <>
+void WilsonKernels<IMPLEMENTATION>::ContractConservedCurrentSiteFwd(const SitePropagator &q_in_1, 
+							  const SitePropagator &q_in_2, 
+							  SitePropagator &q_out, 
+							  DoubledGaugeFieldView &U,
+							  unsigned int sU, 
+							  unsigned int mu, 
+							  bool switch_sign)
+{			
+  assert(0);		
+}			
+template <>		
+void WilsonKernels<IMPLEMENTATION>::ContractConservedCurrentSiteBwd( const SitePropagator &q_in_1,
+							   const SitePropagator &q_in_2,
+							   SitePropagator &q_out,
+							   DoubledGaugeFieldView &U,
+							   unsigned int mu,
+							   unsigned int sU,
+							   bool switch_sign)
+{			
+  assert(0);		
+}
+
+HAND_SPECIALISE_GPARITY(IMPLEMENTATION);
+
+
+template class WilsonKernels<IMPLEMENTATION>; 
+
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonTMFermionInstantiationGparityWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonTMFermionInstantiationGparityWilsonImplF.cc
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/WilsonTMFermionInstantiationGparityWilsonImplF.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/impl.h b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/impl.h
new file mode 100644
index 00000000..43f27bb8
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplF/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION GparityWilsonImplF
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/CayleyFermion5DInstantiationGparityWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/CayleyFermion5DInstantiationGparityWilsonImplFH.cc
new file mode 100644
index 00000000..5130db9c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/CayleyFermion5DInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h>
+
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h>
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dgpu.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class CayleyFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/ContinuedFractionFermion5DInstantiationGparityWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/ContinuedFractionFermion5DInstantiationGparityWilsonImplFH.cc
new file mode 100644
index 00000000..ca0d6cea
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/ContinuedFractionFermion5DInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1,38 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ContinuedFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/DomainWallEOFAFermionInstantiationGparityWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/DomainWallEOFAFermionInstantiationGparityWilsonImplFH.cc
new file mode 100644
index 00000000..f7198131
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/DomainWallEOFAFermionInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class DomainWallEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/MobiusEOFAFermionInstantiationGparityWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/MobiusEOFAFermionInstantiationGparityWilsonImplFH.cc
new file mode 100644
index 00000000..ce7eaac9
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/MobiusEOFAFermionInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class MobiusEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/PartialFractionFermion5DInstantiationGparityWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/PartialFractionFermion5DInstantiationGparityWilsonImplFH.cc
new file mode 100644
index 00000000..757719b6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/PartialFractionFermion5DInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+ 
+#include "impl.h"
+template class PartialFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonCloverFermionInstantiationGparityWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonCloverFermionInstantiationGparityWilsonImplFH.cc
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonCloverFermionInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonFermion5DInstantiationGparityWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonFermion5DInstantiationGparityWilsonImplFH.cc
new file mode 100644
index 00000000..0dac989c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonFermion5DInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonFermionInstantiationGparityWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonFermionInstantiationGparityWilsonImplFH.cc
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonFermionInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonKernelsInstantiationGparityWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonKernelsInstantiationGparityWilsonImplFH.cc
new file mode 100644
index 00000000..75f143cb
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonKernelsInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1,74 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandGparityImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+// Move these
+
+#include "impl.h"
+
+// G-parity requires more specialised implementation.
+template <>
+void WilsonKernels<IMPLEMENTATION>::ContractConservedCurrentSiteFwd(const SitePropagator &q_in_1, 
+							  const SitePropagator &q_in_2, 
+							  SitePropagator &q_out, 
+							  DoubledGaugeFieldView &U,
+							  unsigned int sU, 
+							  unsigned int mu, 
+							  bool switch_sign)
+{			
+  assert(0);		
+}			
+template <>		
+void WilsonKernels<IMPLEMENTATION>::ContractConservedCurrentSiteBwd( const SitePropagator &q_in_1,
+							   const SitePropagator &q_in_2,
+							   SitePropagator &q_out,
+							   DoubledGaugeFieldView &U,
+							   unsigned int mu,
+							   unsigned int sU,
+							   bool switch_sign)
+{			
+  assert(0);		
+}
+
+HAND_SPECIALISE_GPARITY(IMPLEMENTATION);
+
+
+template class WilsonKernels<IMPLEMENTATION>; 
+
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonTMFermionInstantiationGparityWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonTMFermionInstantiationGparityWilsonImplFH.cc
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonTMFermionInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/impl.h b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/impl.h
new file mode 100644
index 00000000..ebcb6e62
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION GparityWilsonImplFH
diff --git a/Grid/qcd/action/fermion/instantiation/ImprovedStaggeredFermion5DInstantiation.cc b/Grid/qcd/action/fermion/instantiation/ImprovedStaggeredFermion5DInstantiation.cc
new file mode 100644
index 00000000..986460b0
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ImprovedStaggeredFermion5DInstantiation.cc
@@ -0,0 +1,41 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ImprovedStaggeredFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ImprovedStaggeredFermion5D.h>
+
+NAMESPACE_BEGIN(Grid);
+  
+// S-direction is INNERMOST and takes no part in the parity.
+const std::vector<int> ImprovedStaggeredFermion5DStatic::directions({1,2,3,4,1,2,3,4,1,2,3,4,1,2,3,4});
+const std::vector<int> ImprovedStaggeredFermion5DStatic::displacements({1, 1, 1, 1, -1, -1, -1, -1, 3, 3, 3, 3, -3, -3, -3, -3});
+  
+NAMESPACE_END(Grid);
+
+
+
diff --git a/Grid/qcd/action/fermion/instantiation/ImprovedStaggeredFermion5DInstantiation.cc.master b/Grid/qcd/action/fermion/instantiation/ImprovedStaggeredFermion5DInstantiation.cc.master
new file mode 100644
index 00000000..572b375c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ImprovedStaggeredFermion5DInstantiation.cc.master
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ImprovedStaggeredFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+
+#include <Grid/qcd/action/fermion/ImprovedStaggeredFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ImprovedStaggeredFermion5DImplementation.h>
+#include <Grid/perfmon/PerfCount.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ImprovedStaggeredFermion5D<IMPLEMENTATION>; 
+  
+NAMESPACE_END(Grid);
+
+
+
+
diff --git a/Grid/qcd/action/fermion/instantiation/ImprovedStaggeredFermionInstantiation.cc b/Grid/qcd/action/fermion/instantiation/ImprovedStaggeredFermionInstantiation.cc
new file mode 100644
index 00000000..a617f6cb
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ImprovedStaggeredFermionInstantiation.cc
@@ -0,0 +1,36 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/ImprovedStaggeredFermion.cc
+
+Copyright (C) 2015
+
+Author: Azusa Yamaguchi, Peter Boyle
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+			   /*  END LEGAL */
+#include <Grid.h>
+
+NAMESPACE_BEGIN(Grid);
+
+const std::vector<int> ImprovedStaggeredFermionStatic::directions({0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3, 0, 1, 2, 3});
+const std::vector<int> ImprovedStaggeredFermionStatic::displacements({1, 1, 1, 1, -1, -1, -1, -1, 3, 3, 3, 3, -3, -3, -3, -3});
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/ImprovedStaggeredFermionInstantiation.cc.master b/Grid/qcd/action/fermion/instantiation/ImprovedStaggeredFermionInstantiation.cc.master
new file mode 100644
index 00000000..2023adc2
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ImprovedStaggeredFermionInstantiation.cc.master
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/ImprovedStaggeredFermion.cc
+
+Copyright (C) 2015
+
+Author: Azusa Yamaguchi, Peter Boyle
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+			   /*  END LEGAL */
+#include <Grid.h>
+#include <Grid/qcd/action/fermion/implementation/ImprovedStaggeredFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ImprovedStaggeredFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/MobiusEOFAFermionInstantiation.cc.master b/Grid/qcd/action/fermion/instantiation/MobiusEOFAFermionInstantiation.cc.master
new file mode 100644
index 00000000..ce7eaac9
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/MobiusEOFAFermionInstantiation.cc.master
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class MobiusEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/PartialFractionFermion5DInstantiation.cc.master b/Grid/qcd/action/fermion/instantiation/PartialFractionFermion5DInstantiation.cc.master
new file mode 100644
index 00000000..757719b6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/PartialFractionFermion5DInstantiation.cc.master
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+ 
+#include "impl.h"
+template class PartialFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/StaggeredImplD/ImprovedStaggeredFermion5DInstantiationStaggeredImplD.cc b/Grid/qcd/action/fermion/instantiation/StaggeredImplD/ImprovedStaggeredFermion5DInstantiationStaggeredImplD.cc
new file mode 100644
index 00000000..572b375c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/StaggeredImplD/ImprovedStaggeredFermion5DInstantiationStaggeredImplD.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ImprovedStaggeredFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+
+#include <Grid/qcd/action/fermion/ImprovedStaggeredFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ImprovedStaggeredFermion5DImplementation.h>
+#include <Grid/perfmon/PerfCount.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ImprovedStaggeredFermion5D<IMPLEMENTATION>; 
+  
+NAMESPACE_END(Grid);
+
+
+
+
diff --git a/Grid/qcd/action/fermion/instantiation/StaggeredImplD/ImprovedStaggeredFermionInstantiationStaggeredImplD.cc b/Grid/qcd/action/fermion/instantiation/StaggeredImplD/ImprovedStaggeredFermionInstantiationStaggeredImplD.cc
new file mode 100644
index 00000000..2023adc2
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/StaggeredImplD/ImprovedStaggeredFermionInstantiationStaggeredImplD.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/ImprovedStaggeredFermion.cc
+
+Copyright (C) 2015
+
+Author: Azusa Yamaguchi, Peter Boyle
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+			   /*  END LEGAL */
+#include <Grid.h>
+#include <Grid/qcd/action/fermion/implementation/ImprovedStaggeredFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ImprovedStaggeredFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/StaggeredImplD/StaggeredKernelsInstantiationStaggeredImplD.cc b/Grid/qcd/action/fermion/instantiation/StaggeredImplD/StaggeredKernelsInstantiationStaggeredImplD.cc
new file mode 100644
index 00000000..c3acf963
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/StaggeredImplD/StaggeredKernelsInstantiationStaggeredImplD.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Azusa Yamaguchi, Peter Boyle
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/StaggeredKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/StaggeredKernelsHand.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class StaggeredKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/StaggeredImplD/impl.h b/Grid/qcd/action/fermion/instantiation/StaggeredImplD/impl.h
new file mode 100644
index 00000000..15644489
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/StaggeredImplD/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION StaggeredImplD
diff --git a/Grid/qcd/action/fermion/instantiation/StaggeredImplF/ImprovedStaggeredFermion5DInstantiationStaggeredImplF.cc b/Grid/qcd/action/fermion/instantiation/StaggeredImplF/ImprovedStaggeredFermion5DInstantiationStaggeredImplF.cc
new file mode 100644
index 00000000..572b375c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/StaggeredImplF/ImprovedStaggeredFermion5DInstantiationStaggeredImplF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ImprovedStaggeredFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+
+#include <Grid/qcd/action/fermion/ImprovedStaggeredFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ImprovedStaggeredFermion5DImplementation.h>
+#include <Grid/perfmon/PerfCount.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ImprovedStaggeredFermion5D<IMPLEMENTATION>; 
+  
+NAMESPACE_END(Grid);
+
+
+
+
diff --git a/Grid/qcd/action/fermion/instantiation/StaggeredImplF/ImprovedStaggeredFermionInstantiationStaggeredImplF.cc b/Grid/qcd/action/fermion/instantiation/StaggeredImplF/ImprovedStaggeredFermionInstantiationStaggeredImplF.cc
new file mode 100644
index 00000000..2023adc2
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/StaggeredImplF/ImprovedStaggeredFermionInstantiationStaggeredImplF.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/ImprovedStaggeredFermion.cc
+
+Copyright (C) 2015
+
+Author: Azusa Yamaguchi, Peter Boyle
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+			   /*  END LEGAL */
+#include <Grid.h>
+#include <Grid/qcd/action/fermion/implementation/ImprovedStaggeredFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ImprovedStaggeredFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/StaggeredImplF/StaggeredKernelsInstantiationStaggeredImplF.cc b/Grid/qcd/action/fermion/instantiation/StaggeredImplF/StaggeredKernelsInstantiationStaggeredImplF.cc
new file mode 100644
index 00000000..c3acf963
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/StaggeredImplF/StaggeredKernelsInstantiationStaggeredImplF.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Azusa Yamaguchi, Peter Boyle
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/StaggeredKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/StaggeredKernelsHand.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class StaggeredKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/StaggeredImplF/impl.h b/Grid/qcd/action/fermion/instantiation/StaggeredImplF/impl.h
new file mode 100644
index 00000000..6569f6ad
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/StaggeredImplF/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION StaggeredImplF
diff --git a/Grid/qcd/action/fermion/instantiation/StaggeredKernelsInstantiation.cc b/Grid/qcd/action/fermion/instantiation/StaggeredKernelsInstantiation.cc
new file mode 100644
index 00000000..5e5175e8
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/StaggeredKernelsInstantiation.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Azusa Yamaguchi, Peter Boyle
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/StaggeredKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/StaggeredKernelsAsm.h>
+
+NAMESPACE_BEGIN(Grid);
+
+int StaggeredKernelsStatic::Opt= StaggeredKernelsStatic::OptGeneric;
+int StaggeredKernelsStatic::Comms = StaggeredKernelsStatic::CommsAndCompute;
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/StaggeredKernelsInstantiation.cc.master b/Grid/qcd/action/fermion/instantiation/StaggeredKernelsInstantiation.cc.master
new file mode 100644
index 00000000..c3acf963
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/StaggeredKernelsInstantiation.cc.master
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Azusa Yamaguchi, Peter Boyle
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/StaggeredKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/StaggeredKernelsHand.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class StaggeredKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/WilsonCloverFermionInstantiationWilsonAdjImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/WilsonCloverFermionInstantiationWilsonAdjImplD.cc
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/WilsonCloverFermionInstantiationWilsonAdjImplD.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/WilsonFermionInstantiationWilsonAdjImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/WilsonFermionInstantiationWilsonAdjImplD.cc
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/WilsonFermionInstantiationWilsonAdjImplD.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/WilsonKernelsInstantiationWilsonAdjImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/WilsonKernelsInstantiationWilsonAdjImplD.cc
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/WilsonKernelsInstantiationWilsonAdjImplD.cc
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/WilsonTMFermionInstantiationWilsonAdjImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/WilsonTMFermionInstantiationWilsonAdjImplD.cc
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/WilsonTMFermionInstantiationWilsonAdjImplD.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/impl.h b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/impl.h
new file mode 100644
index 00000000..1c2fce25
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION WilsonAdjImplD
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/WilsonCloverFermionInstantiationWilsonAdjImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/WilsonCloverFermionInstantiationWilsonAdjImplF.cc
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/WilsonCloverFermionInstantiationWilsonAdjImplF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/WilsonFermionInstantiationWilsonAdjImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/WilsonFermionInstantiationWilsonAdjImplF.cc
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/WilsonFermionInstantiationWilsonAdjImplF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/WilsonKernelsInstantiationWilsonAdjImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/WilsonKernelsInstantiationWilsonAdjImplF.cc
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/WilsonKernelsInstantiationWilsonAdjImplF.cc
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/WilsonTMFermionInstantiationWilsonAdjImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/WilsonTMFermionInstantiationWilsonAdjImplF.cc
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/WilsonTMFermionInstantiationWilsonAdjImplF.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/impl.h b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/impl.h
new file mode 100644
index 00000000..93e4d7d3
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION WilsonAdjImplF
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonCloverFermionInstantiation.cc.master b/Grid/qcd/action/fermion/instantiation/WilsonCloverFermionInstantiation.cc.master
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonCloverFermionInstantiation.cc.master
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonFermion5DInstantiation.cc b/Grid/qcd/action/fermion/instantiation/WilsonFermion5DInstantiation.cc
new file mode 100644
index 00000000..0836849b
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonFermion5DInstantiation.cc
@@ -0,0 +1,41 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonFermion5D.h>
+
+NAMESPACE_BEGIN(Grid);
+
+// S-direction is INNERMOST and takes no part in the parity.
+const std::vector<int> WilsonFermion5DStatic::directions   ({1,2,3,4, 1, 2, 3, 4});
+const std::vector<int> WilsonFermion5DStatic::displacements({1,1,1,1,-1,-1,-1,-1});
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonFermion5DInstantiation.cc.master b/Grid/qcd/action/fermion/instantiation/WilsonFermion5DInstantiation.cc.master
new file mode 100644
index 00000000..0dac989c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonFermion5DInstantiation.cc.master
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonFermionInstantiation.cc b/Grid/qcd/action/fermion/instantiation/WilsonFermionInstantiation.cc
new file mode 100644
index 00000000..a35ef74c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonFermionInstantiation.cc
@@ -0,0 +1,41 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonFermion.h>
+
+NAMESPACE_BEGIN(Grid);
+
+const std::vector<int> WilsonFermionStatic::directions({0, 1, 2, 3, 0, 1, 2, 3});
+const std::vector<int> WilsonFermionStatic::displacements({1, 1, 1, 1, -1, -1, -1, -1});
+int WilsonFermionStatic::HandOptDslash;
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonFermionInstantiation.cc.master b/Grid/qcd/action/fermion/instantiation/WilsonFermionInstantiation.cc.master
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonFermionInstantiation.cc.master
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD/CayleyFermion5DInstantiationWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD/CayleyFermion5DInstantiationWilsonImplD.cc
new file mode 100644
index 00000000..5130db9c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD/CayleyFermion5DInstantiationWilsonImplD.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h>
+
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h>
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dgpu.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class CayleyFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD/ContinuedFractionFermion5DInstantiationWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD/ContinuedFractionFermion5DInstantiationWilsonImplD.cc
new file mode 100644
index 00000000..ca0d6cea
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD/ContinuedFractionFermion5DInstantiationWilsonImplD.cc
@@ -0,0 +1,38 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ContinuedFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD/DomainWallEOFAFermionInstantiationWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD/DomainWallEOFAFermionInstantiationWilsonImplD.cc
new file mode 100644
index 00000000..f7198131
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD/DomainWallEOFAFermionInstantiationWilsonImplD.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class DomainWallEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD/MobiusEOFAFermionInstantiationWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD/MobiusEOFAFermionInstantiationWilsonImplD.cc
new file mode 100644
index 00000000..ce7eaac9
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD/MobiusEOFAFermionInstantiationWilsonImplD.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class MobiusEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD/PartialFractionFermion5DInstantiationWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD/PartialFractionFermion5DInstantiationWilsonImplD.cc
new file mode 100644
index 00000000..757719b6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD/PartialFractionFermion5DInstantiationWilsonImplD.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+ 
+#include "impl.h"
+template class PartialFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonCloverFermionInstantiationWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonCloverFermionInstantiationWilsonImplD.cc
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonCloverFermionInstantiationWilsonImplD.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonFermion5DInstantiationWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonFermion5DInstantiationWilsonImplD.cc
new file mode 100644
index 00000000..0dac989c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonFermion5DInstantiationWilsonImplD.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonFermionInstantiationWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonFermionInstantiationWilsonImplD.cc
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonFermionInstantiationWilsonImplD.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonKernelsInstantiationWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonKernelsInstantiationWilsonImplD.cc
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonKernelsInstantiationWilsonImplD.cc
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonTMFermionInstantiationWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonTMFermionInstantiationWilsonImplD.cc
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonTMFermionInstantiationWilsonImplD.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplD/impl.h b/Grid/qcd/action/fermion/instantiation/WilsonImplD/impl.h
new file mode 100644
index 00000000..401d2774
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplD/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION WilsonImplD
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/CayleyFermion5DInstantiationWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/CayleyFermion5DInstantiationWilsonImplDF.cc
new file mode 100644
index 00000000..5130db9c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/CayleyFermion5DInstantiationWilsonImplDF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h>
+
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h>
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dgpu.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class CayleyFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/ContinuedFractionFermion5DInstantiationWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/ContinuedFractionFermion5DInstantiationWilsonImplDF.cc
new file mode 100644
index 00000000..ca0d6cea
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/ContinuedFractionFermion5DInstantiationWilsonImplDF.cc
@@ -0,0 +1,38 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ContinuedFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/DomainWallEOFAFermionInstantiationWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/DomainWallEOFAFermionInstantiationWilsonImplDF.cc
new file mode 100644
index 00000000..f7198131
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/DomainWallEOFAFermionInstantiationWilsonImplDF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class DomainWallEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/MobiusEOFAFermionInstantiationWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/MobiusEOFAFermionInstantiationWilsonImplDF.cc
new file mode 100644
index 00000000..ce7eaac9
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/MobiusEOFAFermionInstantiationWilsonImplDF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class MobiusEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/PartialFractionFermion5DInstantiationWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/PartialFractionFermion5DInstantiationWilsonImplDF.cc
new file mode 100644
index 00000000..757719b6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/PartialFractionFermion5DInstantiationWilsonImplDF.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+ 
+#include "impl.h"
+template class PartialFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonCloverFermionInstantiationWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonCloverFermionInstantiationWilsonImplDF.cc
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonCloverFermionInstantiationWilsonImplDF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonFermion5DInstantiationWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonFermion5DInstantiationWilsonImplDF.cc
new file mode 100644
index 00000000..0dac989c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonFermion5DInstantiationWilsonImplDF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonFermionInstantiationWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonFermionInstantiationWilsonImplDF.cc
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonFermionInstantiationWilsonImplDF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonKernelsInstantiationWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonKernelsInstantiationWilsonImplDF.cc
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonKernelsInstantiationWilsonImplDF.cc
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonTMFermionInstantiationWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonTMFermionInstantiationWilsonImplDF.cc
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonTMFermionInstantiationWilsonImplDF.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/impl.h b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/impl.h
new file mode 100644
index 00000000..2adc6136
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION WilsonImplDF
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplF/CayleyFermion5DInstantiationWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplF/CayleyFermion5DInstantiationWilsonImplF.cc
new file mode 100644
index 00000000..5130db9c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplF/CayleyFermion5DInstantiationWilsonImplF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h>
+
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h>
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dgpu.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class CayleyFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplF/ContinuedFractionFermion5DInstantiationWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplF/ContinuedFractionFermion5DInstantiationWilsonImplF.cc
new file mode 100644
index 00000000..ca0d6cea
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplF/ContinuedFractionFermion5DInstantiationWilsonImplF.cc
@@ -0,0 +1,38 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ContinuedFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplF/DomainWallEOFAFermionInstantiationWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplF/DomainWallEOFAFermionInstantiationWilsonImplF.cc
new file mode 100644
index 00000000..f7198131
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplF/DomainWallEOFAFermionInstantiationWilsonImplF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class DomainWallEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplF/MobiusEOFAFermionInstantiationWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplF/MobiusEOFAFermionInstantiationWilsonImplF.cc
new file mode 100644
index 00000000..ce7eaac9
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplF/MobiusEOFAFermionInstantiationWilsonImplF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class MobiusEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplF/PartialFractionFermion5DInstantiationWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplF/PartialFractionFermion5DInstantiationWilsonImplF.cc
new file mode 100644
index 00000000..757719b6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplF/PartialFractionFermion5DInstantiationWilsonImplF.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+ 
+#include "impl.h"
+template class PartialFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonCloverFermionInstantiationWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonCloverFermionInstantiationWilsonImplF.cc
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonCloverFermionInstantiationWilsonImplF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonFermion5DInstantiationWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonFermion5DInstantiationWilsonImplF.cc
new file mode 100644
index 00000000..0dac989c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonFermion5DInstantiationWilsonImplF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonFermionInstantiationWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonFermionInstantiationWilsonImplF.cc
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonFermionInstantiationWilsonImplF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonKernelsInstantiationWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonKernelsInstantiationWilsonImplF.cc
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonKernelsInstantiationWilsonImplF.cc
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonTMFermionInstantiationWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonTMFermionInstantiationWilsonImplF.cc
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonTMFermionInstantiationWilsonImplF.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplF/impl.h b/Grid/qcd/action/fermion/instantiation/WilsonImplF/impl.h
new file mode 100644
index 00000000..78966b84
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplF/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION WilsonImplF
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/CayleyFermion5DInstantiationWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/CayleyFermion5DInstantiationWilsonImplFH.cc
new file mode 100644
index 00000000..5130db9c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/CayleyFermion5DInstantiationWilsonImplFH.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h>
+
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h>
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dgpu.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class CayleyFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/ContinuedFractionFermion5DInstantiationWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/ContinuedFractionFermion5DInstantiationWilsonImplFH.cc
new file mode 100644
index 00000000..ca0d6cea
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/ContinuedFractionFermion5DInstantiationWilsonImplFH.cc
@@ -0,0 +1,38 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ContinuedFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/DomainWallEOFAFermionInstantiationWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/DomainWallEOFAFermionInstantiationWilsonImplFH.cc
new file mode 100644
index 00000000..f7198131
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/DomainWallEOFAFermionInstantiationWilsonImplFH.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class DomainWallEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/MobiusEOFAFermionInstantiationWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/MobiusEOFAFermionInstantiationWilsonImplFH.cc
new file mode 100644
index 00000000..ce7eaac9
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/MobiusEOFAFermionInstantiationWilsonImplFH.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class MobiusEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/PartialFractionFermion5DInstantiationWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/PartialFractionFermion5DInstantiationWilsonImplFH.cc
new file mode 100644
index 00000000..757719b6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/PartialFractionFermion5DInstantiationWilsonImplFH.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+ 
+#include "impl.h"
+template class PartialFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonCloverFermionInstantiationWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonCloverFermionInstantiationWilsonImplFH.cc
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonCloverFermionInstantiationWilsonImplFH.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonFermion5DInstantiationWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonFermion5DInstantiationWilsonImplFH.cc
new file mode 100644
index 00000000..0dac989c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonFermion5DInstantiationWilsonImplFH.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonFermionInstantiationWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonFermionInstantiationWilsonImplFH.cc
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonFermionInstantiationWilsonImplFH.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonKernelsInstantiationWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonKernelsInstantiationWilsonImplFH.cc
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonKernelsInstantiationWilsonImplFH.cc
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonTMFermionInstantiationWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonTMFermionInstantiationWilsonImplFH.cc
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonTMFermionInstantiationWilsonImplFH.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/impl.h b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/impl.h
new file mode 100644
index 00000000..e442863d
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION WilsonImplFH
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonKernelsInstantiation.cc b/Grid/qcd/action/fermion/instantiation/WilsonKernelsInstantiation.cc
new file mode 100644
index 00000000..c02da6c8
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonKernelsInstantiation.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+
+NAMESPACE_BEGIN(Grid);
+
+// Move these
+int WilsonKernelsStatic::Opt   = WilsonKernelsStatic::OptGeneric;
+int WilsonKernelsStatic::Comms = WilsonKernelsStatic::CommsAndCompute;
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonKernelsInstantiation.cc.master b/Grid/qcd/action/fermion/instantiation/WilsonKernelsInstantiation.cc.master
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonKernelsInstantiation.cc.master
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonKernelsInstantiationAsm.cc b/Grid/qcd/action/fermion/instantiation/WilsonKernelsInstantiationAsm.cc
new file mode 100644
index 00000000..f6f235c8
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonKernelsInstantiationAsm.cc
@@ -0,0 +1,43 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+
+////////////////////////////////////////////////////////////////////////
+// Include the specialisations for ASM kernels
+////////////////////////////////////////////////////////////////////////
+NAMESPACE_BEGIN(Grid);
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmAvx512.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmQPX.h>
+NAMESPACE_END(Grid);
+
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonKernelsInstantiationGparity.cc.master b/Grid/qcd/action/fermion/instantiation/WilsonKernelsInstantiationGparity.cc.master
new file mode 100644
index 00000000..75f143cb
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonKernelsInstantiationGparity.cc.master
@@ -0,0 +1,74 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandGparityImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+// Move these
+
+#include "impl.h"
+
+// G-parity requires more specialised implementation.
+template <>
+void WilsonKernels<IMPLEMENTATION>::ContractConservedCurrentSiteFwd(const SitePropagator &q_in_1, 
+							  const SitePropagator &q_in_2, 
+							  SitePropagator &q_out, 
+							  DoubledGaugeFieldView &U,
+							  unsigned int sU, 
+							  unsigned int mu, 
+							  bool switch_sign)
+{			
+  assert(0);		
+}			
+template <>		
+void WilsonKernels<IMPLEMENTATION>::ContractConservedCurrentSiteBwd( const SitePropagator &q_in_1,
+							   const SitePropagator &q_in_2,
+							   SitePropagator &q_out,
+							   DoubledGaugeFieldView &U,
+							   unsigned int mu,
+							   unsigned int sU,
+							   bool switch_sign)
+{			
+  assert(0);		
+}
+
+HAND_SPECIALISE_GPARITY(IMPLEMENTATION);
+
+
+template class WilsonKernels<IMPLEMENTATION>; 
+
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTMFermionInstantiation.cc.master b/Grid/qcd/action/fermion/instantiation/WilsonTMFermionInstantiation.cc.master
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTMFermionInstantiation.cc.master
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/WilsonCloverFermionInstantiationWilsonTwoIndexAntiSymmetricImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/WilsonCloverFermionInstantiationWilsonTwoIndexAntiSymmetricImplD.cc
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/WilsonCloverFermionInstantiationWilsonTwoIndexAntiSymmetricImplD.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/WilsonFermionInstantiationWilsonTwoIndexAntiSymmetricImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/WilsonFermionInstantiationWilsonTwoIndexAntiSymmetricImplD.cc
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/WilsonFermionInstantiationWilsonTwoIndexAntiSymmetricImplD.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/WilsonKernelsInstantiationWilsonTwoIndexAntiSymmetricImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/WilsonKernelsInstantiationWilsonTwoIndexAntiSymmetricImplD.cc
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/WilsonKernelsInstantiationWilsonTwoIndexAntiSymmetricImplD.cc
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/WilsonTMFermionInstantiationWilsonTwoIndexAntiSymmetricImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/WilsonTMFermionInstantiationWilsonTwoIndexAntiSymmetricImplD.cc
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/WilsonTMFermionInstantiationWilsonTwoIndexAntiSymmetricImplD.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/impl.h b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/impl.h
new file mode 100644
index 00000000..e2341833
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION WilsonTwoIndexAntiSymmetricImplD
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/WilsonCloverFermionInstantiationWilsonTwoIndexAntiSymmetricImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/WilsonCloverFermionInstantiationWilsonTwoIndexAntiSymmetricImplF.cc
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/WilsonCloverFermionInstantiationWilsonTwoIndexAntiSymmetricImplF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/WilsonFermionInstantiationWilsonTwoIndexAntiSymmetricImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/WilsonFermionInstantiationWilsonTwoIndexAntiSymmetricImplF.cc
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/WilsonFermionInstantiationWilsonTwoIndexAntiSymmetricImplF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/WilsonKernelsInstantiationWilsonTwoIndexAntiSymmetricImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/WilsonKernelsInstantiationWilsonTwoIndexAntiSymmetricImplF.cc
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/WilsonKernelsInstantiationWilsonTwoIndexAntiSymmetricImplF.cc
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/WilsonTMFermionInstantiationWilsonTwoIndexAntiSymmetricImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/WilsonTMFermionInstantiationWilsonTwoIndexAntiSymmetricImplF.cc
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/WilsonTMFermionInstantiationWilsonTwoIndexAntiSymmetricImplF.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/impl.h b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/impl.h
new file mode 100644
index 00000000..86410435
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION WilsonTwoIndexAntiSymmetricImplF
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/WilsonCloverFermionInstantiationWilsonTwoIndexSymmetricImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/WilsonCloverFermionInstantiationWilsonTwoIndexSymmetricImplD.cc
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/WilsonCloverFermionInstantiationWilsonTwoIndexSymmetricImplD.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/WilsonFermionInstantiationWilsonTwoIndexSymmetricImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/WilsonFermionInstantiationWilsonTwoIndexSymmetricImplD.cc
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/WilsonFermionInstantiationWilsonTwoIndexSymmetricImplD.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/WilsonKernelsInstantiationWilsonTwoIndexSymmetricImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/WilsonKernelsInstantiationWilsonTwoIndexSymmetricImplD.cc
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/WilsonKernelsInstantiationWilsonTwoIndexSymmetricImplD.cc
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/WilsonTMFermionInstantiationWilsonTwoIndexSymmetricImplD.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/WilsonTMFermionInstantiationWilsonTwoIndexSymmetricImplD.cc
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/WilsonTMFermionInstantiationWilsonTwoIndexSymmetricImplD.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/impl.h b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/impl.h
new file mode 100644
index 00000000..449c02fc
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION WilsonTwoIndexSymmetricImplD
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/WilsonCloverFermionInstantiationWilsonTwoIndexSymmetricImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/WilsonCloverFermionInstantiationWilsonTwoIndexSymmetricImplF.cc
new file mode 100644
index 00000000..af99dfb6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/WilsonCloverFermionInstantiationWilsonTwoIndexSymmetricImplF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
+
+    Copyright (C) 2017
+
+    Author: paboyle <paboyle@ph.ed.ac.uk>
+    Author: Guido Cossu <guido.cossu@ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+/*  END LEGAL */
+
+#include <Grid/Grid.h>
+#include <Grid/qcd/spin/Dirac.h>
+#include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonCloverFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/WilsonFermionInstantiationWilsonTwoIndexSymmetricImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/WilsonFermionInstantiationWilsonTwoIndexSymmetricImplF.cc
new file mode 100644
index 00000000..6fd9c5ca
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/WilsonFermionInstantiationWilsonTwoIndexSymmetricImplF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/WilsonKernelsInstantiationWilsonTwoIndexSymmetricImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/WilsonKernelsInstantiationWilsonTwoIndexSymmetricImplF.cc
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/WilsonKernelsInstantiationWilsonTwoIndexSymmetricImplF.cc
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/WilsonTMFermionInstantiationWilsonTwoIndexSymmetricImplF.cc b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/WilsonTMFermionInstantiationWilsonTwoIndexSymmetricImplF.cc
new file mode 100644
index 00000000..adfa310c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/WilsonTMFermionInstantiationWilsonTwoIndexSymmetricImplF.cc
@@ -0,0 +1,37 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/WilsonTMFermion.cc
+
+    Copyright (C) 2015
+
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/WilsonTMFermion.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonTMFermionImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonTMFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/impl.h b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/impl.h
new file mode 100644
index 00000000..9c1b14d1
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION WilsonTwoIndexSymmetricImplF
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/CayleyFermion5DInstantiationZWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/CayleyFermion5DInstantiationZWilsonImplD.cc
new file mode 100644
index 00000000..5130db9c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/CayleyFermion5DInstantiationZWilsonImplD.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h>
+
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h>
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dgpu.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class CayleyFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/ContinuedFractionFermion5DInstantiationZWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/ContinuedFractionFermion5DInstantiationZWilsonImplD.cc
new file mode 100644
index 00000000..ca0d6cea
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/ContinuedFractionFermion5DInstantiationZWilsonImplD.cc
@@ -0,0 +1,38 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ContinuedFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/DomainWallEOFAFermionInstantiationZWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/DomainWallEOFAFermionInstantiationZWilsonImplD.cc
new file mode 100644
index 00000000..f7198131
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/DomainWallEOFAFermionInstantiationZWilsonImplD.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class DomainWallEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/MobiusEOFAFermionInstantiationZWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/MobiusEOFAFermionInstantiationZWilsonImplD.cc
new file mode 100644
index 00000000..ce7eaac9
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/MobiusEOFAFermionInstantiationZWilsonImplD.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class MobiusEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/PartialFractionFermion5DInstantiationZWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/PartialFractionFermion5DInstantiationZWilsonImplD.cc
new file mode 100644
index 00000000..757719b6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/PartialFractionFermion5DInstantiationZWilsonImplD.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+ 
+#include "impl.h"
+template class PartialFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/WilsonFermion5DInstantiationZWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/WilsonFermion5DInstantiationZWilsonImplD.cc
new file mode 100644
index 00000000..0dac989c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/WilsonFermion5DInstantiationZWilsonImplD.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/WilsonKernelsInstantiationZWilsonImplD.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/WilsonKernelsInstantiationZWilsonImplD.cc
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/WilsonKernelsInstantiationZWilsonImplD.cc
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/impl.h b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/impl.h
new file mode 100644
index 00000000..d8f93091
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplD/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION ZWilsonImplD
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/CayleyFermion5DInstantiationZWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/CayleyFermion5DInstantiationZWilsonImplDF.cc
new file mode 100644
index 00000000..5130db9c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/CayleyFermion5DInstantiationZWilsonImplDF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h>
+
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h>
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dgpu.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class CayleyFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/ContinuedFractionFermion5DInstantiationZWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/ContinuedFractionFermion5DInstantiationZWilsonImplDF.cc
new file mode 100644
index 00000000..ca0d6cea
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/ContinuedFractionFermion5DInstantiationZWilsonImplDF.cc
@@ -0,0 +1,38 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ContinuedFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/DomainWallEOFAFermionInstantiationZWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/DomainWallEOFAFermionInstantiationZWilsonImplDF.cc
new file mode 100644
index 00000000..f7198131
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/DomainWallEOFAFermionInstantiationZWilsonImplDF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class DomainWallEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/MobiusEOFAFermionInstantiationZWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/MobiusEOFAFermionInstantiationZWilsonImplDF.cc
new file mode 100644
index 00000000..ce7eaac9
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/MobiusEOFAFermionInstantiationZWilsonImplDF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class MobiusEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/PartialFractionFermion5DInstantiationZWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/PartialFractionFermion5DInstantiationZWilsonImplDF.cc
new file mode 100644
index 00000000..757719b6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/PartialFractionFermion5DInstantiationZWilsonImplDF.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+ 
+#include "impl.h"
+template class PartialFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/WilsonFermion5DInstantiationZWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/WilsonFermion5DInstantiationZWilsonImplDF.cc
new file mode 100644
index 00000000..0dac989c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/WilsonFermion5DInstantiationZWilsonImplDF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/WilsonKernelsInstantiationZWilsonImplDF.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/WilsonKernelsInstantiationZWilsonImplDF.cc
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/WilsonKernelsInstantiationZWilsonImplDF.cc
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/impl.h b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/impl.h
new file mode 100644
index 00000000..7daf76ef
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION ZWilsonImplDF
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/CayleyFermion5DInstantiationZWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/CayleyFermion5DInstantiationZWilsonImplF.cc
new file mode 100644
index 00000000..5130db9c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/CayleyFermion5DInstantiationZWilsonImplF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h>
+
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h>
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dgpu.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class CayleyFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/ContinuedFractionFermion5DInstantiationZWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/ContinuedFractionFermion5DInstantiationZWilsonImplF.cc
new file mode 100644
index 00000000..ca0d6cea
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/ContinuedFractionFermion5DInstantiationZWilsonImplF.cc
@@ -0,0 +1,38 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ContinuedFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/DomainWallEOFAFermionInstantiationZWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/DomainWallEOFAFermionInstantiationZWilsonImplF.cc
new file mode 100644
index 00000000..f7198131
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/DomainWallEOFAFermionInstantiationZWilsonImplF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class DomainWallEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/MobiusEOFAFermionInstantiationZWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/MobiusEOFAFermionInstantiationZWilsonImplF.cc
new file mode 100644
index 00000000..ce7eaac9
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/MobiusEOFAFermionInstantiationZWilsonImplF.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class MobiusEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/PartialFractionFermion5DInstantiationZWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/PartialFractionFermion5DInstantiationZWilsonImplF.cc
new file mode 100644
index 00000000..757719b6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/PartialFractionFermion5DInstantiationZWilsonImplF.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+ 
+#include "impl.h"
+template class PartialFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/WilsonFermion5DInstantiationZWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/WilsonFermion5DInstantiationZWilsonImplF.cc
new file mode 100644
index 00000000..0dac989c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/WilsonFermion5DInstantiationZWilsonImplF.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/WilsonKernelsInstantiationZWilsonImplF.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/WilsonKernelsInstantiationZWilsonImplF.cc
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/WilsonKernelsInstantiationZWilsonImplF.cc
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/impl.h b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/impl.h
new file mode 100644
index 00000000..cfb6a73e
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplF/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION ZWilsonImplF
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/CayleyFermion5DInstantiationZWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/CayleyFermion5DInstantiationZWilsonImplFH.cc
new file mode 100644
index 00000000..5130db9c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/CayleyFermion5DInstantiationZWilsonImplFH.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dcache.h>
+
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dvec.h>
+			   //#include <Grid/qcd/action/fermion/implementation/CayleyFermion5Dgpu.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class CayleyFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/ContinuedFractionFermion5DInstantiationZWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/ContinuedFractionFermion5DInstantiationZWilsonImplFH.cc
new file mode 100644
index 00000000..ca0d6cea
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/ContinuedFractionFermion5DInstantiationZWilsonImplFH.cc
@@ -0,0 +1,38 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/ContinuedFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/ContinuedFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class ContinuedFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/DomainWallEOFAFermionInstantiationZWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/DomainWallEOFAFermionInstantiationZWilsonImplFH.cc
new file mode 100644
index 00000000..f7198131
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/DomainWallEOFAFermionInstantiationZWilsonImplFH.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/DomainWallEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/DomainWallEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/DomainWallEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class DomainWallEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/MobiusEOFAFermionInstantiationZWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/MobiusEOFAFermionInstantiationZWilsonImplFH.cc
new file mode 100644
index 00000000..ce7eaac9
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/MobiusEOFAFermionInstantiationZWilsonImplFH.cc
@@ -0,0 +1,44 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/MobiusEOFAFermion.cc
+
+Copyright (C) 2017
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+Author: David Murphy <dmurphy@phys.columbia.edu>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+			   /*  END LEGAL */
+
+#include <Grid/Grid_Eigen_Dense.h>
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/MobiusEOFAFermion.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/MobiusEOFAFermionCache.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class MobiusEOFAFermion<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/PartialFractionFermion5DInstantiationZWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/PartialFractionFermion5DInstantiationZWilsonImplFH.cc
new file mode 100644
index 00000000..757719b6
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/PartialFractionFermion5DInstantiationZWilsonImplFH.cc
@@ -0,0 +1,39 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/qcd/action/fermion/PartialFractionFermion5D.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/PartialFractionFermion5D.h>
+#include <Grid/qcd/action/fermion/implementation/PartialFractionFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+ 
+#include "impl.h"
+template class PartialFractionFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/WilsonFermion5DInstantiationZWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/WilsonFermion5DInstantiationZWilsonImplFH.cc
new file mode 100644
index 00000000..0dac989c
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/WilsonFermion5DInstantiationZWilsonImplFH.cc
@@ -0,0 +1,40 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h>
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonFermion5D<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/WilsonKernelsInstantiationZWilsonImplFH.cc b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/WilsonKernelsInstantiationZWilsonImplFH.cc
new file mode 100644
index 00000000..9af5ed85
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/WilsonKernelsInstantiationZWilsonImplFH.cc
@@ -0,0 +1,47 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>
+
+#ifndef AVX512
+#ifndef QPX
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>; 
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/impl.h b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/impl.h
new file mode 100644
index 00000000..7eb490db
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION ZWilsonImplFH
diff --git a/Grid/qcd/action/fermion/instantiation/generate_instantiations.sh b/Grid/qcd/action/fermion/instantiation/generate_instantiations.sh
new file mode 100755
index 00000000..330dcfa8
--- /dev/null
+++ b/Grid/qcd/action/fermion/instantiation/generate_instantiations.sh
@@ -0,0 +1,104 @@
+#!/bin/sh
+
+STAG_IMPL_LIST=" \
+	   StaggeredImplF \
+	   StaggeredImplD  "
+
+STAG5_IMPL_LIST=""
+
+WILSON_IMPL_LIST=" \
+	   WilsonImplF \
+	   WilsonImplD \
+	   WilsonImplFH \
+	   WilsonImplDF \
+	   WilsonAdjImplF \
+	   WilsonAdjImplD \
+	   WilsonTwoIndexSymmetricImplF \
+	   WilsonTwoIndexSymmetricImplD \
+	   WilsonTwoIndexAntiSymmetricImplF \
+	   WilsonTwoIndexAntiSymmetricImplD \
+	   GparityWilsonImplF \
+	   GparityWilsonImplD \
+	   GparityWilsonImplFH \
+	   GparityWilsonImplDF"
+
+DWF_IMPL_LIST=" \
+	   WilsonImplF \
+	   WilsonImplD \
+	   WilsonImplFH \
+	   WilsonImplDF \
+	   ZWilsonImplF \
+	   ZWilsonImplD \
+	   ZWilsonImplFH \
+	   ZWilsonImplDF "
+
+GDWF_IMPL_LIST=" \
+	   GparityWilsonImplF \
+	   GparityWilsonImplD \
+	   GparityWilsonImplFH \
+	   GparityWilsonImplDF"
+
+
+IMPL_LIST="$STAG_IMPL_LIST  $WILSON_IMPL_LIST $DWF_IMPL_LIST $GDWF_IMPL_LIST"
+
+for impl in $IMPL_LIST
+do
+  echo $impl
+  mkdir -p $impl
+cat > $impl/impl.h <<EOF
+#define IMPLEMENTATION $impl
+EOF
+
+done
+
+CC_LIST="WilsonCloverFermionInstantiation WilsonFermionInstantiation WilsonKernelsInstantiation WilsonTMFermionInstantiation"
+
+for impl in $WILSON_IMPL_LIST
+do
+for f in $CC_LIST
+do
+  ln -f -s ../$f.cc.master $impl/$f$impl.cc 
+done
+done
+
+CC_LIST=" \
+  CayleyFermion5DInstantiation \
+  ContinuedFractionFermion5DInstantiation \
+  DomainWallEOFAFermionInstantiation  \
+  MobiusEOFAFermionInstantiation \
+  PartialFractionFermion5DInstantiation \
+  WilsonFermion5DInstantiation \
+  WilsonKernelsInstantiation "
+
+for impl in $DWF_IMPL_LIST $GDWF_IMPL_LIST
+do
+for f in $CC_LIST
+do
+  ln -f -s ../$f.cc.master $impl/$f$impl.cc 
+done
+done
+
+# overwrite the .cc file in Gparity directories
+for impl in $GDWF_IMPL_LIST
+do
+  ln -f -s ../WilsonKernelsInstantiationGparity.cc.master $impl/WilsonKernelsInstantiation$impl.cc 
+done
+
+
+CC_LIST=" \
+  ImprovedStaggeredFermion5DInstantiation \
+  ImprovedStaggeredFermionInstantiation \
+  StaggeredKernelsInstantiation "
+
+for impl in $STAG_IMPL_LIST
+do
+for f in $CC_LIST
+do
+  ln -f -s ../$f.cc.master $impl/$f$impl.cc 
+done
+done
+
+CC_LIST=" \
+  ImprovedStaggeredFermion5DInstantiation \
+  StaggeredKernelsInstantiation "
+
diff --git a/Grid/qcd/action/gauge/Gauge.h b/Grid/qcd/action/gauge/Gauge.h
index 6f94cf00..b3b34eba 100644
--- a/Grid/qcd/action/gauge/Gauge.h
+++ b/Grid/qcd/action/gauge/Gauge.h
@@ -25,7 +25,7 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef GRID_QCD_GAUGE_H
 #define GRID_QCD_GAUGE_H
 
@@ -34,8 +34,7 @@ directory
 #include <Grid/qcd/action/gauge/WilsonGaugeAction.h>
 #include <Grid/qcd/action/gauge/PlaqPlusRectangleAction.h>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 typedef WilsonGaugeAction<PeriodicGimplR>          WilsonGaugeActionR;
 typedef WilsonGaugeAction<PeriodicGimplF>          WilsonGaugeActionF;
@@ -64,7 +63,6 @@ typedef SymanzikGaugeAction<ConjugateGimplR>        ConjugateSymanzikGaugeAction
 typedef SymanzikGaugeAction<ConjugateGimplF>        ConjugateSymanzikGaugeActionF;
 typedef SymanzikGaugeAction<ConjugateGimplD>        ConjugateSymanzikGaugeActionD;
 
-}}
-
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/gauge/GaugeImplTypes.h b/Grid/qcd/action/gauge/GaugeImplTypes.h
index 52e1e100..b9a5296d 100644
--- a/Grid/qcd/action/gauge/GaugeImplTypes.h
+++ b/Grid/qcd/action/gauge/GaugeImplTypes.h
@@ -29,6 +29,9 @@ directory
 #ifndef GRID_GAUGE_IMPL_TYPES_H
 #define GRID_GAUGE_IMPL_TYPES_H
 
+
+NAMESPACE_BEGIN(Grid);
+
 #define CPS_MD_TIME
 
 #ifdef CPS_MD_TIME
@@ -37,9 +40,6 @@ directory
 #define HMC_MOMENTUM_DENOMINATOR (1.0)
 #endif
 
-namespace Grid {
-namespace QCD {
-
 ////////////////////////////////////////////////////////////////////////
 // Implementation dependent gauge types
 ////////////////////////////////////////////////////////////////////////
@@ -85,13 +85,12 @@ public:
 
 
   // Move this elsewhere? FIXME
-  static inline void AddLink(Field &U, LinkField &W,
-                                  int mu) { // U[mu] += W
-    PARALLEL_FOR_LOOP
-    for (auto ss = 0; ss < U._grid->oSites(); ss++) {
-      U._odata[ss]._internal[mu] =
-          U._odata[ss]._internal[mu] + W._odata[ss]._internal;
-    }
+  static inline void AddLink(Field &U, LinkField &W, int mu) { // U[mu] += W
+    auto U_v = U.View();
+    auto W_v = W.View();
+    thread_for( ss, U.Grid()->oSites(), {
+      U_v[ss](mu) = U_v[ss](mu) + W_v[ss]();
+    });
   }
 
   ///////////////////////////////////////////////////////////
@@ -111,13 +110,12 @@ public:
     //
     //                    =  N \Prod_{x,\mu,a} e^-{1/2 (c_xmua/sqrt{2})^2  }
     // 
-    // Expect c' = cxmua/sqrt(2) to be a unit variance gaussian.
     //
     // Expect cxmua variance sqrt(2).
     //
     // Must scale the momentum by sqrt(2) to invoke CPS and UKQCD conventions
     //
-    LinkField Pmu(P._grid);
+    LinkField Pmu(P.Grid());
     Pmu = Zero();
     for (int mu = 0; mu < Nd; mu++) {
       SU<Nrepresentation>::GaussianFundamentalLieAlgebraMatrix(pRNG, Pmu);
@@ -133,10 +131,13 @@ public:
     //static std::chrono::duration<double> diff;
 
     //auto start = std::chrono::high_resolution_clock::now();
-    parallel_for(int ss=0;ss<P._grid->oSites();ss++){
-      for (int mu = 0; mu < Nd; mu++) 
-        U[ss]._internal[mu] = ProjectOnGroup(Exponentiate(P[ss]._internal[mu], ep, Nexp) * U[ss]._internal[mu]);
-    }
+    auto U_v = U.View();
+    auto P_v = P.View();
+    thread_for(ss, P.Grid()->oSites(),{
+      for (int mu = 0; mu < Nd; mu++) {
+        U_v[ss](mu) = ProjectOnGroup(Exponentiate(P_v[ss](mu), ep, Nexp) * U_v[ss](mu));
+      }
+    });
     
     //auto end = std::chrono::high_resolution_clock::now();
    // diff += end - start;
@@ -144,13 +145,13 @@ public:
   }
 
   static inline RealD FieldSquareNorm(Field& U){
-    LatticeComplex Hloc(U._grid);
-    Hloc = zero;
+    LatticeComplex Hloc(U.Grid());
+    Hloc = Zero();
     for (int mu = 0; mu < Nd; mu++) {
       auto Umu = PeekIndex<LorentzIndex>(U, mu);
       Hloc += trace(Umu * Umu);
     }
-    Complex Hsum = sum(Hloc);
+    auto Hsum = TensorRemove(sum(Hloc));
     return Hsum.real();
   }
 
@@ -176,8 +177,6 @@ typedef GaugeImplTypes<vComplex, SU<Nc>::AdjointDimension> GimplAdjointTypesR;
 typedef GaugeImplTypes<vComplexF, SU<Nc>::AdjointDimension> GimplAdjointTypesF;
 typedef GaugeImplTypes<vComplexD, SU<Nc>::AdjointDimension> GimplAdjointTypesD;
 
-
-} // QCD
-} // Grid
+NAMESPACE_END(Grid);
 
 #endif // GRID_GAUGE_IMPL_TYPES_H
diff --git a/Grid/qcd/action/gauge/GaugeImplementations.h b/Grid/qcd/action/gauge/GaugeImplementations.h
index 2d7464a9..a14aec1b 100644
--- a/Grid/qcd/action/gauge/GaugeImplementations.h
+++ b/Grid/qcd/action/gauge/GaugeImplementations.h
@@ -25,14 +25,13 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef GRID_QCD_GAUGE_IMPLEMENTATIONS_H
 #define GRID_QCD_GAUGE_IMPLEMENTATIONS_H
 
 #include "GaugeImplTypes.h"
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 // Composition with smeared link, bc's etc.. probably need multiple inheritance
 // Variable precision "S" and variable Nc
@@ -42,7 +41,7 @@ public:
 
   ////////////////////////////////////////////////////////////////////////////////////////////////////////////
   // Support needed for the assembly of loops including all boundary condition
-  // effects such as conjugate bcs
+  // effects such as Conjugate bcs
   ////////////////////////////////////////////////////////////////////////////////////////////////////////////
 
   template <class covariant>
@@ -97,7 +96,7 @@ public:
 
   static inline GaugeLinkField
   CovShiftIdentityBackward(const GaugeLinkField &Link, int mu) {
-    GridBase *grid = Link._grid;
+    GridBase *grid = Link.Grid();
     int Lmu = grid->GlobalDimensions()[mu] - 1;
 
     Lattice<iScalar<vInteger>> coor(grid);
@@ -114,7 +113,7 @@ public:
   }
 
   static inline GaugeLinkField ShiftStaple(const GaugeLinkField &Link, int mu) {
-    GridBase *grid = Link._grid;
+    GridBase *grid = Link.Grid();
     int Lmu = grid->GlobalDimensions()[mu] - 1;
 
     Lattice<iScalar<vInteger>> coor(grid);
@@ -141,8 +140,6 @@ typedef ConjugateGaugeImpl<GimplTypesR> ConjugateGimplR; // Real.. whichever pre
 typedef ConjugateGaugeImpl<GimplTypesF> ConjugateGimplF; // Float
 typedef ConjugateGaugeImpl<GimplTypesD> ConjugateGimplD; // Double
 
-
-}
-}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/gauge/Photon.h b/Grid/qcd/action/gauge/Photon.h
index 9afafe6c..465aa8bd 100644
--- a/Grid/qcd/action/gauge/Photon.h
+++ b/Grid/qcd/action/gauge/Photon.h
@@ -27,11 +27,9 @@ Copyright (C) 2015-2018
  See the full license in the file "LICENSE" in the top level distribution directory
  *************************************************************************************/
 /*  END LEGAL */
-#ifndef QCD_PHOTON_ACTION_H
-#define QCD_PHOTON_ACTION_H
+#pragma once
 
-namespace Grid{
-namespace QCD{
+NAMESPACE_BEGIN(Grid);
 
   template <class S>
   class QedGImpl
@@ -119,9 +117,9 @@ namespace QCD{
   void Photon<GImpl>::makeSpatialNorm(LatticeInteger &spNrm)
   {
     LatticeInteger   coor(grid_);
-    std::vector<int> l = grid_->FullDimensions();
+    auto l = grid_->FullDimensions();
 
-    spNrm = zero;
+    spNrm = Zero();
     for(int mu = 0; mu < grid_->Nd() - 1; mu++)
     {
       LatticeCoordinate(coor, mu);
@@ -134,7 +132,7 @@ namespace QCD{
   void Photon<GImpl>::makeKHat(std::vector<GaugeLinkField> &khat)
   {
     const unsigned int nd = grid_->Nd();
-    std::vector<int>   l  = grid_->FullDimensions();
+    auto l  = grid_->FullDimensions();
     Complex            ci(0., 1.);
 
     khat.resize(nd, grid_);
@@ -153,10 +151,10 @@ namespace QCD{
     std::vector<GaugeLinkField> khat;
     GaugeLinkField              lone(grid_);
     const unsigned int          nd = grid_->Nd();
-    std::vector<int>            zm(nd, 0);
+    Coordinate zm(nd, 0);
     ScalarSite                  one = ScalarComplex(1., 0.), z = ScalarComplex(0., 0.);
     
-    out = zero;
+    out = Zero();
     makeKHat(khat);
     for(int mu = 0; mu < nd; mu++)
     {
@@ -175,7 +173,7 @@ namespace QCD{
     {
       case ZmScheme::qedTL:
       {
-        std::vector<int> zm(grid_->Nd(), 0);
+        Coordinate zm(grid_->Nd(), 0);
         ScalarSite       z = ScalarComplex(0., 0.);
         
         pokeSite(z, out, zm);
@@ -208,7 +206,7 @@ namespace QCD{
     LatticeInteger              spNrm(grid_);
     std::vector<GaugeLinkField> khat, a(nd, grid_), aProj(nd, grid_);
 
-    invKHat = zero;
+    invKHat = Zero();
     makeSpatialNorm(spNrm);
     makeKHat(khat);
     for (unsigned int mu = 0; mu < nd; ++mu)
@@ -222,9 +220,9 @@ namespace QCD{
     cst     = ScalarComplex(1., 0.);
     invKHat = where(spNrm == Integer(0), cst, invKHat);
     invKHat = cst/invKHat;
-    cst     = zero;
+    cst     = Zero();
     invKHat = where(spNrm == Integer(0), cst, invKHat);
-    spdiv   = zero;
+    spdiv   = Zero();
     for (unsigned int nu = 0; nu < nd - 1; ++nu)
     {
       spdiv += conjugate(khat[nu])*a[nu];
@@ -272,7 +270,7 @@ namespace QCD{
   void Photon<GImpl>::StochasticWeight(GaugeLinkField &weight)
   {
     const unsigned int nd  = grid_->Nd();
-    std::vector<int>   l   = grid_->FullDimensions();
+    auto  l   = grid_->FullDimensions();
     Integer            vol = 1;
 
     for(unsigned int mu = 0; mu < nd; mu++)
@@ -327,5 +325,5 @@ namespace QCD{
     out = real(out);
   }
   
-}}
-#endif
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/gauge/PlaqPlusRectangleAction.h b/Grid/qcd/action/gauge/PlaqPlusRectangleAction.h
index bdbc8479..639aca19 100644
--- a/Grid/qcd/action/gauge/PlaqPlusRectangleAction.h
+++ b/Grid/qcd/action/gauge/PlaqPlusRectangleAction.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,129 +24,127 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef QCD_PLAQ_PLUS_RECTANGLE_ACTION_H
 #define QCD_PLAQ_PLUS_RECTANGLE_ACTION_H
 
-namespace Grid{
-  namespace QCD{
+NAMESPACE_BEGIN(Grid);  
     
-    ////////////////////////////////////////////////////////////////////////
-    // PlaqPlusRectangleActoin
-    ////////////////////////////////////////////////////////////////////////
-    template<class Gimpl>
-    class PlaqPlusRectangleAction : public Action<typename Gimpl::GaugeField> {
-    public:
+////////////////////////////////////////////////////////////////////////
+// PlaqPlusRectangleActoin
+////////////////////////////////////////////////////////////////////////
+template<class Gimpl>
+class PlaqPlusRectangleAction : public Action<typename Gimpl::GaugeField> {
+public:
 
-      INHERIT_GIMPL_TYPES(Gimpl);
+  INHERIT_GIMPL_TYPES(Gimpl);
 
-    private:
-      RealD c_plaq;
-      RealD c_rect;
+private:
+  RealD c_plaq;
+  RealD c_rect;
 
-    public:
-    PlaqPlusRectangleAction(RealD b,RealD c): c_plaq(b),c_rect(c){};
+public:
+  PlaqPlusRectangleAction(RealD b,RealD c): c_plaq(b),c_rect(c){};
 
-      virtual std::string action_name(){return "PlaqPlusRectangleAction";}
+  virtual std::string action_name(){return "PlaqPlusRectangleAction";}
       
-      virtual void refresh(const GaugeField &U, GridParallelRNG& pRNG) {}; // noop as no pseudoferms
+  virtual void refresh(const GaugeField &U, GridParallelRNG& pRNG) {}; // noop as no pseudoferms
       
-      virtual std::string LogParameters(){
-      	std::stringstream sstream;
-      	sstream << GridLogMessage << "["<<action_name() <<"] c_plaq: " << c_plaq << std::endl;
-      	sstream << GridLogMessage << "["<<action_name() <<"] c_rect: " << c_rect << std::endl;
-      	return sstream.str();
-      }
-
-
-      virtual RealD S(const GaugeField &U) {
-	RealD vol = U._grid->gSites();
-
-	RealD plaq = WilsonLoops<Gimpl>::avgPlaquette(U);
-	RealD rect = WilsonLoops<Gimpl>::avgRectangle(U);
-
-	RealD action=c_plaq*(1.0 -plaq)*(Nd*(Nd-1.0))*vol*0.5
-	            +c_rect*(1.0 -rect)*(Nd*(Nd-1.0))*vol;
-
-	return action;
-      };
-
-      virtual void deriv(const GaugeField &Umu,GaugeField & dSdU) {
-	//extend Ta to include Lorentz indexes
-	RealD factor_p = c_plaq/RealD(Nc)*0.5;
-	RealD factor_r = c_rect/RealD(Nc)*0.5;
-
-	GridBase *grid = Umu._grid;
-
-	std::vector<GaugeLinkField> U (Nd,grid);
-	std::vector<GaugeLinkField> U2(Nd,grid);
-
-	for(int mu=0;mu<Nd;mu++){
-	  U[mu] = PeekIndex<LorentzIndex>(Umu,mu);
-	  WilsonLoops<Gimpl>::RectStapleDouble(U2[mu],U[mu],mu);
-	}
-
-	GaugeLinkField dSdU_mu(grid);
-	GaugeLinkField staple(grid);
-
-	for (int mu=0; mu < Nd; mu++){
-
-	  // Staple in direction mu
-
-	  WilsonLoops<Gimpl>::Staple(staple,Umu,mu);
-
-	  dSdU_mu = Ta(U[mu]*staple)*factor_p;
-
-	  WilsonLoops<Gimpl>::RectStaple(Umu,staple,U2,U,mu);
-
-	  dSdU_mu = dSdU_mu + Ta(U[mu]*staple)*factor_r;
-	  
-	  PokeIndex<LorentzIndex>(dSdU, dSdU_mu, mu);
-	}
-
-      };
-
-    };
-
-    // Convenience for common physically defined cases.
-    //
-    // RBC c1 parameterisation is not really RBC but don't have good
-    // reference and we are happy to change name if prior use of this plaq coeff
-    // parameterisation is made known to us. 
-    template<class Gimpl>
-    class RBCGaugeAction : public PlaqPlusRectangleAction<Gimpl> {
-    public:
-      INHERIT_GIMPL_TYPES(Gimpl);
-      RBCGaugeAction(RealD beta,RealD c1) : PlaqPlusRectangleAction<Gimpl>(beta*(1.0-8.0*c1), beta*c1) {};
-      virtual std::string action_name(){return "RBCGaugeAction";}
-    };
-
-    template<class Gimpl>
-    class IwasakiGaugeAction : public RBCGaugeAction<Gimpl> {
-    public:
-      INHERIT_GIMPL_TYPES(Gimpl);
-      IwasakiGaugeAction(RealD beta) : RBCGaugeAction<Gimpl>(beta,-0.331) {};
-      virtual std::string action_name(){return "IwasakiGaugeAction";}
-    };
-
-    template<class Gimpl>
-    class SymanzikGaugeAction : public RBCGaugeAction<Gimpl> {
-    public:
-      INHERIT_GIMPL_TYPES(Gimpl);
-      SymanzikGaugeAction(RealD beta) : RBCGaugeAction<Gimpl>(beta,-1.0/12.0) {};
-      virtual std::string action_name(){return "SymanzikGaugeAction";}
-    };
-
-    template<class Gimpl>
-    class DBW2GaugeAction : public RBCGaugeAction<Gimpl> {
-    public:
-      INHERIT_GIMPL_TYPES(Gimpl);
-      DBW2GaugeAction(RealD beta) : RBCGaugeAction<Gimpl>(beta,-1.4067) {};
-      virtual std::string action_name(){return "DBW2GaugeAction";}
-    };
-
+  virtual std::string LogParameters(){
+    std::stringstream sstream;
+    sstream << GridLogMessage << "["<<action_name() <<"] c_plaq: " << c_plaq << std::endl;
+    sstream << GridLogMessage << "["<<action_name() <<"] c_rect: " << c_rect << std::endl;
+    return sstream.str();
   }
-}
+
+
+  virtual RealD S(const GaugeField &U) {
+    RealD vol = U.Grid()->gSites();
+
+    RealD plaq = WilsonLoops<Gimpl>::avgPlaquette(U);
+    RealD rect = WilsonLoops<Gimpl>::avgRectangle(U);
+
+    RealD action=c_plaq*(1.0 -plaq)*(Nd*(Nd-1.0))*vol*0.5
+      +c_rect*(1.0 -rect)*(Nd*(Nd-1.0))*vol;
+
+    return action;
+  };
+
+  virtual void deriv(const GaugeField &Umu,GaugeField & dSdU) {
+    //extend Ta to include Lorentz indexes
+    RealD factor_p = c_plaq/RealD(Nc)*0.5;
+    RealD factor_r = c_rect/RealD(Nc)*0.5;
+
+    GridBase *grid = Umu.Grid();
+
+    std::vector<GaugeLinkField> U (Nd,grid);
+    std::vector<GaugeLinkField> U2(Nd,grid);
+
+    for(int mu=0;mu<Nd;mu++){
+      U[mu] = PeekIndex<LorentzIndex>(Umu,mu);
+      WilsonLoops<Gimpl>::RectStapleDouble(U2[mu],U[mu],mu);
+    }
+
+    GaugeLinkField dSdU_mu(grid);
+    GaugeLinkField staple(grid);
+
+    for (int mu=0; mu < Nd; mu++){
+
+      // Staple in direction mu
+
+      WilsonLoops<Gimpl>::Staple(staple,Umu,mu);
+
+      dSdU_mu = Ta(U[mu]*staple)*factor_p;
+
+      WilsonLoops<Gimpl>::RectStaple(Umu,staple,U2,U,mu);
+
+      dSdU_mu = dSdU_mu + Ta(U[mu]*staple)*factor_r;
+	  
+      PokeIndex<LorentzIndex>(dSdU, dSdU_mu, mu);
+    }
+
+  };
+
+};
+
+// Convenience for common physically defined cases.
+//
+// RBC c1 parameterisation is not really RBC but don't have good
+// reference and we are happy to change name if prior use of this plaq coeff
+// parameterisation is made known to us. 
+template<class Gimpl>
+class RBCGaugeAction : public PlaqPlusRectangleAction<Gimpl> {
+public:
+  INHERIT_GIMPL_TYPES(Gimpl);
+  RBCGaugeAction(RealD beta,RealD c1) : PlaqPlusRectangleAction<Gimpl>(beta*(1.0-8.0*c1), beta*c1) {};
+  virtual std::string action_name(){return "RBCGaugeAction";}
+};
+
+template<class Gimpl>
+class IwasakiGaugeAction : public RBCGaugeAction<Gimpl> {
+public:
+  INHERIT_GIMPL_TYPES(Gimpl);
+  IwasakiGaugeAction(RealD beta) : RBCGaugeAction<Gimpl>(beta,-0.331) {};
+  virtual std::string action_name(){return "IwasakiGaugeAction";}
+};
+
+template<class Gimpl>
+class SymanzikGaugeAction : public RBCGaugeAction<Gimpl> {
+public:
+  INHERIT_GIMPL_TYPES(Gimpl);
+  SymanzikGaugeAction(RealD beta) : RBCGaugeAction<Gimpl>(beta,-1.0/12.0) {};
+  virtual std::string action_name(){return "SymanzikGaugeAction";}
+};
+
+template<class Gimpl>
+class DBW2GaugeAction : public RBCGaugeAction<Gimpl> {
+public:
+  INHERIT_GIMPL_TYPES(Gimpl);
+  DBW2GaugeAction(RealD beta) : RBCGaugeAction<Gimpl>(beta,-1.4067) {};
+  virtual std::string action_name(){return "DBW2GaugeAction";}
+};
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/gauge/WilsonGaugeAction.h b/Grid/qcd/action/gauge/WilsonGaugeAction.h
index 77c2424c..40d600d2 100644
--- a/Grid/qcd/action/gauge/WilsonGaugeAction.h
+++ b/Grid/qcd/action/gauge/WilsonGaugeAction.h
@@ -29,19 +29,18 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef QCD_WILSON_GAUGE_ACTION_H
 #define QCD_WILSON_GAUGE_ACTION_H
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 ////////////////////////////////////////////////////////////////////////
 // Wilson Gauge Action .. should I template the Nc etc..
 ////////////////////////////////////////////////////////////////////////
 template <class Gimpl>
 class WilsonGaugeAction : public Action<typename Gimpl::GaugeField> {
- public:  
+public:  
   INHERIT_GIMPL_TYPES(Gimpl);
 
   /////////////////////////// constructors
@@ -60,7 +59,7 @@ class WilsonGaugeAction : public Action<typename Gimpl::GaugeField> {
 
   virtual RealD S(const GaugeField &U) {
     RealD plaq = WilsonLoops<Gimpl>::avgPlaquette(U);
-    RealD vol = U._grid->gSites();
+    RealD vol = U.Grid()->gSites();
     RealD action = beta * (1.0 - plaq) * (Nd * (Nd - 1.0)) * vol * 0.5;
     return action;
   };
@@ -71,25 +70,22 @@ class WilsonGaugeAction : public Action<typename Gimpl::GaugeField> {
 
     RealD factor = 0.5 * beta / RealD(Nc);
 
-    GaugeLinkField Umu(U._grid);
-    GaugeLinkField dSdU_mu(U._grid);
+    GaugeLinkField Umu(U.Grid());
+    GaugeLinkField dSdU_mu(U.Grid());
     for (int mu = 0; mu < Nd; mu++) {
-      Umu = PeekIndex<LorentzIndex>(U, mu);
 
+      Umu = PeekIndex<LorentzIndex>(U, mu);
+      
       // Staple in direction mu
       WilsonLoops<Gimpl>::Staple(dSdU_mu, U, mu);
       dSdU_mu = Ta(Umu * dSdU_mu) * factor;
-
+      
       PokeIndex<LorentzIndex>(dSdU, dSdU_mu, mu);
     }
   }
 private:
   RealD beta;  
-};
-
-
-
-}
-}
+ };
 
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/qcd/action/pseudofermion/Bounds.h b/Grid/qcd/action/pseudofermion/Bounds.h
index 081ebba2..535e1a49 100644
--- a/Grid/qcd/action/pseudofermion/Bounds.h
+++ b/Grid/qcd/action/pseudofermion/Bounds.h
@@ -1,7 +1,6 @@
 #pragma once
 
-namespace Grid{
-  namespace QCD{
+NAMESPACE_BEGIN(Grid);
 
     template<class Field>
     void HighBoundCheck(LinearOperatorBase<Field> &HermOp, 
@@ -20,7 +19,7 @@ namespace Grid{
 						       Field &GaussNoise,
 						       MultiShiftFunction &PowerNegHalf) 
     {
-      GridBase *FermionGrid = GaussNoise._grid;
+      GridBase *FermionGrid = GaussNoise.Grid();
 
       Field X(FermionGrid);
       Field Y(FermionGrid);
@@ -49,5 +48,5 @@ namespace Grid{
       assert( (std::sqrt(Nd/Nx)<tol) && " InverseSqrtBoundsCheck ");
     }
 
-  }
-}
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/action/pseudofermion/EvenOddSchurDifferentiable.h b/Grid/qcd/action/pseudofermion/EvenOddSchurDifferentiable.h
index 90b6dbaa..0d74a560 100644
--- a/Grid/qcd/action/pseudofermion/EvenOddSchurDifferentiable.h
+++ b/Grid/qcd/action/pseudofermion/EvenOddSchurDifferentiable.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,122 +24,121 @@ Author: Guido Cossu <guido.cossu@ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef QCD_EVEN_ODD_SCHUR_DIFFERENTIABLE_H
 #define QCD_EVEN_ODD_SCHUR_DIFFERENTIABLE_H
 
-namespace Grid{
-  namespace QCD{
+NAMESPACE_BEGIN(Grid);  
 
-    // Base even odd HMC on the normal Mee based schur decomposition.
-    //
-    //     M = (Mee Meo) =  (1             0 )   (Mee   0               )  (1 Mee^{-1} Meo)
-    //         (Moe Moo)    (Moe Mee^-1    1 )   (0   Moo-Moe Mee^-1 Meo)  (0   1         )
-    //
-    // Determinant is det of middle factor
-    // This assumes Mee is indept of U.
-    //
-    template<class Impl>
-    class SchurDifferentiableOperator :  public SchurDiagMooeeOperator<FermionOperator<Impl>,typename Impl::FermionField> 
-      {
-      public:
-      INHERIT_IMPL_TYPES(Impl);
+// Base even odd HMC on the normal Mee based schur decomposition.
+//
+//     M = (Mee Meo) =  (1             0 )   (Mee   0               )  (1 Mee^{-1} Meo)
+//         (Moe Moo)    (Moe Mee^-1    1 )   (0   Moo-Moe Mee^-1 Meo)  (0   1         )
+//
+// Determinant is det of middle factor
+// This assumes Mee is indept of U.
+//
+template<class Impl>
+class SchurDifferentiableOperator :  public SchurDiagMooeeOperator<FermionOperator<Impl>,typename Impl::FermionField> 
+{
+public:
+  INHERIT_IMPL_TYPES(Impl);
 
-        typedef FermionOperator<Impl> Matrix;
+  typedef FermionOperator<Impl> Matrix;
 
-        SchurDifferentiableOperator (Matrix &Mat) : SchurDiagMooeeOperator<Matrix,FermionField>(Mat) {};
+  SchurDifferentiableOperator (Matrix &Mat) : SchurDiagMooeeOperator<Matrix,FermionField>(Mat) {};
 
-        void MpcDeriv(GaugeField &Force,const FermionField &U,const FermionField &V) {
+  void MpcDeriv(GaugeField &Force,const FermionField &U,const FermionField &V) {
         
-          GridBase *fgrid   = this->_Mat.FermionGrid();
-          GridBase *fcbgrid = this->_Mat.FermionRedBlackGrid();
+    GridBase *fgrid   = this->_Mat.FermionGrid();
+    GridBase *fcbgrid = this->_Mat.FermionRedBlackGrid();
 
-          FermionField tmp1(fcbgrid);
-          FermionField tmp2(fcbgrid);
+    FermionField tmp1(fcbgrid);
+    FermionField tmp2(fcbgrid);
 
-          conformable(fcbgrid,U._grid);
-          conformable(fcbgrid,V._grid);
+    conformable(fcbgrid,U.Grid());
+    conformable(fcbgrid,V.Grid());
 
-          // Assert the checkerboard?? or code for either
-          assert(U.checkerboard==Odd);
-          assert(V.checkerboard==U.checkerboard);
+    // Assert the checkerboard?? or code for either
+    assert(U.Checkerboard()==Odd);
+    assert(V.Checkerboard()==U.Checkerboard());
 
-          // NOTE Guido: WE DO NOT WANT TO USE THE ucbgrid GRID FOR THE FORCE
-          // it is not conformable with the HMC force field
-	  // Case: Ls vectorised fields
-          // INHERIT FROM THE Force field instead
-          GridRedBlackCartesian* forcecb = new GridRedBlackCartesian(Force._grid);
-          GaugeField ForceO(forcecb);
-          GaugeField ForceE(forcecb);
+    // NOTE Guido: WE DO NOT WANT TO USE THE ucbgrid GRID FOR THE FORCE
+    // it is not conformable with the HMC force field
+    // Case: Ls vectorised fields
+    // INHERIT FROM THE Force field instead
+    GridRedBlackCartesian* forcecb = new GridRedBlackCartesian(Force.Grid());
+    GaugeField ForceO(forcecb);
+    GaugeField ForceE(forcecb);
 
 
-          //  X^dag Der_oe MeeInv Meo Y
-          // Use Mooee as nontrivial but gauge field indept
-          this->_Mat.Meooe   (V,tmp1);      // odd->even -- implicit -0.5 factor to be applied
-	  this->_Mat.MooeeInv(tmp1,tmp2);   // even->even 
-          this->_Mat.MoeDeriv(ForceO,U,tmp2,DaggerNo);
-          //  Accumulate X^dag M_oe MeeInv Der_eo Y
-          this->_Mat.MeooeDag   (U,tmp1);    // even->odd -- implicit -0.5 factor to be applied
-          this->_Mat.MooeeInvDag(tmp1,tmp2); // even->even 
-          this->_Mat.MeoDeriv(ForceE,tmp2,V,DaggerNo);
+    //  X^dag Der_oe MeeInv Meo Y
+    // Use Mooee as nontrivial but gauge field indept
+    this->_Mat.Meooe   (V,tmp1);      // odd->even -- implicit -0.5 factor to be applied
+    this->_Mat.MooeeInv(tmp1,tmp2);   // even->even 
+    this->_Mat.MoeDeriv(ForceO,U,tmp2,DaggerNo);
+    //  Accumulate X^dag M_oe MeeInv Der_eo Y
+    this->_Mat.MeooeDag   (U,tmp1);    // even->odd -- implicit -0.5 factor to be applied
+    this->_Mat.MooeeInvDag(tmp1,tmp2); // even->even 
+    this->_Mat.MeoDeriv(ForceE,tmp2,V,DaggerNo);
           
-          assert(ForceE.checkerboard==Even);
-          assert(ForceO.checkerboard==Odd);
+    assert(ForceE.Checkerboard()==Even);
+    assert(ForceO.Checkerboard()==Odd);
 
-          setCheckerboard(Force,ForceE); 
-          setCheckerboard(Force,ForceO);
-          Force=-Force;
-
-          delete forcecb;
-        }
-
-
-        void MpcDagDeriv(GaugeField &Force,const FermionField &U,const FermionField &V) {
-        
-          GridBase *fgrid   = this->_Mat.FermionGrid();
-          GridBase *fcbgrid = this->_Mat.FermionRedBlackGrid();
-
-          FermionField tmp1(fcbgrid);
-          FermionField tmp2(fcbgrid);
-
-          conformable(fcbgrid,U._grid);
-          conformable(fcbgrid,V._grid);
-
-          // Assert the checkerboard?? or code for either
-          assert(V.checkerboard==Odd);
-          assert(V.checkerboard==V.checkerboard);
-
-          // NOTE Guido: WE DO NOT WANT TO USE THE ucbgrid GRID FOR THE FORCE
-          // it is not conformable with the HMC force field
-          // INHERIT FROM THE Force field instead
-	  GridRedBlackCartesian* forcecb = new GridRedBlackCartesian(Force._grid);
-          GaugeField ForceO(forcecb);
-          GaugeField ForceE(forcecb);
-
-          //  X^dag Der_oe MeeInv Meo Y
-          // Use Mooee as nontrivial but gauge field indept
-          this->_Mat.MeooeDag   (V,tmp1);      // odd->even -- implicit -0.5 factor to be applied
-          this->_Mat.MooeeInvDag(tmp1,tmp2);   // even->even 
-          this->_Mat.MoeDeriv(ForceO,U,tmp2,DaggerYes);
-          
-          //  Accumulate X^dag M_oe MeeInv Der_eo Y
-          this->_Mat.Meooe   (U,tmp1);    // even->odd -- implicit -0.5 factor to be applied
-          this->_Mat.MooeeInv(tmp1,tmp2); // even->even 
-          this->_Mat.MeoDeriv(ForceE,tmp2,V,DaggerYes);
-
-          assert(ForceE.checkerboard==Even);
-          assert(ForceO.checkerboard==Odd);
-
-          setCheckerboard(Force,ForceE); 
-          setCheckerboard(Force,ForceO);
-          Force=-Force;
-
-          delete forcecb;
-        }
-
-    };
+    setCheckerboard(Force,ForceE); 
+    setCheckerboard(Force,ForceO);
+    Force=-Force;
 
+    delete forcecb;
   }
-}
+
+
+  void MpcDagDeriv(GaugeField &Force,const FermionField &U,const FermionField &V) {
+        
+    GridBase *fgrid   = this->_Mat.FermionGrid();
+    GridBase *fcbgrid = this->_Mat.FermionRedBlackGrid();
+
+    FermionField tmp1(fcbgrid);
+    FermionField tmp2(fcbgrid);
+
+    conformable(fcbgrid,U.Grid());
+    conformable(fcbgrid,V.Grid());
+
+    // Assert the checkerboard?? or code for either
+    assert(V.Checkerboard()==Odd);
+    assert(V.Checkerboard()==V.Checkerboard());
+
+    // NOTE Guido: WE DO NOT WANT TO USE THE ucbgrid GRID FOR THE FORCE
+    // it is not conformable with the HMC force field
+    // INHERIT FROM THE Force field instead
+    GridRedBlackCartesian* forcecb = new GridRedBlackCartesian(Force.Grid());
+    GaugeField ForceO(forcecb);
+    GaugeField ForceE(forcecb);
+
+    //  X^dag Der_oe MeeInv Meo Y
+    // Use Mooee as nontrivial but gauge field indept
+    this->_Mat.MeooeDag   (V,tmp1);      // odd->even -- implicit -0.5 factor to be applied
+    this->_Mat.MooeeInvDag(tmp1,tmp2);   // even->even 
+    this->_Mat.MoeDeriv(ForceO,U,tmp2,DaggerYes);
+          
+    //  Accumulate X^dag M_oe MeeInv Der_eo Y
+    this->_Mat.Meooe   (U,tmp1);    // even->odd -- implicit -0.5 factor to be applied
+    this->_Mat.MooeeInv(tmp1,tmp2); // even->even 
+    this->_Mat.MeoDeriv(ForceE,tmp2,V,DaggerYes);
+
+    assert(ForceE.Checkerboard()==Even);
+    assert(ForceO.Checkerboard()==Odd);
+
+    setCheckerboard(Force,ForceE); 
+    setCheckerboard(Force,ForceO);
+    Force=-Force;
+
+    delete forcecb;
+  }
+
+};
+
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/qcd/action/pseudofermion/ExactOneFlavourRatio.h b/Grid/qcd/action/pseudofermion/ExactOneFlavourRatio.h
index c6746a88..9fc0a3b0 100644
--- a/Grid/qcd/action/pseudofermion/ExactOneFlavourRatio.h
+++ b/Grid/qcd/action/pseudofermion/ExactOneFlavourRatio.h
@@ -38,8 +38,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #ifndef QCD_PSEUDOFERMION_EXACT_ONE_FLAVOUR_RATIO_H
 #define QCD_PSEUDOFERMION_EXACT_ONE_FLAVOUR_RATIO_H
 
-namespace Grid{
-namespace QCD{
+NAMESPACE_BEGIN(Grid);
 
   ///////////////////////////////////////////////////////////////
   // Exact one flavour implementation of DWF determinant ratio //
@@ -67,6 +66,13 @@ namespace QCD{
 
     public:
 
+      ExactOneFlavourRatioPseudoFermionAction(AbstractEOFAFermion<Impl>& _Lop, 
+					      AbstractEOFAFermion<Impl>& _Rop,
+					      OperatorFunction<FermionField>& CG, 
+					      Params& p, 
+					      bool use_fc=false) 
+	: ExactOneFlavourRatioPseudoFermionAction(_Lop,_Rop,CG,CG,CG,CG,CG,p,use_fc) {};
+	
       ExactOneFlavourRatioPseudoFermionAction(AbstractEOFAFermion<Impl>& _Lop, 
 					      AbstractEOFAFermion<Impl>& _Rop,
 					      OperatorFunction<FermionField>& HeatbathCG, 
@@ -151,7 +157,7 @@ namespace QCD{
         spProj(eta, tmp[0], -1, Lop.Ls);
         Lop.Omega(tmp[0], tmp[1], -1, 0);
         G5R5(CG_src, tmp[1]);
-        tmp[1] = zero;
+        tmp[1] = Zero();
         for(int k=0; k<param.degree; ++k){
           gamma_l = 1.0 / ( 1.0 + PowerNegHalf.poles[k] );
           Lop.RefreshShiftCoefficients(-gamma_l);
@@ -161,7 +167,7 @@ namespace QCD{
             SolverHB(Lop, CG_src, CG_soln);
             prev_solns.push_back(CG_soln);
           } else {
-            CG_soln = zero; // Just use zero as the initial guess
+            CG_soln = Zero(); // Just use zero as the initial guess
             SolverHB(Lop, CG_src, CG_soln);
           }
           Lop.Dtilde(CG_soln, tmp[0]); // We actually solved Cayley preconditioned system: transform back
@@ -177,7 +183,7 @@ namespace QCD{
         spProj(eta, tmp[0], 1, Rop.Ls);
         Rop.Omega(tmp[0], tmp[1], 1, 0);
         G5R5(CG_src, tmp[1]);
-        tmp[1] = zero;
+        tmp[1] = Zero();
         if(use_heatbath_forecasting){ prev_solns.clear(); } // empirically, LH solns don't help for RH solves
         for(int k=0; k<param.degree; ++k){
           gamma_l = 1.0 / ( 1.0 + PowerNegHalf.poles[k] );
@@ -188,7 +194,7 @@ namespace QCD{
             SolverHB(Rop, CG_src, CG_soln);
             prev_solns.push_back(CG_soln);
           } else {
-            CG_soln = zero;
+            CG_soln = Zero();
             SolverHB(Rop, CG_src, CG_soln);
           }
           Rop.Dtilde(CG_soln, tmp[0]); // We actually solved Cayley preconditioned system: transform back
@@ -223,7 +229,7 @@ namespace QCD{
         spProj(Phi, spProj_Phi, -1, Lop.Ls);
         Lop.Omega(spProj_Phi, tmp[0], -1, 0);
         G5R5(tmp[1], tmp[0]);
-        tmp[0] = zero;
+        tmp[0] = Zero();
         SolverL(Lop, tmp[1], tmp[0]);
         Lop.Dtilde(tmp[0], tmp[1]); // We actually solved Cayley preconditioned system: transform back
         Lop.Omega(tmp[1], tmp[0], -1, 1);
@@ -234,7 +240,7 @@ namespace QCD{
         spProj(Phi, spProj_Phi, 1, Rop.Ls);
         Rop.Omega(spProj_Phi, tmp[0], 1, 0);
         G5R5(tmp[1], tmp[0]);
-        tmp[0] = zero;
+        tmp[0] = Zero();
         SolverR(Rop, tmp[1], tmp[0]);
         Rop.Dtilde(tmp[0], tmp[1]);
         Rop.Omega(tmp[1], tmp[0], 1, 1);
@@ -258,7 +264,7 @@ namespace QCD{
         spProj(Phi, spProj_Phi, -1, Lop.Ls);
         Lop.Omega(spProj_Phi, tmp[0], -1, 0);
         G5R5(tmp[1], tmp[0]);
-        tmp[0] = zero;
+        tmp[0] = Zero();
         SolverL(Lop, tmp[1], tmp[0]);
         Lop.Dtilde(tmp[0], tmp[1]); // We actually solved Cayley preconditioned system: transform back
         Lop.Omega(tmp[1], tmp[0], -1, 1);
@@ -269,7 +275,7 @@ namespace QCD{
         spProj(Phi, spProj_Phi, 1, Rop.Ls);
         Rop.Omega(spProj_Phi, tmp[0], 1, 0);
         G5R5(tmp[1], tmp[0]);
-        tmp[0] = zero;
+        tmp[0] = Zero();
         SolverR(Rop, tmp[1], tmp[0]);
         Rop.Dtilde(tmp[0], tmp[1]);
         Rop.Omega(tmp[1], tmp[0], 1, 1);
@@ -302,7 +308,7 @@ namespace QCD{
         spProj(Phi, spProj_Phi, -1, Lop.Ls);
         Lop.Omega(spProj_Phi, Omega_spProj_Phi, -1, 0);
         G5R5(CG_src, Omega_spProj_Phi);
-        spProj_Phi = zero;
+        spProj_Phi = Zero();
         DerivativeSolverL(Lop, CG_src, spProj_Phi);
         Lop.Dtilde(spProj_Phi, Chi);
         G5R5(g5_R5_Chi, Chi);
@@ -314,7 +320,7 @@ namespace QCD{
         spProj(Phi, spProj_Phi, 1, Rop.Ls);
         Rop.Omega(spProj_Phi, Omega_spProj_Phi, 1, 0);
         G5R5(CG_src, Omega_spProj_Phi);
-        spProj_Phi = zero;
+        spProj_Phi = Zero();
         DerivativeSolverR(Rop, CG_src, spProj_Phi);
         Rop.Dtilde(spProj_Phi, Chi);
         G5R5(g5_R5_Chi, Chi);
@@ -322,6 +328,7 @@ namespace QCD{
         dSdU = dSdU + Rop.k * force;
       };
   };
-}}
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/pseudofermion/OneFlavourEvenOddRational.h b/Grid/qcd/action/pseudofermion/OneFlavourEvenOddRational.h
index 76f69485..56dff94d 100644
--- a/Grid/qcd/action/pseudofermion/OneFlavourEvenOddRational.h
+++ b/Grid/qcd/action/pseudofermion/OneFlavourEvenOddRational.h
@@ -26,12 +26,11 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef QCD_PSEUDOFERMION_ONE_FLAVOUR_EVEN_ODD_RATIONAL_H
 #define QCD_PSEUDOFERMION_ONE_FLAVOUR_EVEN_ODD_RATIONAL_H
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 ///////////////////////////////////////
 // One flavour rational
@@ -45,8 +44,8 @@ namespace QCD {
 
 template <class Impl>
 class OneFlavourEvenOddRationalPseudoFermionAction
-    : public Action<typename Impl::GaugeField> {
- public:
+  : public Action<typename Impl::GaugeField> {
+public:
   INHERIT_IMPL_TYPES(Impl);
 
   typedef OneFlavourRationalParams Params;
@@ -57,7 +56,7 @@ class OneFlavourEvenOddRationalPseudoFermionAction
   MultiShiftFunction PowerQuarter;
   MultiShiftFunction PowerNegQuarter;
 
- private:
+private:
   FermionOperator<Impl> &FermOp;  // the basic operator
 
   // NOT using "Nroots"; IroIro is -- perhaps later, but this wasn't good for us
@@ -67,13 +66,13 @@ class OneFlavourEvenOddRationalPseudoFermionAction
   FermionField PhiEven;  // the pseudo fermion field for this trajectory
   FermionField PhiOdd;   // the pseudo fermion field for this trajectory
 
- public:
+public:
   OneFlavourEvenOddRationalPseudoFermionAction(FermionOperator<Impl> &Op,
                                                Params &p)
-      : FermOp(Op),
-        PhiEven(Op.FermionRedBlackGrid()),
-        PhiOdd(Op.FermionRedBlackGrid()),
-        param(p) {
+    : FermOp(Op),
+      PhiEven(Op.FermionRedBlackGrid()),
+      PhiOdd(Op.FermionRedBlackGrid()),
+      param(p) {
     AlgRemez remez(param.lo, param.hi, param.precision);
 
     // MdagM^(+- 1/2)
@@ -139,7 +138,7 @@ class OneFlavourEvenOddRationalPseudoFermionAction
     //////////////////////////////////////////////////////
 
     assert(FermOp.ConstEE() == 1);
-    PhiEven = zero;
+    PhiEven = Zero();
   };
 
   //////////////////////////////////////////////////////
@@ -166,7 +165,7 @@ class OneFlavourEvenOddRationalPseudoFermionAction
 
     RealD action = norm2(Y);
     std::cout << GridLogMessage << "Pseudofermion action FIXME -- is -1/4 "
-                                   "solve or -1/2 solve faster??? "
+      "solve or -1/2 solve faster??? "
               << action << std::endl;
 
     return action;
@@ -213,7 +212,7 @@ class OneFlavourEvenOddRationalPseudoFermionAction
 
     msCG(Mpc, PhiOdd, MPhi_k);
 
-    dSdU = zero;
+    dSdU = Zero();
     for (int k = 0; k < Npole; k++) {
       RealD ak = PowerNegHalf.residues[k];
 
@@ -229,7 +228,7 @@ class OneFlavourEvenOddRationalPseudoFermionAction
     // dSdU = Ta(dSdU);
   };
 };
-}
-}
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/pseudofermion/OneFlavourEvenOddRationalRatio.h b/Grid/qcd/action/pseudofermion/OneFlavourEvenOddRationalRatio.h
index 84fe4de0..e5f0b602 100644
--- a/Grid/qcd/action/pseudofermion/OneFlavourEvenOddRationalRatio.h
+++ b/Grid/qcd/action/pseudofermion/OneFlavourEvenOddRationalRatio.h
@@ -28,8 +28,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #ifndef QCD_PSEUDOFERMION_ONE_FLAVOUR_EVEN_ODD_RATIONAL_RATIO_H
 #define QCD_PSEUDOFERMION_ONE_FLAVOUR_EVEN_ODD_RATIONAL_RATIO_H
 
-namespace Grid{
-  namespace QCD{
+NAMESPACE_BEGIN(Grid);
 
     ///////////////////////////////////////
     // One flavour rational
@@ -145,7 +144,7 @@ namespace Grid{
 
 	assert(NumOp.ConstEE() == 1);
 	assert(DenOp.ConstEE() == 1);
-	PhiEven = zero;
+	PhiEven = Zero();
 	
       };
 
@@ -245,7 +244,7 @@ namespace Grid{
 
 	RealD ak;
 
-	dSdU = zero;
+	dSdU = Zero();
 
 	// With these building blocks  
 	//  
@@ -282,8 +281,7 @@ namespace Grid{
 
       };
     };
-  }
-}
 
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/pseudofermion/OneFlavourRational.h b/Grid/qcd/action/pseudofermion/OneFlavourRational.h
index 3aa6c780..f6c823c9 100644
--- a/Grid/qcd/action/pseudofermion/OneFlavourRational.h
+++ b/Grid/qcd/action/pseudofermion/OneFlavourRational.h
@@ -28,8 +28,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #ifndef QCD_PSEUDOFERMION_ONE_FLAVOUR_RATIONAL_H
 #define QCD_PSEUDOFERMION_ONE_FLAVOUR_RATIONAL_H
 
-namespace Grid{
-  namespace QCD{
+NAMESPACE_BEGIN(Grid);
 
     ///////////////////////////////////////
     // One flavour rational
@@ -196,7 +195,7 @@ namespace Grid{
 
 	msCG(MdagMOp,Phi,MPhi_k);
 
-	dSdU = zero;
+	dSdU = Zero();
 	for(int k=0;k<Npole;k++){
 
 	  RealD ak = PowerNegHalf.residues[k];
@@ -214,8 +213,7 @@ namespace Grid{
 
       };
     };
-  }
-}
 
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/pseudofermion/OneFlavourRationalRatio.h b/Grid/qcd/action/pseudofermion/OneFlavourRationalRatio.h
index c1252e75..5fae2fe9 100644
--- a/Grid/qcd/action/pseudofermion/OneFlavourRationalRatio.h
+++ b/Grid/qcd/action/pseudofermion/OneFlavourRationalRatio.h
@@ -28,8 +28,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #ifndef QCD_PSEUDOFERMION_ONE_FLAVOUR_RATIONAL_RATIO_H
 #define QCD_PSEUDOFERMION_ONE_FLAVOUR_RATIONAL_RATIO_H
 
-namespace Grid{
-  namespace QCD{
+NAMESPACE_BEGIN(Grid);
 
     ///////////////////////////////////////
     // One flavour rational
@@ -231,7 +230,7 @@ namespace Grid{
 
 	RealD ak;
 
-	dSdU = zero;
+	dSdU = Zero();
 
 	// With these building blocks  
 	//  
@@ -268,8 +267,7 @@ namespace Grid{
 
       };
     };
-  }
-}
 
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/pseudofermion/TwoFlavour.h b/Grid/qcd/action/pseudofermion/TwoFlavour.h
index f189e71c..f905a675 100644
--- a/Grid/qcd/action/pseudofermion/TwoFlavour.h
+++ b/Grid/qcd/action/pseudofermion/TwoFlavour.h
@@ -26,22 +26,21 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef QCD_PSEUDOFERMION_TWO_FLAVOUR_H
 #define QCD_PSEUDOFERMION_TWO_FLAVOUR_H
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 ////////////////////////////////////////////////////////////////////////
 // Two flavour pseudofermion action for any dop
 ////////////////////////////////////////////////////////////////////////
 template <class Impl>
 class TwoFlavourPseudoFermionAction : public Action<typename Impl::GaugeField> {
- public:
+public:
   INHERIT_IMPL_TYPES(Impl);
 
- private:
+private:
   FermionOperator<Impl> &FermOp;  // the basic operator
 
   OperatorFunction<FermionField> &DerivativeSolver;
@@ -50,17 +49,17 @@ class TwoFlavourPseudoFermionAction : public Action<typename Impl::GaugeField> {
 
   FermionField Phi;  // the pseudo fermion field for this trajectory
 
- public:
+public:
   /////////////////////////////////////////////////
   // Pass in required objects.
   /////////////////////////////////////////////////
   TwoFlavourPseudoFermionAction(FermionOperator<Impl> &Op,
                                 OperatorFunction<FermionField> &DS,
                                 OperatorFunction<FermionField> &AS)
-      : FermOp(Op),
-        DerivativeSolver(DS),
-        ActionSolver(AS),
-        Phi(Op.FermionGrid()){};
+    : FermOp(Op),
+      DerivativeSolver(DS),
+      ActionSolver(AS),
+      Phi(Op.FermionGrid()){};
 
 
   virtual std::string action_name(){return "TwoFlavourPseudoFermionAction";}
@@ -111,7 +110,7 @@ class TwoFlavourPseudoFermionAction : public Action<typename Impl::GaugeField> {
     FermionField Y(FermOp.FermionGrid());
 
     MdagMLinearOperator<FermionOperator<Impl>, FermionField> MdagMOp(FermOp);
-    X = zero;
+    X = Zero();
     ActionSolver(MdagMOp, Phi, X);
     MdagMOp.Op(X, Y);
 
@@ -138,7 +137,7 @@ class TwoFlavourPseudoFermionAction : public Action<typename Impl::GaugeField> {
 
     MdagMLinearOperator<FermionOperator<Impl>, FermionField> MdagMOp(FermOp);
 
-    X = zero;
+    X = Zero();
     DerivativeSolver(MdagMOp, Phi, X); // X = (MdagM)^-1 phi    
     MdagMOp.Op(X, Y);                  // Y = M X = (Mdag)^-1 phi
 
@@ -153,7 +152,7 @@ class TwoFlavourPseudoFermionAction : public Action<typename Impl::GaugeField> {
     // not taking here the traceless antihermitian component
   };
 };
-}
-}
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/pseudofermion/TwoFlavourEvenOdd.h b/Grid/qcd/action/pseudofermion/TwoFlavourEvenOdd.h
index 0bbc0ae6..a3cf8f08 100644
--- a/Grid/qcd/action/pseudofermion/TwoFlavourEvenOdd.h
+++ b/Grid/qcd/action/pseudofermion/TwoFlavourEvenOdd.h
@@ -26,164 +26,163 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef QCD_PSEUDOFERMION_TWO_FLAVOUR_EVEN_ODD_H
 #define QCD_PSEUDOFERMION_TWO_FLAVOUR_EVEN_ODD_H
 
-namespace Grid {
-  namespace QCD {
-
-    ////////////////////////////////////////////////////////////////////////
-    // Two flavour pseudofermion action for any EO prec dop
-    ////////////////////////////////////////////////////////////////////////
-    template <class Impl>
-    class TwoFlavourEvenOddPseudoFermionAction
-      : public Action<typename Impl::GaugeField> {
-    public:
-      INHERIT_IMPL_TYPES(Impl);
-
-    private:
-      FermionOperator<Impl> &FermOp;  // the basic operator
-
-      OperatorFunction<FermionField> &DerivativeSolver;
-      OperatorFunction<FermionField> &ActionSolver;
-
-      FermionField PhiOdd;   // the pseudo fermion field for this trajectory
-      FermionField PhiEven;  // the pseudo fermion field for this trajectory
-
-    public:
-      /////////////////////////////////////////////////
-      // Pass in required objects.
-      /////////////////////////////////////////////////
-      TwoFlavourEvenOddPseudoFermionAction(FermionOperator<Impl> &Op,
-					   OperatorFunction<FermionField> &DS,
-					   OperatorFunction<FermionField> &AS)
-	: FermOp(Op),
-	  DerivativeSolver(DS),
-	  ActionSolver(AS),
-	  PhiEven(Op.FermionRedBlackGrid()),
-	  PhiOdd(Op.FermionRedBlackGrid())
-      {};
+NAMESPACE_BEGIN(Grid);
   
-      virtual std::string action_name(){return "TwoFlavourEvenOddPseudoFermionAction";}
+
+////////////////////////////////////////////////////////////////////////
+// Two flavour pseudofermion action for any EO prec dop
+////////////////////////////////////////////////////////////////////////
+template <class Impl>
+class TwoFlavourEvenOddPseudoFermionAction
+  : public Action<typename Impl::GaugeField> {
+public:
+  INHERIT_IMPL_TYPES(Impl);
+
+private:
+  FermionOperator<Impl> &FermOp;  // the basic operator
+
+  OperatorFunction<FermionField> &DerivativeSolver;
+  OperatorFunction<FermionField> &ActionSolver;
+
+  FermionField PhiOdd;   // the pseudo fermion field for this trajectory
+  FermionField PhiEven;  // the pseudo fermion field for this trajectory
+
+public:
+  /////////////////////////////////////////////////
+  // Pass in required objects.
+  /////////////////////////////////////////////////
+  TwoFlavourEvenOddPseudoFermionAction(FermionOperator<Impl> &Op,
+				       OperatorFunction<FermionField> &DS,
+				       OperatorFunction<FermionField> &AS)
+    : FermOp(Op),
+      DerivativeSolver(DS),
+      ActionSolver(AS),
+      PhiEven(Op.FermionRedBlackGrid()),
+      PhiOdd(Op.FermionRedBlackGrid())
+  {};
+  
+  virtual std::string action_name(){return "TwoFlavourEvenOddPseudoFermionAction";}
       
-      virtual std::string LogParameters(){
-	std::stringstream sstream;
-	sstream << GridLogMessage << "["<<action_name()<<"] has no parameters" << std::endl;
-	return sstream.str();
-      }  
+  virtual std::string LogParameters(){
+    std::stringstream sstream;
+    sstream << GridLogMessage << "["<<action_name()<<"] has no parameters" << std::endl;
+    return sstream.str();
+  }  
 
 
-      //////////////////////////////////////////////////////////////////////////////////////
-      // Push the gauge field in to the dops. Assume any BC's and smearing already applied
-      //////////////////////////////////////////////////////////////////////////////////////
-      virtual void refresh(const GaugeField &U, GridParallelRNG& pRNG) {
+  //////////////////////////////////////////////////////////////////////////////////////
+  // Push the gauge field in to the dops. Assume any BC's and smearing already applied
+  //////////////////////////////////////////////////////////////////////////////////////
+  virtual void refresh(const GaugeField &U, GridParallelRNG& pRNG) {
     
-	// P(phi) = e^{- phi^dag (MpcdagMpc)^-1 phi}
-	// Phi = McpDag eta 
-	// P(eta) = e^{- eta^dag eta}
-	//
-	// e^{x^2/2 sig^2} => sig^2 = 0.5.
+    // P(phi) = e^{- phi^dag (MpcdagMpc)^-1 phi}
+    // Phi = McpDag eta 
+    // P(eta) = e^{- eta^dag eta}
+    //
+    // e^{x^2/2 sig^2} => sig^2 = 0.5.
     
-	RealD scale = std::sqrt(0.5);
+    RealD scale = std::sqrt(0.5);
     
-	FermionField eta    (FermOp.FermionGrid());
-	FermionField etaOdd (FermOp.FermionRedBlackGrid());
-	FermionField etaEven(FermOp.FermionRedBlackGrid());
+    FermionField eta    (FermOp.FermionGrid());
+    FermionField etaOdd (FermOp.FermionRedBlackGrid());
+    FermionField etaEven(FermOp.FermionRedBlackGrid());
     
-	gaussian(pRNG,eta);
-	pickCheckerboard(Even,etaEven,eta);
-	pickCheckerboard(Odd,etaOdd,eta);
+    gaussian(pRNG,eta);
+    pickCheckerboard(Even,etaEven,eta);
+    pickCheckerboard(Odd,etaOdd,eta);
     
-	FermOp.ImportGauge(U);
-	SchurDifferentiableOperator<Impl> PCop(FermOp);
+    FermOp.ImportGauge(U);
+    SchurDifferentiableOperator<Impl> PCop(FermOp);
     
     
-	PCop.MpcDag(etaOdd,PhiOdd);
+    PCop.MpcDag(etaOdd,PhiOdd);
     
-	FermOp.MooeeDag(etaEven,PhiEven);
+    FermOp.MooeeDag(etaEven,PhiEven);
     
-	PhiOdd =PhiOdd*scale;
-	PhiEven=PhiEven*scale;
+    PhiOdd =PhiOdd*scale;
+    PhiEven=PhiEven*scale;
     
-      };
+  };
   
-      //////////////////////////////////////////////////////
-      // S = phi^dag (Mdag M)^-1 phi  (odd)
-      //   + phi^dag (Mdag M)^-1 phi  (even)
-      //////////////////////////////////////////////////////
-      virtual RealD S(const GaugeField &U) {
+  //////////////////////////////////////////////////////
+  // S = phi^dag (Mdag M)^-1 phi  (odd)
+  //   + phi^dag (Mdag M)^-1 phi  (even)
+  //////////////////////////////////////////////////////
+  virtual RealD S(const GaugeField &U) {
 	
-	FermOp.ImportGauge(U);
+    FermOp.ImportGauge(U);
 
-	FermionField X(FermOp.FermionRedBlackGrid());
-	FermionField Y(FermOp.FermionRedBlackGrid());
+    FermionField X(FermOp.FermionRedBlackGrid());
+    FermionField Y(FermOp.FermionRedBlackGrid());
 	
-	SchurDifferentiableOperator<Impl> PCop(FermOp);
+    SchurDifferentiableOperator<Impl> PCop(FermOp);
 
-	X=zero;
-	ActionSolver(PCop,PhiOdd,X);
-	PCop.Op(X,Y);
-	RealD action = norm2(Y);
+    X=Zero();
+    ActionSolver(PCop,PhiOdd,X);
+    PCop.Op(X,Y);
+    RealD action = norm2(Y);
 
-	// The EE factorised block; normally can replace with zero if det is constant (gauge field indept)
-	// Only really clover term that creates this.
-	FermOp.MooeeInvDag(PhiEven,Y);
-	action = action + norm2(Y);
+    // The EE factorised block; normally can replace with zero if det is constant (gauge field indept)
+    // Only really clover term that creates this.
+    FermOp.MooeeInvDag(PhiEven,Y);
+    action = action + norm2(Y);
 
-	std::cout << GridLogMessage << "Pseudofermion EO action "<<action<<std::endl;
-	return action;
-      };
+    std::cout << GridLogMessage << "Pseudofermion EO action "<<action<<std::endl;
+    return action;
+  };
 
-      //////////////////////////////////////////////////////
-      //
-      // dS/du = - phi^dag  (Mdag M)^-1 [ Mdag dM + dMdag M ]  (Mdag M)^-1 phi
-      //       = - phi^dag M^-1 dM (MdagM)^-1 phi -  phi^dag (MdagM)^-1 dMdag dM (Mdag)^-1 phi 
-      //
-      //       = - Ydag dM X  - Xdag dMdag Y
-      //
-      //////////////////////////////////////////////////////
-      virtual void deriv(const GaugeField &U,GaugeField & dSdU) {
-	FermOp.ImportGauge(U);
+  //////////////////////////////////////////////////////
+  //
+  // dS/du = - phi^dag  (Mdag M)^-1 [ Mdag dM + dMdag M ]  (Mdag M)^-1 phi
+  //       = - phi^dag M^-1 dM (MdagM)^-1 phi -  phi^dag (MdagM)^-1 dMdag dM (Mdag)^-1 phi 
+  //
+  //       = - Ydag dM X  - Xdag dMdag Y
+  //
+  //////////////////////////////////////////////////////
+  virtual void deriv(const GaugeField &U,GaugeField & dSdU) {
+    FermOp.ImportGauge(U);
 
-	FermionField X(FermOp.FermionRedBlackGrid());
-	FermionField Y(FermOp.FermionRedBlackGrid());
-	GaugeField tmp(FermOp.GaugeGrid());
+    FermionField X(FermOp.FermionRedBlackGrid());
+    FermionField Y(FermOp.FermionRedBlackGrid());
+    GaugeField tmp(FermOp.GaugeGrid());
 
-	SchurDifferentiableOperator<Impl> Mpc(FermOp);
+    SchurDifferentiableOperator<Impl> Mpc(FermOp);
 
-	// Our conventions really make this UdSdU; We do not differentiate wrt Udag here.
-	// So must take dSdU - adj(dSdU) and left multiply by mom to get dS/dt.
+    // Our conventions really make this UdSdU; We do not differentiate wrt Udag here.
+    // So must take dSdU - adj(dSdU) and left multiply by mom to get dS/dt.
 
-	X=zero;
-	DerivativeSolver(Mpc,PhiOdd,X);
-	Mpc.Mpc(X,Y);
-  Mpc.MpcDeriv(tmp , Y, X );    dSdU=tmp;
-  Mpc.MpcDagDeriv(tmp , X, Y);  dSdU=dSdU+tmp;
+    X=Zero();
+    DerivativeSolver(Mpc,PhiOdd,X);
+    Mpc.Mpc(X,Y);
+    Mpc.MpcDeriv(tmp , Y, X );    dSdU=tmp;
+    Mpc.MpcDagDeriv(tmp , X, Y);  dSdU=dSdU+tmp;
 
-	// Treat the EE case. (MdagM)^-1 = Minv Minvdag
-	// Deriv defaults to zero.
-	//        FermOp.MooeeInvDag(PhiOdd,Y);
-	//      FermOp.MooeeInv(Y,X);
-	//	FermOp.MeeDeriv(tmp , Y, X,DaggerNo );    dSdU=tmp;
-	//  FermOp.MeeDeriv(tmp , X, Y,DaggerYes);  dSdU=dSdU+tmp;
+    // Treat the EE case. (MdagM)^-1 = Minv Minvdag
+    // Deriv defaults to zero.
+    //        FermOp.MooeeInvDag(PhiOdd,Y);
+    //      FermOp.MooeeInv(Y,X);
+    //	FermOp.MeeDeriv(tmp , Y, X,DaggerNo );    dSdU=tmp;
+    //  FermOp.MeeDeriv(tmp , X, Y,DaggerYes);  dSdU=dSdU+tmp;
 
-	assert(FermOp.ConstEE() == 1);
+    assert(FermOp.ConstEE() == 1);
 
-	/*
-	  FermOp.MooeeInvDag(PhiOdd,Y);
-	  FermOp.MooeeInv(Y,X);
-	  FermOp.MeeDeriv(tmp , Y, X,DaggerNo );    dSdU=tmp;
-	  FermOp.MeeDeriv(tmp , X, Y,DaggerYes);  dSdU=dSdU+tmp;
-	*/
+    /*
+      FermOp.MooeeInvDag(PhiOdd,Y);
+      FermOp.MooeeInv(Y,X);
+      FermOp.MeeDeriv(tmp , Y, X,DaggerNo );    dSdU=tmp;
+      FermOp.MeeDeriv(tmp , X, Y,DaggerYes);  dSdU=dSdU+tmp;
+    */
 	
-	//dSdU = Ta(dSdU);
+    //dSdU = Ta(dSdU);
 
-      };
+  };
 
-    };
+};
     
-  }
-}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/action/pseudofermion/TwoFlavourEvenOddRatio.h b/Grid/qcd/action/pseudofermion/TwoFlavourEvenOddRatio.h
index e9a8853a..d1d6f336 100644
--- a/Grid/qcd/action/pseudofermion/TwoFlavourEvenOddRatio.h
+++ b/Grid/qcd/action/pseudofermion/TwoFlavourEvenOddRatio.h
@@ -29,8 +29,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #ifndef QCD_PSEUDOFERMION_TWO_FLAVOUR_EVEN_ODD_RATIO_H
 #define QCD_PSEUDOFERMION_TWO_FLAVOUR_EVEN_ODD_RATIO_H
 
-namespace Grid{
-  namespace QCD{
+NAMESPACE_BEGIN(Grid);
 
     ///////////////////////////////////////
     // Two flavour ratio
@@ -118,7 +117,7 @@ namespace Grid{
 
         // Odd det factors
         Mpc.MpcDag(etaOdd,PhiOdd);
-        tmp=zero;
+        tmp=Zero();
         HeatbathSolver(Vpc,PhiOdd,tmp);
         Vpc.Mpc(tmp,PhiOdd);            
 
@@ -146,7 +145,7 @@ namespace Grid{
         FermionField Y(NumOp.FermionRedBlackGrid());
 
         Vpc.MpcDag(PhiOdd,Y);           // Y= Vdag phi
-        X=zero;
+        X=Zero();
         ActionSolver(Mpc,Y,X);          // X= (MdagM)^-1 Vdag phi
         //Mpc.Mpc(X,Y);                   // Y=  Mdag^-1 Vdag phi
         // Multiply by Ydag
@@ -182,13 +181,13 @@ namespace Grid{
         FermionField  Y(NumOp.FermionRedBlackGrid());
 
         // This assignment is necessary to be compliant with the HMC grids
-	GaugeField force(dSdU._grid);
+	GaugeField force(dSdU.Grid());
 
         //Y=Vdag phi
         //X = (Mdag M)^-1 V^dag phi
         //Y = (Mdag)^-1 V^dag  phi
         Vpc.MpcDag(PhiOdd,Y);          // Y= Vdag phi
-        X=zero;
+        X=Zero();
         DerivativeSolver(Mpc,Y,X);     // X= (MdagM)^-1 Vdag phi
         Mpc.Mpc(X,Y);                  // Y=  Mdag^-1 Vdag phi
 
@@ -212,6 +211,5 @@ namespace Grid{
         
       };
     };
-  }
-}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/qcd/action/pseudofermion/TwoFlavourRatio.h b/Grid/qcd/action/pseudofermion/TwoFlavourRatio.h
index bcbf9364..4d72faba 100644
--- a/Grid/qcd/action/pseudofermion/TwoFlavourRatio.h
+++ b/Grid/qcd/action/pseudofermion/TwoFlavourRatio.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,149 +25,149 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef QCD_PSEUDOFERMION_TWO_FLAVOUR_RATIO_H
 #define QCD_PSEUDOFERMION_TWO_FLAVOUR_RATIO_H
 
-namespace Grid{
-  namespace QCD{
+NAMESPACE_BEGIN(Grid);
 
-    ///////////////////////////////////////
-    // Two flavour ratio
-    ///////////////////////////////////////
-    template<class Impl>
-    class TwoFlavourRatioPseudoFermionAction : public Action<typename Impl::GaugeField> {
-    public:
-      INHERIT_IMPL_TYPES(Impl);
+///////////////////////////////////////
+// Two flavour ratio
+///////////////////////////////////////
+template<class Impl>
+class TwoFlavourRatioPseudoFermionAction : public Action<typename Impl::GaugeField> {
+public:
+  INHERIT_IMPL_TYPES(Impl);
 
-    private:
-      FermionOperator<Impl> & NumOp;// the basic operator
-      FermionOperator<Impl> & DenOp;// the basic operator
+private:
+  FermionOperator<Impl> & NumOp;// the basic operator
+  FermionOperator<Impl> & DenOp;// the basic operator
 
-      OperatorFunction<FermionField> &DerivativeSolver;
-      OperatorFunction<FermionField> &ActionSolver;
+  OperatorFunction<FermionField> &DerivativeSolver;
+  OperatorFunction<FermionField> &ActionSolver;
 
-      FermionField Phi; // the pseudo fermion field for this trajectory
+  FermionField Phi; // the pseudo fermion field for this trajectory
 
-    public:
-      TwoFlavourRatioPseudoFermionAction(FermionOperator<Impl>  &_NumOp, 
-					 FermionOperator<Impl>  &_DenOp, 
-					 OperatorFunction<FermionField> & DS,
-					 OperatorFunction<FermionField> & AS
-					 ) : NumOp(_NumOp), DenOp(_DenOp), DerivativeSolver(DS), ActionSolver(AS), Phi(_NumOp.FermionGrid()) {};
+public:
+  TwoFlavourRatioPseudoFermionAction(FermionOperator<Impl>  &_NumOp, 
+				     FermionOperator<Impl>  &_DenOp, 
+				     OperatorFunction<FermionField> & DS,
+				     OperatorFunction<FermionField> & AS
+				     ) : NumOp(_NumOp), DenOp(_DenOp), DerivativeSolver(DS), ActionSolver(AS), Phi(_NumOp.FermionGrid()) {};
       
-      virtual std::string action_name(){return "TwoFlavourRatioPseudoFermionAction";}
+  virtual std::string action_name(){return "TwoFlavourRatioPseudoFermionAction";}
 
-      virtual std::string LogParameters(){
-	std::stringstream sstream;
-	sstream << GridLogMessage << "["<<action_name()<<"] has no parameters" << std::endl;
-	return sstream.str();
-      }  
+  virtual std::string LogParameters(){
+    std::stringstream sstream;
+    sstream << GridLogMessage << "["<<action_name()<<"] has no parameters" << std::endl;
+    return sstream.str();
+  }  
       
-      virtual void refresh(const GaugeField &U, GridParallelRNG& pRNG) {
+  virtual void refresh(const GaugeField &U, GridParallelRNG& pRNG) {
 
-	// P(phi) = e^{- phi^dag V (MdagM)^-1 Vdag phi}
-	//
-	// NumOp == V
-	// DenOp == M
-	//
-	// Take phi = Vdag^{-1} Mdag eta  ; eta = Mdag^{-1} Vdag Phi
-	//
-	// P(eta) = e^{- eta^dag eta}
-	//
-	// e^{x^2/2 sig^2} => sig^2 = 0.5.
-	// 
-	// So eta should be of width sig = 1/sqrt(2) and must multiply by 0.707....
-	//
-	RealD scale = std::sqrt(0.5);
+    // P(phi) = e^{- phi^dag V (MdagM)^-1 Vdag phi}
+    //
+    // NumOp == V
+    // DenOp == M
+    //
+    // Take phi = Vdag^{-1} Mdag eta  ; eta = Mdag^{-1} Vdag Phi
+    //
+    // P(eta) = e^{- eta^dag eta}
+    //
+    // e^{x^2/2 sig^2} => sig^2 = 0.5.
+    // 
+    // So eta should be of width sig = 1/sqrt(2) and must multiply by 0.707....
+    //
+    RealD scale = std::sqrt(0.5);
 
-	FermionField eta(NumOp.FermionGrid());
-	FermionField tmp(NumOp.FermionGrid());
+    FermionField eta(NumOp.FermionGrid());
+    FermionField tmp(NumOp.FermionGrid());
 
-	gaussian(pRNG,eta);
+    gaussian(pRNG,eta);
 
-	NumOp.ImportGauge(U);
-	DenOp.ImportGauge(U);
+    NumOp.ImportGauge(U);
+    DenOp.ImportGauge(U);
 
-	// Note: this hard codes normal equations type solvers; alternate implementation needed for 
-	// non-herm style solvers.
-	MdagMLinearOperator<FermionOperator<Impl> ,FermionField> MdagMOp(NumOp);
+    // Note: this hard codes normal equations type solvers; alternate implementation needed for 
+    // non-herm style solvers.
+    MdagMLinearOperator<FermionOperator<Impl> ,FermionField> MdagMOp(NumOp);
 
-	DenOp.Mdag(eta,Phi);            // Mdag eta
-	tmp = zero;
-	ActionSolver(MdagMOp,Phi,tmp);  // (VdagV)^-1 Mdag eta = V^-1 Vdag^-1 Mdag eta
-	NumOp.M(tmp,Phi);               // Vdag^-1 Mdag eta
+    DenOp.Mdag(eta,Phi);            // Mdag eta
+    tmp = Zero();
+    ActionSolver(MdagMOp,Phi,tmp);  // (VdagV)^-1 Mdag eta = V^-1 Vdag^-1 Mdag eta
+    NumOp.M(tmp,Phi);               // Vdag^-1 Mdag eta
 
-	Phi=Phi*scale;
+    Phi=Phi*scale;
 	
-      };
+  };
 
-      //////////////////////////////////////////////////////
-      // S = phi^dag V (Mdag M)^-1 Vdag phi
-      //////////////////////////////////////////////////////
-      virtual RealD S(const GaugeField &U) {
+  //////////////////////////////////////////////////////
+  // S = phi^dag V (Mdag M)^-1 Vdag phi
+  //////////////////////////////////////////////////////
+  virtual RealD S(const GaugeField &U) {
 
-	NumOp.ImportGauge(U);
-	DenOp.ImportGauge(U);
+    NumOp.ImportGauge(U);
+    DenOp.ImportGauge(U);
 
-	FermionField X(NumOp.FermionGrid());
-	FermionField Y(NumOp.FermionGrid());
+    FermionField X(NumOp.FermionGrid());
+    FermionField Y(NumOp.FermionGrid());
 	
-	MdagMLinearOperator<FermionOperator<Impl> ,FermionField> MdagMOp(DenOp);
+    MdagMLinearOperator<FermionOperator<Impl> ,FermionField> MdagMOp(DenOp);
 
-	NumOp.Mdag(Phi,Y);              // Y= Vdag phi
-	X=zero;
-	ActionSolver(MdagMOp,Y,X);      // X= (MdagM)^-1 Vdag phi
-	DenOp.M(X,Y);                  // Y=  Mdag^-1 Vdag phi
+    NumOp.Mdag(Phi,Y);              // Y= Vdag phi
+    X=Zero();
+    ActionSolver(MdagMOp,Y,X);      // X= (MdagM)^-1 Vdag phi
+    DenOp.M(X,Y);                  // Y=  Mdag^-1 Vdag phi
 
-	RealD action = norm2(Y);
+    RealD action = norm2(Y);
 
-	return action;
-      };
+    return action;
+  };
 
-      //////////////////////////////////////////////////////
-      // dS/du = phi^dag dV (Mdag M)^-1 V^dag  phi
-      //       - phi^dag V (Mdag M)^-1 [ Mdag dM + dMdag M ]  (Mdag M)^-1 V^dag  phi
-      //       + phi^dag V (Mdag M)^-1 dV^dag  phi
-      //////////////////////////////////////////////////////
-      virtual void deriv(const GaugeField &U,GaugeField & dSdU) {
+  //////////////////////////////////////////////////////
+  // dS/du = phi^dag dV (Mdag M)^-1 V^dag  phi
+  //       - phi^dag V (Mdag M)^-1 [ Mdag dM + dMdag M ]  (Mdag M)^-1 V^dag  phi
+  //       + phi^dag V (Mdag M)^-1 dV^dag  phi
+  //////////////////////////////////////////////////////
+  virtual void deriv(const GaugeField &U,GaugeField & dSdU) {
 
-	NumOp.ImportGauge(U);
-	DenOp.ImportGauge(U);
+    NumOp.ImportGauge(U);
+    DenOp.ImportGauge(U);
 
-	MdagMLinearOperator<FermionOperator<Impl> ,FermionField> MdagMOp(DenOp);
+    MdagMLinearOperator<FermionOperator<Impl> ,FermionField> MdagMOp(DenOp);
 
-	FermionField  X(NumOp.FermionGrid());
-	FermionField  Y(NumOp.FermionGrid());
+    FermionField  X(NumOp.FermionGrid());
+    FermionField  Y(NumOp.FermionGrid());
 
-	GaugeField   force(NumOp.GaugeGrid());	
+    GaugeField   force(NumOp.GaugeGrid());	
 
 
-	//Y=Vdag phi
-	//X = (Mdag M)^-1 V^dag phi
-	//Y = (Mdag)^-1 V^dag  phi
-	NumOp.Mdag(Phi,Y);              // Y= Vdag phi
-	X=zero;
-	DerivativeSolver(MdagMOp,Y,X);      // X= (MdagM)^-1 Vdag phi
-	DenOp.M(X,Y);                  // Y=  Mdag^-1 Vdag phi
+    //Y=Vdag phi
+    //X = (Mdag M)^-1 V^dag phi
+    //Y = (Mdag)^-1 V^dag  phi
+    NumOp.Mdag(Phi,Y);              // Y= Vdag phi
+    X=Zero();
+    DerivativeSolver(MdagMOp,Y,X);      // X= (MdagM)^-1 Vdag phi
+    DenOp.M(X,Y);                  // Y=  Mdag^-1 Vdag phi
 
-	// phi^dag V (Mdag M)^-1 dV^dag  phi
-	NumOp.MDeriv(force , X, Phi, DaggerYes );  dSdU=force;
+    // phi^dag V (Mdag M)^-1 dV^dag  phi
+    NumOp.MDeriv(force , X, Phi, DaggerYes );  dSdU=force;
   
-	// phi^dag dV (Mdag M)^-1 V^dag  phi
-	NumOp.MDeriv(force , Phi, X ,DaggerNo  );  dSdU=dSdU+force;
+    // phi^dag dV (Mdag M)^-1 V^dag  phi
+    NumOp.MDeriv(force , Phi, X ,DaggerNo  );  dSdU=dSdU+force;
 
-	//    -    phi^dag V (Mdag M)^-1 Mdag dM   (Mdag M)^-1 V^dag  phi
-	//    -    phi^dag V (Mdag M)^-1 dMdag M   (Mdag M)^-1 V^dag  phi
-	DenOp.MDeriv(force,Y,X,DaggerNo);   dSdU=dSdU-force;
-	DenOp.MDeriv(force,X,Y,DaggerYes);  dSdU=dSdU-force;
+    //    -    phi^dag V (Mdag M)^-1 Mdag dM   (Mdag M)^-1 V^dag  phi
+    //    -    phi^dag V (Mdag M)^-1 dMdag M   (Mdag M)^-1 V^dag  phi
+    DenOp.MDeriv(force,Y,X,DaggerNo);   dSdU=dSdU-force;
+    DenOp.MDeriv(force,X,Y,DaggerYes);  dSdU=dSdU-force;
 
-	dSdU *= -1.0;
-	//dSdU = - Ta(dSdU);
+    dSdU *= -1.0;
+    //dSdU = - Ta(dSdU);
+
+  };
+};
+
+NAMESPACE_END(Grid);
 
-      };
-    };
-  }
-}
 #endif
diff --git a/Grid/qcd/action/scalar/Scalar.h b/Grid/qcd/action/scalar/Scalar.h
index 485a6765..44f7c450 100644
--- a/Grid/qcd/action/scalar/Scalar.h
+++ b/Grid/qcd/action/scalar/Scalar.h
@@ -25,26 +25,24 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
-#ifndef GRID_QCD_SCALAR_H
-#define GRID_QCD_SCALAR_H
+			   /*  END LEGAL */
+
+#pragma once
 
 #include <Grid/qcd/action/scalar/ScalarImpl.h>
 #include <Grid/qcd/action/scalar/ScalarAction.h>
 #include <Grid/qcd/action/scalar/ScalarInteractionAction.h>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
-  typedef ScalarAction<ScalarImplR>                 ScalarActionR;
-  typedef ScalarAction<ScalarImplF>                 ScalarActionF;
-  typedef ScalarAction<ScalarImplD>                 ScalarActionD;
+typedef ScalarAction<ScalarImplR>                 ScalarActionR;
+typedef ScalarAction<ScalarImplF>                 ScalarActionF;
+typedef ScalarAction<ScalarImplD>                 ScalarActionD;
 
-  template <int Colours, int Dimensions> using ScalarAdjActionR = ScalarInteractionAction<ScalarNxNAdjImplR<Colours>, Dimensions>;
-  template <int Colours, int Dimensions> using ScalarAdjActionF = ScalarInteractionAction<ScalarNxNAdjImplF<Colours>, Dimensions>;
-  template <int Colours, int Dimensions> using ScalarAdjActionD = ScalarInteractionAction<ScalarNxNAdjImplD<Colours>, Dimensions>;
+template <int Colours, int Dimensions> using ScalarAdjActionR = ScalarInteractionAction<ScalarNxNAdjImplR<Colours>, Dimensions>;
+template <int Colours, int Dimensions> using ScalarAdjActionF = ScalarInteractionAction<ScalarNxNAdjImplF<Colours>, Dimensions>;
+template <int Colours, int Dimensions> using ScalarAdjActionD = ScalarInteractionAction<ScalarNxNAdjImplD<Colours>, Dimensions>;
   
-}
-}
+NAMESPACE_END(Grid);
+
 
-#endif  // GRID_QCD_SCALAR_H
diff --git a/Grid/qcd/action/scalar/ScalarAction.h b/Grid/qcd/action/scalar/ScalarAction.h
index 2c82d2e3..34fc4fac 100644
--- a/Grid/qcd/action/scalar/ScalarAction.h
+++ b/Grid/qcd/action/scalar/ScalarAction.h
@@ -27,57 +27,54 @@
 
   See the full license in the file "LICENSE" in the top level distribution
 directory
-  *************************************************************************************/
+*************************************************************************************/
 /*  END LEGAL */
 
 #ifndef SCALAR_ACTION_H
 #define SCALAR_ACTION_H
 
-namespace Grid {
-  // FIXME drop the QCD namespace everywhere here
+NAMESPACE_BEGIN(Grid);
 
 template <class Impl>
-class ScalarAction : public QCD::Action<typename Impl::Field> {
- public:
-    INHERIT_FIELD_TYPES(Impl);
+class ScalarAction : public Action<typename Impl::Field> {
+public:
+  INHERIT_FIELD_TYPES(Impl);
 
- private:
-    RealD mass_square;
-    RealD lambda;
+private:
+  RealD mass_square;
+  RealD lambda;
 
- public:
-    ScalarAction(RealD ms, RealD l) : mass_square(ms), lambda(l) {}
+public:
+  ScalarAction(RealD ms, RealD l) : mass_square(ms), lambda(l) {}
 
-    virtual std::string LogParameters() {
-      std::stringstream sstream;
-      sstream << GridLogMessage << "[ScalarAction] lambda      : " << lambda      << std::endl;
-      sstream << GridLogMessage << "[ScalarAction] mass_square : " << mass_square << std::endl;
-      return sstream.str();
-    }
-    virtual std::string action_name() {return "ScalarAction";}
+  virtual std::string LogParameters() {
+    std::stringstream sstream;
+    sstream << GridLogMessage << "[ScalarAction] lambda      : " << lambda      << std::endl;
+    sstream << GridLogMessage << "[ScalarAction] mass_square : " << mass_square << std::endl;
+    return sstream.str();
+  }
+  virtual std::string action_name() {return "ScalarAction";}
 
-    virtual void refresh(const Field &U, GridParallelRNG &pRNG) {}  // noop as no pseudoferms
+  virtual void refresh(const Field &U, GridParallelRNG &pRNG) {}  // noop as no pseudoferms
 
-    virtual RealD S(const Field &p) {
-      return (mass_square * 0.5 + QCD::Nd) * ScalarObs<Impl>::sumphisquared(p) +
-    (lambda / 24.) * ScalarObs<Impl>::sumphifourth(p) +
-    ScalarObs<Impl>::sumphider(p);
-    };
+  virtual RealD S(const Field &p) {
+    return (mass_square * 0.5 + Nd) * ScalarObs<Impl>::sumphisquared(p) +
+      (lambda / 24.) * ScalarObs<Impl>::sumphifourth(p) +
+      ScalarObs<Impl>::sumphider(p);
+  };
 
-    virtual void deriv(const Field &p,
-                       Field &force) {
-      Field tmp(p._grid);
-      Field p2(p._grid);
-      ScalarObs<Impl>::phisquared(p2, p);
-      tmp = -(Cshift(p, 0, -1) + Cshift(p, 0, 1));
-      for (int mu = 1; mu < QCD::Nd; mu++) tmp -= Cshift(p, mu, -1) + Cshift(p, mu, 1);
+  virtual void deriv(const Field &p,
+		     Field &force) {
+    Field tmp(p.Grid());
+    Field p2(p.Grid());
+    ScalarObs<Impl>::phisquared(p2, p);
+    tmp = -(Cshift(p, 0, -1) + Cshift(p, 0, 1));
+    for (int mu = 1; mu < Nd; mu++) tmp -= Cshift(p, mu, -1) + Cshift(p, mu, 1);
 
-      force =+(mass_square + 2. * QCD::Nd) * p + (lambda / 6.) * p2 * p + tmp;
-    }
+    force =+(mass_square + 2. * Nd) * p + (lambda / 6.) * p2 * p + tmp;
+  }
 };
 
-
-
-}  // namespace Grid
+NAMESPACE_END(Grid);
 
 #endif // SCALAR_ACTION_H
diff --git a/Grid/qcd/action/scalar/ScalarImpl.h b/Grid/qcd/action/scalar/ScalarImpl.h
index 55f5049d..febb315e 100644
--- a/Grid/qcd/action/scalar/ScalarImpl.h
+++ b/Grid/qcd/action/scalar/ScalarImpl.h
@@ -1,93 +1,90 @@
-#ifndef SCALAR_IMPL
-#define SCALAR_IMPL
+#pragma once
 
-
-namespace Grid {
-  //namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 template <class S>
 class ScalarImplTypes {
- public:
-    typedef S Simd;
+public:
+  typedef S Simd;
 
-    template <typename vtype>
-    using iImplField = iScalar<iScalar<iScalar<vtype> > >;
+  template <typename vtype>
+  using iImplField = iScalar<iScalar<iScalar<vtype> > >;
 
-    typedef iImplField<Simd> SiteField;
-    typedef SiteField        SitePropagator;
-    typedef SiteField        SiteComplex;
+  typedef iImplField<Simd> SiteField;
+  typedef SiteField        SitePropagator;
+  typedef SiteField        SiteComplex;
+    
+  typedef Lattice<SiteField> Field;
+  typedef Field              ComplexField;
+  typedef Field              FermionField;
+  typedef Field              PropagatorField;
+    
+  static inline void generate_momenta(Field& P, GridParallelRNG& pRNG){
+    gaussian(pRNG, P);
+  }
 
-    typedef Lattice<SiteField> Field;
-    typedef Field              ComplexField;
-    typedef Field              FermionField;
-    typedef Field              PropagatorField;
+  static inline Field projectForce(Field& P){return P;}
 
-    static inline void generate_momenta(Field& P, GridParallelRNG& pRNG){
-      gaussian(pRNG, P);
-    }
+  static inline void update_field(Field& P, Field& U, double ep) {
+    U += P*ep;
+  }
 
-    static inline Field projectForce(Field& P){return P;}
+  static inline RealD FieldSquareNorm(Field& U) {
+    return (- sum(trace(U*U))/2.0);
+  }
 
-    static inline void update_field(Field& P, Field& U, double ep) {
-      U += P*ep;
-    }
+  static inline void HotConfiguration(GridParallelRNG &pRNG, Field &U) {
+    gaussian(pRNG, U);
+  }
 
-    static inline RealD FieldSquareNorm(Field& U) {
-      return (- sum(trace(U*U))/2.0);
-    }
+  static inline void TepidConfiguration(GridParallelRNG &pRNG, Field &U) {
+    gaussian(pRNG, U);
+  }
 
-    static inline void HotConfiguration(GridParallelRNG &pRNG, Field &U) {
-      gaussian(pRNG, U);
-    }
-
-    static inline void TepidConfiguration(GridParallelRNG &pRNG, Field &U) {
-      gaussian(pRNG, U);
-    }
-
-    static inline void ColdConfiguration(GridParallelRNG &pRNG, Field &U) {
-      U = 1.0;
-    }
-
-    static void MomentumSpacePropagator(Field &out, RealD m)
-    {
-      GridBase           *grid = out._grid;
-      Field              kmu(grid), one(grid);
-      const unsigned int nd    = grid->_ndimension;
-      std::vector<int>   &l    = grid->_fdimensions;
-
-      one = Complex(1.0,0.0);
-      out = m*m;
-      for(int mu = 0; mu < nd; mu++)
+  static inline void ColdConfiguration(GridParallelRNG &pRNG, Field &U) {
+    U = 1.0;
+  }
+    
+  static void MomentumSpacePropagator(Field &out, RealD m)
+  {
+    GridBase           *grid = out.Grid();
+    Field              kmu(grid), one(grid);
+    const unsigned int nd    = grid->_ndimension;
+    Coordinate         &l    = grid->_fdimensions;
+      
+    one = Complex(1.0,0.0);
+    out = m*m;
+    for(int mu = 0; mu < nd; mu++)
       {
         Real twoPiL = M_PI*2./l[mu];
-
+        
         LatticeCoordinate(kmu,mu);
         kmu = 2.*sin(.5*twoPiL*kmu);
         out = out + kmu*kmu;
       }
-      out = one/out;
-    }
-
-    static void FreePropagator(const Field &in, Field &out,
-                               const Field &momKernel)
-    {
-      FFT   fft((GridCartesian *)in._grid);
-      Field inFT(in._grid);
-
-      fft.FFT_all_dim(inFT, in, FFT::forward);
-      inFT = inFT*momKernel;
-      fft.FFT_all_dim(out, inFT, FFT::backward);
-    }
-
-    static void FreePropagator(const Field &in, Field &out, RealD m)
-    {
-      Field momKernel(in._grid);
-
-      MomentumSpacePropagator(momKernel, m);
-      FreePropagator(in, out, momKernel);
-    }
-
-  };
+    out = one/out;
+  }
+    
+  static void FreePropagator(const Field &in, Field &out,
+			     const Field &momKernel)
+  {
+    FFT   fft((GridCartesian *)in.Grid());
+    Field inFT(in.Grid());
+      
+    fft.FFT_all_dim(inFT, in, FFT::forward);
+    inFT = inFT*momKernel;
+    fft.FFT_all_dim(out, inFT, FFT::backward);
+  }
+    
+  static void FreePropagator(const Field &in, Field &out, RealD m)
+  {
+    Field momKernel(in.Grid());
+      
+    MomentumSpacePropagator(momKernel, m);
+    FreePropagator(in, out, momKernel);
+  }
+    
+};
 
   #ifdef  USE_FFT_ACCELERATION
   #ifndef FFT_MASS
@@ -95,30 +92,30 @@ class ScalarImplTypes {
   #endif
   #endif
   
-  template <class S, unsigned int N>
-  class ScalarAdjMatrixImplTypes {
-  public:
-    typedef S Simd;
-    typedef QCD::SU<N> Group;
+template <class S, unsigned int N>
+class ScalarAdjMatrixImplTypes {
+public:
+  typedef S Simd;
+  typedef SU<N> Group;
+    
+  template <typename vtype>
+  using iImplField   = iScalar<iScalar<iMatrix<vtype, N>>>;
+  template <typename vtype>
+  using iImplComplex = iScalar<iScalar<iScalar<vtype>>>;
 
-    template <typename vtype>
-    using iImplField   = iScalar<iScalar<iMatrix<vtype, N>>>;
-    template <typename vtype>
-    using iImplComplex = iScalar<iScalar<iScalar<vtype>>>;
-
-    typedef iImplField<Simd>   SiteField;
-    typedef SiteField          SitePropagator;
-    typedef iImplComplex<Simd> SiteComplex;
-
-    typedef Lattice<SiteField>   Field;
-    typedef Lattice<SiteComplex> ComplexField;
-    typedef Field                FermionField;
-    typedef Field                PropagatorField;
+  typedef iImplField<Simd>   SiteField;
+  typedef SiteField          SitePropagator;
+  typedef iImplComplex<Simd> SiteComplex;
+    
+  typedef Lattice<SiteField>   Field;
+  typedef Lattice<SiteComplex> ComplexField;
+  typedef Field                FermionField;
+  typedef Field                PropagatorField;
 
     static void MomentaSquare(ComplexField &out)
     {
-      GridBase *grid = out._grid;
-      const std::vector<int> &l = grid->FullDimensions();
+      GridBase *grid = out.Grid();
+      const Coordinate &l = grid->FullDimensions();
       ComplexField kmu(grid);
 
       for (int mu = 0; mu < grid->Nd(); mu++)
@@ -132,7 +129,7 @@ class ScalarImplTypes {
 
     static void MomentumSpacePropagator(ComplexField &out, RealD m)
     {
-      GridBase *grid = out._grid;
+      GridBase *grid = out.Grid();
       ComplexField one(grid);
       one = Complex(1.0, 0.0);
       out = m * m;
@@ -143,16 +140,16 @@ class ScalarImplTypes {
     static inline void generate_momenta(Field &P, GridParallelRNG &pRNG)
     {
 #ifndef USE_FFT_ACCELERATION
-      Group::GaussianFundamentalLieAlgebraMatrix(pRNG, P);
+    Group::GaussianFundamentalLieAlgebraMatrix(pRNG, P);
 #else
 
-      Field Pgaussian(P._grid), Pp(P._grid);
-      ComplexField p2(P._grid); p2 = zero;
+      Field Pgaussian(P.Grid()), Pp(P.Grid());
+      ComplexField p2(P.Grid()); p2 = zero;
       RealD M = FFT_MASS;
       
       Group::GaussianFundamentalLieAlgebraMatrix(pRNG, Pgaussian);
 
-      FFT theFFT((GridCartesian*)P._grid);
+      FFT theFFT((GridCartesian*)P.Grid());
       theFFT.FFT_all_dim(Pp, Pgaussian, FFT::forward);
       MomentaSquare(p2);
       p2 += M * M;
@@ -161,15 +158,15 @@ class ScalarImplTypes {
       theFFT.FFT_all_dim(P, Pp, FFT::backward);
 
 #endif //USE_FFT_ACCELERATION
-    }
+  }
 
-    static inline Field projectForce(Field& P) {return P;}
+  static inline Field projectForce(Field& P) {return P;}
 
     static inline void update_field(Field &P, Field &U, double ep)
     {
 #ifndef USE_FFT_ACCELERATION
       double t0=usecond(); 
-      U += P * ep;
+    U += P*ep;
       double t1=usecond();
       double total_time = (t1-t0)/1e6;
       std::cout << GridLogIntegrator << "Total time for updating field (s)       : " << total_time << std::endl; 
@@ -180,11 +177,11 @@ class ScalarImplTypes {
       // Transform back -> P'(x)
       // U += P'(x)*ep
 
-      Field Pp(U._grid), P_FFT(U._grid);     
-      static ComplexField p2(U._grid);
+      Field Pp(U.Grid()), P_FFT(U.Grid());     
+      static ComplexField p2(U.Grid());
       RealD M = FFT_MASS;
       
-      FFT theFFT((GridCartesian*)U._grid);
+      FFT theFFT((GridCartesian*)U.Grid());
       theFFT.FFT_all_dim(Pp, P, FFT::forward);
 
       static bool first_call = true;
@@ -193,7 +190,7 @@ class ScalarImplTypes {
         // avoid recomputing
         MomentumSpacePropagator(p2, M);
         first_call = false;
-      }
+  }
       Pp *= p2;
       theFFT.FFT_all_dim(P_FFT, Pp, FFT::backward);
       U += P_FFT * ep;
@@ -204,60 +201,55 @@ class ScalarImplTypes {
     static inline RealD FieldSquareNorm(Field &U)
     {
 #ifndef USE_FFT_ACCELERATION
-      return (TensorRemove(sum(trace(U * U))).real());
+    return (TensorRemove(sum(trace(U*U))).real());
 #else
       // In case of Fourier acceleration we have to:
       // compute U(p)*U(p)/(M^2+p^2))   Parseval theorem
       // 1 FFT needed U(x) -> U(p)
       // M to be passed
 
-      FFT theFFT((GridCartesian*)U._grid);
-      Field Up(U._grid);
+      FFT theFFT((GridCartesian*)U.Grid());
+      Field Up(U.Grid());
 
       theFFT.FFT_all_dim(Up, U, FFT::forward);
       RealD M = FFT_MASS;
-      ComplexField p2(U._grid);
+      ComplexField p2(U.Grid());
       MomentumSpacePropagator(p2, M);
       Field Up2 = Up * p2;
       // from the definition of the DFT we need to divide by the volume
-      return (-TensorRemove(sum(trace(adj(Up) * Up2))).real() / U._grid->gSites());
+      return (-TensorRemove(sum(trace(adj(Up) * Up2))).real() / U.Grid()->gSites());
 #endif //USE_FFT_ACCELERATION
-    }
+  }
 
-    static inline void HotConfiguration(GridParallelRNG &pRNG, Field &U) {
-      Group::GaussianFundamentalLieAlgebraMatrix(pRNG, U);
-    }
+  static inline void HotConfiguration(GridParallelRNG &pRNG, Field &U) {
+    Group::GaussianFundamentalLieAlgebraMatrix(pRNG, U);
+  }
 
-    static inline void TepidConfiguration(GridParallelRNG &pRNG, Field &U) {
-      Group::GaussianFundamentalLieAlgebraMatrix(pRNG, U, 0.01);
-    }
+  static inline void TepidConfiguration(GridParallelRNG &pRNG, Field &U) {
+    Group::GaussianFundamentalLieAlgebraMatrix(pRNG, U, 0.01);
+  }
 
-    static inline void ColdConfiguration(GridParallelRNG &pRNG, Field &U) {
-      U = zero;
-    }
+  static inline void ColdConfiguration(GridParallelRNG &pRNG, Field &U) {
+    U = Zero();
+  }
 
-  };
+};
 
+typedef ScalarImplTypes<vReal> ScalarImplR;
+typedef ScalarImplTypes<vRealF> ScalarImplF;
+typedef ScalarImplTypes<vRealD> ScalarImplD;
+typedef ScalarImplTypes<vComplex> ScalarImplCR;
+typedef ScalarImplTypes<vComplexF> ScalarImplCF;
+typedef ScalarImplTypes<vComplexD> ScalarImplCD;
+    
+// Hardcoding here the size of the matrices
+typedef ScalarAdjMatrixImplTypes<vComplex,  Nc> ScalarAdjImplR;
+typedef ScalarAdjMatrixImplTypes<vComplexF, Nc> ScalarAdjImplF;
+typedef ScalarAdjMatrixImplTypes<vComplexD, Nc> ScalarAdjImplD;
 
+template <int Colours > using ScalarNxNAdjImplR = ScalarAdjMatrixImplTypes<vComplex,   Colours >;
+template <int Colours > using ScalarNxNAdjImplF = ScalarAdjMatrixImplTypes<vComplexF,  Colours >;
+template <int Colours > using ScalarNxNAdjImplD = ScalarAdjMatrixImplTypes<vComplexD,  Colours >;
 
+NAMESPACE_END(Grid);
 
-  typedef ScalarImplTypes<vReal> ScalarImplR;
-  typedef ScalarImplTypes<vRealF> ScalarImplF;
-  typedef ScalarImplTypes<vRealD> ScalarImplD;
-  typedef ScalarImplTypes<vComplex> ScalarImplCR;
-  typedef ScalarImplTypes<vComplexF> ScalarImplCF;
-  typedef ScalarImplTypes<vComplexD> ScalarImplCD;
-
-  // Hardcoding here the size of the matrices
-  typedef ScalarAdjMatrixImplTypes<vComplex,  QCD::Nc> ScalarAdjImplR;
-  typedef ScalarAdjMatrixImplTypes<vComplexF, QCD::Nc> ScalarAdjImplF;
-  typedef ScalarAdjMatrixImplTypes<vComplexD, QCD::Nc> ScalarAdjImplD;
-
-  template <int Colours > using ScalarNxNAdjImplR = ScalarAdjMatrixImplTypes<vComplex,   Colours >;
-  template <int Colours > using ScalarNxNAdjImplF = ScalarAdjMatrixImplTypes<vComplexF,  Colours >;
-  template <int Colours > using ScalarNxNAdjImplD = ScalarAdjMatrixImplTypes<vComplexD,  Colours >;
-
-  //}
-}
-
-#endif
diff --git a/Grid/qcd/action/scalar/ScalarInteractionAction.h b/Grid/qcd/action/scalar/ScalarInteractionAction.h
index 8738b647..3be84480 100644
--- a/Grid/qcd/action/scalar/ScalarInteractionAction.h
+++ b/Grid/qcd/action/scalar/ScalarInteractionAction.h
@@ -27,18 +27,15 @@ directory
   *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef SCALAR_INT_ACTION_H
-#define SCALAR_INT_ACTION_H
+#pragma once
 
 // Note: this action can completely absorb the ScalarAction for real float fields
 // use the scalarObjs to generalise the structure
 
-namespace Grid
-{
-// FIXME drop the QCD namespace everywhere here
+NAMESPACE_BEGIN(Grid);
 
 template <class Impl, int Ndim>
-class ScalarInteractionAction : public QCD::Action<typename Impl::Field>
+class ScalarInteractionAction : public Action<typename Impl::Field>
 {
 public:
   INHERIT_FIELD_TYPES(Impl);
@@ -50,7 +47,7 @@ private:
   const unsigned int N = Impl::Group::Dimension;
 
   typedef typename Field::vector_object vobj;
-  typedef CartesianStencil<vobj, vobj> Stencil;
+  typedef CartesianStencil<vobj, vobj,int> Stencil;
 
   SimpleCompressor<vobj> compressor;
   int npoint = 2 * Ndim;
@@ -84,16 +81,20 @@ public:
 
   virtual RealD S(const Field &p)
   {
-    assert(p._grid->Nd() == Ndim);
-    static Stencil phiStencil(p._grid, npoint, 0, directions, displacements);
+    assert(p.Grid()->Nd() == Ndim);
+    static Stencil phiStencil(p.Grid(), npoint, 0, directions, displacements,0);
     phiStencil.HaloExchange(p, compressor);
-    Field action(p._grid), pshift(p._grid), phisquared(p._grid);
+    Field action(p.Grid()), pshift(p.Grid()), phisquared(p.Grid());
     phisquared = p * p;
     action = (2.0 * Ndim + mass_square) * phisquared - lambda * phisquared * phisquared;
+    
+    
+    auto p_v = p.View();
+    auto action_v = action.View();
     for (int mu = 0; mu < Ndim; mu++)
     {
       //  pshift = Cshift(p, mu, +1);  // not efficient, implement with stencils
-      parallel_for(int i = 0; i < p._grid->oSites(); i++)
+      thread_for(i, p.Grid()->oSites(),
       {
         int permute_type;
         StencilEntry *SE;
@@ -101,25 +102,22 @@ public:
         const vobj *temp, *t_p;
 
         SE = phiStencil.GetEntry(permute_type, mu, i);
-        t_p = &p._odata[i];
+        t_p = &p_v[i];
         if (SE->_is_local)
         {
-          temp = &p._odata[SE->_offset];
-          if (SE->_permute)
-          {
+          temp = &p_v[SE->_offset];
+          if (SE->_permute) {
             permute(temp2, *temp, permute_type);
-            action._odata[i] -= temp2 * (*t_p) + (*t_p) * temp2;
-          }
-          else
-          {
-            action._odata[i] -= (*temp) * (*t_p) + (*t_p) * (*temp);
+            action_v[i] -= temp2 * (*t_p) + (*t_p) * temp2;
+          } else {
+            action_v[i] -= (*temp) * (*t_p) + (*t_p) * (*temp);
           }
         }
         else
         {
-          action._odata[i] -= phiStencil.CommBuf()[SE->_offset] * (*t_p) + (*t_p) * phiStencil.CommBuf()[SE->_offset];
+          action_v[i] -= phiStencil.CommBuf()[SE->_offset] * (*t_p) + (*t_p) * phiStencil.CommBuf()[SE->_offset];
         }
-      }
+      });
       //  action -= pshift*p + p*pshift;
     }
     // NB the trace in the algebra is normalised to 1/2
@@ -130,17 +128,17 @@ public:
   virtual void deriv(const Field &p, Field &force)
   {
     double t0 = usecond();
-    assert(p._grid->Nd() == Ndim);
+    assert(p.Grid()->Nd() == Ndim);
     force = (2. * Ndim + mass_square) * p - 2. * lambda * p * p * p;
     double interm_t = usecond();
 
     // move this outside
-    static Stencil phiStencil(p._grid, npoint, 0, directions, displacements);
+    static Stencil phiStencil(p.Grid(), npoint, 0, directions, displacements,0);
 
     phiStencil.HaloExchange(p, compressor);
     double halo_t = usecond();
-    int chunk = 128;
-    //for (int mu = 0; mu < QCD::Nd; mu++) force -= Cshift(p, mu, -1) + Cshift(p, mu, 1);
+    //    int chunk = 128;
+    //for (int mu = 0; mu < Nd; mu++) force -= Cshift(p, mu, -1) + Cshift(p, mu, 1);
 
     // inverting the order of the loops slows down the code(! g++ 7)
     // cannot try to reduce the number of  force writes by factor npoint...
@@ -148,61 +146,53 @@ public:
     for (int point = 0; point < npoint; point++)
     {
 
-#pragma omp parallel 
-{
-        int permute_type;
-        StencilEntry *SE;
-        const vobj *temp;
+      auto p_v = p.View();
+      auto force_v = force.View();
+            
+      int permute_type;
+      StencilEntry *SE;
+      const vobj *temp;
 
-#pragma omp for schedule(static, chunk)
-      for (int i = 0; i < p._grid->oSites(); i++)
+      thread_for(i, p.Grid()->oSites(),
       {
-        SE = phiStencil.GetEntry(permute_type, point, i);
-        // prefetch next p?
-
-        if (SE->_is_local)
-        {
-          temp = &p._odata[SE->_offset];
-      
-          if (SE->_permute)
-          {
+	SE = phiStencil.GetEntry(permute_type, point, i);
+	// prefetch next p?
+	  
+	if (SE->_is_local) {
+	  temp = &p_v[SE->_offset];
+	    
+          if (SE->_permute) {
             vobj temp2;
             permute(temp2, *temp, permute_type);
-            force._odata[i] -= temp2;
-          }
-          else
-          {
-            force._odata[i] -= *temp; // slow part. Dominated by this read/write (BW)
+            force_v[i] -= temp2;
+          } else {
+            force_v[i] -= *temp; // slow part. Dominated by this read/write (BW)
           }
+        } else {
+          force_v[i] -= phiStencil.CommBuf()[SE->_offset];
         }
-        else
-        {
-          force._odata[i] -= phiStencil.CommBuf()[SE->_offset];
-        }
-      }
-
+      });
     }
-  }
-  force *= N / g;
+    force *= N / g;
 
-  double t1 = usecond();
-  double total_time = (t1 - t0) / 1e6;
-  double interm_time = (interm_t - t0) / 1e6;
-  double halo_time = (halo_t - interm_t) / 1e6;
-  double stencil_time = (t1 - halo_t) / 1e6;
-  std::cout << GridLogIntegrator << "Total time for force computation (s)       : " << total_time << std::endl;
-  std::cout << GridLogIntegrator << "Intermediate time for force computation (s): " << interm_time << std::endl;
-  std::cout << GridLogIntegrator << "Halo time in force computation (s)         : " << halo_time << std::endl;
-  std::cout << GridLogIntegrator << "Stencil time in force computation (s)      : " << stencil_time << std::endl;
-  double flops = p._grid->gSites() * (14 * N * N * N + 18 * N * N + 2);
-  double flops_no_stencil = p._grid->gSites() * (14 * N * N * N + 6 * N * N + 2);
-  double Gflops = flops / (total_time * 1e9);
-  double Gflops_no_stencil = flops_no_stencil / (interm_time * 1e9);
-  std::cout << GridLogIntegrator << "Flops: " << flops << "  - Gflop/s : " << Gflops << std::endl;
-  std::cout << GridLogIntegrator << "Flops NS: " << flops_no_stencil << "  - Gflop/s NS: " << Gflops_no_stencil << std::endl;
-}
+    double t1 = usecond();
+    double total_time = (t1 - t0) / 1e6;
+    double interm_time = (interm_t - t0) / 1e6;
+    double halo_time = (halo_t - interm_t) / 1e6;
+    double stencil_time = (t1 - halo_t) / 1e6;
+    std::cout << GridLogIntegrator << "Total time for force computation (s)       : " << total_time << std::endl;
+    std::cout << GridLogIntegrator << "Intermediate time for force computation (s): " << interm_time << std::endl;
+    std::cout << GridLogIntegrator << "Halo time in force computation (s)         : " << halo_time << std::endl;
+    std::cout << GridLogIntegrator << "Stencil time in force computation (s)      : " << stencil_time << std::endl;
+    double flops = p.Grid()->gSites() * (14 * N * N * N + 18 * N * N + 2);
+    double flops_no_stencil = p.Grid()->gSites() * (14 * N * N * N + 6 * N * N + 2);
+    double Gflops = flops / (total_time * 1e9);
+    double Gflops_no_stencil = flops_no_stencil / (interm_time * 1e9);
+    std::cout << GridLogIntegrator << "Flops: " << flops << "  - Gflop/s : " << Gflops << std::endl;
+    std::cout << GridLogIntegrator << "Flops NS: " << flops_no_stencil << "  - Gflop/s NS: " << Gflops_no_stencil << std::endl;
+  }
 };
 
-} // namespace Grid
+NAMESPACE_END(Grid);
+
 
-#endif // SCALAR_INT_ACTION_H
diff --git a/Grid/qcd/hmc/GenericHMCrunner.h b/Grid/qcd/hmc/GenericHMCrunner.h
index 26fec3d5..c2443dd0 100644
--- a/Grid/qcd/hmc/GenericHMCrunner.h
+++ b/Grid/qcd/hmc/GenericHMCrunner.h
@@ -25,16 +25,14 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 
   See the full license in the file "LICENSE" in the top level distribution
   directory
-  *************************************************************************************/
-/*  END LEGAL */
+*************************************************************************************/
+			   /*  END LEGAL */
 #ifndef GRID_GENERIC_HMC_RUNNER
 #define GRID_GENERIC_HMC_RUNNER
 
 #include <unordered_map>
 
-namespace Grid {
-namespace QCD {
-
+NAMESPACE_BEGIN(Grid);
 
 // very ugly here but possibly resolved if we had a base Reader class
 template < class ReaderClass >
@@ -44,12 +42,11 @@ public:
   virtual void initialize(ReaderClass& ) = 0;
 };
 
-
 template <class Implementation,
           template <typename, typename, typename> class Integrator,
           class RepresentationsPolicy = NoHirep, class ReaderClass = XmlReader>
 class HMCWrapperTemplate: public HMCRunnerBase<ReaderClass> {
- public:
+public:
   INHERIT_FIELD_TYPES(Implementation);
   typedef Implementation ImplPolicy;  // visible from outside
   template <typename S = NoSmearing<Implementation> >
@@ -88,8 +85,8 @@ class HMCWrapperTemplate: public HMCRunnerBase<ReaderClass> {
           arg != "CheckpointStart") {
         std::cout << GridLogError << "Unrecognized option in --StartingType\n";
         std::cout
-            << GridLogError
-            << "Valid [HotStart, ColdStart, TepidStart, CheckpointStart]\n";
+	  << GridLogError
+	  << "Valid [HotStart, ColdStart, TepidStart, CheckpointStart]\n";
         exit(1);
       }
       Parameters.StartingType = arg;
@@ -134,7 +131,7 @@ class HMCWrapperTemplate: public HMCRunnerBase<ReaderClass> {
 
   //////////////////////////////////////////////////////////////////
 
- private:
+private:
   template <class SmearingPolicy>
   void Runner(SmearingPolicy &Smearing) {
     auto UGrid = Resources.GetCartesian();
@@ -160,8 +157,8 @@ class HMCWrapperTemplate: public HMCRunnerBase<ReaderClass> {
     } else if (Parameters.StartingType == "CheckpointStart") {
       // CheckpointRestart
       Resources.GetCheckPointer()->CheckpointRestore(Parameters.StartTrajectory, U,
-                                   Resources.GetSerialRNG(),
-                                   Resources.GetParallelRNG());
+						     Resources.GetSerialRNG(),
+						     Resources.GetParallelRNG());
     }
 
     Smearing.set_Field(U);
@@ -198,22 +195,21 @@ using ConjugateHMCRunnerD = HMCWrapperTemplate<ConjugateGimplD, Integrator>;
 template <class RepresentationsPolicy,
           template <typename, typename, typename> class Integrator>
 using GenericHMCRunnerHirep =
-    HMCWrapperTemplate<PeriodicGimplR, Integrator, RepresentationsPolicy>;
+				     HMCWrapperTemplate<PeriodicGimplR, Integrator, RepresentationsPolicy>;
 
 template <class Implementation, class RepresentationsPolicy, 
           template <typename, typename, typename> class Integrator>
 using GenericHMCRunnerTemplate = HMCWrapperTemplate<Implementation, Integrator, RepresentationsPolicy>;
 
 typedef HMCWrapperTemplate<ScalarImplR, MinimumNorm2, ScalarFields>
-    ScalarGenericHMCRunner;
+ScalarGenericHMCRunner;
 
 typedef HMCWrapperTemplate<ScalarAdjImplR, MinimumNorm2, ScalarMatrixFields>
-    ScalarAdjGenericHMCRunner;
+ScalarAdjGenericHMCRunner;
 
 template <int Colours> 
 using ScalarNxNAdjGenericHMCRunner = HMCWrapperTemplate < ScalarNxNAdjImplR<Colours>, ForceGradient, ScalarNxNMatrixFields<Colours> >;
 
-}  // namespace QCD
-}  // namespace Grid
+NAMESPACE_END(Grid);
 
 #endif  // GRID_GENERIC_HMC_RUNNER
diff --git a/Grid/qcd/hmc/HMC.h b/Grid/qcd/hmc/HMC.h
index 5688bb24..0f933204 100644
--- a/Grid/qcd/hmc/HMC.h
+++ b/Grid/qcd/hmc/HMC.h
@@ -28,30 +28,26 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
-//--------------------------------------------------------------------
-/*! @file HMC.h
- * @brief Classes for Hybrid Monte Carlo update
- *
- * @author Guido Cossu
- */
-//--------------------------------------------------------------------
-#ifndef HMC_INCLUDED
-#define HMC_INCLUDED
+			   /*  END LEGAL */
+			   //--------------------------------------------------------------------
+			   /*! @file HMC.h
+			    * @brief Classes for Hybrid Monte Carlo update
+			    *
+			    * @author Guido Cossu
+			    */
+			   //--------------------------------------------------------------------
+#pragma once
 
 #include <string>
 #include <list>
 
-
-
 #include <Grid/qcd/hmc/integrators/Integrator.h>
 #include <Grid/qcd/hmc/integrators/Integrator_algorithm.h>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 struct HMCparameters: Serializable {
-	GRID_SERIALIZABLE_CLASS_MEMBERS(HMCparameters,
+  GRID_SERIALIZABLE_CLASS_MEMBERS(HMCparameters,
                                   Integer, StartTrajectory,
                                   Integer, Trajectories, /* @brief Number of sweeps in this run */
                                   bool, MetropolisTest,
@@ -71,13 +67,13 @@ struct HMCparameters: Serializable {
 
   template <class ReaderClass >
   HMCparameters(Reader<ReaderClass> & TheReader){
-  	initialize(TheReader);
+    initialize(TheReader);
   }
 
   template < class ReaderClass > 
   void initialize(Reader<ReaderClass> &TheReader){
-  	std::cout << GridLogMessage << "Reading HMC\n";
-  	read(TheReader, "HMC", *this);
+    std::cout << GridLogMessage << "Reading HMC\n";
+    read(TheReader, "HMC", *this);
   }
 
 
@@ -94,20 +90,20 @@ struct HMCparameters: Serializable {
 	
 template <class IntegratorType>
 class HybridMonteCarlo {
- private:
+private:
   const HMCparameters Params;
 
   typedef typename IntegratorType::Field Field;
   typedef std::vector< HmcObservable<Field> * > ObsListType;
   
-  	//pass these from the resource manager
+  //pass these from the resource manager
   GridSerialRNG &sRNG;   
   GridParallelRNG &pRNG; 
 
   Field &Ucur;
   
   IntegratorType &TheIntegrator;
-	ObsListType Observables;
+  ObsListType Observables;
 
   /////////////////////////////////////////////////////////
   // Metropolis step
@@ -179,7 +175,7 @@ class HybridMonteCarlo {
 
   
 
- public:
+public:
   /////////////////////////////////////////
   // Constructor
   /////////////////////////////////////////
@@ -192,7 +188,7 @@ class HybridMonteCarlo {
   void evolve(void) {
     Real DeltaH;
 
-    Field Ucopy(Ucur._grid);
+    Field Ucopy(Ucur.Grid());
 
     Params.print_parameters();
     TheIntegrator.print_actions();
@@ -238,14 +234,9 @@ class HybridMonteCarlo {
 
 };
 
-
-}  // QCD
-}  // Grid
-
-
+NAMESPACE_END(Grid);
 // april 11 2017 merge, Guido, commenting out
 //#include <Grid/parallelIO/NerscIO.h>
 //#include <Grid/qcd/hmc/NerscCheckpointer.h>
 //#include <Grid/qcd/hmc/HmcRunner.h>
 
-#endif
diff --git a/Grid/qcd/hmc/HMCModules.h b/Grid/qcd/hmc/HMCModules.h
index 8b9b0479..4c61a006 100644
--- a/Grid/qcd/hmc/HMCModules.h
+++ b/Grid/qcd/hmc/HMCModules.h
@@ -26,21 +26,19 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef GRID_HMC_MODULES
 #define GRID_HMC_MODULES
 
-
 #include "HMC_GridModules.h"
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 ////////////////////////////////////////////////////////////////////
 struct RNGModuleParameters: Serializable {
   GRID_SERIALIZABLE_CLASS_MEMBERS(RNGModuleParameters,
-  std::string, serial_seeds,
-  std::string, parallel_seeds,);
+				  std::string, serial_seeds,
+				  std::string, parallel_seeds,);
 
   std::vector<int> getSerialSeeds(){return strToVec<int>(serial_seeds);}
   std::vector<int> getParallelSeeds(){return strToVec<int>(parallel_seeds);}
@@ -56,9 +54,9 @@ struct RNGModuleParameters: Serializable {
 
 // Random number generators module
 class RNGModule{
-   GridSerialRNG sRNG_;
-   std::unique_ptr<GridParallelRNG> pRNG_;
-   RNGModuleParameters Params_;
+  GridSerialRNG sRNG_;
+  std::unique_ptr<GridParallelRNG> pRNG_;
+  RNGModuleParameters Params_;
 
 public:
 
@@ -93,19 +91,16 @@ public:
 /// Smearing module
 template <class ImplementationPolicy>
 class SmearingModule{
-   virtual void get_smearing();
+virtual void get_smearing();
 };
 
 template <class ImplementationPolicy>
 class StoutSmearingModule: public SmearingModule<ImplementationPolicy>{
-   SmearedConfiguration<ImplementationPolicy> SmearingPolicy;
+SmearedConfiguration<ImplementationPolicy> SmearingPolicy;
 };
 
 */
 
-
-
-}  // namespace QCD
-}  // namespace Grid
+NAMESPACE_END(Grid);
 
 #endif  // GRID_HMC_MODULES
diff --git a/Grid/qcd/hmc/HMCResourceManager.h b/Grid/qcd/hmc/HMCResourceManager.h
index fcfaeaed..783e4890 100644
--- a/Grid/qcd/hmc/HMCResourceManager.h
+++ b/Grid/qcd/hmc/HMCResourceManager.h
@@ -25,8 +25,8 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 
   See the full license in the file "LICENSE" in the top level distribution
   directory
-  *************************************************************************************/
-/*  END LEGAL */
+*************************************************************************************/
+			   /*  END LEGAL */
 #ifndef HMC_RESOURCE_MANAGER_H
 #define HMC_RESOURCE_MANAGER_H
 
@@ -64,15 +64,14 @@ with this program; if not, write to the Free Software Foundation, Inc.,
     }                                                                    \
   }
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 // HMC Resource manager
 template <class ImplementationPolicy>
 class HMCResourceManager {
-  typedef HMCModuleBase< QCD::BaseHmcCheckpointer<ImplementationPolicy> > CheckpointerBaseModule;
-  typedef HMCModuleBase< QCD::HmcObservable<typename ImplementationPolicy::Field> > ObservableBaseModule;
-  typedef ActionModuleBase< QCD::Action<typename ImplementationPolicy::Field>, GridModule > ActionBaseModule;
+  typedef HMCModuleBase< BaseHmcCheckpointer<ImplementationPolicy> > CheckpointerBaseModule;
+  typedef HMCModuleBase< HmcObservable<typename ImplementationPolicy::Field> > ObservableBaseModule;
+  typedef ActionModuleBase< Action<typename ImplementationPolicy::Field>, GridModule > ActionBaseModule;
 
   // Named storage for grid pairs (std + red-black)
   std::unordered_map<std::string, GridModule> Grids;
@@ -101,7 +100,7 @@ class HMCResourceManager {
   }
 
 
- public:
+public:
   HMCResourceManager() : have_RNG(false), have_CheckPointer(false) {}
 
   template <class ReaderClass, class vector_type = vComplex >
@@ -151,13 +150,13 @@ class HMCResourceManager {
     }
 
     if(!Read.push("Level")){// push must check if the node exist
-         std::cout << "Level not found" << std::endl; 
+      std::cout << "Level not found" << std::endl; 
       exit(1);
     }
     do
-    {
-      fill_ActionsLevel(Read); 
-    }
+      {
+	fill_ActionsLevel(Read); 
+      }
     while(Read.push("Level"));
 
     Read.pop();
@@ -268,10 +267,10 @@ class HMCResourceManager {
 
   RegisterLoadCheckPointerFunction(Binary);
   RegisterLoadCheckPointerFunction(Nersc);
-  #ifdef HAVE_LIME
+#ifdef HAVE_LIME
   RegisterLoadCheckPointerFunction(ILDG);
   RegisterLoadCheckPointerMetadataFunction(Scidac);
-  #endif
+#endif
 
   ////////////////////////////////////////////////////////
   // Observables
@@ -297,7 +296,7 @@ class HMCResourceManager {
 
 
 private:
-   // this private
+  // this private
   template <class ReaderClass >
   void fill_ActionsLevel(ReaderClass &Read){
     // Actions set
@@ -322,7 +321,7 @@ private:
 
 
 };
-}
-}
+
+NAMESPACE_END(Grid);
 
 #endif  // HMC_RESOURCE_MANAGER_H
diff --git a/Grid/qcd/hmc/HMCRunnerModule.h b/Grid/qcd/hmc/HMCRunnerModule.h
index f595d996..fa714a1c 100644
--- a/Grid/qcd/hmc/HMCRunnerModule.h
+++ b/Grid/qcd/hmc/HMCRunnerModule.h
@@ -26,28 +26,28 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef HMC_RUNNER_MODULE
 #define HMC_RUNNER_MODULE
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 // the reader class is necessary here for the automatic initialization of the resources
 // if we had a virtual reader would have been unecessary
 template <class HMCType, class ReaderClass >
 class HMCModule
-    : public Parametrized< QCD::HMCparameters >,
-      public HMCModuleBase< QCD::HMCRunnerBase<ReaderClass> > {
- public:
-  typedef HMCModuleBase< QCD::HMCRunnerBase<ReaderClass> > Base;
+  : public Parametrized< HMCparameters >,
+    public HMCModuleBase< HMCRunnerBase<ReaderClass> > {
+public:
+  typedef HMCModuleBase< HMCRunnerBase<ReaderClass> > Base;
   typedef typename Base::Product Product;
 
   std::unique_ptr<HMCType> HMCPtr;
 
-  HMCModule(QCD::HMCparameters Par) : Parametrized<QCD::HMCparameters>(Par) {}
+  HMCModule(HMCparameters Par) : Parametrized<HMCparameters>(Par) {}
 
   template <class ReaderCl>
-  HMCModule(Reader<ReaderCl>& R) : Parametrized<QCD::HMCparameters>(R, "HMC"){};
+  HMCModule(Reader<ReaderCl>& R) : Parametrized<HMCparameters>(R, "HMC"){};
 
   Product* getPtr() {
     if (!HMCPtr) initialize();
@@ -55,17 +55,17 @@ class HMCModule
     return HMCPtr.get();
   }
 
- private:
+private:
   virtual void initialize() = 0;
 };
 
 // Factory
 template <char const *str, class ReaderClass >
 class HMCRunnerModuleFactory
-    : public Factory < HMCModuleBase< QCD::HMCRunnerBase<ReaderClass> > ,	Reader<ReaderClass> > {
- public:
- 	typedef Reader<ReaderClass> TheReader; 
- 	// use SINGLETON FUNCTOR MACRO HERE
+  : public Factory < HMCModuleBase< HMCRunnerBase<ReaderClass> > ,	Reader<ReaderClass> > {
+public:
+  typedef Reader<ReaderClass> TheReader; 
+  // use SINGLETON FUNCTOR MACRO HERE
   HMCRunnerModuleFactory(const HMCRunnerModuleFactory& e) = delete;
   void operator=(const HMCRunnerModuleFactory& e) = delete;
   static HMCRunnerModuleFactory& getInstance(void) {
@@ -73,10 +73,10 @@ class HMCRunnerModuleFactory
     return e;
   }
 
- private:
+private:
   HMCRunnerModuleFactory(void) = default;
   std::string obj_type() const {
-  	return std::string(str);
+    return std::string(str);
   }
 };
 
@@ -88,50 +88,43 @@ class HMCRunnerModuleFactory
 // macro for these
 
 template < class ImplementationPolicy, class RepresentationPolicy, class ReaderClass >
-class HMCLeapFrog: public HMCModule< QCD::GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, QCD::LeapFrog>, ReaderClass >{
-  typedef HMCModule< QCD::GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, QCD::LeapFrog>, ReaderClass  > HMCBaseMod;
+class HMCLeapFrog: public HMCModule< GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, LeapFrog>, ReaderClass >{
+  typedef HMCModule< GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, LeapFrog>, ReaderClass  > HMCBaseMod;
   using HMCBaseMod::HMCBaseMod;
 
   // aquire resource
   virtual void initialize(){
-    this->HMCPtr.reset(new QCD::GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, QCD::LeapFrog>(this->Par_) );
+    this->HMCPtr.reset(new GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, LeapFrog>(this->Par_) );
   }
 };
 
 template < class ImplementationPolicy, class RepresentationPolicy, class ReaderClass >
-class HMCMinimumNorm2: public HMCModule< QCD::GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, QCD::MinimumNorm2>, ReaderClass  >{
-  typedef HMCModule< QCD::GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, QCD::MinimumNorm2>, ReaderClass  > HMCBaseMod;
+class HMCMinimumNorm2: public HMCModule< GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, MinimumNorm2>, ReaderClass  >{
+  typedef HMCModule< GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, MinimumNorm2>, ReaderClass  > HMCBaseMod;
   using HMCBaseMod::HMCBaseMod;
 
   // aquire resource
   virtual void initialize(){
-    this->HMCPtr.reset(new QCD::GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, QCD::MinimumNorm2>(this->Par_));
+    this->HMCPtr.reset(new GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, MinimumNorm2>(this->Par_));
   }
 };
 
 
 template < class ImplementationPolicy, class RepresentationPolicy, class ReaderClass >
-class HMCForceGradient: public HMCModule< QCD::GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, QCD::ForceGradient>, ReaderClass  >{
-  typedef HMCModule< QCD::GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, QCD::ForceGradient>, ReaderClass   > HMCBaseMod;
+class HMCForceGradient: public HMCModule< GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, ForceGradient>, ReaderClass  >{
+  typedef HMCModule< GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, ForceGradient>, ReaderClass   > HMCBaseMod;
   using HMCBaseMod::HMCBaseMod;
 
   // aquire resource
   virtual void initialize(){
-    this->HMCPtr.reset(new QCD::GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, QCD::ForceGradient>(this->Par_) );
+    this->HMCPtr.reset(new GenericHMCRunnerTemplate<ImplementationPolicy, RepresentationPolicy, ForceGradient>(this->Par_) );
   }
 };
 
 extern char hmc_string[];
 
-
-
-
-
-
 //////////////////////////////////////////////////////////////
 
+NAMESPACE_END(Grid);
 
-
-}
-
-#endif
\ No newline at end of file
+#endif
diff --git a/Grid/qcd/hmc/HMC_GridModules.h b/Grid/qcd/hmc/HMC_GridModules.h
index 0f34e9a7..0c834cf2 100644
--- a/Grid/qcd/hmc/HMC_GridModules.h
+++ b/Grid/qcd/hmc/HMC_GridModules.h
@@ -26,11 +26,11 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef HMC_GRID_MODULES
 #define HMC_GRID_MODULES
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 // Resources
 // Modules for grids
@@ -40,8 +40,8 @@ namespace Grid {
 class GridModuleParameters: Serializable{
 public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(GridModuleParameters,
-  std::string, lattice,
-  std::string, mpi);
+				  std::string, lattice,
+				  std::string, mpi);
 
   std::vector<int> getLattice() const {return strToVec<int>(lattice);}
   std::vector<int> getMpi()     const {return strToVec<int>(mpi);}
@@ -51,7 +51,7 @@ public:
     if (getLattice().size() != getMpi().size() ) {
       std::cout << GridLogError
                 << "Error in GridModuleParameters: lattice and mpi dimensions "
-                   "do not match"
+	"do not match"
                 << std::endl;
       exit(1);
     }
@@ -70,12 +70,12 @@ public:
     write(Writer, name, *this);
   }
 private:
-    std::string name;
+  std::string name;
 };
 
 // Lower level class
 class GridModule {
- public:
+public:
   GridCartesian* get_full() {
     std::cout << GridLogDebug << "Getting cartesian in module"<< std::endl;
     return grid_.get(); }
@@ -88,7 +88,7 @@ class GridModule {
   void show_full_decomposition(){ grid_->show_decomposition(); }
   void show_rb_decomposition(){ rbgrid_->show_decomposition(); }
 
- protected:
+protected:
   std::unique_ptr<GridCartesian> grid_;
   std::unique_ptr<GridRedBlackCartesian> rbgrid_;
 
@@ -104,24 +104,23 @@ class GridFourDimModule : public GridModule
 public:
   GridFourDimModule()
   {
-    using namespace QCD;
     set_full(SpaceTimeGrid::makeFourDimGrid(
-        GridDefaultLatt(), 
-        GridDefaultSimd(4, vector_type::Nsimd()),
-        GridDefaultMpi()));
+					    GridDefaultLatt(), 
+					    GridDefaultSimd(4, vector_type::Nsimd()),
+					    GridDefaultMpi()));
     set_rb(SpaceTimeGrid::makeFourDimRedBlackGrid(grid_.get()));
   }
 
   GridFourDimModule(const std::vector<int> tweak_simd)
   {
-    using namespace QCD;
+    Coordinate _tweak_simd(tweak_simd);
     if (tweak_simd.size() != 4)
-    {
-      std::cout << GridLogError
-                << "Error in GridFourDimModule: SIMD size different from 4" 
-                << std::endl;
-      exit(1);
-    }
+      {
+	std::cout << GridLogError
+		  << "Error in GridFourDimModule: SIMD size different from 4" 
+		  << std::endl;
+	exit(1);
+      }
 
     // Checks that the product agrees with the expectation
     int simd_sum = 1;
@@ -130,48 +129,47 @@ public:
     std::cout << GridLogDebug << "TweakSIMD: " << tweak_simd << "  Sum: " << simd_sum << std::endl;
 
     if (simd_sum == vector_type::Nsimd())
-    {
-      set_full(SpaceTimeGrid::makeFourDimGrid(
-          GridDefaultLatt(), 
-          tweak_simd, 
-          GridDefaultMpi()));
-      set_rb(SpaceTimeGrid::makeFourDimRedBlackGrid(grid_.get()));
-    }
+      {
+	set_full(SpaceTimeGrid::makeFourDimGrid(
+						GridDefaultLatt(), 
+						_tweak_simd, 
+						GridDefaultMpi()));
+	set_rb(SpaceTimeGrid::makeFourDimRedBlackGrid(grid_.get()));
+      }
     else
-    {
-      std::cout << GridLogError 
-                << "Error in GridFourDimModule: SIMD lanes must sum to " 
-                << vector_type::Nsimd() 
-                << std::endl;
-    }
+      {
+	std::cout << GridLogError 
+		  << "Error in GridFourDimModule: SIMD lanes must sum to " 
+		  << vector_type::Nsimd() 
+		  << std::endl;
+      }
   }
 
   GridFourDimModule(const GridModuleParameters Params)
   {
-    using namespace QCD;
     std::vector<int> lattice_v = Params.getLattice();
     std::vector<int> mpi_v = Params.getMpi();
     if (lattice_v.size() == 4)
-    {
-      set_full(SpaceTimeGrid::makeFourDimGrid(
-          lattice_v, 
-          GridDefaultSimd(4, vector_type::Nsimd()),
-          mpi_v));
-      set_rb(SpaceTimeGrid::makeFourDimRedBlackGrid(grid_.get()));
-    }
+      {
+	set_full(SpaceTimeGrid::makeFourDimGrid(
+						lattice_v, 
+						GridDefaultSimd(4, vector_type::Nsimd()),
+						mpi_v));
+	set_rb(SpaceTimeGrid::makeFourDimRedBlackGrid(grid_.get()));
+      }
     else
-    {
-      std::cout << GridLogError
-                << "Error in GridFourDimModule: lattice dimension different from 4"
-                << std::endl;
-      exit(1);
-    }
+      {
+	std::cout << GridLogError
+		  << "Error in GridFourDimModule: lattice dimension different from 4"
+		  << std::endl;
+	exit(1);
+      }
   }
 };
 
 typedef GridFourDimModule<vComplex> GridDefaultFourDimModule;
 
 
-}  // namespace Grid
+NAMESPACE_END(Grid);
 
 #endif  // HMC_GRID_MODULES
diff --git a/Grid/qcd/hmc/HMC_aggregate.h b/Grid/qcd/hmc/HMC_aggregate.h
index 53d4b9c8..e4d2ce83 100644
--- a/Grid/qcd/hmc/HMC_aggregate.h
+++ b/Grid/qcd/hmc/HMC_aggregate.h
@@ -28,25 +28,26 @@ directory
 /*  END LEGAL */
 //--------------------------------------------------------------------
 //--------------------------------------------------------------------
-#ifndef HMC_AGGREGATE_INCLUDED
-#define HMC_AGGREGATE_INCLUDED
+#pragma once
 
 #include <string>
 
 #include <Grid/qcd/observables/hmc_observable.h>
 #include <Grid/qcd/hmc/HMC.h>
 
-
 // annoying location; should move this ?
 #include <Grid/parallelIO/IldgIOtypes.h>
 #include <Grid/parallelIO/IldgIO.h>
 #include <Grid/parallelIO/NerscIO.h>
+NAMESPACE_CHECK(Ildg);
 
 #include <Grid/qcd/hmc/checkpointers/CheckPointers.h>
 #include <Grid/qcd/hmc/HMCModules.h>
 #include <Grid/qcd/modules/mods.h>
+NAMESPACE_CHECK(HMCmodules);
 #include <Grid/qcd/hmc/HMCResourceManager.h>
+NAMESPACE_CHECK(HMCresourcemanager);
 #include <Grid/qcd/hmc/GenericHMCrunner.h>
 #include <Grid/qcd/hmc/HMCRunnerModule.h>
+NAMESPACE_CHECK(HMCrunner);
 
-#endif
diff --git a/Grid/qcd/hmc/checkpointers/BaseCheckpointer.h b/Grid/qcd/hmc/checkpointers/BaseCheckpointer.h
index f4ef252b..3cd05ebc 100644
--- a/Grid/qcd/hmc/checkpointers/BaseCheckpointer.h
+++ b/Grid/qcd/hmc/checkpointers/BaseCheckpointer.h
@@ -25,27 +25,26 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef BASE_CHECKPOINTER
 #define BASE_CHECKPOINTER
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 class CheckpointerParameters : Serializable {
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(CheckpointerParameters, 
-  	std::string, config_prefix, 
-  	std::string, rng_prefix, 
-  	int, saveInterval, 
-  	std::string, format, );
+				  std::string, config_prefix, 
+				  std::string, rng_prefix, 
+				  int, saveInterval, 
+				  std::string, format, );
 
   CheckpointerParameters(std::string cf = "cfg", std::string rn = "rng",
-   		      int savemodulo = 1, const std::string &f = "IEEE64BIG")
-      : config_prefix(cf),
-        rng_prefix(rn),
-        saveInterval(savemodulo),
-        format(f){};
+			 int savemodulo = 1, const std::string &f = "IEEE64BIG")
+    : config_prefix(cf),
+      rng_prefix(rn),
+      saveInterval(savemodulo),
+      format(f){};
 
 
   template <class ReaderClass >
@@ -60,7 +59,7 @@ class CheckpointerParameters : Serializable {
 // Base class for checkpointers
 template <class Impl>
 class BaseHmcCheckpointer : public HmcObservable<typename Impl::Field> {
- public:
+public:
   void build_filenames(int traj, CheckpointerParameters &Params,
                        std::string &conf_file, std::string &rng_file) {
     {
@@ -74,7 +73,7 @@ class BaseHmcCheckpointer : public HmcObservable<typename Impl::Field> {
       os << Params.config_prefix << "." << traj;
       conf_file = os.str();
     }
- 	} 
+  } 
 
   void check_filename(const std::string &filename){
     std::ifstream f(filename.c_str());
@@ -92,6 +91,7 @@ class BaseHmcCheckpointer : public HmcObservable<typename Impl::Field> {
 
 };  // class BaseHmcCheckpointer
 ///////////////////////////////////////////////////////////////////////////////
-}
-}
+
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/qcd/hmc/checkpointers/BinaryCheckpointer.h b/Grid/qcd/hmc/checkpointers/BinaryCheckpointer.h
index 025398eb..ef9e6194 100644
--- a/Grid/qcd/hmc/checkpointers/BinaryCheckpointer.h
+++ b/Grid/qcd/hmc/checkpointers/BinaryCheckpointer.h
@@ -25,7 +25,7 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef BINARY_CHECKPOINTER
 #define BINARY_CHECKPOINTER
 
@@ -33,16 +33,15 @@ directory
 #include <sstream>
 #include <string>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 // Simple checkpointer, only binary file
 template <class Impl>
 class BinaryHmcCheckpointer : public BaseHmcCheckpointer<Impl> {
- private:
+private:
   CheckpointerParameters Params;
 
- public:
+public:
   INHERIT_FIELD_TYPES(Impl);  // Gets the Field type, a Lattice object
 
   // Extract types from the Field
@@ -111,6 +110,6 @@ class BinaryHmcCheckpointer : public BaseHmcCheckpointer<Impl> {
 	      << std::dec << std::endl;
   };
 };
-}
-}
+
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/qcd/hmc/checkpointers/CheckPointerModules.h b/Grid/qcd/hmc/checkpointers/CheckPointerModules.h
index f17cd2c8..13b6cf13 100644
--- a/Grid/qcd/hmc/checkpointers/CheckPointerModules.h
+++ b/Grid/qcd/hmc/checkpointers/CheckPointerModules.h
@@ -25,27 +25,23 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 
 #ifndef CP_MODULES_H
 #define CP_MODULES_H
 
-
-// FIXME  Reorganize QCD namespace 
-
-
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 ////////////////////////////////////////////////////////////////////////
 // Checkpoint module, owns the Checkpointer
 ////////////////////////////////////////////////////////////////////////
 
 template <class ImplementationPolicy>
-class CheckPointerModule: public Parametrized<QCD::CheckpointerParameters>, public HMCModuleBase< QCD::BaseHmcCheckpointer<ImplementationPolicy> >  {
- public:
- 	std::unique_ptr<QCD::BaseHmcCheckpointer<ImplementationPolicy> > CheckPointPtr;
- 	typedef QCD::CheckpointerParameters APar;
-  typedef HMCModuleBase< QCD::BaseHmcCheckpointer<ImplementationPolicy> > Base;
+class CheckPointerModule: public Parametrized<CheckpointerParameters>, public HMCModuleBase< BaseHmcCheckpointer<ImplementationPolicy> >  {
+public:
+  std::unique_ptr<BaseHmcCheckpointer<ImplementationPolicy> > CheckPointPtr;
+  typedef CheckpointerParameters APar;
+  typedef HMCModuleBase< BaseHmcCheckpointer<ImplementationPolicy> > Base;
   typedef typename Base::Product Product;
 
   CheckPointerModule(APar Par): Parametrized<APar>(Par) {}
@@ -53,7 +49,7 @@ class CheckPointerModule: public Parametrized<QCD::CheckpointerParameters>, publ
   CheckPointerModule(Reader<ReaderClass>& Reader) : Parametrized<APar>(Reader){};
 
   virtual void print_parameters(){
-  	std::cout << this->Par_ << std::endl;
+    std::cout << this->Par_ << std::endl;
   }
 
   Product* getPtr() {
@@ -62,19 +58,17 @@ class CheckPointerModule: public Parametrized<QCD::CheckpointerParameters>, publ
     return CheckPointPtr.get();
   }
 
- private:
+private:
   virtual void initialize() = 0;
 
 };
 
-
-
 template <char const *str, class ImplementationPolicy, class ReaderClass >
 class HMC_CPModuleFactory
-    : public Factory < HMCModuleBase< QCD::BaseHmcCheckpointer<ImplementationPolicy> > ,	Reader<ReaderClass> > {
- public:
- 	typedef Reader<ReaderClass> TheReader; 
- 	// use SINGLETON FUNCTOR MACRO HERE
+  : public Factory < HMCModuleBase< BaseHmcCheckpointer<ImplementationPolicy> > ,	Reader<ReaderClass> > {
+public:
+  typedef Reader<ReaderClass> TheReader; 
+  // use SINGLETON FUNCTOR MACRO HERE
   HMC_CPModuleFactory(const HMC_CPModuleFactory& e) = delete;
   void operator=(const HMC_CPModuleFactory& e) = delete;
   static HMC_CPModuleFactory& getInstance(void) {
@@ -82,19 +76,16 @@ class HMC_CPModuleFactory
     return e;
   }
 
- private:
+private:
   HMC_CPModuleFactory(void) = default;
   std::string obj_type() const {
-  	return std::string(str);
+    return std::string(str);
   }
 };
 
-
-
 /////////////////////////////////////////////////////////////////////
 // Concrete classes
 /////////////////////////////////////////////////////////////////////
-namespace QCD{
 
 template<class ImplementationPolicy>
 class BinaryCPModule: public CheckPointerModule< ImplementationPolicy> {
@@ -116,7 +107,7 @@ class NerscCPModule: public CheckPointerModule< ImplementationPolicy> {
 
   // acquire resource
   virtual void initialize(){
-     this->CheckPointPtr.reset(new NerscHmcCheckpointer<ImplementationPolicy>(this->Par_));
+    this->CheckPointPtr.reset(new NerscHmcCheckpointer<ImplementationPolicy>(this->Par_));
   }
 
 };
@@ -131,7 +122,7 @@ class ILDGCPModule: public CheckPointerModule< ImplementationPolicy> {
 
   // acquire resource
   virtual void initialize(){
-     this->CheckPointPtr.reset(new ILDGHmcCheckpointer<ImplementationPolicy>(this->Par_));
+    this->CheckPointPtr.reset(new ILDGHmcCheckpointer<ImplementationPolicy>(this->Par_));
   }
 
 };
@@ -152,21 +143,18 @@ public:
 };
 #endif
 
-
-}// QCD temporarily here
-
-
 extern char cp_string[];
 
 /*
 // use macros?
-static Registrar<QCD::BinaryCPModule<QCD::PeriodicGimplR>, HMC_CPModuleFactory<cp_string, QCD::PeriodicGimplR, XmlReader> > __CPBinarymodXMLInit("Binary");
-static Registrar<QCD::NerscCPModule<QCD::PeriodicGimplR> , HMC_CPModuleFactory<cp_string, QCD::PeriodicGimplR, XmlReader> > __CPNerscmodXMLInit("Nersc");
+static Registrar<BinaryCPModule<PeriodicGimplR>, HMC_CPModuleFactory<cp_string, PeriodicGimplR, XmlReader> > __CPBinarymodXMLInit("Binary");
+static Registrar<NerscCPModule<PeriodicGimplR> , HMC_CPModuleFactory<cp_string, PeriodicGimplR, XmlReader> > __CPNerscmodXMLInit("Nersc");
 
 #ifdef HAVE_LIME
-static Registrar<QCD::ILDGCPModule<QCD::PeriodicGimplR>  , HMC_CPModuleFactory<cp_string, QCD::PeriodicGimplR, XmlReader> > __CPILDGmodXMLInit("ILDG");
+static Registrar<ILDGCPModule<PeriodicGimplR>  , HMC_CPModuleFactory<cp_string, PeriodicGimplR, XmlReader> > __CPILDGmodXMLInit("ILDG");
 #endif
 */
 
-}// Grid
+NAMESPACE_END(Grid);
+
 #endif //CP_MODULES_H
diff --git a/Grid/qcd/hmc/checkpointers/ILDGCheckpointer.h b/Grid/qcd/hmc/checkpointers/ILDGCheckpointer.h
index f7e6b17e..269caa6e 100644
--- a/Grid/qcd/hmc/checkpointers/ILDGCheckpointer.h
+++ b/Grid/qcd/hmc/checkpointers/ILDGCheckpointer.h
@@ -25,7 +25,7 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef ILDG_CHECKPOINTER
 #define ILDG_CHECKPOINTER
 
@@ -35,16 +35,15 @@ directory
 #include <sstream>
 #include <string>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 // Only for Gauge fields
 template <class Implementation>
 class ILDGHmcCheckpointer : public BaseHmcCheckpointer<Implementation> {
- private:
+private:
   CheckpointerParameters Params;
 
- public:
+public:
   INHERIT_GIMPL_TYPES(Implementation);
 
   ILDGHmcCheckpointer(const CheckpointerParameters &Params_) { initialize(Params_); }
@@ -74,7 +73,7 @@ class ILDGHmcCheckpointer : public BaseHmcCheckpointer<Implementation> {
     if ((traj % Params.saveInterval) == 0) {
       std::string config, rng;
       this->build_filenames(traj, Params, config, rng);
-      GridBase *grid = U._grid;
+      GridBase *grid = U.Grid();
       uint32_t nersc_csum,scidac_csuma,scidac_csumb;
       BinaryIO::writeRNG(sRNG, pRNG, rng, 0,nersc_csum,scidac_csuma,scidac_csumb);
       IldgWriter _IldgWriter(grid->IsBoss());
@@ -117,8 +116,8 @@ class ILDGHmcCheckpointer : public BaseHmcCheckpointer<Implementation> {
 	      << std::dec << std::endl;
   };
 };
-}
-}
+
+NAMESPACE_END(Grid);
 
 #endif  // HAVE_LIME
 #endif  // ILDG_CHECKPOINTER
diff --git a/Grid/qcd/hmc/checkpointers/NerscCheckpointer.h b/Grid/qcd/hmc/checkpointers/NerscCheckpointer.h
index d452b994..cfcc44d8 100644
--- a/Grid/qcd/hmc/checkpointers/NerscCheckpointer.h
+++ b/Grid/qcd/hmc/checkpointers/NerscCheckpointer.h
@@ -25,7 +25,7 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef NERSC_CHECKPOINTER
 #define NERSC_CHECKPOINTER
 
@@ -33,16 +33,15 @@ directory
 #include <sstream>
 #include <string>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 // Only for Gauge fields
 template <class Gimpl>
 class NerscHmcCheckpointer : public BaseHmcCheckpointer<Gimpl> {
- private:
+private:
   CheckpointerParameters Params;
 
- public:
+public:
   INHERIT_GIMPL_TYPES(Gimpl);  // only for gauge configurations
 
   NerscHmcCheckpointer(const CheckpointerParameters &Params_) { initialize(Params_); }
@@ -78,6 +77,6 @@ class NerscHmcCheckpointer : public BaseHmcCheckpointer<Gimpl> {
     NerscIO::readConfiguration(U, header, config);
   };
 };
-}
-}
+
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/qcd/hmc/checkpointers/ScidacCheckpointer.h b/Grid/qcd/hmc/checkpointers/ScidacCheckpointer.h
index 063a475c..986585ea 100644
--- a/Grid/qcd/hmc/checkpointers/ScidacCheckpointer.h
+++ b/Grid/qcd/hmc/checkpointers/ScidacCheckpointer.h
@@ -26,8 +26,7 @@ See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
 /*  END LEGAL */
-#ifndef SCIDAC_CHECKPOINTER
-#define SCIDAC_CHECKPOINTER
+#pragma once
 
 #ifdef HAVE_LIME
 
@@ -35,9 +34,7 @@ directory
 #include <sstream>
 #include <string>
 
-namespace Grid {
-namespace QCD {
-
+NAMESPACE_BEGIN(Grid);
 // For generic fields
 template <class Implementation, class Metadata>
 class ScidacHmcCheckpointer : public BaseHmcCheckpointer<Implementation> {
@@ -78,7 +75,7 @@ class ScidacHmcCheckpointer : public BaseHmcCheckpointer<Implementation> {
     if ((traj % Params.saveInterval) == 0) {
       std::string config, rng;
       this->build_filenames(traj, Params, config, rng);
-      GridBase *grid = U._grid;
+      GridBase *grid = U.Grid();
       uint32_t nersc_csum,scidac_csuma,scidac_csumb;
       BinaryIO::writeRNG(sRNG, pRNG, rng, 0,nersc_csum,scidac_csuma,scidac_csumb);
       ScidacWriter _ScidacWriter(grid->IsBoss());
@@ -115,8 +112,8 @@ class ScidacHmcCheckpointer : public BaseHmcCheckpointer<Implementation> {
 	      << std::dec << std::endl;
   };
 };
-}
-}
+NAMESPACE_END(Grid);
+
 
 #endif  // HAVE_LIME
-#endif  // ILDG_CHECKPOINTER
+
diff --git a/Grid/qcd/hmc/integrators/Integrator.h b/Grid/qcd/hmc/integrators/Integrator.h
index 9d4376fc..91af7372 100644
--- a/Grid/qcd/hmc/integrators/Integrator.h
+++ b/Grid/qcd/hmc/integrators/Integrator.h
@@ -27,33 +27,29 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
-//--------------------------------------------------------------------
+			   /*  END LEGAL */
+			   //--------------------------------------------------------------------
 #ifndef INTEGRATOR_INCLUDED
 #define INTEGRATOR_INCLUDED
 
 #include <memory>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 class IntegratorParameters: Serializable {
 public:
-        GRID_SERIALIZABLE_CLASS_MEMBERS(IntegratorParameters,
-                std::string, name,      // name of the integrator
-    unsigned int, MDsteps,  // number of outer steps
-    RealD, trajL,           // trajectory length
-  )
+  GRID_SERIALIZABLE_CLASS_MEMBERS(IntegratorParameters,
+				  std::string, name,      // name of the integrator
+				  unsigned int, MDsteps,  // number of outer steps
+				  RealD, trajL)           // trajectory length
 
   IntegratorParameters(int MDsteps_ = 10, RealD trajL_ = 1.0)
-      : MDsteps(MDsteps_),
-        trajL(trajL_){
-        // empty body constructor
-  };
-
+  : MDsteps(MDsteps_),
+    trajL(trajL_) {};
 
   template <class ReaderClass, typename std::enable_if<isReader<ReaderClass>::value, int >::type = 0 >
-  IntegratorParameters(ReaderClass & Reader){
+  IntegratorParameters(ReaderClass & Reader)
+  {
     std::cout << GridLogMessage << "Reading integrator\n";
     read(Reader, "Integrator", *this);
   }
@@ -69,7 +65,7 @@ public:
 /*! @brief Class for Molecular Dynamics management */
 template <class FieldImplementation, class SmearingPolicy, class RepresentationPolicy>
 class Integrator {
- protected:
+protected:
   typedef typename FieldImplementation::Field MomentaField;  //for readability
   typedef typename FieldImplementation::Field Field;
 
@@ -84,7 +80,8 @@ class Integrator {
 
   const ActionSet<Field, RepresentationPolicy> as;
 
-  void update_P(Field& U, int level, double ep) {
+  void update_P(Field& U, int level, double ep) 
+  {
     t_P[level] += ep;
     update_P(P, U, level, ep);
 
@@ -93,16 +90,17 @@ class Integrator {
 
   // to be used by the actionlevel class to iterate
   // over the representations
-  struct _updateP {
+  struct _updateP 
+  {
     template <class FieldType, class GF, class Repr>
     void operator()(std::vector<Action<FieldType>*> repr_set, Repr& Rep,
                     GF& Mom, GF& U, double ep) {
       for (int a = 0; a < repr_set.size(); ++a) {
-        FieldType forceR(U._grid);
+        FieldType forceR(U.Grid());
         // Implement smearing only for the fundamental representation now
         repr_set.at(a)->deriv(Rep.U, forceR);
         GF force = Rep.RtoFundamentalProject(forceR);  // Ta for the fundamental rep
-        Real force_abs = std::sqrt(norm2(force)/(U._grid->gSites()));
+        Real force_abs = std::sqrt(norm2(force)/(U.Grid()->gSites()));
         std::cout << GridLogIntegrator << "Hirep Force average: " << force_abs << std::endl;
 	Mom -= force * ep* HMC_MOMENTUM_DENOMINATOR;; 
       }
@@ -115,8 +113,8 @@ class Integrator {
 
     for (int a = 0; a < as[level].actions.size(); ++a) {
       double start_full = usecond();
-      Field force(U._grid);
-      conformable(U._grid, Mom._grid);
+      Field force(U.Grid());
+      conformable(U.Grid(), Mom.Grid());
 
       Field& Us = Smearer.get_U(as[level].actions.at(a)->is_smeared);
       double start_force = usecond();
@@ -126,7 +124,7 @@ class Integrator {
       if (as[level].actions.at(a)->is_smeared) Smearer.smeared_force(force);
       force = FieldImplementation::projectForce(force); // Ta for gauge fields
       double end_force = usecond();
-      Real force_abs = std::sqrt(norm2(force)/U._grid->gSites());
+      Real force_abs = std::sqrt(norm2(force)/U.Grid()->gSites());
       std::cout << GridLogIntegrator << "["<<level<<"]["<<a<<"] Force average: " << force_abs << std::endl;
       Mom -= force * ep* HMC_MOMENTUM_DENOMINATOR;; 
       double end_full = usecond();
@@ -139,7 +137,8 @@ class Integrator {
     as[level].apply(update_P_hireps, Representations, Mom, U, ep);
   }
 
-  void update_U(Field& U, double ep) {
+  void update_U(Field& U, double ep) 
+  {
     update_U(P, U, ep);
 
     t_U += ep;
@@ -147,7 +146,8 @@ class Integrator {
     std::cout << GridLogIntegrator << "   " << "[" << fl << "] U " << " dt " << ep << " : t_U " << t_U << std::endl;
   }
   
-  void update_U(MomentaField& Mom, Field& U, double ep) {
+  void update_U(MomentaField& Mom, Field& U, double ep) 
+  {
     // exponential of Mom*U in the gauge fields case
     FieldImplementation::update_field(Mom, U, ep);
 
@@ -160,16 +160,17 @@ class Integrator {
 
   virtual void step(Field& U, int level, int first, int last) = 0;
 
- public:
+public:
   Integrator(GridBase* grid, IntegratorParameters Par,
              ActionSet<Field, RepresentationPolicy>& Aset,
              SmearingPolicy& Sm)
-      : Params(Par),
-        as(Aset),
-        P(grid),
-        levels(Aset.size()),
-        Smearer(Sm),
-        Representations(grid) {
+    : Params(Par),
+      as(Aset),
+      P(grid),
+      levels(Aset.size()),
+      Smearer(Sm),
+      Representations(grid) 
+  {
     t_P.resize(levels, 0.0);
     t_U = 0.0;
     // initialization of smearer delegated outside of Integrator
@@ -179,26 +180,29 @@ class Integrator {
 
   virtual std::string integrator_name() = 0;
 
-  void print_parameters(){
+  void print_parameters()
+  {
     std::cout << GridLogMessage << "[Integrator] Name : "<< integrator_name() << std::endl;
     Params.print_parameters();
   }
 
-  void print_actions(){
-        std::cout << GridLogMessage << ":::::::::::::::::::::::::::::::::::::::::" << std::endl;
-        std::cout << GridLogMessage << "[Integrator] Action summary: "<<std::endl;
-        for (int level = 0; level < as.size(); ++level) {
-                std::cout << GridLogMessage << "[Integrator] ---- Level: "<< level << std::endl;
-                for (int actionID = 0; actionID < as[level].actions.size(); ++actionID) {
-                        std::cout << GridLogMessage << "["<< as[level].actions.at(actionID)->action_name() << "] ID: " << actionID << std::endl;
-                        std::cout << as[level].actions.at(actionID)->LogParameters();
-                }
-        }
-        std::cout << GridLogMessage << ":::::::::::::::::::::::::::::::::::::::::"<< std::endl;
+  void print_actions()
+  {
+    std::cout << GridLogMessage << ":::::::::::::::::::::::::::::::::::::::::" << std::endl;
+    std::cout << GridLogMessage << "[Integrator] Action summary: "<<std::endl;
+    for (int level = 0; level < as.size(); ++level) {
+      std::cout << GridLogMessage << "[Integrator] ---- Level: "<< level << std::endl;
+      for (int actionID = 0; actionID < as[level].actions.size(); ++actionID) {
+	std::cout << GridLogMessage << "["<< as[level].actions.at(actionID)->action_name() << "] ID: " << actionID << std::endl;
+	std::cout << as[level].actions.at(actionID)->LogParameters();
+      }
+    }
+    std::cout << GridLogMessage << ":::::::::::::::::::::::::::::::::::::::::"<< std::endl;
 
   }
 
-  void reverse_momenta(){
+  void reverse_momenta()
+  {
     P *= -1.0;
   }
 
@@ -211,14 +215,15 @@ class Integrator {
       for (int a = 0; a < repr_set.size(); ++a){
         repr_set.at(a)->refresh(Rep.U, pRNG);
       
-      std::cout << GridLogDebug << "Hirep refreshing pseudofermions" << std::endl;
-    }
+	std::cout << GridLogDebug << "Hirep refreshing pseudofermions" << std::endl;
+      }
     }
   } refresh_hireps{};
 
   // Initialization of momenta and actions
-  void refresh(Field& U, GridParallelRNG& pRNG) {
-    assert(P._grid == U._grid);
+  void refresh(Field& U, GridParallelRNG& pRNG) 
+  {
+    assert(P.Grid() == U.Grid());
     std::cout << GridLogIntegrator << "Integrator refresh\n";
 
     FieldImplementation::generate_momenta(P, pRNG);
@@ -262,7 +267,8 @@ class Integrator {
   } S_hireps{};
 
   // Calculate action
-  RealD S(Field& U) {  // here also U not used
+  RealD S(Field& U) 
+  {  // here also U not used
 
     std::cout << GridLogIntegrator << "Integrator action\n";
 
@@ -287,7 +293,8 @@ class Integrator {
     return H;
   }
 
-  void integrate(Field& U) {
+  void integrate(Field& U) 
+  {
     // reset the clocks
     t_U = 0;
     for (int level = 0; level < as.size(); ++level) {
@@ -311,11 +318,9 @@ class Integrator {
 
   }
 
-
-  
-
 };
-}
-}
+
+NAMESPACE_END(Grid);
+
 #endif  // INTEGRATOR_INCLUDED
 
diff --git a/Grid/qcd/hmc/integrators/Integrator_algorithm.h b/Grid/qcd/hmc/integrators/Integrator_algorithm.h
index 47574026..b05c4ea8 100644
--- a/Grid/qcd/hmc/integrators/Integrator_algorithm.h
+++ b/Grid/qcd/hmc/integrators/Integrator_algorithm.h
@@ -39,8 +39,7 @@ directory
 #ifndef INTEGRATOR_ALG_INCLUDED
 #define INTEGRATOR_ALG_INCLUDED
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 /* PAB:
  *
@@ -93,22 +92,17 @@ namespace QCD {
  *  P 1/2                            P 1/2
  */
 
-template <class FieldImplementation, class SmearingPolicy,
-          class RepresentationPolicy =
-              Representations<FundamentalRepresentation> >
-class LeapFrog : public Integrator<FieldImplementation, SmearingPolicy,
-                                   RepresentationPolicy> {
- public:
-  typedef LeapFrog<FieldImplementation, SmearingPolicy, RepresentationPolicy>
-      Algorithm;
+template <class FieldImplementation, class SmearingPolicy, class RepresentationPolicy = Representations<FundamentalRepresentation> >
+class LeapFrog : public Integrator<FieldImplementation, SmearingPolicy, RepresentationPolicy> 
+{
+public:
+  typedef LeapFrog<FieldImplementation, SmearingPolicy, RepresentationPolicy> Algorithm;
   INHERIT_FIELD_TYPES(FieldImplementation);
 
   std::string integrator_name(){return "LeapFrog";}
 
-  LeapFrog(GridBase* grid, IntegratorParameters Par,
-           ActionSet<Field, RepresentationPolicy>& Aset, SmearingPolicy& Sm)
-      : Integrator<FieldImplementation, SmearingPolicy, RepresentationPolicy>(
-            grid, Par, Aset, Sm){};
+  LeapFrog(GridBase* grid, IntegratorParameters Par, ActionSet<Field, RepresentationPolicy>& Aset, SmearingPolicy& Sm)
+    : Integrator<FieldImplementation, SmearingPolicy, RepresentationPolicy>(grid, Par, Aset, Sm){};
 
   void step(Field& U, int level, int _first, int _last) {
     int fl = this->as.size() - 1;
@@ -141,21 +135,17 @@ class LeapFrog : public Integrator<FieldImplementation, SmearingPolicy,
   }
 };
 
-template <class FieldImplementation, class SmearingPolicy,
-          class RepresentationPolicy =
-              Representations<FundamentalRepresentation> >
-class MinimumNorm2 : public Integrator<FieldImplementation, SmearingPolicy,
-                                       RepresentationPolicy> {
- private:
+template <class FieldImplementation, class SmearingPolicy, class RepresentationPolicy = Representations<FundamentalRepresentation> >
+class MinimumNorm2 : public Integrator<FieldImplementation, SmearingPolicy, RepresentationPolicy> 
+{
+private:
   const RealD lambda = 0.1931833275037836;
 
- public:
+public:
   INHERIT_FIELD_TYPES(FieldImplementation);
 
-  MinimumNorm2(GridBase* grid, IntegratorParameters Par,
-               ActionSet<Field, RepresentationPolicy>& Aset, SmearingPolicy& Sm)
-      : Integrator<FieldImplementation, SmearingPolicy, RepresentationPolicy>(
-            grid, Par, Aset, Sm){};
+  MinimumNorm2(GridBase* grid, IntegratorParameters Par, ActionSet<Field, RepresentationPolicy>& Aset, SmearingPolicy& Sm)
+    : Integrator<FieldImplementation, SmearingPolicy, RepresentationPolicy>(grid, Par, Aset, Sm){};
 
   std::string integrator_name(){return "MininumNorm2";}
 
@@ -202,35 +192,32 @@ class MinimumNorm2 : public Integrator<FieldImplementation, SmearingPolicy,
   }
 };
 
-template <class FieldImplementation, class SmearingPolicy,
-          class RepresentationPolicy =
-              Representations<FundamentalRepresentation> >
-class ForceGradient : public Integrator<FieldImplementation, SmearingPolicy,
-                                        RepresentationPolicy> {
- private:
+template <class FieldImplementation, class SmearingPolicy, class RepresentationPolicy = Representations<FundamentalRepresentation> >
+class ForceGradient : public Integrator<FieldImplementation, SmearingPolicy, RepresentationPolicy> 
+{
+private:
   const RealD lambda = 1.0 / 6.0;
-  ;
   const RealD chi = 1.0 / 72.0;
   const RealD xi = 0.0;
   const RealD theta = 0.0;
 
- public:
+public:
   INHERIT_FIELD_TYPES(FieldImplementation);
 
   // Looks like dH scales as dt^4. tested wilson/wilson 2 level.
   ForceGradient(GridBase* grid, IntegratorParameters Par,
                 ActionSet<Field, RepresentationPolicy>& Aset,
                 SmearingPolicy& Sm)
-      : Integrator<FieldImplementation, SmearingPolicy, RepresentationPolicy>(
-            grid, Par, Aset, Sm){};
+    : Integrator<FieldImplementation, SmearingPolicy, RepresentationPolicy>(
+									    grid, Par, Aset, Sm){};
 
   std::string integrator_name(){return "ForceGradient";}
   
   void FG_update_P(Field& U, int level, double fg_dt, double ep) {
-    Field Ufg(U._grid);
-    Field Pfg(U._grid);
+    Field Ufg(U.Grid());
+    Field Pfg(U.Grid());
     Ufg = U;
-    Pfg = zero;
+    Pfg = Zero();
     std::cout << GridLogIntegrator << "FG update " << fg_dt << " " << ep << std::endl;
     // prepare_fg; no prediction/result cache for now
     // could relax CG stopping conditions for the
@@ -284,10 +271,6 @@ class ForceGradient : public Integrator<FieldImplementation, SmearingPolicy,
   }
 };
 
-
-
-
-}
-}
+NAMESPACE_END(Grid);
 
 #endif  // INTEGRATOR_INCLUDED
diff --git a/Grid/qcd/modules/ActionModules.h b/Grid/qcd/modules/ActionModules.h
index fbfbbf33..cef4c036 100644
--- a/Grid/qcd/modules/ActionModules.h
+++ b/Grid/qcd/modules/ActionModules.h
@@ -25,16 +25,16 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef ACTION_MODULES_H
 #define ACTION_MODULES_H
 
-/*
-Define loadable, serializable modules
-for the HMC execution
-*/
+			   /*
+			     Define loadable, serializable modules
+			     for the HMC execution
+			   */
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 //////////////////////////////////////////////
 //              Actions
@@ -51,10 +51,10 @@ public:
 
 template <class ActionType, class APar>
 class ActionModule
-    : public Parametrized<APar>,
-      public ActionModuleBase< QCD::Action<typename ActionType::GaugeField> , QCD::GridModule > {
- public:
-  typedef ActionModuleBase< QCD::Action<typename ActionType::GaugeField>, QCD::GridModule > Base;
+  : public Parametrized<APar>,
+    public ActionModuleBase< Action<typename ActionType::GaugeField> , GridModule > {
+public:
+  typedef ActionModuleBase< Action<typename ActionType::GaugeField>, GridModule > Base;
   typedef typename Base::Product Product;
   typedef APar Parameters;
 
@@ -76,7 +76,7 @@ class ActionModule
     return ActionPtr.get();
   }
 
- private:
+private:
   virtual void initialize() = 0;
 
 };
@@ -85,28 +85,28 @@ class ActionModule
 // Modules
 //////////////////////////
 
-namespace QCD{
+
 
 class PlaqPlusRectangleGaugeActionParameters : Serializable {
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(PlaqPlusRectangleGaugeActionParameters, 
-    RealD, c_plaq,
-    RealD, c_rect);
+				  RealD, c_plaq,
+				  RealD, c_rect);
 
 };
 
 class RBCGaugeActionParameters : Serializable {
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(RBCGaugeActionParameters, 
-    RealD, beta,
-    RealD, c1);
+				  RealD, beta,
+				  RealD, c1);
 
 };
 
 class BetaGaugeActionParameters : Serializable {
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(BetaGaugeActionParameters, 
-    RealD, beta);
+				  RealD, beta);
 };
 
 
@@ -232,17 +232,17 @@ class TwoFlavourFModule: public PseudoFermionModuleBase<Impl, TwoFlavourPseudoFe
   typename Base::operator_type fop_mod;
   typename Base::solver_type   solver_mod;
 
- public:
+public:
   virtual void acquireResource(typename Base::Resource& GridMod){
     fop_mod->AddGridPair(GridMod);
   }
 
-   // constructor
-   template <class ReaderClass>
-   TwoFlavourFModule(Reader<ReaderClass>& R): Base(R) {
+  // constructor
+  template <class ReaderClass>
+  TwoFlavourFModule(Reader<ReaderClass>& R): Base(R) {
     this->getSolverOperator(R, solver_mod, "Solver");
     this->getFermionOperator(R, fop_mod, "Operator");
-   } 
+  } 
 
   // acquire resource
   virtual void initialize() {
@@ -261,17 +261,17 @@ class TwoFlavourEOFModule: public PseudoFermionModuleBase<Impl, TwoFlavourEvenOd
   typename Base::operator_type fop_mod;
   typename Base::solver_type   solver_mod;
 
- public:
+public:
   virtual void acquireResource(typename Base::Resource& GridMod){
     fop_mod->AddGridPair(GridMod);
   }
 
-   // constructor
-   template <class ReaderClass>
-   TwoFlavourEOFModule(Reader<ReaderClass>& R): PseudoFermionModuleBase<Impl, TwoFlavourEvenOddPseudoFermionAction>(R) {
+  // constructor
+  template <class ReaderClass>
+  TwoFlavourEOFModule(Reader<ReaderClass>& R): PseudoFermionModuleBase<Impl, TwoFlavourEvenOddPseudoFermionAction>(R) {
     this->getSolverOperator(R, solver_mod, "Solver");
     this->getFermionOperator(R, fop_mod, "Operator");
-   } 
+  } 
 
   // acquire resource
   virtual void initialize() {
@@ -291,25 +291,25 @@ class TwoFlavourRatioFModule: public PseudoFermionModuleBase<Impl, TwoFlavourRat
   typename Base::operator_type fop_denominator_mod;
   typename Base::solver_type   solver_mod;
 
- public:
+public:
   virtual void acquireResource(typename Base::Resource& GridMod){
     fop_numerator_mod->AddGridPair(GridMod);
     fop_denominator_mod->AddGridPair(GridMod);
   }
 
-   // constructor
-   template <class ReaderClass>
-   TwoFlavourRatioFModule(Reader<ReaderClass>& R): PseudoFermionModuleBase<Impl, TwoFlavourRatioPseudoFermionAction>(R) {
+  // constructor
+  template <class ReaderClass>
+  TwoFlavourRatioFModule(Reader<ReaderClass>& R): PseudoFermionModuleBase<Impl, TwoFlavourRatioPseudoFermionAction>(R) {
     this->getSolverOperator(R, solver_mod, "Solver");
     this->getFermionOperator(R, fop_numerator_mod, "Numerator");
     this->getFermionOperator(R, fop_denominator_mod, "Denominator");
-   } 
+  } 
 
   // acquire resource
   virtual void initialize() {
     // here temporarily assuming that the force and action solver are the same
     this->ActionPtr.reset(new TwoFlavourRatioPseudoFermionAction<Impl>(*(this->fop_numerator_mod->getPtr()), 
-      *(this->fop_denominator_mod->getPtr()), *(this->solver_mod->getPtr()), *(this->solver_mod->getPtr())));
+								       *(this->fop_denominator_mod->getPtr()), *(this->solver_mod->getPtr()), *(this->solver_mod->getPtr())));
   }
 
 };
@@ -323,25 +323,25 @@ class TwoFlavourRatioEOFModule: public PseudoFermionModuleBase<Impl, TwoFlavourE
   typename Base::operator_type fop_denominator_mod;
   typename Base::solver_type   solver_mod;
 
- public:
+public:
   virtual void acquireResource(typename Base::Resource& GridMod){
     fop_numerator_mod->AddGridPair(GridMod);
     fop_denominator_mod->AddGridPair(GridMod);
   }
 
-   // constructor
-   template <class ReaderClass>
-   TwoFlavourRatioEOFModule(Reader<ReaderClass>& R): Base(R) {
+  // constructor
+  template <class ReaderClass>
+  TwoFlavourRatioEOFModule(Reader<ReaderClass>& R): Base(R) {
     this->getSolverOperator(R, solver_mod, "Solver");
     this->getFermionOperator(R, fop_numerator_mod, "Numerator");
     this->getFermionOperator(R, fop_denominator_mod, "Denominator");
-   } 
+  } 
 
   // acquire resource
   virtual void initialize() {
     // here temporarily assuming that the force and action solver are the same
     this->ActionPtr.reset(new TwoFlavourEvenOddRatioPseudoFermionAction<Impl>(*(this->fop_numerator_mod->getPtr()), 
-      *(this->fop_denominator_mod->getPtr()), *(this->solver_mod->getPtr()), *(this->solver_mod->getPtr())));
+									      *(this->fop_denominator_mod->getPtr()), *(this->solver_mod->getPtr()), *(this->solver_mod->getPtr())));
   }
 
 };
@@ -354,16 +354,16 @@ class OneFlavourFModule: public PseudoFermionModuleBase<Impl, OneFlavourRational
 
   typename Base::operator_type fop_mod;
 
- public:
+public:
   virtual void acquireResource(typename Base::Resource& GridMod){
     fop_mod->AddGridPair(GridMod);
   }
 
-   // constructor
-   template <class ReaderClass>
-   OneFlavourFModule(Reader<ReaderClass>& R): Base(R) {
+  // constructor
+  template <class ReaderClass>
+  OneFlavourFModule(Reader<ReaderClass>& R): Base(R) {
     this->getFermionOperator(R, fop_mod, "Operator");
-   } 
+  } 
 
   // acquire resource
   virtual void initialize() {
@@ -375,22 +375,22 @@ class OneFlavourFModule: public PseudoFermionModuleBase<Impl, OneFlavourRational
 template <class Impl >
 class OneFlavourEOFModule: 
   public PseudoFermionModuleBase<Impl, OneFlavourEvenOddRationalPseudoFermionAction, OneFlavourRationalParams>
-  {
+{
   typedef PseudoFermionModuleBase<Impl, OneFlavourEvenOddRationalPseudoFermionAction, OneFlavourRationalParams> Base;
   using Base::Base;
 
   typename Base::operator_type fop_mod;
 
- public:
+public:
   virtual void acquireResource(typename Base::Resource& GridMod){
     fop_mod->AddGridPair(GridMod);
   }
 
-   // constructor
-   template <class ReaderClass>
-   OneFlavourEOFModule(Reader<ReaderClass>& R): Base(R) {
+  // constructor
+  template <class ReaderClass>
+  OneFlavourEOFModule(Reader<ReaderClass>& R): Base(R) {
     this->getFermionOperator(R, fop_mod, "Operator");
-   } 
+  } 
 
   // acquire resource
   virtual void initialize() {
@@ -403,7 +403,7 @@ class OneFlavourEOFModule:
 template <class Impl >
 class OneFlavourRatioFModule: 
   public PseudoFermionModuleBase<Impl, OneFlavourRatioRationalPseudoFermionAction, OneFlavourRationalParams>
-  {
+{
 
   typedef PseudoFermionModuleBase<Impl, OneFlavourRatioRationalPseudoFermionAction, OneFlavourRationalParams> Base;
   using Base::Base;
@@ -411,18 +411,18 @@ class OneFlavourRatioFModule:
   typename Base::operator_type fop_numerator_mod;
   typename Base::operator_type fop_denominator_mod;
 
- public:
+public:
   virtual void acquireResource(typename Base::Resource& GridMod){
     fop_numerator_mod->AddGridPair(GridMod);
     fop_denominator_mod->AddGridPair(GridMod);
   }
 
-   // constructor
-   template <class ReaderClass>
-   OneFlavourRatioFModule(Reader<ReaderClass>& R): Base(R) {
+  // constructor
+  template <class ReaderClass>
+  OneFlavourRatioFModule(Reader<ReaderClass>& R): Base(R) {
     this->getFermionOperator(R, fop_numerator_mod, "Numerator");
     this->getFermionOperator(R, fop_denominator_mod, "Denominator");
-   } 
+  } 
 
   // acquire resource
   virtual void initialize() {
@@ -437,7 +437,7 @@ class OneFlavourRatioFModule:
 template <class Impl >
 class OneFlavourRatioEOFModule: 
   public PseudoFermionModuleBase<Impl, OneFlavourEvenOddRatioRationalPseudoFermionAction, OneFlavourRationalParams>
-  {
+{
 
   typedef PseudoFermionModuleBase<Impl, OneFlavourEvenOddRatioRationalPseudoFermionAction, OneFlavourRationalParams> Base;
   using Base::Base;
@@ -445,18 +445,18 @@ class OneFlavourRatioEOFModule:
   typename Base::operator_type fop_numerator_mod;
   typename Base::operator_type fop_denominator_mod;
 
- public:
+public:
   virtual void acquireResource(typename Base::Resource& GridMod){
     fop_numerator_mod->AddGridPair(GridMod);
     fop_denominator_mod->AddGridPair(GridMod);
   }
 
-   // constructor
-   template <class ReaderClass>
-   OneFlavourRatioEOFModule(Reader<ReaderClass>& R): Base(R) {
+  // constructor
+  template <class ReaderClass>
+  OneFlavourRatioEOFModule(Reader<ReaderClass>& R): Base(R) {
     this->getFermionOperator(R, fop_numerator_mod, "Numerator");
     this->getFermionOperator(R, fop_denominator_mod, "Denominator");
-   } 
+  } 
 
   // acquire resource
   virtual void initialize() {
@@ -467,32 +467,20 @@ class OneFlavourRatioEOFModule:
 
 };
 
-}// QCD temporarily here
-
-
-
-
-
-
-
 ////////////////////////////////////////
 // Factories specialisations
 ////////////////////////////////////////
-
-
-
 // use the same classed defined by Antonin, does not make sense to rewrite
 // Factory is perfectly fine
 // Registar must be changed because I do not want to use the ModuleFactory
-
 // explicit ref to LatticeGaugeField must be changed or put in the factory
-//typedef ActionModuleBase< QCD::Action< QCD::LatticeGaugeField >, QCD::GridModule > HMC_LGTActionModBase;
-//typedef ActionModuleBase< QCD::Action< QCD::LatticeReal >, QCD::GridModule > HMC_ScalarActionModBase;
+//typedef ActionModuleBase< Action< LatticeGaugeField >, GridModule > HMC_LGTActionModBase;
+//typedef ActionModuleBase< Action< LatticeReal >, GridModule > HMC_ScalarActionModBase;
 
 template <char const *str, class Field, class ReaderClass >
 class HMC_ActionModuleFactory
-    : public Factory < ActionModuleBase< QCD::Action< Field >, QCD::GridModule > , Reader<ReaderClass> > {
- public:
+  : public Factory < ActionModuleBase< Action< Field >, GridModule > , Reader<ReaderClass> > {
+public:
   typedef Reader<ReaderClass> TheReader; 
   // use SINGLETON FUNCTOR MACRO HERE
   HMC_ActionModuleFactory(const HMC_ActionModuleFactory& e) = delete;
@@ -502,16 +490,15 @@ class HMC_ActionModuleFactory
     return e;
   }
 
- private:
+private:
   HMC_ActionModuleFactory(void) = default;
-    std::string obj_type() const {
-        return std::string(str);
+  std::string obj_type() const {
+    return std::string(str);
   }
 };
 
-
 extern char gauge_string[];
-} // Grid
 
+NAMESPACE_END(Grid);
 
-#endif //HMC_MODULES_H
\ No newline at end of file
+#endif //HMC_MODULES_H
diff --git a/Grid/qcd/modules/FermionOperatorModules.h b/Grid/qcd/modules/FermionOperatorModules.h
index fc9d96a7..5af1378e 100644
--- a/Grid/qcd/modules/FermionOperatorModules.h
+++ b/Grid/qcd/modules/FermionOperatorModules.h
@@ -25,11 +25,11 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef FERMIONOPERATOR_MODULES_H
 #define FERMIONOPERATOR_MODULES_H
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 ////////////////////////////////////
 //  Fermion operators
@@ -37,19 +37,19 @@ namespace Grid {
 template < class Product>
 class FermionOperatorModuleBase : public HMCModuleBase<Product>{
 public:
-  virtual void AddGridPair(QCD::GridModule&) = 0;
+  virtual void AddGridPair(GridModule&) = 0;
 };
 
 template <template <typename> class FOType, class FermionImpl, class FOPar>
 class FermionOperatorModule
-    : public Parametrized<FOPar>,
-      public FermionOperatorModuleBase<QCD::FermionOperator<FermionImpl> > {
+  : public Parametrized<FOPar>,
+    public FermionOperatorModuleBase<FermionOperator<FermionImpl> > {
 
 protected:
   std::unique_ptr< FOType<FermionImpl> > FOPtr;
-  std::vector< QCD::GridModule* >    GridRefs;
- public:
-  typedef HMCModuleBase< QCD::FermionOperator<FermionImpl> > Base;
+  std::vector< GridModule* >    GridRefs;
+public:
+  typedef HMCModuleBase< FermionOperator<FermionImpl> > Base;
   typedef typename Base::Product Product;
 
   FermionOperatorModule(FOPar Par) : Parametrized<FOPar>(Par) {}
@@ -57,7 +57,7 @@ protected:
   template <class ReaderClass>
   FermionOperatorModule(Reader<ReaderClass>& Reader) : Parametrized<FOPar>(Reader){};
 
-  void AddGridPair(QCD::GridModule &Mod){
+  void AddGridPair(GridModule &Mod){
     if (GridRefs.size()>1){
       std::cout << GridLogError << "Adding too many Grids to the FermionOperatorModule" << std::endl;
       exit(1);
@@ -65,9 +65,9 @@ protected:
     GridRefs.push_back(&Mod);
 
     if (Ls()){
-      GridRefs.push_back(new QCD::GridModule());
-      GridRefs[1]->set_full(QCD::SpaceTimeGrid::makeFiveDimGrid(Ls(),GridRefs[0]->get_full()));
-      GridRefs[1]->set_rb(QCD::SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls(),GridRefs[0]->get_full()));
+      GridRefs.push_back(new GridModule());
+      GridRefs[1]->set_full(SpaceTimeGrid::makeFiveDimGrid(Ls(),GridRefs[0]->get_full()));
+      GridRefs[1]->set_rb(SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls(),GridRefs[0]->get_full()));
     }
   }
 
@@ -85,7 +85,7 @@ protected:
     return FOPtr.get();
   }
 
- private:
+private:
   virtual void initialize() = 0;
 };
 
@@ -94,8 +94,8 @@ protected:
 // Factory
 template <char const *str, class FermionImpl, class ReaderClass >
 class HMC_FermionOperatorModuleFactory
-    : public Factory < FermionOperatorModuleBase<QCD::FermionOperator<FermionImpl> > ,  Reader<ReaderClass> > {
- public:
+  : public Factory < FermionOperatorModuleBase<FermionOperator<FermionImpl> > ,  Reader<ReaderClass> > {
+public:
   // use SINGLETON FUNCTOR MACRO HERE
   typedef Reader<ReaderClass> TheReader;
 
@@ -106,10 +106,10 @@ class HMC_FermionOperatorModuleFactory
     return e;
   }
 
- private:
+private:
   HMC_FermionOperatorModuleFactory(void) = default;
-    std::string obj_type() const {
-        return std::string(str);
+  std::string obj_type() const {
+    return std::string(str);
   }
 };
 
@@ -117,13 +117,13 @@ class HMC_FermionOperatorModuleFactory
 
 
 extern char fermionop_string[];
-namespace QCD{
+
 
 // Modules
 class WilsonFermionParameters : Serializable {
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(WilsonFermionParameters,
-    RealD, mass);
+				  RealD, mass);
 };
 
 
@@ -143,13 +143,13 @@ class WilsonFermionModule: public FermionOperatorModule<WilsonFermion, FermionIm
 
 
 class MobiusFermionParameters : Serializable {
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(MobiusFermionParameters,
-    RealD, mass,
-    RealD, M5,
-    RealD, b,
-    RealD, c,
-    unsigned int, Ls);
+				  RealD, mass,
+				  RealD, M5,
+				  RealD, b,
+				  RealD, c,
+				  unsigned int, Ls);
 };
 
 template <class FermionImpl >
@@ -174,11 +174,11 @@ class MobiusFermionModule: public FermionOperatorModule<MobiusFermion, FermionIm
 
 
 class DomainWallFermionParameters : Serializable {
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(DomainWallFermionParameters,
-    RealD, mass,
-    RealD, M5,
-    unsigned int, Ls);
+				  RealD, mass,
+				  RealD, M5,
+				  unsigned int, Ls);
 };
 
 template <class FermionImpl >
@@ -196,22 +196,22 @@ class DomainWallFermionModule: public FermionOperatorModule<DomainWallFermion, F
     auto GridMod5d = this->GridRefs[1];
     typename FermionImpl::GaugeField U(GridMod->get_full());
     this->FOPtr.reset(new DomainWallFermion<FermionImpl>( U, *(GridMod->get_full()), *(GridMod->get_rb()),
-                                                      *(GridMod5d->get_full()), *(GridMod5d->get_rb()),
-                                                      this->Par_.mass, this->Par_.M5));
+							  *(GridMod5d->get_full()), *(GridMod5d->get_rb()),
+							  this->Par_.mass, this->Par_.M5));
   }
 };
 
 
 class DomainWallEOFAFermionParameters : Serializable {
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(DomainWallEOFAFermionParameters,
-    RealD, mq1,
-    RealD, mq2,
-    RealD, mq3,
-    RealD, shift,
-    int, pm,
-    RealD, M5,
-    unsigned int, Ls);
+				  RealD, mq1,
+				  RealD, mq2,
+				  RealD, mq3,
+				  RealD, shift,
+				  int, pm,
+				  RealD, M5,
+				  unsigned int, Ls);
 };
 
 template <class FermionImpl >
@@ -229,15 +229,12 @@ class DomainWallEOFAFermionModule: public FermionOperatorModule<DomainWallEOFAFe
     auto GridMod5d = this->GridRefs[1];
     typename FermionImpl::GaugeField U(GridMod->get_full());
     this->FOPtr.reset(new DomainWallEOFAFermion<FermionImpl>( U, *(GridMod->get_full()), *(GridMod->get_rb()),
-                                                      *(GridMod5d->get_full()), *(GridMod5d->get_rb()),
-                                                      this->Par_.mq1, this->Par_.mq2, this->Par_.mq3,
-                                                      this->Par_.shift, this->Par_.pm, this->Par_.M5));
+							      *(GridMod5d->get_full()), *(GridMod5d->get_rb()),
+							      this->Par_.mq1, this->Par_.mq2, this->Par_.mq3,
+							      this->Par_.shift, this->Par_.pm, this->Par_.M5));
   }
 };
 
-
-} // QCD
-} // Grid
-
+NAMESPACE_END(Grid);
 
 #endif //FERMIONOPERATOR_MODULES_H
diff --git a/Grid/qcd/modules/Modules.h b/Grid/qcd/modules/Modules.h
index 19290e59..1c1c8889 100644
--- a/Grid/qcd/modules/Modules.h
+++ b/Grid/qcd/modules/Modules.h
@@ -25,23 +25,23 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef HMC_MODULES_H
 #define HMC_MODULES_H
 
-/*
-Define loadable, serializable modules
-for the HMC execution
-*/
+			   /*
+			     Define loadable, serializable modules
+			     for the HMC execution
+			   */
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 // Empty class for no parameters
 class NoParameters{};
 
 
 /*
-Base class for modules with parameters
+  Base class for modules with parameters
 */
 template < class P >
 class Parametrized{
@@ -56,7 +56,7 @@ public:
   }
 
   void set_parameters(Parameters Par){
-        Par_ = Par;
+    Par_ = Par;
   }
 
   void print_parameters(){
@@ -72,7 +72,7 @@ private:
 
 template <>
 class Parametrized<NoParameters>{
-        public:
+public:
   typedef NoParameters Parameters;
 
   Parametrized(Parameters Par){};
@@ -93,7 +93,7 @@ class Parametrized<NoParameters>{
 ////////////////////////////////////////
 template <class Prod>
 class HMCModuleBase {
- public:
+public:
   typedef Prod Product;
 
   virtual Prod* getPtr() = 0;
@@ -110,21 +110,18 @@ class HMCModuleBase {
 
 template <class T, class TheFactory>
 class Registrar {
- public:
+public:
   Registrar(std::string className) {
     // register the class factory function
     TheFactory::getInstance().registerBuilder(className, 
-        [&](typename TheFactory::TheReader Reader)
-        { 
-          return std::unique_ptr<T>(new T(Reader));
-        }
-        );
+					      [&](typename TheFactory::TheReader Reader)
+					      { 
+						return std::unique_ptr<T>(new T(Reader));
+					      }
+					      );
   }
 };
 
+NAMESPACE_END(Grid);
 
-
-}
-
-
-#endif //HMC_MODULES_H
\ No newline at end of file
+#endif //HMC_MODULES_H
diff --git a/Grid/qcd/modules/ObservableModules.h b/Grid/qcd/modules/ObservableModules.h
index fbffc236..87fcbb92 100644
--- a/Grid/qcd/modules/ObservableModules.h
+++ b/Grid/qcd/modules/ObservableModules.h
@@ -25,21 +25,21 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef HMC_OBSERVABLE_MODULES_H
 #define HMC_OBSERVABLE_MODULES_H
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 /////////////////////////////
 // Observables
 /////////////////////////////
 template <class ObservableType, class OPar>
 class ObservableModule
-    : public Parametrized<OPar>,
-      public HMCModuleBase< QCD::HmcObservable<typename ObservableType::Field> > {
- public:
-  typedef HMCModuleBase< QCD::HmcObservable< typename ObservableType::Field> > Base;
+  : public Parametrized<OPar>,
+    public HMCModuleBase< HmcObservable<typename ObservableType::Field> > {
+public:
+  typedef HMCModuleBase< HmcObservable< typename ObservableType::Field> > Base;
   typedef typename Base::Product Product;
   typedef OPar Parameters;
 
@@ -60,7 +60,7 @@ class ObservableModule
     return ObservablePtr.get();
   }
 
- private:
+private:
   virtual void initialize() = 0;
 };
 
@@ -70,13 +70,11 @@ class ObservableModule
 // Modules
 ////////////////
 
-namespace QCD{
-
 //// Observables module
 class PlaquetteObsParameters : Serializable {
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(PlaquetteObsParameters, 
-    std::string, output_prefix);
+				  std::string, output_prefix);
 };
 
 template < class Impl >
@@ -88,7 +86,7 @@ class PlaquetteMod: public ObservableModule<PlaquetteLogger<Impl>, NoParameters>
   virtual void initialize(){
     this->ObservablePtr.reset(new PlaquetteLogger<Impl>());
   }
-  public:
+public:
   PlaquetteMod(): ObsBase(NoParameters()){}
 };
 
@@ -115,25 +113,21 @@ class TopologicalChargeMod: public ObservableModule<TopologicalCharge<Impl>, Top
   virtual void initialize(){
     this->ObservablePtr.reset(new TopologicalCharge<Impl>(this->Par_));
   }
-  public:
+public:
   TopologicalChargeMod(TopologyObsParameters Par): ObsBase(Par){}
   TopologicalChargeMod(): ObsBase(){}
 };
 
-
-}// QCD temporarily here
-
-
 ////////////////////////////////////////
 // Factories specialisations
 ////////////////////////////////////////
 // explicit ref to LatticeGaugeField must be changed or put in the factory
-//typedef HMCModuleBase< QCD::HmcObservable<QCD::LatticeGaugeField> > HMC_ObsModBase;
+//typedef HMCModuleBase< HmcObservable<LatticeGaugeField> > HMC_ObsModBase;
 
 template <char const *str, class Field, class ReaderClass >
 class HMC_ObservablesModuleFactory
-    : public Factory < HMCModuleBase< QCD::HmcObservable<Field> >, Reader<ReaderClass> > {
- public:
+  : public Factory < HMCModuleBase< HmcObservable<Field> >, Reader<ReaderClass> > {
+public:
   typedef Reader<ReaderClass> TheReader; 
   // use SINGLETON FUNCTOR MACRO HERE
   HMC_ObservablesModuleFactory(const HMC_ObservablesModuleFactory& e) = delete;
@@ -143,16 +137,15 @@ class HMC_ObservablesModuleFactory
     return e;
   }
 
- private:
+private:
   HMC_ObservablesModuleFactory(void) = default;
-    std::string obj_type() const {
+  std::string obj_type() const {
     return std::string(str);
   }
 };
 
 extern char observable_string[];
 
-}
+NAMESPACE_END(Grid);
 
-
-#endif //HMC_OBSERVABLE_MODULES_H
\ No newline at end of file
+#endif //HMC_OBSERVABLE_MODULES_H
diff --git a/Grid/qcd/modules/Registration.h b/Grid/qcd/modules/Registration.h
index d31ee22b..ebf8f420 100644
--- a/Grid/qcd/modules/Registration.h
+++ b/Grid/qcd/modules/Registration.h
@@ -25,48 +25,48 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 
 #ifndef MODULES_REGISTRATION_H
 #define MODULES_REGISTRATION_H
 
-// simplify with macros
+			   // simplify with macros
 
-///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-// Actions
-///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-typedef QCD::WilsonGModule<ImplementationPolicy> WilsonGMod;
-typedef QCD::SymanzikGModule<ImplementationPolicy> SymanzikGMod;
-typedef QCD::IwasakiGModule<ImplementationPolicy> IwasakiGMod;
-typedef QCD::DBW2GModule<ImplementationPolicy> DBW2GMod;
-typedef QCD::RBCGModule<ImplementationPolicy> RBCGMod;
-typedef QCD::PlaqPlusRectangleGModule<ImplementationPolicy> PlaqPlusRectangleGMod;
+			   ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+			   // Actions
+			   ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+typedef WilsonGModule<ImplementationPolicy> WilsonGMod;
+typedef SymanzikGModule<ImplementationPolicy> SymanzikGMod;
+typedef IwasakiGModule<ImplementationPolicy> IwasakiGMod;
+typedef DBW2GModule<ImplementationPolicy> DBW2GMod;
+typedef RBCGModule<ImplementationPolicy> RBCGMod;
+typedef PlaqPlusRectangleGModule<ImplementationPolicy> PlaqPlusRectangleGMod;
 
-static Registrar<QCD::WilsonGMod,            HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __WGmodXMLInit("Wilson"); 
-static Registrar<QCD::SymanzikGMod,          HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __SymGmodXMLInit("Symanzik"); 
-static Registrar<QCD::IwasakiGMod,           HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __IwGmodXMLInit("Iwasaki"); 
-static Registrar<QCD::DBW2GMod,              HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __DBW2GmodXMLInit("DBW2"); 
-static Registrar<QCD::RBCGMod,               HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __RBCGmodXMLInit("RBC"); 
-static Registrar<QCD::PlaqPlusRectangleGMod, HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __PPRectGmodXMLInit("PlaqPlusRect"); 
+static Registrar<WilsonGMod,            HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __WGmodXMLInit("Wilson"); 
+static Registrar<SymanzikGMod,          HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __SymGmodXMLInit("Symanzik"); 
+static Registrar<IwasakiGMod,           HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __IwGmodXMLInit("Iwasaki"); 
+static Registrar<DBW2GMod,              HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __DBW2GmodXMLInit("DBW2"); 
+static Registrar<RBCGMod,               HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __RBCGmodXMLInit("RBC"); 
+static Registrar<PlaqPlusRectangleGMod, HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __PPRectGmodXMLInit("PlaqPlusRect"); 
 
 
 // FIXME more general implementation
-static Registrar<QCD::TwoFlavourFModule<FermionImplementationPolicy> ,      
-	HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __TwoFlavourFmodXMLInit("TwoFlavours"); 
-static Registrar<QCD::TwoFlavourRatioFModule<FermionImplementationPolicy> , 
-	HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __TwoFlavourRatioFmodXMLInit("TwoFlavoursRatio"); 
-static Registrar<QCD::TwoFlavourEOFModule<FermionImplementationPolicy> ,    
-	HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __TwoFlavourEOFmodXMLInit("TwoFlavoursEvenOdd"); 
-static Registrar<QCD::TwoFlavourRatioEOFModule<FermionImplementationPolicy>,
-	HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __TwoFlavourRatioEOFmodXMLInit("TwoFlavoursEvenOddRatio"); 
-static Registrar<QCD::OneFlavourFModule<FermionImplementationPolicy> ,      
-	HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __OneFlavourFmodXMLInit("OneFlavour"); 
-static Registrar<QCD::OneFlavourEOFModule<FermionImplementationPolicy> ,    
-	HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __OneFlavourEOFmodXMLInit("OneFlavourEvenOdd"); 
-static Registrar<QCD::OneFlavourRatioFModule<FermionImplementationPolicy> , 
-	HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __OneFlavourRatioFmodXMLInit("OneFlavourRatio"); 
-static Registrar<QCD::OneFlavourRatioEOFModule<FermionImplementationPolicy>,
-	HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __OneFlavourRatioEOFmodXMLInit("OneFlavourEvenOddRatio"); 
+static Registrar<TwoFlavourFModule<FermionImplementationPolicy> ,      
+		 HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __TwoFlavourFmodXMLInit("TwoFlavours"); 
+static Registrar<TwoFlavourRatioFModule<FermionImplementationPolicy> , 
+		 HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __TwoFlavourRatioFmodXMLInit("TwoFlavoursRatio"); 
+static Registrar<TwoFlavourEOFModule<FermionImplementationPolicy> ,    
+		 HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __TwoFlavourEOFmodXMLInit("TwoFlavoursEvenOdd"); 
+static Registrar<TwoFlavourRatioEOFModule<FermionImplementationPolicy>,
+		 HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __TwoFlavourRatioEOFmodXMLInit("TwoFlavoursEvenOddRatio"); 
+static Registrar<OneFlavourFModule<FermionImplementationPolicy> ,      
+		 HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __OneFlavourFmodXMLInit("OneFlavour"); 
+static Registrar<OneFlavourEOFModule<FermionImplementationPolicy> ,    
+		 HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __OneFlavourEOFmodXMLInit("OneFlavourEvenOdd"); 
+static Registrar<OneFlavourRatioFModule<FermionImplementationPolicy> , 
+		 HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __OneFlavourRatioFmodXMLInit("OneFlavourRatio"); 
+static Registrar<OneFlavourRatioEOFModule<FermionImplementationPolicy>,
+		 HMC_ActionModuleFactory<gauge_string, typename ImplementationPolicy::Field, Serialiser> > __OneFlavourRatioEOFmodXMLInit("OneFlavourEvenOddRatio"); 
 
 
 
@@ -78,10 +78,10 @@ static Registrar<QCD::OneFlavourRatioEOFModule<FermionImplementationPolicy>,
 // Now a specific registration with a fermion field
 // here must instantiate CG and CR for every new fermion field type (macro!!)
 
-static Registrar< ConjugateGradientModule<QCD::WilsonFermionR::FermionField>,   
-                  HMC_SolverModuleFactory<solver_string, QCD::WilsonFermionR::FermionField, Serialiser> > __CGWFmodXMLInit("ConjugateGradient"); 
-static Registrar< ConjugateResidualModule<QCD::WilsonFermionR::FermionField>,   
-                  HMC_SolverModuleFactory<solver_string, QCD::WilsonFermionR::FermionField, Serialiser> > __CRWFmodXMLInit("ConjugateResidual"); 
+static Registrar< ConjugateGradientModule<WilsonFermionR::FermionField>,   
+                  HMC_SolverModuleFactory<solver_string, WilsonFermionR::FermionField, Serialiser> > __CGWFmodXMLInit("ConjugateGradient"); 
+static Registrar< ConjugateResidualModule<WilsonFermionR::FermionField>,   
+                  HMC_SolverModuleFactory<solver_string, WilsonFermionR::FermionField, Serialiser> > __CRWFmodXMLInit("ConjugateResidual"); 
 
 // add the staggered, scalar versions here
 
@@ -90,11 +90,11 @@ static Registrar< ConjugateResidualModule<QCD::WilsonFermionR::FermionField>,
 // Fermion operators
 ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
 
-static Registrar< QCD::WilsonFermionModule<FermionImplementationPolicy>,   
+static Registrar< WilsonFermionModule<FermionImplementationPolicy>,   
                   HMC_FermionOperatorModuleFactory<fermionop_string, FermionImplementationPolicy, Serialiser> > __WilsonFOPmodXMLInit("Wilson"); 
-static Registrar< QCD::MobiusFermionModule<FermionImplementationPolicy>,   
+static Registrar< MobiusFermionModule<FermionImplementationPolicy>,   
                   HMC_FermionOperatorModuleFactory<fermionop_string, FermionImplementationPolicy, Serialiser> > __MobiusFOPmodXMLInit("Mobius");
-static Registrar< QCD::DomainWallFermionModule<FermionImplementationPolicy>,   
+static Registrar< DomainWallFermionModule<FermionImplementationPolicy>,   
                   HMC_FermionOperatorModuleFactory<fermionop_string, FermionImplementationPolicy, Serialiser> > __DWFOPmodXMLInit("DomainWall");
 
 
@@ -102,21 +102,19 @@ static Registrar< QCD::DomainWallFermionModule<FermionImplementationPolicy>,
 // Observables
 ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
 
-static Registrar<QCD::PlaquetteMod<ImplementationPolicy>, HMC_ObservablesModuleFactory<observable_string, typename ImplementationPolicy::Field, Serialiser> > __OBSPLmodXMLInit("Plaquette"); 
+static Registrar<PlaquetteMod<ImplementationPolicy>, HMC_ObservablesModuleFactory<observable_string, typename ImplementationPolicy::Field, Serialiser> > __OBSPLmodXMLInit("Plaquette"); 
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
 // Checkpointers
 ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
 
-static Registrar<QCD::BinaryCPModule<ImplementationPolicy>, HMC_CPModuleFactory<cp_string, ImplementationPolicy, Serialiser> > __CPBinarymodXMLInit("Binary");
-static Registrar<QCD::NerscCPModule<ImplementationPolicy> , HMC_CPModuleFactory<cp_string, ImplementationPolicy, Serialiser> > __CPNerscmodXMLInit("Nersc");
+static Registrar<BinaryCPModule<ImplementationPolicy>, HMC_CPModuleFactory<cp_string, ImplementationPolicy, Serialiser> > __CPBinarymodXMLInit("Binary");
+static Registrar<NerscCPModule<ImplementationPolicy> , HMC_CPModuleFactory<cp_string, ImplementationPolicy, Serialiser> > __CPNerscmodXMLInit("Nersc");
 
 #ifdef HAVE_LIME
-static Registrar<QCD::ILDGCPModule<ImplementationPolicy>  , HMC_CPModuleFactory<cp_string, ImplementationPolicy, Serialiser> > __CPILDGmodXMLInit("ILDG");
+static Registrar<ILDGCPModule<ImplementationPolicy>  , HMC_CPModuleFactory<cp_string, ImplementationPolicy, Serialiser> > __CPILDGmodXMLInit("ILDG");
 #endif
 
-
-
 ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
 // Integrators
 ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -126,4 +124,4 @@ static Registrar< HMCForceGradient<ImplementationPolicy, RepresentationPolicy, S
 
 typedef HMCRunnerModuleFactory<hmc_string, Serialiser > HMCModuleFactory;
 
-#endif
\ No newline at end of file
+#endif
diff --git a/Grid/qcd/modules/SolverModules.h b/Grid/qcd/modules/SolverModules.h
index 5efe1b69..65cb91f9 100644
--- a/Grid/qcd/modules/SolverModules.h
+++ b/Grid/qcd/modules/SolverModules.h
@@ -25,11 +25,11 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef SOLVER_MODULES_H
 #define SOLVER_MODULES_H
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 //////////////////////////////////////////////
 //       Operator Functions (Solvers)
@@ -37,9 +37,9 @@ namespace Grid {
 
 template <template <typename> class SolverType, class Field, class SPar>
 class SolverModule
-    : public Parametrized<SPar>,
-      public HMCModuleBase<OperatorFunction<Field> > {
- public:
+  : public Parametrized<SPar>,
+    public HMCModuleBase<OperatorFunction<Field> > {
+public:
   typedef HMCModuleBase< OperatorFunction<Field> > Base;
   typedef typename Base::Product Product;
 
@@ -60,7 +60,7 @@ class SolverModule
     return SolverPtr.get();
   }
 
- private:
+private:
   virtual void initialize() = 0;
 };
 
@@ -68,8 +68,8 @@ class SolverModule
 // Factory
 template <char const *str, class Field, class ReaderClass >
 class HMC_SolverModuleFactory
-    : public Factory < HMCModuleBase<OperatorFunction<Field> > ,  Reader<ReaderClass> > {
- public:
+  : public Factory < HMCModuleBase<OperatorFunction<Field> > ,  Reader<ReaderClass> > {
+public:
   // use SINGLETON FUNCTOR MACRO HERE
   typedef Reader<ReaderClass> TheReader; 
 
@@ -80,20 +80,20 @@ class HMC_SolverModuleFactory
     return e;
   }
 
- private:
+private:
   HMC_SolverModuleFactory(void) = default;
-    std::string obj_type() const {
-        return std::string(str);
+  std::string obj_type() const {
+    return std::string(str);
   }
 };
 
 
 
 class SolverParameters : Serializable {
- public:
+public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(SolverParameters,
-    RealD, tolerance,
-    RealD, max_iterations);
+				  RealD, tolerance,
+				  RealD, max_iterations);
   // add error on no convergence?
 };
 
@@ -101,8 +101,8 @@ class SolverParameters : Serializable {
 class SolverObjName: Serializable {
 public:
   GRID_SERIALIZABLE_CLASS_MEMBERS(SolverObjName, 
-  std::string, name,
-  SolverParameters, parameters);
+				  std::string, name,
+				  SolverParameters, parameters);
 
 };
 
@@ -132,7 +132,7 @@ class ConjugateResidualModule: public SolverModule<ConjugateResidual, Field, Sol
 };
 
 extern char solver_string[];
-} // Grid
 
+NAMESPACE_END(Grid);
 
-#endif //SOLVER_MODULES_H
\ No newline at end of file
+#endif //SOLVER_MODULES_H
diff --git a/Grid/qcd/observables/hmc_observable.h b/Grid/qcd/observables/hmc_observable.h
index fcf11774..c28c376d 100644
--- a/Grid/qcd/observables/hmc_observable.h
+++ b/Grid/qcd/observables/hmc_observable.h
@@ -27,10 +27,9 @@ directory
 *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef HMC_OBSERVABLE_H
-#define HMC_OBSERVABLE_H
+#pragma once
 
-namespace Grid{
+NAMESPACE_BEGIN(Grid);
 
 template <class Field>
 class HmcObservable {
@@ -41,11 +40,14 @@ class HmcObservable {
                                   GridParallelRNG &pRNG) = 0;
 };
 
-}  // namespace Grid
+NAMESPACE_END(Grid);
 
+NAMESPACE_CHECK(HmcObs);
 #include "plaquette.h"
+NAMESPACE_CHECK(Plaq);
 #include "topological_charge.h"
+NAMESPACE_CHECK(Topo);
 #include "polyakov_loop.h"
+NAMESPACE_CHECK(Polyakov);
 
 
-#endif  //  HMC_OBSERVABLE_H
diff --git a/Grid/qcd/observables/plaquette.h b/Grid/qcd/observables/plaquette.h
index c848dcd9..f038e455 100644
--- a/Grid/qcd/observables/plaquette.h
+++ b/Grid/qcd/observables/plaquette.h
@@ -25,18 +25,16 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 
-#ifndef HMC_PLAQUETTE_H
-#define HMC_PLAQUETTE_H
+#pragma once
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 // this is only defined for a gauge theory
 template <class Impl>
 class PlaquetteLogger : public HmcObservable<typename Impl::Field> {
- public:
+public:
   // here forces the Impl to be of gauge fields
   // if not the compiler will complain
   INHERIT_GIMPL_TYPES(Impl);
@@ -54,15 +52,14 @@ class PlaquetteLogger : public HmcObservable<typename Impl::Field> {
     int def_prec = std::cout.precision();
 
     std::cout << GridLogMessage
-        << std::setprecision(std::numeric_limits<Real>::digits10 + 1)
-        << "Plaquette: [ " << traj << " ] "<< plaq << std::endl;
+	      << std::setprecision(std::numeric_limits<Real>::digits10 + 1)
+	      << "Plaquette: [ " << traj << " ] "<< plaq << std::endl;
 
     std::cout.precision(def_prec);
 
   }
 };
 
-}  // namespace QCD
-}  // namespace Grid
+NAMESPACE_END(Grid);
+
 
-#endif  // HMC_PLAQUETTE_H
diff --git a/Grid/qcd/observables/polyakov_loop.h b/Grid/qcd/observables/polyakov_loop.h
index d708b474..0b59f549 100644
--- a/Grid/qcd/observables/polyakov_loop.h
+++ b/Grid/qcd/observables/polyakov_loop.h
@@ -27,11 +27,9 @@ directory
 *************************************************************************************/
 /*  END LEGAL */
 
-#ifndef HMC_POLYAKOV_H
-#define HMC_POLYAKOV_H
+#pragma once
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 // this is only defined for a gauge theory
 template <class Impl>
@@ -62,7 +60,4 @@ class PolyakovLogger : public HmcObservable<typename Impl::Field> {
   }
 };
 
-}  // namespace QCD
-}  // namespace Grid
-
-#endif  // HMC_POLYAKOV_H
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/observables/topological_charge.h b/Grid/qcd/observables/topological_charge.h
index 5af8d77b..4f116496 100644
--- a/Grid/qcd/observables/topological_charge.h
+++ b/Grid/qcd/observables/topological_charge.h
@@ -25,44 +25,42 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 
-#ifndef HMC_TOP_CHARGE_H
-#define HMC_TOP_CHARGE_H
+#pragma once
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 struct TopologySmearingParameters : Serializable {
-    GRID_SERIALIZABLE_CLASS_MEMBERS(TopologySmearingParameters,
-    int, steps,
-    float, step_size,
-    int, meas_interval,
-    float, maxTau);
+  GRID_SERIALIZABLE_CLASS_MEMBERS(TopologySmearingParameters,
+				  int, steps,
+				  float, step_size,
+				  int, meas_interval,
+				  float, maxTau);
 
-    TopologySmearingParameters(int s = 0, float ss = 0.0f, int mi = 0, float mT = 0.0f):
-        steps(s), step_size(ss), meas_interval(mi), maxTau(mT){}
+  TopologySmearingParameters(int s = 0, float ss = 0.0f, int mi = 0, float mT = 0.0f):
+    steps(s), step_size(ss), meas_interval(mi), maxTau(mT){}
 
-    template < class ReaderClass >
-    TopologySmearingParameters(Reader<ReaderClass>& Reader){
-        read(Reader, "Smearing", *this);  
-    }  
+  template < class ReaderClass >
+  TopologySmearingParameters(Reader<ReaderClass>& Reader){
+    read(Reader, "Smearing", *this);  
+  }  
 };
 
 
 
 struct TopologyObsParameters : Serializable {
-    GRID_SERIALIZABLE_CLASS_MEMBERS(TopologyObsParameters,
-      int, interval,
-      bool, do_smearing,
-      TopologySmearingParameters, Smearing);  
+  GRID_SERIALIZABLE_CLASS_MEMBERS(TopologyObsParameters,
+				  int, interval,
+				  bool, do_smearing,
+				  TopologySmearingParameters, Smearing);  
 
-    TopologyObsParameters(int interval = 1, bool smearing = false):
-        interval(interval), Smearing(smearing){}
+  TopologyObsParameters(int interval = 1, bool smearing = false):
+    interval(interval), Smearing(smearing){}
 
-    template <class ReaderClass >
-      TopologyObsParameters(Reader<ReaderClass>& Reader){
-        read(Reader, "TopologyMeasurement", *this);
+  template <class ReaderClass >
+  TopologyObsParameters(Reader<ReaderClass>& Reader){
+    read(Reader, "TopologyMeasurement", *this);
   }
 };
 
@@ -70,53 +68,53 @@ struct TopologyObsParameters : Serializable {
 // this is only defined for a gauge theory
 template <class Impl>
 class TopologicalCharge : public HmcObservable<typename Impl::Field> {
-    TopologyObsParameters Pars;
+  TopologyObsParameters Pars;
 
- public:
-    // here forces the Impl to be of gauge fields
-    // if not the compiler will complain
-    INHERIT_GIMPL_TYPES(Impl);
+public:
+  // here forces the Impl to be of gauge fields
+  // if not the compiler will complain
+  INHERIT_GIMPL_TYPES(Impl);
 
-    // necessary for HmcObservable compatibility
-    typedef typename Impl::Field Field;
+  // necessary for HmcObservable compatibility
+  typedef typename Impl::Field Field;
 
-    TopologicalCharge(int interval = 1, bool do_smearing = false):
-        Pars(interval, do_smearing){}
+  TopologicalCharge(int interval = 1, bool do_smearing = false):
+    Pars(interval, do_smearing){}
     
-    TopologicalCharge(TopologyObsParameters P):Pars(P){
-        std::cout << GridLogDebug << "Creating TopologicalCharge " << std::endl;
-    }
+  TopologicalCharge(TopologyObsParameters P):Pars(P){
+    std::cout << GridLogDebug << "Creating TopologicalCharge " << std::endl;
+  }
 
-    void TrajectoryComplete(int traj,
-                            Field &U,
-                            GridSerialRNG &sRNG,
-                            GridParallelRNG &pRNG) {
+  void TrajectoryComplete(int traj,
+			  Field &U,
+			  GridSerialRNG &sRNG,
+			  GridParallelRNG &pRNG) {
 
     if (traj%Pars.interval == 0){
-        // Smearing
-        Field Usmear = U;
-        int def_prec = std::cout.precision();
+      // Smearing
+      Field Usmear = U;
+      int def_prec = std::cout.precision();
         
-        if (Pars.do_smearing){
-            // using wilson flow by default here
-            WilsonFlow<PeriodicGimplR> WF(Pars.Smearing.steps, Pars.Smearing.step_size, Pars.Smearing.meas_interval);
-            WF.smear_adaptive(Usmear, U, Pars.Smearing.maxTau);
-            Real T0   = WF.energyDensityPlaquette(Usmear);
-            std::cout << GridLogMessage << std::setprecision(std::numeric_limits<Real>::digits10 + 1)
-                      << "T0                : [ " << traj << " ] "<< T0 << std::endl;
-        }
+      if (Pars.do_smearing){
+	// using wilson flow by default here
+	WilsonFlow<PeriodicGimplR> WF(Pars.Smearing.steps, Pars.Smearing.step_size, Pars.Smearing.meas_interval);
+	WF.smear_adaptive(Usmear, U, Pars.Smearing.maxTau);
+	Real T0   = WF.energyDensityPlaquette(Usmear);
+	std::cout << GridLogMessage << std::setprecision(std::numeric_limits<Real>::digits10 + 1)
+		  << "T0                : [ " << traj << " ] "<< T0 << std::endl;
+      }
 
-        Real q    = WilsonLoops<Impl>::TopologicalCharge(Usmear);
-        std::cout << GridLogMessage
-            << std::setprecision(std::numeric_limits<Real>::digits10 + 1)
-            << "Topological Charge: [ " << traj << " ] "<< q << std::endl;
+      Real q    = WilsonLoops<Impl>::TopologicalCharge(Usmear);
+      std::cout << GridLogMessage
+		<< std::setprecision(std::numeric_limits<Real>::digits10 + 1)
+		<< "Topological Charge: [ " << traj << " ] "<< q << std::endl;
 
-        std::cout.precision(def_prec);
-        }
+      std::cout.precision(def_prec);
     }
+  }
 
 };
-}
-}
 
-#endif  //  HMC_TOP_CHARGE_H
+NAMESPACE_END(Grid);
+
+
diff --git a/Grid/qcd/representations/adjoint.h b/Grid/qcd/representations/adjoint.h
index 052cd7a8..ee54b465 100644
--- a/Grid/qcd/representations/adjoint.h
+++ b/Grid/qcd/representations/adjoint.h
@@ -1,23 +1,22 @@
 /*
  *  Policy classes for the HMC
  *  Author: Guido Cossu
-*/
+ */
 
 #ifndef ADJOINT_H
 #define ADJOINT_H
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 /*
-* This is an helper class for the HMC
-* Should contain only the data for the adjoint representation
-* and the facility to convert from the fundamental -> adjoint
-*/
+ * This is an helper class for the HMC
+ * Should contain only the data for the adjoint representation
+ * and the facility to convert from the fundamental -> adjoint
+ */
 
 template <int ncolour>
 class AdjointRep {
- public:
+public:
   // typdef to be used by the Representations class in HMC to get the
   // types for the higher representation fields
   typedef typename SU_Adjoint<ncolour>::LatticeAdjMatrix LatticeMatrix;
@@ -38,13 +37,13 @@ class AdjointRep {
     // where t^a is the generator in the fundamental
     // T_F is 1/2 for the fundamental representation
     conformable(U, Uin);
-    U = zero;
-    LatticeColourMatrix tmp(Uin._grid);
+    U = Zero();
+    LatticeColourMatrix tmp(Uin.Grid());
 
     Vector<typename SU<ncolour>::Matrix> ta(Dimension);
 
     // Debug lines
-    // LatticeMatrix uno(Uin._grid);
+    // LatticeMatrix uno(Uin.Grid());
     // uno = 1.0;
     ////////////////
 
@@ -64,29 +63,29 @@ class AdjointRep {
       // Check matrix U_mu, must be real orthogonal
       // reality
       /*
-      LatticeMatrix Ucheck = U_mu - conjugate(U_mu);
-      std::cout << GridLogMessage << "Reality check: " << norm2(Ucheck) <<
-      std::endl;
+	LatticeMatrix Ucheck = U_mu - conjugate(U_mu);
+	std::cout << GridLogMessage << "Reality check: " << norm2(Ucheck) <<
+	std::endl;
 
-      Ucheck = U_mu * adj(U_mu) - uno;
-      std::cout << GridLogMessage << "orthogonality check: " << norm2(Ucheck) <<
-      std::endl;
+	Ucheck = U_mu * adj(U_mu) - uno;
+	std::cout << GridLogMessage << "orthogonality check: " << norm2(Ucheck) <<
+	std::endl;
       */
     }
   }
 
   LatticeGaugeField RtoFundamentalProject(const LatticeField &in,
                                           Real scale = 1.0) const {
-    LatticeGaugeField out(in._grid);
-    out = zero;
+    LatticeGaugeField out(in.Grid());
+    out = Zero();
 
     for (int mu = 0; mu < Nd; mu++) {
-      LatticeColourMatrix out_mu(in._grid);  // fundamental representation
+      LatticeColourMatrix out_mu(in.Grid());  // fundamental representation
       LatticeMatrix in_mu = peekLorentz(in, mu);
 
-      out_mu = zero;
+      out_mu = Zero();
 
-      typename SU<ncolour>::LatticeAlgebraVector h(in._grid);
+      typename SU<ncolour>::LatticeAlgebraVector h(in.Grid());
       projectOnAlgebra(h, in_mu, double(Nc) * 2.0);  // factor C(r)/C(fund)
       FundamentalLieAlgebraMatrix(h, out_mu);   // apply scale only once
       pokeLorentz(out, out_mu, mu);
@@ -96,21 +95,21 @@ class AdjointRep {
     return out;
   }
 
- private:
+private:
   void projectOnAlgebra(typename SU<ncolour>::LatticeAlgebraVector &h_out,
                         const LatticeMatrix &in, Real scale = 1.0) const {
     SU_Adjoint<ncolour>::projectOnAlgebra(h_out, in, scale);
   }
 
   void FundamentalLieAlgebraMatrix(
-      typename SU<ncolour>::LatticeAlgebraVector &h,
-      typename SU<ncolour>::LatticeMatrix &out, Real scale = 1.0) const {
+				   typename SU<ncolour>::LatticeAlgebraVector &h,
+				   typename SU<ncolour>::LatticeMatrix &out, Real scale = 1.0) const {
     SU<ncolour>::FundamentalLieAlgebraMatrix(h, out, scale);
   }
 };
 
 typedef AdjointRep<Nc> AdjointRepresentation;
-}
-}
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/representations/fundamental.h b/Grid/qcd/representations/fundamental.h
index 9f039a07..4d4779e1 100644
--- a/Grid/qcd/representations/fundamental.h
+++ b/Grid/qcd/representations/fundamental.h
@@ -1,23 +1,21 @@
 /*
  *	Policy classes for the HMC
  *	Author: Guido Cossu
-*/	
+ */	
 
 #ifndef FUNDAMENTAL_H
 #define FUNDAMENTAL_H
 
-
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 /*
-* This is an helper class for the HMC
-* Empty since HMC updates already the fundamental representation 
-*/
+ * This is an helper class for the HMC
+ * Empty since HMC updates already the fundamental representation 
+ */
 
 template <int ncolour>
 class FundamentalRep {
- public:
+public:
   static const int Dimension = ncolour;
   static const bool isFundamental = true;
 
@@ -35,25 +33,20 @@ class FundamentalRep {
 
 };
 
-  template<class Field> 
-  class EmptyRep {
-  public:
-    typedef Field LatticeField;
+template<class Field> 
+class EmptyRep {
+public:
+  typedef Field LatticeField;
 
-    explicit EmptyRep(GridBase* grid) {} //do nothing
-    void update_representation(const LatticeField& Uin) {} // do nothing
-    LatticeField RtoFundamentalProject(const LatticeField& in, Real scale = 1.0) const{}// do nothing
-  };
+  explicit EmptyRep(GridBase* grid) {} //do nothing
+  void update_representation(const LatticeField& Uin) {} // do nothing
+  LatticeField RtoFundamentalProject(const LatticeField& in, Real scale = 1.0) const{}// do nothing
+};
     
 
   
 typedef	 FundamentalRep<Nc> FundamentalRepresentation;
 
-  
-}
-}
-
-
-
+NAMESPACE_END(Grid);  
 
 #endif
diff --git a/Grid/qcd/representations/hmc_types.h b/Grid/qcd/representations/hmc_types.h
index 3fee377e..5e4ab020 100644
--- a/Grid/qcd/representations/hmc_types.h
+++ b/Grid/qcd/representations/hmc_types.h
@@ -9,17 +9,15 @@
 #include <tuple>
 #include <utility>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 // Supported types
 // enum {Fundamental, Adjoint} repr_type;
-
 // Utility to add support to the HMC for representations other than the
 // fundamental
 template <class... Reptypes>
 class Representations {
- public:
+public:
   typedef std::tuple<Reptypes...> Representation_type;
 
   // Size of the tuple, known at compile time
@@ -47,11 +45,11 @@ class Representations {
   
   template <std::size_t I = 0>
   inline typename std::enable_if<(I == tuple_size), void>::type update(
-      LatticeSourceField& U) {}
+								       LatticeSourceField& U) {}
 
   template <std::size_t I = 0>
   inline typename std::enable_if<(I < tuple_size), void>::type update(
-      LatticeSourceField& U) {
+								      LatticeSourceField& U) {
     std::get<I>(rep).update_representation(U);
     update<I + 1>(U);
   }
@@ -77,7 +75,7 @@ struct AccessTypes : AccessTypes<A, TupleClass, N - 1, N - 1, S...> {};
 
 template <template <typename> class A, class TupleClass, size_t... S>
 struct AccessTypes<A, TupleClass, 0, S...> {
- public:
+public:
   typedef typename TupleClass::Representation_Fields Rfields;
 
   template <std::size_t N>
@@ -96,7 +94,7 @@ struct AccessTypes<A, TupleClass, 0, S...> {
               << "\n";
   }
 };
-}
-}
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/representations/two_index.h b/Grid/qcd/representations/two_index.h
index 2c7e8b3a..8dfb5561 100644
--- a/Grid/qcd/representations/two_index.h
+++ b/Grid/qcd/representations/two_index.h
@@ -1,13 +1,12 @@
 /*
  *  Policy classes for the HMC
  *  Authors: Guido Cossu, David Preti
-*/
+ */
 
 #ifndef SUN2INDEX_H_H
 #define SUN2INDEX_H_H
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 /*
  * This is an helper class for the HMC
@@ -23,7 +22,7 @@ namespace QCD {
 
 template <int ncolour, TwoIndexSymmetry S>
 class TwoIndexRep {
- public:
+public:
   // typdef to be used by the Representations class in HMC to get the
   // types for the higher representation fields
   typedef typename SU_TwoIndex<ncolour, S>::LatticeTwoIndexMatrix LatticeMatrix;
@@ -41,8 +40,8 @@ class TwoIndexRep {
     // get the U in TwoIndexRep
     // (U)_{(ij)(lk)} = tr [ adj(e^(ij)) U e^(lk) transpose(U) ]
     conformable(U, Uin);
-    U = zero;
-    LatticeColourMatrix tmp(Uin._grid);
+    U = Zero();
+    LatticeColourMatrix tmp(Uin.Grid());
 
     Vector<typename SU<ncolour>::Matrix> eij(Dimension);
 
@@ -63,16 +62,16 @@ class TwoIndexRep {
 
   LatticeGaugeField RtoFundamentalProject(const LatticeField &in,
                                           Real scale = 1.0) const {
-    LatticeGaugeField out(in._grid);
-    out = zero;
+    LatticeGaugeField out(in.Grid());
+    out = Zero();
 
     for (int mu = 0; mu < Nd; mu++) {
-      LatticeColourMatrix out_mu(in._grid);  // fundamental representation
+      LatticeColourMatrix out_mu(in.Grid());  // fundamental representation
       LatticeMatrix in_mu = peekLorentz(in, mu);
 
-      out_mu = zero;
+      out_mu = Zero();
 
-      typename SU<ncolour>::LatticeAlgebraVector h(in._grid);
+      typename SU<ncolour>::LatticeAlgebraVector h(in.Grid());
       projectOnAlgebra(h, in_mu, double(Nc + 2 * S));  // factor T(r)/T(fund)
       FundamentalLieAlgebraMatrix(h, out_mu);          // apply scale only once
       pokeLorentz(out, out_mu, mu);
@@ -80,21 +79,22 @@ class TwoIndexRep {
     return out;
   }
 
- private:
+private:
   void projectOnAlgebra(typename SU<ncolour>::LatticeAlgebraVector &h_out,
                         const LatticeMatrix &in, Real scale = 1.0) const {
     SU_TwoIndex<ncolour, S>::projectOnAlgebra(h_out, in, scale);
   }
 
   void FundamentalLieAlgebraMatrix(
-      typename SU<ncolour>::LatticeAlgebraVector &h,
-      typename SU<ncolour>::LatticeMatrix &out, Real scale = 1.0) const {
+				   typename SU<ncolour>::LatticeAlgebraVector &h,
+				   typename SU<ncolour>::LatticeMatrix &out, Real scale = 1.0) const {
     SU<ncolour>::FundamentalLieAlgebraMatrix(h, out, scale);
   }
 };
 
 typedef TwoIndexRep<Nc, Symmetric> TwoIndexSymmetricRepresentation;
 typedef TwoIndexRep<Nc, AntiSymmetric> TwoIndexAntiSymmetricRepresentation;
-}
-}
+
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/qcd/smearing/APEsmearing.h b/Grid/qcd/smearing/APEsmearing.h
index d3fe94f6..bec97ee4 100644
--- a/Grid/qcd/smearing/APEsmearing.h
+++ b/Grid/qcd/smearing/APEsmearing.h
@@ -25,134 +25,128 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
-/*!
-  @brief Declaration of Smear_APE class for APE smearing
-*/
+			   /*  END LEGAL */
+			   /*!
+			     @brief Declaration of Smear_APE class for APE smearing
+			   */
 
-#ifndef APE_SMEAR_
-#define APE_SMEAR_
+#pragma once
 
-  namespace Grid {
-  	namespace QCD {
+NAMESPACE_BEGIN(Grid);
+
+/*!  @brief APE type smearing of link variables. */
+template <class Gimpl>
+class Smear_APE: public Smear<Gimpl>{
+private:
+  const std::vector<double> rho;/*!< Array of weights */
+
+  //This member must be private - we do not want to control from outside
+  std::vector<double> set_rho(const double common_rho) const {
+    std::vector<double> res;
+
+    for(int mn=0; mn<Nd*Nd; ++mn) res.push_back(common_rho);
+    for(int mu=0; mu<Nd; ++mu) res[mu + mu*Nd] = 0.0;
+    return res;
+  }
+
+public:
+  // Defines the gauge field types
+  INHERIT_GIMPL_TYPES(Gimpl)
 
 
-    /*!  @brief APE type smearing of link variables. */
-    template <class Gimpl>
-  		class Smear_APE: public Smear<Gimpl>{
-  		private:
-      	const std::vector<double> rho;/*!< Array of weights */
+  // Constructors and destructors
+  Smear_APE(const std::vector<double>& rho_):rho(rho_){} // check vector size
+  Smear_APE(double rho_val):rho(set_rho(rho_val)){}
+  Smear_APE():rho(set_rho(1.0)){}
+  ~Smear_APE(){}
 
-//This member must be private - we do not want to control from outside
-  			std::vector<double> set_rho(const double common_rho) const {
-  				std::vector<double> res;
+  ///////////////////////////////////////////////////////////////////////////////
+  void smear(GaugeField& u_smr, const GaugeField& U)const{
+    GridBase *grid = U.Grid();
+    GaugeLinkField Cup(grid), tmp_stpl(grid);
+    WilsonLoops<Gimpl> WL;
+    u_smr = Zero();
 
-  				for(int mn=0; mn<Nd*Nd; ++mn) res.push_back(common_rho);
-  					for(int mu=0; mu<Nd; ++mu) res[mu + mu*Nd] = 0.0;
-  						return res;
-  				}
+    for(int mu=0; mu<Nd; ++mu){
+      Cup = Zero();
+      for(int nu=0; nu<Nd; ++nu){
+	if (nu != mu) {
+	  // get the staple in direction mu, nu
+	  WL.Staple(tmp_stpl, U, mu, nu);  //nb staple conventions of IroIro and Grid differ by a dagger
+	  Cup += tmp_stpl*rho[mu + Nd * nu];
+	}
+      }
+      // save the Cup link-field on the u_smr gauge-field
+      pokeLorentz(u_smr, adj(Cup), mu); // u_smr[mu] = Cup^dag   see conventions for Staple
+    }
+  }
 
-  			public:
-      // Defines the gauge field types
-  				INHERIT_GIMPL_TYPES(Gimpl)
+  ////////////////////////////////////////////////////////////////////////////////
+  void derivative(GaugeField& SigmaTerm,
+		  const GaugeField& iLambda,
+		  const GaugeField& U)const{
 
-
-      // Constructors and destructors
-  				Smear_APE(const std::vector<double>& rho_):rho(rho_){} // check vector size
-  				Smear_APE(double rho_val):rho(set_rho(rho_val)){}
-  				Smear_APE():rho(set_rho(1.0)){}
-  				~Smear_APE(){}
-
-      ///////////////////////////////////////////////////////////////////////////////
-  				void smear(GaugeField& u_smr, const GaugeField& U)const{
-  					GridBase *grid = U._grid;
-  					GaugeLinkField Cup(grid), tmp_stpl(grid);
-  					WilsonLoops<Gimpl> WL;
-  					u_smr = zero;
-
-  					for(int mu=0; mu<Nd; ++mu){
-  						Cup = zero;
-  						for(int nu=0; nu<Nd; ++nu){
-  							if (nu != mu) {
-  								// get the staple in direction mu, nu
-	      						WL.Staple(tmp_stpl, U, mu, nu);  //nb staple conventions of IroIro and Grid differ by a dagger
-	      						Cup += tmp_stpl*rho[mu + Nd * nu];
-	      					}
-	      				}
-	  					// save the Cup link-field on the u_smr gauge-field
-	  					pokeLorentz(u_smr, adj(Cup), mu); // u_smr[mu] = Cup^dag   see conventions for Staple
-	  				}
-	  			}
-
-////////////////////////////////////////////////////////////////////////////////
-	  			void derivative(GaugeField& SigmaTerm,
-	  				const GaugeField& iLambda,
-	  				const GaugeField& U)const{
-
-	// Reference
-	// Morningstar, Peardon, Phys.Rev.D69,054501(2004)
-	// Equation 75
+    // Reference
+    // Morningstar, Peardon, Phys.Rev.D69,054501(2004)
+    // Equation 75
     // Computing Sigma_mu, derivative of S[fat links] with respect to the thin links
     // Output SigmaTerm
 
-	  				GridBase *grid = U._grid;
+    GridBase *grid = U.Grid();
 
-	  				WilsonLoops<Gimpl> WL;
-	  				GaugeLinkField staple(grid), u_tmp(grid);
-	  				GaugeLinkField iLambda_mu(grid), iLambda_nu(grid);
-	  				GaugeLinkField U_mu(grid), U_nu(grid);
-	  				GaugeLinkField sh_field(grid), temp_Sigma(grid);
-	  				Real rho_munu, rho_numu;
+    WilsonLoops<Gimpl> WL;
+    GaugeLinkField staple(grid), u_tmp(grid);
+    GaugeLinkField iLambda_mu(grid), iLambda_nu(grid);
+    GaugeLinkField U_mu(grid), U_nu(grid);
+    GaugeLinkField sh_field(grid), temp_Sigma(grid);
+    Real rho_munu, rho_numu;
 
-	  				for(int mu = 0; mu < Nd; ++mu){
-	  					U_mu       = peekLorentz(      U, mu);
-	  					iLambda_mu = peekLorentz(iLambda, mu);
+    for(int mu = 0; mu < Nd; ++mu){
+      U_mu       = peekLorentz(      U, mu);
+      iLambda_mu = peekLorentz(iLambda, mu);
 
-	  					for(int nu = 0; nu < Nd; ++nu){
-	  						if(nu==mu) continue;
-	  						U_nu       = peekLorentz(      U, nu);
-	  						iLambda_nu = peekLorentz(iLambda, nu);
+      for(int nu = 0; nu < Nd; ++nu){
+	if(nu==mu) continue;
+	U_nu       = peekLorentz(      U, nu);
+	iLambda_nu = peekLorentz(iLambda, nu);
 
-	  						rho_munu = rho[mu + Nd * nu];
-	  						rho_numu = rho[nu + Nd * mu];
+	rho_munu = rho[mu + Nd * nu];
+	rho_numu = rho[nu + Nd * mu];
 
-	  						WL.StapleUpper(staple, U, mu, nu);
+	WL.StapleUpper(staple, U, mu, nu);
 
-	  						temp_Sigma = -rho_numu*staple*iLambda_nu;  //ok
-	        				//-r_numu*U_nu(x+mu)*Udag_mu(x+nu)*Udag_nu(x)*Lambda_nu(x)
-	  						Gimpl::AddLink(SigmaTerm, temp_Sigma, mu);
+	temp_Sigma = -rho_numu*staple*iLambda_nu;  //ok
+	//-r_numu*U_nu(x+mu)*Udag_mu(x+nu)*Udag_nu(x)*Lambda_nu(x)
+	Gimpl::AddLink(SigmaTerm, temp_Sigma, mu);
 
-	    					sh_field = Cshift(iLambda_nu, mu, 1);// general also for Gparity?
+	sh_field = Cshift(iLambda_nu, mu, 1);// general also for Gparity?
 
-	    					temp_Sigma = rho_numu*sh_field*staple; //ok
-	    					//r_numu*Lambda_nu(mu)*U_nu(x+mu)*Udag_mu(x+nu)*Udag_nu(x)
-	    					Gimpl::AddLink(SigmaTerm, temp_Sigma, mu);
+	temp_Sigma = rho_numu*sh_field*staple; //ok
+	//r_numu*Lambda_nu(mu)*U_nu(x+mu)*Udag_mu(x+nu)*Udag_nu(x)
+	Gimpl::AddLink(SigmaTerm, temp_Sigma, mu);
 
-	    					sh_field = Cshift(iLambda_mu, nu, 1);
+	sh_field = Cshift(iLambda_mu, nu, 1);
 
-	    					temp_Sigma = -rho_munu*staple*U_nu*sh_field*adj(U_nu); //ok
-	    					//-r_munu*U_nu(x+mu)*Udag_mu(x+nu)*Lambda_mu(x+nu)*Udag_nu(x)
-	    					Gimpl::AddLink(SigmaTerm, temp_Sigma, mu);
+	temp_Sigma = -rho_munu*staple*U_nu*sh_field*adj(U_nu); //ok
+	//-r_munu*U_nu(x+mu)*Udag_mu(x+nu)*Lambda_mu(x+nu)*Udag_nu(x)
+	Gimpl::AddLink(SigmaTerm, temp_Sigma, mu);
 
-	    					staple = zero;
-	    					sh_field = Cshift(U_nu, mu, 1);
+	staple = Zero();
+	sh_field = Cshift(U_nu, mu, 1);
 
-	    					temp_Sigma = -rho_munu*adj(sh_field)*adj(U_mu)*iLambda_mu*U_nu;
-	    					temp_Sigma += rho_numu*adj(sh_field)*adj(U_mu)*iLambda_nu*U_nu;
+	temp_Sigma = -rho_munu*adj(sh_field)*adj(U_mu)*iLambda_mu*U_nu;
+	temp_Sigma += rho_numu*adj(sh_field)*adj(U_mu)*iLambda_nu*U_nu;
 
-	    					u_tmp = adj(U_nu)*iLambda_nu;
-	    					sh_field = Cshift(u_tmp, mu, 1);
-	    					temp_Sigma += -rho_numu*sh_field*adj(U_mu)*U_nu;
-	    					sh_field = Cshift(temp_Sigma, nu, -1);
-	    					Gimpl::AddLink(SigmaTerm, sh_field, mu);
+	u_tmp = adj(U_nu)*iLambda_nu;
+	sh_field = Cshift(u_tmp, mu, 1);
+	temp_Sigma += -rho_numu*sh_field*adj(U_mu)*U_nu;
+	sh_field = Cshift(temp_Sigma, nu, -1);
+	Gimpl::AddLink(SigmaTerm, sh_field, mu);
 
-	    				}
-	    			}
-	    		}
-	    	};
+      }
+    }
+  }
+};
 
+NAMESPACE_END(Grid);
 
-
-  }// namespace QCD
-}//namespace Grid
-#endif
diff --git a/Grid/qcd/smearing/BaseSmearing.h b/Grid/qcd/smearing/BaseSmearing.h
index f356cfd5..163da275 100644
--- a/Grid/qcd/smearing/BaseSmearing.h
+++ b/Grid/qcd/smearing/BaseSmearing.h
@@ -29,9 +29,8 @@ directory
 /*
   @brief Declares base smearing class Smear
  */
-#ifndef BASE_SMEAR_
-#define BASE_SMEAR_
-
+#pragma once
+NAMESPACE_BEGIN(Grid);
 template <class Gimpl>
 class Smear{
 public:
@@ -41,4 +40,5 @@ public:
   virtual void smear     (GaugeField&,const GaugeField&)const = 0;
   virtual void derivative(GaugeField&, const GaugeField&,const GaugeField&) const = 0;
 };
-#endif
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/smearing/GaugeConfiguration.h b/Grid/qcd/smearing/GaugeConfiguration.h
index 6fea875b..f4d00c72 100644
--- a/Grid/qcd/smearing/GaugeConfiguration.h
+++ b/Grid/qcd/smearing/GaugeConfiguration.h
@@ -3,31 +3,26 @@
 
   @brief Declares the GaugeConfiguration class
 */
-#ifndef GAUGE_CONFIG_
-#define GAUGE_CONFIG_
+#pragma once
 
-namespace Grid
-{
-
-namespace QCD
-{
+NAMESPACE_BEGIN(Grid);
 
 //trivial class for no smearing
-template <class Impl>
+template< class Impl >
 class NoSmearing
 {
 public:
   INHERIT_FIELD_TYPES(Impl);
 
-  Field *ThinField;
+  Field* ThinField;
 
-  NoSmearing() : ThinField(NULL) {}
+  NoSmearing(): ThinField(NULL) {}
 
-  void set_Field(Field &U) { ThinField = &U; }
+  void set_Field(Field& U) { ThinField = &U; }
 
-  void smeared_force(Field &) const {}
+  void smeared_force(Field&) const {}
 
-  Field &get_SmearedU() { return *ThinField; }
+  Field& get_SmearedU() { return *ThinField; }
 
   Field &get_U(bool smeared = false)
   {
@@ -61,7 +56,7 @@ private:
   //====================================================================
   void fill_smearedSet(GaugeField &U)
   {
-    ThinLinks = &U; // attach the smearing routine to the field U
+    ThinLinks = &U;  // attach the smearing routine to the field U
 
     // check the pointer is not null
     if (ThinLinks == NULL)
@@ -72,7 +67,7 @@ private:
     {
       std::cout << GridLogDebug
                 << "[SmearedConfiguration] Filling SmearedSet\n";
-      GaugeField previous_u(ThinLinks->_grid);
+      GaugeField previous_u(ThinLinks->Grid());
 
       previous_u = *ThinLinks;
       for (int smearLvl = 0; smearLvl < smearingLevels; ++smearLvl)
@@ -88,10 +83,10 @@ private:
     }
   }
   //====================================================================
-  GaugeField AnalyticSmearedForce(const GaugeField &SigmaKPrime,
-                                  const GaugeField &GaugeK) const
+  GaugeField AnalyticSmearedForce(const GaugeField& SigmaKPrime,
+                                  const GaugeField& GaugeK) const 
   {
-    GridBase *grid = GaugeK._grid;
+    GridBase* grid = GaugeK.Grid();
     GaugeField C(grid), SigmaK(grid), iLambda(grid);
     GaugeLinkField iLambda_mu(grid);
     GaugeLinkField iQ(grid), e_iQ(grid);
@@ -99,8 +94,8 @@ private:
     GaugeLinkField GaugeKmu(grid), Cmu(grid);
 
     StoutSmearing.BaseSmear(C, GaugeK);
-    SigmaK = zero;
-    iLambda = zero;
+    SigmaK = Zero();
+    iLambda = Zero();
 
     for (int mu = 0; mu < Nd; mu++)
     {
@@ -113,7 +108,7 @@ private:
       pokeLorentz(iLambda, iLambda_mu, mu);
     }
     StoutSmearing.derivative(SigmaK, iLambda,
-                             GaugeK); // derivative of SmearBase
+                             GaugeK);  // derivative of SmearBase
     return SigmaK;
   }
 
@@ -124,11 +119,11 @@ private:
   }
 
   //====================================================================
-  void set_iLambda(GaugeLinkField &iLambda, GaugeLinkField &e_iQ,
-                   const GaugeLinkField &iQ, const GaugeLinkField &Sigmap,
-                   const GaugeLinkField &GaugeK) const
+  void set_iLambda(GaugeLinkField& iLambda, GaugeLinkField& e_iQ,
+                   const GaugeLinkField& iQ, const GaugeLinkField& Sigmap,
+                   const GaugeLinkField& GaugeK) const 
   {
-    GridBase *grid = iQ._grid;
+    GridBase* grid = iQ.Grid();
     GaugeLinkField iQ2(grid), iQ3(grid), B1(grid), B2(grid), USigmap(grid);
     GaugeLinkField unity(grid);
     unity = 1.0;
@@ -141,7 +136,7 @@ private:
     LatticeComplex r01(grid), r11(grid), r21(grid), r02(grid), r12(grid);
     LatticeComplex r22(grid), tr1(grid), tr2(grid);
     LatticeComplex b10(grid), b11(grid), b12(grid), b20(grid), b21(grid),
-        b22(grid);
+      b22(grid);
     LatticeComplex LatticeUnitComplex(grid);
 
     LatticeUnitComplex = 1.0;
@@ -165,19 +160,19 @@ private:
     e2iu = cos(2.0 * u) + timesI(sin(2.0 * u));
 
     r01 = (2.0 * u + timesI(2.0 * (u2 - w2))) * e2iu +
-          emiu * ((16.0 * u * cosw + 2.0 * u * (3.0 * u2 + w2) * xi0) +
-                  timesI(-8.0 * u2 * cosw + 2.0 * (9.0 * u2 + w2) * xi0));
+      emiu * ((16.0 * u * cosw + 2.0 * u * (3.0 * u2 + w2) * xi0) +
+	      timesI(-8.0 * u2 * cosw + 2.0 * (9.0 * u2 + w2) * xi0));
 
     r11 = (2.0 * LatticeUnitComplex + timesI(4.0 * u)) * e2iu +
-          emiu * ((-2.0 * cosw + (3.0 * u2 - w2) * xi0) +
-                  timesI((2.0 * u * cosw + 6.0 * u * xi0)));
+      emiu * ((-2.0 * cosw + (3.0 * u2 - w2) * xi0) +
+	      timesI((2.0 * u * cosw + 6.0 * u * xi0)));
 
     r21 =
-        2.0 * timesI(e2iu) + emiu * (-3.0 * u * xi0 + timesI(cosw - 3.0 * xi0));
+      2.0 * timesI(e2iu) + emiu * (-3.0 * u * xi0 + timesI(cosw - 3.0 * xi0));
 
     r02 = -2.0 * e2iu +
-          emiu * (-8.0 * u2 * xi0 +
-                  timesI(2.0 * u * (cosw + xi0 + 3.0 * u2 * xi1)));
+      emiu * (-8.0 * u2 * xi0 +
+	      timesI(2.0 * u * (cosw + xi0 + 3.0 * u2 * xi1)));
 
     r12 = emiu * (2.0 * u * xi0 + timesI(-cosw - xi0 + 3.0 * u2 * xi1));
 
@@ -211,19 +206,19 @@ private:
     GaugeLinkField USQ = USigmap * iQ;
 
     GaugeLinkField iGamma = tr1 * iQ - timesI(tr2) * iQ2 +
-                            timesI(f1) * USigmap + f2 * QUS + f2 * USQ;
+      timesI(f1) * USigmap + f2 * QUS + f2 * USQ;
 
     iLambda = Ta(iGamma);
   }
 
   //====================================================================
 public:
-  GaugeField *
+  GaugeField*
       ThinLinks; /* Pointer to the thin links configuration */
 
   /* Standard constructor */
-  SmearedConfiguration(GridCartesian *UGrid, unsigned int Nsmear,
-                       Smear_Stout<Gimpl> &Stout)
+  SmearedConfiguration(GridCartesian* UGrid, unsigned int Nsmear,
+                       Smear_Stout<Gimpl>& Stout)
       : smearingLevels(Nsmear), StoutSmearing(Stout), ThinLinks(NULL)
   {
     for (unsigned int i = 0; i < smearingLevels; ++i)
@@ -232,7 +227,7 @@ public:
 
   /*! For just thin links */
   SmearedConfiguration()
-      : smearingLevels(0), StoutSmearing(), SmearedSet(), ThinLinks(NULL) {}
+    : smearingLevels(0), StoutSmearing(), SmearedSet(), ThinLinks(NULL) {}
 
   // attach the smeared routines to the thin links U and fill the smeared set
   void set_Field(GaugeField &U)
@@ -251,7 +246,7 @@ public:
     {
       double start = usecond();
       GaugeField force = SigmaTilde; // actually = U*SigmaTilde
-      GaugeLinkField tmp_mu(SigmaTilde._grid);
+      GaugeLinkField tmp_mu(SigmaTilde.Grid());
 
       for (int mu = 0; mu < Nd; mu++)
       {
@@ -273,11 +268,11 @@ public:
       double end = usecond();
       double time = (end - start)/ 1e3;
       std::cout << GridLogMessage << "Smearing force in " << time << " ms" << std::endl;  
-    } // if smearingLevels = 0 do nothing
+    }  // if smearingLevels = 0 do nothing
   }
   //====================================================================
 
-  GaugeField &get_SmearedU() { return SmearedSet[smearingLevels - 1]; }
+  GaugeField& get_SmearedU() { return SmearedSet[smearingLevels - 1]; }
 
   GaugeField &get_U(bool smeared = false)
   {
@@ -287,7 +282,7 @@ public:
       if (smearingLevels)
       {
         RealD impl_plaq =
-            WilsonLoops<Gimpl>::avgPlaquette(SmearedSet[smearingLevels - 1]);
+	  WilsonLoops<Gimpl>::avgPlaquette(SmearedSet[smearingLevels - 1]);
         std::cout << GridLogDebug << "getting Usmr Plaq: " << impl_plaq
                   << std::endl;
         return get_SmearedU();
@@ -309,7 +304,6 @@ public:
     }
   }
 };
-}
-}
 
-#endif
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/smearing/Smearing.h b/Grid/qcd/smearing/Smearing.h
index 932acf86..da5ede72 100644
--- a/Grid/qcd/smearing/Smearing.h
+++ b/Grid/qcd/smearing/Smearing.h
@@ -1,5 +1,4 @@
-#ifndef GRID_QCD_SMEARING_H
-#define GRID_QCD_SMEARING_H
+#pragma once
 
 #include <Grid/qcd/smearing/BaseSmearing.h>
 #include <Grid/qcd/smearing/APEsmearing.h>
@@ -7,4 +6,3 @@
 #include <Grid/qcd/smearing/GaugeConfiguration.h>
 #include <Grid/qcd/smearing/WilsonFlow.h>
 
-#endif
diff --git a/Grid/qcd/smearing/StoutSmearing.h b/Grid/qcd/smearing/StoutSmearing.h
index bfc37d0d..f463dc81 100644
--- a/Grid/qcd/smearing/StoutSmearing.h
+++ b/Grid/qcd/smearing/StoutSmearing.h
@@ -2,19 +2,17 @@
   @file stoutSmear.hpp
   @brief Declares Stout smearing class
 */
-#ifndef STOUT_SMEAR_
-#define STOUT_SMEAR_
+#pragma once
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 /*!  @brief Stout smearing of link variable. */
 template <class Gimpl>
 class Smear_Stout : public Smear<Gimpl> {
- private:
+private:
   const Smear<Gimpl>* SmearBase;
 
- public:
+public:
   INHERIT_GIMPL_TYPES(Gimpl)
 
   Smear_Stout(Smear<Gimpl>* base) : SmearBase(base) {
@@ -29,8 +27,8 @@ class Smear_Stout : public Smear<Gimpl> {
   ~Smear_Stout() {}  // delete SmearBase...
 
   void smear(GaugeField& u_smr, const GaugeField& U) const {
-    GaugeField C(U._grid);
-    GaugeLinkField tmp(U._grid), iq_mu(U._grid), Umu(U._grid);
+    GaugeField C(U.Grid());
+    GaugeLinkField tmp(U.Grid()), iq_mu(U.Grid()), Umu(U.Grid());
 
     std::cout << GridLogDebug << "Stout smearing started\n";
 
@@ -41,8 +39,8 @@ class Smear_Stout : public Smear<Gimpl> {
       tmp = peekLorentz(C, mu);
       Umu = peekLorentz(U, mu);
       iq_mu = Ta(
-          tmp *
-          adj(Umu));  // iq_mu = Ta(Omega_mu) to match the signs with the paper
+		 tmp *
+		 adj(Umu));  // iq_mu = Ta(Omega_mu) to match the signs with the paper
       exponentiate_iQ(tmp, iq_mu);
       pokeLorentz(u_smr, tmp * Umu, mu);  // u_smr = exp(iQ_mu)*U_mu
     }
@@ -71,7 +69,7 @@ class Smear_Stout : public Smear<Gimpl> {
     // the i sign is coming from outside
     // input matrix is anti-hermitian NOT hermitian
 
-    GridBase* grid = iQ._grid;
+    GridBase* grid = iQ.Grid();
     GaugeLinkField unity(grid);
     unity = 1.0;
 
@@ -93,7 +91,7 @@ class Smear_Stout : public Smear<Gimpl> {
     Complex one_over_three = 1.0 / 3.0;
     Complex one_over_two = 1.0 / 2.0;
 
-    GridBase* grid = u._grid;
+    GridBase* grid = u.Grid();
     LatticeComplex c0(grid), c1(grid), tmp(grid), c0max(grid), theta(grid);
 
     // sign in c0 from the conventions on the Ta
@@ -105,14 +103,14 @@ class Smear_Stout : public Smear<Gimpl> {
     c0max = 2.0 * pow(tmp, 1.5);
 
     theta = acos(c0 / c0max) *
-            one_over_three;  // divide by three here, now leave as it is
+      one_over_three;  // divide by three here, now leave as it is
     u = sqrt(tmp) * cos(theta);
     w = sqrt(c1) * sin(theta);
   }
 
   void set_fj(LatticeComplex& f0, LatticeComplex& f1, LatticeComplex& f2,
               const LatticeComplex& u, const LatticeComplex& w) const {
-    GridBase* grid = u._grid;
+    GridBase* grid = u.Grid();
     LatticeComplex xi0(grid), u2(grid), w2(grid), cosw(grid);
     LatticeComplex fden(grid);
     LatticeComplex h0(grid), h1(grid), h2(grid);
@@ -130,7 +128,7 @@ class Smear_Stout : public Smear<Gimpl> {
     e2iu = cos(2.0 * u) + timesI(sin(2.0 * u));
 
     h0 = e2iu * (u2 - w2) +
-         emiu * ((8.0 * u2 * cosw) + (2.0 * u * (3.0 * u2 + w2) * ixi0));
+      emiu * ((8.0 * u2 * cosw) + (2.0 * u * (3.0 * u2 + w2) * ixi0));
     h1 = e2iu * (2.0 * u) - emiu * ((2.0 * u * cosw) - (3.0 * u2 - w2) * ixi0);
     h2 = e2iu - emiu * (cosw + (3.0 * u) * ixi0);
 
@@ -154,7 +152,6 @@ class Smear_Stout : public Smear<Gimpl> {
     return cos(w) / (w * w) - sin(w) / (w * w * w);
   }
 };
-}
-}
 
-#endif
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/smearing/WilsonFlow.h b/Grid/qcd/smearing/WilsonFlow.h
index 1c0855ab..19fd94e2 100644
--- a/Grid/qcd/smearing/WilsonFlow.h
+++ b/Grid/qcd/smearing/WilsonFlow.h
@@ -25,59 +25,57 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 
-#ifndef WILSONFLOW_H
-#define WILSONFLOW_H
+#pragma once
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 template <class Gimpl>
 class WilsonFlow: public Smear<Gimpl>{
-    unsigned int Nstep;
-    unsigned int measure_interval;
-    mutable RealD epsilon, taus;
+  unsigned int Nstep;
+  unsigned int measure_interval;
+  mutable RealD epsilon, taus;
 
 
-    mutable WilsonGaugeAction<Gimpl> SG;
+  mutable WilsonGaugeAction<Gimpl> SG;
 
-    void evolve_step(typename Gimpl::GaugeField&) const;
-    void evolve_step_adaptive(typename Gimpl::GaugeField&, RealD);
-    RealD tau(unsigned int t)const {return epsilon*(t+1.0); }
+  void evolve_step(typename Gimpl::GaugeField&) const;
+  void evolve_step_adaptive(typename Gimpl::GaugeField&, RealD);
+  RealD tau(unsigned int t)const {return epsilon*(t+1.0); }
 
- public:
-    INHERIT_GIMPL_TYPES(Gimpl)
+public:
+  INHERIT_GIMPL_TYPES(Gimpl)
 
-    explicit WilsonFlow(unsigned int Nstep, RealD epsilon, unsigned int interval = 1):
-        Nstep(Nstep),
-        epsilon(epsilon),
-        measure_interval(interval),
-        SG(WilsonGaugeAction<Gimpl>(3.0)) {
-            // WilsonGaugeAction with beta 3.0
-            assert(epsilon > 0.0);
-            LogMessage();
-    }
+  explicit WilsonFlow(unsigned int Nstep, RealD epsilon, unsigned int interval = 1):
+  Nstep(Nstep),
+    epsilon(epsilon),
+    measure_interval(interval),
+    SG(WilsonGaugeAction<Gimpl>(3.0)) {
+    // WilsonGaugeAction with beta 3.0
+    assert(epsilon > 0.0);
+    LogMessage();
+  }
 
-    void LogMessage() {
-        std::cout << GridLogMessage
-            << "[WilsonFlow] Nstep   : " << Nstep << std::endl;
-        std::cout << GridLogMessage
-            << "[WilsonFlow] epsilon : " << epsilon << std::endl;
-        std::cout << GridLogMessage
-            << "[WilsonFlow] full trajectory : " << Nstep * epsilon << std::endl;
-    }
+  void LogMessage() {
+    std::cout << GridLogMessage
+	      << "[WilsonFlow] Nstep   : " << Nstep << std::endl;
+    std::cout << GridLogMessage
+	      << "[WilsonFlow] epsilon : " << epsilon << std::endl;
+    std::cout << GridLogMessage
+	      << "[WilsonFlow] full trajectory : " << Nstep * epsilon << std::endl;
+  }
 
-    virtual void smear(GaugeField&, const GaugeField&) const;
+  virtual void smear(GaugeField&, const GaugeField&) const;
 
-    virtual void derivative(GaugeField&, const GaugeField&, const GaugeField&) const {
-        assert(0);
-        // undefined for WilsonFlow
-    }
+  virtual void derivative(GaugeField&, const GaugeField&, const GaugeField&) const {
+    assert(0);
+    // undefined for WilsonFlow
+  }
 
-    void smear_adaptive(GaugeField&, const GaugeField&, RealD maxTau);
-    RealD energyDensityPlaquette(unsigned int step, const GaugeField& U) const;
-    RealD energyDensityPlaquette(const GaugeField& U) const;
+  void smear_adaptive(GaugeField&, const GaugeField&, RealD maxTau);
+  RealD energyDensityPlaquette(unsigned int step, const GaugeField& U) const;
+  RealD energyDensityPlaquette(const GaugeField& U) const;
 };
 
 
@@ -86,74 +84,74 @@ class WilsonFlow: public Smear<Gimpl>{
 ////////////////////////////////////////////////////////////////////////////////
 template <class Gimpl>
 void WilsonFlow<Gimpl>::evolve_step(typename Gimpl::GaugeField &U) const{
-    GaugeField Z(U._grid);
-    GaugeField tmp(U._grid);
-    SG.deriv(U, Z);
-    Z *= 0.25;                                  // Z0 = 1/4 * F(U)
-    Gimpl::update_field(Z, U, -2.0*epsilon);    // U = W1 = exp(ep*Z0)*W0
+  GaugeField Z(U.Grid());
+  GaugeField tmp(U.Grid());
+  SG.deriv(U, Z);
+  Z *= 0.25;                                  // Z0 = 1/4 * F(U)
+  Gimpl::update_field(Z, U, -2.0*epsilon);    // U = W1 = exp(ep*Z0)*W0
 
-    Z *= -17.0/8.0;
-    SG.deriv(U, tmp); Z += tmp;                 // -17/32*Z0 +Z1
-    Z *= 8.0/9.0;                               // Z = -17/36*Z0 +8/9*Z1
-    Gimpl::update_field(Z, U, -2.0*epsilon);    // U_= W2 = exp(ep*Z)*W1
+  Z *= -17.0/8.0;
+  SG.deriv(U, tmp); Z += tmp;                 // -17/32*Z0 +Z1
+  Z *= 8.0/9.0;                               // Z = -17/36*Z0 +8/9*Z1
+  Gimpl::update_field(Z, U, -2.0*epsilon);    // U_= W2 = exp(ep*Z)*W1
 
-    Z *= -4.0/3.0;
-    SG.deriv(U, tmp); Z += tmp;                 // 4/3*(17/36*Z0 -8/9*Z1) +Z2
-    Z *= 3.0/4.0;                               // Z = 17/36*Z0 -8/9*Z1 +3/4*Z2
-    Gimpl::update_field(Z, U, -2.0*epsilon);    // V(t+e) = exp(ep*Z)*W2
+  Z *= -4.0/3.0;
+  SG.deriv(U, tmp); Z += tmp;                 // 4/3*(17/36*Z0 -8/9*Z1) +Z2
+  Z *= 3.0/4.0;                               // Z = 17/36*Z0 -8/9*Z1 +3/4*Z2
+  Gimpl::update_field(Z, U, -2.0*epsilon);    // V(t+e) = exp(ep*Z)*W2
 }
 
 template <class Gimpl>
 void WilsonFlow<Gimpl>::evolve_step_adaptive(typename Gimpl::GaugeField &U, RealD maxTau) {
-    if (maxTau - taus < epsilon){
-        epsilon = maxTau-taus;
-    }
-    //std::cout << GridLogMessage << "Integration epsilon : " << epsilon << std::endl;
-    GaugeField Z(U._grid);
-    GaugeField Zprime(U._grid);
-    GaugeField tmp(U._grid), Uprime(U._grid);
-    Uprime = U;
-    SG.deriv(U, Z);
-    Zprime = -Z;
-    Z *= 0.25;                                  // Z0 = 1/4 * F(U)
-    Gimpl::update_field(Z, U, -2.0*epsilon);    // U = W1 = exp(ep*Z0)*W0
+  if (maxTau - taus < epsilon){
+    epsilon = maxTau-taus;
+  }
+  //std::cout << GridLogMessage << "Integration epsilon : " << epsilon << std::endl;
+  GaugeField Z(U.Grid());
+  GaugeField Zprime(U.Grid());
+  GaugeField tmp(U.Grid()), Uprime(U.Grid());
+  Uprime = U;
+  SG.deriv(U, Z);
+  Zprime = -Z;
+  Z *= 0.25;                                  // Z0 = 1/4 * F(U)
+  Gimpl::update_field(Z, U, -2.0*epsilon);    // U = W1 = exp(ep*Z0)*W0
 
-    Z *= -17.0/8.0;
-    SG.deriv(U, tmp); Z += tmp;                 // -17/32*Z0 +Z1
-    Zprime += 2.0*tmp;
-    Z *= 8.0/9.0;                               // Z = -17/36*Z0 +8/9*Z1
-    Gimpl::update_field(Z, U, -2.0*epsilon);    // U_= W2 = exp(ep*Z)*W1
+  Z *= -17.0/8.0;
+  SG.deriv(U, tmp); Z += tmp;                 // -17/32*Z0 +Z1
+  Zprime += 2.0*tmp;
+  Z *= 8.0/9.0;                               // Z = -17/36*Z0 +8/9*Z1
+  Gimpl::update_field(Z, U, -2.0*epsilon);    // U_= W2 = exp(ep*Z)*W1
     
 
-    Z *= -4.0/3.0;
-    SG.deriv(U, tmp); Z += tmp;                 // 4/3*(17/36*Z0 -8/9*Z1) +Z2
-    Z *= 3.0/4.0;                               // Z = 17/36*Z0 -8/9*Z1 +3/4*Z2
-    Gimpl::update_field(Z, U, -2.0*epsilon);    // V(t+e) = exp(ep*Z)*W2
+  Z *= -4.0/3.0;
+  SG.deriv(U, tmp); Z += tmp;                 // 4/3*(17/36*Z0 -8/9*Z1) +Z2
+  Z *= 3.0/4.0;                               // Z = 17/36*Z0 -8/9*Z1 +3/4*Z2
+  Gimpl::update_field(Z, U, -2.0*epsilon);    // V(t+e) = exp(ep*Z)*W2
 
-    // Ramos 
-    Gimpl::update_field(Zprime, Uprime, -2.0*epsilon); // V'(t+e) = exp(ep*Z')*W0
-    // Compute distance as norm^2 of the difference
-    GaugeField diffU = U - Uprime;
-    RealD diff = norm2(diffU);
-    // adjust integration step
+  // Ramos 
+  Gimpl::update_field(Zprime, Uprime, -2.0*epsilon); // V'(t+e) = exp(ep*Z')*W0
+  // Compute distance as norm^2 of the difference
+  GaugeField diffU = U - Uprime;
+  RealD diff = norm2(diffU);
+  // adjust integration step
     
-    taus += epsilon;
-    //std::cout << GridLogMessage << "Adjusting integration step with distance: " << diff << std::endl;
+  taus += epsilon;
+  //std::cout << GridLogMessage << "Adjusting integration step with distance: " << diff << std::endl;
     
-    epsilon = epsilon*0.95*std::pow(1e-4/diff,1./3.);
-    //std::cout << GridLogMessage << "New epsilon : " << epsilon << std::endl;
+  epsilon = epsilon*0.95*std::pow(1e-4/diff,1./3.);
+  //std::cout << GridLogMessage << "New epsilon : " << epsilon << std::endl;
 
 }
 
 template <class Gimpl>
 RealD WilsonFlow<Gimpl>::energyDensityPlaquette(unsigned int step, const GaugeField& U) const {
-    RealD td = tau(step);
-    return 2.0 * td * td * SG.S(U)/U._grid->gSites();
+  RealD td = tau(step);
+  return 2.0 * td * td * SG.S(U)/U.Grid()->gSites();
 }
 
 template <class Gimpl>
 RealD WilsonFlow<Gimpl>::energyDensityPlaquette(const GaugeField& U) const {
-    return 2.0 * taus * taus * SG.S(U)/U._grid->gSites();
+  return 2.0 * taus * taus * SG.S(U)/U.Grid()->gSites();
 }
 
 
@@ -163,51 +161,48 @@ RealD WilsonFlow<Gimpl>::energyDensityPlaquette(const GaugeField& U) const {
 
 template <class Gimpl>
 void WilsonFlow<Gimpl>::smear(GaugeField& out, const GaugeField& in) const {
-    out = in;
-    for (unsigned int step = 1; step <= Nstep; step++) {
-        auto start = std::chrono::high_resolution_clock::now();
-        evolve_step(out);
-        auto end = std::chrono::high_resolution_clock::now();
-        std::chrono::duration<double> diff = end - start;
-        #ifdef WF_TIMING
-        std::cout << "Time to evolve " << diff.count() << " s\n";
-        #endif
-        std::cout << GridLogMessage << "[WilsonFlow] Energy density (plaq) : "
+  out = in;
+  for (unsigned int step = 1; step <= Nstep; step++) {
+    auto start = std::chrono::high_resolution_clock::now();
+    evolve_step(out);
+    auto end = std::chrono::high_resolution_clock::now();
+    std::chrono::duration<double> diff = end - start;
+#ifdef WF_TIMING
+    std::cout << "Time to evolve " << diff.count() << " s\n";
+#endif
+    std::cout << GridLogMessage << "[WilsonFlow] Energy density (plaq) : "
 		  << step << "  " << tau(step) << "  " 
-		  << energyDensityPlaquette(step,out) << std::endl;
-         if( step % measure_interval == 0){
-         std::cout << GridLogMessage << "[WilsonFlow] Top. charge           : "
-            << step << "  " 
-            << WilsonLoops<PeriodicGimplR>::TopologicalCharge(out) << std::endl;
-        }
+	      << energyDensityPlaquette(step,out) << std::endl;
+    if( step % measure_interval == 0){
+      std::cout << GridLogMessage << "[WilsonFlow] Top. charge           : "
+		<< step << "  " 
+		<< WilsonLoops<PeriodicGimplR>::TopologicalCharge(out) << std::endl;
     }
+  }
 }
 
 template <class Gimpl>
 void WilsonFlow<Gimpl>::smear_adaptive(GaugeField& out, const GaugeField& in, RealD maxTau){
-    out = in;
-    taus = epsilon;
-    unsigned int step = 0;
-    do{
-        step++;
-        //std::cout << GridLogMessage << "Evolution time :"<< taus << std::endl;
-        evolve_step_adaptive(out, maxTau);
-        std::cout << GridLogMessage << "[WilsonFlow] Energy density (plaq) : "
+  out = in;
+  taus = epsilon;
+  unsigned int step = 0;
+  do{
+    step++;
+    //std::cout << GridLogMessage << "Evolution time :"<< taus << std::endl;
+    evolve_step_adaptive(out, maxTau);
+    std::cout << GridLogMessage << "[WilsonFlow] Energy density (plaq) : "
 		  << step << "  " << taus << "  "
-		  << energyDensityPlaquette(out) << std::endl;
-         if( step % measure_interval == 0){
-         std::cout << GridLogMessage << "[WilsonFlow] Top. charge           : "
-            << step << "  " 
-            << WilsonLoops<PeriodicGimplR>::TopologicalCharge(out) << std::endl;
-        }
-    } while (taus < maxTau);
+	      << energyDensityPlaquette(out) << std::endl;
+    if( step % measure_interval == 0){
+      std::cout << GridLogMessage << "[WilsonFlow] Top. charge           : "
+		<< step << "  " 
+		<< WilsonLoops<PeriodicGimplR>::TopologicalCharge(out) << std::endl;
+    }
+  } while (taus < maxTau);
 
 
 
 }
 
+NAMESPACE_END(Grid);
 
-}  // namespace QCD
-}  // namespace Grid
-
-#endif   // WILSONFLOW_H
diff --git a/Grid/qcd/spin/Dirac.h b/Grid/qcd/spin/Dirac.h
index ed22ec97..d03e0939 100644
--- a/Grid/qcd/spin/Dirac.h
+++ b/Grid/qcd/spin/Dirac.h
@@ -28,35 +28,35 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 
 See the full license in the file "LICENSE" in the top level distribution directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef GRID_QCD_DIRAC_H
 #define GRID_QCD_DIRAC_H
 
-// Gamma matrices using the code generated by the Mathematica notebook 
-// gamma-gen/gamma-gen.nb in Gamma.cc & Gamma.h
-////////////////////////////////////////////////////////////////////////////////
+			   // Gamma matrices using the code generated by the Mathematica notebook 
+			   // gamma-gen/gamma-gen.nb in Gamma.cc & Gamma.h
+			   ////////////////////////////////////////////////////////////////////////////////
 #include <Grid/qcd/spin/Gamma.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-// Dirac algebra adjoint operator (not in QCD:: to overload other adj)
-inline QCD::Gamma adj(const QCD::Gamma &g)
+// Dirac algebra adjoint operator (not in  to overload other adj)
+accelerator_inline Gamma adj(const Gamma &g)
 {
-   return QCD::Gamma (QCD::Gamma::adj[g.g]);
+  return Gamma (Gamma::adj[g.g]);
 }
 
-namespace QCD {
+
 
 // Dirac algebra mutliplication operator
-inline Gamma operator*(const Gamma &g1, const Gamma &g2)
+accelerator_inline Gamma operator*(const Gamma &g1, const Gamma &g2)
 {
-   return Gamma (Gamma::mul[g1.g][g2.g]);
+  return Gamma (Gamma::mul[g1.g][g2.g]);
 }
 
 // general left multiply
 template<class vtype> 
-inline auto operator*(const Gamma &G, const iScalar<vtype> &arg)
-->typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype>>::type 
+accelerator_inline auto operator*(const Gamma &G, const iScalar<vtype> &arg)
+  ->typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype>>::type 
 {
   iScalar<vtype> ret;
   ret._internal=G*arg._internal;
@@ -64,8 +64,8 @@ inline auto operator*(const Gamma &G, const iScalar<vtype> &arg)
 }
 
 template<class vtype,int N>
-inline auto operator*(const Gamma &G, const iVector<vtype, N> &arg)
-->typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N>>::type 
+accelerator_inline auto operator*(const Gamma &G, const iVector<vtype, N> &arg)
+  ->typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N>>::type 
 {
   iVector<vtype,N> ret;
   for(int i=0;i<N;i++){
@@ -75,21 +75,21 @@ inline auto operator*(const Gamma &G, const iVector<vtype, N> &arg)
 }
 
 template<class vtype, int N>
-inline auto operator*(const Gamma &G, const iMatrix<vtype, N> &arg) 
-->typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N>>::type 
+accelerator_inline auto operator*(const Gamma &G, const iMatrix<vtype, N> &arg) 
+  ->typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N>>::type 
 {
   iMatrix<vtype,N> ret;
   for(int i=0;i<N;i++){
-  for(int j=0;j<N;j++){
-    ret._internal[i][j]=G*arg._internal[i][j];
-  }}
+    for(int j=0;j<N;j++){
+      ret._internal[i][j]=G*arg._internal[i][j];
+    }}
   return ret;
 }
 
 // general right multiply
 template<class vtype>
-inline auto operator*(const iScalar<vtype> &arg, const Gamma &G)
-->typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype>>::type 
+accelerator_inline auto operator*(const iScalar<vtype> &arg, const Gamma &G)
+  ->typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype>>::type 
 {
   iScalar<vtype> ret;
   ret._internal=arg._internal*G;
@@ -97,14 +97,14 @@ inline auto operator*(const iScalar<vtype> &arg, const Gamma &G)
 }
 
 template<class vtype, int N>
-inline auto operator * (const iMatrix<vtype, N> &arg, const Gamma &G) 
-->typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N>>::type 
+accelerator_inline auto operator * (const iMatrix<vtype, N> &arg, const Gamma &G) 
+  ->typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N>>::type 
 {
   iMatrix<vtype,N> ret;
   for(int i=0;i<N;i++){
-  for(int j=0;j<N;j++){
-    ret._internal[i][j]=arg._internal[i][j]*G;
-  }}
+    for(int j=0;j<N;j++){
+      ret._internal[i][j]=arg._internal[i][j]*G;
+    }}
   return ret;
 }
 
@@ -122,8 +122,8 @@ public:
 
 // vector multiply
 template<class vtype> 
-inline auto operator*(const GammaL &gl, const iVector<vtype, Ns> &arg)
-->typename std::enable_if<matchGridTensorIndex<iVector<vtype, Ns>, SpinorIndex>::value, iVector<vtype, Ns>>::type
+accelerator_inline auto operator*(const GammaL &gl, const iVector<vtype, Ns> &arg)
+  ->typename std::enable_if<matchGridTensorIndex<iVector<vtype, Ns>, SpinorIndex>::value, iVector<vtype, Ns>>::type
 {
   iVector<vtype, Ns> buf;
   
@@ -137,45 +137,45 @@ inline auto operator*(const GammaL &gl, const iVector<vtype, Ns> &arg)
 
 // matrix left multiply
 template<class vtype> 
-inline auto operator*(const GammaL &gl, const iMatrix<vtype, Ns> &arg)
-->typename std::enable_if<matchGridTensorIndex<iMatrix<vtype, Ns>, SpinorIndex>::value, iMatrix<vtype, Ns>>::type
+accelerator_inline auto operator*(const GammaL &gl, const iMatrix<vtype, Ns> &arg)
+  ->typename std::enable_if<matchGridTensorIndex<iMatrix<vtype, Ns>, SpinorIndex>::value, iMatrix<vtype, Ns>>::type
 {
   iMatrix<vtype, Ns> buf;
   
   for(unsigned int i = 0; i < Ns; ++i)
-  {
-    buf(0, i) = 0.;
-    buf(1, i) = 0.;
-    buf(2, i) = 2.*arg(2, i);
-    buf(3, i) = 2.*arg(3, i);
-  }
+    {
+      buf(0, i) = 0.;
+      buf(1, i) = 0.;
+      buf(2, i) = 2.*arg(2, i);
+      buf(3, i) = 2.*arg(3, i);
+    }
   
   return gl.gamma*buf;
 };
 
 // matrix right multiply
 template<class vtype> 
-inline auto operator*(const iMatrix<vtype, Ns> &arg, const GammaL &gl)
-->typename std::enable_if<matchGridTensorIndex<iMatrix<vtype, Ns>, SpinorIndex>::value, iMatrix<vtype, Ns>>::type
+accelerator_inline auto operator*(const iMatrix<vtype, Ns> &arg, const GammaL &gl)
+  ->typename std::enable_if<matchGridTensorIndex<iMatrix<vtype, Ns>, SpinorIndex>::value, iMatrix<vtype, Ns>>::type
 {
   iMatrix<vtype, Ns> buf;
   
   buf = arg*gl.gamma;
   for(unsigned int i = 0; i < Ns; ++i)
-  {
-    buf(i, 0) = 0.;
-    buf(i, 1) = 0.;
-    buf(i, 2) = 2.*buf(i, 2);
-    buf(i, 3) = 2.*buf(i, 3);
-  }
+    {
+      buf(i, 0) = 0.;
+      buf(i, 1) = 0.;
+      buf(i, 2) = 2.*buf(i, 2);
+      buf(i, 3) = 2.*buf(i, 3);
+    }
   
   return buf;
 };
 
 //general left multiply
 template<class vtype> 
-inline auto operator*(const GammaL &gl, const iScalar<vtype> &arg)
-->typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype>>::type 
+accelerator_inline auto operator*(const GammaL &gl, const iScalar<vtype> &arg)
+  ->typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype>>::type 
 {
   iScalar<vtype> ret;
   ret._internal=gl*arg._internal;
@@ -183,8 +183,8 @@ inline auto operator*(const GammaL &gl, const iScalar<vtype> &arg)
 }
 
 template<class vtype,int N>
-inline auto operator*(const GammaL &gl, const iVector<vtype, N> &arg)
-->typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N>>::type 
+accelerator_inline auto operator*(const GammaL &gl, const iVector<vtype, N> &arg)
+  ->typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N>>::type 
 {
   iVector<vtype,N> ret;
   for(int i=0;i<N;i++){
@@ -194,21 +194,21 @@ inline auto operator*(const GammaL &gl, const iVector<vtype, N> &arg)
 }
 
 template<class vtype, int N>
-inline auto operator*(const GammaL &gl, const iMatrix<vtype, N> &arg) 
-->typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N>>::type 
+accelerator_inline auto operator*(const GammaL &gl, const iMatrix<vtype, N> &arg) 
+  ->typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N>>::type 
 {
   iMatrix<vtype,N> ret;
   for(int i=0;i<N;i++){
-  for(int j=0;j<N;j++){
-    ret._internal[i][j]=gl*arg._internal[i][j];
-  }}
+    for(int j=0;j<N;j++){
+      ret._internal[i][j]=gl*arg._internal[i][j];
+    }}
   return ret;
 }
 
 //general right multiply
 template<class vtype>
-inline auto operator*(const iScalar<vtype> &arg, const GammaL &gl)
-->typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype>>::type 
+accelerator_inline auto operator*(const iScalar<vtype> &arg, const GammaL &gl)
+  ->typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype>>::type 
 {
   iScalar<vtype> ret;
   ret._internal=arg._internal*gl;
@@ -216,17 +216,17 @@ inline auto operator*(const iScalar<vtype> &arg, const GammaL &gl)
 }
 
 template<class vtype, int N>
-inline auto operator * (const iMatrix<vtype, N> &arg, const GammaL &gl) 
-->typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N>>::type 
+accelerator_inline auto operator * (const iMatrix<vtype, N> &arg, const GammaL &gl) 
+  ->typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N>>::type 
 {
   iMatrix<vtype,N> ret;
   for(int i=0;i<N;i++){
-  for(int j=0;j<N;j++){
-    ret._internal[i][j]=arg._internal[i][j]*gl;
-  }}
+    for(int j=0;j<N;j++){
+      ret._internal[i][j]=arg._internal[i][j]*gl;
+    }}
   return ret;
 }
 
-}}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/spin/Gamma.cc b/Grid/qcd/spin/Gamma.cc
index 6cc881fb..de56d713 100644
--- a/Grid/qcd/spin/Gamma.cc
+++ b/Grid/qcd/spin/Gamma.cc
@@ -2,14 +2,13 @@
 
 #include <Grid/Grid.h>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 const std::array<const Gamma, 4> Gamma::gmu = {{
-  Gamma(Gamma::Algebra::GammaX),
-  Gamma(Gamma::Algebra::GammaY),
-  Gamma(Gamma::Algebra::GammaZ),
-  Gamma(Gamma::Algebra::GammaT)}};
+    Gamma(Gamma::Algebra::GammaX),
+    Gamma(Gamma::Algebra::GammaY),
+    Gamma(Gamma::Algebra::GammaZ),
+    Gamma(Gamma::Algebra::GammaT)}};
 
 const std::array<const Gamma, 16> Gamma::gall = {{
   Gamma(Gamma::Algebra::Identity),
@@ -30,1129 +29,1129 @@ const std::array<const Gamma, 16> Gamma::gall = {{
   Gamma(Gamma::Algebra::SigmaZT)}};
 
 const std::array<const char *, Gamma::nGamma> Gamma::name = {{
-  "-Gamma5      ",
-  "Gamma5       ",
-  "-GammaT      ",
-  "GammaT       ",
-  "-GammaTGamma5",
-  "GammaTGamma5 ",
-  "-GammaX      ",
-  "GammaX       ",
-  "-GammaXGamma5",
-  "GammaXGamma5 ",
-  "-GammaY      ",
-  "GammaY       ",
-  "-GammaYGamma5",
-  "GammaYGamma5 ",
-  "-GammaZ      ",
-  "GammaZ       ",
-  "-GammaZGamma5",
-  "GammaZGamma5 ",
-  "-Identity    ",
-  "Identity     ",
-  "-SigmaXT     ",
-  "SigmaXT      ",
-  "-SigmaXY     ",
-  "SigmaXY      ",
-  "-SigmaXZ     ",
-  "SigmaXZ      ",
-  "-SigmaYT     ",
-  "SigmaYT      ",
-  "-SigmaYZ     ",
-  "SigmaYZ      ",
-  "-SigmaZT     ",
-  "SigmaZT      "}};
+    "-Gamma5      ",
+    "Gamma5       ",
+    "-GammaT      ",
+    "GammaT       ",
+    "-GammaTGamma5",
+    "GammaTGamma5 ",
+    "-GammaX      ",
+    "GammaX       ",
+    "-GammaXGamma5",
+    "GammaXGamma5 ",
+    "-GammaY      ",
+    "GammaY       ",
+    "-GammaYGamma5",
+    "GammaYGamma5 ",
+    "-GammaZ      ",
+    "GammaZ       ",
+    "-GammaZGamma5",
+    "GammaZGamma5 ",
+    "-Identity    ",
+    "Identity     ",
+    "-SigmaXT     ",
+    "SigmaXT      ",
+    "-SigmaXY     ",
+    "SigmaXY      ",
+    "-SigmaXZ     ",
+    "SigmaXZ      ",
+    "-SigmaYT     ",
+    "SigmaYT      ",
+    "-SigmaYZ     ",
+    "SigmaYZ      ",
+    "-SigmaZT     ",
+    "SigmaZT      "}};
 
 const std::array<Gamma::Algebra, Gamma::nGamma> Gamma::adj = {{
-  Gamma::Algebra::MinusGamma5,
-  Gamma::Algebra::Gamma5,
-  Gamma::Algebra::MinusGammaT,
-  Gamma::Algebra::GammaT,
-  Gamma::Algebra::GammaTGamma5,
-  Gamma::Algebra::MinusGammaTGamma5,
-  Gamma::Algebra::MinusGammaX,
-  Gamma::Algebra::GammaX,
-  Gamma::Algebra::GammaXGamma5,
-  Gamma::Algebra::MinusGammaXGamma5,
-  Gamma::Algebra::MinusGammaY,
-  Gamma::Algebra::GammaY,
-  Gamma::Algebra::GammaYGamma5,
-  Gamma::Algebra::MinusGammaYGamma5,
-  Gamma::Algebra::MinusGammaZ,
-  Gamma::Algebra::GammaZ,
-  Gamma::Algebra::GammaZGamma5,
-  Gamma::Algebra::MinusGammaZGamma5,
-  Gamma::Algebra::MinusIdentity,
-  Gamma::Algebra::Identity,
-  Gamma::Algebra::SigmaXT,
-  Gamma::Algebra::MinusSigmaXT,
-  Gamma::Algebra::SigmaXY,
-  Gamma::Algebra::MinusSigmaXY,
-  Gamma::Algebra::SigmaXZ,
-  Gamma::Algebra::MinusSigmaXZ,
-  Gamma::Algebra::SigmaYT,
-  Gamma::Algebra::MinusSigmaYT,
-  Gamma::Algebra::SigmaYZ,
-  Gamma::Algebra::MinusSigmaYZ,
-  Gamma::Algebra::SigmaZT,
-  Gamma::Algebra::MinusSigmaZT}};
+    Gamma::Algebra::MinusGamma5,
+    Gamma::Algebra::Gamma5,
+    Gamma::Algebra::MinusGammaT,
+    Gamma::Algebra::GammaT,
+    Gamma::Algebra::GammaTGamma5,
+    Gamma::Algebra::MinusGammaTGamma5,
+    Gamma::Algebra::MinusGammaX,
+    Gamma::Algebra::GammaX,
+    Gamma::Algebra::GammaXGamma5,
+    Gamma::Algebra::MinusGammaXGamma5,
+    Gamma::Algebra::MinusGammaY,
+    Gamma::Algebra::GammaY,
+    Gamma::Algebra::GammaYGamma5,
+    Gamma::Algebra::MinusGammaYGamma5,
+    Gamma::Algebra::MinusGammaZ,
+    Gamma::Algebra::GammaZ,
+    Gamma::Algebra::GammaZGamma5,
+    Gamma::Algebra::MinusGammaZGamma5,
+    Gamma::Algebra::MinusIdentity,
+    Gamma::Algebra::Identity,
+    Gamma::Algebra::SigmaXT,
+    Gamma::Algebra::MinusSigmaXT,
+    Gamma::Algebra::SigmaXY,
+    Gamma::Algebra::MinusSigmaXY,
+    Gamma::Algebra::SigmaXZ,
+    Gamma::Algebra::MinusSigmaXZ,
+    Gamma::Algebra::SigmaYT,
+    Gamma::Algebra::MinusSigmaYT,
+    Gamma::Algebra::SigmaYZ,
+    Gamma::Algebra::MinusSigmaYZ,
+    Gamma::Algebra::SigmaZT,
+    Gamma::Algebra::MinusSigmaZT}};
 
 const std::array<std::array<Gamma::Algebra, Gamma::nGamma>, Gamma::nGamma> Gamma::mul = {{
-  {{Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY}}
-  ,
-  {{Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY}}
-  ,
-  {{Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ}}
-  ,
-  {{Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ}}
-  ,
-  {{Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5}}
-  ,
-  {{Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5}}
-  ,
-  {{Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5}}
-  ,
-  {{Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5}}
-  ,
-  {{Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY}}
-  ,
-  {{Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY}}
-  ,
-  {{Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5}}
-  ,
-  {{Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5}}
-  ,
-  {{Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX}}
-  ,
-  {{Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX}}
-  ,
-  {{Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT}}
-  ,
-  {{Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT}}
-  ,
-  {{Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5}}
-  ,
-  {{Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5}}
-  ,
-  {{Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT}}
-  ,
-  {{Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT}}
-  ,
-  {{Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ}}
-  ,
-  {{Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ}}
-  ,
-  {{Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5}}
-  ,
-  {{Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5}}
-  ,
-  {{Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT}}
-  ,
-  {{Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT}}
-  ,
-  {{Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ}}
-  ,
-  {{Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ}}
-  ,
-  {{Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT}}
-  ,
-  {{Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT}}
-  ,
-  {{Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusIdentity,
-   Gamma::Algebra::Identity}}
-  ,
-  {{Gamma::Algebra::SigmaXY,
-   Gamma::Algebra::MinusSigmaXY,
-   Gamma::Algebra::MinusGammaZ,
-   Gamma::Algebra::GammaZ,
-   Gamma::Algebra::MinusGammaZGamma5,
-   Gamma::Algebra::GammaZGamma5,
-   Gamma::Algebra::GammaYGamma5,
-   Gamma::Algebra::MinusGammaYGamma5,
-   Gamma::Algebra::GammaY,
-   Gamma::Algebra::MinusGammaY,
-   Gamma::Algebra::MinusGammaXGamma5,
-   Gamma::Algebra::GammaXGamma5,
-   Gamma::Algebra::MinusGammaX,
-   Gamma::Algebra::GammaX,
-   Gamma::Algebra::GammaT,
-   Gamma::Algebra::MinusGammaT,
-   Gamma::Algebra::GammaTGamma5,
-   Gamma::Algebra::MinusGammaTGamma5,
-   Gamma::Algebra::MinusSigmaZT,
-   Gamma::Algebra::SigmaZT,
-   Gamma::Algebra::MinusSigmaXZ,
-   Gamma::Algebra::SigmaXZ,
-   Gamma::Algebra::MinusGamma5,
-   Gamma::Algebra::Gamma5,
-   Gamma::Algebra::SigmaXT,
-   Gamma::Algebra::MinusSigmaXT,
-   Gamma::Algebra::MinusSigmaYZ,
-   Gamma::Algebra::SigmaYZ,
-   Gamma::Algebra::SigmaYT,
-   Gamma::Algebra::MinusSigmaYT,
-   Gamma::Algebra::Identity,
-   Gamma::Algebra::MinusIdentity}}
-}};
+    {{Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY}}
+    ,
+    {{Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY}}
+    ,
+    {{Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ}}
+    ,
+    {{Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ}}
+    ,
+    {{Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5}}
+    ,
+    {{Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5}}
+    ,
+    {{Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5}}
+    ,
+    {{Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5}}
+    ,
+    {{Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY}}
+    ,
+    {{Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY}}
+    ,
+    {{Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5}}
+    ,
+    {{Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5}}
+    ,
+    {{Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX}}
+    ,
+    {{Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX}}
+    ,
+    {{Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT}}
+    ,
+    {{Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT}}
+    ,
+    {{Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5}}
+    ,
+    {{Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5}}
+    ,
+    {{Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT}}
+    ,
+    {{Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT}}
+    ,
+    {{Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ}}
+    ,
+    {{Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ}}
+    ,
+    {{Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5}}
+    ,
+    {{Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5}}
+    ,
+    {{Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT}}
+    ,
+    {{Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT}}
+    ,
+    {{Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ}}
+    ,
+    {{Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ}}
+    ,
+    {{Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT}}
+    ,
+    {{Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT}}
+    ,
+    {{Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusIdentity,
+      Gamma::Algebra::Identity}}
+    ,
+    {{Gamma::Algebra::SigmaXY,
+      Gamma::Algebra::MinusSigmaXY,
+      Gamma::Algebra::MinusGammaZ,
+      Gamma::Algebra::GammaZ,
+      Gamma::Algebra::MinusGammaZGamma5,
+      Gamma::Algebra::GammaZGamma5,
+      Gamma::Algebra::GammaYGamma5,
+      Gamma::Algebra::MinusGammaYGamma5,
+      Gamma::Algebra::GammaY,
+      Gamma::Algebra::MinusGammaY,
+      Gamma::Algebra::MinusGammaXGamma5,
+      Gamma::Algebra::GammaXGamma5,
+      Gamma::Algebra::MinusGammaX,
+      Gamma::Algebra::GammaX,
+      Gamma::Algebra::GammaT,
+      Gamma::Algebra::MinusGammaT,
+      Gamma::Algebra::GammaTGamma5,
+      Gamma::Algebra::MinusGammaTGamma5,
+      Gamma::Algebra::MinusSigmaZT,
+      Gamma::Algebra::SigmaZT,
+      Gamma::Algebra::MinusSigmaXZ,
+      Gamma::Algebra::SigmaXZ,
+      Gamma::Algebra::MinusGamma5,
+      Gamma::Algebra::Gamma5,
+      Gamma::Algebra::SigmaXT,
+      Gamma::Algebra::MinusSigmaXT,
+      Gamma::Algebra::MinusSigmaYZ,
+      Gamma::Algebra::SigmaYZ,
+      Gamma::Algebra::SigmaYT,
+      Gamma::Algebra::MinusSigmaYT,
+      Gamma::Algebra::Identity,
+      Gamma::Algebra::MinusIdentity}}
+  }};
 
-}}
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/spin/Gamma.h b/Grid/qcd/spin/Gamma.h
index 5655bb8c..e9778ffc 100644
--- a/Grid/qcd/spin/Gamma.h
+++ b/Grid/qcd/spin/Gamma.h
@@ -5,8 +5,7 @@
 
 #include <array>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 class Gamma {
   public:
@@ -51,11 +50,11 @@ class Gamma {
     static const std::array<const Gamma, 16>                     gall;
     Algebra                                                      g;
   public:
-    Gamma(Algebra initg): g(initg) {}  
+  accelerator Gamma(Algebra initg): g(initg) {}  
 };
 
 template<class vtype>
-inline void multMinusGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = -rhs(0);
   ret(1) = -rhs(1);
@@ -64,7 +63,7 @@ inline void multMinusGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &r
 };
 
 template<class vtype>
-inline void lmultMinusGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -76,7 +75,7 @@ inline void lmultMinusGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &
 };
 
 template<class vtype>
-inline void rmultMinusGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -88,7 +87,7 @@ inline void rmultMinusGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &
 };
 
 template<class vtype>
-inline void multGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = rhs(0);
   ret(1) = rhs(1);
@@ -97,7 +96,7 @@ inline void multGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void lmultGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -109,7 +108,7 @@ inline void lmultGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void rmultGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -121,7 +120,7 @@ inline void rmultGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void multMinusGammaT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusGammaT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = -rhs(2);
   ret(1) = -rhs(3);
@@ -130,7 +129,7 @@ inline void multMinusGammaT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &r
 };
 
 template<class vtype>
-inline void lmultMinusGammaT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusGammaT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -142,7 +141,7 @@ inline void lmultMinusGammaT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &
 };
 
 template<class vtype>
-inline void rmultMinusGammaT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusGammaT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -154,7 +153,7 @@ inline void rmultMinusGammaT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &
 };
 
 template<class vtype>
-inline void multGammaT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multGammaT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = rhs(2);
   ret(1) = rhs(3);
@@ -163,7 +162,7 @@ inline void multGammaT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void lmultGammaT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultGammaT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -175,7 +174,7 @@ inline void lmultGammaT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void rmultGammaT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultGammaT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -187,7 +186,7 @@ inline void rmultGammaT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void multMinusGammaTGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusGammaTGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = rhs(2);
   ret(1) = rhs(3);
@@ -196,7 +195,7 @@ inline void multMinusGammaTGamma5(iVector<vtype, Ns> &ret, const iVector<vtype,
 };
 
 template<class vtype>
-inline void lmultMinusGammaTGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusGammaTGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -208,7 +207,7 @@ inline void lmultMinusGammaTGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype,
 };
 
 template<class vtype>
-inline void rmultMinusGammaTGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusGammaTGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -220,7 +219,7 @@ inline void rmultMinusGammaTGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype,
 };
 
 template<class vtype>
-inline void multGammaTGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multGammaTGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = -rhs(2);
   ret(1) = -rhs(3);
@@ -229,7 +228,7 @@ inline void multGammaTGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &
 };
 
 template<class vtype>
-inline void lmultGammaTGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultGammaTGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -241,7 +240,7 @@ inline void lmultGammaTGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void rmultGammaTGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultGammaTGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -253,7 +252,7 @@ inline void rmultGammaTGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void multMinusGammaX(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusGammaX(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesMinusI(rhs(3));
   ret(1) = timesMinusI(rhs(2));
@@ -262,7 +261,7 @@ inline void multMinusGammaX(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &r
 };
 
 template<class vtype>
-inline void lmultMinusGammaX(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusGammaX(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -274,7 +273,7 @@ inline void lmultMinusGammaX(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &
 };
 
 template<class vtype>
-inline void rmultMinusGammaX(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusGammaX(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -286,7 +285,7 @@ inline void rmultMinusGammaX(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &
 };
 
 template<class vtype>
-inline void multGammaX(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multGammaX(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesI(rhs(3));
   ret(1) = timesI(rhs(2));
@@ -295,7 +294,7 @@ inline void multGammaX(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void lmultGammaX(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultGammaX(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -307,7 +306,7 @@ inline void lmultGammaX(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void rmultGammaX(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultGammaX(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -319,7 +318,7 @@ inline void rmultGammaX(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void multMinusGammaXGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusGammaXGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesI(rhs(3));
   ret(1) = timesI(rhs(2));
@@ -328,7 +327,7 @@ inline void multMinusGammaXGamma5(iVector<vtype, Ns> &ret, const iVector<vtype,
 };
 
 template<class vtype>
-inline void lmultMinusGammaXGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusGammaXGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -340,7 +339,7 @@ inline void lmultMinusGammaXGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype,
 };
 
 template<class vtype>
-inline void rmultMinusGammaXGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusGammaXGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -352,7 +351,7 @@ inline void rmultMinusGammaXGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype,
 };
 
 template<class vtype>
-inline void multGammaXGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multGammaXGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesMinusI(rhs(3));
   ret(1) = timesMinusI(rhs(2));
@@ -361,7 +360,7 @@ inline void multGammaXGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &
 };
 
 template<class vtype>
-inline void lmultGammaXGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultGammaXGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -373,7 +372,7 @@ inline void lmultGammaXGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void rmultGammaXGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultGammaXGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -385,7 +384,7 @@ inline void rmultGammaXGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void multMinusGammaY(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusGammaY(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = rhs(3);
   ret(1) = -rhs(2);
@@ -394,7 +393,7 @@ inline void multMinusGammaY(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &r
 };
 
 template<class vtype>
-inline void lmultMinusGammaY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusGammaY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -406,7 +405,7 @@ inline void lmultMinusGammaY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &
 };
 
 template<class vtype>
-inline void rmultMinusGammaY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusGammaY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -418,7 +417,7 @@ inline void rmultMinusGammaY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &
 };
 
 template<class vtype>
-inline void multGammaY(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multGammaY(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = -rhs(3);
   ret(1) = rhs(2);
@@ -427,7 +426,7 @@ inline void multGammaY(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void lmultGammaY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultGammaY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -439,7 +438,7 @@ inline void lmultGammaY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void rmultGammaY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultGammaY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -451,7 +450,7 @@ inline void rmultGammaY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void multMinusGammaYGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusGammaYGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = -rhs(3);
   ret(1) = rhs(2);
@@ -460,7 +459,7 @@ inline void multMinusGammaYGamma5(iVector<vtype, Ns> &ret, const iVector<vtype,
 };
 
 template<class vtype>
-inline void lmultMinusGammaYGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusGammaYGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -472,7 +471,7 @@ inline void lmultMinusGammaYGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype,
 };
 
 template<class vtype>
-inline void rmultMinusGammaYGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusGammaYGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -484,7 +483,7 @@ inline void rmultMinusGammaYGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype,
 };
 
 template<class vtype>
-inline void multGammaYGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multGammaYGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = rhs(3);
   ret(1) = -rhs(2);
@@ -493,7 +492,7 @@ inline void multGammaYGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &
 };
 
 template<class vtype>
-inline void lmultGammaYGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultGammaYGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -505,7 +504,7 @@ inline void lmultGammaYGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void rmultGammaYGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultGammaYGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -517,7 +516,7 @@ inline void rmultGammaYGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void multMinusGammaZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusGammaZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesMinusI(rhs(2));
   ret(1) = timesI(rhs(3));
@@ -526,7 +525,7 @@ inline void multMinusGammaZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &r
 };
 
 template<class vtype>
-inline void lmultMinusGammaZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusGammaZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -538,7 +537,7 @@ inline void lmultMinusGammaZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &
 };
 
 template<class vtype>
-inline void rmultMinusGammaZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusGammaZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -550,7 +549,7 @@ inline void rmultMinusGammaZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &
 };
 
 template<class vtype>
-inline void multGammaZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multGammaZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesI(rhs(2));
   ret(1) = timesMinusI(rhs(3));
@@ -559,7 +558,7 @@ inline void multGammaZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void lmultGammaZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultGammaZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -571,7 +570,7 @@ inline void lmultGammaZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void rmultGammaZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultGammaZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -583,7 +582,7 @@ inline void rmultGammaZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void multMinusGammaZGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusGammaZGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesI(rhs(2));
   ret(1) = timesMinusI(rhs(3));
@@ -592,7 +591,7 @@ inline void multMinusGammaZGamma5(iVector<vtype, Ns> &ret, const iVector<vtype,
 };
 
 template<class vtype>
-inline void lmultMinusGammaZGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusGammaZGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -604,7 +603,7 @@ inline void lmultMinusGammaZGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype,
 };
 
 template<class vtype>
-inline void rmultMinusGammaZGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusGammaZGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -616,7 +615,7 @@ inline void rmultMinusGammaZGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype,
 };
 
 template<class vtype>
-inline void multGammaZGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multGammaZGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesMinusI(rhs(2));
   ret(1) = timesI(rhs(3));
@@ -625,7 +624,7 @@ inline void multGammaZGamma5(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &
 };
 
 template<class vtype>
-inline void lmultGammaZGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultGammaZGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -637,7 +636,7 @@ inline void lmultGammaZGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void rmultGammaZGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultGammaZGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -649,7 +648,7 @@ inline void rmultGammaZGamma5(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void multMinusIdentity(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusIdentity(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = -rhs(0);
   ret(1) = -rhs(1);
@@ -658,7 +657,7 @@ inline void multMinusIdentity(iVector<vtype, Ns> &ret, const iVector<vtype, Ns>
 };
 
 template<class vtype>
-inline void lmultMinusIdentity(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusIdentity(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -670,7 +669,7 @@ inline void lmultMinusIdentity(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void rmultMinusIdentity(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusIdentity(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -682,7 +681,7 @@ inline void rmultMinusIdentity(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void multIdentity(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multIdentity(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = rhs(0);
   ret(1) = rhs(1);
@@ -691,7 +690,7 @@ inline void multIdentity(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void lmultIdentity(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultIdentity(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -703,7 +702,7 @@ inline void lmultIdentity(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs
 };
 
 template<class vtype>
-inline void rmultIdentity(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultIdentity(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -715,7 +714,7 @@ inline void rmultIdentity(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs
 };
 
 template<class vtype>
-inline void multMinusSigmaXT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusSigmaXT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesMinusI(rhs(1));
   ret(1) = timesMinusI(rhs(0));
@@ -724,7 +723,7 @@ inline void multMinusSigmaXT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &
 };
 
 template<class vtype>
-inline void lmultMinusSigmaXT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusSigmaXT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -736,7 +735,7 @@ inline void lmultMinusSigmaXT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void rmultMinusSigmaXT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusSigmaXT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -748,7 +747,7 @@ inline void rmultMinusSigmaXT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void multSigmaXT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multSigmaXT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesI(rhs(1));
   ret(1) = timesI(rhs(0));
@@ -757,7 +756,7 @@ inline void multSigmaXT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void lmultSigmaXT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultSigmaXT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -769,7 +768,7 @@ inline void lmultSigmaXT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void rmultSigmaXT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultSigmaXT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -781,7 +780,7 @@ inline void rmultSigmaXT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void multMinusSigmaXY(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusSigmaXY(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesI(rhs(0));
   ret(1) = timesMinusI(rhs(1));
@@ -790,7 +789,7 @@ inline void multMinusSigmaXY(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &
 };
 
 template<class vtype>
-inline void lmultMinusSigmaXY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusSigmaXY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -802,7 +801,7 @@ inline void lmultMinusSigmaXY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void rmultMinusSigmaXY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusSigmaXY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -814,7 +813,7 @@ inline void rmultMinusSigmaXY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void multSigmaXY(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multSigmaXY(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesMinusI(rhs(0));
   ret(1) = timesI(rhs(1));
@@ -823,7 +822,7 @@ inline void multSigmaXY(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void lmultSigmaXY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultSigmaXY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -835,7 +834,7 @@ inline void lmultSigmaXY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void rmultSigmaXY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultSigmaXY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -847,7 +846,7 @@ inline void rmultSigmaXY(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void multMinusSigmaXZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusSigmaXZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = rhs(1);
   ret(1) = -rhs(0);
@@ -856,7 +855,7 @@ inline void multMinusSigmaXZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &
 };
 
 template<class vtype>
-inline void lmultMinusSigmaXZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusSigmaXZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -868,7 +867,7 @@ inline void lmultMinusSigmaXZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void rmultMinusSigmaXZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusSigmaXZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -880,7 +879,7 @@ inline void rmultMinusSigmaXZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void multSigmaXZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multSigmaXZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = -rhs(1);
   ret(1) = rhs(0);
@@ -889,7 +888,7 @@ inline void multSigmaXZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void lmultSigmaXZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultSigmaXZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -901,7 +900,7 @@ inline void lmultSigmaXZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void rmultSigmaXZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultSigmaXZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -913,7 +912,7 @@ inline void rmultSigmaXZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void multMinusSigmaYT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusSigmaYT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = rhs(1);
   ret(1) = -rhs(0);
@@ -922,7 +921,7 @@ inline void multMinusSigmaYT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &
 };
 
 template<class vtype>
-inline void lmultMinusSigmaYT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusSigmaYT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -934,7 +933,7 @@ inline void lmultMinusSigmaYT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void rmultMinusSigmaYT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusSigmaYT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -946,7 +945,7 @@ inline void rmultMinusSigmaYT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void multSigmaYT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multSigmaYT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = -rhs(1);
   ret(1) = rhs(0);
@@ -955,7 +954,7 @@ inline void multSigmaYT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void lmultSigmaYT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultSigmaYT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -967,7 +966,7 @@ inline void lmultSigmaYT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void rmultSigmaYT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultSigmaYT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -979,7 +978,7 @@ inline void rmultSigmaYT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void multMinusSigmaYZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusSigmaYZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesI(rhs(1));
   ret(1) = timesI(rhs(0));
@@ -988,7 +987,7 @@ inline void multMinusSigmaYZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &
 };
 
 template<class vtype>
-inline void lmultMinusSigmaYZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusSigmaYZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -1000,7 +999,7 @@ inline void lmultMinusSigmaYZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void rmultMinusSigmaYZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusSigmaYZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -1012,7 +1011,7 @@ inline void rmultMinusSigmaYZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void multSigmaYZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multSigmaYZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesMinusI(rhs(1));
   ret(1) = timesMinusI(rhs(0));
@@ -1021,7 +1020,7 @@ inline void multSigmaYZ(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void lmultSigmaYZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultSigmaYZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -1033,7 +1032,7 @@ inline void lmultSigmaYZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void rmultSigmaYZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultSigmaYZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -1045,7 +1044,7 @@ inline void rmultSigmaYZ(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void multMinusSigmaZT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multMinusSigmaZT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesMinusI(rhs(0));
   ret(1) = timesI(rhs(1));
@@ -1054,7 +1053,7 @@ inline void multMinusSigmaZT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &
 };
 
 template<class vtype>
-inline void lmultMinusSigmaZT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultMinusSigmaZT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -1066,7 +1065,7 @@ inline void lmultMinusSigmaZT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void rmultMinusSigmaZT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultMinusSigmaZT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -1078,7 +1077,7 @@ inline void rmultMinusSigmaZT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns>
 };
 
 template<class vtype>
-inline void multSigmaZT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
+accelerator_inline void multSigmaZT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 {
   ret(0) = timesI(rhs(0));
   ret(1) = timesMinusI(rhs(1));
@@ -1087,7 +1086,7 @@ inline void multSigmaZT(iVector<vtype, Ns> &ret, const iVector<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void lmultSigmaZT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void lmultSigmaZT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -1099,7 +1098,7 @@ inline void lmultSigmaZT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype>
-inline void rmultSigmaZT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
+accelerator_inline void rmultSigmaZT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 {
   for(unsigned int i = 0; i < Ns; ++i)
   {
@@ -1111,7 +1110,7 @@ inline void rmultSigmaZT(iMatrix<vtype, Ns> &ret, const iMatrix<vtype, Ns> &rhs)
 };
 
 template<class vtype> 
-inline auto operator*(const Gamma &G, const iVector<vtype, Ns> &arg)
+accelerator_inline auto operator*(const Gamma &G, const iVector<vtype, Ns> &arg)
 ->typename std::enable_if<matchGridTensorIndex<iVector<vtype, Ns>, SpinorIndex>::value, iVector<vtype, Ns>>::type
 {
   iVector<vtype, Ns> ret;
@@ -1189,7 +1188,7 @@ inline auto operator*(const Gamma &G, const iVector<vtype, Ns> &arg)
 }
 
 template<class vtype> 
-inline auto operator*(const Gamma &G, const iMatrix<vtype, Ns> &arg)
+accelerator_inline auto operator*(const Gamma &G, const iMatrix<vtype, Ns> &arg)
 ->typename std::enable_if<matchGridTensorIndex<iMatrix<vtype, Ns>, SpinorIndex>::value, iMatrix<vtype, Ns>>::type
 {
   iMatrix<vtype, Ns> ret;
@@ -1267,7 +1266,7 @@ inline auto operator*(const Gamma &G, const iMatrix<vtype, Ns> &arg)
 }
 
 template<class vtype> 
-inline auto operator*(const iMatrix<vtype, Ns> &arg, const Gamma &G)
+accelerator_inline auto operator*(const iMatrix<vtype, Ns> &arg, const Gamma &G)
 ->typename std::enable_if<matchGridTensorIndex<iMatrix<vtype, Ns>, SpinorIndex>::value, iMatrix<vtype, Ns>>::type
 {
   iMatrix<vtype, Ns> ret;
@@ -1344,6 +1343,6 @@ inline auto operator*(const iMatrix<vtype, Ns> &arg, const Gamma &G)
   return ret;
 }
 
-}}
+NAMESPACE_END(Grid);
 
 #endif // GRID_QCD_GAMMA_H
diff --git a/Grid/qcd/spin/TwoSpinor.h b/Grid/qcd/spin/TwoSpinor.h
index 76cdbcf8..924594ab 100644
--- a/Grid/qcd/spin/TwoSpinor.h
+++ b/Grid/qcd/spin/TwoSpinor.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,1067 +24,1068 @@ Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_QCD_TWOSPIN_H
 #define GRID_QCD_TWOSPIN_H
-namespace Grid{
-namespace QCD {
 
-  //////////////////////////////////////////////////////////////////////////////////////////////////////
-  // Normalisation alert; the g5   project is 1/2(1+-G5) 
-  //                      the xyzt projects are (1+-Gxyzt)
-  //
-  // * xyzt project
-  //
-  // This is because this is how the Wilson operator is normally written as
-  // (m+4r) - \frac{1}{2} D_{hop}
-  // and / or
-  // 1 - \frac{1}{2 m+8r} D_{hop} = 1 - kappa D_{hop}
-  //
-  // Note that the free, critical hopping parameter kappa is then 1/8 th for r=1.
-  //
-  // However, the xyzt 2spin "projectors" are not really projectors because they do not
-  // square to 1, however the ChiralProjector is a true projector.
-  //
-  // For this reason there is NO provision in Grid of a four spinor result from the
-  // xyzt projectors. They are intended to be used only in combination with "reconstruct" in the
-  // wilson dirac operator and closely related actions.
-  //
-  // I also do NOT provide lattice wide operators of these, since the dirac operator is best implemented
-  // via Stencils and single site variants will be needed only for the cache friendly high perf dirac operator.
-  //
-  // * chiral project
-  //
-  // Both four spinor and two spinor result variants are provided.
-  //
-  // The four spinor project will be recursively provided to Lattice wide routines, and likely used in 
-  // the domain wall and mobius implementations.
-  //
-  //////////////////////////////////////////////////////////////////////////////////////////////////////
+NAMESPACE_BEGIN(Grid);
 
-  /* Gx
-   *  0 0  0  i    [0]+-i[3]
-   *  0 0  i  0    [1]+-i[2]
-   *  0 -i 0  0
-   * -i 0  0  0
-   */
 
-  // To fail is not to err (Cryptic clue: suggest to Google SFINAE ;) )
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spProjXp (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
-    {
-      hspin(0)=fspin(0)+timesI(fspin(3));
-      hspin(1)=fspin(1)+timesI(fspin(2));
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spProjXm (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
-    {
-      hspin(0)=fspin(0)-timesI(fspin(3));
-      hspin(1)=fspin(1)-timesI(fspin(2));
-    }
+//////////////////////////////////////////////////////////////////////////////////////////////////////
+// Normalisation alert; the g5   project is 1/2(1+-G5) 
+//                      the xyzt projects are (1+-Gxyzt)
+//
+// * xyzt project
+//
+// This is because this is how the Wilson operator is normally written as
+// (m+4r) - \frac{1}{2} D_{hop}
+// and / or
+// 1 - \frac{1}{2 m+8r} D_{hop} = 1 - kappa D_{hop}
+//
+// Note that the free, critical hopping parameter kappa is then 1/8 th for r=1.
+//
+// However, the xyzt 2spin "projectors" are not really projectors because they do not
+// square to 1, however the ChiralProjector is a true projector.
+//
+// For this reason there is NO provision in Grid of a four spinor result from the
+// xyzt projectors. They are intended to be used only in combination with "reconstruct" in the
+// wilson dirac operator and closely related actions.
+//
+// I also do NOT provide lattice wide operators of these, since the dirac operator is best implemented
+// via Stencils and single site variants will be needed only for the cache friendly high perf dirac operator.
+//
+// * chiral project
+//
+// Both four spinor and two spinor result variants are provided.
+//
+// The four spinor project will be recursively provided to Lattice wide routines, and likely used in 
+// the domain wall and mobius implementations.
+//
+//////////////////////////////////////////////////////////////////////////////////////////////////////
 
-      //  0 0  0  -1  [0] -+ [3]
-      //  0 0  1  0   [1] +- [2]
-      //  0 1  0  0
-      // -1 0  0  0
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spProjYp (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
-    {
-      hspin(0)=fspin(0)-fspin(3);
-      hspin(1)=fspin(1)+fspin(2);
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spProjYm (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
-    {
-      hspin(0)=fspin(0)+fspin(3);
-      hspin(1)=fspin(1)-fspin(2);
-    }
-	    /*Gz
-	     *  0 0  i  0   [0]+-i[2]
-	     *  0 0  0 -i   [1]-+i[3]
-	     * -i 0  0  0
-	     *  0 i  0  0
-	     */
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spProjZp (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
-    {
-      hspin(0)=fspin(0)+timesI(fspin(2));
-      hspin(1)=fspin(1)-timesI(fspin(3));
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spProjZm (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      hspin(0)=fspin(0)-timesI(fspin(2));
-      hspin(1)=fspin(1)+timesI(fspin(3));
-    }
-	    /*Gt
-	     *  0 0  1  0 [0]+-[2]
-	     *  0 0  0  1 [1]+-[3]
-	     *  1 0  0  0
-	     *  0 1  0  0
-	     */
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spProjTp (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      hspin(0)=fspin(0)+fspin(2);
-      hspin(1)=fspin(1)+fspin(3);
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spProjTm (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      hspin(0)=fspin(0)-fspin(2);
-      hspin(1)=fspin(1)-fspin(3);
-    }
-	    /*G5
-	     *  1 0  0  0 
-	     *  0 1  0  0 
-	     *  0 0 -1  0
-	     *  0 0  0 -1
-	     */
+/* Gx
+ *  0 0  0  i    [0]+-i[3]
+ *  0 0  i  0    [1]+-i[2]
+ *  0 -i 0  0
+ * -i 0  0  0
+ */
 
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spProj5p (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      hspin(0)=fspin(0);
-      hspin(1)=fspin(1);
-    }
+// To fail is not to err (Cryptic clue: suggest to Google SFINAE ;) )
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spProjXp (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
+{
+  hspin(0)=fspin(0)+timesI(fspin(3));
+  hspin(1)=fspin(1)+timesI(fspin(2));
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spProjXm (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
+{
+  hspin(0)=fspin(0)-timesI(fspin(3));
+  hspin(1)=fspin(1)-timesI(fspin(2));
+}
 
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spProj5m (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      hspin(0)=fspin(2);
-      hspin(1)=fspin(3);
-    }
+//  0 0  0  -1  [0] -+ [3]
+//  0 0  1  0   [1] +- [2]
+//  0 1  0  0
+// -1 0  0  0
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spProjYp (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
+{
+  hspin(0)=fspin(0)-fspin(3);
+  hspin(1)=fspin(1)+fspin(2);
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spProjYm (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
+{
+  hspin(0)=fspin(0)+fspin(3);
+  hspin(1)=fspin(1)-fspin(2);
+}
+/*Gz
+ *  0 0  i  0   [0]+-i[2]
+ *  0 0  0 -i   [1]-+i[3]
+ * -i 0  0  0
+ *  0 i  0  0
+ */
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spProjZp (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
+{
+  hspin(0)=fspin(0)+timesI(fspin(2));
+  hspin(1)=fspin(1)-timesI(fspin(3));
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spProjZm (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  hspin(0)=fspin(0)-timesI(fspin(2));
+  hspin(1)=fspin(1)+timesI(fspin(3));
+}
+/*Gt
+ *  0 0  1  0 [0]+-[2]
+ *  0 0  0  1 [1]+-[3]
+ *  1 0  0  0
+ *  0 1  0  0
+ */
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spProjTp (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  hspin(0)=fspin(0)+fspin(2);
+  hspin(1)=fspin(1)+fspin(3);
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spProjTm (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  hspin(0)=fspin(0)-fspin(2);
+  hspin(1)=fspin(1)-fspin(3);
+}
+/*G5
+ *  1 0  0  0 
+ *  0 1  0  0 
+ *  0 0 -1  0
+ *  0 0  0 -1
+ */
+
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spProj5p (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  hspin(0)=fspin(0);
+  hspin(1)=fspin(1);
+}
+
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spProj5m (iVector<vtype,Nhs> &hspin,const iVector<vtype,Ns> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  hspin(0)=fspin(2);
+  hspin(1)=fspin(3);
+}
   
-  //  template<class vtype> strong_inline void fspProj5p (iVector<vtype,Ns> &rfspin,const iVector<vtype,Ns> &fspin)
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spProj5p (iVector<vtype,Ns> &rfspin,const iVector<vtype,Ns> &fspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      rfspin(0)=fspin(0);
-      rfspin(1)=fspin(1);
-      rfspin(2)=zero;
-      rfspin(3)=zero;
-    }
-  //  template<class vtype> strong_inline void fspProj5m (iVector<vtype,Ns> &rfspin,const iVector<vtype,Ns> &fspin)
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spProj5m (iVector<vtype,Ns> &rfspin,const iVector<vtype,Ns> &fspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      rfspin(0)=zero;
-      rfspin(1)=zero;
-      rfspin(2)=fspin(2);
-      rfspin(3)=fspin(3);
-    }
+//  template<class vtype> accelerator_inline void fspProj5p (iVector<vtype,Ns> &rfspin,const iVector<vtype,Ns> &fspin)
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spProj5p (iVector<vtype,Ns> &rfspin,const iVector<vtype,Ns> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  rfspin(0)=fspin(0);
+  rfspin(1)=fspin(1);
+  rfspin(2)=Zero();
+  rfspin(3)=Zero();
+}
+//  template<class vtype> accelerator_inline void fspProj5m (iVector<vtype,Ns> &rfspin,const iVector<vtype,Ns> &fspin)
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spProj5m (iVector<vtype,Ns> &rfspin,const iVector<vtype,Ns> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  rfspin(0)=Zero();
+  rfspin(1)=Zero();
+  rfspin(2)=fspin(2);
+  rfspin(3)=fspin(3);
+}
 
-  ////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-  // Reconstruction routines to move back again to four spin
-  ////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-  /* Gx
-   *  0 0  0  i    [0]+-i[3]
-   *  0 0  i  0    [1]+-i[2]
-   *  0 -i 0  0  -i[1]+-[2]   == -i ([0]+-i[3]) = -i (1)
-   * -i 0  0  0  -i[0]+-[3]   == -i ([1]+-i[2]) = -i (0)
-   */
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spReconXp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)=hspin(0);
-      fspin(1)=hspin(1);
-      fspin(2)=timesMinusI(hspin(1));
-      fspin(3)=timesMinusI(hspin(0));
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spReconXm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)=hspin(0);
-      fspin(1)=hspin(1);
-      fspin(2)=timesI(hspin(1));
-      fspin(3)=timesI(hspin(0));
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void accumReconXp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)+=hspin(0);
-      fspin(1)+=hspin(1);
-      fspin(2)-=timesI(hspin(1));
-      fspin(3)-=timesI(hspin(0));
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void accumReconXm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)+=hspin(0);
-      fspin(1)+=hspin(1);
-      fspin(2)+=timesI(hspin(1));
-      fspin(3)+=timesI(hspin(0));
-    }
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// Reconstruction routines to move back again to four spin
+////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+/* Gx
+ *  0 0  0  i    [0]+-i[3]
+ *  0 0  i  0    [1]+-i[2]
+ *  0 -i 0  0  -i[1]+-[2]   == -i ([0]+-i[3]) = -i (1)
+ * -i 0  0  0  -i[0]+-[3]   == -i ([1]+-i[2]) = -i (0)
+ */
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spReconXp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)=hspin(0);
+  fspin(1)=hspin(1);
+  fspin(2)=timesMinusI(hspin(1));
+  fspin(3)=timesMinusI(hspin(0));
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spReconXm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)=hspin(0);
+  fspin(1)=hspin(1);
+  fspin(2)=timesI(hspin(1));
+  fspin(3)=timesI(hspin(0));
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void accumReconXp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)+=hspin(0);
+  fspin(1)+=hspin(1);
+  fspin(2)-=timesI(hspin(1));
+  fspin(3)-=timesI(hspin(0));
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void accumReconXm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)+=hspin(0);
+  fspin(1)+=hspin(1);
+  fspin(2)+=timesI(hspin(1));
+  fspin(3)+=timesI(hspin(0));
+}
 
-      //  0 0  0  -1  [0] -+ [3]
-      //  0 0  1  0   [1] +- [2]
-      //  0 1  0  0              == 1(1)
-      // -1 0  0  0              ==-1(0)
+//  0 0  0  -1  [0] -+ [3]
+//  0 0  1  0   [1] +- [2]
+//  0 1  0  0              == 1(1)
+// -1 0  0  0              ==-1(0)
 
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spReconYp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)=hspin(0);
-      fspin(1)=hspin(1);
-      fspin(2)= hspin(1);
-      fspin(3)=-hspin(0);//Unary minus?
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spReconYm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)=hspin(0);
-      fspin(1)=hspin(1);
-      fspin(2)=-hspin(1);
-      fspin(3)= hspin(0);
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void accumReconYp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)+=hspin(0);
-      fspin(1)+=hspin(1);
-      fspin(2)+=hspin(1);
-      fspin(3)-=hspin(0);
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void accumReconYm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)+=hspin(0);
-      fspin(1)+=hspin(1);
-      fspin(2)-=hspin(1);
-      fspin(3)+=hspin(0);
-    }
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spReconYp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)=hspin(0);
+  fspin(1)=hspin(1);
+  fspin(2)= hspin(1);
+  fspin(3)=-hspin(0);//Unary minus?
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spReconYm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)=hspin(0);
+  fspin(1)=hspin(1);
+  fspin(2)=-hspin(1);
+  fspin(3)= hspin(0);
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void accumReconYp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)+=hspin(0);
+  fspin(1)+=hspin(1);
+  fspin(2)+=hspin(1);
+  fspin(3)-=hspin(0);
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void accumReconYm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)+=hspin(0);
+  fspin(1)+=hspin(1);
+  fspin(2)-=hspin(1);
+  fspin(3)+=hspin(0);
+}
 
-	    /*Gz
-	     *  0 0  i  0   [0]+-i[2]
-	     *  0 0  0 -i   [1]-+i[3]
-	     * -i 0  0  0     => -i (0)
-	     *  0 i  0  0     =>  i (1)
-	     */
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spReconZp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)=hspin(0);
-      fspin(1)=hspin(1);
-      fspin(2)=timesMinusI(hspin(0));
-      fspin(3)=timesI(hspin(1));
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spReconZm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)=hspin(0);
-      fspin(1)=hspin(1);
-      fspin(2)=     timesI(hspin(0));
-      fspin(3)=timesMinusI(hspin(1));
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void accumReconZp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)+=hspin(0);
-      fspin(1)+=hspin(1);
-      fspin(2)-=timesI(hspin(0));
-      fspin(3)+=timesI(hspin(1));
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void accumReconZm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)+=hspin(0);
-      fspin(1)+=hspin(1);
-      fspin(2)+=timesI(hspin(0));
-      fspin(3)-=timesI(hspin(1));
-    }
-	    /*Gt
-	     *  0 0  1  0 [0]+-[2]
-	     *  0 0  0  1 [1]+-[3]
-	     *  1 0  0  0    => (0)
-	     *  0 1  0  0    => (1)
-	     */
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spReconTp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)=hspin(0);
-      fspin(1)=hspin(1);
-      fspin(2)=hspin(0);
-      fspin(3)=hspin(1);
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spReconTm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)=hspin(0);
-      fspin(1)=hspin(1);
-      fspin(2)=-hspin(0);
-      fspin(3)=-hspin(1);
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void accumReconTp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)+=hspin(0);
-      fspin(1)+=hspin(1);
-      fspin(2)+=hspin(0);
-      fspin(3)+=hspin(1);
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void accumReconTm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)+=hspin(0);
-      fspin(1)+=hspin(1);
-      fspin(2)-=hspin(0);
-      fspin(3)-=hspin(1);
-    }
-	    /*G5
-	     *  1 0  0  0 
-	     *  0 1  0  0 
-	     *  0 0 -1  0
-	     *  0 0  0 -1
-	     */
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spRecon5p (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)=hspin(0)+hspin(0); // add is lower latency than mul
-      fspin(1)=hspin(1)+hspin(1); // probably no measurable diffence though
-      fspin(2)=zero;
-      fspin(3)=zero;
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void spRecon5m (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)=zero;
-      fspin(1)=zero;
-      fspin(2)=hspin(0)+hspin(0);
-      fspin(3)=hspin(1)+hspin(1);
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void accumRecon5p (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(0)+=hspin(0)+hspin(0);
-      fspin(1)+=hspin(1)+hspin(1);
-    }
-  template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> strong_inline void accumRecon5m (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
-    {
-      //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
-      fspin(2)+=hspin(0)+hspin(0);
-      fspin(3)+=hspin(1)+hspin(1);
-    }
+/*Gz
+ *  0 0  i  0   [0]+-i[2]
+ *  0 0  0 -i   [1]-+i[3]
+ * -i 0  0  0     => -i (0)
+ *  0 i  0  0     =>  i (1)
+ */
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spReconZp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)=hspin(0);
+  fspin(1)=hspin(1);
+  fspin(2)=timesMinusI(hspin(0));
+  fspin(3)=timesI(hspin(1));
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spReconZm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)=hspin(0);
+  fspin(1)=hspin(1);
+  fspin(2)=     timesI(hspin(0));
+  fspin(3)=timesMinusI(hspin(1));
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void accumReconZp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)+=hspin(0);
+  fspin(1)+=hspin(1);
+  fspin(2)-=timesI(hspin(0));
+  fspin(3)+=timesI(hspin(1));
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void accumReconZm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)+=hspin(0);
+  fspin(1)+=hspin(1);
+  fspin(2)+=timesI(hspin(0));
+  fspin(3)-=timesI(hspin(1));
+}
+/*Gt
+ *  0 0  1  0 [0]+-[2]
+ *  0 0  0  1 [1]+-[3]
+ *  1 0  0  0    => (0)
+ *  0 1  0  0    => (1)
+ */
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spReconTp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)=hspin(0);
+  fspin(1)=hspin(1);
+  fspin(2)=hspin(0);
+  fspin(3)=hspin(1);
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spReconTm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)=hspin(0);
+  fspin(1)=hspin(1);
+  fspin(2)=-hspin(0);
+  fspin(3)=-hspin(1);
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void accumReconTp (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)+=hspin(0);
+  fspin(1)+=hspin(1);
+  fspin(2)+=hspin(0);
+  fspin(3)+=hspin(1);
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void accumReconTm (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)+=hspin(0);
+  fspin(1)+=hspin(1);
+  fspin(2)-=hspin(0);
+  fspin(3)-=hspin(1);
+}
+/*G5
+ *  1 0  0  0 
+ *  0 1  0  0 
+ *  0 0 -1  0
+ *  0 0  0 -1
+ */
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spRecon5p (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)=hspin(0)+hspin(0); // add is lower latency than mul
+  fspin(1)=hspin(1)+hspin(1); // probably no measurable diffence though
+  fspin(2)=Zero();
+  fspin(3)=Zero();
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void spRecon5m (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)=Zero();
+  fspin(1)=Zero();
+  fspin(2)=hspin(0)+hspin(0);
+  fspin(3)=hspin(1)+hspin(1);
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void accumRecon5p (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(0)+=hspin(0)+hspin(0);
+  fspin(1)+=hspin(1)+hspin(1);
+}
+template<class vtype,IfSpinor<iVector<vtype,Ns> > = 0> accelerator_inline void accumRecon5m (iVector<vtype,Ns> &fspin,const iVector<vtype,Nhs> &hspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,Ns>,SpinorIndex>::value,iVector<vtype,Ns> >::type *SFINAE;
+  fspin(2)+=hspin(0)+hspin(0);
+  fspin(3)+=hspin(1)+hspin(1);
+}
 
-  //////////////////////////////////////////////////////////////////////////////////////////////
-  // Recursively apply these until we hit the spin index
-  //////////////////////////////////////////////////////////////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////////////////////////
+// Recursively apply these until we hit the spin index
+//////////////////////////////////////////////////////////////////////////////////////////////
 
-  //////////
-  // Xp
-  //////////
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spProjXp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spProjXp(hspin._internal[i],fspin._internal[i]);
-    }
+//////////
+// Xp
+//////////
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spProjXp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spProjXp(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype> strong_inline void spProjXp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    spProjXp(hspin._internal,fspin._internal);
-  }
-  template<class rtype,class vtype,int N> strong_inline void spProjXp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype> accelerator_inline void spProjXp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  spProjXp(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N> accelerator_inline void spProjXp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spProjXp(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  template<class rtype,class vtype> strong_inline void spReconXp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    spReconXp(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void spReconXp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  spReconXp(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spReconXp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  for(int i=0;i<N;i++) {
+    spReconXp(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spReconXp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    for(int i=0;i<N;i++) {
-      spReconXp(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spReconXp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spReconXp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spReconXp(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
-  template<class rtype,class vtype> strong_inline void accumReconXp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    accumReconXp(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void accumReconXp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  accumReconXp(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void accumReconXp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  for(int i=0;i<N;i++) {
+    accumReconXp(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void accumReconXp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    for(int i=0;i<N;i++) {
-      accumReconXp(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void accumReconXp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void accumReconXp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       accumReconXp(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
 
-  ////////
-  // Xm
-  ////////
-  template<class rtype,class vtype> strong_inline void spProjXm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spProjXm(hspin._internal,fspin._internal);
+////////
+// Xm
+////////
+template<class rtype,class vtype> accelerator_inline void spProjXm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spProjXm(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spProjXm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spProjXm(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spProjXm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spProjXm(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spProjXm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spProjXm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spProjXm(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  template<class rtype,class vtype> strong_inline void spReconXm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spReconXm(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void spReconXm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spReconXm(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spReconXm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spReconXm(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spReconXm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spReconXm(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spReconXm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spReconXm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spReconXm(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
-  template<class rtype,class vtype> strong_inline void accumReconXm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    accumReconXm(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void accumReconXm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  accumReconXm(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void accumReconXm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    accumReconXm(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void accumReconXm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      accumReconXm(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void accumReconXm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void accumReconXm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       accumReconXm(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
 
-  ////////
-  // Yp
-  ////////
-  template<class rtype,class vtype> strong_inline void spProjYp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spProjYp(hspin._internal,fspin._internal);
+////////
+// Yp
+////////
+template<class rtype,class vtype> accelerator_inline void spProjYp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spProjYp(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spProjYp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spProjYp(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spProjYp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spProjYp(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spProjYp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spProjYp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spProjYp(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  template<class rtype,class vtype> strong_inline void spReconYp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spReconYp(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void spReconYp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spReconYp(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spReconYp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spReconYp(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spReconYp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spReconYp(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spReconYp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spReconYp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spReconYp(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
-  template<class rtype,class vtype> strong_inline void accumReconYp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    accumReconYp(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void accumReconYp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  accumReconYp(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void accumReconYp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    accumReconYp(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void accumReconYp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      accumReconYp(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void accumReconYp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void accumReconYp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       accumReconYp(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  ////////
-  // Ym
-  ////////
-  template<class rtype,class vtype> strong_inline void spProjYm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spProjYm(hspin._internal,fspin._internal);
+////////
+// Ym
+////////
+template<class rtype,class vtype> accelerator_inline void spProjYm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spProjYm(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spProjYm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spProjYm(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spProjYm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spProjYm(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spProjYm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spProjYm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spProjYm(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  template<class rtype,class vtype> strong_inline void spReconYm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spReconYm(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void spReconYm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spReconYm(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spReconYm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,const iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spReconYm(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spReconYm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,const iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spReconYm(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spReconYm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spReconYm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spReconYm(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
-  template<class rtype,class vtype> strong_inline void accumReconYm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    accumReconYm(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void accumReconYm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  accumReconYm(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void accumReconYm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    accumReconYm(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void accumReconYm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      accumReconYm(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void accumReconYm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void accumReconYm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       accumReconYm(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  ////////
-  // Zp
-  ////////
-  template<class rtype,class vtype> strong_inline void spProjZp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spProjZp(hspin._internal,fspin._internal);
+////////
+// Zp
+////////
+template<class rtype,class vtype> accelerator_inline void spProjZp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spProjZp(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spProjZp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spProjZp(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spProjZp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spProjZp(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spProjZp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spProjZp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spProjZp(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  template<class rtype,class vtype> strong_inline void spReconZp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spReconZp(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void spReconZp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spReconZp(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spReconZp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spReconZp(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spReconZp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spReconZp(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spReconZp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spReconZp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spReconZp(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
-  template<class rtype,class vtype> strong_inline void accumReconZp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    accumReconZp(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void accumReconZp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  accumReconZp(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void accumReconZp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    accumReconZp(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void accumReconZp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      accumReconZp(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void accumReconZp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void accumReconZp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       accumReconZp(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  ////////
-  // Zm
-  ////////
-  template<class rtype,class vtype> strong_inline void spProjZm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spProjZm(hspin._internal,fspin._internal);
+////////
+// Zm
+////////
+template<class rtype,class vtype> accelerator_inline void spProjZm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spProjZm(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spProjZm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spProjZm(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spProjZm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spProjZm(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spProjZm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spProjZm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spProjZm(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  template<class rtype,class vtype> strong_inline void spReconZm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spReconZm(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void spReconZm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spReconZm(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spReconZm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spReconZm(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spReconZm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spReconZm(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spReconZm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spReconZm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spReconZm(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
-  template<class rtype,class vtype> strong_inline void accumReconZm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    accumReconZm(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void accumReconZm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  accumReconZm(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void accumReconZm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    accumReconZm(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void accumReconZm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      accumReconZm(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void accumReconZm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void accumReconZm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       accumReconZm(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  ////////
-  // Tp
-  ////////
-  template<class rtype,class vtype> strong_inline void spProjTp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spProjTp(hspin._internal,fspin._internal);
+////////
+// Tp
+////////
+template<class rtype,class vtype> accelerator_inline void spProjTp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spProjTp(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spProjTp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spProjTp(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spProjTp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spProjTp(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spProjTp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spProjTp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spProjTp(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  template<class rtype,class vtype> strong_inline void spReconTp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spReconTp(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void spReconTp (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spReconTp(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spReconTp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spReconTp(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spReconTp (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spReconTp(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spReconTp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spReconTp (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spReconTp(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
-  template<class rtype,class vtype> strong_inline void accumReconTp (iScalar<rtype> &hspin, iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    accumReconTp(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void accumReconTp (iScalar<rtype> &hspin, iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  accumReconTp(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void accumReconTp (iVector<rtype,N> &hspin, const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    accumReconTp(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void accumReconTp (iVector<rtype,N> &hspin, const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      accumReconTp(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void accumReconTp (iMatrix<rtype,N> &hspin, const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void accumReconTp (iMatrix<rtype,N> &hspin, const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       accumReconTp(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  ////////
-  // Tm
-  ////////
-  template<class rtype,class vtype> strong_inline void spProjTm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spProjTm(hspin._internal,fspin._internal);
+////////
+// Tm
+////////
+template<class rtype,class vtype> accelerator_inline void spProjTm (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spProjTm(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spProjTm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spProjTm(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spProjTm (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spProjTm(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spProjTm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spProjTm (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spProjTm(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  template<class rtype,class vtype> strong_inline void spReconTm (iScalar<rtype> &hspin, const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spReconTm(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void spReconTm (iScalar<rtype> &hspin, const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spReconTm(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spReconTm (iVector<rtype,N> &hspin, const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spReconTm(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spReconTm (iVector<rtype,N> &hspin, const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spReconTm(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spReconTm (iMatrix<rtype,N> &hspin, const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spReconTm (iMatrix<rtype,N> &hspin, const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spReconTm(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
-  template<class rtype,class vtype> strong_inline void accumReconTm (iScalar<rtype> &hspin, const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    accumReconTm(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void accumReconTm (iScalar<rtype> &hspin, const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  accumReconTm(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void accumReconTm (iVector<rtype,N> &hspin, const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    accumReconTm(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void accumReconTm (iVector<rtype,N> &hspin, const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      accumReconTm(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void accumReconTm (iMatrix<rtype,N> &hspin, const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void accumReconTm (iMatrix<rtype,N> &hspin, const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       accumReconTm(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  ////////
-  // 5p
-  ////////
-  template<class rtype,class vtype> strong_inline void spProj5p (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spProj5p(hspin._internal,fspin._internal);
+////////
+// 5p
+////////
+template<class rtype,class vtype> accelerator_inline void spProj5p (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spProj5p(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spProj5p (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spProj5p(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spProj5p (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spProj5p(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spProj5p (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spProj5p (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spProj5p(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
-  template<class rtype,class vtype> strong_inline void spRecon5p (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spRecon5p(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void spRecon5p (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spRecon5p(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spRecon5p (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spRecon5p(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spRecon5p (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spRecon5p(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spRecon5p (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spRecon5p (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spRecon5p(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
-  template<class rtype,class vtype> strong_inline void accumRecon5p (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    accumRecon5p(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void accumRecon5p (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  accumRecon5p(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void accumRecon5p (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    accumRecon5p(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void accumRecon5p (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      accumRecon5p(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void accumRecon5p (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void accumRecon5p (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       accumRecon5p(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
-  // four spinor projectors for chiral proj
-  //  template<class vtype> strong_inline void fspProj5p (iScalar<vtype> &hspin,const iScalar<vtype> &fspin)
-  template<class vtype> strong_inline void spProj5p (iScalar<vtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spProj5p(hspin._internal,fspin._internal);
+// four spinor projectors for chiral proj
+//  template<class vtype> accelerator_inline void fspProj5p (iScalar<vtype> &hspin,const iScalar<vtype> &fspin)
+template<class vtype> accelerator_inline void spProj5p (iScalar<vtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spProj5p(hspin._internal,fspin._internal);
+}
+//  template<class vtype,int N> accelerator_inline void fspProj5p (iVector<vtype,N> &hspin,iVector<vtype,N> &fspin)
+template<class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spProj5p (iVector<vtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spProj5p(hspin._internal[i],fspin._internal[i]);
   }
-  //  template<class vtype,int N> strong_inline void fspProj5p (iVector<vtype,N> &hspin,iVector<vtype,N> &fspin)
-  template<class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spProj5p (iVector<vtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spProj5p(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  //  template<class vtype,int N> strong_inline void fspProj5p (iMatrix<vtype,N> &hspin,iMatrix<vtype,N> &fspin)
-  template<class vtype,int N> strong_inline void spProj5p (iMatrix<vtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+//  template<class vtype,int N> accelerator_inline void fspProj5p (iMatrix<vtype,N> &hspin,iMatrix<vtype,N> &fspin)
+template<class vtype,int N> accelerator_inline void spProj5p (iMatrix<vtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spProj5p(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  ////////
-  // 5m
-  ////////
+////////
+// 5m
+////////
 
-  template<class rtype,class vtype> strong_inline void spProj5m (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    spProj5m(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void spProj5m (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  spProj5m(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<rtype,N> > = 0> accelerator_inline void spProj5m (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  for(int i=0;i<N;i++) {
+    spProj5m(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<rtype,N> > = 0> strong_inline void spProj5m (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    for(int i=0;i<N;i++) {
-      spProj5m(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spProj5m (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spProj5m (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spProj5m(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
-  template<class rtype,class vtype> strong_inline void spRecon5m (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spRecon5m(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void spRecon5m (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spRecon5m(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spRecon5m (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spRecon5m(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spRecon5m (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spRecon5m(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void spRecon5m (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void spRecon5m (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spRecon5m(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
-  template<class rtype,class vtype> strong_inline void accumRecon5m (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    accumRecon5m(hspin._internal,fspin._internal);
+template<class rtype,class vtype> accelerator_inline void accumRecon5m (iScalar<rtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  accumRecon5m(hspin._internal,fspin._internal);
+}
+template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void accumRecon5m (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    accumRecon5m(hspin._internal[i],fspin._internal[i]);
   }
-  template<class rtype,class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void accumRecon5m (iVector<rtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      accumRecon5m(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  template<class rtype,class vtype,int N> strong_inline void accumRecon5m (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+template<class rtype,class vtype,int N> accelerator_inline void accumRecon5m (iMatrix<rtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       accumRecon5m(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
+}
 
 
-  // four spinor projectors for chiral proj
-  //  template<class vtype> strong_inline void fspProj5m (iScalar<vtype> &hspin,const iScalar<vtype> &fspin)
-  template<class vtype> strong_inline void spProj5m (iScalar<vtype> &hspin,const iScalar<vtype> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
-    spProj5m(hspin._internal,fspin._internal);
+// four spinor projectors for chiral proj
+//  template<class vtype> accelerator_inline void fspProj5m (iScalar<vtype> &hspin,const iScalar<vtype> &fspin)
+template<class vtype> accelerator_inline void spProj5m (iScalar<vtype> &hspin,const iScalar<vtype> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,SpinorIndex>::notvalue,iScalar<vtype> >::type *temp;
+  spProj5m(hspin._internal,fspin._internal);
+}
+//  template<class vtype,int N> accelerator_inline void fspProj5m (iVector<vtype,N> &hspin,iVector<vtype,N> &fspin)
+template<class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> accelerator_inline void spProj5m (iVector<vtype,N> &hspin,const iVector<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++) {
+    spProj5m(hspin._internal[i],fspin._internal[i]);
   }
-  //  template<class vtype,int N> strong_inline void fspProj5m (iVector<vtype,N> &hspin,iVector<vtype,N> &fspin)
-  template<class vtype,int N,IfNotSpinor<iVector<vtype,N> > = 0> strong_inline void spProj5m (iVector<vtype,N> &hspin,const iVector<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iVector<vtype,N>,SpinorIndex>::notvalue,iVector<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++) {
-      spProj5m(hspin._internal[i],fspin._internal[i]);
-    }
-  }
-  //  template<class vtype,int N> strong_inline void fspProj5m (iMatrix<vtype,N> &hspin,iMatrix<vtype,N> &fspin)
-  template<class vtype,int N> strong_inline void spProj5m (iMatrix<vtype,N> &hspin,const iMatrix<vtype,N> &fspin)
-  {
-    //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
-    for(int i=0;i<N;i++){ 
+}
+//  template<class vtype,int N> accelerator_inline void fspProj5m (iMatrix<vtype,N> &hspin,iMatrix<vtype,N> &fspin)
+template<class vtype,int N> accelerator_inline void spProj5m (iMatrix<vtype,N> &hspin,const iMatrix<vtype,N> &fspin)
+{
+  //typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,SpinorIndex>::notvalue,iMatrix<vtype,N> >::type *temp;
+  for(int i=0;i<N;i++){ 
     for(int j=0;j<N;j++){
       spProj5m(hspin._internal[i][j],fspin._internal[i][j]);
     }}
-  }
-}   //namespace QCD
-} // Grid
+}
+
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/qcd/utils/A2Autils.h b/Grid/qcd/utils/A2Autils.h
index 97188ffe..f50bb370 100644
--- a/Grid/qcd/utils/A2Autils.h
+++ b/Grid/qcd/utils/A2Autils.h
@@ -1,9 +1,8 @@
 #pragma once
 //#include <Grid/Hadrons/Global.hpp>
-#include <Grid/Eigen/unsupported/CXX11/Tensor>
+#include <Grid/Grid_Eigen_Tensor.h>
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 #undef DELTA_F_EQ_2
 
@@ -122,7 +121,7 @@ void A2Autils<FImpl>::MesonField(TensorType &mat,
   int Lblock = mat.dimension(3); 
   int Rblock = mat.dimension(4);
 
-  GridBase *grid = lhs_wi[0]._grid;
+  GridBase *grid = lhs_wi[0].Grid();
   
   const int    Nd = grid->_ndimension;
   const int Nsimd = grid->Nsimd();
@@ -142,14 +141,14 @@ void A2Autils<FImpl>::MesonField(TensorType &mat,
   int MFlvol = ld*Lblock*Rblock*Nmom;
 
   Vector<SpinMatrix_v > lvSum(MFrvol);
-  parallel_for (int r = 0; r < MFrvol; r++){
-    lvSum[r] = zero;
-  }
+  thread_for( r, MFrvol,{
+    lvSum[r] = Zero();
+  });
 
   Vector<SpinMatrix_s > lsSum(MFlvol);             
-  parallel_for (int r = 0; r < MFlvol; r++){
+  thread_for(r,MFlvol,{
     lsSum[r]=scalar_type(0.0);
-  }
+  });
 
   int e1=    grid->_slice_nblock[orthogdim];
   int e2=    grid->_slice_block [orthogdim];
@@ -157,7 +156,7 @@ void A2Autils<FImpl>::MesonField(TensorType &mat,
 
   // potentially wasting cores here if local time extent too small
   if (t_kernel) *t_kernel = -usecond();
-  parallel_for(int r=0;r<rd;r++){
+  thread_for(r,rd,{
 
     int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
 
@@ -168,12 +167,14 @@ void A2Autils<FImpl>::MesonField(TensorType &mat,
 
 	for(int i=0;i<Lblock;i++){
 
-	  auto left = conjugate(lhs_wi[i]._odata[ss]);
+	  auto lhs_v = lhs_wi[i].View();
+	  auto left = conjugate(lhs_v[ss]);
 
 	  for(int j=0;j<Rblock;j++){
 
 	    SpinMatrix_v vv;
-	    auto right = rhs_vj[j]._odata[ss];
+	    auto rhs_v = rhs_vj[j].View();
+	    auto right = rhs_v[ss];
 	    for(int s1=0;s1<Ns;s1++){
 	    for(int s2=0;s2<Ns;s2++){
 	      vv()(s1,s2)() = left()(s2)(0) * right()(s1)(0)
@@ -185,7 +186,8 @@ void A2Autils<FImpl>::MesonField(TensorType &mat,
 	    int base = Nmom*i+Nmom*Lblock*j+Nmom*Lblock*Rblock*r;
 	    for ( int m=0;m<Nmom;m++){
 	      int idx = m+base;
-	      auto phase = mom[m]._odata[ss];
+	      auto mom_v = mom[m].View();
+	      auto phase = mom_v[ss];
 	      mac(&lvSum[idx],&vv,&phase);
 	    }
 	  
@@ -193,14 +195,13 @@ void A2Autils<FImpl>::MesonField(TensorType &mat,
 	}
       }
     }
-  }
-
+  });
 
   // Sum across simd lanes in the plane, breaking out orthog dir.
-  parallel_for(int rt=0;rt<rd;rt++){
+  thread_for(rt,rd,{
 
-    std::vector<int> icoor(Nd);
-    std::vector<SpinMatrix_s> extracted(Nsimd);               
+    Coordinate icoor(Nd);
+    ExtractBuffer<SpinMatrix_s> extracted(Nsimd);               
 
     for(int i=0;i<Lblock;i++){
     for(int j=0;j<Rblock;j++){
@@ -222,7 +223,7 @@ void A2Autils<FImpl>::MesonField(TensorType &mat,
 
       }
     }}}
-  }
+  });
   if (t_kernel) *t_kernel += usecond();
   assert(mat.dimension(0) == Nmom);
   assert(mat.dimension(1) == Ngamma);
@@ -231,8 +232,7 @@ void A2Autils<FImpl>::MesonField(TensorType &mat,
   // ld loop and local only??
   int pd = grid->_processors[orthogdim];
   int pc = grid->_processor_coor[orthogdim];
-  parallel_for_nest2(int lt=0;lt<ld;lt++)
-  {
+  thread_for_collapse(2,lt,ld,{
     for(int pt=0;pt<pd;pt++){
       int t = lt + pt*ld;
       if (pt == pc){
@@ -260,7 +260,7 @@ void A2Autils<FImpl>::MesonField(TensorType &mat,
 	}
       }
     }
-  }
+  });
 
   ////////////////////////////////////////////////////////////////////
   // This global sum is taking as much as 50% of time on 16 nodes
@@ -311,7 +311,7 @@ void A2Autils<FImpl>::PionFieldXX(Eigen::Tensor<ComplexD,3> &mat,
   int Lblock = mat.dimension(1); 
   int Rblock = mat.dimension(2);
 
-  GridBase *grid = wi[0]._grid;
+  GridBase *grid = wi[0].Grid();
   
   const int    nd = grid->_ndimension;
   const int Nsimd = grid->Nsimd();
@@ -329,20 +329,20 @@ void A2Autils<FImpl>::PionFieldXX(Eigen::Tensor<ComplexD,3> &mat,
   int MFlvol = ld*Lblock*Rblock;
 
   Vector<vector_type > lvSum(MFrvol);
-  parallel_for (int r = 0; r < MFrvol; r++){
-    lvSum[r] = zero;
-  }
+  thread_for(r,MFrvol,{
+    lvSum[r] = Zero();
+  });
 
   Vector<scalar_type > lsSum(MFlvol);             
-  parallel_for (int r = 0; r < MFlvol; r++){
+  thread_for(r,MFlvol,{
     lsSum[r]=scalar_type(0.0);
-  }
+  });
 
   int e1=    grid->_slice_nblock[orthogdim];
   int e2=    grid->_slice_block [orthogdim];
   int stride=grid->_slice_stride[orthogdim];
 
-  parallel_for(int r=0;r<rd;r++){
+  thread_for(r,rd,{
 
     int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
 
@@ -353,7 +353,8 @@ void A2Autils<FImpl>::PionFieldXX(Eigen::Tensor<ComplexD,3> &mat,
 
 	for(int i=0;i<Lblock;i++){
 
-	  auto w = conjugate(wi[i]._odata[ss]);
+	  auto wi_v = wi[i].View();
+	  auto w = conjugate(wi_v[ss]);
 	  if (g5) {
 	    w()(2)(0) = - w()(2)(0);
 	    w()(2)(1) = - w()(2)(1);
@@ -363,8 +364,9 @@ void A2Autils<FImpl>::PionFieldXX(Eigen::Tensor<ComplexD,3> &mat,
 	    w()(3)(2) = - w()(3)(2);
 	  }
 	  for(int j=0;j<Rblock;j++){
-
-	    auto v = vj[j]._odata[ss];
+	    
+	    auto vj_v=vj[j].View();
+	    auto v  = vj_v[ss];
 	    auto vv = v()(0)(0);
 
 	    vv =      w()(0)(0) * v()(0)(0)// Gamma5 Dirac basis explicitly written out
@@ -386,14 +388,14 @@ void A2Autils<FImpl>::PionFieldXX(Eigen::Tensor<ComplexD,3> &mat,
 	}
       }
     }
-  }
+  });
 
   // Sum across simd lanes in the plane, breaking out orthog dir.
-  parallel_for(int rt=0;rt<rd;rt++){
+  thread_for(rt,rd,{
 
-    std::vector<int> icoor(nd);
+      Coordinate icoor(nd);
     iScalar<vector_type> temp; 
-    std::vector<iScalar<scalar_type> > extracted(Nsimd);               
+    ExtractBuffer<iScalar<scalar_type> > extracted(Nsimd);               
 
     for(int i=0;i<Lblock;i++){
     for(int j=0;j<Rblock;j++){
@@ -415,14 +417,13 @@ void A2Autils<FImpl>::PionFieldXX(Eigen::Tensor<ComplexD,3> &mat,
 
       }
     }}
-  }
+  });
 
   assert(mat.dimension(0) == Nt);
   // ld loop and local only??
   int pd = grid->_processors[orthogdim];
   int pc = grid->_processor_coor[orthogdim];
-  parallel_for_nest2(int lt=0;lt<ld;lt++)
-  {
+  thread_for_collapse(2,lt,ld,{
     for(int pt=0;pt<pd;pt++){
       int t = lt + pt*ld;
       if (pt == pc){
@@ -441,7 +442,7 @@ void A2Autils<FImpl>::PionFieldXX(Eigen::Tensor<ComplexD,3> &mat,
 	}
       }
     }
-  }
+  });
 
   grid->GlobalSumVector(&mat(0,0,0),Nt*Lblock*Rblock);
 }
@@ -456,7 +457,7 @@ void A2Autils<FImpl>::PionFieldWVmom(Eigen::Tensor<ComplexD,4> &mat,
   int Lblock = mat.dimension(2); 
   int Rblock = mat.dimension(3);
 
-  GridBase *grid = wi[0]._grid;
+  GridBase *grid = wi[0].Grid();
   
   const int    nd = grid->_ndimension;
   const int Nsimd = grid->Nsimd();
@@ -475,20 +476,20 @@ void A2Autils<FImpl>::PionFieldWVmom(Eigen::Tensor<ComplexD,4> &mat,
   int MFlvol = ld*Lblock*Rblock*Nmom;
 
   Vector<vector_type > lvSum(MFrvol);
-  parallel_for (int r = 0; r < MFrvol; r++){
-    lvSum[r] = zero;
-  }
+  thread_for(r,MFrvol,{
+    lvSum[r] = Zero();
+  });
 
   Vector<scalar_type > lsSum(MFlvol);             
-  parallel_for (int r = 0; r < MFlvol; r++){
+  thread_for(r,MFlvol,{
     lsSum[r]=scalar_type(0.0);
-  }
+  });
 
   int e1=    grid->_slice_nblock[orthogdim];
   int e2=    grid->_slice_block [orthogdim];
   int stride=grid->_slice_stride[orthogdim];
 
-  parallel_for(int r=0;r<rd;r++){
+  thread_for(r,rd,{
 
     int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
 
@@ -499,11 +500,13 @@ void A2Autils<FImpl>::PionFieldWVmom(Eigen::Tensor<ComplexD,4> &mat,
 
 	for(int i=0;i<Lblock;i++){
 
-	  auto w = conjugate(wi[i]._odata[ss]);
+	  auto wi_v = wi[i].View();
+	  auto w = conjugate(wi_v[ss]);
 
 	  for(int j=0;j<Rblock;j++){
-
-	    auto v = vj[j]._odata[ss];
+	    
+	    auto vj_v = vj[j].View();
+	    auto v = vj_v[ss];
 
 	    auto vv = w()(0)(0) * v()(0)(0)// Gamma5 Dirac basis explicitly written out
 	      +       w()(0)(1) * v()(0)(1)
@@ -523,22 +526,23 @@ void A2Autils<FImpl>::PionFieldWVmom(Eigen::Tensor<ComplexD,4> &mat,
 	    int base = Nmom*i+Nmom*Lblock*j+Nmom*Lblock*Rblock*r;
 	    for ( int m=0;m<Nmom;m++){
 	      int idx = m+base;
-	      auto phase = mom[m]._odata[ss];
+	      auto mom_v = mom[m].View();
+	      auto phase = mom_v[ss];
 	      mac(&lvSum[idx],&vv,&phase()()());
 	    }
 	  }
 	}
       }
     }
-  }
+  });
 
 
   // Sum across simd lanes in the plane, breaking out orthog dir.
-  parallel_for(int rt=0;rt<rd;rt++){
+  thread_for(rt,rd,{
 
-    std::vector<int> icoor(nd);
+    Coordinate icoor(nd);
     iScalar<vector_type> temp; 
-    std::vector<iScalar<scalar_type> > extracted(Nsimd);               
+    ExtractBuffer<iScalar<scalar_type> > extracted(Nsimd);               
 
     for(int i=0;i<Lblock;i++){
     for(int j=0;j<Rblock;j++){
@@ -561,15 +565,14 @@ void A2Autils<FImpl>::PionFieldWVmom(Eigen::Tensor<ComplexD,4> &mat,
 
       }
     }}}
-  }
+  });
 
   assert(mat.dimension(0) == Nmom);
   assert(mat.dimension(1) == Nt);
-
+ 
   int pd = grid->_processors[orthogdim];
   int pc = grid->_processor_coor[orthogdim];
-  parallel_for_nest2(int lt=0;lt<ld;lt++)
-  {
+  thread_for_collapse(2,lt,ld,{
     for(int pt=0;pt<pd;pt++){
       int t = lt + pt*ld;
       if (pt == pc){
@@ -592,7 +595,7 @@ void A2Autils<FImpl>::PionFieldWVmom(Eigen::Tensor<ComplexD,4> &mat,
 	}
       }
     }
-  }
+  });
 
   grid->GlobalSumVector(&mat(0,0,0,0),Nmom*Nt*Lblock*Rblock);
 }
@@ -660,7 +663,7 @@ void A2Autils<FImpl>::AslashField(TensorType &mat,
     int Lblock = mat.dimension(3); 
     int Rblock = mat.dimension(4);
 
-    GridBase *grid = lhs_wi[0]._grid;
+    GridBase *grid = lhs_wi[0].Grid();
     
     const int    Nd = grid->_ndimension;
     const int Nsimd = grid->Nsimd();
@@ -680,16 +683,16 @@ void A2Autils<FImpl>::AslashField(TensorType &mat,
     int MFlvol = ld*Lblock*Rblock*Nem;
 
     Vector<vector_type> lvSum(MFrvol);
-    parallel_for (int r = 0; r < MFrvol; r++)
+    thread_for(r,MFrvol,
     {
-        lvSum[r] = zero;
-    }
+      lvSum[r] = Zero();
+    });
 
     Vector<scalar_type> lsSum(MFlvol);             
-    parallel_for (int r = 0; r < MFlvol; r++)
+    thread_for(r,MFlvol,
     {
         lsSum[r] = scalar_type(0.0);
-    }
+    });
 
     int e1=    grid->_slice_nblock[orthogdim];
     int e2=    grid->_slice_block [orthogdim];
@@ -698,7 +701,7 @@ void A2Autils<FImpl>::AslashField(TensorType &mat,
     // Nested parallelism would be ok
     // Wasting cores here. Test case r
     if (t_kernel) *t_kernel = -usecond();
-    parallel_for(int r=0;r<rd;r++)
+    thread_for(r,rd,
     {
         int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
 
@@ -709,17 +712,19 @@ void A2Autils<FImpl>::AslashField(TensorType &mat,
 
             for(int i=0;i<Lblock;i++)
             {
-                auto left = conjugate(lhs_wi[i]._odata[ss]);
+  	        auto wi_v = lhs_wi[i].View();
+                auto left = conjugate(wi_v[ss]);
 
                 for(int j=0;j<Rblock;j++)
                 {
                     SpinMatrix_v vv;
-                    auto right = rhs_vj[j]._odata[ss];
+		    auto vj_v  = rhs_vj[j].View();
+                    auto right = vj_v[ss];
 
                     for(int s1=0;s1<Ns;s1++)
                     for(int s2=0;s2<Ns;s2++)
                     {
-                        vv()(s1,s2)() = left()(s2)(0) * right()(s1)(0)
+		          vv()(s1,s2)() = left()(s2)(0) * right()(s1)(0)
                                         + left()(s2)(1) * right()(s1)(1)
                                         + left()(s2)(2) * right()(s1)(2);
                     }
@@ -729,9 +734,11 @@ void A2Autils<FImpl>::AslashField(TensorType &mat,
 
                     for ( int m=0;m<Nem;m++)
                     {
+  		        auto emB0_v = emB0[m].View();
+  		        auto emB1_v = emB1[m].View();
                         int idx  = m+base;
-                        auto b0  = emB0[m]._odata[ss];
-                        auto b1  = emB1[m]._odata[ss];
+                        auto b0  = emB0_v[ss];
+                        auto b1  = emB1_v[ss];
                         auto cb0 = conjugate(b0);
                         auto cb1 = conjugate(b1);
 
@@ -743,13 +750,13 @@ void A2Autils<FImpl>::AslashField(TensorType &mat,
                 }
             }
         }
-    }
+    });
 
     // Sum across simd lanes in the plane, breaking out orthog dir.
-    parallel_for(int rt=0;rt<rd;rt++)
+    thread_for(rt,rd,
     {
-        std::vector<int> icoor(Nd);
-        std::vector<scalar_type> extracted(Nsimd);               
+        Coordinate icoor(Nd);
+        ExtractBuffer<scalar_type> extracted(Nsimd);               
 
         for(int i=0;i<Lblock;i++)
         for(int j=0;j<Rblock;j++)
@@ -769,13 +776,13 @@ void A2Autils<FImpl>::AslashField(TensorType &mat,
                 lsSum[ij_ldx]=lsSum[ij_ldx]+extracted[idx];
             }
         }
-    }
+    });
     if (t_kernel) *t_kernel += usecond();
 
     // ld loop and local only??
     int pd = grid->_processors[orthogdim];
     int pc = grid->_processor_coor[orthogdim];
-    parallel_for_nest2(int lt=0;lt<ld;lt++)
+    thread_for_collapse(2,lt,ld,
     {
         for(int pt=0;pt<pd;pt++)
         {
@@ -803,7 +810,7 @@ void A2Autils<FImpl>::AslashField(TensorType &mat,
                 }
             }
         }
-    }
+    });
     if (t_gsum) *t_gsum = -usecond();
     grid->GlobalSumVector(&mat(0,0,0,0,0),Nem*Nt*Lblock*Rblock);
     if (t_gsum) *t_gsum += usecond();
@@ -967,9 +974,9 @@ void A2Autils<FImpl>::ContractWWVV(std::vector<PropagatorField> &WWVV,
 				   const FermionField *vs,
 				   const FermionField *vd)
 {
-  GridBase *grid = vs[0]._grid;
+  GridBase *grid = vs[0].Grid();
 
-  int nd    = grid->_ndimension;
+  //  int nd    = grid->_ndimension;
   int Nsimd = grid->Nsimd();
   int N_t   = WW_sd.dimension(0);
   int N_s = WW_sd.dimension(1); 
@@ -978,42 +985,44 @@ void A2Autils<FImpl>::ContractWWVV(std::vector<PropagatorField> &WWVV,
   int d_unroll = 32;// Empirical optimisation
 
   for(int t=0;t<N_t;t++){
-    WWVV[t] = zero;
+    WWVV[t] = Zero();
   }
 
-  parallel_for(int ss=0;ss<grid->oSites();ss++){
+  thread_for(ss,grid->oSites(),{
     for(int d_o=0;d_o<N_d;d_o+=d_unroll){
       for(int t=0;t<N_t;t++){
       for(int s=0;s<N_s;s++){
-	auto tmp1 = vs[s]._odata[ss];
-  vobj tmp2 = zero;
-  vobj tmp3 = zero;
-
+	auto vs_v = vs[s].View();
+	auto tmp1 = vs_v[ss];
+	vobj tmp2 = Zero();
+	vobj tmp3 = Zero();
 	for(int d=d_o;d<MIN(d_o+d_unroll,N_d);d++){
+	  auto vd_v = vd[d].View();
 	  Scalar_v coeff = WW_sd(t,s,d);
-	  tmp3 = conjugate(vd[d]._odata[ss]);
+	  tmp3 = conjugate(vd_v[ss]);
 	  mac(&tmp2, &coeff, &tmp3);
-  }
+	}
 
 	//////////////////////////
 	// Fast outer product of tmp1 with a sum of terms suppressed by d_unroll
 	//////////////////////////
+	auto WWVV_v = WWVV[t].View();
 	for(int s1=0;s1<Ns;s1++){
 	for(int s2=0;s2<Ns;s2++){
-	  WWVV[t]._odata[ss]()(s1,s2)(0,0) += tmp1()(s1)(0)*tmp2()(s2)(0);
-	  WWVV[t]._odata[ss]()(s1,s2)(0,1) += tmp1()(s1)(0)*tmp2()(s2)(1);
-	  WWVV[t]._odata[ss]()(s1,s2)(0,2) += tmp1()(s1)(0)*tmp2()(s2)(2);
-	  WWVV[t]._odata[ss]()(s1,s2)(1,0) += tmp1()(s1)(1)*tmp2()(s2)(0);
-	  WWVV[t]._odata[ss]()(s1,s2)(1,1) += tmp1()(s1)(1)*tmp2()(s2)(1);
-	  WWVV[t]._odata[ss]()(s1,s2)(1,2) += tmp1()(s1)(1)*tmp2()(s2)(2);
-	  WWVV[t]._odata[ss]()(s1,s2)(2,0) += tmp1()(s1)(2)*tmp2()(s2)(0);
-	  WWVV[t]._odata[ss]()(s1,s2)(2,1) += tmp1()(s1)(2)*tmp2()(s2)(1);
-	  WWVV[t]._odata[ss]()(s1,s2)(2,2) += tmp1()(s1)(2)*tmp2()(s2)(2);
+	  WWVV_v[ss]()(s1,s2)(0,0) += tmp1()(s1)(0)*tmp2()(s2)(0);
+	  WWVV_v[ss]()(s1,s2)(0,1) += tmp1()(s1)(0)*tmp2()(s2)(1);
+	  WWVV_v[ss]()(s1,s2)(0,2) += tmp1()(s1)(0)*tmp2()(s2)(2);
+	  WWVV_v[ss]()(s1,s2)(1,0) += tmp1()(s1)(1)*tmp2()(s2)(0);
+	  WWVV_v[ss]()(s1,s2)(1,1) += tmp1()(s1)(1)*tmp2()(s2)(1);
+	  WWVV_v[ss]()(s1,s2)(1,2) += tmp1()(s1)(1)*tmp2()(s2)(2);
+	  WWVV_v[ss]()(s1,s2)(2,0) += tmp1()(s1)(2)*tmp2()(s2)(0);
+	  WWVV_v[ss]()(s1,s2)(2,1) += tmp1()(s1)(2)*tmp2()(s2)(1);
+	  WWVV_v[ss]()(s1,s2)(2,2) += tmp1()(s1)(2)*tmp2()(s2)(2);
 	}}
 
       }}
     }
-  }
+  });
 }
 
 
@@ -1028,17 +1037,21 @@ void A2Autils<FImpl>::ContractFourQuarkColourDiagonal(const PropagatorField &WWV
   assert(gamma0.size()==gamma1.size());
   int Ng = gamma0.size();
 
-  GridBase *grid = WWVV0._grid;
+  GridBase *grid = WWVV0.Grid();
 
-  parallel_for(int ss=0;ss<grid->oSites();ss++){
+  auto WWVV0_v = WWVV0.View();
+  auto WWVV1_v = WWVV1.View();
+  auto O_trtr_v= O_trtr.View();
+  auto O_fig8_v= O_fig8.View();
+  thread_for(ss,grid->oSites(),{
 
     typedef typename ComplexField::vector_object vobj;
 
     vobj v_trtr;
     vobj v_fig8;
 
-    auto VV0 = WWVV0._odata[ss];
-    auto VV1 = WWVV1._odata[ss];
+    auto VV0 = WWVV0_v[ss];
+    auto VV1 = WWVV1_v[ss];
     
     for(int g=0;g<Ng;g++){
 
@@ -1046,15 +1059,15 @@ void A2Autils<FImpl>::ContractFourQuarkColourDiagonal(const PropagatorField &WWV
       v_fig8 = trace(VV0 * gamma0[g] * VV1 * gamma1[g]);
 
       if ( g==0 ) {
-	O_trtr._odata[ss] = v_trtr; 
-	O_fig8._odata[ss] = v_fig8;
+	O_trtr_v[ss] = v_trtr; 
+	O_fig8_v[ss] = v_fig8;
       } else { 
-	O_trtr._odata[ss]+= v_trtr; 
-	O_fig8._odata[ss]+= v_fig8;
+	O_trtr_v[ss]+= v_trtr; 
+	O_fig8_v[ss]+= v_fig8;
       }
       
     }
-  }
+  });
 }
 
 template<class FImpl>
@@ -1068,22 +1081,27 @@ void A2Autils<FImpl>::ContractFourQuarkColourMix(const PropagatorField &WWVV0,
   assert(gamma0.size()==gamma1.size());
   int Ng = gamma0.size();
 
-  GridBase *grid = WWVV0._grid;
+  GridBase *grid = WWVV0.Grid();
 
-  parallel_for(int ss=0;ss<grid->oSites();ss++){
+  auto WWVV0_v = WWVV0.View();
+  auto WWVV1_v = WWVV1.View();
+  auto O_trtr_v= O_trtr.View();
+  auto O_fig8_v= O_fig8.View();
+
+  thread_for(ss,grid->oSites(),{
 
     typedef typename ComplexField::vector_object vobj;
 
-    auto VV0 = WWVV0._odata[ss];
-    auto VV1 = WWVV1._odata[ss];
+    auto VV0 = WWVV0_v[ss];
+    auto VV1 = WWVV1_v[ss];
     
     for(int g=0;g<Ng;g++){
 
       auto VV0G = VV0 * gamma0[g];  // Spin multiply
       auto VV1G = VV1 * gamma1[g];
 
-      vobj v_trtr=zero;
-      vobj v_fig8=zero;
+      vobj v_trtr=Zero();
+      vobj v_fig8=Zero();
 
       /////////////////////////////////////////
       // Colour mixed
@@ -1134,15 +1152,15 @@ Bag [8,4]  fig8 (-227.58,3.58808e-17) trtr (-32.5776,1.83286e-17)     //  - 1602
       }}}}
 
       if ( g==0 ) {
-	O_trtr._odata[ss] = v_trtr; 
-	O_fig8._odata[ss] = v_fig8;
+	O_trtr_v[ss] = v_trtr; 
+	O_fig8_v[ss] = v_fig8;
       } else { 
-	O_trtr._odata[ss]+= v_trtr; 
-	O_fig8._odata[ss]+= v_fig8;
+	O_trtr_v[ss]+= v_trtr; 
+	O_fig8_v[ss]+= v_fig8;
       }
       
     }
-  }
+  });
 }
 
 #ifdef DELTA_F_EQ_2
@@ -1164,7 +1182,7 @@ void A2Autils<FImpl>::DeltaFeq2(int dt_min,int dt_max,
 				const FermionField *vd,
 				int orthogdim)
 {
-  GridBase *grid = vs[0]._grid;
+  GridBase *grid = vs[0].Grid();
 
   LOG(Message) << "Computing A2A DeltaF=2 graph" << std::endl;
 
@@ -1216,32 +1234,32 @@ void A2Autils<FImpl>::DeltaFeq2(int dt_min,int dt_max,
     denom_P(t) =ComplexD(0.0);
   }
 
-  ComplexField D0(grid);   D0 = zero; // <P|A0> correlator from each wall
-  ComplexField D1(grid);   D1 = zero;
+  ComplexField D0(grid);   D0 = Zero(); // <P|A0> correlator from each wall
+  ComplexField D1(grid);   D1 = Zero();
 
-  ComplexField O1_trtr(grid);  O1_trtr = zero;
-  ComplexField O2_trtr(grid);  O2_trtr = zero;
-  ComplexField O3_trtr(grid);  O3_trtr = zero;
-  ComplexField O4_trtr(grid);  O4_trtr = zero;
-  ComplexField O5_trtr(grid);  O5_trtr = zero;
+  ComplexField O1_trtr(grid);  O1_trtr = Zero();
+  ComplexField O2_trtr(grid);  O2_trtr = Zero();
+  ComplexField O3_trtr(grid);  O3_trtr = Zero();
+  ComplexField O4_trtr(grid);  O4_trtr = Zero();
+  ComplexField O5_trtr(grid);  O5_trtr = Zero();
 
-  ComplexField O1_fig8(grid);  O1_fig8 = zero;
-  ComplexField O2_fig8(grid);  O2_fig8 = zero;
-  ComplexField O3_fig8(grid);  O3_fig8 = zero;
-  ComplexField O4_fig8(grid);  O4_fig8 = zero;
-  ComplexField O5_fig8(grid);  O5_fig8 = zero;
+  ComplexField O1_fig8(grid);  O1_fig8 = Zero();
+  ComplexField O2_fig8(grid);  O2_fig8 = Zero();
+  ComplexField O3_fig8(grid);  O3_fig8 = Zero();
+  ComplexField O4_fig8(grid);  O4_fig8 = Zero();
+  ComplexField O5_fig8(grid);  O5_fig8 = Zero();
 
-  ComplexField VV_trtr(grid);  VV_trtr = zero;
-  ComplexField AA_trtr(grid);  AA_trtr = zero;
-  ComplexField SS_trtr(grid);  SS_trtr = zero;
-  ComplexField PP_trtr(grid);  PP_trtr = zero;
-  ComplexField TT_trtr(grid);  TT_trtr = zero;
+  ComplexField VV_trtr(grid);  VV_trtr = Zero();
+  ComplexField AA_trtr(grid);  AA_trtr = Zero();
+  ComplexField SS_trtr(grid);  SS_trtr = Zero();
+  ComplexField PP_trtr(grid);  PP_trtr = Zero();
+  ComplexField TT_trtr(grid);  TT_trtr = Zero();
 
-  ComplexField VV_fig8(grid);  VV_fig8 = zero;
-  ComplexField AA_fig8(grid);  AA_fig8 = zero;
-  ComplexField SS_fig8(grid);  SS_fig8 = zero;
-  ComplexField PP_fig8(grid);  PP_fig8 = zero;
-  ComplexField TT_fig8(grid);  TT_fig8 = zero;
+  ComplexField VV_fig8(grid);  VV_fig8 = Zero();
+  ComplexField AA_fig8(grid);  AA_fig8 = Zero();
+  ComplexField SS_fig8(grid);  SS_fig8 = Zero();
+  ComplexField PP_fig8(grid);  PP_fig8 = Zero();
+  ComplexField TT_fig8(grid);  TT_fig8 = Zero();
 
   //////////////////////////////////////////////////
   // Used to store appropriate correlation funcs
@@ -1376,5 +1394,5 @@ void A2Autils<FImpl>::DeltaFeq2(int dt_min,int dt_max,
 }
 #endif 
 
-}}
+NAMESPACE_END(Grid);
 
diff --git a/Grid/qcd/utils/CovariantCshift.h b/Grid/qcd/utils/CovariantCshift.h
index 2f7561f8..ed96f3bf 100644
--- a/Grid/qcd/utils/CovariantCshift.h
+++ b/Grid/qcd/utils/CovariantCshift.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,13 +24,13 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef QCD_UTILS_COVARIANT_CSHIFT_H
 #define QCD_UTILS_COVARIANT_CSHIFT_H
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
+
 ////////////////////////////////////////////////////////////////////////
 // Low performance implementation of CovariantCshift API
 ////////////////////////////////////////////////////////////////////////
@@ -39,8 +39,8 @@ namespace QCD {
 namespace PeriodicBC { 
 
   template<class covariant,class gauge> Lattice<covariant> CovShiftForward(const Lattice<gauge> &Link, 
-									    int mu,
-									    const Lattice<covariant> &field)
+									   int mu,
+									   const Lattice<covariant> &field)
   {
     return Link*Cshift(field,mu,1);// moves towards negative mu
   }
@@ -48,7 +48,7 @@ namespace PeriodicBC {
 									    int mu,
 									    const Lattice<covariant> &field)
   {
-    Lattice<covariant> tmp(field._grid);
+    Lattice<covariant> tmp(field.Grid());
     tmp = adj(Link)*field;
     return Cshift(tmp,mu,-1);// moves towards positive mu
   }
@@ -73,21 +73,21 @@ namespace ConjugateBC {
   //                      U  U^* U^* U^T U^adj =  U  (U U U^dag U^T )^*
   //                                           =  U  (U U U^dag)^* ( U^T )^*
   //
-  // So covariant shift rule: conjugate inward shifted plane when crossing boundary applies.
+  // So covariant shift rule: Conjugate inward shifted plane when crossing boundary applies.
   //
-  // This conjugate should be applied to BOTH the link and the covariant field on backward shift
+  // This Conjugate should be applied to BOTH the link and the covariant field on backward shift
   // boundary wrap.
   // 
   //      |  |              
   // xxxxxxxxxxxxxxxxx         
-  //      |  | <---- this link is conjugated, and the path leading into it. Segment crossing in and out is double conjugated.
+  //      |  | <---- this link is Conjugated, and the path leading into it. Segment crossing in and out is double Conjugated.
   //       -- 
   //    ------->
   template<class covariant,class gauge> Lattice<covariant> CovShiftForward(const Lattice<gauge> &Link, 
-									    int mu,
-									    const Lattice<covariant> &field)
+									   int mu,
+									   const Lattice<covariant> &field)
   {
-    GridBase * grid = Link._grid;
+    GridBase * grid = Link.Grid();
 
     int Lmu = grid->GlobalDimensions()[mu]-1;
 
@@ -106,7 +106,7 @@ namespace ConjugateBC {
 									    int mu,
 									    const Lattice<covariant> &field)
   {
-    GridBase * grid = field._grid;
+    GridBase * grid = field.Grid();
 
     int Lmu = grid->GlobalDimensions()[mu]-1;
 
@@ -122,9 +122,8 @@ namespace ConjugateBC {
     return Cshift(tmp,mu,-1);// moves towards positive mu
   }
 
-
 }
 
 
-}}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/qcd/utils/CovariantLaplacian.h b/Grid/qcd/utils/CovariantLaplacian.h
index 0c99b03e..0e0620a7 100644
--- a/Grid/qcd/utils/CovariantLaplacian.h
+++ b/Grid/qcd/utils/CovariantLaplacian.h
@@ -25,13 +25,10 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
+#pragma once 
 
-#ifndef COVARIANT_LAPLACIAN_H
-#define COVARIANT_LAPLACIAN_H
-
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 struct LaplacianParams : Serializable {
   GRID_SERIALIZABLE_CLASS_MEMBERS(LaplacianParams, 
@@ -80,19 +77,19 @@ class LaplacianAdjointField: public Metric<typename Impl::Field> {
   MultiShiftFunction PowerHalf;    
   MultiShiftFunction PowerInvHalf;    
 
- public:
+public:
   INHERIT_GIMPL_TYPES(Impl);
 
   LaplacianAdjointField(GridBase* grid, OperatorFunction<GaugeField>& S, LaplacianParams& p, const RealD k = 1.0)
-      : U(Nd, grid), Solver(S), param(p), kappa(k){
-        AlgRemez remez(param.lo,param.hi,param.precision);
-        std::cout<<GridLogMessage << "Generating degree "<<param.degree<<" for x^(1/2)"<<std::endl;
-        remez.generateApprox(param.degree,1,2);
-        PowerHalf.Init(remez,param.tolerance,false);
-        PowerInvHalf.Init(remez,param.tolerance,true);
+    : U(Nd, grid), Solver(S), param(p), kappa(k){
+    AlgRemez remez(param.lo,param.hi,param.precision);
+    std::cout<<GridLogMessage << "Generating degree "<<param.degree<<" for x^(1/2)"<<std::endl;
+    remez.generateApprox(param.degree,1,2);
+    PowerHalf.Init(remez,param.tolerance,false);
+    PowerInvHalf.Init(remez,param.tolerance,true);
         
 
-      };
+  };
 
   void Mdir(const GaugeField&, GaugeField&, int, int){ assert(0);}
   void Mdiag(const GaugeField&, GaugeField&){ assert(0);}
@@ -109,14 +106,14 @@ class LaplacianAdjointField: public Metric<typename Impl::Field> {
     //GaugeField herm = in + adj(in);
     //std::cout << "AHermiticity: " << norm2(herm) << std::endl;
 
-    GaugeLinkField tmp(in._grid);
-    GaugeLinkField tmp2(in._grid);
-    GaugeLinkField sum(in._grid);
+    GaugeLinkField tmp(in.Grid());
+    GaugeLinkField tmp2(in.Grid());
+    GaugeLinkField sum(in.Grid());
 
     for (int nu = 0; nu < Nd; nu++) {
-      sum = zero;
+      sum = Zero();
       GaugeLinkField in_nu = PeekIndex<LorentzIndex>(in, nu);
-      GaugeLinkField out_nu(out._grid);
+      GaugeLinkField out_nu(out.Grid());
       for (int mu = 0; mu < Nd; mu++) {
         tmp = U[mu] * Cshift(in_nu, mu, +1) * adj(U[mu]);
         tmp2 = adj(U[mu]) * in_nu * U[mu];
@@ -132,8 +129,8 @@ class LaplacianAdjointField: public Metric<typename Impl::Field> {
     RealD factor = -kappa / (double(4 * Nd));
     
     for (int mu = 0; mu < Nd; mu++){
-      GaugeLinkField der_mu(der._grid);
-      der_mu = zero;
+      GaugeLinkField der_mu(der.Grid());
+      der_mu = Zero();
       for (int nu = 0; nu < Nd; nu++){
         GaugeLinkField in_nu = PeekIndex<LorentzIndex>(in, nu);
         der_mu += U[mu] * Cshift(in_nu, mu, 1) * adj(U[mu]) * in_nu;
@@ -151,8 +148,8 @@ class LaplacianAdjointField: public Metric<typename Impl::Field> {
     RealD factor = -kappa / (double(4 * Nd));
 
     for (int mu = 0; mu < Nd; mu++) {
-      GaugeLinkField der_mu(der._grid);
-      der_mu = zero;
+      GaugeLinkField der_mu(der.Grid());
+      der_mu = Zero();
       for (int nu = 0; nu < Nd; nu++) {
         GaugeLinkField left_nu = PeekIndex<LorentzIndex>(left, nu);
         GaugeLinkField right_nu = PeekIndex<LorentzIndex>(right, nu);
@@ -169,7 +166,7 @@ class LaplacianAdjointField: public Metric<typename Impl::Field> {
   }
 
   void MSquareRoot(GaugeField& P){
-    GaugeField Gp(P._grid);
+    GaugeField Gp(P.Grid());
     HermitianLinearOperator<LaplacianAdjointField<Impl>,GaugeField> HermOp(*this);
     ConjugateGradientMultiShift<GaugeField> msCG(param.MaxIter,PowerHalf);
     msCG(HermOp,P,Gp);
@@ -177,7 +174,7 @@ class LaplacianAdjointField: public Metric<typename Impl::Field> {
   }
 
   void MInvSquareRoot(GaugeField& P){
-    GaugeField Gp(P._grid);
+    GaugeField Gp(P.Grid());
     HermitianLinearOperator<LaplacianAdjointField<Impl>,GaugeField> HermOp(*this);
     ConjugateGradientMultiShift<GaugeField> msCG(param.MaxIter,PowerInvHalf);
     msCG(HermOp,P,Gp);
@@ -186,12 +183,9 @@ class LaplacianAdjointField: public Metric<typename Impl::Field> {
 
 
 
- private:
+private:
   RealD kappa;
   std::vector<GaugeLinkField> U;
 };
 
-}
-}
-
-#endif
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/utils/CovariantSmearing.h b/Grid/qcd/utils/CovariantSmearing.h
index eae5ed71..7feddea9 100644
--- a/Grid/qcd/utils/CovariantSmearing.h
+++ b/Grid/qcd/utils/CovariantSmearing.h
@@ -43,7 +43,7 @@ public:
 			    T& chi, 
 			    const Real& width, int Iterations, int orthog)
   {
-    GridBase *grid = chi._grid;
+    GridBase *grid = chi.Grid();
     T psi(grid);
 
     ////////////////////////////////////////////////////////////////////////////////////
diff --git a/Grid/qcd/utils/GaugeFix.h b/Grid/qcd/utils/GaugeFix.h
index 67f84f14..2b3384da 100644
--- a/Grid/qcd/utils/GaugeFix.h
+++ b/Grid/qcd/utils/GaugeFix.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     grid` physics library, www.github.com/paboyle/Grid 
 
@@ -22,19 +22,19 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 //#include <Grid/Grid.h>
 
 #ifndef GRID_QCD_GAUGE_FIX_H
 #define GRID_QCD_GAUGE_FIX_H
-namespace Grid {
-namespace QCD {
+
+NAMESPACE_BEGIN(Grid);
 
 
 template <class Gimpl> 
 class FourierAcceleratedGaugeFixer  : public Gimpl {
- public:
+public:
   INHERIT_GIMPL_TYPES(Gimpl);
 
   typedef typename Gimpl::GaugeLinkField GaugeMat;
@@ -47,7 +47,7 @@ class FourierAcceleratedGaugeFixer  : public Gimpl {
     }
   }
   static void DmuAmu(const std::vector<GaugeMat> &A,GaugeMat &dmuAmu,int orthog) {
-    dmuAmu=zero;
+    dmuAmu=Zero();
     for(int mu=0;mu<Nd;mu++){
       if ( mu != orthog ) {
 	dmuAmu = dmuAmu + A[mu] - Cshift(A[mu],mu,-1);
@@ -56,13 +56,13 @@ class FourierAcceleratedGaugeFixer  : public Gimpl {
   }  
 
   static void SteepestDescentGaugeFix(GaugeLorentz &Umu,Real & alpha,int maxiter,Real Omega_tol, Real Phi_tol,bool Fourier=false,int orthog=-1) {
-    GridBase *grid = Umu._grid;
+    GridBase *grid = Umu.Grid();
     GaugeMat xform(grid);
     SteepestDescentGaugeFix(Umu,xform,alpha,maxiter,Omega_tol,Phi_tol,Fourier,orthog);
   }
   static void SteepestDescentGaugeFix(GaugeLorentz &Umu,GaugeMat &xform,Real & alpha,int maxiter,Real Omega_tol, Real Phi_tol,bool Fourier=false,int orthog=-1) {
 
-    GridBase *grid = Umu._grid;
+    GridBase *grid = Umu.Grid();
 
     Real org_plaq      =WilsonLoops<Gimpl>::avgPlaquette(Umu);
     Real org_link_trace=WilsonLoops<Gimpl>::linkTrace(Umu); 
@@ -72,7 +72,6 @@ class FourierAcceleratedGaugeFixer  : public Gimpl {
     xform=1.0;
 
     std::vector<GaugeMat> U(Nd,grid);
-
     GaugeMat dmuAmu(grid);
 
     {
@@ -125,7 +124,7 @@ class FourierAcceleratedGaugeFixer  : public Gimpl {
     }
   };
   static Real SteepestDescentStep(std::vector<GaugeMat> &U,GaugeMat &xform,Real & alpha, GaugeMat & dmuAmu,int orthog) {
-    GridBase *grid = U[0]._grid;
+    GridBase *grid = U[0].Grid();
 
     std::vector<GaugeMat> A(Nd,grid);
     GaugeMat g(grid);
@@ -145,14 +144,14 @@ class FourierAcceleratedGaugeFixer  : public Gimpl {
 
   static Real FourierAccelSteepestDescentStep(std::vector<GaugeMat> &U,GaugeMat &xform,Real & alpha, GaugeMat & dmuAmu,int orthog) {
 
-    GridBase *grid = U[0]._grid;
+    GridBase *grid = U[0].Grid();
 
     Real vol = grid->gSites();
 
     FFT theFFT((GridCartesian *)grid);
 
     LatticeComplex  Fp(grid);
-    LatticeComplex  psq(grid); psq=zero;
+    LatticeComplex  psq(grid); psq=Zero();
     LatticeComplex  pmu(grid); 
     LatticeComplex   one(grid); one = Complex(1.0,0.0);
 
@@ -172,8 +171,8 @@ class FourierAcceleratedGaugeFixer  : public Gimpl {
     // Work out Fp = psq_max/ psq...
     // Avoid singularities in Fp
     //////////////////////////////////
-    std::vector<int> latt_size = grid->GlobalDimensions();
-    std::vector<int> coor(grid->_ndimension,0);
+    Coordinate latt_size = grid->GlobalDimensions();
+    Coordinate coor(grid->_ndimension,0);
     for(int mu=0;mu<Nd;mu++) {
       if ( mu != orthog ) { 
 	Real TwoPiL =  M_PI * 2.0/ latt_size[mu];
@@ -212,7 +211,7 @@ class FourierAcceleratedGaugeFixer  : public Gimpl {
   }
 
   static void ExpiAlphaDmuAmu(const std::vector<GaugeMat> &A,GaugeMat &g,Real & alpha, GaugeMat &dmuAmu,int orthog) {
-    GridBase *grid = g._grid;
+    GridBase *grid = g.Grid();
     Complex cialpha(0.0,-alpha);
     GaugeMat ciadmam(grid);
     DmuAmu(A,dmuAmu,orthog);
@@ -221,6 +220,6 @@ class FourierAcceleratedGaugeFixer  : public Gimpl {
   }  
 };
 
-}
-}
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/qcd/utils/LinalgUtils.h b/Grid/qcd/utils/LinalgUtils.h
index 04a224e5..5974f4ed 100644
--- a/Grid/qcd/utils/LinalgUtils.h
+++ b/Grid/qcd/utils/LinalgUtils.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,13 +25,12 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef GRID_QCD_LINALG_UTILS_H
-#define GRID_QCD_LINALG_UTILS_H
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+NAMESPACE_BEGIN(Grid);
 
-namespace Grid{
-namespace QCD{
 ////////////////////////////////////////////////////////////////////////
 //This file brings additional linear combination assist that is helpful
 //to QCD such as chiral projectors and spin matrices applied to one of the inputs.
@@ -42,170 +41,200 @@ namespace QCD{
 template<class vobj,class Coeff>
 void axpibg5x(Lattice<vobj> &z,const Lattice<vobj> &x,Coeff a,Coeff b)
 {
-  z.checkerboard = x.checkerboard;
+  z.Checkerboard() = x.Checkerboard();
   conformable(x,z);
 
-  GridBase *grid=x._grid;
+  GridBase *grid=x.Grid();
 
   Gamma G5(Gamma::Algebra::Gamma5);
-  parallel_for(int ss=0;ss<grid->oSites();ss++){
-    vobj tmp;
-    tmp = a*x._odata[ss];
-    tmp = tmp + G5*(b*timesI(x._odata[ss]));
-    vstream(z._odata[ss],tmp);
-  }
+  auto x_v = x.View();
+  auto z_v = z.View();
+  accelerator_for( ss, x_v.size(),vobj::Nsimd(), {
+    auto tmp = a*x_v(ss) + G5*(b*timesI(x_v(ss)));
+    coalescedWrite(z_v[ss],tmp);
+  });
 }
 
 template<class vobj,class Coeff> 
 void axpby_ssp(Lattice<vobj> &z, Coeff a,const Lattice<vobj> &x,Coeff b,const Lattice<vobj> &y,int s,int sp)
 {
-  z.checkerboard = x.checkerboard;
+  z.Checkerboard() = x.Checkerboard();
   conformable(x,y);
   conformable(x,z);
-  GridBase *grid=x._grid;
+  GridBase *grid=x.Grid();
   int Ls = grid->_rdimensions[0];
-  parallel_for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
-    vobj tmp = a*x._odata[ss+s]+b*y._odata[ss+sp];
-    vstream(z._odata[ss+s],tmp);
-  }
+  auto x_v = x.View();
+  auto y_v = y.View();
+  auto z_v = z.View();
+  // FIXME -- need a new class of accelerator_loop to implement this
+  //
+  uint64_t nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,vobj::Nsimd(),{
+    uint64_t ss = sss*Ls;
+    auto tmp = a*x_v(ss+s)+b*y_v(ss+sp);
+    coalescedWrite(z_v[ss+s],tmp);
+  });
 }
 
 template<class vobj,class Coeff> 
 void ag5xpby_ssp(Lattice<vobj> &z,Coeff a,const Lattice<vobj> &x,Coeff b,const Lattice<vobj> &y,int s,int sp)
 {
-  z.checkerboard = x.checkerboard;
+  z.Checkerboard() = x.Checkerboard();
   conformable(x,y);
   conformable(x,z);
-  GridBase *grid=x._grid;
+  GridBase *grid=x.Grid();
   int Ls = grid->_rdimensions[0];
-
   Gamma G5(Gamma::Algebra::Gamma5);
-  parallel_for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
-    vobj tmp;
-    tmp = G5*x._odata[ss+s]*a;
-    tmp = tmp + b*y._odata[ss+sp];
-    vstream(z._odata[ss+s],tmp);
-  }
+  auto x_v = x.View();
+  auto y_v = y.View();
+  auto z_v = z.View();
+  uint64_t nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,vobj::Nsimd(),{
+    uint64_t ss = sss*Ls;
+    auto tmp = G5*x_v(ss+s)*a + b*y_v(ss+sp);
+    coalescedWrite(z_v[ss+s],tmp);
+  });
 }
 
 template<class vobj,class Coeff> 
 void axpbg5y_ssp(Lattice<vobj> &z,Coeff a,const Lattice<vobj> &x,Coeff b,const Lattice<vobj> &y,int s,int sp)
 {
-  z.checkerboard = x.checkerboard;
+  z.Checkerboard() = x.Checkerboard();
   conformable(x,y);
   conformable(x,z);
-  GridBase *grid=x._grid;
+  GridBase *grid=x.Grid();
   int Ls = grid->_rdimensions[0];
+  auto x_v = x.View();
+  auto y_v = y.View();
+  auto z_v = z.View();
   Gamma G5(Gamma::Algebra::Gamma5);
-  parallel_for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
-    vobj tmp;
-    tmp = G5*y._odata[ss+sp]*b;
-    tmp = tmp + a*x._odata[ss+s];
-    vstream(z._odata[ss+s],tmp);
-  }
+  uint64_t nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,vobj::Nsimd(),{
+    uint64_t ss = sss*Ls;
+    auto tmp = G5*y_v(ss+sp)*b + a*x_v(ss+s);
+    coalescedWrite(z_v[ss+s],tmp);
+  });
 }
 
 template<class vobj,class Coeff> 
 void ag5xpbg5y_ssp(Lattice<vobj> &z,Coeff a,const Lattice<vobj> &x,Coeff b,const Lattice<vobj> &y,int s,int sp)
 {
-  z.checkerboard = x.checkerboard;
+  z.Checkerboard() = x.Checkerboard();
   conformable(x,y);
   conformable(x,z);
-  GridBase *grid=x._grid;
+  GridBase *grid=x.Grid();
   int Ls = grid->_rdimensions[0];
 
+  auto x_v = x.View();
+  auto y_v = y.View();
+  auto z_v = z.View();
   Gamma G5(Gamma::Algebra::Gamma5);
-  parallel_for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
-    vobj tmp1;
-    vobj tmp2;
-    tmp1 = a*x._odata[ss+s]+b*y._odata[ss+sp];
-    tmp2 = G5*tmp1;
-    vstream(z._odata[ss+s],tmp2);
-  }
+  uint64_t nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,vobj::Nsimd(),{
+    uint64_t ss = sss*Ls;
+    auto tmp1 = a*x_v(ss+s)+b*y_v(ss+sp);
+    auto tmp2 = G5*tmp1;
+    coalescedWrite(z_v[ss+s],tmp2);
+  });
 }
 
 template<class vobj,class Coeff> 
 void axpby_ssp_pminus(Lattice<vobj> &z,Coeff a,const Lattice<vobj> &x,Coeff b,const Lattice<vobj> &y,int s,int sp)
 {
-  z.checkerboard = x.checkerboard;
+  z.Checkerboard() = x.Checkerboard();
   conformable(x,y);
   conformable(x,z);
-  GridBase *grid=x._grid;
+  GridBase *grid=x.Grid();
   int Ls = grid->_rdimensions[0];
-  parallel_for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
-    vobj tmp;
-    spProj5m(tmp,y._odata[ss+sp]);
-    tmp = a*x._odata[ss+s]+b*tmp;
-    vstream(z._odata[ss+s],tmp);
-  }
+
+  auto x_v = x.View();
+  auto y_v = y.View();
+  auto z_v = z.View();
+  uint64_t nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,vobj::Nsimd(),{
+    uint64_t ss = sss*Ls;
+    decltype(coalescedRead(y_v[ss+sp])) tmp;
+    spProj5m(tmp,y_v(ss+sp));
+    tmp = a*x_v(ss+s)+b*tmp;
+    coalescedWrite(z_v[ss+s],tmp);
+  });
 }
 
 template<class vobj,class Coeff> 
 void axpby_ssp_pplus(Lattice<vobj> &z,Coeff a,const Lattice<vobj> &x,Coeff b,const Lattice<vobj> &y,int s,int sp)
 {
-  z.checkerboard = x.checkerboard;
+  z.Checkerboard() = x.Checkerboard();
   conformable(x,y);
   conformable(x,z);
-  GridBase *grid=x._grid;
+  GridBase *grid=x.Grid();
   int Ls = grid->_rdimensions[0];
-  parallel_for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
-    vobj tmp;
-    spProj5p(tmp,y._odata[ss+sp]);
-    tmp = a*x._odata[ss+s]+b*tmp;
-    vstream(z._odata[ss+s],tmp);
-  }
+  auto x_v = x.View();
+  auto y_v = y.View();
+  auto z_v = z.View();
+  uint64_t nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,vobj::Nsimd(),{
+    uint64_t ss = sss*Ls;
+    decltype(coalescedRead(y_v[ss+sp])) tmp;
+    spProj5p(tmp,y_v(ss+sp));
+    tmp = a*x_v(ss+s)+b*tmp;
+    coalescedWrite(z_v[ss+s],tmp);
+  });
 }
 
 template<class vobj> 
 void G5R5(Lattice<vobj> &z,const Lattice<vobj> &x)
 {
-  GridBase *grid=x._grid;
-  z.checkerboard = x.checkerboard;
+  GridBase *grid=x.Grid();
+  z.Checkerboard() = x.Checkerboard();
   conformable(x,z);
   int Ls = grid->_rdimensions[0];
   Gamma G5(Gamma::Algebra::Gamma5);
-  parallel_for(int ss=0;ss<grid->oSites();ss+=Ls) {
-    vobj tmp;
+  auto x_v = x.View();
+  auto z_v = z.View();
+  uint64_t nloop = grid->oSites()/Ls;
+  accelerator_for(sss,nloop,vobj::Nsimd(),{
+    uint64_t ss = sss*Ls;
     for(int s=0;s<Ls;s++){
       int sp = Ls-1-s;
-      tmp = G5*x._odata[ss+s];
-      vstream(z._odata[ss+sp],tmp);
+      coalescedWrite(z_v[ss+sp],G5*x_v(ss+s));
     }
-  }
-}
+  });
 }
 
 // I explicitly need these outside the QCD namespace
 template<typename vobj>
 void G5C(Lattice<vobj> &z, const Lattice<vobj> &x)
 {
-  GridBase *grid = x._grid;
-  z.checkerboard = x.checkerboard;
+  GridBase *grid = x.Grid();
+  z.Checkerboard() = x.Checkerboard();
   conformable(x, z);
 
-  QCD::Gamma G5(QCD::Gamma::Algebra::Gamma5);
+  Gamma G5(Gamma::Algebra::Gamma5);
   z = G5 * x;
 }
 
 template<class CComplex, int nbasis>
 void G5C(Lattice<iVector<CComplex, nbasis>> &z, const Lattice<iVector<CComplex, nbasis>> &x)
 {
-  GridBase *grid = x._grid;
-  z.checkerboard = x.checkerboard;
+  GridBase *grid = x.Grid();
+  z.Checkerboard() = x.Checkerboard();
   conformable(x, z);
 
   static_assert(nbasis % 2 == 0, "");
   int nb = nbasis / 2;
 
-  parallel_for(int ss = 0; ss < grid->oSites(); ss++) {
+  auto z_v = z.View();
+  auto x_v = x.View();
+  accelerator_for(ss,grid->oSites(),CComplex::Nsimd(),
+  {
     for(int n = 0; n < nb; ++n) {
-      z._odata[ss](n) = x._odata[ss](n);
+      coalescedWrite(z_v[ss](n), x_v(ss)(n));
     }
     for(int n = nb; n < nbasis; ++n) {
-      z._odata[ss](n) = -x._odata[ss](n);
+      coalescedWrite(z_v[ss](n), -x_v(ss)(n));
     }
-  }
+  });
 }
 
-}
-#endif 
+NAMESPACE_END(Grid);
+
diff --git a/Grid/qcd/utils/Metric.h b/Grid/qcd/utils/Metric.h
index 60a9bfc5..10d06de8 100644
--- a/Grid/qcd/utils/Metric.h
+++ b/Grid/qcd/utils/Metric.h
@@ -25,13 +25,11 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
-//--------------------------------------------------------------------
-#ifndef METRIC_H
-#define METRIC_H
+			   /*  END LEGAL */
+			   //--------------------------------------------------------------------
+#pragma once
 
-namespace Grid{
-namespace QCD{
+NAMESPACE_BEGIN(Grid);
 
 template <typename Field> 
 class Metric{
@@ -64,10 +62,10 @@ public:
     // do nothing
   }
   virtual void MDeriv(const Field& in, Field& out){
-    out = zero;
+    out = Zero();
   }
   virtual void MDeriv(const Field& left, const Field& right, Field& out){
-    out = zero;
+    out = Zero();
   }
 
 };
@@ -108,7 +106,7 @@ public:
     if (1) {
       // Auxiliary momenta
       // do nothing if trivial, so hide in the metric
-      MomentaField AuxMomTemp(Mom._grid);
+      MomentaField AuxMomTemp(Mom.Grid());
       Implementation::generate_momenta(AuxMom, pRNG);
       Implementation::generate_momenta(AuxField, pRNG);
       // Modify the distribution with the metric
@@ -119,11 +117,11 @@ public:
 
   // Correct
   RealD MomentaAction(){
-    MomentaField inv(Mom._grid);
-    inv = zero;
+    MomentaField inv(Mom.Grid());
+    inv = Zero();
     M.Minv(Mom, inv);
-    LatticeComplex Hloc(Mom._grid);
-    Hloc = zero;
+    LatticeComplex Hloc(Mom.Grid());
+    Hloc = Zero();
     for (int mu = 0; mu < Nd; mu++) {
       // This is not very general
       // hide in the metric
@@ -147,7 +145,7 @@ public:
       }
     }
 
-    Complex Hsum = sum(Hloc);
+    auto Hsum = TensorRemove(sum(Hloc));
     return Hsum.real();
   }
 
@@ -156,9 +154,9 @@ public:
 
     // Compute the derivative of the kinetic term
     // with respect to the gauge field
-    MomentaField MDer(in._grid);
-    MomentaField X(in._grid);
-    X = zero;
+    MomentaField MDer(in.Grid());
+    MomentaField X(in.Grid());
+    X = Zero();
     M.Minv(in, X);  // X = G in
     M.MDeriv(X, MDer);  // MDer = U * dS/dU
     der = Implementation::projectForce(MDer);  // Ta if gauge fields
@@ -166,27 +164,27 @@ public:
   }
 
   void AuxiliaryFieldsDerivative(MomentaField& der){
-    der = zero;
+    der = Zero();
     if (1){
-    // Auxiliary fields
-    MomentaField der_temp(der._grid);
-    MomentaField X(der._grid);
-    X=zero;
-    //M.M(AuxMom, X); // X = M Aux
-    // Two derivative terms
-    // the Mderiv need separation of left and right terms
-    M.MDeriv(AuxMom, der); 
+      // Auxiliary fields
+      MomentaField der_temp(der.Grid());
+      MomentaField X(der.Grid());
+      X=Zero();
+      //M.M(AuxMom, X); // X = M Aux
+      // Two derivative terms
+      // the Mderiv need separation of left and right terms
+      M.MDeriv(AuxMom, der); 
 
 
-    // this one should not be necessary (identical to the previous one)
-    //M.MDeriv(X, AuxMom, der_temp); der += der_temp;
+      // this one should not be necessary (identical to the previous one)
+      //M.MDeriv(X, AuxMom, der_temp); der += der_temp;
 
-    der = -1.0*Implementation::projectForce(der);
+      der = -1.0*Implementation::projectForce(der);
     }
   }
 
   void DerivativeP(MomentaField& der){
-    der = zero;
+    der = Zero();
     M.Minv(Mom, der);
     // is the projection necessary here?
     // no for fields in the algebra
@@ -201,8 +199,8 @@ public:
 
   void update_auxiliary_fields(RealD ep){
     if (1) {
-      MomentaField tmp(AuxMom._grid);
-      MomentaField tmp2(AuxMom._grid);
+      MomentaField tmp(AuxMom.Grid());
+      MomentaField tmp2(AuxMom.Grid());
       M.M(AuxMom, tmp);
       // M.M(tmp, tmp2);
       AuxField += ep * tmp;  // M^2 AuxMom
@@ -212,15 +210,5 @@ public:
 
 };
 
+NAMESPACE_END(Grid);
 
-
-
-
-
-
-
-}
-}
-
-
-#endif //METRIC_H
\ No newline at end of file
diff --git a/Grid/qcd/utils/SUn.h b/Grid/qcd/utils/SUn.h
index 5cb564bd..7ad80d00 100644
--- a/Grid/qcd/utils/SUn.h
+++ b/Grid/qcd/utils/SUn.h
@@ -28,16 +28,15 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef QCD_UTIL_SUN_H
 #define QCD_UTIL_SUN_H
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 template <int ncolour>
 class SU {
- public:
+public:
   static const int Dimension = ncolour;
   static const int AdjointDimension = ncolour * ncolour - 1;
   static int su2subgroups(void) { return (ncolour * (ncolour - 1)) / 2; }
@@ -48,7 +47,7 @@ class SU {
   using iSU2Matrix = iScalar<iScalar<iMatrix<vtype, 2> > >;
   template <typename vtype>
   using iSUnAlgebraVector =
-      iScalar<iScalar<iVector<vtype, AdjointDimension> > >;
+    iScalar<iScalar<iVector<vtype, AdjointDimension> > >;
 
   //////////////////////////////////////////////////////////////////////////////////////////////////
   // Types can be accessed as SU<2>::Matrix , SU<2>::vSUnMatrix,
@@ -163,7 +162,7 @@ class SU {
   
   template <class cplx>
   static void generatorSigmaY(int su2Index, iSUnMatrix<cplx> &ta) {
-    ta = zero;
+    ta = Zero();
     int i1, i2;
     su2SubGroupIndex(i1, i2, su2Index);
     ta()()(i1, i2) = 1.0;
@@ -173,7 +172,7 @@ class SU {
   
   template <class cplx>
   static void generatorSigmaX(int su2Index, iSUnMatrix<cplx> &ta) {
-    ta = zero;
+    ta = Zero();
     cplx i(0.0, 1.0);
     int i1, i2;
     su2SubGroupIndex(i1, i2, su2Index);
@@ -185,7 +184,7 @@ class SU {
   template <class cplx>
   static void generatorDiagonal(int diagIndex, iSUnMatrix<cplx> &ta) {
     // diag ({1, 1, ..., 1}(k-times), -k, 0, 0, ...)
-    ta = zero;
+    ta = Zero();
     int k = diagIndex + 1;                  // diagIndex starts from 0
     for (int i = 0; i <= diagIndex; i++) {  // k iterations
       ta()()(i, i) = 1.0;
@@ -218,28 +217,32 @@ class SU {
                          Lattice<iSU2Matrix<vcplx> > &subgroup,
                          const Lattice<iSUnMatrix<vcplx> > &source,
                          int su2_index) {
-    GridBase *grid(source._grid);
+    GridBase *grid(source.Grid());
     conformable(subgroup, source);
     conformable(subgroup, Determinant);
     int i0, i1;
     su2SubGroupIndex(i0, i1, su2_index);
+    auto subgroup_v = subgroup.View();
+    auto source_v   = source.View();
+    auto Determinant_v = Determinant.View();
 
-    parallel_for (int ss = 0; ss < grid->oSites(); ss++) {
-      subgroup._odata[ss]()()(0, 0) = source._odata[ss]()()(i0, i0);
-      subgroup._odata[ss]()()(0, 1) = source._odata[ss]()()(i0, i1);
-      subgroup._odata[ss]()()(1, 0) = source._odata[ss]()()(i1, i0);
-      subgroup._odata[ss]()()(1, 1) = source._odata[ss]()()(i1, i1);
+    thread_for(ss, grid->oSites(), {
 
-      iSU2Matrix<vcplx> Sigma = subgroup._odata[ss];
+      subgroup_v[ss]()()(0, 0) = source_v[ss]()()(i0, i0);
+      subgroup_v[ss]()()(0, 1) = source_v[ss]()()(i0, i1);
+      subgroup_v[ss]()()(1, 0) = source_v[ss]()()(i1, i0);
+      subgroup_v[ss]()()(1, 1) = source_v[ss]()()(i1, i1);
+
+      iSU2Matrix<vcplx> Sigma = subgroup_v[ss];
 
       Sigma = Sigma - adj(Sigma) + trace(adj(Sigma));
 
-      subgroup._odata[ss] = Sigma;
+      subgroup_v[ss] = Sigma;
 
       // this should be purely real
-      Determinant._odata[ss] =
-          Sigma()()(0, 0) * Sigma()()(1, 1) - Sigma()()(0, 1) * Sigma()()(1, 0);
-    }
+      Determinant_v[ss] =
+	Sigma()()(0, 0) * Sigma()()(1, 1) - Sigma()()(0, 1) * Sigma()()(1, 0);
+    });
   }
 
   //////////////////////////////////////////////////////////////////////////////////////////
@@ -248,18 +251,21 @@ class SU {
   template <class vcplx>
   static void su2Insert(const Lattice<iSU2Matrix<vcplx> > &subgroup,
                         Lattice<iSUnMatrix<vcplx> > &dest, int su2_index) {
-    GridBase *grid(dest._grid);
+    GridBase *grid(dest.Grid());
     conformable(subgroup, dest);
     int i0, i1;
     su2SubGroupIndex(i0, i1, su2_index);
 
     dest = 1.0;  // start out with identity
-    parallel_for (int ss = 0; ss < grid->oSites(); ss++) {
-      dest._odata[ss]()()(i0, i0) = subgroup._odata[ss]()()(0, 0);
-      dest._odata[ss]()()(i0, i1) = subgroup._odata[ss]()()(0, 1);
-      dest._odata[ss]()()(i1, i0) = subgroup._odata[ss]()()(1, 0);
-      dest._odata[ss]()()(i1, i1) = subgroup._odata[ss]()()(1, 1);
-    }
+    auto dest_v = dest.View();
+    auto subgroup_v = subgroup.View();
+    thread_for(ss, grid->oSites(),
+    {
+      dest_v[ss]()()(i0, i0) = subgroup_v[ss]()()(0, 0);
+      dest_v[ss]()()(i0, i1) = subgroup_v[ss]()()(0, 1);
+      dest_v[ss]()()(i1, i0) = subgroup_v[ss]()()(1, 0);
+      dest_v[ss]()()(i1, i1) = subgroup_v[ss]()()(1, 1);
+    });
   }
 
   ///////////////////////////////////////////////
@@ -272,16 +278,14 @@ class SU {
   // in action.
   //
   ///////////////////////////////////////////////
-  static void SubGroupHeatBath(
-      GridSerialRNG &sRNG, GridParallelRNG &pRNG,
-      RealD beta,  // coeff multiplying staple in action (with no 1/Nc)
-      LatticeMatrix &link,
-      const LatticeMatrix &barestaple,  // multiplied by action coeffs so th
-      int su2_subgroup, int nheatbath, LatticeInteger &wheremask) {
-    GridBase *grid = link._grid;
+  static void SubGroupHeatBath(GridSerialRNG &sRNG, GridParallelRNG &pRNG,
+			       RealD beta,  // coeff multiplying staple in action (with no 1/Nc)
+			       LatticeMatrix &link,
+			       const LatticeMatrix &barestaple,  // multiplied by action coeffs so th
+			       int su2_subgroup, int nheatbath, LatticeInteger &wheremask) 
+  {
+    GridBase *grid = link.Grid();
 
-    int ntrials = 0;
-    int nfails = 0;
     const RealD twopi = 2.0 * M_PI;
 
     LatticeMatrix staple(grid);
@@ -292,8 +296,7 @@ class SU {
     V = link * staple;
 
     // Subgroup manipulation in the lie algebra space
-    LatticeSU2Matrix u(
-        grid);  // Kennedy pendleton "u" real projected normalised Sigma
+    LatticeSU2Matrix u(grid);  // Kennedy pendleton "u" real projected normalised Sigma
     LatticeSU2Matrix uinv(grid);
     LatticeSU2Matrix ua(grid);  // a in pauli form
     LatticeSU2Matrix b(grid);   // rotated matrix after hb
@@ -302,11 +305,11 @@ class SU {
     LatticeComplex ones(grid);
     ones = 1.0;
     LatticeComplex zeros(grid);
-    zeros = zero;
+    zeros = Zero();
     LatticeReal rones(grid);
     rones = 1.0;
     LatticeReal rzeros(grid);
-    rzeros = zero;
+    rzeros = Zero();
     LatticeComplex udet(grid);  // determinant of real(staple)
     LatticeInteger mask_true(grid);
     mask_true = 1;
@@ -314,41 +317,41 @@ class SU {
     mask_false = 0;
 
     /*
-  PLB 156 P393 (1985) (Kennedy and Pendleton)
+      PLB 156 P393 (1985) (Kennedy and Pendleton)
 
-  Note: absorb "beta" into the def of sigma compared to KP paper; staple
-  passed to this routine has "beta" already multiplied in
+      Note: absorb "beta" into the def of sigma compared to KP paper; staple
+      passed to this routine has "beta" already multiplied in
 
-  Action linear in links h and of form:
+      Action linear in links h and of form:
 
       beta S = beta  Sum_p (1 - 1/Nc Re Tr Plaq )
 
-  Writing Sigma = 1/Nc (beta Sigma') where sum over staples is "Sigma' "
+      Writing Sigma = 1/Nc (beta Sigma') where sum over staples is "Sigma' "
 
-       beta S = const - beta/Nc Re Tr h Sigma'
-              = const - Re Tr h Sigma
+      beta S = const - beta/Nc Re Tr h Sigma'
+      = const - Re Tr h Sigma
 
-  Decompose h and Sigma into (1, sigma_j) ; h_i real, h^2=1, Sigma_i complex
-  arbitrary.
+      Decompose h and Sigma into (1, sigma_j) ; h_i real, h^2=1, Sigma_i complex
+      arbitrary.
 
       Tr h Sigma = h_i Sigma_j Tr (sigma_i sigma_j)  = h_i Sigma_j 2 delta_ij
-   Re Tr h Sigma = 2 h_j Re Sigma_j
+      Re Tr h Sigma = 2 h_j Re Sigma_j
 
-  Normalised re Sigma_j = xi u_j
+      Normalised re Sigma_j = xi u_j
 
-  With u_j a unit vector and U can be in SU(2);
+      With u_j a unit vector and U can be in SU(2);
 
-  Re Tr h Sigma = 2 h_j Re Sigma_j = 2 xi (h.u)
+      Re Tr h Sigma = 2 h_j Re Sigma_j = 2 xi (h.u)
 
-  4xi^2 = Det [ Sig - Sig^dag  + 1 Tr Sigdag]
-   u   = 1/2xi [ Sig - Sig^dag  + 1 Tr Sigdag]
+      4xi^2 = Det [ Sig - Sig^dag  + 1 Tr Sigdag]
+      u   = 1/2xi [ Sig - Sig^dag  + 1 Tr Sigdag]
 
-   xi = sqrt(Det)/2;
+      xi = sqrt(Det)/2;
 
-  Write a= u h in SU(2); a has pauli decomp a_j;
+      Write a= u h in SU(2); a has pauli decomp a_j;
 
-  Note: Product b' xi is unvariant because scaling Sigma leaves
-        normalised vector "u" fixed; Can rescale Sigma so b' = 1.
+      Note: Product b' xi is unvariant because scaling Sigma leaves
+      normalised vector "u" fixed; Can rescale Sigma so b' = 1.
     */
 
     ////////////////////////////////////////////////////////
@@ -386,7 +389,7 @@ class SU {
 
     xi = 0.5 * sqrt(udet);  // 4xi^2 = Det [ Sig - Sig^dag  + 1 Tr Sigdag]
     u = 0.5 * u *
-        pow(xi, -1.0);  //  u   = 1/2xi [ Sig - Sig^dag  + 1 Tr Sigdag]
+      pow(xi, -1.0);  //  u   = 1/2xi [ Sig - Sig^dag  + 1 Tr Sigdag]
 
     // Debug test for sanity
     uinv = adj(u);
@@ -394,36 +397,36 @@ class SU {
     assert(norm2(b) < 1.0e-4);
 
     /*
-  Measure: Haar measure dh has d^4a delta(1-|a^2|)
-  In polars:
-    da = da0 r^2 sin theta dr dtheta dphi delta( 1 - r^2 -a0^2)
-       = da0 r^2 sin theta dr dtheta dphi delta( (sqrt(1-a0^) - r)(sqrt(1-a0^) +
-  r) )
-       = da0 r/2 sin theta dr dtheta dphi delta( (sqrt(1-a0^) - r) )
+      Measure: Haar measure dh has d^4a delta(1-|a^2|)
+      In polars:
+      da = da0 r^2 sin theta dr dtheta dphi delta( 1 - r^2 -a0^2)
+      = da0 r^2 sin theta dr dtheta dphi delta( (sqrt(1-a0^) - r)(sqrt(1-a0^) +
+      r) )
+      = da0 r/2 sin theta dr dtheta dphi delta( (sqrt(1-a0^) - r) )
 
-  Action factor Q(h) dh  = e^-S[h]  dh =  e^{  xi Tr uh} dh    // beta enters
-  through xi
-                                       =  e^{2 xi (h.u)} dh
-                                       =  e^{2 xi h0u0}.e^{2 xi h1u1}.e^{2 xi
-  h2u2}.e^{2 xi h3u3} dh
+      Action factor Q(h) dh  = e^-S[h]  dh =  e^{  xi Tr uh} dh    // beta enters
+      through xi
+      =  e^{2 xi (h.u)} dh
+      =  e^{2 xi h0u0}.e^{2 xi h1u1}.e^{2 xi
+      h2u2}.e^{2 xi h3u3} dh
 
-  Therefore for each site, take xi for that site
-  i) generate  |a0|<1 with dist
-     (1-a0^2)^0.5 e^{2 xi a0 } da0
+      Therefore for each site, take xi for that site
+      i) generate  |a0|<1 with dist
+      (1-a0^2)^0.5 e^{2 xi a0 } da0
 
-  Take alpha = 2 xi  = 2 xi [ recall 2 beta/Nc unmod staple norm]; hence 2.0/Nc
-  factor in Chroma ]
-  A. Generate two uniformly distributed pseudo-random numbers R and R', R'',
-  R''' in the unit interval;
-  B. Set X = -(ln R)/alpha, X' =-(ln R')/alpha;
-  C. Set C = cos^2(2pi R"), with R" another uniform random number in [0,1] ;
-  D. Set A = XC;
-  E. Let d  = X'+A;
-  F. If R'''^2 :> 1 - 0.5 d,  go back to A;
-  G. Set a0 = 1 - d;
+      Take alpha = 2 xi  = 2 xi [ recall 2 beta/Nc unmod staple norm]; hence 2.0/Nc
+      factor in Chroma ]
+      A. Generate two uniformly distributed pseudo-random numbers R and R', R'',
+      R''' in the unit interval;
+      B. Set X = -(ln R)/alpha, X' =-(ln R')/alpha;
+      C. Set C = cos^2(2pi R"), with R" another uniform random number in [0,1] ;
+      D. Set A = XC;
+      E. Let d  = X'+A;
+      F. If R'''^2 :> 1 - 0.5 d,  go back to A;
+      G. Set a0 = 1 - d;
 
-  Note that in step D setting B ~ X - A and using B in place of A in step E will
-  generate a second independent a 0 value.
+      Note that in step D setting B ~ X - A and using B in place of A in step E will
+      generate a second independent a 0 value.
     */
 
     /////////////////////////////////////////////////////////
@@ -435,13 +438,13 @@ class SU {
     RealD numSites = sum(rtmp);
     RealD numAccepted;
     LatticeInteger Accepted(grid);
-    Accepted = zero;
+    Accepted = Zero();
     LatticeInteger newlyAccepted(grid);
 
     std::vector<LatticeReal> xr(4, grid);
     std::vector<LatticeReal> a(4, grid);
     LatticeReal d(grid);
-    d = zero;
+    d = Zero();
     LatticeReal alpha(grid);
 
     //    std::cout<<GridLogMessage<<"xi "<<xi <<std::endl;
@@ -478,7 +481,7 @@ class SU {
       LatticeInteger ione(grid);
       ione = 1;
       LatticeInteger izero(grid);
-      izero = zero;
+      izero = Zero();
 
       newlyAccepted = where(xrsq < thresh, ione, izero);
       Accepted = where(newlyAccepted, newlyAccepted, Accepted);
@@ -493,7 +496,7 @@ class SU {
     } while ((numAccepted < numSites) && (hit < nheatbath));
 
     // G. Set a0 = 1 - d;
-    a[0] = zero;
+    a[0] = Zero();
     a[0] = where(wheremask, 1.0 - d, a[0]);
 
     //////////////////////////////////////////
@@ -517,7 +520,7 @@ class SU {
     a[2] = a123mag * sin_theta * sin(phi);
     a[3] = a123mag * cos_theta;
 
-    ua = toComplex(a[0]) * ident + toComplex(a[1]) * pauli1 +
+    ua = toComplex(a[0]) * ident  + toComplex(a[1]) * pauli1 +
          toComplex(a[2]) * pauli2 + toComplex(a[3]) * pauli3;
 
     b = 1.0;
@@ -531,7 +534,7 @@ class SU {
     // Debug Checks
     // SU2 check
     LatticeSU2Matrix check(grid);  // rotated matrix after hb
-    u = zero;
+    u = Zero();
     check = ua * adj(ua) - 1.0;
     check = where(Accepted, check, u);
     assert(norm2(check) < 1.0e-4);
@@ -541,7 +544,7 @@ class SU {
     assert(norm2(check) < 1.0e-4);
 
     LatticeMatrix Vcheck(grid);
-    Vcheck = zero;
+    Vcheck = Zero();
     Vcheck = where(Accepted, V * adj(V) - 1.0, Vcheck);
     //    std::cout<<GridLogMessage << "SU3 check " <<norm2(Vcheck)<<std::endl;
     assert(norm2(Vcheck) < 1.0e-4);
@@ -607,7 +610,7 @@ class SU {
   template <typename LatticeMatrixType>
   static void LieRandomize(GridParallelRNG &pRNG, LatticeMatrixType &out,
                            double scale = 1.0) {
-    GridBase *grid = out._grid;
+    GridBase *grid = out.Grid();
 
     typedef typename LatticeMatrixType::vector_type vector_type;
     typedef typename LatticeMatrixType::scalar_type scalar_type;
@@ -616,16 +619,16 @@ class SU {
 
     typedef Lattice<vTComplexType> LatticeComplexType;
     typedef typename GridTypeMapper<
-        typename LatticeMatrixType::vector_object>::scalar_object MatrixType;
+      typename LatticeMatrixType::vector_object>::scalar_object MatrixType;
 
     LatticeComplexType ca(grid);
     LatticeMatrixType lie(grid);
     LatticeMatrixType la(grid);
     ComplexD ci(0.0, scale);
-    ComplexD cone(1.0, 0.0);
+    //    ComplexD cone(1.0, 0.0);
     MatrixType ta;
 
-    lie = zero;
+    lie = Zero();
     for (int a = 0; a < AdjointDimension; a++) {
       random(pRNG, ca);
 
@@ -644,13 +647,13 @@ class SU {
   static void GaussianFundamentalLieAlgebraMatrix(GridParallelRNG &pRNG,
                                                   LatticeMatrix &out,
                                                   Real scale = 1.0) {
-    GridBase *grid = out._grid;
+    GridBase *grid = out.Grid();
     LatticeReal ca(grid);
     LatticeMatrix la(grid);
     Complex ci(0.0, scale);
     Matrix ta;
 
-    out = zero;
+    out = Zero();
     for (int a = 0; a < AdjointDimension; a++) {
       gaussian(pRNG, ca);
       generator(a, ta);
@@ -664,11 +667,11 @@ class SU {
                                           LatticeMatrix &out,
                                           Real scale = 1.0) {
     conformable(h, out);
-    GridBase *grid = out._grid;
+    GridBase *grid = out.Grid();
     LatticeMatrix la(grid);
     Matrix ta;
 
-    out = zero;
+    out = Zero();
     for (int a = 0; a < AdjointDimension; a++) {
       generator(a, ta);
       la = peekColour(h, a) * timesI(ta) * scale;
@@ -687,10 +690,11 @@ class SU {
 /*
  * Adjoint rep gauge xform
  */
+
   template<typename GaugeField,typename GaugeMat>
   static void GaugeTransform( GaugeField &Umu, GaugeMat &g){
-    GridBase *grid = Umu._grid;
-    conformable(grid,g._grid);
+    GridBase *grid = Umu.Grid();
+    conformable(grid,g.Grid());
 
     GaugeMat U(grid);
     GaugeMat ag(grid); ag = adj(g);
@@ -702,8 +706,8 @@ class SU {
     }
   }
   template<typename GaugeMat>
-    static void GaugeTransform( std::vector<GaugeMat> &U, GaugeMat &g){
-    GridBase *grid = g._grid;
+  static void GaugeTransform( std::vector<GaugeMat> &U, GaugeMat &g){
+    GridBase *grid = g.Grid();
     GaugeMat ag(grid); ag = adj(g);
     for(int mu=0;mu<Nd;mu++){
       U[mu] = g*U[mu]*Cshift(ag, mu, 1);
@@ -719,7 +723,7 @@ class SU {
   // inverse operation: FundamentalLieAlgebraMatrix
   static void projectOnAlgebra(LatticeAlgebraVector &h_out, const LatticeMatrix &in, Real scale = 1.0) {
     conformable(h_out, in);
-    h_out = zero;
+    h_out = Zero();
     Matrix Ta;
 
     for (int a = 0; a < AdjointDimension; a++) {
@@ -735,7 +739,7 @@ class SU {
     typedef iSUnMatrix<vector_type> vMatrixType;
     typedef Lattice<vMatrixType> LatticeMatrixType;
 
-    LatticeMatrixType Umu(out._grid);
+    LatticeMatrixType Umu(out.Grid());
     for (int mu = 0; mu < Nd; mu++) {
       LieRandomize(pRNG, Umu, 1.0);
       PokeIndex<LorentzIndex>(out, Umu, mu);
@@ -747,7 +751,7 @@ class SU {
     typedef iSUnMatrix<vector_type> vMatrixType;
     typedef Lattice<vMatrixType> LatticeMatrixType;
 
-    LatticeMatrixType Umu(out._grid);
+    LatticeMatrixType Umu(out.Grid());
     for(int mu=0;mu<Nd;mu++){
       LieRandomize(pRNG,Umu,0.01);
       PokeIndex<LorentzIndex>(out,Umu,mu);
@@ -759,7 +763,7 @@ class SU {
     typedef iSUnMatrix<vector_type> vMatrixType;
     typedef Lattice<vMatrixType> LatticeMatrixType;
 
-    LatticeMatrixType Umu(out._grid);
+    LatticeMatrixType Umu(out.Grid());
     Umu=1.0;
     for(int mu=0;mu<Nd;mu++){
       PokeIndex<LorentzIndex>(out,Umu,mu);
@@ -778,7 +782,7 @@ class SU {
   static void taExp(const LatticeMatrixType &x, LatticeMatrixType &ex) {
     typedef typename LatticeMatrixType::scalar_type ComplexType;
 
-    LatticeMatrixType xn(x._grid);
+    LatticeMatrixType xn(x.Grid());
     RealD nfac = 1.0;
 
     xn = x;
@@ -801,6 +805,5 @@ typedef SU<5> SU5;
 
 typedef SU<Nc> FundamentalMatrices;
 
-}
-}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/qcd/utils/SUnAdjoint.h b/Grid/qcd/utils/SUnAdjoint.h
index 9d9b77bd..1d530373 100644
--- a/Grid/qcd/utils/SUnAdjoint.h
+++ b/Grid/qcd/utils/SUnAdjoint.h
@@ -22,17 +22,16 @@
 //
 ////////////////////////////////////////////////////////////////////////
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 template <int ncolour>
 class SU_Adjoint : public SU<ncolour> {
- public:
+public:
   static const int Dimension = ncolour * ncolour - 1;
 
   template <typename vtype>
   using iSUnAdjointMatrix =
-      iScalar<iScalar<iMatrix<vtype, Dimension > > >;
+    iScalar<iScalar<iMatrix<vtype, Dimension > > >;
 
   // Actually the adjoint matrices are real...
   // Consider this overhead... FIXME
@@ -49,11 +48,11 @@ class SU_Adjoint : public SU<ncolour> {
   typedef Lattice<vAMatrixD> LatticeAdjMatrixD;
 
   typedef Lattice<iVector<iScalar<iMatrix<vComplex, Dimension> >, Nd> >
-      LatticeAdjField;
+  LatticeAdjField;
   typedef Lattice<iVector<iScalar<iMatrix<vComplexF, Dimension> >, Nd> >
-      LatticeAdjFieldF;
+  LatticeAdjFieldF;
   typedef Lattice<iVector<iScalar<iMatrix<vComplexD, Dimension> >, Nd> >
-      LatticeAdjFieldD;
+  LatticeAdjFieldD;
 
 
 
@@ -62,7 +61,7 @@ class SU_Adjoint : public SU<ncolour> {
   static void generator(int Index, iSUnAdjointMatrix<cplx> &iAdjTa) {
     // returns i(T_Adj)^index necessary for the projectors
     // see definitions above
-    iAdjTa = zero;
+    iAdjTa = Zero();
     Vector<typename SU<ncolour>::template iSUnMatrix<cplx> > ta(ncolour * ncolour - 1);
     typename SU<ncolour>::template iSUnMatrix<cplx> tmp;
 
@@ -73,7 +72,7 @@ class SU_Adjoint : public SU<ncolour> {
       tmp = ta[a] * ta[Index] - ta[Index] * ta[a];
       for (int b = 0; b < (ncolour * ncolour - 1); b++) {
         typename SU<ncolour>::template iSUnMatrix<cplx> tmp1 =
-            2.0 * tmp * ta[b];  // 2.0 from the normalization
+	  2.0 * tmp * ta[b];  // 2.0 from the normalization
         Complex iTr = TensorRemove(timesI(trace(tmp1)));
         //iAdjTa()()(b, a) = iTr;
         iAdjTa()()(a, b) = iTr;
@@ -112,14 +111,14 @@ class SU_Adjoint : public SU<ncolour> {
   }
 
   static void AdjointLieAlgebraMatrix(
-      const typename SU<ncolour>::LatticeAlgebraVector &h,
-      LatticeAdjMatrix &out, Real scale = 1.0) {
+				      const typename SU<ncolour>::LatticeAlgebraVector &h,
+				      LatticeAdjMatrix &out, Real scale = 1.0) {
     conformable(h, out);
-    GridBase *grid = out._grid;
+    GridBase *grid = out.Grid();
     LatticeAdjMatrix la(grid);
     AMatrix iTa;
 
-    out = zero;
+    out = Zero();
     for (int a = 0; a < Dimension; a++) {
       generator(a, iTa);
       la = peekColour(h, a) * iTa;
@@ -131,7 +130,7 @@ class SU_Adjoint : public SU<ncolour> {
   // Projects the algebra components a lattice matrix (of dimension ncol*ncol -1 )
   static void projectOnAlgebra(typename SU<ncolour>::LatticeAlgebraVector &h_out, const LatticeAdjMatrix &in, Real scale = 1.0) {
     conformable(h_out, in);
-    h_out = zero;
+    h_out = Zero();
     AMatrix iTa;
     Real coefficient = - 1.0/(ncolour) * scale;// 1/Nc for the normalization of the trace in the adj rep
 
@@ -146,15 +145,15 @@ class SU_Adjoint : public SU<ncolour> {
   static void projector(typename SU<ncolour>::LatticeAlgebraVector &h_out, const LatticeAdjMatrix &in, Real scale = 1.0) {
     conformable(h_out, in);
     static std::vector<AMatrix> iTa(Dimension);  // to store the generators
-    h_out = zero;
+    h_out = Zero();
     static bool precalculated = false; 
     if (!precalculated){
       precalculated = true;
-        for (int a = 0; a < Dimension; a++) generator(a, iTa[a]);
+      for (int a = 0; a < Dimension; a++) generator(a, iTa[a]);
     }
 
     Real coefficient = -1.0 / (ncolour) * scale;  // 1/Nc for the normalization of
-                                                // the trace in the adj rep
+    // the trace in the adj rep
 
     for (int a = 0; a < Dimension; a++) {
       auto tmp = real(trace(iTa[a] * in)) * coefficient; 
@@ -176,7 +175,7 @@ typedef SU_Adjoint<4> SU4Adjoint;
 typedef SU_Adjoint<5> SU5Adjoint;
 
 typedef SU_Adjoint<Nc> AdjointMatrices;
-}
-}
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/utils/SUnTwoIndex.h b/Grid/qcd/utils/SUnTwoIndex.h
index 0c09f5e2..cd440c26 100644
--- a/Grid/qcd/utils/SUnTwoIndex.h
+++ b/Grid/qcd/utils/SUnTwoIndex.h
@@ -26,8 +26,7 @@
 #define QCD_UTIL_SUN2INDEX_H
 
 
-namespace Grid {
-namespace QCD {
+NAMESPACE_BEGIN(Grid);
 
 enum TwoIndexSymmetry { Symmetric = 1, AntiSymmetric = -1 };
 
@@ -35,7 +34,7 @@ inline Real delta(int a, int b) { return (a == b) ? 1.0 : 0.0; }
 
 template <int ncolour, TwoIndexSymmetry S>
 class SU_TwoIndex : public SU<ncolour> {
- public:
+public:
   static const int Dimension = ncolour * (ncolour + S) / 2;
   static const int NumGenerators = SU<ncolour>::AdjointDimension;
 
@@ -55,11 +54,11 @@ class SU_TwoIndex : public SU<ncolour> {
   typedef Lattice<vTIMatrixD> LatticeTwoIndexMatrixD;
 
   typedef Lattice<iVector<iScalar<iMatrix<vComplex, Dimension> >, Nd> >
-      LatticeTwoIndexField;
+  LatticeTwoIndexField;
   typedef Lattice<iVector<iScalar<iMatrix<vComplexF, Dimension> >, Nd> >
-      LatticeTwoIndexFieldF;
+  LatticeTwoIndexFieldF;
   typedef Lattice<iVector<iScalar<iMatrix<vComplexD, Dimension> >, Nd> >
-      LatticeTwoIndexFieldD;
+  LatticeTwoIndexFieldD;
 
   template <typename vtype>
   using iSUnMatrix = iScalar<iScalar<iMatrix<vtype, ncolour> > >;
@@ -72,7 +71,7 @@ class SU_TwoIndex : public SU<ncolour> {
   static void base(int Index, iSUnMatrix<cplx> &eij) {
     // returns (e)^(ij)_{kl} necessary for change of base U_F -> U_R
     assert(Index < NumGenerators);
-    eij = zero;
+    eij = Zero();
 
     // for the linearisation of the 2 indexes 
     static int a[ncolour * (ncolour - 1) / 2][2]; // store the a <-> i,j
@@ -98,18 +97,18 @@ class SU_TwoIndex : public SU<ncolour> {
 
   template <class cplx>
   static void baseDiagonal(int Index, iSUnMatrix<cplx> &eij) {
-    eij = zero;
+    eij = Zero();
     eij()()(Index - ncolour * (ncolour - 1) / 2,
             Index - ncolour * (ncolour - 1) / 2) = 1.0;
   }
 
   template <class cplx>
   static void baseOffDiagonal(int i, int j, iSUnMatrix<cplx> &eij) {
-    eij = zero;
+    eij = Zero();
     for (int k = 0; k < ncolour; k++)
       for (int l = 0; l < ncolour; l++)
         eij()()(l, k) = delta(i, k) * delta(j, l) +
-                        S * delta(j, k) * delta(i, l);
+	  S * delta(j, k) * delta(i, l);
 
     RealD nrm = 1. / std::sqrt(2.0);
     eij = eij * nrm;
@@ -128,10 +127,10 @@ class SU_TwoIndex : public SU<ncolour> {
   template <class cplx>
   static void generator(int Index, iSUnTwoIndexMatrix<cplx> &i2indTa) {
     Vector<typename SU<ncolour>::template iSUnMatrix<cplx> > ta(
-        ncolour * ncolour - 1);
+								ncolour * ncolour - 1);
     Vector<typename SU<ncolour>::template iSUnMatrix<cplx> > eij(Dimension);
     typename SU<ncolour>::template iSUnMatrix<cplx> tmp;
-    i2indTa = zero;
+    i2indTa = Zero();
     
     for (int a = 0; a < ncolour * ncolour - 1; a++)
       SU<ncolour>::generator(a, ta[a]);
@@ -142,7 +141,7 @@ class SU_TwoIndex : public SU<ncolour> {
       tmp = transpose(ta[Index]) * adj(eij[a]) + adj(eij[a]) * ta[Index];
       for (int b = 0; b < Dimension; b++) {
         typename SU<ncolour>::template iSUnMatrix<cplx> tmp1 =
-            tmp * eij[b]; 
+	  tmp * eij[b]; 
         Complex iTr = TensorRemove(timesI(trace(tmp1)));
         i2indTa()()(a, b) = iTr;
       }
@@ -197,14 +196,14 @@ class SU_TwoIndex : public SU<ncolour> {
   }
 
   static void TwoIndexLieAlgebraMatrix(
-      const typename SU<ncolour>::LatticeAlgebraVector &h,
-      LatticeTwoIndexMatrix &out, Real scale = 1.0) {
+				       const typename SU<ncolour>::LatticeAlgebraVector &h,
+				       LatticeTwoIndexMatrix &out, Real scale = 1.0) {
     conformable(h, out);
-    GridBase *grid = out._grid;
+    GridBase *grid = out.Grid();
     LatticeTwoIndexMatrix la(grid);
     TIMatrix i2indTa;
 
-    out = zero;
+    out = Zero();
     for (int a = 0; a < ncolour * ncolour - 1; a++) {
       generator(a, i2indTa);
       la = peekColour(h, a) * i2indTa;
@@ -216,10 +215,10 @@ class SU_TwoIndex : public SU<ncolour> {
   // Projects the algebra components 
   // of a lattice matrix ( of dimension ncol*ncol -1 )
   static void projectOnAlgebra(
-      typename SU<ncolour>::LatticeAlgebraVector &h_out,
-      const LatticeTwoIndexMatrix &in, Real scale = 1.0) {
+			       typename SU<ncolour>::LatticeAlgebraVector &h_out,
+			       const LatticeTwoIndexMatrix &in, Real scale = 1.0) {
     conformable(h_out, in);
-    h_out = zero;
+    h_out = Zero();
     TIMatrix i2indTa;
     Real coefficient = -2.0 / (ncolour + 2 * S) * scale;
     // 2/(Nc +/- 2) for the normalization of the trace in the two index rep
@@ -237,7 +236,7 @@ class SU_TwoIndex : public SU<ncolour> {
     conformable(h_out, in);
     // to store the generators
     static std::vector<TIMatrix> i2indTa(ncolour * ncolour -1); 
-    h_out = zero;
+    h_out = Zero();
     static bool precalculated = false;
     if (!precalculated) {
       precalculated = true;
@@ -245,8 +244,8 @@ class SU_TwoIndex : public SU<ncolour> {
     }
 
     Real coefficient =
-        -2.0 / (ncolour + 2 * S) * scale;  // 2/(Nc +/- 2) for the normalization
-                                           // of the trace in the two index rep
+      -2.0 / (ncolour + 2 * S) * scale;  // 2/(Nc +/- 2) for the normalization
+    // of the trace in the two index rep
 
     for (int a = 0; a < ncolour * ncolour - 1; a++) {
       auto tmp = real(trace(i2indTa[a] * in)) * coefficient;
@@ -269,8 +268,6 @@ typedef SU_TwoIndex<3, AntiSymmetric> SU3TwoIndexAntiSymm;
 typedef SU_TwoIndex<4, AntiSymmetric> SU4TwoIndexAntiSymm;
 typedef SU_TwoIndex<5, AntiSymmetric> SU5TwoIndexAntiSymm;
 
-
-}
-}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/utils/ScalarObjs.h b/Grid/qcd/utils/ScalarObjs.h
index a284af40..91a617d8 100644
--- a/Grid/qcd/utils/ScalarObjs.h
+++ b/Grid/qcd/utils/ScalarObjs.h
@@ -28,15 +28,13 @@ directory
 /*  END LEGAL */
 #ifndef SCALAR_OBJS_H
 #define SCALAR_OBJS_H
-namespace Grid {
 
-  // FIXME drop the QCD namespace in Nd
-  
+NAMESPACE_BEGIN(Grid);
 
 // Scalar field obs
 template <class Impl>
 class ScalarObs {
- public:
+public:
   //////////////////////////////////////////////////
   // squared field
   //////////////////////////////////////////////////
@@ -61,7 +59,7 @@ class ScalarObs {
   static void phider(typename Impl::Field &fsq,
                      const typename Impl::Field &f) {
     fsq = Cshift(f, 0, -1) * f;
-    for (int mu = 1; mu < QCD::Nd; mu++) fsq += Cshift(f, mu, -1) * f;
+    for (int mu = 1; mu < Nd; mu++) fsq += Cshift(f, mu, -1) * f;
   }
 
   //////////////////////////////////////////////////
@@ -69,28 +67,27 @@ class ScalarObs {
   //////////////////////////////////////////////////
 
   static RealD sumphider(const typename Impl::Field &f) {
-    typename Impl::Field tmp(f._grid);
+    typename Impl::Field tmp(f.Grid());
     tmp = Cshift(f, 0, -1) * f;
-    for (int mu = 1; mu < QCD::Nd; mu++) {
+    for (int mu = 1; mu < Nd; mu++) {
       tmp += Cshift(f, mu, -1) * f;
     }
     return -sum(trace(tmp));
   }
 
   static RealD sumphisquared(const typename Impl::Field &f) {
-    typename Impl::Field tmp(f._grid);
+    typename Impl::Field tmp(f.Grid());
     tmp = f * f;
     return sum(trace(tmp));
   }
 
   static RealD sumphifourth(const typename Impl::Field &f) {
-    typename Impl::Field tmp(f._grid);
+    typename Impl::Field tmp(f.Grid());
     phifourth(tmp, f);
     return sum(trace(tmp));
   }
 };
 
-
-}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/utils/SpaceTimeGrid.cc b/Grid/qcd/utils/SpaceTimeGrid.cc
index b2b5d9c8..31e8c79a 100644
--- a/Grid/qcd/utils/SpaceTimeGrid.cc
+++ b/Grid/qcd/utils/SpaceTimeGrid.cc
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,18 +23,17 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #include <Grid/GridCore.h>
 #include <Grid/GridQCDcore.h>
 
-namespace Grid { 
-  namespace QCD {
+NAMESPACE_BEGIN(Grid); 
 
 /////////////////////////////////////////////////////////////////
 // Public interface
 /////////////////////////////////////////////////////////////////
-GridCartesian *SpaceTimeGrid::makeFourDimGrid(const std::vector<int> & latt,const std::vector<int> &simd,const std::vector<int> &mpi)
+GridCartesian *SpaceTimeGrid::makeFourDimGrid(const Coordinate & latt,const Coordinate &simd,const Coordinate &mpi)
 {
   return new GridCartesian(latt,simd,mpi); 
 }
@@ -42,23 +41,23 @@ GridRedBlackCartesian *SpaceTimeGrid::makeFourDimRedBlackGrid(const GridCartesia
 {
   return new GridRedBlackCartesian(FourDimGrid); 
 }
-GridCartesian *SpaceTimeGrid::makeFourDimDWFGrid(const std::vector<int> & latt,const std::vector<int> &mpi)
+GridCartesian *SpaceTimeGrid::makeFourDimDWFGrid(const Coordinate & latt,const Coordinate &mpi)
 {
-  std::vector<int> simd(4,1);
+  Coordinate simd(4,1);
   return makeFourDimGrid(latt,simd,mpi);
 }
 GridCartesian         *SpaceTimeGrid::makeFiveDimGrid(int Ls,const GridCartesian *FourDimGrid)
 {
   int N4=FourDimGrid->_ndimension;
 
-  std::vector<int> latt5(1,Ls);
-  std::vector<int> simd5(1,1);
-  std::vector<int>  mpi5(1,1);
+  Coordinate latt5(1,Ls);
+  Coordinate simd5(1,1);
+  Coordinate  mpi5(1,1);
   
   for(int d=0;d<N4;d++){
     latt5.push_back(FourDimGrid->_fdimensions[d]);
     simd5.push_back(FourDimGrid->_simd_layout[d]);
-     mpi5.push_back(FourDimGrid->_processors[d]);
+    mpi5.push_back(FourDimGrid->_processors[d]);
   }
   return new GridCartesian(latt5,simd5,mpi5,*FourDimGrid); 
 }
@@ -68,9 +67,9 @@ GridRedBlackCartesian *SpaceTimeGrid::makeFiveDimRedBlackGrid(int Ls,const GridC
 {
   int N4=FourDimGrid->_ndimension;
   int cbd=1;
-  std::vector<int>   cb5(1,0);
+  Coordinate   cb5(1,0);
   for(int d=0;d<N4;d++){
-      cb5.push_back(  1);
+    cb5.push_back(  1);
   }
   GridCartesian *tmp = makeFiveDimGrid(Ls,FourDimGrid);
   GridRedBlackCartesian *ret = new GridRedBlackCartesian(tmp,cb5,cbd); 
@@ -84,14 +83,14 @@ GridCartesian         *SpaceTimeGrid::makeFiveDimDWFGrid(int Ls,const GridCartes
   int N4    = FourDimGrid->_ndimension;
   int nsimd = FourDimGrid->Nsimd();
 
-  std::vector<int> latt5(1,Ls);
-  std::vector<int> simd5(1,nsimd);
-  std::vector<int>  mpi5(1,1);
+  Coordinate latt5(1,Ls);
+  Coordinate simd5(1,nsimd);
+  Coordinate  mpi5(1,1);
   
   for(int d=0;d<N4;d++){
     latt5.push_back(FourDimGrid->_fdimensions[d]);
     simd5.push_back(1);
-     mpi5.push_back(FourDimGrid->_processors[d]);
+    mpi5.push_back(FourDimGrid->_processors[d]);
   }
   return new GridCartesian(latt5,simd5,mpi5,*FourDimGrid); 
 }
@@ -103,9 +102,9 @@ GridRedBlackCartesian *SpaceTimeGrid::makeFiveDimDWFRedBlackGrid(int Ls,const Gr
 {
   int N4=FourDimGrid->_ndimension;
   int cbd=1;
-  std::vector<int>   cb5(1,0);
+  Coordinate   cb5(1,0);
   for(int d=0;d<N4;d++){
-      cb5.push_back(1);
+    cb5.push_back(1);
   }
   GridCartesian *tmp         = makeFiveDimDWFGrid(Ls,FourDimGrid);
   GridRedBlackCartesian *ret = new GridRedBlackCartesian(tmp,cb5,cbd); 
@@ -113,5 +112,4 @@ GridRedBlackCartesian *SpaceTimeGrid::makeFiveDimDWFRedBlackGrid(int Ls,const Gr
   return ret;
 }
 
-
-}}
+NAMESPACE_END(Grid);
diff --git a/Grid/qcd/utils/SpaceTimeGrid.h b/Grid/qcd/utils/SpaceTimeGrid.h
index 61613e4d..6f8a4532 100644
--- a/Grid/qcd/utils/SpaceTimeGrid.h
+++ b/Grid/qcd/utils/SpaceTimeGrid.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,17 +23,17 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_QCD_SPACE_TIME_GRID_H
 #define GRID_QCD_SPACE_TIME_GRID_H
-namespace Grid {
-namespace QCD {
+
+NAMESPACE_BEGIN(Grid);
 
 class SpaceTimeGrid {
- public:
+public:
 
-  static GridCartesian         *makeFourDimGrid(const std::vector<int> & latt,const std::vector<int> &simd,const std::vector<int> &mpi);
+  static GridCartesian         *makeFourDimGrid(const Coordinate & latt,const Coordinate &simd,const Coordinate &mpi);
   static GridRedBlackCartesian *makeFourDimRedBlackGrid       (const GridCartesian *FourDimGrid);
 
   static GridCartesian         *makeFiveDimGrid        (int Ls,const GridCartesian *FourDimGrid);
@@ -41,10 +41,10 @@ class SpaceTimeGrid {
 
   static GridCartesian         *makeFiveDimDWFGrid        (int Ls,const GridCartesian *FourDimGrid);
   static GridRedBlackCartesian *makeFiveDimDWFRedBlackGrid(int Ls,const GridCartesian *FourDimGrid);
-  static GridCartesian         *makeFourDimDWFGrid        (const std::vector<int> & latt,const std::vector<int> &mpi);
+  static GridCartesian         *makeFourDimDWFGrid        (const Coordinate & latt,const Coordinate &mpi);
 
 };
 
-}}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/qcd/utils/WilsonLoops.h b/Grid/qcd/utils/WilsonLoops.h
index d4790df2..0367c9fa 100644
--- a/Grid/qcd/utils/WilsonLoops.h
+++ b/Grid/qcd/utils/WilsonLoops.h
@@ -33,8 +33,8 @@ directory
 /*  END LEGAL */
 #ifndef QCD_UTILS_WILSON_LOOPS_H
 #define QCD_UTILS_WILSON_LOOPS_H
-namespace Grid {
-namespace QCD {
+
+NAMESPACE_BEGIN(Grid);
 
 // Common wilson loop observables
 template <class Gimpl> class WilsonLoops : public Gimpl {
@@ -57,16 +57,16 @@ public:
     // purpose of deriving
     // from Gimpl.
     /*
-    plaq = Gimpl::CovShiftBackward(
-        U[mu], mu, Gimpl::CovShiftBackward(
-                       U[nu], nu, Gimpl::CovShiftForward(U[mu], mu, U[nu])));
-                       */
+      plaq = Gimpl::CovShiftBackward(
+      U[mu], mu, Gimpl::CovShiftBackward(
+      U[nu], nu, Gimpl::CovShiftForward(U[mu], mu, U[nu])));
+    */
     // _
     //|< _|
     plaq = Gimpl::CovShiftForward(U[mu],mu,
-           Gimpl::CovShiftForward(U[nu],nu,
-           Gimpl::CovShiftBackward(U[mu],mu,
-           Gimpl::CovShiftIdentityBackward(U[nu], nu))));
+				  Gimpl::CovShiftForward(U[nu],nu,
+							 Gimpl::CovShiftBackward(U[mu],mu,
+										 Gimpl::CovShiftIdentityBackward(U[nu], nu))));
 
 
 
@@ -78,7 +78,7 @@ public:
   static void traceDirPlaquette(ComplexField &plaq,
                                 const std::vector<GaugeMat> &U, const int mu,
                                 const int nu) {
-    GaugeMat sp(U[0]._grid);
+    GaugeMat sp(U[0].Grid());
     dirPlaquette(sp, U, mu, nu);
     plaq = trace(sp);
   }
@@ -87,8 +87,8 @@ public:
   //////////////////////////////////////////////////
   static void sitePlaquette(ComplexField &Plaq,
                             const std::vector<GaugeMat> &U) {
-    ComplexField sitePlaq(U[0]._grid);
-    Plaq = zero;
+    ComplexField sitePlaq(U[0].Grid());
+    Plaq = Zero();
     for (int mu = 1; mu < Nd; mu++) {
       for (int nu = 0; nu < mu; nu++) {
         traceDirPlaquette(sitePlaq, U, mu, nu);
@@ -100,13 +100,13 @@ public:
   // sum over all x,y,z,t and over all planes of plaquette
   //////////////////////////////////////////////////
   static RealD sumPlaquette(const GaugeLorentz &Umu) {
-    std::vector<GaugeMat> U(Nd, Umu._grid);
+    std::vector<GaugeMat> U(Nd, Umu.Grid());
     // inefficient here
     for (int mu = 0; mu < Nd; mu++) {
       U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
     }
 
-    ComplexField Plaq(Umu._grid);
+    ComplexField Plaq(Umu.Grid());
 
     sitePlaquette(Plaq, U);
     auto Tp = sum(Plaq);
@@ -120,7 +120,7 @@ public:
   //////////////////////////////////////////////////
   static RealD avgPlaquette(const GaugeLorentz &Umu) {
     RealD sumplaq = sumPlaquette(Umu);
-    double vol = Umu._grid->gSites();
+    double vol = Umu.Grid()->gSites();
     double faces = (1.0 * Nd * (Nd - 1)) / 2.0;
     return sumplaq / vol / faces / Nc; // Nd , Nc dependent... FIXME
   }
@@ -130,12 +130,12 @@ public:
   // average over all x,y,z the temporal loop
   //////////////////////////////////////////////////
   static ComplexD avgPolyakovLoop(const GaugeField &Umu) {  //assume Nd=4
-    GaugeMat Ut(Umu._grid), P(Umu._grid);
+    GaugeMat Ut(Umu.Grid()), P(Umu.Grid());
     ComplexD out;
-    int T = Umu._grid->GlobalDimensions()[3];
-    int X = Umu._grid->GlobalDimensions()[0];
-    int Y = Umu._grid->GlobalDimensions()[1];
-    int Z = Umu._grid->GlobalDimensions()[2];
+    int T = Umu.Grid()->GlobalDimensions()[3];
+    int X = Umu.Grid()->GlobalDimensions()[0];
+    int Y = Umu.Grid()->GlobalDimensions()[1];
+    int Z = Umu.Grid()->GlobalDimensions()[2];
 
     Ut = peekLorentz(Umu,3); //Select temporal direction
     P = Ut;
@@ -151,10 +151,10 @@ public:
   // average over traced single links
   //////////////////////////////////////////////////
   static RealD linkTrace(const GaugeLorentz &Umu) {
-    std::vector<GaugeMat> U(Nd, Umu._grid);
+    std::vector<GaugeMat> U(Nd, Umu.Grid());
 
-    ComplexField Tr(Umu._grid);
-    Tr = zero;
+    ComplexField Tr(Umu.Grid());
+    Tr = Zero();
     for (int mu = 0; mu < Nd; mu++) {
       U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
       Tr = Tr + trace(U[mu]);
@@ -163,7 +163,7 @@ public:
     auto Tp = sum(Tr);
     auto p = TensorRemove(Tp);
 
-    double vol = Umu._grid->gSites();
+    double vol = Umu.Grid()->gSites();
 
     return p.real() / vol / 4.0 / 3.0;
   };
@@ -174,13 +174,13 @@ public:
   static void Staple(GaugeMat &staple, const GaugeLorentz &Umu, int mu,
                      int nu) {
 
-    GridBase *grid = Umu._grid;
+    GridBase *grid = Umu.Grid();
 
     std::vector<GaugeMat> U(Nd, grid);
     for (int d = 0; d < Nd; d++) {
       U[d] = PeekIndex<LorentzIndex>(Umu, d);
     }
-    staple = zero;
+    staple = Zero();
 
     if (nu != mu) {
 
@@ -194,11 +194,11 @@ public:
       //
 
       staple += Gimpl::ShiftStaple(
-          Gimpl::CovShiftForward(
-              U[nu], nu,
-              Gimpl::CovShiftBackward(
-                  U[mu], mu, Gimpl::CovShiftIdentityBackward(U[nu], nu))),
-          mu);
+				   Gimpl::CovShiftForward(
+							  U[nu], nu,
+							  Gimpl::CovShiftBackward(
+										  U[mu], mu, Gimpl::CovShiftIdentityBackward(U[nu], nu))),
+				   mu);
 
       //  __
       // |
@@ -206,23 +206,23 @@ public:
       //
       //
       staple += Gimpl::ShiftStaple(
-          Gimpl::CovShiftBackward(U[nu], nu,
-                                  Gimpl::CovShiftBackward(U[mu], mu, U[nu])),
-          mu);
+				   Gimpl::CovShiftBackward(U[nu], nu,
+							   Gimpl::CovShiftBackward(U[mu], mu, U[nu])),
+				   mu);
     }
   }
 
 
-// For the force term
+  // For the force term
 /*
-static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
-    GridBase *grid = Umu._grid;
+  static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
+    GridBase *grid = Umu.Grid();
     std::vector<GaugeMat> U(Nd, grid);
     for (int d = 0; d < Nd; d++) {
       // this operation is taking too much time
       U[d] = PeekIndex<LorentzIndex>(Umu, d);
     }
-    staple = zero;
+    staple = Zero();
     GaugeMat tmp1(grid);
     GaugeMat tmp2(grid);
 
@@ -237,20 +237,20 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
       }
     }
     staple = U[mu]*staple;
-}
+  }
 */
   //////////////////////////////////////////////////
   // the sum over all staples on each site
   //////////////////////////////////////////////////
   static void Staple(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
 
-    GridBase *grid = Umu._grid;
+    GridBase *grid = Umu.Grid();
 
     std::vector<GaugeMat> U(Nd, grid);
     for (int d = 0; d < Nd; d++) {
       U[d] = PeekIndex<LorentzIndex>(Umu, d);
     }
-    staple = zero;
+    staple = Zero();
 
     for (int nu = 0; nu < Nd; nu++) {
 
@@ -266,11 +266,11 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
         //
      
         staple += Gimpl::ShiftStaple(
-            Gimpl::CovShiftForward(
-                U[nu], nu,
-                Gimpl::CovShiftBackward(
-                    U[mu], mu, Gimpl::CovShiftIdentityBackward(U[nu], nu))),
-            mu);
+				     Gimpl::CovShiftForward(
+							    U[nu], nu,
+							    Gimpl::CovShiftBackward(
+										    U[mu], mu, Gimpl::CovShiftIdentityBackward(U[nu], nu))),
+				     mu);
 
         //  __
         // |
@@ -279,8 +279,8 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
         //
 
         staple += Gimpl::ShiftStaple(
-            Gimpl::CovShiftBackward(U[nu], nu,
-                                    Gimpl::CovShiftBackward(U[mu], mu, U[nu])), mu);
+				     Gimpl::CovShiftBackward(U[nu], nu,
+							     Gimpl::CovShiftBackward(U[mu], mu, U[nu])), mu);
       }
     }
   }
@@ -291,7 +291,7 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   static void StapleUpper(GaugeMat &staple, const GaugeLorentz &Umu, int mu,
                           int nu) {
     if (nu != mu) {
-      GridBase *grid = Umu._grid;
+      GridBase *grid = Umu.Grid();
 
       std::vector<GaugeMat> U(Nd, grid);
       for (int d = 0; d < Nd; d++) {
@@ -308,11 +308,11 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
       //
 
       staple = Gimpl::ShiftStaple(
-          Gimpl::CovShiftForward(
-              U[nu], nu,
-              Gimpl::CovShiftBackward(
-                  U[mu], mu, Gimpl::CovShiftIdentityBackward(U[nu], nu))),
-          mu);
+				  Gimpl::CovShiftForward(
+							 U[nu], nu,
+							 Gimpl::CovShiftBackward(
+										 U[mu], mu, Gimpl::CovShiftIdentityBackward(U[nu], nu))),
+				  mu);
     }
   }
 
@@ -322,7 +322,7 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   static void StapleLower(GaugeMat &staple, const GaugeLorentz &Umu, int mu,
                           int nu) {
     if (nu != mu) {
-      GridBase *grid = Umu._grid;
+      GridBase *grid = Umu.Grid();
 
       std::vector<GaugeMat> U(Nd, grid);
       for (int d = 0; d < Nd; d++) {
@@ -339,7 +339,7 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
       //
       //
       staple = Gimpl::ShiftStaple(
-          Gimpl::CovShiftBackward(U[nu], nu,
+				  Gimpl::CovShiftBackward(U[nu], nu,
                                   Gimpl::CovShiftBackward(U[mu], mu, U[nu])),
           mu);
 
@@ -350,18 +350,18 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   //  Field Strength
   //////////////////////////////////////////////////////
   static void FieldStrength(GaugeMat &FS, const GaugeLorentz &Umu, int mu, int nu){
-      // Fmn +--<--+  Ut +--<--+
-      //     |     |     |     |
+    // Fmn +--<--+  Ut +--<--+
+    //     |     |     |     |
       //  (x)+-->--+     +-->--+(x)  - h.c.
-      //     |     |     |     |
-      //     +--<--+     +--<--+
+    //     |     |     |     |
+    //     +--<--+     +--<--+
 
-      GaugeMat Vup(Umu._grid), Vdn(Umu._grid);
-      StapleUpper(Vup, Umu, mu, nu);
-      StapleLower(Vdn, Umu, mu, nu);
-      GaugeMat v = Vup - Vdn;
-      GaugeMat u = PeekIndex<LorentzIndex>(Umu, mu);  // some redundant copies
-      GaugeMat vu = v*u;
+    GaugeMat Vup(Umu.Grid()), Vdn(Umu.Grid());
+    StapleUpper(Vup, Umu, mu, nu);
+    StapleLower(Vdn, Umu, mu, nu);
+    GaugeMat v = Vup - Vdn;
+    GaugeMat u = PeekIndex<LorentzIndex>(Umu, mu);  // some redundant copies
+    GaugeMat vu = v*u;
       //FS = 0.25*Ta(u*v + Cshift(vu, mu, -1));
       FS = (u*v + Cshift(vu, mu, -1));
       FS = 0.125*(FS - adj(FS));
@@ -371,13 +371,13 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
     // 4d topological charge
     assert(Nd==4);
     // Bx = -iF(y,z), By = -iF(z,y), Bz = -iF(x,y)
-    GaugeMat Bx(U._grid), By(U._grid), Bz(U._grid);
+    GaugeMat Bx(U.Grid()), By(U.Grid()), Bz(U.Grid());
     FieldStrength(Bx, U, Ydir, Zdir);
     FieldStrength(By, U, Zdir, Xdir);
     FieldStrength(Bz, U, Xdir, Ydir);
 
     // Ex = -iF(t,x), Ey = -iF(t,y), Ez = -iF(t,z)
-    GaugeMat Ex(U._grid), Ey(U._grid), Ez(U._grid);
+    GaugeMat Ex(U.Grid()), Ey(U.Grid()), Ez(U.Grid());
     FieldStrength(Ex, U, Tdir, Xdir);
     FieldStrength(Ey, U, Tdir, Ydir);
     FieldStrength(Ez, U, Tdir, Zdir);
@@ -396,26 +396,26 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   static void dirRectangle(GaugeMat &rect, const std::vector<GaugeMat> &U,
                            const int mu, const int nu) {
     rect = Gimpl::CovShiftForward(
-               U[mu], mu, Gimpl::CovShiftForward(U[mu], mu, U[nu])) * // ->->|
-           adj(Gimpl::CovShiftForward(
-               U[nu], nu, Gimpl::CovShiftForward(U[mu], mu, U[mu])));
+				  U[mu], mu, Gimpl::CovShiftForward(U[mu], mu, U[nu])) * // ->->|
+      adj(Gimpl::CovShiftForward(
+				 U[nu], nu, Gimpl::CovShiftForward(U[mu], mu, U[mu])));
     rect = rect +
-           Gimpl::CovShiftForward(
-               U[mu], mu, Gimpl::CovShiftForward(U[nu], nu, U[nu])) * // ->||
-               adj(Gimpl::CovShiftForward(
-                   U[nu], nu, Gimpl::CovShiftForward(U[nu], nu, U[mu])));
+      Gimpl::CovShiftForward(
+			     U[mu], mu, Gimpl::CovShiftForward(U[nu], nu, U[nu])) * // ->||
+      adj(Gimpl::CovShiftForward(
+				 U[nu], nu, Gimpl::CovShiftForward(U[nu], nu, U[mu])));
   }
   static void traceDirRectangle(ComplexField &rect,
                                 const std::vector<GaugeMat> &U, const int mu,
                                 const int nu) {
-    GaugeMat sp(U[0]._grid);
+    GaugeMat sp(U[0].Grid());
     dirRectangle(sp, U, mu, nu);
     rect = trace(sp);
   }
   static void siteRectangle(ComplexField &Rect,
                             const std::vector<GaugeMat> &U) {
-    ComplexField siteRect(U[0]._grid);
-    Rect = zero;
+    ComplexField siteRect(U[0].Grid());
+    Rect = Zero();
     for (int mu = 1; mu < Nd; mu++) {
       for (int nu = 0; nu < mu; nu++) {
         traceDirRectangle(siteRect, U, mu, nu);
@@ -428,13 +428,13 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   // sum over all x,y,z,t and over all planes of plaquette
   //////////////////////////////////////////////////
   static RealD sumRectangle(const GaugeLorentz &Umu) {
-    std::vector<GaugeMat> U(Nd, Umu._grid);
+    std::vector<GaugeMat> U(Nd, Umu.Grid());
 
     for (int mu = 0; mu < Nd; mu++) {
       U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
     }
 
-    ComplexField Rect(Umu._grid);
+    ComplexField Rect(Umu.Grid());
 
     siteRectangle(Rect, U);
 
@@ -449,7 +449,7 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
 
     RealD sumrect = sumRectangle(Umu);
 
-    double vol = Umu._grid->gSites();
+    double vol = Umu.Grid()->gSites();
 
     double faces = (1.0 * Nd * (Nd - 1)); // 2 distinct orientations summed
 
@@ -473,9 +473,9 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   static void RectStapleOptimised(GaugeMat &Stap, std::vector<GaugeMat> &U2,
                                   std::vector<GaugeMat> &U, int mu) {
 
-    Stap = zero;
+    Stap = Zero();
 
-    GridBase *grid = U[0]._grid;
+    GridBase *grid = U[0].Grid();
 
     GaugeMat Staple2x1(grid);
     GaugeMat tmp(grid);
@@ -552,14 +552,14 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
 
   static void RectStapleUnoptimised(GaugeMat &Stap, const GaugeLorentz &Umu,
                                     int mu) {
-    GridBase *grid = Umu._grid;
+    GridBase *grid = Umu.Grid();
 
     std::vector<GaugeMat> U(Nd, grid);
     for (int d = 0; d < Nd; d++) {
       U[d] = PeekIndex<LorentzIndex>(Umu, d);
     }
 
-    Stap = zero;
+    Stap = Zero();
 
     for (int nu = 0; nu < Nd; nu++) {
       if (nu != mu) {
@@ -567,52 +567,52 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
         //          |    __ |
         //
         Stap += Gimpl::ShiftStaple(
-            Gimpl::CovShiftForward(
-                U[mu], mu,
-                Gimpl::CovShiftForward(
-                    U[nu], nu,
-                    Gimpl::CovShiftBackward(
-                        U[mu], mu,
-                        Gimpl::CovShiftBackward(
-                            U[mu], mu,
-                            Gimpl::CovShiftIdentityBackward(U[nu], nu))))),
-            mu);
+				   Gimpl::CovShiftForward(
+							  U[mu], mu,
+							  Gimpl::CovShiftForward(
+										 U[nu], nu,
+										 Gimpl::CovShiftBackward(
+													 U[mu], mu,
+													 Gimpl::CovShiftBackward(
+																 U[mu], mu,
+																 Gimpl::CovShiftIdentityBackward(U[nu], nu))))),
+				   mu);
 
         //              __
         //          |__ __ |
 
         Stap += Gimpl::ShiftStaple(
-            Gimpl::CovShiftForward(
-                U[mu], mu,
-                Gimpl::CovShiftBackward(
-                    U[nu], nu,
-                    Gimpl::CovShiftBackward(
-                        U[mu], mu, Gimpl::CovShiftBackward(U[mu], mu, U[nu])))),
-            mu);
+				   Gimpl::CovShiftForward(
+							  U[mu], mu,
+							  Gimpl::CovShiftBackward(
+										  U[nu], nu,
+										  Gimpl::CovShiftBackward(
+													  U[mu], mu, Gimpl::CovShiftBackward(U[mu], mu, U[nu])))),
+				   mu);
 
         //           __
         //          |__ __ |
 
         Stap += Gimpl::ShiftStaple(
-            Gimpl::CovShiftBackward(
-                U[nu], nu,
-                Gimpl::CovShiftBackward(
-                    U[mu], mu,
-                    Gimpl::CovShiftBackward(
-                        U[mu], mu, Gimpl::CovShiftForward(U[nu], nu, U[mu])))),
-            mu);
+				   Gimpl::CovShiftBackward(
+							   U[nu], nu,
+							   Gimpl::CovShiftBackward(
+										   U[mu], mu,
+										   Gimpl::CovShiftBackward(
+													   U[mu], mu, Gimpl::CovShiftForward(U[nu], nu, U[mu])))),
+				   mu);
 
         //           __ ___
         //          |__    |
 
         Stap += Gimpl::ShiftStaple(
-            Gimpl::CovShiftForward(
-                U[nu], nu,
-                Gimpl::CovShiftBackward(
-                    U[mu], mu,
-                    Gimpl::CovShiftBackward(
-                        U[mu], mu, Gimpl::CovShiftBackward(U[nu], nu, U[mu])))),
-            mu);
+				   Gimpl::CovShiftForward(
+							  U[nu], nu,
+							  Gimpl::CovShiftBackward(
+										  U[mu], mu,
+										  Gimpl::CovShiftBackward(
+													  U[mu], mu, Gimpl::CovShiftBackward(U[nu], nu, U[mu])))),
+				   mu);
 
         //       --
         //      |  |
@@ -620,16 +620,16 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
         //      |  |
 
         Stap += Gimpl::ShiftStaple(
-            Gimpl::CovShiftForward(
-                U[nu], nu,
-                Gimpl::CovShiftForward(
-                    U[nu], nu,
-                    Gimpl::CovShiftBackward(
-                        U[mu], mu,
-                        Gimpl::CovShiftBackward(
-                            U[nu], nu,
-                            Gimpl::CovShiftIdentityBackward(U[nu], nu))))),
-            mu);
+				   Gimpl::CovShiftForward(
+							  U[nu], nu,
+							  Gimpl::CovShiftForward(
+										 U[nu], nu,
+										 Gimpl::CovShiftBackward(
+													 U[mu], mu,
+													 Gimpl::CovShiftBackward(
+																 U[nu], nu,
+																 Gimpl::CovShiftIdentityBackward(U[nu], nu))))),
+				   mu);
 
         //      |  |
         //
@@ -637,13 +637,13 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
         //       --
 
         Stap += Gimpl::ShiftStaple(
-            Gimpl::CovShiftBackward(
-                U[nu], nu,
-                Gimpl::CovShiftBackward(
-                    U[nu], nu,
-                    Gimpl::CovShiftBackward(
-                        U[mu], mu, Gimpl::CovShiftForward(U[nu], nu, U[nu])))),
-            mu);
+				   Gimpl::CovShiftBackward(
+							   U[nu], nu,
+							   Gimpl::CovShiftBackward(
+										   U[nu], nu,
+										   Gimpl::CovShiftBackward(
+													   U[mu], mu, Gimpl::CovShiftForward(U[nu], nu, U[nu])))),
+				   mu);
       }
     }
   }
@@ -679,7 +679,7 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
                                 const std::vector<GaugeMat> &U,
                                 const int Rmu, const int Rnu,
                                 const int mu, const int nu) {
-    GaugeMat sp(U[0]._grid);
+    GaugeMat sp(U[0].Grid());
     wilsonLoop(sp, U, Rmu, Rnu, mu, nu);
     wl = trace(sp);
   }
@@ -689,9 +689,9 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   static void siteWilsonLoop(LatticeComplex &Wl,
                             const std::vector<GaugeMat> &U,
                             const int R1, const int R2) {
-    LatticeComplex siteWl(U[0]._grid);
-    Wl = zero;
-    for (int mu = 1; mu < U[0]._grid->_ndimension; mu++) {
+    LatticeComplex siteWl(U[0].Grid());
+    Wl = Zero();
+    for (int mu = 1; mu < U[0].Grid()->_ndimension; mu++) {
       for (int nu = 0; nu < mu; nu++) {
         traceWilsonLoop(siteWl, U, R1, R2, mu, nu);
         Wl = Wl + siteWl;
@@ -707,11 +707,11 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   static void siteTimelikeWilsonLoop(LatticeComplex &Wl,
                             const std::vector<GaugeMat> &U,
                             const int R1, const int R2) {
-    LatticeComplex siteWl(U[0]._grid);
+    LatticeComplex siteWl(U[0].Grid());
 
-    int ndim = U[0]._grid->_ndimension;
+    int ndim = U[0].Grid()->_ndimension;
 
-    Wl = zero;
+    Wl = Zero();
     for (int nu = 0; nu < ndim - 1; nu++) {
       traceWilsonLoop(siteWl, U, R1, R2, ndim-1, nu);
       Wl = Wl + siteWl;
@@ -723,10 +723,10 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   static void siteSpatialWilsonLoop(LatticeComplex &Wl,
                             const std::vector<GaugeMat> &U,
                             const int R1, const int R2) {
-    LatticeComplex siteWl(U[0]._grid);
+    LatticeComplex siteWl(U[0].Grid());
 
-    Wl = zero;
-    for (int mu = 1; mu < U[0]._grid->_ndimension - 1; mu++) {
+    Wl = Zero();
+    for (int mu = 1; mu < U[0].Grid()->_ndimension - 1; mu++) {
       for (int nu = 0; nu < mu; nu++) {
         traceWilsonLoop(siteWl, U, R1, R2, mu, nu);
         Wl = Wl + siteWl;
@@ -740,13 +740,13 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   //////////////////////////////////////////////////
   static Real sumWilsonLoop(const GaugeLorentz &Umu,
                             const int R1, const int R2) {
-    std::vector<GaugeMat> U(4, Umu._grid);
+    std::vector<GaugeMat> U(4, Umu.Grid());
 
-    for (int mu = 0; mu < Umu._grid->_ndimension; mu++) {
+    for (int mu = 0; mu < Umu.Grid()->_ndimension; mu++) {
       U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
     }
 
-    LatticeComplex Wl(Umu._grid);
+    LatticeComplex Wl(Umu.Grid());
 
     siteWilsonLoop(Wl, U, R1, R2);
 
@@ -759,13 +759,13 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   //////////////////////////////////////////////////
   static Real sumTimelikeWilsonLoop(const GaugeLorentz &Umu,
                             const int R1, const int R2) {
-    std::vector<GaugeMat> U(4, Umu._grid);
+    std::vector<GaugeMat> U(4, Umu.Grid());
 
-    for (int mu = 0; mu < Umu._grid->_ndimension; mu++) {
+    for (int mu = 0; mu < Umu.Grid()->_ndimension; mu++) {
       U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
     }
 
-    LatticeComplex Wl(Umu._grid);
+    LatticeComplex Wl(Umu.Grid());
 
     siteTimelikeWilsonLoop(Wl, U, R1, R2);
 
@@ -778,13 +778,13 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   //////////////////////////////////////////////////
   static Real sumSpatialWilsonLoop(const GaugeLorentz &Umu,
                             const int R1, const int R2) {
-    std::vector<GaugeMat> U(4, Umu._grid);
+    std::vector<GaugeMat> U(4, Umu.Grid());
 
-    for (int mu = 0; mu < Umu._grid->_ndimension; mu++) {
+    for (int mu = 0; mu < Umu.Grid()->_ndimension; mu++) {
       U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
     }
 
-    LatticeComplex Wl(Umu._grid);
+    LatticeComplex Wl(Umu.Grid());
 
     siteSpatialWilsonLoop(Wl, U, R1, R2);
 
@@ -797,9 +797,9 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   //////////////////////////////////////////////////
   static Real avgWilsonLoop(const GaugeLorentz &Umu,
                             const int R1, const int R2) {
-    int ndim = Umu._grid->_ndimension;
+    int ndim = Umu.Grid()->_ndimension;
     Real sumWl = sumWilsonLoop(Umu, R1, R2);
-    Real vol = Umu._grid->gSites();
+    Real vol = Umu.Grid()->gSites();
     Real faces = 1.0 * ndim * (ndim - 1);
     return sumWl / vol / faces / Nc; // Nc dependent... FIXME
   }
@@ -808,9 +808,9 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   //////////////////////////////////////////////////
   static Real avgTimelikeWilsonLoop(const GaugeLorentz &Umu,
                             const int R1, const int R2) {
-    int ndim = Umu._grid->_ndimension;
+    int ndim = Umu.Grid()->_ndimension;
     Real sumWl = sumTimelikeWilsonLoop(Umu, R1, R2);
-    Real vol = Umu._grid->gSites();
+    Real vol = Umu.Grid()->gSites();
     Real faces = 1.0 * (ndim - 1);
     return sumWl / vol / faces / Nc; // Nc dependent... FIXME
   }
@@ -819,9 +819,9 @@ static void StapleMult(GaugeMat &staple, const GaugeLorentz &Umu, int mu) {
   //////////////////////////////////////////////////
   static Real avgSpatialWilsonLoop(const GaugeLorentz &Umu,
                             const int R1, const int R2) {
-    int ndim = Umu._grid->_ndimension;
+    int ndim = Umu.Grid()->_ndimension;
     Real sumWl = sumSpatialWilsonLoop(Umu, R1, R2);
-    Real vol = Umu._grid->gSites();
+    Real vol = Umu.Grid()->gSites();
     Real faces = 1.0 * (ndim - 1) * (ndim - 2);
     return sumWl / vol / faces / Nc; // Nc dependent... FIXME
   }
@@ -831,7 +831,7 @@ typedef WilsonLoops<PeriodicGimplR> ColourWilsonLoops;
 typedef WilsonLoops<PeriodicGimplR> U1WilsonLoops;
 typedef WilsonLoops<PeriodicGimplR> SU2WilsonLoops;
 typedef WilsonLoops<PeriodicGimplR> SU3WilsonLoops;
-}
-}
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/serialisation/BaseIO.h b/Grid/serialisation/BaseIO.h
index e9d5bedf..a9147c49 100644
--- a/Grid/serialisation/BaseIO.h
+++ b/Grid/serialisation/BaseIO.h
@@ -33,7 +33,7 @@ Author: Guido Cossu <guido.cossu@ed.ac.uk>
 #include <type_traits>
 #include <Grid/tensors/Tensors.h>
 #include <Grid/serialisation/VectorUtils.h>
-#include <Grid/Eigen/unsupported/CXX11/Tensor>
+#include <Grid/Grid_Eigen_Tensor.h>
 
 namespace Grid {
   namespace EigenIO {
diff --git a/Grid/serialisation/BinaryIO.cc b/Grid/serialisation/BinaryIO.cc
index a4d26b69..8ddf3c6c 100644
--- a/Grid/serialisation/BinaryIO.cc
+++ b/Grid/serialisation/BinaryIO.cc
@@ -28,7 +28,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     /*  END LEGAL */
 #include <Grid/GridCore.h>
 
-using namespace Grid;
+NAMESPACE_BEGIN(Grid);
 
 // Writer implementation ///////////////////////////////////////////////////////
 BinaryWriter::BinaryWriter(const std::string &fileName)
@@ -68,3 +68,5 @@ void BinaryReader::readDefault(const std::string &s, std::string &output)
     output.resize(sz);
     file_.read((char *)output.data(), sz);
 }
+
+NAMESPACE_END(Grid);
diff --git a/Grid/serialisation/BinaryIO.h b/Grid/serialisation/BinaryIO.h
index bf011454..9ba34c16 100644
--- a/Grid/serialisation/BinaryIO.h
+++ b/Grid/serialisation/BinaryIO.h
@@ -37,8 +37,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <vector>
 #include <cassert>
 
-namespace Grid {
-  
+NAMESPACE_BEGIN(Grid);
+
   class BinaryWriter: public Writer<BinaryWriter>
   {
   public:
@@ -163,6 +163,6 @@ namespace Grid {
     for( auto i = 0; i < count; ++i)
       read("", buf[i]);
   }
-}
 
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/serialisation/Hdf5Type.h b/Grid/serialisation/Hdf5Type.h
index 8634f35b..64dda349 100644
--- a/Grid/serialisation/Hdf5Type.h
+++ b/Grid/serialisation/Hdf5Type.h
@@ -49,14 +49,14 @@ namespace Grid
   DEFINE_HDF5_NATIVE_TYPES;
   
   template <typename R>
-  class Hdf5Type<std::complex<R>>
+  class Hdf5Type<Grid::complex<R> >
   {
   public:
     static inline const H5NS::DataType & type(void)
     {
       if (typePtr_ == nullptr)
       {
-        typePtr_.reset(new H5NS::CompType(sizeof(std::complex<R>)));
+        typePtr_.reset(new H5NS::CompType(sizeof(Grid::complex<R>)));
         typePtr_->insertMember("re", 0,         Hdf5Type<R>::type());
         typePtr_->insertMember("im", sizeof(R), Hdf5Type<R>::type());
       }
@@ -69,7 +69,8 @@ namespace Grid
   };
   
   template <typename R>
-  std::unique_ptr<H5NS::CompType> Hdf5Type<std::complex<R>>::typePtr_ = nullptr;
+  std::unique_ptr<H5NS::CompType> Hdf5Type<Grid::complex<R>>::typePtr_ = nullptr;
+
 }
 
 #undef HDF5_NATIVE_TYPE
diff --git a/Grid/serialisation/JSON_IO.cc b/Grid/serialisation/JSON_IO.cc
index 98edc375..aca8bab3 100644
--- a/Grid/serialisation/JSON_IO.cc
+++ b/Grid/serialisation/JSON_IO.cc
@@ -26,8 +26,9 @@
     *************************************************************************************/
     /*  END LEGAL */
 #include <Grid/Grid.h>
+#ifndef __NVCC__
 
-using namespace Grid;
+NAMESPACE_BEGIN(Grid);
 
 // Writer implementation ///////////////////////////////////////////////////////
 JSONWriter::JSONWriter(const std::string &fileName)
@@ -73,20 +74,16 @@ void JSONWriter::delete_comma()
 // compiles fine with clang
 // have to wrap in the Grid namespace
 // annoying, but necessary for TravisCI
-namespace Grid
+void JSONWriter::writeDefault(const std::string &s,	const std::string &x)
 {
-  void JSONWriter::writeDefault(const std::string &s,	const std::string &x)
-  {
     //std::cout << "JSONWriter::writeDefault(string) : " << s <<  std::endl;
-    std::ostringstream os;
-    os << std::boolalpha << x;
-    if (s.size())
-      ss_ << "\""<< s << "\" : \"" << os.str() << "\" ," ;
-    else
-     ss_ << os.str() << " ," ;
-  }
-}// namespace Grid 
-
+  std::ostringstream os;
+  os << std::boolalpha << x;
+  if (s.size())
+    ss_ << "\""<< s << "\" : \"" << os.str() << "\" ," ;
+  else
+    ss_ << os.str() << " ," ;
+}
 
 // Reader implementation ///////////////////////////////////////////////////////
 JSONReader::JSONReader(const std::string &fileName)
@@ -151,7 +148,7 @@ bool JSONReader::nextElement(const std::string &s)
   //}
 
   jcur_ = *it_; 
-  //cout << "JSONReader::nextElement(string) : " << s << " : "<< jcur_ << endl;
+  //cout << "JSONReader::nextElement(std::string) : " << s << " : "<< jcur_ << endl;
   //return true;
 
     return false;
@@ -171,3 +168,5 @@ void JSONReader::readDefault(const std::string &s, std::string &output)
     output = jcur_;    
   }
 }
+NAMESPACE_END(Grid);
+#endif
diff --git a/Grid/serialisation/JSON_IO.h b/Grid/serialisation/JSON_IO.h
index 00287447..b5255de3 100644
--- a/Grid/serialisation/JSON_IO.h
+++ b/Grid/serialisation/JSON_IO.h
@@ -54,6 +54,18 @@ namespace Grid
     void pop(void);
     template <typename U>
     void writeDefault(const std::string &s, const U &x);
+#ifdef __NVCC__
+    void writeDefault(const std::string &s, const Grid::ComplexD &x) 
+    { 
+      std::complex<double> z(real(x),imag(x));
+      writeDefault(s,z);
+    }
+    void writeDefault(const std::string &s, const Grid::ComplexF &x) 
+    { 
+      std::complex<float> z(real(x),imag(x));
+      writeDefault(s,z);
+    }
+#endif
     template <typename U>
     void writeDefault(const std::string &s, const std::complex<U> &x);
     template <typename U>
@@ -89,6 +101,21 @@ namespace Grid
     void readDefault(const std::string &s, std::vector<U> &output);
     template <typename U, typename P>
     void readDefault(const std::string &s, std::pair<U,P> &output);
+#ifdef __NVCC__
+    void readDefault(const std::string &s, ComplexD &output)
+    { 
+      std::complex<double> z;
+      readDefault(s,z);
+      output = ComplexD(real(z),imag(z));
+    }
+    void readDefault(const std::string &s, ComplexF &output)
+    { 
+      std::complex<float> z;
+      readDefault(s,z);
+      output = ComplexD(real(z),imag(z));
+    }
+#endif
+
   private:
     json                jobject_; // main object
     json                jcur_;  // current json object
diff --git a/Grid/serialisation/MacroMagic.h b/Grid/serialisation/MacroMagic.h
index ce305673..7866327e 100644
--- a/Grid/serialisation/MacroMagic.h
+++ b/Grid/serialisation/MacroMagic.h
@@ -42,8 +42,6 @@ THE SOFTWARE.
 */
 ///////////////////////////////////////////
 
-#define strong_inline __attribute__((always_inline)) inline
-
 #ifndef MAX
 #define MAX(x,y) ((x)>(y)?(x):(y))
 #define MIN(x,y) ((x)>(y)?(y):(x))
@@ -109,10 +107,11 @@ THE SOFTWARE.
 //////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
 
 #define GRID_MACRO_MEMBER(A,B)        A B;
+
 #define GRID_MACRO_COMP_MEMBER(A,B) result = (result and CompareMember(lhs. B, rhs. B));
 #define GRID_MACRO_OS_WRITE_MEMBER(A,B) os<< #A <<" " #B << " = "; WriteMember( os, obj. B ); os << " ; " <<std::endl;
-#define GRID_MACRO_READ_MEMBER(A,B) Grid::read(RD,#B,obj. B);
-#define GRID_MACRO_WRITE_MEMBER(A,B) Grid::write(WR,#B,obj. B);
+#define GRID_MACRO_READ_MEMBER(A,B) ::Grid::read(RD,#B,obj. B);
+#define GRID_MACRO_WRITE_MEMBER(A,B) ::Grid::write(WR,#B,obj. B);
 
 #define GRID_SERIALIZABLE_CLASS_MEMBERS(cname,...)\
   std::string SerialisableClassName(void) const {return std::string(#cname);}	\
@@ -127,7 +126,7 @@ template <typename T>\
 static inline void read(Reader<T> &RD,const std::string &s, cname &obj){	\
   if (!push(RD,s))\
   {\
-    std::cout << Grid::GridLogWarning << "IO: Cannot open node '" << s << "'" << std::endl;\
+    std::cout << ::Grid::GridLogWarning << "IO: Cannot open node '" << s << "'" << std::endl; \
     return;\
   };\
   GRID_MACRO_EVAL(GRID_MACRO_MAP(GRID_MACRO_READ_MEMBER,__VA_ARGS__))	\
@@ -147,12 +146,12 @@ friend inline bool operator==(const cname &lhs, const cname &rhs) {\
 
 #define GRID_ENUM_TYPE(obj) std::remove_reference<decltype(obj)>::type
 #define GRID_MACRO_ENUMVAL(A,B) A = B,
-#define GRID_MACRO_ENUMCASE(A,B) case GRID_ENUM_TYPE(obj)::A: Grid::write(WR,s,#A); break;
+#define GRID_MACRO_ENUMCASE(A,B) case GRID_ENUM_TYPE(obj)::A: ::Grid::write(WR,s,#A); break;
 #define GRID_MACRO_ENUMTEST(A,B) else if (buf == #A) {obj = GRID_ENUM_TYPE(obj)::A;}
 #define GRID_MACRO_ENUMCASEIO(A,B) case GRID_ENUM_TYPE(obj)::A: os << #A; break;
 
 #define GRID_SERIALIZABLE_ENUM(name,undefname,...)\
-class name: public Grid::Serializable\
+class name: public ::Grid::Serializable	  \
 {\
 public:\
   enum\
@@ -161,28 +160,28 @@ public:\
     undefname = -1\
   };\
 public:\
-  name(void): value_(undefname) {};\
-  name(int value): value_(value) {};\
+  accelerator name(void)     : value_(undefname) {};		\
+  accelerator name(int value): value_(value) {};			\
   template <typename T>\
-  static inline void write(Grid::Writer<T> &WR,const std::string &s, const name &obj)\
+  static inline void write(::Grid::Writer<T> &WR,const std::string &s, const name &obj) \
   {\
     switch (obj.value_)\
     {\
       GRID_MACRO_EVAL(GRID_MACRO_MAP(GRID_MACRO_ENUMCASE,__VA_ARGS__))\
-      default: Grid::write(WR,s,#undefname); break;\
+      default: ::Grid::write(WR,s,#undefname); break;\
     }\
   }\
   \
   template <typename T>\
-  static inline void read(Grid::Reader<T> &RD,const std::string &s, name &obj)\
+  static inline void read(::Grid::Reader<T> &RD,const std::string &s, name &obj)\
   {\
     std::string buf;\
-    Grid::read(RD, s, buf);\
+    ::Grid::read(RD, s, buf);\
     if (buf == #undefname) {obj = name::undefname;}\
     GRID_MACRO_EVAL(GRID_MACRO_MAP(GRID_MACRO_ENUMTEST,__VA_ARGS__))\
     else {obj = name::undefname;}\
   }\
-  inline operator int(void) const\
+  accelerator_inline operator int(void) const\
   {\
     return value_;\
   }\
diff --git a/Grid/serialisation/Serialisation.h b/Grid/serialisation/Serialisation.h
index 0e1c7531..c95226b1 100644
--- a/Grid/serialisation/Serialisation.h
+++ b/Grid/serialisation/Serialisation.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,8 +24,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_SERIALISATION_READER_H
 #define GRID_SERIALISATION_READER_H
 
@@ -36,7 +36,9 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include "BinaryIO.h"
 #include "TextIO.h"
 #include "XmlIO.h"
+#ifndef GRID_NVCC
 #include "JSON_IO.h"
+#endif
 
 #ifdef HAVE_HDF5
 #include "Hdf5IO.h"
@@ -50,8 +52,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 //////////////////////////////////////////
 // Select the default serialiser use ifdef's
 //////////////////////////////////////////
-namespace Grid {
-  typedef XmlReader DefaultReader;
-  typedef XmlWriter DefaultWriter;
-}
+NAMESPACE_BEGIN(Grid);
+typedef XmlReader DefaultReader;
+typedef XmlWriter DefaultWriter;
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/serialisation/TextIO.cc b/Grid/serialisation/TextIO.cc
index 77c45fd9..ee1d01cd 100644
--- a/Grid/serialisation/TextIO.cc
+++ b/Grid/serialisation/TextIO.cc
@@ -29,11 +29,9 @@
     /*  END LEGAL */
 #include <Grid/GridCore.h>
 
-using namespace Grid;
-
 #define GRID_TEXT_INDENT 2      //number of spaces for indentation of levels
 
-
+NAMESPACE_BEGIN(Grid);
 // Writer implementation ///////////////////////////////////////////////////////
 TextWriter::TextWriter(const std::string &fileName)
 : file_(fileName, std::ios::out)
@@ -103,3 +101,4 @@ void TextReader::readDefault(const std::string &s, std::string &output)
     output.clear();
     getline(file_, output);
 }
+NAMESPACE_END(Grid);
diff --git a/Grid/serialisation/TextIO.h b/Grid/serialisation/TextIO.h
index a6f2d4cf..1709dbab 100644
--- a/Grid/serialisation/TextIO.h
+++ b/Grid/serialisation/TextIO.h
@@ -112,6 +112,9 @@ namespace Grid
   }
 
   // Reader template implementation ////////////////////////////////////////////
+  template <>
+  void TextReader::readDefault(const std::string &s, std::string &output);
+
   template <typename U>
   void TextReader::readDefault(const std::string &s, U &output)
   {
@@ -121,9 +124,6 @@ namespace Grid
     fromString(output, buf);
   }
   
-  template <>
-  void TextReader::readDefault(const std::string &s, std::string &output);
-  
   template <typename U>
   void TextReader::readDefault(const std::string &s, std::vector<U> &output)
   {
@@ -153,6 +153,7 @@ namespace Grid
     for( auto &x : buf )
       read(s, x);
   }
+
 }
 
 #endif
diff --git a/Grid/serialisation/XmlIO.cc b/Grid/serialisation/XmlIO.cc
index 96f07f34..fae52ac1 100644
--- a/Grid/serialisation/XmlIO.cc
+++ b/Grid/serialisation/XmlIO.cc
@@ -28,9 +28,9 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     /*  END LEGAL */
 #include <Grid/GridCore.h>
 
-using namespace Grid;
+NAMESPACE_BEGIN(Grid);
 
-void Grid::xmlCheckParse(const pugi::xml_parse_result &result, const std::string name)
+void xmlCheckParse(const pugi::xml_parse_result &result, const std::string name)
 {
   if (!result) 
   {
@@ -115,7 +115,7 @@ XmlReader::XmlReader(const std::string &s,  const bool isBuffer,
     xmlCheckParse(result, "file '" + fileName_ + "'");
   }
   if ( toplev == std::string("") ) {
-  node_ = doc_;
+    node_ = doc_;
   } else { 
     node_ = doc_.child(toplev.c_str());
   }
@@ -186,3 +186,4 @@ void XmlReader::readDefault(const std::string &s, std::string &output)
     output = ""; 
   }
 }
+NAMESPACE_END(Grid);
diff --git a/Grid/serialisation/XmlIO.h b/Grid/serialisation/XmlIO.h
index 40d5f2bb..cb9a49e7 100644
--- a/Grid/serialisation/XmlIO.h
+++ b/Grid/serialisation/XmlIO.h
@@ -48,7 +48,7 @@ namespace Grid
   class XmlWriter: public Writer<XmlWriter>
   {    
   public:
-    XmlWriter(const std::string &fileName, std::string toplev = std::string("grid") );
+    XmlWriter(const std::string &fileName,std::string toplev = std::string("grid") );
     virtual ~XmlWriter(void);
     void push(const std::string &s);
     void pushXmlString(const std::string &s);
@@ -79,11 +79,11 @@ namespace Grid
     bool nextElement(const std::string &s = "");
     template <typename U>
     void readDefault(const std::string &s, U &output);
-    template <typename U>
-    void readDefault(const std::string &s, std::vector<U> &output);
+    template <typename U>    void readDefault(const std::string &s, std::vector<U> &output);
     template <typename U>
     void readMultiDim(const std::string &s, std::vector<U> &buf, std::vector<size_t> &dim);
     void readCurrentSubtree(std::string &s);
+
   private:
     void checkParse(const pugi::xml_parse_result &result, const std::string name);
   private:
@@ -109,9 +109,9 @@ namespace Grid
   {
     std::ostringstream os;
     
-    if (getPrecision())
+    if (this->getPrecision())
     {
-      os.precision(getPrecision());
+      os.precision(this->getPrecision());
     }
     if (isScientific())
     {
@@ -166,6 +166,7 @@ namespace Grid
   }
 
   // Reader template implementation ////////////////////////////////////////////
+  template <> void XmlReader::readDefault(const std::string &s, std::string &output);
   template <typename U>
   void XmlReader::readDefault(const std::string &s, U &output)
   {
@@ -175,9 +176,6 @@ namespace Grid
     fromString(output, buf);
   }
   
-  template <>
-  void XmlReader::readDefault(const std::string &s, std::string &output);
-  
   template <typename U>
   void XmlReader::readDefault(const std::string &s, std::vector<U> &output)
   {
diff --git a/Grid/simd/BGQQPX.h b/Grid/simd/BGQQPX.h
index 34888ab7..c263e6c7 100644
--- a/Grid/simd/BGQQPX.h
+++ b/Grid/simd/BGQQPX.h
@@ -1,4 +1,4 @@
-   /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,8 +23,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_ASM_BGQ_QPX_H
 #define GRID_ASM_BGQ_QPX_H
 
@@ -127,20 +127,20 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 /*********************************************************
  * Macro sequences encoding QCD
  *********************************************************/
-#define LOCK_GAUGE(dir)							\
-  {									\
-    uint8_t *byte_addr = (uint8_t *)&U._odata[sU](dir);			\
-    for(int i=0;i< 18*2*BYTES_PER_WORD*8;i+=32){			\
-      CACHE_LOCK(&byte_addr[i]);					\
-    }									\
+#define LOCK_GAUGE(dir)					\
+  {							\
+    uint8_t *byte_addr = (uint8_t *)&U[sU](dir);	\
+    for(int i=0;i< 18*2*BYTES_PER_WORD*8;i+=32){	\
+      CACHE_LOCK(&byte_addr[i]);			\
+    }							\
   }
 
-#define UNLOCK_GAUGE(dir)						\
-  {									\
-    uint8_t *byte_addr = (uint8_t *)&U._odata[sU](dir);			\
-    for(int i=0;i< 18*2*BYTES_PER_WORD*8;i+=32){			\
-      CACHE_UNLOCK(&byte_addr[i]);					\
-    }									\
+#define UNLOCK_GAUGE(dir)				\
+  {							\
+    uint8_t *byte_addr = (uint8_t *)&U[sU](dir);	\
+    for(int i=0;i< 18*2*BYTES_PER_WORD*8;i+=32){	\
+      CACHE_UNLOCK(&byte_addr[i]);			\
+    }							\
   }
 
 #define MAYBEPERM(A,B)
@@ -150,14 +150,14 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define PERMUTE_DIR1 
 #define PERMUTE_DIR0 
 
-#define MULT_2SPIN_DIR_PFXP(A,p) MULT_2SPIN(&U._odata[sU](A),p)
-#define MULT_2SPIN_DIR_PFYP(A,p) MULT_2SPIN(&U._odata[sU](A),p)
-#define MULT_2SPIN_DIR_PFZP(A,p) MULT_2SPIN(&U._odata[sU](A),p)
-#define MULT_2SPIN_DIR_PFTP(A,p) MULT_2SPIN(&U._odata[sU](A),p)
-#define MULT_2SPIN_DIR_PFXM(A,p) MULT_2SPIN(&U._odata[sU](A),p)
-#define MULT_2SPIN_DIR_PFYM(A,p) MULT_2SPIN(&U._odata[sU](A),p)
-#define MULT_2SPIN_DIR_PFZM(A,p) MULT_2SPIN(&U._odata[sU](A),p)
-#define MULT_2SPIN_DIR_PFTM(A,p) MULT_2SPIN(&U._odata[sU](A),p)
+#define MULT_2SPIN_DIR_PFXP(A,p) MULT_2SPIN(&U[sU](A),p)
+#define MULT_2SPIN_DIR_PFYP(A,p) MULT_2SPIN(&U[sU](A),p)
+#define MULT_2SPIN_DIR_PFZP(A,p) MULT_2SPIN(&U[sU](A),p)
+#define MULT_2SPIN_DIR_PFTP(A,p) MULT_2SPIN(&U[sU](A),p)
+#define MULT_2SPIN_DIR_PFXM(A,p) MULT_2SPIN(&U[sU](A),p)
+#define MULT_2SPIN_DIR_PFYM(A,p) MULT_2SPIN(&U[sU](A),p)
+#define MULT_2SPIN_DIR_PFZM(A,p) MULT_2SPIN(&U[sU](A),p)
+#define MULT_2SPIN_DIR_PFTM(A,p) MULT_2SPIN(&U[sU](A),p)
 
 #define MULT_SPIN(ptr,p) {					\
     uint64_t ub = ((uint64_t)base);				\
@@ -165,369 +165,369 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
          VLOAD(%0,%3,U0)					\
          VLOAD(%1,%3,U1)					\
          VLOAD(%2,%3,U2)					\
-	 VMUL_RR_RI(U0,Chi_00,UChi_00)					\
-	 VMUL_RR_RI(U1,Chi_00,UChi_01)					\
-	 VMUL_RR_RI(U2,Chi_00,UChi_02)					\
-	 VMUL_RR_RI(U0,Chi_10,UChi_10)					\
-	 VMUL_RR_RI(U1,Chi_10,UChi_11)					\
-	 VMUL_RR_RI(U2,Chi_10,UChi_12)					\
-	 VMADD_MII_IR(U0,Chi_00,UChi_00,UChi_00)			\
-	 VMADD_MII_IR(U1,Chi_00,UChi_01,UChi_01)			\
-	 VMADD_MII_IR(U2,Chi_00,UChi_02,UChi_02)			\
-	 VMADD_MII_IR(U0,Chi_10,UChi_10,UChi_10)			\
-	 VMADD_MII_IR(U1,Chi_10,UChi_11,UChi_11)			\
-	 VMADD_MII_IR(U2,Chi_10,UChi_12,UChi_12)			\
-	 : : "r" (0), "r" (32*3), "r" (32*6), "r" (ub ));		\
-    asm (								\
-         VLOAD(%0,%3,U0)						\
-         VLOAD(%1,%3,U1)						\
-         VLOAD(%2,%3,U2)						\
-	 VMADD_RR_RI(U0,Chi_01,UChi_00,UChi_00)				\
-	 VMADD_RR_RI(U1,Chi_01,UChi_01,UChi_01)				\
-	 VMADD_RR_RI(U2,Chi_01,UChi_02,UChi_02)				\
-	 VMADD_RR_RI(U0,Chi_11,UChi_10,UChi_10)				\
-	 VMADD_RR_RI(U1,Chi_11,UChi_11,UChi_11)				\
-	 VMADD_RR_RI(U2,Chi_11,UChi_12,UChi_12)				\
-	 VMADD_MII_IR(U0,Chi_01,UChi_00,UChi_00)			\
-	 VMADD_MII_IR(U1,Chi_01,UChi_01,UChi_01)			\
-	 VMADD_MII_IR(U2,Chi_01,UChi_02,UChi_02)			\
-	 VMADD_MII_IR(U0,Chi_11,UChi_10,UChi_10)			\
-	 VMADD_MII_IR(U1,Chi_11,UChi_11,UChi_11)			\
-	 VMADD_MII_IR(U2,Chi_11,UChi_12,UChi_12)			\
-	 : : "r" (32), "r" (32*4), "r" (32*7), "r" (ub ));		\
-    asm (								\
-         VLOAD(%0,%3,U0)						\
-         VLOAD(%1,%3,U1)						\
-         VLOAD(%2,%3,U2)						\
-	 VMADD_RR_RI(U0,Chi_02,UChi_00,UChi_00)				\
-	 VMADD_RR_RI(U1,Chi_02,UChi_01,UChi_01)				\
-	 VMADD_RR_RI(U2,Chi_02,UChi_02,UChi_02)				\
-	 VMADD_RR_RI(U0,Chi_12,UChi_10,UChi_10)				\
-	 VMADD_RR_RI(U1,Chi_12,UChi_11,UChi_11)				\
-	 VMADD_RR_RI(U2,Chi_12,UChi_12,UChi_12)				\
-	 VMADD_MII_IR(U0,Chi_02,UChi_00,UChi_00)			\
-	 VMADD_MII_IR(U1,Chi_02,UChi_01,UChi_01)			\
-	 VMADD_MII_IR(U2,Chi_02,UChi_02,UChi_02)			\
-	 VMADD_MII_IR(U0,Chi_12,UChi_10,UChi_10)			\
-	 VMADD_MII_IR(U1,Chi_12,UChi_11,UChi_11)			\
-	 VMADD_MII_IR(U2,Chi_12,UChi_12,UChi_12)			\
-	 : : "r" (32*2), "r" (32*5), "r" (32*8), "r" (ub ));		\
+	 VMUL_RR_RI(U0,Chi_00,UChi_00)				\
+	 VMUL_RR_RI(U1,Chi_00,UChi_01)				\
+	 VMUL_RR_RI(U2,Chi_00,UChi_02)				\
+	 VMUL_RR_RI(U0,Chi_10,UChi_10)				\
+	 VMUL_RR_RI(U1,Chi_10,UChi_11)				\
+	 VMUL_RR_RI(U2,Chi_10,UChi_12)				\
+	 VMADD_MII_IR(U0,Chi_00,UChi_00,UChi_00)		\
+	 VMADD_MII_IR(U1,Chi_00,UChi_01,UChi_01)		\
+	 VMADD_MII_IR(U2,Chi_00,UChi_02,UChi_02)		\
+	 VMADD_MII_IR(U0,Chi_10,UChi_10,UChi_10)		\
+	 VMADD_MII_IR(U1,Chi_10,UChi_11,UChi_11)		\
+	 VMADD_MII_IR(U2,Chi_10,UChi_12,UChi_12)		\
+	 : : "r" (0), "r" (32*3), "r" (32*6), "r" (ub ));	\
+    asm (							\
+         VLOAD(%0,%3,U0)					\
+         VLOAD(%1,%3,U1)					\
+         VLOAD(%2,%3,U2)					\
+	 VMADD_RR_RI(U0,Chi_01,UChi_00,UChi_00)			\
+	 VMADD_RR_RI(U1,Chi_01,UChi_01,UChi_01)			\
+	 VMADD_RR_RI(U2,Chi_01,UChi_02,UChi_02)			\
+	 VMADD_RR_RI(U0,Chi_11,UChi_10,UChi_10)			\
+	 VMADD_RR_RI(U1,Chi_11,UChi_11,UChi_11)			\
+	 VMADD_RR_RI(U2,Chi_11,UChi_12,UChi_12)			\
+	 VMADD_MII_IR(U0,Chi_01,UChi_00,UChi_00)		\
+	 VMADD_MII_IR(U1,Chi_01,UChi_01,UChi_01)		\
+	 VMADD_MII_IR(U2,Chi_01,UChi_02,UChi_02)		\
+	 VMADD_MII_IR(U0,Chi_11,UChi_10,UChi_10)		\
+	 VMADD_MII_IR(U1,Chi_11,UChi_11,UChi_11)		\
+	 VMADD_MII_IR(U2,Chi_11,UChi_12,UChi_12)		\
+	 : : "r" (32), "r" (32*4), "r" (32*7), "r" (ub ));	\
+    asm (							\
+         VLOAD(%0,%3,U0)					\
+         VLOAD(%1,%3,U1)					\
+         VLOAD(%2,%3,U2)					\
+	 VMADD_RR_RI(U0,Chi_02,UChi_00,UChi_00)			\
+	 VMADD_RR_RI(U1,Chi_02,UChi_01,UChi_01)			\
+	 VMADD_RR_RI(U2,Chi_02,UChi_02,UChi_02)			\
+	 VMADD_RR_RI(U0,Chi_12,UChi_10,UChi_10)			\
+	 VMADD_RR_RI(U1,Chi_12,UChi_11,UChi_11)			\
+	 VMADD_RR_RI(U2,Chi_12,UChi_12,UChi_12)			\
+	 VMADD_MII_IR(U0,Chi_02,UChi_00,UChi_00)		\
+	 VMADD_MII_IR(U1,Chi_02,UChi_01,UChi_01)		\
+	 VMADD_MII_IR(U2,Chi_02,UChi_02,UChi_02)		\
+	 VMADD_MII_IR(U0,Chi_12,UChi_10,UChi_10)		\
+	 VMADD_MII_IR(U1,Chi_12,UChi_11,UChi_11)		\
+	 VMADD_MII_IR(U2,Chi_12,UChi_12,UChi_12)		\
+	 : : "r" (32*2), "r" (32*5), "r" (32*8), "r" (ub ));	\
   }
 
-#define SAVE_RESULT(base,basep) {\
-    uint64_t ub = ((uint64_t)base)  - 32;				\
-    asm("mr %0,"REP";\n\t"						\
-	"li " IMM ",32;\n\t"						\
-	VSTORE(IMM,REP,psi_00)						\
-	VSTORE(IMM,REP,psi_01)						\
-	VSTORE(IMM,REP,psi_02)						\
-	VSTORE(IMM,REP,psi_10)						\
-	VSTORE(IMM,REP,psi_11)						\
-	VSTORE(IMM,REP,psi_12)						\
-	VSTORE(IMM,REP,psi_20)						\
-	VSTORE(IMM,REP,psi_21)						\
-	VSTORE(IMM,REP,psi_22)						\
-	VSTORE(IMM,REP,psi_30)						\
-	VSTORE(IMM,REP,psi_31)						\
-	VSTORE(IMM,REP,psi_32)						\
-	);								\
-}
+#define SAVE_RESULT(base,basep) {		\
+    uint64_t ub = ((uint64_t)base)  - 32;	\
+    asm("mr %0,"REP";\n\t"			\
+	"li " IMM ",32;\n\t"			\
+	VSTORE(IMM,REP,psi_00)			\
+	VSTORE(IMM,REP,psi_01)			\
+	VSTORE(IMM,REP,psi_02)			\
+	VSTORE(IMM,REP,psi_10)			\
+	VSTORE(IMM,REP,psi_11)			\
+	VSTORE(IMM,REP,psi_12)			\
+	VSTORE(IMM,REP,psi_20)			\
+	VSTORE(IMM,REP,psi_21)			\
+	VSTORE(IMM,REP,psi_22)			\
+	VSTORE(IMM,REP,psi_30)			\
+	VSTORE(IMM,REP,psi_31)			\
+	VSTORE(IMM,REP,psi_32)			\
+	);					\
+  }
 
 /*
  *Annoying BG/Q loads with no immediat indexing and big performance hit
  *when second miss to a L1 line occurs
  */
-#define LOAD_CHI(base) {						\
-    uint64_t ub = ((uint64_t)base)  - 64;				\
-    asm("mr  %0,"REP";\n\t"						\
-	"li " IMM ",64;\n\t"				    		\
-	VLOAD(IMM,REP,Chi_00)						\
-	VLOAD(IMM,REP,Chi_02)						\
-	VLOAD(IMM,REP,Chi_11) : : "r" (ub) 	     );			\
-    ub = ((uint64_t)base)  - 32;					\
-    asm("mr  %0,"REP";\n\t"						\
-	"li IMM,64;\n\t"				    		\
-	VLOAD(IMM,REP,Chimu_01)						\
-	VLOAD(IMM,REP,Chimu_10)						\
-	VLOAD(IMM,REP,Chimu_12)	: : "r" (ub) 	     );			\
+#define LOAD_CHI(base) {				\
+    uint64_t ub = ((uint64_t)base)  - 64;		\
+    asm("mr  %0,"REP";\n\t"				\
+	"li " IMM ",64;\n\t"				\
+	VLOAD(IMM,REP,Chi_00)				\
+	VLOAD(IMM,REP,Chi_02)				\
+	VLOAD(IMM,REP,Chi_11) : : "r" (ub) 	     );	\
+    ub = ((uint64_t)base)  - 32;			\
+    asm("mr  %0,"REP";\n\t"				\
+	"li IMM,64;\n\t"				\
+	VLOAD(IMM,REP,Chimu_01)				\
+	VLOAD(IMM,REP,Chimu_10)				\
+	VLOAD(IMM,REP,Chimu_12)	: : "r" (ub) 	     );	\
   }
 
-#define LOAD_CHIMU(base) {						\
-    uint64_t ub = ((uint64_t)base)  - 64;				\
-    asm("mr  %0,"REP";\n\t"						\
-	"li IMM,64;\n\t"				    		\
-	VLOAD(IMM,REP,Chimu_00)						\
-	VLOAD(IMM,REP,Chimu_02)						\
-	VLOAD(IMM,REP,Chimu_11)						\
-	VLOAD(IMM,REP,Chimu_20)						\
-	VLOAD(IMM,REP,Chimu_22)						\
-	VLOAD(IMM,REP,Chimu_31) : : "r" (ub) 	     );			\
-    ub = ((uint64_t)base)  - 32;					\
-    asm("mr  %0,"REP";\n\t"						\
-	"li IMM,64;\n\t"				    		\
-	VLOAD(IMM,REP,Chimu_01)						\
-	VLOAD(IMM,REP,Chimu_10)						\
-	VLOAD(IMM,REP,Chimu_12)						\
-	VLOAD(IMM,REP,Chimu_21)						\
-	VLOAD(IMM,REP,Chimu_30)						\
-	VLOAD(IMM,REP,Chimu_32)	: : "r" (ub) 	     );			\
+#define LOAD_CHIMU(base) {				\
+    uint64_t ub = ((uint64_t)base)  - 64;		\
+    asm("mr  %0,"REP";\n\t"				\
+	"li IMM,64;\n\t"				\
+	VLOAD(IMM,REP,Chimu_00)				\
+	VLOAD(IMM,REP,Chimu_02)				\
+	VLOAD(IMM,REP,Chimu_11)				\
+	VLOAD(IMM,REP,Chimu_20)				\
+	VLOAD(IMM,REP,Chimu_22)				\
+	VLOAD(IMM,REP,Chimu_31) : : "r" (ub) 	     );	\
+    ub = ((uint64_t)base)  - 32;			\
+    asm("mr  %0,"REP";\n\t"				\
+	"li IMM,64;\n\t"				\
+	VLOAD(IMM,REP,Chimu_01)				\
+	VLOAD(IMM,REP,Chimu_10)				\
+	VLOAD(IMM,REP,Chimu_12)				\
+	VLOAD(IMM,REP,Chimu_21)				\
+	VLOAD(IMM,REP,Chimu_30)				\
+	VLOAD(IMM,REP,Chimu_32)	: : "r" (ub) 	     );	\
   }
 
 //      hspin(0)=fspin(0)+timesI(fspin(3));
 //      hspin(1)=fspin(1)+timesI(fspin(2));
-#define XP_PROJMEM(base) {					\
+#define XP_PROJMEM(base) {						\
+    LOAD_CHIMU(base);							\
+    asm (								\
+         VONE(one)							\
+	 VMADD_MII_IR(one,Chimu_30,Chimu_00,Chi_00)			\
+	 VMADD_MII_IR(one,Chimu_31,Chimu_01,Chi_01)			\
+	 VMADD_MII_IR(one,Chimu_32,Chimu_02,Chi_02)			\
+	 VMADD_MII_IR(one,Chimu_20,Chimu_10,Chi_10)			\
+	 VMADD_MII_IR(one,Chimu_21,Chimu_11,Chi_11)			\
+	 VMADD_MII_IR(one,Chimu_22,Chimu_12,Chi_12)			\
+								);	\
+  }
+
+#define XM_PROJMEM(base) {					\
     LOAD_CHIMU(base);						\
     asm (							\
          VONE(one)						\
-	 VMADD_MII_IR(one,Chimu_30,Chimu_00,Chi_00)		\
-	 VMADD_MII_IR(one,Chimu_31,Chimu_01,Chi_01)		\
-	 VMADD_MII_IR(one,Chimu_32,Chimu_02,Chi_02)		\
-	 VMADD_MII_IR(one,Chimu_20,Chimu_10,Chi_10)		\
-	 VMADD_MII_IR(one,Chimu_21,Chimu_11,Chi_11)		\
-	 VMADD_MII_IR(one,Chimu_22,Chimu_12,Chi_12)		\
-							);	\
-  }
-
-#define XM_PROJMEM(base) {				\
-    LOAD_CHIMU(base);					\
-    asm (						\
-         VONE(one)						\
-	 VMADD_II_MIR(one,Chimu_30,Chimu_00,Chi_00)	\
-	 VMADD_II_MIR(one,Chimu_31,Chimu_01,Chi_01)	\
-	 VMADD_II_MIR(one,Chimu_32,Chimu_02,Chi_02)	\
-	 VMADD_II_MIR(one,Chimu_20,Chimu_10,Chi_10)	\
-	 VMADD_II_MIR(one,Chimu_21,Chimu_11,Chi_11)	\
-	 VMADD_II_MIR(one,Chimu_22,Chimu_12,Chi_12)	\
+	 VMADD_II_MIR(one,Chimu_30,Chimu_00,Chi_00)		\
+	 VMADD_II_MIR(one,Chimu_31,Chimu_01,Chi_01)		\
+	 VMADD_II_MIR(one,Chimu_32,Chimu_02,Chi_02)		\
+	 VMADD_II_MIR(one,Chimu_20,Chimu_10,Chi_10)		\
+	 VMADD_II_MIR(one,Chimu_21,Chimu_11,Chi_11)		\
+	 VMADD_II_MIR(one,Chimu_22,Chimu_12,Chi_12)		\
 							);	\
   }
 
 //      hspin(0)=fspin(0)-fspin(3);
 //      hspin(1)=fspin(1)+fspin(2);
-#define YP_PROJMEM(base) {  \
-    LOAD_CHIMU(base);						\
-    asm (							\
-	 VSUB(Chimu_00,Chimu_00,Chi_30)		\
-	 VSUB(Chimu_01,Chimu_01,Chi_31)		\
-	 VSUB(Chimu_02,Chimu_02,Chi_32)		\
-	 VADD(Chimu_10,Chimu_10,Chi_20)		\
-	 VADD(Chimu_11,Chimu_11,Chi_21)		\
-	 VADD(Chimu_12,Chimu_12,Chi_22)		\
-							);	\
+#define YP_PROJMEM(base) {						\
+    LOAD_CHIMU(base);							\
+    asm (								\
+	 VSUB(Chimu_00,Chimu_00,Chi_30)					\
+	 VSUB(Chimu_01,Chimu_01,Chi_31)					\
+	 VSUB(Chimu_02,Chimu_02,Chi_32)					\
+	 VADD(Chimu_10,Chimu_10,Chi_20)					\
+	 VADD(Chimu_11,Chimu_11,Chi_21)					\
+	 VADD(Chimu_12,Chimu_12,Chi_22)					\
+								);	\
   }
 
-#define YM_PROJMEM(base) {			\
-    LOAD_CHIMU(base);						\
-    asm (							\
-	 VADD(Chimu_00,Chimu_00,Chi_30)		\
-	 VADD(Chimu_01,Chimu_01,Chi_31)		\
-	 VADD(Chimu_02,Chimu_02,Chi_32)		\
-	 VSUB(Chimu_10,Chimu_10,Chi_20)		\
-	 VSUB(Chimu_11,Chimu_11,Chi_21)		\
-	 VSUB(Chimu_12,Chimu_12,Chi_22)		\
-							);	\
+#define YM_PROJMEM(base) {						\
+    LOAD_CHIMU(base);							\
+    asm (								\
+	 VADD(Chimu_00,Chimu_00,Chi_30)					\
+	 VADD(Chimu_01,Chimu_01,Chi_31)					\
+	 VADD(Chimu_02,Chimu_02,Chi_32)					\
+	 VSUB(Chimu_10,Chimu_10,Chi_20)					\
+	 VSUB(Chimu_11,Chimu_11,Chi_21)					\
+	 VSUB(Chimu_12,Chimu_12,Chi_22)					\
+								);	\
   }
 
-	    /*Gz
-	     *  0 0  i  0   [0]+-i[2]
-	     *  0 0  0 -i   [1]-+i[3]
-	     * -i 0  0  0
-	     *  0 i  0  0
-	     */
-#define ZP_PROJMEM(base) {  \
-    LOAD_CHIMU(base);						\
-    asm (							\
-         VONE(one)						\
-	 VMADD_MII_IR(one,Chimu_20,Chimu_00,Chi_00)		\
-	 VMADD_MII_IR(one,Chimu_21,Chimu_01,Chi_01)		\
-	 VMADD_MII_IR(one,Chimu_22,Chimu_02,Chi_02)		\
-	 VMADD_II_MIR(one,Chimu_30,Chimu_10,Chi_10)		\
-	 VMADD_II_MIR(one,Chimu_31,Chimu_11,Chi_11)		\
-	 VMADD_II_MIR(one,Chimu_32,Chimu_12,Chi_12)		\
-							);	\
+/*Gz
+ *  0 0  i  0   [0]+-i[2]
+ *  0 0  0 -i   [1]-+i[3]
+ * -i 0  0  0
+ *  0 i  0  0
+ */
+#define ZP_PROJMEM(base) {						\
+    LOAD_CHIMU(base);							\
+    asm (								\
+         VONE(one)							\
+	 VMADD_MII_IR(one,Chimu_20,Chimu_00,Chi_00)			\
+	 VMADD_MII_IR(one,Chimu_21,Chimu_01,Chi_01)			\
+	 VMADD_MII_IR(one,Chimu_22,Chimu_02,Chi_02)			\
+	 VMADD_II_MIR(one,Chimu_30,Chimu_10,Chi_10)			\
+	 VMADD_II_MIR(one,Chimu_31,Chimu_11,Chi_11)			\
+	 VMADD_II_MIR(one,Chimu_32,Chimu_12,Chi_12)			\
+								);	\
   }
 
-#define ZM_PROJMEM(base) {  \
-    LOAD_CHIMU(base);						\
-    asm (							\
-         VONE(one)						\
-	 VMADD_II_MIR(one,Chimu_20,Chimu_00,Chi_00)		\
-	 VMADD_II_MIR(one,Chimu_21,Chimu_01,Chi_01)		\
-	 VMADD_II_MIR(one,Chimu_22,Chimu_02,Chi_02)		\
-	 VMADD_MII_IR(one,Chimu_30,Chimu_10,Chi_10)		\
-	 VMADD_MII_IR(one,Chimu_31,Chimu_11,Chi_11)		\
-	 VMADD_MII_IR(one,Chimu_32,Chimu_12,Chi_12)		\
-							);	\
+#define ZM_PROJMEM(base) {						\
+    LOAD_CHIMU(base);							\
+    asm (								\
+         VONE(one)							\
+	 VMADD_II_MIR(one,Chimu_20,Chimu_00,Chi_00)			\
+	 VMADD_II_MIR(one,Chimu_21,Chimu_01,Chi_01)			\
+	 VMADD_II_MIR(one,Chimu_22,Chimu_02,Chi_02)			\
+	 VMADD_MII_IR(one,Chimu_30,Chimu_10,Chi_10)			\
+	 VMADD_MII_IR(one,Chimu_31,Chimu_11,Chi_11)			\
+	 VMADD_MII_IR(one,Chimu_32,Chimu_12,Chi_12)			\
+								);	\
   }
-	    /*Gt
-	     *  0 0  1  0 [0]+-[2]
-	     *  0 0  0  1 [1]+-[3]
-	     *  1 0  0  0
-	     *  0 1  0  0
-	     */
-#define TP_PROJMEM(base) {  \
-    LOAD_CHIMU(base);						\
-    asm (							\
-	 VADD(Chimu_00,Chimu_00,Chi_20)		\
-	 VADD(Chimu_01,Chimu_01,Chi_21)		\
-	 VADD(Chimu_02,Chimu_02,Chi_22)		\
-	 VADD(Chimu_10,Chimu_10,Chi_30)		\
-	 VADD(Chimu_11,Chimu_11,Chi_31)		\
-	 VADD(Chimu_12,Chimu_12,Chi_32)		\
-							);	\
+/*Gt
+ *  0 0  1  0 [0]+-[2]
+ *  0 0  0  1 [1]+-[3]
+ *  1 0  0  0
+ *  0 1  0  0
+ */
+#define TP_PROJMEM(base) {						\
+    LOAD_CHIMU(base);							\
+    asm (								\
+	 VADD(Chimu_00,Chimu_00,Chi_20)					\
+	 VADD(Chimu_01,Chimu_01,Chi_21)					\
+	 VADD(Chimu_02,Chimu_02,Chi_22)					\
+	 VADD(Chimu_10,Chimu_10,Chi_30)					\
+	 VADD(Chimu_11,Chimu_11,Chi_31)					\
+	 VADD(Chimu_12,Chimu_12,Chi_32)					\
+								);	\
   }
 
-#define TM_PROJMEM(base) {  \
-    LOAD_CHIMU(base);						\
-    asm (							\
-	 VSUB(Chimu_00,Chimu_00,Chi_20)		\
-	 VSUB(Chimu_01,Chimu_01,Chi_21)		\
-	 VSUB(Chimu_02,Chimu_02,Chi_22)		\
-	 VSUB(Chimu_10,Chimu_10,Chi_30)		\
-	 VSUB(Chimu_11,Chimu_11,Chi_31)		\
-	 VSUB(Chimu_12,Chimu_12,Chi_32)		\
-							);	\
+#define TM_PROJMEM(base) {						\
+    LOAD_CHIMU(base);							\
+    asm (								\
+	 VSUB(Chimu_00,Chimu_00,Chi_20)					\
+	 VSUB(Chimu_01,Chimu_01,Chi_21)					\
+	 VSUB(Chimu_02,Chimu_02,Chi_22)					\
+	 VSUB(Chimu_10,Chimu_10,Chi_30)					\
+	 VSUB(Chimu_11,Chimu_11,Chi_31)					\
+	 VSUB(Chimu_12,Chimu_12,Chi_32)					\
+								);	\
   }
 
 /*
-      fspin(0)=hspin(0);
-      fspin(1)=hspin(1);
-      fspin(2)=timesMinusI(hspin(1));
-      fspin(3)=timesMinusI(hspin(0));
+  fspin(0)=hspin(0);
+  fspin(1)=hspin(1);
+  fspin(2)=timesMinusI(hspin(1));
+  fspin(3)=timesMinusI(hspin(0));
 
-      fspin(0)+=hspin(0);
-      fspin(1)+=hspin(1);
-      fspin(2)-=timesI(hspin(1));
-      fspin(3)-=timesI(hspin(0));
- */
-#define XP_RECON {				\
-    asm(\
-	VONE(one)\
-	VMOV(psi_00,UChi_00) 	VMOV(psi_01,UChi_01)	VMOV(psi_02,UChi_02)\
-	VMOV(psi_10,UChi_10) 	VMOV(psi_11,UChi_11)	VMOV(psi_12,UChi_12)\
-	VZERO(psi_20)	VZERO(psi_21)	VZERO(psi_22) \
-	VZERO(psi_30) 	VZERO(psi_31)   VZERO(psi_32) \
-	VMADD_II_MIR(one,UChi_10,psi_20,psi_20)	      \
-	VMADD_II_MIR(one,UChi_11,psi_21,psi_21)	      \
-	VMADD_II_MIR(one,UChi_12,psi_22,psi_22)	      \
-	VMADD_II_MIR(one,UChi_00,psi_30,psi_30)	      \
-	VMADD_II_MIR(one,UChi_01,psi_31,psi_31)	      \
-	VMADD_II_MIR(one,UChi_02,psi_32,psi_32)	      \
-	);		     \
+  fspin(0)+=hspin(0);
+  fspin(1)+=hspin(1);
+  fspin(2)-=timesI(hspin(1));
+  fspin(3)-=timesI(hspin(0));
+*/
+#define XP_RECON {							\
+    asm(								\
+	VONE(one)							\
+	VMOV(psi_00,UChi_00) 	VMOV(psi_01,UChi_01)	VMOV(psi_02,UChi_02) \
+	VMOV(psi_10,UChi_10) 	VMOV(psi_11,UChi_11)	VMOV(psi_12,UChi_12) \
+	VZERO(psi_20)	VZERO(psi_21)	VZERO(psi_22)			\
+	VZERO(psi_30) 	VZERO(psi_31)   VZERO(psi_32)			\
+	VMADD_II_MIR(one,UChi_10,psi_20,psi_20)				\
+	VMADD_II_MIR(one,UChi_11,psi_21,psi_21)				\
+	VMADD_II_MIR(one,UChi_12,psi_22,psi_22)				\
+	VMADD_II_MIR(one,UChi_00,psi_30,psi_30)				\
+	VMADD_II_MIR(one,UChi_01,psi_31,psi_31)				\
+	VMADD_II_MIR(one,UChi_02,psi_32,psi_32)				\
+	);								\
   }
 
-#define XM_RECON {				\
-    asm(\
-	VONE(one)\
-	VMOV(psi_00,UChi_00) 	VMOV(psi_01,UChi_01)	VMOV(psi_02,UChi_02)\
-	VMOV(psi_10,UChi_10) 	VMOV(psi_11,UChi_11)	VMOV(psi_12,UChi_12)\
-	VZERO(psi_20)	VZERO(psi_21)	VZERO(psi_22) \
-	VZERO(psi_30) 	VZERO(psi_31)   VZERO(psi_32) \
-	VMADD_MII_IR(one,UChi_10,psi_20,psi_20)	      \
-	VMADD_MII_IR(one,UChi_11,psi_21,psi_21)	      \
-	VMADD_MII_IR(one,UChi_12,psi_22,psi_22)	      \
-	VMADD_MII_IR(one,UChi_00,psi_30,psi_30)	      \
-	VMADD_MII_IR(one,UChi_01,psi_31,psi_31)	      \
-	VMADD_MII_IR(one,UChi_02,psi_32,psi_32)	      \
-	);		     \
+#define XM_RECON {							\
+    asm(								\
+	VONE(one)							\
+	VMOV(psi_00,UChi_00) 	VMOV(psi_01,UChi_01)	VMOV(psi_02,UChi_02) \
+	VMOV(psi_10,UChi_10) 	VMOV(psi_11,UChi_11)	VMOV(psi_12,UChi_12) \
+	VZERO(psi_20)	VZERO(psi_21)	VZERO(psi_22)			\
+	VZERO(psi_30) 	VZERO(psi_31)   VZERO(psi_32)			\
+	VMADD_MII_IR(one,UChi_10,psi_20,psi_20)				\
+	VMADD_MII_IR(one,UChi_11,psi_21,psi_21)				\
+	VMADD_MII_IR(one,UChi_12,psi_22,psi_22)				\
+	VMADD_MII_IR(one,UChi_00,psi_30,psi_30)				\
+	VMADD_MII_IR(one,UChi_01,psi_31,psi_31)				\
+	VMADD_MII_IR(one,UChi_02,psi_32,psi_32)				\
+	);								\
   }
 
-#define XP_RECON_ACCUM {				\
-    asm(\
-	VONE(one)\
+#define XP_RECON_ACCUM {						\
+    asm(								\
+	VONE(one)							\
 	VADD(psi_00,UChi_00,psi_00) 	VADD(psi_01,UChi_01,psi_01)	VADD(psi_02,UChi_02,psi_02) \
 	VADD(psi_10,UChi_10,psi_10) 	VADD(psi_11,UChi_11,psi_11)	VADD(psi_12,UChi_12,psi_12) \
-	VMADD_II_MIR(one,UChi_10,psi_20,psi_20)	      \
-	VMADD_II_MIR(one,UChi_11,psi_21,psi_21)	      \
-	VMADD_II_MIR(one,UChi_12,psi_22,psi_22)	      \
-	VMADD_II_MIR(one,UChi_00,psi_30,psi_30)	      \
-	VMADD_II_MIR(one,UChi_01,psi_31,psi_31)	      \
-	VMADD_II_MIR(one,UChi_02,psi_32,psi_32)	      \
-	);		     \
+	VMADD_II_MIR(one,UChi_10,psi_20,psi_20)				\
+	VMADD_II_MIR(one,UChi_11,psi_21,psi_21)				\
+	VMADD_II_MIR(one,UChi_12,psi_22,psi_22)				\
+	VMADD_II_MIR(one,UChi_00,psi_30,psi_30)				\
+	VMADD_II_MIR(one,UChi_01,psi_31,psi_31)				\
+	VMADD_II_MIR(one,UChi_02,psi_32,psi_32)				\
+	);								\
   }
 
-#define XM_RECON_ACCUM {				\
-    asm(\
-	VONE(one)\
+#define XM_RECON_ACCUM {						\
+    asm(								\
+	VONE(one)							\
 	VADD(psi_00,UChi_00,psi_00) 	VADD(psi_01,UChi_01,psi_01)	VADD(psi_02,UChi_02,psi_02) \
 	VADD(psi_10,UChi_10,psi_10) 	VADD(psi_11,UChi_11,psi_11)	VADD(psi_12,UChi_12,psi_12) \
-	VMADD_MII_IR(one,UChi_10,psi_20,psi_20)	      \
-	VMADD_MII_IR(one,UChi_11,psi_21,psi_21)	      \
-	VMADD_MII_IR(one,UChi_12,psi_22,psi_22)	      \
-	VMADD_MII_IR(one,UChi_00,psi_30,psi_30)	      \
-	VMADD_MII_IR(one,UChi_01,psi_31,psi_31)	      \
-	VMADD_MII_IR(one,UChi_02,psi_32,psi_32)	      \
-	);		     \
+	VMADD_MII_IR(one,UChi_10,psi_20,psi_20)				\
+	VMADD_MII_IR(one,UChi_11,psi_21,psi_21)				\
+	VMADD_MII_IR(one,UChi_12,psi_22,psi_22)				\
+	VMADD_MII_IR(one,UChi_00,psi_30,psi_30)				\
+	VMADD_MII_IR(one,UChi_01,psi_31,psi_31)				\
+	VMADD_MII_IR(one,UChi_02,psi_32,psi_32)				\
+	);								\
   }
 
 //      fspin(2)+=hspin(1);
 //      fspin(3)-=hspin(0);
-#define YP_RECON_ACCUM {\
-    asm(\
+#define YP_RECON_ACCUM {						\
+    asm(								\
 	VADD(psi_00,UChi_00,psi_00) 	VADD(psi_01,UChi_01,psi_01)	VADD(psi_02,UChi_02,psi_02) \
 	VADD(psi_10,UChi_10,psi_10) 	VADD(psi_11,UChi_11,psi_11)	VADD(psi_12,UChi_12,psi_12) \
 	VADD(psi_20,UChi_10,psi_20) 	VADD(psi_21,UChi_11,psi_21)	VADD(psi_22,UChi_12,psi_22) \
 	VSUB(psi_30,UChi_00,psi_30) 	VSUB(psi_31,UChi_01,psi_31)	VSUB(psi_32,UChi_02,psi_32) \
-	);\
- }
-#define YM_RECON_ACCUM {\
-    asm(\
+	);								\
+  }
+#define YM_RECON_ACCUM {						\
+    asm(								\
 	VADD(psi_00,UChi_00,psi_00) 	VADD(psi_01,UChi_01,psi_01)	VADD(psi_02,UChi_02,psi_02) \
 	VADD(psi_10,UChi_10,psi_10) 	VADD(psi_11,UChi_11,psi_11)	VADD(psi_12,UChi_12,psi_12) \
 	VSUB(psi_20,UChi_10,psi_20) 	VSUB(psi_21,UChi_11,psi_21)	VSUB(psi_22,UChi_12,psi_22) \
 	VADD(psi_30,UChi_00,psi_30) 	VADD(psi_31,UChi_01,psi_31)	VADD(psi_32,UChi_02,psi_32) \
-	);\
- }
+	);								\
+  }
 
 //      fspin(2)-=timesI(hspin(0));
 //      fspin(3)+=timesI(hspin(1));
-#define ZP_RECON_ACCUM {\
-    asm(\
-	VONE(one)\
+#define ZP_RECON_ACCUM {						\
+    asm(								\
+	VONE(one)							\
 	VADD(psi_00,UChi_00,psi_00) 	VADD(psi_01,UChi_01,psi_01)	VADD(psi_02,UChi_02,psi_02) \
 	VADD(psi_10,UChi_10,psi_10) 	VADD(psi_11,UChi_11,psi_11)	VADD(psi_12,UChi_12,psi_12) \
-	VMADD_II_MIR(one,UChi_00,psi_20,psi_20)	      \
-	VMADD_II_MIR(one,UChi_01,psi_21,psi_21)	      \
-	VMADD_II_MIR(one,UChi_02,psi_22,psi_22)	      \
-	VMADD_MII_IR(one,UChi_10,psi_30,psi_30)	      \
-	VMADD_MII_IR(one,UChi_11,psi_31,psi_31)	      \
-	VMADD_MII_IR(one,UChi_12,psi_32,psi_32)	      \
-	);\
- }
+	VMADD_II_MIR(one,UChi_00,psi_20,psi_20)				\
+	VMADD_II_MIR(one,UChi_01,psi_21,psi_21)				\
+	VMADD_II_MIR(one,UChi_02,psi_22,psi_22)				\
+	VMADD_MII_IR(one,UChi_10,psi_30,psi_30)				\
+	VMADD_MII_IR(one,UChi_11,psi_31,psi_31)				\
+	VMADD_MII_IR(one,UChi_12,psi_32,psi_32)				\
+	);								\
+  }
 
-#define ZM_RECON_ACCUM {\
-    asm(\
-	VONE(one)\
+#define ZM_RECON_ACCUM {						\
+    asm(								\
+	VONE(one)							\
 	VADD(psi_00,UChi_00,psi_00) 	VADD(psi_01,UChi_01,psi_01)	VADD(psi_02,UChi_02,psi_02) \
 	VADD(psi_10,UChi_10,psi_10) 	VADD(psi_11,UChi_11,psi_11)	VADD(psi_12,UChi_12,psi_12) \
-	VMADD_MII_IR(one,UChi_00,psi_20,psi_20)	      \
-	VMADD_MII_IR(one,UChi_01,psi_21,psi_21)	      \
-	VMADD_MII_IR(one,UChi_02,psi_22,psi_22)	      \
-	VMADD_II_MIR(one,UChi_10,psi_30,psi_30)	      \
-	VMADD_II_MIR(one,UChi_11,psi_31,psi_31)	      \
-	VMADD_II_MIR(one,UChi_12,psi_32,psi_32)	      \
-	);\
- }
+	VMADD_MII_IR(one,UChi_00,psi_20,psi_20)				\
+	VMADD_MII_IR(one,UChi_01,psi_21,psi_21)				\
+	VMADD_MII_IR(one,UChi_02,psi_22,psi_22)				\
+	VMADD_II_MIR(one,UChi_10,psi_30,psi_30)				\
+	VMADD_II_MIR(one,UChi_11,psi_31,psi_31)				\
+	VMADD_II_MIR(one,UChi_12,psi_32,psi_32)				\
+	);								\
+  }
 
 //      fspin(2)+=hspin(0);
 //      fspin(3)+=hspin(1);
-#define TP_RECON_ACCUM {\
-    asm(\
+#define TP_RECON_ACCUM {						\
+    asm(								\
 	VADD(psi_00,UChi_00,psi_00) 	VADD(psi_01,UChi_01,psi_01)	VADD(psi_02,UChi_02,psi_02) \
 	VADD(psi_10,UChi_10,psi_10) 	VADD(psi_11,UChi_11,psi_11)	VADD(psi_12,UChi_12,psi_12) \
 	VADD(psi_20,UChi_00,psi_20) 	VADD(psi_21,UChi_01,psi_21)	VADD(psi_22,UChi_02,psi_22) \
 	VADD(psi_30,UChi_10,psi_30) 	VADD(psi_31,UChi_11,psi_31)	VADD(psi_32,UChi_12,psi_32) \
-	);\
- }
+	);								\
+  }
 
-#define TM_RECON_ACCUM {\
-    asm(\
-	VONE(one)\
+#define TM_RECON_ACCUM {						\
+    asm(								\
+	VONE(one)							\
 	VADD(psi_00,UChi_00,psi_00) 	VADD(psi_01,UChi_01,psi_01)	VADD(psi_02,UChi_02,psi_02) \
 	VADD(psi_10,UChi_10,psi_10) 	VADD(psi_11,UChi_11,psi_11)	VADD(psi_12,UChi_12,psi_12) \
 	VSUB(psi_20,UChi_00,psi_20) 	VSUB(psi_21,UChi_01,psi_21)	VSUB(psi_22,UChi_02,psi_22) \
 	VSUB(psi_30,UChi_10,psi_30) 	VSUB(psi_31,UChi_11,psi_31)	VSUB(psi_32,UChi_12,psi_32) \
-	);\
- }
+	);								\
+  }
 
 uint64_t GetPFInfo(int nent,int plocal);
 uint64_t GetInfo(int ptype,int local,int perm,int Xp,int ent,int plocal); 
@@ -540,7 +540,7 @@ void testme(int osites,int ssU)
   int local,perm, ptype;
   uint64_t base;
   uint64_t basep;
-  const uint64_t plocal =(uint64_t) & in._odata[0];
+  const uint64_t plocal =(uint64_t) & in[0];
 
   //  vComplexF isigns[2] = { signs[0], signs[1] };
   //COMPLEX_TYPE is vComplexF of vComplexD depending 
@@ -551,244 +551,244 @@ void testme(int osites,int ssU)
   int nmax=osites;
   for(int site=0;site<Ns;site++) {
     int sU =ssU;
-  int ssn=ssU+1; 
-  if(ssn>=nmax) ssn=0;
-  int sUn=ssn;
-  for(int s=0;s<Ls;s++) {
-  ss =sU*Ls+s;
-  ssn=sUn*Ls+s; 
-  ////////////////////////////////
-  // Xp
-  ////////////////////////////////
-  int  ent=ss*8;// 2*Ndim
-  int nent=ssn*8;
+    int ssn=ssU+1; 
+    if(ssn>=nmax) ssn=0;
+    int sUn=ssn;
+    for(int s=0;s<Ls;s++) {
+      ss =sU*Ls+s;
+      ssn=sUn*Ls+s; 
+      ////////////////////////////////
+      // Xp
+      ////////////////////////////////
+      int  ent=ss*8;// 2*Ndim
+      int nent=ssn*8;
 
-  PF_GAUGE(Xp); 
-  base  = GetInfo(ptype,local,perm,Xp,ent,plocal); ent++;
-  PREFETCH1_CHIMU(base);
+      PF_GAUGE(Xp); 
+      base  = GetInfo(ptype,local,perm,Xp,ent,plocal); ent++;
+      PREFETCH1_CHIMU(base);
 
-  basep = GetPFInfo(nent,plocal); nent++;
-  if ( local ) {
-    LOAD64(%r10,isigns);
+      basep = GetPFInfo(nent,plocal); nent++;
+      if ( local ) {
+	LOAD64(%r10,isigns);
 #ifdef KERNEL_DAG
-    XP_PROJMEM(base);
+	XP_PROJMEM(base);
 #else 
-    XM_PROJMEM(base);
+	XM_PROJMEM(base);
 #endif
-    MAYBEPERM(PERMUTE_DIR3,perm);
-  } else { 
-    LOAD_CHI(base);
-  }
-  base = GetInfo(ptype,local,perm,Yp,ent,plocal); ent++;
-  PREFETCH_CHIMU(base);
-  {
-    MULT_2SPIN_DIR_PFXP(Xp,basep);
-  }
-  LOAD64(%r10,isigns);
+	MAYBEPERM(PERMUTE_DIR3,perm);
+      } else { 
+	LOAD_CHI(base);
+      }
+      base = GetInfo(ptype,local,perm,Yp,ent,plocal); ent++;
+      PREFETCH_CHIMU(base);
+      {
+	MULT_2SPIN_DIR_PFXP(Xp,basep);
+      }
+      LOAD64(%r10,isigns);
 #ifdef KERNEL_DAG
-  XP_RECON;
+      XP_RECON;
 #else
-  XM_RECON;
+      XM_RECON;
 #endif
-  ////////////////////////////////
-  // Yp
-  ////////////////////////////////
-  basep = GetPFInfo(nent,plocal); nent++;
-  if ( local ) {
-    LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+      ////////////////////////////////
+      // Yp
+      ////////////////////////////////
+      basep = GetPFInfo(nent,plocal); nent++;
+      if ( local ) {
+	LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
 #ifdef KERNEL_DAG
-    YP_PROJMEM(base);
+	YP_PROJMEM(base);
 #else
-    YM_PROJMEM(base);
+	YM_PROJMEM(base);
 #endif
-    MAYBEPERM(PERMUTE_DIR2,perm);
-  } else { 
-    LOAD_CHI(base);
-  }
-  base  = GetInfo(ptype,local,perm,Zp,ent,plocal); ent++;
-  PREFETCH_CHIMU(base);
-  {
-    MULT_2SPIN_DIR_PFYP(Yp,basep);
-  }
-  LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+	MAYBEPERM(PERMUTE_DIR2,perm);
+      } else { 
+	LOAD_CHI(base);
+      }
+      base  = GetInfo(ptype,local,perm,Zp,ent,plocal); ent++;
+      PREFETCH_CHIMU(base);
+      {
+	MULT_2SPIN_DIR_PFYP(Yp,basep);
+      }
+      LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
 #ifdef KERNEL_DAG
-  YP_RECON_ACCUM;
+      YP_RECON_ACCUM;
 #else
-  YM_RECON_ACCUM;
+      YM_RECON_ACCUM;
 #endif
 
-  ////////////////////////////////
-  // Zp
-  ////////////////////////////////
-  basep = GetPFInfo(nent,plocal); nent++;
-  if ( local ) {
-    LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+      ////////////////////////////////
+      // Zp
+      ////////////////////////////////
+      basep = GetPFInfo(nent,plocal); nent++;
+      if ( local ) {
+	LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
 #ifdef KERNEL_DAG
-    ZP_PROJMEM(base);
+	ZP_PROJMEM(base);
 #else
-    ZM_PROJMEM(base);
+	ZM_PROJMEM(base);
 #endif
-    MAYBEPERM(PERMUTE_DIR1,perm);
-  } else { 
-    LOAD_CHI(base);
-  }
-  base  = GetInfo(ptype,local,perm,Tp,ent,plocal); ent++;
-  PREFETCH_CHIMU(base);
-  {
-    MULT_2SPIN_DIR_PFZP(Zp,basep);
-  }
-  LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+	MAYBEPERM(PERMUTE_DIR1,perm);
+      } else { 
+	LOAD_CHI(base);
+      }
+      base  = GetInfo(ptype,local,perm,Tp,ent,plocal); ent++;
+      PREFETCH_CHIMU(base);
+      {
+	MULT_2SPIN_DIR_PFZP(Zp,basep);
+      }
+      LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
 #ifdef KERNEL_DAG
-  ZP_RECON_ACCUM;
+      ZP_RECON_ACCUM;
 #else
-  ZM_RECON_ACCUM;
+      ZM_RECON_ACCUM;
 #endif
 
-  ////////////////////////////////
-  // Tp
-  ////////////////////////////////
-  basep = GetPFInfo(nent,plocal); nent++;
-  if ( local ) {
-    LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+      ////////////////////////////////
+      // Tp
+      ////////////////////////////////
+      basep = GetPFInfo(nent,plocal); nent++;
+      if ( local ) {
+	LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
 #ifdef KERNEL_DAG
-    TP_PROJMEM(base);
+	TP_PROJMEM(base);
 #else
-    TM_PROJMEM(base);
+	TM_PROJMEM(base);
 #endif
-    MAYBEPERM(PERMUTE_DIR0,perm);
-  } else { 
-    LOAD_CHI(base);
-  }
-  base = GetInfo(ptype,local,perm,Xm,ent,plocal); ent++;
-  PREFETCH_CHIMU(base);
-  {
-    MULT_2SPIN_DIR_PFTP(Tp,basep);
-  }
-  LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+	MAYBEPERM(PERMUTE_DIR0,perm);
+      } else { 
+	LOAD_CHI(base);
+      }
+      base = GetInfo(ptype,local,perm,Xm,ent,plocal); ent++;
+      PREFETCH_CHIMU(base);
+      {
+	MULT_2SPIN_DIR_PFTP(Tp,basep);
+      }
+      LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
 #ifdef KERNEL_DAG
-  TP_RECON_ACCUM;
+      TP_RECON_ACCUM;
 #else
-  TM_RECON_ACCUM;
+      TM_RECON_ACCUM;
 #endif
 
-  ////////////////////////////////
-  // Xm
-  ////////////////////////////////
+      ////////////////////////////////
+      // Xm
+      ////////////////////////////////
 #ifndef STREAM_STORE
-  basep= (uint64_t) &out._odata[ss];
+      basep= (uint64_t) &out[ss];
 #endif
-  //  basep= GetPFInfo(nent,plocal); nent++;
-  if ( local ) {
-    LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+      //  basep= GetPFInfo(nent,plocal); nent++;
+      if ( local ) {
+	LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
 #ifdef KERNEL_DAG
-    XM_PROJMEM(base);
+	XM_PROJMEM(base);
 #else
-    XP_PROJMEM(base);
+	XP_PROJMEM(base);
 #endif
-    MAYBEPERM(PERMUTE_DIR3,perm);
-  } else { 
-    LOAD_CHI(base);
-  }
-  base = GetInfo(ptype,local,perm,Ym,ent,plocal); ent++;
-  PREFETCH_CHIMU(base);
-  {
-    MULT_2SPIN_DIR_PFXM(Xm,basep);
-  }
-  LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+	MAYBEPERM(PERMUTE_DIR3,perm);
+      } else { 
+	LOAD_CHI(base);
+      }
+      base = GetInfo(ptype,local,perm,Ym,ent,plocal); ent++;
+      PREFETCH_CHIMU(base);
+      {
+	MULT_2SPIN_DIR_PFXM(Xm,basep);
+      }
+      LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
 #ifdef KERNEL_DAG
-  XM_RECON_ACCUM;
+      XM_RECON_ACCUM;
 #else
-  XP_RECON_ACCUM;
+      XP_RECON_ACCUM;
 #endif
 
-  ////////////////////////////////
-  // Ym
-  ////////////////////////////////
-  basep= GetPFInfo(nent,plocal); nent++;
-  if ( local ) {
-    LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+      ////////////////////////////////
+      // Ym
+      ////////////////////////////////
+      basep= GetPFInfo(nent,plocal); nent++;
+      if ( local ) {
+	LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
 #ifdef KERNEL_DAG
-    YM_PROJMEM(base);
+	YM_PROJMEM(base);
 #else
-    YP_PROJMEM(base);
+	YP_PROJMEM(base);
 #endif
-    MAYBEPERM(PERMUTE_DIR2,perm);
-  } else { 
-    LOAD_CHI(base);
-  }
-  base = GetInfo(ptype,local,perm,Zm,ent,plocal); ent++;
-  PREFETCH_CHIMU(base);
-  {
-    MULT_2SPIN_DIR_PFYM(Ym,basep);
-  }
-  LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+	MAYBEPERM(PERMUTE_DIR2,perm);
+      } else { 
+	LOAD_CHI(base);
+      }
+      base = GetInfo(ptype,local,perm,Zm,ent,plocal); ent++;
+      PREFETCH_CHIMU(base);
+      {
+	MULT_2SPIN_DIR_PFYM(Ym,basep);
+      }
+      LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
 #ifdef KERNEL_DAG
-  YM_RECON_ACCUM;
+      YM_RECON_ACCUM;
 #else
-  YP_RECON_ACCUM;
+      YP_RECON_ACCUM;
 #endif
 
-  ////////////////////////////////
-  // Zm
-  ////////////////////////////////
-  basep= GetPFInfo(nent,plocal); nent++;
-  if ( local ) {
-    LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+      ////////////////////////////////
+      // Zm
+      ////////////////////////////////
+      basep= GetPFInfo(nent,plocal); nent++;
+      if ( local ) {
+	LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
 #ifdef KERNEL_DAG
-    ZM_PROJMEM(base);
+	ZM_PROJMEM(base);
 #else
-    ZP_PROJMEM(base);
+	ZP_PROJMEM(base);
 #endif
-    MAYBEPERM(PERMUTE_DIR1,perm);
-  } else { 
-    LOAD_CHI(base);
-  }
-  base = GetInfo(ptype,local,perm,Tm,ent,plocal); ent++;
-  PREFETCH_CHIMU(base);
-  {
-    MULT_2SPIN_DIR_PFZM(Zm,basep);
-  }
-  LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+	MAYBEPERM(PERMUTE_DIR1,perm);
+      } else { 
+	LOAD_CHI(base);
+      }
+      base = GetInfo(ptype,local,perm,Tm,ent,plocal); ent++;
+      PREFETCH_CHIMU(base);
+      {
+	MULT_2SPIN_DIR_PFZM(Zm,basep);
+      }
+      LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
 #ifdef KERNEL_DAG
-  ZM_RECON_ACCUM;
+      ZM_RECON_ACCUM;
 #else
-  ZP_RECON_ACCUM;
+      ZP_RECON_ACCUM;
 #endif
 
-  ////////////////////////////////
-  // Tm
-  ////////////////////////////////
-  basep= GetPFInfo(nent,plocal); nent++;
-  if ( local ) {
-    LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+      ////////////////////////////////
+      // Tm
+      ////////////////////////////////
+      basep= GetPFInfo(nent,plocal); nent++;
+      if ( local ) {
+	LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
 #ifdef KERNEL_DAG
-    TM_PROJMEM(base);
+	TM_PROJMEM(base);
 #else
-    TP_PROJMEM(base);
+	TP_PROJMEM(base);
 #endif
-    MAYBEPERM(PERMUTE_DIR0,perm);
-  } else { 
-    LOAD_CHI(base);
-  }
-  base= (uint64_t) &out._odata[ss];
+	MAYBEPERM(PERMUTE_DIR0,perm);
+      } else { 
+	LOAD_CHI(base);
+      }
+      base= (uint64_t) &out[ss];
 #ifndef STREAM_STORE
-  PREFETCH_CHIMU(base);
+      PREFETCH_CHIMU(base);
 #endif
-  {
-    MULT_2SPIN_DIR_PFTM(Tm,basep);
-  }
-  LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+      {
+	MULT_2SPIN_DIR_PFTM(Tm,basep);
+      }
+      LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
 #ifdef KERNEL_DAG
-  TM_RECON_ACCUM;
+      TM_RECON_ACCUM;
 #else
-  TP_RECON_ACCUM;
+      TP_RECON_ACCUM;
 #endif
 
-  basep= GetPFInfo(nent,plocal); nent++;
-  SAVE_RESULT(base,basep);
+      basep= GetPFInfo(nent,plocal); nent++;
+      SAVE_RESULT(base,basep);
   
-  }
-  ssU++;
+    }
+    ssU++;
   }
 }
 
diff --git a/Grid/simd/Grid_avx.h b/Grid/simd/Grid_avx.h
index f4634432..ad9800fb 100644
--- a/Grid/simd/Grid_avx.h
+++ b/Grid/simd/Grid_avx.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid
 
@@ -27,8 +27,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #include <immintrin.h>
 #ifdef AVXFMA4
 #include <x86intrin.h>
@@ -38,527 +38,527 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define _mm256_set_m128i(hi,lo) _mm256_insertf128_si256(_mm256_castsi128_si256(lo),(hi),1)
 #endif
 
-namespace Grid {
-namespace Optimization {
+NAMESPACE_BEGIN(Grid);
+NAMESPACE_BEGIN(Optimization);
 
-  template<class vtype>
-  union uconv {
-    __m256 f;
-    vtype v;
-  };
+template<class vtype>
+union uconv {
+  __m256 f;
+  vtype v;
+};
 
-  union u256f {
-    __m256 v;
-    float f[8];
-  };
+union u256f {
+  __m256 v;
+  float f[8];
+};
 
-  union u256d {
-    __m256d v;
-    double f[4];
-  };
+union u256d {
+  __m256d v;
+  double f[4];
+};
 
- struct Vsplat{
-    // Complex float
-    inline __m256 operator()(float a, float b) {
-      return _mm256_set_ps(b,a,b,a,b,a,b,a);
-    }
-    // Real float
-    inline __m256 operator()(float a){
-      return _mm256_set_ps(a,a,a,a,a,a,a,a);
-    }
-    //Complex double
-    inline __m256d operator()(double a, double b){
-      return _mm256_set_pd(b,a,b,a);
-    }
-    //Real double
-    inline __m256d operator()(double a){
-      return _mm256_set_pd(a,a,a,a);
-    }
-    //Integer
-    inline __m256i operator()(Integer a){
-      return _mm256_set1_epi32(a);
-    }
-  };
+struct Vsplat{
+  // Complex float
+  inline __m256 operator()(float a, float b) {
+    return _mm256_set_ps(b,a,b,a,b,a,b,a);
+  }
+  // Real float
+  inline __m256 operator()(float a){
+    return _mm256_set_ps(a,a,a,a,a,a,a,a);
+  }
+  //Complex double
+  inline __m256d operator()(double a, double b){
+    return _mm256_set_pd(b,a,b,a);
+  }
+  //Real double
+  inline __m256d operator()(double a){
+    return _mm256_set_pd(a,a,a,a);
+  }
+  //Integer
+  inline __m256i operator()(Integer a){
+    return _mm256_set1_epi32(a);
+  }
+};
 
-  struct Vstore{
-    //Float
-    inline void operator()(__m256 a, float* F){
-      _mm256_store_ps(F,a);
-    }
-    //Double
-    inline void operator()(__m256d a, double* D){
-      _mm256_store_pd(D,a);
-    }
-    //Integer
-    inline void operator()(__m256i a, Integer* I){
-      _mm256_store_si256((__m256i*)I,a);
-    }
+struct Vstore{
+  //Float
+  inline void operator()(__m256 a, float* F){
+    _mm256_store_ps(F,a);
+  }
+  //Double
+  inline void operator()(__m256d a, double* D){
+    _mm256_store_pd(D,a);
+  }
+  //Integer
+  inline void operator()(__m256i a, Integer* I){
+    _mm256_store_si256((__m256i*)I,a);
+  }
 
-  };
+};
 
-  struct Vstream{
-    //Float
-    inline void operator()(float * a, __m256 b){
-      _mm256_stream_ps(a,b);
-    }
-    //Double
-    inline void operator()(double * a, __m256d b){
-      _mm256_stream_pd(a,b);
-    }
+struct Vstream{
+  //Float
+  inline void operator()(float * a, __m256 b){
+    _mm256_stream_ps(a,b);
+  }
+  //Double
+  inline void operator()(double * a, __m256d b){
+    _mm256_stream_pd(a,b);
+  }
 
 
-  };
+};
 
-  struct Vset{
-    // Complex float
-    inline __m256 operator()(Grid::ComplexF *a){
-      return _mm256_set_ps(a[3].imag(),a[3].real(),a[2].imag(),a[2].real(),a[1].imag(),a[1].real(),a[0].imag(),a[0].real());
-    }
-    // Complex double
-    inline __m256d operator()(Grid::ComplexD *a){
-      return _mm256_set_pd(a[1].imag(),a[1].real(),a[0].imag(),a[0].real());
-    }
-    // Real float
-    inline __m256 operator()(float *a){
-      return _mm256_set_ps(a[7],a[6],a[5],a[4],a[3],a[2],a[1],a[0]);
-    }
-    // Real double
-    inline __m256d operator()(double *a){
-      return _mm256_set_pd(a[3],a[2],a[1],a[0]);
-    }
-    // Integer
-    inline __m256i operator()(Integer *a){
-      return _mm256_set_epi32(a[7],a[6],a[5],a[4],a[3],a[2],a[1],a[0]);
-    }
+struct Vset{
+  // Complex float
+  inline __m256 operator()(Grid::ComplexF *a){
+    return _mm256_set_ps(a[3].imag(),a[3].real(),a[2].imag(),a[2].real(),a[1].imag(),a[1].real(),a[0].imag(),a[0].real());
+  }
+  // Complex double
+  inline __m256d operator()(Grid::ComplexD *a){
+    return _mm256_set_pd(a[1].imag(),a[1].real(),a[0].imag(),a[0].real());
+  }
+  // Real float
+  inline __m256 operator()(float *a){
+    return _mm256_set_ps(a[7],a[6],a[5],a[4],a[3],a[2],a[1],a[0]);
+  }
+  // Real double
+  inline __m256d operator()(double *a){
+    return _mm256_set_pd(a[3],a[2],a[1],a[0]);
+  }
+  // Integer
+  inline __m256i operator()(Integer *a){
+    return _mm256_set_epi32(a[7],a[6],a[5],a[4],a[3],a[2],a[1],a[0]);
+  }
 
-  };
+};
 
-  template <typename Out_type, typename In_type>
-  struct Reduce{
-    // Need templated class to overload output type
-    // General form must generate error if compiled
-    inline Out_type operator()(In_type in){
-      printf("Error, using wrong Reduce function\n");
-      exit(1);
-      return 0;
-    }
-  };
+template <typename Out_type, typename In_type>
+struct Reduce{
+  // Need templated class to overload output type
+  // General form must generate error if compiled
+  inline Out_type operator()(In_type in){
+    printf("Error, using wrong Reduce function\n");
+    exit(1);
+    return 0;
+  }
+};
 
-  /////////////////////////////////////////////////////
-  // Arithmetic operations
-  /////////////////////////////////////////////////////
-  struct Sum{
-    //Complex/Real float
-    inline __m256 operator()(__m256 a, __m256 b){
-      return _mm256_add_ps(a,b);
-    }
-    //Complex/Real double
-    inline __m256d operator()(__m256d a, __m256d b){
-      return _mm256_add_pd(a,b);
-    }
-    //Integer
-    inline __m256i operator()(__m256i a, __m256i b){
+/////////////////////////////////////////////////////
+// Arithmetic operations
+/////////////////////////////////////////////////////
+struct Sum{
+  //Complex/Real float
+  inline __m256 operator()(__m256 a, __m256 b){
+    return _mm256_add_ps(a,b);
+  }
+  //Complex/Real double
+  inline __m256d operator()(__m256d a, __m256d b){
+    return _mm256_add_pd(a,b);
+  }
+  //Integer
+  inline __m256i operator()(__m256i a, __m256i b){
 #if defined (AVX1) || defined (AVXFMA) || defined (AVXFMA4)
-          __m128i a0,a1;
-          __m128i b0,b1;
-          a0 = _mm256_extractf128_si256(a,0);
-          b0 = _mm256_extractf128_si256(b,0);
-          a1 = _mm256_extractf128_si256(a,1);
-          b1 = _mm256_extractf128_si256(b,1);
-          a0 = _mm_add_epi32(a0,b0);
-          a1 = _mm_add_epi32(a1,b1);
-          return _mm256_set_m128i(a1,a0);
+    __m128i a0,a1;
+    __m128i b0,b1;
+    a0 = _mm256_extractf128_si256(a,0);
+    b0 = _mm256_extractf128_si256(b,0);
+    a1 = _mm256_extractf128_si256(a,1);
+    b1 = _mm256_extractf128_si256(b,1);
+    a0 = _mm_add_epi32(a0,b0);
+    a1 = _mm_add_epi32(a1,b1);
+    return _mm256_set_m128i(a1,a0);
 #endif
 #if defined (AVX2)
-            return _mm256_add_epi32(a,b);
+    return _mm256_add_epi32(a,b);
 #endif
-    }
-  };
+  }
+};
 
-  struct Sub{
-    //Complex/Real float
-    inline __m256 operator()(__m256 a, __m256 b){
-      return _mm256_sub_ps(a,b);
-    }
-    //Complex/Real double
-    inline __m256d operator()(__m256d a, __m256d b){
-      return _mm256_sub_pd(a,b);
-    }
-    //Integer
-    inline __m256i operator()(__m256i a, __m256i b){
+struct Sub{
+  //Complex/Real float
+  inline __m256 operator()(__m256 a, __m256 b){
+    return _mm256_sub_ps(a,b);
+  }
+  //Complex/Real double
+  inline __m256d operator()(__m256d a, __m256d b){
+    return _mm256_sub_pd(a,b);
+  }
+  //Integer
+  inline __m256i operator()(__m256i a, __m256i b){
 #if defined (AVX1) || defined (AVXFMA) || defined (AVXFMA4)
-          __m128i a0,a1;
-          __m128i b0,b1;
-          a0 = _mm256_extractf128_si256(a,0);
-          b0 = _mm256_extractf128_si256(b,0);
-          a1 = _mm256_extractf128_si256(a,1);
-          b1 = _mm256_extractf128_si256(b,1);
-          a0 = _mm_sub_epi32(a0,b0);
-          a1 = _mm_sub_epi32(a1,b1);
-          return _mm256_set_m128i(a1,a0);
+    __m128i a0,a1;
+    __m128i b0,b1;
+    a0 = _mm256_extractf128_si256(a,0);
+    b0 = _mm256_extractf128_si256(b,0);
+    a1 = _mm256_extractf128_si256(a,1);
+    b1 = _mm256_extractf128_si256(b,1);
+    a0 = _mm_sub_epi32(a0,b0);
+    a1 = _mm_sub_epi32(a1,b1);
+    return _mm256_set_m128i(a1,a0);
 #endif
 #if defined (AVX2)
-            return _mm256_sub_epi32(a,b);
+    return _mm256_sub_epi32(a,b);
 #endif
 
-    }
-  };
+  }
+};
 
-  struct MultRealPart{
-    inline __m256 operator()(__m256 a, __m256 b){
-      __m256 ymm0;
-      ymm0  = _mm256_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ymm0 <- ar ar,
-      return  _mm256_mul_ps(ymm0,b);                       // ymm0 <- ar bi, ar br
-    }
-    inline __m256d operator()(__m256d a, __m256d b){
-      __m256d ymm0;
-      ymm0 = _mm256_shuffle_pd(a,a,0x0); // ymm0 <- ar ar, ar,ar b'00,00
-      return _mm256_mul_pd(ymm0,b);      // ymm0 <- ar bi, ar br
-    }
-  };
-  struct MaddRealPart{
-    inline __m256 operator()(__m256 a, __m256 b, __m256 c){
-      __m256 ymm0 =  _mm256_moveldup_ps(a); // ymm0 <- ar ar,
-      return _mm256_add_ps(_mm256_mul_ps( ymm0, b),c);                         
-    }
-    inline __m256d operator()(__m256d a, __m256d b, __m256d c){
-      __m256d ymm0 = _mm256_shuffle_pd( a, a, 0x0 );
-      return _mm256_add_pd(_mm256_mul_pd( ymm0, b),c);                         
-    }
-  };
+struct MultRealPart{
+  inline __m256 operator()(__m256 a, __m256 b){
+    __m256 ymm0;
+    ymm0  = _mm256_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ymm0 <- ar ar,
+    return  _mm256_mul_ps(ymm0,b);                       // ymm0 <- ar bi, ar br
+  }
+  inline __m256d operator()(__m256d a, __m256d b){
+    __m256d ymm0;
+    ymm0 = _mm256_shuffle_pd(a,a,0x0); // ymm0 <- ar ar, ar,ar b'00,00
+    return _mm256_mul_pd(ymm0,b);      // ymm0 <- ar bi, ar br
+  }
+};
+struct MaddRealPart{
+  inline __m256 operator()(__m256 a, __m256 b, __m256 c){
+    __m256 ymm0 =  _mm256_moveldup_ps(a); // ymm0 <- ar ar,
+    return _mm256_add_ps(_mm256_mul_ps( ymm0, b),c);                         
+  }
+  inline __m256d operator()(__m256d a, __m256d b, __m256d c){
+    __m256d ymm0 = _mm256_shuffle_pd( a, a, 0x0 );
+    return _mm256_add_pd(_mm256_mul_pd( ymm0, b),c);                         
+  }
+};
 
-  struct MultComplex{
-    // Complex float
-    inline __m256 operator()(__m256 a, __m256 b){
+struct MultComplex{
+  // Complex float
+  inline __m256 operator()(__m256 a, __m256 b){
 #if defined (AVX1)
-      __m256 ymm0,ymm1,ymm2;
-      ymm0 = _mm256_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ymm0 <- ar ar,
-      ymm0 = _mm256_mul_ps(ymm0,b);                       // ymm0 <- ar bi, ar br
-      // FIXME AVX2 could MAC
-      ymm1 = _mm256_shuffle_ps(b,b,_MM_SELECT_FOUR_FOUR(2,3,0,1)); // ymm1 <- br,bi
-      ymm2 = _mm256_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(3,3,1,1)); // ymm2 <- ai,ai
-      ymm1 = _mm256_mul_ps(ymm1,ymm2);                    // ymm1 <- br ai, ai bi
-      return _mm256_addsub_ps(ymm0,ymm1);
+    __m256 ymm0,ymm1,ymm2;
+    ymm0 = _mm256_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ymm0 <- ar ar,
+    ymm0 = _mm256_mul_ps(ymm0,b);                       // ymm0 <- ar bi, ar br
+    // FIXME AVX2 could MAC
+    ymm1 = _mm256_shuffle_ps(b,b,_MM_SELECT_FOUR_FOUR(2,3,0,1)); // ymm1 <- br,bi
+    ymm2 = _mm256_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(3,3,1,1)); // ymm2 <- ai,ai
+    ymm1 = _mm256_mul_ps(ymm1,ymm2);                    // ymm1 <- br ai, ai bi
+    return _mm256_addsub_ps(ymm0,ymm1);
 #endif
 #if defined (AVXFMA4)
-      __m256 a_real = _mm256_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ar ar,
-      __m256 a_imag = _mm256_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(3,3,1,1)); // ai ai
-      __m256 tmp = _mm256_shuffle_ps( b,b, _MM_SELECT_FOUR_FOUR(2,3,0,1));
-      a_imag = _mm256_mul_ps( a_imag,tmp  );  // (Ai, Ai) * (Bi, Br) = Ai Bi, Ai Br
-      return _mm256_maddsub_ps( a_real, b, a_imag ); // Ar Br , Ar Bi   +- Ai Bi             = ArBr-AiBi , ArBi+AiBr
+    __m256 a_real = _mm256_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ar ar,
+    __m256 a_imag = _mm256_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(3,3,1,1)); // ai ai
+    __m256 tmp = _mm256_shuffle_ps( b,b, _MM_SELECT_FOUR_FOUR(2,3,0,1));
+    a_imag = _mm256_mul_ps( a_imag,tmp  );  // (Ai, Ai) * (Bi, Br) = Ai Bi, Ai Br
+    return _mm256_maddsub_ps( a_real, b, a_imag ); // Ar Br , Ar Bi   +- Ai Bi             = ArBr-AiBi , ArBi+AiBr
 #endif
 #if defined (AVX2)  || defined (AVXFMA)
-      __m256 a_real = _mm256_moveldup_ps( a ); // Ar Ar
-      __m256 a_imag = _mm256_movehdup_ps( a ); // Ai Ai
-      a_imag = _mm256_mul_ps( a_imag, _mm256_shuffle_ps( b,b, _MM_SELECT_FOUR_FOUR(2,3,0,1) ));  // (Ai, Ai) * (Bi, Br) = Ai Bi, Ai Br
-      return _mm256_fmaddsub_ps( a_real, b, a_imag ); // Ar Br , Ar Bi   +- Ai Bi             = ArBr-AiBi , ArBi+AiBr
+    __m256 a_real = _mm256_moveldup_ps( a ); // Ar Ar
+    __m256 a_imag = _mm256_movehdup_ps( a ); // Ai Ai
+    a_imag = _mm256_mul_ps( a_imag, _mm256_shuffle_ps( b,b, _MM_SELECT_FOUR_FOUR(2,3,0,1) ));  // (Ai, Ai) * (Bi, Br) = Ai Bi, Ai Br
+    return _mm256_fmaddsub_ps( a_real, b, a_imag ); // Ar Br , Ar Bi   +- Ai Bi             = ArBr-AiBi , ArBi+AiBr
 #endif
-    }
-    // Complex double
-    inline __m256d operator()(__m256d a, __m256d b) {
-      // Multiplication of (ak+ibk)*(ck+idk)
-      // a + i b can be stored as a data structure
-      // From intel optimisation reference guide
-      /*
-	movsldup xmm0, Src1; load real parts into the destination,
-	; a1, a1, a0, a0
-	movaps xmm1, src2; load the 2nd pair of complex values, ; i.e. d1, c1, d0, c0
-	mulps xmm0, xmm1; temporary results, a1d1, a1c1, a0d0, ; a0c0
-	shufps xmm1, xmm1, b1; reorder the real and imaginary ; parts, c1, d1, c0, d0
-	movshdup xmm2, Src1; load the imaginary parts into the ; destination, b1, b1, b0, b0
-	mulps xmm2, xmm1; temporary results, b1c1, b1d1, b0c0, ; b0d0
-	addsubps xmm0, xmm2; b1c1+a1d1, a1c1 -b1d1, b0c0+a0d
-	VSHUFPD (VEX.256 encoded version)
-	IF IMM0[0] = 0
-	THEN DEST[63:0]=SRC1[63:0] ELSE DEST[63:0]=SRC1[127:64] FI;
-	IF IMM0[1] = 0
-	THEN DEST[127:64]=SRC2[63:0] ELSE DEST[127:64]=SRC2[127:64] FI;
-	IF IMM0[2] = 0
-	THEN DEST[191:128]=SRC1[191:128] ELSE DEST[191:128]=SRC1[255:192] FI;
-	IF IMM0[3] = 0
-	THEN DEST[255:192]=SRC2[191:128] ELSE DEST[255:192]=SRC2[255:192] FI; // Ox5 r<->i   ; 0xC unchanged
-      */
+  }
+  // Complex double
+  inline __m256d operator()(__m256d a, __m256d b) {
+    // Multiplication of (ak+ibk)*(ck+idk)
+    // a + i b can be stored as a data structure
+    // From intel optimisation reference guide
+    /*
+      movsldup xmm0, Src1; load real parts into the destination,
+      ; a1, a1, a0, a0
+      movaps xmm1, src2; load the 2nd pair of complex values, ; i.e. d1, c1, d0, c0
+      mulps xmm0, xmm1; temporary results, a1d1, a1c1, a0d0, ; a0c0
+      shufps xmm1, xmm1, b1; reorder the real and imaginary ; parts, c1, d1, c0, d0
+      movshdup xmm2, Src1; load the imaginary parts into the ; destination, b1, b1, b0, b0
+      mulps xmm2, xmm1; temporary results, b1c1, b1d1, b0c0, ; b0d0
+      addsubps xmm0, xmm2; b1c1+a1d1, a1c1 -b1d1, b0c0+a0d
+      VSHUFPD (VEX.256 encoded version)
+      IF IMM0[0] = 0
+      THEN DEST[63:0]=SRC1[63:0] ELSE DEST[63:0]=SRC1[127:64] FI;
+      IF IMM0[1] = 0
+      THEN DEST[127:64]=SRC2[63:0] ELSE DEST[127:64]=SRC2[127:64] FI;
+      IF IMM0[2] = 0
+      THEN DEST[191:128]=SRC1[191:128] ELSE DEST[191:128]=SRC1[255:192] FI;
+      IF IMM0[3] = 0
+      THEN DEST[255:192]=SRC2[191:128] ELSE DEST[255:192]=SRC2[255:192] FI; // Ox5 r<->i   ; 0xC unchanged
+    */
 #if defined (AVX1)
-      __m256d ymm0,ymm1,ymm2;
-      ymm0 = _mm256_shuffle_pd(a,a,0x0); // ymm0 <- ar ar, ar,ar b'00,00
-      ymm0 = _mm256_mul_pd(ymm0,b);      // ymm0 <- ar bi, ar br
-      ymm1 = _mm256_shuffle_pd(b,b,0x5); // ymm1 <- br,bi  b'01,01
-      ymm2 = _mm256_shuffle_pd(a,a,0xF); // ymm2 <- ai,ai  b'11,11
-      ymm1 = _mm256_mul_pd(ymm1,ymm2);   // ymm1 <- br ai, ai bi
-      return _mm256_addsub_pd(ymm0,ymm1);
+    __m256d ymm0,ymm1,ymm2;
+    ymm0 = _mm256_shuffle_pd(a,a,0x0); // ymm0 <- ar ar, ar,ar b'00,00
+    ymm0 = _mm256_mul_pd(ymm0,b);      // ymm0 <- ar bi, ar br
+    ymm1 = _mm256_shuffle_pd(b,b,0x5); // ymm1 <- br,bi  b'01,01
+    ymm2 = _mm256_shuffle_pd(a,a,0xF); // ymm2 <- ai,ai  b'11,11
+    ymm1 = _mm256_mul_pd(ymm1,ymm2);   // ymm1 <- br ai, ai bi
+    return _mm256_addsub_pd(ymm0,ymm1);
 #endif
 #if defined (AVXFMA4)
-      __m256d a_real = _mm256_shuffle_pd(a,a,0x0);//arar
-      __m256d a_imag = _mm256_shuffle_pd(a,a,0xF);//aiai
-      a_imag = _mm256_mul_pd( a_imag, _mm256_permute_pd( b, 0x5 ) );  // (Ai, Ai) * (Bi, Br) = Ai Bi, Ai Br
-      return _mm256_maddsub_pd( a_real, b, a_imag ); // Ar Br , Ar Bi   +- Ai Bi             = ArBr-AiBi , ArBi+AiBr
+    __m256d a_real = _mm256_shuffle_pd(a,a,0x0);//arar
+    __m256d a_imag = _mm256_shuffle_pd(a,a,0xF);//aiai
+    a_imag = _mm256_mul_pd( a_imag, _mm256_permute_pd( b, 0x5 ) );  // (Ai, Ai) * (Bi, Br) = Ai Bi, Ai Br
+    return _mm256_maddsub_pd( a_real, b, a_imag ); // Ar Br , Ar Bi   +- Ai Bi             = ArBr-AiBi , ArBi+AiBr
 #endif
 #if defined (AVX2) || defined (AVXFMA)
-      __m256d a_real = _mm256_movedup_pd( a ); // Ar Ar
-      __m256d a_imag = _mm256_shuffle_pd(a,a,0xF);//aiai
-      a_imag = _mm256_mul_pd( a_imag, _mm256_permute_pd( b, 0x5 ) );  // (Ai, Ai) * (Bi, Br) = Ai Bi, Ai Br
-      return _mm256_fmaddsub_pd( a_real, b, a_imag ); // Ar Br , Ar Bi   +- Ai Bi             = ArBr-AiBi , ArBi+AiBr
+    __m256d a_real = _mm256_movedup_pd( a ); // Ar Ar
+    __m256d a_imag = _mm256_shuffle_pd(a,a,0xF);//aiai
+    a_imag = _mm256_mul_pd( a_imag, _mm256_permute_pd( b, 0x5 ) );  // (Ai, Ai) * (Bi, Br) = Ai Bi, Ai Br
+    return _mm256_fmaddsub_pd( a_real, b, a_imag ); // Ar Br , Ar Bi   +- Ai Bi             = ArBr-AiBi , ArBi+AiBr
 #endif
-    }
+  }
 
 
-  };
+};
 
 #if 0
-  struct ComplexDot {
-
-    inline void Prep(__m256 ari,__m256 &air) {
-      cdotRIperm(ari,air);
-    }
-    inline void Mul(__m256 ari,__m256 air,__m256 b,__m256 &riir,__m256 &iirr) {
-      riir=air*b;
-      iirr=arr*b;
-    };
-    inline void Madd(__m256 ari,__m256 air,__m256 b,__m256 &riir,__m256 &iirr) {
-      mac(riir,air,b);
-      mac(iirr,ari,b);
-    }
-    inline void End(__m256 ari,__m256 &air) {
-      //      cdotRI
-    }
+struct ComplexDot {
 
+  inline void Prep(__m256 ari,__m256 &air) {
+    cdotRIperm(ari,air);
+  }
+  inline void Mul(__m256 ari,__m256 air,__m256 b,__m256 &riir,__m256 &iirr) {
+    riir=air*b;
+    iirr=arr*b;
   };
+  inline void Madd(__m256 ari,__m256 air,__m256 b,__m256 &riir,__m256 &iirr) {
+    mac(riir,air,b);
+    mac(iirr,ari,b);
+  }
+  inline void End(__m256 ari,__m256 &air) {
+    //      cdotRI
+  }
+
+};
 #endif
 
-  struct Mult{
+struct Mult{
 
-    inline void mac(__m256 &a, __m256 b, __m256 c){
+  inline void mac(__m256 &a, __m256 b, __m256 c){
 #if defined (AVX1)
-      a= _mm256_add_ps(_mm256_mul_ps(b,c),a);
+    a= _mm256_add_ps(_mm256_mul_ps(b,c),a);
 #endif
 #if defined (AVXFMA4)
-      a= _mm256_macc_ps(b,c,a);
+    a= _mm256_macc_ps(b,c,a);
 #endif
 #if defined (AVX2) || defined (AVXFMA)
-      a= _mm256_fmadd_ps( b, c, a);
+    a= _mm256_fmadd_ps( b, c, a);
 #endif
-    }
+  }
 
-    inline void mac(__m256d &a, __m256d b, __m256d c){
+  inline void mac(__m256d &a, __m256d b, __m256d c){
 #if defined (AVX1)
-      a= _mm256_add_pd(_mm256_mul_pd(b,c),a);
+    a= _mm256_add_pd(_mm256_mul_pd(b,c),a);
 #endif
 #if defined (AVXFMA4)
-      a= _mm256_macc_pd(b,c,a);
+    a= _mm256_macc_pd(b,c,a);
 #endif
 #if defined (AVX2) || defined (AVXFMA)
-      a= _mm256_fmadd_pd( b, c, a);
+    a= _mm256_fmadd_pd( b, c, a);
 #endif
-    }
+  }
 
-    // Real float
-    inline __m256 operator()(__m256 a, __m256 b){
-      return _mm256_mul_ps(a,b);
-    }
-    // Real double
-    inline __m256d operator()(__m256d a, __m256d b){
-      return _mm256_mul_pd(a,b);
-    }
-    // Integer
-    inline __m256i operator()(__m256i a, __m256i b){
+  // Real float
+  inline __m256 operator()(__m256 a, __m256 b){
+    return _mm256_mul_ps(a,b);
+  }
+  // Real double
+  inline __m256d operator()(__m256d a, __m256d b){
+    return _mm256_mul_pd(a,b);
+  }
+  // Integer
+  inline __m256i operator()(__m256i a, __m256i b){
 #if defined (AVX1) || defined (AVXFMA)
-      __m128i a0,a1;
-      __m128i b0,b1;
-      a0 = _mm256_extractf128_si256(a,0);
-      b0 = _mm256_extractf128_si256(b,0);
-      a1 = _mm256_extractf128_si256(a,1);
-      b1 = _mm256_extractf128_si256(b,1);
-      a0 = _mm_mullo_epi32(a0,b0);
-      a1 = _mm_mullo_epi32(a1,b1);
-      return _mm256_set_m128i(a1,a0);
+    __m128i a0,a1;
+    __m128i b0,b1;
+    a0 = _mm256_extractf128_si256(a,0);
+    b0 = _mm256_extractf128_si256(b,0);
+    a1 = _mm256_extractf128_si256(a,1);
+    b1 = _mm256_extractf128_si256(b,1);
+    a0 = _mm_mullo_epi32(a0,b0);
+    a1 = _mm_mullo_epi32(a1,b1);
+    return _mm256_set_m128i(a1,a0);
 #endif
 #if defined (AVX2)
-      return _mm256_mullo_epi32(a,b);
+    return _mm256_mullo_epi32(a,b);
 #endif
 
-    }
+  }
+};
+
+struct Div {
+  // Real float
+  inline __m256 operator()(__m256 a, __m256 b) {
+    return _mm256_div_ps(a, b);
+  }
+  // Real double
+  inline __m256d operator()(__m256d a, __m256d b){
+    return _mm256_div_pd(a,b);
+  }
+};
+
+
+struct Conj{
+  // Complex single
+  inline __m256 operator()(__m256 in){
+    return _mm256_xor_ps(_mm256_addsub_ps(_mm256_setzero_ps(),in), _mm256_set1_ps(-0.f));
+  }
+  // Complex double
+  inline __m256d operator()(__m256d in){
+    return _mm256_xor_pd(_mm256_addsub_pd(_mm256_setzero_pd(),in), _mm256_set1_pd(-0.f));
+  }
+  // do not define for integer input
+};
+
+struct TimesMinusI{
+  //Complex single
+  inline __m256 operator()(__m256 in, __m256 ret){
+    __m256 tmp =_mm256_addsub_ps(_mm256_setzero_ps(),in);   // r,-i
+    return _mm256_shuffle_ps(tmp,tmp,_MM_SELECT_FOUR_FOUR(2,3,0,1)); //-i,r
+  }
+  //Complex double
+  inline __m256d operator()(__m256d in, __m256d ret){
+    __m256d tmp = _mm256_addsub_pd(_mm256_setzero_pd(),in); // r,-i
+    return _mm256_shuffle_pd(tmp,tmp,0x5);
+  }
+};
+
+struct TimesI{
+  //Complex single
+  inline __m256 operator()(__m256 in, __m256 ret){
+    __m256 tmp =_mm256_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1)); // i,r
+    return _mm256_addsub_ps(_mm256_setzero_ps(),tmp);          // i,-r
+  }
+  //Complex double
+  inline __m256d operator()(__m256d in, __m256d ret){
+    __m256d tmp = _mm256_shuffle_pd(in,in,0x5);
+    return _mm256_addsub_pd(_mm256_setzero_pd(),tmp); // i,-r
+  }
+};
+
+//////////////////////////////////////////////
+// Some Template specialization
+//////////////////////////////////////////////
+
+struct Permute{
+
+  static inline __m256 Permute0(__m256 in){
+    return _mm256_permute2f128_ps(in,in,0x01); //ABCD EFGH -> EFGH ABCD
+  };
+  static inline __m256 Permute1(__m256 in){
+    return _mm256_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(1,0,3,2)); //ABCD EFGH -> CDAB GHEF
+  };
+  static inline __m256 Permute2(__m256 in){
+    return _mm256_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1)); //ABCD EFGH -> BADC FEHG
+  };
+  static inline __m256 Permute3(__m256 in){
+    return in;
   };
 
-  struct Div {
-    // Real float
-    inline __m256 operator()(__m256 a, __m256 b) {
-      return _mm256_div_ps(a, b);
-    }
-    // Real double
-    inline __m256d operator()(__m256d a, __m256d b){
-      return _mm256_div_pd(a,b);
-    }
+  static inline __m256d Permute0(__m256d in){
+    return _mm256_permute2f128_pd(in,in,0x01); //AB CD -> CD AB
   };
-
-
-  struct Conj{
-    // Complex single
-    inline __m256 operator()(__m256 in){
-      return _mm256_xor_ps(_mm256_addsub_ps(_mm256_setzero_ps(),in), _mm256_set1_ps(-0.f));
-    }
-    // Complex double
-    inline __m256d operator()(__m256d in){
-      return _mm256_xor_pd(_mm256_addsub_pd(_mm256_setzero_pd(),in), _mm256_set1_pd(-0.f));
-    }
-    // do not define for integer input
+  static inline __m256d Permute1(__m256d in){ //AB CD -> BA DC
+    return _mm256_shuffle_pd(in,in,0x5);
   };
-
-  struct TimesMinusI{
-    //Complex single
-    inline __m256 operator()(__m256 in, __m256 ret){
-      __m256 tmp =_mm256_addsub_ps(_mm256_setzero_ps(),in);   // r,-i
-      return _mm256_shuffle_ps(tmp,tmp,_MM_SELECT_FOUR_FOUR(2,3,0,1)); //-i,r
-    }
-    //Complex double
-    inline __m256d operator()(__m256d in, __m256d ret){
-      __m256d tmp = _mm256_addsub_pd(_mm256_setzero_pd(),in); // r,-i
-      return _mm256_shuffle_pd(tmp,tmp,0x5);
-    }
+  static inline __m256d Permute2(__m256d in){
+    return in;
   };
-
-  struct TimesI{
-    //Complex single
-    inline __m256 operator()(__m256 in, __m256 ret){
-      __m256 tmp =_mm256_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1)); // i,r
-      return _mm256_addsub_ps(_mm256_setzero_ps(),tmp);          // i,-r
-    }
-    //Complex double
-    inline __m256d operator()(__m256d in, __m256d ret){
-      __m256d tmp = _mm256_shuffle_pd(in,in,0x5);
-      return _mm256_addsub_pd(_mm256_setzero_pd(),tmp); // i,-r
-    }
-  };
-
-  //////////////////////////////////////////////
-  // Some Template specialization
-  //////////////////////////////////////////////
-
-  struct Permute{
-
-    static inline __m256 Permute0(__m256 in){
-      return _mm256_permute2f128_ps(in,in,0x01); //ABCD EFGH -> EFGH ABCD
-    };
-    static inline __m256 Permute1(__m256 in){
-      return _mm256_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(1,0,3,2)); //ABCD EFGH -> CDAB GHEF
-    };
-    static inline __m256 Permute2(__m256 in){
-      return _mm256_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1)); //ABCD EFGH -> BADC FEHG
-    };
-    static inline __m256 Permute3(__m256 in){
-      return in;
-    };
-
-    static inline __m256d Permute0(__m256d in){
-      return _mm256_permute2f128_pd(in,in,0x01); //AB CD -> CD AB
-    };
-    static inline __m256d Permute1(__m256d in){ //AB CD -> BA DC
-      return _mm256_shuffle_pd(in,in,0x5);
-    };
-    static inline __m256d Permute2(__m256d in){
-      return in;
-    };
-    static inline __m256d Permute3(__m256d in){
-      return in;
-    };
+  static inline __m256d Permute3(__m256d in){
+    return in;
   };
+};
 #define USE_FP16
-  struct PrecisionChange {
-    static inline __m256i StoH (__m256 a,__m256 b) {
-      __m256i h;
+struct PrecisionChange {
+  static inline __m256i StoH (__m256 a,__m256 b) {
+    __m256i h;
 #ifdef USE_FP16
-      __m128i ha = _mm256_cvtps_ph(a,0);
-      __m128i hb = _mm256_cvtps_ph(b,0);
-      h =(__m256i) _mm256_castps128_ps256((__m128)ha);
-      h =(__m256i) _mm256_insertf128_ps((__m256)h,(__m128)hb,1);
+    __m128i ha = _mm256_cvtps_ph(a,0);
+    __m128i hb = _mm256_cvtps_ph(b,0);
+    h =(__m256i) _mm256_castps128_ps256((__m128)ha);
+    h =(__m256i) _mm256_insertf128_ps((__m256)h,(__m128)hb,1);
 #else 
-      assert(0);
+    assert(0);
 #endif
-      return h;
-    }
-    static inline void  HtoS (__m256i h,__m256 &sa,__m256 &sb) {
+    return h;
+  }
+  static inline void  HtoS (__m256i h,__m256 &sa,__m256 &sb) {
 #ifdef USE_FP16
-      sa = _mm256_cvtph_ps((__m128i)_mm256_extractf128_ps((__m256)h,0));
-      sb = _mm256_cvtph_ps((__m128i)_mm256_extractf128_ps((__m256)h,1));
+    sa = _mm256_cvtph_ps((__m128i)_mm256_extractf128_ps((__m256)h,0));
+    sb = _mm256_cvtph_ps((__m128i)_mm256_extractf128_ps((__m256)h,1));
 #else 
-      assert(0);
+    assert(0);
 #endif
-    }
-    static inline __m256 DtoS (__m256d a,__m256d b) {
-      __m128 sa = _mm256_cvtpd_ps(a);
-      __m128 sb = _mm256_cvtpd_ps(b);
-      __m256 s = _mm256_castps128_ps256(sa);
-      s = _mm256_insertf128_ps(s,sb,1);
-      return s;
-    }
-    static inline void StoD (__m256 s,__m256d &a,__m256d &b) {
-      a = _mm256_cvtps_pd(_mm256_extractf128_ps(s,0));
-      b = _mm256_cvtps_pd(_mm256_extractf128_ps(s,1));
-    }
-    static inline __m256i DtoH (__m256d a,__m256d b,__m256d c,__m256d d) {
-      __m256 sa,sb;
-      sa = DtoS(a,b);
-      sb = DtoS(c,d);
-      return StoH(sa,sb);
-    }
-    static inline void HtoD (__m256i h,__m256d &a,__m256d &b,__m256d &c,__m256d &d) {
-      __m256 sa,sb;
-      HtoS(h,sa,sb);
-      StoD(sa,a,b);
-      StoD(sb,c,d);
-    }
+  }
+  static inline __m256 DtoS (__m256d a,__m256d b) {
+    __m128 sa = _mm256_cvtpd_ps(a);
+    __m128 sb = _mm256_cvtpd_ps(b);
+    __m256 s = _mm256_castps128_ps256(sa);
+    s = _mm256_insertf128_ps(s,sb,1);
+    return s;
+  }
+  static inline void StoD (__m256 s,__m256d &a,__m256d &b) {
+    a = _mm256_cvtps_pd(_mm256_extractf128_ps(s,0));
+    b = _mm256_cvtps_pd(_mm256_extractf128_ps(s,1));
+  }
+  static inline __m256i DtoH (__m256d a,__m256d b,__m256d c,__m256d d) {
+    __m256 sa,sb;
+    sa = DtoS(a,b);
+    sb = DtoS(c,d);
+    return StoH(sa,sb);
+  }
+  static inline void HtoD (__m256i h,__m256d &a,__m256d &b,__m256d &c,__m256d &d) {
+    __m256 sa,sb;
+    HtoS(h,sa,sb);
+    StoD(sa,a,b);
+    StoD(sb,c,d);
+  }
+};
+struct Exchange{
+  // 3210 ordering
+  static inline void Exchange0(__m256 &out1,__m256 &out2,__m256 in1,__m256 in2){
+    //Invertible
+    //AB CD ->  AC BD
+    //AC BD ->  AB CD
+    out1= _mm256_permute2f128_ps(in1,in2,0x20);
+    out2= _mm256_permute2f128_ps(in1,in2,0x31);
+  };
+  static inline void Exchange1(__m256 &out1,__m256 &out2,__m256 in1,__m256 in2){
+    //Invertible
+    // ABCD EFGH  ->ABEF CDGH
+    // ABEF CDGH  ->ABCD EFGH
+    out1= _mm256_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(1,0,1,0));
+    out2= _mm256_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(3,2,3,2));
+  };
+  static inline void Exchange2(__m256 &out1,__m256 &out2,__m256 in1,__m256 in2){
+    // Invertible ? 
+    // ABCD EFGH -> ACEG BDFH
+    // ACEG BDFH -> AEBF CGDH
+    //      out1= _mm256_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(2,0,2,0));
+    //      out2= _mm256_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(3,1,3,1));
+    // Bollocks; need 
+    // AECG BFDH -> ABCD EFGH
+    out1= _mm256_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(2,0,2,0)); /*ACEG*/
+    out2= _mm256_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(3,1,3,1)); /*BDFH*/
+    out1= _mm256_shuffle_ps(out1,out1,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
+    out2= _mm256_shuffle_ps(out2,out2,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
+  };
+  static inline void Exchange3(__m256 &out1,__m256 &out2,__m256 in1,__m256 in2){
+    assert(0);
+    return;
   };
-  struct Exchange{
-    // 3210 ordering
-    static inline void Exchange0(__m256 &out1,__m256 &out2,__m256 in1,__m256 in2){
-      //Invertible
-      //AB CD ->  AC BD
-      //AC BD ->  AB CD
-      out1= _mm256_permute2f128_ps(in1,in2,0x20);
-      out2= _mm256_permute2f128_ps(in1,in2,0x31);
-    };
-    static inline void Exchange1(__m256 &out1,__m256 &out2,__m256 in1,__m256 in2){
-      //Invertible
-      // ABCD EFGH  ->ABEF CDGH
-      // ABEF CDGH  ->ABCD EFGH
-      out1= _mm256_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(1,0,1,0));
-      out2= _mm256_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(3,2,3,2));
-    };
-    static inline void Exchange2(__m256 &out1,__m256 &out2,__m256 in1,__m256 in2){
-      // Invertible ? 
-      // ABCD EFGH -> ACEG BDFH
-      // ACEG BDFH -> AEBF CGDH
-      //      out1= _mm256_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(2,0,2,0));
-      //      out2= _mm256_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(3,1,3,1));
-      // Bollocks; need 
-      // AECG BFDH -> ABCD EFGH
-      out1= _mm256_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(2,0,2,0)); /*ACEG*/
-      out2= _mm256_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(3,1,3,1)); /*BDFH*/
-      out1= _mm256_shuffle_ps(out1,out1,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
-      out2= _mm256_shuffle_ps(out2,out2,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
-    };
-    static inline void Exchange3(__m256 &out1,__m256 &out2,__m256 in1,__m256 in2){
-      assert(0);
-      return;
-    };
 
-    static inline void Exchange0(__m256d &out1,__m256d &out2,__m256d in1,__m256d in2){
-      out1= _mm256_permute2f128_pd(in1,in2,0x20);
-      out2= _mm256_permute2f128_pd(in1,in2,0x31);
-      return;
-    };
-    static inline void Exchange1(__m256d &out1,__m256d &out2,__m256d in1,__m256d in2){
-      out1= _mm256_shuffle_pd(in1,in2,0x0);
-      out2= _mm256_shuffle_pd(in1,in2,0xF);
-    };
-    static inline void Exchange2(__m256d &out1,__m256d &out2,__m256d in1,__m256d in2){
-      assert(0);
-      return;
-    };
-    static inline void Exchange3(__m256d &out1,__m256d &out2,__m256d in1,__m256d in2){
-      assert(0);
-      return;
-    };
+  static inline void Exchange0(__m256d &out1,__m256d &out2,__m256d in1,__m256d in2){
+    out1= _mm256_permute2f128_pd(in1,in2,0x20);
+    out2= _mm256_permute2f128_pd(in1,in2,0x31);
+    return;
   };
+  static inline void Exchange1(__m256d &out1,__m256d &out2,__m256d in1,__m256d in2){
+    out1= _mm256_shuffle_pd(in1,in2,0x0);
+    out2= _mm256_shuffle_pd(in1,in2,0xF);
+  };
+  static inline void Exchange2(__m256d &out1,__m256d &out2,__m256d in1,__m256d in2){
+    assert(0);
+    return;
+  };
+  static inline void Exchange3(__m256d &out1,__m256d &out2,__m256d in1,__m256d in2){
+    assert(0);
+    return;
+  };
+};
 
 
 #if defined (AVX2)
@@ -567,203 +567,203 @@ namespace Optimization {
 #endif
 
 #if defined (AVX1) || defined (AVXFMA)
-#define _mm256_alignr_epi32_grid(ret,a,b,n) {	\
-    __m128 aa, bb;				\
-						\
-    aa  = _mm256_extractf128_ps(a,1);		\
-    bb  = _mm256_extractf128_ps(b,1);		\
+#define _mm256_alignr_epi32_grid(ret,a,b,n) {				\
+    __m128 aa, bb;							\
+									\
+    aa  = _mm256_extractf128_ps(a,1);					\
+    bb  = _mm256_extractf128_ps(b,1);					\
     aa  = (__m128)_mm_alignr_epi8((__m128i)aa,(__m128i)bb,(n*4)%16);	\
-    ret = _mm256_insertf128_ps(ret,aa,1);	\
-						\
-    aa  = _mm256_extractf128_ps(a,0);		\
-    bb  = _mm256_extractf128_ps(b,0);		\
+    ret = _mm256_insertf128_ps(ret,aa,1);				\
+									\
+    aa  = _mm256_extractf128_ps(a,0);					\
+    bb  = _mm256_extractf128_ps(b,0);					\
     aa  = (__m128)_mm_alignr_epi8((__m128i)aa,(__m128i)bb,(n*4)%16);	\
-    ret = _mm256_insertf128_ps(ret,aa,0);	\
+    ret = _mm256_insertf128_ps(ret,aa,0);				\
   }
 
-#define _mm256_alignr_epi64_grid(ret,a,b,n) {	\
-    __m128d aa, bb;				\
-						\
-    aa  = _mm256_extractf128_pd(a,1);		\
-    bb  = _mm256_extractf128_pd(b,1);		\
+#define _mm256_alignr_epi64_grid(ret,a,b,n) {				\
+    __m128d aa, bb;							\
+									\
+    aa  = _mm256_extractf128_pd(a,1);					\
+    bb  = _mm256_extractf128_pd(b,1);					\
     aa  = (__m128d)_mm_alignr_epi8((__m128i)aa,(__m128i)bb,(n*8)%16);	\
-    ret = _mm256_insertf128_pd(ret,aa,1);	\
-						\
-    aa  = _mm256_extractf128_pd(a,0);		\
-    bb  = _mm256_extractf128_pd(b,0);		\
+    ret = _mm256_insertf128_pd(ret,aa,1);				\
+									\
+    aa  = _mm256_extractf128_pd(a,0);					\
+    bb  = _mm256_extractf128_pd(b,0);					\
     aa  = (__m128d)_mm_alignr_epi8((__m128i)aa,(__m128i)bb,(n*8)%16);	\
-    ret = _mm256_insertf128_pd(ret,aa,0);	\
+    ret = _mm256_insertf128_pd(ret,aa,0);				\
   }
 
 #endif
 
-  struct Rotate{
+struct Rotate{
 
-    static inline __m256 rotate(__m256 in,int n){
-      switch(n){
-      case 0: return tRotate<0>(in);break;
-      case 1: return tRotate<1>(in);break;
-      case 2: return tRotate<2>(in);break;
-      case 3: return tRotate<3>(in);break;
-      case 4: return tRotate<4>(in);break;
-      case 5: return tRotate<5>(in);break;
-      case 6: return tRotate<6>(in);break;
-      case 7: return tRotate<7>(in);break;
-      default: assert(0);
-      }
+  static inline __m256 rotate(__m256 in,int n){
+    switch(n){
+    case 0: return tRotate<0>(in);break;
+    case 1: return tRotate<1>(in);break;
+    case 2: return tRotate<2>(in);break;
+    case 3: return tRotate<3>(in);break;
+    case 4: return tRotate<4>(in);break;
+    case 5: return tRotate<5>(in);break;
+    case 6: return tRotate<6>(in);break;
+    case 7: return tRotate<7>(in);break;
+    default: assert(0);
     }
-    static inline __m256d rotate(__m256d in,int n){
-      switch(n){
-      case 0: return tRotate<0>(in);break;
-      case 1: return tRotate<1>(in);break;
-      case 2: return tRotate<2>(in);break;
-      case 3: return tRotate<3>(in);break;
-      default: assert(0);
-      }
+  }
+  static inline __m256d rotate(__m256d in,int n){
+    switch(n){
+    case 0: return tRotate<0>(in);break;
+    case 1: return tRotate<1>(in);break;
+    case 2: return tRotate<2>(in);break;
+    case 3: return tRotate<3>(in);break;
+    default: assert(0);
     }
+  }
 
 
-    template<int n>
-    static inline __m256 tRotate(__m256 in){
-      __m256 tmp = Permute::Permute0(in);
-      __m256 ret;
-      if ( n > 3 ) {
-          _mm256_alignr_epi32_grid(ret,in,tmp,n);
-      } else {
-          _mm256_alignr_epi32_grid(ret,tmp,in,n);
-      }
-      return ret;
+  template<int n>
+  static inline __m256 tRotate(__m256 in){
+    __m256 tmp = Permute::Permute0(in);
+    __m256 ret;
+    if ( n > 3 ) {
+      _mm256_alignr_epi32_grid(ret,in,tmp,n);
+    } else {
+      _mm256_alignr_epi32_grid(ret,tmp,in,n);
     }
+    return ret;
+  }
 
-    template<int n>
-    static inline __m256d tRotate(__m256d in){
-      __m256d tmp = Permute::Permute0(in);
-      __m256d ret;
-      if ( n > 1 ) {
-	_mm256_alignr_epi64_grid(ret,in,tmp,n);
-      } else {
-        _mm256_alignr_epi64_grid(ret,tmp,in,n);
-      }
-      return ret;
-    };
-
+  template<int n>
+  static inline __m256d tRotate(__m256d in){
+    __m256d tmp = Permute::Permute0(in);
+    __m256d ret;
+    if ( n > 1 ) {
+      _mm256_alignr_epi64_grid(ret,in,tmp,n);
+    } else {
+      _mm256_alignr_epi64_grid(ret,tmp,in,n);
+    }
+    return ret;
   };
 
-  //Complex float Reduce
-  template<>
-    inline Grid::ComplexF Reduce<Grid::ComplexF, __m256>::operator()(__m256 in){
-    __m256 v1,v2;
-    v1=Optimization::Permute::Permute0(in); // avx 256; quad complex single
-    v1= _mm256_add_ps(v1,in);
-    v2=Optimization::Permute::Permute1(v1);
-    v1 = _mm256_add_ps(v1,v2);
-    u256f conv; conv.v = v1;
-    return Grid::ComplexF(conv.f[0],conv.f[1]);
-  }
-
-  //Real float Reduce
-  template<>
-  inline Grid::RealF Reduce<Grid::RealF, __m256>::operator()(__m256 in){
-    __m256 v1,v2;
-    v1 = Optimization::Permute::Permute0(in); // avx 256; octo-double
-    v1 = _mm256_add_ps(v1,in);
-    v2 = Optimization::Permute::Permute1(v1);
-    v1 = _mm256_add_ps(v1,v2);
-    v2 = Optimization::Permute::Permute2(v1);
-    v1 = _mm256_add_ps(v1,v2);
-    u256f conv; conv.v=v1;
-    return conv.f[0];
-  }
-
-
-  //Complex double Reduce
-  template<>
-  inline Grid::ComplexD Reduce<Grid::ComplexD, __m256d>::operator()(__m256d in){
-    __m256d v1;
-    v1 = Optimization::Permute::Permute0(in); // sse 128; paired complex single
-    v1 = _mm256_add_pd(v1,in);
-    u256d conv; conv.v = v1;
-    return Grid::ComplexD(conv.f[0],conv.f[1]);
-  }
-
-  //Real double Reduce
-  template<>
-  inline Grid::RealD Reduce<Grid::RealD, __m256d>::operator()(__m256d in){
-    __m256d v1,v2;
-    v1 = Optimization::Permute::Permute0(in); // avx 256; quad double
-    v1 = _mm256_add_pd(v1,in);
-    v2 = Optimization::Permute::Permute1(v1);
-    v1 = _mm256_add_pd(v1,v2);
-    u256d conv; conv.v = v1;
-    return conv.f[0];
-  }
-
-  //Integer Reduce
-  template<>
-  inline Integer Reduce<Integer, __m256i>::operator()(__m256i in){
-    __m128i ret;
-#if defined (AVX2)
-    // AVX2 horizontal adds within upper and lower halves of register; use
-    // SSE to add upper and lower halves for result.
-    __m256i v1, v2;
-    __m128i u1, u2;
-    v1  = _mm256_hadd_epi32(in, in);
-    v2  = _mm256_hadd_epi32(v1, v1);
-    u1  = _mm256_castsi256_si128(v2);      // upper half
-    u2  = _mm256_extracti128_si256(v2, 1); // lower half
-    ret = _mm_add_epi32(u1, u2);
-#else
-    // No AVX horizontal add; extract upper and lower halves of register & use
-    // SSE intrinsics.
-    __m128i u1, u2, u3;
-    u1  = _mm256_extractf128_si256(in, 0); // upper half
-    u2  = _mm256_extractf128_si256(in, 1); // lower half
-    u3  = _mm_add_epi32(u1, u2);
-    u1  = _mm_hadd_epi32(u3, u3);
-    ret = _mm_hadd_epi32(u1, u1);
-#endif
-    return _mm_cvtsi128_si32(ret);
-  }
+};
 
+//Complex float Reduce
+template<>
+inline Grid::ComplexF Reduce<Grid::ComplexF, __m256>::operator()(__m256 in){
+  __m256 v1,v2;
+  v1=Optimization::Permute::Permute0(in); // avx 256; quad complex single
+  v1= _mm256_add_ps(v1,in);
+  v2=Optimization::Permute::Permute1(v1);
+  v1 = _mm256_add_ps(v1,v2);
+  u256f conv; conv.v = v1;
+  return Grid::ComplexF(conv.f[0],conv.f[1]);
 }
 
+//Real float Reduce
+template<>
+inline Grid::RealF Reduce<Grid::RealF, __m256>::operator()(__m256 in){
+  __m256 v1,v2;
+  v1 = Optimization::Permute::Permute0(in); // avx 256; octo-double
+  v1 = _mm256_add_ps(v1,in);
+  v2 = Optimization::Permute::Permute1(v1);
+  v1 = _mm256_add_ps(v1,v2);
+  v2 = Optimization::Permute::Permute2(v1);
+  v1 = _mm256_add_ps(v1,v2);
+  u256f conv; conv.v=v1;
+  return conv.f[0];
+}
+
+
+//Complex double Reduce
+template<>
+inline Grid::ComplexD Reduce<Grid::ComplexD, __m256d>::operator()(__m256d in){
+  __m256d v1;
+  v1 = Optimization::Permute::Permute0(in); // sse 128; paired complex single
+  v1 = _mm256_add_pd(v1,in);
+  u256d conv; conv.v = v1;
+  return Grid::ComplexD(conv.f[0],conv.f[1]);
+}
+
+//Real double Reduce
+template<>
+inline Grid::RealD Reduce<Grid::RealD, __m256d>::operator()(__m256d in){
+  __m256d v1,v2;
+  v1 = Optimization::Permute::Permute0(in); // avx 256; quad double
+  v1 = _mm256_add_pd(v1,in);
+  v2 = Optimization::Permute::Permute1(v1);
+  v1 = _mm256_add_pd(v1,v2);
+  u256d conv; conv.v = v1;
+  return conv.f[0];
+}
+
+//Integer Reduce
+template<>
+inline Integer Reduce<Integer, __m256i>::operator()(__m256i in){
+  __m128i ret;
+#if defined (AVX2)
+  // AVX2 horizontal adds within upper and lower halves of register; use
+  // SSE to add upper and lower halves for result.
+  __m256i v1, v2;
+  __m128i u1, u2;
+  v1  = _mm256_hadd_epi32(in, in);
+  v2  = _mm256_hadd_epi32(v1, v1);
+  u1  = _mm256_castsi256_si128(v2);      // upper half
+  u2  = _mm256_extracti128_si256(v2, 1); // lower half
+  ret = _mm_add_epi32(u1, u2);
+#else
+  // No AVX horizontal add; extract upper and lower halves of register & use
+  // SSE intrinsics.
+  __m128i u1, u2, u3;
+  u1  = _mm256_extractf128_si256(in, 0); // upper half
+  u2  = _mm256_extractf128_si256(in, 1); // lower half
+  u3  = _mm_add_epi32(u1, u2);
+  u1  = _mm_hadd_epi32(u3, u3);
+  ret = _mm_hadd_epi32(u1, u1);
+#endif
+  return _mm_cvtsi128_si32(ret);
+}
+
+NAMESPACE_END(Optimization);
+
 //////////////////////////////////////////////////////////////////////////////////////
 // Here assign types
 
-  typedef __m256i SIMD_Htype;  // Single precision type
-  typedef __m256  SIMD_Ftype; // Single precision type
-  typedef __m256d SIMD_Dtype; // Double precision type
-  typedef __m256i SIMD_Itype; // Integer type
+typedef __m256i SIMD_Htype;  // Single precision type
+typedef __m256  SIMD_Ftype; // Single precision type
+typedef __m256d SIMD_Dtype; // Double precision type
+typedef __m256i SIMD_Itype; // Integer type
 
-  // prefecthing
-  inline void v_prefetch0(int size, const char *ptr){
-    for(int i=0;i<size;i+=64){ //  Define L1 linesize above
-      _mm_prefetch(ptr+i+4096,_MM_HINT_T1);
-      _mm_prefetch(ptr+i+512,_MM_HINT_T0);
-    }
-  }
-  inline void prefetch_HINT_T0(const char *ptr){
-    _mm_prefetch(ptr, _MM_HINT_T0);
+// prefecthing
+inline void v_prefetch0(int size, const char *ptr){
+  for(int i=0;i<size;i+=64){ //  Define L1 linesize above
+    _mm_prefetch(ptr+i+4096,_MM_HINT_T1);
+    _mm_prefetch(ptr+i+512,_MM_HINT_T0);
   }
+}
+inline void prefetch_HINT_T0(const char *ptr){
+  _mm_prefetch(ptr, _MM_HINT_T0);
+}
 
-  // Function name aliases
-  typedef Optimization::Vsplat   VsplatSIMD;
-  typedef Optimization::Vstore   VstoreSIMD;
-  typedef Optimization::Vset     VsetSIMD;
-  typedef Optimization::Vstream  VstreamSIMD;
+// Function name aliases
+typedef Optimization::Vsplat   VsplatSIMD;
+typedef Optimization::Vstore   VstoreSIMD;
+typedef Optimization::Vset     VsetSIMD;
+typedef Optimization::Vstream  VstreamSIMD;
 
-  template <typename S, typename T> using ReduceSIMD = Optimization::Reduce<S, T>;
+template <typename S, typename T> using ReduceSIMD = Optimization::Reduce<S, T>;
 
-  // Arithmetic operations
-  typedef Optimization::Sum         SumSIMD;
-  typedef Optimization::Sub         SubSIMD;
-  typedef Optimization::Div         DivSIMD;
-  typedef Optimization::Mult        MultSIMD;
-  typedef Optimization::MultComplex  MultComplexSIMD;
-  typedef Optimization::MultRealPart MultRealPartSIMD;
-  typedef Optimization::MaddRealPart MaddRealPartSIMD;
-  typedef Optimization::Conj        ConjSIMD;
-  typedef Optimization::TimesMinusI TimesMinusISIMD;
-  typedef Optimization::TimesI      TimesISIMD;
+// Arithmetic operations
+typedef Optimization::Sum         SumSIMD;
+typedef Optimization::Sub         SubSIMD;
+typedef Optimization::Div         DivSIMD;
+typedef Optimization::Mult        MultSIMD;
+typedef Optimization::MultComplex  MultComplexSIMD;
+typedef Optimization::MultRealPart MultRealPartSIMD;
+typedef Optimization::MaddRealPart MaddRealPartSIMD;
+typedef Optimization::Conj        ConjSIMD;
+typedef Optimization::TimesMinusI TimesMinusISIMD;
+typedef Optimization::TimesI      TimesISIMD;
 
-}  // namespace Grid
+NAMESPACE_END(Grid)
diff --git a/Grid/simd/Grid_avx512.h b/Grid/simd/Grid_avx512.h
index 7546b22d..839d4554 100644
--- a/Grid/simd/Grid_avx512.h
+++ b/Grid/simd/Grid_avx512.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,537 +25,529 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #include <immintrin.h>
 
+NAMESPACE_BEGIN(Grid);
+NAMESPACE_BEGIN(Optimization);
 
-namespace Grid{
-namespace Optimization {
+union u512f {
+  __m512 v;
+  float f[16];
+};
 
-  union u512f {
-    __m512 v;
-    float f[16];
-  };
-
-  union u512d {
-    __m512d v;
-    double f[8];
-  };
+union u512d {
+  __m512d v;
+  double f[8];
+};
   
-  struct Vsplat{
-    //Complex float
-    inline __m512 operator()(float a, float b){
-      return _mm512_set_ps(b,a,b,a,b,a,b,a,b,a,b,a,b,a,b,a);
-    }
-    // Real float
-    inline __m512 operator()(float a){
-      return _mm512_set1_ps(a);
-    }
-    //Complex double
-    inline __m512d operator()(double a, double b){
-      return _mm512_set_pd(b,a,b,a,b,a,b,a);
-    }
-    //Real double
-    inline __m512d operator()(double a){
-      return _mm512_set1_pd(a);
-    }
-    //Integer
-    inline __m512i operator()(Integer a){
-      return _mm512_set1_epi32(a);
-    }
-  };
+struct Vsplat{
+  //Complex float
+  inline __m512 operator()(float a, float b){
+    return _mm512_set_ps(b,a,b,a,b,a,b,a,b,a,b,a,b,a,b,a);
+  }
+  // Real float
+  inline __m512 operator()(float a){
+    return _mm512_set1_ps(a);
+  }
+  //Complex double
+  inline __m512d operator()(double a, double b){
+    return _mm512_set_pd(b,a,b,a,b,a,b,a);
+  }
+  //Real double
+  inline __m512d operator()(double a){
+    return _mm512_set1_pd(a);
+  }
+  //Integer
+  inline __m512i operator()(Integer a){
+    return _mm512_set1_epi32(a);
+  }
+};
 
-  struct Vstore{
-    //Float 
-    inline void operator()(__m512 a, float* F){
-      _mm512_store_ps(F,a);
-    }
-    //Double
-    inline void operator()(__m512d a, double* D){
-      _mm512_store_pd(D,a);
-    }
-    //Integer
-    inline void operator()(__m512i a, Integer* I){
-      _mm512_store_si512((__m512i *)I,a);
-    }
+struct Vstore{
+  //Float 
+  inline void operator()(__m512 a, float* F){
+    _mm512_store_ps(F,a);
+  }
+  //Double
+  inline void operator()(__m512d a, double* D){
+    _mm512_store_pd(D,a);
+  }
+  //Integer
+  inline void operator()(__m512i a, Integer* I){
+    _mm512_store_si512((__m512i *)I,a);
+  }
 
-  };
+};
 
+struct Vstream{
+  //Float
+  inline void operator()(float * a, __m512 b){
+    _mm512_stream_ps(a,b);
+    //      _mm512_store_ps(a,b);
+  }
+  //Double
+  inline void operator()(double * a, __m512d b){
+    _mm512_stream_pd(a,b);
+    //      _mm512_store_pd(a,b);
+  }
 
-  struct Vstream{
-    //Float
-    inline void operator()(float * a, __m512 b){
-      _mm512_stream_ps(a,b);
-      //      _mm512_store_ps(a,b);
-    }
-    //Double
-    inline void operator()(double * a, __m512d b){
-      _mm512_stream_pd(a,b);
-      //      _mm512_store_pd(a,b);
-    }
+};
 
-  };
-
-
-
-  struct Vset{
-    // Complex float 
-    inline __m512 operator()(Grid::ComplexF *a){
-      return _mm512_set_ps(a[7].imag(),a[7].real(),a[6].imag(),a[6].real(),
-			   a[5].imag(),a[5].real(),a[4].imag(),a[4].real(),
-			   a[3].imag(),a[3].real(),a[2].imag(),a[2].real(),
-			   a[1].imag(),a[1].real(),a[0].imag(),a[0].real());
-    }
-    // Complex double 
-    inline __m512d operator()(Grid::ComplexD *a){
-      return _mm512_set_pd(a[3].imag(),a[3].real(),a[2].imag(),a[2].real(),
-			   a[1].imag(),a[1].real(),a[0].imag(),a[0].real());
-    }
-    // Real float 
-    inline __m512 operator()(float *a){
-      return _mm512_set_ps( a[15],a[14],a[13],a[12],a[11],a[10],a[9],a[8],
-			    a[7],a[6],a[5],a[4],a[3],a[2],a[1],a[0]);
-    }
-    // Real double
-    inline __m512d operator()(double *a){
-      return _mm512_set_pd(a[7],a[6],a[5],a[4],a[3],a[2],a[1],a[0]);
-    }
-    // Integer
-    inline __m512i operator()(Integer *a){
-      return _mm512_set_epi32( a[15],a[14],a[13],a[12],a[11],a[10],a[9],a[8],
-			       a[7],a[6],a[5],a[4],a[3],a[2],a[1],a[0]);
-    }
-
-
-  };
-
-  template <typename Out_type, typename In_type>
-  struct Reduce{
-    //Need templated class to overload output type
-    //General form must generate error if compiled
-    inline Out_type operator()(In_type in){
-      printf("Error, using wrong Reduce function\n");
-      exit(1);
-      return 0;
-    }
-  };
+struct Vset{
+  // Complex float 
+  inline __m512 operator()(Grid::ComplexF *a){
+    return _mm512_set_ps(a[7].imag(),a[7].real(),a[6].imag(),a[6].real(),
+			 a[5].imag(),a[5].real(),a[4].imag(),a[4].real(),
+			 a[3].imag(),a[3].real(),a[2].imag(),a[2].real(),
+			 a[1].imag(),a[1].real(),a[0].imag(),a[0].real());
+  }
+  // Complex double 
+  inline __m512d operator()(Grid::ComplexD *a){
+    return _mm512_set_pd(a[3].imag(),a[3].real(),a[2].imag(),a[2].real(),
+			 a[1].imag(),a[1].real(),a[0].imag(),a[0].real());
+  }
+  // Real float 
+  inline __m512 operator()(float *a){
+    return _mm512_set_ps( a[15],a[14],a[13],a[12],a[11],a[10],a[9],a[8],
+			  a[7],a[6],a[5],a[4],a[3],a[2],a[1],a[0]);
+  }
+  // Real double
+  inline __m512d operator()(double *a){
+    return _mm512_set_pd(a[7],a[6],a[5],a[4],a[3],a[2],a[1],a[0]);
+  }
+  // Integer
+  inline __m512i operator()(Integer *a){
+    return _mm512_set_epi32( a[15],a[14],a[13],a[12],a[11],a[10],a[9],a[8],
+			     a[7],a[6],a[5],a[4],a[3],a[2],a[1],a[0]);
+  }
 
+};
 
+template <typename Out_type, typename In_type>
+struct Reduce{
+  //Need templated class to overload output type
+  //General form must generate error if compiled
+  inline Out_type operator()(In_type in){
+    printf("Error, using wrong Reduce function\n");
+    exit(1);
+    return 0;
+  }
+};
  
 
-  /////////////////////////////////////////////////////
-  // Arithmetic operations
-  /////////////////////////////////////////////////////
-  struct Sum{
-    //Complex/Real float
-    inline __m512 operator()(__m512 a, __m512 b){
-      return _mm512_add_ps(a,b);
-    }
-    //Complex/Real double
-    inline __m512d operator()(__m512d a, __m512d b){
-      return _mm512_add_pd(a,b);
-    }
-    //Integer
-    inline __m512i operator()(__m512i a, __m512i b){
-      return _mm512_add_epi32(a,b);
-    }
-  };
+/////////////////////////////////////////////////////
+// Arithmetic operations
+/////////////////////////////////////////////////////
+struct Sum{
+  //Complex/Real float
+  inline __m512 operator()(__m512 a, __m512 b){
+    return _mm512_add_ps(a,b);
+  }
+  //Complex/Real double
+  inline __m512d operator()(__m512d a, __m512d b){
+    return _mm512_add_pd(a,b);
+  }
+  //Integer
+  inline __m512i operator()(__m512i a, __m512i b){
+    return _mm512_add_epi32(a,b);
+  }
+};
 
-  struct Sub{
-    //Complex/Real float
-    inline __m512 operator()(__m512 a, __m512 b){
-      return _mm512_sub_ps(a,b);
-    }
-    //Complex/Real double
-    inline __m512d operator()(__m512d a, __m512d b){
-      return _mm512_sub_pd(a,b);
-    }
-    //Integer
-    inline __m512i operator()(__m512i a, __m512i b){
-      return _mm512_sub_epi32(a,b);
-    }
-  };
+struct Sub{
+  //Complex/Real float
+  inline __m512 operator()(__m512 a, __m512 b){
+    return _mm512_sub_ps(a,b);
+  }
+  //Complex/Real double
+  inline __m512d operator()(__m512d a, __m512d b){
+    return _mm512_sub_pd(a,b);
+  }
+  //Integer
+  inline __m512i operator()(__m512i a, __m512i b){
+    return _mm512_sub_epi32(a,b);
+  }
+};
 
-  // Note, we can beat the shuf overhead in chain with two temporaries
-  // Ar Ai , Br Bi,  Ai Ar  // one shuf
-  //tmpr Ar Br,  Ai Bi    // Mul/Mac/Mac
-  //tmpi Br Ai,  Bi Ar    // Mul/Mac/Mac
-  // add tmpi,shuf(tmpi)
-  // sub tmpr,shuf(tmpi)
-  // shuf(tmpr,tmpi).    // Could drop/trade for write mask
+// Note, we can beat the shuf overhead in chain with two temporaries
+// Ar Ai , Br Bi,  Ai Ar  // one shuf
+//tmpr Ar Br,  Ai Bi    // Mul/Mac/Mac
+//tmpi Br Ai,  Bi Ar    // Mul/Mac/Mac
+// add tmpi,shuf(tmpi)
+// sub tmpr,shuf(tmpi)
+// shuf(tmpr,tmpi).    // Could drop/trade for write mask
 
-  // Gives
-  //  2mul,4 mac +add+sub = 8 flop type insns
-  //  3shuf + 2 (+shuf)   = 5/6 simd perm and 1/2 the load.
+// Gives
+//  2mul,4 mac +add+sub = 8 flop type insns
+//  3shuf + 2 (+shuf)   = 5/6 simd perm and 1/2 the load.
 
-  struct MultRealPart{
-    inline __m512 operator()(__m512 a, __m512 b){
-      __m512 ymm0;
-      ymm0 = _mm512_moveldup_ps(a); // ymm0 <- ar ar,
-      return _mm512_mul_ps(ymm0,b);                       // ymm0 <- ar bi, ar br
-    }
-    inline __m512d operator()(__m512d a, __m512d b){
-      __m512d ymm0;
-      ymm0 = _mm512_shuffle_pd(a,a,0x00); // ymm0 <- ar ar, ar,ar b'00,00
-      return _mm512_mul_pd(ymm0,b);      // ymm0 <- ar bi, ar br
-    }
-  };
-  struct MaddRealPart{
-    inline __m512 operator()(__m512 a, __m512 b, __m512 c){
-      __m512 ymm0 =  _mm512_moveldup_ps(a); // ymm0 <- ar ar,
-      return _mm512_fmadd_ps( ymm0, b, c);                         
-    }
-    inline __m512d operator()(__m512d a, __m512d b, __m512d c){
-      __m512d ymm0 = _mm512_shuffle_pd( a, a, 0x00 );
-      return _mm512_fmadd_pd( ymm0, b, c);                         
-    }
-  };
+struct MultRealPart{
+  inline __m512 operator()(__m512 a, __m512 b){
+    __m512 ymm0;
+    ymm0 = _mm512_moveldup_ps(a); // ymm0 <- ar ar,
+    return _mm512_mul_ps(ymm0,b);                       // ymm0 <- ar bi, ar br
+  }
+  inline __m512d operator()(__m512d a, __m512d b){
+    __m512d ymm0;
+    ymm0 = _mm512_shuffle_pd(a,a,0x00); // ymm0 <- ar ar, ar,ar b'00,00
+    return _mm512_mul_pd(ymm0,b);      // ymm0 <- ar bi, ar br
+  }
+};
+struct MaddRealPart{
+  inline __m512 operator()(__m512 a, __m512 b, __m512 c){
+    __m512 ymm0 =  _mm512_moveldup_ps(a); // ymm0 <- ar ar,
+    return _mm512_fmadd_ps( ymm0, b, c);                         
+  }
+  inline __m512d operator()(__m512d a, __m512d b, __m512d c){
+    __m512d ymm0 = _mm512_shuffle_pd( a, a, 0x00 );
+    return _mm512_fmadd_pd( ymm0, b, c);                         
+  }
+};
 
-  struct MultComplex{
-    // Complex float
-    inline __m512 operator()(__m512 a, __m512 b){
-      // dup, dup, perm, mul, madd
-      __m512 a_real = _mm512_moveldup_ps( a ); // Ar Ar
-      __m512 a_imag = _mm512_movehdup_ps( a ); // Ai Ai
-      a_imag = _mm512_mul_ps( a_imag, _mm512_permute_ps( b, 0xB1 ) );  // (Ai, Ai) * (Bi, Br) = Ai Bi, Ai Br
-      return _mm512_fmaddsub_ps( a_real, b, a_imag ); // Ar Br , Ar Bi   +- Ai Bi             = ArBr-AiBi , ArBi+AiBr
-    }
-    // Complex double
-    inline __m512d operator()(__m512d a, __m512d b){
-      __m512d a_real = _mm512_shuffle_pd( a, a, 0x00 );
-      __m512d a_imag = _mm512_shuffle_pd( a, a, 0xFF );
-      a_imag = _mm512_mul_pd( a_imag, _mm512_permute_pd( b, 0x55 ) ); 
-      return _mm512_fmaddsub_pd( a_real, b, a_imag );
-    }
-  };
+struct MultComplex{
+  // Complex float
+  inline __m512 operator()(__m512 a, __m512 b){
+    // dup, dup, perm, mul, madd
+    __m512 a_real = _mm512_moveldup_ps( a ); // Ar Ar
+    __m512 a_imag = _mm512_movehdup_ps( a ); // Ai Ai
+    a_imag = _mm512_mul_ps( a_imag, _mm512_permute_ps( b, 0xB1 ) );  // (Ai, Ai) * (Bi, Br) = Ai Bi, Ai Br
+    return _mm512_fmaddsub_ps( a_real, b, a_imag ); // Ar Br , Ar Bi   +- Ai Bi             = ArBr-AiBi , ArBi+AiBr
+  }
+  // Complex double
+  inline __m512d operator()(__m512d a, __m512d b){
+    __m512d a_real = _mm512_shuffle_pd( a, a, 0x00 );
+    __m512d a_imag = _mm512_shuffle_pd( a, a, 0xFF );
+    a_imag = _mm512_mul_pd( a_imag, _mm512_permute_pd( b, 0x55 ) ); 
+    return _mm512_fmaddsub_pd( a_real, b, a_imag );
+  }
+};
   
-  struct Mult{
+struct Mult{
 
-    inline void mac(__m512 &a, __m512 b, __m512 c){         
-       a= _mm512_fmadd_ps( b, c, a);                         
-    }
-    inline void mac(__m512d &a, __m512d b, __m512d c){
-      a= _mm512_fmadd_pd( b, c, a);                   
-    }                                             
-    // Real float
-    inline __m512 operator()(__m512 a, __m512 b){
-      return _mm512_mul_ps(a,b);
-    }
-    // Real double
-    inline __m512d operator()(__m512d a, __m512d b){
-      return _mm512_mul_pd(a,b);
-    }
-    // Integer
-    inline __m512i operator()(__m512i a, __m512i b){
-      return _mm512_mullo_epi32(a,b);
-    }
-  };
+  inline void mac(__m512 &a, __m512 b, __m512 c){         
+    a= _mm512_fmadd_ps( b, c, a);                         
+  }
+  inline void mac(__m512d &a, __m512d b, __m512d c){
+    a= _mm512_fmadd_pd( b, c, a);                   
+  }                                             
+  // Real float
+  inline __m512 operator()(__m512 a, __m512 b){
+    return _mm512_mul_ps(a,b);
+  }
+  // Real double
+  inline __m512d operator()(__m512d a, __m512d b){
+    return _mm512_mul_pd(a,b);
+  }
+  // Integer
+  inline __m512i operator()(__m512i a, __m512i b){
+    return _mm512_mullo_epi32(a,b);
+  }
+};
 
-  struct Div{
-    // Real float
-    inline __m512 operator()(__m512 a, __m512 b){
-      return _mm512_div_ps(a,b);
-    }
-    // Real double
-    inline __m512d operator()(__m512d a, __m512d b){
-      return _mm512_div_pd(a,b);
-    }
-  };
+struct Div{
+  // Real float
+  inline __m512 operator()(__m512 a, __m512 b){
+    return _mm512_div_ps(a,b);
+  }
+  // Real double
+  inline __m512d operator()(__m512d a, __m512d b){
+    return _mm512_div_pd(a,b);
+  }
+};
 
 
-  struct Conj{
-    // Complex single
-    inline __m512 operator()(__m512 in){
-      return _mm512_mask_sub_ps(in,0xaaaa,_mm512_setzero_ps(),in); // Zero out 0+real 0-imag  
-    }
-    // Complex double
-    inline __m512d operator()(__m512d in){
-      return _mm512_mask_sub_pd(in, 0xaa,_mm512_setzero_pd(), in);
-    }
-    // do not define for integer input
-  };
+struct Conj{
+  // Complex single
+  inline __m512 operator()(__m512 in){
+    return _mm512_mask_sub_ps(in,0xaaaa,_mm512_setzero_ps(),in); // Zero out 0+real 0-imag  
+  }
+  // Complex double
+  inline __m512d operator()(__m512d in){
+    return _mm512_mask_sub_pd(in, 0xaa,_mm512_setzero_pd(), in);
+  }
+  // do not define for integer input
+};
 
-  struct TimesMinusI{
-    //Complex single
-    inline __m512 operator()(__m512 in, __m512 ret){
-      //__m512 tmp = _mm512_mask_sub_ps(in,0xaaaa,_mm512_setzero_ps(),in); // real -imag 
-      //return _mm512_shuffle_ps(tmp,tmp,_MM_SELECT_FOUR_FOUR(2,3,1,0));   // 0x4E??
-      __m512 tmp = _mm512_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
-      return _mm512_mask_sub_ps(tmp,0xaaaa,_mm512_setzero_ps(),tmp);
-    }
-    //Complex double
-    inline __m512d operator()(__m512d in, __m512d ret){
-      //__m512d tmp = _mm512_mask_sub_pd(in,0xaa,_mm512_setzero_pd(),in); // real -imag 
-      //return _mm512_shuffle_pd(tmp,tmp,0x55);
-      __m512d tmp = _mm512_shuffle_pd(in,in,0x55);
-      return _mm512_mask_sub_pd(tmp,0xaa,_mm512_setzero_pd(),tmp);
-    } 
-  };
+struct TimesMinusI{
+  //Complex single
+  inline __m512 operator()(__m512 in, __m512 ret){
+    //__m512 tmp = _mm512_mask_sub_ps(in,0xaaaa,_mm512_setzero_ps(),in); // real -imag 
+    //return _mm512_shuffle_ps(tmp,tmp,_MM_SELECT_FOUR_FOUR(2,3,1,0));   // 0x4E??
+    __m512 tmp = _mm512_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
+    return _mm512_mask_sub_ps(tmp,0xaaaa,_mm512_setzero_ps(),tmp);
+  }
+  //Complex double
+  inline __m512d operator()(__m512d in, __m512d ret){
+    //__m512d tmp = _mm512_mask_sub_pd(in,0xaa,_mm512_setzero_pd(),in); // real -imag 
+    //return _mm512_shuffle_pd(tmp,tmp,0x55);
+    __m512d tmp = _mm512_shuffle_pd(in,in,0x55);
+    return _mm512_mask_sub_pd(tmp,0xaa,_mm512_setzero_pd(),tmp);
+  } 
+};
 
-  struct TimesI{
-    //Complex single
-    inline __m512 operator()(__m512 in, __m512 ret){
-      __m512 tmp = _mm512_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
-      return _mm512_mask_sub_ps(tmp,0x5555,_mm512_setzero_ps(),tmp); 
-    }
-    //Complex double
-    inline __m512d operator()(__m512d in, __m512d ret){
-      __m512d tmp = _mm512_shuffle_pd(in,in,0x55);
-      return _mm512_mask_sub_pd(tmp,0x55,_mm512_setzero_pd(),tmp); 
-    }
-
-
-  };
+struct TimesI{
+  //Complex single
+  inline __m512 operator()(__m512 in, __m512 ret){
+    __m512 tmp = _mm512_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
+    return _mm512_mask_sub_ps(tmp,0x5555,_mm512_setzero_ps(),tmp); 
+  }
+  //Complex double
+  inline __m512d operator()(__m512d in, __m512d ret){
+    __m512d tmp = _mm512_shuffle_pd(in,in,0x55);
+    return _mm512_mask_sub_pd(tmp,0x55,_mm512_setzero_pd(),tmp); 
+  }
 
 
+};
   
-  // Gpermute utilities consider coalescing into 1 Gpermute
-  struct Permute{
+// Gpermute utilities consider coalescing into 1 Gpermute
+struct Permute{
     
-    static inline __m512 Permute0(__m512 in){
-      return _mm512_shuffle_f32x4(in,in,_MM_SELECT_FOUR_FOUR(1,0,3,2));
-    };
-    static inline __m512 Permute1(__m512 in){
-      return _mm512_shuffle_f32x4(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
-    };
-    static inline __m512 Permute2(__m512 in){
-      return _mm512_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(1,0,3,2));
-    };
-    static inline __m512 Permute3(__m512 in){
-      return _mm512_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
-    };
-
-    static inline __m512d Permute0(__m512d in){
-      return _mm512_shuffle_f64x2(in,in,_MM_SELECT_FOUR_FOUR(1,0,3,2));
-    };
-    static inline __m512d Permute1(__m512d in){
-      return _mm512_shuffle_f64x2(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
-    };
-    static inline __m512d Permute2(__m512d in){
-      return _mm512_shuffle_pd(in,in,0x55);
-    };
-    static inline __m512d Permute3(__m512d in){
-      return in;
-    };
-
+  static inline __m512 Permute0(__m512 in){
+    return _mm512_shuffle_f32x4(in,in,_MM_SELECT_FOUR_FOUR(1,0,3,2));
   };
+  static inline __m512 Permute1(__m512 in){
+    return _mm512_shuffle_f32x4(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
+  };
+  static inline __m512 Permute2(__m512 in){
+    return _mm512_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(1,0,3,2));
+  };
+  static inline __m512 Permute3(__m512 in){
+    return _mm512_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
+  };
+
+  static inline __m512d Permute0(__m512d in){
+    return _mm512_shuffle_f64x2(in,in,_MM_SELECT_FOUR_FOUR(1,0,3,2));
+  };
+  static inline __m512d Permute1(__m512d in){
+    return _mm512_shuffle_f64x2(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
+  };
+  static inline __m512d Permute2(__m512d in){
+    return _mm512_shuffle_pd(in,in,0x55);
+  };
+  static inline __m512d Permute3(__m512d in){
+    return in;
+  };
+
+};
 #define USE_FP16
-  struct PrecisionChange {
-    static inline __m512i StoH (__m512 a,__m512 b) {
-      __m512i h;
+struct PrecisionChange {
+  static inline __m512i StoH (__m512 a,__m512 b) {
+    __m512i h;
 #ifdef USE_FP16
-      __m256i ha = _mm512_cvtps_ph(a,0);
-      __m256i hb = _mm512_cvtps_ph(b,0);
-      h =(__m512i) _mm512_castps256_ps512((__m256)ha);
-      h =(__m512i) _mm512_insertf64x4((__m512d)h,(__m256d)hb,1);
+    __m256i ha = _mm512_cvtps_ph(a,0);
+    __m256i hb = _mm512_cvtps_ph(b,0);
+    h =(__m512i) _mm512_castps256_ps512((__m256)ha);
+    h =(__m512i) _mm512_insertf64x4((__m512d)h,(__m256d)hb,1);
 #else
-      assert(0);
+    assert(0);
 #endif
-      return h;
-    }
-    static inline void  HtoS (__m512i h,__m512 &sa,__m512 &sb) {
+    return h;
+  }
+
+  static inline void  HtoS (__m512i h,__m512 &sa,__m512 &sb) {
 #ifdef USE_FP16
-      sa = _mm512_cvtph_ps((__m256i)_mm512_extractf64x4_pd((__m512d)h,0));
-      sb = _mm512_cvtph_ps((__m256i)_mm512_extractf64x4_pd((__m512d)h,1));
+    sa = _mm512_cvtph_ps((__m256i)_mm512_extractf64x4_pd((__m512d)h,0));
+    sb = _mm512_cvtph_ps((__m256i)_mm512_extractf64x4_pd((__m512d)h,1));
 #else
-      assert(0);
+    assert(0);
 #endif
-    }
-    static inline __m512 DtoS (__m512d a,__m512d b) {
-      __m256 sa = _mm512_cvtpd_ps(a);
-      __m256 sb = _mm512_cvtpd_ps(b);
-      __m512 s = _mm512_castps256_ps512(sa);
-      s =(__m512) _mm512_insertf64x4((__m512d)s,(__m256d)sb,1);
-      return s;
-    }
-    static inline void StoD (__m512 s,__m512d &a,__m512d &b) {
-      a = _mm512_cvtps_pd((__m256)_mm512_extractf64x4_pd((__m512d)s,0));
-      b = _mm512_cvtps_pd((__m256)_mm512_extractf64x4_pd((__m512d)s,1));
-    }
-    static inline __m512i DtoH (__m512d a,__m512d b,__m512d c,__m512d d) {
-      __m512 sa,sb;
-      sa = DtoS(a,b);
-      sb = DtoS(c,d);
-      return StoH(sa,sb);
-    }
-    static inline void HtoD (__m512i h,__m512d &a,__m512d &b,__m512d &c,__m512d &d) {
-      __m512 sa,sb;
-      HtoS(h,sa,sb);
-      StoD(sa,a,b);
-      StoD(sb,c,d);
-    }
+  }
+
+  static inline __m512 DtoS (__m512d a,__m512d b) {
+    __m256 sa = _mm512_cvtpd_ps(a);
+    __m256 sb = _mm512_cvtpd_ps(b);
+    __m512 s = _mm512_castps256_ps512(sa);
+    s =(__m512) _mm512_insertf64x4((__m512d)s,(__m256d)sb,1);
+    return s;
+  }
+
+  static inline void StoD (__m512 s,__m512d &a,__m512d &b) {
+    a = _mm512_cvtps_pd((__m256)_mm512_extractf64x4_pd((__m512d)s,0));
+    b = _mm512_cvtps_pd((__m256)_mm512_extractf64x4_pd((__m512d)s,1));
+  }
+
+  static inline __m512i DtoH (__m512d a,__m512d b,__m512d c,__m512d d) {
+    __m512 sa,sb;
+    sa = DtoS(a,b);
+    sb = DtoS(c,d);
+    return StoH(sa,sb);
+  }
+
+  static inline void HtoD (__m512i h,__m512d &a,__m512d &b,__m512d &c,__m512d &d) {
+    __m512 sa,sb;
+    HtoS(h,sa,sb);
+    StoD(sa,a,b);
+    StoD(sb,c,d);
+  }
+};
+// On extracting face: Ah Al , Bh Bl -> Ah Bh, Al Bl
+// On merging buffers: Ah,Bh , Al Bl -> Ah Al, Bh, Bl
+// The operation is its own inverse
+struct Exchange{
+  // 3210 ordering
+  static inline void Exchange0(__m512 &out1,__m512 &out2,__m512 in1,__m512 in2){
+    out1= _mm512_shuffle_f32x4(in1,in2,_MM_SELECT_FOUR_FOUR(1,0,1,0));
+    out2= _mm512_shuffle_f32x4(in1,in2,_MM_SELECT_FOUR_FOUR(3,2,3,2));
+  };
+  static inline void Exchange1(__m512 &out1,__m512 &out2,__m512 in1,__m512 in2){
+    out1= _mm512_shuffle_f32x4(in1,in2,_MM_SELECT_FOUR_FOUR(2,0,2,0));
+    out2= _mm512_shuffle_f32x4(in1,in2,_MM_SELECT_FOUR_FOUR(3,1,3,1));
+    out1= _mm512_shuffle_f32x4(out1,out1,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
+    out2= _mm512_shuffle_f32x4(out2,out2,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
+  };
+  static inline void Exchange2(__m512 &out1,__m512 &out2,__m512 in1,__m512 in2){
+    out1= _mm512_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(1,0,1,0));
+    out2= _mm512_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(3,2,3,2));
+  };
+  static inline void Exchange3(__m512 &out1,__m512 &out2,__m512 in1,__m512 in2){
+    out1= _mm512_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(2,0,2,0));
+    out2= _mm512_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(3,1,3,1));
+    out1= _mm512_shuffle_ps(out1,out1,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
+    out2= _mm512_shuffle_ps(out2,out2,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
   };
-  // On extracting face: Ah Al , Bh Bl -> Ah Bh, Al Bl
-  // On merging buffers: Ah,Bh , Al Bl -> Ah Al, Bh, Bl
-  // The operation is its own inverse
-  struct Exchange{
-    // 3210 ordering
-    static inline void Exchange0(__m512 &out1,__m512 &out2,__m512 in1,__m512 in2){
-      out1= _mm512_shuffle_f32x4(in1,in2,_MM_SELECT_FOUR_FOUR(1,0,1,0));
-      out2= _mm512_shuffle_f32x4(in1,in2,_MM_SELECT_FOUR_FOUR(3,2,3,2));
-    };
-    static inline void Exchange1(__m512 &out1,__m512 &out2,__m512 in1,__m512 in2){
-      out1= _mm512_shuffle_f32x4(in1,in2,_MM_SELECT_FOUR_FOUR(2,0,2,0));
-      out2= _mm512_shuffle_f32x4(in1,in2,_MM_SELECT_FOUR_FOUR(3,1,3,1));
-      out1= _mm512_shuffle_f32x4(out1,out1,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
-      out2= _mm512_shuffle_f32x4(out2,out2,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
-    };
-    static inline void Exchange2(__m512 &out1,__m512 &out2,__m512 in1,__m512 in2){
-      out1= _mm512_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(1,0,1,0));
-      out2= _mm512_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(3,2,3,2));
-    };
-    static inline void Exchange3(__m512 &out1,__m512 &out2,__m512 in1,__m512 in2){
-      out1= _mm512_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(2,0,2,0));
-      out2= _mm512_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(3,1,3,1));
-      out1= _mm512_shuffle_ps(out1,out1,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
-      out2= _mm512_shuffle_ps(out2,out2,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
-    };
  
-    static inline void Exchange0(__m512d &out1,__m512d &out2,__m512d in1,__m512d in2){
-      out1= _mm512_shuffle_f64x2(in1,in2,_MM_SELECT_FOUR_FOUR(1,0,1,0));
-      out2= _mm512_shuffle_f64x2(in1,in2,_MM_SELECT_FOUR_FOUR(3,2,3,2));
-    };
-    static inline void Exchange1(__m512d &out1,__m512d &out2,__m512d in1,__m512d in2){
-      out1= _mm512_shuffle_f64x2(in1,in2,_MM_SELECT_FOUR_FOUR(2,0,2,0));
-      out2= _mm512_shuffle_f64x2(in1,in2,_MM_SELECT_FOUR_FOUR(3,1,3,1));
-      out1= _mm512_shuffle_f64x2(out1,out1,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
-      out2= _mm512_shuffle_f64x2(out2,out2,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
-    };
-    static inline void Exchange2(__m512d &out1,__m512d &out2,__m512d in1,__m512d in2){
-      out1 = _mm512_shuffle_pd(in1,in2,0x00);
-      out2 = _mm512_shuffle_pd(in1,in2,0xFF);
-    };
-    static inline void Exchange3(__m512d &out1,__m512d &out2,__m512d in1,__m512d in2){
-      assert(0);
-      return;
-    };
+  static inline void Exchange0(__m512d &out1,__m512d &out2,__m512d in1,__m512d in2){
+    out1= _mm512_shuffle_f64x2(in1,in2,_MM_SELECT_FOUR_FOUR(1,0,1,0));
+    out2= _mm512_shuffle_f64x2(in1,in2,_MM_SELECT_FOUR_FOUR(3,2,3,2));
+  };
+  static inline void Exchange1(__m512d &out1,__m512d &out2,__m512d in1,__m512d in2){
+    out1= _mm512_shuffle_f64x2(in1,in2,_MM_SELECT_FOUR_FOUR(2,0,2,0));
+    out2= _mm512_shuffle_f64x2(in1,in2,_MM_SELECT_FOUR_FOUR(3,1,3,1));
+    out1= _mm512_shuffle_f64x2(out1,out1,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
+    out2= _mm512_shuffle_f64x2(out2,out2,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
+  };
+  static inline void Exchange2(__m512d &out1,__m512d &out2,__m512d in1,__m512d in2){
+    out1 = _mm512_shuffle_pd(in1,in2,0x00);
+    out2 = _mm512_shuffle_pd(in1,in2,0xFF);
+  };
+  static inline void Exchange3(__m512d &out1,__m512d &out2,__m512d in1,__m512d in2){
+    assert(0);
+    return;
+  };
+};
+
+
+struct Rotate{
+
+  static inline __m512 rotate(__m512 in,int n){ 
+    switch(n){
+    case 0: return tRotate<0>(in);break;
+    case 1: return tRotate<1>(in);break;
+    case 2: return tRotate<2>(in);break;
+    case 3: return tRotate<3>(in);break;
+    case 4: return tRotate<4>(in);break;
+    case 5: return tRotate<5>(in);break;
+    case 6: return tRotate<6>(in);break;
+    case 7: return tRotate<7>(in);break;
+
+    case 8 : return tRotate<8>(in);break;
+    case 9 : return tRotate<9>(in);break;
+    case 10: return tRotate<10>(in);break;
+    case 11: return tRotate<11>(in);break;
+    case 12: return tRotate<12>(in);break;
+    case 13: return tRotate<13>(in);break;
+    case 14: return tRotate<14>(in);break;
+    case 15: return tRotate<15>(in);break;
+    default: assert(0);
+    }
+  }
+  static inline __m512d rotate(__m512d in,int n){ 
+    switch(n){
+    case 0: return tRotate<0>(in);break;
+    case 1: return tRotate<1>(in);break;
+    case 2: return tRotate<2>(in);break;
+    case 3: return tRotate<3>(in);break;
+    case 4: return tRotate<4>(in);break;
+    case 5: return tRotate<5>(in);break;
+    case 6: return tRotate<6>(in);break;
+    case 7: return tRotate<7>(in);break;
+    default: assert(0);
+    }
+  }
+
+  template<int n> static inline __m512 tRotate(__m512 in){ 
+    return (__m512)_mm512_alignr_epi32((__m512i)in,(__m512i)in,n);          
   };
 
-
-  struct Rotate{
-
-    static inline __m512 rotate(__m512 in,int n){ 
-      switch(n){
-      case 0: return tRotate<0>(in);break;
-      case 1: return tRotate<1>(in);break;
-      case 2: return tRotate<2>(in);break;
-      case 3: return tRotate<3>(in);break;
-      case 4: return tRotate<4>(in);break;
-      case 5: return tRotate<5>(in);break;
-      case 6: return tRotate<6>(in);break;
-      case 7: return tRotate<7>(in);break;
-
-      case 8 : return tRotate<8>(in);break;
-      case 9 : return tRotate<9>(in);break;
-      case 10: return tRotate<10>(in);break;
-      case 11: return tRotate<11>(in);break;
-      case 12: return tRotate<12>(in);break;
-      case 13: return tRotate<13>(in);break;
-      case 14: return tRotate<14>(in);break;
-      case 15: return tRotate<15>(in);break;
-      default: assert(0);
-      }
-    }
-    static inline __m512d rotate(__m512d in,int n){ 
-      switch(n){
-      case 0: return tRotate<0>(in);break;
-      case 1: return tRotate<1>(in);break;
-      case 2: return tRotate<2>(in);break;
-      case 3: return tRotate<3>(in);break;
-      case 4: return tRotate<4>(in);break;
-      case 5: return tRotate<5>(in);break;
-      case 6: return tRotate<6>(in);break;
-      case 7: return tRotate<7>(in);break;
-      default: assert(0);
-      }
-    }
-
-    template<int n> static inline __m512 tRotate(__m512 in){ 
-      return (__m512)_mm512_alignr_epi32((__m512i)in,(__m512i)in,n);          
-    };
-
-    template<int n> static inline __m512d tRotate(__m512d in){ 
-      return (__m512d)_mm512_alignr_epi64((__m512i)in,(__m512i)in,n);          
-    };
-
+  template<int n> static inline __m512d tRotate(__m512d in){ 
+    return (__m512d)_mm512_alignr_epi64((__m512i)in,(__m512i)in,n);          
   };
 
-  //////////////////////////////////////////////
-  // Some Template specialization
+};
 
-  // Hack for CLANG until mm512_reduce_add_ps etc... are implemented in GCC and Clang releases
-  //Complex float Reduce
-  template<>
-  inline Grid::ComplexF Reduce<Grid::ComplexF, __m512>::operator()(__m512 in){
-    return Grid::ComplexF(_mm512_mask_reduce_add_ps(0x5555, in),_mm512_mask_reduce_add_ps(0xAAAA, in));
-  }
-  //Real float Reduce
-  template<>
-  inline Grid::RealF Reduce<Grid::RealF, __m512>::operator()(__m512 in){
-    return _mm512_reduce_add_ps(in);
-  }
-  
-  //Complex double Reduce
-  template<>
-  inline Grid::ComplexD Reduce<Grid::ComplexD, __m512d>::operator()(__m512d in){
-    return Grid::ComplexD(_mm512_mask_reduce_add_pd(0x55, in),_mm512_mask_reduce_add_pd(0xAA, in));
-  }
-  
-  //Real double Reduce
-  template<>
-  inline Grid::RealD Reduce<Grid::RealD, __m512d>::operator()(__m512d in){
-    return _mm512_reduce_add_pd(in);
-  }
+//////////////////////////////////////////////
+// Some Template specialization
 
-  //Integer Reduce
-  template<>
-  inline Integer Reduce<Integer, __m512i>::operator()(__m512i in){
-    return _mm512_reduce_add_epi32(in);
-  }
-  
+// Hack for CLANG until mm512_reduce_add_ps etc... are implemented in GCC and Clang releases
+//Complex float Reduce
+template<>
+inline Grid::ComplexF Reduce<Grid::ComplexF, __m512>::operator()(__m512 in){
+  return Grid::ComplexF(_mm512_mask_reduce_add_ps(0x5555, in),_mm512_mask_reduce_add_ps(0xAAAA, in));
 }
+//Real float Reduce
+template<>
+inline Grid::RealF Reduce<Grid::RealF, __m512>::operator()(__m512 in){
+  return _mm512_reduce_add_ps(in);
+}
+  
+//Complex double Reduce
+template<>
+inline Grid::ComplexD Reduce<Grid::ComplexD, __m512d>::operator()(__m512d in){
+  return Grid::ComplexD(_mm512_mask_reduce_add_pd(0x55, in),_mm512_mask_reduce_add_pd(0xAA, in));
+}
+  
+//Real double Reduce
+template<>
+inline Grid::RealD Reduce<Grid::RealD, __m512d>::operator()(__m512d in){
+  return _mm512_reduce_add_pd(in);
+}
+
+//Integer Reduce
+template<>
+inline Integer Reduce<Integer, __m512i>::operator()(__m512i in){
+  return _mm512_reduce_add_epi32(in);
+}
+  
+NAMESPACE_END(Optimization);
 
 //////////////////////////////////////////////////////////////////////////////////////
 // Here assign types 
 
+typedef __m512i SIMD_Htype;  // Single precision type
+typedef __m512  SIMD_Ftype;  // Single precision type
+typedef __m512d SIMD_Dtype; // Double precision type
+typedef __m512i SIMD_Itype; // Integer type
 
-  typedef __m512i SIMD_Htype;  // Single precision type
-  typedef __m512  SIMD_Ftype;  // Single precision type
-  typedef __m512d SIMD_Dtype; // Double precision type
-  typedef __m512i SIMD_Itype; // Integer type
-
-  // prefecth
-  inline void v_prefetch0(int size, const char *ptr){
-    for(int i=0;i<size;i+=64){ //  Define L1 linesize above
-      _mm_prefetch(ptr+i+4096,_MM_HINT_T1);
-      _mm_prefetch(ptr+i+512,_MM_HINT_T0);
-    }
+// prefecth
+inline void v_prefetch0(int size, const char *ptr){
+  for(int i=0;i<size;i+=64){ //  Define L1 linesize above
+    _mm_prefetch(ptr+i+4096,_MM_HINT_T1);
+    _mm_prefetch(ptr+i+512,_MM_HINT_T0);
   }
-  inline void prefetch_HINT_T0(const char *ptr){
-    _mm_prefetch(ptr,_MM_HINT_T0);
-  }
-
-
-  
-  // Function name aliases
-  typedef Optimization::Vsplat   VsplatSIMD;
-  typedef Optimization::Vstore   VstoreSIMD;
-  typedef Optimization::Vset     VsetSIMD;
-  typedef Optimization::Vstream  VstreamSIMD;
-  template <typename S, typename T> using ReduceSIMD = Optimization::Reduce<S,T>;
-
-
-  // Arithmetic operations
-  typedef Optimization::Sum         SumSIMD;
-  typedef Optimization::Sub         SubSIMD;
-  typedef Optimization::Mult        MultSIMD;
-  typedef Optimization::Div         DivSIMD;
-  typedef Optimization::MultComplex MultComplexSIMD;
-  typedef Optimization::MultRealPart MultRealPartSIMD;
-  typedef Optimization::MaddRealPart MaddRealPartSIMD;
-  typedef Optimization::Conj        ConjSIMD;
-  typedef Optimization::TimesMinusI TimesMinusISIMD;
-  typedef Optimization::TimesI      TimesISIMD;
-
 }
+inline void prefetch_HINT_T0(const char *ptr){
+  _mm_prefetch(ptr,_MM_HINT_T0);
+}
+  
+// Function name aliases
+typedef Optimization::Vsplat   VsplatSIMD;
+typedef Optimization::Vstore   VstoreSIMD;
+typedef Optimization::Vset     VsetSIMD;
+typedef Optimization::Vstream  VstreamSIMD;
+template <typename S, typename T> using ReduceSIMD = Optimization::Reduce<S,T>;
+
+// Arithmetic operations
+typedef Optimization::Sum         SumSIMD;
+typedef Optimization::Sub         SubSIMD;
+typedef Optimization::Mult        MultSIMD;
+typedef Optimization::Div         DivSIMD;
+typedef Optimization::MultComplex MultComplexSIMD;
+typedef Optimization::MultRealPart MultRealPartSIMD;
+typedef Optimization::MaddRealPart MaddRealPartSIMD;
+typedef Optimization::Conj        ConjSIMD;
+typedef Optimization::TimesMinusI TimesMinusISIMD;
+typedef Optimization::TimesI      TimesISIMD;
+
+NAMESPACE_END(Grid);
diff --git a/Grid/simd/Grid_generic.h b/Grid/simd/Grid_generic.h
index e1d5f894..35cb3bc3 100644
--- a/Grid/simd/Grid_generic.h
+++ b/Grid/simd/Grid_generic.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,506 +25,508 @@ Author: Antonin Portelli <antonin.portelli@me.com>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 
 #include "Grid_generic_types.h"
 
-namespace Grid {
-namespace Optimization {
+NAMESPACE_BEGIN(Grid);
+NAMESPACE_BEGIN(Optimization);
   
-  struct Vsplat{
-    // Complex
-    template <typename T>
-    inline vec<T> operator()(T a, T b){
-      vec<T> out;
+struct Vsplat{
+  // Complex
+  template <typename T>
+  accelerator_inline vec<T> operator()(T a, T b){
+    vec<T> out;
       
-      VECTOR_FOR(i, W<T>::r, 2)
+    VECTOR_FOR(i, W<T>::r, 2)
       {
         out.v[i]   = a;
         out.v[i+1] = b;
       }
 
-      return out;
-    }
+    return out;
+  }
     
-    // Real
-    template <typename T>
-    inline vec<T> operator()(T a){
-      vec<T> out;
+  // Real
+  template <typename T>
+  accelerator_inline vec<T> operator()(T a){
+    vec<T> out;
       
-      VECTOR_FOR(i, W<T>::r, 1)
+    VECTOR_FOR(i, W<T>::r, 1)
       {
         out.v[i] = a;
       }
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
 
-  struct Vstore{
-    // Real
-    template <typename T>
-    inline void operator()(vec<T> a, T *D){
-      *((vec<T> *)D) = a;
-    }
-  };
+struct Vstore{
+  // Real
+  template <typename T>
+  accelerator_inline void operator()(vec<T> a, T *D){
+    *((vec<T> *)D) = a;
+  }
+};
 
-  struct Vstream{
-    // Real
-    template <typename T>
-    inline void operator()(T * a, vec<T> b){
-      *((vec<T> *)a) = b;
-    }
-  };
+struct Vstream{
+  // Real
+  template <typename T>
+  accelerator_inline void operator()(T * a, vec<T> b){
+    *((vec<T> *)a) = b;
+  }
+};
 
-  struct Vset{
-    // Complex
-    template <typename T>
-    inline vec<T> operator()(std::complex<T> *a){
-      vec<T> out;
+struct Vset{
+  // Complex
+  template <typename T>
+  accelerator_inline vec<T> operator()(std::complex<T> *a){
+    vec<T> out;
       
-      VECTOR_FOR(i, W<T>::c, 1)
+    VECTOR_FOR(i, W<T>::c, 1)
       {
         out.v[2*i]   = a[i].real();
         out.v[2*i+1] = a[i].imag();
       }
       
-      return out;
-    }
+    return out;
+  }
     
-    // Real
-    template <typename T>
-    inline vec<T> operator()(T *a){
-      vec<T> out;
+  // Real
+  template <typename T>
+  accelerator_inline vec<T> operator()(T *a){
+    vec<T> out;
       
-      out = *((vec<T> *)a);
+    out = *((vec<T> *)a);
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
 
-  /////////////////////////////////////////////////////
-  // Arithmetic operations
-  /////////////////////////////////////////////////////
-  struct Sum{
-    // Complex/Real
-    template <typename T>
-    inline vec<T> operator()(vec<T> a, vec<T> b){
-      vec<T> out;
+/////////////////////////////////////////////////////
+// Arithmetic operations
+/////////////////////////////////////////////////////
+struct Sum{
+  // Complex/Real
+  template <typename T>
+  accelerator_inline vec<T> operator()(vec<T> a, vec<T> b){
+    vec<T> out;
       
-      VECTOR_FOR(i, W<T>::r, 1)
+    VECTOR_FOR(i, W<T>::r, 1)
       {
         out.v[i] = a.v[i] + b.v[i];
       }
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
 
-  struct Sub{
-    // Complex/Real
-    template <typename T>
-    inline vec<T> operator()(vec<T> a, vec<T> b){
-      vec<T> out;
+struct Sub{
+  // Complex/Real
+  template <typename T>
+  accelerator_inline vec<T> operator()(vec<T> a, vec<T> b){
+    vec<T> out;
       
-      VECTOR_FOR(i, W<T>::r, 1)
+    VECTOR_FOR(i, W<T>::r, 1)
       {
         out.v[i] = a.v[i] - b.v[i];
       }
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
 
-  struct Mult{
-    // Real
-    template <typename T>
-    inline vec<T> operator()(vec<T> a, vec<T> b){
-      vec<T> out;
+struct Mult{
+  // Real
+  template <typename T>
+  accelerator_inline vec<T> operator()(vec<T> a, vec<T> b){
+    vec<T> out;
       
-      VECTOR_FOR(i, W<T>::r, 1)
+    VECTOR_FOR(i, W<T>::r, 1)
       {
         out.v[i] = a.v[i]*b.v[i];
       }
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
   
-  #define cmul(a, b, c, i)\
-  c[i]   = a[i]*b[i]   - a[i+1]*b[i+1];\
+#define cmul(a, b, c, i)			\
+  c[i]   = a[i]*b[i]   - a[i+1]*b[i+1];		\
   c[i+1] = a[i]*b[i+1] + a[i+1]*b[i];
 
-  struct MultRealPart{
-    template <typename T>
-    inline vec<T> operator()(vec<T> a, vec<T> b){
-      vec<T> out;
+struct MultRealPart{
+  template <typename T>
+  accelerator_inline vec<T> operator()(vec<T> a, vec<T> b){
+    vec<T> out;
       
-      VECTOR_FOR(i, W<T>::c, 1)
+    VECTOR_FOR(i, W<T>::c, 1)
       {
-         out.v[2*i]   = a.v[2*i]*b.v[2*i];
-         out.v[2*i+1] = a.v[2*i]*b.v[2*i+1];
+	out.v[2*i]   = a.v[2*i]*b.v[2*i];
+	out.v[2*i+1] = a.v[2*i]*b.v[2*i+1];
       }      
-      return out;
-    }
-  };
+    return out;
+  }
+};
 
-  struct MaddRealPart{
-    template <typename T>
-    inline vec<T> operator()(vec<T> a, vec<T> b, vec<T> c){
-      vec<T> out;
+struct MaddRealPart{
+  template <typename T>
+  accelerator_inline vec<T> operator()(vec<T> a, vec<T> b, vec<T> c){
+    vec<T> out;
       
-      VECTOR_FOR(i, W<T>::c, 1)
+    VECTOR_FOR(i, W<T>::c, 1)
       {
-         out.v[2*i]   = a.v[2*i]*b.v[2*i] + c.v[2*i];
-         out.v[2*i+1] = a.v[2*i]*b.v[2*i+1] + c.v[2*i+1];
+	out.v[2*i]   = a.v[2*i]*b.v[2*i] + c.v[2*i];
+	out.v[2*i+1] = a.v[2*i]*b.v[2*i+1] + c.v[2*i+1];
       }      
-      return out;
-    }
-  };
+    return out;
+  }
+};
   
-  struct MultComplex{
-    // Complex
-    template <typename T>
-    inline vec<T> operator()(vec<T> a, vec<T> b){
-      vec<T> out;
+struct MultComplex{
+  // Complex
+  template <typename T>
+  accelerator_inline vec<T> operator()(vec<T> a, vec<T> b){
+    vec<T> out;
       
-      VECTOR_FOR(i, W<T>::c, 1)
+    VECTOR_FOR(i, W<T>::c, 1)
       {
         cmul(a.v, b.v, out.v, 2*i);
       }      
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
   
-  #undef cmul
+#undef cmul
 
-  struct Div{
-    // Real
-    template <typename T>
-    inline vec<T> operator()(vec<T> a, vec<T> b){
-      vec<T> out;
+struct Div{
+  // Real
+  template <typename T>
+  accelerator_inline vec<T> operator()(vec<T> a, vec<T> b){
+    vec<T> out;
       
-      VECTOR_FOR(i, W<T>::r, 1)
+    VECTOR_FOR(i, W<T>::r, 1)
       {
         out.v[i] = a.v[i]/b.v[i];
       }
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
   
-  #define conj(a, b, i)\
-  b[i]   = a[i];\
+#define conj(a, b, i)				\
+  b[i]   = a[i];				\
   b[i+1] = -a[i+1];
   
-  struct Conj{
-    // Complex
-    template <typename T>
-    inline vec<T> operator()(vec<T> a){
-      vec<T> out;
+struct Conj{
+  // Complex
+  template <typename T>
+  accelerator_inline vec<T> operator()(vec<T> a){
+    vec<T> out;
       
-      VECTOR_FOR(i, W<T>::c, 1)
+    VECTOR_FOR(i, W<T>::c, 1)
       {
         conj(a.v, out.v, 2*i);
       }
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
   
-  #undef conj
+#undef conj
 
-  #define timesmi(a, b, i)\
-  b[i]   = a[i+1];\
+#define timesmi(a, b, i)			\
+  b[i]   = a[i+1];				\
   b[i+1] = -a[i];
   
-  struct TimesMinusI{
-    // Complex
-    template <typename T>
-    inline vec<T> operator()(vec<T> a, vec<T> b){
-      vec<T> out;
+struct TimesMinusI{
+  // Complex
+  template <typename T>
+  accelerator_inline vec<T> operator()(vec<T> a, vec<T> b){
+    vec<T> out;
       
-      VECTOR_FOR(i, W<T>::c, 1)
+    VECTOR_FOR(i, W<T>::c, 1)
       {
         timesmi(a.v, out.v, 2*i);
       }
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
 
-  #undef timesmi
+#undef timesmi
   
-  #define timesi(a, b, i)\
-  b[i]   = -a[i+1];\
+#define timesi(a, b, i)				\
+  b[i]   = -a[i+1];				\
   b[i+1] = a[i];
   
-  struct TimesI{
-    // Complex
-    template <typename T>
-    inline vec<T> operator()(vec<T> a, vec<T> b){
-      vec<T> out;
+struct TimesI{
+  // Complex
+  template <typename T>
+  accelerator_inline vec<T> operator()(vec<T> a, vec<T> b){
+    vec<T> out;
       
-      VECTOR_FOR(i, W<T>::c, 1)
+    VECTOR_FOR(i, W<T>::c, 1)
       {
         timesi(a.v, out.v, 2*i);
       }
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
   
-  #undef timesi
+#undef timesi
 
-  struct PrecisionChange {
-    static inline vech StoH (const vecf &a,const vecf &b) {
-      vech ret;
+struct PrecisionChange {
+  static accelerator_inline vech StoH (const vecf &a,const vecf &b) {
+    vech ret; 
+    const int nf = W<float>::r;
 #ifdef USE_FP16
-      vech *ha = (vech *)&a;
-      vech *hb = (vech *)&b;
-      const int nf = W<float>::r;
-      //      VECTOR_FOR(i, nf,1){ ret.v[i]    = ( (uint16_t *) &a.v[i])[1] ; }
-      //      VECTOR_FOR(i, nf,1){ ret.v[i+nf] = ( (uint16_t *) &b.v[i])[1] ; }
-      VECTOR_FOR(i, nf,1){ ret.v[i]    = ha->v[2*i+1]; }
-      VECTOR_FOR(i, nf,1){ ret.v[i+nf] = hb->v[2*i+1]; }
+    vech *ha = (vech *)&a;
+    vech *hb = (vech *)&b;
+    //      VECTOR_FOR(i, nf,1){ ret.v[i]    = ( (uint16_t *) &a.v[i])[1] ; }
+    //      VECTOR_FOR(i, nf,1){ ret.v[i+nf] = ( (uint16_t *) &b.v[i])[1] ; }
+    VECTOR_FOR(i, nf,1){ ret.v[i]    = ha->v[2*i+1]; }
+    VECTOR_FOR(i, nf,1){ ret.v[i+nf] = hb->v[2*i+1]; }
 #else
-      assert(0);
+    VECTOR_FOR(i, nf,1){ ret.v[i]=0; }
+    assert(0);
 #endif
-      return ret;
-    }
-    static inline void  HtoS (vech h,vecf &sa,vecf &sb) {
+    return ret;
+  }
+  static accelerator_inline void  HtoS (vech h,vecf &sa,vecf &sb) {
 #ifdef USE_FP16
-      const int nf = W<float>::r;
-      const int nh = W<uint16_t>::r;
-      vech *ha = (vech *)&sa;
-      vech *hb = (vech *)&sb;
-      VECTOR_FOR(i, nf, 1){ sb.v[i]= sa.v[i] = 0; }
-      //      VECTOR_FOR(i, nf, 1){ ( (uint16_t *) (&sa.v[i]))[1] = h.v[i];}
-      //      VECTOR_FOR(i, nf, 1){ ( (uint16_t *) (&sb.v[i]))[1] = h.v[i+nf];}
-      VECTOR_FOR(i, nf, 1){ ha->v[2*i+1]=h.v[i]; }
-      VECTOR_FOR(i, nf, 1){ hb->v[2*i+1]=h.v[i+nf]; }
+    const int nf = W<float>::r;
+    const int nh = W<uint16_t>::r;
+    vech *ha = (vech *)&sa;
+    vech *hb = (vech *)&sb;
+    VECTOR_FOR(i, nf, 1){ sb.v[i]= sa.v[i] = 0; }
+    //      VECTOR_FOR(i, nf, 1){ ( (uint16_t *) (&sa.v[i]))[1] = h.v[i];}
+    //      VECTOR_FOR(i, nf, 1){ ( (uint16_t *) (&sb.v[i]))[1] = h.v[i+nf];}
+    VECTOR_FOR(i, nf, 1){ ha->v[2*i+1]=h.v[i]; }
+    VECTOR_FOR(i, nf, 1){ hb->v[2*i+1]=h.v[i+nf]; }
 #else
-      assert(0);
+    assert(0);
 #endif
-    }
-    static inline vecf DtoS (vecd a,vecd b) {
-      const int nd = W<double>::r;
-      const int nf = W<float>::r;
-      vecf ret;
-      VECTOR_FOR(i, nd,1){ ret.v[i]    = a.v[i] ; }
-      VECTOR_FOR(i, nd,1){ ret.v[i+nd] = b.v[i] ; }
-      return ret;
-    }
-    static inline void StoD (vecf s,vecd &a,vecd &b) {
-      const int nd = W<double>::r;
-      VECTOR_FOR(i, nd,1){ a.v[i] = s.v[i] ; }
-      VECTOR_FOR(i, nd,1){ b.v[i] = s.v[i+nd] ; }
-    }
-    static inline vech DtoH (vecd a,vecd b,vecd c,vecd d) {
-      vecf sa,sb;
-      sa = DtoS(a,b);
-      sb = DtoS(c,d);
-      return StoH(sa,sb);
-    }
-    static inline void HtoD (vech h,vecd &a,vecd &b,vecd &c,vecd &d) {
-      vecf sa,sb;
-      HtoS(h,sa,sb);
-      StoD(sa,a,b);
-      StoD(sb,c,d);
-    }
+  }
+  static accelerator_inline vecf DtoS (vecd a,vecd b) {
+    const int nd = W<double>::r;
+    vecf ret;
+    VECTOR_FOR(i, nd,1){ ret.v[i]    = a.v[i] ; }
+    VECTOR_FOR(i, nd,1){ ret.v[i+nd] = b.v[i] ; }
+    return ret;
+  }
+  static accelerator_inline void StoD (vecf s,vecd &a,vecd &b) {
+    const int nd = W<double>::r;
+    VECTOR_FOR(i, nd,1){ a.v[i] = s.v[i] ; }
+    VECTOR_FOR(i, nd,1){ b.v[i] = s.v[i+nd] ; }
+  }
+  static accelerator_inline vech DtoH (vecd a,vecd b,vecd c,vecd d) {
+    vecf sa,sb;
+    sa = DtoS(a,b);
+    sb = DtoS(c,d);
+    return StoH(sa,sb);
+  }
+  static accelerator_inline void HtoD (vech h,vecd &a,vecd &b,vecd &c,vecd &d) {
+    vecf sa,sb;
+    HtoS(h,sa,sb);
+    StoD(sa,a,b);
+    StoD(sb,c,d);
+  }
+};
+
+//////////////////////////////////////////////
+// Exchange support
+struct Exchange{
+
+  template <typename T,int n>
+  static accelerator_inline void ExchangeN(vec<T> &out1,vec<T> &out2,vec<T> &in1,vec<T> &in2){
+    const int w = W<T>::r;
+    unsigned int mask = w >> (n + 1);
+    //      std::cout << " Exchange "<<n<<" nsimd "<<w<<" mask 0x" <<std::hex<<mask<<std::dec<<std::endl;
+    VECTOR_FOR(i, w, 1) {	
+      int j1 = i&(~mask);
+      if  ( (i&mask) == 0 ) { out1.v[i]=in1.v[j1];}
+      else                  { out1.v[i]=in2.v[j1];}
+      int j2 = i|mask;
+      if  ( (i&mask) == 0 ) { out2.v[i]=in1.v[j2];}
+      else                  { out2.v[i]=in2.v[j2];}
+    }      
+  }
+  template <typename T>
+  static accelerator_inline void Exchange0(vec<T> &out1,vec<T> &out2,vec<T> &in1,vec<T> &in2){
+    ExchangeN<T,0>(out1,out2,in1,in2);
   };
-
-  //////////////////////////////////////////////
-  // Exchange support
-  struct Exchange{
-
-    template <typename T,int n>
-    static inline void ExchangeN(vec<T> &out1,vec<T> &out2,vec<T> &in1,vec<T> &in2){
-      const int w = W<T>::r;
-      unsigned int mask = w >> (n + 1);
-      //      std::cout << " Exchange "<<n<<" nsimd "<<w<<" mask 0x" <<std::hex<<mask<<std::dec<<std::endl;
-      VECTOR_FOR(i, w, 1) {	
-	int j1 = i&(~mask);
-	if  ( (i&mask) == 0 ) { out1.v[i]=in1.v[j1];}
-	else                  { out1.v[i]=in2.v[j1];}
-	int j2 = i|mask;
-	if  ( (i&mask) == 0 ) { out2.v[i]=in1.v[j2];}
-	else                  { out2.v[i]=in2.v[j2];}
-      }      
-    }
-    template <typename T>
-    static inline void Exchange0(vec<T> &out1,vec<T> &out2,vec<T> &in1,vec<T> &in2){
-      ExchangeN<T,0>(out1,out2,in1,in2);
-    };
-    template <typename T>
-    static inline void Exchange1(vec<T> &out1,vec<T> &out2,vec<T> &in1,vec<T> &in2){
-      ExchangeN<T,1>(out1,out2,in1,in2);
-    };
-    template <typename T>
-    static inline void Exchange2(vec<T> &out1,vec<T> &out2,vec<T> &in1,vec<T> &in2){
-      ExchangeN<T,2>(out1,out2,in1,in2);
-    };
-    template <typename T>
-    static inline void Exchange3(vec<T> &out1,vec<T> &out2,vec<T> &in1,vec<T> &in2){
-      ExchangeN<T,3>(out1,out2,in1,in2);
-    };
+  template <typename T>
+  static accelerator_inline void Exchange1(vec<T> &out1,vec<T> &out2,vec<T> &in1,vec<T> &in2){
+    ExchangeN<T,1>(out1,out2,in1,in2);
   };
+  template <typename T>
+  static accelerator_inline void Exchange2(vec<T> &out1,vec<T> &out2,vec<T> &in1,vec<T> &in2){
+    ExchangeN<T,2>(out1,out2,in1,in2);
+  };
+  template <typename T>
+  static accelerator_inline void Exchange3(vec<T> &out1,vec<T> &out2,vec<T> &in1,vec<T> &in2){
+    ExchangeN<T,3>(out1,out2,in1,in2);
+  };
+};
 
 
-  //////////////////////////////////////////////
-  // Some Template specialization
-  #define perm(a, b, n, w)\
-  unsigned int _mask = w >> (n + 1);\
-  VECTOR_FOR(i, w, 1)\
-  {\
-    b[i] = a[i^_mask];\
+//////////////////////////////////////////////
+// Some Template specialization
+#define perm(a, b, n, w)			\
+  unsigned int _mask = w >> (n + 1);		\
+  VECTOR_FOR(i, w, 1)				\
+  {						\
+    b[i] = a[i^_mask];				\
   }
   
-  #define DECL_PERMUTE_N(n)\
-  template <typename T>\
-  static inline vec<T> Permute##n(vec<T> in) {\
-    vec<T> out;\
-    perm(in.v, out.v, n, W<T>::r);\
-    return out;\
+#define DECL_PERMUTE_N(n)			\
+  template <typename T>				\
+  static accelerator_inline vec<T> Permute##n(vec<T> in) {	\
+    vec<T> out;					\
+    perm(in.v, out.v, n, W<T>::r);		\
+    return out;					\
   }
   
-  struct Permute{
-    DECL_PERMUTE_N(0);
-    DECL_PERMUTE_N(1);
-    DECL_PERMUTE_N(2);
-    DECL_PERMUTE_N(3);
-  };
+struct Permute{
+  DECL_PERMUTE_N(0);
+  DECL_PERMUTE_N(1);
+  DECL_PERMUTE_N(2);
+  DECL_PERMUTE_N(3);
+};
   
-  #undef perm
-  #undef DECL_PERMUTE_N
+#undef perm
+#undef DECL_PERMUTE_N
   
-  #define rot(a, b, n, w)\
-  VECTOR_FOR(i, w, 1)\
-  {\
-    b[i] = a[(i + n)%w];\
+#define rot(a, b, n, w)				\
+  VECTOR_FOR(i, w, 1)				\
+  {						\
+    b[i] = a[(i + n)%w];			\
   }
   
-  struct Rotate{
+struct Rotate{
       
-    template <int n, typename T> static inline vec<T> tRotate(vec<T> in){
-      return rotate(in, n);
-    }
+  template <int n, typename T> static accelerator_inline vec<T> tRotate(vec<T> in){
+    return rotate(in, n);
+  }
     
-    template <typename T>
-    static inline vec<T> rotate(vec<T> in, int n){
-      vec<T> out;
+  template <typename T>
+  static accelerator_inline vec<T> rotate(vec<T> in, int n){
+    vec<T> out;
       
-      rot(in.v, out.v, n, W<T>::r);
+    rot(in.v, out.v, n, W<T>::r);
       
-      return out;
+    return out;
+  }
+};
+
+#undef rot
+  
+#define acc(v, a, off, step, n)			\
+  for (unsigned int i = off; i < n; i += step)	\
+    {						\
+      a += v[i];				\
     }
-  };
-
-  #undef rot
   
-  #define acc(v, a, off, step, n)\
-  for (unsigned int i = off; i < n; i += step)\
-  {\
-    a += v[i];\
+template <typename Out_type, typename In_type>
+struct Reduce{
+  //Need templated class to overload output type
+  //General form must generate error if compiled
+  accelerator_inline Out_type operator()(In_type in){
+    printf("Error, using wrong Reduce function\n");
+    exit(1);
+    return 0;
   }
+};
   
-  template <typename Out_type, typename In_type>
-  struct Reduce{
-    //Need templated class to overload output type
-    //General form must generate error if compiled
-    inline Out_type operator()(In_type in){
-      printf("Error, using wrong Reduce function\n");
-      exit(1);
-      return 0;
-    }
-  };
-  
-  //Complex float Reduce
-  template <>
-  inline Grid::ComplexF Reduce<Grid::ComplexF, vecf>::operator()(vecf in){
-    float a = 0.f, b = 0.f;
+//Complex float Reduce
+template <>
+accelerator_inline Grid::ComplexF Reduce<Grid::ComplexF, vecf>::operator()(vecf in){
+  float a = 0.f, b = 0.f;
     
-    acc(in.v, a, 0, 2, W<float>::r);
-    acc(in.v, b, 1, 2, W<float>::r);
+  acc(in.v, a, 0, 2, W<float>::r);
+  acc(in.v, b, 1, 2, W<float>::r);
     
-    return Grid::ComplexF(a, b);
-  }
-  
-  //Real float Reduce
-  template<>
-  inline Grid::RealF Reduce<Grid::RealF, vecf>::operator()(vecf in){
-    float a = 0.;
-    
-    acc(in.v, a, 0, 1, W<float>::r);
-    
-    return a;
-  }
-  
-  //Complex double Reduce
-  template<>
-  inline Grid::ComplexD Reduce<Grid::ComplexD, vecd>::operator()(vecd in){
-    double a = 0., b = 0.;
-    
-    acc(in.v, a, 0, 2, W<double>::r);
-    acc(in.v, b, 1, 2, W<double>::r);
-    
-    return Grid::ComplexD(a, b);
-  }
-  
-  //Real double Reduce
-  template<>
-  inline Grid::RealD Reduce<Grid::RealD, vecd>::operator()(vecd in){
-    double a = 0.f;
-    
-    acc(in.v, a, 0, 1, W<double>::r);
-    
-    return a;
-  }
-
-  //Integer Reduce
-  template<>
-  inline Integer Reduce<Integer, veci>::operator()(veci in){
-    Integer a = 0;
-    
-    acc(in.v, a, 0, 1, W<Integer>::r);
-    
-    return a;
-  }
-
-  #undef acc  // EIGEN compatibility
+  return Grid::ComplexF(a, b);
 }
+  
+//Real float Reduce
+template<>
+accelerator_inline Grid::RealF Reduce<Grid::RealF, vecf>::operator()(vecf in){
+  float a = 0.;
+    
+  acc(in.v, a, 0, 1, W<float>::r);
+    
+  return a;
+}
+  
+//Complex double Reduce
+template<>
+accelerator_inline Grid::ComplexD Reduce<Grid::ComplexD, vecd>::operator()(vecd in){
+  double a = 0., b = 0.;
+    
+  acc(in.v, a, 0, 2, W<double>::r);
+  acc(in.v, b, 1, 2, W<double>::r);
+    
+  return Grid::ComplexD(a, b);
+}
+  
+//Real double Reduce
+template<>
+accelerator_inline Grid::RealD Reduce<Grid::RealD, vecd>::operator()(vecd in){
+  double a = 0.f;
+    
+  acc(in.v, a, 0, 1, W<double>::r);
+    
+  return a;
+}
+
+//Integer Reduce
+template<>
+accelerator_inline Integer Reduce<Integer, veci>::operator()(veci in){
+  Integer a = 0;
+    
+  acc(in.v, a, 0, 1, W<Integer>::r);
+    
+  return a;
+}
+
+#undef acc  // EIGEN compatibility
+NAMESPACE_END(Optimization)
 
 //////////////////////////////////////////////////////////////////////////////////////
 // Here assign types 
 
-  typedef Optimization::vech SIMD_Htype; // Reduced precision type
-  typedef Optimization::vecf SIMD_Ftype; // Single precision type
-  typedef Optimization::vecd SIMD_Dtype; // Double precision type
-  typedef Optimization::veci SIMD_Itype; // Integer type
+typedef Optimization::vech SIMD_Htype; // Reduced precision type
+typedef Optimization::vecf SIMD_Ftype; // Single precision type
+typedef Optimization::vecd SIMD_Dtype; // Double precision type
+typedef Optimization::veci SIMD_Itype; // Integer type
 
-  // prefetch utilities
-  inline void v_prefetch0(int size, const char *ptr){};
-  inline void prefetch_HINT_T0(const char *ptr){};
+// prefetch utilities
+accelerator_inline void v_prefetch0(int size, const char *ptr){};
+accelerator_inline void prefetch_HINT_T0(const char *ptr){};
 
-  // Function name aliases
-  typedef Optimization::Vsplat   VsplatSIMD;
-  typedef Optimization::Vstore   VstoreSIMD;
-  typedef Optimization::Vset     VsetSIMD;
-  typedef Optimization::Vstream  VstreamSIMD;
-  template <typename S, typename T> using ReduceSIMD = Optimization::Reduce<S,T>;
+// Function name aliases
+typedef Optimization::Vsplat   VsplatSIMD;
+typedef Optimization::Vstore   VstoreSIMD;
+typedef Optimization::Vset     VsetSIMD;
+typedef Optimization::Vstream  VstreamSIMD;
+template <typename S, typename T> using ReduceSIMD = Optimization::Reduce<S,T>;
+
+// Arithmetic operations
+typedef Optimization::Sum         SumSIMD;
+typedef Optimization::Sub         SubSIMD;
+typedef Optimization::Div         DivSIMD;
+typedef Optimization::Mult        MultSIMD;
+typedef Optimization::MultComplex MultComplexSIMD;
+typedef Optimization::MultRealPart MultRealPartSIMD;
+typedef Optimization::MaddRealPart MaddRealPartSIMD;
+typedef Optimization::Conj        ConjSIMD;
+typedef Optimization::TimesMinusI TimesMinusISIMD;
+typedef Optimization::TimesI      TimesISIMD;
+
+NAMESPACE_END(Grid)
 
-  // Arithmetic operations
-  typedef Optimization::Sum         SumSIMD;
-  typedef Optimization::Sub         SubSIMD;
-  typedef Optimization::Div         DivSIMD;
-  typedef Optimization::Mult        MultSIMD;
-  typedef Optimization::MultComplex MultComplexSIMD;
-  typedef Optimization::MultRealPart MultRealPartSIMD;
-  typedef Optimization::MaddRealPart MaddRealPartSIMD;
-  typedef Optimization::Conj        ConjSIMD;
-  typedef Optimization::TimesMinusI TimesMinusISIMD;
-  typedef Optimization::TimesI      TimesISIMD;
-}
diff --git a/Grid/simd/Grid_generic_types.h b/Grid/simd/Grid_generic_types.h
index 642f6ffe..5965df73 100644
--- a/Grid/simd/Grid_generic_types.h
+++ b/Grid/simd/Grid_generic_types.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,62 +24,89 @@ Author: Antonin Portelli <antonin.portelli@me.com>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 
 static_assert(GEN_SIMD_WIDTH % 16u == 0, "SIMD vector size is not an integer multiple of 16 bytes");
 
-//#define VECTOR_LOOPS
+#undef VECTOR_LOOPS
 
 // playing with compiler pragmas
+
 #ifdef VECTOR_LOOPS
 #ifdef __clang__
-#define VECTOR_FOR(i, w, inc)\
-_Pragma("clang loop unroll(full) vectorize(enable) interleave(enable) vectorize_width(w)")\
-for (unsigned int i = 0; i < w; i += inc)
+#define VECTOR_FOR(i, w, inc)						\
+  _Pragma("clang loop unroll(full) vectorize(enable) interleave(enable) vectorize_width(w)") \
+  for (unsigned int i = 0; i < w; i += inc)
 #elif defined __INTEL_COMPILER
-#define VECTOR_FOR(i, w, inc)\
-_Pragma("simd vectorlength(w*8)")\
-for (unsigned int i = 0; i < w; i += inc)
+#define VECTOR_FOR(i, w, inc)			\
+  _Pragma("simd vectorlength(w*8)")		\
+  for (unsigned int i = 0; i < w; i += inc)
 #else
-#define VECTOR_FOR(i, w, inc)\
-for (unsigned int i = 0; i < w; i += inc)
+#define VECTOR_FOR(i, w, inc)			\
+  for (unsigned int i = 0; i < w; i += inc)
 #endif
 #else
-#define VECTOR_FOR(i, w, inc)\
-for (unsigned int i = 0; i < w; i += inc)
+#define VECTOR_FOR(i, w, inc)			\
+  for (unsigned int i = 0; i < w; i += inc)
 #endif
 
-namespace Grid {
-namespace Optimization {
+NAMESPACE_BEGIN(Grid);
+NAMESPACE_BEGIN(Optimization);
 
-  // type traits giving the number of elements for each vector type
-  template <typename T> struct W;
-  template <> struct W<double> {
-    constexpr static unsigned int c = GEN_SIMD_WIDTH/16u;
-    constexpr static unsigned int r = GEN_SIMD_WIDTH/8u;
-  };
-  template <> struct W<float> {
-    constexpr static unsigned int c = GEN_SIMD_WIDTH/8u;
-    constexpr static unsigned int r = GEN_SIMD_WIDTH/4u;
-  };
-  template <> struct W<Integer> {
-    constexpr static unsigned int r = GEN_SIMD_WIDTH/4u;
-  };
-  template <> struct W<uint16_t> {
-    constexpr static unsigned int c = GEN_SIMD_WIDTH/4u;
-    constexpr static unsigned int r = GEN_SIMD_WIDTH/2u;
-  };
-  
-  // SIMD vector types
-  template <typename T>
-  struct vec {
-    alignas(GEN_SIMD_WIDTH) T v[W<T>::r];
-  };
+#ifdef GENERIC_SCALAR
 
-  typedef vec<float>     vecf;
-  typedef vec<double>    vecd;
-  typedef vec<uint16_t>  vech; // half precision comms
-  typedef vec<Integer>   veci;
+// type traits giving the number of elements for each vector type
+template <typename T> struct W;
+template <> struct W<double> {
+  constexpr static unsigned int c = 1;
+  constexpr static unsigned int r = 2;
+};
+template <> struct W<float> {
+  constexpr static unsigned int c = 1;
+  constexpr static unsigned int r = 2;
+};
+template <> struct W<Integer> {
+  constexpr static unsigned int r = 1;
+};
+template <> struct W<uint16_t> {
+  constexpr static unsigned int c = 1;
+  constexpr static unsigned int r = 2;
+};
+// SIMD vector types
+template <typename T>
+struct vec {
+  T v[W<T>::r];
+};
+#else 
+// type traits giving the number of elements for each vector type
+template <typename T> struct W;
+template <> struct W<double> {
+  constexpr static unsigned int c = GEN_SIMD_WIDTH/16u;
+  constexpr static unsigned int r = GEN_SIMD_WIDTH/8u;
+};
+template <> struct W<float> {
+  constexpr static unsigned int c = GEN_SIMD_WIDTH/8u;
+  constexpr static unsigned int r = GEN_SIMD_WIDTH/4u;
+};
+template <> struct W<Integer> {
+  constexpr static unsigned int r = GEN_SIMD_WIDTH/4u;
+};
+template <> struct W<uint16_t> {
+  constexpr static unsigned int c = GEN_SIMD_WIDTH/4u;
+  constexpr static unsigned int r = GEN_SIMD_WIDTH/2u;
+};
+// SIMD vector types
+template <typename T>
+struct vec {
+  alignas(GEN_SIMD_WIDTH) T v[W<T>::r];
+};
+#endif
+
+typedef vec<float>     vecf;
+typedef vec<double>    vecd;
+typedef vec<uint16_t>  vech; // half precision comms
+typedef vec<Integer>   veci;
   
-}}
+NAMESPACE_END(Optimization);
+NAMESPACE_END(Grid);
diff --git a/Grid/simd/Grid_gpu_vec.h b/Grid/simd/Grid_gpu_vec.h
new file mode 100644
index 00000000..471fbccc
--- /dev/null
+++ b/Grid/simd/Grid_gpu_vec.h
@@ -0,0 +1,792 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/simd/Grid_gpu.h
+
+    Copyright (C) 2018
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+//----------------------------------------------------------------------
+/*! @file Grid_gpu.h
+  @brief Optimization libraries for GPU
+  Use float4, double2 
+*/
+//----------------------------------------------------------------------
+
+#include <cuda_fp16.h>
+
+namespace Grid {
+
+#define COALESCE_GRANULARITY ( GEN_SIMD_WIDTH )
+
+template<class pair>
+class GpuComplex {
+public:
+  pair z;
+  typedef decltype(z.x) real;
+public: 
+  accelerator_inline GpuComplex() = default;
+  accelerator_inline GpuComplex(real re,real im) { z.x=re; z.y=im; };
+  accelerator_inline GpuComplex(const GpuComplex &zz) { z = zz.z;};
+  friend accelerator_inline  GpuComplex operator+(const GpuComplex &lhs,const GpuComplex &rhs) { 
+    GpuComplex r ; 
+    r.z.x = lhs.z.x + rhs.z.x; 
+    r.z.y = lhs.z.y + rhs.z.y; 
+    return r; 
+  }
+  friend accelerator_inline GpuComplex operator-(const GpuComplex &lhs,const GpuComplex &rhs) { 
+    GpuComplex r ; 
+    r.z.x = lhs.z.x - rhs.z.x; 
+    r.z.y = lhs.z.y - rhs.z.y; 
+    return r; 
+  }
+  friend accelerator_inline GpuComplex operator*(const GpuComplex &lhs,const GpuComplex &rhs) { 
+    GpuComplex r ; 
+    r.z.x= lhs.z.x*rhs.z.x - lhs.z.y*rhs.z.y; // rr-ii
+    r.z.y= lhs.z.x*rhs.z.y + lhs.z.y*rhs.z.x; // ri+ir
+    return r;
+  }
+  friend accelerator_inline GpuComplex real_mult(const GpuComplex &l,const GpuComplex &r) 
+  {
+    GpuComplex ret;
+    ret.z.x = l.z.x*r.z.x;
+    ret.z.y = l.z.x*r.z.y;
+    return ret;
+  }
+  friend std::ostream& operator<< (std::ostream& stream, const GpuComplex o){
+    stream << "("<< o.z.x << ","<< o.z.y <<")";
+    return stream;
+  }
+};
+
+template<int _N, class _datum>
+struct GpuVector {
+  _datum v[_N];
+  static const int N = _N;
+  typedef _datum datum;
+};
+
+
+template<int N,class datum>
+inline accelerator GpuVector<N,datum> operator*(const GpuVector<N,datum> l,const GpuVector<N,datum> r) {
+  GpuVector<N,datum> ret;
+  for(int i=0;i<N;i++) { 
+    ret.v[i] = l.v[i]*r.v[i];
+  }
+  return ret;
+}
+template<int N,class datum>
+inline accelerator GpuVector<N,datum> operator-(const GpuVector<N,datum> l,const GpuVector<N,datum> r) {
+  GpuVector<N,datum> ret;
+  for(int i=0;i<N;i++) { 
+    ret.v[i] = l.v[i]-r.v[i];
+  }
+  return ret;
+}
+template<int N,class datum>
+inline accelerator GpuVector<N,datum> operator+(const GpuVector<N,datum> l,const GpuVector<N,datum> r) {
+  GpuVector<N,datum> ret;
+  for(int i=0;i<N;i++) { 
+    ret.v[i] = l.v[i]+r.v[i];
+  }
+  return ret;
+}
+template<int N,class datum>
+inline accelerator GpuVector<N,datum> operator/(const GpuVector<N,datum> l,const GpuVector<N,datum> r) {
+  GpuVector<N,datum> ret;
+  for(int i=0;i<N;i++) { 
+    ret.v[i] = l.v[i]/r.v[i];
+  }
+  return ret;
+}
+
+constexpr int NSIMD_RealH    = COALESCE_GRANULARITY / sizeof(half);
+constexpr int NSIMD_ComplexH = COALESCE_GRANULARITY / sizeof(half2);
+constexpr int NSIMD_RealF    = COALESCE_GRANULARITY / sizeof(float);
+constexpr int NSIMD_ComplexF = COALESCE_GRANULARITY / sizeof(float2);
+constexpr int NSIMD_RealD    = COALESCE_GRANULARITY / sizeof(double);
+constexpr int NSIMD_ComplexD = COALESCE_GRANULARITY / sizeof(double2);
+constexpr int NSIMD_Integer  = COALESCE_GRANULARITY / sizeof(Integer);
+
+typedef GpuComplex<half2  > GpuComplexH;
+typedef GpuComplex<float2 > GpuComplexF;
+typedef GpuComplex<double2> GpuComplexD;
+
+typedef GpuVector<NSIMD_RealH   , half        > GpuVectorRH;
+typedef GpuVector<NSIMD_ComplexH, GpuComplexH > GpuVectorCH;
+typedef GpuVector<NSIMD_RealF,    float       > GpuVectorRF;
+typedef GpuVector<NSIMD_ComplexF, GpuComplexF > GpuVectorCF;
+typedef GpuVector<NSIMD_RealD,    double      > GpuVectorRD;
+typedef GpuVector<NSIMD_ComplexD, GpuComplexD > GpuVectorCD;
+typedef GpuVector<NSIMD_Integer,  Integer     > GpuVectorI;
+
+accelerator_inline float half2float(half h)
+{
+  float f;
+#ifdef __CUDA_ARCH__
+  f = __half2float(h);
+#else 
+  //f = __half2float(h);
+  __half_raw hr(h);
+  Grid_half hh; 
+  hh.x = hr.x;
+  f=  sfw_half_to_float(hh);
+#endif
+  return f;
+}
+accelerator_inline half float2half(float f)
+{
+  half h;
+#ifdef __CUDA_ARCH__
+  h = __float2half(f);
+#else
+  Grid_half hh = sfw_float_to_half(f);
+  __half_raw hr;  
+  hr.x = hh.x;
+  h = __half(hr);
+#endif
+  return h;
+}
+
+namespace Optimization {
+
+  struct Vsplat{
+    //Complex float
+    accelerator_inline GpuVectorCF operator()(float a, float b){
+      GpuVectorCF ret;
+      for(int i=0;i<GpuVectorCF::N;i++){
+	ret.v[i] = typename GpuVectorCF::datum(a,b);
+      }
+      return ret;
+    }
+    // Real float
+    accelerator_inline GpuVectorRF operator()(float a){
+      GpuVectorRF ret;
+      for(int i=0;i<GpuVectorRF::N;i++){
+	ret.v[i] = typename GpuVectorRF::datum(a);
+      }
+      return ret;
+    }
+    //Complex double
+    accelerator_inline GpuVectorCD operator()(double a, double b){
+      GpuVectorCD ret;
+      for(int i=0;i<GpuVectorCD::N;i++){
+	ret.v[i] = typename GpuVectorCD::datum(a,b);
+      }
+      return ret;
+    }
+    //Real double
+    accelerator_inline GpuVectorRD operator()(double a){
+      GpuVectorRD ret; 
+      for(int i=0;i<GpuVectorRD::N;i++){
+	ret.v[i] = typename GpuVectorRD::datum(a);
+      }
+      return ret;
+    }
+    //Integer
+    accelerator_inline GpuVectorI operator()(Integer a){
+      GpuVectorI ret;
+      for(int i=0;i<GpuVectorI::N;i++){
+	ret.v[i] = typename GpuVectorI::datum(a);
+      }
+      return ret;
+    }
+  };
+
+  struct Vstore{
+    template<int N,class datum,class P>
+    accelerator_inline void operator()(GpuVector<N,datum> a, P* Fp){
+      GpuVector<N,datum> *vF = (GpuVector<N,datum> *)Fp;
+      *vF = a;
+    }
+  };
+
+  struct Vstream{
+    template<int N,class datum, class P>
+    accelerator_inline void operator()(P* F,GpuVector<N,datum> a){
+      GpuVector<N,datum> *vF = (GpuVector<N,datum> *)F;
+      *vF = a;
+    }
+  };
+
+  struct Vset{
+    // Complex float 
+    accelerator_inline GpuVectorCF operator()(Grid::ComplexF *a){
+      typedef GpuVectorCF vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i] = vec::datum(a[i].real(),a[i].imag());
+      }
+      return ret;
+    }
+    // Complex double 
+    accelerator_inline GpuVectorCD operator()(Grid::ComplexD *a){
+      typedef GpuVectorCD vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i] = vec::datum(a[i].real(),a[i].imag());
+      }
+      return ret;
+    }
+    // Real float 
+    accelerator_inline GpuVectorRF operator()(float *a){
+      typedef GpuVectorRF vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i] = vec::datum(a[i]);
+      }
+      return ret;
+    }
+    // Real double
+    accelerator_inline GpuVectorRD operator()(double *a){
+      typedef GpuVectorRD vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i] = vec::datum(a[i]);
+      }
+      return ret;
+    }
+    // Integer
+    accelerator_inline GpuVectorI operator()(Integer *a){
+      typedef GpuVectorI vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i] = vec::datum(a[i]);
+      }
+      return ret;
+    }
+  };
+
+  template <typename Out_type, typename In_type>
+  struct Reduce{
+    //Need templated class to overload output type
+    //General form must generate error if compiled
+    accelerator_inline Out_type operator()(In_type in){
+      printf("Error, using wrong Reduce function\n");
+      exit(1);
+      return 0;
+    }
+  };
+
+  /////////////////////////////////////////////////////
+  // Arithmetic operations
+  /////////////////////////////////////////////////////
+  struct Sum{
+    //Real float
+    accelerator_inline GpuVectorRF operator()(GpuVectorRF a,GpuVectorRF b){
+      return a+b;
+    }
+    accelerator_inline GpuVectorRD operator()(GpuVectorRD a,GpuVectorRD b){
+      return a+b;
+    }
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF a,GpuVectorCF b){
+      return a+b;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD a,GpuVectorCD b){
+      return a+b;
+    }
+    accelerator_inline GpuVectorI operator()(GpuVectorI a,GpuVectorI b){
+      return a+b;
+    }
+  };
+
+  struct Sub{
+    accelerator_inline GpuVectorRF operator()(GpuVectorRF a,GpuVectorRF b){
+      return a-b;
+    }
+    accelerator_inline GpuVectorRD operator()(GpuVectorRD a,GpuVectorRD b){
+      return a-b;
+    }
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF a,GpuVectorCF b){
+      return a-b;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD a,GpuVectorCD b){
+      return a-b;
+    }
+    accelerator_inline GpuVectorI operator()(GpuVectorI a,GpuVectorI b){
+      return a-b;
+    }
+  };
+
+  struct MultRealPart{
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF a,GpuVectorCF b){
+      typedef GpuVectorCF vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i] = real_mult(a.v[i],b.v[i]);
+      }
+      return ret;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD a,GpuVectorCD b){
+      typedef GpuVectorCD vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i] = real_mult(a.v[i],b.v[i]);
+      }
+      return ret;
+    }
+  };
+
+  struct MaddRealPart{
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF a,GpuVectorCF b,GpuVectorCF c){
+      typedef GpuVectorCF vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i] = real_mult(a.v[i],b.v[i]) +c.v[i];
+      }
+      return ret;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD a,GpuVectorCD b,GpuVectorCD c){
+      typedef GpuVectorCD vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i] = real_mult(a.v[i],b.v[i]) +c.v[i];
+      }
+      return ret;
+    }
+  };
+
+  struct MultComplex{
+
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF a,GpuVectorCF b){
+      return a*b;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD a,GpuVectorCD b){
+      return a*b;
+    }
+  };
+
+  struct Mult{
+    accelerator_inline void mac(GpuVectorRF &a, GpuVectorRF b, GpuVectorRF c){
+      a= a+b*c;
+    }
+    accelerator_inline void mac(GpuVectorRD &a, GpuVectorRD b, GpuVectorRD c){
+      a= a+b*c;
+    }
+    // Real float
+    accelerator_inline GpuVectorRF operator()(GpuVectorRF a, GpuVectorRF b){
+      return a*b;
+    }
+    // Real double
+    accelerator_inline GpuVectorRD operator()(GpuVectorRD a, GpuVectorRD b){
+      return a*b;
+    }
+    accelerator_inline GpuVectorI operator()(GpuVectorI a, GpuVectorI b){
+      return a*b;
+    }
+  };
+
+  struct Div{
+    // Real float
+    accelerator_inline GpuVectorRF operator()(GpuVectorRF a, GpuVectorRF b){
+      return a/b;
+    }
+    accelerator_inline GpuVectorRD operator()(GpuVectorRD a, GpuVectorRD b){
+      return a/b;
+    }
+    // Danger -- element wise divide fro complex, not complex div. 
+    // See Grid_vector_types.h lines around 735, applied after "toReal"
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF a, GpuVectorCF b){
+      GpuVectorCF ret;
+      for(int i=0;i< GpuVectorCF::N;i++){
+	ret.v[i].z.x = a.v[i].z.x / b.v[i].z.x;
+	ret.v[i].z.y = a.v[i].z.y / b.v[i].z.y;
+      }
+      return ret;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD a, GpuVectorCD b){
+      GpuVectorCD ret;
+      for(int i=0;i< GpuVectorCD::N;i++){
+	ret.v[i].z.x = a.v[i].z.x / b.v[i].z.x;
+	ret.v[i].z.y = a.v[i].z.y / b.v[i].z.y;
+      }
+      return ret;
+    }
+  };
+
+
+  struct Conj{
+    // Complex single
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF in){
+      typedef GpuVectorCF vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i].z.x = in.v[i].z.x;
+	ret.v[i].z.y =-in.v[i].z.y;
+      }
+      return ret;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD in){
+      typedef GpuVectorCD vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i].z.x = in.v[i].z.x;
+	ret.v[i].z.y =-in.v[i].z.y;
+      }
+      return ret;
+    }
+  };
+
+  struct TimesMinusI{
+    //Complex single
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF in,GpuVectorCF dummy){
+      typedef GpuVectorCF vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i].z.x = in.v[i].z.y;
+	ret.v[i].z.y =-in.v[i].z.x;
+      }
+      return ret;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD in,GpuVectorCD dummy){
+      typedef GpuVectorCD vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i].z.x = in.v[i].z.y;
+	ret.v[i].z.y =-in.v[i].z.x;
+      }
+      return ret;
+    }
+  };
+
+  struct TimesI{
+    //Complex single
+    accelerator_inline GpuVectorCF operator()(GpuVectorCF in,GpuVectorCF dummy){
+      typedef GpuVectorCF vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i].z.x =-in.v[i].z.y;
+	ret.v[i].z.y = in.v[i].z.x;
+      }
+      return ret;
+    }
+    accelerator_inline GpuVectorCD operator()(GpuVectorCD in,GpuVectorCD dummy){
+      typedef GpuVectorCD vec;
+      vec ret;
+      for(int i=0;i<vec::N;i++){
+	ret.v[i].z.x =-in.v[i].z.y;
+	ret.v[i].z.y = in.v[i].z.x;
+      }
+      return ret;
+    }
+  };
+
+  struct Permute{
+
+    template <int n,typename vec>				       
+    static accelerator_inline vec PermuteN(vec in) {   
+      vec out;					
+      unsigned int _mask = vec::N >> (n + 1);	
+      for(int i=0;i<vec::N;i++) {
+	out.v[i] = in.v[i^_mask];
+      }
+      return out;	
+    }
+    
+    template <typename vec>  static accelerator_inline vec Permute0(vec in) { return PermuteN<0,vec>(in);  }
+    template <typename vec>  static accelerator_inline vec Permute1(vec in) { return PermuteN<1,vec>(in);  }
+    template <typename vec>  static accelerator_inline vec Permute2(vec in) { return PermuteN<2,vec>(in);  }
+    template <typename vec>  static accelerator_inline vec Permute3(vec in) { return PermuteN<3,vec>(in);  }
+    
+  };
+  
+  struct PrecisionChange {
+
+    ////////////////////////////////////////////////////////////////////////////////////
+    // Single / Half
+    ////////////////////////////////////////////////////////////////////////////////////
+    static accelerator_inline GpuVectorCH StoH (GpuVectorCF a,GpuVectorCF b) {
+      int N = GpuVectorCF::N;
+      GpuVectorCH h;
+      for(int i=0;i<N;i++) {
+        h.v[i  ].z.x = float2half(a.v[i].z.x);
+        h.v[i  ].z.y = float2half(a.v[i].z.y);
+	h.v[i+N].z.x = float2half(b.v[i].z.x);
+	h.v[i+N].z.y = float2half(b.v[i].z.y);
+      }
+      return h;
+    }
+    static accelerator_inline void  HtoS (GpuVectorCH h,GpuVectorCF &sa,GpuVectorCF &sb) {
+      int N = GpuVectorCF::N;
+      for(int i=0;i<N;i++) {
+	sa.v[i].z.x = half2float(h.v[i  ].z.x);
+	sa.v[i].z.y = half2float(h.v[i  ].z.y);
+	sb.v[i].z.x = half2float(h.v[i+N].z.x);
+	sb.v[i].z.y = half2float(h.v[i+N].z.y);
+      }
+    }
+    static accelerator_inline GpuVectorRH StoH (GpuVectorRF a,GpuVectorRF b) {
+      int N = GpuVectorRF::N;
+      GpuVectorRH h;
+      for(int i=0;i<N;i++) {
+        h.v[i  ] = float2half(a.v[i]);
+	h.v[i+N] = float2half(b.v[i]);
+      }
+      return h;
+    }
+    static accelerator_inline void  HtoS (GpuVectorRH h,GpuVectorRF &sa,GpuVectorRF &sb) {
+      int N = GpuVectorRF::N;
+      for(int i=0;i<N;i++) {
+	sa.v[i] = half2float(h.v[i  ]);
+	sb.v[i] = half2float(h.v[i+N]);
+      }
+    }
+
+    ////////////////////////////////////////////////////////////////////////////////////
+    // Double Single
+    ////////////////////////////////////////////////////////////////////////////////////
+    static accelerator_inline GpuVectorCF DtoS (GpuVectorCD a,GpuVectorCD b) {
+      int N = GpuVectorCD::N;
+      GpuVectorCF h;
+      for(int i=0;i<N;i++) {
+        h.v[i  ].z.x = a.v[i].z.x;
+        h.v[i  ].z.y = a.v[i].z.y;
+	h.v[i+N].z.x = b.v[i].z.x;
+	h.v[i+N].z.y = b.v[i].z.y;
+      }
+      return h;
+    }
+
+    static accelerator_inline void  StoD (GpuVectorCF h,GpuVectorCD &sa,GpuVectorCD &sb) {
+      int N = GpuVectorCD::N;
+      for(int i=0;i<N;i++) {
+	sa.v[i].z.x = h.v[i  ].z.x;
+	sa.v[i].z.y = h.v[i  ].z.y;
+	sb.v[i].z.x = h.v[i+N].z.x;
+	sb.v[i].z.y = h.v[i+N].z.y;
+      }
+    }
+
+    static accelerator_inline GpuVectorRF DtoS (GpuVectorRD a,GpuVectorRD b) {
+      int N = GpuVectorRD::N;
+      GpuVectorRF h;
+      for(int i=0;i<N;i++) {
+        h.v[i  ] = a.v[i];
+	h.v[i+N] = b.v[i];
+      }
+      return h;
+    }
+
+    static accelerator_inline void  StoD (GpuVectorRF h,GpuVectorRD &sa,GpuVectorRD &sb) {
+      int N = GpuVectorRD::N;
+      for(int i=0;i<N;i++) {
+	sa.v[i] = h.v[i  ];
+	sb.v[i] = h.v[i+N];
+      }
+    }
+
+    ////////////////////////////////////////////////////////////////////////////////////
+    // Double Half
+    ////////////////////////////////////////////////////////////////////////////////////
+    static accelerator_inline GpuVectorCH DtoH (GpuVectorCD a,GpuVectorCD b,GpuVectorCD c,GpuVectorCD d) {
+      GpuVectorCF sa,sb;
+      sa = DtoS(a,b);
+      sb = DtoS(c,d);
+      return StoH(sa,sb);
+    }
+    static accelerator_inline void HtoD (GpuVectorCH h,GpuVectorCD &a,GpuVectorCD &b,GpuVectorCD &c,GpuVectorCD &d) {
+      GpuVectorCF sa,sb;
+      HtoS(h,sa,sb);
+      StoD(sa,a,b);
+      StoD(sb,c,d);
+    }
+    static accelerator_inline GpuVectorRH DtoH (GpuVectorRD a,GpuVectorRD b,GpuVectorRD c,GpuVectorRD d) {
+      GpuVectorRF sa,sb;
+      sa = DtoS(a,b);
+      sb = DtoS(c,d);
+      return StoH(sa,sb);
+    }
+    static accelerator_inline void HtoD (GpuVectorRH h,GpuVectorRD &a,GpuVectorRD &b,GpuVectorRD &c,GpuVectorRD &d) {
+      GpuVectorRF sa,sb;
+      HtoS(h,sa,sb);
+      StoD(sa,a,b);
+      StoD(sb,c,d);
+    }
+  };
+
+struct Exchange{
+
+  template <typename vec,int n>
+  static accelerator_inline void ExchangeN(vec &out1,vec &out2,vec &in1,vec &in2){
+    unsigned int mask = vec::N >> (n + 1);
+    for(int i=0;i<vec::N;i++) {
+      int j1 = i&(~mask);
+      if  ( (i&mask) == 0 ) { out1.v[i]=in1.v[j1];}
+      else                  { out1.v[i]=in2.v[j1];}
+      int j2 = i|mask;
+      if  ( (i&mask) == 0 ) { out2.v[i]=in1.v[j2];}
+      else                  { out2.v[i]=in2.v[j2];}
+    }      
+  }
+  template <typename vec>
+  static accelerator_inline void Exchange0(vec &out1,vec &out2,vec &in1,vec &in2){
+    ExchangeN<vec,0>(out1,out2,in1,in2);
+  };
+  template <typename vec>
+  static accelerator_inline void Exchange1(vec &out1,vec &out2,vec &in1,vec &in2){
+    ExchangeN<vec,1>(out1,out2,in1,in2);
+  };
+  template <typename vec>
+  static accelerator_inline void Exchange2(vec &out1,vec &out2,vec &in1,vec &in2){
+    ExchangeN<vec,2>(out1,out2,in1,in2);
+  };
+  template <typename vec>
+  static accelerator_inline void Exchange3(vec &out1,vec &out2,vec &in1,vec &in2){
+    ExchangeN<vec,3>(out1,out2,in1,in2);
+  };
+
+};
+
+struct Rotate{
+
+  template <int n, typename vec> static accelerator_inline vec tRotate(vec in){
+    return rotate(in, n);
+  }
+    
+  template <typename vec>
+  static accelerator_inline vec rotate_template(vec in, int n){
+    vec out;
+    for(int i=0;i<vec::N;i++){
+      out.v[i] = in.v[(i + n)%vec::N];
+    }
+    return out;
+  }
+
+  typedef GpuVectorRH  SIMD_Htype; // Single precision type
+  typedef GpuVectorRF  SIMD_Ftype; // Single precision type
+  typedef GpuVectorRD  SIMD_Dtype; // Double precision type
+  typedef GpuVectorI   SIMD_Itype; // Integer type
+
+  typedef GpuVectorCH  SIMD_CHtype; // Single precision type
+  typedef GpuVectorCF  SIMD_CFtype; // Single precision type
+  typedef GpuVectorCD  SIMD_CDtype; // Double precision type
+
+  static accelerator_inline GpuVectorRH rotate(GpuVectorRH in, int n){ return rotate_template(in,n);}
+  static accelerator_inline GpuVectorRF rotate(GpuVectorRF in, int n){ return rotate_template(in,n);}
+  static accelerator_inline GpuVectorRD rotate(GpuVectorRD in, int n){ return rotate_template(in,n);}
+  static accelerator_inline GpuVectorI  rotate(GpuVectorI  in, int n){ return rotate_template(in,n);}
+  static accelerator_inline GpuVectorCH rotate(GpuVectorCH in, int n){ return rotate_template(in,n/2);} // Measure in complex not float
+  static accelerator_inline GpuVectorCF rotate(GpuVectorCF in, int n){ return rotate_template(in,n/2);}
+  static accelerator_inline GpuVectorCD rotate(GpuVectorCD in, int n){ return rotate_template(in,n/2);}
+
+};
+
+//////////////////////////////////////////////
+// Some Template specialization
+
+  //Complex float Reduce
+  template<>
+  accelerator_inline Grid::ComplexF 
+  Reduce<Grid::ComplexF, GpuVectorCF>::operator()(GpuVectorCF in)
+  {
+    GpuComplexF greduce = in.v[0];
+    for(int i=1;i<GpuVectorCF::N;i++) {
+      greduce = greduce+in.v[i];
+    }
+    Grid::ComplexF ret(greduce.z.x,greduce.z.y);
+    return ret;
+  }
+
+  template<>
+  accelerator_inline Grid::ComplexD
+  Reduce<Grid::ComplexD, GpuVectorCD>::operator()(GpuVectorCD in)
+  {
+    GpuComplexD greduce = in.v[0];
+    for(int i=1;i<GpuVectorCD::N;i++) {
+      greduce = greduce+in.v[i];
+    }
+    Grid::ComplexD ret(greduce.z.x,greduce.z.y);
+    return ret;
+  }
+
+  // Real
+  template<>
+  accelerator_inline Grid::RealF 
+  Reduce<RealF, GpuVectorRF>::operator()(GpuVectorRF in)
+  {
+    RealF ret = in.v[0];
+    for(int i=1;i<GpuVectorRF::N;i++) {
+      ret = ret+in.v[i];
+    }
+    return ret;
+  }
+
+  template<>
+  accelerator_inline Grid::RealD 
+  Reduce<RealD, GpuVectorRD>::operator()(GpuVectorRD in)
+  {
+    RealD ret = in.v[0];
+    for(int i=1;i<GpuVectorRD::N;i++) {
+      ret = ret+in.v[i];
+    }
+    return ret;
+  }
+
+  template<>
+  accelerator_inline Integer
+  Reduce<Integer, GpuVectorI>::operator()(GpuVectorI in)
+  {
+    Integer ret = in.v[0];
+    for(int i=1;i<GpuVectorI::N;i++) {
+      ret = ret+in.v[i];
+    }
+    return ret;
+  }
+
+}// End optimizatoin
+
+//////////////////////////////////////////////////////////////////////////////////////
+// Here assign types 
+//////////////////////////////////////////////////////////////////////////////////////
+  typedef GpuVectorRH  SIMD_Htype; // Single precision type
+  typedef GpuVectorRF  SIMD_Ftype; // Single precision type
+  typedef GpuVectorRD  SIMD_Dtype; // Double precision type
+  typedef GpuVectorI   SIMD_Itype; // Integer type
+
+  typedef GpuVectorCH  SIMD_CHtype; // Single precision type
+  typedef GpuVectorCF  SIMD_CFtype; // Single precision type
+  typedef GpuVectorCD  SIMD_CDtype; // Double precision type
+
+  // prefetch utilities
+  accelerator_inline void v_prefetch0(int size, const char *ptr){};
+  accelerator_inline void prefetch_HINT_T0(const char *ptr){};
+
+  // Function name aliases
+  typedef Optimization::Vsplat   VsplatSIMD;
+  typedef Optimization::Vstore   VstoreSIMD;
+  typedef Optimization::Vset     VsetSIMD;
+  typedef Optimization::Vstream  VstreamSIMD;
+  template <typename S, typename T> using ReduceSIMD = Optimization::Reduce<S,T>;
+
+  // Arithmetic operations
+  typedef Optimization::Sum         SumSIMD;
+  typedef Optimization::Sub         SubSIMD;
+  typedef Optimization::Div         DivSIMD;
+  typedef Optimization::Mult        MultSIMD;
+  typedef Optimization::MultComplex MultComplexSIMD;
+  typedef Optimization::MultRealPart MultRealPartSIMD;
+  typedef Optimization::MaddRealPart MaddRealPartSIMD;
+  typedef Optimization::Conj        ConjSIMD;
+  typedef Optimization::TimesMinusI TimesMinusISIMD;
+  typedef Optimization::TimesI      TimesISIMD;
+
+}
diff --git a/Grid/simd/Grid_neon.h b/Grid/simd/Grid_neon.h
index 9f757fbf..f53575cb 100644
--- a/Grid/simd/Grid_neon.h
+++ b/Grid/simd/Grid_neon.h
@@ -25,8 +25,8 @@
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 
 /*
 
@@ -45,555 +45,551 @@
 #include "Grid_generic_types.h"
 #include <arm_neon.h>
 
-namespace Grid {
-namespace Optimization {
+NAMESPACE_BEGIN(Grid);
+NAMESPACE_BEGIN(Optimization);
 
-  template<class vtype>
-  union uconv {
-    float32x4_t f;
-    vtype v;
-  };
-  union u128f {
-    float32x4_t v;
-    float f[4];
-  };
-  union u128d {
-    float64x2_t v;
-    double f[2];
-  };
-  // half precision
-  union u128h {
-    float16x8_t v;
-    uint16_t f[8];
-  };
+template<class vtype>
+union uconv {
+  float32x4_t f;
+  vtype v;
+};
+union u128f {
+  float32x4_t v;
+  float f[4];
+};
+union u128d {
+  float64x2_t v;
+  double f[2];
+};
+// half precision
+union u128h {
+  float16x8_t v;
+  uint16_t f[8];
+};
 
-  struct Vsplat{
-    //Complex float
-    inline float32x4_t operator()(float a, float b){
-      float tmp[4]={a,b,a,b};
-      return vld1q_f32(tmp);
-    }
-    // Real float
-    inline float32x4_t operator()(float a){
-      return vdupq_n_f32(a);
-    }
-    //Complex double
-    inline float64x2_t operator()(double a, double b){
-      double tmp[2]={a,b};
-      return vld1q_f64(tmp);
-    }
-    //Real double
-    inline float64x2_t operator()(double a){
-      return vdupq_n_f64(a);
-    }
-    //Integer
-    inline uint32x4_t operator()(Integer a){
-      return vdupq_n_u32(a);
-    }
-  };
-
-  struct Vstore{
-    //Float
-    inline void operator()(float32x4_t a, float* F){
-      vst1q_f32(F, a);
-    }
-    //Double
-    inline void operator()(float64x2_t a, double* D){
-      vst1q_f64(D, a);
-    }
-    //Integer
-    inline void operator()(uint32x4_t a, Integer* I){
-      vst1q_u32(I, a);
-    }
-
-  };
-
-  struct Vstream{ // N:equivalents to _mm_stream_p* in NEON?
-    //Float // N:generic
-    inline void operator()(float * a, float32x4_t b){
-      memcpy(a,&b,4*sizeof(float));
-    }
-    //Double // N:generic
-    inline void operator()(double * a, float64x2_t b){
-      memcpy(a,&b,2*sizeof(double));
-    }
-
-
-  };
-
-  // Nils: Vset untested; not used currently in Grid at all;
-  // git commit 4a8c4ccfba1d05159348d21a9698028ea847e77b
-  struct Vset{
-    // Complex float
-    inline float32x4_t operator()(Grid::ComplexF *a){
-      float tmp[4]={a[1].imag(),a[1].real(),a[0].imag(),a[0].real()};
-      return vld1q_f32(tmp);
-    }
-    // Complex double
-    inline float64x2_t operator()(Grid::ComplexD *a){
-      double tmp[2]={a[0].imag(),a[0].real()};
-      return vld1q_f64(tmp);
-    }
-    // Real float
-    inline float32x4_t operator()(float *a){
-      float tmp[4]={a[3],a[2],a[1],a[0]};
-      return vld1q_f32(tmp);
-    }
-    // Real double
-    inline float64x2_t operator()(double *a){
-      double tmp[2]={a[1],a[0]};
-      return vld1q_f64(tmp);
-    }
-    // Integer
-    inline uint32x4_t operator()(Integer *a){
-      return vld1q_dup_u32(a);
-    }
-  };
-
-  template <typename Out_type, typename In_type>
-  struct Reduce{
-    //Need templated class to overload output type
-    //General form must generate error if compiled
-      inline Out_type operator()(In_type in){
-      printf("Error, using wrong Reduce function\n");
-      exit(1);
-      return 0;
-    }
-  };
-
-  /////////////////////////////////////////////////////
-  // Arithmetic operations
-  /////////////////////////////////////////////////////
-  struct Sum{
-    //Complex/Real float
-    inline float32x4_t operator()(float32x4_t a, float32x4_t b){
-      return vaddq_f32(a,b);
-    }
-    //Complex/Real double
-    inline float64x2_t operator()(float64x2_t a, float64x2_t b){
-      return vaddq_f64(a,b);
-    }
-    //Integer
-    inline uint32x4_t operator()(uint32x4_t a, uint32x4_t b){
-      return vaddq_u32(a,b);
-    }
-  };
-
-  struct Sub{
-    //Complex/Real float
-    inline float32x4_t operator()(float32x4_t a, float32x4_t b){
-      return vsubq_f32(a,b);
-    }
-    //Complex/Real double
-    inline float64x2_t operator()(float64x2_t a, float64x2_t b){
-      return vsubq_f64(a,b);
-    }
-    //Integer
-    inline uint32x4_t operator()(uint32x4_t a, uint32x4_t b){
-      return vsubq_u32(a,b);
-    }
-  };
-
-  struct MultRealPart{
-    inline float32x4_t operator()(float32x4_t a, float32x4_t b){
-      float32x4_t re = vtrn1q_f32(a, a);
-      return vmulq_f32(re, b);
-    }
-    inline float64x2_t operator()(float64x2_t a, float64x2_t b){
-      float64x2_t re = vzip1q_f64(a, a);
-      return vmulq_f64(re, b);
-    }
-  };
-
-  struct MaddRealPart{
-    inline float32x4_t operator()(float32x4_t a, float32x4_t b, float32x4_t c){
-      float32x4_t re = vtrn1q_f32(a, a);
-      return vfmaq_f32(c, re, b);
-    }
-    inline float64x2_t operator()(float64x2_t a, float64x2_t b, float64x2_t c){
-      float64x2_t re = vzip1q_f64(a, a);
-      return vfmaq_f64(c, re, b);
-    }
-  };
-
-  struct Div{
-    // Real float
-    inline float32x4_t operator()(float32x4_t a, float32x4_t b){
-      return vdivq_f32(a, b);
-    }
-    // Real double
-    inline float64x2_t operator()(float64x2_t a, float64x2_t b){
-      return vdivq_f64(a, b);
-    }
-  };
-
-  struct MultComplex{
-    // Complex float
-    inline float32x4_t operator()(float32x4_t a, float32x4_t b){
-
-      float32x4_t r0, r1, r2, r3, r4;
-
-      // a = ar ai Ar Ai
-      // b = br bi Br Bi
-      // collect real/imag part, negate bi and Bi
-      r0 = vtrn1q_f32(b, b);       //  br  br  Br  Br
-      r1 = vnegq_f32(b);           // -br -bi -Br -Bi
-      r2 = vtrn2q_f32(b, r1);      //  bi -bi  Bi -Bi
-
-      // the fun part
-      r3 = vmulq_f32(r2, a);       //  bi*ar -bi*ai ...
-      r4 = vrev64q_f32(r3);        // -bi*ai  bi*ar ...
-
-      // fma(a,b,c) = a+b*c
-      return vfmaq_f32(r4, r0, a); //  ar*br-ai*bi ai*br+ar*bi ...
-
-      // no fma, use mul and add
-      // float32x4_t r5;
-      // r5 = vmulq_f32(r0, a);
-      // return vaddq_f32(r4, r5);
-    }
-    // Complex double
-    inline float64x2_t operator()(float64x2_t a, float64x2_t b){
-
-      float64x2_t r0, r1, r2, r3, r4;
-
-      // b = br bi
-      // collect real/imag part, negate bi
-      r0 = vtrn1q_f64(b, b);       //  br  br
-      r1 = vnegq_f64(b);           // -br -bi
-      r2 = vtrn2q_f64(b, r1);      //  bi -bi
-
-      // the fun part
-      r3 = vmulq_f64(r2, a);       //  bi*ar -bi*ai
-      r4 = vextq_f64(r3,r3,1);     // -bi*ai  bi*ar
-
-      // fma(a,b,c) = a+b*c
-      return vfmaq_f64(r4, r0, a); //  ar*br-ai*bi ai*br+ar*bi
-
-      // no fma, use mul and add
-      // float64x2_t r5;
-      // r5 = vmulq_f64(r0, a);
-      // return vaddq_f64(r4, r5);
-    }
-  };
-
-  struct Mult{
-    // Real float
-    inline float32x4_t mac(float32x4_t a, float32x4_t b, float32x4_t c){
-      //return vaddq_f32(vmulq_f32(b,c),a);
-      return vfmaq_f32(a, b, c);
-    }
-    inline float64x2_t mac(float64x2_t a, float64x2_t b, float64x2_t c){
-      //return vaddq_f64(vmulq_f64(b,c),a);
-      return vfmaq_f64(a, b, c);
-    }
-    inline float32x4_t operator()(float32x4_t a, float32x4_t b){
-      return vmulq_f32(a,b);
-    }
-    // Real double
-    inline float64x2_t operator()(float64x2_t a, float64x2_t b){
-      return vmulq_f64(a,b);
-    }
-    // Integer
-    inline uint32x4_t operator()(uint32x4_t a, uint32x4_t b){
-      return vmulq_u32(a,b);
-    }
-  };
-
-  struct Conj{
-    // Complex single
-    inline float32x4_t operator()(float32x4_t in){
-      // ar ai br bi -> ar -ai br -bi
-      float32x4_t r0, r1;
-      r0 = vnegq_f32(in);        // -ar -ai -br -bi
-      r1 = vrev64q_f32(r0);      // -ai -ar -bi -br
-      return vtrn1q_f32(in, r1); //  ar -ai  br -bi
-    }
-    // Complex double
-    inline float64x2_t operator()(float64x2_t in){
-
-      float64x2_t r0, r1;
-      r0 = vextq_f64(in, in, 1);    //  ai  ar
-      r1 = vnegq_f64(r0);           // -ai -ar
-      return vextq_f64(r0, r1, 1);  //  ar -ai
-    }
-    // do not define for integer input
-  };
-
-  struct TimesMinusI{
-    //Complex single
-    inline float32x4_t operator()(float32x4_t in, float32x4_t ret){
-      // ar ai br bi -> ai -ar ai -br
-      float32x4_t r0, r1;
-      r0 = vnegq_f32(in);        // -ar -ai -br -bi
-      r1 = vrev64q_f32(in);      //  ai  ar  bi  br
-      return vtrn1q_f32(r1, r0); //  ar -ai  br -bi
-    }
-    //Complex double
-    inline float64x2_t operator()(float64x2_t in, float64x2_t ret){
-      // a ib -> b -ia
-      float64x2_t tmp;
-      tmp = vnegq_f64(in);
-      return vextq_f64(in, tmp, 1);
-    }
-  };
-
-  struct TimesI{
-    //Complex single
-    inline float32x4_t operator()(float32x4_t in, float32x4_t ret){
-      // ar ai br bi -> -ai ar -bi br
-      float32x4_t r0, r1;
-      r0 = vnegq_f32(in);        // -ar -ai -br -bi
-      r1 = vrev64q_f32(r0);      // -ai -ar -bi -br
-      return vtrn1q_f32(r1, in); // -ai  ar -bi  br
-    }
-    //Complex double
-    inline float64x2_t operator()(float64x2_t in, float64x2_t ret){
-      // a ib -> -b ia
-      float64x2_t tmp;
-      tmp = vnegq_f64(in);
-      return vextq_f64(tmp, in, 1);
-    }
-  };
-
-  struct Permute{
-
-    static inline float32x4_t Permute0(float32x4_t in){ // N:ok
-      // AB CD -> CD AB
-      return vextq_f32(in, in, 2);
-    };
-    static inline float32x4_t Permute1(float32x4_t in){ // N:ok
-      // AB CD -> BA DC
-      return vrev64q_f32(in);
-    };
-    static inline float32x4_t Permute2(float32x4_t in){ // N:not used by Boyle
-      return in;
-    };
-    static inline float32x4_t Permute3(float32x4_t in){ // N:not used by Boyle
-      return in;
-    };
-
-    static inline float64x2_t Permute0(float64x2_t in){ // N:ok
-      // AB -> BA
-      return vextq_f64(in, in, 1);
-    };
-    static inline float64x2_t Permute1(float64x2_t in){ // N:not used by Boyle
-      return in;
-    };
-    static inline float64x2_t Permute2(float64x2_t in){ // N:not used by Boyle
-      return in;
-    };
-    static inline float64x2_t Permute3(float64x2_t in){ // N:not used by Boyle
-      return in;
-    };
-
-  };
-
-  struct Rotate{
-
-    static inline float32x4_t rotate(float32x4_t in,int n){ // N:ok
-      switch(n){
-      case 0: // AB CD -> AB CD
-        return tRotate<0>(in);
-        break;
-      case 1: // AB CD -> BC DA
-        return tRotate<1>(in);
-        break;
-      case 2: // AB CD -> CD AB
-        return tRotate<2>(in);
-        break;
-      case 3: // AB CD -> DA BC
-        return tRotate<3>(in);
-        break;
-      default: assert(0);
-      }
-    }
-    static inline float64x2_t rotate(float64x2_t in,int n){ // N:ok
-      switch(n){
-      case 0: // AB -> AB
-        return tRotate<0>(in);
-        break;
-      case 1: // AB -> BA
-        return tRotate<1>(in);
-        break;
-      default: assert(0);
-      }
-    }
-
-    template<int n> static inline float32x4_t tRotate(float32x4_t in){ return vextq_f32(in,in,n%4); };
-    template<int n> static inline float64x2_t tRotate(float64x2_t in){ return vextq_f64(in,in,n%2); };
-
-  };
-
-  struct PrecisionChange {
-
-    static inline float16x8_t StoH (const float32x4_t &a,const float32x4_t &b) {
-      float16x4_t h = vcvt_f16_f32(a);
-      return vcvt_high_f16_f32(h, b);
-    }
-    static inline void  HtoS (float16x8_t h,float32x4_t &sa,float32x4_t &sb) {
-      sb = vcvt_high_f32_f16(h);
-      // there is no direct conversion from lower float32x4_t to float64x2_t
-      // vextq_f16 not supported by clang 3.8 / 4.0 / arm clang
-      // float16x8_t h1 = vextq_f16(h, h, 4); // correct, but not supported by clang
-      // workaround for clang
-      uint32x4_t h1u = reinterpret_cast<uint32x4_t>(h);
-      float16x8_t h1 = reinterpret_cast<float16x8_t>(vextq_u32(h1u, h1u, 2));
-      sa = vcvt_high_f32_f16(h1);
-    }
-    static inline float32x4_t DtoS (float64x2_t a,float64x2_t b) {
-      float32x2_t s = vcvt_f32_f64(a);
-      return vcvt_high_f32_f64(s, b);
-
-    }
-    static inline void StoD (float32x4_t s,float64x2_t &a,float64x2_t &b) {
-      b = vcvt_high_f64_f32(s);
-      // there is no direct conversion from lower float32x4_t to float64x2_t
-      float32x4_t s1 = vextq_f32(s, s, 2);
-      a = vcvt_high_f64_f32(s1);
-
-    }
-    static inline float16x8_t DtoH (float64x2_t a,float64x2_t b,float64x2_t c,float64x2_t d) {
-      float32x4_t s1 = DtoS(a, b);
-      float32x4_t s2 = DtoS(c, d);
-      return StoH(s1, s2);
-    }
-    static inline void HtoD (float16x8_t h,float64x2_t &a,float64x2_t &b,float64x2_t &c,float64x2_t &d) {
-      float32x4_t s1, s2;
-      HtoS(h, s1, s2);
-      StoD(s1, a, b);
-      StoD(s2, c, d);
-    }
-  };
-
-  //////////////////////////////////////////////
-  // Exchange support
-
-  struct Exchange{
-    static inline void Exchange0(float32x4_t &out1,float32x4_t &out2,float32x4_t in1,float32x4_t in2){
-      // in1: ABCD -> out1: ABEF
-      // in2: EFGH -> out2: CDGH
-
-      // z: CDAB
-      float32x4_t z = vextq_f32(in1, in1, 2);
-      // out1: ABEF
-      out1 = vextq_f32(z, in2, 2);
-
-      // z: GHEF
-      z = vextq_f32(in2, in2, 2);
-      // out2: CDGH
-      out2 = vextq_f32(in1, z, 2);
-    };
-
-    static inline void Exchange1(float32x4_t &out1,float32x4_t &out2,float32x4_t in1,float32x4_t in2){
-      // in1: ABCD -> out1: AECG
-      // in2: EFGH -> out2: BFDH
-      out1 = vtrn1q_f32(in1, in2);
-      out2 = vtrn2q_f32(in1, in2);
-    };
-    static inline void Exchange2(float32x4_t &out1,float32x4_t &out2,float32x4_t in1,float32x4_t in2){
-      assert(0);
-      return;
-    };
-    static inline void Exchange3(float32x4_t &out1,float32x4_t &out2,float32x4_t in1,float32x4_t in2){
-      assert(0);
-      return;
-    };
-    // double precision
-    static inline void Exchange0(float64x2_t &out1,float64x2_t &out2,float64x2_t in1,float64x2_t in2){
-      // in1: AB -> out1: AC
-      // in2: CD -> out2: BD
-      out1 = vzip1q_f64(in1, in2);
-      out2 = vzip2q_f64(in1, in2);
-    };
-    static inline void Exchange1(float64x2_t &out1,float64x2_t &out2,float64x2_t in1,float64x2_t in2){
-      assert(0);
-      return;
-    };
-    static inline void Exchange2(float64x2_t &out1,float64x2_t &out2,float64x2_t in1,float64x2_t in2){
-      assert(0);
-      return;
-    };
-    static inline void Exchange3(float64x2_t &out1,float64x2_t &out2,float64x2_t in1,float64x2_t in2){
-      assert(0);
-      return;
-    };
-  };
-
-  //////////////////////////////////////////////
-  // Some Template specialization
-
-
-  //Complex float Reduce
-  template<>
-  inline Grid::ComplexF Reduce<Grid::ComplexF, float32x4_t>::operator()(float32x4_t in){
-    float32x4_t v1; // two complex
-    v1 = Optimization::Permute::Permute0(in);
-    v1 = vaddq_f32(v1,in);
-    u128f conv;    conv.v=v1;
-    return Grid::ComplexF(conv.f[0],conv.f[1]);
+struct Vsplat{
+  //Complex float
+  inline float32x4_t operator()(float a, float b){
+    float tmp[4]={a,b,a,b};
+    return vld1q_f32(tmp);
   }
-  //Real float Reduce
-  template<>
-  inline Grid::RealF Reduce<Grid::RealF, float32x4_t>::operator()(float32x4_t in){
-    return vaddvq_f32(in);
+  // Real float
+  inline float32x4_t operator()(float a){
+    return vdupq_n_f32(a);
+  }
+  //Complex double
+  inline float64x2_t operator()(double a, double b){
+    double tmp[2]={a,b};
+    return vld1q_f64(tmp);
+  }
+  //Real double
+  inline float64x2_t operator()(double a){
+    return vdupq_n_f64(a);
+  }
+  //Integer
+  inline uint32x4_t operator()(Integer a){
+    return vdupq_n_u32(a);
+  }
+};
+
+struct Vstore{
+  //Float
+  inline void operator()(float32x4_t a, float* F){
+    vst1q_f32(F, a);
+  }
+  //Double
+  inline void operator()(float64x2_t a, double* D){
+    vst1q_f64(D, a);
+  }
+  //Integer
+  inline void operator()(uint32x4_t a, Integer* I){
+    vst1q_u32(I, a);
   }
 
+};
 
-  //Complex double Reduce
-  template<>
-  inline Grid::ComplexD Reduce<Grid::ComplexD, float64x2_t>::operator()(float64x2_t in){
-    u128d conv; conv.v = in;
-    return Grid::ComplexD(conv.f[0],conv.f[1]);
+struct Vstream{ // N:equivalents to _mm_stream_p* in NEON?
+  //Float // N:generic
+  inline void operator()(float * a, float32x4_t b){
+    memcpy(a,&b,4*sizeof(float));
+  }
+  //Double // N:generic
+  inline void operator()(double * a, float64x2_t b){
+    memcpy(a,&b,2*sizeof(double));
+  }
+};
+
+// Nils: Vset untested; not used currently in Grid at all;
+// git commit 4a8c4ccfba1d05159348d21a9698028ea847e77b
+struct Vset{
+  // Complex float
+  inline float32x4_t operator()(Grid::ComplexF *a){
+    float tmp[4]={a[1].imag(),a[1].real(),a[0].imag(),a[0].real()};
+    return vld1q_f32(tmp);
+  }
+  // Complex double
+  inline float64x2_t operator()(Grid::ComplexD *a){
+    double tmp[2]={a[0].imag(),a[0].real()};
+    return vld1q_f64(tmp);
+  }
+  // Real float
+  inline float32x4_t operator()(float *a){
+    float tmp[4]={a[3],a[2],a[1],a[0]};
+    return vld1q_f32(tmp);
+  }
+  // Real double
+  inline float64x2_t operator()(double *a){
+    double tmp[2]={a[1],a[0]};
+    return vld1q_f64(tmp);
+  }
+  // Integer
+  inline uint32x4_t operator()(Integer *a){
+    return vld1q_dup_u32(a);
+  }
+};
+
+template <typename Out_type, typename In_type>
+struct Reduce{
+  //Need templated class to overload output type
+  //General form must generate error if compiled
+  inline Out_type operator()(In_type in){
+    printf("Error, using wrong Reduce function\n");
+    exit(1);
+    return 0;
+  }
+};
+
+/////////////////////////////////////////////////////
+// Arithmetic operations
+/////////////////////////////////////////////////////
+struct Sum{
+  //Complex/Real float
+  inline float32x4_t operator()(float32x4_t a, float32x4_t b){
+    return vaddq_f32(a,b);
+  }
+  //Complex/Real double
+  inline float64x2_t operator()(float64x2_t a, float64x2_t b){
+    return vaddq_f64(a,b);
+  }
+  //Integer
+  inline uint32x4_t operator()(uint32x4_t a, uint32x4_t b){
+    return vaddq_u32(a,b);
+  }
+};
+
+struct Sub{
+  //Complex/Real float
+  inline float32x4_t operator()(float32x4_t a, float32x4_t b){
+    return vsubq_f32(a,b);
+  }
+  //Complex/Real double
+  inline float64x2_t operator()(float64x2_t a, float64x2_t b){
+    return vsubq_f64(a,b);
+  }
+  //Integer
+  inline uint32x4_t operator()(uint32x4_t a, uint32x4_t b){
+    return vsubq_u32(a,b);
+  }
+};
+
+struct MultRealPart{
+  inline float32x4_t operator()(float32x4_t a, float32x4_t b){
+    float32x4_t re = vtrn1q_f32(a, a);
+    return vmulq_f32(re, b);
+  }
+  inline float64x2_t operator()(float64x2_t a, float64x2_t b){
+    float64x2_t re = vzip1q_f64(a, a);
+    return vmulq_f64(re, b);
+  }
+};
+
+struct MaddRealPart{
+  inline float32x4_t operator()(float32x4_t a, float32x4_t b, float32x4_t c){
+    float32x4_t re = vtrn1q_f32(a, a);
+    return vfmaq_f32(c, re, b);
+  }
+  inline float64x2_t operator()(float64x2_t a, float64x2_t b, float64x2_t c){
+    float64x2_t re = vzip1q_f64(a, a);
+    return vfmaq_f64(c, re, b);
+  }
+};
+
+struct Div{
+  // Real float
+  inline float32x4_t operator()(float32x4_t a, float32x4_t b){
+    return vdivq_f32(a, b);
+  }
+  // Real double
+  inline float64x2_t operator()(float64x2_t a, float64x2_t b){
+    return vdivq_f64(a, b);
+  }
+};
+
+struct MultComplex{
+  // Complex float
+  inline float32x4_t operator()(float32x4_t a, float32x4_t b){
+
+    float32x4_t r0, r1, r2, r3, r4;
+
+    // a = ar ai Ar Ai
+    // b = br bi Br Bi
+    // collect real/imag part, negate bi and Bi
+    r0 = vtrn1q_f32(b, b);       //  br  br  Br  Br
+    r1 = vnegq_f32(b);           // -br -bi -Br -Bi
+    r2 = vtrn2q_f32(b, r1);      //  bi -bi  Bi -Bi
+
+    // the fun part
+    r3 = vmulq_f32(r2, a);       //  bi*ar -bi*ai ...
+    r4 = vrev64q_f32(r3);        // -bi*ai  bi*ar ...
+
+    // fma(a,b,c) = a+b*c
+    return vfmaq_f32(r4, r0, a); //  ar*br-ai*bi ai*br+ar*bi ...
+
+    // no fma, use mul and add
+    // float32x4_t r5;
+    // r5 = vmulq_f32(r0, a);
+    // return vaddq_f32(r4, r5);
+  }
+  // Complex double
+  inline float64x2_t operator()(float64x2_t a, float64x2_t b){
+
+    float64x2_t r0, r1, r2, r3, r4;
+
+    // b = br bi
+    // collect real/imag part, negate bi
+    r0 = vtrn1q_f64(b, b);       //  br  br
+    r1 = vnegq_f64(b);           // -br -bi
+    r2 = vtrn2q_f64(b, r1);      //  bi -bi
+
+    // the fun part
+    r3 = vmulq_f64(r2, a);       //  bi*ar -bi*ai
+    r4 = vextq_f64(r3,r3,1);     // -bi*ai  bi*ar
+
+    // fma(a,b,c) = a+b*c
+    return vfmaq_f64(r4, r0, a); //  ar*br-ai*bi ai*br+ar*bi
+
+    // no fma, use mul and add
+    // float64x2_t r5;
+    // r5 = vmulq_f64(r0, a);
+    // return vaddq_f64(r4, r5);
+  }
+};
+
+struct Mult{
+  // Real float
+  inline float32x4_t mac(float32x4_t a, float32x4_t b, float32x4_t c){
+    //return vaddq_f32(vmulq_f32(b,c),a);
+    return vfmaq_f32(a, b, c);
+  }
+  inline float64x2_t mac(float64x2_t a, float64x2_t b, float64x2_t c){
+    //return vaddq_f64(vmulq_f64(b,c),a);
+    return vfmaq_f64(a, b, c);
+  }
+  inline float32x4_t operator()(float32x4_t a, float32x4_t b){
+    return vmulq_f32(a,b);
+  }
+  // Real double
+  inline float64x2_t operator()(float64x2_t a, float64x2_t b){
+    return vmulq_f64(a,b);
+  }
+  // Integer
+  inline uint32x4_t operator()(uint32x4_t a, uint32x4_t b){
+    return vmulq_u32(a,b);
+  }
+};
+
+struct Conj{
+  // Complex single
+  inline float32x4_t operator()(float32x4_t in){
+    // ar ai br bi -> ar -ai br -bi
+    float32x4_t r0, r1;
+    r0 = vnegq_f32(in);        // -ar -ai -br -bi
+    r1 = vrev64q_f32(r0);      // -ai -ar -bi -br
+    return vtrn1q_f32(in, r1); //  ar -ai  br -bi
+  }
+  // Complex double
+  inline float64x2_t operator()(float64x2_t in){
+
+    float64x2_t r0, r1;
+    r0 = vextq_f64(in, in, 1);    //  ai  ar
+    r1 = vnegq_f64(r0);           // -ai -ar
+    return vextq_f64(r0, r1, 1);  //  ar -ai
+  }
+  // do not define for integer input
+};
+
+struct TimesMinusI{
+  //Complex single
+  inline float32x4_t operator()(float32x4_t in, float32x4_t ret){
+    // ar ai br bi -> ai -ar ai -br
+    float32x4_t r0, r1;
+    r0 = vnegq_f32(in);        // -ar -ai -br -bi
+    r1 = vrev64q_f32(in);      //  ai  ar  bi  br
+    return vtrn1q_f32(r1, r0); //  ar -ai  br -bi
+  }
+  //Complex double
+  inline float64x2_t operator()(float64x2_t in, float64x2_t ret){
+    // a ib -> b -ia
+    float64x2_t tmp;
+    tmp = vnegq_f64(in);
+    return vextq_f64(in, tmp, 1);
+  }
+};
+
+struct TimesI{
+  //Complex single
+  inline float32x4_t operator()(float32x4_t in, float32x4_t ret){
+    // ar ai br bi -> -ai ar -bi br
+    float32x4_t r0, r1;
+    r0 = vnegq_f32(in);        // -ar -ai -br -bi
+    r1 = vrev64q_f32(r0);      // -ai -ar -bi -br
+    return vtrn1q_f32(r1, in); // -ai  ar -bi  br
+  }
+  //Complex double
+  inline float64x2_t operator()(float64x2_t in, float64x2_t ret){
+    // a ib -> -b ia
+    float64x2_t tmp;
+    tmp = vnegq_f64(in);
+    return vextq_f64(tmp, in, 1);
+  }
+};
+
+struct Permute{
+
+  static inline float32x4_t Permute0(float32x4_t in){ // N:ok
+    // AB CD -> CD AB
+    return vextq_f32(in, in, 2);
+  };
+  static inline float32x4_t Permute1(float32x4_t in){ // N:ok
+    // AB CD -> BA DC
+    return vrev64q_f32(in);
+  };
+  static inline float32x4_t Permute2(float32x4_t in){ // N:not used by Boyle
+    return in;
+  };
+  static inline float32x4_t Permute3(float32x4_t in){ // N:not used by Boyle
+    return in;
+  };
+
+  static inline float64x2_t Permute0(float64x2_t in){ // N:ok
+    // AB -> BA
+    return vextq_f64(in, in, 1);
+  };
+  static inline float64x2_t Permute1(float64x2_t in){ // N:not used by Boyle
+    return in;
+  };
+  static inline float64x2_t Permute2(float64x2_t in){ // N:not used by Boyle
+    return in;
+  };
+  static inline float64x2_t Permute3(float64x2_t in){ // N:not used by Boyle
+    return in;
+  };
+
+};
+
+struct Rotate{
+
+  static inline float32x4_t rotate(float32x4_t in,int n){ // N:ok
+    switch(n){
+    case 0: // AB CD -> AB CD
+      return tRotate<0>(in);
+      break;
+    case 1: // AB CD -> BC DA
+      return tRotate<1>(in);
+      break;
+    case 2: // AB CD -> CD AB
+      return tRotate<2>(in);
+      break;
+    case 3: // AB CD -> DA BC
+      return tRotate<3>(in);
+      break;
+    default: assert(0);
+    }
+  }
+  static inline float64x2_t rotate(float64x2_t in,int n){ // N:ok
+    switch(n){
+    case 0: // AB -> AB
+      return tRotate<0>(in);
+      break;
+    case 1: // AB -> BA
+      return tRotate<1>(in);
+      break;
+    default: assert(0);
+    }
   }
 
-  //Real double Reduce
-  template<>
-  inline Grid::RealD Reduce<Grid::RealD, float64x2_t>::operator()(float64x2_t in){
-    return vaddvq_f64(in);
-  }
+  template<int n> static inline float32x4_t tRotate(float32x4_t in){ return vextq_f32(in,in,n%4); };
+  template<int n> static inline float64x2_t tRotate(float64x2_t in){ return vextq_f64(in,in,n%2); };
 
-  //Integer Reduce
-  template<>
-  inline Integer Reduce<Integer, uint32x4_t>::operator()(uint32x4_t in){
-    return vaddvq_u32(in);
+};
+
+struct PrecisionChange {
+
+  static inline float16x8_t StoH (const float32x4_t &a,const float32x4_t &b) {
+    float16x4_t h = vcvt_f16_f32(a);
+    return vcvt_high_f16_f32(h, b);
   }
+  static inline void  HtoS (float16x8_t h,float32x4_t &sa,float32x4_t &sb) {
+    sb = vcvt_high_f32_f16(h);
+    // there is no direct conversion from lower float32x4_t to float64x2_t
+    // vextq_f16 not supported by clang 3.8 / 4.0 / arm clang
+    // float16x8_t h1 = vextq_f16(h, h, 4); // correct, but not supported by clang
+    // workaround for clang
+    uint32x4_t h1u = reinterpret_cast<uint32x4_t>(h);
+    float16x8_t h1 = reinterpret_cast<float16x8_t>(vextq_u32(h1u, h1u, 2));
+    sa = vcvt_high_f32_f16(h1);
+  }
+  static inline float32x4_t DtoS (float64x2_t a,float64x2_t b) {
+    float32x2_t s = vcvt_f32_f64(a);
+    return vcvt_high_f32_f64(s, b);
+
+  }
+  static inline void StoD (float32x4_t s,float64x2_t &a,float64x2_t &b) {
+    b = vcvt_high_f64_f32(s);
+    // there is no direct conversion from lower float32x4_t to float64x2_t
+    float32x4_t s1 = vextq_f32(s, s, 2);
+    a = vcvt_high_f64_f32(s1);
+
+  }
+  static inline float16x8_t DtoH (float64x2_t a,float64x2_t b,float64x2_t c,float64x2_t d) {
+    float32x4_t s1 = DtoS(a, b);
+    float32x4_t s2 = DtoS(c, d);
+    return StoH(s1, s2);
+  }
+  static inline void HtoD (float16x8_t h,float64x2_t &a,float64x2_t &b,float64x2_t &c,float64x2_t &d) {
+    float32x4_t s1, s2;
+    HtoS(h, s1, s2);
+    StoD(s1, a, b);
+    StoD(s2, c, d);
+  }
+};
+
+//////////////////////////////////////////////
+// Exchange support
+
+struct Exchange{
+  static inline void Exchange0(float32x4_t &out1,float32x4_t &out2,float32x4_t in1,float32x4_t in2){
+    // in1: ABCD -> out1: ABEF
+    // in2: EFGH -> out2: CDGH
+
+    // z: CDAB
+    float32x4_t z = vextq_f32(in1, in1, 2);
+    // out1: ABEF
+    out1 = vextq_f32(z, in2, 2);
+
+    // z: GHEF
+    z = vextq_f32(in2, in2, 2);
+    // out2: CDGH
+    out2 = vextq_f32(in1, z, 2);
+  };
+
+  static inline void Exchange1(float32x4_t &out1,float32x4_t &out2,float32x4_t in1,float32x4_t in2){
+    // in1: ABCD -> out1: AECG
+    // in2: EFGH -> out2: BFDH
+    out1 = vtrn1q_f32(in1, in2);
+    out2 = vtrn2q_f32(in1, in2);
+  };
+  static inline void Exchange2(float32x4_t &out1,float32x4_t &out2,float32x4_t in1,float32x4_t in2){
+    assert(0);
+    return;
+  };
+  static inline void Exchange3(float32x4_t &out1,float32x4_t &out2,float32x4_t in1,float32x4_t in2){
+    assert(0);
+    return;
+  };
+  // double precision
+  static inline void Exchange0(float64x2_t &out1,float64x2_t &out2,float64x2_t in1,float64x2_t in2){
+    // in1: AB -> out1: AC
+    // in2: CD -> out2: BD
+    out1 = vzip1q_f64(in1, in2);
+    out2 = vzip2q_f64(in1, in2);
+  };
+  static inline void Exchange1(float64x2_t &out1,float64x2_t &out2,float64x2_t in1,float64x2_t in2){
+    assert(0);
+    return;
+  };
+  static inline void Exchange2(float64x2_t &out1,float64x2_t &out2,float64x2_t in1,float64x2_t in2){
+    assert(0);
+    return;
+  };
+  static inline void Exchange3(float64x2_t &out1,float64x2_t &out2,float64x2_t in1,float64x2_t in2){
+    assert(0);
+    return;
+  };
+};
+
+//////////////////////////////////////////////
+// Some Template specialization
+
+
+//Complex float Reduce
+template<>
+inline Grid::ComplexF Reduce<Grid::ComplexF, float32x4_t>::operator()(float32x4_t in){
+  float32x4_t v1; // two complex
+  v1 = Optimization::Permute::Permute0(in);
+  v1 = vaddq_f32(v1,in);
+  u128f conv;    conv.v=v1;
+  return Grid::ComplexF(conv.f[0],conv.f[1]);
 }
+//Real float Reduce
+template<>
+inline Grid::RealF Reduce<Grid::RealF, float32x4_t>::operator()(float32x4_t in){
+  return vaddvq_f32(in);
+}
+
+
+//Complex double Reduce
+template<>
+inline Grid::ComplexD Reduce<Grid::ComplexD, float64x2_t>::operator()(float64x2_t in){
+  u128d conv; conv.v = in;
+  return Grid::ComplexD(conv.f[0],conv.f[1]);
+}
+
+//Real double Reduce
+template<>
+inline Grid::RealD Reduce<Grid::RealD, float64x2_t>::operator()(float64x2_t in){
+  return vaddvq_f64(in);
+}
+
+//Integer Reduce
+template<>
+inline Integer Reduce<Integer, uint32x4_t>::operator()(uint32x4_t in){
+  return vaddvq_u32(in);
+}
+
+NAMESPACE_END(Optimization);
 
 //////////////////////////////////////////////////////////////////////////////////////
 // Here assign types
 
 // typedef Optimization::vech SIMD_Htype; // Reduced precision type
-  typedef float16x8_t  SIMD_Htype; // Half precision type
-  typedef float32x4_t  SIMD_Ftype; // Single precision type
-  typedef float64x2_t  SIMD_Dtype; // Double precision type
-  typedef uint32x4_t   SIMD_Itype; // Integer type
+typedef float16x8_t  SIMD_Htype; // Half precision type
+typedef float32x4_t  SIMD_Ftype; // Single precision type
+typedef float64x2_t  SIMD_Dtype; // Double precision type
+typedef uint32x4_t   SIMD_Itype; // Integer type
 
-  inline void v_prefetch0(int size, const char *ptr){};  // prefetch utilities
-  inline void prefetch_HINT_T0(const char *ptr){};
+inline void v_prefetch0(int size, const char *ptr){};  // prefetch utilities
+inline void prefetch_HINT_T0(const char *ptr){};
 
 
-  // Function name aliases
-  typedef Optimization::Vsplat   VsplatSIMD;
-  typedef Optimization::Vstore   VstoreSIMD;
-  typedef Optimization::Vset     VsetSIMD;
-  typedef Optimization::Vstream  VstreamSIMD;
-  template <typename S, typename T> using ReduceSIMD = Optimization::Reduce<S,T>;
+// Function name aliases
+typedef Optimization::Vsplat   VsplatSIMD;
+typedef Optimization::Vstore   VstoreSIMD;
+typedef Optimization::Vset     VsetSIMD;
+typedef Optimization::Vstream  VstreamSIMD;
+template <typename S, typename T> using ReduceSIMD = Optimization::Reduce<S,T>;
 
+// Arithmetic operations
+typedef Optimization::Sum         SumSIMD;
+typedef Optimization::Sub         SubSIMD;
+typedef Optimization::Div         DivSIMD;
+typedef Optimization::Mult        MultSIMD;
+typedef Optimization::MultComplex MultComplexSIMD;
+typedef Optimization::MultRealPart MultRealPartSIMD;
+typedef Optimization::MaddRealPart MaddRealPartSIMD;
+typedef Optimization::Conj        ConjSIMD;
+typedef Optimization::TimesMinusI TimesMinusISIMD;
+typedef Optimization::TimesI      TimesISIMD;
 
-
-
-  // Arithmetic operations
-  typedef Optimization::Sum         SumSIMD;
-  typedef Optimization::Sub         SubSIMD;
-  typedef Optimization::Div         DivSIMD;
-  typedef Optimization::Mult        MultSIMD;
-  typedef Optimization::MultComplex MultComplexSIMD;
-  typedef Optimization::MultRealPart MultRealPartSIMD;
-  typedef Optimization::MaddRealPart MaddRealPartSIMD;
-  typedef Optimization::Conj        ConjSIMD;
-  typedef Optimization::TimesMinusI TimesMinusISIMD;
-  typedef Optimization::TimesI      TimesISIMD;
-
-}
+NAMESPACE_END(Grid);
 
diff --git a/Grid/simd/Grid_qpx.h b/Grid/simd/Grid_qpx.h
index 8de7bde8..d93dbc5b 100644
--- a/Grid/simd/Grid_qpx.h
+++ b/Grid/simd/Grid_qpx.h
@@ -25,14 +25,14 @@
  51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  
  See the full license in the file "LICENSE" in the top level distribution directory
- ******************************************************************************/
+******************************************************************************/
 
 #ifndef GEN_SIMD_WIDTH
 #define GEN_SIMD_WIDTH 32u
 #endif
 #include "Grid_generic_types.h" // Definitions for simulated integer SIMD.
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 #ifdef QPX
 #include <spi/include/kernel/location.h>
@@ -41,549 +41,551 @@ namespace Grid {
 #include <hwi/include/bqc/A2_inlines.h>
 #endif
 
-namespace Optimization {
-  typedef struct 
-  {
-    float v0,v1,v2,v3;
-  } vector4float;
+NAMESPACE_BEGIN(Optimization);
 
-  inline std::ostream & operator<<(std::ostream& stream, const vector4double a)
-  {
-    stream << "{"<<vec_extract(a,0)<<","<<vec_extract(a,1)<<","<<vec_extract(a,2)<<","<<vec_extract(a,3)<<"}";
-    return stream;
-  };
+typedef struct 
+{
+  float v0,v1,v2,v3;
+} vector4float;
 
-  inline std::ostream & operator<<(std::ostream& stream, const vector4float a)
-  {
-    stream << "{"<< a.v0 <<","<< a.v1 <<","<< a.v2 <<","<< a.v3 <<"}";
-    return stream;
-  };
+inline std::ostream & operator<<(std::ostream& stream, const vector4double a)
+{
+  stream << "{"<<vec_extract(a,0)<<","<<vec_extract(a,1)<<","<<vec_extract(a,2)<<","<<vec_extract(a,3)<<"}";
+  return stream;
+};
+
+inline std::ostream & operator<<(std::ostream& stream, const vector4float a)
+{
+  stream << "{"<< a.v0 <<","<< a.v1 <<","<< a.v2 <<","<< a.v3 <<"}";
+  return stream;
+};
   
-  struct Vsplat{
-    //Complex float
-    inline vector4float operator()(float a, float b){
-      return (vector4float){a, b, a, b};
-    }
-    // Real float
-    inline vector4float operator()(float a){
-      return (vector4float){a, a, a, a};
-    }
-    //Complex double
-    inline vector4double operator()(double a, double b){
-      return (vector4double){a, b, a, b};
-    }
-    //Real double
-    inline vector4double operator()(double a){
-      return (vector4double){a, a, a, a};
-    }
-    //Integer
-    inline veci operator()(Integer a){
-      veci out;
+struct Vsplat{
+  //Complex float
+  inline vector4float operator()(float a, float b){
+    return (vector4float){a, b, a, b};
+  }
+  // Real float
+  inline vector4float operator()(float a){
+    return (vector4float){a, a, a, a};
+  }
+  //Complex double
+  inline vector4double operator()(double a, double b){
+    return (vector4double){a, b, a, b};
+  }
+  //Real double
+  inline vector4double operator()(double a){
+    return (vector4double){a, a, a, a};
+  }
+  //Integer
+  inline veci operator()(Integer a){
+    veci out;
       
-      VECTOR_FOR(i, W<Integer>::r, 1)
+    VECTOR_FOR(i, W<Integer>::r, 1)
       {
         out.v[i] = a;
       }
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
   
-  struct Vstore{
-    //Float
-    inline void operator()(vector4double a, float *f){
-      vec_st(a, 0, f);
-    }
+struct Vstore{
+  //Float
+  inline void operator()(vector4double a, float *f){
+    vec_st(a, 0, f);
+  }
 
-    inline void operator()(vector4double a, vector4float &f){
-      vec_st(a, 0, (float *)(&f));
-    }
+  inline void operator()(vector4double a, vector4float &f){
+    vec_st(a, 0, (float *)(&f));
+  }
 
-    inline void operator()(vector4float a, float *f){
-      f[0] = a.v0;
-      f[1] = a.v1;
-      f[2] = a.v2;
-      f[3] = a.v3;
-    }
+  inline void operator()(vector4float a, float *f){
+    f[0] = a.v0;
+    f[1] = a.v1;
+    f[2] = a.v2;
+    f[3] = a.v3;
+  }
 
-    //Double
-    inline void operator()(vector4double a, double *d){
-      vec_st(a, 0, d);
-    }
+  //Double
+  inline void operator()(vector4double a, double *d){
+    vec_st(a, 0, d);
+  }
 
-    //Integer
-    inline void operator()(veci a, Integer *i){
-      *((veci *)i) = a;
-    }
-  };
+  //Integer
+  inline void operator()(veci a, Integer *i){
+    *((veci *)i) = a;
+  }
+};
   
-  struct Vstream{
-    //Float
-    inline void operator()(float *f, vector4double a){
-      vec_st(a, 0, f);
-    }
+struct Vstream{
+  //Float
+  inline void operator()(float *f, vector4double a){
+    vec_st(a, 0, f);
+  }
 
-    inline void operator()(vector4float f, vector4double a){
-      vec_st(a, 0, (float *)(&f));
-    }
+  inline void operator()(vector4float f, vector4double a){
+    vec_st(a, 0, (float *)(&f));
+  }
 
-    inline void operator()(float *f, vector4float a){
-      f[0] = a.v0;
-      f[1] = a.v1;
-      f[2] = a.v2;
-      f[3] = a.v3;
-    }
-    //Double
-    inline void operator()(double *d, vector4double a){
-      vec_st(a, 0, d);
-    }
+  inline void operator()(float *f, vector4float a){
+    f[0] = a.v0;
+    f[1] = a.v1;
+    f[2] = a.v2;
+    f[3] = a.v3;
+  }
+  //Double
+  inline void operator()(double *d, vector4double a){
+    vec_st(a, 0, d);
+  }
 
-  };
+};
   
-  struct Vset{
-    // Complex float
-    inline vector4float operator()(Grid::ComplexF *a){
-      return (vector4float){a[0].real(), a[0].imag(), a[1].real(), a[1].imag()};
-    }
-    // Complex double
-    inline vector4double operator()(Grid::ComplexD *a){
-      return vec_ld(0, (double *)a);
-    }
+struct Vset{
+  // Complex float
+  inline vector4float operator()(Grid::ComplexF *a){
+    return (vector4float){a[0].real(), a[0].imag(), a[1].real(), a[1].imag()};
+  }
+  // Complex double
+  inline vector4double operator()(Grid::ComplexD *a){
+    return vec_ld(0, (double *)a);
+  }
 
-    // Real float
-    inline vector4float operator()(float *a){
-      return (vector4float){a[0], a[1], a[2], a[3]};
-    }
+  // Real float
+  inline vector4float operator()(float *a){
+    return (vector4float){a[0], a[1], a[2], a[3]};
+  }
 
-    inline vector4double operator()(vector4float a){
-      return vec_ld(0, (float *)(&a));
-    }
+  inline vector4double operator()(vector4float a){
+    return vec_ld(0, (float *)(&a));
+  }
 
-    // Real double
-    inline vector4double operator()(double *a){
-      return vec_ld(0, a);
-    }
-    // Integer
-    inline veci operator()(Integer *a){
-      veci out;
+  // Real double
+  inline vector4double operator()(double *a){
+    return vec_ld(0, a);
+  }
+  // Integer
+  inline veci operator()(Integer *a){
+    veci out;
       
-      out = *((veci *)a);
+    out = *((veci *)a);
       
-      return out;
-    }    
-  };
+    return out;
+  }    
+};
   
-  template <typename Out_type, typename In_type>
-  struct Reduce{
-    //Need templated class to overload output type
-    //General form must generate error if compiled
-    inline Out_type operator()(In_type in){
-      printf("Error, using wrong Reduce function\n");
-      exit(1);
-      return 0;
-    }
-  };
+template <typename Out_type, typename In_type>
+struct Reduce{
+  //Need templated class to overload output type
+  //General form must generate error if compiled
+  inline Out_type operator()(In_type in){
+    printf("Error, using wrong Reduce function\n");
+    exit(1);
+    return 0;
+  }
+};
   
-  /////////////////////////////////////////////////////
-  // Arithmetic operations
-  /////////////////////////////////////////////////////
+/////////////////////////////////////////////////////
+// Arithmetic operations
+/////////////////////////////////////////////////////
 
-  #define FLOAT_WRAP_3(fn, pref)\
+#define FLOAT_WRAP_3(fn, pref)						\
   pref vector4float fn(vector4float a, vector4float b, vector4float c)	\
-  {\
-    vector4double ad, bd, rd, cd;			\
-    vector4float  r;\
-    \
-    ad = Vset()(a);\
-    bd = Vset()(b);\
-    cd = Vset()(c);\
-    rd = fn(ad, bd, cd);				\
-    Vstore()(rd, r);\
-    \
-    return r;\
+  {									\
+    vector4double ad, bd, rd, cd;					\
+    vector4float  r;							\
+									\
+    ad = Vset()(a);							\
+    bd = Vset()(b);							\
+    cd = Vset()(c);							\
+    rd = fn(ad, bd, cd);						\
+    Vstore()(rd, r);							\
+									\
+    return r;								\
   }
 
-  #define FLOAT_WRAP_2(fn, pref)\
-  pref vector4float fn(vector4float a, vector4float b)\
-  {\
-    vector4double ad, bd, rd;\
-    vector4float  r;\
-    \
-    ad = Vset()(a);\
-    bd = Vset()(b);\
-    rd = fn(ad, bd);\
-    Vstore()(rd, r);\
-    \
-    return r;\
+#define FLOAT_WRAP_2(fn, pref)				\
+  pref vector4float fn(vector4float a, vector4float b)	\
+  {							\
+    vector4double ad, bd, rd;				\
+    vector4float  r;					\
+							\
+    ad = Vset()(a);					\
+    bd = Vset()(b);					\
+    rd = fn(ad, bd);					\
+    Vstore()(rd, r);					\
+							\
+    return r;						\
   }
 
-  #define FLOAT_WRAP_1(fn, pref)\
-  pref vector4float fn(vector4float a)\
-  {\
-    vector4double ad, rd;\
-    vector4float  r;\
-    \
-    ad = Vset()(a);\
-    rd = fn(ad);\
-    Vstore()(rd, r);\
-    \
-    return r;\
+#define FLOAT_WRAP_1(fn, pref)			\
+  pref vector4float fn(vector4float a)		\
+  {						\
+    vector4double ad, rd;			\
+    vector4float  r;				\
+						\
+    ad = Vset()(a);				\
+    rd = fn(ad);				\
+    Vstore()(rd, r);				\
+						\
+    return r;					\
   }
 
-  struct Sum{
-    //Complex/Real double
-    inline vector4double operator()(vector4double a, vector4double b){
-      return vec_add(a, b);
-    }
+struct Sum{
+  //Complex/Real double
+  inline vector4double operator()(vector4double a, vector4double b){
+    return vec_add(a, b);
+  }
 
-    //Complex/Real float
-    FLOAT_WRAP_2(operator(), inline)
+  //Complex/Real float
+  FLOAT_WRAP_2(operator(), inline)
 
-    //Integer
-    inline veci operator()(veci a, veci b){
-      veci out;
+  //Integer
+  inline veci operator()(veci a, veci b){
+    veci out;
       
-      VECTOR_FOR(i, W<Integer>::r, 1)
+    VECTOR_FOR(i, W<Integer>::r, 1)
       {
         out.v[i] = a.v[i] + b.v[i];
       }
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
   
-  struct Sub{
-    //Complex/Real double
-    inline vector4double operator()(vector4double a, vector4double b){
-      return vec_sub(a, b);
-    }
+struct Sub{
+  //Complex/Real double
+  inline vector4double operator()(vector4double a, vector4double b){
+    return vec_sub(a, b);
+  }
 
-    //Complex/Real float
-    FLOAT_WRAP_2(operator(), inline)
+  //Complex/Real float
+  FLOAT_WRAP_2(operator(), inline)
 
-    //Integer
-    inline veci operator()(veci a, veci b){
-      veci out;
+  //Integer
+  inline veci operator()(veci a, veci b){
+    veci out;
       
-      VECTOR_FOR(i, W<Integer>::r, 1)
+    VECTOR_FOR(i, W<Integer>::r, 1)
       {
         out.v[i] = a.v[i] - b.v[i];
       }
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
   
-  struct MultRealPart{
-    // Complex double
-    inline vector4double operator()(vector4double a, vector4double b){
-  //      return vec_xmul(b, a);
-        return vec_xmul(a, b);
-    }
-    FLOAT_WRAP_2(operator(), inline)
-  };
-  struct MaddRealPart{
-    // Complex double
-    inline vector4double operator()(vector4double a, vector4double b,vector4double c){
-      return vec_xmadd(a, b, c);
-    }
-    FLOAT_WRAP_3(operator(), inline)
-  };
-  struct MultComplex{
-    // Complex double
-    inline vector4double operator()(vector4double a, vector4double b){
-      return vec_xxnpmadd(a, b, vec_xmul(b, a));
-    }
+struct MultRealPart{
+  // Complex double
+  inline vector4double operator()(vector4double a, vector4double b){
+    //      return vec_xmul(b, a);
+    return vec_xmul(a, b);
+  }
+  FLOAT_WRAP_2(operator(), inline)
+};
+struct MaddRealPart{
+  // Complex double
+  inline vector4double operator()(vector4double a, vector4double b,vector4double c){
+    return vec_xmadd(a, b, c);
+  }
+  FLOAT_WRAP_3(operator(), inline)
+};
+struct MultComplex{
+  // Complex double
+  inline vector4double operator()(vector4double a, vector4double b){
+    return vec_xxnpmadd(a, b, vec_xmul(b, a));
+  }
 
-    // Complex float
-    FLOAT_WRAP_2(operator(), inline)
-  };
+  // Complex float
+  FLOAT_WRAP_2(operator(), inline)
+};
   
-  struct Mult{
-    // Real double
-    inline vector4double operator()(vector4double a, vector4double b){
-      return vec_mul(a, b);
-    }
+struct Mult{
+  // Real double
+  inline vector4double operator()(vector4double a, vector4double b){
+    return vec_mul(a, b);
+  }
 
-    // Real float
-    FLOAT_WRAP_2(operator(), inline)
+  // Real float
+  FLOAT_WRAP_2(operator(), inline)
 
-    // Integer
-    inline veci operator()(veci a, veci b){
-      veci out;
+  // Integer
+  inline veci operator()(veci a, veci b){
+    veci out;
       
-      VECTOR_FOR(i, W<Integer>::r, 1)
+    VECTOR_FOR(i, W<Integer>::r, 1)
       {
         out.v[i] = a.v[i]*b.v[i];
       }
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
 
-  struct Div{
-    // Real double
-    inline vector4double operator()(vector4double a, vector4double b){
-      return vec_swdiv(a, b);
-    }
+struct Div{
+  // Real double
+  inline vector4double operator()(vector4double a, vector4double b){
+    return vec_swdiv(a, b);
+  }
 
-    // Real float
-    FLOAT_WRAP_2(operator(), inline)
+  // Real float
+  FLOAT_WRAP_2(operator(), inline)
 
-    // Integer
-    inline veci operator()(veci a, veci b){
-      veci out;
+  // Integer
+  inline veci operator()(veci a, veci b){
+    veci out;
       
-      VECTOR_FOR(i, W<Integer>::r, 1)
+    VECTOR_FOR(i, W<Integer>::r, 1)
       {
         out.v[i] = a.v[i]/b.v[i];
       }
       
-      return out;
-    }
-  };
+    return out;
+  }
+};
 
-  struct Conj{
-    // Complex double
-    inline vector4double operator()(vector4double v){
-      return vec_mul(v, (vector4double){1., -1., 1., -1.});
-    }
+struct Conj{
+  // Complex double
+  inline vector4double operator()(vector4double v){
+    return vec_mul(v, (vector4double){1., -1., 1., -1.});
+  }
 
-    // Complex float
-    FLOAT_WRAP_1(operator(), inline)
-  };
+  // Complex float
+  FLOAT_WRAP_1(operator(), inline)
+};
   
-  struct TimesMinusI{
-    //Complex double
-    inline vector4double operator()(vector4double v, vector4double ret){
-      return vec_xxcpnmadd(v, (vector4double){1., 1., 1., 1.},
-                               (vector4double){0., 0., 0., 0.});
-    }
+struct TimesMinusI{
+  //Complex double
+  inline vector4double operator()(vector4double v, vector4double ret){
+    return vec_xxcpnmadd(v, (vector4double){1., 1., 1., 1.},
+			 (vector4double){0., 0., 0., 0.});
+  }
 
-    // Complex float
-    FLOAT_WRAP_2(operator(), inline)
-  };
+  // Complex float
+  FLOAT_WRAP_2(operator(), inline)
+};
   
-  struct TimesI{
-    //Complex double
-    inline vector4double operator()(vector4double v, vector4double ret){
-      return vec_xxcpnmadd(v, (vector4double){-1., -1., -1., -1.},
-                              (vector4double){0., 0., 0., 0.});
-    }
+struct TimesI{
+  //Complex double
+  inline vector4double operator()(vector4double v, vector4double ret){
+    return vec_xxcpnmadd(v, (vector4double){-1., -1., -1., -1.},
+			 (vector4double){0., 0., 0., 0.});
+  }
 
-    // Complex float
-    FLOAT_WRAP_2(operator(), inline)
-  };
+  // Complex float
+  FLOAT_WRAP_2(operator(), inline)
+};
 #define USE_FP16
-  struct PrecisionChange {
-    static inline vech StoH (const vector4float &a, const vector4float &b) {
-      vech ret;
-      std::cout << GridLogError << "QPX single to half precision conversion not yet supported." << std::endl;
-      assert(0);
-      return ret;
-    }
-    static inline void  HtoS (vech h, vector4float &sa, vector4float &sb) {
-      std::cout << GridLogError << "QPX half to single precision conversion not yet supported." << std::endl;
-      assert(0);
-    }
-    static inline vector4float DtoS (vector4double a, vector4double b) {
-      vector4float ret;
-      std::cout << GridLogError << "QPX double to single precision conversion not yet supported." << std::endl;
-      assert(0);
-      return ret;
-    }
-    static inline void StoD (vector4float s, vector4double &a, vector4double &b) {
-      std::cout << GridLogError << "QPX single to double precision conversion not yet supported." << std::endl;
-      assert(0);
-    }
-    static inline vech DtoH (vector4double a, vector4double b, 
-                             vector4double c, vector4double d) {
-      vech ret;
-      std::cout << GridLogError << "QPX double to half precision conversion not yet supported." << std::endl;
-      assert(0);
-      return ret;
-    }
-    static inline void HtoD (vech h, vector4double &a, vector4double &b, 
-                                     vector4double &c, vector4double &d) {
-      std::cout << GridLogError << "QPX half to double precision conversion not yet supported." << std::endl;
-      assert(0);
-    }
-  };
+struct PrecisionChange {
+  static inline vech StoH (const vector4float &a, const vector4float &b) {
+    vech ret;
+    std::cout << GridLogError << "QPX single to half precision conversion not yet supported." << std::endl;
+    assert(0);
+    return ret;
+  }
+  static inline void  HtoS (vech h, vector4float &sa, vector4float &sb) {
+    std::cout << GridLogError << "QPX half to single precision conversion not yet supported." << std::endl;
+    assert(0);
+  }
+  static inline vector4float DtoS (vector4double a, vector4double b) {
+    vector4float ret;
+    std::cout << GridLogError << "QPX double to single precision conversion not yet supported." << std::endl;
+    assert(0);
+    return ret;
+  }
+  static inline void StoD (vector4float s, vector4double &a, vector4double &b) {
+    std::cout << GridLogError << "QPX single to double precision conversion not yet supported." << std::endl;
+    assert(0);
+  }
+  static inline vech DtoH (vector4double a, vector4double b, 
+			   vector4double c, vector4double d) {
+    vech ret;
+    std::cout << GridLogError << "QPX double to half precision conversion not yet supported." << std::endl;
+    assert(0);
+    return ret;
+  }
+  static inline void HtoD (vech h, vector4double &a, vector4double &b, 
+			   vector4double &c, vector4double &d) {
+    std::cout << GridLogError << "QPX half to double precision conversion not yet supported." << std::endl;
+    assert(0);
+  }
+};
 
-  //////////////////////////////////////////////
-  // Exchange support
-#define FLOAT_WRAP_EXCHANGE(fn) \
+//////////////////////////////////////////////
+// Exchange support
+#define FLOAT_WRAP_EXCHANGE(fn)					\
   static inline void fn(vector4float &out1, vector4float &out2, \
-                        vector4float in1,  vector4float in2) \
-  { \
-    vector4double out1d, out2d, in1d, in2d; \
-    in1d  = Vset()(in1);   \
-    in2d  = Vset()(in2);   \
-    fn(out1d, out2d, in1d, in2d); \
-    Vstore()(out1d, out1); \
-    Vstore()(out2d, out2); \
+                        vector4float in1,  vector4float in2)	\
+  {								\
+    vector4double out1d, out2d, in1d, in2d;			\
+    in1d  = Vset()(in1);					\
+    in2d  = Vset()(in2);					\
+    fn(out1d, out2d, in1d, in2d);				\
+    Vstore()(out1d, out1);					\
+    Vstore()(out2d, out2);					\
   }
 
-  struct Exchange{
+struct Exchange{
 
-    // double precision
-    static inline void Exchange0(vector4double &out1, vector4double &out2,
-                                 vector4double in1,  vector4double in2) {
-      out1 = vec_perm(in1, in2, vec_gpci(0145));
-      out2 = vec_perm(in1, in2, vec_gpci(02367));
-    }
-    static inline void Exchange1(vector4double &out1, vector4double &out2,
-                                 vector4double in1,  vector4double in2) {
-      out1 = vec_perm(in1, in2, vec_gpci(0426));
-      out2 = vec_perm(in1, in2, vec_gpci(01537));
-    }
-    static inline void Exchange2(vector4double &out1, vector4double &out2,
-                                 vector4double in1,  vector4double in2) {
-      assert(0);
-    }
-    static inline void Exchange3(vector4double &out1, vector4double &out2,
-                                 vector4double in1,  vector4double in2) {
-      assert(0);
-    }
+  // double precision
+  static inline void Exchange0(vector4double &out1, vector4double &out2,
+			       vector4double in1,  vector4double in2) {
+    out1 = vec_perm(in1, in2, vec_gpci(0145));
+    out2 = vec_perm(in1, in2, vec_gpci(02367));
+  }
+  static inline void Exchange1(vector4double &out1, vector4double &out2,
+			       vector4double in1,  vector4double in2) {
+    out1 = vec_perm(in1, in2, vec_gpci(0426));
+    out2 = vec_perm(in1, in2, vec_gpci(01537));
+  }
+  static inline void Exchange2(vector4double &out1, vector4double &out2,
+			       vector4double in1,  vector4double in2) {
+    assert(0);
+  }
+  static inline void Exchange3(vector4double &out1, vector4double &out2,
+			       vector4double in1,  vector4double in2) {
+    assert(0);
+  }
 
-    // single precision
-    FLOAT_WRAP_EXCHANGE(Exchange0);
-    FLOAT_WRAP_EXCHANGE(Exchange1);
-    FLOAT_WRAP_EXCHANGE(Exchange2);
-    FLOAT_WRAP_EXCHANGE(Exchange3);
+  // single precision
+  FLOAT_WRAP_EXCHANGE(Exchange0);
+  FLOAT_WRAP_EXCHANGE(Exchange1);
+  FLOAT_WRAP_EXCHANGE(Exchange2);
+  FLOAT_WRAP_EXCHANGE(Exchange3);
+};
+
+struct Permute{
+  //Complex double
+  static inline vector4double Permute0(vector4double v){ //0123 -> 2301
+    return vec_perm(v, v, vec_gpci(02301));
+  };
+  static inline vector4double Permute1(vector4double v){ //0123 -> 1032
+    return vec_perm(v, v, vec_gpci(01032));
+  };
+  static inline vector4double Permute2(vector4double v){
+    return v;
+  };
+  static inline vector4double Permute3(vector4double v){
+    return v;
   };
 
-  struct Permute{
-    //Complex double
-    static inline vector4double Permute0(vector4double v){ //0123 -> 2301
-      return vec_perm(v, v, vec_gpci(02301));
-    };
-    static inline vector4double Permute1(vector4double v){ //0123 -> 1032
-      return vec_perm(v, v, vec_gpci(01032));
-    };
-    static inline vector4double Permute2(vector4double v){
-      return v;
-    };
-    static inline vector4double Permute3(vector4double v){
-      return v;
-    };
+  // Complex float
+  FLOAT_WRAP_1(Permute0, static inline)
+  FLOAT_WRAP_1(Permute1, static inline)
+  FLOAT_WRAP_1(Permute2, static inline)
+  FLOAT_WRAP_1(Permute3, static inline)
+};
+  
+struct Rotate{
 
-    // Complex float
-    FLOAT_WRAP_1(Permute0, static inline)
-    FLOAT_WRAP_1(Permute1, static inline)
-    FLOAT_WRAP_1(Permute2, static inline)
-    FLOAT_WRAP_1(Permute3, static inline)
+  template<int n> static inline vector4double tRotate(vector4double v){ 
+    if ( n==1 ) return vec_perm(v, v, vec_gpci(01230));
+    if ( n==2 ) return vec_perm(v, v, vec_gpci(02301));
+    if ( n==3 ) return vec_perm(v, v, vec_gpci(03012));
+    return v;
   };
-  
-  struct Rotate{
-
-    template<int n> static inline vector4double tRotate(vector4double v){ 
-      if ( n==1 ) return vec_perm(v, v, vec_gpci(01230));
-      if ( n==2 ) return vec_perm(v, v, vec_gpci(02301));
-      if ( n==3 ) return vec_perm(v, v, vec_gpci(03012));
-      return v;
-    };
-    template<int n> static inline vector4float tRotate(vector4float a)	
-    {					       
-      vector4double ad, rd;
-      vector4float  r;
-      ad = Vset()(a);
-      rd = tRotate<n>(ad);
-      Vstore()(rd, r);
-      return r;
-    };
-
-    static inline vector4double rotate(vector4double v, int n){
-      switch(n){
-        case 0:
-          return v;
-          break;
-        case 1:
-          return tRotate<1>(v);
-          break;
-        case 2:
-          return tRotate<2>(v);
-          break;
-        case 3:
-          return tRotate<3>(v);
-          break;
-        default: assert(0);
-      }
-    }
-
-    static inline vector4float rotate(vector4float v, int n){
-      vector4double vd, rd;
-      vector4float  r;
-      vd = Vset()(v);
-      rd = rotate(vd, n);
-      Vstore()(rd, r);
-      return r;
-    }
+  template<int n> static inline vector4float tRotate(vector4float a)	
+  {					       
+    vector4double ad, rd;
+    vector4float  r;
+    ad = Vset()(a);
+    rd = tRotate<n>(ad);
+    Vstore()(rd, r);
+    return r;
   };
-  
-  //Complex float Reduce
-  template<>
-  inline Grid::ComplexF
-  Reduce<Grid::ComplexF, vector4float>::operator()(vector4float v) { //2 complex
-    vector4float v1,v2;
-    
-    v1 = Optimization::Permute::Permute0(v);
-    v1 = Optimization::Sum()(v1, v);
-    
-    return Grid::ComplexF(v1.v0, v1.v1);
-  }
-  //Real float Reduce
-  template<>
-  inline Grid::RealF
-  Reduce<Grid::RealF, vector4float>::operator()(vector4float v){ //4 floats
-    vector4float v1,v2;
-    
-    v1 = Optimization::Permute::Permute0(v);
-    v1 = Optimization::Sum()(v1, v);
-    v2 = Optimization::Permute::Permute1(v1);
-    v1 = Optimization::Sum()(v1, v2);
-    
-    return v1.v0;
-  }
-  
-  
-  //Complex double Reduce
-  template<>
-  inline Grid::ComplexD
-  Reduce<Grid::ComplexD, vector4double>::operator()(vector4double v){ //2 complex
-    vector4double v1;
-    
-    v1 = Optimization::Permute::Permute0(v);
-    v1 = vec_add(v1, v);
-    
-    return Grid::ComplexD(vec_extract(v1, 0), vec_extract(v1, 1));
-  }
-  
-  //Real double Reduce
-  template<>
-  inline Grid::RealD
-  Reduce<Grid::RealD, vector4double>::operator()(vector4double v){ //4 doubles
-    vector4double v1,v2;
-    
-    v1 = Optimization::Permute::Permute0(v);
-    v1 = vec_add(v1, v);
-    v2 = Optimization::Permute::Permute1(v1);
-    v1 = vec_add(v1, v2);
 
-    return vec_extract(v1, 0);
-  }
-  
-  //Integer Reduce
-  template<>
-  inline Integer Reduce<Integer, veci>::operator()(veci in){
-    Integer a = 0;
-    for (unsigned int i = 0; i < W<Integer>::r; ++i)
-    {
-        a += in.v[i];
+  static inline vector4double rotate(vector4double v, int n){
+    switch(n){
+    case 0:
+      return v;
+      break;
+    case 1:
+      return tRotate<1>(v);
+      break;
+    case 2:
+      return tRotate<2>(v);
+      break;
+    case 3:
+      return tRotate<3>(v);
+      break;
+    default: assert(0);
     }
-    return a;
   }
+
+  static inline vector4float rotate(vector4float v, int n){
+    vector4double vd, rd;
+    vector4float  r;
+    vd = Vset()(v);
+    rd = rotate(vd, n);
+    Vstore()(rd, r);
+    return r;
+  }
+};
+  
+//Complex float Reduce
+template<>
+inline Grid::ComplexF
+Reduce<Grid::ComplexF, vector4float>::operator()(vector4float v) { //2 complex
+  vector4float v1,v2;
+    
+  v1 = Optimization::Permute::Permute0(v);
+  v1 = Optimization::Sum()(v1, v);
+    
+  return Grid::ComplexF(v1.v0, v1.v1);
 }
+//Real float Reduce
+template<>
+inline Grid::RealF
+Reduce<Grid::RealF, vector4float>::operator()(vector4float v){ //4 floats
+  vector4float v1,v2;
+    
+  v1 = Optimization::Permute::Permute0(v);
+  v1 = Optimization::Sum()(v1, v);
+  v2 = Optimization::Permute::Permute1(v1);
+  v1 = Optimization::Sum()(v1, v2);
+    
+  return v1.v0;
+}
+  
+  
+//Complex double Reduce
+template<>
+inline Grid::ComplexD
+Reduce<Grid::ComplexD, vector4double>::operator()(vector4double v){ //2 complex
+  vector4double v1;
+    
+  v1 = Optimization::Permute::Permute0(v);
+  v1 = vec_add(v1, v);
+    
+  return Grid::ComplexD(vec_extract(v1, 0), vec_extract(v1, 1));
+}
+  
+//Real double Reduce
+template<>
+inline Grid::RealD
+Reduce<Grid::RealD, vector4double>::operator()(vector4double v){ //4 doubles
+  vector4double v1,v2;
+    
+  v1 = Optimization::Permute::Permute0(v);
+  v1 = vec_add(v1, v);
+  v2 = Optimization::Permute::Permute1(v1);
+  v1 = vec_add(v1, v2);
+
+  return vec_extract(v1, 0);
+}
+  
+//Integer Reduce
+template<>
+inline Integer Reduce<Integer, veci>::operator()(veci in){
+  Integer a = 0;
+  for (unsigned int i = 0; i < W<Integer>::r; ++i)
+    {
+      a += in.v[i];
+    }
+  return a;
+}
+
+NAMESPACE_END(Optimization);
 
 ////////////////////////////////////////////////////////////////////////////////
 // Here assign types
@@ -596,7 +598,6 @@ typedef Optimization::veci         SIMD_Itype; // Integer type
 inline void v_prefetch0(int size, const char *ptr){};
 inline void prefetch_HINT_T0(const char *ptr){};
 
-
 // Function name aliases
 typedef Optimization::Vsplat   VsplatSIMD;
 typedef Optimization::Vstore   VstoreSIMD;
@@ -615,5 +616,5 @@ typedef Optimization::MaddRealPart MaddRealPartSIMD;
 typedef Optimization::Conj        ConjSIMD;
 typedef Optimization::TimesMinusI TimesMinusISIMD;
 typedef Optimization::TimesI      TimesISIMD;
-  
-}
+ 
+NAMESPACE_END(Grid)
diff --git a/Grid/simd/Grid_sse4.h b/Grid/simd/Grid_sse4.h
index 0b1f9ffb..eb76427e 100644
--- a/Grid/simd/Grid_sse4.h
+++ b/Grid/simd/Grid_sse4.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,8 +25,8 @@ Author: neo <cossu@post.kek.jp>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 //----------------------------------------------------------------------
 /*! @file Grid_sse4.h
   @brief Optimization libraries for SSE4 instructions set
@@ -38,580 +38,505 @@ Author: neo <cossu@post.kek.jp>
 
 #include <pmmintrin.h>
 
-namespace Grid {
-namespace Optimization {
+NAMESPACE_BEGIN(Grid);
+NAMESPACE_BEGIN(Optimization);
 
-  template<class vtype>
-  union uconv {
-    __m128 f;
-    vtype v;
-  };
+template<class vtype>
+union uconv {
+  __m128 f;
+  vtype v;
+};
 
-  union u128f {
-    __m128 v;
-    float f[4];
-  };
-  union u128d {
-    __m128d v;
-    double f[2];
-  };
+union u128f {
+  __m128 v;
+  float f[4];
+};
+union u128d {
+  __m128d v;
+  double f[2];
+};
   
-  struct Vsplat{
-    //Complex float
-    inline __m128 operator()(float a, float b){
-      return _mm_set_ps(b,a,b,a);
-    }
-    // Real float
-    inline __m128 operator()(float a){
-      return _mm_set_ps(a,a,a,a);
-    }
-    //Complex double
-    inline __m128d operator()(double a, double b){
-      return _mm_set_pd(b,a);
-    }
-    //Real double
-    inline __m128d operator()(double a){
-      return _mm_set_pd(a,a);
-    }
-    //Integer
-    inline __m128i operator()(Integer a){
-      return _mm_set1_epi32(a);
-    }
+struct Vsplat{
+  //Complex float
+  inline __m128 operator()(float a, float b){
+    return _mm_set_ps(b,a,b,a);
+  }
+  // Real float
+  inline __m128 operator()(float a){
+    return _mm_set_ps(a,a,a,a);
+  }
+  //Complex double
+  inline __m128d operator()(double a, double b){
+    return _mm_set_pd(b,a);
+  }
+  //Real double
+  inline __m128d operator()(double a){
+    return _mm_set_pd(a,a);
+  }
+  //Integer
+  inline __m128i operator()(Integer a){
+    return _mm_set1_epi32(a);
+  }
+};
+
+struct Vstore{
+  //Float 
+  inline void operator()(__m128 a, float* F){
+    _mm_store_ps(F,a);
+  }
+  //Double
+  inline void operator()(__m128d a, double* D){
+    _mm_store_pd(D,a);
+  }
+  //Integer
+  inline void operator()(__m128i a, Integer* I){
+    _mm_store_si128((__m128i *)I,a);
+  }
+
+};
+
+struct Vstream{
+  //Float
+  inline void operator()(float * a, __m128 b){
+    _mm_stream_ps(a,b);
+  }
+  //Double
+  inline void operator()(double * a, __m128d b){
+    _mm_stream_pd(a,b);
+  }
+
+
+};
+
+struct Vset{
+  // Complex float 
+  inline __m128 operator()(Grid::ComplexF *a){
+    return _mm_set_ps(a[1].imag(), a[1].real(),a[0].imag(),a[0].real());
+  }
+  // Complex double 
+  inline __m128d operator()(Grid::ComplexD *a){
+    return _mm_set_pd(a[0].imag(),a[0].real());
+  }
+  // Real float 
+  inline __m128 operator()(float *a){
+    return _mm_set_ps(a[3],a[2],a[1],a[0]);
+  }
+  // Real double
+  inline __m128d operator()(double *a){
+    return _mm_set_pd(a[1],a[0]);
+  }
+  // Integer
+  inline __m128i operator()(Integer *a){
+    return _mm_set_epi32(a[3],a[2],a[1],a[0]);
+  }
+
+
+};
+
+template <typename Out_type, typename In_type>
+struct Reduce{
+  //Need templated class to overload output type
+  //General form must generate error if compiled
+  inline Out_type operator()(In_type in){
+    printf("Error, using wrong Reduce function\n");
+    exit(1);
+    return 0;
+  }
+};
+
+/////////////////////////////////////////////////////
+// Arithmetic operations
+/////////////////////////////////////////////////////
+struct Sum{
+  //Complex/Real float
+  inline __m128 operator()(__m128 a, __m128 b){
+    return _mm_add_ps(a,b);
+  }
+  //Complex/Real double
+  inline __m128d operator()(__m128d a, __m128d b){
+    return _mm_add_pd(a,b);
+  }
+  //Integer
+  inline __m128i operator()(__m128i a, __m128i b){
+    return _mm_add_epi32(a,b);
+  }
+};
+
+struct Sub{
+  //Complex/Real float
+  inline __m128 operator()(__m128 a, __m128 b){
+    return _mm_sub_ps(a,b);
+  }
+  //Complex/Real double
+  inline __m128d operator()(__m128d a, __m128d b){
+    return _mm_sub_pd(a,b);
+  }
+  //Integer
+  inline __m128i operator()(__m128i a, __m128i b){
+    return _mm_sub_epi32(a,b);
+  }
+};
+
+struct MultRealPart{
+  inline __m128 operator()(__m128 a, __m128 b){
+    __m128 ymm0;
+    ymm0  = _mm_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ymm0 <- ar ar,
+    return  _mm_mul_ps(ymm0,b);                       // ymm0 <- ar bi, ar br
+  }
+  inline __m128d operator()(__m128d a, __m128d b){
+    __m128d ymm0;
+    ymm0 = _mm_shuffle_pd(a,a,0x0); // ymm0 <- ar ar, ar,ar b'00,00
+    return _mm_mul_pd(ymm0,b);      // ymm0 <- ar bi, ar br
+  }
+};
+struct MaddRealPart{
+  inline __m128 operator()(__m128 a, __m128 b, __m128 c){
+    __m128 ymm0 =  _mm_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ymm0 <- ar ar,
+    return _mm_add_ps(_mm_mul_ps( ymm0, b),c);                         
+  }
+  inline __m128d operator()(__m128d a, __m128d b, __m128d c){
+    __m128d ymm0 = _mm_shuffle_pd( a, a, 0x0 );
+    return _mm_add_pd(_mm_mul_pd( ymm0, b),c);                         
+  }
+};
+
+struct MultComplex{
+  // Complex float
+  inline __m128 operator()(__m128 a, __m128 b){
+    __m128 ymm0,ymm1,ymm2;
+    ymm0 = _mm_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ymm0 <- ar ar,
+    ymm0 = _mm_mul_ps(ymm0,b);                       // ymm0 <- ar bi, ar br
+    ymm1 = _mm_shuffle_ps(b,b,_MM_SELECT_FOUR_FOUR(2,3,0,1)); // ymm1 <- br,bi
+    ymm2 = _mm_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(3,3,1,1)); // ymm2 <- ai,ai
+    ymm1 = _mm_mul_ps(ymm1,ymm2);                    // ymm1 <- br ai, ai bi
+    return _mm_addsub_ps(ymm0,ymm1);    
+  }
+  // Complex double
+  inline __m128d operator()(__m128d a, __m128d b){
+    __m128d ymm0,ymm1,ymm2;
+    ymm0 = _mm_shuffle_pd(a,a,0x0);   // ymm0 <- ar ar,
+    ymm0 = _mm_mul_pd(ymm0,b);        // ymm0 <- ar bi, ar br
+    ymm1 = _mm_shuffle_pd(b,b,0x1);   // ymm1 <- br,bi   b01
+    ymm2 = _mm_shuffle_pd(a,a,0x3);   // ymm2 <- ai,ai   b11
+    ymm1 = _mm_mul_pd(ymm1,ymm2);     // ymm1 <- br ai, ai bi
+    return _mm_addsub_pd(ymm0,ymm1);  
+  }
+};
+
+struct Mult{
+
+  inline void mac(__m128 &a, __m128 b, __m128 c){
+    a= _mm_add_ps(_mm_mul_ps(b,c),a);
+  }
+
+  inline void mac(__m128d &a, __m128d b, __m128d c){
+    a= _mm_add_pd(_mm_mul_pd(b,c),a);
+  }
+
+  // Real float
+  inline __m128 operator()(__m128 a, __m128 b){
+    return _mm_mul_ps(a,b);
+  }
+  // Real double
+  inline __m128d operator()(__m128d a, __m128d b){
+    return _mm_mul_pd(a,b);
+  }
+  // Integer
+  inline __m128i operator()(__m128i a, __m128i b){
+    return _mm_mullo_epi32(a,b);
+  }
+};
+
+struct Div{
+  // Real float
+  inline __m128 operator()(__m128 a, __m128 b){
+    return _mm_div_ps(a,b);
+  }
+  // Real double
+  inline __m128d operator()(__m128d a, __m128d b){
+    return _mm_div_pd(a,b);
+  }
+};
+
+
+struct Conj{
+  // Complex single
+  inline __m128 operator()(__m128 in){
+    return _mm_xor_ps(_mm_addsub_ps(_mm_setzero_ps(),in), _mm_set1_ps(-0.f));
+  }
+  // Complex double
+  inline __m128d operator()(__m128d in){
+    return _mm_xor_pd(_mm_addsub_pd(_mm_setzero_pd(),in), _mm_set1_pd(-0.f));//untested
+  }
+  // do not define for integer input
+};
+
+struct TimesMinusI{
+  //Complex single
+  inline __m128 operator()(__m128 in, __m128 ret){
+    __m128 tmp =_mm_addsub_ps(_mm_setzero_ps(),in); // r,-i
+    return _mm_shuffle_ps(tmp,tmp,_MM_SELECT_FOUR_FOUR(2,3,0,1));
+  }
+  //Complex double
+  inline __m128d operator()(__m128d in, __m128d ret){
+    __m128d tmp =_mm_addsub_pd(_mm_setzero_pd(),in); // r,-i
+    return _mm_shuffle_pd(tmp,tmp,0x1);
+  }
+
+
+};
+
+struct TimesI{
+  //Complex single
+  inline __m128 operator()(__m128 in, __m128 ret){
+    __m128 tmp =_mm_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
+    return _mm_addsub_ps(_mm_setzero_ps(),tmp); // r,-i
+  }
+  //Complex double
+  inline __m128d operator()(__m128d in, __m128d ret){
+    __m128d tmp = _mm_shuffle_pd(in,in,0x1);
+    return _mm_addsub_pd(_mm_setzero_pd(),tmp); // r,-i
+  }
+};
+
+struct Permute{
+
+  static inline __m128 Permute0(__m128 in){
+    return _mm_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(1,0,3,2)); //AB CD -> CD AB
+  };
+  static inline __m128 Permute1(__m128 in){
+    return _mm_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1)); //AB CD -> BA DC
+  };
+  static inline __m128 Permute2(__m128 in){
+    return in;
+  };
+  static inline __m128 Permute3(__m128 in){
+    return in;
   };
 
-  struct Vstore{
-    //Float 
-    inline void operator()(__m128 a, float* F){
-      _mm_store_ps(F,a);
-    }
-    //Double
-    inline void operator()(__m128d a, double* D){
-      _mm_store_pd(D,a);
-    }
-    //Integer
-    inline void operator()(__m128i a, Integer* I){
-      _mm_store_si128((__m128i *)I,a);
-    }
-
+  static inline __m128d Permute0(__m128d in){ //AB -> BA
+    return _mm_shuffle_pd(in,in,0x1);
   };
-
-  struct Vstream{
-    //Float
-    inline void operator()(float * a, __m128 b){
-      _mm_stream_ps(a,b);
-    }
-    //Double
-    inline void operator()(double * a, __m128d b){
-      _mm_stream_pd(a,b);
-    }
-
-
+  static inline __m128d Permute1(__m128d in){
+    return in;
   };
-
-  struct Vset{
-    // Complex float 
-    inline __m128 operator()(Grid::ComplexF *a){
-      return _mm_set_ps(a[1].imag(), a[1].real(),a[0].imag(),a[0].real());
-    }
-    // Complex double 
-    inline __m128d operator()(Grid::ComplexD *a){
-      return _mm_set_pd(a[0].imag(),a[0].real());
-    }
-    // Real float 
-    inline __m128 operator()(float *a){
-      return _mm_set_ps(a[3],a[2],a[1],a[0]);
-    }
-    // Real double
-    inline __m128d operator()(double *a){
-      return _mm_set_pd(a[1],a[0]);
-    }
-    // Integer
-    inline __m128i operator()(Integer *a){
-      return _mm_set_epi32(a[3],a[2],a[1],a[0]);
-    }
-
-
+  static inline __m128d Permute2(__m128d in){
+    return in;
   };
-
-  template <typename Out_type, typename In_type>
-  struct Reduce{
-    //Need templated class to overload output type
-    //General form must generate error if compiled
-    inline Out_type operator()(In_type in){
-      printf("Error, using wrong Reduce function\n");
-      exit(1);
-      return 0;
-    }
+  static inline __m128d Permute3(__m128d in){
+    return in;
   };
-
-  /////////////////////////////////////////////////////
-  // Arithmetic operations
-  /////////////////////////////////////////////////////
-  struct Sum{
-    //Complex/Real float
-    inline __m128 operator()(__m128 a, __m128 b){
-      return _mm_add_ps(a,b);
-    }
-    //Complex/Real double
-    inline __m128d operator()(__m128d a, __m128d b){
-      return _mm_add_pd(a,b);
-    }
-    //Integer
-    inline __m128i operator()(__m128i a, __m128i b){
-      return _mm_add_epi32(a,b);
-    }
-  };
-
-  struct Sub{
-    //Complex/Real float
-    inline __m128 operator()(__m128 a, __m128 b){
-      return _mm_sub_ps(a,b);
-    }
-    //Complex/Real double
-    inline __m128d operator()(__m128d a, __m128d b){
-      return _mm_sub_pd(a,b);
-    }
-    //Integer
-    inline __m128i operator()(__m128i a, __m128i b){
-      return _mm_sub_epi32(a,b);
-    }
-  };
-
-  struct MultRealPart{
-    inline __m128 operator()(__m128 a, __m128 b){
-      __m128 ymm0;
-      ymm0  = _mm_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ymm0 <- ar ar,
-      return  _mm_mul_ps(ymm0,b);                       // ymm0 <- ar bi, ar br
-    }
-    inline __m128d operator()(__m128d a, __m128d b){
-      __m128d ymm0;
-      ymm0 = _mm_shuffle_pd(a,a,0x0); // ymm0 <- ar ar, ar,ar b'00,00
-      return _mm_mul_pd(ymm0,b);      // ymm0 <- ar bi, ar br
-    }
-  };
-  struct MaddRealPart{
-    inline __m128 operator()(__m128 a, __m128 b, __m128 c){
-      __m128 ymm0 =  _mm_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ymm0 <- ar ar,
-      return _mm_add_ps(_mm_mul_ps( ymm0, b),c);                         
-    }
-    inline __m128d operator()(__m128d a, __m128d b, __m128d c){
-      __m128d ymm0 = _mm_shuffle_pd( a, a, 0x0 );
-      return _mm_add_pd(_mm_mul_pd( ymm0, b),c);                         
-    }
-  };
-
-  struct MultComplex{
-    // Complex float
-    inline __m128 operator()(__m128 a, __m128 b){
-      __m128 ymm0,ymm1,ymm2;
-      ymm0 = _mm_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(2,2,0,0)); // ymm0 <- ar ar,
-      ymm0 = _mm_mul_ps(ymm0,b);                       // ymm0 <- ar bi, ar br
-      ymm1 = _mm_shuffle_ps(b,b,_MM_SELECT_FOUR_FOUR(2,3,0,1)); // ymm1 <- br,bi
-      ymm2 = _mm_shuffle_ps(a,a,_MM_SELECT_FOUR_FOUR(3,3,1,1)); // ymm2 <- ai,ai
-      ymm1 = _mm_mul_ps(ymm1,ymm2);                    // ymm1 <- br ai, ai bi
-      return _mm_addsub_ps(ymm0,ymm1);    
-    }
-    // Complex double
-    inline __m128d operator()(__m128d a, __m128d b){
-      __m128d ymm0,ymm1,ymm2;
-      ymm0 = _mm_shuffle_pd(a,a,0x0);   // ymm0 <- ar ar,
-      ymm0 = _mm_mul_pd(ymm0,b);        // ymm0 <- ar bi, ar br
-      ymm1 = _mm_shuffle_pd(b,b,0x1);   // ymm1 <- br,bi   b01
-      ymm2 = _mm_shuffle_pd(a,a,0x3);   // ymm2 <- ai,ai   b11
-      ymm1 = _mm_mul_pd(ymm1,ymm2);     // ymm1 <- br ai, ai bi
-      return _mm_addsub_pd(ymm0,ymm1);  
-    }
-  };
-
-  struct Mult{
-
-    inline void mac(__m128 &a, __m128 b, __m128 c){
-      a= _mm_add_ps(_mm_mul_ps(b,c),a);
-    }
-
-    inline void mac(__m128d &a, __m128d b, __m128d c){
-      a= _mm_add_pd(_mm_mul_pd(b,c),a);
-    }
-
-    // Real float
-    inline __m128 operator()(__m128 a, __m128 b){
-      return _mm_mul_ps(a,b);
-    }
-    // Real double
-    inline __m128d operator()(__m128d a, __m128d b){
-      return _mm_mul_pd(a,b);
-    }
-    // Integer
-    inline __m128i operator()(__m128i a, __m128i b){
-      return _mm_mullo_epi32(a,b);
-    }
-  };
-
-  struct Div{
-    // Real float
-    inline __m128 operator()(__m128 a, __m128 b){
-      return _mm_div_ps(a,b);
-    }
-    // Real double
-    inline __m128d operator()(__m128d a, __m128d b){
-      return _mm_div_pd(a,b);
-    }
-  };
-
-
-  struct Conj{
-    // Complex single
-    inline __m128 operator()(__m128 in){
-      return _mm_xor_ps(_mm_addsub_ps(_mm_setzero_ps(),in), _mm_set1_ps(-0.f));
-    }
-    // Complex double
-    inline __m128d operator()(__m128d in){
-      return _mm_xor_pd(_mm_addsub_pd(_mm_setzero_pd(),in), _mm_set1_pd(-0.f));//untested
-    }
-    // do not define for integer input
-  };
-
-  struct TimesMinusI{
-    //Complex single
-    inline __m128 operator()(__m128 in, __m128 ret){
-      __m128 tmp =_mm_addsub_ps(_mm_setzero_ps(),in); // r,-i
-      return _mm_shuffle_ps(tmp,tmp,_MM_SELECT_FOUR_FOUR(2,3,0,1));
-    }
-    //Complex double
-    inline __m128d operator()(__m128d in, __m128d ret){
-      __m128d tmp =_mm_addsub_pd(_mm_setzero_pd(),in); // r,-i
-      return _mm_shuffle_pd(tmp,tmp,0x1);
-    }
-
-
-  };
-
-  struct TimesI{
-    //Complex single
-    inline __m128 operator()(__m128 in, __m128 ret){
-      __m128 tmp =_mm_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1));
-      return _mm_addsub_ps(_mm_setzero_ps(),tmp); // r,-i
-    }
-    //Complex double
-    inline __m128d operator()(__m128d in, __m128d ret){
-      __m128d tmp = _mm_shuffle_pd(in,in,0x1);
-      return _mm_addsub_pd(_mm_setzero_pd(),tmp); // r,-i
-    }
-  };
-
-  struct Permute{
-
-    static inline __m128 Permute0(__m128 in){
-      return _mm_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(1,0,3,2)); //AB CD -> CD AB
-    };
-    static inline __m128 Permute1(__m128 in){
-      return _mm_shuffle_ps(in,in,_MM_SELECT_FOUR_FOUR(2,3,0,1)); //AB CD -> BA DC
-    };
-    static inline __m128 Permute2(__m128 in){
-      return in;
-    };
-    static inline __m128 Permute3(__m128 in){
-      return in;
-    };
-
-    static inline __m128d Permute0(__m128d in){ //AB -> BA
-      return _mm_shuffle_pd(in,in,0x1);
-    };
-    static inline __m128d Permute1(__m128d in){
-      return in;
-    };
-    static inline __m128d Permute2(__m128d in){
-      return in;
-    };
-    static inline __m128d Permute3(__m128d in){
-      return in;
-    };
-  };
-
+};
   
 #define _my_alignr_epi32(a,b,n) _mm_alignr_epi8(a,b,(n*4)%16)
 #define _my_alignr_epi64(a,b,n) _mm_alignr_epi8(a,b,(n*8)%16)
 
 #ifdef SFW_FP16
 
-  struct Grid_half {
-    Grid_half(){}
-    Grid_half(uint16_t raw) : x(raw) {}
-    uint16_t x;
-  };
-  union FP32 {
-    unsigned int u;
-    float f;
-  };
-
-  // PAB - Lifted and adapted from Eigen, which is GPL V2
-  inline float sfw_half_to_float(Grid_half h) {
-    const FP32 magic = { 113 << 23 };
-    const unsigned int shifted_exp = 0x7c00 << 13; // exponent mask after shift
-    FP32 o;
-    o.u = (h.x & 0x7fff) << 13;             // exponent/mantissa bits
-    unsigned int exp = shifted_exp & o.u;   // just the exponent
-    o.u += (127 - 15) << 23;                // exponent adjust
-    // handle exponent special cases
-    if (exp == shifted_exp) {     // Inf/NaN?
-      o.u += (128 - 16) << 23;    // extra exp adjust
-    } else if (exp == 0) {        // Zero/Denormal?
-      o.u += 1 << 23;             // extra exp adjust
-      o.f -= magic.f;             // renormalize
-    }
-    o.u |= (h.x & 0x8000) << 16;    // sign bit
-    return o.f;
-  }
-  inline Grid_half sfw_float_to_half(float ff) {
-    FP32 f; f.f = ff;
-    const FP32 f32infty = { 255 << 23 };
-    const FP32 f16max = { (127 + 16) << 23 };
-    const FP32 denorm_magic = { ((127 - 15) + (23 - 10) + 1) << 23 };
-    unsigned int sign_mask = 0x80000000u;
-    Grid_half o;
-    
-    o.x = static_cast<unsigned short>(0x0u);
-    unsigned int sign = f.u & sign_mask;
-    f.u ^= sign;
-    // NOTE all the integer compares in this function can be safely
-    // compiled into signed compares since all operands are below
-    // 0x80000000. Important if you want fast straight SSE2 code
-    // (since there's no unsigned PCMPGTD).
-    if (f.u >= f16max.u) {  // result is Inf or NaN (all exponent bits set)
-      o.x = (f.u > f32infty.u) ? 0x7e00 : 0x7c00; // NaN->qNaN and Inf->Inf
-    } else {  // (De)normalized number or zero
-      if (f.u < (113 << 23)) {  // resulting FP16 is subnormal or zero
-	// use a magic value to align our 10 mantissa bits at the bottom of
-	// the float. as long as FP addition is round-to-nearest-even this
-	// just works.
-	f.f += denorm_magic.f;
-	// and one integer subtract of the bias later, we have our final float!
-	o.x = static_cast<unsigned short>(f.u - denorm_magic.u);
-      } else {
-	unsigned int mant_odd = (f.u >> 13) & 1; // resulting mantissa is odd
-	
-	// update exponent, rounding bias part 1
-	f.u += ((unsigned int)(15 - 127) << 23) + 0xfff;
-	// rounding bias part 2
-	f.u += mant_odd;
-	// take the bits!
-	o.x = static_cast<unsigned short>(f.u >> 13);
-      }
-    } 
-    o.x |= static_cast<unsigned short>(sign >> 16);
-    return o;
-  }
-  static inline __m128i Grid_mm_cvtps_ph(__m128 f,int discard) {
-    __m128i ret=(__m128i)_mm_setzero_ps();
-    float *fp = (float *)&f;
-    Grid_half *hp = (Grid_half *)&ret;
-    hp[0] = sfw_float_to_half(fp[0]);
-    hp[1] = sfw_float_to_half(fp[1]);
-    hp[2] = sfw_float_to_half(fp[2]);
-    hp[3] = sfw_float_to_half(fp[3]);
-    return ret;
-  }
-  static inline __m128 Grid_mm_cvtph_ps(__m128i h,int discard) {
-    __m128 ret=_mm_setzero_ps();
-    float *fp = (float *)&ret;
-    Grid_half  *hp = (Grid_half *)&h;
-    fp[0] = sfw_half_to_float(hp[0]);
-    fp[1] = sfw_half_to_float(hp[1]);
-    fp[2] = sfw_half_to_float(hp[2]);
-    fp[3] = sfw_half_to_float(hp[3]);
-    return ret;
-  }
+static inline __m128i Grid_mm_cvtps_ph(__m128 f,int discard) {
+  __m128i ret=(__m128i)_mm_setzero_ps();
+  float *fp = (float *)&f;
+  Grid_half *hp = (Grid_half *)&ret;
+  hp[0] = sfw_float_to_half(fp[0]);
+  hp[1] = sfw_float_to_half(fp[1]);
+  hp[2] = sfw_float_to_half(fp[2]);
+  hp[3] = sfw_float_to_half(fp[3]);
+  return ret;
+}
+static inline __m128 Grid_mm_cvtph_ps(__m128i h,int discard) {
+  __m128 ret=_mm_setzero_ps();
+  float *fp = (float *)&ret;
+  Grid_half  *hp = (Grid_half *)&h;
+  fp[0] = sfw_half_to_float(hp[0]);
+  fp[1] = sfw_half_to_float(hp[1]);
+  fp[2] = sfw_half_to_float(hp[2]);
+  fp[3] = sfw_half_to_float(hp[3]);
+  return ret;
+}
 #else 
 #define Grid_mm_cvtps_ph _mm_cvtps_ph
 #define Grid_mm_cvtph_ps _mm_cvtph_ps
 #endif
-  struct PrecisionChange {
-    static inline __m128i StoH (__m128 a,__m128 b) {
-      __m128i ha = Grid_mm_cvtps_ph(a,0);
-      __m128i hb = Grid_mm_cvtps_ph(b,0);
-      __m128i h =(__m128i) _mm_shuffle_ps((__m128)ha,(__m128)hb,_MM_SELECT_FOUR_FOUR(1,0,1,0));
-      return h;
-    }
-    static inline void  HtoS (__m128i h,__m128 &sa,__m128 &sb) {
-      sa = Grid_mm_cvtph_ps(h,0); 
-      h =  (__m128i)_my_alignr_epi32((__m128i)h,(__m128i)h,2);
-      sb = Grid_mm_cvtph_ps(h,0);
-    }
-    static inline __m128 DtoS (__m128d a,__m128d b) {
-      __m128 sa = _mm_cvtpd_ps(a);
-      __m128 sb = _mm_cvtpd_ps(b);
-      __m128 s = _mm_shuffle_ps(sa,sb,_MM_SELECT_FOUR_FOUR(1,0,1,0));
-      return s;
-    }
-    static inline void StoD (__m128 s,__m128d &a,__m128d &b) {
-      a = _mm_cvtps_pd(s);
-      s = (__m128)_my_alignr_epi32((__m128i)s,(__m128i)s,2);
-      b = _mm_cvtps_pd(s);
-    }
-    static inline __m128i DtoH (__m128d a,__m128d b,__m128d c,__m128d d) {
-      __m128 sa,sb;
-      sa = DtoS(a,b);
-      sb = DtoS(c,d);
-      return StoH(sa,sb);
-    }
-    static inline void HtoD (__m128i h,__m128d &a,__m128d &b,__m128d &c,__m128d &d) {
-      __m128 sa,sb;
-      HtoS(h,sa,sb);
-      StoD(sa,a,b);
-      StoD(sb,c,d);
-    }
+struct PrecisionChange {
+  static inline __m128i StoH (__m128 a,__m128 b) {
+    __m128i ha = Grid_mm_cvtps_ph(a,0);
+    __m128i hb = Grid_mm_cvtps_ph(b,0);
+    __m128i h =(__m128i) _mm_shuffle_ps((__m128)ha,(__m128)hb,_MM_SELECT_FOUR_FOUR(1,0,1,0));
+    return h;
+  }
+  static inline void  HtoS (__m128i h,__m128 &sa,__m128 &sb) {
+    sa = Grid_mm_cvtph_ps(h,0); 
+    h =  (__m128i)_my_alignr_epi32((__m128i)h,(__m128i)h,2);
+    sb = Grid_mm_cvtph_ps(h,0);
+  }
+  static inline __m128 DtoS (__m128d a,__m128d b) {
+    __m128 sa = _mm_cvtpd_ps(a);
+    __m128 sb = _mm_cvtpd_ps(b);
+    __m128 s = _mm_shuffle_ps(sa,sb,_MM_SELECT_FOUR_FOUR(1,0,1,0));
+    return s;
+  }
+  static inline void StoD (__m128 s,__m128d &a,__m128d &b) {
+    a = _mm_cvtps_pd(s);
+    s = (__m128)_my_alignr_epi32((__m128i)s,(__m128i)s,2);
+    b = _mm_cvtps_pd(s);
+  }
+  static inline __m128i DtoH (__m128d a,__m128d b,__m128d c,__m128d d) {
+    __m128 sa,sb;
+    sa = DtoS(a,b);
+    sb = DtoS(c,d);
+    return StoH(sa,sb);
+  }
+  static inline void HtoD (__m128i h,__m128d &a,__m128d &b,__m128d &c,__m128d &d) {
+    __m128 sa,sb;
+    HtoS(h,sa,sb);
+    StoD(sa,a,b);
+    StoD(sb,c,d);
+  }
+};
+
+struct Exchange{
+  // 3210 ordering
+  static inline void Exchange0(__m128 &out1,__m128 &out2,__m128 in1,__m128 in2){
+    out1= _mm_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(1,0,1,0));
+    out2= _mm_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(3,2,3,2));
+  };
+  static inline void Exchange1(__m128 &out1,__m128 &out2,__m128 in1,__m128 in2){
+    out1= _mm_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(2,0,2,0)); /*ACEG*/
+    out2= _mm_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(3,1,3,1)); /*BDFH*/
+    out1= _mm_shuffle_ps(out1,out1,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
+    out2= _mm_shuffle_ps(out2,out2,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
+  };
+  static inline void Exchange2(__m128 &out1,__m128 &out2,__m128 in1,__m128 in2){
+    assert(0);
+    return;
+  };
+  static inline void Exchange3(__m128 &out1,__m128 &out2,__m128 in1,__m128 in2){
+    assert(0);
+    return;
   };
 
-  struct Exchange{
-    // 3210 ordering
-    static inline void Exchange0(__m128 &out1,__m128 &out2,__m128 in1,__m128 in2){
-      out1= _mm_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(1,0,1,0));
-      out2= _mm_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(3,2,3,2));
-    };
-    static inline void Exchange1(__m128 &out1,__m128 &out2,__m128 in1,__m128 in2){
-      out1= _mm_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(2,0,2,0)); /*ACEG*/
-      out2= _mm_shuffle_ps(in1,in2,_MM_SELECT_FOUR_FOUR(3,1,3,1)); /*BDFH*/
-      out1= _mm_shuffle_ps(out1,out1,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
-      out2= _mm_shuffle_ps(out2,out2,_MM_SELECT_FOUR_FOUR(3,1,2,0)); /*AECG*/
-    };
-    static inline void Exchange2(__m128 &out1,__m128 &out2,__m128 in1,__m128 in2){
-      assert(0);
-      return;
-    };
-    static inline void Exchange3(__m128 &out1,__m128 &out2,__m128 in1,__m128 in2){
-      assert(0);
-      return;
-    };
-
-    static inline void Exchange0(__m128d &out1,__m128d &out2,__m128d in1,__m128d in2){
-      out1= _mm_shuffle_pd(in1,in2,0x0);
-      out2= _mm_shuffle_pd(in1,in2,0x3);
-    };
-    static inline void Exchange1(__m128d &out1,__m128d &out2,__m128d in1,__m128d in2){
-      assert(0);
-      return;
-    };
-    static inline void Exchange2(__m128d &out1,__m128d &out2,__m128d in1,__m128d in2){
-      assert(0);
-      return;
-    };
-    static inline void Exchange3(__m128d &out1,__m128d &out2,__m128d in1,__m128d in2){
-      assert(0);
-      return;
-    };
+  static inline void Exchange0(__m128d &out1,__m128d &out2,__m128d in1,__m128d in2){
+    out1= _mm_shuffle_pd(in1,in2,0x0);
+    out2= _mm_shuffle_pd(in1,in2,0x3);
   };
-
-  struct Rotate{
-
-    static inline __m128 rotate(__m128 in,int n){ 
-      switch(n){
-      case 0: return tRotate<0>(in);break;
-      case 1: return tRotate<1>(in);break;
-      case 2: return tRotate<2>(in);break;
-      case 3: return tRotate<3>(in);break;
-      default: assert(0);
-      }
-    }
-    static inline __m128d rotate(__m128d in,int n){ 
-      switch(n){
-      case 0: return tRotate<0>(in);break;
-      case 1: return tRotate<1>(in);break;
-      default: assert(0);
-      }
-    }
-
-    template<int n> static inline __m128  tRotate(__m128  in){ return (__m128)_my_alignr_epi32((__m128i)in,(__m128i)in,n); };
-    template<int n> static inline __m128d tRotate(__m128d in){ return (__m128d)_my_alignr_epi64((__m128i)in,(__m128i)in,n); };
-
+  static inline void Exchange1(__m128d &out1,__m128d &out2,__m128d in1,__m128d in2){
+    assert(0);
+    return;
   };
-  //////////////////////////////////////////////
-  // Some Template specialization
+  static inline void Exchange2(__m128d &out1,__m128d &out2,__m128d in1,__m128d in2){
+    assert(0);
+    return;
+  };
+  static inline void Exchange3(__m128d &out1,__m128d &out2,__m128d in1,__m128d in2){
+    assert(0);
+    return;
+  };
+};
 
+struct Rotate{
 
-  //Complex float Reduce
-  template<>
-  inline Grid::ComplexF Reduce<Grid::ComplexF, __m128>::operator()(__m128 in){
-    __m128 v1; // two complex
-    v1= Optimization::Permute::Permute0(in); 
-    v1= _mm_add_ps(v1,in);
-    u128f conv;    conv.v=v1;
-    return Grid::ComplexF(conv.f[0],conv.f[1]);
+  static inline __m128 rotate(__m128 in,int n){ 
+    switch(n){
+    case 0: return tRotate<0>(in);break;
+    case 1: return tRotate<1>(in);break;
+    case 2: return tRotate<2>(in);break;
+    case 3: return tRotate<3>(in);break;
+    default: assert(0);
+    }
   }
-  //Real float Reduce
-  template<>
-  inline Grid::RealF Reduce<Grid::RealF, __m128>::operator()(__m128 in){
-    __m128 v1,v2; // quad single
-    v1= Optimization::Permute::Permute0(in); 
-    v1= _mm_add_ps(v1,in);
-    v2= Optimization::Permute::Permute1(v1); 
-    v1 = _mm_add_ps(v1,v2);
-    u128f conv; conv.v=v1;
-    return conv.f[0];
-  }
-  
-  
-  //Complex double Reduce
-  template<>
-  inline Grid::ComplexD Reduce<Grid::ComplexD, __m128d>::operator()(__m128d in){
-    u128d conv; conv.v = in;
-    return Grid::ComplexD(conv.f[0],conv.f[1]);
-  }
-  
-  //Real double Reduce
-  template<>
-  inline Grid::RealD Reduce<Grid::RealD, __m128d>::operator()(__m128d in){
-    __m128d v1;
-    v1 = Optimization::Permute::Permute0(in); 
-    v1 = _mm_add_pd(v1,in);
-    u128d conv; conv.v = v1;
-    return conv.f[0];
+  static inline __m128d rotate(__m128d in,int n){ 
+    switch(n){
+    case 0: return tRotate<0>(in);break;
+    case 1: return tRotate<1>(in);break;
+    default: assert(0);
+    }
   }
 
-  //Integer Reduce
-  template<>
-  inline Integer Reduce<Integer, __m128i>::operator()(__m128i in){
-    __m128i v1 = _mm_hadd_epi32(in, in);
-    __m128i v2 = _mm_hadd_epi32(v1, v1);
-    return _mm_cvtsi128_si32(v2);
-  }
+  template<int n> static inline __m128  tRotate(__m128  in){ return (__m128)_my_alignr_epi32((__m128i)in,(__m128i)in,n); };
+  template<int n> static inline __m128d tRotate(__m128d in){ return (__m128d)_my_alignr_epi64((__m128i)in,(__m128i)in,n); };
+
+};
+//////////////////////////////////////////////
+// Some Template specialization
+
+
+//Complex float Reduce
+template<>
+inline Grid::ComplexF Reduce<Grid::ComplexF, __m128>::operator()(__m128 in){
+  __m128 v1; // two complex
+  v1= Optimization::Permute::Permute0(in); 
+  v1= _mm_add_ps(v1,in);
+  u128f conv;    conv.v=v1;
+  return Grid::ComplexF(conv.f[0],conv.f[1]);
+}
+//Real float Reduce
+template<>
+inline Grid::RealF Reduce<Grid::RealF, __m128>::operator()(__m128 in){
+  __m128 v1,v2; // quad single
+  v1= Optimization::Permute::Permute0(in); 
+  v1= _mm_add_ps(v1,in);
+  v2= Optimization::Permute::Permute1(v1); 
+  v1 = _mm_add_ps(v1,v2);
+  u128f conv; conv.v=v1;
+  return conv.f[0];
+}
+  
+//Complex double Reduce
+template<>
+inline Grid::ComplexD Reduce<Grid::ComplexD, __m128d>::operator()(__m128d in){
+  u128d conv; conv.v = in;
+  return Grid::ComplexD(conv.f[0],conv.f[1]);
+}
+  
+//Real double Reduce
+template<>
+inline Grid::RealD Reduce<Grid::RealD, __m128d>::operator()(__m128d in){
+  __m128d v1;
+  v1 = Optimization::Permute::Permute0(in); 
+  v1 = _mm_add_pd(v1,in);
+  u128d conv; conv.v = v1;
+  return conv.f[0];
 }
 
-
+//Integer Reduce
+template<>
+inline Integer Reduce<Integer, __m128i>::operator()(__m128i in){
+  __m128i v1 = _mm_hadd_epi32(in, in);
+  __m128i v2 = _mm_hadd_epi32(v1, v1);
+  return _mm_cvtsi128_si32(v2);
+}
+NAMESPACE_END(Optimization);
 
 //////////////////////////////////////////////////////////////////////////////////////
 // Here assign types 
+typedef __m128i SIMD_Htype;  // Single precision type
+typedef __m128  SIMD_Ftype;  // Single precision type
+typedef __m128d SIMD_Dtype; // Double precision type
+typedef __m128i SIMD_Itype; // Integer type
 
-  typedef __m128i SIMD_Htype;  // Single precision type
-  typedef __m128  SIMD_Ftype;  // Single precision type
-  typedef __m128d SIMD_Dtype; // Double precision type
-  typedef __m128i SIMD_Itype; // Integer type
-
-  // prefetch utilities
-  inline void v_prefetch0(int size, const char *ptr){};
-  inline void prefetch_HINT_T0(const char *ptr){
-    _mm_prefetch(ptr,_MM_HINT_T0);
-  }
-
-  // Function name aliases
-  typedef Optimization::Vsplat   VsplatSIMD;
-  typedef Optimization::Vstore   VstoreSIMD;
-  typedef Optimization::Vset     VsetSIMD;
-  typedef Optimization::Vstream  VstreamSIMD;
-  template <typename S, typename T> using ReduceSIMD = Optimization::Reduce<S,T>;
-
- 
-
-
-  // Arithmetic operations
-  typedef Optimization::Sum         SumSIMD;
-  typedef Optimization::Sub         SubSIMD;
-  typedef Optimization::Div         DivSIMD;
-  typedef Optimization::Mult        MultSIMD;
-  typedef Optimization::MultComplex MultComplexSIMD;
-  typedef Optimization::MultRealPart MultRealPartSIMD;
-  typedef Optimization::MaddRealPart MaddRealPartSIMD;
-  typedef Optimization::Conj        ConjSIMD;
-  typedef Optimization::TimesMinusI TimesMinusISIMD;
-  typedef Optimization::TimesI      TimesISIMD;
-
+// prefetch utilities
+inline void v_prefetch0(int size, const char *ptr){};
+inline void prefetch_HINT_T0(const char *ptr){
+  _mm_prefetch(ptr,_MM_HINT_T0);
 }
+
+// Function name aliases
+typedef Optimization::Vsplat   VsplatSIMD;
+typedef Optimization::Vstore   VstoreSIMD;
+typedef Optimization::Vset     VsetSIMD;
+typedef Optimization::Vstream  VstreamSIMD;
+template <typename S, typename T> using ReduceSIMD = Optimization::Reduce<S,T>;
+
+// Arithmetic operations
+typedef Optimization::Sum         SumSIMD;
+typedef Optimization::Sub         SubSIMD;
+typedef Optimization::Div         DivSIMD;
+typedef Optimization::Mult        MultSIMD;
+typedef Optimization::MultComplex MultComplexSIMD;
+typedef Optimization::MultRealPart MultRealPartSIMD;
+typedef Optimization::MaddRealPart MaddRealPartSIMD;
+typedef Optimization::Conj        ConjSIMD;
+typedef Optimization::TimesMinusI TimesMinusISIMD;
+typedef Optimization::TimesI      TimesISIMD;
+
+NAMESPACE_END(Grid);
diff --git a/Grid/simd/Grid_vector_types.h b/Grid/simd/Grid_vector_types.h
index 707af211..e2b1fd07 100644
--- a/Grid/simd/Grid_vector_types.h
+++ b/Grid/simd/Grid_vector_types.h
@@ -1,33 +1,33 @@
 /*************************************************************************************
 
-Grid physics library, www.github.com/paboyle/Grid
+    Grid physics library, www.github.com/paboyle/Grid 
 
-Source file: ./lib/simd/Grid_vector_type.h
+    Source file: ./lib/simd/Grid_vector_types.h
 
-Copyright (C) 2015
+    Copyright (C) 2015
 
 Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
 Author: Guido Cossu <cossu@iroiro-pc.kek.jp>
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: neo <cossu@post.kek.jp>
+Author: paboyle <paboyle@ph.ed.ac.uk>
 Author: Michael Marshall <michael.marshall@ed.ac.au>
 
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
 
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
 
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-See the full license in the file "LICENSE" in the top level distribution
-directory
+    See the full license in the file "LICENSE" in the top level distribution directory
 *************************************************************************************/
 /*  END LEGAL */
 //---------------------------------------------------------------------------
@@ -39,6 +39,77 @@ directory
 #ifndef GRID_VECTOR_TYPES
 #define GRID_VECTOR_TYPES
 
+// PAB - Lifted and adapted from Eigen, which is GPL V2
+struct Grid_half {
+  accelerator Grid_half(){}
+  accelerator Grid_half(uint16_t raw) : x(raw) {}
+  uint16_t x;
+};
+union FP32 {
+  unsigned int u;
+  float f;
+};
+accelerator_inline float sfw_half_to_float(Grid_half h) {
+  const FP32 magic = { 113 << 23 };
+  const unsigned int shifted_exp = 0x7c00 << 13; // exponent mask after shift
+  FP32 o;
+  o.u = (h.x & 0x7fff) << 13;             // exponent/mantissa bits
+  unsigned int exp = shifted_exp & o.u;   // just the exponent
+  o.u += (127 - 15) << 23;                // exponent adjust
+  // handle exponent special cases
+  if (exp == shifted_exp) {     // Inf/NaN?
+    o.u += (128 - 16) << 23;    // extra exp adjust
+  } else if (exp == 0) {        // Zero/Denormal?
+    o.u += 1 << 23;             // extra exp adjust
+    o.f -= magic.f;             // renormalize
+  }
+  o.u |= (h.x & 0x8000) << 16;    // sign bit
+  return o.f;
+}
+accelerator_inline Grid_half sfw_float_to_half(float ff) {
+  FP32 f; f.f = ff;
+  const FP32 f32infty = { 255 << 23 };
+  const FP32 f16max = { (127 + 16) << 23 };
+  const FP32 denorm_magic = { ((127 - 15) + (23 - 10) + 1) << 23 };
+  unsigned int sign_mask = 0x80000000u;
+  Grid_half o;
+    
+  o.x = static_cast<unsigned short>(0x0u);
+  unsigned int sign = f.u & sign_mask;
+  f.u ^= sign;
+  // NOTE all the integer compares in this function can be safely
+  // compiled into signed compares since all operands are below
+  // 0x80000000. Important if you want fast straight SSE2 code
+  // (since there's no unsigned PCMPGTD).
+  if (f.u >= f16max.u) {  // result is Inf or NaN (all exponent bits set)
+    o.x = (f.u > f32infty.u) ? 0x7e00 : 0x7c00; // NaN->qNaN and Inf->Inf
+  } else {  // (De)normalized number or zero
+    if (f.u < (113 << 23)) {  // resulting FP16 is subnormal or zero
+      // use a magic value to align our 10 mantissa bits at the bottom of
+      // the float. as long as FP addition is round-to-nearest-even this
+      // just works.
+      f.f += denorm_magic.f;
+      // and one integer subtract of the bias later, we have our final float!
+      o.x = static_cast<unsigned short>(f.u - denorm_magic.u);
+    } else {
+      unsigned int mant_odd = (f.u >> 13) & 1; // resulting mantissa is odd
+	
+      // update exponent, rounding bias part 1
+      f.u += ((unsigned int)(15 - 127) << 23) + 0xfff;
+      // rounding bias part 2
+      f.u += mant_odd;
+      // take the bits!
+      o.x = static_cast<unsigned short>(f.u >> 13);
+    }
+  } 
+  o.x |= static_cast<unsigned short>(sign >> 16);
+  return o;
+}
+
+
+#ifdef GPU_VEC
+#include "Grid_gpu_vec.h"
+#endif
 #ifdef GEN
 #include "Grid_generic.h"
 #endif
@@ -61,9 +132,8 @@ directory
 #include "Grid_qpx.h"
 #endif
 
-#include "l1p.h"
+NAMESPACE_BEGIN(Grid);
 
-namespace Grid {
 
 //////////////////////////////////////
 // To take the floating point type of real/complex type
@@ -73,7 +143,7 @@ struct RealPart {
   typedef T type;
 };
 template <typename T>
-struct RealPart<std::complex<T> > {
+struct RealPart<complex<T> > {
   typedef T type;
 };
 
@@ -117,27 +187,25 @@ template <typename T1,typename T2> using IfNotSame    = Invoke<std::enable_if<!s
 // need explicit declaration of types when used since
 // clang cannot automatically determine the output type sometimes
 template <class Out, class Input1, class Input2, class Input3, class Operation>
-Out trinary(Input1 src_1, Input2 src_2, Input3 src_3, Operation op) {
+Out accelerator_inline  trinary(Input1 src_1, Input2 src_2, Input3 src_3, Operation op) {
   return op(src_1, src_2, src_3);
 }
-
 template <class Out, class Input1, class Input2, class Operation>
-Out binary(Input1 src_1, Input2 src_2, Operation op) {
+Out accelerator_inline binary(Input1 src_1, Input2 src_2, Operation op) {
   return op(src_1, src_2);
 }
-
 template <class Out, class Input, class Operation>
-Out unary(Input src, Operation op) {
+Out accelerator_inline  unary(Input src, Operation op) {
   return op(src);
 }
 ///////////////////////////////////////////////
 
 /*
   @brief Grid_simd class for the SIMD vector type operations
- */
+*/
 template <class Scalar_type, class Vector_type>
 class Grid_simd {
- public:
+public:
   typedef typename RealPart<Scalar_type>::type Real;
   typedef Vector_type vector_type;
   typedef Scalar_type scalar_type;
@@ -145,161 +213,164 @@ class Grid_simd {
   typedef union conv_t_union {
     Vector_type v;
     Scalar_type s[sizeof(Vector_type) / sizeof(Scalar_type)];
-    conv_t_union(){};
+    accelerator_inline conv_t_union(){};
   } conv_t;
 
   Vector_type v;
 
-  static inline constexpr int Nsimd(void) {
+  static accelerator_inline constexpr int Nsimd(void) {
+    static_assert( (sizeof(Vector_type) / sizeof(Scalar_type) >= 1), " size mismatch " );
     return sizeof(Vector_type) / sizeof(Scalar_type);
   }
 
-  Grid_simd &operator=(const Grid_simd &&rhs) {
+  accelerator_inline Grid_simd &operator=(const Grid_simd &&rhs) {
     v = rhs.v;
     return *this;
   };
-  Grid_simd &operator=(const Grid_simd &rhs) {
+  accelerator_inline Grid_simd &operator=(const Grid_simd &rhs) {
     v = rhs.v;
     return *this;
   };  // faster than not declaring it and leaving to the compiler
-  Grid_simd() = default;
-  Grid_simd(const Grid_simd &rhs) : v(rhs.v){};  // compiles in movaps
-  Grid_simd(const Grid_simd &&rhs) : v(rhs.v){};
+
+
+  accelerator Grid_simd() = default;
+  accelerator_inline Grid_simd(const Grid_simd &rhs) : v(rhs.v){};  // compiles in movaps
+  accelerator_inline Grid_simd(const Grid_simd &&rhs) : v(rhs.v){};
+  accelerator_inline Grid_simd(const Real a) { vsplat(*this, Scalar_type(a)); };
+  // Enable if complex type
+  template <typename S = Scalar_type> accelerator_inline
+  Grid_simd(const typename std::enable_if<is_complex<S>::value, S>::type a) {
+      vsplat(*this, a);
+  };
 
   /////////////////////////////
   // Constructors
   /////////////////////////////
-  Grid_simd &operator=(Zero &z) {
+  accelerator_inline Grid_simd &  operator=(const Zero &z) {
     vzero(*this);
     return (*this);
   }
 
-  // Enable if complex type
-  template <typename S = Scalar_type>
-  Grid_simd(const typename std::enable_if<is_complex<S>::value, S>::type a) {
-    vsplat(*this, a);
-  };
 
-  Grid_simd(const Real a) { vsplat(*this, Scalar_type(a)); };
 
   ///////////////////////////////////////////////
   // mac, mult, sub, add, adj
   ///////////////////////////////////////////////
 
-  // FIXME -- alias this to an inline MAC struct.
-  friend inline void mac(Grid_simd *__restrict__ y,
-                         const Grid_simd *__restrict__ a,
-                         const Grid_simd *__restrict__ x) {
+  // FIXME -- alias this to an accelerator_inline MAC struct.
+  friend accelerator_inline void mac(Grid_simd *__restrict__ y,
+				     const Grid_simd *__restrict__ a,
+				     const Grid_simd *__restrict__ x) {
     *y = (*a) * (*x) + (*y);
   };
-
-  friend inline void mult(Grid_simd *__restrict__ y,
-                          const Grid_simd *__restrict__ l,
-                          const Grid_simd *__restrict__ r) {
+  
+  friend accelerator_inline void mult(Grid_simd *__restrict__ y,
+				      const Grid_simd *__restrict__ l,
+				      const Grid_simd *__restrict__ r) {
     *y = (*l) * (*r);
   }
 
-  friend inline void sub(Grid_simd *__restrict__ y,
-                         const Grid_simd *__restrict__ l,
-                         const Grid_simd *__restrict__ r) {
+  friend accelerator_inline void sub(Grid_simd *__restrict__ y,
+				     const Grid_simd *__restrict__ l,
+				     const Grid_simd *__restrict__ r) {
     *y = (*l) - (*r);
   }
-  friend inline void add(Grid_simd *__restrict__ y,
-                         const Grid_simd *__restrict__ l,
-                         const Grid_simd *__restrict__ r) {
+  friend accelerator_inline void add(Grid_simd *__restrict__ y,
+				     const Grid_simd *__restrict__ l,
+				     const Grid_simd *__restrict__ r) {
     *y = (*l) + (*r);
   }
-  friend inline void mac(Grid_simd *__restrict__ y,
-                         const Scalar_type *__restrict__ a,
-                         const Grid_simd *__restrict__ x) {
+  friend accelerator_inline void mac(Grid_simd *__restrict__ y,
+				     const Scalar_type *__restrict__ a,
+				     const Grid_simd *__restrict__ x) {
     *y = (*a) * (*x) + (*y);
   };
-  friend inline void mult(Grid_simd *__restrict__ y,
-                          const Scalar_type *__restrict__ l,
-                          const Grid_simd *__restrict__ r) {
+  friend accelerator_inline void mult(Grid_simd *__restrict__ y,
+				      const Scalar_type *__restrict__ l,
+				      const Grid_simd *__restrict__ r) {
     *y = (*l) * (*r);
   }
-  friend inline void sub(Grid_simd *__restrict__ y,
-                         const Scalar_type *__restrict__ l,
-                         const Grid_simd *__restrict__ r) {
+  friend accelerator_inline void sub(Grid_simd *__restrict__ y,
+				     const Scalar_type *__restrict__ l,
+				     const Grid_simd *__restrict__ r) {
     *y = (*l) - (*r);
   }
-  friend inline void add(Grid_simd *__restrict__ y,
-                         const Scalar_type *__restrict__ l,
-                         const Grid_simd *__restrict__ r) {
+  friend accelerator_inline void add(Grid_simd *__restrict__ y,
+				     const Scalar_type *__restrict__ l,
+				     const Grid_simd *__restrict__ r) {
     *y = (*l) + (*r);
   }
 
-  friend inline void mac(Grid_simd *__restrict__ y,
-                         const Grid_simd *__restrict__ a,
-                         const Scalar_type *__restrict__ x) {
+  friend accelerator_inline void mac(Grid_simd *__restrict__ y,
+				     const Grid_simd *__restrict__ a,
+				     const Scalar_type *__restrict__ x) {
     *y = (*a) * (*x) + (*y);
   };
-  friend inline void mult(Grid_simd *__restrict__ y,
-                          const Grid_simd *__restrict__ l,
-                          const Scalar_type *__restrict__ r) {
+  friend accelerator_inline void mult(Grid_simd *__restrict__ y,
+				      const Grid_simd *__restrict__ l,
+				      const Scalar_type *__restrict__ r) {
     *y = (*l) * (*r);
   }
-  friend inline void sub(Grid_simd *__restrict__ y,
-                         const Grid_simd *__restrict__ l,
-                         const Scalar_type *__restrict__ r) {
+  friend accelerator_inline void sub(Grid_simd *__restrict__ y,
+				     const Grid_simd *__restrict__ l,
+				     const Scalar_type *__restrict__ r) {
     *y = (*l) - (*r);
   }
-  friend inline void add(Grid_simd *__restrict__ y,
-                         const Grid_simd *__restrict__ l,
-                         const Scalar_type *__restrict__ r) {
+  friend accelerator_inline void add(Grid_simd *__restrict__ y,
+				     const Grid_simd *__restrict__ l,
+				     const Scalar_type *__restrict__ r) {
     *y = (*l) + (*r);
   }
 
   ////////////////////////////////////////////////////////////////////////
   // FIXME:  gonna remove these load/store, get, set, prefetch
   ////////////////////////////////////////////////////////////////////////
-  friend inline void vset(Grid_simd &ret, Scalar_type *a) {
+  friend accelerator_inline void vset(Grid_simd &ret, Scalar_type *a) {
     ret.v = unary<Vector_type>(a, VsetSIMD());
   }
 
   ///////////////////////
   // Vstore
   ///////////////////////
-  friend inline void vstore(const Grid_simd &ret, Scalar_type *a) {
-    binary<void>(ret.v, (Real *)a, VstoreSIMD());
+  friend accelerator_inline void vstore(const Grid_simd &ret, Scalar_type *a) {
+    binary<void>(ret.v, (Real *) a, VstoreSIMD());
   }
 
   ///////////////////////
   // Vprefetch
   ///////////////////////
-  friend inline void vprefetch(const Grid_simd &v) {
+  friend accelerator_inline void vprefetch(const Grid_simd &v) {
     prefetch_HINT_T0((const char *)&v.v);
   }
 
   ///////////////////////
   // Reduce
   ///////////////////////
-  friend inline Scalar_type Reduce(const Grid_simd &in) {
+  friend accelerator_inline Scalar_type Reduce(const Grid_simd &in) {
     return unary<Scalar_type>(in.v, ReduceSIMD<Scalar_type, Vector_type>());
   }
 
   ////////////////////////////
   // operator scalar * simd
   ////////////////////////////
-  friend inline Grid_simd operator*(const Scalar_type &a, Grid_simd b) {
+  friend accelerator_inline Grid_simd operator*(const Scalar_type &a, Grid_simd b) {
     Grid_simd va;
     vsplat(va, a);
     return va * b;
   }
-  friend inline Grid_simd operator*(Grid_simd b, const Scalar_type &a) {
+  friend accelerator_inline Grid_simd operator*(Grid_simd b, const Scalar_type &a) {
     return a * b;
   }
 
   //////////////////////////////////
   // Divides
   //////////////////////////////////
-  friend inline Grid_simd operator/(const Scalar_type &a, Grid_simd b) {
+  friend accelerator_inline Grid_simd operator/(const Scalar_type &a, Grid_simd b) {
     Grid_simd va;
     vsplat(va, a);
     return va / b;
   }
-  friend inline Grid_simd operator/(Grid_simd b, const Scalar_type &a) {
+  friend accelerator_inline Grid_simd operator/(Grid_simd b, const Scalar_type &a) {
     Grid_simd va;
     vsplat(va, a);
     return b / a;
@@ -308,23 +379,23 @@ class Grid_simd {
   ///////////////////////
   // Unary negation
   ///////////////////////
-  friend inline Grid_simd operator-(const Grid_simd &r) {
+  friend accelerator_inline Grid_simd operator-(const Grid_simd &r) {
     Grid_simd ret;
     vzero(ret);
     ret = ret - r;
     return ret;
   }
   // *=,+=,-= operators
-  inline Grid_simd &operator*=(const Grid_simd &r) {
+  accelerator_inline Grid_simd &operator*=(const Grid_simd &r) {
     *this = (*this) * r;
     return *this;
     // return (*this)*r; ?
   }
-  inline Grid_simd &operator+=(const Grid_simd &r) {
+  accelerator_inline Grid_simd &operator+=(const Grid_simd &r) {
     *this = *this + r;
     return *this;
   }
-  inline Grid_simd &operator-=(const Grid_simd &r) {
+  accelerator_inline Grid_simd &operator-=(const Grid_simd &r) {
     *this = *this - r;
     return *this;
   }
@@ -337,7 +408,7 @@ class Grid_simd {
   ///////////////////////////////////////
 
   template <class functor>
-  friend inline Grid_simd SimdApply(const functor &func, const Grid_simd &v) {
+  friend accelerator_inline Grid_simd SimdApply(const functor &func, const Grid_simd &v) {
     Grid_simd ret;
     Grid_simd::conv_t conv;
     Grid_simd::scalar_type s;
@@ -351,7 +422,7 @@ class Grid_simd {
     return ret;
   }
   template <class functor>
-  friend inline Grid_simd SimdApplyBinop(const functor &func,
+  friend accelerator_inline Grid_simd SimdApplyBinop(const functor &func,
                                          const Grid_simd &x,
                                          const Grid_simd &y) {
     Grid_simd ret;
@@ -373,7 +444,7 @@ class Grid_simd {
   // Exchange 
   // Al Ah , Bl Bh -> Al Bl Ah,Bh
   ///////////////////////
-  friend inline void exchange(Grid_simd &out1,Grid_simd &out2,Grid_simd in1,Grid_simd in2,int n)
+  friend accelerator_inline void exchange(Grid_simd &out1,Grid_simd &out2,Grid_simd in1,Grid_simd in2,int n)
   {
     if       (n==3) {
       Optimization::Exchange::Exchange3(out1.v,out2.v,in1.v,in2.v);
@@ -385,16 +456,16 @@ class Grid_simd {
       Optimization::Exchange::Exchange0(out1.v,out2.v,in1.v,in2.v);
     }
   }
-  friend inline void exchange0(Grid_simd &out1,Grid_simd &out2,Grid_simd in1,Grid_simd in2){    
+  friend accelerator_inline void exchange0(Grid_simd &out1,Grid_simd &out2,Grid_simd in1,Grid_simd in2){    
     Optimization::Exchange::Exchange0(out1.v,out2.v,in1.v,in2.v);
   }
-  friend inline void exchange1(Grid_simd &out1,Grid_simd &out2,Grid_simd in1,Grid_simd in2){    
+  friend accelerator_inline void exchange1(Grid_simd &out1,Grid_simd &out2,Grid_simd in1,Grid_simd in2){    
     Optimization::Exchange::Exchange1(out1.v,out2.v,in1.v,in2.v);
   }
-  friend inline void exchange2(Grid_simd &out1,Grid_simd &out2,Grid_simd in1,Grid_simd in2){    
+  friend accelerator_inline void exchange2(Grid_simd &out1,Grid_simd &out2,Grid_simd in1,Grid_simd in2){    
     Optimization::Exchange::Exchange2(out1.v,out2.v,in1.v,in2.v);
   }
-  friend inline void exchange3(Grid_simd &out1,Grid_simd &out2,Grid_simd in1,Grid_simd in2){    
+  friend accelerator_inline void exchange3(Grid_simd &out1,Grid_simd &out2,Grid_simd in1,Grid_simd in2){    
     Optimization::Exchange::Exchange3(out1.v,out2.v,in1.v,in2.v);
   }
   ////////////////////////////////////////////////////////////////////
@@ -402,19 +473,19 @@ class Grid_simd {
   // all subtypes; may not be a good assumption, but could
   // add the vector width as a template param for BG/Q for example
   ////////////////////////////////////////////////////////////////////
-  friend inline void permute0(Grid_simd &y, Grid_simd b) {
+  friend accelerator_inline void permute0(Grid_simd &y, Grid_simd b) {
     y.v = Optimization::Permute::Permute0(b.v);
   }
-  friend inline void permute1(Grid_simd &y, Grid_simd b) {
+  friend accelerator_inline void permute1(Grid_simd &y, Grid_simd b) {
     y.v = Optimization::Permute::Permute1(b.v);
   }
-  friend inline void permute2(Grid_simd &y, Grid_simd b) {
+  friend accelerator_inline void permute2(Grid_simd &y, Grid_simd b) {
     y.v = Optimization::Permute::Permute2(b.v);
   }
-  friend inline void permute3(Grid_simd &y, Grid_simd b) {
+  friend accelerator_inline void permute3(Grid_simd &y, Grid_simd b) {
     y.v = Optimization::Permute::Permute3(b.v);
   }
-  friend inline void permute(Grid_simd &y, Grid_simd b, int perm) {
+  friend accelerator_inline void permute(Grid_simd &y, Grid_simd b, int perm) {
     if (perm & RotateBit) {
       int dist = perm & 0xF;
       y = rotate(b, dist);
@@ -429,60 +500,90 @@ class Grid_simd {
   ///////////////////////////////
   // Getting single lanes
   ///////////////////////////////
-  inline Scalar_type getlane(int lane) {
+  accelerator_inline Scalar_type getlane(int lane) {
     return ((Scalar_type*)&v)[lane];
   }
 
-  inline void putlane(const Scalar_type &S, int lane){
+  accelerator_inline void putlane(const Scalar_type &S, int lane){
     ((Scalar_type*)&v)[lane] = S;
   }
-
-
-  
 };  // end of Grid_simd class definition
 
-inline void permute(ComplexD &y,ComplexD b, int perm) {  y=b; }
-inline void permute(ComplexF &y,ComplexF b, int perm) {  y=b; }
-inline void permute(RealD &y,RealD b, int perm) {  y=b; }
-inline void permute(RealF &y,RealF b, int perm) {  y=b; }
+///////////////////////////////
+// Define available types
+///////////////////////////////
+
+typedef Grid_simd<float  , SIMD_Ftype> vRealF;
+typedef Grid_simd<double , SIMD_Dtype> vRealD;
+typedef Grid_simd<Integer, SIMD_Itype> vInteger;
+typedef Grid_simd<uint16_t,SIMD_Htype> vRealH;
+
+#ifdef GPU_VEC
+typedef Grid_simd<complex<uint16_t>, SIMD_CHtype> vComplexH;
+typedef Grid_simd<complex<float>   , SIMD_CFtype> vComplexF;
+typedef Grid_simd<complex<double>  , SIMD_CDtype> vComplexD;
+#else
+typedef Grid_simd<complex<uint16_t>, SIMD_Htype> vComplexH;
+typedef Grid_simd<complex<float>   , SIMD_Ftype> vComplexF;
+typedef Grid_simd<complex<double>  , SIMD_Dtype> vComplexD;
+#endif
+
+/////////////////////////////////////////
+// Pointer type to use on extractLane
+/////////////////////////////////////////
+template<class _scalar> class ExtractTypeMap                      { public: typedef _scalar extract_type;};
+#ifdef GPU_VEC
+template<>              class ExtractTypeMap< complex<uint16_t> > { public: typedef   half2 extract_type;};
+template<>              class ExtractTypeMap< complex<   float> > { public: typedef  float2 extract_type;};
+template<>              class ExtractTypeMap< complex<  double> > { public: typedef double2 extract_type;};
+#endif
+
+/////////////////////////////////////////
+// Permute
+/////////////////////////////////////////
+
+accelerator_inline void permute(ComplexD &y,ComplexD b, int perm) {  y=b; }
+accelerator_inline void permute(ComplexF &y,ComplexF b, int perm) {  y=b; }
+accelerator_inline void permute(RealD &y,RealD b, int perm) {  y=b; }
+accelerator_inline void permute(RealF &y,RealF b, int perm) {  y=b; }
 
 ////////////////////////////////////////////////////////////////////
 // General rotate
 ////////////////////////////////////////////////////////////////////
 template <class S, class V, IfNotComplex<S> = 0>
-inline Grid_simd<S, V> rotate(Grid_simd<S, V> b, int nrot) {
+accelerator_inline Grid_simd<S, V> rotate(Grid_simd<S, V> b, int nrot) {
   nrot = nrot % Grid_simd<S, V>::Nsimd();
   Grid_simd<S, V> ret;
   ret.v = Optimization::Rotate::rotate(b.v, nrot);
   return ret;
 }
 template <class S, class V, IfComplex<S> = 0>
-inline Grid_simd<S, V> rotate(Grid_simd<S, V> b, int nrot) {
+accelerator_inline Grid_simd<S, V> rotate(Grid_simd<S, V> b, int nrot) {
   nrot = nrot % Grid_simd<S, V>::Nsimd();
   Grid_simd<S, V> ret;
   ret.v = Optimization::Rotate::rotate(b.v, 2 * nrot);
   return ret;
 }
 template <class S, class V, IfNotComplex<S> =0> 
-inline void rotate( Grid_simd<S,V> &ret,Grid_simd<S,V> b,int nrot)
+accelerator_inline void rotate( Grid_simd<S,V> &ret,Grid_simd<S,V> b,int nrot)
 {
   nrot = nrot % Grid_simd<S,V>::Nsimd();
   ret.v = Optimization::Rotate::rotate(b.v,nrot);
 }
 template <class S, class V, IfComplex<S> =0> 
-inline void rotate(Grid_simd<S,V> &ret,Grid_simd<S,V> b,int nrot)
+accelerator_inline void rotate(Grid_simd<S,V> &ret,Grid_simd<S,V> b,int nrot)
 {
   nrot = nrot % Grid_simd<S,V>::Nsimd();
   ret.v = Optimization::Rotate::rotate(b.v,2*nrot);
 }
 
 template <class S, class V> 
-inline void vbroadcast(Grid_simd<S,V> &ret,const Grid_simd<S,V> &src,int lane){
+accelerator_inline void vbroadcast(Grid_simd<S,V> &ret,const Grid_simd<S,V> &src,int lane){
   S* typepun =(S*) &src;
   vsplat(ret,typepun[lane]);
 }    
 template <class S, class V, IfComplex<S> =0> 
-inline void rbroadcast(Grid_simd<S,V> &ret,const Grid_simd<S,V> &src,int lane){
+accelerator_inline void rbroadcast(Grid_simd<S,V> &ret,const Grid_simd<S,V> &src,int lane){
   S* typepun =(S*) &src;
   ret.v = unary<V>(real(typepun[lane]), VsplatSIMD());
 }    
@@ -495,24 +596,24 @@ inline void rbroadcast(Grid_simd<S,V> &ret,const Grid_simd<S,V> &src,int lane){
 
 // this is only for the complex version
 template <class S, class V, IfComplex<S> = 0, class ABtype>
-inline void vsplat(Grid_simd<S, V> &ret, ABtype a, ABtype b) {
+accelerator_inline void vsplat(Grid_simd<S, V> &ret, ABtype a, ABtype b) {
   ret.v = binary<V>(a, b, VsplatSIMD());
 }
 
 // overload if complex
 template <class S, class V>
-inline void vsplat(Grid_simd<S, V> &ret, EnableIf<is_complex<S>, S> c) {
+accelerator_inline void vsplat(Grid_simd<S, V> &ret, EnableIf<is_complex<S>, S> c) {
   vsplat(ret, real(c), imag(c));
 }
 template <class S, class V>
-inline void rsplat(Grid_simd<S, V> &ret, EnableIf<is_complex<S>, S> c) {
+accelerator_inline void rsplat(Grid_simd<S, V> &ret, EnableIf<is_complex<S>, S> c) {
   vsplat(ret, real(c), real(c));
 }
 
 // if real fill with a, if complex fill with a in the real part (first function
 // above)
 template <class S, class V>
-inline void vsplat(Grid_simd<S, V> &ret, NotEnableIf<is_complex<S>, S> a) {
+accelerator_inline void vsplat(Grid_simd<S, V> &ret, NotEnableIf<is_complex<S>, S> a) {
   ret.v = unary<V>(a, VsplatSIMD());
 }
 //////////////////////////
@@ -522,56 +623,56 @@ inline void vsplat(Grid_simd<S, V> &ret, NotEnableIf<is_complex<S>, S> a) {
 ///////////////////////////////////////////////
 // For complex types
 template <class S, class V, IfComplex<S> = 0>
-inline void vone(Grid_simd<S, V> &ret) {
+accelerator_inline void vone(Grid_simd<S, V> &ret) {
   vsplat(ret, S(1.0, 0.0));
 }
 template <class S, class V, IfComplex<S> = 0>
-inline void vzero(Grid_simd<S, V> &ret) {
+accelerator_inline void vzero(Grid_simd<S, V> &ret) {
   vsplat(ret, S(0.0, 0.0));
 }  // use xor?
 template <class S, class V, IfComplex<S> = 0>
-inline void vcomplex_i(Grid_simd<S, V> &ret) {
+accelerator_inline void vcomplex_i(Grid_simd<S, V> &ret) {
   vsplat(ret, S(0.0, 1.0));
 }
 
 template <class S, class V, IfComplex<S> = 0>
-inline void visign(Grid_simd<S, V> &ret) {
+accelerator_inline void visign(Grid_simd<S, V> &ret) {
   vsplat(ret, S(1.0, -1.0));
 }
 template <class S, class V, IfComplex<S> = 0>
-inline void vrsign(Grid_simd<S, V> &ret) {
+accelerator_inline void vrsign(Grid_simd<S, V> &ret) {
   vsplat(ret, S(-1.0, 1.0));
 }
 
 // if not complex overload here
 template <class S, class V, IfReal<S> = 0>
-inline void vone(Grid_simd<S, V> &ret) {
+accelerator_inline void vone(Grid_simd<S, V> &ret) {
   vsplat(ret, S(1.0));
 }
 template <class S, class V, IfReal<S> = 0>
-inline void vzero(Grid_simd<S, V> &ret) {
+accelerator_inline void vzero(Grid_simd<S, V> &ret) {
   vsplat(ret, S(0.0));
 }
 
 // For integral types
 template <class S, class V, IfInteger<S> = 0>
-inline void vone(Grid_simd<S, V> &ret) {
+accelerator_inline void vone(Grid_simd<S, V> &ret) {
   vsplat(ret, 1);
 }
 template <class S, class V, IfInteger<S> = 0>
-inline void vzero(Grid_simd<S, V> &ret) {
+accelerator_inline void vzero(Grid_simd<S, V> &ret) {
   vsplat(ret, 0);
 }
 template <class S, class V, IfInteger<S> = 0>
-inline void vtrue(Grid_simd<S, V> &ret) {
+accelerator_inline void vtrue(Grid_simd<S, V> &ret) {
   vsplat(ret, 0xFFFFFFFF);
 }
 template <class S, class V, IfInteger<S> = 0>
-inline void vfalse(Grid_simd<S, V> &ret) {
+accelerator_inline void vfalse(Grid_simd<S, V> &ret) {
   vsplat(ret, 0);
 }
 template <class S, class V>
-inline void zeroit(Grid_simd<S, V> &z) {
+accelerator_inline void zeroit(Grid_simd<S, V> &z) {
   vzero(z);
 }
 
@@ -579,16 +680,16 @@ inline void zeroit(Grid_simd<S, V> &z) {
 // Vstream
 ///////////////////////
 template <class S, class V, IfReal<S> = 0>
-inline void vstream(Grid_simd<S, V> &out, const Grid_simd<S, V> &in) {
+accelerator_inline void vstream(Grid_simd<S, V> &out, const Grid_simd<S, V> &in) {
   binary<void>((S *)&out.v, in.v, VstreamSIMD());
 }
 template <class S, class V, IfComplex<S> = 0>
-inline void vstream(Grid_simd<S, V> &out, const Grid_simd<S, V> &in) {
+accelerator_inline void vstream(Grid_simd<S, V> &out, const Grid_simd<S, V> &in) {
   typedef typename S::value_type T;
   binary<void>((T *)&out.v, in.v, VstreamSIMD());
 }
 template <class S, class V, IfInteger<S> = 0>
-inline void vstream(Grid_simd<S, V> &out, const Grid_simd<S, V> &in) {
+accelerator_inline void vstream(Grid_simd<S, V> &out, const Grid_simd<S, V> &in) {
   out = in;
 }
 
@@ -596,14 +697,14 @@ inline void vstream(Grid_simd<S, V> &out, const Grid_simd<S, V> &in) {
 // Arithmetic operator overloads +,-,*
 ////////////////////////////////////
 template <class S, class V>
-inline Grid_simd<S, V> operator+(Grid_simd<S, V> a, Grid_simd<S, V> b) {
+accelerator_inline Grid_simd<S, V> operator+(Grid_simd<S, V> a, Grid_simd<S, V> b) {
   Grid_simd<S, V> ret;
   ret.v = binary<V>(a.v, b.v, SumSIMD());
   return ret;
 };
 
 template <class S, class V>
-inline Grid_simd<S, V> operator-(Grid_simd<S, V> a, Grid_simd<S, V> b) {
+accelerator_inline Grid_simd<S, V> operator-(Grid_simd<S, V> a, Grid_simd<S, V> b) {
   Grid_simd<S, V> ret;
   ret.v = binary<V>(a.v, b.v, SubSIMD());
   return ret;
@@ -611,13 +712,13 @@ inline Grid_simd<S, V> operator-(Grid_simd<S, V> a, Grid_simd<S, V> b) {
 
 // Distinguish between complex types and others
 template <class S, class V, IfComplex<S> = 0>
-inline Grid_simd<S, V> real_mult(Grid_simd<S, V> a, Grid_simd<S, V> b) {
+accelerator_inline Grid_simd<S, V> real_mult(Grid_simd<S, V> a, Grid_simd<S, V> b) {
   Grid_simd<S, V> ret;
   ret.v = binary<V>(a.v, b.v, MultRealPartSIMD());
   return ret;
 };
 template <class S, class V, IfComplex<S> = 0>
-inline Grid_simd<S, V> real_madd(Grid_simd<S, V> a, Grid_simd<S, V> b, Grid_simd<S,V> c) {
+accelerator_inline Grid_simd<S, V> real_madd(Grid_simd<S, V> a, Grid_simd<S, V> b, Grid_simd<S,V> c) {
   Grid_simd<S, V> ret;
   ret.v = trinary<V>(a.v, b.v, c.v, MaddRealPartSIMD());
   return ret;
@@ -626,7 +727,7 @@ inline Grid_simd<S, V> real_madd(Grid_simd<S, V> a, Grid_simd<S, V> b, Grid_simd
 
 // Distinguish between complex types and others
 template <class S, class V, IfComplex<S> = 0>
-inline Grid_simd<S, V> operator*(Grid_simd<S, V> a, Grid_simd<S, V> b) {
+accelerator_inline Grid_simd<S, V> operator*(Grid_simd<S, V> a, Grid_simd<S, V> b) {
   Grid_simd<S, V> ret;
   ret.v = binary<V>(a.v, b.v, MultComplexSIMD());
   return ret;
@@ -634,7 +735,7 @@ inline Grid_simd<S, V> operator*(Grid_simd<S, V> a, Grid_simd<S, V> b) {
 
 // Real/Integer types
 template <class S, class V, IfNotComplex<S> = 0>
-inline Grid_simd<S, V> operator*(Grid_simd<S, V> a, Grid_simd<S, V> b) {
+accelerator_inline Grid_simd<S, V> operator*(Grid_simd<S, V> a, Grid_simd<S, V> b) {
   Grid_simd<S, V> ret;
   ret.v = binary<V>(a.v, b.v, MultSIMD());
   return ret;
@@ -642,7 +743,7 @@ inline Grid_simd<S, V> operator*(Grid_simd<S, V> a, Grid_simd<S, V> b) {
 
 // Distinguish between complex types and others
 template <class S, class V, IfComplex<S> = 0>
-inline Grid_simd<S, V> operator/(Grid_simd<S, V> a, Grid_simd<S, V> b) {
+accelerator_inline Grid_simd<S, V> operator/(Grid_simd<S, V> a, Grid_simd<S, V> b) {
   typedef Grid_simd<S, V> simd;
 
   simd ret;
@@ -652,17 +753,17 @@ inline Grid_simd<S, V> operator/(Grid_simd<S, V> a, Grid_simd<S, V> b) {
   ret = a * conjugate(b) ;
   den = b * conjugate(b) ;
 
-  
-  auto real_den = toReal(den);
-
-  ret.v=binary<V>(ret.v, real_den.v, DivSIMD());
-
+  // duplicates real part
+  auto real_den  = toReal(den);
+  simd zden;
+  memcpy((void *)&zden.v,(void *)&real_den.v,sizeof(zden));
+  ret.v=binary<V>(ret.v, zden.v, DivSIMD());
   return ret;
 };
 
 // Real/Integer types
 template <class S, class V, IfNotComplex<S> = 0>
-inline Grid_simd<S, V> operator/(Grid_simd<S, V> a, Grid_simd<S, V> b) {
+accelerator_inline Grid_simd<S, V> operator/(Grid_simd<S, V> a, Grid_simd<S, V> b) {
   Grid_simd<S, V> ret;
   ret.v = binary<V>(a.v, b.v, DivSIMD());
   return ret;
@@ -672,18 +773,18 @@ inline Grid_simd<S, V> operator/(Grid_simd<S, V> a, Grid_simd<S, V> b) {
 // Conjugate
 ///////////////////////
 template <class S, class V, IfComplex<S> = 0>
-inline Grid_simd<S, V> conjugate(const Grid_simd<S, V> &in) {
+accelerator_inline Grid_simd<S, V> conjugate(const Grid_simd<S, V> &in) {
   Grid_simd<S, V> ret;
   ret.v = unary<V>(in.v, ConjSIMD());
   return ret;
 }
 template <class S, class V, IfNotComplex<S> = 0>
-inline Grid_simd<S, V> conjugate(const Grid_simd<S, V> &in) {
+accelerator_inline Grid_simd<S, V> conjugate(const Grid_simd<S, V> &in) {
   return in;  // for real objects
 }
 // Suppress adj for integer types... // odd; why conjugate above but not adj??
 template <class S, class V, IfNotInteger<S> = 0>
-inline Grid_simd<S, V> adj(const Grid_simd<S, V> &in) {
+accelerator_inline Grid_simd<S, V> adj(const Grid_simd<S, V> &in) {
   return conjugate(in);
 }
 
@@ -691,17 +792,17 @@ inline Grid_simd<S, V> adj(const Grid_simd<S, V> &in) {
 // timesMinusI
 ///////////////////////
 template <class S, class V, IfComplex<S> = 0>
-inline void timesMinusI(Grid_simd<S, V> &ret, const Grid_simd<S, V> &in) {
+accelerator_inline void timesMinusI(Grid_simd<S, V> &ret, const Grid_simd<S, V> &in) {
   ret.v = binary<V>(in.v, ret.v, TimesMinusISIMD());
 }
 template <class S, class V, IfComplex<S> = 0>
-inline Grid_simd<S, V> timesMinusI(const Grid_simd<S, V> &in) {
+accelerator_inline Grid_simd<S, V> timesMinusI(const Grid_simd<S, V> &in) {
   Grid_simd<S, V> ret;
   timesMinusI(ret, in);
   return ret;
 }
 template <class S, class V, IfNotComplex<S> = 0>
-inline Grid_simd<S, V> timesMinusI(const Grid_simd<S, V> &in) {
+accelerator_inline Grid_simd<S, V> timesMinusI(const Grid_simd<S, V> &in) {
   return in;
 }
 
@@ -709,37 +810,34 @@ inline Grid_simd<S, V> timesMinusI(const Grid_simd<S, V> &in) {
 // timesI
 ///////////////////////
 template <class S, class V, IfComplex<S> = 0>
-inline void timesI(Grid_simd<S, V> &ret, const Grid_simd<S, V> &in) {
+accelerator_inline void timesI(Grid_simd<S, V> &ret, const Grid_simd<S, V> &in) {
   ret.v = binary<V>(in.v, ret.v, TimesISIMD());
 }
 template <class S, class V, IfComplex<S> = 0>
-inline Grid_simd<S, V> timesI(const Grid_simd<S, V> &in) {
+accelerator_inline Grid_simd<S, V> timesI(const Grid_simd<S, V> &in) {
   Grid_simd<S, V> ret;
   timesI(ret, in);
   return ret;
 }
 template <class S, class V, IfNotComplex<S> = 0>
-inline Grid_simd<S, V> timesI(const Grid_simd<S, V> &in) {
+accelerator_inline Grid_simd<S, V> timesI(const Grid_simd<S, V> &in) {
   return in;
 }
 
 /////////////////////
 // Inner, outer
 /////////////////////
-
 template <class S, class V>
-inline Grid_simd<S, V> innerProduct(const Grid_simd<S, V> &l,
-                                    const Grid_simd<S, V> &r) {
+accelerator_inline Grid_simd<S, V> innerProduct(const Grid_simd<S, V> &l,const Grid_simd<S, V> &r) {
   return conjugate(l) * r;
 }
 template <class S, class V>
-inline Grid_simd<S, V> outerProduct(const Grid_simd<S, V> &l,
-                                    const Grid_simd<S, V> &r) {
+accelerator_inline Grid_simd<S, V> outerProduct(const Grid_simd<S, V> &l,const Grid_simd<S, V> &r) {
   return l * conjugate(r);
 }
 
 template <class S, class V>
-inline Grid_simd<S, V> trace(const Grid_simd<S, V> &arg) {
+accelerator_inline Grid_simd<S, V> trace(const Grid_simd<S, V> &arg) {
   return arg;
 }
 
@@ -748,25 +846,33 @@ inline Grid_simd<S, V> trace(const Grid_simd<S, V> &arg) {
 // insert real into complex and zero imag;
 ////////////////////////////////////////////////////////////
 
+
+template <class T> struct toRealMapper {};
+template<> struct toRealMapper<vComplexF> {  typedef vRealF Realified; };
+template<> struct toRealMapper<vComplexD> {  typedef vRealD Realified; };
 // real = toReal( complex )
-template <class S, class V, IfReal<S> = 0>
-inline Grid_simd<S, V> toReal(const Grid_simd<std::complex<S>, V> &in) {
-  typedef Grid_simd<S, V> simd;
-  simd ret;
-  typename simd::conv_t conv;
-  conv.v = in.v;  // copy the vector content (bytewise)
-  for (int i = 0; i < simd::Nsimd(); i += 2) {
+template <class Csimd>  // must be a real arg
+accelerator_inline typename toRealMapper<Csimd>::Realified toReal(const Csimd &in) {
+  typedef typename toRealMapper<Csimd>::Realified Rsimd;
+  Rsimd ret;
+  typename Rsimd::conv_t conv;
+  memcpy((void *)&conv.v,(void *)&in.v,sizeof(conv.v));
+  for (int i = 0; i < Rsimd::Nsimd(); i += 2) {
     conv.s[i + 1] = conv.s[i];  // duplicate (r,r);(r,r);(r,r); etc...
   }
-  ret.v = conv.v;
+  memcpy((void *)&ret.v,(void *)&conv.v,sizeof(ret.v));
   return ret;
 }
 
+template <class T> struct toComplexMapper {};
+template<> struct toComplexMapper<vRealF> {  typedef vComplexF Complexified; };
+template<> struct toComplexMapper<vRealD> {  typedef vComplexD Complexified; };
+
 // complex = toComplex( real )
-template <class R, class V, IfReal<R> = 0>  // must be a real arg
-inline Grid_simd<std::complex<R>, V> toComplex(const Grid_simd<R, V> &in) {
-  typedef Grid_simd<R, V> Rsimd;
-  typedef Grid_simd<std::complex<R>, V> Csimd;
+template <class Rsimd>  // must be a real arg
+accelerator_inline typename toComplexMapper<Rsimd>::Complexified toComplex(const Rsimd &in) {
+
+  typedef typename toComplexMapper<Rsimd>::Complexified   Csimd;
   typename Rsimd::conv_t conv;  // address as real
 
   conv.v = in.v;
@@ -778,24 +884,12 @@ inline Grid_simd<std::complex<R>, V> toComplex(const Grid_simd<R, V> &in) {
     conv.s[i + 1] = 0.0;  // zero imaginary parts
   }
   Csimd ret;
-  ret.v = conv.v;
+  memcpy((void *)&ret.v,(void *)&conv.v,sizeof(ret.v));
   return ret;
 }
 
-///////////////////////////////
-// Define available types
-///////////////////////////////
-typedef Grid_simd<float, SIMD_Ftype> vRealF;
-typedef Grid_simd<double, SIMD_Dtype> vRealD;
-typedef Grid_simd<std::complex<float>, SIMD_Ftype> vComplexF;
-typedef Grid_simd<std::complex<double>, SIMD_Dtype> vComplexD;
-typedef Grid_simd<Integer, SIMD_Itype> vInteger;
 
-// Half precision; no arithmetic support
-typedef Grid_simd<uint16_t, SIMD_Htype>               vRealH;
-typedef Grid_simd<std::complex<uint16_t>, SIMD_Htype> vComplexH;
-
-inline void precisionChange(vRealF    *out,vRealD    *in,int nvec)
+accelerator_inline void precisionChange(vRealF    *out,vRealD    *in,int nvec)
 {
   assert((nvec&0x1)==0);
   for(int m=0;m*2<nvec;m++){
@@ -803,7 +897,7 @@ inline void precisionChange(vRealF    *out,vRealD    *in,int nvec)
     out[m].v=Optimization::PrecisionChange::DtoS(in[n].v,in[n+1].v);
   }
 }
-inline void precisionChange(vRealH    *out,vRealD    *in,int nvec)
+accelerator_inline void precisionChange(vRealH    *out,vRealD    *in,int nvec)
 {
   assert((nvec&0x3)==0);
   for(int m=0;m*4<nvec;m++){
@@ -811,7 +905,7 @@ inline void precisionChange(vRealH    *out,vRealD    *in,int nvec)
     out[m].v=Optimization::PrecisionChange::DtoH(in[n].v,in[n+1].v,in[n+2].v,in[n+3].v);
   }
 }
-inline void precisionChange(vRealH    *out,vRealF    *in,int nvec)
+accelerator_inline void precisionChange(vRealH    *out,vRealF    *in,int nvec)
 {
   assert((nvec&0x1)==0);
   for(int m=0;m*2<nvec;m++){
@@ -819,7 +913,7 @@ inline void precisionChange(vRealH    *out,vRealF    *in,int nvec)
     out[m].v=Optimization::PrecisionChange::StoH(in[n].v,in[n+1].v);
   }
 }
-inline void precisionChange(vRealD    *out,vRealF    *in,int nvec)
+accelerator_inline void precisionChange(vRealD    *out,vRealF    *in,int nvec)
 {
   assert((nvec&0x1)==0);
   for(int m=0;m*2<nvec;m++){
@@ -827,7 +921,7 @@ inline void precisionChange(vRealD    *out,vRealF    *in,int nvec)
     Optimization::PrecisionChange::StoD(in[m].v,out[n].v,out[n+1].v);
   }
 }
-inline void precisionChange(vRealD    *out,vRealH    *in,int nvec)
+accelerator_inline void precisionChange(vRealD    *out,vRealH    *in,int nvec)
 {
   assert((nvec&0x3)==0);
   for(int m=0;m*4<nvec;m++){
@@ -835,7 +929,7 @@ inline void precisionChange(vRealD    *out,vRealH    *in,int nvec)
     Optimization::PrecisionChange::HtoD(in[m].v,out[n].v,out[n+1].v,out[n+2].v,out[n+3].v);
   }
 }
-inline void precisionChange(vRealF    *out,vRealH    *in,int nvec)
+accelerator_inline void precisionChange(vRealF    *out,vRealH    *in,int nvec)
 {
   assert((nvec&0x1)==0);
   for(int m=0;m*2<nvec;m++){
@@ -843,21 +937,24 @@ inline void precisionChange(vRealF    *out,vRealH    *in,int nvec)
     Optimization::PrecisionChange::HtoS(in[m].v,out[n].v,out[n+1].v);
   }
 }
-inline void precisionChange(vComplexF *out,vComplexD *in,int nvec){ precisionChange((vRealF *)out,(vRealD *)in,nvec);}
-inline void precisionChange(vComplexH *out,vComplexD *in,int nvec){ precisionChange((vRealH *)out,(vRealD *)in,nvec);}
-inline void precisionChange(vComplexH *out,vComplexF *in,int nvec){ precisionChange((vRealH *)out,(vRealF *)in,nvec);}
-inline void precisionChange(vComplexD *out,vComplexF *in,int nvec){ precisionChange((vRealD *)out,(vRealF *)in,nvec);}
-inline void precisionChange(vComplexD *out,vComplexH *in,int nvec){ precisionChange((vRealD *)out,(vRealH *)in,nvec);}
-inline void precisionChange(vComplexF *out,vComplexH *in,int nvec){ precisionChange((vRealF *)out,(vRealH *)in,nvec);}
+accelerator_inline void precisionChange(vComplexF *out,vComplexD *in,int nvec){ precisionChange((vRealF *)out,(vRealD *)in,nvec);}
+accelerator_inline void precisionChange(vComplexH *out,vComplexD *in,int nvec){ precisionChange((vRealH *)out,(vRealD *)in,nvec);}
+accelerator_inline void precisionChange(vComplexH *out,vComplexF *in,int nvec){ precisionChange((vRealH *)out,(vRealF *)in,nvec);}
+accelerator_inline void precisionChange(vComplexD *out,vComplexF *in,int nvec){ precisionChange((vRealD *)out,(vRealF *)in,nvec);}
+accelerator_inline void precisionChange(vComplexD *out,vComplexH *in,int nvec){ precisionChange((vRealD *)out,(vRealH *)in,nvec);}
+accelerator_inline void precisionChange(vComplexF *out,vComplexH *in,int nvec){ precisionChange((vRealF *)out,(vRealH *)in,nvec);}
 
 // Check our vector types are of an appropriate size.
+
 #if defined QPX
 static_assert(2*sizeof(SIMD_Ftype) == sizeof(SIMD_Dtype), "SIMD vector lengths incorrect");
 static_assert(2*sizeof(SIMD_Ftype) == sizeof(SIMD_Itype), "SIMD vector lengths incorrect");
 #else
+#ifndef GENERIC_SCALAR
 static_assert(sizeof(SIMD_Ftype) == sizeof(SIMD_Dtype), "SIMD vector lengths incorrect");
 static_assert(sizeof(SIMD_Ftype) == sizeof(SIMD_Itype), "SIMD vector lengths incorrect");
 #endif
+#endif
 
 /////////////////////////////////////////
 // Some traits to recognise the types
@@ -874,6 +971,7 @@ template <> struct is_simd<vInteger>   : public std::true_type {};
 
 template <typename T> using IfSimd    = Invoke<std::enable_if<is_simd<T>::value, int> >;
 template <typename T> using IfNotSimd = Invoke<std::enable_if<!is_simd<T>::value, unsigned> >;
-}
+
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/simd/Grid_vector_unops.h b/Grid/simd/Grid_vector_unops.h
index 2244566f..d225699b 100644
--- a/Grid/simd/Grid_vector_unops.h
+++ b/Grid/simd/Grid_vector_unops.h
@@ -28,152 +28,152 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
+			   /*  END LEGAL */
 #ifndef GRID_VECTOR_UNOPS
 #define GRID_VECTOR_UNOPS
 
 #include <cmath>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 template <class scalar>
 struct SqrtRealFunctor {
-  scalar operator()(const scalar &a) const { return sqrt(real(a)); }
+  accelerator scalar operator()(const scalar &a) const { return sqrt(real(a)); }
 };
 
 template <class scalar>
 struct RSqrtRealFunctor {
-  scalar operator()(const scalar &a) const {
+  accelerator scalar operator()(const scalar &a) const {
     return scalar(1.0 / sqrt(real(a)));
   }
 };
 
 template <class scalar>
 struct CosRealFunctor {
-  scalar operator()(const scalar &a) const { return cos(real(a)); }
+  accelerator scalar operator()(const scalar &a) const { return cos(real(a)); }
 };
 
 template <class scalar>
 struct SinRealFunctor {
-  scalar operator()(const scalar &a) const { return sin(real(a)); }
+  accelerator scalar operator()(const scalar &a) const { return sin(real(a)); }
 };
 
 template <class scalar>
 struct AcosRealFunctor {
-  scalar operator()(const scalar &a) const { return acos(real(a)); }
+  accelerator scalar operator()(const scalar &a) const { return acos(real(a)); }
 };
 
 template <class scalar>
 struct AsinRealFunctor {
-  scalar operator()(const scalar &a) const { return asin(real(a)); }
+  accelerator scalar operator()(const scalar &a) const { return asin(real(a)); }
 };
 template <class scalar>
 struct LogRealFunctor {
-  scalar operator()(const scalar &a) const { return log(real(a)); }
+  accelerator scalar operator()(const scalar &a) const { return log(real(a)); }
 };
 template <class scalar>
 struct ExpFunctor {
-  scalar operator()(const scalar &a) const { return exp(a); }
+  accelerator scalar operator()(const scalar &a) const { return exp(a); }
 };
 template <class scalar>
 struct NotFunctor {
-  scalar operator()(const scalar &a) const { return (!a); }
+  accelerator scalar operator()(const scalar &a) const { return (!a); }
 };
 template <class scalar>
 struct AbsRealFunctor {
-  scalar operator()(const scalar &a) const { return std::abs(real(a)); }
+  accelerator scalar operator()(const scalar &a) const { return std::abs(real(a)); }
 };
 template <class scalar>
 struct PowRealFunctor {
   double y;
-  PowRealFunctor(double _y) : y(_y){};
-  scalar operator()(const scalar &a) const { return pow(real(a), y); }
+  accelerator PowRealFunctor(double _y) : y(_y){};
+  accelerator scalar operator()(const scalar &a) const { return pow(real(a), y); }
 };
 
 template <class scalar>
 struct ModIntFunctor {
   Integer y;
-  ModIntFunctor(Integer _y) : y(_y){};
-  scalar operator()(const scalar &a) const { return Integer(a) % y; }
+  accelerator ModIntFunctor(Integer _y) : y(_y){};
+  accelerator scalar operator()(const scalar &a) const { return Integer(a) % y; }
 };
 
 template <class scalar>
 struct DivIntFunctor {
   Integer y;
-  DivIntFunctor(Integer _y) : y(_y){};
-  scalar operator()(const scalar &a) const { return Integer(a) / y; }
+  accelerator DivIntFunctor(Integer _y) : y(_y){};
+  accelerator scalar operator()(const scalar &a) const { return Integer(a) / y; }
 };
 
 template <class scalar>
 struct RealFunctor {
-  scalar operator()(const scalar &a) const { return std::real(a); }
+  accelerator scalar operator()(const scalar &a) const { return real(a); }
 };
 template <class scalar>
 struct ImagFunctor {
-  scalar operator()(const scalar &a) const { return std::imag(a); }
+  accelerator scalar operator()(const scalar &a) const { return imag(a); }
 };
 template <class S, class V>
-inline Grid_simd<S, V> real(const Grid_simd<S, V> &r) {
+accelerator_inline Grid_simd<S, V> real(const Grid_simd<S, V> &r) {
   return SimdApply(RealFunctor<S>(), r);
 }
 template <class S, class V>
-inline Grid_simd<S, V> imag(const Grid_simd<S, V> &r) {
+accelerator_inline Grid_simd<S, V> imag(const Grid_simd<S, V> &r) {
   return SimdApply(ImagFunctor<S>(), r);
 }
 template <class S, class V>
-inline Grid_simd<S, V> sqrt(const Grid_simd<S, V> &r) {
+accelerator_inline Grid_simd<S, V> sqrt(const Grid_simd<S, V> &r) {
   return SimdApply(SqrtRealFunctor<S>(), r);
 }
 template <class S, class V>
-inline Grid_simd<S, V> rsqrt(const Grid_simd<S, V> &r) {
+accelerator_inline Grid_simd<S, V> rsqrt(const Grid_simd<S, V> &r) {
   return SimdApply(RSqrtRealFunctor<S>(), r);
 }
 template <class Scalar>
-inline Scalar rsqrt(const Scalar &r) {
+accelerator_inline Scalar rsqrt(const Scalar &r) {
   return (RSqrtRealFunctor<Scalar>(), r);
 }
 template <class S, class V>
-inline Grid_simd<S, V> cos(const Grid_simd<S, V> &r) {
+accelerator_inline Grid_simd<S, V> cos(const Grid_simd<S, V> &r) {
   return SimdApply(CosRealFunctor<S>(), r);
 }
 template <class S, class V>
-inline Grid_simd<S, V> sin(const Grid_simd<S, V> &r) {
+accelerator_inline Grid_simd<S, V> sin(const Grid_simd<S, V> &r) {
   return SimdApply(SinRealFunctor<S>(), r);
 }
 template <class S, class V>
-inline Grid_simd<S, V> acos(const Grid_simd<S, V> &r) {
+accelerator_inline Grid_simd<S, V> acos(const Grid_simd<S, V> &r) {
   return SimdApply(AcosRealFunctor<S>(), r);
 }
 template <class S, class V>
-inline Grid_simd<S, V> asin(const Grid_simd<S, V> &r) {
+accelerator_inline Grid_simd<S, V> asin(const Grid_simd<S, V> &r) {
   return SimdApply(AsinRealFunctor<S>(), r);
 }
 template <class S, class V>
-inline Grid_simd<S, V> log(const Grid_simd<S, V> &r) {
+accelerator_inline Grid_simd<S, V> log(const Grid_simd<S, V> &r) {
   return SimdApply(LogRealFunctor<S>(), r);
 }
 template <class S, class V>
-inline Grid_simd<S, V> abs(const Grid_simd<S, V> &r) {
+accelerator_inline Grid_simd<S, V> abs(const Grid_simd<S, V> &r) {
   return SimdApply(AbsRealFunctor<S>(), r);
 }
 template <class S, class V>
-inline Grid_simd<S, V> exp(const Grid_simd<S, V> &r) {
+accelerator_inline Grid_simd<S, V> exp(const Grid_simd<S, V> &r) {
   return SimdApply(ExpFunctor<S>(), r);
 }
 template <class S, class V>
-inline Grid_simd<S, V> Not(const Grid_simd<S, V> &r) {
+accelerator_inline Grid_simd<S, V> Not(const Grid_simd<S, V> &r) {
   return SimdApply(NotFunctor<S>(), r);
 }
 template <class S, class V>
-inline Grid_simd<S, V> pow(const Grid_simd<S, V> &r, double y) {
+accelerator_inline Grid_simd<S, V> pow(const Grid_simd<S, V> &r, double y) {
   return SimdApply(PowRealFunctor<S>(y), r);
 }
 template <class S, class V>
-inline Grid_simd<S, V> mod(const Grid_simd<S, V> &r, Integer y) {
+accelerator_inline Grid_simd<S, V> mod(const Grid_simd<S, V> &r, Integer y) {
   return SimdApply(ModIntFunctor<S>(y), r);
 }
 template <class S, class V>
-inline Grid_simd<S, V> div(const Grid_simd<S, V> &r, Integer y) {
+accelerator_inline Grid_simd<S, V> div(const Grid_simd<S, V> &r, Integer y) {
   return SimdApply(DivIntFunctor<S>(y), r);
 }
 ////////////////////////////////////////////////////////////////////////////
@@ -181,43 +181,43 @@ inline Grid_simd<S, V> div(const Grid_simd<S, V> &r, Integer y) {
 ////////////////////////////////////////////////////////////////////////////
 template <class scalar>
 struct AndFunctor {
-  scalar operator()(const scalar &x, const scalar &y) const { return x & y; }
+  accelerator scalar operator()(const scalar &x, const scalar &y) const { return x & y; }
 };
 template <class scalar>
 struct OrFunctor {
-  scalar operator()(const scalar &x, const scalar &y) const { return x | y; }
+  accelerator scalar operator()(const scalar &x, const scalar &y) const { return x | y; }
 };
 template <class scalar>
 struct AndAndFunctor {
-  scalar operator()(const scalar &x, const scalar &y) const { return x && y; }
+  accelerator scalar operator()(const scalar &x, const scalar &y) const { return x && y; }
 };
 template <class scalar>
 struct OrOrFunctor {
-  scalar operator()(const scalar &x, const scalar &y) const { return x || y; }
+  accelerator scalar operator()(const scalar &x, const scalar &y) const { return x || y; }
 };
 
 ////////////////////////////////
 // Calls to simd binop functors
 ////////////////////////////////
 template <class S, class V>
-inline Grid_simd<S, V> operator&(const Grid_simd<S, V> &x,
-                                 const Grid_simd<S, V> &y) {
+accelerator_inline Grid_simd<S, V> operator&(const Grid_simd<S, V> &x,
+				 const Grid_simd<S, V> &y) {
   return SimdApplyBinop(AndFunctor<S>(), x, y);
 }
 template <class S, class V>
-inline Grid_simd<S, V> operator&&(const Grid_simd<S, V> &x,
-                                  const Grid_simd<S, V> &y) {
+accelerator_inline Grid_simd<S, V> operator&&(const Grid_simd<S, V> &x,
+				  const Grid_simd<S, V> &y) {
   return SimdApplyBinop(AndAndFunctor<S>(), x, y);
 }
 template <class S, class V>
-inline Grid_simd<S, V> operator|(const Grid_simd<S, V> &x,
-                                 const Grid_simd<S, V> &y) {
+accelerator_inline Grid_simd<S, V> operator|(const Grid_simd<S, V> &x,
+				 const Grid_simd<S, V> &y) {
   return SimdApplyBinop(OrFunctor<S>(), x, y);
 }
 template <class S, class V>
-inline Grid_simd<S, V> operator||(const Grid_simd<S, V> &x,
-                                  const Grid_simd<S, V> &y) {
+accelerator_inline Grid_simd<S, V> operator||(const Grid_simd<S, V> &x,
+				  const Grid_simd<S, V> &y) {
   return SimdApplyBinop(OrOrFunctor<S>(), x, y);
 }
-}
+NAMESPACE_END(Grid);
 #endif
diff --git a/Grid/simd/IBM_qpx.h b/Grid/simd/IBM_qpx.h
index df91d8e6..610d758b 100644
--- a/Grid/simd/IBM_qpx.h
+++ b/Grid/simd/IBM_qpx.h
@@ -1,4 +1,4 @@
-   /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,8 +23,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_ASM_BGQ_QPX_H
 #define GRID_ASM_BGQ_QPX_H
 
@@ -153,32 +153,32 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 /*********************************************************
  * Macro sequences encoding QCD
  *********************************************************/
-#define LOCK_GAUGE(dir)							\
-  {									\
-    uint64_t byte_addr = (uint64_t)&U._odata[sU];			\
-    int count = (sizeof(U._odata[0])+63)/64;				\
-    asm (" mtctr %0 \n"							\
-	 " mr " HASH(REP) ", %1\n"					\
-	 " li " HASH(IMM) ", 64\n"					\
+#define LOCK_GAUGE(dir)						\
+  {								\
+    uint64_t byte_addr = (uint64_t)&U[sU];		\
+    int count = (sizeof(U[0])+63)/64;			\
+    asm (" mtctr %0 \n"						\
+	 " mr " HASH(REP) ", %1\n"				\
+	 " li " HASH(IMM) ", 64\n"				\
 	 "0:\n"							\
-	 LOCK_SET							\
-	 "  add " HASH(REP) "," HASH(IMM) "," HASH(REP) "\n"		\
+	 LOCK_SET						\
+	 "  add " HASH(REP) "," HASH(IMM) "," HASH(REP) "\n"	\
 	 "  bdnz 0b\n"						\
-	 : : "b" (count), "b" (byte_addr) );					\
+	 : : "b" (count), "b" (byte_addr) );			\
   }
 
-#define UNLOCK_GAUGE(dir)						\
-  {									\
-    uint64_t byte_addr = (uint64_t)&U._odata[sU];			\
-    int count = (sizeof(U._odata[0])+63)/64;				\
-    asm (" mtctr %0 \n"							\
-	 " mr " HASH(REP) ", %1\n"					\
-	 " li " HASH(IMM) ", 64\n"					\
-	 "0:\n"								\
-	 LOCK_CLEAR							\
-	 "  add " HASH(REP) "," HASH(IMM) "," HASH(REP) "\n"		\
+#define UNLOCK_GAUGE(dir)					\
+  {								\
+    uint64_t byte_addr = (uint64_t)&U[sU];		\
+    int count = (sizeof(U[0])+63)/64;			\
+    asm (" mtctr %0 \n"						\
+	 " mr " HASH(REP) ", %1\n"				\
+	 " li " HASH(IMM) ", 64\n"				\
+	 "0:\n"							\
+	 LOCK_CLEAR						\
+	 "  add " HASH(REP) "," HASH(IMM) "," HASH(REP) "\n"	\
 	 "  bdnz 0b\n"						\
-	 : : "b" (count), "b" (byte_addr) );					\
+	 : : "b" (count), "b" (byte_addr) );			\
   }
 
 #define ZERO_PSI				\
@@ -201,11 +201,11 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define MULT_2SPIN_QPXf(ptr,p)    MULT_2SPIN_QPX_INTERNAL(ptr,p,VLOAD,16) 
 
 #define MULT_2SPIN_QPX_INTERNAL(ptr,p,ULOAD,USKIP) {			\
-    uint64_t ub = ((uint64_t)ptr);				\
-    asm (							\
-         ULOAD(%0,%3,U0)					\
-         ULOAD(%1,%3,U1)					\
-         ULOAD(%2,%3,U2)					\
+    uint64_t ub = ((uint64_t)ptr);					\
+    asm (								\
+         ULOAD(%0,%3,U0)						\
+         ULOAD(%1,%3,U1)						\
+         ULOAD(%2,%3,U2)						\
 	 VMUL_RR_RI(UChi_00,U0,Chi_00)					\
 	 VMUL_RR_RI(UChi_01,U1,Chi_00)					\
 	 VMUL_RR_RI(UChi_02,U2,Chi_00)					\
@@ -235,7 +235,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 	 VMADD_MII_IR(UChi_10,U0,Chi_11,UChi_10)			\
 	 VMADD_MII_IR(UChi_11,U1,Chi_11,UChi_11)			\
 	 VMADD_MII_IR(UChi_12,U2,Chi_11,UChi_12)			\
-	 : : "b" (USKIP*1), "b" (USKIP*4), "b" (USKIP*7), "b" (ub ));		\
+	 : : "b" (USKIP*1), "b" (USKIP*4), "b" (USKIP*7), "b" (ub ));	\
     asm (								\
          ULOAD(%0,%3,U0)						\
          ULOAD(%1,%3,U1)						\
@@ -252,30 +252,30 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 	 VMADD_MII_IR(UChi_10,U0,Chi_12,UChi_10)			\
 	 VMADD_MII_IR(UChi_11,U1,Chi_12,UChi_11)			\
 	 VMADD_MII_IR(UChi_12,U2,Chi_12,UChi_12)			\
-	 : : "b" (USKIP*2), "b" (USKIP*5), "b" (USKIP*8), "b" (ub ));		\
+	 : : "b" (USKIP*2), "b" (USKIP*5), "b" (USKIP*8), "b" (ub ));	\
   }
 
 
-#define MULT_2SPIN_DIR_PF(A,p) MULT_2SPIN_PF(&U._odata[sU](A),p)
+#define MULT_2SPIN_DIR_PF(A,p) MULT_2SPIN_PF(&U[sU](A),p)
 #define MULT_2SPIN_PF(ptr,pf) MULT_2SPIN(ptr,pf)
 
-#define SAVE_RESULT(base,basep) {\
-    uint64_t ub = ((uint64_t)base)  - (VSIZE);			\
-    asm("mr " HASH(REP)  ", %0;\n"					\
-	"li " HASH(IMM)      "," HASH(VSIZE)" ;\n"				\
-	VSTOREu(IMM,REP,psi_00)						\
-	VSTOREu(IMM,REP,psi_01)						\
-	VSTOREu(IMM,REP,psi_02)						\
-	VSTOREu(IMM,REP,psi_10)						\
-	VSTOREu(IMM,REP,psi_11)						\
-	VSTOREu(IMM,REP,psi_12)						\
-	VSTOREu(IMM,REP,psi_20)						\
-	VSTOREu(IMM,REP,psi_21)						\
-	VSTOREu(IMM,REP,psi_22)						\
-	VSTOREu(IMM,REP,psi_30)						\
-	VSTOREu(IMM,REP,psi_31)						\
-	VSTOREu(IMM,REP,psi_32)						\
-	: : "b" (ub) : HASH(pIMM), HASH(pREP) );				\
+#define SAVE_RESULT(base,basep) {			\
+    uint64_t ub = ((uint64_t)base)  - (VSIZE);		\
+    asm("mr " HASH(REP)  ", %0;\n"			\
+	"li " HASH(IMM)      "," HASH(VSIZE)" ;\n"	\
+	VSTOREu(IMM,REP,psi_00)				\
+	VSTOREu(IMM,REP,psi_01)				\
+	VSTOREu(IMM,REP,psi_02)				\
+	VSTOREu(IMM,REP,psi_10)				\
+	VSTOREu(IMM,REP,psi_11)				\
+	VSTOREu(IMM,REP,psi_12)				\
+	VSTOREu(IMM,REP,psi_20)				\
+	VSTOREu(IMM,REP,psi_21)				\
+	VSTOREu(IMM,REP,psi_22)				\
+	VSTOREu(IMM,REP,psi_30)				\
+	VSTOREu(IMM,REP,psi_31)				\
+	VSTOREu(IMM,REP,psi_32)				\
+	: : "b" (ub) : HASH(pIMM), HASH(pREP) );	\
   }
 
 
@@ -295,7 +295,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 	"li  " HASH(IMM) ",(2*" HASH(VSIZE) ");\n"			\
 	VLOADu(IMM,REP,Chi_01)						\
 	VLOADu(IMM,REP,Chi_10)						\
-	VLOADu(IMM,REP,Chi_12)	: : "b" (ub)  : HASH(pIMM), HASH(pREP) );	\
+	VLOADu(IMM,REP,Chi_12)	: : "b" (ub)  : HASH(pIMM), HASH(pREP) ); \
   }
 
 #define LOAD_CHIMU(base) {						\
@@ -316,27 +316,27 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 	VLOADu(IMM,REP,Chi_12)						\
 	VLOADu(IMM,REP,Chi_21)						\
 	VLOADu(IMM,REP,Chi_30)						\
-	VLOADu(IMM,REP,Chi_32)	: : "b" (ub)  : HASH(pIMM), HASH(pREP) );	\
+	VLOADu(IMM,REP,Chi_32)	: : "b" (ub)  : HASH(pIMM), HASH(pREP) ); \
   }
 
 //      hspin(0)=fspin(0)+timesI(fspin(3));
 //      hspin(1)=fspin(1)+timesI(fspin(2));
-#define XP_PROJMEM(base) {					\
-    LOAD_CHIMU(base);						\
-    asm (							\
-         VONE(one)						\
-	 VMADD_MII_IR(Chi_00,one,Chi_30,Chi_00)			\
-	 VMADD_MII_IR(Chi_01,one,Chi_31,Chi_01)			\
-	 VMADD_MII_IR(Chi_02,one,Chi_32,Chi_02)			\
-	 VMADD_MII_IR(Chi_10,one,Chi_20,Chi_10)			\
-	 VMADD_MII_IR(Chi_11,one,Chi_21,Chi_11)			\
-	 VMADD_MII_IR(Chi_12,one,Chi_22,Chi_12)			\
-							);	\
+#define XP_PROJMEM(base) {						\
+    LOAD_CHIMU(base);							\
+    asm (								\
+         VONE(one)							\
+	 VMADD_MII_IR(Chi_00,one,Chi_30,Chi_00)				\
+	 VMADD_MII_IR(Chi_01,one,Chi_31,Chi_01)				\
+	 VMADD_MII_IR(Chi_02,one,Chi_32,Chi_02)				\
+	 VMADD_MII_IR(Chi_10,one,Chi_20,Chi_10)				\
+	 VMADD_MII_IR(Chi_11,one,Chi_21,Chi_11)				\
+	 VMADD_MII_IR(Chi_12,one,Chi_22,Chi_12)				\
+								);	\
   }
 
-#define XM_PROJMEM(base) {				\
-    LOAD_CHIMU(base);					\
-    asm (						\
+#define XM_PROJMEM(base) {					\
+    LOAD_CHIMU(base);						\
+    asm (							\
          VONE(one)						\
 	 VMADD_II_MIR(Chi_00,one,Chi_30,Chi_00)			\
 	 VMADD_II_MIR(Chi_01,one,Chi_31,Chi_01)			\
@@ -349,19 +349,19 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 //      hspin(0)=fspin(0)-fspin(3);
 //      hspin(1)=fspin(1)+fspin(2);
-#define YP_PROJMEM(base) {  \
-    LOAD_CHIMU(base);						\
-    asm (							\
-	 VSUB(Chi_00,Chi_00,Chi_30)				\
-	 VSUB(Chi_01,Chi_01,Chi_31)				\
-	 VSUB(Chi_02,Chi_02,Chi_32)				\
-	 VADD(Chi_10,Chi_10,Chi_20)				\
-	 VADD(Chi_11,Chi_11,Chi_21)				\
-	 VADD(Chi_12,Chi_12,Chi_22)				\
-							);	\
+#define YP_PROJMEM(base) {						\
+    LOAD_CHIMU(base);							\
+    asm (								\
+	 VSUB(Chi_00,Chi_00,Chi_30)					\
+	 VSUB(Chi_01,Chi_01,Chi_31)					\
+	 VSUB(Chi_02,Chi_02,Chi_32)					\
+	 VADD(Chi_10,Chi_10,Chi_20)					\
+	 VADD(Chi_11,Chi_11,Chi_21)					\
+	 VADD(Chi_12,Chi_12,Chi_22)					\
+								);	\
   }
 
-#define YM_PROJMEM(base) {			\
+#define YM_PROJMEM(base) {					\
     LOAD_CHIMU(base);						\
     asm (							\
 	 VADD(Chi_00,Chi_00,Chi_30)				\
@@ -372,162 +372,162 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 	 VSUB(Chi_12,Chi_12,Chi_22)			);	\
   }
 
-	    /*Gz
-	     *  0 0  i  0   [0]+-i[2]
-	     *  0 0  0 -i   [1]-+i[3]
-	     * -i 0  0  0
-	     *  0 i  0  0
-	     */
-#define ZP_PROJMEM(base) {  \
-    LOAD_CHIMU(base);						\
-    asm (							\
-         VONE(one)						\
-	 VMADD_MII_IR(Chi_00,one,Chi_20,Chi_00)			\
-	 VMADD_MII_IR(Chi_01,one,Chi_21,Chi_01)			\
-	 VMADD_MII_IR(Chi_02,one,Chi_22,Chi_02)			\
-	 VMADD_II_MIR(Chi_10,one,Chi_30,Chi_10)			\
-	 VMADD_II_MIR(Chi_11,one,Chi_31,Chi_11)			\
-	 VMADD_II_MIR(Chi_12,one,Chi_32,Chi_12)			\
-							);	\
+/*Gz
+ *  0 0  i  0   [0]+-i[2]
+ *  0 0  0 -i   [1]-+i[3]
+ * -i 0  0  0
+ *  0 i  0  0
+ */
+#define ZP_PROJMEM(base) {						\
+    LOAD_CHIMU(base);							\
+    asm (								\
+         VONE(one)							\
+	 VMADD_MII_IR(Chi_00,one,Chi_20,Chi_00)				\
+	 VMADD_MII_IR(Chi_01,one,Chi_21,Chi_01)				\
+	 VMADD_MII_IR(Chi_02,one,Chi_22,Chi_02)				\
+	 VMADD_II_MIR(Chi_10,one,Chi_30,Chi_10)				\
+	 VMADD_II_MIR(Chi_11,one,Chi_31,Chi_11)				\
+	 VMADD_II_MIR(Chi_12,one,Chi_32,Chi_12)				\
+								);	\
   }
 
-#define ZM_PROJMEM(base) {  \
-    LOAD_CHIMU(base);						\
-    asm (							\
-         VONE(one)						\
-	 VMADD_II_MIR(Chi_00,one,Chi_20,Chi_00)			\
-	 VMADD_II_MIR(Chi_01,one,Chi_21,Chi_01)			\
-	 VMADD_II_MIR(Chi_02,one,Chi_22,Chi_02)			\
-	 VMADD_MII_IR(Chi_10,one,Chi_30,Chi_10)			\
-	 VMADD_MII_IR(Chi_11,one,Chi_31,Chi_11)			\
-	 VMADD_MII_IR(Chi_12,one,Chi_32,Chi_12)			\
-							);	\
+#define ZM_PROJMEM(base) {						\
+    LOAD_CHIMU(base);							\
+    asm (								\
+         VONE(one)							\
+	 VMADD_II_MIR(Chi_00,one,Chi_20,Chi_00)				\
+	 VMADD_II_MIR(Chi_01,one,Chi_21,Chi_01)				\
+	 VMADD_II_MIR(Chi_02,one,Chi_22,Chi_02)				\
+	 VMADD_MII_IR(Chi_10,one,Chi_30,Chi_10)				\
+	 VMADD_MII_IR(Chi_11,one,Chi_31,Chi_11)				\
+	 VMADD_MII_IR(Chi_12,one,Chi_32,Chi_12)				\
+								);	\
   }
-	    /*Gt
-	     *  0 0  1  0 [0]+-[2]
-	     *  0 0  0  1 [1]+-[3]
-	     *  1 0  0  0
-	     *  0 1  0  0
-	     */
-#define TP_PROJMEM(base) {  \
-    LOAD_CHIMU(base);						\
-    asm (							\
-	 VADD(Chi_00,Chi_00,Chi_20)				\
-	 VADD(Chi_01,Chi_01,Chi_21)				\
-	 VADD(Chi_02,Chi_02,Chi_22)				\
-	 VADD(Chi_10,Chi_10,Chi_30)				\
-	 VADD(Chi_11,Chi_11,Chi_31)				\
-	 VADD(Chi_12,Chi_12,Chi_32)				\
-							);	\
+/*Gt
+ *  0 0  1  0 [0]+-[2]
+ *  0 0  0  1 [1]+-[3]
+ *  1 0  0  0
+ *  0 1  0  0
+ */
+#define TP_PROJMEM(base) {						\
+    LOAD_CHIMU(base);							\
+    asm (								\
+	 VADD(Chi_00,Chi_00,Chi_20)					\
+	 VADD(Chi_01,Chi_01,Chi_21)					\
+	 VADD(Chi_02,Chi_02,Chi_22)					\
+	 VADD(Chi_10,Chi_10,Chi_30)					\
+	 VADD(Chi_11,Chi_11,Chi_31)					\
+	 VADD(Chi_12,Chi_12,Chi_32)					\
+								);	\
   }
 
-#define TM_PROJMEM(base) {  \
-    LOAD_CHIMU(base);						\
-    asm (							\
-	 VSUB(Chi_00,Chi_00,Chi_20)				\
-	 VSUB(Chi_01,Chi_01,Chi_21)				\
-	 VSUB(Chi_02,Chi_02,Chi_22)				\
-	 VSUB(Chi_10,Chi_10,Chi_30)				\
-	 VSUB(Chi_11,Chi_11,Chi_31)				\
-	 VSUB(Chi_12,Chi_12,Chi_32)				\
-							);	\
+#define TM_PROJMEM(base) {						\
+    LOAD_CHIMU(base);							\
+    asm (								\
+	 VSUB(Chi_00,Chi_00,Chi_20)					\
+	 VSUB(Chi_01,Chi_01,Chi_21)					\
+	 VSUB(Chi_02,Chi_02,Chi_22)					\
+	 VSUB(Chi_10,Chi_10,Chi_30)					\
+	 VSUB(Chi_11,Chi_11,Chi_31)					\
+	 VSUB(Chi_12,Chi_12,Chi_32)					\
+								);	\
   }
 
 /*
-      fspin(0)=hspin(0);
-      fspin(1)=hspin(1);
-      fspin(2)=timesMinusI(hspin(1));
-      fspin(3)=timesMinusI(hspin(0));
+  fspin(0)=hspin(0);
+  fspin(1)=hspin(1);
+  fspin(2)=timesMinusI(hspin(1));
+  fspin(3)=timesMinusI(hspin(0));
 
-      fspin(0)+=hspin(0);
-      fspin(1)+=hspin(1);
-      fspin(2)-=timesI(hspin(1));
-      fspin(3)-=timesI(hspin(0));
- */
-#define XP_RECON {				\
-    asm(\
-	VONE(one)\
-	VMOV(psi_00,UChi_00) 	VMOV(psi_01,UChi_01)	VMOV(psi_02,UChi_02)\
-	VMOV(psi_10,UChi_10) 	VMOV(psi_11,UChi_11)	VMOV(psi_12,UChi_12)\
-	VZERO(psi_20)	VZERO(psi_21)	VZERO(psi_22) \
-	VZERO(psi_30) 	VZERO(psi_31)   VZERO(psi_32) \
-	VMADD_II_MIR(psi_20,one,UChi_10,psi_20)	      \
-	VMADD_II_MIR(psi_21,one,UChi_11,psi_21)	      \
-	VMADD_II_MIR(psi_22,one,UChi_12,psi_22)	      \
-	VMADD_II_MIR(psi_30,one,UChi_00,psi_30)	      \
-	VMADD_II_MIR(psi_31,one,UChi_01,psi_31)	      \
-	VMADD_II_MIR(psi_32,one,UChi_02,psi_32)	      \
-	);		     \
+  fspin(0)+=hspin(0);
+  fspin(1)+=hspin(1);
+  fspin(2)-=timesI(hspin(1));
+  fspin(3)-=timesI(hspin(0));
+*/
+#define XP_RECON {							\
+    asm(								\
+	VONE(one)							\
+	VMOV(psi_00,UChi_00) 	VMOV(psi_01,UChi_01)	VMOV(psi_02,UChi_02) \
+	VMOV(psi_10,UChi_10) 	VMOV(psi_11,UChi_11)	VMOV(psi_12,UChi_12) \
+	VZERO(psi_20)	VZERO(psi_21)	VZERO(psi_22)			\
+	VZERO(psi_30) 	VZERO(psi_31)   VZERO(psi_32)			\
+	VMADD_II_MIR(psi_20,one,UChi_10,psi_20)				\
+	VMADD_II_MIR(psi_21,one,UChi_11,psi_21)				\
+	VMADD_II_MIR(psi_22,one,UChi_12,psi_22)				\
+	VMADD_II_MIR(psi_30,one,UChi_00,psi_30)				\
+	VMADD_II_MIR(psi_31,one,UChi_01,psi_31)				\
+	VMADD_II_MIR(psi_32,one,UChi_02,psi_32)				\
+	);								\
   }
 
-#define XM_RECON {				\
-    asm(\
-	VONE(one)\
-	VMOV(psi_00,UChi_00) 	VMOV(psi_01,UChi_01)	VMOV(psi_02,UChi_02)\
-	VMOV(psi_10,UChi_10) 	VMOV(psi_11,UChi_11)	VMOV(psi_12,UChi_12)\
-	VZERO(psi_20)	VZERO(psi_21)	VZERO(psi_22) \
-	VZERO(psi_30) 	VZERO(psi_31)   VZERO(psi_32) \
-	VMADD_MII_IR(psi_20,one,UChi_10,psi_20)	      \
-	VMADD_MII_IR(psi_21,one,UChi_11,psi_21)	      \
-	VMADD_MII_IR(psi_22,one,UChi_12,psi_22)	      \
-	VMADD_MII_IR(psi_30,one,UChi_00,psi_30)	      \
-	VMADD_MII_IR(psi_31,one,UChi_01,psi_31)	      \
-	VMADD_MII_IR(psi_32,one,UChi_02,psi_32)	      \
-	);		     \
+#define XM_RECON {							\
+    asm(								\
+	VONE(one)							\
+	VMOV(psi_00,UChi_00) 	VMOV(psi_01,UChi_01)	VMOV(psi_02,UChi_02) \
+	VMOV(psi_10,UChi_10) 	VMOV(psi_11,UChi_11)	VMOV(psi_12,UChi_12) \
+	VZERO(psi_20)	VZERO(psi_21)	VZERO(psi_22)			\
+	VZERO(psi_30) 	VZERO(psi_31)   VZERO(psi_32)			\
+	VMADD_MII_IR(psi_20,one,UChi_10,psi_20)				\
+	VMADD_MII_IR(psi_21,one,UChi_11,psi_21)				\
+	VMADD_MII_IR(psi_22,one,UChi_12,psi_22)				\
+	VMADD_MII_IR(psi_30,one,UChi_00,psi_30)				\
+	VMADD_MII_IR(psi_31,one,UChi_01,psi_31)				\
+	VMADD_MII_IR(psi_32,one,UChi_02,psi_32)				\
+	);								\
   }
 
-#define XP_RECON_ACCUM {				\
-    asm(\
-	VONE(one)\
+#define XP_RECON_ACCUM {						\
+    asm(								\
+	VONE(one)							\
 	VADD(psi_00,psi_00,UChi_00) 	VADD(psi_01,psi_01,UChi_01)	VADD(psi_02,psi_02,UChi_02) \
 	VADD(psi_10,psi_10,UChi_10) 	VADD(psi_11,psi_11,UChi_11)	VADD(psi_12,psi_12,UChi_12) \
-	VMADD_II_MIR(psi_20,one,UChi_10,psi_20)	      \
-	VMADD_II_MIR(psi_21,one,UChi_11,psi_21)	      \
-	VMADD_II_MIR(psi_22,one,UChi_12,psi_22)	      \
-	VMADD_II_MIR(psi_30,one,UChi_00,psi_30)	      \
-	VMADD_II_MIR(psi_31,one,UChi_01,psi_31)	      \
-	VMADD_II_MIR(psi_32,one,UChi_02,psi_32)	      \
-	);		     \
+	VMADD_II_MIR(psi_20,one,UChi_10,psi_20)				\
+	VMADD_II_MIR(psi_21,one,UChi_11,psi_21)				\
+	VMADD_II_MIR(psi_22,one,UChi_12,psi_22)				\
+	VMADD_II_MIR(psi_30,one,UChi_00,psi_30)				\
+	VMADD_II_MIR(psi_31,one,UChi_01,psi_31)				\
+	VMADD_II_MIR(psi_32,one,UChi_02,psi_32)				\
+	);								\
   }
 
-#define XM_RECON_ACCUM {				\
-    asm(\
-	VONE(one)\
+#define XM_RECON_ACCUM {						\
+    asm(								\
+	VONE(one)							\
 	VADD(psi_00,psi_00,UChi_00) 	VADD(psi_01,psi_01,UChi_01)	VADD(psi_02,psi_02,UChi_02) \
 	VADD(psi_10,psi_10,UChi_10) 	VADD(psi_11,psi_11,UChi_11)	VADD(psi_12,psi_12,UChi_12) \
-	VMADD_MII_IR(psi_20,one,UChi_10,psi_20)	      \
-	VMADD_MII_IR(psi_21,one,UChi_11,psi_21)	      \
-	VMADD_MII_IR(psi_22,one,UChi_12,psi_22)	      \
-	VMADD_MII_IR(psi_30,one,UChi_00,psi_30)	      \
-	VMADD_MII_IR(psi_31,one,UChi_01,psi_31)	      \
-	VMADD_MII_IR(psi_32,one,UChi_02,psi_32)	      \
-	);		     \
+	VMADD_MII_IR(psi_20,one,UChi_10,psi_20)				\
+	VMADD_MII_IR(psi_21,one,UChi_11,psi_21)				\
+	VMADD_MII_IR(psi_22,one,UChi_12,psi_22)				\
+	VMADD_MII_IR(psi_30,one,UChi_00,psi_30)				\
+	VMADD_MII_IR(psi_31,one,UChi_01,psi_31)				\
+	VMADD_MII_IR(psi_32,one,UChi_02,psi_32)				\
+	);								\
   }
 
 //      fspin(2)+=hspin(1);
 //      fspin(3)-=hspin(0);
-#define YP_RECON_ACCUM {\
-    asm(\
+#define YP_RECON_ACCUM {						\
+    asm(								\
 	VADD(psi_00,psi_00,UChi_00) 	VADD(psi_01,psi_01,UChi_01)	VADD(psi_02,psi_02,UChi_02) \
 	VADD(psi_10,psi_10,UChi_10) 	VADD(psi_11,psi_11,UChi_11)	VADD(psi_12,psi_12,UChi_12) \
 	VADD(psi_20,psi_20,UChi_10) 	VADD(psi_21,psi_21,UChi_11)	VADD(psi_22,psi_22,UChi_12) \
 	VSUB(psi_30,psi_30,UChi_00) 	VSUB(psi_31,psi_31,UChi_01)	VSUB(psi_32,psi_32,UChi_02) \
-	);\
- }
-#define YM_RECON_ACCUM {\
-    asm(\
+	);								\
+  }
+#define YM_RECON_ACCUM {						\
+    asm(								\
 	VADD(psi_00,psi_00,UChi_00) 	VADD(psi_01,psi_01,UChi_01)	VADD(psi_02,psi_02,UChi_02) \
 	VADD(psi_10,psi_10,UChi_10) 	VADD(psi_11,psi_11,UChi_11)	VADD(psi_12,psi_12,UChi_12) \
 	VSUB(psi_20,psi_20,UChi_10) 	VSUB(psi_21,psi_21,UChi_11)	VSUB(psi_22,psi_22,UChi_12) \
 	VADD(psi_30,psi_30,UChi_00) 	VADD(psi_31,psi_31,UChi_01)	VADD(psi_32,psi_32,UChi_02) \
-	);\
- }
+	);								\
+  }
 
 //      fspin(2)-=timesI(hspin(0));
 //      fspin(3)+=timesI(hspin(1));
-#define ZP_RECON_ACCUM {\
-    asm(\
-	VONE(one)\
+#define ZP_RECON_ACCUM {						\
+    asm(								\
+	VONE(one)							\
 	VADD(psi_00,psi_00,UChi_00) 	VADD(psi_01,psi_01,UChi_01)	VADD(psi_02,psi_02,UChi_02) \
 	VADD(psi_10,psi_10,UChi_10) 	VADD(psi_11,psi_11,UChi_11)	VADD(psi_12,psi_12,UChi_12) \
 	VMADD_II_MIR(psi_20,one,UChi_00,psi_20)				\
@@ -536,12 +536,12 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 	VMADD_MII_IR(psi_30,one,UChi_10,psi_30)				\
 	VMADD_MII_IR(psi_31,one,UChi_11,psi_31)				\
 	VMADD_MII_IR(psi_32,one,UChi_12,psi_32)				\
-	);\
- }
+	);								\
+  }
 
-#define ZM_RECON_ACCUM {\
-    asm(\
-	VONE(one)\
+#define ZM_RECON_ACCUM {						\
+    asm(								\
+	VONE(one)							\
 	VADD(psi_00,psi_00,UChi_00) 	VADD(psi_01,psi_01,UChi_01)	VADD(psi_02,psi_02,UChi_02) \
 	VADD(psi_10,psi_10,UChi_10) 	VADD(psi_11,psi_11,UChi_11)	VADD(psi_12,psi_12,UChi_12) \
 	VMADD_MII_IR(psi_20,one,UChi_00,psi_20)				\
@@ -550,37 +550,37 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 	VMADD_II_MIR(psi_30,one,UChi_10,psi_30)				\
 	VMADD_II_MIR(psi_31,one,UChi_11,psi_31)				\
 	VMADD_II_MIR(psi_32,one,UChi_12,psi_32)				\
-	);\
- }
+	);								\
+  }
 
 //      fspin(2)+=hspin(0);
 //      fspin(3)+=hspin(1);
-#define TP_RECON_ACCUM {\
-    asm(\
+#define TP_RECON_ACCUM {						\
+    asm(								\
 	VADD(psi_00,psi_00,UChi_00) 	VADD(psi_01,psi_01,UChi_01)	VADD(psi_02,psi_02,UChi_02) \
 	VADD(psi_10,psi_10,UChi_10) 	VADD(psi_11,psi_11,UChi_11)	VADD(psi_12,psi_12,UChi_12) \
 	VADD(psi_20,psi_20,UChi_00) 	VADD(psi_21,psi_21,UChi_01)	VADD(psi_22,psi_22,UChi_02) \
 	VADD(psi_30,psi_30,UChi_10) 	VADD(psi_31,psi_31,UChi_11)	VADD(psi_32,psi_32,UChi_12) \
-	);\
- }
+	);								\
+  }
 
-#define TM_RECON_ACCUM {\
-    asm(\
+#define TM_RECON_ACCUM {						\
+    asm(								\
 	VADD(psi_00,psi_00,UChi_00) 	VADD(psi_01,psi_01,UChi_01)	VADD(psi_02,psi_02,UChi_02) \
 	VADD(psi_10,psi_10,UChi_10) 	VADD(psi_11,psi_11,UChi_11)	VADD(psi_12,psi_12,UChi_12) \
 	VSUB(psi_20,psi_20,UChi_00) 	VSUB(psi_21,psi_21,UChi_01)	VSUB(psi_22,psi_22,UChi_02) \
 	VSUB(psi_30,psi_30,UChi_10) 	VSUB(psi_31,psi_31,UChi_11)	VSUB(psi_32,psi_32,UChi_12) \
-	);\
- }
+	);								\
+  }
 
 
 #define ADD_RESULTi(PTR,pf)						\
   LOAD_CHIMU(PTR)							\
   asm(									\
-  VADD(psi_00,chi_00,psi_00)  VADD(psi_01,chi_01,psi_01)  VADD(psi_02,chi_02,psi_02) \
-  VADD(psi_10,chi_10,psi_10)  VADD(psi_11,chi_11,psi_11)  VADD(psi_12,chi_12,psi_12) \
-  VADD(psi_20,chi_20,psi_20)  VADD(psi_21,chi_21,psi_21)  VADD(psi_22,chi_22,psi_22) \
-  VADD(psi_30,chi_30,psi_30)  VADD(psi_31,chi_31,psi_31)  VADD(psi_32,chi_32,psi_32) ); \
+      VADD(psi_00,chi_00,psi_00)  VADD(psi_01,chi_01,psi_01)  VADD(psi_02,chi_02,psi_02) \
+      VADD(psi_10,chi_10,psi_10)  VADD(psi_11,chi_11,psi_11)  VADD(psi_12,chi_12,psi_12) \
+      VADD(psi_20,chi_20,psi_20)  VADD(psi_21,chi_21,psi_21)  VADD(psi_22,chi_22,psi_22) \
+      VADD(psi_30,chi_30,psi_30)  VADD(psi_31,chi_31,psi_31)  VADD(psi_32,chi_32,psi_32) ); \
   SAVE_RESULT(PTR,pf);
 
 
@@ -590,9 +590,9 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 #define PERMUTE_DIR0 {							\
     asm(								\
-	VPERMI(perm_reg)							\
-	VPERM(Chi_00,perm_reg)   VPERM(Chi_01,perm_reg)   VPERM(Chi_02,perm_reg)	\
-	VPERM(Chi_10,perm_reg)   VPERM(Chi_11,perm_reg)   VPERM(Chi_12,perm_reg) );	\
+	VPERMI(perm_reg)						\
+	VPERM(Chi_00,perm_reg)   VPERM(Chi_01,perm_reg)   VPERM(Chi_02,perm_reg) \
+	VPERM(Chi_10,perm_reg)   VPERM(Chi_11,perm_reg)   VPERM(Chi_12,perm_reg) ); \
   }
 
 #endif
diff --git a/Grid/simd/IBM_qpx_double.h b/Grid/simd/IBM_qpx_double.h
index 60709102..a69b1552 100644
--- a/Grid/simd/IBM_qpx_double.h
+++ b/Grid/simd/IBM_qpx_double.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,8 +23,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 // No guard; ok multi-include
 #undef VSIZE
 #undef VLOAD
diff --git a/Grid/simd/IBM_qpx_single.h b/Grid/simd/IBM_qpx_single.h
index ab903ea7..28bab041 100644
--- a/Grid/simd/IBM_qpx_single.h
+++ b/Grid/simd/IBM_qpx_single.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,8 +23,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 // No guard; ok multi-include
 #undef VSIZE
 #undef VLOAD
diff --git a/Grid/simd/Intel512common.h b/Grid/simd/Intel512common.h
index e69e541c..c9472918 100644
--- a/Grid/simd/Intel512common.h
+++ b/Grid/simd/Intel512common.h
@@ -1,4 +1,4 @@
-   /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,8 +23,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_ASM_INTEL_COMMON_512_H
 #define GRID_ASM_INTEL_COMMON_512_H
 
@@ -36,10 +36,10 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 // Opcodes common 
 ////////////////////////////////////////////////////////////////////////////////////////////////////
-#define MASK_REGS \
-  __asm__ ("mov     $0xAAAA, %%eax \n"\
-           "kmovw    %%eax, %%k6 \n"\
-           "mov     $0x5555, %%eax \n"\
+#define MASK_REGS					\
+  __asm__ ("mov     $0xAAAA, %%eax \n"			\
+           "kmovw    %%eax, %%k6 \n"			\
+           "mov     $0x5555, %%eax \n"			\
            "kmovw    %%eax, %%k7 \n" : : : "%eax");
 
 //#define label(B) __asm__ ( __func__ _LINE__ #B ":\n" );
@@ -47,44 +47,44 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define VZEROf(A)       "vpxorq " #A ","  #A "," #A ";\n"
 #define VZEROd(A)       "vpxorq " #A ","  #A "," #A ";\n"
 
-#define VTIMESIf(A,DEST, Z) \
-  VTIMESI0f(A,DEST, Z) \
-  VTIMESI1f(A,DEST, Z) \
+#define VTIMESIf(A,DEST, Z)			\
+  VTIMESI0f(A,DEST, Z)				\
+  VTIMESI1f(A,DEST, Z)				\
   VTIMESI2f(A,DEST, Z) 
 
-#define VTIMESId(A,DEST, Z) \
-  VTIMESI0d(A,DEST, Z) \
-  VTIMESI1d(A,DEST, Z) \
+#define VTIMESId(A,DEST, Z)			\
+  VTIMESI0d(A,DEST, Z)				\
+  VTIMESI1d(A,DEST, Z)				\
   VTIMESI2d(A,DEST, Z) 
 
-#define VTIMESMINUSIf(A,DEST, Z) \
-        VTIMESMINUSI0f(A,DEST, Z) \
-        VTIMESMINUSI1f(A,DEST, Z) \
-        VTIMESMINUSI2f(A,DEST, Z) 
+#define VTIMESMINUSIf(A,DEST, Z)		\
+  VTIMESMINUSI0f(A,DEST, Z)			\
+  VTIMESMINUSI1f(A,DEST, Z)			\
+  VTIMESMINUSI2f(A,DEST, Z) 
 
-#define VTIMESMINUSId(A,DEST, Z) \
-        VTIMESMINUSI0d(A,DEST, Z) \
-        VTIMESMINUSI1d(A,DEST, Z) \
-        VTIMESMINUSI2d(A,DEST, Z) 
+#define VTIMESMINUSId(A,DEST, Z)		\
+  VTIMESMINUSI0d(A,DEST, Z)			\
+  VTIMESMINUSI1d(A,DEST, Z)			\
+  VTIMESMINUSI2d(A,DEST, Z) 
 
 #define VACCTIMESIf(A,ACC,tmp)			\
- VACCTIMESI0f(A,ACC,tmp)			\
- VACCTIMESI1f(A,ACC,tmp)			\
- VACCTIMESI2f(A,ACC,tmp)			
+  VACCTIMESI0f(A,ACC,tmp)			\
+  VACCTIMESI1f(A,ACC,tmp)			\
+  VACCTIMESI2f(A,ACC,tmp)			
 
 #define VACCTIMESId(A,ACC,tmp)			\
- VACCTIMESI0d(A,ACC,tmp)			\
- VACCTIMESI1d(A,ACC,tmp)			\
- VACCTIMESI2d(A,ACC,tmp)			
+  VACCTIMESI0d(A,ACC,tmp)			\
+  VACCTIMESI1d(A,ACC,tmp)			\
+  VACCTIMESI2d(A,ACC,tmp)			
 
-#define VACCTIMESMINUSIf(A,ACC,tmp)			\
-  VACCTIMESMINUSI0f(A,ACC,tmp)				\
-  VACCTIMESMINUSI1f(A,ACC,tmp)				\
+#define VACCTIMESMINUSIf(A,ACC,tmp)		\
+  VACCTIMESMINUSI0f(A,ACC,tmp)			\
+  VACCTIMESMINUSI1f(A,ACC,tmp)			\
   VACCTIMESMINUSI2f(A,ACC,tmp)			
 
-#define VACCTIMESMINUSId(A,ACC,tmp)			\
-  VACCTIMESMINUSI0d(A,ACC,tmp)				\
-  VACCTIMESMINUSI1d(A,ACC,tmp)				\
+#define VACCTIMESMINUSId(A,ACC,tmp)		\
+  VACCTIMESMINUSI0d(A,ACC,tmp)			\
+  VACCTIMESMINUSI1d(A,ACC,tmp)			\
   VACCTIMESMINUSI2d(A,ACC,tmp)			
 
 #define LOAD64a(A,ptr)  "movq %0, %" #A :  : "r"(ptr)  : #A  
diff --git a/Grid/simd/Intel512double.h b/Grid/simd/Intel512double.h
index 632b5639..f0b7fa0d 100644
--- a/Grid/simd/Intel512double.h
+++ b/Grid/simd/Intel512double.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,8 +23,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 // No guard can be multiply included as undef clearage
 #undef VZERO
 #undef VMOV
diff --git a/Grid/simd/Intel512imci.h b/Grid/simd/Intel512imci.h
index 7176890f..655c001d 100644
--- a/Grid/simd/Intel512imci.h
+++ b/Grid/simd/Intel512imci.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,8 +23,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_ASM_AV512_H
 #define GRID_ASM_AV512_H
 
@@ -44,32 +44,32 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define ZENDf(Criir,Ciirr, tmp) ZEND1f(Criir,Ciirr, tmp) ZEND2f(Criir,Ciirr, tmp)
 #define ZENDd(Criir,Ciirr, tmp) ZEND1d(Criir,Ciirr, tmp) ZEND2d(Criir,Ciirr, tmp)
 
-#define ZMULMEM2SPf(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)\
-  VSHUFMEMf(O,P,tmp) \
-  VMULMEMf(O,P,B,Biirr) \
-  VMULMEMf(O,P,C,Ciirr) \
-  VMULf(tmp,B,Briir) \
+#define ZMULMEM2SPf(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)	\
+  VSHUFMEMf(O,P,tmp)						\
+  VMULMEMf(O,P,B,Biirr)						\
+  VMULMEMf(O,P,C,Ciirr)						\
+  VMULf(tmp,B,Briir)						\
   VMULf(tmp,C,Criir)
 
-#define ZMULMEM2SPd(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)\
-  VSHUFMEMd(O,P,tmp)  \
+#define ZMULMEM2SPd(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)	\
+  VSHUFMEMd(O,P,tmp)						\
   VMULMEMd(O,P,B,Biirr)  \ 
-  VMULMEMd(O,P,C,Ciirr)  \
-  VMULd(tmp,B,Briir)  \
-  VMULd(tmp,C,Criir) 
+VMULMEMd(O,P,C,Ciirr)  \
+VMULd(tmp,B,Briir)  \
+VMULd(tmp,C,Criir) 
 
-#define ZMADDMEM2SPf(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)\
-  VSHUFMEMf(O,P,tmp) \
-  VMADDMEMf(O,P,B,Biirr) \
-  VMADDMEMf(O,P,C,Ciirr) \
-  VMADDf(tmp,B,Briir) \
+#define ZMADDMEM2SPf(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)	\
+  VSHUFMEMf(O,P,tmp)						\
+  VMADDMEMf(O,P,B,Biirr)					\
+  VMADDMEMf(O,P,C,Ciirr)					\
+  VMADDf(tmp,B,Briir)						\
   VMADDf(tmp,C,Criir)
 
 #define ZMADDMEM2SPd(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)	\
-  VSHUFMEMd(O,P,tmp) \
-  VMADDMEMd(O,P,B,Biirr) \
-  VMADDMEMd(O,P,C,Ciirr) \
-  VMADDd(tmp,B,Briir) \
+  VSHUFMEMd(O,P,tmp)						\
+  VMADDMEMd(O,P,B,Biirr)					\
+  VMADDMEMd(O,P,C,Ciirr)					\
+  VMADDd(tmp,B,Briir)						\
   VMADDd(tmp,C,Criir)
 
 #define ZEND1d(Criir,Ciirr, tmp) "vaddpd  " #Criir "{cdab} ," #Criir "," #Criir"{%k6}"  ";\n"
@@ -106,7 +106,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define VACCTIMESMINUSI1f(A,ACC,tmp)  "vsubps  " #A "{cdab}," #ACC "," #ACC"{%k7}" ";\n"
 #define VACCTIMESMINUSI2f(A,ACC,tmp)  "vaddps  " #A "{cdab}," #ACC "," #ACC"{%k6}" ";\n"
 
-	   // Acc = Acc - i A
+// Acc = Acc - i A
 #define VACCTIMESMINUSI0d(A,ACC,tmp)  
 #define VACCTIMESMINUSI1d(A,ACC,tmp)  "vsubpd  " #A "{cdab}," #ACC "," #ACC"{%k7}" ";\n"
 #define VACCTIMESMINUSI2d(A,ACC,tmp)  "vaddpd  " #A "{cdab}," #ACC "," #ACC"{%k6}" ";\n"
diff --git a/Grid/simd/Intel512single.h b/Grid/simd/Intel512single.h
index ed135651..5c1e4135 100644
--- a/Grid/simd/Intel512single.h
+++ b/Grid/simd/Intel512single.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,8 +23,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 // No guard can be multiply included as undef clearge of macros
 #undef VZERO
 #undef VMOV
diff --git a/Grid/simd/Intel512wilson.h b/Grid/simd/Intel512wilson.h
index 64142a2e..af53a63d 100644
--- a/Grid/simd/Intel512wilson.h
+++ b/Grid/simd/Intel512wilson.h
@@ -23,8 +23,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_ASM_INTEL_512_QCD_H
 #define GRID_ASM_INTEL_512_QCD_H
 
@@ -101,7 +101,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define LOCK_GAUGE(dir) 
 #define UNLOCK_GAUGE(dir) 
 
-//  const SiteSpinor * ptr = & in._odata[offset];	
+//  const SiteSpinor * ptr = & in[offset];	
 #define LOAD_CHIMU(PTR)	 LOAD64(%r8,PTR) __asm__ ( LOAD_CHIMUi );
 #define LOAD_CHI(PTR)	 LOAD64(%r8,PTR) __asm__ ( LOAD_CHIi );
 #define SAVE_UCHI(PTR)	 SAVE_UCHIi(PTR)
@@ -109,18 +109,18 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define SAVE_RESULT(PT,R) SAVE_RESULTi(PT,R)
 #define ADD_RESULT(PT,R) ADD_RESULTi(PT,R)
 
-#define ZERO_PSI			\
-  asm( VZERO(psi_00)			\
-       VZERO(psi_01)			\
-       VZERO(psi_02)			\
-       VZERO(psi_10)			\
-       VZERO(psi_11)			\
-       VZERO(psi_12)			\
-       VZERO(psi_20)			\
-       VZERO(psi_21)			\
-       VZERO(psi_22)			\
-       VZERO(psi_30)			\
-       VZERO(psi_31)			\
+#define ZERO_PSI				\
+  asm( VZERO(psi_00)				\
+       VZERO(psi_01)				\
+       VZERO(psi_02)				\
+       VZERO(psi_10)				\
+       VZERO(psi_11)				\
+       VZERO(psi_12)				\
+       VZERO(psi_20)				\
+       VZERO(psi_21)				\
+       VZERO(psi_22)				\
+       VZERO(psi_30)				\
+       VZERO(psi_31)				\
        VZERO(psi_32));
 
 #define LOAD_CHIMUi				\
@@ -143,43 +143,43 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
   VLOAD(10,%r8,Chimu_31)			\
   VLOAD(11,%r8,Chimu_32)		
 
-#define SHUF_CHIMU23i\
-	   VSHUFMEM(6,%r8,Chimu_20)		\
-	   VSHUFMEM(7,%r8,Chimu_21)		\
-	   VSHUFMEM(8,%r8,Chimu_22)		\
-	   VSHUFMEM(9,%r8,Chimu_30)		\
-	   VSHUFMEM(10,%r8,Chimu_31)		\
-	   VSHUFMEM(11,%r8,Chimu_32)		
+#define SHUF_CHIMU23i				\
+  VSHUFMEM(6,%r8,Chimu_20)			\
+  VSHUFMEM(7,%r8,Chimu_21)			\
+  VSHUFMEM(8,%r8,Chimu_22)			\
+  VSHUFMEM(9,%r8,Chimu_30)			\
+  VSHUFMEM(10,%r8,Chimu_31)			\
+  VSHUFMEM(11,%r8,Chimu_32)		
 
 #define LOAD_CHIi				\
-  VLOAD(0,%r8,Chi_00)					\
-  VLOAD(1,%r8,Chi_01)					\
-  VLOAD(2,%r8,Chi_02)					\
-  VLOAD(3,%r8,Chi_10)					\
-  VLOAD(4,%r8,Chi_11)					\
+  VLOAD(0,%r8,Chi_00)				\
+  VLOAD(1,%r8,Chi_01)				\
+  VLOAD(2,%r8,Chi_02)				\
+  VLOAD(3,%r8,Chi_10)				\
+  VLOAD(4,%r8,Chi_11)				\
   VLOAD(5,%r8,Chi_12)	
 
-#define SAVE_UCHIi(PTR)				\
-  LOAD64(%r8,PTR)				\
-  __asm__ (					\
-  VSTORE(0,%r8,UChi_00)				\
-  VSTORE(1,%r8,UChi_01)				\
-  VSTORE(2,%r8,UChi_02)				\
-  VSTORE(3,%r8,UChi_10)				\
-  VSTORE(4,%r8,UChi_11)				\
-  VSTORE(5,%r8,UChi_12)				);
+#define SAVE_UCHIi(PTR)							\
+  LOAD64(%r8,PTR)							\
+  __asm__ (								\
+	   VSTORE(0,%r8,UChi_00)					\
+	   VSTORE(1,%r8,UChi_01)					\
+	   VSTORE(2,%r8,UChi_02)					\
+	   VSTORE(3,%r8,UChi_10)					\
+	   VSTORE(4,%r8,UChi_11)					\
+	   VSTORE(5,%r8,UChi_12)				);
 
-#define SAVE_CHIi(PTR)				\
-  LOAD64(%r8,PTR)				\
-  __asm__ (					\
-  VSTORE(0,%r8,Chi_00)				\
-  VSTORE(1,%r8,Chi_01)				\
-  VSTORE(2,%r8,Chi_02)				\
-  VSTORE(3,%r8,Chi_10)				\
-  VSTORE(4,%r8,Chi_11)				\
-  VSTORE(5,%r8,Chi_12)				);
+#define SAVE_CHIi(PTR)						\
+  LOAD64(%r8,PTR)						\
+  __asm__ (							\
+	   VSTORE(0,%r8,Chi_00)					\
+	   VSTORE(1,%r8,Chi_01)					\
+	   VSTORE(2,%r8,Chi_02)					\
+	   VSTORE(3,%r8,Chi_10)					\
+	   VSTORE(4,%r8,Chi_11)					\
+	   VSTORE(5,%r8,Chi_12)				);
 
-#define MULT_2SPIN_DIR_PF(A,p) MULT_2SPIN_PF(&U._odata[sU](A),p)
+#define MULT_2SPIN_DIR_PF(A,p) MULT_2SPIN_PF(&U[sU](A),p)
 #define MULT_2SPIN_PF(ptr,pf) MULT_2SPIN(ptr,pf)
 
 //////////////////////////////////////////////////////////////////
@@ -187,352 +187,352 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 //////////////////////////////////////////////////////////////////
 //      hspin(0)=fspin(0)+timesI(fspin(3));
 //      hspin(1)=fspin(1)+timesI(fspin(2));
-#define XP_PROJMEM(PTR) \
-  LOAD64(%r8,PTR)							\
-  __asm__ (								\
+#define XP_PROJMEM(PTR)						\
+  LOAD64(%r8,PTR)						\
+  __asm__ (							\
 	   LOAD_CHIi						\
-	   SHUF_CHIMU23i						\
-	   VACCTIMESI1(Chi_00,Chi_00,Chimu_30)		\
-	   VACCTIMESI1(Chi_01,Chi_01,Chimu_31)		\
-	   VACCTIMESI1(Chi_02,Chi_02,Chimu_32)		\
-	   VACCTIMESI1(Chi_10,Chi_10,Chimu_20)		\
-	   VACCTIMESI1(Chi_11,Chi_11,Chimu_21)		\
-	   VACCTIMESI1(Chi_12,Chi_12,Chimu_22)		\
-	   VACCTIMESI2(Chi_00,Chi_00,Chimu_30)		\
-	   VACCTIMESI2(Chi_01,Chi_01,Chimu_31)		\
-	   VACCTIMESI2(Chi_02,Chi_02,Chimu_32)		\
-	   VACCTIMESI2(Chi_10,Chi_10,Chimu_20)		\
-	   VACCTIMESI2(Chi_11,Chi_11,Chimu_21)		\
+	   SHUF_CHIMU23i					\
+	   VACCTIMESI1(Chi_00,Chi_00,Chimu_30)			\
+	   VACCTIMESI1(Chi_01,Chi_01,Chimu_31)			\
+	   VACCTIMESI1(Chi_02,Chi_02,Chimu_32)			\
+	   VACCTIMESI1(Chi_10,Chi_10,Chimu_20)			\
+	   VACCTIMESI1(Chi_11,Chi_11,Chimu_21)			\
+	   VACCTIMESI1(Chi_12,Chi_12,Chimu_22)			\
+	   VACCTIMESI2(Chi_00,Chi_00,Chimu_30)			\
+	   VACCTIMESI2(Chi_01,Chi_01,Chimu_31)			\
+	   VACCTIMESI2(Chi_02,Chi_02,Chimu_32)			\
+	   VACCTIMESI2(Chi_10,Chi_10,Chimu_20)			\
+	   VACCTIMESI2(Chi_11,Chi_11,Chimu_21)			\
 	   VACCTIMESI2(Chi_12,Chi_12,Chimu_22)		);
 
 
-#define YP_PROJMEM(ptr) \
-  LOAD64(%r8,ptr)		\
-  __asm__ (					\
-  LOAD_CHIMU01i					\
-  VSUBMEM(9,%r8 ,Chimu_00,Chi_00)		\
-  VSUBMEM(10,%r8,Chimu_01,Chi_01)		\
-  VSUBMEM(11,%r8,Chimu_02,Chi_02)		\
-  VADDMEM(6,%r8,Chimu_10,Chi_10)		\
-  VADDMEM(7,%r8,Chimu_11,Chi_11)		\
-  VADDMEM(8,%r8,Chimu_12,Chi_12)		);
+#define YP_PROJMEM(ptr)						\
+  LOAD64(%r8,ptr)						\
+  __asm__ (							\
+	   LOAD_CHIMU01i					\
+	   VSUBMEM(9,%r8 ,Chimu_00,Chi_00)			\
+	   VSUBMEM(10,%r8,Chimu_01,Chi_01)			\
+	   VSUBMEM(11,%r8,Chimu_02,Chi_02)			\
+	   VADDMEM(6,%r8,Chimu_10,Chi_10)			\
+	   VADDMEM(7,%r8,Chimu_11,Chi_11)			\
+	   VADDMEM(8,%r8,Chimu_12,Chi_12)		);
 
-#define ZP_PROJMEM(PTR) \
-  LOAD64(%r8,PTR)							\
-  __asm__ (								\
+#define ZP_PROJMEM(PTR)						\
+  LOAD64(%r8,PTR)						\
+  __asm__ (							\
 	   LOAD_CHIi						\
-	   SHUF_CHIMU23i						\
-	   VACCTIMESI1(Chi_00,Chi_00,Chimu_20)				\
-	   VACCTIMESI1(Chi_01,Chi_01,Chimu_21)		   	        \
-	   VACCTIMESI1(Chi_02,Chi_02,Chimu_22)				\
-	   VACCTIMESMINUSI1(Chi_10,Chi_10,Chimu_30)			\
-	   VACCTIMESMINUSI1(Chi_11,Chi_11,Chimu_31)			\
-	   VACCTIMESMINUSI1(Chi_12,Chi_12,Chimu_32)			\
-	   VACCTIMESI2(Chi_00,Chi_00,Chimu_20)				\
-	   VACCTIMESI2(Chi_01,Chi_01,Chimu_21)				\
-	   VACCTIMESI2(Chi_02,Chi_02,Chimu_22)				\
+	   SHUF_CHIMU23i					\
+	   VACCTIMESI1(Chi_00,Chi_00,Chimu_20)			\
+	   VACCTIMESI1(Chi_01,Chi_01,Chimu_21)			\
+	   VACCTIMESI1(Chi_02,Chi_02,Chimu_22)			\
+	   VACCTIMESMINUSI1(Chi_10,Chi_10,Chimu_30)		\
+	   VACCTIMESMINUSI1(Chi_11,Chi_11,Chimu_31)		\
+	   VACCTIMESMINUSI1(Chi_12,Chi_12,Chimu_32)		\
+	   VACCTIMESI2(Chi_00,Chi_00,Chimu_20)			\
+	   VACCTIMESI2(Chi_01,Chi_01,Chimu_21)			\
+	   VACCTIMESI2(Chi_02,Chi_02,Chimu_22)			\
 	   VACCTIMESMINUSI2(Chi_10,Chi_10,Chimu_30)		\
 	   VACCTIMESMINUSI2(Chi_11,Chi_11,Chimu_31)		\
 	   VACCTIMESMINUSI2(Chi_12,Chi_12,Chimu_32)	);
 
 
-#define TP_PROJMEM(ptr)				\
-  LOAD64(%r8,ptr)				\
-  __asm__ (					\
-	   LOAD_CHIMU01i			\
-	   VADDMEM(6,%r8 ,Chimu_00,Chi_00)	\
-	   VADDMEM(7,%r8,Chimu_01,Chi_01)	\
-	   VADDMEM(8,%r8,Chimu_02,Chi_02)	\
-	   VADDMEM(9,%r8,Chimu_10,Chi_10)	\
-	   VADDMEM(10,%r8,Chimu_11,Chi_11)	\
+#define TP_PROJMEM(ptr)					\
+  LOAD64(%r8,ptr)					\
+  __asm__ (						\
+	   LOAD_CHIMU01i				\
+	   VADDMEM(6,%r8 ,Chimu_00,Chi_00)		\
+	   VADDMEM(7,%r8,Chimu_01,Chi_01)		\
+	   VADDMEM(8,%r8,Chimu_02,Chi_02)		\
+	   VADDMEM(9,%r8,Chimu_10,Chi_10)		\
+	   VADDMEM(10,%r8,Chimu_11,Chi_11)		\
 	   VADDMEM(11,%r8,Chimu_12,Chi_12)	);
 
 //      hspin(0)=fspin(0)-timesI(fspin(3))
 //      hspin(1)=fspin(1)-timesI(fspin(2))
-#define XM_PROJMEM(PTR) \
-  LOAD64(%r8,PTR)\
-  __asm__ (								\
-	   LOAD_CHIi \
-	   SHUF_CHIMU23i						\
-	   VACCTIMESMINUSI1(Chi_00,Chi_00,Chimu_30)\
-	   VACCTIMESMINUSI1(Chi_01,Chi_01,Chimu_31)\
-	   VACCTIMESMINUSI1(Chi_02,Chi_02,Chimu_32)\
-	   VACCTIMESMINUSI1(Chi_10,Chi_10,Chimu_20)\
-	   VACCTIMESMINUSI1(Chi_11,Chi_11,Chimu_21)\
-	   VACCTIMESMINUSI1(Chi_12,Chi_12,Chimu_22)\
-	   VACCTIMESMINUSI2(Chi_00,Chi_00,Chimu_30)\
-	   VACCTIMESMINUSI2(Chi_01,Chi_01,Chimu_31)\
-	   VACCTIMESMINUSI2(Chi_02,Chi_02,Chimu_32)\
-	   VACCTIMESMINUSI2(Chi_10,Chi_10,Chimu_20)\
-	   VACCTIMESMINUSI2(Chi_11,Chi_11,Chimu_21)\
+#define XM_PROJMEM(PTR)					\
+  LOAD64(%r8,PTR)					\
+  __asm__ (						\
+	   LOAD_CHIi					\
+	   SHUF_CHIMU23i				\
+	   VACCTIMESMINUSI1(Chi_00,Chi_00,Chimu_30)	\
+	   VACCTIMESMINUSI1(Chi_01,Chi_01,Chimu_31)	\
+	   VACCTIMESMINUSI1(Chi_02,Chi_02,Chimu_32)	\
+	   VACCTIMESMINUSI1(Chi_10,Chi_10,Chimu_20)	\
+	   VACCTIMESMINUSI1(Chi_11,Chi_11,Chimu_21)	\
+	   VACCTIMESMINUSI1(Chi_12,Chi_12,Chimu_22)	\
+	   VACCTIMESMINUSI2(Chi_00,Chi_00,Chimu_30)	\
+	   VACCTIMESMINUSI2(Chi_01,Chi_01,Chimu_31)	\
+	   VACCTIMESMINUSI2(Chi_02,Chi_02,Chimu_32)	\
+	   VACCTIMESMINUSI2(Chi_10,Chi_10,Chimu_20)	\
+	   VACCTIMESMINUSI2(Chi_11,Chi_11,Chimu_21)	\
 	   VACCTIMESMINUSI2(Chi_12,Chi_12,Chimu_22) );
 
-#define YM_PROJMEM(ptr)				\
-  LOAD64(%r8,ptr)				\
-  __asm__ (					\
-  LOAD_CHIMU01i					\
-  VADDMEM(9,%r8 ,Chimu_00,Chi_00)		\
-  VADDMEM(10,%r8,Chimu_01,Chi_01)		\
-  VADDMEM(11,%r8,Chimu_02,Chi_02)		\
-  VSUBMEM(6,%r8,Chimu_10,Chi_10)		\
-  VSUBMEM(7,%r8,Chimu_11,Chi_11)		\
-  VSUBMEM(8,%r8,Chimu_12,Chi_12)			);
-
-#define ZM_PROJMEM(PTR) \
-  LOAD64(%r8,PTR)							\
+#define YM_PROJMEM(ptr)							\
+  LOAD64(%r8,ptr)							\
   __asm__ (								\
-           LOAD_CHIi \
-	   SHUF_CHIMU23i						\
-	   VACCTIMESMINUSI1(Chi_00,Chi_00,Chimu_20)\
-	   VACCTIMESMINUSI1(Chi_01,Chi_01,Chimu_21)\
-	   VACCTIMESMINUSI1(Chi_02,Chi_02,Chimu_22)\
-	   VACCTIMESI1(Chi_10,Chi_10,Chimu_30)\
-	   VACCTIMESI1(Chi_11,Chi_11,Chimu_31)\
-	   VACCTIMESI1(Chi_12,Chi_12,Chimu_32)\
-	   VACCTIMESMINUSI2(Chi_00,Chi_00,Chimu_20)\
-	   VACCTIMESMINUSI2(Chi_01,Chi_01,Chimu_21)\
-	   VACCTIMESMINUSI2(Chi_02,Chi_02,Chimu_22)\
-	   VACCTIMESI2(Chi_10,Chi_10,Chimu_30)\
-	   VACCTIMESI2(Chi_11,Chi_11,Chimu_31)\
+	   LOAD_CHIMU01i						\
+	   VADDMEM(9,%r8 ,Chimu_00,Chi_00)				\
+	   VADDMEM(10,%r8,Chimu_01,Chi_01)				\
+	   VADDMEM(11,%r8,Chimu_02,Chi_02)				\
+	   VSUBMEM(6,%r8,Chimu_10,Chi_10)				\
+	   VSUBMEM(7,%r8,Chimu_11,Chi_11)				\
+	   VSUBMEM(8,%r8,Chimu_12,Chi_12)			);
+
+#define ZM_PROJMEM(PTR)					\
+  LOAD64(%r8,PTR)					\
+  __asm__ (						\
+           LOAD_CHIi					\
+	   SHUF_CHIMU23i				\
+	   VACCTIMESMINUSI1(Chi_00,Chi_00,Chimu_20)	\
+	   VACCTIMESMINUSI1(Chi_01,Chi_01,Chimu_21)	\
+	   VACCTIMESMINUSI1(Chi_02,Chi_02,Chimu_22)	\
+	   VACCTIMESI1(Chi_10,Chi_10,Chimu_30)		\
+	   VACCTIMESI1(Chi_11,Chi_11,Chimu_31)		\
+	   VACCTIMESI1(Chi_12,Chi_12,Chimu_32)		\
+	   VACCTIMESMINUSI2(Chi_00,Chi_00,Chimu_20)	\
+	   VACCTIMESMINUSI2(Chi_01,Chi_01,Chimu_21)	\
+	   VACCTIMESMINUSI2(Chi_02,Chi_02,Chimu_22)	\
+	   VACCTIMESI2(Chi_10,Chi_10,Chimu_30)		\
+	   VACCTIMESI2(Chi_11,Chi_11,Chimu_31)		\
 	   VACCTIMESI2(Chi_12,Chi_12,Chimu_32) );
 
-#define TM_PROJMEM(ptr)				\
-  LOAD64(%r8,ptr)				\
-  __asm__ (					\
-  LOAD_CHIMU01i					\
-  VSUBMEM(6,%r8,Chimu_00,Chi_00)		\
-  VSUBMEM(7,%r8,Chimu_01,Chi_01)		\
-  VSUBMEM(8,%r8,Chimu_02,Chi_02)		\
-  VSUBMEM(9,%r8,Chimu_10,Chi_10)		\
-  VSUBMEM(10,%r8,Chimu_11,Chi_11)		\
-  VSUBMEM(11,%r8,Chimu_12,Chi_12)		);
+#define TM_PROJMEM(ptr)						\
+  LOAD64(%r8,ptr)						\
+  __asm__ (							\
+	   LOAD_CHIMU01i					\
+	   VSUBMEM(6,%r8,Chimu_00,Chi_00)			\
+	   VSUBMEM(7,%r8,Chimu_01,Chi_01)			\
+	   VSUBMEM(8,%r8,Chimu_02,Chi_02)			\
+	   VSUBMEM(9,%r8,Chimu_10,Chi_10)			\
+	   VSUBMEM(10,%r8,Chimu_11,Chi_11)			\
+	   VSUBMEM(11,%r8,Chimu_12,Chi_12)		);
 
 //      fspin(0)=hspin(0)
 //      fspin(1)=hspin(1)
 //      fspin(2)=timesMinusI(hspin(1))
 //      fspin(3)=timesMinusI(hspin(0))
-#define XP_RECON __asm__ (			\
-			  VZERO(TMP)		\
+#define XP_RECON __asm__ (					\
+			  VZERO(TMP)				\
 			  VTIMESMINUSI0(UChi_00,psi_30,TMP)	\
 			  VTIMESMINUSI0(UChi_10,psi_20,TMP)	\
 			  VTIMESMINUSI0(UChi_01,psi_31,TMP)	\
 			  VTIMESMINUSI0(UChi_11,psi_21,TMP)	\
-			  VTIMESMINUSI0(UChi_02,psi_32,TMP)   \
+			  VTIMESMINUSI0(UChi_02,psi_32,TMP)	\
 			  VTIMESMINUSI0(UChi_12,psi_22,TMP)	\
-			  VMOV(UChi_00,psi_00)	\
-			  VMOV(UChi_10,psi_10)	\
-			  VMOV(UChi_01,psi_01)	\
-			  VMOV(UChi_11,psi_11)	\
-			  VMOV(UChi_02,psi_02)	\
-			  VMOV(UChi_12,psi_12)	\
+			  VMOV(UChi_00,psi_00)			\
+			  VMOV(UChi_10,psi_10)			\
+			  VMOV(UChi_01,psi_01)			\
+			  VMOV(UChi_11,psi_11)			\
+			  VMOV(UChi_02,psi_02)			\
+			  VMOV(UChi_12,psi_12)			\
 			  VTIMESMINUSI1(UChi_10,psi_20,TMP)	\
 			  VTIMESMINUSI1(UChi_11,psi_21,TMP)	\
 			  VTIMESMINUSI1(UChi_12,psi_22,TMP)	\
 			  VTIMESMINUSI1(UChi_00,psi_30,TMP)	\
 			  VTIMESMINUSI1(UChi_01,psi_31,TMP)	\
-			  VTIMESMINUSI1(UChi_02,psi_32,TMP)   \
+			  VTIMESMINUSI1(UChi_02,psi_32,TMP)	\
 			  VTIMESMINUSI2(UChi_10,psi_20,TMP)	\
 			  VTIMESMINUSI2(UChi_11,psi_21,TMP)	\
 			  VTIMESMINUSI2(UChi_12,psi_22,TMP)	\
 			  VTIMESMINUSI2(UChi_00,psi_30,TMP)	\
 			  VTIMESMINUSI2(UChi_01,psi_31,TMP)	\
-			  VTIMESMINUSI2(UChi_02,psi_32,TMP)   \
+			  VTIMESMINUSI2(UChi_02,psi_32,TMP)	\
 						);
-  // NB could save 6 ops using addsub => 12 cycles
-#define XP_RECON_ACCUM __asm__ ( \
-  VZERO(TMP)\
-  VACCTIMESMINUSI0(UChi_00,psi_30,Z3)\
-  VACCTIMESMINUSI0(UChi_10,psi_20,Z0)\
-  VACCTIMESMINUSI0(UChi_01,psi_31,Z4)\
-  VACCTIMESMINUSI0(UChi_11,psi_21,Z1)\
-  VACCTIMESMINUSI0(UChi_02,psi_32,Z5)\
-  VACCTIMESMINUSI0(UChi_12,psi_22,Z2)\
-  VADD(UChi_00,psi_00,psi_00)\
-  VADD(UChi_10,psi_10,psi_10)\
-  VADD(UChi_01,psi_01,psi_01)\
-  VADD(UChi_11,psi_11,psi_11)\
-  VADD(UChi_02,psi_02,psi_02)\
-  VADD(UChi_12,psi_12,psi_12)\
-  VACCTIMESMINUSI1(UChi_00,psi_30,Z3)\
-  VACCTIMESMINUSI1(UChi_10,psi_20,Z0)\
-  VACCTIMESMINUSI1(UChi_01,psi_31,Z4)\
-  VACCTIMESMINUSI1(UChi_11,psi_21,Z1)\
-  VACCTIMESMINUSI1(UChi_02,psi_32,Z5)\
-  VACCTIMESMINUSI1(UChi_12,psi_22,Z2)\
-  VACCTIMESMINUSI2(UChi_10,psi_20,Z0)\
-  VACCTIMESMINUSI2(UChi_11,psi_21,Z1)\
-  VACCTIMESMINUSI2(UChi_12,psi_22,Z2)\
-  VACCTIMESMINUSI2(UChi_00,psi_30,Z3)\
-  VACCTIMESMINUSI2(UChi_01,psi_31,Z4)\
-  VACCTIMESMINUSI2(UChi_02,psi_32,Z5)\
+// NB could save 6 ops using addsub => 12 cycles
+#define XP_RECON_ACCUM __asm__ (					\
+				VZERO(TMP)				\
+				VACCTIMESMINUSI0(UChi_00,psi_30,Z3)	\
+				VACCTIMESMINUSI0(UChi_10,psi_20,Z0)	\
+				VACCTIMESMINUSI0(UChi_01,psi_31,Z4)	\
+				VACCTIMESMINUSI0(UChi_11,psi_21,Z1)	\
+				VACCTIMESMINUSI0(UChi_02,psi_32,Z5)	\
+				VACCTIMESMINUSI0(UChi_12,psi_22,Z2)	\
+				VADD(UChi_00,psi_00,psi_00)		\
+				VADD(UChi_10,psi_10,psi_10)		\
+				VADD(UChi_01,psi_01,psi_01)		\
+				VADD(UChi_11,psi_11,psi_11)		\
+				VADD(UChi_02,psi_02,psi_02)		\
+				VADD(UChi_12,psi_12,psi_12)		\
+				VACCTIMESMINUSI1(UChi_00,psi_30,Z3)	\
+				VACCTIMESMINUSI1(UChi_10,psi_20,Z0)	\
+				VACCTIMESMINUSI1(UChi_01,psi_31,Z4)	\
+				VACCTIMESMINUSI1(UChi_11,psi_21,Z1)	\
+				VACCTIMESMINUSI1(UChi_02,psi_32,Z5)	\
+				VACCTIMESMINUSI1(UChi_12,psi_22,Z2)	\
+				VACCTIMESMINUSI2(UChi_10,psi_20,Z0)	\
+				VACCTIMESMINUSI2(UChi_11,psi_21,Z1)	\
+				VACCTIMESMINUSI2(UChi_12,psi_22,Z2)	\
+				VACCTIMESMINUSI2(UChi_00,psi_30,Z3)	\
+				VACCTIMESMINUSI2(UChi_01,psi_31,Z4)	\
+				VACCTIMESMINUSI2(UChi_02,psi_32,Z5)	\
 				 );
 
-#define XM_RECON __asm__ ( \
-  VZERO(TMP)\
-  VTIMESI0(UChi_00,psi_30,TMP)\
-  VTIMESI0(UChi_10,psi_20,TMP)\
-  VTIMESI0(UChi_01,psi_31,TMP)\
-  VTIMESI0(UChi_11,psi_21,TMP)\
-  VTIMESI0(UChi_02,psi_32,TMP)\
-  VTIMESI0(UChi_12,psi_22,TMP)\
-  VMOV(UChi_00,psi_00)\
-  VMOV(UChi_10,psi_10)\
-  VMOV(UChi_01,psi_01)\
-  VMOV(UChi_11,psi_11)\
-  VMOV(UChi_02,psi_02)\
-  VMOV(UChi_12,psi_12)\
-  VTIMESI1(UChi_00,psi_30,TMP)\
-  VTIMESI1(UChi_10,psi_20,TMP)\
-  VTIMESI1(UChi_01,psi_31,TMP)\
-  VTIMESI1(UChi_11,psi_21,TMP)\
-  VTIMESI1(UChi_02,psi_32,TMP)\
-  VTIMESI1(UChi_12,psi_22,TMP)\
-  VTIMESI2(UChi_10,psi_20,TMP)\
-  VTIMESI2(UChi_11,psi_21,TMP)\
-  VTIMESI2(UChi_12,psi_22,TMP)\
-  VTIMESI2(UChi_00,psi_30,TMP)\
-  VTIMESI2(UChi_01,psi_31,TMP)\
-  VTIMESI2(UChi_02,psi_32,TMP)\
+#define XM_RECON __asm__ (				\
+			  VZERO(TMP)			\
+			  VTIMESI0(UChi_00,psi_30,TMP)	\
+			  VTIMESI0(UChi_10,psi_20,TMP)	\
+			  VTIMESI0(UChi_01,psi_31,TMP)	\
+			  VTIMESI0(UChi_11,psi_21,TMP)	\
+			  VTIMESI0(UChi_02,psi_32,TMP)	\
+			  VTIMESI0(UChi_12,psi_22,TMP)	\
+			  VMOV(UChi_00,psi_00)		\
+			  VMOV(UChi_10,psi_10)		\
+			  VMOV(UChi_01,psi_01)		\
+			  VMOV(UChi_11,psi_11)		\
+			  VMOV(UChi_02,psi_02)		\
+			  VMOV(UChi_12,psi_12)		\
+			  VTIMESI1(UChi_00,psi_30,TMP)	\
+			  VTIMESI1(UChi_10,psi_20,TMP)	\
+			  VTIMESI1(UChi_01,psi_31,TMP)	\
+			  VTIMESI1(UChi_11,psi_21,TMP)	\
+			  VTIMESI1(UChi_02,psi_32,TMP)	\
+			  VTIMESI1(UChi_12,psi_22,TMP)	\
+			  VTIMESI2(UChi_10,psi_20,TMP)	\
+			  VTIMESI2(UChi_11,psi_21,TMP)	\
+			  VTIMESI2(UChi_12,psi_22,TMP)	\
+			  VTIMESI2(UChi_00,psi_30,TMP)	\
+			  VTIMESI2(UChi_01,psi_31,TMP)	\
+			  VTIMESI2(UChi_02,psi_32,TMP)	\
 			   );
 
-#define XM_RECON_ACCUM __asm__ ( \
-  VACCTIMESI0(UChi_10,psi_20,Z0)\
-  VACCTIMESI0(UChi_00,psi_30,Z3)\
-  VACCTIMESI0(UChi_11,psi_21,Z1)\
-  VACCTIMESI0(UChi_01,psi_31,Z4)\
-  VACCTIMESI0(UChi_12,psi_22,Z2)\
-  VACCTIMESI0(UChi_02,psi_32,Z5)\
-  \
-  VADD(UChi_10,psi_10,psi_10)\
-  VADD(UChi_00,psi_00,psi_00)\
-  VADD(UChi_11,psi_11,psi_11)\
-  VADD(UChi_01,psi_01,psi_01)\
-  VADD(UChi_12,psi_12,psi_12)\
-  VADD(UChi_02,psi_02,psi_02)\
-  \
-  VACCTIMESI1(UChi_10,psi_20,Z0)\
-  VACCTIMESI1(UChi_00,psi_30,Z3)\
-  VACCTIMESI1(UChi_11,psi_21,Z1)\
-  VACCTIMESI1(UChi_01,psi_31,Z4)\
-  VACCTIMESI1(UChi_12,psi_22,Z2)\
-  VACCTIMESI1(UChi_02,psi_32,Z5)\
-  VACCTIMESI2(UChi_10,psi_20,Z0)\
-  VACCTIMESI2(UChi_11,psi_21,Z1)\
-  VACCTIMESI2(UChi_12,psi_22,Z2)\
-  VACCTIMESI2(UChi_00,psi_30,Z3)\
-  VACCTIMESI2(UChi_01,psi_31,Z4)\
-  VACCTIMESI2(UChi_02,psi_32,Z5)\
+#define XM_RECON_ACCUM __asm__ (				\
+				VACCTIMESI0(UChi_10,psi_20,Z0)	\
+				VACCTIMESI0(UChi_00,psi_30,Z3)	\
+				VACCTIMESI0(UChi_11,psi_21,Z1)	\
+				VACCTIMESI0(UChi_01,psi_31,Z4)	\
+				VACCTIMESI0(UChi_12,psi_22,Z2)	\
+				VACCTIMESI0(UChi_02,psi_32,Z5)	\
+								\
+				VADD(UChi_10,psi_10,psi_10)	\
+				VADD(UChi_00,psi_00,psi_00)	\
+				VADD(UChi_11,psi_11,psi_11)	\
+				VADD(UChi_01,psi_01,psi_01)	\
+				VADD(UChi_12,psi_12,psi_12)	\
+				VADD(UChi_02,psi_02,psi_02)	\
+								\
+				VACCTIMESI1(UChi_10,psi_20,Z0)	\
+				VACCTIMESI1(UChi_00,psi_30,Z3)	\
+				VACCTIMESI1(UChi_11,psi_21,Z1)	\
+				VACCTIMESI1(UChi_01,psi_31,Z4)	\
+				VACCTIMESI1(UChi_12,psi_22,Z2)	\
+				VACCTIMESI1(UChi_02,psi_32,Z5)	\
+				VACCTIMESI2(UChi_10,psi_20,Z0)	\
+				VACCTIMESI2(UChi_11,psi_21,Z1)	\
+				VACCTIMESI2(UChi_12,psi_22,Z2)	\
+				VACCTIMESI2(UChi_00,psi_30,Z3)	\
+				VACCTIMESI2(UChi_01,psi_31,Z4)	\
+				VACCTIMESI2(UChi_02,psi_32,Z5)	\
 				 );
 
-#define YP_RECON_ACCUM __asm__ ( \
-  VADD(UChi_00,psi_00,psi_00)\
-  VADD(UChi_10,psi_10,psi_10)\
-  VADD(UChi_01,psi_01,psi_01)\
-  VADD(UChi_11,psi_11,psi_11)\
-  VADD(UChi_02,psi_02,psi_02)\
-  VADD(UChi_12,psi_12,psi_12)\
-  VADD(UChi_10,psi_20,psi_20)\
-  VADD(UChi_11,psi_21,psi_21)\
-  VADD(UChi_12,psi_22,psi_22)\
-  VSUB(UChi_00,psi_30,psi_30)\
-  VSUB(UChi_01,psi_31,psi_31)\
-  VSUB(UChi_02,psi_32,psi_32) );
+#define YP_RECON_ACCUM __asm__ (				\
+				VADD(UChi_00,psi_00,psi_00)	\
+				VADD(UChi_10,psi_10,psi_10)	\
+				VADD(UChi_01,psi_01,psi_01)	\
+				VADD(UChi_11,psi_11,psi_11)	\
+				VADD(UChi_02,psi_02,psi_02)	\
+				VADD(UChi_12,psi_12,psi_12)	\
+				VADD(UChi_10,psi_20,psi_20)	\
+				VADD(UChi_11,psi_21,psi_21)	\
+				VADD(UChi_12,psi_22,psi_22)	\
+				VSUB(UChi_00,psi_30,psi_30)	\
+				VSUB(UChi_01,psi_31,psi_31)	\
+				VSUB(UChi_02,psi_32,psi_32) );
 
-#define YM_RECON_ACCUM __asm__ ( \
-  VADD(UChi_00,psi_00,psi_00)\
-  VADD(UChi_10,psi_10,psi_10)\
-  VADD(UChi_01,psi_01,psi_01)\
-  VADD(UChi_11,psi_11,psi_11)\
-  VADD(UChi_02,psi_02,psi_02)\
-  VADD(UChi_12,psi_12,psi_12)\
-  VSUB(UChi_10,psi_20,psi_20)\
-  VSUB(UChi_11,psi_21,psi_21)\
-  VSUB(UChi_12,psi_22,psi_22)\
-  VADD(UChi_00,psi_30,psi_30)\
-  VADD(UChi_01,psi_31,psi_31)\
-  VADD(UChi_02,psi_32,psi_32) );
+#define YM_RECON_ACCUM __asm__ (				\
+				VADD(UChi_00,psi_00,psi_00)	\
+				VADD(UChi_10,psi_10,psi_10)	\
+				VADD(UChi_01,psi_01,psi_01)	\
+				VADD(UChi_11,psi_11,psi_11)	\
+				VADD(UChi_02,psi_02,psi_02)	\
+				VADD(UChi_12,psi_12,psi_12)	\
+				VSUB(UChi_10,psi_20,psi_20)	\
+				VSUB(UChi_11,psi_21,psi_21)	\
+				VSUB(UChi_12,psi_22,psi_22)	\
+				VADD(UChi_00,psi_30,psi_30)	\
+				VADD(UChi_01,psi_31,psi_31)	\
+				VADD(UChi_02,psi_32,psi_32) );
 
-#define ZP_RECON_ACCUM __asm__ ( \
-  VACCTIMESMINUSI0(UChi_00,psi_20,Z0)\
-  VACCTIMESI0(UChi_10,psi_30,Z3)\
-  VACCTIMESMINUSI0(UChi_01,psi_21,Z1)\
-  VACCTIMESI0(UChi_11,psi_31,Z4)\
-  VACCTIMESMINUSI0(UChi_02,psi_22,Z2)\
-  VACCTIMESI0(UChi_12,psi_32,Z5)\
-  VADD(UChi_00,psi_00,psi_00)\
-  VADD(UChi_10,psi_10,psi_10)\
-  VADD(UChi_01,psi_01,psi_01)\
-  VADD(UChi_11,psi_11,psi_11)\
-  VADD(UChi_02,psi_02,psi_02)\
-  VADD(UChi_12,psi_12,psi_12)\
-  VACCTIMESMINUSI1(UChi_00,psi_20,Z0)\
-  VACCTIMESI1(UChi_10,psi_30,Z3)\
-  VACCTIMESMINUSI1(UChi_01,psi_21,Z1)\
-  VACCTIMESI1(UChi_11,psi_31,Z4)\
-  VACCTIMESMINUSI1(UChi_02,psi_22,Z2)\
-  VACCTIMESI1(UChi_12,psi_32,Z5)\
-  VACCTIMESMINUSI2(UChi_00,psi_20,Z0)\
-  VACCTIMESMINUSI2(UChi_01,psi_21,Z1)\
-  VACCTIMESMINUSI2(UChi_02,psi_22,Z2)\
-  VACCTIMESI2(UChi_10,psi_30,Z3)\
-  VACCTIMESI2(UChi_11,psi_31,Z4)\
-  VACCTIMESI2(UChi_12,psi_32,Z5)\
+#define ZP_RECON_ACCUM __asm__ (					\
+				VACCTIMESMINUSI0(UChi_00,psi_20,Z0)	\
+				VACCTIMESI0(UChi_10,psi_30,Z3)		\
+				VACCTIMESMINUSI0(UChi_01,psi_21,Z1)	\
+				VACCTIMESI0(UChi_11,psi_31,Z4)		\
+				VACCTIMESMINUSI0(UChi_02,psi_22,Z2)	\
+				VACCTIMESI0(UChi_12,psi_32,Z5)		\
+				VADD(UChi_00,psi_00,psi_00)		\
+				VADD(UChi_10,psi_10,psi_10)		\
+				VADD(UChi_01,psi_01,psi_01)		\
+				VADD(UChi_11,psi_11,psi_11)		\
+				VADD(UChi_02,psi_02,psi_02)		\
+				VADD(UChi_12,psi_12,psi_12)		\
+				VACCTIMESMINUSI1(UChi_00,psi_20,Z0)	\
+				VACCTIMESI1(UChi_10,psi_30,Z3)		\
+				VACCTIMESMINUSI1(UChi_01,psi_21,Z1)	\
+				VACCTIMESI1(UChi_11,psi_31,Z4)		\
+				VACCTIMESMINUSI1(UChi_02,psi_22,Z2)	\
+				VACCTIMESI1(UChi_12,psi_32,Z5)		\
+				VACCTIMESMINUSI2(UChi_00,psi_20,Z0)	\
+				VACCTIMESMINUSI2(UChi_01,psi_21,Z1)	\
+				VACCTIMESMINUSI2(UChi_02,psi_22,Z2)	\
+				VACCTIMESI2(UChi_10,psi_30,Z3)		\
+				VACCTIMESI2(UChi_11,psi_31,Z4)		\
+				VACCTIMESI2(UChi_12,psi_32,Z5)		\
 				 );
 
-#define ZM_RECON_ACCUM __asm__ ( \
-  VACCTIMESI0(UChi_00,psi_20,Z0)\
-  VACCTIMESMINUSI0(UChi_10,psi_30,Z3)\
-  VACCTIMESI0(UChi_01,psi_21,Z1)\
-  VACCTIMESMINUSI0(UChi_11,psi_31,Z4)\
-  VACCTIMESI0(UChi_02,psi_22,Z2)\
-  VACCTIMESMINUSI0(UChi_12,psi_32,Z5)\
-  VADD(UChi_00,psi_00,psi_00)\
-  VADD(UChi_10,psi_10,psi_10)\
-  VADD(UChi_01,psi_01,psi_01)\
-  VADD(UChi_11,psi_11,psi_11)\
-  VADD(UChi_02,psi_02,psi_02)\
-  VADD(UChi_12,psi_12,psi_12)\
-  VACCTIMESI1(UChi_00,psi_20,Z0)\
-  VACCTIMESMINUSI1(UChi_10,psi_30,Z3)\
-  VACCTIMESI1(UChi_01,psi_21,Z1)\
-  VACCTIMESMINUSI1(UChi_11,psi_31,Z4)\
-  VACCTIMESI1(UChi_02,psi_22,Z2)\
-  VACCTIMESMINUSI1(UChi_12,psi_32,Z5)\
-  VACCTIMESI2(UChi_00,psi_20,Z0)\
-  VACCTIMESI2(UChi_01,psi_21,Z1)\
-  VACCTIMESI2(UChi_02,psi_22,Z2)\
-  VACCTIMESMINUSI2(UChi_10,psi_30,Z3)\
-  VACCTIMESMINUSI2(UChi_11,psi_31,Z4)\
-  VACCTIMESMINUSI2(UChi_12,psi_32,Z5)\
+#define ZM_RECON_ACCUM __asm__ (					\
+				VACCTIMESI0(UChi_00,psi_20,Z0)		\
+				VACCTIMESMINUSI0(UChi_10,psi_30,Z3)	\
+				VACCTIMESI0(UChi_01,psi_21,Z1)		\
+				VACCTIMESMINUSI0(UChi_11,psi_31,Z4)	\
+				VACCTIMESI0(UChi_02,psi_22,Z2)		\
+				VACCTIMESMINUSI0(UChi_12,psi_32,Z5)	\
+				VADD(UChi_00,psi_00,psi_00)		\
+				VADD(UChi_10,psi_10,psi_10)		\
+				VADD(UChi_01,psi_01,psi_01)		\
+				VADD(UChi_11,psi_11,psi_11)		\
+				VADD(UChi_02,psi_02,psi_02)		\
+				VADD(UChi_12,psi_12,psi_12)		\
+				VACCTIMESI1(UChi_00,psi_20,Z0)		\
+				VACCTIMESMINUSI1(UChi_10,psi_30,Z3)	\
+				VACCTIMESI1(UChi_01,psi_21,Z1)		\
+				VACCTIMESMINUSI1(UChi_11,psi_31,Z4)	\
+				VACCTIMESI1(UChi_02,psi_22,Z2)		\
+				VACCTIMESMINUSI1(UChi_12,psi_32,Z5)	\
+				VACCTIMESI2(UChi_00,psi_20,Z0)		\
+				VACCTIMESI2(UChi_01,psi_21,Z1)		\
+				VACCTIMESI2(UChi_02,psi_22,Z2)		\
+				VACCTIMESMINUSI2(UChi_10,psi_30,Z3)	\
+				VACCTIMESMINUSI2(UChi_11,psi_31,Z4)	\
+				VACCTIMESMINUSI2(UChi_12,psi_32,Z5)	\
 				 );
 
-#define TP_RECON_ACCUM __asm__ ( \
-  VADD(UChi_00,psi_00,psi_00)\
-  VADD(UChi_10,psi_10,psi_10)\
-  VADD(UChi_01,psi_01,psi_01)\
-  VADD(UChi_11,psi_11,psi_11)\
-  VADD(UChi_02,psi_02,psi_02)\
-  VADD(UChi_12,psi_12,psi_12)\
-  VADD(UChi_00,psi_20,psi_20)\
-  VADD(UChi_10,psi_30,psi_30)\
-  VADD(UChi_01,psi_21,psi_21)\
-  VADD(UChi_11,psi_31,psi_31)\
-  VADD(UChi_02,psi_22,psi_22)\
-  VADD(UChi_12,psi_32,psi_32) );
+#define TP_RECON_ACCUM __asm__ (				\
+				VADD(UChi_00,psi_00,psi_00)	\
+				VADD(UChi_10,psi_10,psi_10)	\
+				VADD(UChi_01,psi_01,psi_01)	\
+				VADD(UChi_11,psi_11,psi_11)	\
+				VADD(UChi_02,psi_02,psi_02)	\
+				VADD(UChi_12,psi_12,psi_12)	\
+				VADD(UChi_00,psi_20,psi_20)	\
+				VADD(UChi_10,psi_30,psi_30)	\
+				VADD(UChi_01,psi_21,psi_21)	\
+				VADD(UChi_11,psi_31,psi_31)	\
+				VADD(UChi_02,psi_22,psi_22)	\
+				VADD(UChi_12,psi_32,psi_32) );
 
-#define TM_RECON_ACCUM __asm__ ( \
-  VADD(UChi_00,psi_00,psi_00)\
-  VADD(UChi_10,psi_10,psi_10)\
-  VADD(UChi_01,psi_01,psi_01)\
-  VADD(UChi_11,psi_11,psi_11)\
-  VADD(UChi_02,psi_02,psi_02)\
-  VADD(UChi_12,psi_12,psi_12)\
-  VSUB(UChi_00,psi_20,psi_20)\
-  VSUB(UChi_10,psi_30,psi_30)\
-  VSUB(UChi_01,psi_21,psi_21)\
-  VSUB(UChi_11,psi_31,psi_31)\
-  VSUB(UChi_02,psi_22,psi_22)\
-  VSUB(UChi_12,psi_32,psi_32) );
+#define TM_RECON_ACCUM __asm__ (				\
+				VADD(UChi_00,psi_00,psi_00)	\
+				VADD(UChi_10,psi_10,psi_10)	\
+				VADD(UChi_01,psi_01,psi_01)	\
+				VADD(UChi_11,psi_11,psi_11)	\
+				VADD(UChi_02,psi_02,psi_02)	\
+				VADD(UChi_12,psi_12,psi_12)	\
+				VSUB(UChi_00,psi_20,psi_20)	\
+				VSUB(UChi_10,psi_30,psi_30)	\
+				VSUB(UChi_01,psi_21,psi_21)	\
+				VSUB(UChi_11,psi_31,psi_31)	\
+				VSUB(UChi_02,psi_22,psi_22)	\
+				VSUB(UChi_12,psi_32,psi_32) );
 
 #define AVX512_PF_L1
 #define AVX512_PF_L2_GAUGE
@@ -560,27 +560,27 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define VPREFETCH_G2(A,B)  VPREFETCH2(A,B)
 #endif
 
-#define PF_GAUGE(A) \
-  LOAD64(%r8,&U._odata[sU](A))						\
+#define PF_GAUGE(A)							\
+  LOAD64(%r8,&U[sU](A))						\
   __asm__ (								\
 	   VPREFETCH_G1(0,%r8) VPREFETCH_G1(1,%r8)			\
 	   VPREFETCH_G1(2,%r8) VPREFETCH_G1(3,%r8)			\
 									);
 
-#define SAVE_RESULTi(PTR,pf)			\
-	   LOAD64(%r8,PTR)			\
-	   LOAD64(%r9,pf)			\
-  __asm__ (					\
-	   VSTORE(0,%r8,psi_00)	VPREFETCH_M1(0,%r9)	\
-	   VSTORE(1,%r8,psi_01)	VPREFETCH_M1(1,%r9)	\
-	   VSTORE(2,%r8,psi_02)	VPREFETCH_M1(2,%r9)	\
-	   VSTORE(3,%r8,psi_10)	VPREFETCH_M1(3,%r9)	\
-	   VSTORE(4,%r8,psi_11)	VPREFETCH_M1(4,%r9)	\
-	   VSTORE(5,%r8,psi_12)	VPREFETCH_M1(5,%r9)	\
-	   VSTORE(6,%r8,psi_20)	VPREFETCH_M1(6,%r9)	\
-	   VSTORE(7,%r8,psi_21)	VPREFETCH_M1(7,%r9)	\
-	   VSTORE(8,%r8,psi_22)	VPREFETCH_M1(8,%r9)	\
-	   VSTORE(9,%r8,psi_30)	VPREFETCH_M1(9,%r9)	\
+#define SAVE_RESULTi(PTR,pf)					\
+  LOAD64(%r8,PTR)						\
+  LOAD64(%r9,pf)						\
+  __asm__ (							\
+	   VSTORE(0,%r8,psi_00)	VPREFETCH_M1(0,%r9)		\
+	   VSTORE(1,%r8,psi_01)	VPREFETCH_M1(1,%r9)		\
+	   VSTORE(2,%r8,psi_02)	VPREFETCH_M1(2,%r9)		\
+	   VSTORE(3,%r8,psi_10)	VPREFETCH_M1(3,%r9)		\
+	   VSTORE(4,%r8,psi_11)	VPREFETCH_M1(4,%r9)		\
+	   VSTORE(5,%r8,psi_12)	VPREFETCH_M1(5,%r9)		\
+	   VSTORE(6,%r8,psi_20)	VPREFETCH_M1(6,%r9)		\
+	   VSTORE(7,%r8,psi_21)	VPREFETCH_M1(7,%r9)		\
+	   VSTORE(8,%r8,psi_22)	VPREFETCH_M1(8,%r9)		\
+	   VSTORE(9,%r8,psi_30)	VPREFETCH_M1(9,%r9)		\
 	   VSTORE(10,%r8,psi_31)	VPREFETCH_M1(10,%r9)	\
 	   VSTORE(11,%r8,psi_32) 	VPREFETCH_M1(11,%r9)	\
 						);
@@ -596,343 +596,343 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 
 #define ADD_RESULTia(PTR,pf)                                            \
-  LOAD64(%r8,PTR)                                                      \
+  LOAD64(%r8,PTR)							\
   __asm__ (								\
-	   VADDMEM(0,%r8,psi_00,psi_00)				\
-	   VADDMEM(1,%r8,psi_01,psi_01)				\
-	   VADDMEM(2,%r8,psi_02,psi_02)                           \
-	   VADDMEM(3,%r8,psi_10,psi_10)                           \
-	   VADDMEM(4,%r8,psi_11,psi_11)                           \
-	   VADDMEM(5,%r8,psi_12,psi_12)                           \
-	   VADDMEM(6,%r8,psi_20,psi_20)                           \
-	   VADDMEM(7,%r8,psi_21,psi_21)                           \
-	   VADDMEM(8,%r8,psi_22,psi_22)                           \
-	   VADDMEM(9,%r8,psi_30,psi_30)                           \
-	   VADDMEM(10,%r8,psi_31,psi_31)                           \
+	   VADDMEM(0,%r8,psi_00,psi_00)					\
+	   VADDMEM(1,%r8,psi_01,psi_01)					\
+	   VADDMEM(2,%r8,psi_02,psi_02)					\
+	   VADDMEM(3,%r8,psi_10,psi_10)					\
+	   VADDMEM(4,%r8,psi_11,psi_11)					\
+	   VADDMEM(5,%r8,psi_12,psi_12)					\
+	   VADDMEM(6,%r8,psi_20,psi_20)					\
+	   VADDMEM(7,%r8,psi_21,psi_21)					\
+	   VADDMEM(8,%r8,psi_22,psi_22)					\
+	   VADDMEM(9,%r8,psi_30,psi_30)					\
+	   VADDMEM(10,%r8,psi_31,psi_31)				\
 	   VADDMEM(11,%r8,psi_32,psi_32)				\
-	   VSTORE(0,%r8,psi_00)                                      \
-	   VSTORE(1,%r8,psi_01)                                      \
-	   VSTORE(2,%r8,psi_02)                                      \
-	   VSTORE(3,%r8,psi_10)                                      \
-	   VSTORE(4,%r8,psi_11)                                      \
-	   VSTORE(5,%r8,psi_12)                                      \
-	   VSTORE(6,%r8,psi_20)                                      \
-	   VSTORE(7,%r8,psi_21)                                      \
-	   VSTORE(8,%r8,psi_22)                                      \
-	   VSTORE(9,%r8,psi_30)                                      \
-	   VSTORE(10,%r8,psi_31)                                     \
-	   VSTORE(11,%r8,psi_32)                                     \
+	   VSTORE(0,%r8,psi_00)						\
+	   VSTORE(1,%r8,psi_01)						\
+	   VSTORE(2,%r8,psi_02)						\
+	   VSTORE(3,%r8,psi_10)						\
+	   VSTORE(4,%r8,psi_11)						\
+	   VSTORE(5,%r8,psi_12)						\
+	   VSTORE(6,%r8,psi_20)						\
+	   VSTORE(7,%r8,psi_21)						\
+	   VSTORE(8,%r8,psi_22)						\
+	   VSTORE(9,%r8,psi_30)						\
+	   VSTORE(10,%r8,psi_31)					\
+	   VSTORE(11,%r8,psi_32)					\
 									);
 
 
 #ifdef AVX512_PF_L2_TABLE
-#define PREFETCH_CHIMU(A)					\
-  LOAD64(%r9,A)							\
-  __asm__ (							\
-	   VPREFETCH_P1(0,%r9)					\
-	   VPREFETCH_P1(1,%r9)					\
-	   VPREFETCH_P1(2,%r9)					\
-	   VPREFETCH_P1(3,%r9)					\
-	   VPREFETCH_P1(4,%r9)					\
-	   VPREFETCH_P1(5,%r9)					\
-	   VPREFETCH_P1(6,%r9)					\
-	   VPREFETCH_P1(7,%r9)					\
-	   VPREFETCH_P1(8,%r9)					\
-	   VPREFETCH_P1(9,%r9)					\
-	   VPREFETCH_P1(10,%r9)					\
+#define PREFETCH_CHIMU(A)			\
+  LOAD64(%r9,A)					\
+  __asm__ (					\
+	   VPREFETCH_P1(0,%r9)			\
+	   VPREFETCH_P1(1,%r9)			\
+	   VPREFETCH_P1(2,%r9)			\
+	   VPREFETCH_P1(3,%r9)			\
+	   VPREFETCH_P1(4,%r9)			\
+	   VPREFETCH_P1(5,%r9)			\
+	   VPREFETCH_P1(6,%r9)			\
+	   VPREFETCH_P1(7,%r9)			\
+	   VPREFETCH_P1(8,%r9)			\
+	   VPREFETCH_P1(9,%r9)			\
+	   VPREFETCH_P1(10,%r9)			\
 	   VPREFETCH_P1(11,%r9));
 
 #else
 #define PREFETCH_CHIMU(A)
 #endif
 
-#define PREFETCH1_CHIMU(A) \
-  LOAD64(%r9,A)							\
-  __asm__ (							\
-	   VPREFETCH_P1(0,%r9)					\
-	   VPREFETCH_P1(1,%r9)					\
-	   VPREFETCH_P1(2,%r9)					\
-	   VPREFETCH_P1(3,%r9)					\
-	   VPREFETCH_P1(4,%r9)					\
-	   VPREFETCH_P1(5,%r9)					\
-	   VPREFETCH_P1(6,%r9)					\
-	   VPREFETCH_P1(7,%r9)					\
-	   VPREFETCH_P1(8,%r9)					\
-	   VPREFETCH_P1(9,%r9)					\
-	   VPREFETCH_P1(10,%r9)					\
+#define PREFETCH1_CHIMU(A)			\
+  LOAD64(%r9,A)					\
+  __asm__ (					\
+	   VPREFETCH_P1(0,%r9)			\
+	   VPREFETCH_P1(1,%r9)			\
+	   VPREFETCH_P1(2,%r9)			\
+	   VPREFETCH_P1(3,%r9)			\
+	   VPREFETCH_P1(4,%r9)			\
+	   VPREFETCH_P1(5,%r9)			\
+	   VPREFETCH_P1(6,%r9)			\
+	   VPREFETCH_P1(7,%r9)			\
+	   VPREFETCH_P1(8,%r9)			\
+	   VPREFETCH_P1(9,%r9)			\
+	   VPREFETCH_P1(10,%r9)			\
 	   VPREFETCH_P1(11,%r9));
 
-#define PERMUTE_DIR0 __asm__ (			\
-  VPERM0(Chi_00,Chi_00)	\
-  VPERM0(Chi_01,Chi_01)	\
-  VPERM0(Chi_02,Chi_02)	\
-  VPERM0(Chi_10,Chi_10)	\
-  VPERM0(Chi_11,Chi_11)	\
-  VPERM0(Chi_12,Chi_12) );
+#define PERMUTE_DIR0 __asm__ (				\
+			      VPERM0(Chi_00,Chi_00)	\
+			      VPERM0(Chi_01,Chi_01)	\
+			      VPERM0(Chi_02,Chi_02)	\
+			      VPERM0(Chi_10,Chi_10)	\
+			      VPERM0(Chi_11,Chi_11)	\
+			      VPERM0(Chi_12,Chi_12) );
 
-#define PERMUTE_DIR1 __asm__ (	\
-  VPERM1(Chi_00,Chi_00)	\
-  VPERM1(Chi_01,Chi_01)	\
-  VPERM1(Chi_02,Chi_02)	\
-  VPERM1(Chi_10,Chi_10)	\
-  VPERM1(Chi_11,Chi_11)	\
-  VPERM1(Chi_12,Chi_12));
+#define PERMUTE_DIR1 __asm__ (				\
+			      VPERM1(Chi_00,Chi_00)	\
+			      VPERM1(Chi_01,Chi_01)	\
+			      VPERM1(Chi_02,Chi_02)	\
+			      VPERM1(Chi_10,Chi_10)	\
+			      VPERM1(Chi_11,Chi_11)	\
+			      VPERM1(Chi_12,Chi_12));
 
-#define PERMUTE_DIR2 __asm__ (	\
-  VPERM2(Chi_00,Chi_00)	\
-  VPERM2(Chi_01,Chi_01)	\
-  VPERM2(Chi_02,Chi_02)	\
-  VPERM2(Chi_10,Chi_10)	\
-  VPERM2(Chi_11,Chi_11)	\
-  VPERM2(Chi_12,Chi_12) );
+#define PERMUTE_DIR2 __asm__ (				\
+			      VPERM2(Chi_00,Chi_00)	\
+			      VPERM2(Chi_01,Chi_01)	\
+			      VPERM2(Chi_02,Chi_02)	\
+			      VPERM2(Chi_10,Chi_10)	\
+			      VPERM2(Chi_11,Chi_11)	\
+			      VPERM2(Chi_12,Chi_12) );
 
-#define PERMUTE_DIR3 __asm__ (	\
-  VPERM3(Chi_00,Chi_00)	\
-  VPERM3(Chi_01,Chi_01)	\
-  VPERM3(Chi_02,Chi_02)	\
-  VPERM3(Chi_10,Chi_10)	\
-  VPERM3(Chi_11,Chi_11)	\
-  VPERM3(Chi_12,Chi_12) );
+#define PERMUTE_DIR3 __asm__ (				\
+			      VPERM3(Chi_00,Chi_00)	\
+			      VPERM3(Chi_01,Chi_01)	\
+			      VPERM3(Chi_02,Chi_02)	\
+			      VPERM3(Chi_10,Chi_10)	\
+			      VPERM3(Chi_11,Chi_11)	\
+			      VPERM3(Chi_12,Chi_12) );
 
 
 #define MULT_ADDSUB_2SPIN(ptr,pf)					\
-  LOAD64(%r8,ptr)						\
-  LOAD64(%r9,pf)						\
-	   __asm__ (						\
-	   VPREFETCH_G2(9,%r8)				   \
-	   VPREFETCH_G2(10,%r8)					   \
-	   VPREFETCH_G2(11,%r8)					   \
-	   VPREFETCH_G2(12,%r8)					   \
-	   VPREFETCH_G2(13,%r8)					   \
-	   VPREFETCH_G2(14,%r8)					   \
-	   VPREFETCH_G2(15,%r8)					   \
-	   VPREFETCH_G2(16,%r8)					   \
-	   VPREFETCH_G2(17,%r8)					   \
-	   VSHUF(Chi_00,T1)				\
-	   VMOVIDUP(0,%r8,Z0 )					\
-           VMOVIDUP(3,%r8,Z1 )					\
+  LOAD64(%r8,ptr)							\
+  LOAD64(%r9,pf)							\
+  __asm__ (								\
+	   VPREFETCH_G2(9,%r8)						\
+	   VPREFETCH_G2(10,%r8)						\
+	   VPREFETCH_G2(11,%r8)						\
+	   VPREFETCH_G2(12,%r8)						\
+	   VPREFETCH_G2(13,%r8)						\
+	   VPREFETCH_G2(14,%r8)						\
+	   VPREFETCH_G2(15,%r8)						\
+	   VPREFETCH_G2(16,%r8)						\
+	   VPREFETCH_G2(17,%r8)						\
+	   VSHUF(Chi_00,T1)						\
+	   VMOVIDUP(0,%r8,Z0 )						\
+           VMOVIDUP(3,%r8,Z1 )						\
            VMOVIDUP(6,%r8,Z2 )	          VSHUF(Chi_10,T2)		\
 	   /*6*/							\
-           VMUL(Z0,T1,UChi_00)            VMOVRDUP(0,%r8,Z3 )	\
-           VMUL(Z0,T2,UChi_10)            VMOVRDUP(3,%r8,Z4 )	\
-           VMUL(Z1,T1,UChi_01)            VMOVRDUP(6,%r8,Z5 )	\
-           VMUL(Z1,T2,UChi_11)            VMOVIDUP(1,%r8,Z0 )	\
-           VMUL(Z2,T1,UChi_02)            VMOVIDUP(4,%r8,Z1 )	\
-           VMUL(Z2,T2,UChi_12)            VMOVIDUP(7,%r8,Z2 )	\
-	   VPREFETCH_M1(0,%r9)					   \
-	   VPREFETCH_M1(1,%r9)					   \
-	   VPREFETCH_M1(2,%r9)					   \
-	   VPREFETCH_M1(3,%r9)					   \
-	   /*18*/						\
-           VMADDSUB(Z3,Chi_00,UChi_00)    VSHUF(Chi_01,T1)	\
-           VMADDSUB(Z3,Chi_10,UChi_10)				\
-           VMADDSUB(Z4,Chi_00,UChi_01)    VMOVRDUP(1,%r8,Z3 )	\
-           VMADDSUB(Z4,Chi_10,UChi_11)    VSHUF(Chi_11,T2)	\
-           VMADDSUB(Z5,Chi_00,UChi_02)    VMOVRDUP(4,%r8,Z4 )	\
-           VMADDSUB(Z5,Chi_10,UChi_12)				\
-	   VPREFETCH_M1(4,%r9)					   \
-	   VPREFETCH_M1(5,%r9)					   \
-	   VPREFETCH_M1(6,%r9)					   \
-	   VPREFETCH_M1(7,%r9)					   \
-	   /*28*/						\
-           VMADDSUB(Z0,T1,UChi_00)        VMOVRDUP(7,%r8,Z5 )	\
-           VMADDSUB(Z0,T2,UChi_10)				\
-           VMADDSUB(Z1,T1,UChi_01)        VMOVIDUP(2,%r8,Z0 )	\
-           VMADDSUB(Z1,T2,UChi_11)				\
-           VMADDSUB(Z2,T1,UChi_02)        VMOVIDUP(5,%r8,Z1 )	\
-           VMADDSUB(Z2,T2,UChi_12)        VMOVIDUP(8,%r8,Z2 )	\
-	   VPREFETCH2(12,%r9)					   \
-	   VPREFETCH2(13,%r9)					   \
-	   VPREFETCH2(14,%r9)					   \
-	   VPREFETCH2(15,%r9)					   \
-	   VPREFETCH2(16,%r9)					   \
-	   VPREFETCH2(17,%r9)					   \
-	   VPREFETCH2(18,%r9)					   \
-	   VPREFETCH2(19,%r9)					   \
-	   VPREFETCH2(20,%r9)					   \
-	   VPREFETCH2(21,%r9)					   \
-	   VPREFETCH2(22,%r9)					   \
-	   VPREFETCH2(23,%r9)					   \
-           /*38*/						\
-           VMADDSUB(Z3,Chi_01,UChi_00)    VSHUF(Chi_02,T1)	\
-           VMADDSUB(Z3,Chi_11,UChi_10)				\
-           VMADDSUB(Z4,Chi_01,UChi_01)    VMOVRDUP(2,%r8,Z3 )	\
-           VMADDSUB(Z4,Chi_11,UChi_11)    VSHUF(Chi_12,T2)	\
-           VMADDSUB(Z5,Chi_01,UChi_02)    VMOVRDUP(5,%r8,Z4 )	\
-           VMADDSUB(Z5,Chi_11,UChi_12)				\
-	   VPREFETCH_M1(9,%r8)				   \
-	   VPREFETCH_M1(10,%r8)					   \
-	   VPREFETCH_M1(11,%r8)					   \
-	   VPREFETCH_M1(12,%r8)					   \
-	   VPREFETCH_M1(13,%r8)					   \
-	   VPREFETCH_M1(14,%r8)					   \
-	   VPREFETCH_M1(15,%r8)					   \
-	   VPREFETCH_M1(16,%r8)					   \
-	   VPREFETCH_M1(17,%r8)					   \
-	   /*48*/						\
-           VMADDSUB(Z0,T1,UChi_00)        VMOVRDUP(8,%r8,Z5 ) \
-           VMADDSUB(Z0,T2,UChi_10)			      \
-           VMADDSUB(Z1,T1,UChi_01)			      \
-           VMADDSUB(Z1,T2,UChi_11)			      \
-           VMADDSUB(Z2,T1,UChi_02)			      \
-           VMADDSUB(Z2,T2,UChi_12)			      \
-	   VPREFETCH_M1(8,%r9)					   \
-	   VPREFETCH_M1(9,%r9)					   \
-	   VPREFETCH_M1(10,%r9)					   \
-	   VPREFETCH_M1(11,%r9)					   \
-	   /*55*/					      \
-           VMADDSUB(Z3,Chi_02,UChi_00)			      \
-           VMADDSUB(Z3,Chi_12,UChi_10)			      \
-           VMADDSUB(Z4,Chi_02,UChi_01)			      \
-           VMADDSUB(Z4,Chi_12,UChi_11)			      \
-           VMADDSUB(Z5,Chi_02,UChi_02)			      \
-           VMADDSUB(Z5,Chi_12,UChi_12)			      \
+           VMUL(Z0,T1,UChi_00)            VMOVRDUP(0,%r8,Z3 )		\
+           VMUL(Z0,T2,UChi_10)            VMOVRDUP(3,%r8,Z4 )		\
+           VMUL(Z1,T1,UChi_01)            VMOVRDUP(6,%r8,Z5 )		\
+           VMUL(Z1,T2,UChi_11)            VMOVIDUP(1,%r8,Z0 )		\
+           VMUL(Z2,T1,UChi_02)            VMOVIDUP(4,%r8,Z1 )		\
+           VMUL(Z2,T2,UChi_12)            VMOVIDUP(7,%r8,Z2 )		\
+	   VPREFETCH_M1(0,%r9)						\
+	   VPREFETCH_M1(1,%r9)						\
+	   VPREFETCH_M1(2,%r9)						\
+	   VPREFETCH_M1(3,%r9)						\
+	   /*18*/							\
+           VMADDSUB(Z3,Chi_00,UChi_00)    VSHUF(Chi_01,T1)		\
+           VMADDSUB(Z3,Chi_10,UChi_10)					\
+           VMADDSUB(Z4,Chi_00,UChi_01)    VMOVRDUP(1,%r8,Z3 )		\
+           VMADDSUB(Z4,Chi_10,UChi_11)    VSHUF(Chi_11,T2)		\
+           VMADDSUB(Z5,Chi_00,UChi_02)    VMOVRDUP(4,%r8,Z4 )		\
+           VMADDSUB(Z5,Chi_10,UChi_12)					\
+	   VPREFETCH_M1(4,%r9)						\
+	   VPREFETCH_M1(5,%r9)						\
+	   VPREFETCH_M1(6,%r9)						\
+	   VPREFETCH_M1(7,%r9)						\
+	   /*28*/							\
+           VMADDSUB(Z0,T1,UChi_00)        VMOVRDUP(7,%r8,Z5 )		\
+           VMADDSUB(Z0,T2,UChi_10)					\
+           VMADDSUB(Z1,T1,UChi_01)        VMOVIDUP(2,%r8,Z0 )		\
+           VMADDSUB(Z1,T2,UChi_11)					\
+           VMADDSUB(Z2,T1,UChi_02)        VMOVIDUP(5,%r8,Z1 )		\
+           VMADDSUB(Z2,T2,UChi_12)        VMOVIDUP(8,%r8,Z2 )		\
+	   VPREFETCH2(12,%r9)						\
+	   VPREFETCH2(13,%r9)						\
+	   VPREFETCH2(14,%r9)						\
+	   VPREFETCH2(15,%r9)						\
+	   VPREFETCH2(16,%r9)						\
+	   VPREFETCH2(17,%r9)						\
+	   VPREFETCH2(18,%r9)						\
+	   VPREFETCH2(19,%r9)						\
+	   VPREFETCH2(20,%r9)						\
+	   VPREFETCH2(21,%r9)						\
+	   VPREFETCH2(22,%r9)						\
+	   VPREFETCH2(23,%r9)						\
+           /*38*/							\
+           VMADDSUB(Z3,Chi_01,UChi_00)    VSHUF(Chi_02,T1)		\
+           VMADDSUB(Z3,Chi_11,UChi_10)					\
+           VMADDSUB(Z4,Chi_01,UChi_01)    VMOVRDUP(2,%r8,Z3 )		\
+           VMADDSUB(Z4,Chi_11,UChi_11)    VSHUF(Chi_12,T2)		\
+           VMADDSUB(Z5,Chi_01,UChi_02)    VMOVRDUP(5,%r8,Z4 )		\
+           VMADDSUB(Z5,Chi_11,UChi_12)					\
+	   VPREFETCH_M1(9,%r8)						\
+	   VPREFETCH_M1(10,%r8)						\
+	   VPREFETCH_M1(11,%r8)						\
+	   VPREFETCH_M1(12,%r8)						\
+	   VPREFETCH_M1(13,%r8)						\
+	   VPREFETCH_M1(14,%r8)						\
+	   VPREFETCH_M1(15,%r8)						\
+	   VPREFETCH_M1(16,%r8)						\
+	   VPREFETCH_M1(17,%r8)						\
+	   /*48*/							\
+           VMADDSUB(Z0,T1,UChi_00)        VMOVRDUP(8,%r8,Z5 )		\
+           VMADDSUB(Z0,T2,UChi_10)					\
+           VMADDSUB(Z1,T1,UChi_01)					\
+           VMADDSUB(Z1,T2,UChi_11)					\
+           VMADDSUB(Z2,T1,UChi_02)					\
+           VMADDSUB(Z2,T2,UChi_12)					\
+	   VPREFETCH_M1(8,%r9)						\
+	   VPREFETCH_M1(9,%r9)						\
+	   VPREFETCH_M1(10,%r9)						\
+	   VPREFETCH_M1(11,%r9)						\
+	   /*55*/							\
+           VMADDSUB(Z3,Chi_02,UChi_00)					\
+           VMADDSUB(Z3,Chi_12,UChi_10)					\
+           VMADDSUB(Z4,Chi_02,UChi_01)					\
+           VMADDSUB(Z4,Chi_12,UChi_11)					\
+           VMADDSUB(Z5,Chi_02,UChi_02)					\
+           VMADDSUB(Z5,Chi_12,UChi_12)					\
 	   /*61 insns*/							);
 
 
-#define MULT_ADDSUB_2SPIN_LS(ptr,pf)				   \
-  LOAD64(%r8,ptr)						   \
-  LOAD64(%r9,pf)						   \
-  __asm__ (							   \
-           VSHUF(Chi_00,T1)      VSHUF(Chi_10,T2)		   \
-           VMULIDUP(0,%r8,T1,UChi_00) VMULIDUP(0,%r8,T2,UChi_10)   \
-           VMULIDUP(3,%r8,T1,UChi_01) VMULIDUP(3,%r8,T2,UChi_11)   \
-           VMULIDUP(6,%r8,T1,UChi_02) VMULIDUP(6,%r8,T2,UChi_12)   \
-	   VPREFETCH_M1(0,%r9)					   \
-	   VPREFETCH_M1(1,%r9)					   \
-	   VPREFETCH_M1(2,%r9)					   \
-	   VPREFETCH_M1(3,%r9)					   \
-	   /*8*/						   \
-           VSHUF(Chi_01,T1)	  VSHUF(Chi_11,T2)	       	   \
+#define MULT_ADDSUB_2SPIN_LS(ptr,pf)					\
+  LOAD64(%r8,ptr)							\
+  LOAD64(%r9,pf)							\
+  __asm__ (								\
+           VSHUF(Chi_00,T1)      VSHUF(Chi_10,T2)			\
+           VMULIDUP(0,%r8,T1,UChi_00) VMULIDUP(0,%r8,T2,UChi_10)	\
+           VMULIDUP(3,%r8,T1,UChi_01) VMULIDUP(3,%r8,T2,UChi_11)	\
+           VMULIDUP(6,%r8,T1,UChi_02) VMULIDUP(6,%r8,T2,UChi_12)	\
+	   VPREFETCH_M1(0,%r9)						\
+	   VPREFETCH_M1(1,%r9)						\
+	   VPREFETCH_M1(2,%r9)						\
+	   VPREFETCH_M1(3,%r9)						\
+	   /*8*/							\
+           VSHUF(Chi_01,T1)	  VSHUF(Chi_11,T2)			\
            VMADDSUBRDUP(0,%r8,Chi_00,UChi_00) VMADDSUBRDUP(0,%r8,Chi_10,UChi_10) \
            VMADDSUBRDUP(3,%r8,Chi_00,UChi_01) VMADDSUBRDUP(3,%r8,Chi_10,UChi_11) \
            VMADDSUBRDUP(6,%r8,Chi_00,UChi_02) VMADDSUBRDUP(6,%r8,Chi_10,UChi_12) \
-	   VPREFETCH_M1(4,%r9)					   \
-	   VPREFETCH_M1(5,%r9)					   \
-	   VPREFETCH_M1(6,%r9)					   \
-	   VPREFETCH_M1(7,%r9)					   \
-	   /*16*/					  	   \
-           VMADDSUBIDUP(1,%r8,T1,UChi_00)     VMADDSUBIDUP(1,%r8,T2,UChi_10)	   \
+	   VPREFETCH_M1(4,%r9)						\
+	   VPREFETCH_M1(5,%r9)						\
+	   VPREFETCH_M1(6,%r9)						\
+	   VPREFETCH_M1(7,%r9)						\
+	   /*16*/							\
+           VMADDSUBIDUP(1,%r8,T1,UChi_00)     VMADDSUBIDUP(1,%r8,T2,UChi_10) \
            VMADDSUBIDUP(4,%r8,T1,UChi_01)     VMADDSUBIDUP(4,%r8,T2,UChi_11) \
            VMADDSUBIDUP(7,%r8,T1,UChi_02)     VMADDSUBIDUP(7,%r8,T2,UChi_12) \
-	   VPREFETCH_M1(8,%r9)					   \
-	   VPREFETCH_M1(9,%r9)					   \
-	   VPREFETCH_M1(10,%r9)					   \
-	   VPREFETCH_M1(11,%r9)					   \
-           /*22*/						   \
-           VSHUF(Chi_02,T1)    VSHUF(Chi_12,T2)	                   \
+	   VPREFETCH_M1(8,%r9)						\
+	   VPREFETCH_M1(9,%r9)						\
+	   VPREFETCH_M1(10,%r9)						\
+	   VPREFETCH_M1(11,%r9)						\
+           /*22*/							\
+           VSHUF(Chi_02,T1)    VSHUF(Chi_12,T2)				\
            VMADDSUBRDUP(1,%r8,Chi_01,UChi_00) VMADDSUBRDUP(1,%r8,Chi_11,UChi_10) \
            VMADDSUBRDUP(4,%r8,Chi_01,UChi_01) VMADDSUBRDUP(4,%r8,Chi_11,UChi_11) \
            VMADDSUBRDUP(7,%r8,Chi_01,UChi_02) VMADDSUBRDUP(7,%r8,Chi_11,UChi_12) \
-	   VPREFETCH_M2(12,%r9)					   \
-	   VPREFETCH_M2(13,%r9)					   \
-	   VPREFETCH_M2(14,%r9)					   \
-	   VPREFETCH_M2(15,%r9)					   \
-	   /*30*/						   \
-           VMADDSUBIDUP(2,%r8,T1,UChi_00)     VMADDSUBIDUP(2,%r8,T2,UChi_10)	   \
-           VMADDSUBIDUP(5,%r8,T1,UChi_01)     VMADDSUBIDUP(5,%r8,T2,UChi_11)     \
-	   VPREFETCH_M2(16,%r9)					   \
-	   VPREFETCH_M2(17,%r9)					   \
-	   VPREFETCH_M2(18,%r9)					   \
-	   VPREFETCH_M2(19,%r9)					   \
-           VMADDSUBIDUP(8,%r8,T1,UChi_02)     VMADDSUBIDUP(8,%r8,T2,UChi_12)     \
-	   /*36*/					           \
+	   VPREFETCH_M2(12,%r9)						\
+	   VPREFETCH_M2(13,%r9)						\
+	   VPREFETCH_M2(14,%r9)						\
+	   VPREFETCH_M2(15,%r9)						\
+	   /*30*/							\
+           VMADDSUBIDUP(2,%r8,T1,UChi_00)     VMADDSUBIDUP(2,%r8,T2,UChi_10) \
+           VMADDSUBIDUP(5,%r8,T1,UChi_01)     VMADDSUBIDUP(5,%r8,T2,UChi_11) \
+	   VPREFETCH_M2(16,%r9)						\
+	   VPREFETCH_M2(17,%r9)						\
+	   VPREFETCH_M2(18,%r9)						\
+	   VPREFETCH_M2(19,%r9)						\
+           VMADDSUBIDUP(8,%r8,T1,UChi_02)     VMADDSUBIDUP(8,%r8,T2,UChi_12) \
+	   /*36*/							\
            VMADDSUBRDUP(2,%r8,Chi_02,UChi_00) VMADDSUBRDUP(2,%r8,Chi_12,UChi_10) \
            VMADDSUBRDUP(5,%r8,Chi_02,UChi_01) VMADDSUBRDUP(5,%r8,Chi_12,UChi_11) \
            VMADDSUBRDUP(8,%r8,Chi_02,UChi_02) VMADDSUBRDUP(8,%r8,Chi_12,UChi_12) \
-	   VPREFETCH_M2(20,%r9)					   \
-	   VPREFETCH_M2(21,%r9)					   \
-	   VPREFETCH_M2(22,%r9)					   \
-	   VPREFETCH_M2(23,%r9)					   \
-	   VPREFETCH_G1(2,%r8)					   \
-	   VPREFETCH_G1(3,%r8)					   \
-	   VPREFETCH_G2(4,%r8)					   \
-	   VPREFETCH_G2(5,%r8)					   \
-	   VPREFETCH_G2(6,%r8)					   \
-	   VPREFETCH_G2(7,%r8)					   \
+	   VPREFETCH_M2(20,%r9)						\
+	   VPREFETCH_M2(21,%r9)						\
+	   VPREFETCH_M2(22,%r9)						\
+	   VPREFETCH_M2(23,%r9)						\
+	   VPREFETCH_G1(2,%r8)						\
+	   VPREFETCH_G1(3,%r8)						\
+	   VPREFETCH_G2(4,%r8)						\
+	   VPREFETCH_G2(5,%r8)						\
+	   VPREFETCH_G2(6,%r8)						\
+	   VPREFETCH_G2(7,%r8)						\
 	   /*42 insns*/						);
 
-#define MULT_ADDSUB_2SPIN_LSNOPF(ptr,pf)				   \
-  LOAD64(%r8,ptr)						   \
-  LOAD64(%r9,pf)						   \
-  __asm__ (							   \
-           VSHUF(Chi_00,T1)      VSHUF(Chi_10,T2)		   \
-           VMULIDUP(0,%r8,T1,UChi_00) VMULIDUP(0,%r8,T2,UChi_10)   \
-           VMULIDUP(3,%r8,T1,UChi_01) VMULIDUP(3,%r8,T2,UChi_11)   \
-           VMULIDUP(6,%r8,T1,UChi_02) VMULIDUP(6,%r8,T2,UChi_12)   \
-	   /*8*/						   \
-           VSHUF(Chi_01,T1)	  VSHUF(Chi_11,T2)	       	   \
+#define MULT_ADDSUB_2SPIN_LSNOPF(ptr,pf)				\
+  LOAD64(%r8,ptr)							\
+  LOAD64(%r9,pf)							\
+  __asm__ (								\
+           VSHUF(Chi_00,T1)      VSHUF(Chi_10,T2)			\
+           VMULIDUP(0,%r8,T1,UChi_00) VMULIDUP(0,%r8,T2,UChi_10)	\
+           VMULIDUP(3,%r8,T1,UChi_01) VMULIDUP(3,%r8,T2,UChi_11)	\
+           VMULIDUP(6,%r8,T1,UChi_02) VMULIDUP(6,%r8,T2,UChi_12)	\
+	   /*8*/							\
+           VSHUF(Chi_01,T1)	  VSHUF(Chi_11,T2)			\
            VMADDSUBRDUP(0,%r8,Chi_00,UChi_00) VMADDSUBRDUP(0,%r8,Chi_10,UChi_10) \
            VMADDSUBRDUP(3,%r8,Chi_00,UChi_01) VMADDSUBRDUP(3,%r8,Chi_10,UChi_11) \
            VMADDSUBRDUP(6,%r8,Chi_00,UChi_02) VMADDSUBRDUP(6,%r8,Chi_10,UChi_12) \
-	   /*16*/					  	   \
-           VMADDSUBIDUP(1,%r8,T1,UChi_00)     VMADDSUBIDUP(1,%r8,T2,UChi_10)	   \
+	   /*16*/							\
+           VMADDSUBIDUP(1,%r8,T1,UChi_00)     VMADDSUBIDUP(1,%r8,T2,UChi_10) \
            VMADDSUBIDUP(4,%r8,T1,UChi_01)     VMADDSUBIDUP(4,%r8,T2,UChi_11) \
            VMADDSUBIDUP(7,%r8,T1,UChi_02)     VMADDSUBIDUP(7,%r8,T2,UChi_12) \
-           /*22*/						   \
-           VSHUF(Chi_02,T1)    VSHUF(Chi_12,T2)	                   \
+           /*22*/							\
+           VSHUF(Chi_02,T1)    VSHUF(Chi_12,T2)				\
            VMADDSUBRDUP(1,%r8,Chi_01,UChi_00) VMADDSUBRDUP(1,%r8,Chi_11,UChi_10) \
            VMADDSUBRDUP(4,%r8,Chi_01,UChi_01) VMADDSUBRDUP(4,%r8,Chi_11,UChi_11) \
            VMADDSUBRDUP(7,%r8,Chi_01,UChi_02) VMADDSUBRDUP(7,%r8,Chi_11,UChi_12) \
-	   /*30*/						   \
-           VMADDSUBIDUP(2,%r8,T1,UChi_00)     VMADDSUBIDUP(2,%r8,T2,UChi_10)	   \
-           VMADDSUBIDUP(5,%r8,T1,UChi_01)     VMADDSUBIDUP(5,%r8,T2,UChi_11)     \
-           VMADDSUBIDUP(8,%r8,T1,UChi_02)     VMADDSUBIDUP(8,%r8,T2,UChi_12)     \
-	   /*36*/					           \
+	   /*30*/							\
+           VMADDSUBIDUP(2,%r8,T1,UChi_00)     VMADDSUBIDUP(2,%r8,T2,UChi_10) \
+           VMADDSUBIDUP(5,%r8,T1,UChi_01)     VMADDSUBIDUP(5,%r8,T2,UChi_11) \
+           VMADDSUBIDUP(8,%r8,T1,UChi_02)     VMADDSUBIDUP(8,%r8,T2,UChi_12) \
+	   /*36*/							\
            VMADDSUBRDUP(2,%r8,Chi_02,UChi_00) VMADDSUBRDUP(2,%r8,Chi_12,UChi_10) \
            VMADDSUBRDUP(5,%r8,Chi_02,UChi_01) VMADDSUBRDUP(5,%r8,Chi_12,UChi_11) \
            VMADDSUBRDUP(8,%r8,Chi_02,UChi_02) VMADDSUBRDUP(8,%r8,Chi_12,UChi_12) \
-	   /*	   VPREFETCH1(2,%r8)*/				   \
-	   /*	   VPREFETCH1(3,%r8)*/				   \
+	   /*	   VPREFETCH1(2,%r8)*/					\
+	   /*	   VPREFETCH1(3,%r8)*/					\
 	   /*42 insns*/						);
 
 
 #define Z6 Chi_00
-#define MULT_ADDSUB_2SPIN_NEW(ptr,pf)			       \
-  LOAD64(%r8,ptr)					       \
-  __asm__ (							  \
-   VSHUFMEM(0,%r8,Z0)					          \
-   VRDUP(Chi_00,T1)           VIDUP(Chi_00,Chi_00)	          \
-   VRDUP(Chi_10,T2)           VIDUP(Chi_10,Chi_10)		  \
-   VMUL(Z0,Chi_00,Z1)         VMUL(Z0,Chi_10,Z2)		  \
-   VSHUFMEM(3,%r8,Z0)						  \
-   VMUL(Z0,Chi_00,Z3)         VMUL(Z0,Chi_10,Z4)		  \
-   VSHUFMEM(6,%r8,Z0)						  \
-   VMUL(Z0,Chi_00,Z5)         VMUL(Z0,Chi_10,Z6)		  \
-   VMULMEM(0,%r8,T1,UChi_00)  VMULMEM(0,%r8,T2,UChi_10)		  \
-   VMULMEM(3,%r8,T1,UChi_01)  VMULMEM(3,%r8,T2,UChi_11)		  \
-   VMULMEM(6,%r8,T1,UChi_02)  VMULMEM(6,%r8,T2,UChi_12)		  \
-   /*11 cycles*/						  \
-   VSHUFMEM(1,%r8,Z0)						  \
-   VRDUP(Chi_01,T1)           VIDUP(Chi_01,Chi_01)		  \
-   VRDUP(Chi_11,T2)           VIDUP(Chi_11,Chi_11)		  \
-   VMADD(Z0,Chi_01,Z1)        VMADD(Z0,Chi_11,Z2)		  \
-   VSHUFMEM(4,%r8,Z0)						  \
-   VMADD(Z0,Chi_01,Z3)        VMADD(Z0,Chi_11,Z4)		  \
-   VSHUFMEM(7,%r8,Z0)						  \
-   VMADD(Z0,Chi_01,Z5)        VMADD(Z0,Chi_11,Z6)		  \
-   VMADDMEM(1,%r8,T1,UChi_00) VMADDMEM(1,%r8,T2,UChi_10)	  \
-   VMADDMEM(4,%r8,T1,UChi_01) VMADDMEM(4,%r8,T2,UChi_11)	  \
-   VMADDMEM(7,%r8,T1,UChi_02) VMADDMEM(7,%r8,T2,UChi_12)	  \
-   /*22 cycles*/						  \
-   VSHUFMEM(2,%r8,Z0)						  \
-   VRDUP(Chi_02,T1)        VIDUP(Chi_02,Chi_02)			  \
-   VRDUP(Chi_12,T2)        VIDUP(Chi_12,Chi_12)			  \
-   VMADD(Z0,Chi_02,Z1)        VMADD(Z0,Chi_12,Z2)		  \
-   VSHUFMEM(5,%r8,Z0)						  \
-   VMADD(Z0,Chi_02,Z3)        VMADD(Z0,Chi_12,Z4)		  \
-   VSHUFMEM(8,%r8,Z0)						  \
-   VMADD(Z0,Chi_02,Z5)        VMADD(Z0,Chi_12,Z6)		  \
-   /*33 cycles*/						  \
-   VMADDSUBMEM(2,%r8,T1,Z1)   VMADDSUBMEM(2,%r8,T2,Z2)		  \
-   VMADDSUBMEM(5,%r8,T1,Z3)   VMADDSUBMEM(5,%r8,T2,Z4)	          \
-   VMADDSUBMEM(8,%r8,T1,Z5)   VMADDSUBMEM(8,%r8,T2,Z6)	       \
-  /*stall*/						       \
-  /*stall*/						       \
-  /*stall*/						       \
-  VADD(Z1,UChi_00,UChi_00)   VADD(Z2,UChi_10,UChi_10)	       \
-  VADD(Z3,UChi_01,UChi_01)   VADD(Z4,UChi_11,UChi_11)	       \
-  VADD(Z5,UChi_02,UChi_02)   VADD(Z6,UChi_12,UChi_12) )
+#define MULT_ADDSUB_2SPIN_NEW(ptr,pf)					\
+  LOAD64(%r8,ptr)							\
+  __asm__ (								\
+	   VSHUFMEM(0,%r8,Z0)						\
+	   VRDUP(Chi_00,T1)           VIDUP(Chi_00,Chi_00)		\
+	   VRDUP(Chi_10,T2)           VIDUP(Chi_10,Chi_10)		\
+	   VMUL(Z0,Chi_00,Z1)         VMUL(Z0,Chi_10,Z2)		\
+	   VSHUFMEM(3,%r8,Z0)						\
+	   VMUL(Z0,Chi_00,Z3)         VMUL(Z0,Chi_10,Z4)		\
+	   VSHUFMEM(6,%r8,Z0)						\
+	   VMUL(Z0,Chi_00,Z5)         VMUL(Z0,Chi_10,Z6)		\
+	   VMULMEM(0,%r8,T1,UChi_00)  VMULMEM(0,%r8,T2,UChi_10)		\
+	   VMULMEM(3,%r8,T1,UChi_01)  VMULMEM(3,%r8,T2,UChi_11)		\
+	   VMULMEM(6,%r8,T1,UChi_02)  VMULMEM(6,%r8,T2,UChi_12)		\
+	   /*11 cycles*/						\
+	   VSHUFMEM(1,%r8,Z0)						\
+	   VRDUP(Chi_01,T1)           VIDUP(Chi_01,Chi_01)		\
+	   VRDUP(Chi_11,T2)           VIDUP(Chi_11,Chi_11)		\
+	   VMADD(Z0,Chi_01,Z1)        VMADD(Z0,Chi_11,Z2)		\
+	   VSHUFMEM(4,%r8,Z0)						\
+	   VMADD(Z0,Chi_01,Z3)        VMADD(Z0,Chi_11,Z4)		\
+	   VSHUFMEM(7,%r8,Z0)						\
+	   VMADD(Z0,Chi_01,Z5)        VMADD(Z0,Chi_11,Z6)		\
+	   VMADDMEM(1,%r8,T1,UChi_00) VMADDMEM(1,%r8,T2,UChi_10)	\
+	   VMADDMEM(4,%r8,T1,UChi_01) VMADDMEM(4,%r8,T2,UChi_11)	\
+	   VMADDMEM(7,%r8,T1,UChi_02) VMADDMEM(7,%r8,T2,UChi_12)	\
+	   /*22 cycles*/						\
+	   VSHUFMEM(2,%r8,Z0)						\
+	   VRDUP(Chi_02,T1)        VIDUP(Chi_02,Chi_02)			\
+	   VRDUP(Chi_12,T2)        VIDUP(Chi_12,Chi_12)			\
+	   VMADD(Z0,Chi_02,Z1)        VMADD(Z0,Chi_12,Z2)		\
+	   VSHUFMEM(5,%r8,Z0)						\
+	   VMADD(Z0,Chi_02,Z3)        VMADD(Z0,Chi_12,Z4)		\
+	   VSHUFMEM(8,%r8,Z0)						\
+	   VMADD(Z0,Chi_02,Z5)        VMADD(Z0,Chi_12,Z6)		\
+	   /*33 cycles*/						\
+	   VMADDSUBMEM(2,%r8,T1,Z1)   VMADDSUBMEM(2,%r8,T2,Z2)		\
+	   VMADDSUBMEM(5,%r8,T1,Z3)   VMADDSUBMEM(5,%r8,T2,Z4)		\
+	   VMADDSUBMEM(8,%r8,T1,Z5)   VMADDSUBMEM(8,%r8,T2,Z6)		\
+	   /*stall*/							\
+	   /*stall*/							\
+	   /*stall*/							\
+	   VADD(Z1,UChi_00,UChi_00)   VADD(Z2,UChi_10,UChi_10)		\
+	   VADD(Z3,UChi_01,UChi_01)   VADD(Z4,UChi_11,UChi_11)		\
+	   VADD(Z5,UChi_02,UChi_02)   VADD(Z6,UChi_12,UChi_12) )
 
 
 #endif
diff --git a/Grid/simd/Simd.h b/Grid/simd/Simd.h
index 3f2b10dc..bc8cd2fd 100644
--- a/Grid/simd/Simd.h
+++ b/Grid/simd/Simd.h
@@ -31,6 +31,10 @@ directory
 #ifndef GRID_SIMD_H
 #define GRID_SIMD_H
 
+#ifdef GRID_NVCC
+#include <thrust/complex.h>
+#endif
+
 ////////////////////////////////////////////////////////////////////////
 // Define scalar and vector floating point types
 //
@@ -41,7 +45,6 @@ directory
 // Vector types are arch dependent
 ////////////////////////////////////////////////////////////////////////
 
-
 #define _MM_SELECT_FOUR_FOUR(A,B,C,D) ((A<<6)|(B<<4)|(C<<2)|(D))
 #define _MM_SELECT_FOUR_FOUR_STRING(A,B,C,D) "((" #A "<<6)|(" #B "<<4)|(" #C "<<2)|(" #D "))"
 #define _MM_SELECT_EIGHT_TWO(A,B,C,D,E,F,G,H) ((A<<7)|(B<<6)|(C<<5)|(D<<4)|(E<<3)|(F<<2)|(G<<4)|(H))
@@ -50,206 +53,244 @@ directory
 
 #define RotateBit (0x100)
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  typedef uint32_t Integer;
+typedef uint32_t Integer;
 
-  typedef  float  RealF;
-  typedef  double RealD;
+typedef  float  RealF;
+typedef  double RealD;
 #ifdef GRID_DEFAULT_PRECISION_DOUBLE
-  typedef RealD   Real;
+typedef RealD   Real;
 #else
-  typedef RealF  Real;
+typedef RealF  Real;
 #endif
 
-  typedef std::complex<RealF> ComplexF;
-  typedef std::complex<RealD> ComplexD;
-  typedef std::complex<Real>  Complex;
+#ifdef GRID_NVCC
+typedef thrust::complex<RealF> ComplexF;
+typedef thrust::complex<RealD> ComplexD;
+typedef thrust::complex<Real>  Complex;
+template<class T> using complex = thrust::complex<T>;
 
-  inline RealF adj(const RealF  & r){ return r; }
-  inline RealF conjugate(const RealF  & r){ return r; }
-  inline RealF real(const RealF  & r){ return r; }
+accelerator_inline ComplexD pow(const ComplexD& r,RealD y){ return(thrust::pow(r,(double)y)); }
+accelerator_inline ComplexF pow(const ComplexF& r,RealF y){ return(thrust::pow(r,(float)y)); }
+#else 
+typedef std::complex<RealF> ComplexF;
+typedef std::complex<RealD> ComplexD;
+typedef std::complex<Real>  Complex;
+template<class T> using complex = std::complex<T>;
 
-  inline RealD adj(const RealD  & r){ return r; }
-  inline RealD conjugate(const RealD  & r){ return r; }
-  inline RealD real(const RealD  & r){ return r; }
+accelerator_inline ComplexD pow(const ComplexD& r,RealD y){ return(std::pow(r,y)); }
+accelerator_inline ComplexF pow(const ComplexF& r,RealF y){ return(std::pow(r,y)); }
+#endif
 
-  inline RealD sqrt(const RealD  & r){ return std::sqrt(r); }
+//accelerator_inline RealD pow(const RealD& r,RealD y){ return(std::pow(r,y)); }
+//accelerator_inline RealD sqrt(const RealD  & r){ return std::sqrt(r); }
 
-  inline ComplexD conjugate(const ComplexD& r){ return(conj(r)); }
-  inline ComplexD adj(const ComplexD& r){ return(conjugate(r)); }
-  inline ComplexF conjugate(const ComplexF& r ){ return(conj(r)); }
-  inline ComplexF adj(const ComplexF& r ){ return(conjugate(r)); }
+// This comes from ::pow already from math.h and CUDA
+// Calls either Grid::pow for complex, or std::pow for real
+// Problem is CUDA math_functions is exposing ::pow, and I can't define
 
-  inline ComplexD innerProduct(const ComplexD & l, const ComplexD & r) { return conjugate(l)*r; }
-  inline ComplexF innerProduct(const ComplexF & l, const ComplexF & r) { return conjugate(l)*r; }
-  inline RealD innerProduct(const RealD & l, const RealD & r) { return l*r; }
-  inline RealF innerProduct(const RealF & l, const RealF & r) { return l*r; }
+using std::abs;
+using std::pow;
+using std::sqrt;
 
-  inline ComplexD Reduce(const ComplexD& r){ return r; }
-  inline ComplexF Reduce(const ComplexF& r){ return r; }
-  inline RealD Reduce(const RealD& r){ return r; }
-  inline RealF Reduce(const RealF& r){ return r; }
+accelerator_inline RealF    conjugate(const RealF  & r){ return r; }
+accelerator_inline RealD    conjugate(const RealD  & r){ return r; }
+accelerator_inline ComplexD conjugate(const ComplexD& r){ return(conj(r)); }
+accelerator_inline ComplexF conjugate(const ComplexF& r ){ return(conj(r)); }
 
-  inline RealD toReal(const ComplexD& r){ return real(r); }
-  inline RealF toReal(const ComplexF& r){ return real(r); }
-  inline RealD toReal(const RealD& r){ return r; }
-  inline RealF toReal(const RealF& r){ return r; }
+accelerator_inline RealF    adj(const RealF  & r){ return r; }
+accelerator_inline RealD    adj(const RealD  & r){ return r; }
+accelerator_inline ComplexD adj(const ComplexD& r){ return(conjugate(r)); }
+accelerator_inline ComplexF adj(const ComplexF& r ){ return(conjugate(r)); }
 
+accelerator_inline RealF real(const RealF  & r){ return r; }
+accelerator_inline RealD real(const RealD  & r){ return r; }
+accelerator_inline RealF real(const ComplexF  & r){ return r.real(); }
+accelerator_inline RealD real(const ComplexD  & r){ return r.real(); }
+
+accelerator_inline RealF imag(const ComplexF  & r){ return r.imag(); }
+accelerator_inline RealD imag(const ComplexD  & r){ return r.imag(); }
+
+accelerator_inline ComplexD innerProduct(const ComplexD & l, const ComplexD & r) { return conjugate(l)*r; }
+accelerator_inline ComplexF innerProduct(const ComplexF & l, const ComplexF & r) { return conjugate(l)*r; }
+accelerator_inline RealD innerProduct(const RealD & l, const RealD & r) { return l*r; }
+accelerator_inline RealF innerProduct(const RealF & l, const RealF & r) { return l*r; }
+
+accelerator_inline ComplexD Reduce(const ComplexD& r){ return r; }
+accelerator_inline ComplexF Reduce(const ComplexF& r){ return r; }
+accelerator_inline RealD Reduce(const RealD& r){ return r; }
+accelerator_inline RealF Reduce(const RealF& r){ return r; }
+
+accelerator_inline RealD toReal(const ComplexD& r){ return r.real(); }
+accelerator_inline RealF toReal(const ComplexF& r){ return r.real(); }
+accelerator_inline RealD toReal(const RealD& r){ return r; }
+accelerator_inline RealF toReal(const RealF& r){ return r; }
   
+////////////////////////////////////////////////////////////////////////////////
+//Provide support functions for basic real and complex data types required by Grid
+//Single and double precision versions. Should be able to template this once only.
+////////////////////////////////////////////////////////////////////////////////
+accelerator_inline void mac (ComplexD * __restrict__ y,const ComplexD * __restrict__ a,const ComplexD *__restrict__ x){ *y = (*a) * (*x)+(*y); };
+accelerator_inline void mult(ComplexD * __restrict__ y,const ComplexD * __restrict__ l,const ComplexD *__restrict__ r){ *y = (*l) * (*r);}
+accelerator_inline void sub (ComplexD * __restrict__ y,const ComplexD * __restrict__ l,const ComplexD *__restrict__ r){ *y = (*l) - (*r);}
+accelerator_inline void add (ComplexD * __restrict__ y,const ComplexD * __restrict__ l,const ComplexD *__restrict__ r){ *y = (*l) + (*r);}
+// conjugate already supported for complex
   
-  ////////////////////////////////////////////////////////////////////////////////
-  //Provide support functions for basic real and complex data types required by Grid
-  //Single and double precision versions. Should be able to template this once only.
-  ////////////////////////////////////////////////////////////////////////////////
-  inline void mac (ComplexD * __restrict__ y,const ComplexD * __restrict__ a,const ComplexD *__restrict__ x){ *y = (*a) * (*x)+(*y); };
-  inline void mult(ComplexD * __restrict__ y,const ComplexD * __restrict__ l,const ComplexD *__restrict__ r){ *y = (*l) * (*r);}
-  inline void sub (ComplexD * __restrict__ y,const ComplexD * __restrict__ l,const ComplexD *__restrict__ r){ *y = (*l) - (*r);}
-  inline void add (ComplexD * __restrict__ y,const ComplexD * __restrict__ l,const ComplexD *__restrict__ r){ *y = (*l) + (*r);}
-  // conjugate already supported for complex
+accelerator_inline void mac (ComplexF * __restrict__ y,const ComplexF * __restrict__ a,const ComplexF *__restrict__ x){ *y = (*a) * (*x)+(*y); }
+accelerator_inline void mult(ComplexF * __restrict__ y,const ComplexF * __restrict__ l,const ComplexF *__restrict__ r){ *y = (*l) * (*r); }
+accelerator_inline void sub (ComplexF * __restrict__ y,const ComplexF * __restrict__ l,const ComplexF *__restrict__ r){ *y = (*l) - (*r); }
+accelerator_inline void add (ComplexF * __restrict__ y,const ComplexF * __restrict__ l,const ComplexF *__restrict__ r){ *y = (*l) + (*r); }
   
-  inline void mac (ComplexF * __restrict__ y,const ComplexF * __restrict__ a,const ComplexF *__restrict__ x){ *y = (*a) * (*x)+(*y); }
-  inline void mult(ComplexF * __restrict__ y,const ComplexF * __restrict__ l,const ComplexF *__restrict__ r){ *y = (*l) * (*r); }
-  inline void sub (ComplexF * __restrict__ y,const ComplexF * __restrict__ l,const ComplexF *__restrict__ r){ *y = (*l) - (*r); }
-  inline void add (ComplexF * __restrict__ y,const ComplexF * __restrict__ l,const ComplexF *__restrict__ r){ *y = (*l) + (*r); }
-  
-  //conjugate already supported for complex
-  
-  inline ComplexF timesI(const ComplexF &r)     { return(r*ComplexF(0.0,1.0));}
-  inline ComplexD timesI(const ComplexD &r)     { return(r*ComplexD(0.0,1.0));}
-  inline ComplexF timesMinusI(const ComplexF &r){ return(r*ComplexF(0.0,-1.0));}
-  inline ComplexD timesMinusI(const ComplexD &r){ return(r*ComplexD(0.0,-1.0));}
+//conjugate already supported for complex
+accelerator_inline ComplexF timesI(const ComplexF &r)     { return(r*ComplexF(0.0,1.0));}
+accelerator_inline ComplexD timesI(const ComplexD &r)     { return(r*ComplexD(0.0,1.0));}
+accelerator_inline ComplexF timesMinusI(const ComplexF &r){ return(r*ComplexF(0.0,-1.0));}
+accelerator_inline ComplexD timesMinusI(const ComplexD &r){ return(r*ComplexD(0.0,-1.0));}
 
-  // define projections to real and imaginay parts
-  inline ComplexF projReal(const ComplexF &r){return( ComplexF(std::real(r), 0.0));}
-  inline ComplexD projReal(const ComplexD &r){return( ComplexD(std::real(r), 0.0));}
-  inline ComplexF projImag(const ComplexF &r){return (ComplexF(std::imag(r), 0.0 ));}
-  inline ComplexD projImag(const ComplexD &r){return (ComplexD(std::imag(r), 0.0));}
+// define projections to real and imaginay parts
+accelerator_inline ComplexF projReal(const ComplexF &r){return( ComplexF(r.real(), 0.0));}
+accelerator_inline ComplexD projReal(const ComplexD &r){return( ComplexD(r.real(), 0.0));}
+accelerator_inline ComplexF projImag(const ComplexF &r){return (ComplexF(r.imag(), 0.0 ));}
+accelerator_inline ComplexD projImag(const ComplexD &r){return (ComplexD(r.imag(), 0.0));}
 
-  // define auxiliary functions for complex computations
-  inline void timesI(ComplexF &ret,const ComplexF &r)     { ret = timesI(r);}
-  inline void timesI(ComplexD &ret,const ComplexD &r)     { ret = timesI(r);}
-  inline void timesMinusI(ComplexF &ret,const ComplexF &r){ ret = timesMinusI(r);}
-  inline void timesMinusI(ComplexD &ret,const ComplexD &r){ ret = timesMinusI(r);}
+// define auxiliary functions for complex computations
+accelerator_inline void timesI(ComplexF &ret,const ComplexF &r)     { ret = timesI(r);}
+accelerator_inline void timesI(ComplexD &ret,const ComplexD &r)     { ret = timesI(r);}
+accelerator_inline void timesMinusI(ComplexF &ret,const ComplexF &r){ ret = timesMinusI(r);}
+accelerator_inline void timesMinusI(ComplexD &ret,const ComplexD &r){ ret = timesMinusI(r);}
   
-  inline void mac (RealD * __restrict__ y,const RealD * __restrict__ a,const RealD *__restrict__ x){  *y = (*a) * (*x)+(*y);}
-  inline void mult(RealD * __restrict__ y,const RealD * __restrict__ l,const RealD *__restrict__ r){ *y = (*l) * (*r);}
-  inline void sub (RealD * __restrict__ y,const RealD * __restrict__ l,const RealD *__restrict__ r){ *y = (*l) - (*r);}
-  inline void add (RealD * __restrict__ y,const RealD * __restrict__ l,const RealD *__restrict__ r){ *y = (*l) + (*r);}
+accelerator_inline void mac (RealD * __restrict__ y,const RealD * __restrict__ a,const RealD *__restrict__ x){ *y = (*a) * (*x)+(*y);}
+accelerator_inline void mult(RealD * __restrict__ y,const RealD * __restrict__ l,const RealD *__restrict__ r){ *y = (*l) * (*r);}
+accelerator_inline void sub (RealD * __restrict__ y,const RealD * __restrict__ l,const RealD *__restrict__ r){ *y = (*l) - (*r);}
+accelerator_inline void add (RealD * __restrict__ y,const RealD * __restrict__ l,const RealD *__restrict__ r){ *y = (*l) + (*r);}
   
-  inline void mac (RealF * __restrict__ y,const RealF * __restrict__ a,const RealF *__restrict__ x){  *y = (*a) * (*x)+(*y); }
-  inline void mult(RealF * __restrict__ y,const RealF * __restrict__ l,const RealF *__restrict__ r){ *y = (*l) * (*r); }
-  inline void sub (RealF * __restrict__ y,const RealF * __restrict__ l,const RealF *__restrict__ r){ *y = (*l) - (*r); }
-  inline void add (RealF * __restrict__ y,const RealF * __restrict__ l,const RealF *__restrict__ r){ *y = (*l) + (*r); }
+accelerator_inline void mac (RealF * __restrict__ y,const RealF * __restrict__ a,const RealF *__restrict__ x){  *y = (*a) * (*x)+(*y); }
+accelerator_inline void mult(RealF * __restrict__ y,const RealF * __restrict__ l,const RealF *__restrict__ r){ *y = (*l) * (*r); }
+accelerator_inline void sub (RealF * __restrict__ y,const RealF * __restrict__ l,const RealF *__restrict__ r){ *y = (*l) - (*r); }
+accelerator_inline void add (RealF * __restrict__ y,const RealF * __restrict__ l,const RealF *__restrict__ r){ *y = (*l) + (*r); }
   
-  inline void vstream(ComplexF &l, const ComplexF &r){ l=r;}
-  inline void vstream(ComplexD &l, const ComplexD &r){ l=r;}
-  inline void vstream(RealF &l, const RealF &r){ l=r;}
-  inline void vstream(RealD &l, const RealD &r){ l=r;}
+accelerator_inline void vstream(ComplexF &l, const ComplexF &r){ l=r;}
+accelerator_inline void vstream(ComplexD &l, const ComplexD &r){ l=r;}
+accelerator_inline void vstream(RealF &l, const RealF &r){ l=r;}
+accelerator_inline void vstream(RealD &l, const RealD &r){ l=r;}
   
+accelerator_inline ComplexD toComplex(const RealD &in) { return ComplexD(in);}
+accelerator_inline ComplexF toComplex(const RealF &in) { return ComplexF(in);}
   
-  class Zero{};
-  static Zero zero;
-  template<class itype> inline void zeroit(itype &arg){ arg=zero;};
-  template<>            inline void zeroit(ComplexF &arg){ arg=0; };
-  template<>            inline void zeroit(ComplexD &arg){ arg=0; };
-  template<>            inline void zeroit(RealF &arg){ arg=0; };
-  template<>            inline void zeroit(RealD &arg){ arg=0; };
-  
+class Zero{};
+//static Zero Zero();
+template<class itype> accelerator_inline void zeroit(itype &arg)   { arg=Zero();};
+template<>            accelerator_inline void zeroit(ComplexF &arg){ arg=0; };
+template<>            accelerator_inline void zeroit(ComplexD &arg){ arg=0; };
+template<>            accelerator_inline void zeroit(RealF &arg)   { arg=0; };
+template<>            accelerator_inline void zeroit(RealD &arg)   { arg=0; };
 
-  //////////////////////////////////////////////////////////
-  // Permute
-  // Permute 0 every ABCDEFGH -> BA DC FE HG
-  // Permute 1 every ABCDEFGH -> CD AB GH EF
-  // Permute 2 every ABCDEFGH -> EFGH ABCD
-  // Permute 3 possible on longer iVector lengths (512bit = 8 double = 16 single)
-  // Permute 4 possible on half precision @512bit vectors.
-  //
-  // Defined inside SIMD specialization files
-  //////////////////////////////////////////////////////////
-  template<class VectorSIMD>
-    inline void Gpermute(VectorSIMD &y,const VectorSIMD &b,int perm);
+// More limited Integer support  
+accelerator_inline Integer Reduce(const Integer& r){ return r; }
+accelerator_inline void mac (Integer * __restrict__ y,const Integer * __restrict__ a,const Integer *__restrict__ x){  *y = (*a) * (*x)+(*y); }
+accelerator_inline void mult(Integer * __restrict__ y,const Integer * __restrict__ l,const Integer *__restrict__ r){ *y = (*l) * (*r); }
+accelerator_inline void sub (Integer * __restrict__ y,const Integer * __restrict__ l,const Integer *__restrict__ r){ *y = (*l) - (*r); }
+accelerator_inline void add (Integer * __restrict__ y,const Integer * __restrict__ l,const Integer *__restrict__ r){ *y = (*l) + (*r); }
+accelerator_inline void vstream(Integer &l, const RealD &r){ l=r;}
+template<>            accelerator_inline void zeroit(Integer &arg)   { arg=0; };
 
-};
+accelerator_inline Integer mod (Integer a,Integer y) { return a%y;}
+accelerator_inline Integer div (Integer a,Integer y) { return a/y;}
+//accelerator_inline Integer abs (Integer &a) { return a%y;}
+
+//////////////////////////////////////////////////////////
+// Permute
+// Permute 0 every ABCDEFGH -> BA DC FE HG
+// Permute 1 every ABCDEFGH -> CD AB GH EF
+// Permute 2 every ABCDEFGH -> EFGH ABCD
+// Permute 3 possible on longer iVector lengths (512bit = 8 double = 16 single)
+// Permute 4 possible on half precision @512bit vectors.
+//
+// Defined inside SIMD specialization files
+//////////////////////////////////////////////////////////
+template<class VectorSIMD>
+accelerator_inline void Gpermute(VectorSIMD &y,const VectorSIMD &b,int perm);
+
+NAMESPACE_END(Grid);
 
 #include <Grid/simd/Grid_vector_types.h>
 #include <Grid/simd/Grid_vector_unops.h>
 
-namespace Grid {
-  // Default precision
+NAMESPACE_BEGIN(Grid);
+// Default precision
 #ifdef GRID_DEFAULT_PRECISION_DOUBLE
-  typedef vRealD vReal;
-  typedef vComplexD vComplex;
+typedef vRealD vReal;
+typedef vComplexD vComplex;
 #else
-  typedef vRealF vReal;
-  typedef vComplexF vComplex;
+typedef vRealF vReal;
+typedef vComplexF vComplex;
 #endif
 
  
-  inline std::ostream& operator<< (std::ostream& stream, const vComplexF &o){
-    int nn=vComplexF::Nsimd();
-    std::vector<ComplexF,alignedAllocator<ComplexF> > buf(nn);
-    vstore(o,&buf[0]);
-    stream<<"<";
-    for(int i=0;i<nn;i++){
-      stream<<buf[i];
-      if(i<nn-1) stream<<",";
-    }
-    stream<<">";
-    return stream;
+inline std::ostream& operator<< (std::ostream& stream, const vComplexF &o){
+  int nn=vComplexF::Nsimd();
+  std::vector<ComplexF,alignedAllocator<ComplexF> > buf(nn);
+  vstore(o,&buf[0]);
+  stream<<"<";
+  for(int i=0;i<nn;i++){
+    stream<<buf[i];
+    if(i<nn-1) stream<<",";
   }
- 
-  inline std::ostream& operator<< (std::ostream& stream, const vComplexD &o){
-    int nn=vComplexD::Nsimd();
-    std::vector<ComplexD,alignedAllocator<ComplexD> > buf(nn);
-    vstore(o,&buf[0]);
-    stream<<"<";
-    for(int i=0;i<nn;i++){
-      stream<<buf[i];
-      if(i<nn-1) stream<<",";
-    }
-    stream<<">";
-    return stream;
-  }
-
-  inline std::ostream& operator<< (std::ostream& stream, const vRealF &o){
-    int nn=vRealF::Nsimd();
-    std::vector<RealF,alignedAllocator<RealF> > buf(nn);
-    vstore(o,&buf[0]);
-    stream<<"<";
-    for(int i=0;i<nn;i++){
-      stream<<buf[i];
-      if(i<nn-1) stream<<",";
-    }
-    stream<<">";
-    return stream;
-  }
-
-  inline std::ostream& operator<< (std::ostream& stream, const vRealD &o){
-    int nn=vRealD::Nsimd();
-    std::vector<RealD,alignedAllocator<RealD> > buf(nn);
-    vstore(o,&buf[0]);
-    stream<<"<";
-    for(int i=0;i<nn;i++){
-      stream<<buf[i];
-      if(i<nn-1) stream<<",";
-    }
-    stream<<">";
-    return stream;
-  }
-  inline std::ostream& operator<< (std::ostream& stream, const vInteger &o){
-    int nn=vInteger::Nsimd();
-    std::vector<Integer,alignedAllocator<Integer> > buf(nn);
-    vstore(o,&buf[0]);
-    stream<<"<";
-    for(int i=0;i<nn;i++){
-      stream<<buf[i];
-      if(i<nn-1) stream<<",";
-    }
-    stream<<">";
-    return stream;
-  }
-
-
+  stream<<">";
+  return stream;
 }
+ 
+inline std::ostream& operator<< (std::ostream& stream, const vComplexD &o){
+  int nn=vComplexD::Nsimd();
+  std::vector<ComplexD,alignedAllocator<ComplexD> > buf(nn);
+  vstore(o,&buf[0]);
+  stream<<"<";
+  for(int i=0;i<nn;i++){
+    stream<<buf[i];
+    if(i<nn-1) stream<<",";
+  }
+  stream<<">";
+  return stream;
+}
+
+inline std::ostream& operator<< (std::ostream& stream, const vRealF &o){
+  int nn=vRealF::Nsimd();
+  std::vector<RealF,alignedAllocator<RealF> > buf(nn);
+  vstore(o,&buf[0]);
+  stream<<"<";
+  for(int i=0;i<nn;i++){
+    stream<<buf[i];
+    if(i<nn-1) stream<<",";
+  }
+  stream<<">";
+  return stream;
+}
+
+inline std::ostream& operator<< (std::ostream& stream, const vRealD &o){
+  int nn=vRealD::Nsimd();
+  std::vector<RealD,alignedAllocator<RealD> > buf(nn);
+  vstore(o,&buf[0]);
+  stream<<"<";
+  for(int i=0;i<nn;i++){
+    stream<<buf[i];
+    if(i<nn-1) stream<<",";
+  }
+  stream<<">";
+  return stream;
+}
+inline std::ostream& operator<< (std::ostream& stream, const vInteger &o){
+  int nn=vInteger::Nsimd();
+  std::vector<Integer,alignedAllocator<Integer> > buf(nn);
+  vstore(o,&buf[0]);
+  stream<<"<";
+  for(int i=0;i<nn;i++){
+    stream<<buf[i];
+    if(i<nn-1) stream<<",";
+  }
+  stream<<">";
+  return stream;
+}
+
+NAMESPACE_END(Grid)
 #endif
diff --git a/Grid/simd/l1p.h b/Grid/simd/l1p.h
index 8e43fdbd..f13d69b0 100644
--- a/Grid/simd/l1p.h
+++ b/Grid/simd/l1p.h
@@ -1,26 +1,26 @@
 #pragma once
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 // L1p optimisation 
-inline void bgq_l1p_optimisation(int mode)
+accelerator_inline void bgq_l1p_optimisation(int mode)
 {
 #ifdef QPX
 #undef L1P_CFG_PF_USR
 #define L1P_CFG_PF_USR  (0x3fde8000108ll)   /*  (64 bit reg, 23 bits wide, user/unpriv) */
-
+  
   uint64_t cfg_pf_usr;
   if ( mode ) { 
     cfg_pf_usr =
-        L1P_CFG_PF_USR_ifetch_depth(0)       
+      L1P_CFG_PF_USR_ifetch_depth(0)       
       | L1P_CFG_PF_USR_ifetch_max_footprint(1)   
       | L1P_CFG_PF_USR_pf_stream_est_on_dcbt 
       | L1P_CFG_PF_USR_pf_stream_establish_enable
       | L1P_CFG_PF_USR_pf_stream_optimistic
       | L1P_CFG_PF_USR_pf_adaptive_throttle(0xF) ;
     //    if ( sizeof(Float) == sizeof(double) ) {
-      cfg_pf_usr |=  L1P_CFG_PF_USR_dfetch_depth(2)| L1P_CFG_PF_USR_dfetch_max_footprint(3)   ;
-      //    } else {
-      //      cfg_pf_usr |=  L1P_CFG_PF_USR_dfetch_depth(1)| L1P_CFG_PF_USR_dfetch_max_footprint(2)   ;
-      //    }
+    cfg_pf_usr |=  L1P_CFG_PF_USR_dfetch_depth(2)| L1P_CFG_PF_USR_dfetch_max_footprint(3)   ;
+    //    } else {
+    //      cfg_pf_usr |=  L1P_CFG_PF_USR_dfetch_depth(1)| L1P_CFG_PF_USR_dfetch_max_footprint(2)   ;
+    //    }
   } else { 
     cfg_pf_usr = L1P_CFG_PF_USR_dfetch_depth(1)
       | L1P_CFG_PF_USR_dfetch_max_footprint(2)   
@@ -34,4 +34,4 @@ inline void bgq_l1p_optimisation(int mode)
   *((uint64_t *)L1P_CFG_PF_USR) = cfg_pf_usr;
 #endif
 }
-}
+NAMESPACE_END(Grid)
diff --git a/Grid/stencil/GeneralLocalStencil.h b/Grid/stencil/GeneralLocalStencil.h
new file mode 100644
index 00000000..f14d8096
--- /dev/null
+++ b/Grid/stencil/GeneralLocalStencil.h
@@ -0,0 +1,141 @@
+/*************************************************************************************
+
+     Grid physics library, www.github.com/paboyle/Grid 
+
+     Source file: ./lib/GeneralLocalStencil.h
+
+     Copyright (C) 2019
+
+ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+     This program is free software; you can redistribute it and/or modify
+     it under the terms of the GNU General Public License as published by
+     the Free Software Foundation; either version 2 of the License, or
+     (at your option) any later version.
+
+     This program is distributed in the hope that it will be useful,
+     but WITHOUT ANY WARRANTY; without even the implied warranty of
+     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+     GNU General Public License for more details.
+
+     You should have received a copy of the GNU General Public License along
+     with this program; if not, write to the Free Software Foundation, Inc.,
+     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+     See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+NAMESPACE_BEGIN(Grid);
+
+// Share with Cartesian Stencil
+struct GeneralStencilEntry { 
+  uint64_t _offset;            // 4 bytes 
+  uint8_t _permute;            // 1 bytes // Horrible alignment properties
+};
+// Could pack to 8 + 4 + 4 = 128 bit and use 
+
+class GeneralLocalStencilView {
+ public:
+  ////////////////////////////////////////
+  // Basic Grid and stencil info
+  ////////////////////////////////////////
+  int                               _npoints; // Move to template param?
+  GeneralStencilEntry*  _entries_p;
+
+  accelerator_inline GeneralStencilEntry * GetEntry(int point,int osite) { 
+    return & this->_entries_p[point+this->_npoints*osite]; 
+  }
+
+};
+////////////////////////////////////////
+// The Stencil Class itself
+////////////////////////////////////////
+class GeneralLocalStencil : public GeneralLocalStencilView {
+public:
+  typedef GeneralLocalStencilView View_type;
+
+protected:
+  GridBase *                        _grid;
+
+public: 
+  GridBase *Grid(void) const { return _grid; }
+
+  View_type View(void) const {
+    View_type accessor(*( (View_type *) this));
+    return accessor;
+  }
+
+  // Resident in managed memory
+  Vector<GeneralStencilEntry>  _entries; 
+
+  GeneralLocalStencil(GridBase *grid, const std::vector<Coordinate> &shifts)
+  {
+    int npoints = shifts.size();
+    int osites  = grid->oSites();
+    
+    this->_grid    = grid;
+    this->_npoints = npoints;
+    this->_entries.resize(npoints* osites);
+    this->_entries_p = &_entries[0];
+
+
+    Coordinate Coor;
+    Coordinate NbrCoor;
+    for(Integer site=0;site<osites;site++){
+      for(Integer ii=0;ii<npoints;ii++){
+	Integer lex = site*npoints+ii;
+	GeneralStencilEntry SE;
+	////////////////////////////////////////////////
+	// Outer index of neighbour Offset calculation
+	////////////////////////////////////////////////
+	grid->oCoorFromOindex(Coor,site);
+	for(int d=0;d<Coor.size();d++){
+	  int rd = grid->_rdimensions[d];
+	  NbrCoor[d] = (Coor[d] + shifts[ii][d] + rd )%rd;
+	}
+	SE._offset      = grid->oIndexReduced(NbrCoor);
+
+	////////////////////////////////////////////////
+	// Inner index permute calculation
+	// Simpler version using icoor calculation
+	////////////////////////////////////////////////
+	SE._permute =0;
+	for(int d=0;d<Coor.size();d++){
+
+	  int fd = grid->_fdimensions[d];
+	  int rd = grid->_rdimensions[d];
+	  int ly = grid->_simd_layout[d];
+
+	  assert((ly==1)||(ly==2));
+
+	  int shift = (shifts[ii][d]+fd)%fd;  // make it strictly positive 0.. L-1
+	  int x = Coor[d];                // x in [0... rd-1] as an oSite 
+
+	  int permute_dim  = grid->PermuteDim(d);
+	  int permute_slice=0;
+	  if(permute_dim){    
+	    int  num = shift%rd; // Slice within dest osite cell of slice zero
+	    int wrap = shift/rd; // Number of osite local volume cells crossed through
+                                  // x+num < rd dictates whether we are in same permute state as slice 0
+	    if ( x< rd-num ) permute_slice=wrap;
+	    else             permute_slice=(wrap+1)%ly;
+	  }
+	  if ( permute_slice ) {
+	    int ptype       =grid->PermuteType(d);
+	    uint8_t mask    =grid->Nsimd() >> (ptype + 1);		
+	    SE._permute    |= mask;
+	  }
+	}	
+	////////////////////////////////////////////////
+	// Store in look up table
+	////////////////////////////////////////////////
+	this->_entries[lex] = SE;
+      }
+    }      
+  }
+  
+};
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/stencil/Lebesgue.cc b/Grid/stencil/Lebesgue.cc
index 2880e4b6..656ecca8 100644
--- a/Grid/stencil/Lebesgue.cc
+++ b/Grid/stencil/Lebesgue.cc
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,12 +24,12 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #include <Grid/GridCore.h>
 #include <algorithm>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 int LebesgueOrder::UseLebesgueOrder;
 #ifdef KNL
@@ -62,19 +62,19 @@ LebesgueOrder::LebesgueOrder(GridBase *_grid)
 }
 void LebesgueOrder::ThreadInterleave(void)
 {
-  std::vector<IndexInteger> reorder = _LebesgueReorder;
-  std::vector<IndexInteger> throrder;
+  Vector<IndexInteger> reorder = _LebesgueReorder;
+  Vector<IndexInteger> throrder;
   int vol = _LebesgueReorder.size();
   int threads = GridThread::GetThreads();
   int blockbits=3;
-  int blocklen = 8;
-  int msk      = 0x7;
+  //  int blocklen = 8;
+  //  int msk      = 0x7;
 
   for(int t=0;t<threads;t++){
     for(int ss=0;ss<vol;ss++){
-       if ( ( ss >> blockbits) % threads == t ) { 
-         throrder.push_back(reorder[ss]);
-       }
+      if ( ( ss >> blockbits) % threads == t ) { 
+	throrder.push_back(reorder[ss]);
+      }
     }
   }
   _LebesgueReorder = throrder;
@@ -100,9 +100,9 @@ void LebesgueOrder::CartesianBlocking(void)
   assert(ND==4);
   assert(ND==Block.size());
 
-  std::vector<IndexInteger> dims(ND);
-  std::vector<IndexInteger> xo(ND,0);
-  std::vector<IndexInteger> xi(ND,0);
+  Coordinate dims(ND);
+  Coordinate xo(ND,0);
+  Coordinate xi(ND,0);
 
   for(IndexInteger mu=0;mu<ND;mu++){
     dims[mu] = grid->_rdimensions[mu];
@@ -112,9 +112,9 @@ void LebesgueOrder::CartesianBlocking(void)
 };
 
 void LebesgueOrder::IterateO(int ND,int dim,
-	      std::vector<IndexInteger> & xo,
-	      std::vector<IndexInteger> & xi,
-	      std::vector<IndexInteger> &dims)
+			     Coordinate & xo,
+			     Coordinate & xi,
+			     Coordinate &dims)
 {
   for(xo[dim]=0;xo[dim]<dims[dim];xo[dim]+=Block[dim]){
     if ( dim > 0 ) {
@@ -126,21 +126,21 @@ void LebesgueOrder::IterateO(int ND,int dim,
 };
 
 void LebesgueOrder::IterateI(int ND,
-	      int dim,
-	      std::vector<IndexInteger> & xo,
-	      std::vector<IndexInteger> & xi,
-	      std::vector<IndexInteger> &dims)
+			     int dim,
+			     Coordinate & xo,
+			     Coordinate & xi,
+			     Coordinate &dims)
 {
-  std::vector<IndexInteger> x(ND);
+  Coordinate x(ND);
   for(xi[dim]=0;xi[dim]<std::min(dims[dim]-xo[dim],Block[dim]);xi[dim]++){
     if ( dim > 0 ) {
       IterateI(ND,dim-1,xo,xi,dims);
     } else {
       for(int d=0;d<ND;d++){
 	x[d]=xi[d]+xo[d];
-//	std::cout << x[d]<<" ";
+	//	std::cout << x[d]<<" ";
       }
-//      std::cout << "\n";
+      //      std::cout << "\n";
       IndexInteger index;
       Lexicographic::IndexFromCoor(x,index,grid->_rdimensions);
       _LebesgueReorder.push_back(index);
@@ -157,8 +157,8 @@ void LebesgueOrder::ZGraph(void)
   const IndexInteger one=1;
 
   IndexInteger ND = grid->_ndimension;
-  std::vector<IndexInteger> dims(ND);
-  std::vector<IndexInteger> adims(ND);
+  Coordinate dims(ND);
+  Coordinate adims(ND);
   std::vector<std::vector<IndexInteger> > bitlist(ND);
   
   for(IndexInteger mu=0;mu<ND;mu++){
@@ -227,16 +227,17 @@ void LebesgueOrder::ZGraph(void)
   assert( _LebesgueReorder.size() == vol );
 
   /*
-  std::vector<int> coor(4);
-  for(IndexInteger asite=0;asite<vol;asite++){
+    std::vector<int> coor(4);
+    for(IndexInteger asite=0;asite<vol;asite++){
     grid->oCoorFromOindex (coor,_LebesgueReorder[asite]);
-      std::cout << " site "<<asite << "->" << _LebesgueReorder[asite]<< " = ["
-		<< coor[0]<<","
-		<< coor[1]<<","
-		<< coor[2]<<","
-		<< coor[3]<<"]"
-		<<std::endl;
-  }
+    std::cout << " site "<<asite << "->" << _LebesgueReorder[asite]<< " = ["
+    << coor[0]<<","
+    << coor[1]<<","
+    << coor[2]<<","
+    << coor[3]<<"]"
+    <<std::endl;
+    }
   */
 }
-}
+NAMESPACE_END(Grid);
+
diff --git a/Grid/stencil/Lebesgue.h b/Grid/stencil/Lebesgue.h
index 7db0cc6b..25fa772e 100644
--- a/Grid/stencil/Lebesgue.h
+++ b/Grid/stencil/Lebesgue.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,57 +24,58 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_LEBESGUE_H
 #define GRID_LEBESGUE_H
 
 #include<vector>
 
 // Lebesgue, Morton, Z-graph ordering assistance
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
   
-  class LebesgueOrder { 
-  public:
+class LebesgueOrder { 
+public:
 
-    typedef int32_t IndexInteger;
-    
-    static int UseLebesgueOrder;
-    GridBase *grid;
+  typedef int32_t IndexInteger;
+  static int UseLebesgueOrder;
+  GridBase *grid;
 
-  public:
-    LebesgueOrder(GridBase *_grid);
+public:
+  LebesgueOrder(GridBase *_grid);
 
-    inline IndexInteger Reorder(IndexInteger ss) { 
-      return _LebesgueReorder[ss] ;
-    };
+  inline IndexInteger Reorder(IndexInteger ss) { 
+    return _LebesgueReorder[ss] ;
+  };
 
-    ////////////////////////////
-    // Space filling fractal for cache oblivious
-    ////////////////////////////
-    void ZGraph(void);
-    IndexInteger alignup(IndexInteger n);
+  ////////////////////////////
+  // Space filling fractal for cache oblivious
+  ////////////////////////////
+  void ZGraph(void);
+  IndexInteger alignup(IndexInteger n);
 
-    /////////////////////////////////
-    // Cartesian stencil blocking strategy
-    /////////////////////////////////
-    static std::vector<int> Block;
-    void NoBlocking(void);
-    void CartesianBlocking(void);
-    void IterateO(int ND,int dim,
-		  std::vector<IndexInteger> & xo,
-		  std::vector<IndexInteger> & xi,
-		  std::vector<IndexInteger> &dims);
-    void IterateI(int ND,int dim,
-		  std::vector<IndexInteger> & xo,
-		  std::vector<IndexInteger> & xi,
-		  std::vector<IndexInteger> &dims);
+  /////////////////////////////////
+  // Cartesian stencil blocking strategy
+  /////////////////////////////////
+  static std::vector<int> Block;
+  void NoBlocking(void);
+  void CartesianBlocking(void);
+  void IterateO(int ND,int dim,
+		Coordinate & xo,
+		Coordinate & xi,
+		Coordinate &dims);
+  void IterateI(int ND,int dim,
+		Coordinate & xo,
+		Coordinate & xi,
+		Coordinate &dims);
 
-    void ThreadInterleave(void);
+  void ThreadInterleave(void);
 
-  private:
-    std::vector<IndexInteger> _LebesgueReorder;
+private:
+  Vector<IndexInteger> _LebesgueReorder;
+
+};    
+
+NAMESPACE_END(Grid);
 
-  };    
-}
 #endif
diff --git a/Grid/stencil/SimpleCompressor.h b/Grid/stencil/SimpleCompressor.h
index 58c3bcd2..be7c89c0 100644
--- a/Grid/stencil/SimpleCompressor.h
+++ b/Grid/stencil/SimpleCompressor.h
@@ -1,29 +1,30 @@
 #ifndef _STENCIL_SIMPLE_COMPRESSOR_H_
 #define _STENCIL_SIMPLE_COMPRESSOR_H_
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 template<class vobj>
 class SimpleCompressor {
 public:
   void Point(int) {};
-  inline int  CommDatumSize(void) { return sizeof(vobj); }
-  inline bool DecompressionStep(void) { return false; }
-  inline void Compress(vobj *buf,int o,const vobj &in) { buf[o]=in; }
-  inline void Exchange(vobj *mp,vobj *vp0,vobj *vp1,Integer type,Integer o){
+  accelerator_inline int  CommDatumSize(void) { return sizeof(vobj); }
+  accelerator_inline bool DecompressionStep(void) { return false; }
+  template<class cobj> accelerator_inline void Compress(cobj *buf,int o,const cobj &in) { buf[o]=in; }
+  accelerator_inline void Exchange(vobj *mp,vobj *vp0,vobj *vp1,Integer type,Integer o){
     exchange(mp[2*o],mp[2*o+1],vp0[o],vp1[o],type);
   }
-  inline void Decompress(vobj *out,vobj *in, int o){ assert(0); }
-  inline void CompressExchange(vobj *out0,vobj *out1,const vobj *in,
+  accelerator_inline void Decompress(vobj *out,vobj *in, int o){ assert(0); }
+  accelerator_inline void CompressExchange(vobj *out0,vobj *out1,const vobj *in,
 			       int j,int k, int m,int type){
     exchange(out0[j],out1[j],in[k],in[m],type);
   }
   // For cshift. Cshift should drop compressor coupling altogether 
   // because I had to decouple the code from the Stencil anyway
-  inline vobj operator() (const vobj &arg) {
+  accelerator_inline vobj operator() (const vobj &arg) {
     return arg;
   }
 };
 
-}
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/stencil/Stencil.cc b/Grid/stencil/Stencil.cc
index e04a5360..5b1bb2ea 100644
--- a/Grid/stencil/Stencil.cc
+++ b/Grid/stencil/Stencil.cc
@@ -1,4 +1,4 @@
-   /*************************************************************************************
+/*************************************************************************************
 
      Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,14 +23,14 @@
      51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
      See the full license in the file "LICENSE" in the top level distribution directory
-     *************************************************************************************/
-     /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #include <Grid/GridCore.h>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 void Gather_plane_table_compute (GridBase *grid,int dimension,int plane,int cbmask,
-					int off,std::vector<std::pair<int,int> > & table)
+				 int off,Vector<std::pair<int,int> > & table)
 {
   table.resize(0);
 
@@ -52,18 +52,19 @@ void Gather_plane_table_compute (GridBase *grid,int dimension,int plane,int cbma
       }
     }
   } else { 
-     int bo=0;
-     table.resize(e1*e2/2);
-     for(int n=0;n<e1;n++){
-       for(int b=0;b<e2;b++){
-	 int o  = n*stride;
-	 int ocb=1<<grid->CheckerBoardFromOindexTable(o+b);
-	 if ( ocb &cbmask ) {
-	   table[bo]=std::pair<int,int>(bo,o+b); bo++;
-	 }
-       }
-     }
+    int bo=0;
+    table.resize(e1*e2/2);
+    for(int n=0;n<e1;n++){
+      for(int b=0;b<e2;b++){
+	int o  = n*stride;
+	int ocb=1<<grid->CheckerBoardFromOindexTable(o+b);
+	if ( ocb &cbmask ) {
+	  table[bo]=std::pair<int,int>(bo,o+b); bo++;
+	}
+      }
+    }
   }
 }
 
-}
+NAMESPACE_END(Grid);
+
diff --git a/Grid/stencil/Stencil.h b/Grid/stencil/Stencil.h
index 57952a26..b361a19d 100644
--- a/Grid/stencil/Stencil.h
+++ b/Grid/stencil/Stencil.h
@@ -23,51 +23,59 @@
      51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
      See the full license in the file "LICENSE" in the top level distribution directory
-     *************************************************************************************/
-     /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_STENCIL_H
 #define GRID_STENCIL_H
 
+#define STENCIL_MAX (16)
+
 #include <Grid/stencil/SimpleCompressor.h>   // subdir aggregate
 #include <Grid/stencil/Lebesgue.h>   // subdir aggregate
 
- //////////////////////////////////////////////////////////////////////////////////////////
- // Must not lose sight that goal is to be able to construct really efficient
- // gather to a point stencil code. CSHIFT is not the best way, so need
- // additional stencil support.
- //
- // Stencil based code will exchange haloes and use a table lookup for neighbours.
- // This will be done with generality to allow easier efficient implementations.
- // Overlap of comms and compute is enabled by tabulating off-node connected,
- // 
- // Generic services
- // 0) Prebuild neighbour tables
- // 1) Compute sizes of all haloes/comms buffers; allocate them.
- // 2) Gather all faces, and communicate.
- // 3) Loop over result sites, giving nbr index/offnode info for each
- // 
- //////////////////////////////////////////////////////////////////////////////////////////
+//////////////////////////////////////////////////////////////////////////////////////////
+// Must not lose sight that goal is to be able to construct really efficient
+// gather to a point stencil code. CSHIFT is not the best way, so need
+// additional stencil support.
+//
+// Stencil based code will exchange haloes and use a table lookup for neighbours.
+// This will be done with generality to allow easier efficient implementations.
+// Overlap of comms and compute is enabled by tabulating off-node connected,
+// 
+// Generic services
+// 0) Prebuild neighbour tables
+// 1) Compute sizes of all haloes/comms buffers; allocate them.
+// 2) Gather all faces, and communicate.
+// 3) Loop over result sites, giving nbr index/offnode info for each
+// 
+//////////////////////////////////////////////////////////////////////////////////////////
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 ///////////////////////////////////////////////////////////////////
 // Gather for when there *is* need to SIMD split with compression
 ///////////////////////////////////////////////////////////////////
+
 void Gather_plane_table_compute (GridBase *grid,int dimension,int plane,int cbmask,
-					int off,std::vector<std::pair<int,int> > & table);
+				 int off,Vector<std::pair<int,int> > & table);
 
 template<class vobj,class cobj,class compressor> 
-void Gather_plane_simple_table (std::vector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,cobj *buffer,compressor &compress, int off,int so)   __attribute__((noinline));
+void Gather_plane_simple_table (Vector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,cobj *buffer,compressor &compress, int off,int so)   __attribute__((noinline));
 
 template<class vobj,class cobj,class compressor> 
-void Gather_plane_simple_table (std::vector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,cobj *buffer,compressor &compress, int off,int so)
+void Gather_plane_simple_table (Vector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,cobj *buffer,compressor &compress, int off,int so)
 {
   int num=table.size();
-  parallel_for(int i=0;i<num;i++){
-    compress.Compress(&buffer[off],table[i].first,rhs._odata[so+table[i].second]);
-  }
-// Further optimisatoin: i) streaming store the result
-//                       ii) software prefetch the first element of the next table entry
+  std::pair<int,int> *table_v = & table[0];
+  auto rhs_v = rhs.View();
+  accelerator_forNB( i,num, vobj::Nsimd(), {
+    typedef decltype(coalescedRead(buffer[0])) compressed_t;
+    compressed_t   tmp_c;
+    uint64_t o = table_v[i].first;
+    compress.Compress(&tmp_c,0,rhs_v(so+table_v[i].second));
+    coalescedWrite(buffer[off+o],tmp_c);
+  });
+// Further optimisatoin: i) software prefetch the first element of the next table entry, prefetch the table
 }
 
 ///////////////////////////////////////////////////////////////////
@@ -75,42 +83,112 @@ void Gather_plane_simple_table (std::vector<std::pair<int,int> >& table,const La
 ///////////////////////////////////////////////////////////////////
 template<class cobj,class vobj,class compressor>
 void Gather_plane_exchange_table(const Lattice<vobj> &rhs,
-				 std::vector<cobj *> pointers,int dimension,int plane,int cbmask,compressor &compress,int type) __attribute__((noinline));
+				 Vector<cobj *> pointers,int dimension,int plane,int cbmask,compressor &compress,int type) __attribute__((noinline));
 
 template<class cobj,class vobj,class compressor>
-void Gather_plane_exchange_table(std::vector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,
-				 std::vector<cobj *> pointers,int dimension,int plane,int cbmask,
+void Gather_plane_exchange_table(Vector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,
+				 Vector<cobj *> pointers,int dimension,int plane,int cbmask,
 				 compressor &compress,int type)
 {
   assert( (table.size()&0x1)==0);
   int num=table.size()/2;
-  int so  = plane*rhs._grid->_ostride[dimension]; // base offset for start of plane 
-  parallel_for(int j=0;j<num;j++){
-    compress.CompressExchange(&pointers[0][0],&pointers[1][0],&rhs._odata[0],
-			      j,so+table[2*j].second,so+table[2*j+1].second,type);
-  }
+  int so  = plane*rhs.Grid()->_ostride[dimension]; // base offset for start of plane 
+
+  auto rhs_v = rhs.View();
+  auto p0=&pointers[0][0];
+  auto p1=&pointers[1][0];
+  auto tp=&table[0];
+  accelerator_forNB(j, num, 1, { 
+      compress.CompressExchange(p0,p1, &rhs_v[0], j,
+			      so+tp[2*j  ].second,
+			      so+tp[2*j+1].second,
+			      type);
+  });
 }
 
- struct StencilEntry { 
-   uint64_t _offset;
-   uint64_t _byte_offset;
-   uint16_t _is_local;
-   uint16_t _permute;
-   uint16_t _around_the_world; //256 bits, 32 bytes, 1/2 cacheline
-   uint16_t _pad;
- };
+struct StencilEntry { 
+#ifdef GRID_NVCC
+  uint64_t _byte_offset;       // 8 bytes 
+  uint32_t _offset;            // 4 bytes 
+#else
+  uint64_t _byte_offset;       // 8 bytes 
+  uint64_t _offset;            // 8 bytes (8 ever required?)
+#endif
+  uint8_t _is_local;           // 1 bytes 
+  uint8_t _permute;            // 1 bytes
+  uint8_t _around_the_world;   // 1 bytes
+  uint8_t _pad;   // 1 bytes
+};
+// Could pack to 8 + 4 + 4 = 128 bit and use 
 
+template<class vobj,class cobj,class Parameters>
+class CartesianStencilView {
+ public:
+  typedef AcceleratorVector<int,STENCIL_MAX> StencilVector;
+
+  // Stencil runs along coordinate axes only; NO diagonal fill in.
+  ////////////////////////////////////////
+  // Basic Grid and stencil info
+  ////////////////////////////////////////
+  int                               _checkerboard;
+  int                               _npoints; // Move to template param?
+  StencilVector _directions;
+  StencilVector _distances;
+  StencilVector _comm_buf_size;
+  StencilVector _permute_type;
+  StencilVector same_node;
+  Coordinate                         _simd_layout;
+  Parameters    parameters;
+  StencilEntry*  _entries_p;
+  cobj* u_recv_buf_p;
+  cobj* u_send_buf_p;
+
+  accelerator_inline cobj *CommBuf(void) { return u_recv_buf_p; }
+
+  accelerator_inline int GetNodeLocal(int osite,int point) { 
+    return this->_entries_p[point+this->_npoints*osite]._is_local;
+  }
+  accelerator_inline StencilEntry * GetEntry(int &ptype,int point,int osite) { 
+    ptype = this->_permute_type[point]; return & this->_entries_p[point+this->_npoints*osite]; 
+  }
+
+  accelerator_inline uint64_t GetInfo(int &ptype,int &local,int &perm,int point,int ent,uint64_t base) {
+    uint64_t cbase = (uint64_t)&u_recv_buf_p[0];
+    local = this->_entries_p[ent]._is_local;
+    perm  = this->_entries_p[ent]._permute;
+    if (perm)  ptype = this->_permute_type[point]; 
+    if (local) {
+      return  base + this->_entries_p[ent]._byte_offset;
+    } else {
+      return cbase + this->_entries_p[ent]._byte_offset;
+    }
+  }
+
+  accelerator_inline uint64_t GetPFInfo(int ent,uint64_t base) {
+    uint64_t cbase = (uint64_t)&u_recv_buf_p[0];
+    int local = this->_entries_p[ent]._is_local;
+    if (local) return  base + this->_entries_p[ent]._byte_offset;
+    else       return cbase + this->_entries_p[ent]._byte_offset;
+  }
+
+  accelerator_inline void iCoorFromIindex(Coordinate &coor,int lane) 
+  {
+    Lexicographic::CoorFromIndex(coor,lane,this->_simd_layout);
+  }
+
+};
 ////////////////////////////////////////
 // The Stencil Class itself
 ////////////////////////////////////////
-template<class vobj,class cobj>
-class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal fill in.
- public:
+template<class vobj,class cobj,class Parameters>
+class CartesianStencil : public CartesianStencilView<vobj,cobj,Parameters> { // Stencil runs along coordinate axes only; NO diagonal fill in.
+public:
 
   typedef typename cobj::vector_type vector_type;
   typedef typename cobj::scalar_type scalar_type;
   typedef typename cobj::scalar_object scalar_object;
-
+  typedef CartesianStencilView<vobj,cobj,Parameters> View_type;
+  typedef typename View_type::StencilVector StencilVector;
   ///////////////////////////////////////////
   // Helper structs
   ///////////////////////////////////////////
@@ -123,8 +201,8 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
   };
   struct Merge {
     cobj * mpointer;
-    std::vector<scalar_object *> rpointers;
-    std::vector<cobj *> vpointers;
+    Vector<scalar_object *> rpointers;
+    Vector<cobj *> vpointers;
     Integer buffer_size;
     Integer type;
   };
@@ -134,26 +212,30 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     Integer buffer_size;
   };
   
-  ////////////////////////////////////////
-  // Basic Grid and stencil info
-  ////////////////////////////////////////
-  int face_table_computed;
-  std::vector<std::vector<std::pair<int,int> > > face_table ;
 
-  
-  int                               _checkerboard;
-  int                               _npoints; // Move to template param?
+protected:
   GridBase *                        _grid;
-  
-  // npoints of these
-  std::vector<int>                  _directions;
-  std::vector<int>                  _distances;
-  std::vector<int>                  _comm_buf_size;
-  std::vector<int>                  _permute_type;
-  std::vector<int> same_node;
-  std::vector<int> surface_list;
 
-  Vector<StencilEntry>  _entries;
+public: 
+  GridBase *Grid(void) const { return _grid; }
+
+  ////////////////////////////////////////////////////////////////////////
+  // Needed to conveniently communicate gparity parameters into GPU memory
+  // without adding parameters. Perhaps a template parameter to StenciView is
+  // required to pass general parameters.
+  // Generalise as required later if needed
+  ////////////////////////////////////////////////////////////////////////
+
+  View_type View(void) const {
+    View_type accessor(*( (View_type *) this));
+    return accessor;
+  }
+  
+  int face_table_computed;
+  std::vector<Vector<std::pair<int,int> > > face_table ;
+  Vector<int> surface_list;
+
+  Vector<StencilEntry>  _entries; // Resident in managed memory
   std::vector<Packet> Packets;
   std::vector<Merge> Mergers;
   std::vector<Merge> MergersSHM;
@@ -166,14 +248,11 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
   // Vectors that live on the symmetric heap in case of SHMEM
   // These are used; either SHM objects or refs to the above symmetric heap vectors
   // depending on comms target
-  cobj* u_recv_buf_p;
-  cobj* u_send_buf_p;
-  std::vector<cobj *> u_simd_send_buf;
-  std::vector<cobj *> u_simd_recv_buf;
+  Vector<cobj *> u_simd_send_buf;
+  Vector<cobj *> u_simd_recv_buf;
 
   int u_comm_offset;
   int _unified_buffer_size;
-  cobj *CommBuf(void) { return u_recv_buf_p; }
 
   /////////////////////////////////////////
   // Timing info; ugly; possibly temporary
@@ -188,10 +267,12 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
   double mergetime;
   double decompresstime;
   double comms_bytes;
+  double shm_bytes;
   double splicetime;
   double nosplicetime;
   double calls;
   std::vector<double> comm_bytes_thr;
+  std::vector<double> shm_bytes_thr;
   std::vector<double> comm_time_thr;
   std::vector<double> comm_enter_thr;
   std::vector<double> comm_leave_thr;
@@ -201,9 +282,8 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
   ////////////////////////////////////////
   inline int SameNode(int point) { 
 
-    int dimension    = _directions[point];
-    int displacement = _distances[point];
-
+    int dimension    = this->_directions[point];
+    int displacement = this->_distances[point];
 
     int pd              = _grid->_processors[dimension];
     int fd              = _grid->_fdimensions[dimension];
@@ -219,47 +299,23 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 
     int nbr_proc;
     if (displacement>0) nbr_proc = 1;
-    else                nbr_proc = pd-1;
+    else                 nbr_proc = pd-1;
 
     // FIXME  this logic needs to be sorted for three link term
     //    assert( (displacement==1) || (displacement==-1));
     // Present hack only works for >= 4^4 subvol per node
     _grid->ShiftedRanks(dimension,nbr_proc,xmit_to_rank,recv_from_rank); 
 
-    void *shm = (void *) _grid->ShmBufferTranslate(recv_from_rank,u_recv_buf_p);
+    void *shm = (void *) _grid->ShmBufferTranslate(recv_from_rank,this->u_recv_buf_p);
 
     if ( shm==NULL ) return 0;
 
     return 1;
   }
-  inline int GetNodeLocal(int osite,int point) { 
-    return _entries[point+_npoints*osite]._is_local;
-  }
-  inline StencilEntry * GetEntry(int &ptype,int point,int osite) { 
-    ptype = _permute_type[point]; return & _entries[point+_npoints*osite]; 
-  }
-
-  inline uint64_t GetInfo(int &ptype,int &local,int &perm,int point,int ent,uint64_t base) {
-    uint64_t cbase = (uint64_t)&u_recv_buf_p[0];
-    local = _entries[ent]._is_local;
-    perm  = _entries[ent]._permute;
-    if (perm)  ptype = _permute_type[point]; 
-    if (local) {
-      return  base + _entries[ent]._byte_offset;
-    } else {
-      return cbase + _entries[ent]._byte_offset;
-    }
-  }
-
-  inline uint64_t GetPFInfo(int ent,uint64_t base) {
-    uint64_t cbase = (uint64_t)&u_recv_buf_p[0];
-    int local = _entries[ent]._is_local;
-    if (local) return  base + _entries[ent]._byte_offset;
-    else       return cbase + _entries[ent]._byte_offset;
-  }
 
   //////////////////////////////////////////
   // Comms packet queue for asynch thread
+  // Use OpenMP Tasks for cleaner ???
   //////////////////////////////////////////
   void CommunicateThreaded()
   {
@@ -281,6 +337,8 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 						      Packets[i].from_rank,
 						      Packets[i].bytes,i);
 	comm_bytes_thr[mythread] += bytes;
+	shm_bytes_thr[mythread] += 2*Packets[i].bytes-bytes; // Send + Recv.
+
       }
       comm_leave_thr[mythread]= usecond();
       comm_time_thr[mythread] += comm_leave_thr[mythread] - comm_enter_thr[mythread];
@@ -298,11 +356,13 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
       double t0 = comm_enter_thr[t];
       double t1 = comm_leave_thr[t];
       comms_bytes+=comm_bytes_thr[t];
+      shm_bytes  +=shm_bytes_thr[t];
 
       comm_enter_thr[t] = 0.0;
       comm_leave_thr[t] = 0.0;
       comm_time_thr[t]   = 0.0;
       comm_bytes_thr[t]=0;
+      shm_bytes_thr[t]=0;
 
       if ( first == 0.0 ) first = t0;                   // first is t0
       if ( (t0 > 0.0) && ( t0 < first ) ) first = t0;   // min time seen
@@ -312,17 +372,22 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     }
     commtime+= last-first;
   }
+  ////////////////////////////////////////////////////////////////////////
+  // Non blocking send and receive. Necessarily parallel.
+  ////////////////////////////////////////////////////////////////////////
   void CommunicateBegin(std::vector<std::vector<CommsRequest_t> > &reqs)
   {
     reqs.resize(Packets.size());
     commtime-=usecond();
     for(int i=0;i<Packets.size();i++){
-      comms_bytes+=_grid->StencilSendToRecvFromBegin(reqs[i],
+      uint64_t bytes=_grid->StencilSendToRecvFromBegin(reqs[i],
 						     Packets[i].send_buf,
 						     Packets[i].to_rank,
 						     Packets[i].recv_buf,
 						     Packets[i].from_rank,
 						     Packets[i].bytes,i);
+      comms_bytes+=bytes;
+      shm_bytes  +=2*Packets[i].bytes-bytes;
     }
   }
 
@@ -333,47 +398,44 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     }
     commtime+=usecond();
   }
+  ////////////////////////////////////////////////////////////////////////
+  // Blocking send and receive. Either sequential or parallel.
+  ////////////////////////////////////////////////////////////////////////
   void Communicate(void)
   {
-#ifdef GRID_OMP
-#pragma omp parallel 
-    {
-      // must be called in parallel region
-      int mythread  = omp_get_thread_num();
-      int maxthreads= omp_get_max_threads();
-      int nthreads = CartesianCommunicator::nCommThreads;
-      assert(nthreads <= maxthreads);
-
-      if (nthreads == -1) nthreads = 1;
-#else
-      int mythread = 0;
-      int nthreads = 1;
-#endif
-      if (mythread < nthreads) {
-	for (int i = mythread; i < Packets.size(); i += nthreads) {
-	  double start = usecond();
-	  comm_bytes_thr[mythread] += _grid->StencilSendToRecvFrom(Packets[i].send_buf,
-								   Packets[i].to_rank,
-								   Packets[i].recv_buf,
-								   Packets[i].from_rank,
-								   Packets[i].bytes,i);
-	  comm_time_thr[mythread] += usecond() - start;
+    if ( CartesianCommunicator::CommunicatorPolicy == CartesianCommunicator::CommunicatorPolicySequential ){
+      thread_region {
+	// must be called in parallel region
+	int mythread  = thread_num();
+	int maxthreads= thread_max();
+	int nthreads = CartesianCommunicator::nCommThreads;
+	assert(nthreads <= maxthreads);
+	if (nthreads == -1) nthreads = 1;
+	if (mythread < nthreads) {
+	  for (int i = mythread; i < Packets.size(); i += nthreads) {
+	    double start = usecond();
+	    uint64_t bytes= _grid->StencilSendToRecvFrom(Packets[i].send_buf,
+							 Packets[i].to_rank,
+							 Packets[i].recv_buf,
+							 Packets[i].from_rank,
+							 Packets[i].bytes,i);
+	    comm_bytes_thr[mythread] += bytes;
+	    shm_bytes_thr[mythread]  += Packets[i].bytes - bytes;
+	    comm_time_thr[mythread]  += usecond() - start;
+	  }
 	}
       }
-#ifdef GRID_OMP
+    } else { // Concurrent and non-threaded asynch calls to MPI
+      std::vector<std::vector<CommsRequest_t> > reqs;
+      this->CommunicateBegin(reqs);
+      this->CommunicateComplete(reqs);
     }
-#endif
-  }
+  }    
   
   template<class compressor> void HaloExchange(const Lattice<vobj> &source,compressor &compress) 
   {
-    std::vector<std::vector<CommsRequest_t> > reqs;
     Prepare();
     HaloGather(source,compress);
-    // Concurrent
-    //CommunicateBegin(reqs);
-    //CommunicateComplete(reqs);
-    // Sequential, possibly threaded
     Communicate();
     CommsMergeSHM(compress); 
     CommsMerge(compress); 
@@ -381,8 +443,8 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
   
   template<class compressor> int HaloGatherDir(const Lattice<vobj> &source,compressor &compress,int point,int & face_idx)
   {
-    int dimension    = _directions[point];
-    int displacement = _distances[point];
+    int dimension    = this->_directions[point];
+    int displacement = this->_distances[point];
     
     int fd = _grid->_fdimensions[dimension];
     int rd = _grid->_rdimensions[dimension];
@@ -390,29 +452,29 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     // Map to always positive shift modulo global full dimension.
     int shift = (displacement+fd)%fd;
 
-    assert (source.checkerboard== _checkerboard);
+    assert (source.Checkerboard()== this->_checkerboard);
     
     // the permute type
     int simd_layout     = _grid->_simd_layout[dimension];
     int comm_dim        = _grid->_processors[dimension] >1 ;
     int splice_dim      = _grid->_simd_layout[dimension]>1 && (comm_dim);
 
-    int same_node = 1;
+    int is_same_node = 1;
     // Gather phase
     int sshift [2];
     if ( comm_dim ) {
-      sshift[0] = _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,Even);
-      sshift[1] = _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,Odd);
+      sshift[0] = _grid->CheckerBoardShiftForCB(this->_checkerboard,dimension,shift,Even);
+      sshift[1] = _grid->CheckerBoardShiftForCB(this->_checkerboard,dimension,shift,Odd);
       if ( sshift[0] == sshift[1] ) {
 	if (splice_dim) {
 	  splicetime-=usecond();
 	  auto tmp  = GatherSimd(source,dimension,shift,0x3,compress,face_idx);
-	  same_node = same_node && tmp;
+	  is_same_node = is_same_node && tmp;
 	  splicetime+=usecond();
 	} else { 
 	  nosplicetime-=usecond();
 	  auto tmp  = Gather(source,dimension,shift,0x3,compress,face_idx);
-	  same_node = same_node && tmp;
+	  is_same_node = is_same_node && tmp;
 	  nosplicetime+=usecond();
 	}
       } else {
@@ -422,18 +484,18 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 	  // both with block stride loop iteration
 	  auto tmp1 =  GatherSimd(source,dimension,shift,0x1,compress,face_idx);
 	  auto tmp2 =  GatherSimd(source,dimension,shift,0x2,compress,face_idx);
-	  same_node = same_node && tmp1 && tmp2;
+	  is_same_node = is_same_node && tmp1 && tmp2;
 	  splicetime+=usecond();
 	} else {
 	  nosplicetime-=usecond();
 	  auto tmp1 = Gather(source,dimension,shift,0x1,compress,face_idx);
 	  auto tmp2 = Gather(source,dimension,shift,0x2,compress,face_idx);
-	  same_node = same_node && tmp1 && tmp2;
+	  is_same_node = is_same_node && tmp1 && tmp2;
 	  nosplicetime+=usecond();
 	}
       }
     }
-    return same_node;
+    return is_same_node;
   }
   
   template<class compressor>
@@ -443,21 +505,22 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     _grid->StencilBarrier();// Synch shared memory on a single nodes
     mpi3synctime_g+=usecond();
 
-    // conformable(source._grid,_grid);
-    assert(source._grid==_grid);
+    // conformable(source.Grid(),_grid);
+    assert(source.Grid()==_grid);
     halogtime-=usecond();
     
     u_comm_offset=0;
     
     // Gather all comms buffers
     int face_idx=0;
-    for(int point = 0 ; point < _npoints; point++) {
+    for(int point = 0 ; point < this->_npoints; point++) {
       compress.Point(point);
       HaloGatherDir(source,compress,point,face_idx);
     }
     face_table_computed=1;
-    
     assert(u_comm_offset==_unified_buffer_size);
+
+    accelerator_barrier();
     halogtime+=usecond();
   }
  
@@ -489,7 +552,7 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     d.buffer_size = buffer_size;
     dv.push_back(d);
   }
-  void AddMerge(cobj *merge_p,std::vector<cobj *> &rpointers,Integer buffer_size,Integer type,std::vector<Merge> &mv) {
+  void AddMerge(cobj *merge_p,Vector<cobj *> &rpointers,Integer buffer_size,Integer type,std::vector<Merge> &mv) {
     Merge m;
     m.type     = type;
     m.mpointer = merge_p;
@@ -514,20 +577,23 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 
     mergetime-=usecond();
     for(int i=0;i<mm.size();i++){	
-      parallel_for(int o=0;o<mm[i].buffer_size/2;o++){
-	decompress.Exchange(mm[i].mpointer,
-			    mm[i].vpointers[0],
-			    mm[i].vpointers[1],
-			    mm[i].type,o);
-      }
+      auto mp = &mm[i].mpointer[0];
+      auto vp0= &mm[i].vpointers[0][0];
+      auto vp1= &mm[i].vpointers[1][0];
+      auto type= mm[i].type;
+      accelerator_forNB(o,mm[i].buffer_size/2,1,{
+	  decompress.Exchange(mp,vp0,vp1,type,o);
+      });
     }
     mergetime+=usecond();
 
     decompresstime-=usecond();
     for(int i=0;i<dd.size();i++){	
-      parallel_for(int o=0;o<dd[i].buffer_size;o++){
-	decompress.Decompress(dd[i].kernel_p,dd[i].mpi_p,o);
-      }      
+      auto kp = dd[i].kernel_p;
+      auto mp = dd[i].mpi_p;
+      accelerator_forNB(o,dd[i].buffer_size,1,{
+	decompress.Decompress(kp,mp,o);
+      });
     }
     decompresstime+=usecond();
   }
@@ -551,13 +617,13 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     // find same node for SHM
     // Here we know the distance is 1 for WilsonStencil
     for(int point=0;point<this->_npoints;point++){
-      same_node[point] = this->SameNode(point);
+      this->same_node[point] = this->SameNode(point);
     }
     
     for(int site = 0 ;site< vol4;site++){
       int local = 1;
       for(int point=0;point<this->_npoints;point++){
-	if( (!this->GetNodeLocal(site*Ls,point)) && (!same_node[point]) ){ 
+	if( (!this->GetNodeLocal(site*Ls,point)) && (!this->same_node[point]) ){ 
 	  local = 0;
 	}
       }
@@ -567,30 +633,39 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     }
   }
 
- CartesianStencil(GridBase *grid,
-		  int npoints,
-		  int checkerboard,
-		  const std::vector<int> &directions,
-		  const std::vector<int> &distances) 
-   : _permute_type(npoints), 
-    _comm_buf_size(npoints),
-    comm_bytes_thr(npoints), 
-    comm_enter_thr(npoints),
-    comm_leave_thr(npoints), 
-    comm_time_thr(npoints),
-    same_node(npoints)
+  CartesianStencil(GridBase *grid,
+		   int npoints,
+		   int checkerboard,
+		   const std::vector<int> &directions,
+		   const std::vector<int> &distances,
+		   Parameters p) 
+    : shm_bytes_thr(npoints), 
+      comm_bytes_thr(npoints), 
+      comm_enter_thr(npoints),
+      comm_leave_thr(npoints), 
+      comm_time_thr(npoints)
   {
     face_table_computed=0;
-    _npoints = npoints;
     _grid    = grid;
-    _directions = directions;
-    _distances  = distances;
+    this->parameters=p;
+    /////////////////////////////////////
+    // Initialise the base
+    /////////////////////////////////////
+    this->_npoints = npoints;
+    this->_comm_buf_size.resize(npoints),
+    this->_permute_type.resize(npoints), 
+    this->_simd_layout = _grid->_simd_layout; // copy simd_layout to give access to Accelerator Kernels
+    this->_directions = StencilVector(directions);
+    this->_distances  = StencilVector(distances);
+    this->same_node.resize(npoints);
+
     _unified_buffer_size=0;
     surface_list.resize(0);
 
     int osites  = _grid->oSites();
     
-    _entries.resize(_npoints* osites);
+    _entries.resize(this->_npoints* osites);
+    this->_entries_p = &_entries[0];
     for(int ii=0;ii<npoints;ii++){
       
       int i = ii; // reverse direction to get SIMD comms done first
@@ -602,9 +677,9 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
       
       int fd = _grid->_fdimensions[dimension];
       int rd = _grid->_rdimensions[dimension];
-      _permute_type[point]=_grid->PermuteType(dimension);
+      this->_permute_type[point]=_grid->PermuteType(dimension);
       
-      _checkerboard = checkerboard;
+      this->_checkerboard = checkerboard;
       
       //////////////////////////
       // the permute type
@@ -624,8 +699,8 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
       // live in lattice or a comms buffer.
       //////////////////////////
       if ( !comm_dim ) {
-	sshift[0] = _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,Even);
-	sshift[1] = _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,Odd);
+	sshift[0] = _grid->CheckerBoardShiftForCB(this->_checkerboard,dimension,shift,Even);
+	sshift[1] = _grid->CheckerBoardShiftForCB(this->_checkerboard,dimension,shift,Odd);
 	
 	if ( sshift[0] == sshift[1] ) {
 	  Local(point,dimension,shift,0x3);
@@ -636,8 +711,8 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
       } else { 
 	// All permute extract done in comms phase prior to Stencil application
 	//        So tables are the same whether comm_dim or splice_dim
-	sshift[0] = _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,Even);
-	sshift[1] = _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,Odd);
+	sshift[0] = _grid->CheckerBoardShiftForCB(this->_checkerboard,dimension,shift,Even);
+	sshift[1] = _grid->CheckerBoardShiftForCB(this->_checkerboard,dimension,shift,Odd);
 	if ( sshift[0] == sshift[1] ) {
 	  Comms(point,dimension,shift,0x3);
 	} else {
@@ -654,12 +729,13 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 
     _grid->ShmBufferFreeAll();
 
-    u_simd_send_buf.resize(Nsimd);
-    u_simd_recv_buf.resize(Nsimd);
-    u_send_buf_p=(cobj *)_grid->ShmBufferMalloc(_unified_buffer_size*sizeof(cobj));
-    u_recv_buf_p=(cobj *)_grid->ShmBufferMalloc(_unified_buffer_size*sizeof(cobj));
+    int maxl=2;
+    u_simd_send_buf.resize(maxl);
+    u_simd_recv_buf.resize(maxl);
+    this->u_send_buf_p=(cobj *)_grid->ShmBufferMalloc(_unified_buffer_size*sizeof(cobj));
+    this->u_recv_buf_p=(cobj *)_grid->ShmBufferMalloc(_unified_buffer_size*sizeof(cobj));
 
-    for(int l=0;l<2;l++){
+    for(int l=0;l<maxl;l++){
       u_simd_recv_buf[l] = (cobj *)_grid->ShmBufferMalloc(_unified_buffer_size*sizeof(cobj));
       u_simd_send_buf[l] = (cobj *)_grid->ShmBufferMalloc(_unified_buffer_size*sizeof(cobj));
     }
@@ -683,12 +759,12 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     
     for(int x=0;x<rd;x++){       
       
-      int o   = 0;
+      //      int o   = 0;
       int bo  = x * _grid->_ostride[dimension];
       
       int cb= (cbmask==0x2)? Odd : Even;
       
-      int sshift = _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,cb);
+      int sshift = _grid->CheckerBoardShiftForCB(this->_checkerboard,dimension,shift,cb);
       int sx     = (x+sshift)%rd;
       
       int wraparound=0;
@@ -701,7 +777,7 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
       
       int permute_slice=0;
       if(permute_dim){
-	int wrap = sshift/rd;
+	int wrap = sshift/rd; wrap=wrap % ly; // but it is local anyway
 	int  num = sshift%rd;
 	if ( x< rd-num ) permute_slice=wrap;
 	else permute_slice = (wrap+1)%ly;
@@ -732,12 +808,12 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     // done in reduced dims, so SIMD factored
     int buffer_size = _grid->_slice_nblock[dimension]*_grid->_slice_block[dimension]; 
 
-    _comm_buf_size[point] = buffer_size; // Size of _one_ plane. Multiple planes may be gathered and
+    this->_comm_buf_size[point] = buffer_size; // Size of _one_ plane. Multiple planes may be gathered and
 
     // send to one or more remote nodes.
     
     int cb= (cbmask==0x2)? Odd : Even;
-    int sshift= _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,cb);
+    int sshift= _grid->CheckerBoardShiftForCB(this->_checkerboard,dimension,shift,cb);
     
     for(int x=0;x<rd;x++){       
       
@@ -783,9 +859,9 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 	int words = buffer_size;
 	if (cbmask != 0x3) words=words>>1;
 	
-	int rank           = grid->_processor;
-	int recv_from_rank;
-	int xmit_to_rank;
+	//	int rank           = grid->_processor;
+	//	int recv_from_rank;
+	//	int xmit_to_rank;
 	
 	int unified_buffer_offset = _unified_buffer_size;
 	_unified_buffer_size    += words;
@@ -809,7 +885,7 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
       // Simple block stride gather of SIMD objects
       for(int n=0;n<_grid->_slice_nblock[dimension];n++){
 	for(int b=0;b<_grid->_slice_block[dimension];b++){
-	  int idx=point+(lo+o+b)*_npoints;
+	  int idx=point+(lo+o+b)*this->_npoints;
 	  _entries[idx]._offset  =ro+o+b;
 	  _entries[idx]._permute=permute;
 	  _entries[idx]._is_local=1;
@@ -830,7 +906,7 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 	  int ocb=1<<_grid->CheckerBoardFromOindex(o+b);
 	  
 	  if ( ocb&cbmask ) {
-	    int idx = point+(lo+o+b)*_npoints;
+	    int idx = point+(lo+o+b)*this->_npoints;
 	    _entries[idx]._offset =ro+o+b;
 	    _entries[idx]._is_local=1;
 	    _entries[idx]._permute=permute;
@@ -857,7 +933,7 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
       // Simple block stride gather of SIMD objects
       for(int n=0;n<_grid->_slice_nblock[dimension];n++){
 	for(int b=0;b<_grid->_slice_block[dimension];b++){
-	  int idx=point+(so+o+b)*_npoints;
+	  int idx=point+(so+o+b)*this->_npoints;
 	  _entries[idx]._offset  =offset+(bo++);
 	  _entries[idx]._is_local=0;
 	  _entries[idx]._permute=0;
@@ -877,7 +953,7 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 	  
 	  int ocb=1<<_grid->CheckerBoardFromOindex(o+b);// Could easily be a table lookup
 	  if ( ocb & cbmask ) {
-	    int idx = point+(so+o+b)*_npoints;
+	    int idx = point+(so+o+b)*this->_npoints;
 	    _entries[idx]._offset  =offset+(bo++);
 	    _entries[idx]._is_local=0;
 	    _entries[idx]._permute =0;
@@ -895,9 +971,8 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     typedef typename cobj::vector_type vector_type;
     typedef typename cobj::scalar_type scalar_type;
     
-    GridBase *grid=_grid;
-    assert(rhs._grid==_grid);
-    //	  conformable(_grid,rhs._grid);
+    assert(rhs.Grid()==_grid);
+    //	  conformable(_grid,rhs.Grid());
     
     int fd              = _grid->_fdimensions[dimension];
     int rd              = _grid->_rdimensions[dimension];
@@ -908,11 +983,11 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     assert(comm_dim==1);
     assert(shift>=0);
     assert(shift<fd);
-    
+
     int buffer_size = _grid->_slice_nblock[dimension]*_grid->_slice_block[dimension];
     
     int cb= (cbmask==0x2)? Odd : Even;
-    int sshift= _grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,cb);
+    int sshift= _grid->CheckerBoardShiftForCB(rhs.Checkerboard(),dimension,shift,cb);
     
     int shm_receive_only = 1;
     for(int x=0;x<rd;x++){       
@@ -927,13 +1002,13 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 	
 	int bytes =  words * compress.CommDatumSize();
 	
-	int so  = sx*rhs._grid->_ostride[dimension]; // base offset for start of plane 
+	int so  = sx*rhs.Grid()->_ostride[dimension]; // base offset for start of plane 
 	if ( !face_table_computed ) {
 	  face_table.resize(face_idx+1);
 	  Gather_plane_table_compute ((GridBase *)_grid,dimension,sx,cbmask,u_comm_offset,face_table[face_idx]);
 	}
 	
-      	int rank           = _grid->_processor;
+	//      	int rank           = _grid->_processor;
 	int recv_from_rank;
 	int xmit_to_rank;
 	_grid->ShiftedRanks(dimension,comm_proc,xmit_to_rank,recv_from_rank);
@@ -949,16 +1024,16 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 	if ( compress.DecompressionStep() ) {
 	  recv_buf=u_simd_recv_buf[0];
 	} else {
-	  recv_buf=u_recv_buf_p;
+	  recv_buf=this->u_recv_buf_p;
 	}
 
 	send_buf = (cobj *)_grid->ShmBufferTranslate(xmit_to_rank,recv_buf);
 	if ( send_buf==NULL ) { 
-	  send_buf = u_send_buf_p;
+	  send_buf = this->u_send_buf_p;
 	} 
 	
 	// Find out if we get the direct copy.
-	void *success = (void *) _grid->ShmBufferTranslate(recv_from_rank,u_send_buf_p);
+	void *success = (void *) _grid->ShmBufferTranslate(recv_from_rank,this->u_send_buf_p);
 	if (success==NULL) { 
 	  // we found a packet that comes from MPI and contributes to this leg of stencil
 	  shm_receive_only = 0;
@@ -972,11 +1047,11 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 	if ( compress.DecompressionStep() ) {
 	  
 	  if ( shm_receive_only ) { // Early decompress before MPI is finished is possible
-	    AddDecompress(&u_recv_buf_p[u_comm_offset],
+	    AddDecompress(&this->u_recv_buf_p[u_comm_offset],
 			  &recv_buf[u_comm_offset],
 			  words,DecompressionsSHM);
 	  } else { // Decompress after MPI is finished
-	    AddDecompress(&u_recv_buf_p[u_comm_offset],
+	    AddDecompress(&this->u_recv_buf_p[u_comm_offset],
 			  &recv_buf[u_comm_offset],
 			  words,Decompressions);
 	  }
@@ -989,7 +1064,7 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 
 	} else {
 	  AddPacket((void *)&send_buf[u_comm_offset],
-		    (void *)&u_recv_buf_p[u_comm_offset],
+		    (void *)&this->u_recv_buf_p[u_comm_offset],
 		    xmit_to_rank,
 		    recv_from_rank,
 		    bytes);
@@ -1018,6 +1093,7 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     assert(shift>=0);
     assert(shift<fd);
 
+
     int permute_type=_grid->PermuteType(dimension);
     //    std::cout << "SimdNew permute type "<<permute_type<<std::endl;
 
@@ -1025,10 +1101,10 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     // Simd direction uses an extract/merge pair
     ///////////////////////////////////////////////
     int buffer_size = _grid->_slice_nblock[dimension]*_grid->_slice_block[dimension];
-    int words = sizeof(cobj)/sizeof(vector_type);
+    //    int words = sizeof(cobj)/sizeof(vector_type);
     
     assert(cbmask==0x3); // Fixme think there is a latent bug if not true
-    
+                         // This assert will trap it if ever hit. Not hit normally so far
     int reduced_buffer_size = buffer_size;
     if (cbmask != 0x3) reduced_buffer_size=buffer_size>>1;
 
@@ -1036,15 +1112,15 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     int bytes = (reduced_buffer_size*datum_bytes)/simd_layout;
     assert(bytes*simd_layout == reduced_buffer_size*datum_bytes);
 
-    std::vector<cobj *> rpointers(maxl);
-    std::vector<cobj *> spointers(maxl);
+    Vector<cobj *> rpointers(maxl);
+    Vector<cobj *> spointers(maxl);
 
     ///////////////////////////////////////////
     // Work out what to send where
     ///////////////////////////////////////////
     
     int cb    = (cbmask==0x2)? Odd : Even;
-    int sshift= _grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,cb);
+    int sshift= _grid->CheckerBoardShiftForCB(rhs.Checkerboard(),dimension,shift,cb);
     
     // loop over outer coord planes orthog to dim
     int shm_receive_only = 1;
@@ -1065,7 +1141,9 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 	  Gather_plane_table_compute ((GridBase *)_grid,dimension,sx,cbmask,u_comm_offset,face_table[face_idx]);
 	}
 	gathermtime-=usecond();
+
 	Gather_plane_exchange_table(face_table[face_idx],rhs,spointers,dimension,sx,cbmask,compress,permute_type);  face_idx++;
+
 	gathermtime+=usecond();
 	//spointers[0] -- low
 	//spointers[1] -- high
@@ -1099,8 +1177,8 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 	    if (shm==NULL) { 
 	      shm = rp;
 	      // we found a packet that comes from MPI and contributes to this shift.
-	      // same_node is only used in the WilsonStencil, and gets set for this point in the stencil.
-	      // Kernel will add the exterior_terms except if same_node.
+	      // is_same_node is only used in the WilsonStencil, and gets set for this point in the stencil.
+	      // Kernel will add the exterior_terms except if is_same_node.
 	      shm_receive_only = 0;
 	      // leg of stencil
 	    }
@@ -1119,9 +1197,9 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 	}
 
 	if ( shm_receive_only ) { 
-	  AddMerge(&u_recv_buf_p[u_comm_offset],rpointers,reduced_buffer_size,permute_type,MergersSHM);
+	  AddMerge(&this->u_recv_buf_p[u_comm_offset],rpointers,reduced_buffer_size,permute_type,MergersSHM);
 	} else {
-	  AddMerge(&u_recv_buf_p[u_comm_offset],rpointers,reduced_buffer_size,permute_type,Mergers);
+	  AddMerge(&this->u_recv_buf_p[u_comm_offset],rpointers,reduced_buffer_size,permute_type,Mergers);
 	}
 
 	u_comm_offset     +=buffer_size;
@@ -1136,11 +1214,12 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     mpi3synctime=0.;
     mpi3synctime_g=0.;
     shmmergetime=0.;
-    for(int i=0;i<_npoints;i++){
+    for(int i=0;i<this->_npoints;i++){
       comm_time_thr[i]=0;
       comm_bytes_thr[i]=0;
       comm_enter_thr[i]=0;
       comm_leave_thr[i]=0;
+      shm_bytes_thr[i]=0;
     }
     halogtime = 0.;
     mergetime = 0.;
@@ -1149,6 +1228,7 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
     splicetime = 0.;
     nosplicetime = 0.;
     comms_bytes = 0.;
+    shm_bytes = 0.;
     calls = 0.;
   };
   
@@ -1165,6 +1245,7 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
       if ( comm_time_thr[i]>0.0 ) {
 	threaded = 1;
 	comms_bytes += comm_bytes_thr[i];
+	shm_bytes   += shm_bytes_thr[i];
 	if (t < comm_time_thr[i]) t = comm_time_thr[i];
       }
     }
@@ -1184,6 +1265,22 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
 	std::cout << GridLogMessage << " Stencil " << comms_bytes/commtime/1000. << " GB/s per rank"<<std::endl;
 	std::cout << GridLogMessage << " Stencil " << comms_bytes/commtime/1000.*NP/NN << " GB/s per node"<<std::endl;
       }
+      if(shm_bytes>1.0){
+	PRINTIT(shm_bytes); // X bytes + R bytes
+	                    // Double this to include spin projection overhead with 2:1 ratio in wilson
+	auto gatheralltime = gathertime+gathermtime;
+	std::cout << GridLogMessage << " Stencil SHM " << (shm_bytes)/gatheralltime/1000. << " GB/s per rank"<<std::endl;
+	std::cout << GridLogMessage << " Stencil SHM " << (shm_bytes)/gatheralltime/1000.*NP/NN << " GB/s per node"<<std::endl;
+
+	auto all_bytes = comms_bytes+shm_bytes;
+	std::cout << GridLogMessage << " Stencil SHM all" << (all_bytes)/gatheralltime/1000. << " GB/s per rank"<<std::endl;
+	std::cout << GridLogMessage << " Stencil SHM all" << (all_bytes)/gatheralltime/1000.*NP/NN << " GB/s per node"<<std::endl;
+
+	auto membytes = (shm_bytes + comms_bytes/2) // read/write
+	              + (shm_bytes+comms_bytes)/2 * sizeof(vobj)/sizeof(cobj);
+	std::cout << GridLogMessage << " Stencil SHM mem " << (membytes)/gatheralltime/1000. << " GB/s per rank"<<std::endl;
+	std::cout << GridLogMessage << " Stencil SHM mem " << (membytes)/gatheralltime/1000.*NP/NN << " GB/s per node"<<std::endl;
+      }
       PRINTIT(mpi3synctime);
       PRINTIT(mpi3synctime_g);
       PRINTIT(shmmergetime);
@@ -1195,5 +1292,6 @@ class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal
   };
   
 };
-}
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/tensors/Tensor_SIMT.h b/Grid/tensors/Tensor_SIMT.h
new file mode 100644
index 00000000..c18b8484
--- /dev/null
+++ b/Grid/tensors/Tensor_SIMT.h
@@ -0,0 +1,109 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/tensors/Tensor_SIMT.h
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once 
+
+#include <string.h>
+
+NAMESPACE_BEGIN(Grid);
+
+//accelerator_inline void SIMTsynchronise(void) 
+accelerator_inline void synchronise(void) 
+{
+#ifdef __CUDA_ARCH__
+//  __syncthreads();
+  __syncwarp();
+#endif
+  return;
+}
+
+#ifndef __CUDA_ARCH__
+//////////////////////////////////////////
+// Trivial mapping of vectors on host
+//////////////////////////////////////////
+accelerator_inline int SIMTlane(int Nsimd) { return 0; } // CUDA specific
+
+template<class vobj> accelerator_inline
+vobj coalescedRead(const vobj & __restrict__ vec,int lane=0)
+{
+  return vec;
+}
+template<class vobj> accelerator_inline
+vobj coalescedReadPermute(const vobj & __restrict__ vec,int ptype,int doperm,int lane=0)
+{
+  if ( doperm ) {
+    vobj ret;
+    permute(ret,vec, ptype);
+    return ret;
+  } else { 
+    return vec;
+  }
+}
+template<class vobj> accelerator_inline
+void coalescedWrite(vobj & __restrict__ vec,const vobj & __restrict__ extracted,int lane=0)
+{
+  //  vstream(vec, extracted);
+  vec = extracted;
+}
+template<class vobj> accelerator_inline
+void coalescedWriteNonTemporal(vobj & __restrict__ vec,const vobj & __restrict__ extracted,int lane=0)
+{
+  vstream(vec, extracted);
+}
+#else
+accelerator_inline int SIMTlane(int Nsimd) { return threadIdx.y; } // CUDA specific
+
+//////////////////////////////////////////
+// Extract and insert slices on the GPU
+//////////////////////////////////////////
+template<class vobj> accelerator_inline
+typename vobj::scalar_object coalescedRead(const vobj & __restrict__ vec,int lane=SIMTlane(vobj::Nsimd()))
+{
+  return extractLane(lane,vec);
+}
+template<class vobj> accelerator_inline
+typename vobj::scalar_object coalescedReadPermute(const vobj & __restrict__ vec,int ptype,int doperm,int lane=SIMTlane(vobj::Nsimd()))
+{
+  int mask = vobj::Nsimd() >> (ptype + 1);		
+  int plane= doperm ? lane ^ mask : lane;
+  return extractLane(plane,vec);
+}
+template<class vobj> accelerator_inline
+void coalescedWrite(vobj & __restrict__ vec,const typename vobj::scalar_object & __restrict__ extracted,int lane=SIMTlane(vobj::Nsimd()))
+{
+  insertLane(lane,vec,extracted);
+}
+template<class vobj> accelerator_inline
+void coalescedWriteNonTemporal(vobj & __restrict__ vec,const vobj & __restrict__ extracted,int lane=0)
+{
+  insertLane(lane,vec,extracted);
+}
+#endif
+
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/tensors/Tensor_Ta.h b/Grid/tensors/Tensor_Ta.h
index 45a2257f..1ef9fc23 100644
--- a/Grid/tensors/Tensor_Ta.h
+++ b/Grid/tensors/Tensor_Ta.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,102 +24,101 @@ Author: neo <cossu@post.kek.jp>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_TA_H
 #define GRID_MATH_TA_H
 
 
-namespace Grid {
-
-  /////////////////////////////////////////////// 
-  // Ta function for scalar, vector, matrix
-  /////////////////////////////////////////////// 
-  /*
-  inline ComplexF Ta( const ComplexF &arg){    return arg;}
-  inline ComplexD Ta( const ComplexD &arg){    return arg;}
-  inline RealF Ta( const RealF &arg){    return arg;}
-  inline RealD Ta( const RealD &arg){    return arg;}
-  */
-
-  template<class vtype> inline iScalar<vtype> Ta(const iScalar<vtype>&r)
-    {
-      iScalar<vtype> ret;
-      ret._internal = Ta(r._internal);
-      return ret;
-    }
-  template<class vtype,int N> inline iVector<vtype,N> Ta(const iVector<vtype,N>&r)
-    {
-      iVector<vtype,N> ret;
-      for(int i=0;i<N;i++){
-        ret._internal[i] = Ta(r._internal[i]);
-      }
-      return ret;
-    }
-  template<class vtype,int N> inline iMatrix<vtype,N> Ta(const iMatrix<vtype,N> &arg)
-    {
-      iMatrix<vtype,N> ret;
-
-      double factor = (1.0/(double)N);
-      ret= (arg - adj(arg))*0.5;
-      ret=ret - (trace(ret)*factor);
-      return ret;
-    }
-
-
-  /////////////////////////////////////////////// 
-  // ProjectOnGroup function for scalar, vector, matrix 
-  // Projects on orthogonal, unitary group
-  /////////////////////////////////////////////// 
-
-
-  template<class vtype> inline iScalar<vtype> ProjectOnGroup(const iScalar<vtype>&r)
-    {
-      iScalar<vtype> ret;
-      ret._internal = ProjectOnGroup(r._internal);
-      return ret;
-    }
-  template<class vtype,int N> inline iVector<vtype,N> ProjectOnGroup(const iVector<vtype,N>&r)
-    {
-      iVector<vtype,N> ret;
-      for(int i=0;i<N;i++){
-        ret._internal[i] = ProjectOnGroup(r._internal[i]);
-      }
-      return ret;
-    }
-  template<class vtype,int N, typename std::enable_if< GridTypeMapper<vtype>::TensorLevel == 0 >::type * =nullptr> 
-    inline iMatrix<vtype,N> ProjectOnGroup(const iMatrix<vtype,N> &arg)
-    {
-      // need a check for the group type?
-      iMatrix<vtype,N> ret(arg);
-      vtype nrm;
-      vtype inner;
-      for(int c1=0;c1<N;c1++){
-	zeroit(inner);	
-	for(int c2=0;c2<N;c2++)
-	  inner += innerProduct(ret._internal[c1][c2],ret._internal[c1][c2]);
-
-	nrm = rsqrt(inner);
-	for(int c2=0;c2<N;c2++)
-	  ret._internal[c1][c2]*= nrm;
-      
-	for (int b=c1+1; b<N; ++b){
-	  decltype(ret._internal[b][b]*ret._internal[b][b]) pr;
-	  zeroit(pr);
-	  for(int c=0; c<N; ++c)
-	    pr += conjugate(ret._internal[c1][c])*ret._internal[b][c];
-	  
-	  for(int c=0; c<N; ++c){
-	    ret._internal[b][c] -= pr * ret._internal[c1][c];
-	  }
-	}
-	  
-      }
-      // assuming the determinant is ok
-      return ret;
-    }
-
+NAMESPACE_BEGIN(Grid);
 
+/////////////////////////////////////////////// 
+// Ta function for scalar, vector, matrix
+/////////////////////////////////////////////// 
+/*
+  accelerator_inline ComplexF Ta( const ComplexF &arg){    return arg;}
+  accelerator_inline ComplexD Ta( const ComplexD &arg){    return arg;}
+  accelerator_inline RealF Ta( const RealF &arg){    return arg;}
+  accelerator_inline RealD Ta( const RealD &arg){    return arg;}
+*/
 
+template<class vtype> accelerator_inline iScalar<vtype> Ta(const iScalar<vtype>&r)
+{
+  iScalar<vtype> ret;
+  ret._internal = Ta(r._internal);
+  return ret;
 }
+template<class vtype,int N> accelerator_inline iVector<vtype,N> Ta(const iVector<vtype,N>&r)
+{
+  iVector<vtype,N> ret;
+  for(int i=0;i<N;i++){
+    ret._internal[i] = Ta(r._internal[i]);
+  }
+  return ret;
+}
+template<class vtype,int N> accelerator_inline iMatrix<vtype,N> Ta(const iMatrix<vtype,N> &arg)
+{
+  iMatrix<vtype,N> ret;
+
+  double factor = (1.0/(double)N);
+  ret= (arg - adj(arg))*0.5;
+  ret=ret - (trace(ret)*factor);
+  return ret;
+}
+
+
+/////////////////////////////////////////////// 
+// ProjectOnGroup function for scalar, vector, matrix 
+// Projects on orthogonal, unitary group
+/////////////////////////////////////////////// 
+
+
+template<class vtype> accelerator_inline iScalar<vtype> ProjectOnGroup(const iScalar<vtype>&r)
+{
+  iScalar<vtype> ret;
+  ret._internal = ProjectOnGroup(r._internal);
+  return ret;
+}
+template<class vtype,int N> accelerator_inline iVector<vtype,N> ProjectOnGroup(const iVector<vtype,N>&r)
+{
+  iVector<vtype,N> ret;
+  for(int i=0;i<N;i++){
+    ret._internal[i] = ProjectOnGroup(r._internal[i]);
+  }
+  return ret;
+}
+template<class vtype,int N, typename std::enable_if< GridTypeMapper<vtype>::TensorLevel == 0 >::type * =nullptr> 
+accelerator_inline iMatrix<vtype,N> ProjectOnGroup(const iMatrix<vtype,N> &arg)
+{
+  // need a check for the group type?
+  iMatrix<vtype,N> ret(arg);
+  vtype nrm;
+  vtype inner;
+  for(int c1=0;c1<N;c1++){
+    zeroit(inner);	
+    for(int c2=0;c2<N;c2++)
+      inner += innerProduct(ret._internal[c1][c2],ret._internal[c1][c2]);
+
+    nrm = rsqrt(inner);
+    for(int c2=0;c2<N;c2++)
+      ret._internal[c1][c2]*= nrm;
+      
+    for (int b=c1+1; b<N; ++b){
+      decltype(ret._internal[b][b]*ret._internal[b][b]) pr;
+      zeroit(pr);
+      for(int c=0; c<N; ++c)
+	pr += conjugate(ret._internal[c1][c])*ret._internal[b][c];
+	  
+      for(int c=0; c<N; ++c){
+	ret._internal[b][c] -= pr * ret._internal[c1][c];
+      }
+    }
+	  
+  }
+  // assuming the determinant is ok
+  return ret;
+}
+
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/tensors/Tensor_arith.h b/Grid/tensors/Tensor_arith.h
index a07a8305..cf75a8e8 100644
--- a/Grid/tensors/Tensor_arith.h
+++ b/Grid/tensors/Tensor_arith.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,8 +23,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_ARITH_H
 #define GRID_MATH_ARITH_H
 
diff --git a/Grid/tensors/Tensor_arith_add.h b/Grid/tensors/Tensor_arith_add.h
index 6e01fb19..63ba9a49 100644
--- a/Grid/tensors/Tensor_arith_add.h
+++ b/Grid/tensors/Tensor_arith_add.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,123 +24,119 @@ Author: neo <cossu@post.kek.jp>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_ARITH_ADD_H
 #define GRID_MATH_ARITH_ADD_H
 
-namespace Grid {
-
-    
-    ///////////////////////////////////////////////////////////////////////////////////////////////////
-    /////////////////////////////////////////// ADD         ///////////////////////////////////////////
-    ///////////////////////////////////////////////////////////////////////////////////////////////////
+NAMESPACE_BEGIN(Grid);
     
+///////////////////////////////////////////////////////////////////////////////////////////////////
+/////////////////////////////////////////// ADD         ///////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////////////////////////
 
 // ADD is simple for now; cannot mix types and straightforward template
 // Scalar +/- Scalar
 // Vector +/- Vector
 // Matrix +/- Matrix
-  template<class vtype,class ltype,class rtype> strong_inline void add(iScalar<vtype> * __restrict__ ret,
-								const iScalar<ltype> * __restrict__ lhs,
-								const iScalar<rtype> * __restrict__ rhs)
-  {
-    add(&ret->_internal,&lhs->_internal,&rhs->_internal);
-  }
-  template<class vtype,class ltype,class rtype,int N> strong_inline void add(iVector<vtype,N> * __restrict__ ret,
-								      const iVector<ltype,N> * __restrict__ lhs,
-								      const iVector<rtype,N> * __restrict__ rhs)
-  {
-    for(int c=0;c<N;c++){
-      ret->_internal[c]=lhs->_internal[c]+rhs->_internal[c];
-    }
-    return;
+template<class vtype,class ltype,class rtype> accelerator_inline void add(iScalar<vtype> * __restrict__ ret,
+									  const iScalar<ltype> * __restrict__ lhs,
+									  const iScalar<rtype> * __restrict__ rhs)
+{
+  add(&ret->_internal,&lhs->_internal,&rhs->_internal);
+}
+template<class vtype,class ltype,class rtype,int N> accelerator_inline void add(iVector<vtype,N> * __restrict__ ret,
+									   const iVector<ltype,N> * __restrict__ lhs,
+									   const iVector<rtype,N> * __restrict__ rhs)
+{
+  for(int c=0;c<N;c++){
+    ret->_internal[c]=lhs->_internal[c]+rhs->_internal[c];
   }
+  return;
+}
   
-  template<class vtype,class ltype,class rtype, int N> strong_inline  void add(iMatrix<vtype,N> * __restrict__ ret,
-									const iMatrix<ltype,N> * __restrict__ lhs,
-									const iMatrix<rtype,N> * __restrict__ rhs)
-  {
-    for(int c2=0;c2<N;c2++){
-      for(int c1=0;c1<N;c1++){
-        add(&ret->_internal[c1][c2],&lhs->_internal[c1][c2],&rhs->_internal[c1][c2]);
-      }}
-    return;
-  }
-  template<class vtype,class ltype,class rtype, int N> strong_inline  void add(iMatrix<vtype,N> * __restrict__ ret,
-									const iScalar<ltype>   * __restrict__ lhs,
-									const iMatrix<rtype,N> * __restrict__ rhs)
-  {
-    for(int c2=0;c2<N;c2++){
-      for(int c1=0;c1<N;c1++){
-	if ( c1==c2)
-	  add(&ret->_internal[c1][c2],&lhs->_internal,&rhs->_internal[c1][c2]);
-	else
-	  ret->_internal[c1][c2]=lhs->_internal[c1][c2];
-      }}
-    return;
-  }
-  template<class vtype,class ltype,class rtype, int N> strong_inline  void add(iMatrix<vtype,N> * __restrict__ ret,
-									const iMatrix<ltype,N> * __restrict__ lhs,
-									const iScalar<rtype>   * __restrict__ rhs)
-  {
-    for(int c2=0;c2<N;c2++){
-      for(int c1=0;c1<N;c1++){
-        if ( c1==c2)
-	  add(&ret->_internal[c1][c2],&lhs->_internal[c1][c2],&rhs->_internal);
-        else
-	  ret->_internal[c1][c2]=lhs->_internal[c1][c2];
-      }}
-    return;
-  }
-
-
-  // + operator for scalar, vector, matrix
-  template<class ltype,class rtype>
-    //strong_inline auto operator + (iScalar<ltype>& lhs,iScalar<rtype>&& rhs) -> iScalar<decltype(lhs._internal + rhs._internal)>
-    strong_inline auto operator + (const iScalar<ltype>& lhs,const iScalar<rtype>& rhs) -> iScalar<decltype(lhs._internal + rhs._internal)>
-  {
-    typedef iScalar<decltype(lhs._internal+rhs._internal)> ret_t;
-    ret_t ret;
-    add(&ret,&lhs,&rhs);
-    return ret;
-  }
-  template<class ltype,class rtype,int N>
-    strong_inline auto operator + (const iVector<ltype,N>& lhs,const iVector<rtype,N>& rhs) ->iVector<decltype(lhs._internal[0]+rhs._internal[0]),N>
-    {
-      typedef iVector<decltype(lhs._internal[0]+rhs._internal[0]),N> ret_t;
-    ret_t ret;
-    add(&ret,&lhs,&rhs);
-    return ret;
-    }
-  template<class ltype,class rtype,int N>
-    strong_inline auto operator + (const iMatrix<ltype,N>& lhs,const iMatrix<rtype,N>& rhs) ->iMatrix<decltype(lhs._internal[0][0]+rhs._internal[0][0]),N>
-    {
-      typedef iMatrix<decltype(lhs._internal[0][0]+rhs._internal[0][0]),N> ret_t;
-      ret_t ret;
-      add(&ret,&lhs,&rhs);
-      return ret;
-    }
-  template<class ltype,class rtype,int N>
-strong_inline auto operator + (const iScalar<ltype>& lhs,const iMatrix<rtype,N>& rhs)->iMatrix<decltype(lhs._internal+rhs._internal[0][0]),N>
-    {
-      typedef iMatrix<decltype(lhs._internal+rhs._internal[0][0]),N> ret_t;
-      ret_t ret;
-      add(&ret,&lhs,&rhs);
-      return ret;
-    }
-
-  template<class ltype,class rtype,int N>
-    strong_inline auto operator + (const iMatrix<ltype,N>& lhs,const iScalar<rtype>& rhs)->iMatrix<decltype(lhs._internal[0][0]+rhs._internal),N>
-    {
-      typedef iMatrix<decltype(lhs._internal[0][0]+rhs._internal),N> ret_t;
-      ret_t ret;
-      add(&ret,&lhs,&rhs);
-      return ret;
-    }
-
-
-
+template<class vtype,class ltype,class rtype, int N> accelerator_inline  void add(iMatrix<vtype,N> * __restrict__ ret,
+									     const iMatrix<ltype,N> * __restrict__ lhs,
+									     const iMatrix<rtype,N> * __restrict__ rhs)
+{
+  for(int c2=0;c2<N;c2++){
+    for(int c1=0;c1<N;c1++){
+      add(&ret->_internal[c1][c2],&lhs->_internal[c1][c2],&rhs->_internal[c1][c2]);
+    }}
+  return;
+}
+template<class vtype,class ltype,class rtype, int N> accelerator_inline  void add(iMatrix<vtype,N> * __restrict__ ret,
+									     const iScalar<ltype>   * __restrict__ lhs,
+									     const iMatrix<rtype,N> * __restrict__ rhs)
+{
+  for(int c2=0;c2<N;c2++){
+    for(int c1=0;c1<N;c1++){
+      if ( c1==c2)
+	add(&ret->_internal[c1][c2],&lhs->_internal,&rhs->_internal[c1][c2]);
+      else
+	ret->_internal[c1][c2]=lhs->_internal[c1][c2];
+    }}
+  return;
+}
+template<class vtype,class ltype,class rtype, int N> accelerator_inline  void add(iMatrix<vtype,N> * __restrict__ ret,
+									     const iMatrix<ltype,N> * __restrict__ lhs,
+									     const iScalar<rtype>   * __restrict__ rhs)
+{
+  for(int c2=0;c2<N;c2++){
+    for(int c1=0;c1<N;c1++){
+      if ( c1==c2)
+	add(&ret->_internal[c1][c2],&lhs->_internal[c1][c2],&rhs->_internal);
+      else
+	ret->_internal[c1][c2]=lhs->_internal[c1][c2];
+    }}
+  return;
 }
 
+
+// + operator for scalar, vector, matrix
+template<class ltype,class rtype>
+accelerator_inline auto operator + (const iScalar<ltype>& lhs,const iScalar<rtype>& rhs) -> iScalar<decltype(lhs._internal + rhs._internal)>
+{
+  typedef iScalar<decltype(lhs._internal+rhs._internal)> ret_t;
+  ret_t ret ;
+  add(&ret,&lhs,&rhs);
+  return ret;
+}
+template<class ltype,class rtype,int N>
+accelerator_inline auto operator + (const iVector<ltype,N>& lhs,const iVector<rtype,N>& rhs) ->iVector<decltype(lhs._internal[0]+rhs._internal[0]),N>
+{
+  typedef iVector<decltype(lhs._internal[0]+rhs._internal[0]),N> ret_t;
+  ret_t ret ;
+  add(&ret,&lhs,&rhs);
+  return ret;
+}
+template<class ltype,class rtype,int N>
+accelerator_inline auto operator + (const iMatrix<ltype,N>& lhs,const iMatrix<rtype,N>& rhs) ->iMatrix<decltype(lhs._internal[0][0]+rhs._internal[0][0]),N>
+{
+  typedef iMatrix<decltype(lhs._internal[0][0]+rhs._internal[0][0]),N> ret_t;
+  ret_t ret ;
+  add(&ret,&lhs,&rhs);
+  return ret;
+}
+template<class ltype,class rtype,int N>
+accelerator_inline auto operator + (const iScalar<ltype>& lhs,const iMatrix<rtype,N>& rhs)->iMatrix<decltype(lhs._internal+rhs._internal[0][0]),N>
+{
+  typedef iMatrix<decltype(lhs._internal+rhs._internal[0][0]),N> ret_t;
+  ret_t ret ;
+  add(&ret,&lhs,&rhs);
+  return ret;
+}
+
+template<class ltype,class rtype,int N>
+accelerator_inline auto operator + (const iMatrix<ltype,N>& lhs,const iScalar<rtype>& rhs)->iMatrix<decltype(lhs._internal[0][0]+rhs._internal),N>
+{
+  typedef iMatrix<decltype(lhs._internal[0][0]+rhs._internal),N> ret_t;
+  ret_t ret ;
+  add(&ret,&lhs,&rhs);
+  return ret;
+}
+
+NAMESPACE_END(Grid);
+
+
 #endif
diff --git a/Grid/tensors/Tensor_arith_mac.h b/Grid/tensors/Tensor_arith_mac.h
index c053224b..f5d16f13 100644
--- a/Grid/tensors/Tensor_arith_mac.h
+++ b/Grid/tensors/Tensor_arith_mac.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,86 +23,86 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_ARITH_MAC_H
 #define GRID_MATH_ARITH_MAC_H
 
-namespace Grid {
-
+NAMESPACE_BEGIN(Grid);
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 /////////////////////////////////////////// MAC         ///////////////////////////////////////////
 ///////////////////////////////////////////////////////////////////////////////////////////////////
-    ///////////////////////////
+///////////////////////////
 
-    ///////////////////////////
-    // Legal multiplication table
-    ///////////////////////////
-    // scal x scal = scal
-    // mat x  mat  = mat
-    // mat  x scal = mat
-    // scal x mat  = mat
-    // mat  x vec  = vec
-    // vec  x scal = vec
-    // scal x vec  = vec
-    ///////////////////////////
+///////////////////////////
+// Legal multiplication table
+///////////////////////////
+// scal x scal = scal
+// mat x  mat  = mat
+// mat  x scal = mat
+// scal x mat  = mat
+// mat  x vec  = vec
+// vec  x scal = vec
+// scal x vec  = vec
+///////////////////////////
 template<class rtype,class vtype,class mtype>
-strong_inline  void mac(iScalar<rtype> * __restrict__ ret,const iScalar<vtype> * __restrict__ lhs,const iScalar<mtype> * __restrict__ rhs)
+accelerator_inline  void mac(iScalar<rtype> * __restrict__ ret,const iScalar<vtype> * __restrict__ lhs,const iScalar<mtype> * __restrict__ rhs)
 {
-    mac(&ret->_internal,&lhs->_internal,&rhs->_internal);
+  mac(&ret->_internal,&lhs->_internal,&rhs->_internal);
 }
 template<class rrtype,class ltype,class rtype,int N>
-strong_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iMatrix<rtype,N> * __restrict__ rhs){
-    for(int c3=0;c3<N;c3++){
+accelerator_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iMatrix<rtype,N> * __restrict__ rhs){
+  for(int c3=0;c3<N;c3++){
     for(int c1=0;c1<N;c1++){
-    for(int c2=0;c2<N;c2++){
+      for(int c2=0;c2<N;c2++){
         mac(&ret->_internal[c1][c2],&lhs->_internal[c1][c3],&rhs->_internal[c3][c2]);
-    }}}
-    return;
+      }}}
+  return;
 }
 
 template<class rrtype,class ltype,class rtype,int N>
-strong_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iScalar<rtype> * __restrict__ rhs){
-    for(int c1=0;c1<N;c1++){
+accelerator_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iScalar<rtype> * __restrict__ rhs){
+  for(int c1=0;c1<N;c1++){
     for(int c2=0;c2<N;c2++){
-        mac(&ret->_internal[c1][c2],&lhs->_internal[c1][c2],&rhs->_internal);
+      mac(&ret->_internal[c1][c2],&lhs->_internal[c1][c2],&rhs->_internal);
     }}
-    return;
+  return;
 }
 template<class rrtype,class ltype,class rtype,int N>
-strong_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iScalar<ltype> * __restrict__ lhs,const iMatrix<rtype,N> * __restrict__ rhs){
-    for(int c1=0;c1<N;c1++){
+accelerator_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iScalar<ltype> * __restrict__ lhs,const iMatrix<rtype,N> * __restrict__ rhs){
+  for(int c1=0;c1<N;c1++){
     for(int c2=0;c2<N;c2++){
-        mac(&ret->_internal[c1][c2],&lhs->_internal,&rhs->_internal[c1][c2]);
+      mac(&ret->_internal[c1][c2],&lhs->_internal,&rhs->_internal[c1][c2]);
     }}
-    return;
+  return;
 }
 template<class rrtype,class ltype,class rtype,int N>
-strong_inline void mac(iVector<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iVector<rtype,N> * __restrict__ rhs)
+accelerator_inline void mac(iVector<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iVector<rtype,N> * __restrict__ rhs)
 {
-    for(int c1=0;c1<N;c1++){
+  for(int c1=0;c1<N;c1++){
     for(int c2=0;c2<N;c2++){
-        mac(&ret->_internal[c1],&lhs->_internal[c1][c2],&rhs->_internal[c2]);
+      mac(&ret->_internal[c1],&lhs->_internal[c1][c2],&rhs->_internal[c2]);
     }}
-    return;
+  return;
 }
 template<class rrtype,class ltype,class rtype,int N>
-strong_inline void mac(iVector<rrtype,N> * __restrict__ ret,const iScalar<ltype> * __restrict__ lhs,const iVector<rtype,N> * __restrict__ rhs)
+accelerator_inline void mac(iVector<rrtype,N> * __restrict__ ret,const iScalar<ltype> * __restrict__ lhs,const iVector<rtype,N> * __restrict__ rhs)
 {
-    for(int c1=0;c1<N;c1++){
-        mac(&ret->_internal[c1],&lhs->_internal,&rhs->_internal[c1]);
-    }
-    return;
+  for(int c1=0;c1<N;c1++){
+    mac(&ret->_internal[c1],&lhs->_internal,&rhs->_internal[c1]);
+  }
+  return;
 }
 template<class rrtype,class ltype,class rtype,int N>
-strong_inline void mac(iVector<rrtype,N> * __restrict__ ret,const iVector<ltype,N> * __restrict__ lhs,const iScalar<rtype> * __restrict__ rhs)
+accelerator_inline void mac(iVector<rrtype,N> * __restrict__ ret,const iVector<ltype,N> * __restrict__ lhs,const iScalar<rtype> * __restrict__ rhs)
 {
-    for(int c1=0;c1<N;c1++){
-        mac(&ret->_internal[c1],&lhs->_internal[c1],&rhs->_internal);
-    }
-    return;
-}
+  for(int c1=0;c1<N;c1++){
+    mac(&ret->_internal[c1],&lhs->_internal[c1],&rhs->_internal);
+  }
+  return;
 }
+NAMESPACE_END(Grid);
+
 
 #endif
diff --git a/Grid/tensors/Tensor_arith_mul.h b/Grid/tensors/Tensor_arith_mul.h
index a474db9c..74dd3bf2 100644
--- a/Grid/tensors/Tensor_arith_mul.h
+++ b/Grid/tensors/Tensor_arith_mul.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,25 +23,24 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_ARITH_MUL_H
 #define GRID_MATH_ARITH_MUL_H
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-
-    ///////////////////////////////////////////////////////////////////////////////////////////////////
-    /////////////////////////////////////////// MUL         ///////////////////////////////////////////
-    ///////////////////////////////////////////////////////////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////////////////////////
+/////////////////////////////////////////// MUL         ///////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////////////////////////
     
 template<class rtype,class vtype,class mtype>
-strong_inline void mult(iScalar<rtype> * __restrict__ ret,const iScalar<mtype> * __restrict__ lhs,const iScalar<vtype> * __restrict__ rhs){
-    mult(&ret->_internal,&lhs->_internal,&rhs->_internal);
+accelerator_inline void mult(iScalar<rtype> * __restrict__ ret,const iScalar<mtype> * __restrict__ lhs,const iScalar<vtype> * __restrict__ rhs){
+  mult(&ret->_internal,&lhs->_internal,&rhs->_internal);
 }
 
 template<class rrtype,class ltype,class rtype,int N>
-strong_inline void mult(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iMatrix<rtype,N> * __restrict__ rhs){
+accelerator_inline void mult(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iMatrix<rtype,N> * __restrict__ rhs){
   for(int c1=0;c1<N;c1++){
     for(int c2=0;c2<N;c2++){
       mult(&ret->_internal[c1][c2],&lhs->_internal[c1][0],&rhs->_internal[0][c2]);
@@ -58,196 +57,196 @@ strong_inline void mult(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype
 }
 
 template<class rrtype,class ltype,class rtype,int N>
-strong_inline void mult(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iScalar<rtype> * __restrict__ rhs){
-    for(int c2=0;c2<N;c2++){
+accelerator_inline void mult(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iScalar<rtype> * __restrict__ rhs){
+  for(int c2=0;c2<N;c2++){
     for(int c1=0;c1<N;c1++){
-        mult(&ret->_internal[c1][c2],&lhs->_internal[c1][c2],&rhs->_internal);
+      mult(&ret->_internal[c1][c2],&lhs->_internal[c1][c2],&rhs->_internal);
     }}
-    return;
+  return;
 }
 
 template<class rrtype,class ltype,class rtype, int N>
-strong_inline void mult(iMatrix<rrtype,N> * __restrict__ ret,const iScalar<ltype>   * __restrict__ lhs,const iMatrix<rtype,N> * __restrict__ rhs){
-    for(int c2=0;c2<N;c2++){
+accelerator_inline void mult(iMatrix<rrtype,N> * __restrict__ ret,const iScalar<ltype>   * __restrict__ lhs,const iMatrix<rtype,N> * __restrict__ rhs){
+  for(int c2=0;c2<N;c2++){
     for(int c1=0;c1<N;c1++){
-        mult(&ret->_internal[c1][c2],&lhs->_internal,&rhs->_internal[c1][c2]);
+      mult(&ret->_internal[c1][c2],&lhs->_internal,&rhs->_internal[c1][c2]);
     }}
-    return;
+  return;
 }
 // Matrix left multiplies vector
 template<class rtype,class vtype,class mtype,int N>
-strong_inline void mult(iVector<rtype,N> * __restrict__ ret,const iMatrix<mtype,N> * __restrict__ lhs,const iVector<vtype,N> * __restrict__ rhs)
+accelerator_inline void mult(iVector<rtype,N> * __restrict__ ret,const iMatrix<mtype,N> * __restrict__ lhs,const iVector<vtype,N> * __restrict__ rhs)
 {
-    for(int c1=0;c1<N;c1++){
-        mult(&ret->_internal[c1],&lhs->_internal[c1][0],&rhs->_internal[0]);
-        for(int c2=1;c2<N;c2++){
-            mac(&ret->_internal[c1],&lhs->_internal[c1][c2],&rhs->_internal[c2]);
-        }
+  for(int c1=0;c1<N;c1++){
+    mult(&ret->_internal[c1],&lhs->_internal[c1][0],&rhs->_internal[0]);
+    for(int c2=1;c2<N;c2++){
+      mac(&ret->_internal[c1],&lhs->_internal[c1][c2],&rhs->_internal[c2]);
     }
-    return;
+  }
+  return;
 }
 template<class rtype,class vtype,class mtype,int N>
-strong_inline void mult(iVector<rtype,N> * __restrict__ ret,
-                 const iScalar<mtype>   * __restrict__ lhs,
-                 const iVector<vtype,N> * __restrict__ rhs){
-    for(int c1=0;c1<N;c1++){
-        mult(&ret->_internal[c1],&lhs->_internal,&rhs->_internal[c1]);
-    }
+accelerator_inline void mult(iVector<rtype,N> * __restrict__ ret,
+			const iScalar<mtype>   * __restrict__ lhs,
+			const iVector<vtype,N> * __restrict__ rhs){
+  for(int c1=0;c1<N;c1++){
+    mult(&ret->_internal[c1],&lhs->_internal,&rhs->_internal[c1]);
+  }
 }
 template<class rtype,class vtype,class mtype,int N>
-strong_inline void mult(iVector<rtype,N> * __restrict__ ret,
-                 const iVector<vtype,N> * __restrict__ rhs,
-                 const iScalar<mtype> * __restrict__ lhs){
-    for(int c1=0;c1<N;c1++){
-        mult(&ret->_internal[c1],&rhs->_internal[c1],&lhs->_internal);
-    }                 
+accelerator_inline void mult(iVector<rtype,N> * __restrict__ ret,
+			const iVector<vtype,N> * __restrict__ rhs,
+			const iScalar<mtype> * __restrict__ lhs){
+  for(int c1=0;c1<N;c1++){
+    mult(&ret->_internal[c1],&rhs->_internal[c1],&lhs->_internal);
+  }                 
 }
     
 
 
-template<class rtype,class vtype,class mtype,int N> strong_inline
+template<class rtype,class vtype,class mtype,int N> accelerator_inline
 iVector<rtype,N> operator * (const iMatrix<mtype,N>& lhs,const iVector<vtype,N>& rhs)
 {
-    iVector<rtype,N> ret;
-    mult(&ret,&lhs,&rhs);
-    return ret;
+  iVector<rtype,N> ret;
+  mult(&ret,&lhs,&rhs);
+  return ret;
 }
 
-template<class rtype,class vtype,class mtype,int N> strong_inline
+template<class rtype,class vtype,class mtype,int N> accelerator_inline
 iVector<rtype,N> operator * (const iScalar<mtype>& lhs,const iVector<vtype,N>& rhs)
 {
-    iVector<rtype,N> ret;
-    mult(&ret,&lhs,&rhs);
-    return ret;
+  iVector<rtype,N> ret;
+  mult(&ret,&lhs,&rhs);
+  return ret;
 }
 
-template<class rtype,class vtype,class mtype,int N> strong_inline
+template<class rtype,class vtype,class mtype,int N> accelerator_inline
 iVector<rtype,N> operator * (const iVector<mtype,N>& lhs,const iScalar<vtype>& rhs)
 {
-    iVector<rtype,N> ret;
-    mult(&ret,&lhs,&rhs);
-    return ret;
+  iVector<rtype,N> ret;
+  mult(&ret,&lhs,&rhs);
+  return ret;
 }
 
 //////////////////////////////////////////////////////////////////
 // Divide by scalar
 //////////////////////////////////////////////////////////////////
-template<class rtype,class vtype> strong_inline
+template<class rtype,class vtype> accelerator_inline
 iScalar<rtype> operator / (const iScalar<rtype>& lhs,const iScalar<vtype>& rhs)
 {
-    iScalar<rtype> ret;
-    ret._internal = lhs._internal/rhs._internal;
-    return ret;
+  iScalar<rtype> ret;
+  ret._internal = lhs._internal/rhs._internal;
+  return ret;
 }
-template<class rtype,class vtype,int N> strong_inline
+template<class rtype,class vtype,int N> accelerator_inline
 iVector<rtype,N> operator / (const iVector<rtype,N>& lhs,const iScalar<vtype>& rhs)
 {
-    iVector<rtype,N> ret;
-    for(int i=0;i<N;i++){
-      ret._internal[i] = lhs._internal[i]/rhs._internal;
-    }
-    return ret;
+  iVector<rtype,N> ret;
+  for(int i=0;i<N;i++){
+    ret._internal[i] = lhs._internal[i]/rhs._internal;
+  }
+  return ret;
 }
-template<class rtype,class vtype,int N> strong_inline
+template<class rtype,class vtype,int N> accelerator_inline
 iMatrix<rtype,N> operator / (const iMatrix<rtype,N>& lhs,const iScalar<vtype>& rhs)
 {
-    iMatrix<rtype,N> ret;
-    for(int i=0;i<N;i++){
+  iMatrix<rtype,N> ret;
+  for(int i=0;i<N;i++){
     for(int j=0;j<N;j++){
       ret._internal[i][j] = lhs._internal[i][j]/rhs._internal;
     }}
-    return ret;
+  return ret;
 }
     
-    //////////////////////////////////////////////////////////////////
-    // Glue operators to mult routines. Must resolve return type cleverly from typeof(internal)
-    // since nesting matrix<scalar> x matrix<matrix>-> matrix<matrix>
-    // while         matrix<scalar> x matrix<scalar>-> matrix<scalar>
-    // so return type depends on argument types in nasty way.
-    //////////////////////////////////////////////////////////////////
-    // scal x scal = scal
-    // mat x  mat  = mat
-    // mat  x scal = mat
-    // scal x mat  = mat
-    // mat  x vec  = vec
-    // vec  x scal = vec
-    // scal x vec  = vec
-    //
-    // We can special case scalar_type ??
+//////////////////////////////////////////////////////////////////
+// Glue operators to mult routines. Must resolve return type cleverly from typeof(internal)
+// since nesting matrix<scalar> x matrix<matrix>-> matrix<matrix>
+// while         matrix<scalar> x matrix<scalar>-> matrix<scalar>
+// so return type depends on argument types in nasty way.
+//////////////////////////////////////////////////////////////////
+// scal x scal = scal
+// mat x  mat  = mat
+// mat  x scal = mat
+// scal x mat  = mat
+// mat  x vec  = vec
+// vec  x scal = vec
+// scal x vec  = vec
+//
+// We can special case scalar_type ??
 template<class l,class r>
-strong_inline auto operator * (const iScalar<l>& lhs,const iScalar<r>& rhs) -> iScalar<decltype(lhs._internal * rhs._internal)>
+accelerator_inline auto operator * (const iScalar<l>& lhs,const iScalar<r>& rhs) -> iScalar<decltype(lhs._internal * rhs._internal)>
 {
-    typedef iScalar<decltype(lhs._internal*rhs._internal)> ret_t;
-    ret_t ret;
-    mult(&ret,&lhs,&rhs);
-    return ret;
+  typedef iScalar<decltype(lhs._internal*rhs._internal)> ret_t;
+  ret_t ret;
+  mult(&ret,&lhs,&rhs);
+  return ret;
 }
-template<class l,class r,int N> strong_inline
+template<class l,class r,int N> accelerator_inline
 auto operator * (const iMatrix<l,N>& lhs,const iMatrix<r,N>& rhs) -> iMatrix<decltype(lhs._internal[0][0]*rhs._internal[0][0]),N>
 {
-    typedef decltype(lhs._internal[0][0]*rhs._internal[0][0]) ret_t;
-    iMatrix<ret_t,N> ret;
-    mult(&ret,&lhs,&rhs);
-    return ret;
+  typedef decltype(lhs._internal[0][0]*rhs._internal[0][0]) ret_t;
+  iMatrix<ret_t,N> ret;
+  mult(&ret,&lhs,&rhs);
+  return ret;
 }
-template<class l,class r, int N> strong_inline
+template<class l,class r, int N> accelerator_inline
 auto operator * (const iMatrix<r,N>& lhs,const iScalar<l>& rhs) -> iMatrix<decltype(lhs._internal[0][0]*rhs._internal),N>
 {
-    typedef decltype(lhs._internal[0][0]*rhs._internal) ret_t;
+  typedef decltype(lhs._internal[0][0]*rhs._internal) ret_t;
         
-    iMatrix<ret_t,N> ret;
-    for(int c1=0;c1<N;c1++){
+  iMatrix<ret_t,N> ret;
+  for(int c1=0;c1<N;c1++){
     for(int c2=0;c2<N;c2++){
-        mult(&ret._internal[c1][c2],&lhs._internal[c1][c2],&rhs._internal);
+      mult(&ret._internal[c1][c2],&lhs._internal[c1][c2],&rhs._internal);
     }}
-    return ret;
+  return ret;
 }
-template<class l,class r,int N> strong_inline
+template<class l,class r,int N> accelerator_inline
 auto operator * (const iScalar<l>& lhs,const iMatrix<r,N>& rhs) -> iMatrix<decltype(lhs._internal*rhs._internal[0][0]),N>
 {
-    typedef decltype(lhs._internal*rhs._internal[0][0]) ret_t;
-    iMatrix<ret_t,N> ret;
-    for(int c1=0;c1<N;c1++){
+  typedef decltype(lhs._internal*rhs._internal[0][0]) ret_t;
+  iMatrix<ret_t,N> ret;
+  for(int c1=0;c1<N;c1++){
     for(int c2=0;c2<N;c2++){
-        mult(&ret._internal[c1][c2],&lhs._internal,&rhs._internal[c1][c2]);
+      mult(&ret._internal[c1][c2],&lhs._internal,&rhs._internal[c1][c2]);
     }}
-    return ret;
+  return ret;
 }
-template<class l,class r,int N> strong_inline
+template<class l,class r,int N> accelerator_inline
 auto operator * (const iMatrix<l,N>& lhs,const iVector<r,N>& rhs) -> iVector<decltype(lhs._internal[0][0]*rhs._internal[0]),N>
 {
-    typedef decltype(lhs._internal[0][0]*rhs._internal[0]) ret_t;
-    iVector<ret_t,N> ret;
-    for(int c1=0;c1<N;c1++){
-        mult(&ret._internal[c1],&lhs._internal[c1][0],&rhs._internal[0]);
-        for(int c2=1;c2<N;c2++){
-            mac(&ret._internal[c1],&lhs._internal[c1][c2],&rhs._internal[c2]);
-        }
+  typedef decltype(lhs._internal[0][0]*rhs._internal[0]) ret_t;
+  iVector<ret_t,N> ret;
+  for(int c1=0;c1<N;c1++){
+    mult(&ret._internal[c1],&lhs._internal[c1][0],&rhs._internal[0]);
+    for(int c2=1;c2<N;c2++){
+      mac(&ret._internal[c1],&lhs._internal[c1][c2],&rhs._internal[c2]);
     }
-    return ret;
+  }
+  return ret;
 }
-template<class l,class r,int N> strong_inline
+template<class l,class r,int N> accelerator_inline
 auto operator * (const iScalar<l>& lhs,const iVector<r,N>& rhs) -> iVector<decltype(lhs._internal*rhs._internal[0]),N>
 {
-    typedef decltype(lhs._internal*rhs._internal[0]) ret_t;
-    iVector<ret_t,N> ret;
-    for(int c1=0;c1<N;c1++){
-        mult(&ret._internal[c1],&lhs._internal,&rhs._internal[c1]);
-    }
-    return ret;
+  typedef decltype(lhs._internal*rhs._internal[0]) ret_t;
+  iVector<ret_t,N> ret;
+  for(int c1=0;c1<N;c1++){
+    mult(&ret._internal[c1],&lhs._internal,&rhs._internal[c1]);
+  }
+  return ret;
 }
-template<class l,class r,int N> strong_inline
+template<class l,class r,int N> accelerator_inline
 auto operator * (const iVector<l,N>& lhs,const iScalar<r>& rhs) -> iVector<decltype(lhs._internal[0]*rhs._internal),N>
 {
-    typedef decltype(lhs._internal[0]*rhs._internal) ret_t;
-    iVector<ret_t,N> ret;
-    for(int c1=0;c1<N;c1++){
-        mult(&ret._internal[c1],&lhs._internal[c1],&rhs._internal);
-    }
-    return ret;
+  typedef decltype(lhs._internal[0]*rhs._internal) ret_t;
+  iVector<ret_t,N> ret;
+  for(int c1=0;c1<N;c1++){
+    mult(&ret._internal[c1],&lhs._internal[c1],&rhs._internal);
+  }
+  return ret;
 }
 
+NAMESPACE_END(Grid);
 
-}
 
 #endif
diff --git a/Grid/tensors/Tensor_arith_scalar.h b/Grid/tensors/Tensor_arith_scalar.h
index d1087570..a9bb0310 100644
--- a/Grid/tensors/Tensor_arith_scalar.h
+++ b/Grid/tensors/Tensor_arith_scalar.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,71 +24,70 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_ARITH_SCALAR_H
 #define GRID_MATH_ARITH_SCALAR_H
 
-namespace Grid {
-
+NAMESPACE_BEGIN(Grid);
 
 //////////////////////////////////////////////////////////////////////////////////////////
 // Must support native C++ types Integer, Complex, Real
 //////////////////////////////////////////////////////////////////////////////////////////
 
 // multiplication by fundamental scalar type
-template<class l> strong_inline iScalar<l> operator * (const iScalar<l>& lhs,const typename iScalar<l>::scalar_type rhs) 
+template<class l> accelerator_inline iScalar<l> operator * (const iScalar<l>& lhs,const typename iScalar<l>::scalar_type rhs) 
 {
   typename iScalar<l>::tensor_reduced srhs; srhs=rhs;
   return lhs*srhs;
 }
-template<class l> strong_inline iScalar<l> operator * (const typename iScalar<l>::scalar_type lhs,const iScalar<l>& rhs) {  return rhs*lhs; }
+template<class l> accelerator_inline iScalar<l> operator * (const typename iScalar<l>::scalar_type lhs,const iScalar<l>& rhs) {  return rhs*lhs; }
 
-template<class l,int N> strong_inline iVector<l,N> operator * (const iVector<l,N>& lhs,const typename iScalar<l>::scalar_type rhs) 
+template<class l,int N> accelerator_inline iVector<l,N> operator * (const iVector<l,N>& lhs,const typename iScalar<l>::scalar_type rhs) 
 {
   typename iVector<l,N>::tensor_reduced srhs; srhs=rhs;
   return lhs*srhs;
 }
-template<class l,int N> strong_inline iVector<l,N> operator * (const typename iScalar<l>::scalar_type lhs,const iVector<l,N>& rhs) {  return rhs*lhs; }
+template<class l,int N> accelerator_inline iVector<l,N> operator * (const typename iScalar<l>::scalar_type lhs,const iVector<l,N>& rhs) {  return rhs*lhs; }
 
-template<class l,int N> strong_inline iMatrix<l,N> operator * (const iMatrix<l,N>& lhs,const typename iScalar<l>::scalar_type &rhs) 
+template<class l,int N> accelerator_inline iMatrix<l,N> operator * (const iMatrix<l,N>& lhs,const typename iScalar<l>::scalar_type &rhs) 
 {
   typename iMatrix<l,N>::tensor_reduced srhs; srhs=rhs;
   return lhs*srhs;
 }
-template<class l,int N> strong_inline iMatrix<l,N> operator * (const typename iScalar<l>::scalar_type & lhs,const iMatrix<l,N>& rhs) {  return rhs*lhs; }
+template<class l,int N> accelerator_inline iMatrix<l,N> operator * (const typename iScalar<l>::scalar_type & lhs,const iMatrix<l,N>& rhs) {  return rhs*lhs; }
 
 ////////////////////////////////////////////////////////////////////
 // Double support; cast to "scalar_type" through constructor
 ////////////////////////////////////////////////////////////////////
-template<class l> strong_inline iScalar<l> operator * (const iScalar<l>& lhs,double rhs) 
+template<class l> accelerator_inline iScalar<l> operator * (const iScalar<l>& lhs,double rhs) 
 {
   typename iScalar<l>::scalar_type t; t=rhs;
   typename iScalar<l>::tensor_reduced srhs;srhs=t;
   return lhs*srhs;
 }
-template<class l> strong_inline iScalar<l> operator * (double lhs,const iScalar<l>& rhs) {  return rhs*lhs; }
+template<class l> accelerator_inline iScalar<l> operator * (double lhs,const iScalar<l>& rhs) {  return rhs*lhs; }
 
-template<class l,int N> strong_inline iVector<l,N> operator * (const iVector<l,N>& lhs,double rhs) 
+template<class l,int N> accelerator_inline iVector<l,N> operator * (const iVector<l,N>& lhs,double rhs) 
 {
   typename iScalar<l>::scalar_type t;t=rhs;
   typename iScalar<l>::tensor_reduced srhs;srhs=t;
   return lhs*srhs;
 }
-template<class l,int N> strong_inline iVector<l,N> operator * (double lhs,const iVector<l,N>& rhs) {  return rhs*lhs; }
+template<class l,int N> accelerator_inline iVector<l,N> operator * (double lhs,const iVector<l,N>& rhs) {  return rhs*lhs; }
 
-template<class l,int N> strong_inline iMatrix<l,N> operator * (const iMatrix<l,N>& lhs,double rhs) 
+template<class l,int N> accelerator_inline iMatrix<l,N> operator * (const iMatrix<l,N>& lhs,double rhs) 
 {
   typename iScalar<l>::scalar_type t;t=rhs;
   typename iScalar<l>::tensor_reduced srhs;srhs=t;
   return lhs*srhs;
 }
-template<class l,int N> strong_inline iMatrix<l,N> operator * (double lhs,const iMatrix<l,N>& rhs) {  return rhs*lhs; }
+template<class l,int N> accelerator_inline iMatrix<l,N> operator * (double lhs,const iMatrix<l,N>& rhs) {  return rhs*lhs; }
 
 ////////////////////////////////////////////////////////////////////
 // Complex support; cast to "scalar_type" through constructor
 ////////////////////////////////////////////////////////////////////
-template<class l> strong_inline iScalar<l> operator * (const iScalar<l>& lhs,ComplexD rhs) 
+template<class l> accelerator_inline iScalar<l> operator * (const iScalar<l>& lhs,ComplexD rhs) 
 {
   typename iScalar<l>::scalar_type t;t=rhs;
   typename iScalar<l>::tensor_reduced srhs;srhs=t;
@@ -96,131 +95,131 @@ template<class l> strong_inline iScalar<l> operator * (const iScalar<l>& lhs,Com
   
   return lhs*srhs;
 }
-template<class l> strong_inline iScalar<l> operator * (ComplexD lhs,const iScalar<l>& rhs) {  return rhs*lhs; }
+template<class l> accelerator_inline iScalar<l> operator * (ComplexD lhs,const iScalar<l>& rhs) {  return rhs*lhs; }
 
-template<class l,int N> strong_inline iVector<l,N> operator * (const iVector<l,N>& lhs,ComplexD rhs) 
+template<class l,int N> accelerator_inline iVector<l,N> operator * (const iVector<l,N>& lhs,ComplexD rhs) 
 {
   typename iScalar<l>::scalar_type t;t=rhs;
   typename iScalar<l>::tensor_reduced srhs;srhs=t;
   return lhs*srhs;
 }
-template<class l,int N> strong_inline iVector<l,N> operator * (ComplexD lhs,const iVector<l,N>& rhs) {  return rhs*lhs; }
+template<class l,int N> accelerator_inline iVector<l,N> operator * (ComplexD lhs,const iVector<l,N>& rhs) {  return rhs*lhs; }
 
-template<class l,int N> strong_inline iMatrix<l,N> operator * (const iMatrix<l,N>& lhs,ComplexD rhs) 
+template<class l,int N> accelerator_inline iMatrix<l,N> operator * (const iMatrix<l,N>& lhs,ComplexD rhs) 
 {
   typename iScalar<l>::scalar_type t;t=rhs;
   typename iScalar<l>::tensor_reduced srhs;srhs=t;
   return lhs*srhs;
 }
-template<class l,int N> strong_inline iMatrix<l,N> operator * (ComplexD lhs,const iMatrix<l,N>& rhs) {  return rhs*lhs; }
+template<class l,int N> accelerator_inline iMatrix<l,N> operator * (ComplexD lhs,const iMatrix<l,N>& rhs) {  return rhs*lhs; }
 
 ////////////////////////////////////////////////////////////////////
 // Integer support; cast to "scalar_type" through constructor
 ////////////////////////////////////////////////////////////////////
-template<class l> strong_inline iScalar<l> operator * (const iScalar<l>& lhs,Integer rhs) 
+template<class l> accelerator_inline iScalar<l> operator * (const iScalar<l>& lhs,Integer rhs) 
 {
   typename iScalar<l>::scalar_type t;  t=rhs;
   typename iScalar<l>::tensor_reduced srhs; srhs=t;
   return lhs*srhs;
 }
-template<class l> strong_inline iScalar<l> operator * (Integer lhs,const iScalar<l>& rhs) {  return rhs*lhs; }
+template<class l> accelerator_inline iScalar<l> operator * (Integer lhs,const iScalar<l>& rhs) {  return rhs*lhs; }
 
-template<class l,int N> strong_inline iVector<l,N> operator * (const iVector<l,N>& lhs,Integer rhs) 
+template<class l,int N> accelerator_inline iVector<l,N> operator * (const iVector<l,N>& lhs,Integer rhs) 
 {
   typename iScalar<l>::scalar_type t;t=rhs;
   typename iScalar<l>::tensor_reduced srhs;srhs=t;
   return lhs*srhs;
 }
-template<class l,int N> strong_inline iVector<l,N> operator * (Integer lhs,const iVector<l,N>& rhs) {  return rhs*lhs; }
+template<class l,int N> accelerator_inline iVector<l,N> operator * (Integer lhs,const iVector<l,N>& rhs) {  return rhs*lhs; }
 
-template<class l,int N> strong_inline iMatrix<l,N> operator * (const iMatrix<l,N>& lhs,Integer rhs) 
+template<class l,int N> accelerator_inline iMatrix<l,N> operator * (const iMatrix<l,N>& lhs,Integer rhs) 
 {
   typename iScalar<l>::scalar_type t;t=rhs;
   typename iScalar<l>::tensor_reduced srhs;srhs=t;
   return lhs*srhs;
 }
-template<class l,int N> strong_inline iMatrix<l,N> operator * (Integer lhs,const iMatrix<l,N>& rhs) {  return rhs*lhs; }
+template<class l,int N> accelerator_inline iMatrix<l,N> operator * (Integer lhs,const iMatrix<l,N>& rhs) {  return rhs*lhs; }
 
 
 
 ///////////////////////////////////////////////////////////////////////////////////////////////
 // addition by fundamental scalar type applies to matrix(down diag) and scalar
 ///////////////////////////////////////////////////////////////////////////////////////////////
-template<class l> strong_inline iScalar<l> operator + (const iScalar<l>& lhs,const typename iScalar<l>::scalar_type rhs) 
+template<class l> accelerator_inline iScalar<l> operator + (const iScalar<l>& lhs,const typename iScalar<l>::scalar_type rhs) 
 {
   typename iScalar<l>::tensor_reduced srhs; srhs=rhs;
   return lhs+srhs;
 }
-template<class l> strong_inline iScalar<l> operator + (const typename iScalar<l>::scalar_type lhs,const iScalar<l>& rhs) {  return rhs+lhs; }
+template<class l> accelerator_inline iScalar<l> operator + (const typename iScalar<l>::scalar_type lhs,const iScalar<l>& rhs) {  return rhs+lhs; }
 
-template<class l,int N> strong_inline iMatrix<l,N> operator + (const iMatrix<l,N>& lhs,const typename iScalar<l>::scalar_type rhs) 
+template<class l,int N> accelerator_inline iMatrix<l,N> operator + (const iMatrix<l,N>& lhs,const typename iScalar<l>::scalar_type rhs) 
 {
   typename iMatrix<l,N>::tensor_reduced srhs; srhs=rhs;
   return lhs+srhs;
 }
-template<class l,int N> strong_inline iMatrix<l,N> operator + (const typename iScalar<l>::scalar_type lhs,const iMatrix<l,N>& rhs) {  return rhs+lhs; }
+template<class l,int N> accelerator_inline iMatrix<l,N> operator + (const typename iScalar<l>::scalar_type lhs,const iMatrix<l,N>& rhs) {  return rhs+lhs; }
 
 ////////////////////////////////////////////////////////////////////
 // Double support; cast to "scalar_type" through constructor
 ////////////////////////////////////////////////////////////////////
-template<class l> strong_inline iScalar<l> operator + (const iScalar<l>& lhs,double rhs) 
+template<class l> accelerator_inline iScalar<l> operator + (const iScalar<l>& lhs,double rhs) 
 {
   typename iScalar<l>::scalar_type t; t=rhs;
   typename iScalar<l>::tensor_reduced srhs; srhs=t;
   return lhs+srhs;
 }
-template<class l> strong_inline iScalar<l> operator + (double lhs,const iScalar<l>& rhs) {  return rhs+lhs; }
+template<class l> accelerator_inline iScalar<l> operator + (double lhs,const iScalar<l>& rhs) {  return rhs+lhs; }
 
-template<class l,int N> strong_inline iMatrix<l,N> operator + (const iMatrix<l,N>& lhs,double rhs) 
+template<class l,int N> accelerator_inline iMatrix<l,N> operator + (const iMatrix<l,N>& lhs,double rhs) 
 {
   typename iScalar<l>::scalar_type t;t=rhs;
   typename iScalar<l>::tensor_reduced srhs;srhs=t;
   return lhs+srhs;
 }
-template<class l,int N> strong_inline iMatrix<l,N> operator + (double lhs,const iMatrix<l,N>& rhs) {  return rhs+lhs; }
+template<class l,int N> accelerator_inline iMatrix<l,N> operator + (double lhs,const iMatrix<l,N>& rhs) {  return rhs+lhs; }
 
 
 // Integer support cast to scalar type through constructor
 
 
-template<class l> strong_inline iScalar<l> operator + (const iScalar<l>& lhs,Integer rhs) 
+template<class l> accelerator_inline iScalar<l> operator + (const iScalar<l>& lhs,Integer rhs) 
 {
   typename iScalar<l>::scalar_type t; t=rhs;
   typename iScalar<l>::tensor_reduced srhs; srhs=t;
   return lhs+srhs;
 }
 
-template<class l> strong_inline iScalar<l> operator + (Integer lhs,const iScalar<l>& rhs) {  return rhs+lhs; }
+template<class l> accelerator_inline iScalar<l> operator + (Integer lhs,const iScalar<l>& rhs) {  return rhs+lhs; }
 
-template<class l,int N> strong_inline iMatrix<l,N> operator + (const iMatrix<l,N>& lhs,Integer rhs) 
+template<class l,int N> accelerator_inline iMatrix<l,N> operator + (const iMatrix<l,N>& lhs,Integer rhs) 
 {
   typename iScalar<l>::scalar_type t;t=rhs;
   typename iScalar<l>::tensor_reduced srhs;srhs=t;
   return lhs+srhs;
 }
-template<class l,int N> strong_inline iMatrix<l,N> operator + (Integer lhs,const iMatrix<l,N>& rhs) {  return rhs+lhs; }
+template<class l,int N> accelerator_inline iMatrix<l,N> operator + (Integer lhs,const iMatrix<l,N>& rhs) {  return rhs+lhs; }
 
 
 ///////////////////////////////////////////////////////////////////////////////////////////////
 // subtraction of fundamental scalar type applies to matrix(down diag) and scalar
 ///////////////////////////////////////////////////////////////////////////////////////////////
-template<class l> strong_inline iScalar<l> operator - (const iScalar<l>& lhs,const typename iScalar<l>::scalar_type rhs) 
+template<class l> accelerator_inline iScalar<l> operator - (const iScalar<l>& lhs,const typename iScalar<l>::scalar_type rhs) 
 {
   typename iScalar<l>::tensor_reduced srhs; srhs=rhs;
   return lhs-srhs;
 }
-template<class l> strong_inline iScalar<l> operator - (const typename iScalar<l>::scalar_type lhs,const iScalar<l>& rhs) 
+template<class l> accelerator_inline iScalar<l> operator - (const typename iScalar<l>::scalar_type lhs,const iScalar<l>& rhs) 
 {
   typename iScalar<l>::tensor_reduced slhs;slhs=lhs;
   return slhs-rhs;
 }
 
-template<class l,int N> strong_inline iMatrix<l,N> operator - (const iMatrix<l,N>& lhs,const typename iScalar<l>::scalar_type rhs) 
+template<class l,int N> accelerator_inline iMatrix<l,N> operator - (const iMatrix<l,N>& lhs,const typename iScalar<l>::scalar_type rhs) 
 {
   typename iScalar<l>::tensor_reduced srhs; srhs=rhs;
   return lhs-srhs;
 }
-template<class l,int N> strong_inline iMatrix<l,N> operator - (const typename iScalar<l>::scalar_type lhs,const iMatrix<l,N>& rhs) 
+template<class l,int N> accelerator_inline iMatrix<l,N> operator - (const typename iScalar<l>::scalar_type lhs,const iMatrix<l,N>& rhs) 
 {
   typename iScalar<l>::tensor_reduced slhs;slhs=lhs;
   return slhs-rhs;
@@ -229,26 +228,26 @@ template<class l,int N> strong_inline iMatrix<l,N> operator - (const typename iS
 ////////////////////////////////////////////////////////////////////
 // Double support; cast to "scalar_type" through constructor
 ////////////////////////////////////////////////////////////////////
-template<class l> strong_inline iScalar<l> operator - (const iScalar<l>& lhs,double rhs) 
+template<class l> accelerator_inline iScalar<l> operator - (const iScalar<l>& lhs,double rhs) 
 {
   typename iScalar<l>::scalar_type t; t=rhs;
   typename iScalar<l>::tensor_reduced srhs; srhs=t;
   return lhs-srhs;
 }
-template<class l> strong_inline iScalar<l> operator - (double lhs,const iScalar<l>& rhs) 
+template<class l> accelerator_inline iScalar<l> operator - (double lhs,const iScalar<l>& rhs) 
 {
   typename iScalar<l>::scalar_type t(lhs);
   typename iScalar<l>::tensor_reduced slhs;slhs=t;
   return slhs-rhs;
 }
 
-template<class l,int N> strong_inline iMatrix<l,N> operator - (const iMatrix<l,N>& lhs,double rhs) 
+template<class l,int N> accelerator_inline iMatrix<l,N> operator - (const iMatrix<l,N>& lhs,double rhs) 
 {
   typename iScalar<l>::scalar_type t;t=rhs;
   typename iScalar<l>::tensor_reduced srhs;srhs=t;
   return lhs-srhs;
 }
-template<class l,int N> strong_inline iMatrix<l,N> operator - (double lhs,const iMatrix<l,N>& rhs) 
+template<class l,int N> accelerator_inline iMatrix<l,N> operator - (double lhs,const iMatrix<l,N>& rhs) 
 {
   typename iScalar<l>::scalar_type t(lhs);
   typename iScalar<l>::tensor_reduced slhs;slhs=t;
@@ -258,31 +257,31 @@ template<class l,int N> strong_inline iMatrix<l,N> operator - (double lhs,const
 ////////////////////////////////////////////////////////////////////
 // Integer support; cast to "scalar_type" through constructor
 ////////////////////////////////////////////////////////////////////
-template<class l> strong_inline iScalar<l> operator - (const iScalar<l>& lhs,Integer rhs) 
+template<class l> accelerator_inline iScalar<l> operator - (const iScalar<l>& lhs,Integer rhs) 
 {
   typename iScalar<l>::scalar_type t; t=rhs;
   typename iScalar<l>::tensor_reduced srhs; srhs=t;
   return lhs-srhs;
 }
-template<class l> strong_inline iScalar<l> operator - (Integer lhs,const iScalar<l>& rhs) 
+template<class l> accelerator_inline iScalar<l> operator - (Integer lhs,const iScalar<l>& rhs) 
 {
   typename iScalar<l>::scalar_type t;t=lhs;
   typename iScalar<l>::tensor_reduced slhs;slhs=t;
   return slhs-rhs;
 }
-template<class l,int N> strong_inline iMatrix<l,N> operator - (const iMatrix<l,N>& lhs,Integer rhs) 
+template<class l,int N> accelerator_inline iMatrix<l,N> operator - (const iMatrix<l,N>& lhs,Integer rhs) 
 {
   typename iScalar<l>::scalar_type t;t=rhs;
   typename iScalar<l>::tensor_reduced srhs;srhs=t;
   return lhs-srhs;
 }
-template<class l,int N> strong_inline iMatrix<l,N> operator - (Integer lhs,const iMatrix<l,N>& rhs) 
+template<class l,int N> accelerator_inline iMatrix<l,N> operator - (Integer lhs,const iMatrix<l,N>& rhs) 
 {
   typename iScalar<l>::scalar_type t;t=lhs;
   typename iScalar<l>::tensor_reduced slhs;slhs=t;
   return slhs-rhs;
 }
 
+NAMESPACE_END(Grid);
 
-}
 #endif
diff --git a/Grid/tensors/Tensor_arith_sub.h b/Grid/tensors/Tensor_arith_sub.h
index 8646bdfb..37d00cea 100644
--- a/Grid/tensors/Tensor_arith_sub.h
+++ b/Grid/tensors/Tensor_arith_sub.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,17 +24,16 @@ Author: neo <cossu@post.kek.jp>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_ARITH_SUB_H
 #define GRID_MATH_ARITH_SUB_H
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-
-    ///////////////////////////////////////////////////////////////////////////////////////////////////
-    /////////////////////////////////////////// SUB         ///////////////////////////////////////////
-    ///////////////////////////////////////////////////////////////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////////////////////////
+/////////////////////////////////////////// SUB         ///////////////////////////////////////////
+///////////////////////////////////////////////////////////////////////////////////////////////////
     
 
 // SUB is simple for now; cannot mix types and straightforward template
@@ -42,103 +41,102 @@ namespace Grid {
 // Vector +/- Vector
 // Matrix +/- Matrix
 // Matrix /- scalar
-template<class vtype,class ltype,class rtype> strong_inline void sub(iScalar<vtype> * __restrict__ ret,
-                                                              const iScalar<ltype> * __restrict__ lhs,
-                                                              const iScalar<rtype> * __restrict__ rhs)
+template<class vtype,class ltype,class rtype> accelerator_inline void sub(iScalar<vtype> * __restrict__ ret,
+								     const iScalar<ltype> * __restrict__ lhs,
+								     const iScalar<rtype> * __restrict__ rhs)
 {
-    sub(&ret->_internal,&lhs->_internal,&rhs->_internal);
+  sub(&ret->_internal,&lhs->_internal,&rhs->_internal);
 }
 
-template<class vtype,class ltype,class rtype,int N> strong_inline void sub(iVector<vtype,N> * __restrict__ ret,
-                                                                    const iVector<ltype,N> * __restrict__ lhs,
-                                                                    const iVector<rtype,N> * __restrict__ rhs)
+template<class vtype,class ltype,class rtype,int N> accelerator_inline void sub(iVector<vtype,N> * __restrict__ ret,
+									   const iVector<ltype,N> * __restrict__ lhs,
+									   const iVector<rtype,N> * __restrict__ rhs)
 {
-    for(int c=0;c<N;c++){
-        ret->_internal[c]=lhs->_internal[c]-rhs->_internal[c];
-    }
-    return;
+  for(int c=0;c<N;c++){
+    ret->_internal[c]=lhs->_internal[c]-rhs->_internal[c];
+  }
+  return;
 }
-template<class vtype,class ltype,class rtype, int N> strong_inline void sub(iMatrix<vtype,N> * __restrict__ ret,
-                                                                     const iMatrix<ltype,N> * __restrict__ lhs,
-                                                                     const iMatrix<rtype,N> * __restrict__ rhs){
-    for(int c2=0;c2<N;c2++){
+template<class vtype,class ltype,class rtype, int N> accelerator_inline void sub(iMatrix<vtype,N> * __restrict__ ret,
+									    const iMatrix<ltype,N> * __restrict__ lhs,
+									    const iMatrix<rtype,N> * __restrict__ rhs){
+  for(int c2=0;c2<N;c2++){
     for(int c1=0;c1<N;c1++){
-        sub(&ret->_internal[c1][c2],&lhs->_internal[c1][c2],&rhs->_internal[c1][c2]);
+      sub(&ret->_internal[c1][c2],&lhs->_internal[c1][c2],&rhs->_internal[c1][c2]);
     }}
-    return;
+  return;
 }
-template<class vtype,class ltype,class rtype, int N> strong_inline void sub(iMatrix<vtype,N> * __restrict__ ret,
-                                                                     const iScalar<ltype> * __restrict__ lhs,
-                                                                     const iMatrix<rtype,N> * __restrict__ rhs){
-    for(int c2=0;c2<N;c2++){
+template<class vtype,class ltype,class rtype, int N> accelerator_inline void sub(iMatrix<vtype,N> * __restrict__ ret,
+									    const iScalar<ltype> * __restrict__ lhs,
+									    const iMatrix<rtype,N> * __restrict__ rhs){
+  for(int c2=0;c2<N;c2++){
     for(int c1=0;c1<N;c1++){
-        if ( c1==c2) {
-            sub(&ret->_internal[c1][c2],&lhs->_internal,&rhs->_internal[c1][c2]);
-        } else {
-            // Fails -- need unary minus. Catalogue other unops?
-            ret->_internal[c1][c2]=zero;
-            ret->_internal[c1][c2]=ret->_internal[c1][c2]-rhs->_internal[c1][c2];
+      if ( c1==c2) {
+	sub(&ret->_internal[c1][c2],&lhs->_internal,&rhs->_internal[c1][c2]);
+      } else {
+	// Fails -- need unary minus. Catalogue other unops?
+	ret->_internal[c1][c2]=Zero();
+	ret->_internal[c1][c2]=ret->_internal[c1][c2]-rhs->_internal[c1][c2];
 
-        }
+      }
     }}
-    return;
+  return;
 }
-template<class vtype,class ltype,class rtype, int N> strong_inline void sub(iMatrix<vtype,N> * __restrict__ ret,
-                                                                     const iMatrix<ltype,N> * __restrict__ lhs,
-                                                                     const iScalar<rtype> * __restrict__ rhs){
-    for(int c2=0;c2<N;c2++){
+template<class vtype,class ltype,class rtype, int N> accelerator_inline void sub(iMatrix<vtype,N> * __restrict__ ret,
+									    const iMatrix<ltype,N> * __restrict__ lhs,
+									    const iScalar<rtype> * __restrict__ rhs){
+  for(int c2=0;c2<N;c2++){
     for(int c1=0;c1<N;c1++){
-        if ( c1==c2)
-            sub(&ret->_internal[c1][c2],&lhs->_internal[c1][c2],&rhs->_internal);
-        else
-            ret->_internal[c1][c2]=lhs->_internal[c1][c2];
+      if ( c1==c2)
+	sub(&ret->_internal[c1][c2],&lhs->_internal[c1][c2],&rhs->_internal);
+      else
+	ret->_internal[c1][c2]=lhs->_internal[c1][c2];
     }}
-    return;
+  return;
 }
 
-    // - operator for scalar, vector, matrix
-template<class ltype,class rtype> strong_inline auto
+// - operator for scalar, vector, matrix
+template<class ltype,class rtype> accelerator_inline auto
 operator - (const iScalar<ltype>& lhs, const iScalar<rtype>& rhs) -> iScalar<decltype(lhs._internal - rhs._internal)>
 {
-    typedef iScalar<decltype(lhs._internal-rhs._internal)> ret_t;
-    ret_t ret;
-    sub(&ret,&lhs,&rhs);
-    return ret;
+  typedef iScalar<decltype(lhs._internal-rhs._internal)> ret_t;
+  ret_t ret;
+  sub(&ret,&lhs,&rhs);
+  return ret;
 }
 template<class ltype,class rtype,int N>
-strong_inline auto operator - (const iVector<ltype,N>& lhs,const iVector<rtype,N>& rhs) ->iVector<decltype(lhs._internal[0]-rhs._internal[0]),N>
+accelerator_inline auto operator - (const iVector<ltype,N>& lhs,const iVector<rtype,N>& rhs) ->iVector<decltype(lhs._internal[0]-rhs._internal[0]),N>
 {
-    typedef iVector<decltype(lhs._internal[0]-rhs._internal[0]),N> ret_t;
-    ret_t ret;
-    sub(&ret,&lhs,&rhs);
-    return ret;
+  typedef iVector<decltype(lhs._internal[0]-rhs._internal[0]),N> ret_t;
+  ret_t ret;
+  sub(&ret,&lhs,&rhs);
+  return ret;
 }
 template<class ltype,class rtype,int N>
-strong_inline auto operator - (const iMatrix<ltype,N>& lhs,const iMatrix<rtype,N>& rhs) ->iMatrix<decltype(lhs._internal[0][0]-rhs._internal[0][0]),N>
+accelerator_inline auto operator - (const iMatrix<ltype,N>& lhs,const iMatrix<rtype,N>& rhs) ->iMatrix<decltype(lhs._internal[0][0]-rhs._internal[0][0]),N>
 {
-    typedef iMatrix<decltype(lhs._internal[0][0]-rhs._internal[0][0]),N> ret_t;
-    ret_t ret;
-    sub(&ret,&lhs,&rhs);
-    return ret;
+  typedef iMatrix<decltype(lhs._internal[0][0]-rhs._internal[0][0]),N> ret_t;
+  ret_t ret;
+  sub(&ret,&lhs,&rhs);
+  return ret;
 }
 template<class ltype,class rtype,int N>
-strong_inline auto operator - (const iScalar<ltype>& lhs,const iMatrix<rtype,N>& rhs)->iMatrix<decltype(lhs._internal-rhs._internal[0][0]),N>
+accelerator_inline auto operator - (const iScalar<ltype>& lhs,const iMatrix<rtype,N>& rhs)->iMatrix<decltype(lhs._internal-rhs._internal[0][0]),N>
 {
-    typedef iMatrix<decltype(lhs._internal-rhs._internal[0][0]),N> ret_t;
-    ret_t ret;
-    sub(&ret,&lhs,&rhs);
-    return ret;
+  typedef iMatrix<decltype(lhs._internal-rhs._internal[0][0]),N> ret_t;
+  ret_t ret;
+  sub(&ret,&lhs,&rhs);
+  return ret;
 }
 template<class ltype,class rtype,int N>
-strong_inline auto operator - (const iMatrix<ltype,N>& lhs,const iScalar<rtype>& rhs)->iMatrix<decltype(lhs._internal[0][0]-rhs._internal),N>
+accelerator_inline auto operator - (const iMatrix<ltype,N>& lhs,const iScalar<rtype>& rhs)->iMatrix<decltype(lhs._internal[0][0]-rhs._internal),N>
 {
-    typedef iMatrix<decltype(lhs._internal[0][0]-rhs._internal),N> ret_t;
-    ret_t ret;
-    sub(&ret,&lhs,&rhs);
-    return ret;
+  typedef iMatrix<decltype(lhs._internal[0][0]-rhs._internal),N> ret_t;
+  ret_t ret;
+  sub(&ret,&lhs,&rhs);
+  return ret;
 }
 
-
-}
+NAMESPACE_END(Grid);
 
 #endif
diff --git a/Grid/tensors/Tensor_class.h b/Grid/tensors/Tensor_class.h
index d59640df..75e42721 100644
--- a/Grid/tensors/Tensor_class.h
+++ b/Grid/tensors/Tensor_class.h
@@ -21,11 +21,10 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 See the full license in the file "LICENSE" in the top level distribution
 directory
 *************************************************************************************/
-/*  END LEGAL */
-#ifndef GRID_MATH_TENSORS_H
-#define GRID_MATH_TENSORS_H
+			   /*  END LEGAL */
+#pragma once 
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
 ///////////////////////////////////////////////////
 // Scalar, Vector, Matrix objects.
@@ -48,9 +47,11 @@ class GridTensorBase {};
   using element = vtype; \
   using scalar_type     = typename Traits::scalar_type; \
   using vector_type     = typename Traits::vector_type; \
+  using scalar_typeD    = typename Traits::scalar_typeD; \
   using vector_typeD    = typename Traits::vector_typeD; \
   using tensor_reduced  = typename Traits::tensor_reduced; \
   using scalar_object   = typename Traits::scalar_object; \
+  using scalar_objectD  = typename Traits::scalar_objectD; \
   using Complexified    = typename Traits::Complexified; \
   using Realified       = typename Traits::Realified; \
   using DoublePrecision = typename Traits::DoublePrecision; \
@@ -58,229 +59,220 @@ class GridTensorBase {};
 
 template <class vtype>
 class iScalar {
- public:
+public:
   vtype _internal;
 
   using Traits = GridTypeMapper<iScalar<vtype> >;
   GridVector_CopyTraits;
 
+  static accelerator_inline constexpr int Nsimd(void) { return sizeof(vector_type)/sizeof(scalar_type); } 
+
   // Scalar no action
-  //  template<int Level> using tensor_reduce_level = typename
-  //  iScalar<GridTypeMapper<vtype>::tensor_reduce_level<Level> >;
-  iScalar() = default;
-  /*
-  iScalar(const iScalar<vtype> &copyme)=default;
-  iScalar(iScalar<vtype> &&copyme)=default;
-  iScalar<vtype> & operator= (const iScalar<vtype> &copyme) = default;
-  iScalar<vtype> & operator= (iScalar<vtype> &&copyme) = default;
-  */
+  accelerator iScalar() = default;
 
-  //  template<int N=0>
-  //  iScalar(EnableIf<isSIMDvectorized<vector_type>, vector_type> s) : _internal(s){};  // recurse down and hit the constructor for vector_type
-
-  iScalar(scalar_type s) : _internal(s){};  // recurse down and hit the constructor for vector_type
-
-  iScalar(const Zero &z) { *this = zero; };
-
-  iScalar<vtype> &operator=(const Zero &hero) {
-    zeroit(*this);
-    return *this;
-  }
-  friend strong_inline void vstream(iScalar<vtype> &out,
-                                    const iScalar<vtype> &in) {
-    vstream(out._internal, in._internal);
-  }
-  friend strong_inline void vbroadcast(iScalar<vtype> &out,const iScalar<vtype> &in,int lane){
-    vbroadcast(out._internal,in._internal,lane);
-  }
-  friend strong_inline void zeroit(iScalar<vtype> &that){
+  friend accelerator_inline void zeroit(iScalar<vtype> &that){
     zeroit(that._internal);
   }
-  friend strong_inline void prefetch(iScalar<vtype> &that) {
+
+  accelerator_inline iScalar(scalar_type s) : _internal(s){};  // recurse down and hit the constructor for vector_type
+
+  accelerator_inline iScalar(const Zero &z) { zeroit(*this); };
+
+  accelerator_inline iScalar<vtype> &operator=(const Zero &hero) {
+    zeroit(*this);  return *this;
+  }
+  friend accelerator_inline void vstream(iScalar<vtype> &out, const iScalar<vtype> &in) {
+    vstream(out._internal, in._internal);
+  }
+  friend accelerator_inline void vbroadcast(iScalar<vtype> &out,const iScalar<vtype> &in,int lane){
+    vbroadcast(out._internal,in._internal,lane);
+  }
+  friend accelerator_inline void prefetch(iScalar<vtype> &that) {
     prefetch(that._internal);
   }
-  friend strong_inline void permute(iScalar<vtype> &out,
-                                    const iScalar<vtype> &in, int permutetype) {
+  friend accelerator_inline void permute(iScalar<vtype> &out, const iScalar<vtype> &in, int permutetype) {
     permute(out._internal, in._internal, permutetype);
   }
-  friend strong_inline void rotate(iScalar<vtype> &out,const iScalar<vtype> &in,int rot){
+  friend accelerator_inline void rotate(iScalar<vtype> &out,const iScalar<vtype> &in,int rot){
     rotate(out._internal,in._internal,rot);
   }
-  friend strong_inline void exchange(iScalar<vtype> &out1,iScalar<vtype> &out2,
-				     const iScalar<vtype> &in1,const iScalar<vtype> &in2,int type){
-    exchange(out1._internal,out2._internal,
-	      in1._internal, in2._internal,type);
+  friend accelerator_inline void exchange(iScalar<vtype> &out1,iScalar<vtype> &out2,
+				     const iScalar<vtype> &in1,const iScalar<vtype> &in2,int type)
+  {
+    exchange(out1._internal,out2._internal,in1._internal, in2._internal,type);
   }
 
   // Unary negation
-  friend strong_inline iScalar<vtype> operator-(const iScalar<vtype> &r) {
+  friend accelerator_inline iScalar<vtype> operator-(const iScalar<vtype> &r) {
     iScalar<vtype> ret;
     ret._internal = -r._internal;
     return ret;
   }
   // *=,+=,-= operators inherit from corresponding "*,-,+" behaviour
-  strong_inline iScalar<vtype> &operator*=(const iScalar<vtype> &r) {
+  accelerator_inline iScalar<vtype> &operator*=(const iScalar<vtype> &r) {
     *this = (*this) * r;
     return *this;
   }
-  strong_inline iScalar<vtype> &operator-=(const iScalar<vtype> &r) {
+  accelerator_inline iScalar<vtype> &operator-=(const iScalar<vtype> &r) {
     *this = (*this) - r;
     return *this;
   }
-  strong_inline iScalar<vtype> &operator+=(const iScalar<vtype> &r) {
+  accelerator_inline iScalar<vtype> &operator+=(const iScalar<vtype> &r) {
     *this = (*this) + r;
     return *this;
   }
-  strong_inline vtype &operator()(void) { return _internal; }
-  strong_inline const vtype &operator()(void) const { return _internal; }
+  accelerator_inline vtype &operator()(void) { return _internal; }
+  accelerator_inline const vtype &operator()(void) const { return _internal; }
 
   // Type casts meta programmed, must be pure scalar to match TensorRemove
-  template <class U = vtype, class V = scalar_type, IfComplex<V> = 0, IfNotSimd<U> = 0>
+  template <class U = vtype, class V = scalar_type, IfComplex<V> = 0, IfNotSimd<U> = 0> accelerator_inline
   operator ComplexF() const {
     return (TensorRemove(_internal));
-  };
-  template <class U = vtype, class V = scalar_type, IfComplex<V> = 0, IfNotSimd<U> = 0>
+  }
+  template <class U = vtype, class V = scalar_type, IfComplex<V> = 0, IfNotSimd<U> = 0> accelerator_inline
   operator ComplexD() const {
     return (TensorRemove(_internal));
-  };
-  //  template<class U=vtype,class V=scalar_type,IfComplex<V> = 0,IfNotSimd<U> =
-  //  0> operator RealD    () const { return(real(TensorRemove(_internal))); }
-  template <class U = vtype, class V = scalar_type, IfReal<V> = 0,IfNotSimd<U> = 0>
+  }
+  template <class U = vtype, class V = scalar_type, IfReal<V> = 0,IfNotSimd<U> = 0> accelerator_inline
   operator RealD() const {
     return TensorRemove(_internal);
   }
-  template <class U = vtype, class V = scalar_type, IfInteger<V> = 0, IfNotSimd<U> = 0>
+  template <class U = vtype, class V = scalar_type, IfInteger<V> = 0, IfNotSimd<U> = 0> accelerator_inline
   operator Integer() const {
     return Integer(TensorRemove(_internal));
   }
 
   // convert from a something to a scalar via constructor of something arg
   template <class T, typename std::enable_if<!isGridTensor<T>::value, T>::type * = nullptr>
-  strong_inline iScalar<vtype> operator=(T arg) {
+  accelerator_inline iScalar<vtype> operator=(T arg) {
     _internal = arg;
     return *this;
   }
 
   // Convert elements
   template <class ttype>
-  strong_inline iScalar<vtype> operator=(iScalar<ttype> &&arg) {
+  accelerator_inline iScalar<vtype> operator=(const iScalar<ttype> &arg) {
     _internal = arg._internal;
     return *this;
   }
 
+  // Host only
   friend std::ostream &operator<<(std::ostream &stream,const iScalar<vtype> &o) {
     stream << "S {" << o._internal << "}";
     return stream;
   };
-
   strong_inline const scalar_type * begin() const { return reinterpret_cast<const scalar_type *>(&_internal); }
   strong_inline       scalar_type * begin()       { return reinterpret_cast<      scalar_type *>(&_internal); }
   strong_inline const scalar_type * end()   const { return begin() + Traits::count; }
   strong_inline       scalar_type * end()         { return begin() + Traits::count; }
 };
+
 ///////////////////////////////////////////////////////////
 // Allows to turn scalar<scalar<scalar<double>>>> back to double.
 ///////////////////////////////////////////////////////////
 template <class T>
-strong_inline typename std::enable_if<!isGridTensor<T>::value, T>::type
+accelerator_inline typename std::enable_if<!isGridTensor<T>::value, T>::type
 TensorRemove(T arg) {
   return arg;
 }
 template <class vtype>
-strong_inline auto TensorRemove(iScalar<vtype> arg)
-    -> decltype(TensorRemove(arg._internal)) {
+accelerator_inline auto TensorRemove(iScalar<vtype> arg)
+  -> decltype(TensorRemove(arg._internal)) {
   return TensorRemove(arg._internal);
 }
 
 template <class vtype, int N>
 class iVector {
- public:
+public:
   vtype _internal[N];
-
+  
   using Traits = GridTypeMapper<iVector<vtype, N> >;
+
   GridVector_CopyTraits;
 
-  template <class T, typename std::enable_if<!isGridTensor<T>::value, T>::type
-                         * = nullptr>
-  strong_inline auto operator=(T arg) -> iVector<vtype, N> {
+  static accelerator_inline constexpr int Nsimd(void) { return sizeof(vector_type)/sizeof(scalar_type); } 
+
+  template <class T, typename std::enable_if<!isGridTensor<T>::value, T>::type * = nullptr>
+  accelerator_inline auto operator=(T arg) -> iVector<vtype, N> {
     zeroit(*this);
     for (int i = 0; i < N; i++) _internal[i] = arg;
     return *this;
   }
 
-  iVector(const Zero &z) { *this = zero; };
-  iVector() = default;
-  /*
-  iVector(const iVector<vtype,N> &copyme)=default;
-  iVector(iVector<vtype,N> &&copyme)=default;
-  iVector<vtype,N> & operator= (const iVector<vtype,N> &copyme) = default;
-  iVector<vtype,N> & operator= (iVector<vtype,N> &&copyme) = default;
-  */
+  accelerator iVector() = default;
+  accelerator_inline iVector(const Zero &z) { zeroit(*this); };
 
-  iVector<vtype, N> &operator=(const Zero &hero) {
+  template<class other>
+  accelerator_inline iVector<vtype, N> &operator=(const iVector<other,N> &him)
+  {
+    for (int i = 0; i < N; i++) {
+      _internal[i] = him._internal[i];
+    }    
+    return *this;
+  } 
+  accelerator_inline iVector<vtype, N> &operator=(const Zero &hero) {
     zeroit(*this);
     return *this;
   }
-  friend strong_inline void zeroit(iVector<vtype, N> &that) {
+  friend accelerator_inline void zeroit(iVector<vtype, N> &that) {
     for (int i = 0; i < N; i++) {
       zeroit(that._internal[i]);
     }
   }
-  friend strong_inline void prefetch(iVector<vtype, N> &that) {
+  friend accelerator_inline void prefetch(iVector<vtype, N> &that) {
     for (int i = 0; i < N; i++) prefetch(that._internal[i]);
   }
-  friend strong_inline void vstream(iVector<vtype, N> &out,
-                                    const iVector<vtype, N> &in) {
+  friend accelerator_inline void vstream(iVector<vtype, N> &out, const iVector<vtype, N> &in) {
     for (int i = 0; i < N; i++) {
       vstream(out._internal[i], in._internal[i]);
     }
   }
-  friend strong_inline void vbroadcast(iVector<vtype,N> &out,const iVector<vtype,N> &in,int lane){
+  friend accelerator_inline void vbroadcast(iVector<vtype,N> &out,const iVector<vtype,N> &in,int lane){
     for(int i=0;i<N;i++){
       vbroadcast(out._internal[i],in._internal[i],lane);
     }
   }
-  friend strong_inline void permute(iVector<vtype,N> &out,const iVector<vtype,N> &in,int permutetype){
+  friend accelerator_inline void permute(iVector<vtype,N> &out,const iVector<vtype,N> &in,int permutetype){
     for(int i=0;i<N;i++){
       permute(out._internal[i],in._internal[i],permutetype);
     }
   }
-  friend strong_inline void rotate(iVector<vtype,N> &out,const iVector<vtype,N> &in,int rot){
+  friend accelerator_inline void rotate(iVector<vtype,N> &out,const iVector<vtype,N> &in,int rot){
     for(int i=0;i<N;i++){
       rotate(out._internal[i],in._internal[i],rot);
     }
   }
-  friend strong_inline void exchange(iVector<vtype,N> &out1,iVector<vtype,N> &out2,
+  friend accelerator_inline void exchange(iVector<vtype,N> &out1,iVector<vtype,N> &out2,
 				     const iVector<vtype,N> &in1,const iVector<vtype,N> &in2,int type){
     for(int i=0;i<N;i++){
-      exchange(out1._internal[i],out2._internal[i],
-	        in1._internal[i], in2._internal[i],type);
+      exchange(out1._internal[i],out2._internal[i],in1._internal[i], in2._internal[i],type);
     }
   }
 
   // Unary negation
-  friend strong_inline iVector<vtype, N> operator-(const iVector<vtype, N> &r) {
+  friend accelerator_inline iVector<vtype, N> operator-(const iVector<vtype, N> &r) {
     iVector<vtype, N> ret;
     for (int i = 0; i < N; i++) ret._internal[i] = -r._internal[i];
     return ret;
   }
   // *=,+=,-= operators inherit from corresponding "*,-,+" behaviour
-  strong_inline iVector<vtype, N> &operator*=(const iScalar<vtype> &r) {
+  accelerator_inline iVector<vtype, N> &operator*=(const iScalar<vtype> &r) {
     *this = (*this) * r;
     return *this;
   }
-  strong_inline iVector<vtype, N> &operator-=(const iVector<vtype, N> &r) {
+  accelerator_inline iVector<vtype, N> &operator-=(const iVector<vtype, N> &r) {
     *this = (*this) - r;
     return *this;
   }
-  strong_inline iVector<vtype, N> &operator+=(const iVector<vtype, N> &r) {
+  accelerator_inline iVector<vtype, N> &operator+=(const iVector<vtype, N> &r) {
     *this = (*this) + r;
     return *this;
   }
-  strong_inline vtype &operator()(int i) { return _internal[i]; }
-  strong_inline const vtype &operator()(int i) const { return _internal[i]; }
-  friend std::ostream &operator<<(std::ostream &stream,
-                                  const iVector<vtype, N> &o) {
+  accelerator_inline vtype &operator()(int i) { return _internal[i]; }
+  accelerator_inline const vtype &operator()(int i) const { return _internal[i]; }
+
+  // Host
+  friend std::ostream &operator<<(std::ostream &stream, const iVector<vtype, N> &o) {
     stream << "V<" << N << ">{";
     for (int i = 0; i < N; i++) {
       stream << o._internal[i];
@@ -297,133 +289,132 @@ class iVector {
   strong_inline       scalar_type * begin()       { return reinterpret_cast<      scalar_type *>(_internal); }
   strong_inline const scalar_type * end()   const { return begin() + Traits::count; }
   strong_inline       scalar_type * end()         { return begin() + Traits::count; }
+
 };
 
 template <class vtype, int N>
 class iMatrix {
- public:
+public:
   vtype _internal[N][N];
 
   using Traits = GridTypeMapper<iMatrix<vtype, N> >;
+
   GridVector_CopyTraits;
 
-  iMatrix(const Zero &z) { *this = zero; };
-  iMatrix() = default;
+  static accelerator_inline constexpr int Nsimd(void) { return sizeof(vector_type)/sizeof(scalar_type); } 
 
-  iMatrix &operator=(const iMatrix &rhs) {
+  accelerator_inline iMatrix(const Zero &z) { zeroit(*this); };
+  accelerator iMatrix() = default;
+
+  // Allow for type conversion.
+  template<class other>
+  accelerator_inline iMatrix &operator=(const iMatrix<other,N> &rhs) {
     for (int i = 0; i < N; i++)
-      for (int j = 0; j < N; j++) vstream(_internal[i][j], rhs._internal[i][j]);
+      for (int j = 0; j < N; j++) 
+	_internal[i][j] = rhs._internal[i][j];
     return *this;
   };
 
-  iMatrix(scalar_type s) {
+  accelerator_inline iMatrix(scalar_type s) {
     (*this) = s;
   };  // recurse down and hit the constructor for vector_type
 
-  /*
-  iMatrix(const iMatrix<vtype,N> &copyme)=default;
-  iMatrix(iMatrix<vtype,N> &&copyme)=default;
-  iMatrix<vtype,N> & operator= (const iMatrix<vtype,N> &copyme) = default;
-  iMatrix<vtype,N> & operator= (iMatrix<vtype,N> &&copyme) = default;
-  */
-
-  iMatrix<vtype, N> &operator=(const Zero &hero) {
+  accelerator_inline iMatrix<vtype, N> &operator=(const Zero &hero) {
     zeroit(*this);
     return *this;
   }
-  template <class T, typename std::enable_if<!isGridTensor<T>::value, T>::type
-                         * = nullptr>
-  strong_inline auto operator=(T arg) -> iMatrix<vtype, N> {
+  template <class T, typename std::enable_if<!isGridTensor<T>::value, T>::type * = nullptr>
+  accelerator_inline auto operator=(T arg) -> iMatrix<vtype, N> {
     zeroit(*this);
     for (int i = 0; i < N; i++) _internal[i][i] = arg;
     return *this;
   }
 
-  friend strong_inline void zeroit(iMatrix<vtype,N> &that){
+  friend accelerator_inline void zeroit(iMatrix<vtype,N> &that){
     for(int i=0;i<N;i++){
       for(int j=0;j<N;j++){
 	zeroit(that._internal[i][j]);
     }}
   }
-  friend strong_inline void prefetch(iMatrix<vtype,N> &that){
-    for(int i=0;i<N;i++) 
-    for(int j=0;j<N;j++) 
-      prefetch(that._internal[i][j]);
+  friend accelerator_inline void prefetch(iMatrix<vtype,N> &that){
+    for(int i=0;i<N;i++) {
+      for(int j=0;j<N;j++) { 
+	prefetch(that._internal[i][j]);
+    }}
   }
-  friend strong_inline void vstream(iMatrix<vtype,N> &out,const iMatrix<vtype,N> &in){
-      for(int i=0;i<N;i++){
+  friend accelerator_inline void vstream(iMatrix<vtype,N> &out,const iMatrix<vtype,N> &in){
+    for(int i=0;i<N;i++){
       for(int j=0;j<N;j++){
 	vstream(out._internal[i][j],in._internal[i][j]);
-      }}
+    }}
   }
-  friend strong_inline void vbroadcast(iMatrix<vtype,N> &out,const iMatrix<vtype,N> &in,int lane){
-      for(int i=0;i<N;i++){
+  friend accelerator_inline void vbroadcast(iMatrix<vtype,N> &out,const iMatrix<vtype,N> &in,int lane){
+    for(int i=0;i<N;i++){
       for(int j=0;j<N;j++){
 	vbroadcast(out._internal[i][j],in._internal[i][j],lane);
-      }}
+    }}
   }
 
-  friend strong_inline void permute(iMatrix<vtype,N> &out,const iMatrix<vtype,N> &in,int permutetype){
+  friend accelerator_inline void permute(iMatrix<vtype,N> &out,const iMatrix<vtype,N> &in,int permutetype){
     for(int i=0;i<N;i++){
       for(int j=0;j<N;j++){
 	permute(out._internal[i][j],in._internal[i][j],permutetype);
     }}
   }
-  friend strong_inline void rotate(iMatrix<vtype,N> &out,const iMatrix<vtype,N> &in,int rot){
+  friend accelerator_inline void rotate(iMatrix<vtype,N> &out,const iMatrix<vtype,N> &in,int rot){
     for(int i=0;i<N;i++){
       for(int j=0;j<N;j++){
-	rotate(out._internal[i][j],in._internal[i][j],rot);
+      rotate(out._internal[i][j],in._internal[i][j],rot);
     }}
   }
-  friend strong_inline void exchange(iMatrix<vtype,N> &out1,iMatrix<vtype,N> &out2,
-				     const iMatrix<vtype,N> &in1,const iMatrix<vtype,N> &in2,int type){
+  friend accelerator_inline void exchange(iMatrix<vtype,N> &out1,iMatrix<vtype,N> &out2,
+					  const iMatrix<vtype,N> &in1,const iMatrix<vtype,N> &in2,int type){
     for(int i=0;i<N;i++){
       for(int j=0;j<N;j++){
-	exchange(out1._internal[i][j],out2._internal[i][j],
-		  in1._internal[i][j], in2._internal[i][j],type);
+	exchange(out1._internal[i][j],out2._internal[i][j],in1._internal[i][j], in2._internal[i][j],type);
     }}
   }
-
+  
   // Unary negation
-  friend strong_inline iMatrix<vtype, N> operator-(const iMatrix<vtype, N> &r) {
+  friend accelerator_inline iMatrix<vtype, N> operator-(const iMatrix<vtype, N> &r) {
     iMatrix<vtype, N> ret;
     for (int i = 0; i < N; i++) {
       for (int j = 0; j < N; j++) {
-        ret._internal[i][j] = -r._internal[i][j];
-      }
-    }
+	ret._internal[i][j] = -r._internal[i][j];
+    }}
     return ret;
   }
   // *=,+=,-= operators inherit from corresponding "*,-,+" behaviour
   template <class T>
-  strong_inline iMatrix<vtype, N> &operator*=(const T &r) {
+  accelerator_inline iMatrix<vtype, N> &operator*=(const T &r) {
     *this = (*this) * r;
     return *this;
   }
   template <class T>
-  strong_inline iMatrix<vtype, N> &operator-=(const T &r) {
+  accelerator_inline iMatrix<vtype, N> &operator-=(const T &r) {
     *this = (*this) - r;
     return *this;
   }
   template <class T>
-  strong_inline iMatrix<vtype, N> &operator+=(const T &r) {
+  accelerator_inline iMatrix<vtype, N> &operator+=(const T &r) {
     *this = (*this) + r;
     return *this;
   }
 
   // returns an lvalue reference
-  strong_inline vtype &operator()(int i, int j) { return _internal[i][j]; }
-  strong_inline const vtype &operator()(int i, int j) const {
+  accelerator_inline vtype &operator()(int i, int j) { return _internal[i][j]; }
+  accelerator_inline const vtype &operator()(int i, int j) const {
     return _internal[i][j];
   }
-  friend std::ostream &operator<<(std::ostream &stream,
-                                  const iMatrix<vtype, N> &o) {
+  
+  // Host function only
+  friend std::ostream &operator<<(std::ostream &stream, const iMatrix<vtype, N> &o) {
     stream << "M<" << N << ">{";
     for (int i = 0; i < N; i++) {
       stream << "{";
       for (int j = 0; j < N; j++) {
-        stream << o._internal[i][j];
-        if (i < N - 1) stream << ",";
+	stream << o._internal[i][j];
+	if (i < N - 1) stream << ",";
       }
       stream << "}";
       if (i != N - 1) stream << "\n\t\t";
@@ -442,17 +433,17 @@ class iMatrix {
   strong_inline       scalar_type * end()         { return begin() + Traits::count; }
 };
 
-template <class v>
+template <class v> accelerator_inline
 void vprefetch(const iScalar<v> &vv) {
   vprefetch(vv._internal);
 }
-template <class v, int N>
+template <class v, int N> accelerator_inline
 void vprefetch(const iVector<v, N> &vv) {
   for (int i = 0; i < N; i++) {
     vprefetch(vv._internal[i]);
   }
 }
-template <class v, int N>
+template <class v, int N> accelerator_inline
 void vprefetch(const iMatrix<v, N> &vv) {
   for (int i = 0; i < N; i++) {
     for (int j = 0; j < N; j++) {
@@ -460,5 +451,6 @@ void vprefetch(const iMatrix<v, N> &vv) {
     }
   }
 }
-}
-#endif
+
+NAMESPACE_END(Grid);
+
diff --git a/Grid/tensors/Tensor_determinant.h b/Grid/tensors/Tensor_determinant.h
index a58a8074..a3862c94 100644
--- a/Grid/tensors/Tensor_determinant.h
+++ b/Grid/tensors/Tensor_determinant.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,50 +23,51 @@ Author: neo <cossu@post.kek.jp>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_DET_H
 #define GRID_MATH_DET_H
-namespace Grid {
-  /////////////////////////////////////////////// 
-  // Determinant function for scalar, vector, matrix
-  /////////////////////////////////////////////// 
-  inline ComplexF Determinant( const ComplexF &arg){    return arg;}
-  inline ComplexD Determinant( const ComplexD &arg){    return arg;}
-  inline RealF Determinant( const RealF &arg){    return arg;}
-  inline RealD Determinant( const RealD &arg){    return arg;}
-
-  template<class vtype> inline auto Determinant(const iScalar<vtype>&r) -> iScalar<decltype(Determinant(r._internal))>
-    {
-      iScalar<decltype(Determinant(r._internal))> ret;
-      ret._internal = Determinant(r._internal);
-      return ret;
-    }
-
-  template<class vtype,int N, typename std::enable_if< GridTypeMapper<vtype>::TensorLevel == 0 >::type * =nullptr> 
-    inline iScalar<vtype> Determinant(const iMatrix<vtype,N> &arg)
-    {
-      iMatrix<vtype,N> ret(arg);
-      iScalar<vtype> det = vtype(1.0);
-      /* Conversion of matrix to upper triangular */
-      for(int i = 0; i < N; i++){
-        for(int j = 0; j < N; j++){
-	  if(j>i){
-	    vtype ratio = ret._internal[j][i]/ret._internal[i][i];
-	    for(int k = 0; k < N; k++){
-	      ret._internal[j][k] -= ratio * ret._internal[i][k];
-	    }
-	  }
-        }
-      }      
-
-      for(int i = 0; i < N; i++)
-	det *= ret._internal[i][i];   
-
-      return det;
-    }
 
+NAMESPACE_BEGIN(Grid);
 
+/////////////////////////////////////////////// 
+// Determinant function for scalar, vector, matrix
+/////////////////////////////////////////////// 
+accelerator_inline ComplexF Determinant( const ComplexF &arg){    return arg;}
+accelerator_inline ComplexD Determinant( const ComplexD &arg){    return arg;}
+accelerator_inline RealF Determinant( const RealF &arg){    return arg;}
+accelerator_inline RealD Determinant( const RealD &arg){    return arg;}
 
+template<class vtype> accelerator_inline auto Determinant(const iScalar<vtype>&r) -> iScalar<decltype(Determinant(r._internal))>
+{
+  iScalar<decltype(Determinant(r._internal))> ret;
+  ret._internal = Determinant(r._internal);
+  return ret;
 }
+
+template<class vtype,int N, typename std::enable_if< GridTypeMapper<vtype>::TensorLevel == 0 >::type * =nullptr> 
+accelerator_inline iScalar<vtype> Determinant(const iMatrix<vtype,N> &arg)
+{
+  iMatrix<vtype,N> ret(arg);
+  iScalar<vtype> det = vtype(1.0);
+  /* Conversion of matrix to upper triangular */
+  for(int i = 0; i < N; i++){
+    for(int j = 0; j < N; j++){
+      if(j>i){
+	vtype ratio = ret._internal[j][i]/ret._internal[i][i];
+	for(int k = 0; k < N; k++){
+	  ret._internal[j][k] -= ratio * ret._internal[i][k];
+	}
+      }
+    }
+  }      
+
+  for(int i = 0; i < N; i++)
+    det *= ret._internal[i][i];   
+
+  return det;
+}
+
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/tensors/Tensor_exp.h b/Grid/tensors/Tensor_exp.h
index f7eee8f0..11d37f9c 100644
--- a/Grid/tensors/Tensor_exp.h
+++ b/Grid/tensors/Tensor_exp.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,122 +23,120 @@ Author: neo <cossu@post.kek.jp>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_EXP_H
 #define GRID_MATH_EXP_H
 
 #define DEFAULT_MAT_EXP 12
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  /////////////////////////////////////////////// 
-  // Exponentiate function for scalar, vector, matrix
-  /////////////////////////////////////////////// 
+/////////////////////////////////////////////// 
+// Exponentiate function for scalar, vector, matrix
+/////////////////////////////////////////////// 
 
 
-  template<class vtype> inline iScalar<vtype> Exponentiate(const iScalar<vtype>&r, RealD alpha ,  Integer Nexp = DEFAULT_MAT_EXP)
-    {
-      iScalar<vtype> ret;
-      ret._internal = Exponentiate(r._internal, alpha, Nexp);
-      return ret;
-    }
+template<class vtype> accelerator_inline iScalar<vtype> Exponentiate(const iScalar<vtype>&r, RealD alpha ,  Integer Nexp = DEFAULT_MAT_EXP)
+{
+  iScalar<vtype> ret;
+  ret._internal = Exponentiate(r._internal, alpha, Nexp);
+  return ret;
+}
 
-template<class vtype, int N> inline iVector<vtype, N> Exponentiate(const iVector<vtype,N>&r, RealD alpha ,  Integer Nexp = DEFAULT_MAT_EXP)
-    {
-      iVector<vtype, N> ret;
-      for (int i = 0; i < N; i++)
-        ret._internal[i] = Exponentiate(r._internal[i], alpha, Nexp);
-      return ret;
-    }
+template<class vtype, int N> accelerator_inline iVector<vtype, N> Exponentiate(const iVector<vtype,N>&r, RealD alpha ,  Integer Nexp = DEFAULT_MAT_EXP)
+{
+  iVector<vtype, N> ret;
+  for (int i = 0; i < N; i++)
+    ret._internal[i] = Exponentiate(r._internal[i], alpha, Nexp);
+  return ret;
+}
 
 
 
-    // Specialisation: Cayley-Hamilton exponential for SU(3)
-    template<class vtype, typename std::enable_if< GridTypeMapper<vtype>::TensorLevel == 0>::type * =nullptr> 
-    inline iMatrix<vtype,3> Exponentiate(const iMatrix<vtype,3> &arg, RealD alpha  , Integer Nexp = DEFAULT_MAT_EXP )
-    {
-    // for SU(3) 2x faster than the std implementation using Nexp=12
-    // notice that it actually computes
-    // exp ( input matrix )
-    // the i sign is coming from outside
-    // input matrix is anti-hermitian NOT hermitian
-      typedef iMatrix<vtype,3> mat;
-      typedef iScalar<vtype> scalar;
-      mat unit(1.0);
-      mat temp(unit);
-      const Complex one_over_three = 1.0 / 3.0;
-      const Complex one_over_two = 1.0 / 2.0;
+// Specialisation: Cayley-Hamilton exponential for SU(3)
+#ifndef GRID_NVCC
+template<class vtype, typename std::enable_if< GridTypeMapper<vtype>::TensorLevel == 0>::type * =nullptr> 
+accelerator_inline iMatrix<vtype,3> Exponentiate(const iMatrix<vtype,3> &arg, RealD alpha  , Integer Nexp = DEFAULT_MAT_EXP )
+{
+  // for SU(3) 2x faster than the std implementation using Nexp=12
+  // notice that it actually computes
+  // exp ( input matrix )
+  // the i sign is coming from outside
+  // input matrix is anti-hermitian NOT hermitian
+  typedef iMatrix<vtype,3> mat;
+  typedef iScalar<vtype> scalar;
+  mat unit(1.0);
+  const Complex one_over_three = 1.0 / 3.0;
+  const Complex one_over_two = 1.0 / 2.0;
 
-      scalar c0, c1, tmp, c0max, theta, u, w;
-      scalar xi0, u2, w2, cosw;
-      scalar fden, h0, h1, h2;
-      scalar e2iu, emiu, ixi0, qt;
-      scalar f0, f1, f2;
-      scalar unity(1.0);
+  scalar c0, c1, tmp, c0max, theta, u, w;
+  scalar xi0, u2, w2, cosw;
+  scalar fden, h0, h1, h2;
+  scalar e2iu, emiu, ixi0;
+  scalar f0, f1, f2;
+  scalar unity(1.0);
       
-      mat iQ2 = arg*arg*alpha*alpha;
-      mat iQ3 = arg*iQ2*alpha;   
-      // sign in c0 from the conventions on the Ta
-      scalar imQ3, reQ2;
-      imQ3 = imag( trace(iQ3) );
-      reQ2 = real( trace(iQ2) );
-      c0 = -imQ3 * one_over_three;  
-      c1 = -reQ2 * one_over_two;
+  mat iQ2 = arg*arg*alpha*alpha;
+  mat iQ3 = arg*iQ2*alpha;   
+  // sign in c0 from the conventions on the Ta
+  scalar imQ3, reQ2;
+  imQ3 = imag( trace(iQ3) );
+  reQ2 = real( trace(iQ2) );
+  c0 = -imQ3 * one_over_three;  
+  c1 = -reQ2 * one_over_two;
 
-      // Cayley Hamilton checks to machine precision, tested
-      tmp = c1 * one_over_three;
-      c0max = 2.0 * pow(tmp, 1.5);
+  // Cayley Hamilton checks to machine precision, tested
+  tmp = c1 * one_over_three;
+  c0max = 2.0 * pow(tmp, 1.5);
 
-      theta = acos(c0 / c0max) * one_over_three;
-      u = sqrt(tmp) * cos(theta);
-      w = sqrt(c1) * sin(theta);
+  theta = acos(c0 / c0max) * one_over_three;
+  u = sqrt(tmp) * cos(theta);
+  w = sqrt(c1) * sin(theta);
 
-      xi0 = sin(w) / w;
-      u2 = u * u;
-      w2 = w * w;
-      cosw = cos(w);
+  xi0 = sin(w) / w;
+  u2 = u * u;
+  w2 = w * w;
+  cosw = cos(w);
 
-      ixi0 = timesI(xi0);
-      emiu = cos(u) - timesI(sin(u));
-      e2iu = cos(2.0 * u) + timesI(sin(2.0 * u));
+  ixi0 = timesI(xi0);
+  emiu = cos(u) - timesI(sin(u));
+  e2iu = cos(2.0 * u) + timesI(sin(2.0 * u));
 
-      h0 = e2iu * (u2 - w2) +
-           emiu * ((8.0 * u2 * cosw) + (2.0 * u * (3.0 * u2 + w2) * ixi0));
-      h1 = e2iu * (2.0 * u) - emiu * ((2.0 * u * cosw) - (3.0 * u2 - w2) * ixi0);
-      h2 = e2iu - emiu * (cosw + (3.0 * u) * ixi0);
+  h0 = e2iu * (u2 - w2) +
+    emiu * ((8.0 * u2 * cosw) + (2.0 * u * (3.0 * u2 + w2) * ixi0));
+  h1 = e2iu * (2.0 * u) - emiu * ((2.0 * u * cosw) - (3.0 * u2 - w2) * ixi0);
+  h2 = e2iu - emiu * (cosw + (3.0 * u) * ixi0);
 
-      fden = unity / (9.0 * u2 - w2);  // reals
-      f0 = h0 * fden;
-      f1 = h1 * fden;
-      f2 = h2 * fden;
-
-      return (f0 * unit + timesMinusI(f1) * arg*alpha - f2 * iQ2);
-    }
+  fden = unity / (9.0 * u2 - w2);  // reals
+  f0 = h0 * fden;
+  f1 = h1 * fden;
+  f2 = h2 * fden;
 
+  return (f0 * unit + timesMinusI(f1) * arg*alpha - f2 * iQ2);
+}
+#endif
 
 
 // General exponential
 template<class vtype,int N, typename std::enable_if< GridTypeMapper<vtype>::TensorLevel == 0 >::type * =nullptr> 
-    inline iMatrix<vtype,N> Exponentiate(const iMatrix<vtype,N> &arg, RealD alpha  , Integer Nexp = DEFAULT_MAT_EXP )
-    {
-    // notice that it actually computes
-    // exp ( input matrix )
-    // the i sign is coming from outside
-    // input matrix is anti-hermitian NOT hermitian
-      typedef iMatrix<vtype,N> mat;
-      mat unit(1.0);
-      mat temp(unit);
-      for(int i=Nexp; i>=1;--i){
-	      temp *= alpha/RealD(i);
-	      temp = unit + temp*arg;
-      }
-      return temp;
-
-    }
-
-
-
+accelerator_inline iMatrix<vtype,N> Exponentiate(const iMatrix<vtype,N> &arg, RealD alpha  , Integer Nexp = DEFAULT_MAT_EXP )
+{
+  // notice that it actually computes
+  // exp ( input matrix )
+  // the i sign is coming from outside
+  // input matrix is anti-hermitian NOT hermitian
+  typedef iMatrix<vtype,N> mat;
+  mat unit(1.0);
+  mat temp(unit);
+  for(int i=Nexp; i>=1;--i){
+    temp *= alpha/RealD(i);
+    temp = unit + temp*arg;
+  }
+  return temp;
 
 }
+
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/tensors/Tensor_extract_merge.h b/Grid/tensors/Tensor_extract_merge.h
index a32d3785..f1ded209 100644
--- a/Grid/tensors/Tensor_extract_merge.h
+++ b/Grid/tensors/Tensor_extract_merge.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -27,267 +27,186 @@ Author: Christopher Kelly <ckelly@phys.columbia.edu>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef GRID_EXTRACT_H
-#define GRID_EXTRACT_H
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once 
+
+#include <string.h>
+
+//#pragma GCC optimize("no-strict-aliasing")
+
+NAMESPACE_BEGIN(Grid);
+
 /////////////////////////////////////////////////////////////////
 // Generic extract/merge/permute
 /////////////////////////////////////////////////////////////////
 
-namespace Grid{
+template<class __T> using ExtractPointerArray = AcceleratorVector<__T *,GRID_MAX_SIMD>;
+template<class __T> using ExtractBuffer       = AcceleratorVector<__T  ,GRID_MAX_SIMD>;
 
-////////////////////////////////////////////////////////////////////////////////////////////////
-// Extract/merge a fundamental vector type, to pointer array with offset
-////////////////////////////////////////////////////////////////////////////////////////////////
-
-template<class vsimd,class scalar>
-inline void extract(typename std::enable_if<!isGridTensor<vsimd>::value, const vsimd >::type * y, 
-		    std::vector<scalar *> &extracted,int offset){
-  // FIXME: bounce off memory is painful
-  static const int Nsimd=sizeof(vsimd)/sizeof(scalar);
-  int Nextr=extracted.size();
-  int s=Nsimd/Nextr;
-
-  scalar*buf = (scalar *)y;
-  for(int i=0;i<Nextr;i++){
-    extracted[i][offset] = buf[i*s];
-  }
-};
-////////////////////////////////////////////////////////////////////////
-// Merge simd vector from array of scalars to pointer array with offset
-////////////////////////////////////////////////////////////////////////
-template<class vsimd,class scalar>
-inline void merge(typename std::enable_if<!isGridTensor<vsimd>::value, vsimd >::type * y, 
-		  std::vector<scalar *> &extracted,int offset){
-
-  static const int Nsimd=sizeof(vsimd)/sizeof(scalar);
-
-  int Nextr=extracted.size();
-  int s=Nsimd/Nextr; // can have sparse occupation of simd vector if simd_layout does not fill it
-                     // replicate n-fold. Use to allow Integer masks to 
-                     // predicate floating point of various width assignments and maintain conformable.
-  scalar *buf =(scalar *) y;
-  for(int i=0;i<Nextr;i++){
-    for(int ii=0;ii<s;ii++){
-      buf[i*s+ii]=extracted[i][offset];
-    }
-  }
-};
-
-////////////////////////////////////////////////////////////////////////////////////////////////
-// Extract a fundamental vector type to scalar array 
-////////////////////////////////////////////////////////////////////////////////////////////////
-template<class vsimd,class scalar>
-inline void extract(typename std::enable_if<!isGridTensor<vsimd>::value, const vsimd >::type  &y,std::vector<scalar> &extracted){
-
-  int Nextr=extracted.size();
-  int Nsimd=vsimd::Nsimd();
-  int s=Nsimd/Nextr;
-
-  scalar *buf = (scalar *)&y;
-  for(int i=0;i<Nextr;i++){
-    extracted[i]=buf[i*s];
-#ifdef PARANOID
-    for(int ii=1;ii<s;ii++){
-      if ( buf[i*s]!=buf[i*s+ii] ){
-	std::cout<<GridLogMessage << " SIMD extract failure splat = "<<s<<" ii "<<ii<<" " <<Nextr<<" "<< Nsimd<<" "<<std::endl;
-	for(int vv=0;vv<Nsimd;vv++) {
-	  std::cout<<GridLogMessage<< buf[vv]<<" ";
-	}
-	std::cout<<GridLogMessage<<std::endl;
-	assert(0);
-      }
-      assert(buf[i*s]==buf[i*s+ii]);
-    }
-#endif
-  }
-
-};
-
-////////////////////////////////////////////////////////////////////////
-// Merge simd vector from array of scalars
-////////////////////////////////////////////////////////////////////////
-template<class vsimd,class scalar>
-inline void merge(typename std::enable_if<!isGridTensor<vsimd>::value, vsimd >::type  &y,std::vector<scalar> &extracted){
-  int Nextr=extracted.size();
-  static const int Nsimd=vsimd::Nsimd();
-  int s=Nsimd/Nextr;
-  scalar *buf = (scalar *)&y;
-
-  for(int i=0;i<Nextr;i++){
-    for(int ii=0;ii<s;ii++){
-      buf[i*s+ii]=extracted[i]; // replicates value
-    }
-  }
-};
+//void extract(const vobj &vec,ExtractBuffer<typename vobj::scalar_object> &extracted);
+//void extract(const vobj &vec,ExtractPointerArray<sobj> &extracted, int offset);
+//void   merge(vobj &vec,ExtractBuffer<typename vobj::scalar_object> &extracted)
+//void   merge(vobj &vec,ExtractPointerArray<typename vobj::scalar_object> &extracted)
 
 ////////////////////////////////////////////////////////////////////////
 // Extract to contiguous array scalar object
 ////////////////////////////////////////////////////////////////////////
-template<class vobj> inline void extract(const vobj &vec,std::vector<typename vobj::scalar_object> &extracted)
+template<class vobj,class sobj> accelerator
+void extract(const vobj &vec,ExtractBuffer<sobj> &extracted)
 {
-  typedef typename vobj::scalar_type scalar_type ;
-  typedef typename vobj::vector_type vector_type ;
-
-  static const int Nsimd=sizeof(vector_type)/sizeof(scalar_type);
-  static const int words=sizeof(vobj)/sizeof(vector_type);
-  int Nextr=extracted.size();
-  int s=Nsimd/Nextr;
-
-  std::vector<scalar_type *> pointers(Nextr);
-  for(int i=0;i<Nextr;i++) 
-    pointers[i] =(scalar_type *)& extracted[i];
-
-  vector_type *vp = (vector_type *)&vec;
-  for(int w=0;w<words;w++){
-    extract<vector_type,scalar_type>(&vp[w],pointers,w);
-  }
-}
-////////////////////////////////////////////////////////////////////////
-// Extract to a bunch of scalar object pointers, with offset
-////////////////////////////////////////////////////////////////////////
-template<class vobj> inline 
-void extract(const vobj &vec,std::vector<typename vobj::scalar_object *> &extracted, int offset)
-{
-  typedef typename vobj::scalar_type scalar_type ;
-  typedef typename vobj::vector_type vector_type ;
-
-  static const int words=sizeof(vobj)/sizeof(vector_type);
-  static const int Nsimd=vobj::vector_type::Nsimd();
-
-  int Nextr=extracted.size();
-  int s = Nsimd/Nextr;
-  scalar_type * vp = (scalar_type *)&vec;
+  typedef typename GridTypeMapper<sobj>::scalar_type sobj_scalar_type;
+  typedef typename GridTypeMapper<vobj>::scalar_type scalar_type;
+  typedef typename GridTypeMapper<vobj>::vector_type vector_type;
 
+  const int words=sizeof(vobj)/sizeof(vector_type);
+  const int Nsimd=vector_type::Nsimd();
+  const int Nextr=extracted.size();
+  const int s=Nsimd/Nextr;
+  sobj_scalar_type *sp = (sobj_scalar_type *) &extracted[0];
+  scalar_type *vp = (scalar_type *)&vec;
+  scalar_type      vtmp;
+  sobj_scalar_type stmp;
   for(int w=0;w<words;w++){
     for(int i=0;i<Nextr;i++){
-      scalar_type * pointer = (scalar_type *)& extracted[i][offset];
-      pointer[w] = vp[i*s+w*Nsimd];
+      memcpy((char *)&vtmp,(char *)&vp[w*Nsimd+i*s],sizeof(vtmp));
+      stmp = vtmp;
+      memcpy((char *)&sp[i*words+w],(char *)&stmp,sizeof(stmp));
     }
   }
+  return;
+}
+
+////////////////////////////////////////////////////////////////////////
+// Merge a contiguous array of scalar objects
+////////////////////////////////////////////////////////////////////////
+template<class vobj,class sobj> accelerator
+void   merge(vobj &vec,ExtractBuffer<sobj> &extracted)
+{
+  typedef typename GridTypeMapper<sobj>::scalar_type sobj_scalar_type;
+  typedef typename GridTypeMapper<vobj>::scalar_type scalar_type;
+  typedef typename GridTypeMapper<vobj>::vector_type vector_type;
+
+  const int words=sizeof(vobj)/sizeof(vector_type);
+  const int Nsimd=vector_type::Nsimd();
+  const int Nextr = extracted.size();
+  const int s=Nsimd/Nextr;
+
+  sobj_scalar_type *sp = (sobj_scalar_type *)&extracted[0];
+  scalar_type *vp = (scalar_type *)&vec;
+  scalar_type      vtmp;
+  sobj_scalar_type stmp;
+  for(int w=0;w<words;w++){
+    for(int i=0;i<Nextr;i++){
+      for(int ii=0;ii<s;ii++){
+	memcpy((char *)&stmp,(char *)&sp[i*words+w],sizeof(stmp));
+	vtmp = stmp;
+	memcpy((char *)&vp[w*Nsimd+i*s+ii],(char *)&vtmp,sizeof(vtmp));
+      }
+    }
+  }
+}
+
+////////////////////////////////////////////////////////////////////////
+// Extract/Insert a single lane  
+////////////////////////////////////////////////////////////////////////
+template<class vobj> accelerator_inline
+typename vobj::scalar_object extractLane(int lane, const vobj & __restrict__ vec)
+{
+  typedef typename vobj::scalar_type   scalar_type;
+  typedef typename vobj::scalar_object scalar_object;
+  typedef typename vobj::vector_type   vector_type;
+  typedef typename ExtractTypeMap<scalar_type>::extract_type extract_type;
+  typedef extract_type * pointer;
+
+  constexpr int words=sizeof(vobj)/sizeof(vector_type);
+  constexpr int Nsimd=vector_type::Nsimd();
+
+  scalar_object extracted;
+  pointer __restrict__  sp = (pointer)&extracted; // Type pun
+  pointer __restrict__  vp = (pointer)&vec;
+  for(int w=0;w<words;w++){
+    sp[w]=vp[w*Nsimd+lane];
+  }
+  return extracted;
+}
+
+template<class vobj> accelerator_inline
+void insertLane(int lane, vobj & __restrict__ vec,const typename vobj::scalar_object & __restrict__ extracted)
+{
+  typedef typename vobj::vector_type vector_type;
+  typedef typename vector_type::scalar_type scalar_type;
+  typedef typename ExtractTypeMap<scalar_type>::extract_type extract_type;
+  typedef extract_type * pointer;
+
+  constexpr int words=sizeof(vobj)/sizeof(vector_type);
+  constexpr int Nsimd=vector_type::Nsimd();
+
+  pointer __restrict__ sp = (pointer)&extracted;
+  pointer __restrict__ vp = (pointer)&vec;
+  for(int w=0;w<words;w++){
+    vp[w*Nsimd+lane]=sp[w];
+  }
 }
 
 ////////////////////////////////////////////////////////////////////////
 // Extract to a bunch of scalar object pointers of different scalar type, with offset. Useful for precision change
 ////////////////////////////////////////////////////////////////////////
-template<class vobj, class sobj> inline 
-void extract1(const vobj &vec,std::vector<sobj*> &extracted, int offset)
+template<class vobj, class sobj> accelerator
+void extract(const vobj &vec,ExtractPointerArray<sobj> &extracted, int offset)
 {
-  typedef typename vobj::scalar_type vobj_scalar_type ;
-  typedef typename vobj::vector_type vobj_vector_type ;
+  typedef typename GridTypeMapper<sobj>::scalar_type sobj_scalar_type;
+  typedef typename GridTypeMapper<vobj>::scalar_type scalar_type;
+  typedef typename GridTypeMapper<vobj>::vector_type vector_type;
 
-  typedef typename sobj::scalar_type sobj_scalar_type ;
-  
-  static const int words=sizeof(vobj)/sizeof(vobj_vector_type);
-  static const int Nsimd=vobj_vector_type::Nsimd();
-
-  int Nextr=extracted.size();
-  int s = Nsimd/Nextr;
-  vobj_scalar_type * vp = (vobj_scalar_type *)&vec;
+  const int words=sizeof(vobj)/sizeof(vector_type);
+  const int Nsimd=vector_type::Nsimd();
+  const int Nextr=extracted.size();
+  const int s = Nsimd/Nextr;
 
+  scalar_type * vp = (scalar_type *)&vec;
+  scalar_type      vtmp;
+  sobj_scalar_type stmp;
   for(int w=0;w<words;w++){
     for(int i=0;i<Nextr;i++){
       sobj_scalar_type * pointer = (sobj_scalar_type *)& extracted[i][offset];
-      pointer[w] = vp[i*s+w*Nsimd];
+      memcpy((char *)&vtmp,(char *)&vp[w*Nsimd+i*s],sizeof(vtmp));
+      stmp = vtmp;
+      memcpy((char *)&pointer[w],(char *)&stmp,sizeof(stmp)); // may do a precision conversion
     }
   }
 }
 
-  
 ////////////////////////////////////////////////////////////////////////
-// Merge a contiguous array of scalar objects
+// Merge bunch of scalar object pointers of different scalar type, with offset. Useful for precision change
 ////////////////////////////////////////////////////////////////////////
-template<class vobj> inline 
-void merge(vobj &vec,std::vector<typename vobj::scalar_object> &extracted)
+template<class vobj, class sobj> accelerator
+void merge(vobj &vec,ExtractPointerArray<sobj> &extracted, int offset)
 {
-  typedef typename vobj::scalar_type scalar_type ;
-  typedef typename vobj::vector_type vector_type ;
-  
-  static const int Nsimd=sizeof(vector_type)/sizeof(scalar_type);
-  static const int words=sizeof(vobj)/sizeof(vector_type);
+  typedef typename GridTypeMapper<sobj>::scalar_type sobj_scalar_type;
+  typedef typename GridTypeMapper<vobj>::scalar_type scalar_type;
+  typedef typename GridTypeMapper<vobj>::vector_type vector_type;
 
-  int Nextr = extracted.size();
-  int splat=Nsimd/Nextr;
-
-  std::vector<scalar_type *> pointers(Nextr);
-  for(int i=0;i<Nextr;i++) 
-    pointers[i] =(scalar_type *)& extracted[i];
-  
-  vector_type *vp = (vector_type *)&vec;
-  for(int w=0;w<words;w++){
-    merge<vector_type,scalar_type>(&vp[w],pointers,w);
-  }
-}
-
-////////////////////////////////////////////////////////////////////////
-// Merge a bunch of different scalar object pointers, with offset
-////////////////////////////////////////////////////////////////////////
-template<class vobj> inline 
-void merge(vobj &vec,std::vector<typename vobj::scalar_object *> &extracted,int offset)
-{
-  typedef typename vobj::scalar_type scalar_type ;
-  typedef typename vobj::vector_type vector_type ;
-  
-  const int Nsimd=sizeof(vector_type)/sizeof(scalar_type);
   const int words=sizeof(vobj)/sizeof(vector_type);
+  const int Nsimd=vector_type::Nsimd();
+  const int Nextr=extracted.size();
+  const int s = Nsimd/Nextr;
 
-  int Nextr=extracted.size();
-  int s=Nsimd/Nextr;
-
-  scalar_type *pointer;
-  scalar_type *vp = (scalar_type *)&vec;
-
-  //  assert( (((uint64_t)vp)&(sizeof(scalar_type)-1)) == 0);
-
+  scalar_type * vp = (scalar_type *)&vec;
+  scalar_type      vtmp;
+  sobj_scalar_type stmp;
   for(int w=0;w<words;w++){
     for(int i=0;i<Nextr;i++){
+      sobj_scalar_type * pointer = (sobj_scalar_type *)& extracted[i][offset];
       for(int ii=0;ii<s;ii++){
-	pointer=(scalar_type *)&extracted[i][offset];
-	vp[w*Nsimd+i*s+ii] = pointer[w];
+	memcpy((char *)&stmp,(char *)&pointer[w],sizeof(stmp));
+	vtmp=stmp;
+	memcpy((char *)&vp[w*Nsimd+i*s+ii],(char *)&vtmp,sizeof(vtmp));
       }
     }
   }
- }
-
-template<class vobj> inline void merge1(vobj &vec,std::vector<typename vobj::scalar_object *> &extracted,int offset)
-{
-  typedef typename vobj::scalar_type scalar_type ;
-  typedef typename vobj::vector_type vector_type ;
-  
-  static const int Nsimd=vobj::vector_type::Nsimd();
-  static const int words=sizeof(vobj)/sizeof(vector_type);
-
-  scalar_type *vp = (scalar_type *)&vec;
-
-  //  assert( (((uint64_t)vp)&(sizeof(scalar_type)-1)) == 0);
-
-  for(int w=0;w<words;w++){
-  for(int i=0;i<Nsimd;i++){
-      vp[w*Nsimd+i] = ((scalar_type *)&extracted[i][offset])[w];
-  }}
 }
 
-template<class vobj> inline void merge2(vobj &vec,std::vector<typename vobj::scalar_object *> &extracted,int offset)
-{
-  typedef typename vobj::scalar_type scalar_type ;
-  typedef typename vobj::vector_type vector_type ;
-  
-  const int Nsimd=vobj::vector_type::Nsimd();
-  const int words=sizeof(vobj)/sizeof(vector_type);
 
-  scalar_type *pointer;
-  scalar_type *vp = (scalar_type *)&vec;
-  //  assert( (((uint64_t)vp)&(sizeof(scalar_type)-1)) == 0);
+NAMESPACE_END(Grid);
 
-  for(int w=0;w<words;w++){
-    for(int i=0;i<Nsimd;i++){
-      pointer=(scalar_type *)&extracted[i][offset];
-      vp[w*Nsimd+i] =pointer[w];
-    }
-  }
-}
-
-}
-
-#endif
diff --git a/Grid/tensors/Tensor_index.h b/Grid/tensors/Tensor_index.h
index 500f6c8d..553801bc 100644
--- a/Grid/tensors/Tensor_index.h
+++ b/Grid/tensors/Tensor_index.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,8 +23,8 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_TENSOR_INDEX_H
 #define GRID_TENSOR_INDEX_H
 
@@ -35,49 +35,49 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 // trace of a different index can distribute across the vector index in a replicated way
 // but we do not trace a vector index.
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-  /* Needed?
-template<int Level> inline ComplexF traceIndex(const ComplexF arg) { return arg;}
-template<int Level> inline ComplexD traceIndex(const ComplexD arg) { return arg;}
-template<int Level> inline RealF traceIndex(const RealF arg) { return arg;}
-template<int Level> inline RealD traceIndex(const RealD arg) { return arg;}
-  */
+/* Needed?
+   template<int Level> accelerator_inline ComplexF traceIndex(const ComplexF arg) { return arg;}
+   template<int Level> accelerator_inline ComplexD traceIndex(const ComplexD arg) { return arg;}
+   template<int Level> accelerator_inline RealF traceIndex(const RealF arg) { return arg;}
+   template<int Level> accelerator_inline RealD traceIndex(const RealD arg) { return arg;}
+*/
 template<int Level> 
 class TensorIndexRecursion {
 
- public:
+public:
 
   ////////////////////////////////////////////////////
   // Type Queries
   ////////////////////////////////////////////////////
-  template<class vtype>       static inline int indexRank(const iScalar<vtype> tmp)  { return TensorIndexRecursion<Level-1>::indexRank(tmp._internal);  }
-  template<class vtype,int N> static inline int indexRank(const iVector<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::indexRank(tmp._internal[0]);  }
-  template<class vtype,int N> static inline int indexRank(const iMatrix<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::indexRank(tmp._internal[0][0]);  }
+  template<class vtype>       static accelerator_inline int indexRank(const iScalar<vtype> tmp)  { return TensorIndexRecursion<Level-1>::indexRank(tmp._internal);  }
+  template<class vtype,int N> static accelerator_inline int indexRank(const iVector<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::indexRank(tmp._internal[0]);  }
+  template<class vtype,int N> static accelerator_inline int indexRank(const iMatrix<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::indexRank(tmp._internal[0][0]);  }
 
-  template<class vtype>       static inline int isScalar(const iScalar<vtype> tmp)  { return TensorIndexRecursion<Level-1>::isScalar(tmp._internal);  }
-  template<class vtype,int N> static inline int isScalar(const iVector<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::isScalar(tmp._internal[0]);  }
-  template<class vtype,int N> static inline int isScalar(const iMatrix<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::isScalar(tmp._internal[0][0]);  }
+  template<class vtype>       static accelerator_inline int isScalar(const iScalar<vtype> tmp)  { return TensorIndexRecursion<Level-1>::isScalar(tmp._internal);  }
+  template<class vtype,int N> static accelerator_inline int isScalar(const iVector<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::isScalar(tmp._internal[0]);  }
+  template<class vtype,int N> static accelerator_inline int isScalar(const iMatrix<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::isScalar(tmp._internal[0][0]);  }
 
-  template<class vtype>       static inline int isVector(const iScalar<vtype> tmp)  { return TensorIndexRecursion<Level-1>::isVector(tmp._internal);  }
-  template<class vtype,int N> static inline int isVector(const iVector<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::isVector(tmp._internal[0]);  }
-  template<class vtype,int N> static inline int isVector(const iMatrix<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::isVector(tmp._internal[0][0]);  }
-
-  template<class vtype>       static inline int isMatrix(const iScalar<vtype> tmp)  { return TensorIndexRecursion<Level-1>::isMatrix(tmp._internal);  }
-  template<class vtype,int N> static inline int isMatrix(const iVector<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::isMatrix(tmp._internal[0]);  }
-  template<class vtype,int N> static inline int isMatrix(const iMatrix<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::isMatrix(tmp._internal[0][0]);  }
+  template<class vtype>       static accelerator_inline int isVector(const iScalar<vtype> tmp)  { return TensorIndexRecursion<Level-1>::isVector(tmp._internal);  }
+  template<class vtype,int N> static accelerator_inline int isVector(const iVector<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::isVector(tmp._internal[0]);  }
+  template<class vtype,int N> static accelerator_inline int isVector(const iMatrix<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::isVector(tmp._internal[0][0]);  }
+  
+  template<class vtype>       static accelerator_inline int isMatrix(const iScalar<vtype> tmp)  { return TensorIndexRecursion<Level-1>::isMatrix(tmp._internal);  }
+  template<class vtype,int N> static accelerator_inline int isMatrix(const iVector<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::isMatrix(tmp._internal[0]);  }
+  template<class vtype,int N> static accelerator_inline int isMatrix(const iMatrix<vtype,N> tmp){ return TensorIndexRecursion<Level-1>::isMatrix(tmp._internal[0][0]);  }
   ////////////////////////////////////////////////////
   // Trace
   ////////////////////////////////////////////////////
   template<class vtype>
-  static auto traceIndex(const iScalar<vtype> arg) ->  iScalar<decltype(TensorIndexRecursion<Level-1>::traceIndex(arg._internal))> 
+  static accelerator_inline auto traceIndex(const iScalar<vtype> arg) ->  iScalar<decltype(TensorIndexRecursion<Level-1>::traceIndex(arg._internal))> 
   {
     iScalar<decltype(TensorIndexRecursion<Level-1>::traceIndex(arg._internal))> ret;
     ret._internal = TensorIndexRecursion<Level-1>::traceIndex(arg._internal);
     return ret;
   }
   template<class vtype,int N>
-  static auto traceIndex(const iVector<vtype,N> arg) ->  iVector<decltype(TensorIndexRecursion<Level-1>::traceIndex(arg._internal[0])),N> 
+  static accelerator_inline auto traceIndex(const iVector<vtype,N> arg) ->  iVector<decltype(TensorIndexRecursion<Level-1>::traceIndex(arg._internal[0])),N> 
   {
     iVector<decltype(TensorIndexRecursion<Level-1>::traceIndex(arg._internal[0])),N> ret;
     for(int i=0;i<N;i++){
@@ -86,27 +86,27 @@ class TensorIndexRecursion {
     return ret;
   }
   template<class vtype,int N>
-  static auto traceIndex(const iMatrix<vtype,N> arg) ->  iMatrix<decltype(TensorIndexRecursion<Level-1>::traceIndex(arg._internal[0][0])),N> 
+  static accelerator_inline auto traceIndex(const iMatrix<vtype,N> arg) ->  iMatrix<decltype(TensorIndexRecursion<Level-1>::traceIndex(arg._internal[0][0])),N> 
   {
     iMatrix<decltype(TensorIndexRecursion<Level-1>::traceIndex(arg._internal[0][0])),N> ret;
     for(int i=0;i<N;i++){
-    for(int j=0;j<N;j++){
-      ret._internal[i][j] = TensorIndexRecursion<Level-1>::traceIndex(arg._internal[i][j]);
-    }}
+      for(int j=0;j<N;j++){
+	ret._internal[i][j] = TensorIndexRecursion<Level-1>::traceIndex(arg._internal[i][j]);
+      }}
     return ret;
   }
   ////////////////////////////////////////////
   // Recursion for peeking a specific index
   ////////////////////////////////////////////
   template<class vtype>
-  static auto peekIndex(const iScalar<vtype> arg,int i) ->  iScalar<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal,0))> 
+  static accelerator_inline auto peekIndex(const iScalar<vtype> arg,int i) ->  iScalar<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal,0))> 
   {
     iScalar<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal,0))> ret;
     ret._internal = TensorIndexRecursion<Level-1>::peekIndex(arg._internal,i);
     return ret;
   }
   template<class vtype>
-  static auto peekIndex(const iScalar<vtype> arg,int i,int j) ->  iScalar<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal,0,0))> 
+  static accelerator_inline auto peekIndex(const iScalar<vtype> arg,int i,int j) ->  iScalar<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal,0,0))> 
   {
     iScalar<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal,0,0))> ret;
     ret._internal = TensorIndexRecursion<Level-1>::peekIndex(arg._internal,i,j);
@@ -114,7 +114,7 @@ class TensorIndexRecursion {
   }
 
   template<class vtype,int N>
-  static auto peekIndex(const iVector<vtype,N> arg,int ii) ->  iVector<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal[0],0)),N> 
+  static accelerator_inline auto peekIndex(const iVector<vtype,N> arg,int ii) ->  iVector<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal[0],0)),N> 
   {
     iVector<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal[0],0)),N> ret;
     for(int i=0;i<N;i++){
@@ -123,7 +123,8 @@ class TensorIndexRecursion {
     return ret;
   }
   template<class vtype,int N>
-  static auto peekIndex(const iVector<vtype,N> arg,int ii,int jj) ->  iVector<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal[0],0,0)),N> 
+  static accelerator_inline auto peekIndex(const iVector<vtype,N> arg,int ii,int jj) 
+    ->  iVector<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal[0],0,0)),N> 
   {
     iVector<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal[0],0,0)),N> ret;
     for(int i=0;i<N;i++){
@@ -131,85 +132,86 @@ class TensorIndexRecursion {
     }
     return ret;
   }
-
+  
   template<class vtype,int N>
-  static auto peekIndex(const iMatrix<vtype,N> arg,int ii) ->  iMatrix<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal[0][0],0)),N> 
+  static accelerator_inline auto peekIndex(const iMatrix<vtype,N> arg,int ii) ->  iMatrix<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal[0][0],0)),N> 
   {
     iMatrix<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal[0][0],0)),N> ret;
     for(int i=0;i<N;i++){
-    for(int j=0;j<N;j++){
-      ret._internal[i][j] = TensorIndexRecursion<Level-1>::peekIndex(arg._internal[i][j],ii);
-    }}
+      for(int j=0;j<N;j++){
+	ret._internal[i][j] = TensorIndexRecursion<Level-1>::peekIndex(arg._internal[i][j],ii);
+      }}
     return ret;
   }
   template<class vtype,int N>
-  static auto peekIndex(const iMatrix<vtype,N> arg,int ii,int jj) ->  iMatrix<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal[0][0],0,0)),N> 
+  static accelerator_inline auto peekIndex(const iMatrix<vtype,N> arg,int ii,int jj) 
+    ->  iMatrix<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal[0][0],0,0)),N> 
   {
     iMatrix<decltype(TensorIndexRecursion<Level-1>::peekIndex(arg._internal[0][0],0,0)),N> ret;
     for(int i=0;i<N;i++){
-    for(int j=0;j<N;j++){
-      ret._internal[i][j] = TensorIndexRecursion<Level-1>::peekIndex(arg._internal[i][j],ii,jj);
-    }}
+      for(int j=0;j<N;j++){
+	ret._internal[i][j] = TensorIndexRecursion<Level-1>::peekIndex(arg._internal[i][j],ii,jj);
+      }}
     return ret;
   }
   ////////////////////////////////////////////
   // Recursion for poking a specific index
   ////////////////////////////////////////////
-
-  template<class vtype> inline static 
-    void pokeIndex(iScalar<vtype> &ret, const iScalar<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal,0))> &arg, int i)
-    {
-      TensorIndexRecursion<Level-1>::pokeIndex(ret._internal,arg._internal,i);
+  
+  template<class vtype> accelerator_inline static 
+  void pokeIndex(iScalar<vtype> &ret, const iScalar<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal,0))> &arg, int i)
+  {
+    TensorIndexRecursion<Level-1>::pokeIndex(ret._internal,arg._internal,i);
+  }
+  template<class vtype> accelerator_inline static 
+  void pokeIndex(iScalar<vtype> &ret, const iScalar<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal,0,0))> &arg, int i,int j)
+  {
+    TensorIndexRecursion<Level-1>::pokeIndex(ret._internal,arg._internal,i,j);
+  }
+  
+  template<class vtype,int N> accelerator_inline static 
+  void pokeIndex(iVector<vtype,N> &ret, const iVector<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal[0],0)),N> &arg, int i)
+  {
+    for(int ii=0;ii<N;ii++){
+      TensorIndexRecursion<Level-1>::pokeIndex(ret._internal[ii],arg._internal[ii],i);
     }
-  template<class vtype> inline static 
-    void pokeIndex(iScalar<vtype> &ret, const iScalar<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal,0,0))> &arg, int i,int j)
-    {
-      TensorIndexRecursion<Level-1>::pokeIndex(ret._internal,arg._internal,i,j);
+  }
+  template<class vtype,int N> accelerator_inline static 
+  void pokeIndex(iVector<vtype,N> &ret, const iVector<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal[0],0,0)),N> &arg, int i,int j)
+  {
+    for(int ii=0;ii<N;ii++){
+      TensorIndexRecursion<Level-1>::pokeIndex(ret._internal[ii],arg._internal[ii],i,j);
     }
-
-  template<class vtype,int N> inline static 
-    void pokeIndex(iVector<vtype,N> &ret, const iVector<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal[0],0)),N> &arg, int i)
-    {
-      for(int ii=0;ii<N;ii++){
-	TensorIndexRecursion<Level-1>::pokeIndex(ret._internal[ii],arg._internal[ii],i);
-      }
-    }
-  template<class vtype,int N> inline static 
-    void pokeIndex(iVector<vtype,N> &ret, const iVector<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal[0],0,0)),N> &arg, int i,int j)
-    {
-      for(int ii=0;ii<N;ii++){
-	TensorIndexRecursion<Level-1>::pokeIndex(ret._internal[ii],arg._internal[ii],i,j);
-      }
-    }
-
-  template<class vtype,int N> inline static 
-    void pokeIndex(iMatrix<vtype,N> &ret, const iMatrix<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal[0][0],0)),N> &arg, int i)
-    {
-      for(int ii=0;ii<N;ii++){
+  }
+  
+  template<class vtype,int N> accelerator_inline static 
+  void pokeIndex(iMatrix<vtype,N> &ret, const iMatrix<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal[0][0],0)),N> &arg, int i)
+  {
+    for(int ii=0;ii<N;ii++){
       for(int jj=0;jj<N;jj++){
 	TensorIndexRecursion<Level-1>::pokeIndex(ret._internal[ii][jj],arg._internal[ii][jj],i);
       }}
-    }
-  template<class vtype,int N> inline static 
-    void pokeIndex(iMatrix<vtype,N> &ret, const iMatrix<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal[0][0],0,0)),N> &arg, int i,int j)
-    {
-      for(int ii=0;ii<N;ii++){
+  }
+  template<class vtype,int N> accelerator_inline static 
+  void pokeIndex(iMatrix<vtype,N> &ret, const iMatrix<decltype(TensorIndexRecursion<Level-1>::peekIndex(ret._internal[0][0],0,0)),N> &arg, int i,int j)
+  {
+    for(int ii=0;ii<N;ii++){
       for(int jj=0;jj<N;jj++){
 	TensorIndexRecursion<Level-1>::pokeIndex(ret._internal[ii][jj],arg._internal[ii][jj],i,j);
       }}
-    }
+  }
 
   ////////////////////////////////////////////
   // Recursion for transposing a specific index
   ////////////////////////////////////////////
-  template<class vtype>
+  template<class vtype> accelerator_inline 
   static auto transposeIndex(const iScalar<vtype> arg) ->  iScalar<vtype> 
   {
     iScalar<vtype> ret;
     ret._internal = TensorIndexRecursion<Level-1>::transposeIndex(arg._internal);
     return ret;
   }
-  template<class vtype,int N>
+  template<class vtype,int N> accelerator_inline 
   static auto transposeIndex(const iVector<vtype,N> arg) ->  iVector<vtype,N> 
   {
     iVector<vtype,N> ret;
@@ -218,14 +220,14 @@ class TensorIndexRecursion {
     }
     return ret;
   }
-  template<class vtype,int N>
+  template<class vtype,int N> accelerator_inline 
   static auto transposeIndex(const iMatrix<vtype,N> arg) ->  iMatrix<vtype,N> 
   {
     iMatrix<vtype,N> ret;
     for(int i=0;i<N;i++){
-    for(int j=0;j<N;j++){
-      ret._internal[i][j] = TensorIndexRecursion<Level-1>::transposeIndex(arg._internal[i][j]);
-    }}
+      for(int j=0;j<N;j++){
+	ret._internal[i][j] = TensorIndexRecursion<Level-1>::transposeIndex(arg._internal[i][j]);
+      }}
     return ret;
   }
 };
@@ -236,51 +238,51 @@ class TensorIndexRecursion {
 #define RemoveCRV(a) typename std::remove_const<typename std::remove_reference<decltype(a)>::type>::type
 template<>
 class TensorIndexRecursion<0> {
- public:
+public:
   ////////////////////////////////////////////////////
   // Type Queries
   ////////////////////////////////////////////////////
-  template<class vtype>       static inline int indexRank(const iScalar<vtype> tmp)  { return 1; }
-  template<class vtype,int N> static inline int indexRank(const iVector<vtype,N> tmp){ return N; }
-  template<class vtype,int N> static inline int indexRank(const iMatrix<vtype,N> tmp){ return N; }
+  template<class vtype>       static accelerator_inline int indexRank(const iScalar<vtype> tmp)  { return 1; }
+  template<class vtype,int N> static accelerator_inline int indexRank(const iVector<vtype,N> tmp){ return N; }
+  template<class vtype,int N> static accelerator_inline int indexRank(const iMatrix<vtype,N> tmp){ return N; }
 
-  template<class vtype>       static inline int isScalar(const iScalar<vtype> tmp)  { return true;}
-  template<class vtype,int N> static inline int isScalar(const iVector<vtype,N> tmp){ return false;}
-  template<class vtype,int N> static inline int isScalar(const iMatrix<vtype,N> tmp){ return false;}
+  template<class vtype>       static accelerator_inline int isScalar(const iScalar<vtype> tmp)  { return true;}
+  template<class vtype,int N> static accelerator_inline int isScalar(const iVector<vtype,N> tmp){ return false;}
+  template<class vtype,int N> static accelerator_inline int isScalar(const iMatrix<vtype,N> tmp){ return false;}
 
-  template<class vtype>       static inline int isVector(const iScalar<vtype> tmp)  { return false;}
-  template<class vtype,int N> static inline int isVector(const iVector<vtype,N> tmp){ return true;}
-  template<class vtype,int N> static inline int isVector(const iMatrix<vtype,N> tmp){ return false;}
+  template<class vtype>       static accelerator_inline int isVector(const iScalar<vtype> tmp)  { return false;}
+  template<class vtype,int N> static accelerator_inline int isVector(const iVector<vtype,N> tmp){ return true;}
+  template<class vtype,int N> static accelerator_inline int isVector(const iMatrix<vtype,N> tmp){ return false;}
 
-  template<class vtype>       static inline int isMatrix(const iScalar<vtype> tmp)  { return false;}
-  template<class vtype,int N> static inline int isMatrix(const iVector<vtype,N> tmp){ return false;}
-  template<class vtype,int N> static inline int isMatrix(const iMatrix<vtype,N> tmp){ return true;}
+  template<class vtype>       static accelerator_inline int isMatrix(const iScalar<vtype> tmp)  { return false;}
+  template<class vtype,int N> static accelerator_inline int isMatrix(const iVector<vtype,N> tmp){ return false;}
+  template<class vtype,int N> static accelerator_inline int isMatrix(const iMatrix<vtype,N> tmp){ return true;}
 
   /////////////////////////////////////////
   // Ends recursion for trace (scalar/vector/matrix)
   /////////////////////////////////////////
-  template<class vtype>
+  template<class vtype> accelerator_inline 
   static auto traceIndex(const iScalar<vtype> arg) ->  iScalar<RemoveCRV(arg._internal)>
   {
     iScalar<RemoveCRV(arg._internal)> ret;
     ret._internal = arg._internal;
     return ret;
   }
-  template<class vtype,int N>
+  template<class vtype,int N> accelerator_inline 
   static auto traceIndex(const iVector<vtype,N> arg) ->  iScalar<RemoveCRV(arg._internal[0])>
   {
     iScalar<RemoveCRV(arg._internal[0])> ret;
-    ret._internal=zero;
+    ret._internal=Zero();
     for(int i=0;i<N;i++){
       ret._internal = ret._internal+ arg._internal[i];
     }
     return ret;
   }
-  template<class vtype,int N>
+  template<class vtype,int N> accelerator_inline 
   static auto traceIndex(const iMatrix<vtype,N> arg) ->  iScalar<RemoveCRV(arg._internal[0][0])> 
   {
     iScalar<RemoveCRV(arg._internal[0][0])> ret;
-    ret=zero;
+    zeroit(ret);
     for(int i=0;i<N;i++){
       ret._internal = ret._internal+arg._internal[i][i];
     }
@@ -289,35 +291,35 @@ class TensorIndexRecursion<0> {
   /////////////////////////////////////////
   // Ends recursion for transpose scalar/matrix ; no way to terminate on vector
   /////////////////////////////////////////
-  template<class vtype>
+  template<class vtype> accelerator_inline 
   static auto transposeIndex(const iScalar<vtype> arg) ->  iScalar<vtype>
   {
     iScalar<vtype> ret;
     ret._internal = arg._internal;
     return ret;
   }
-  template<class vtype,int N>
+  template<class vtype,int N> accelerator_inline 
   static auto transposeIndex(const iMatrix<vtype,N> arg)  ->  iMatrix<vtype,N> 
   {
     iMatrix<vtype,N> ret;
-    ret=zero;
+    ret=Zero();
     for(int i=0;i<N;i++){
-    for(int j=0;j<N;j++){
-      ret._internal[i][j] = ret._internal[i][j]+arg._internal[i][j];
-    }}
+      for(int j=0;j<N;j++){
+	ret._internal[i][j] = ret._internal[i][j]+arg._internal[i][j];
+      }}
     return ret;
   }
   ////////////////////////////////////////////////////////////////////////////////////////////////////////////////
   // End recursion for peeking a specific index; single index on vector, double index on matrix
   ////////////////////////////////////////////////////////////////////////////////////////////////////////////////
-  template<class vtype,int N>
+  template<class vtype,int N> accelerator_inline 
   static auto peekIndex(const iVector<vtype,N> arg,int ii) ->  iScalar<vtype> 
   {
     iScalar<vtype> ret;
     ret._internal = arg._internal[ii];
     return ret;
   }
-  template<class vtype,int N>
+  template<class vtype,int N> accelerator_inline 
   static auto peekIndex(const iMatrix<vtype,N> arg,int ii,int jj) ->  iScalar<vtype>
   {
     iScalar<vtype> ret;
@@ -325,77 +327,77 @@ class TensorIndexRecursion<0> {
     return ret;
   }
   // Vector poke, one index
-  template<class vtype,int N> inline static 
-    void pokeIndex(iVector<vtype,N> &ret, const iScalar<vtype> &arg,int i)
-    {
-      ret._internal[i] = arg._internal;
-    }
+  template<class vtype,int N> accelerator_inline static 
+  void pokeIndex(iVector<vtype,N> &ret, const iScalar<vtype> &arg,int i)
+  {
+    ret._internal[i] = arg._internal;
+  }
   // Matrix poke two indices
-  template<class vtype,int N> inline static 
-    void pokeIndex(iMatrix<vtype,N> &ret, const iScalar<vtype> &arg,int i,int j)
-    {
-      ret._internal[i][j] = arg._internal;
-    }
-
+  template<class vtype,int N> accelerator_inline static 
+  void pokeIndex(iMatrix<vtype,N> &ret, const iScalar<vtype> &arg,int i,int j)
+  {
+    ret._internal[i][j] = arg._internal;
+  }
+  
 };
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////////
 // External wrappers
 ////////////////////////////////////////////////////////////////////////////////////////////////////////
-template<int Level,class vtype> inline int indexRank(void)
+template<int Level,class vtype> accelerator_inline int indexRank(void)
 {
   vtype tmp;
   return TensorIndexRecursion<Level>::indexRank(tmp);
 }
-template<int Level,class vtype> inline int isScalar(void)
+template<int Level,class vtype> accelerator_inline int isScalar(void)
 {
   vtype tmp;
   return TensorIndexRecursion<Level>::isScalar(tmp);
 }
-template<int Level,class vtype> inline int isVector(void)
+template<int Level,class vtype> accelerator_inline int isVector(void)
 {
   vtype tmp;
   return TensorIndexRecursion<Level>::isVector(tmp);
 }
-template<int Level,class vtype> inline int isMatrix(void)
+template<int Level,class vtype> accelerator_inline int isMatrix(void)
 {
   vtype tmp;
   return TensorIndexRecursion<Level>::isMatrix(tmp);
 }
 
-template<int Level,class vtype> inline auto traceIndex (const vtype &arg) -> RemoveCRV(TensorIndexRecursion<Level>::traceIndex(arg))
+template<int Level,class vtype> accelerator_inline auto traceIndex (const vtype &arg) -> RemoveCRV(TensorIndexRecursion<Level>::traceIndex(arg))
 {
   RemoveCRV(TensorIndexRecursion<Level>::traceIndex(arg)) ret;
   ret=TensorIndexRecursion<Level>::traceIndex(arg);
   return ret;
 }
-template<int Level,class vtype> inline auto transposeIndex (const vtype &arg) -> RemoveCRV(TensorIndexRecursion<Level>::transposeIndex(arg))
+template<int Level,class vtype> accelerator_inline auto transposeIndex (const vtype &arg) -> RemoveCRV(TensorIndexRecursion<Level>::transposeIndex(arg))
 {
   RemoveCRV(TensorIndexRecursion<Level>::transposeIndex(arg)) ret;
   ret=TensorIndexRecursion<Level>::transposeIndex(arg);
   return ret;
 }
 
-template<int Level,class vtype> inline auto peekIndex (const vtype &arg,int i) -> RemoveCRV(TensorIndexRecursion<Level>::peekIndex(arg,0))
+template<int Level,class vtype> accelerator_inline auto peekIndex (const vtype &arg,int i) -> RemoveCRV(TensorIndexRecursion<Level>::peekIndex(arg,0))
 {
   RemoveCRV(TensorIndexRecursion<Level>::peekIndex(arg,0)) ret;
   ret=TensorIndexRecursion<Level>::peekIndex(arg,i);
   return ret;
 }
-template<int Level,class vtype> inline auto peekIndex (const vtype &arg,int i,int j) -> RemoveCRV(TensorIndexRecursion<Level>::peekIndex(arg,0,0))
+template<int Level,class vtype> accelerator_inline auto peekIndex (const vtype &arg,int i,int j) -> RemoveCRV(TensorIndexRecursion<Level>::peekIndex(arg,0,0))
 {
   RemoveCRV(TensorIndexRecursion<Level>::peekIndex(arg,0,0)) ret;
   ret=TensorIndexRecursion<Level>::peekIndex(arg,i,j);
   return ret;
 }
 
-template<int Level,class vtype> inline 
+template<int Level,class vtype> accelerator_inline 
 void pokeIndex (vtype &ret,const decltype(TensorIndexRecursion<Level>::peekIndex(ret,0)) &arg,int i) 
 {
   TensorIndexRecursion<Level>::pokeIndex(ret,arg,i);
 }
 
-template<int Level,class vtype> inline 
+template<int Level,class vtype> accelerator_inline 
 void pokeIndex (vtype &ret,const decltype(TensorIndexRecursion<Level>::peekIndex(ret,0,0)) &arg,int i,int j) 
 {
   TensorIndexRecursion<Level>::pokeIndex(ret,arg,i,j);
@@ -404,5 +406,6 @@ void pokeIndex (vtype &ret,const decltype(TensorIndexRecursion<Level>::peekIndex
 
 #undef RemoveCRV
 
-}
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/tensors/Tensor_inner.h b/Grid/tensors/Tensor_inner.h
index 46185652..03f72966 100644
--- a/Grid/tensors/Tensor_inner.h
+++ b/Grid/tensors/Tensor_inner.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,40 +24,137 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_INNER_H
 #define GRID_MATH_INNER_H
-namespace Grid {
-  ///////////////////////////////////////////////////////////////////////////////////////
-  // innerProduct Scalar x Scalar -> Scalar
-  // innerProduct Vector x Vector -> Scalar
-  // innerProduct Matrix x Matrix -> Scalar
-  ///////////////////////////////////////////////////////////////////////////////////////
-  template<class sobj> inline RealD norm2(const sobj &arg){
-    auto nrm = innerProductD(arg,arg);
-    RealD ret = real(nrm);
-    return ret;
+
+NAMESPACE_BEGIN(Grid);
+
+///////////////////////////////////////////////////////////////////////////////////////
+// innerProduct Scalar x Scalar -> Scalar
+// innerProduct Vector x Vector -> Scalar
+// innerProduct Matrix x Matrix -> Scalar
+///////////////////////////////////////////////////////////////////////////////////////
+template<class sobj> accelerator_inline RealD norm2(const sobj &arg){
+  auto nrm = innerProductD(arg,arg);
+  RealD ret = real(nrm);
+  return ret;
+}
+
+////////////////////////////////////////////////////////////////
+// FIXME horizontal sum: single promote to double and sum 
+// Not yet used; make sum_cpu in LatticeReduction.h call these
+// GPU sum_gpu is using a double precision promotion.
+////////////////////////////////////////////////////////////////
+#if 0
+accelerator_inline ComplexD ReduceD(const ComplexF &l) { return l; };
+accelerator_inline ComplexD ReduceD(const ComplexD &l) { return l; };
+accelerator_inline RealD    ReduceD(const RealD    &l) { return l; };
+accelerator_inline RealD    ReduceD(const RealF    &l) { return l; };
+
+accelerator_inline ComplexD ReduceD(const vComplexD &l) { return Reduce(l); };
+accelerator_inline RealD    ReduceD(const vRealD    &l) { return Reduce(l); };
+accelerator_inline ComplexD ReduceD(const vComplexF  &l) 
+{ 
+  vComplexD la,lb;
+  Optimization::PrecisionChange::StoD(l.v,la.v,lb.v);
+  return Reduce(la)+Reduce(lb); 
+}
+accelerator_inline RealD    ReduceD(const vRealF    &l) 
+{
+  vRealD la,lb;
+  Optimization::PrecisionChange::StoD(l.v,la.v,lb.v);
+  return Reduce(la)+Reduce(lb); 
+};
+///////////////////////////////////////////////////////
+// Now do it for vector, matrix, scalar
+///////////////////////////////////////////////////////
+template<class l,int N> accelerator_inline
+auto ReduceD (const iVector<l,N>& lhs) -> iVector<decltype(ReduceD(lhs._internal[0])),N>
+{
+  typedef decltype(ReduceD(lhs._internal[0])) ret_t;
+  iVector<ret_t,N> ret;
+  for(int c1=0;c1<N;c1++){
+    ret._internal[c1] = ReduceD(lhs._internal[c1]);
   }
-  //////////////////////////////////////
-  // If single promote to double and sum 2x
-  //////////////////////////////////////
+  return ret;
+}
+template<class l,int N> accelerator_inline
+auto ReduceD (const iMatrix<l,N>& lhs) -> iMatrix<decltype(ReduceD(lhs._internal[0][0])),N>
+{
+  typedef decltype(ReduceD(lhs._internal[0][0])) ret_t;
+  iMatrix<ret_t,N> ret;
+  for(int c1=0;c1<N;c1++){
+  for(int c2=0;c2<N;c2++){
+    ret._internal[c1][c2]=ReduceD(lhs._internal[c1][c2]);
+  }}
+  return ret;
+}
+template<class l> accelerator_inline
+auto ReduceD (const iScalar<l>& lhs) -> iScalar<decltype(ReduceD(lhs._internal))>
+{
+  typedef decltype(ReduceD(lhs._internal)) ret_t;
+  iScalar<ret_t> ret;
+  ret._internal = ReduceD(lhs._internal);
+  return ret;
+}
+#endif
+///////////////////////////////////////////////////////
+// Now do Reduce for vector, matrix, scalar
+///////////////////////////////////////////////////////
+template<class l,int N> accelerator_inline
+auto Reduce (const iVector<l,N>& lhs) -> iVector<decltype(Reduce(lhs._internal[0])),N>
+{
+  typedef decltype(Reduce(lhs._internal[0])) ret_t;
+  iVector<ret_t,N> ret;
+  for(int c1=0;c1<N;c1++){
+    ret._internal[c1] = Reduce(lhs._internal[c1]);
+  }
+  return ret;
+}
+template<class l,int N> accelerator_inline
+auto Reduce (const iMatrix<l,N>& lhs) -> iMatrix<decltype(Reduce(lhs._internal[0][0])),N>
+{
+  typedef decltype(Reduce(lhs._internal[0][0])) ret_t;
+  iMatrix<ret_t,N> ret;
+  for(int c1=0;c1<N;c1++){
+  for(int c2=0;c2<N;c2++){
+    ret._internal[c1][c2]=Reduce(lhs._internal[c1][c2]);
+  }}
+  return ret;
+}
+template<class l> accelerator_inline
+auto Reduce (const iScalar<l>& lhs) -> iScalar<decltype(Reduce(lhs._internal))>
+{
+  typedef decltype(Reduce(lhs._internal)) ret_t;
+  iScalar<ret_t> ret;
+  ret._internal = Reduce(lhs._internal);
+  return ret;
+}
 
-inline ComplexD innerProductD(const ComplexF &l,const ComplexF &r){  return innerProduct(l,r); }
-inline ComplexD innerProductD(const ComplexD &l,const ComplexD &r){  return innerProduct(l,r); }
-inline RealD    innerProductD(const RealD    &l,const RealD    &r){  return innerProduct(l,r); }
-inline RealD    innerProductD(const RealF    &l,const RealF    &r){  return innerProduct(l,r); }
 
-inline vComplexD innerProductD(const vComplexD &l,const vComplexD &r){  return innerProduct(l,r); }
-inline vRealD    innerProductD(const vRealD    &l,const vRealD    &r){  return innerProduct(l,r); }
-inline vComplexD innerProductD(const vComplexF &l,const vComplexF &r){  
+
+//////////////////////////////////////
+// innerProductD : if single promote to double and evaluate with sum 2x
+//////////////////////////////////////
+accelerator_inline ComplexD innerProductD(const ComplexF &l,const ComplexF &r){  return innerProduct(l,r); }
+accelerator_inline ComplexD innerProductD(const ComplexD &l,const ComplexD &r){  return innerProduct(l,r); }
+accelerator_inline RealD    innerProductD(const RealD    &l,const RealD    &r){  return innerProduct(l,r); }
+accelerator_inline RealD    innerProductD(const RealF    &l,const RealF    &r){  return innerProduct(l,r); }
+
+accelerator_inline vComplexD innerProductD(const vComplexD &l,const vComplexD &r){  return innerProduct(l,r); }
+accelerator_inline vRealD    innerProductD(const vRealD    &l,const vRealD    &r){  return innerProduct(l,r); }
+accelerator_inline vComplexD innerProductD(const vComplexF &l,const vComplexF &r)
+{  
   vComplexD la,lb;
   vComplexD ra,rb;
   Optimization::PrecisionChange::StoD(l.v,la.v,lb.v);
   Optimization::PrecisionChange::StoD(r.v,ra.v,rb.v);
   return innerProduct(la,ra) + innerProduct(lb,rb); 
 }
-inline vRealD innerProductD(const vRealF &l,const vRealF &r){  
+accelerator_inline vRealD innerProductD(const vRealF &l,const vRealF &r)
+{  
   vRealD la,lb;
   vRealD ra,rb;
   Optimization::PrecisionChange::StoD(l.v,la.v,lb.v);
@@ -65,73 +162,73 @@ inline vRealD innerProductD(const vRealF &l,const vRealF &r){
   return innerProduct(la,ra) + innerProduct(lb,rb); 
 }
 
-  template<class l,class r,int N> inline
-  auto innerProductD (const iVector<l,N>& lhs,const iVector<r,N>& rhs) -> iScalar<decltype(innerProductD(lhs._internal[0],rhs._internal[0]))>
-  {
-    typedef decltype(innerProductD(lhs._internal[0],rhs._internal[0])) ret_t;
-    iScalar<ret_t> ret;
-    ret=zero;
-    for(int c1=0;c1<N;c1++){
-      ret._internal += innerProductD(lhs._internal[c1],rhs._internal[c1]);
-    }
-    return ret;
+// Now do it for vector, matrix, scalar
+template<class l,class r,int N> accelerator_inline
+auto innerProductD (const iVector<l,N>& lhs,const iVector<r,N>& rhs) -> iScalar<decltype(innerProductD(lhs._internal[0],rhs._internal[0]))>
+{
+  typedef decltype(innerProductD(lhs._internal[0],rhs._internal[0])) ret_t;
+  iScalar<ret_t> ret;
+  zeroit(ret);
+  for(int c1=0;c1<N;c1++){
+    ret._internal += innerProductD(lhs._internal[c1],rhs._internal[c1]);
   }
-  template<class l,class r,int N> inline
-  auto innerProductD (const iMatrix<l,N>& lhs,const iMatrix<r,N>& rhs) -> iScalar<decltype(innerProductD(lhs._internal[0][0],rhs._internal[0][0]))>
-  {
-    typedef decltype(innerProductD(lhs._internal[0][0],rhs._internal[0][0])) ret_t;
-    iScalar<ret_t> ret;
-    iScalar<ret_t> tmp;
-    ret=zero;
-    for(int c1=0;c1<N;c1++){
+  return ret;
+}
+template<class l,class r,int N> accelerator_inline
+auto innerProductD (const iMatrix<l,N>& lhs,const iMatrix<r,N>& rhs) -> iScalar<decltype(innerProductD(lhs._internal[0][0],rhs._internal[0][0]))>
+{
+  typedef decltype(innerProductD(lhs._internal[0][0],rhs._internal[0][0])) ret_t;
+  iScalar<ret_t> ret;
+  ret=Zero();
+  for(int c1=0;c1<N;c1++){
     for(int c2=0;c2<N;c2++){
       ret._internal+=innerProductD(lhs._internal[c1][c2],rhs._internal[c1][c2]);
-    }}
-    return ret;
+  }}
+  return ret;
+}
+template<class l,class r> accelerator_inline
+auto innerProductD (const iScalar<l>& lhs,const iScalar<r>& rhs) -> iScalar<decltype(innerProductD(lhs._internal,rhs._internal))>
+{
+  typedef decltype(innerProductD(lhs._internal,rhs._internal)) ret_t;
+  iScalar<ret_t> ret;
+  ret._internal = innerProductD(lhs._internal,rhs._internal);
+  return ret;
+}
+//////////////////////
+// Keep same precison
+//////////////////////
+template<class l,class r,int N> accelerator_inline
+auto innerProduct (const iVector<l,N>& lhs,const iVector<r,N>& rhs) -> iScalar<decltype(innerProduct(lhs._internal[0],rhs._internal[0]))>
+{
+  typedef decltype(innerProduct(lhs._internal[0],rhs._internal[0])) ret_t;
+  iScalar<ret_t> ret;
+  ret=Zero();
+  for(int c1=0;c1<N;c1++){
+    ret._internal += innerProduct(lhs._internal[c1],rhs._internal[c1]);
   }
-  template<class l,class r> inline
-  auto innerProductD (const iScalar<l>& lhs,const iScalar<r>& rhs) -> iScalar<decltype(innerProductD(lhs._internal,rhs._internal))>
-  {
-    typedef decltype(innerProductD(lhs._internal,rhs._internal)) ret_t;
-    iScalar<ret_t> ret;
-    ret._internal = innerProductD(lhs._internal,rhs._internal);
-    return ret;
-  }
-  //////////////////////
-  // Keep same precison
-  //////////////////////
-  template<class l,class r,int N> inline
-  auto innerProduct (const iVector<l,N>& lhs,const iVector<r,N>& rhs) -> iScalar<decltype(innerProduct(lhs._internal[0],rhs._internal[0]))>
-  {
-    typedef decltype(innerProduct(lhs._internal[0],rhs._internal[0])) ret_t;
-    iScalar<ret_t> ret;
-    ret=zero;
-    for(int c1=0;c1<N;c1++){
-      ret._internal += innerProduct(lhs._internal[c1],rhs._internal[c1]);
-    }
-    return ret;
-  }
-  template<class l,class r,int N> inline
-  auto innerProduct (const iMatrix<l,N>& lhs,const iMatrix<r,N>& rhs) -> iScalar<decltype(innerProduct(lhs._internal[0][0],rhs._internal[0][0]))>
-  {
-    typedef decltype(innerProduct(lhs._internal[0][0],rhs._internal[0][0])) ret_t;
-    iScalar<ret_t> ret;
-    iScalar<ret_t> tmp;
-    ret=zero;
-    for(int c1=0;c1<N;c1++){
+  return ret;
+}
+template<class l,class r,int N> accelerator_inline
+auto innerProduct (const iMatrix<l,N>& lhs,const iMatrix<r,N>& rhs) -> iScalar<decltype(innerProduct(lhs._internal[0][0],rhs._internal[0][0]))>
+{
+  typedef decltype(innerProduct(lhs._internal[0][0],rhs._internal[0][0])) ret_t;
+  iScalar<ret_t> ret;
+  ret=Zero();
+  for(int c1=0;c1<N;c1++){
     for(int c2=0;c2<N;c2++){
       ret._internal+=innerProduct(lhs._internal[c1][c2],rhs._internal[c1][c2]);
     }}
-    return ret;
-  }
-  template<class l,class r> inline
-  auto innerProduct (const iScalar<l>& lhs,const iScalar<r>& rhs) -> iScalar<decltype(innerProduct(lhs._internal,rhs._internal))>
-  {
-    typedef decltype(innerProduct(lhs._internal,rhs._internal)) ret_t;
-    iScalar<ret_t> ret;
-    ret._internal = innerProduct(lhs._internal,rhs._internal);
-    return ret;
-  }
-
+  return ret;
 }
+template<class l,class r> accelerator_inline
+auto innerProduct (const iScalar<l>& lhs,const iScalar<r>& rhs) -> iScalar<decltype(innerProduct(lhs._internal,rhs._internal))>
+{
+  typedef decltype(innerProduct(lhs._internal,rhs._internal)) ret_t;
+  iScalar<ret_t> ret;
+  ret._internal = innerProduct(lhs._internal,rhs._internal);
+  return ret;
+}
+
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/tensors/Tensor_logical.h b/Grid/tensors/Tensor_logical.h
index 58b2b03b..e45a8464 100644
--- a/Grid/tensors/Tensor_logical.h
+++ b/Grid/tensors/Tensor_logical.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,32 +23,31 @@ Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef GRID_TENSOR_LOGICAL_H
-#define GRID_TENSOR_LOGICAL_H
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
-#define LOGICAL_BINOP(Op)\
-template<class v> strong_inline iScalar<v> operator Op (const iScalar<v>& lhs,const iScalar<v>& rhs) \
-{\
-  iScalar<v> ret;\
-  ret._internal = lhs._internal Op rhs._internal ;\
-  return ret;\
-}\
-template<class l> strong_inline iScalar<l> operator Op (const iScalar<l>& lhs,Integer rhs) \
-{\
-  typename iScalar<l>::scalar_type t; t=rhs;\
-  typename iScalar<l>::tensor_reduced srhs; srhs=t;\
-  return lhs Op srhs;\
-}\
-template<class l> strong_inline iScalar<l> operator Op (Integer lhs,const iScalar<l>& rhs) \
-{\
-  typename iScalar<l>::scalar_type t;t=lhs;\
-  typename iScalar<l>::tensor_reduced slhs;slhs=t;\
-  return slhs Op rhs;\
-}
+#define LOGICAL_BINOP(Op)						\
+  template<class v> accelerator_inline iScalar<v> operator Op (const iScalar<v>& lhs,const iScalar<v>& rhs) \
+  {									\
+    iScalar<v> ret;							\
+    ret._internal = lhs._internal Op rhs._internal ;			\
+    return ret;								\
+  }									\
+  template<class l> accelerator_inline iScalar<l> operator Op (const iScalar<l>& lhs,Integer rhs) \
+  {									\
+    typename iScalar<l>::scalar_type t; t=rhs;				\
+    typename iScalar<l>::tensor_reduced srhs; srhs=t;			\
+    return lhs Op srhs;							\
+  }									\
+  template<class l> accelerator_inline iScalar<l> operator Op (Integer lhs,const iScalar<l>& rhs) \
+  {									\
+    typename iScalar<l>::scalar_type t;t=lhs;				\
+    typename iScalar<l>::tensor_reduced slhs;slhs=t;			\
+    return slhs Op rhs;							\
+  }
 
 LOGICAL_BINOP(|);
 LOGICAL_BINOP(&);
@@ -56,13 +55,13 @@ LOGICAL_BINOP(||);
 LOGICAL_BINOP(&&);
 
 template <class T>
-strong_inline bool operator==(const iScalar<T> &t1, const iScalar<T> &t2)
+accelerator_inline bool operator==(const iScalar<T> &t1, const iScalar<T> &t2)
 {
   return (t1._internal == t2._internal);
 }
 
 template <class T, int N>
-strong_inline bool operator==(const iVector<T, N> &t1, const iVector<T, N> &t2)
+accelerator_inline bool operator==(const iVector<T, N> &t1, const iVector<T, N> &t2)
 {
   bool res = true;
 
@@ -75,7 +74,7 @@ strong_inline bool operator==(const iVector<T, N> &t1, const iVector<T, N> &t2)
 }
 
 template <class T, int N>
-strong_inline bool operator==(const iMatrix<T, N> &t1, const iMatrix<T, N> &t2)
+accelerator_inline bool operator==(const iMatrix<T, N> &t1, const iMatrix<T, N> &t2)
 {
   bool res = true;
 
@@ -88,5 +87,6 @@ strong_inline bool operator==(const iMatrix<T, N> &t1, const iMatrix<T, N> &t2)
   return res;
 }
 
-}
-#endif
+NAMESPACE_END(Grid);
+
+
diff --git a/Grid/tensors/Tensor_outer.h b/Grid/tensors/Tensor_outer.h
index 6429a190..4902c22f 100644
--- a/Grid/tensors/Tensor_outer.h
+++ b/Grid/tensors/Tensor_outer.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,57 +23,51 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_OUTER_H
 #define GRID_MATH_OUTER_H
-namespace Grid {
-    ///////////////////////////////////////////////////////////////////////////////////////
-    // outerProduct Scalar x Scalar -> Scalar
-    //              Vector x Vector -> Matrix
-    ///////////////////////////////////////////////////////////////////////////////////////
 
-template<class l,class r,int N> inline
+NAMESPACE_BEGIN(Grid);
+
+///////////////////////////////////////////////////////////////////////////////////////
+// outerProduct Scalar x Scalar -> Scalar
+//              Vector x Vector -> Matrix
+///////////////////////////////////////////////////////////////////////////////////////
+
+template<class l,class r,int N> accelerator_inline
 auto outerProduct (const iVector<l,N>& lhs,const iVector<r,N>& rhs) -> iMatrix<decltype(outerProduct(lhs._internal[0],rhs._internal[0])),N>
 {
-    typedef decltype(outerProduct(lhs._internal[0],rhs._internal[0])) ret_t;
-    iMatrix<ret_t,N> ret;
-    for(int c1=0;c1<N;c1++){
+  typedef decltype(outerProduct(lhs._internal[0],rhs._internal[0])) ret_t;
+  iMatrix<ret_t,N> ret;
+  for(int c1=0;c1<N;c1++){
     for(int c2=0;c2<N;c2++){
-        ret._internal[c1][c2] = outerProduct(lhs._internal[c1],rhs._internal[c2]);
-    }}
-    return ret;
+      ret._internal[c1][c2] = outerProduct(lhs._internal[c1],rhs._internal[c2]);
+  }}
+  return ret;
 }
 
 
-template<class l,class r> inline
+template<class l,class r> accelerator_inline
 auto outerProduct (const iScalar<l>& lhs,const iScalar<r>& rhs) -> iScalar<decltype(outerProduct(lhs._internal,rhs._internal))>
 {
-    typedef decltype(outerProduct(lhs._internal,rhs._internal)) ret_t;
-    iScalar<ret_t> ret;
-    ret._internal = outerProduct(lhs._internal,rhs._internal);
-    return ret;
+  typedef decltype(outerProduct(lhs._internal,rhs._internal)) ret_t;
+  iScalar<ret_t> ret;
+  ret._internal = outerProduct(lhs._internal,rhs._internal);
+  return ret;
 }
 
-  
-inline ComplexF outerProduct(const ComplexF &l, const ComplexF& r)
+template<class CC,IfComplex<CC> = 0>
+accelerator_inline CC outerProduct(const CC &l, const CC& r)
 {
-  std::cout << "outer product taking conj "<<r<<" "<<conj(r)<<std::endl;
   return l*conj(r);
 }
-inline ComplexD outerProduct(const ComplexD &l, const ComplexD& r)
-{
-  std::cout << "outer product taking conj "<<r<<" "<<conj(r)<<std::endl;
-  return l*conj(r);
-}
-inline RealF outerProduct(const RealF &l, const RealF& r)
-{
-  return l*r;
-}
-inline RealD outerProduct(const RealD &l, const RealD& r)
+template<class RR,IfReal<RR> = 0>
+accelerator_inline RR outerProduct(const RR &l, const RR& r)
 {
   return l*r;
 }
 
-}
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/tensors/Tensor_reality.h b/Grid/tensors/Tensor_reality.h
index 8a183b88..ca1d52ca 100644
--- a/Grid/tensors/Tensor_reality.h
+++ b/Grid/tensors/Tensor_reality.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,22 +24,23 @@ Author: neo <cossu@post.kek.jp>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_REALITY_H
 #define GRID_MATH_REALITY_H
-namespace Grid {
+
+NAMESPACE_BEGIN(Grid);
 
 /////////////////////////////////////////////// 
 // multiply by I; make recursive.
 /////////////////////////////////////////////// 
-template<class vtype> inline iScalar<vtype> timesI(const iScalar<vtype>&r) 
+template<class vtype> accelerator_inline iScalar<vtype> timesI(const iScalar<vtype>&r) 
 {
-    iScalar<vtype> ret;
-    timesI(ret._internal,r._internal);
-    return ret;
+  iScalar<vtype> ret;
+  timesI(ret._internal,r._internal);
+  return ret;
 }
-template<class vtype,int N> inline iVector<vtype,N> timesI(const iVector<vtype,N>&r) 
+template<class vtype,int N> accelerator_inline iVector<vtype,N> timesI(const iVector<vtype,N>&r) 
 {
   iVector<vtype,N> ret;
   for(int i=0;i<N;i++){
@@ -47,42 +48,42 @@ template<class vtype,int N> inline iVector<vtype,N> timesI(const iVector<vtype,N
   }
   return ret;
 }
-template<class vtype,int N> inline iMatrix<vtype,N> timesI(const iMatrix<vtype,N>&r)
+template<class vtype,int N> accelerator_inline iMatrix<vtype,N> timesI(const iMatrix<vtype,N>&r)
 {
   iMatrix<vtype,N> ret;
   for(int i=0;i<N;i++){
-  for(int j=0;j<N;j++){
-    timesI(ret._internal[i][j],r._internal[i][j]);
-  }}
+    for(int j=0;j<N;j++){
+      timesI(ret._internal[i][j],r._internal[i][j]);
+    }}
   return ret;
 }
 
-template<class vtype> inline void timesI(iScalar<vtype> &ret,const iScalar<vtype>&r) 
+template<class vtype> accelerator_inline void timesI(iScalar<vtype> &ret,const iScalar<vtype>&r) 
 {
   timesI(ret._internal,r._internal);
 }
-template<class vtype,int N> inline void timesI(iVector<vtype,N> &ret,const iVector<vtype,N>&r) 
+template<class vtype,int N> accelerator_inline void timesI(iVector<vtype,N> &ret,const iVector<vtype,N>&r) 
 {
   for(int i=0;i<N;i++){
     timesI(ret._internal[i],r._internal[i]);
   }
 }
-template<class vtype,int N> inline void  timesI(iMatrix<vtype,N> &ret,const iMatrix<vtype,N>&r)
+template<class vtype,int N> accelerator_inline void  timesI(iMatrix<vtype,N> &ret,const iMatrix<vtype,N>&r)
 {
   for(int i=0;i<N;i++){
-  for(int j=0;j<N;j++){
-    timesI(ret._internal[i][j],r._internal[i][j]);
-  }}
+    for(int j=0;j<N;j++){
+      timesI(ret._internal[i][j],r._internal[i][j]);
+    }}
 }
 
 
-template<class vtype> inline iScalar<vtype> timesMinusI(const iScalar<vtype>&r) 
+template<class vtype> accelerator_inline iScalar<vtype> timesMinusI(const iScalar<vtype>&r) 
 {
-    iScalar<vtype> ret;
-    timesMinusI(ret._internal,r._internal);
-    return ret;
+  iScalar<vtype> ret;
+  timesMinusI(ret._internal,r._internal);
+  return ret;
 }
-template<class vtype,int N> inline iVector<vtype,N> timesMinusI(const iVector<vtype,N>&r) 
+template<class vtype,int N> accelerator_inline iVector<vtype,N> timesMinusI(const iVector<vtype,N>&r) 
 {
   iVector<vtype,N> ret;
   for(int i=0;i<N;i++){
@@ -90,45 +91,45 @@ template<class vtype,int N> inline iVector<vtype,N> timesMinusI(const iVector<vt
   }
   return ret;
 }
-template<class vtype,int N> inline iMatrix<vtype,N> timesMinusI(const iMatrix<vtype,N>&r)
+template<class vtype,int N> accelerator_inline iMatrix<vtype,N> timesMinusI(const iMatrix<vtype,N>&r)
 {
   iMatrix<vtype,N> ret;
   for(int i=0;i<N;i++){
-  for(int j=0;j<N;j++){
-    timesMinusI(ret._internal[i][j],r._internal[i][j]);
-  }}
+    for(int j=0;j<N;j++){
+      timesMinusI(ret._internal[i][j],r._internal[i][j]);
+    }}
   return ret;
 }
 
-template<class vtype>  inline void timesMinusI(iScalar<vtype> &ret,const iScalar<vtype>&r) 
+template<class vtype>  accelerator_inline void timesMinusI(iScalar<vtype> &ret,const iScalar<vtype>&r) 
 {
   timesMinusI(ret._internal,r._internal);
 }
-template<class vtype,int N> inline void timesMinusI(iVector<vtype,N> &ret,const iVector<vtype,N>&r) 
+template<class vtype,int N> accelerator_inline void timesMinusI(iVector<vtype,N> &ret,const iVector<vtype,N>&r) 
 {
   for(int i=0;i<N;i++){
     timesMinusI(ret._internal[i],r._internal[i]);
   }
 }
-template<class vtype,int N> inline void  timesMinusI(iMatrix<vtype,N> &ret,const iMatrix<vtype,N>&r)
+template<class vtype,int N> accelerator_inline void  timesMinusI(iMatrix<vtype,N> &ret,const iMatrix<vtype,N>&r)
 {
   for(int i=0;i<N;i++){
-  for(int j=0;j<N;j++){
-    timesMinusI(ret._internal[i][j],r._internal[i][j]);
-  }}
+    for(int j=0;j<N;j++){
+      timesMinusI(ret._internal[i][j],r._internal[i][j]);
+    }}
 }
 
 
 /////////////////////////////////////////////// 
 // Conj function for scalar, vector, matrix
 /////////////////////////////////////////////// 
-template<class vtype> inline iScalar<vtype> conjugate(const iScalar<vtype>&r)
+template<class vtype> accelerator_inline iScalar<vtype> conjugate(const iScalar<vtype>&r)
 {
-    iScalar<vtype> ret;
-    ret._internal = conjugate(r._internal);
-    return ret;
+  iScalar<vtype> ret;
+  ret._internal = conjugate(r._internal);
+  return ret;
 }
-template<class vtype,int N> inline iVector<vtype,N> conjugate(const iVector<vtype,N>&r)
+template<class vtype,int N> accelerator_inline iVector<vtype,N> conjugate(const iVector<vtype,N>&r)
 {
   iVector<vtype,N> ret;
   for(int i=0;i<N;i++){
@@ -136,41 +137,41 @@ template<class vtype,int N> inline iVector<vtype,N> conjugate(const iVector<vtyp
   }
   return ret;
 }
-template<class vtype,int N> inline iMatrix<vtype,N> conjugate(const iMatrix<vtype,N>&r)
+template<class vtype,int N> accelerator_inline iMatrix<vtype,N> conjugate(const iMatrix<vtype,N>&r)
 {
   iMatrix<vtype,N> ret;
   for(int i=0;i<N;i++){
-  for(int j=0;j<N;j++){
-    ret._internal[i][j] = conjugate(r._internal[i][j]);
-  }}
+    for(int j=0;j<N;j++){
+      ret._internal[i][j] = conjugate(r._internal[i][j]);
+    }}
   return ret;
 }
 
 /////////////////////////////////////////////// 
 // Adj function for scalar, vector, matrix
 /////////////////////////////////////////////// 
-template<class vtype> inline iScalar<vtype> adj(const iScalar<vtype>&r)
+template<class vtype> accelerator_inline iScalar<vtype> adj(const iScalar<vtype>&r)
 {
-    iScalar<vtype> ret;
-    ret._internal = adj(r._internal);
-    return ret;
+  iScalar<vtype> ret;
+  ret._internal = adj(r._internal);
+  return ret;
 }
-template<class vtype,int N> inline iVector<vtype,N> adj(const iVector<vtype,N>&r)
+template<class vtype,int N> accelerator_inline iVector<vtype,N> adj(const iVector<vtype,N>&r)
 {
-    iVector<vtype,N> ret;
-    for(int i=0;i<N;i++){
-        ret._internal[i] = adj(r._internal[i]);
-    }
-    return ret;
+  iVector<vtype,N> ret;
+  for(int i=0;i<N;i++){
+    ret._internal[i] = adj(r._internal[i]);
+  }
+  return ret;
 }
-template<class vtype,int N> inline iMatrix<vtype,N> adj(const iMatrix<vtype,N> &arg)
+template<class vtype,int N> accelerator_inline iMatrix<vtype,N> adj(const iMatrix<vtype,N> &arg)
 {
-    iMatrix<vtype,N> ret;
-    for(int c1=0;c1<N;c1++){
+  iMatrix<vtype,N> ret;
+  for(int c1=0;c1<N;c1++){
     for(int c2=0;c2<N;c2++){
-        ret._internal[c1][c2]=adj(arg._internal[c2][c1]);
+      ret._internal[c1][c2]=adj(arg._internal[c2][c1]);
     }}
-    return ret;
+  return ret;
 }
 
 
@@ -182,54 +183,54 @@ template<class vtype,int N> inline iMatrix<vtype,N> adj(const iMatrix<vtype,N> &
 // Can only take the real/imag part of scalar objects, since
 // lattice objects of different complex nature are non-conformable.
 /////////////////////////////////////////////////////////////////
-template<class itype> inline auto real(const iScalar<itype> &z) -> iScalar<decltype(real(z._internal))>
+template<class itype> accelerator_inline auto real(const iScalar<itype> &z) -> iScalar<decltype(real(z._internal))>
 {
-    iScalar<decltype(real(z._internal))> ret;
-    ret._internal = real(z._internal);
-    return ret;
+  iScalar<decltype(real(z._internal))> ret;
+  ret._internal = real(z._internal);
+  return ret;
 }
-template<class itype,int N> inline auto real(const iMatrix<itype,N> &z) -> iMatrix<decltype(real(z._internal[0][0])),N>
+template<class itype,int N> accelerator_inline auto real(const iMatrix<itype,N> &z) -> iMatrix<decltype(real(z._internal[0][0])),N>
 {
-    iMatrix<decltype(real(z._internal[0][0])),N> ret;
-    for(int c1=0;c1<N;c1++){
+  iMatrix<decltype(real(z._internal[0][0])),N> ret;
+  for(int c1=0;c1<N;c1++){
     for(int c2=0;c2<N;c2++){
-        ret._internal[c1][c2] = real(z._internal[c1][c2]);
+      ret._internal[c1][c2] = real(z._internal[c1][c2]);
     }}
-    return ret;
+  return ret;
 }
-template<class itype,int N> inline auto real(const iVector<itype,N> &z) -> iVector<decltype(real(z._internal[0])),N>
+template<class itype,int N> accelerator_inline auto real(const iVector<itype,N> &z) -> iVector<decltype(real(z._internal[0])),N>
 {
-    iVector<decltype(real(z._internal[0])),N> ret;
-    for(int c1=0;c1<N;c1++){
-        ret._internal[c1] = real(z._internal[c1]);
-    }
-    return ret;
+  iVector<decltype(real(z._internal[0])),N> ret;
+  for(int c1=0;c1<N;c1++){
+    ret._internal[c1] = real(z._internal[c1]);
+  }
+  return ret;
 }
     
-template<class itype> inline auto imag(const iScalar<itype> &z) -> iScalar<decltype(imag(z._internal))>
+template<class itype> accelerator_inline auto imag(const iScalar<itype> &z) -> iScalar<decltype(imag(z._internal))>
 {
-    iScalar<decltype(imag(z._internal))> ret;
-    ret._internal = imag(z._internal);
-    return ret;
+  iScalar<decltype(imag(z._internal))> ret;
+  ret._internal = imag(z._internal);
+  return ret;
 }
-template<class itype,int N> inline auto imag(const iMatrix<itype,N> &z) -> iMatrix<decltype(imag(z._internal[0][0])),N>
+template<class itype,int N> accelerator_inline auto imag(const iMatrix<itype,N> &z) -> iMatrix<decltype(imag(z._internal[0][0])),N>
 {
-    iMatrix<decltype(imag(z._internal[0][0])),N> ret;
-    for(int c1=0;c1<N;c1++){
+  iMatrix<decltype(imag(z._internal[0][0])),N> ret;
+  for(int c1=0;c1<N;c1++){
     for(int c2=0;c2<N;c2++){
-        ret._internal[c1][c2] = imag(z._internal[c1][c2]);
+      ret._internal[c1][c2] = imag(z._internal[c1][c2]);
     }}
-    return ret;
+  return ret;
 }
-template<class itype,int N> inline auto imag(const iVector<itype,N> &z) -> iVector<decltype(imag(z._internal[0])),N>
+template<class itype,int N> accelerator_inline auto imag(const iVector<itype,N> &z) -> iVector<decltype(imag(z._internal[0])),N>
 {
-    iVector<decltype(imag(z._internal[0])),N> ret;
-    for(int c1=0;c1<N;c1++){
-        ret._internal[c1] = imag(z._internal[c1]);
-    }
-    return ret;
+  iVector<decltype(imag(z._internal[0])),N> ret;
+  for(int c1=0;c1<N;c1++){
+    ret._internal[c1] = imag(z._internal[c1]);
+  }
+  return ret;
 }
 
+NAMESPACE_END(Grid);
 
-}
 #endif
diff --git a/Grid/tensors/Tensor_trace.h b/Grid/tensors/Tensor_trace.h
index 1bab2a4e..6aa398f9 100644
--- a/Grid/tensors/Tensor_trace.h
+++ b/Grid/tensors/Tensor_trace.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,43 +24,44 @@ Author: neo <cossu@post.kek.jp>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_TRACE_H
 #define GRID_MATH_TRACE_H
-namespace Grid {
+
+NAMESPACE_BEGIN(Grid);
 
 //////////////////////////////////////////////////////////////////
 // Traces: both all indices and a specific index. Indices must be
 // either scalar or matrix
 /////////////////////////////////////////////////////////////////
 
-inline ComplexF trace( const ComplexF &arg){    return arg;}
-inline ComplexD trace( const ComplexD &arg){    return arg;}
-inline RealF trace( const RealF &arg){    return arg;}
-inline RealD trace( const RealD &arg){    return arg;}
+accelerator_inline ComplexF trace( const ComplexF &arg){    return arg;}
+accelerator_inline ComplexD trace( const ComplexD &arg){    return arg;}
+accelerator_inline RealF trace( const RealF &arg){    return arg;}
+accelerator_inline RealD trace( const RealD &arg){    return arg;}
 
 template<class vtype,int N>
-inline auto trace(const iMatrix<vtype,N> &arg) -> iScalar<decltype(trace(arg._internal[0][0]))>
+accelerator_inline auto trace(const iMatrix<vtype,N> &arg) -> iScalar<decltype(trace(arg._internal[0][0]))>
 {
-    iScalar<decltype( trace(arg._internal[0][0] )) > ret;
-    zeroit(ret._internal);
-    for(int i=0;i<N;i++){
-        ret._internal=ret._internal+trace(arg._internal[i][i]);
-    }
-    return ret;
+  iScalar<decltype( trace(arg._internal[0][0] )) > ret;
+  zeroit(ret._internal);
+  for(int i=0;i<N;i++){
+    ret._internal=ret._internal+trace(arg._internal[i][i]);
+  }
+  return ret;
 }
 
 template<class vtype>
-inline auto trace(const iScalar<vtype> &arg) -> iScalar<decltype(trace(arg._internal))>
+accelerator_inline auto trace(const iScalar<vtype> &arg) -> iScalar<decltype(trace(arg._internal))>
 {
-    iScalar<decltype(trace(arg._internal))> ret;
-    ret._internal=trace(arg._internal);
-    return ret;
+  iScalar<decltype(trace(arg._internal))> ret;
+  ret._internal=trace(arg._internal);
+  return ret;
 }
 
 template<class vtype,int N>
-  inline auto trace(const iVector<vtype,N> &arg) -> iVector<decltype(trace(arg._internal[0])),N>
+accelerator_inline auto trace(const iVector<vtype,N> &arg) -> iVector<decltype(trace(arg._internal[0])),N>
 {
   iVector<decltype(trace(arg._internal[0])),N> ret;
   for(int i=0;i<N;i++){
@@ -69,6 +70,6 @@ template<class vtype,int N>
   return ret;
 }
 
+NAMESPACE_END(Grid);
 
-}
 #endif
diff --git a/Grid/tensors/Tensor_traits.h b/Grid/tensors/Tensor_traits.h
index 45a1d807..9067d43d 100644
--- a/Grid/tensors/Tensor_traits.h
+++ b/Grid/tensors/Tensor_traits.h
@@ -20,12 +20,11 @@ Author: Michael Marshall <michael.marshall@ed.ac.au>
     See the full license in the file "LICENSE" in the top level distribution directory
     *************************************************************************************/
     /*  END LEGAL */
-#ifndef GRID_MATH_TRAITS_H
-#define GRID_MATH_TRAITS_H
+#pragma once 
 
 #include <type_traits>
 
-namespace Grid {
+NAMESPACE_BEGIN(Grid);
 
   // Forward declarations
   template<class T>        class iScalar;
@@ -73,50 +72,60 @@ namespace Grid {
 
   template<> struct GridTypeMapper<RealF> : public GridTypeMapper_Base {
     typedef RealF scalar_type;
+    typedef RealD scalar_typeD;
     typedef RealF vector_type;
     typedef RealD vector_typeD;
     typedef RealF tensor_reduced ;
     typedef RealF scalar_object;
+    typedef RealD scalar_objectD;
     typedef ComplexF Complexified;
     typedef RealF Realified;
     typedef RealD DoublePrecision;
   };
   template<> struct GridTypeMapper<RealD> : public GridTypeMapper_Base {
     typedef RealD scalar_type;
+    typedef RealD scalar_typeD;
     typedef RealD vector_type;
     typedef RealD vector_typeD;
     typedef RealD tensor_reduced;
     typedef RealD scalar_object;
+    typedef RealD scalar_objectD;
     typedef ComplexD Complexified;
     typedef RealD Realified;
     typedef RealD DoublePrecision;
   };
   template<> struct GridTypeMapper<ComplexF> : public GridTypeMapper_Base {
     typedef ComplexF scalar_type;
+    typedef ComplexD scalar_typeD;
     typedef ComplexF vector_type;
     typedef ComplexD vector_typeD;
     typedef ComplexF tensor_reduced;
     typedef ComplexF scalar_object;
+    typedef ComplexD scalar_objectD;
     typedef ComplexF Complexified;
     typedef RealF Realified;
     typedef ComplexD DoublePrecision;
   };
   template<> struct GridTypeMapper<ComplexD> : public GridTypeMapper_Base {
     typedef ComplexD scalar_type;
+    typedef ComplexD scalar_typeD;
     typedef ComplexD vector_type;
     typedef ComplexD vector_typeD;
     typedef ComplexD tensor_reduced;
     typedef ComplexD scalar_object;
+    typedef ComplexD scalar_objectD;
     typedef ComplexD Complexified;
     typedef RealD Realified;
     typedef ComplexD DoublePrecision;
   };
   template<> struct GridTypeMapper<Integer> : public GridTypeMapper_Base {
     typedef Integer scalar_type;
+    typedef Integer scalar_typeD;
     typedef Integer vector_type;
     typedef Integer vector_typeD;
     typedef Integer tensor_reduced;
     typedef Integer scalar_object;
+    typedef Integer scalar_objectD;
     typedef void Complexified;
     typedef void Realified;
     typedef void DoublePrecision;
@@ -124,20 +133,24 @@ namespace Grid {
 
   template<> struct GridTypeMapper<vRealF> : public GridTypeMapper_Base {
     typedef RealF  scalar_type;
+    typedef RealD  scalar_typeD;
     typedef vRealF vector_type;
     typedef vRealD vector_typeD;
     typedef vRealF tensor_reduced;
     typedef RealF  scalar_object;
+    typedef RealD  scalar_objectD;
     typedef vComplexF Complexified;
     typedef vRealF Realified;
     typedef vRealD DoublePrecision;
   };
   template<> struct GridTypeMapper<vRealD> : public GridTypeMapper_Base {
     typedef RealD  scalar_type;
+    typedef RealD  scalar_typeD;
     typedef vRealD vector_type;
     typedef vRealD vector_typeD;
     typedef vRealD tensor_reduced;
     typedef RealD  scalar_object;
+    typedef RealD  scalar_objectD;
     typedef vComplexD Complexified;
     typedef vRealD Realified;
     typedef vRealD DoublePrecision;
@@ -145,10 +158,12 @@ namespace Grid {
   template<> struct GridTypeMapper<vRealH> : public GridTypeMapper_Base {
     // Fixme this is incomplete until Grid supports fp16 or bfp16 arithmetic types
     typedef RealF  scalar_type;
+    typedef RealD  scalar_typeD;
     typedef vRealH vector_type;
     typedef vRealD vector_typeD;
     typedef vRealH tensor_reduced;
     typedef RealF  scalar_object;
+    typedef RealD  scalar_objectD;
     typedef vComplexH Complexified;
     typedef vRealH Realified;
     typedef vRealD DoublePrecision;
@@ -156,40 +171,48 @@ namespace Grid {
   template<> struct GridTypeMapper<vComplexH> : public GridTypeMapper_Base {
     // Fixme this is incomplete until Grid supports fp16 or bfp16 arithmetic types
     typedef ComplexF  scalar_type;
+    typedef ComplexD  scalar_typeD;
     typedef vComplexH vector_type;
     typedef vComplexD vector_typeD;
     typedef vComplexH tensor_reduced;
     typedef ComplexF  scalar_object;
+    typedef ComplexD  scalar_objectD;
     typedef vComplexH Complexified;
     typedef vRealH Realified;
     typedef vComplexD DoublePrecision;
   };
   template<> struct GridTypeMapper<vComplexF> : public GridTypeMapper_Base {
     typedef ComplexF  scalar_type;
+    typedef ComplexD  scalar_typeD;
     typedef vComplexF vector_type;
     typedef vComplexD vector_typeD;
     typedef vComplexF tensor_reduced;
     typedef ComplexF  scalar_object;
+    typedef ComplexD  scalar_objectD;
     typedef vComplexF Complexified;
     typedef vRealF Realified;
     typedef vComplexD DoublePrecision;
   };
   template<> struct GridTypeMapper<vComplexD> : public GridTypeMapper_Base {
     typedef ComplexD  scalar_type;
+    typedef ComplexD  scalar_typeD;
     typedef vComplexD vector_type;
     typedef vComplexD vector_typeD;
     typedef vComplexD tensor_reduced;
     typedef ComplexD  scalar_object;
+    typedef ComplexD  scalar_objectD;
     typedef vComplexD Complexified;
     typedef vRealD Realified;
     typedef vComplexD DoublePrecision;
   };
   template<> struct GridTypeMapper<vInteger> : public GridTypeMapper_Base {
     typedef  Integer scalar_type;
+    typedef  Integer scalar_typeD;
     typedef vInteger vector_type;
     typedef vInteger vector_typeD;
     typedef vInteger tensor_reduced;
     typedef  Integer scalar_object;
+    typedef  Integer scalar_objectD;
     typedef void Complexified;
     typedef void Realified;
     typedef void DoublePrecision;
@@ -199,6 +222,7 @@ namespace Grid {
   using BaseTraits   = GridTypeMapper<T>; \
   using scalar_type  = typename BaseTraits::scalar_type; \
   using vector_type  = typename BaseTraits::vector_type; \
+  using scalar_typeD = typename BaseTraits::scalar_typeD; \
   using vector_typeD = typename BaseTraits::vector_typeD; \
   static constexpr int TensorLevel = BaseTraits::TensorLevel + 1
 
@@ -206,6 +230,7 @@ namespace Grid {
     GridTypeMapper_RepeatedTypes;
     using tensor_reduced  = iScalar<typename BaseTraits::tensor_reduced>;
     using scalar_object   = iScalar<typename BaseTraits::scalar_object>;
+    using scalar_objectD  = iScalar<typename BaseTraits::scalar_objectD>;
     using Complexified    = iScalar<typename BaseTraits::Complexified>;
     using Realified       = iScalar<typename BaseTraits::Realified>;
     using DoublePrecision = iScalar<typename BaseTraits::DoublePrecision>;
@@ -219,6 +244,7 @@ namespace Grid {
     GridTypeMapper_RepeatedTypes;
     using tensor_reduced  = iScalar<typename BaseTraits::tensor_reduced>;
     using scalar_object   = iVector<typename BaseTraits::scalar_object,   N>;
+    using scalar_objectD  = iVector<typename BaseTraits::scalar_objectD,  N>;
     using Complexified    = iVector<typename BaseTraits::Complexified,    N>;
     using Realified       = iVector<typename BaseTraits::Realified,       N>;
     using DoublePrecision = iVector<typename BaseTraits::DoublePrecision, N>;
@@ -232,6 +258,7 @@ namespace Grid {
     GridTypeMapper_RepeatedTypes;
     using tensor_reduced  = iScalar<typename BaseTraits::tensor_reduced>;
     using scalar_object   = iMatrix<typename BaseTraits::scalar_object,   N>;
+    using scalar_objectD  = iMatrix<typename BaseTraits::scalar_objectD,  N>;
     using Complexified    = iMatrix<typename BaseTraits::Complexified,    N>;
     using Realified       = iMatrix<typename BaseTraits::Realified,       N>;
     using DoublePrecision = iMatrix<typename BaseTraits::DoublePrecision, N>;
@@ -281,7 +308,5 @@ namespace Grid {
 
     enum { value = sizeof(real_scalar_type)/sizeof(float) };
   };
-}
-
-#endif
+NAMESPACE_END(Grid);
 
diff --git a/Grid/tensors/Tensor_transpose.h b/Grid/tensors/Tensor_transpose.h
index c12d4f09..3fb26335 100644
--- a/Grid/tensors/Tensor_transpose.h
+++ b/Grid/tensors/Tensor_transpose.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -23,63 +23,62 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_TRANSPOSE_H
 #define GRID_MATH_TRANSPOSE_H
-namespace Grid {
-
 
+NAMESPACE_BEGIN(Grid);
 
 /////////////////////////////////////////////////////////////////
 // Transpose all indices
 /////////////////////////////////////////////////////////////////
 
-inline ComplexD transpose(ComplexD &rhs){  return rhs;}
-inline ComplexF transpose(ComplexF &rhs){  return rhs;}
-inline RealD transpose(RealD &rhs){  return rhs;}
-inline RealF transpose(RealF &rhs){  return rhs;}
+accelerator_inline ComplexD transpose(ComplexD &rhs){  return rhs;}
+accelerator_inline ComplexF transpose(ComplexF &rhs){  return rhs;}
+accelerator_inline RealD transpose(RealD &rhs){  return rhs;}
+accelerator_inline RealF transpose(RealF &rhs){  return rhs;}
 
 template<class vtype,int N>
-  inline typename std::enable_if<isGridTensor<vtype>::value, iMatrix<vtype,N> >::type 
-  transpose(iMatrix<vtype,N> arg)
-  {
-    iMatrix<vtype,N> ret;
-    for(int i=0;i<N;i++){
-      for(int j=0;j<N;j++){
-	ret._internal[i][j] = transpose(arg._internal[j][i]); // NB recurses
-      }}
-    return ret;
-  }
+accelerator_inline typename std::enable_if<isGridTensor<vtype>::value, iMatrix<vtype,N> >::type 
+transpose(iMatrix<vtype,N> arg)
+{
+  iMatrix<vtype,N> ret;
+  for(int i=0;i<N;i++){
+    for(int j=0;j<N;j++){
+      ret._internal[i][j] = transpose(arg._internal[j][i]); // NB recurses
+    }}
+  return ret;
+}
 template<class vtype,int N>
-  inline typename std::enable_if<isGridTensor<vtype>::notvalue, iMatrix<vtype,N> >::type 
-  transpose(iMatrix<vtype,N> arg)
-  {
-    iMatrix<vtype,N> ret;
-    for(int i=0;i<N;i++){
-      for(int j=0;j<N;j++){
-	ret._internal[i][j] = arg._internal[j][i]; // Stop recursion if not a tensor type
-      }}
-    return ret;
-  }
+accelerator_inline typename std::enable_if<isGridTensor<vtype>::notvalue, iMatrix<vtype,N> >::type 
+transpose(iMatrix<vtype,N> arg)
+{
+  iMatrix<vtype,N> ret;
+  for(int i=0;i<N;i++){
+    for(int j=0;j<N;j++){
+      ret._internal[i][j] = arg._internal[j][i]; // Stop recursion if not a tensor type
+    }}
+  return ret;
+}
 
 template<class vtype>
-  inline typename std::enable_if<isGridTensor<vtype>::value, iScalar<vtype> >::type 
-  transpose(iScalar<vtype> arg)
-  {
-    iScalar<vtype> ret;
-    ret._internal = transpose(arg._internal); // NB recurses
-    return ret;
-  }
+accelerator_inline typename std::enable_if<isGridTensor<vtype>::value, iScalar<vtype> >::type 
+transpose(iScalar<vtype> arg)
+{
+  iScalar<vtype> ret;
+  ret._internal = transpose(arg._internal); // NB recurses
+  return ret;
+}
 
 template<class vtype>
-  inline typename std::enable_if<isGridTensor<vtype>::notvalue, iScalar<vtype> >::type 
-  transpose(iScalar<vtype> arg)
-  {
-    iScalar<vtype> ret;
-    ret._internal = arg._internal; // NB recursion stops
-    return ret;
-  }
+accelerator_inline typename std::enable_if<isGridTensor<vtype>::notvalue, iScalar<vtype> >::type 
+transpose(iScalar<vtype> arg)
+{
+  iScalar<vtype> ret;
+  ret._internal = arg._internal; // NB recursion stops
+  return ret;
+}
 
 
 ////////////////////////////////////////////////////////////////////////////////////////////
@@ -87,19 +86,19 @@ template<class vtype>
 // to that of adj; which is easiers?
 ////////////////////////////////////////////////////////////////////////////////////////////
 #if 0
-template<int Level,class vtype,int N> inline 
-  typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,Level>::value, iMatrix<vtype,N> >::type 
+template<int Level,class vtype,int N> accelerator_inline 
+typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,Level>::value, iMatrix<vtype,N> >::type 
 transposeIndex (const iMatrix<vtype,N> &arg)
 {
   iMatrix<vtype,N> ret;
   for(int i=0;i<N;i++){
     for(int j=0;j<N;j++){
       ret._internal[i][j] = arg._internal[j][i]; 
-  }}
+    }}
   return ret;
 }
 // or not
-template<int Level,class vtype,int N> inline 
+template<int Level,class vtype,int N> accelerator_inline 
 typename std::enable_if<matchGridTensorIndex<iMatrix<vtype,N>,Level>::notvalue, iMatrix<vtype,N> >::type 
 transposeIndex (const iMatrix<vtype,N> &arg)
 {
@@ -107,10 +106,10 @@ transposeIndex (const iMatrix<vtype,N> &arg)
   for(int i=0;i<N;i++){
     for(int j=0;j<N;j++){
       ret._internal[i][j] = transposeIndex<Level>(arg._internal[i][j]); 
-  }}
+    }}
   return ret;
 }
-template<int Level,class vtype> inline 
+template<int Level,class vtype> accelerator_inline 
 typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,Level>::notvalue, iScalar<vtype> >::type 
 transposeIndex (const iScalar<vtype> &arg)
 {
@@ -118,7 +117,7 @@ transposeIndex (const iScalar<vtype> &arg)
   ret._internal=transposeIndex<Level>(arg._internal);
   return ret;
 }
-template<int Level,class vtype> inline 
+template<int Level,class vtype> accelerator_inline 
 typename std::enable_if<matchGridTensorIndex<iScalar<vtype>,Level>::value, iScalar<vtype> >::type 
 transposeIndex (const iScalar<vtype> &arg)
 {
@@ -126,5 +125,6 @@ transposeIndex (const iScalar<vtype> &arg)
 }
 #endif
 
-}
+NAMESPACE_END(Grid);
+
 #endif
diff --git a/Grid/tensors/Tensor_unary.h b/Grid/tensors/Tensor_unary.h
index dd05a4a7..3adc2897 100644
--- a/Grid/tensors/Tensor_unary.h
+++ b/Grid/tensors/Tensor_unary.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,62 +25,63 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_TENSOR_UNARY_H
 #define GRID_TENSOR_UNARY_H
-namespace Grid {
 
-#define UNARY(func)\
-template<class obj> inline auto func(const iScalar<obj> &z) -> iScalar<obj>\
-{\
-    iScalar<obj> ret;\
-    ret._internal = func( (z._internal));\
-    return ret;\
-}\
-template<class obj,int N> inline auto func(const iVector<obj,N> &z) -> iVector<obj,N>\
-{\
-    iVector<obj,N> ret;\
-    for(int c1=0;c1<N;c1++){\
-      ret._internal[c1] = func( (z._internal[c1]));\
-    }\
-    return ret;\
-}\
-template<class obj,int N> inline auto func(const iMatrix<obj,N> &z) -> iMatrix<obj,N>\
-{\
-    iMatrix<obj,N> ret;\
-    for(int c1=0;c1<N;c1++){\
-    for(int c2=0;c2<N;c2++){\
-      ret._internal[c1][c2] = func( (z._internal[c1][c2]));\
-    }}\
-    return ret;\
-}
+NAMESPACE_BEGIN(Grid);
+
+#define UNARY(func)							\
+  template<class obj> accelerator_inline auto func(const iScalar<obj> &z) -> iScalar<obj> \
+  {									\
+    iScalar<obj> ret;							\
+    ret._internal = func( (z._internal));				\
+    return ret;								\
+  }									\
+  template<class obj,int N> accelerator_inline auto func(const iVector<obj,N> &z) -> iVector<obj,N>	\
+  {									\
+    iVector<obj,N> ret;							\
+    for(int c1=0;c1<N;c1++){						\
+      ret._internal[c1] = func( (z._internal[c1]));			\
+    }									\
+    return ret;								\
+  }									\
+  template<class obj,int N> accelerator_inline auto func(const iMatrix<obj,N> &z) -> iMatrix<obj,N>	\
+  {									\
+    iMatrix<obj,N> ret;							\
+    for(int c1=0;c1<N;c1++){						\
+      for(int c2=0;c2<N;c2++){						\
+	ret._internal[c1][c2] = func( (z._internal[c1][c2]));		\
+      }}								\
+    return ret;								\
+  }
 
 
 #define BINARY_RSCALAR(func,scal)					\
-template<class obj> inline iScalar<obj> func(const iScalar<obj> &z,scal y)	\
-{\
-    iScalar<obj> ret;\
-    ret._internal = func(z._internal,y);	\
-    return ret;\
-}\
- template<class obj,int N> inline iVector<obj,N> func(const iVector<obj,N> &z,scal y) \
-{\
-    iVector<obj,N> ret;\
-    for(int c1=0;c1<N;c1++){\
-      ret._internal[c1] = func(z._internal[c1],y);	\
-    }\
-    return ret;\
-}\
- template<class obj,int N> inline  iMatrix<obj,N> func(const iMatrix<obj,N> &z, scal y) \
-{\
-    iMatrix<obj,N> ret;\
-    for(int c1=0;c1<N;c1++){\
-    for(int c2=0;c2<N;c2++){\
-      ret._internal[c1][c2] = func(z._internal[c1][c2],y);	\
-    }}\
-    return ret;\
-}
+  template<class obj> accelerator_inline iScalar<obj> func(const iScalar<obj> &z,scal y) \
+  {									\
+    iScalar<obj> ret;							\
+    ret._internal = func(z._internal,y);				\
+    return ret;								\
+  }									\
+  template<class obj,int N> accelerator_inline iVector<obj,N> func(const iVector<obj,N> &z,scal y) \
+  {									\
+    iVector<obj,N> ret;							\
+    for(int c1=0;c1<N;c1++){						\
+      ret._internal[c1] = func(z._internal[c1],y);			\
+    }									\
+    return ret;								\
+  }									\
+  template<class obj,int N> accelerator_inline  iMatrix<obj,N> func(const iMatrix<obj,N> &z, scal y) \
+  {									\
+    iMatrix<obj,N> ret;							\
+    for(int c1=0;c1<N;c1++){						\
+      for(int c2=0;c2<N;c2++){						\
+	ret._internal[c1][c2] = func(z._internal[c1][c2],y);		\
+      }}								\
+    return ret;								\
+  }
 
 UNARY(sqrt);
 UNARY(rsqrt);
@@ -94,13 +95,13 @@ UNARY(abs);
 UNARY(Not);
 
 
-template<class obj> inline auto toReal(const iScalar<obj> &z) -> typename iScalar<obj>::Realified
+template<class obj> accelerator_inline auto toReal(const iScalar<obj> &z) -> typename iScalar<obj>::Realified
 {
   typename iScalar<obj>::Realified ret;
   ret._internal = toReal(z._internal);
   return ret;
 }
- template<class obj,int N> inline auto toReal(const iVector<obj,N> &z) -> typename iVector<obj,N>::Realified
+template<class obj,int N> accelerator_inline auto toReal(const iVector<obj,N> &z) -> typename iVector<obj,N>::Realified
 {
   typename iVector<obj,N>::Realified ret;
   for(int c1=0;c1<N;c1++){  
@@ -108,23 +109,23 @@ template<class obj> inline auto toReal(const iScalar<obj> &z) -> typename iScala
   }
   return ret;
 }
-template<class obj,int N> inline auto toReal(const iMatrix<obj,N> &z) -> typename iMatrix<obj,N>::Realified
+template<class obj,int N> accelerator_inline auto toReal(const iMatrix<obj,N> &z) -> typename iMatrix<obj,N>::Realified
 {
   typename iMatrix<obj,N>::Realified ret;
   for(int c1=0;c1<N;c1++){
-  for(int c2=0;c2<N;c2++){
-    ret._internal[c1][c2] = toReal(z._internal[c1][c2]);
-  }}
+    for(int c2=0;c2<N;c2++){
+      ret._internal[c1][c2] = toReal(z._internal[c1][c2]);
+    }}
   return ret;
 }
 
-template<class obj> inline auto toComplex(const iScalar<obj> &z) -> typename iScalar<obj>::Complexified
+template<class obj> accelerator_inline auto toComplex(const iScalar<obj> &z) -> typename iScalar<obj>::Complexified
 {
   typename iScalar<obj>::Complexified ret;
   ret._internal = toComplex(z._internal);
   return ret;
 }
- template<class obj,int N> inline auto toComplex(const iVector<obj,N> &z) -> typename iVector<obj,N>::Complexified
+template<class obj,int N> accelerator_inline auto toComplex(const iVector<obj,N> &z) -> typename iVector<obj,N>::Complexified
 {
   typename iVector<obj,N>::Complexified ret;
   for(int c1=0;c1<N;c1++){  
@@ -132,23 +133,24 @@ template<class obj> inline auto toComplex(const iScalar<obj> &z) -> typename iSc
   }
   return ret;
 }
-template<class obj,int N> inline auto toComplex(const iMatrix<obj,N> &z) -> typename iMatrix<obj,N>::Complexified
+template<class obj,int N> accelerator_inline auto toComplex(const iMatrix<obj,N> &z) -> typename iMatrix<obj,N>::Complexified
 {
   typename iMatrix<obj,N>::Complexified ret;
   for(int c1=0;c1<N;c1++){
-  for(int c2=0;c2<N;c2++){
-    ret._internal[c1][c2] = toComplex(z._internal[c1][c2]);
-  }}
+    for(int c2=0;c2<N;c2++){
+      ret._internal[c1][c2] = toComplex(z._internal[c1][c2]);
+    }}
   return ret;
 }
 
 BINARY_RSCALAR(div,Integer);
 BINARY_RSCALAR(mod,Integer);
 BINARY_RSCALAR(pow,RealD);
+BINARY_RSCALAR(pow,RealF);
 
 #undef UNARY
 #undef BINARY_RSCALAR
 
+NAMESPACE_END(Grid);
 
-}
 #endif
diff --git a/Grid/tensors/Tensors.h b/Grid/tensors/Tensors.h
index b94f22e3..43072a24 100644
--- a/Grid/tensors/Tensors.h
+++ b/Grid/tensors/Tensors.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -25,8 +25,8 @@ Author: neo <cossu@post.kek.jp>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_MATH_H
 #define GRID_MATH_H
 
@@ -48,4 +48,7 @@ Author: neo <cossu@post.kek.jp>
 #include <Grid/tensors/Tensor_extract_merge.h>
 #include <Grid/tensors/Tensor_logical.h>
 
+// Maybe expect different thread on each SIMD lane
+#include <Grid/tensors/Tensor_SIMT.h>
+
 #endif
diff --git a/Grid/threads/Pragmas.h b/Grid/threads/Pragmas.h
new file mode 100644
index 00000000..2ed33d67
--- /dev/null
+++ b/Grid/threads/Pragmas.h
@@ -0,0 +1,136 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/Threads.h
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
+
+#ifndef MAX
+#define MAX(x,y) ((x)>(y)?(x):(y))
+#define MIN(x,y) ((x)>(y)?(y):(x))
+#endif
+
+#define strong_inline     __attribute__((always_inline)) inline
+#define UNROLL  _Pragma("unroll")
+
+//////////////////////////////////////////////////////////////////////////////////
+// New primitives; explicit host thread calls, and accelerator data parallel calls
+//////////////////////////////////////////////////////////////////////////////////
+
+#ifdef _OPENMP
+#define GRID_OMP
+#include <omp.h>
+#endif
+
+#ifdef GRID_OMP
+#define DO_PRAGMA_(x) _Pragma (#x)
+#define DO_PRAGMA(x) DO_PRAGMA_(x)
+#define thread_num(a) omp_get_thread_num()
+#define thread_max(a) omp_get_max_threads()
+#else 
+#define DO_PRAGMA_(x) 
+#define DO_PRAGMA(x) 
+#define thread_num(a) (0)
+#define thread_max(a) (1)
+#endif
+
+#define naked_for(i,num,...) for ( uint64_t i=0;i<num;i++) { __VA_ARGS__ } ;
+#define naked_foreach(i,container,...) for ( uint64_t i=container.begin();i<container.end();i++) { __VA_ARGS__ } ;
+#define thread_for( i, num, ... )                           DO_PRAGMA(omp parallel for schedule(static)) naked_for(i,num,{__VA_ARGS__});
+#define thread_foreach( i, num, ... )                       DO_PRAGMA(omp parallel for schedule(static)) naked_foreach(i,num,{__VA_ARGS__});
+#define thread_for_in_region( i, num, ... )                 DO_PRAGMA(omp for schedule(static))          naked_for(i,num,{__VA_ARGS__});
+#define thread_for_collapse2( i, num, ... )                 DO_PRAGMA(omp parallel for collapse(2))      naked_for(i,num,{__VA_ARGS__});
+#define thread_for_collapse( N , i, num, ... )              DO_PRAGMA(omp parallel for collapse ( N ) )  naked_for(i,num,{__VA_ARGS__});
+#define thread_for_collapse_in_region( N , i, num, ... )    DO_PRAGMA(omp for collapse ( N ))            naked_for(i,num,{__VA_ARGS__});
+#define thread_region                                       DO_PRAGMA(omp parallel)
+#define thread_critical                                     DO_PRAGMA(omp critical)
+
+
+//////////////////////////////////////////////////////////////////////////////////
+// Accelerator primitives; fall back to threading
+//////////////////////////////////////////////////////////////////////////////////
+#ifdef __NVCC__
+#define GRID_NVCC
+#endif
+
+#ifdef GRID_NVCC
+
+extern uint32_t gpu_threads;
+
+#define accelerator        __host__ __device__
+#define accelerator_inline __host__ __device__ inline
+
+template<typename lambda>  __global__
+void LambdaApplySIMT(uint64_t Isites, uint64_t Osites, lambda Lambda)
+{
+  uint64_t isite = threadIdx.y;
+  uint64_t osite = threadIdx.x+blockDim.x*blockIdx.x;
+  if ( (osite <Osites) && (isite<Isites) ) {
+    Lambda(isite,osite);
+  }
+}
+
+/////////////////////////////////////////////////////////////////
+// Internal only really... but need to call when 
+/////////////////////////////////////////////////////////////////
+#define accelerator_barrier(dummy)				\
+  {								\
+    cudaDeviceSynchronize();					\
+    cudaError err = cudaGetLastError();				\
+    if ( cudaSuccess != err ) {					\
+      printf("Cuda error %s \n", cudaGetErrorString( err )); \
+      puts(__FILE__); \
+      printf("Line %d\n",__LINE__);					\
+      exit(0);							\
+    }								\
+  }
+
+// Copy the for_each_n style ; Non-blocking variant
+#define accelerator_forNB( iterator, num, nsimd, ... )			\
+  {									\
+    typedef uint64_t Iterator;						\
+    auto lambda = [=] accelerator (Iterator lane,Iterator iterator) mutable { \
+      __VA_ARGS__;							\
+    };									\
+    dim3 cu_threads(gpu_threads,nsimd);					\
+    dim3 cu_blocks ((num+gpu_threads-1)/gpu_threads);			\
+    LambdaApplySIMT<<<cu_blocks,cu_threads>>>(nsimd,num,lambda);	\
+  }
+
+// Copy the for_each_n style ; Non-blocking variant (default
+#define accelerator_for( iterator, num, nsimd, ... )		\
+  accelerator_forNB(iterator, num, nsimd, { __VA_ARGS__ } );	\
+  accelerator_barrier(dummy);
+
+#else
+
+#define accelerator 
+#define accelerator_inline strong_inline
+#define accelerator_for(iterator,num,nsimd, ... )   thread_for(iterator, num, { __VA_ARGS__ });
+#define accelerator_forNB(iterator,num,nsimd, ... ) thread_for(iterator, num, { __VA_ARGS__ });
+#define accelerator_barrier(dummy) 
+
+#endif
diff --git a/Grid/threads/Threads.h b/Grid/threads/Threads.h
index 9041d01e..29cae060 100644
--- a/Grid/threads/Threads.h
+++ b/Grid/threads/Threads.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -24,50 +24,17 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef GRID_THREADS_H
-#define GRID_THREADS_H
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once 
 
-#ifdef _OPENMP
-#define GRID_OMP
-#endif
 
-#define UNROLL  _Pragma("unroll")
-
-#ifdef GRID_OMP
-#include <omp.h>
-
-#define PARALLEL_FOR_LOOP        _Pragma("omp parallel for schedule(static)")
-#define PARALLEL_FOR_LOOP_INTERN _Pragma("omp for schedule(static)")
-#define PARALLEL_NESTED_LOOP2 _Pragma("omp parallel for collapse(2)")
-#define PARALLEL_NESTED_LOOP5 _Pragma("omp parallel for collapse(5)")
-#define PARALLEL_REGION       _Pragma("omp parallel")
-#define PARALLEL_CRITICAL     _Pragma("omp critical")
-#else
-#define PARALLEL_FOR_LOOP
-#define PARALLEL_FOR_LOOP_INTERN
-#define PARALLEL_FOR_LOOP_REDUCE(op, var)
-#define PARALLEL_NESTED_LOOP2
-#define PARALLEL_NESTED_LOOP5
-#define PARALLEL_REGION
-#define PARALLEL_CRITICAL
-#endif
-
-#define parallel_region    PARALLEL_REGION
-#define parallel_for       PARALLEL_FOR_LOOP for
-#define parallel_for_internal PARALLEL_FOR_LOOP_INTERN for
-#define parallel_for_nest2 PARALLEL_NESTED_LOOP2 for
-#define parallel_for_nest5 PARALLEL_NESTED_LOOP5 for
-#define parallel_critical PARALLEL_CRITICAL
-
-namespace Grid {
-
-  // Introduce a class to gain deterministic bit reproducible reduction.
-  // make static; perhaps just a namespace is required.
+// Introduce a class to gain deterministic bit reproducible reduction.
+// make static; perhaps just a namespace is required.
+NAMESPACE_BEGIN(Grid);
 
 class GridThread {
- public:
+public:
   static int _threads;
   static int _hyperthreads;
   static int _cores;
@@ -89,7 +56,6 @@ class GridThread {
   };
   static void SetMaxThreads(void) { 
 #ifdef GRID_OMP
-    //    setenv("KMP_AFFINITY","balanced",1);
     _threads = omp_get_max_threads();
     omp_set_num_threads(_threads);
 #else 
@@ -134,11 +100,11 @@ class GridThread {
   
   template<class obj> static void ThreadSum( std::vector<obj> &sum_array,obj &val,int me){
     sum_array[me] = val;
-    val=zero;
+    val=Zero();
     ThreadBarrier();
     for(int i=0;i<_threads;i++) val+= sum_array[i];
     ThreadBarrier();
-  };
+  }
 
   static void bcopy(const void *src, void *dst, size_t len) {
 #ifdef GRID_OMP
@@ -158,5 +124,5 @@ class GridThread {
 
 };
 
-}
-#endif
+NAMESPACE_END(Grid);
+
diff --git a/Grid/util/Coordinate.h b/Grid/util/Coordinate.h
new file mode 100644
index 00000000..c319a14d
--- /dev/null
+++ b/Grid/util/Coordinate.h
@@ -0,0 +1,111 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./lib/cartesian/Coordinate.h
+
+    Copyright (C) 2018
+
+    Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once 
+
+////////////////////////////////////////////////////////////////////////////
+// Provide a stack resident container for coordinates, or extents
+////////////////////////////////////////////////////////////////////////////
+NAMESPACE_BEGIN(Grid);
+template<class _T,int MaxEntries>
+class AcceleratorVector { 
+public:
+
+  typedef _T  value;
+  typedef int size_type;
+  typedef value & reference;
+  typedef const value & const_reference;
+  typedef value * pointer;
+  typedef const value * const_pointer;
+
+private:
+  value _data[MaxEntries];
+  size_type _size;
+  
+public:
+  accelerator_inline reference       operator[](size_type __n)       { return _data[__n];}
+  accelerator_inline const_reference operator[](size_type __n) const { return _data[__n];}
+  accelerator_inline size_type size(void) const { return _size; };
+  accelerator_inline void  clear(void) { resize(0);}
+  accelerator_inline void  resize(size_type sz) {
+    assert(sz>=0);
+    assert(sz<=MaxEntries);
+    _size = sz;
+  }
+  accelerator_inline void  resize(size_type sz,const value &val) {
+    assert(sz>=0);
+    assert(sz<=MaxEntries);
+    _size = sz;
+    for(int s=0;s<sz;s++) _data[s]=val;
+  }
+  accelerator_inline pointer begin(void)                   { return &_data[0]; } 
+  accelerator_inline const_pointer begin(void) const       { return &_data[0]; } 
+  accelerator_inline pointer end  (void)                   { return &_data[_size]; } 
+  accelerator_inline const_pointer end  (void) const       { return &_data[_size]; } 
+  accelerator_inline void push_back(const value &val)      { resize(_size+1); _data[_size-1] = val;}
+  accelerator_inline AcceleratorVector()                   { _size = 0; }
+  accelerator_inline AcceleratorVector(size_type sz)           { resize(sz); }
+  accelerator_inline AcceleratorVector(size_type sz,const value &val) { resize(sz,val); }
+  AcceleratorVector(const std::vector<value> &copyme) { 
+    resize(copyme.size());
+    for(int s=0;s<_size;s++){
+      _data[s] = copyme[s];
+    }
+  }
+  std::vector<value> toVector(void) const { 
+    auto clone =*this;
+    std::vector<value> ret;
+    std::copy(clone.begin(),clone.end(),std::back_inserter(ret));
+    return ret;
+  }
+};
+
+////////////////////////////////////////////////////////////////
+// Coordinate class, maxdims = 8 for now.
+////////////////////////////////////////////////////////////////
+#define GRID_MAX_LATTICE_DIMENSION (8)
+#define GRID_MAX_SIMD              (16)
+
+static constexpr int MaxDims = GRID_MAX_LATTICE_DIMENSION;
+
+typedef AcceleratorVector<int,MaxDims> Coordinate;
+
+template<class T,int _ndim>
+inline std::ostream & operator<<(std::ostream &os, const AcceleratorVector<T,_ndim> &v)
+{
+  os << "[";
+  for(int s=0;s<v.size();s++) {
+    os << v[s] << " ";
+  }
+  if (v.size() > 0) {
+    os << "\b";
+  }
+  os << "]";
+  return os;
+}
+
+NAMESPACE_END(Grid);
diff --git a/Grid/util/Init.cc b/Grid/util/Init.cc
index e15f9e4f..ad5eac47 100644
--- a/Grid/util/Init.cc
+++ b/Grid/util/Init.cc
@@ -26,8 +26,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 /****************************************************************************/
 /* pab: Signal magic. Processor state dump is x86-64 specific               */
 /****************************************************************************/
@@ -57,8 +57,9 @@ static int
 feenableexcept (unsigned int excepts)
 {
   static fenv_t fenv;
-  unsigned int new_excepts = excepts & FE_ALL_EXCEPT,
-    old_excepts;  // previous masks
+  unsigned int new_excepts = excepts & FE_ALL_EXCEPT;
+  unsigned int old_excepts;  // previous masks
+  int iold_excepts;  // previous masks
 
   if ( fegetenv (&fenv) ) return -1;
   old_excepts = fenv.__control & FE_ALL_EXCEPT;
@@ -67,40 +68,43 @@ feenableexcept (unsigned int excepts)
   fenv.__control &= ~new_excepts;
   fenv.__mxcsr   &= ~(new_excepts << 7);
 
-  return ( fesetenv (&fenv) ? -1 : old_excepts );
+  iold_excepts  = (int) old_excepts;
+  return ( fesetenv (&fenv) ? -1 : iold_excepts );
 }
 #endif
 
-namespace Grid {
+uint32_t gpu_threads=8;
 
+NAMESPACE_BEGIN(Grid);
 
 //////////////////////////////////////////////////////
 // Convenience functions to access stadard command line arg
 // driven parallelism controls
 //////////////////////////////////////////////////////
-static std::vector<int> Grid_default_latt;
-static std::vector<int> Grid_default_mpi;
+static Coordinate Grid_default_latt;
+static Coordinate Grid_default_mpi;
 
 int GridThread::_threads =1;
 int GridThread::_hyperthreads=1;
 int GridThread::_cores=1;
 
-const std::vector<int> &GridDefaultLatt(void)     {return Grid_default_latt;};
-const std::vector<int> &GridDefaultMpi(void)      {return Grid_default_mpi;};
-const std::vector<int> GridDefaultSimd(int dims,int nsimd)
+
+const Coordinate &GridDefaultLatt(void)     {return Grid_default_latt;};
+const Coordinate &GridDefaultMpi(void)      {return Grid_default_mpi;};
+const Coordinate GridDefaultSimd(int dims,int nsimd)
 {
-    std::vector<int> layout(dims);
-    int nn=nsimd;
-    for(int d=dims-1;d>=0;d--){
-      if ( nn>=2) {
-	layout[d]=2;
-	nn/=2;
-      } else {
-	layout[d]=1;
-      }
+  Coordinate layout(dims);
+  int nn=nsimd;
+  for(int d=dims-1;d>=0;d--){
+    if ( nn>=2) {
+      layout[d]=2;
+      nn/=2;
+    } else {
+      layout[d]=1;
     }
-    assert(nn==1);
-    return layout;
+  }
+  assert(nn==1);
+  return layout;
 }
 
 ////////////////////////////////////////////////////////////
@@ -119,7 +123,7 @@ bool GridCmdOptionExists(char** begin, char** end, const std::string& option)
 {
   return std::find(begin, end, option) != end;
 }
-  // Comma separated list
+// Comma separated list
 void GridCmdOptionCSL(std::string str,std::vector<std::string> & vec)
 {
   size_t pos = 0;
@@ -137,7 +141,8 @@ void GridCmdOptionCSL(std::string str,std::vector<std::string> & vec)
   return;
 }
 
-void GridCmdOptionIntVector(std::string &str,std::vector<int> & vec)
+template<class VectorInt>
+void GridCmdOptionIntVector(std::string &str,VectorInt & vec)
 {
   vec.resize(0);
   std::stringstream ss(str);
@@ -159,11 +164,11 @@ void GridCmdOptionInt(std::string &str,int & val)
 
 
 void GridParseLayout(char **argv,int argc,
-		     std::vector<int> &latt,
-		     std::vector<int> &mpi)
+		     Coordinate &latt_c,
+		     Coordinate &mpi_c)
 {
-  mpi =std::vector<int>({1,1,1,1});
-  latt=std::vector<int>({8,8,8,8});
+  auto mpi =std::vector<int>({1,1,1,1});
+  auto latt=std::vector<int>({8,8,8,8});
 
   GridThread::SetMaxThreads();
 
@@ -187,15 +192,40 @@ void GridParseLayout(char **argv,int argc,
     assert(ompthreads.size()==1);
     GridThread::SetThreads(ompthreads[0]);
   }
+  if( GridCmdOptionExists(argv,argv+argc,"--gpu-threads") ){
+    std::vector<int> gputhreads(0);
+#ifndef GRID_NVCC
+    std::cout << GridLogWarning << "'--gpu-threads' option used but Grid was"
+              << " not compiled with GPU support" << std::endl;
+#endif
+    arg= GridCmdOptionPayload(argv,argv+argc,"--gpu-threads");
+    GridCmdOptionIntVector(arg,gputhreads);
+    assert(gputhreads.size()==1);
+    gpu_threads=gputhreads[0];
+  }
+
   if( GridCmdOptionExists(argv,argv+argc,"--cores") ){
     int cores;
     arg= GridCmdOptionPayload(argv,argv+argc,"--cores");
     GridCmdOptionInt(arg,cores);
     GridThread::SetCores(cores);
   }
+  // Copy back into coordinate format
+  int nd = mpi.size();
+  assert(latt.size()==nd);
+  latt_c.resize(nd);
+   mpi_c.resize(nd);
+  for(int d=0;d<nd;d++){
+    latt_c[d] = latt[d];
+     mpi_c[d] = mpi[d];
+  } 
 }
 
-std::string GridCmdVectorIntToString(const std::vector<int> & vec){
+template<class VectorInt>
+std::string GridCmdVectorIntToString(const VectorInt & vec_in){
+  int sz = vec_in.size();
+  std::vector<int> vec(sz);
+  for(int s=0;s<sz;s++) vec[s] = vec_in[s];
   std::ostringstream oss;
   std::copy(vec.begin(), vec.end(),std::ostream_iterator<int>(oss, " "));
   return oss.str();
@@ -203,63 +233,16 @@ std::string GridCmdVectorIntToString(const std::vector<int> & vec){
 /////////////////////////////////////////////////////////
 // Reinit guard
 /////////////////////////////////////////////////////////
-static int Grid_is_initialised = 0;
 static MemoryStats dbgMemStats;
+static int Grid_is_initialised;
 
-void Grid_init(int *argc,char ***argv)
+/////////////////////////////////////////////////////////
+// Reinit guard
+/////////////////////////////////////////////////////////
+void GridBanner(void)
 {
-  GridLogger::GlobalStopWatch.Start();
-
-  std::string arg;
-
-  ////////////////////////////////////
-  // Shared memory block size
-  ////////////////////////////////////
-  if( GridCmdOptionExists(*argv,*argv+*argc,"--shm") ){
-    int MB;
-    arg= GridCmdOptionPayload(*argv,*argv+*argc,"--shm");
-    GridCmdOptionInt(arg,MB);
-    uint64_t MB64 = MB;
-    GlobalSharedMemory::MAX_MPI_SHM_BYTES = MB64*1024LL*1024LL;
-  }
-
-  if( GridCmdOptionExists(*argv,*argv+*argc,"--shm-hugepages") ){
-    GlobalSharedMemory::Hugepages = 1;
-  }
-
-
-  if( GridCmdOptionExists(*argv,*argv+*argc,"--debug-signals") ){
-    Grid_debug_handler_init();
-  }
-
-  CartesianCommunicator::Init(argc,argv);
-
-  if( !GridCmdOptionExists(*argv,*argv+*argc,"--debug-stdout") ){
-    Grid_quiesce_nodes();
-  } else { 
-    FILE *fp;
-    std::ostringstream fname;
-    fname<<"Grid.stdout.";
-    fname<<CartesianCommunicator::RankWorld();
-    fp=freopen(fname.str().c_str(),"w",stdout);
-    assert(fp!=(FILE *)NULL);
-
-    std::ostringstream ename;
-    ename<<"Grid.stderr.";
-    ename<<CartesianCommunicator::RankWorld();
-    fp=freopen(ename.str().c_str(),"w",stderr);
-    assert(fp!=(FILE *)NULL);
-  }
-
-  if( GridCmdOptionExists(*argv,*argv+*argc,"--debug-mem") ){
-    MemoryProfiler::debug = true;
-    MemoryProfiler::stats = &dbgMemStats;
-  }
-
-  ////////////////////////////////////
-  // Banner
-  ////////////////////////////////////
-  if ( CartesianCommunicator::RankWorld() == 0 ) { 
+  static int printed =0;
+  if( !printed ) {
     std::cout <<std::endl;
     std::cout  << "__|__|__|__|__|__|__|__|__|__|__|__|__|__|__"<<std::endl; 
     std::cout  << "__|__|__|__|__|__|__|__|__|__|__|__|__|__|__"<<std::endl; 
@@ -289,19 +272,175 @@ void Grid_init(int *argc,char ***argv)
     std::cout << "MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the"<<std::endl;
     std::cout << "GNU General Public License for more details."<<std::endl;
     printHash();
-  #ifdef GRID_BUILD_REF
-  #define _GRID_BUILD_STR(x) #x
-  #define GRID_BUILD_STR(x) _GRID_BUILD_STR(x)
+#ifdef GRID_BUILD_REF
+#define _GRID_BUILD_STR(x) #x
+#define GRID_BUILD_STR(x) _GRID_BUILD_STR(x)
     std::cout << "Build " << GRID_BUILD_STR(GRID_BUILD_REF) << std::endl;
-  #endif
+#endif
     std::cout << std::endl;
+    printed=1;
+  }
+}
+#ifdef GRID_NVCC
+cudaDeviceProp *gpu_props;
+#endif
+void GridGpuInit(void)
+{
+#ifdef GRID_NVCC
+  int nDevices = 1;
+  cudaGetDeviceCount(&nDevices);
+  gpu_props = new cudaDeviceProp[nDevices];
+
+  char * localRankStr = NULL;
+  int rank = 0, device = 0, world_rank=0; 
+#define ENV_LOCAL_RANK_OMPI    "OMPI_COMM_WORLD_LOCAL_RANK"
+#define ENV_LOCAL_RANK_MVAPICH "MV2_COMM_WORLD_LOCAL_RANK"
+#define ENV_RANK_OMPI          "OMPI_COMM_WORLD_RANK"
+#define ENV_RANK_MVAPICH       "MV2_COMM_WORLD_RANK"
+  // We extract the local rank initialization using an environment variable
+  if ((localRankStr = getenv(ENV_LOCAL_RANK_OMPI)) != NULL)
+  {
+    rank = atoi(localRankStr);		
+    device = rank %nDevices;
+  }
+  if ((localRankStr = getenv(ENV_LOCAL_RANK_MVAPICH)) != NULL)
+  {
+    rank = atoi(localRankStr);		
+    device = rank %nDevices;
+  }
+  if ((localRankStr = getenv(ENV_RANK_OMPI   )) != NULL) { world_rank = atoi(localRankStr);}
+  if ((localRankStr = getenv(ENV_RANK_MVAPICH)) != NULL) { world_rank = atoi(localRankStr);}
+
+  cudaSetDevice(device);
+  if ( world_rank == 0 ) {
+    GridBanner();
+    printf("GpuInit: ================================================\n");
+    printf("GpuInit: Setting up Cuda Device map before first MPI call\n",nDevices);
+    printf("GpuInit: ================================================\n");
+    printf("GpuInit: Cuda reports %d GPUs on MPI rank 0\n",nDevices);
   }
 
+  for (int i = 0; i < nDevices; i++) {
+
+#define GPU_PROP_FMT(canMapHostMemory,FMT)     printf("GpuInit:   " #canMapHostMemory ": " FMT" \n",prop.canMapHostMemory);
+#define GPU_PROP(canMapHostMemory)             GPU_PROP_FMT(canMapHostMemory,"%d");
+    
+    //      cudaGetDeviceProperties(&prop, i);
+    cudaGetDeviceProperties(&gpu_props[i], i);
+    if ( world_rank == 0) {
+      cudaDeviceProp prop; 
+      prop = gpu_props[i];
+      printf("GpuInit: ========================\n");
+      printf("GpuInit: Device Number    : %d\n", i);
+      printf("GpuInit: ========================\n");
+      printf("GpuInit: Device identifier: %s\n", prop.name);
+      //      printf("GpuInit:   Peak Memory Bandwidth (GB/s): %f\n",(float)2.0*prop.memoryClockRate*(prop.memoryBusWidth/8)/1.0e6);
+      GPU_PROP(managedMemory);
+      GPU_PROP(isMultiGpuBoard);
+      GPU_PROP(warpSize);
+#if 0
+      GPU_PROP(unifiedAddressing);
+      GPU_PROP(l2CacheSize);
+      GPU_PROP(singleToDoublePrecisionPerfRatio);
+#endif
+    }
+  }
+  if ( world_rank == 0 ) {
+    printf("GpuInit: ================================================\n");
+  }
+#endif
+}
+
+void Grid_init(int *argc,char ***argv)
+{
+
+  assert(Grid_is_initialised == 0);
+
+  GridLogger::GlobalStopWatch.Start();
+
+  std::string arg;
+
+  //////////////////////////////////////////////////////////
+  // Early intialisation necessities without rank knowledge
+  //////////////////////////////////////////////////////////
+  GridGpuInit(); // Must come first to set device prior to MPI init
+
+  if( GridCmdOptionExists(*argv,*argv+*argc,"--shm") ){
+    int MB;
+    arg= GridCmdOptionPayload(*argv,*argv+*argc,"--shm");
+    GridCmdOptionInt(arg,MB);
+    uint64_t MB64 = MB;
+    GlobalSharedMemory::MAX_MPI_SHM_BYTES = MB64*1024LL*1024LL;
+  }
+
+  if( GridCmdOptionExists(*argv,*argv+*argc,"--hypercube") ){
+    int enable;
+    arg= GridCmdOptionPayload(*argv,*argv+*argc,"--hypercube");
+    GridCmdOptionInt(arg,enable);
+    GlobalSharedMemory::HPEhypercube = enable;
+  }
+
+  if( GridCmdOptionExists(*argv,*argv+*argc,"--shm-hugepages") ){
+    GlobalSharedMemory::Hugepages = 1;
+  }
+
+
+  if( GridCmdOptionExists(*argv,*argv+*argc,"--debug-signals") ){
+    Grid_debug_handler_init();
+  }
+
+  //////////////////////////////////////////////////////////
+  // MPI initialisation
+  //////////////////////////////////////////////////////////
+  CartesianCommunicator::Init(argc,argv);
+
+  ////////////////////////////////////
+  // Banner after MPI (unless GPU)
+  ////////////////////////////////////
+  if ( CartesianCommunicator::RankWorld() == 0 ) { 
+    GridBanner();
+  }
+
+  /////////////////////////////////////////////////////////////////
+  // Rank information can be used to control who logs
+  /////////////////////////////////////////////////////////////////
+  if( !GridCmdOptionExists(*argv,*argv+*argc,"--debug-stdout") ){
+    Grid_quiesce_nodes();
+  } else { 
+    FILE *fp;
+    std::ostringstream fname;
+    fname<<"Grid.stdout.";
+    fname<<CartesianCommunicator::RankWorld();
+    fp=freopen(fname.str().c_str(),"w",stdout);
+    assert(fp!=(FILE *)NULL);
+
+    std::ostringstream ename;
+    ename<<"Grid.stderr.";
+    ename<<CartesianCommunicator::RankWorld();
+    fp=freopen(ename.str().c_str(),"w",stderr);
+    assert(fp!=(FILE *)NULL);
+  }
+  ////////////////////////////////////////////////////
+  // OK to use GridLogMessage etc from here on
+  ////////////////////////////////////////////////////
+  std::cout << GridLogMessage << "================================================ "<<std::endl;
+  std::cout << GridLogMessage << "MPI is initialised and logging filters activated "<<std::endl;
+  std::cout << GridLogMessage << "================================================ "<<std::endl;
+
+  std::cout << GridLogMessage << "Requested "<< GlobalSharedMemory::MAX_MPI_SHM_BYTES <<" byte stencil comms buffers "<<std::endl;
+  if ( GlobalSharedMemory::Hugepages) {
+    std::cout << GridLogMessage << "Mapped stencil comms buffers as MAP_HUGETLB "<<std::endl;
+  }
+
+
+  if( GridCmdOptionExists(*argv,*argv+*argc,"--debug-mem") ){
+    MemoryProfiler::debug = true;
+    MemoryProfiler::stats = &dbgMemStats;
+  }
 
   ////////////////////////////////////
   // Logging
   ////////////////////////////////////
-
   std::vector<std::string> logstreams;
   std::string defaultLog("Error,Warning,Message,Performance");
   GridCmdOptionCSL(defaultLog,logstreams);
@@ -359,23 +498,23 @@ void Grid_init(int *argc,char ***argv)
   ////////////////////////////////////
 
   if( GridCmdOptionExists(*argv,*argv+*argc,"--dslash-unroll") ){
-    QCD::WilsonKernelsStatic::Opt=QCD::WilsonKernelsStatic::OptHandUnroll;
-    QCD::StaggeredKernelsStatic::Opt=QCD::StaggeredKernelsStatic::OptHandUnroll;
+    WilsonKernelsStatic::Opt=WilsonKernelsStatic::OptHandUnroll;
+    StaggeredKernelsStatic::Opt=StaggeredKernelsStatic::OptHandUnroll;
   }
   if( GridCmdOptionExists(*argv,*argv+*argc,"--dslash-asm") ){
-    QCD::WilsonKernelsStatic::Opt=QCD::WilsonKernelsStatic::OptInlineAsm;
-    QCD::StaggeredKernelsStatic::Opt=QCD::StaggeredKernelsStatic::OptInlineAsm;
+    WilsonKernelsStatic::Opt=WilsonKernelsStatic::OptInlineAsm;
+    StaggeredKernelsStatic::Opt=StaggeredKernelsStatic::OptInlineAsm;
   }
   if( GridCmdOptionExists(*argv,*argv+*argc,"--dslash-generic") ){
-    QCD::WilsonKernelsStatic::Opt=QCD::WilsonKernelsStatic::OptGeneric;
-    QCD::StaggeredKernelsStatic::Opt=QCD::StaggeredKernelsStatic::OptGeneric;
+    WilsonKernelsStatic::Opt=WilsonKernelsStatic::OptGeneric;
+    StaggeredKernelsStatic::Opt=StaggeredKernelsStatic::OptGeneric;
   }
   if( GridCmdOptionExists(*argv,*argv+*argc,"--comms-overlap") ){
-    QCD::WilsonKernelsStatic::Comms = QCD::WilsonKernelsStatic::CommsAndCompute;
-    QCD::StaggeredKernelsStatic::Comms = QCD::StaggeredKernelsStatic::CommsAndCompute;
+    WilsonKernelsStatic::Comms = WilsonKernelsStatic::CommsAndCompute;
+    StaggeredKernelsStatic::Comms = StaggeredKernelsStatic::CommsAndCompute;
   } else {
-    QCD::WilsonKernelsStatic::Comms = QCD::WilsonKernelsStatic::CommsThenCompute;
-    QCD::StaggeredKernelsStatic::Comms = QCD::StaggeredKernelsStatic::CommsThenCompute;
+    WilsonKernelsStatic::Comms = WilsonKernelsStatic::CommsThenCompute;
+    StaggeredKernelsStatic::Comms = StaggeredKernelsStatic::CommsThenCompute;
   }
   if( GridCmdOptionExists(*argv,*argv+*argc,"--comms-concurrent") ){
     CartesianCommunicator::SetCommunicatorPolicy(CartesianCommunicator::CommunicatorPolicyConcurrent);
@@ -407,10 +546,6 @@ void Grid_init(int *argc,char ***argv)
 		  Grid_default_latt,
 		  Grid_default_mpi);
 
-  std::cout << GridLogMessage << "Requesting "<< GlobalSharedMemory::MAX_MPI_SHM_BYTES <<" byte stencil comms buffers "<<std::endl;
-  if ( GlobalSharedMemory::Hugepages) {
-    std::cout << GridLogMessage << "Mapped stencil comms buffers as MAP_HUGETLB "<<std::endl;
-  }
 
   if( GridCmdOptionExists(*argv,*argv+*argc,"--decomposition") ){
     std::cout<<GridLogMessage<<"Grid Default Decomposition patterns\n";
@@ -421,8 +556,6 @@ void Grid_init(int *argc,char ***argv)
     std::cout<<GridLogMessage<<"\tvComplexF      : "<<sizeof(vComplexF)*8 <<"bits ; " <<GridCmdVectorIntToString(GridDefaultSimd(4,vComplexF::Nsimd()))<<std::endl;
     std::cout<<GridLogMessage<<"\tvComplexD      : "<<sizeof(vComplexD)*8 <<"bits ; " <<GridCmdVectorIntToString(GridDefaultSimd(4,vComplexD::Nsimd()))<<std::endl;
   }
-
-
   Grid_is_initialised = 1;
 }
 
@@ -436,13 +569,14 @@ void Grid_finalize(void)
 #if defined (GRID_COMMS_SHMEM)
   shmem_finalize();
 #endif
+  Grid_is_initialised = 0;
 }
 
 void GridLogLayout() {
-    std::cout << GridLogMessage << "Grid Layout\n";
-    std::cout << GridLogMessage << "\tGlobal lattice size  : "<< GridCmdVectorIntToString(GridDefaultLatt()) << std::endl;
-    std::cout << GridLogMessage << "\tOpenMP threads       : "<< GridThread::GetThreads() <<std::endl;
-    std::cout << GridLogMessage << "\tMPI tasks            : "<< GridCmdVectorIntToString(GridDefaultMpi()) << std::endl;
+  std::cout << GridLogMessage << "Grid Layout\n";
+  std::cout << GridLogMessage << "\tGlobal lattice size  : "<< GridCmdVectorIntToString(GridDefaultLatt()) << std::endl;
+  std::cout << GridLogMessage << "\tOpenMP threads       : "<< GridThread::GetThreads() <<std::endl;
+  std::cout << GridLogMessage << "\tMPI tasks            : "<< GridCmdVectorIntToString(GridDefaultMpi()) << std::endl;
 }
 
 void * Grid_backtrace_buffer[_NBACKTRACE];
@@ -490,9 +624,14 @@ void Grid_sa_signal_handler(int sig,siginfo_t *si,void * ptr)
   return;
 };
 
+void Grid_exit_handler(void)
+{
+  BACKTRACEFP(stdout);
+  fflush(stdout);
+}
 void Grid_debug_handler_init(void)
 {
-  struct sigaction sa,osa;
+  struct sigaction sa;
   sigemptyset (&sa.sa_mask);
   sa.sa_sigaction= Grid_sa_signal_handler;
   sa.sa_flags    = SA_SIGINFO;
@@ -505,5 +644,8 @@ void Grid_debug_handler_init(void)
   sigaction(SIGFPE,&sa,NULL);
   sigaction(SIGKILL,&sa,NULL);
   sigaction(SIGILL,&sa,NULL);
+
+  atexit(Grid_exit_handler);
 }
-}
+
+NAMESPACE_END(Grid);
diff --git a/Grid/util/Init.h b/Grid/util/Init.h
index c8236dc6..f7f032ba 100644
--- a/Grid/util/Init.h
+++ b/Grid/util/Init.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid
 
@@ -24,44 +24,47 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#ifndef GRID_INIT_H
-#define GRID_INIT_H
-
-namespace Grid {
-
-  void Grid_init(int *argc,char ***argv);
-  void Grid_finalize(void);
-
-  // internal, controled with --handle
-  void Grid_sa_signal_handler(int sig,siginfo_t *si,void * ptr);
-  void Grid_debug_handler_init(void);
-  void Grid_quiesce_nodes(void);
-  void Grid_unquiesce_nodes(void);
-
-  const std::vector<int> GridDefaultSimd(int dims,int nsimd);
-  const std::vector<int> &GridDefaultLatt(void);
-  const std::vector<int> &GridDefaultMpi(void);
-  const int              &GridThreads(void)  ;
-  void                    GridSetThreads(int t) ;
-  void GridLogTimestamp(int);
-  void GridLogLayout();
-
-  // Common parsing chores
-  std::string GridCmdOptionPayload(char ** begin, char ** end, const std::string & option);
-  bool        GridCmdOptionExists(char** begin, char** end, const std::string& option);
-  std::string GridCmdVectorIntToString(const std::vector<int> & vec);
-  void GridCmdOptionCSL(std::string str,std::vector<std::string> & vec);
-  void GridCmdOptionIntVector(std::string &str,std::vector<int> & vec);
+*************************************************************************************/
+/*  END LEGAL */
+#pragma once
 
 
-  void GridParseLayout(char **argv,int argc,
-		       std::vector<int> &latt,
-		       std::vector<int> &simd,
-		       std::vector<int> &mpi);
+NAMESPACE_BEGIN(Grid);
 
-  void printHash(void);
+void Grid_init(int *argc,char ***argv);
+void Grid_finalize(void);
+
+// internal, controled with --handle
+void Grid_sa_signal_handler(int sig,siginfo_t *si,void * ptr);
+void Grid_debug_handler_init(void);
+void Grid_quiesce_nodes(void);
+void Grid_unquiesce_nodes(void);
+
+const Coordinate  GridDefaultSimd(int dims,int nsimd);
+const Coordinate &GridDefaultLatt(void);
+const Coordinate &GridDefaultMpi(void);
+const int        &GridThreads(void)  ;
+void              GridSetThreads(int t) ;
+void GridLogTimestamp(int);
+void GridLogLayout();
+
+// Common parsing chores
+std::string GridCmdOptionPayload(char ** begin, char ** end, const std::string & option);
+bool        GridCmdOptionExists(char** begin, char** end, const std::string& option);
+template<class VectorInt>
+std::string GridCmdVectorIntToString(const VectorInt & vec);
+void GridCmdOptionCSL(std::string str,std::vector<std::string> & vec);
+template<class VectorInt>
+void GridCmdOptionIntVector(std::string &str,VectorInt & vec);
+
+
+void GridParseLayout(char **argv,int argc,
+		     std::vector<int> &latt,
+		     std::vector<int> &simd,
+		     std::vector<int> &mpi);
+
+void printHash(void);
+
+
+NAMESPACE_END(Grid);
 
-};
-#endif
diff --git a/Grid/util/Lexicographic.h b/Grid/util/Lexicographic.h
index f5c55b74..ff98c6fc 100644
--- a/Grid/util/Lexicographic.h
+++ b/Grid/util/Lexicographic.h
@@ -7,7 +7,8 @@ namespace Grid{
   class Lexicographic {
   public:
 
-    static inline void CoorFromIndex (std::vector<int>& coor,int index,const std::vector<int> &dims){
+    template<class coor_t>
+    static accelerator_inline void CoorFromIndex (coor_t& coor,int index,const coor_t &dims){
       int nd= dims.size();
       coor.resize(nd);
       for(int d=0;d<nd;d++){
@@ -16,7 +17,8 @@ namespace Grid{
       }
     }
 
-    static inline void IndexFromCoor (const std::vector<int>& coor,int &index,const std::vector<int> &dims){
+    template<class coor_t>
+    static accelerator_inline void IndexFromCoor (const coor_t& coor,int &index,const coor_t &dims){
       int nd=dims.size();
       int stride=1;
       index=0;
@@ -26,7 +28,8 @@ namespace Grid{
       }
     }
 
-    static inline void IndexFromCoorReversed (const std::vector<int>& coor,int &index,const std::vector<int> &dims){
+    template<class coor_t>
+    static inline void IndexFromCoorReversed (const coor_t& coor,int &index,const coor_t &dims){
       int nd=dims.size();
       int stride=1;
       index=0;
@@ -35,7 +38,8 @@ namespace Grid{
 	stride=stride*dims[d];
       }
     }
-    static inline void CoorFromIndexReversed (std::vector<int>& coor,int index,const std::vector<int> &dims){
+    template<class coor_t>
+    static inline void CoorFromIndexReversed (coor_t& coor,int index,const coor_t &dims){
       int nd= dims.size();
       coor.resize(nd);
       for(int d=nd-1;d>=0;d--){
diff --git a/Grid/util/Util.h b/Grid/util/Util.h
index 0a6802a0..85c7596f 100644
--- a/Grid/util/Util.h
+++ b/Grid/util/Util.h
@@ -1,5 +1,6 @@
 #ifndef GRID_UTIL_H
 #define GRID_UTIL_H
+#include <Grid/util/Coordinate.h>
 #include <Grid/util/Lexicographic.h>
 #include <Grid/util/Init.h>
 #endif
diff --git a/HMC/Mobius2p1fEOFA.cc b/HMC/Mobius2p1fEOFA.cc
index 997e76ab..35fb19cb 100644
--- a/HMC/Mobius2p1fEOFA.cc
+++ b/HMC/Mobius2p1fEOFA.cc
@@ -50,6 +50,8 @@ namespace Grid{
     typedef typename FermionOperatorD::FermionField FieldD;
     typedef typename FermionOperatorF::FermionField FieldF;
 
+    using OperatorFunction<FieldD>::operator();
+
     RealD   Tolerance;
     RealD   InnerTolerance; //Initial tolerance for inner CG. Defaults to Tolerance but can be changed
     Integer MaxInnerIterations;
@@ -117,7 +119,7 @@ namespace Grid{
       typedef typename FermionOperatorD::GaugeLinkField GaugeLinkFieldD;
 
       GridBase * GridPtrF = SinglePrecGrid4;
-      GridBase * GridPtrD = FermOpD.Umu._grid;
+      GridBase * GridPtrD = FermOpD.Umu.Grid();
       GaugeFieldF     U_f  (GridPtrF);
       GaugeLinkFieldF Umu_f(GridPtrF);
       //      std::cout << " Dim gauge field "<<GridPtrF->Nd()<<std::endl; // 4d
@@ -241,10 +243,10 @@ int main(int argc, char **argv) {
   auto FGrid     = SpaceTimeGrid::makeFiveDimGrid(Ls,GridPtr);
   auto FrbGrid   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,GridPtr);
 
-  std::vector<int> latt  = GridDefaultLatt();
-  std::vector<int> mpi   = GridDefaultMpi();
-  std::vector<int> simdF = GridDefaultSimd(Nd,vComplexF::Nsimd());
-  std::vector<int> simdD = GridDefaultSimd(Nd,vComplexD::Nsimd());
+  Coordinate latt  = GridDefaultLatt();
+  Coordinate mpi   = GridDefaultMpi();
+  Coordinate simdF = GridDefaultSimd(Nd,vComplexF::Nsimd());
+  Coordinate simdD = GridDefaultSimd(Nd,vComplexD::Nsimd());
   auto GridPtrF   = SpaceTimeGrid::makeFourDimGrid(latt,simdF,mpi);
   auto GridRBPtrF = SpaceTimeGrid::makeFourDimRedBlackGrid(GridPtrF);
   auto FGridF     = SpaceTimeGrid::makeFiveDimGrid(Ls,GridPtrF);
@@ -350,7 +352,7 @@ int main(int argc, char **argv) {
     EOFA(Strange_Op_L, Strange_Op_R, 
 	 ActionCG,
 	 ActionCG, ActionCG,
-	 DerivativeCG, DerivativeCG,
+	 DerivativeCG, DerivativeCG, 
 	 OFRp, true);
 #endif
   Level1.push_back(&EOFA);
diff --git a/HMC/Mobius2p1fEOFA_F1.cc b/HMC/Mobius2p1fEOFA_F1.cc
new file mode 100644
index 00000000..3d51b16c
--- /dev/null
+++ b/HMC/Mobius2p1fEOFA_F1.cc
@@ -0,0 +1,446 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: 
+
+Copyright (C) 2015-2016
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Guido Cossu
+Author: David Murphy
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include <Grid/Grid.h>
+
+#ifdef GRID_DEFAULT_PRECISION_DOUBLE
+#define MIXED_PRECISION
+#endif
+
+namespace Grid{ 
+  namespace QCD{
+
+  /*
+   * Need a plan for gauge field update for mixed precision in HMC                      (2x speed up)
+   *    -- Store the single prec action operator.
+   *    -- Clone the gauge field from the operator function argument.
+   *    -- Build the mixed precision operator dynamically from the passed operator and single prec clone.
+   */
+
+  template<class FermionOperatorD, class FermionOperatorF, class SchurOperatorD, class  SchurOperatorF> 
+  class MixedPrecisionConjugateGradientOperatorFunction : public OperatorFunction<typename FermionOperatorD::FermionField> {
+  public:
+    typedef typename FermionOperatorD::FermionField FieldD;
+    typedef typename FermionOperatorF::FermionField FieldF;
+
+    using OperatorFunction<FieldD>::operator();
+
+    RealD   Tolerance;
+    RealD   InnerTolerance; //Initial tolerance for inner CG. Defaults to Tolerance but can be changed
+    Integer MaxInnerIterations;
+    Integer MaxOuterIterations;
+    GridBase* SinglePrecGrid4; //Grid for single-precision fields
+    GridBase* SinglePrecGrid5; //Grid for single-precision fields
+    RealD OuterLoopNormMult; //Stop the outer loop and move to a final double prec solve when the residual is OuterLoopNormMult * Tolerance
+
+    FermionOperatorF &FermOpF;
+    FermionOperatorD &FermOpD;;
+    SchurOperatorF &LinOpF;
+    SchurOperatorD &LinOpD;
+
+    Integer TotalInnerIterations; //Number of inner CG iterations
+    Integer TotalOuterIterations; //Number of restarts
+    Integer TotalFinalStepIterations; //Number of CG iterations in final patch-up step
+
+    MixedPrecisionConjugateGradientOperatorFunction(RealD tol, 
+						    Integer maxinnerit, 
+						    Integer maxouterit, 
+						    GridBase* _sp_grid4, 
+						    GridBase* _sp_grid5, 
+						    FermionOperatorF &_FermOpF,
+						    FermionOperatorD &_FermOpD,
+						    SchurOperatorF   &_LinOpF,
+						    SchurOperatorD   &_LinOpD): 
+      LinOpF(_LinOpF),
+      LinOpD(_LinOpD),
+      FermOpF(_FermOpF),
+      FermOpD(_FermOpD),
+      Tolerance(tol), 
+      InnerTolerance(tol), 
+      MaxInnerIterations(maxinnerit), 
+      MaxOuterIterations(maxouterit), 
+      SinglePrecGrid4(_sp_grid4),
+      SinglePrecGrid5(_sp_grid5),
+      OuterLoopNormMult(100.) 
+    { 
+      /* Debugging instances of objects; references are stored
+      std::cout << GridLogMessage << " Mixed precision CG wrapper LinOpF " <<std::hex<< &LinOpF<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper LinOpD " <<std::hex<< &LinOpD<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper FermOpF " <<std::hex<< &FermOpF<<std::dec <<std::endl;
+      std::cout << GridLogMessage << " Mixed precision CG wrapper FermOpD " <<std::hex<< &FermOpD<<std::dec <<std::endl;
+      */
+    };
+
+    void operator()(LinearOperatorBase<FieldD> &LinOpU, const FieldD &src, FieldD &psi) {
+
+      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() "<<std::endl;
+
+      SchurOperatorD * SchurOpU = static_cast<SchurOperatorD *>(&LinOpU);
+      
+      //      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() FermOpU " <<std::hex<< &(SchurOpU->_Mat)<<std::dec <<std::endl;
+      //      std::cout << GridLogMessage << " Mixed precision CG wrapper operator() FermOpD " <<std::hex<< &(LinOpD._Mat) <<std::dec <<std::endl;
+      // Assumption made in code to extract gauge field
+      // We could avoid storing LinopD reference alltogether ?
+      assert(&(SchurOpU->_Mat)==&(LinOpD._Mat));
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Must snarf a single precision copy of the gauge field in Linop_d argument
+      ////////////////////////////////////////////////////////////////////////////////////
+      typedef typename FermionOperatorF::GaugeField GaugeFieldF;
+      typedef typename FermionOperatorF::GaugeLinkField GaugeLinkFieldF;
+      typedef typename FermionOperatorD::GaugeField GaugeFieldD;
+      typedef typename FermionOperatorD::GaugeLinkField GaugeLinkFieldD;
+
+      GridBase * GridPtrF = SinglePrecGrid4;
+      GridBase * GridPtrD = FermOpD.Umu.Grid();
+      GaugeFieldF     U_f  (GridPtrF);
+      GaugeLinkFieldF Umu_f(GridPtrF);
+      //      std::cout << " Dim gauge field "<<GridPtrF->Nd()<<std::endl; // 4d
+      //      std::cout << " Dim gauge field "<<GridPtrD->Nd()<<std::endl; // 4d
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Moving this to a Clone method of fermion operator would allow to duplicate the 
+      // physics parameters and decrease gauge field copies
+      ////////////////////////////////////////////////////////////////////////////////////
+      GaugeLinkFieldD Umu_d(GridPtrD);
+      for(int mu=0;mu<Nd*2;mu++){ 
+	Umu_d = PeekIndex<LorentzIndex>(FermOpD.Umu, mu);
+	precisionChange(Umu_f,Umu_d);
+	PokeIndex<LorentzIndex>(FermOpF.Umu, Umu_f, mu);
+      }
+      pickCheckerboard(Even,FermOpF.UmuEven,FermOpF.Umu);
+      pickCheckerboard(Odd ,FermOpF.UmuOdd ,FermOpF.Umu);
+
+      ////////////////////////////////////////////////////////////////////////////////////
+      // Make a mixed precision conjugate gradient
+      ////////////////////////////////////////////////////////////////////////////////////
+      MixedPrecisionConjugateGradient<FieldD,FieldF> MPCG(Tolerance,MaxInnerIterations,MaxOuterIterations,SinglePrecGrid5,LinOpF,LinOpD);
+      std::cout << GridLogMessage << "Calling mixed precision Conjugate Gradient" <<std::endl;
+      MPCG(src,psi);
+    }
+  };
+}};
+
+int main(int argc, char **argv) {
+  using namespace Grid;
+  using namespace Grid::QCD;
+
+  Grid_init(&argc, &argv);
+  int threads = GridThread::GetThreads();
+  // here make a routine to print all the relevant information on the run
+  std::cout << GridLogMessage << "Grid is setup to use " << threads << " threads" << std::endl;
+
+   // Typedefs to simplify notation
+  typedef WilsonImplR FermionImplPolicy;
+  typedef MobiusFermionR FermionAction;
+  typedef MobiusFermionF FermionActionF;
+  typedef MobiusEOFAFermionR FermionEOFAAction;
+  typedef MobiusEOFAFermionF FermionEOFAActionF;
+  typedef typename FermionAction::FermionField FermionField;
+  typedef typename FermionActionF::FermionField FermionFieldF;
+
+  typedef Grid::XmlReader       Serialiser;
+  
+  //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
+
+  //  typedef GenericHMCRunner<LeapFrog> HMCWrapper; 
+  //  typedef GenericHMCRunner<MinimumNorm2> HMCWrapper; 
+  typedef GenericHMCRunner<ForceGradient> HMCWrapper; 
+
+  HMCparameters HMCparams;
+  {
+    XmlReader  HMCrd("HMCparameters.xml");
+    read(HMCrd,"HMCparameters",HMCparams);
+    std::cout << GridLogMessage<< HMCparams <<std::endl;
+  }
+  HMCWrapper TheHMC(HMCparams);
+
+  // Grid from the command line arguments --grid and --mpi
+  TheHMC.Resources.AddFourDimGrid("gauge"); // use default simd lanes decomposition
+  
+  CheckpointerParameters CPparams;
+  CPparams.config_prefix = "ckpoint_EODWF_lat";
+  CPparams.rng_prefix    = "ckpoint_EODWF_rng";
+  CPparams.saveInterval  = 1;
+  CPparams.format        = "IEEE64BIG";
+  TheHMC.Resources.LoadNerscCheckpointer(CPparams);
+
+  RNGModuleParameters RNGpar;
+  RNGpar.serial_seeds = "1 2 3 4 5";
+  RNGpar.parallel_seeds = "6 7 8 9 10";
+  TheHMC.Resources.SetRNGSeeds(RNGpar);
+
+  // Construct observables
+  // here there is too much indirection 
+  typedef PlaquetteMod<HMCWrapper::ImplPolicy> PlaqObs;
+  TheHMC.Resources.AddObservable<PlaqObs>();
+  //////////////////////////////////////////////
+
+  const int Ls      = 12;
+  Real beta         = 2.31;
+  Real light_mass   = 0.002144;
+  Real strange_mass = 0.02144;
+  Real pv_mass      = 1.0;
+  RealD M5  = 1.8;
+  RealD b   = 1.5; 
+  RealD c   = 0.5;
+
+  // Copied from paper
+  std::vector<Real> hasenbusch({ 0.005, 0.017, 0.07, 0.18, 0.45 }); // Paper values from F1 incorrect run
+  //  std::vector<Real> hasenbusch({ 0.004, 0.016, 0.07, 0.18, 0.45 }); // Paper values from F1 incorrect run 120-130
+  //  std::vector<Real> hasenbusch({ 0.004, 0.015, 0.07, 0.18, 0.45 }); // Paper values from F1 incorrect run 110-120
+  //  std::vector<Real> hasenbusch({ 0.005, 0.017, 0.07, 0.18, 0.45 }); // Paper values from F1 incorrect run
+  //  std::vector<Real> hasenbusch({  0.01, 0.07, 0.18, 0.45 }); // Experiment?
+
+  ///////////////////////////////////////////////////////////////////////////////////////////////
+  //Bad choices with large dH. Equalising force L2 norm was not wise.
+  ///////////////////////////////////////////////////////////////////////////////////////////////
+  //std::vector<Real> hasenbusch({ 0.03, 0.2, 0.3, 0.5, 0.8 }); 
+  //std::vector<Real> hasenbusch({ 0.05, 0.2, 0.4, 0.6, 0.8 });
+
+  auto GridPtr   = TheHMC.Resources.GetCartesian();
+  auto GridRBPtr = TheHMC.Resources.GetRBCartesian();
+  auto FGrid     = SpaceTimeGrid::makeFiveDimGrid(Ls,GridPtr);
+  auto FrbGrid   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,GridPtr);
+
+  Coordinate latt  = GridDefaultLatt();
+  Coordinate mpi   = GridDefaultMpi();
+  Coordinate simdF = GridDefaultSimd(Nd,vComplexF::Nsimd());
+  Coordinate simdD = GridDefaultSimd(Nd,vComplexD::Nsimd());
+  auto GridPtrF   = SpaceTimeGrid::makeFourDimGrid(latt,simdF,mpi);
+  auto GridRBPtrF = SpaceTimeGrid::makeFourDimRedBlackGrid(GridPtrF);
+  auto FGridF     = SpaceTimeGrid::makeFiveDimGrid(Ls,GridPtrF);
+  auto FrbGridF   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,GridPtrF);
+
+  IwasakiGaugeActionR GaugeAction(beta);
+
+  // temporarily need a gauge field
+  LatticeGaugeField U(GridPtr);
+  LatticeGaugeFieldF UF(GridPtrF);
+
+  // These lines are unecessary if BC are all periodic
+  std::vector<Complex> boundary = {1,1,1,-1};
+  FermionAction::ImplParams Params(boundary);
+  FermionActionF::ImplParams ParamsF(boundary);
+  
+  double ActionStoppingCondition     = 1e-10;
+  double DerivativeStoppingCondition = 1e-7;
+  double MaxCGIterations = 30000;
+
+  ////////////////////////////////////
+  // Collect actions
+  ////////////////////////////////////
+  ActionLevel<HMCWrapper::Field> Level1(1);
+  ActionLevel<HMCWrapper::Field> Level2(16);
+
+  ////////////////////////////////////
+  // Strange action
+  ////////////////////////////////////
+  typedef SchurDiagMooeeOperator<FermionActionF,FermionFieldF> LinearOperatorF;
+  typedef SchurDiagMooeeOperator<FermionAction ,FermionField > LinearOperatorD;
+  typedef SchurDiagMooeeOperator<FermionEOFAActionF,FermionFieldF> LinearOperatorEOFAF;
+  typedef SchurDiagMooeeOperator<FermionEOFAAction ,FermionField > LinearOperatorEOFAD;
+
+  typedef MixedPrecisionConjugateGradientOperatorFunction<MobiusFermionD,MobiusFermionF,LinearOperatorD,LinearOperatorF> MxPCG;
+  typedef MixedPrecisionConjugateGradientOperatorFunction<MobiusEOFAFermionD,MobiusEOFAFermionF,LinearOperatorEOFAD,LinearOperatorEOFAF> MxPCG_EOFA;
+
+  // DJM: setup for EOFA ratio (Mobius)
+  OneFlavourRationalParams OFRp;
+  OFRp.lo       = 0.1; // How do I know this on F1?
+  OFRp.hi       = 25.0;
+  OFRp.MaxIter  = 10000;
+  OFRp.tolerance= 1.0e-9;
+  OFRp.degree   = 12;
+  OFRp.precision= 50;
+
+  
+  MobiusEOFAFermionR Strange_Op_L (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionF Strange_Op_LF(UF, *FGridF, *FrbGridF, *GridPtrF, *GridRBPtrF, strange_mass, strange_mass, pv_mass, 0.0, -1, M5, b, c);
+  MobiusEOFAFermionR Strange_Op_R (U , *FGrid , *FrbGrid , *GridPtr , *GridRBPtr , pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
+  MobiusEOFAFermionF Strange_Op_RF(UF, *FGridF, *FrbGridF, *GridPtrF, *GridRBPtrF, pv_mass, strange_mass,      pv_mass, -1.0, 1, M5, b, c);
+
+  ConjugateGradient<FermionField>      ActionCG(ActionStoppingCondition,MaxCGIterations);
+  ConjugateGradient<FermionField>  DerivativeCG(DerivativeStoppingCondition,MaxCGIterations);
+#ifdef MIXED_PRECISION
+  const int MX_inner = 5000;
+
+  // Mixed precision EOFA
+  LinearOperatorEOFAD Strange_LinOp_L (Strange_Op_L);
+  LinearOperatorEOFAD Strange_LinOp_R (Strange_Op_R);
+  LinearOperatorEOFAF Strange_LinOp_LF(Strange_Op_LF);
+  LinearOperatorEOFAF Strange_LinOp_RF(Strange_Op_RF);
+
+  MxPCG_EOFA ActionCGL(ActionStoppingCondition,
+		       MX_inner,
+		       MaxCGIterations,
+		       GridPtrF,
+		       FrbGridF,
+		       Strange_Op_LF,Strange_Op_L,
+		       Strange_LinOp_LF,Strange_LinOp_L);
+
+  MxPCG_EOFA DerivativeCGL(DerivativeStoppingCondition,
+			   MX_inner,
+			   MaxCGIterations,
+			   GridPtrF,
+			   FrbGridF,
+			   Strange_Op_LF,Strange_Op_L,
+			   Strange_LinOp_LF,Strange_LinOp_L);
+  
+  MxPCG_EOFA ActionCGR(ActionStoppingCondition,
+		       MX_inner,
+		       MaxCGIterations,
+		       GridPtrF,
+		       FrbGridF,
+		       Strange_Op_RF,Strange_Op_R,
+		       Strange_LinOp_RF,Strange_LinOp_R);
+  
+  MxPCG_EOFA DerivativeCGR(DerivativeStoppingCondition,
+			   MX_inner,
+			   MaxCGIterations,
+			   GridPtrF,
+			   FrbGridF,
+			   Strange_Op_RF,Strange_Op_R,
+			   Strange_LinOp_RF,Strange_LinOp_R);
+
+  ExactOneFlavourRatioPseudoFermionAction<FermionImplPolicy> 
+    EOFA(Strange_Op_L, Strange_Op_R, 
+	 ActionCG, 
+	 ActionCGL, ActionCGR,
+	 DerivativeCGL, DerivativeCGR,
+	 OFRp, true);
+#else
+  ExactOneFlavourRatioPseudoFermionAction<FermionImplPolicy> 
+    EOFA(Strange_Op_L, Strange_Op_R, 
+	 ActionCG, 
+	 ActionCG, ActionCG,
+	 ActionCG, ActionCG,
+	 //	 DerivativeCG, DerivativeCG,
+	 OFRp, true);
+#endif
+  Level1.push_back(&EOFA);
+
+  ////////////////////////////////////
+  // up down action
+  ////////////////////////////////////
+  std::vector<Real> light_den;
+  std::vector<Real> light_num;
+
+  int n_hasenbusch = hasenbusch.size();
+  light_den.push_back(light_mass);
+  for(int h=0;h<n_hasenbusch;h++){
+    light_den.push_back(hasenbusch[h]);
+    light_num.push_back(hasenbusch[h]);
+  }
+  light_num.push_back(pv_mass);
+
+  //////////////////////////////////////////////////////////////
+  // Forced to replicate the MxPCG and DenominatorsF etc.. because
+  // there is no convenient way to "Clone" physics params from double op
+  // into single op for any operator pair.
+  // Same issue prevents using MxPCG in the Heatbath step
+  //////////////////////////////////////////////////////////////
+  std::vector<FermionAction *> Numerators;
+  std::vector<FermionAction *> Denominators;
+  std::vector<TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy> *> Quotients;
+  std::vector<MxPCG *> ActionMPCG;
+  std::vector<MxPCG *> MPCG;
+  std::vector<FermionActionF *> DenominatorsF;
+  std::vector<LinearOperatorD *> LinOpD;
+  std::vector<LinearOperatorF *> LinOpF; 
+
+  for(int h=0;h<n_hasenbusch+1;h++){
+
+    std::cout << GridLogMessage << " 2f quotient Action  "<< light_num[h] << " / " << light_den[h]<< std::endl;
+
+    Numerators.push_back  (new FermionAction(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,light_num[h],M5,b,c, Params));
+    Denominators.push_back(new FermionAction(U,*FGrid,*FrbGrid,*GridPtr,*GridRBPtr,light_den[h],M5,b,c, Params));
+
+#ifdef MIXED_PRECISION
+    ////////////////////////////////////////////////////////////////////////////
+    // Mixed precision CG for 2f force
+    ////////////////////////////////////////////////////////////////////////////
+    double DerivativeStoppingConditionLoose = 3e-7;
+
+    DenominatorsF.push_back(new FermionActionF(UF,*FGridF,*FrbGridF,*GridPtrF,*GridRBPtrF,light_den[h],M5,b,c, ParamsF));
+    LinOpD.push_back(new LinearOperatorD(*Denominators[h]));
+    LinOpF.push_back(new LinearOperatorF(*DenominatorsF[h]));
+
+    double conv  = DerivativeStoppingCondition;
+    if (h<3) conv= DerivativeStoppingConditionLoose; // Relax on first two hasenbusch factors
+    MPCG.push_back(new MxPCG(conv,
+			     MX_inner,
+			     MaxCGIterations,
+			     GridPtrF,
+			     FrbGridF,
+			     *DenominatorsF[h],*Denominators[h],
+			     *LinOpF[h], *LinOpD[h]) );
+
+    ActionMPCG.push_back(new MxPCG(ActionStoppingCondition,
+				   MX_inner,
+				   MaxCGIterations,
+				   GridPtrF,
+				   FrbGridF,
+				   *DenominatorsF[h],*Denominators[h],
+				   *LinOpF[h], *LinOpD[h]) );
+
+    // Heatbath not mixed yet. As inverts numerators not so important as raised mass.
+    Quotients.push_back (new TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],*MPCG[h],*ActionMPCG[h],ActionCG));
+#else
+    ////////////////////////////////////////////////////////////////////////////
+    // Standard CG for 2f force
+    ////////////////////////////////////////////////////////////////////////////
+    Quotients.push_back   (new TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy>(*Numerators[h],*Denominators[h],DerivativeCG,ActionCG));
+#endif
+
+  }
+
+  for(int h=0;h<n_hasenbusch+1;h++){
+    Level1.push_back(Quotients[h]);
+  }
+
+  /////////////////////////////////////////////////////////////
+  // Gauge action
+  /////////////////////////////////////////////////////////////
+  Level2.push_back(&GaugeAction);
+  TheHMC.TheAction.push_back(Level1);
+  TheHMC.TheAction.push_back(Level2);
+  std::cout << GridLogMessage << " Action complete "<< std::endl;
+
+  /////////////////////////////////////////////////////////////
+  // HMC parameters are serialisable
+
+  std::cout << GridLogMessage << " Running the HMC "<< std::endl;
+  TheHMC.Run();  // no smearing
+
+  Grid_finalize();
+} // main
+
+
+
diff --git a/Hadrons/A2AMatrix.hpp b/Hadrons/A2AMatrix.hpp
index 29b30a7a..033f4a3c 100644
--- a/Hadrons/A2AMatrix.hpp
+++ b/Hadrons/A2AMatrix.hpp
@@ -167,10 +167,12 @@ public:
     template <typename C, typename MatLeft, typename MatRight>
     static inline void accTrMul(C &acc, const MatLeft &a, const MatRight &b)
     {
-        if ((MatLeft::Options == Eigen::RowMajor) and
-            (MatRight::Options == Eigen::ColMajor))
+        const int RowMajor = Eigen::RowMajor;
+        const int ColMajor = Eigen::ColMajor;
+        if ((MatLeft::Options  == RowMajor) and
+            (MatRight::Options == ColMajor))
         {
-            parallel_for (unsigned int r = 0; r < a.rows(); ++r)
+  	  thread_for(r,a.rows(),
             {
                 C tmp;
 #ifdef USE_MKL
@@ -178,15 +180,15 @@ public:
 #else
                 tmp = a.row(r).conjugate().dot(b.col(r));
 #endif
-                parallel_critical
+                thread_critical
                 {
                     acc += tmp;
                 }
-            }
+            });
         }
         else
-        {
-            parallel_for (unsigned int c = 0; c < a.cols(); ++c)
+	  {
+            thread_for(c,a.cols(),
             {
                 C tmp;
 #ifdef USE_MKL 
@@ -194,11 +196,11 @@ public:
 #else
                 tmp = a.col(c).conjugate().dot(b.row(c));
 #endif
-                parallel_critical
+                thread_critical
                 {
                     acc += tmp;
                 }
-            }
+            });
         }
     }
 
@@ -218,18 +220,20 @@ public:
                            const Mat<ComplexD, Opts...> &b)
     {
         static const ComplexD one(1., 0.), zero(0., 0.);
+        const int RowMajor = Eigen::RowMajor;
+        const int ColMajor = Eigen::ColMajor;
 
         if ((res.rows() != a.rows()) or (res.cols() != b.cols()))
         {
             res.resize(a.rows(), b.cols());
         }
-        if (Mat<ComplexD, Opts...>::Options == Eigen::RowMajor)
+        if (Mat<ComplexD, Opts...>::Options == RowMajor)
         {
             cblas_zgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, a.rows(), b.cols(),
                         a.cols(), &one, a.data(), a.cols(), b.data(), b.cols(), &zero,
                         res.data(), res.cols());
         }
-        else if (Mat<ComplexD, Opts...>::Options == Eigen::ColMajor)
+        else if (Mat<ComplexD, Opts...>::Options == ColMajor)
         {
             cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, a.rows(), b.cols(),
                         a.cols(), &one, a.data(), a.rows(), b.data(), b.rows(), &zero,
@@ -243,18 +247,20 @@ public:
                            const Mat<ComplexF, Opts...> &b)
     {
         static const ComplexF one(1., 0.), zero(0., 0.);
+        const int RowMajor = Eigen::RowMajor;
+        const int ColMajor = Eigen::ColMajor;
 
         if ((res.rows() != a.rows()) or (res.cols() != b.cols()))
         {
             res.resize(a.rows(), b.cols());
         }
-        if (Mat<ComplexF, Opts...>::Options == Eigen::RowMajor)
+        if (Mat<ComplexF, Opts...>::Options == RowMajor)
         {
             cblas_cgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, a.rows(), b.cols(),
                         a.cols(), &one, a.data(), a.cols(), b.data(), b.cols(), &zero,
                         res.data(), res.cols());
         }
-        else if (Mat<ComplexF, Opts...>::Options == Eigen::ColMajor)
+        else if (Mat<ComplexF, Opts...>::Options == ColMajor)
         {
             cblas_cgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, a.rows(), b.cols(),
                         a.cols(), &one, a.data(), a.rows(), b.data(), b.rows(), &zero,
@@ -281,22 +287,25 @@ private:
                                     unsigned int &bInc, const unsigned int aRow, 
                                     const MatLeft &a, const MatRight &b)
     {
-        if (MatLeft::Options == Eigen::RowMajor)
+        const int RowMajor = Eigen::RowMajor;
+        const int ColMajor = Eigen::ColMajor;
+
+        if (MatLeft::Options == RowMajor)
         {
             aPt  = a.data() + aRow*a.cols();
             aInc = 1;
         }
-        else if (MatLeft::Options == Eigen::ColMajor)
+        else if (MatLeft::Options == ColMajor)
         {
             aPt  = a.data() + aRow;
             aInc = a.rows();
         }
-        if (MatRight::Options == Eigen::RowMajor)
+        if (MatRight::Options == RowMajor)
         {
             bPt  = b.data() + aRow;
             bInc = b.cols();
         }
-        else if (MatRight::Options == Eigen::ColMajor)
+        else if (MatRight::Options == ColMajor)
         {
             bPt  = b.data() + aRow*b.rows();
             bInc = 1;
@@ -309,22 +318,24 @@ private:
                                     unsigned int &bInc, const unsigned int aCol, 
                                     const MatLeft &a, const MatRight &b)
     {
-        if (MatLeft::Options == Eigen::RowMajor)
+        const int RowMajor = Eigen::RowMajor;
+        const int ColMajor = Eigen::ColMajor;
+        if (MatLeft::Options == RowMajor)
         {
             aPt  = a.data() + aCol;
             aInc = a.cols();
         }
-        else if (MatLeft::Options == Eigen::ColMajor)
+        else if (MatLeft::Options == ColMajor)
         {
             aPt  = a.data() + aCol*a.rows();
             aInc = 1;
         }
-        if (MatRight::Options == Eigen::RowMajor)
+        if (MatRight::Options == RowMajor)
         {
             bPt  = b.data() + aCol*b.cols();
             bInc = 1;
         }
-        else if (MatRight::Options == Eigen::ColMajor)
+        else if (MatRight::Options == ColMajor)
         {
             bPt  = b.data() + aCol;
             bInc = b.rows();
@@ -466,7 +477,7 @@ void A2AMatrixIo<T>::saveBlock(const T *data,
                          block  = {1, 1, 1}; 
     H5NS::DataSpace      memspace(count.size(), count.data()), dataspace;
     H5NS::DataSet        dataset;
-    size_t               shift;
+    //    size_t               shift;
 
     push(reader, dataname_);
     auto &group = reader.getGroup();
@@ -646,14 +657,15 @@ void A2AMatrixBlockComputation<T, Field, MetadataType, TIo>
             bytes    += kernel.bytes(N_iii, N_jjj);
 
             START_TIMER("cache copy");
-            parallel_for_nest5(int e =0;e<next_;e++)
-            for(int s =0;s< nstr_;s++)
-            for(int t =0;t< nt_;t++)
-            for(int iii=0;iii< N_iii;iii++)
-            for(int jjj=0;jjj< N_jjj;jjj++)
-            {
+            thread_for_collapse( 5,e,next_,{
+              for(int s =0;s< nstr_;s++)
+              for(int t =0;t< nt_;t++)
+              for(int iii=0;iii< N_iii;iii++)
+              for(int jjj=0;jjj< N_jjj;jjj++)
+              {
                 mBlock(e,s,t,ii+iii,jj+jjj) = mCacheBlock(e,s,t,iii,jjj);
-            }
+              }
+            });
             STOP_TIMER("cache copy");
         }
 
diff --git a/Hadrons/A2AVectors.hpp b/Hadrons/A2AVectors.hpp
index ecc1497f..7cbd19ef 100644
--- a/Hadrons/A2AVectors.hpp
+++ b/Hadrons/A2AVectors.hpp
@@ -128,7 +128,7 @@ template <typename FImpl>
 void A2AVectorsSchurDiagTwo<FImpl>::makeLowModeV(FermionField &vout, const FermionField &evec, const Real &eval)
 {
     src_o_ = evec;
-    src_o_.checkerboard = Odd;
+    src_o_.Checkerboard() = Odd;
     pickCheckerboard(Even, sol_e_, vout);
     pickCheckerboard(Odd, sol_o_, vout);
 
@@ -136,25 +136,25 @@ void A2AVectorsSchurDiagTwo<FImpl>::makeLowModeV(FermionField &vout, const Fermi
     // v_ie = -(1/eval_i) * MeeInv Meo MooInv evec_i
     /////////////////////////////////////////////////////
     action_.MooeeInv(src_o_, tmp_);
-    assert(tmp_.checkerboard == Odd);
+    assert(tmp_.Checkerboard() == Odd);
     action_.Meooe(tmp_, sol_e_);
-    assert(sol_e_.checkerboard == Even);
+    assert(sol_e_.Checkerboard() == Even);
     action_.MooeeInv(sol_e_, tmp_);
-    assert(tmp_.checkerboard == Even);
+    assert(tmp_.Checkerboard() == Even);
     sol_e_ = (-1.0 / eval) * tmp_;
-    assert(sol_e_.checkerboard == Even);
+    assert(sol_e_.Checkerboard() == Even);
 
     /////////////////////////////////////////////////////
     // v_io = (1/eval_i) * MooInv evec_i
     /////////////////////////////////////////////////////
     action_.MooeeInv(src_o_, tmp_);
-    assert(tmp_.checkerboard == Odd);
+    assert(tmp_.Checkerboard() == Odd);
     sol_o_ = (1.0 / eval) * tmp_;
-    assert(sol_o_.checkerboard == Odd);
+    assert(sol_o_.Checkerboard() == Odd);
     setCheckerboard(vout, sol_e_);
-    assert(sol_e_.checkerboard == Even);
+    assert(sol_e_.Checkerboard() == Even);
     setCheckerboard(vout, sol_o_);
-    assert(sol_o_.checkerboard == Odd);
+    assert(sol_o_.Checkerboard() == Odd);
 }
 
 template <typename FImpl>
@@ -168,7 +168,7 @@ template <typename FImpl>
 void A2AVectorsSchurDiagTwo<FImpl>::makeLowModeW(FermionField &wout, const FermionField &evec, const Real &eval)
 {
     src_o_ = evec;
-    src_o_.checkerboard = Odd;
+    src_o_.Checkerboard() = Odd;
     pickCheckerboard(Even, sol_e_, wout);
     pickCheckerboard(Odd, sol_o_, wout);
 
@@ -176,22 +176,22 @@ void A2AVectorsSchurDiagTwo<FImpl>::makeLowModeW(FermionField &wout, const Fermi
     // w_ie = - MeeInvDag MoeDag Doo evec_i
     /////////////////////////////////////////////////////
     op_.Mpc(src_o_, tmp_);
-    assert(tmp_.checkerboard == Odd);
+    assert(tmp_.Checkerboard() == Odd);
     action_.MeooeDag(tmp_, sol_e_);
-    assert(sol_e_.checkerboard == Even);
+    assert(sol_e_.Checkerboard() == Even);
     action_.MooeeInvDag(sol_e_, tmp_);
-    assert(tmp_.checkerboard == Even);
+    assert(tmp_.Checkerboard() == Even);
     sol_e_ = (-1.0) * tmp_;
 
     /////////////////////////////////////////////////////
     // w_io = Doo evec_i
     /////////////////////////////////////////////////////
     op_.Mpc(src_o_, sol_o_);
-    assert(sol_o_.checkerboard == Odd);
+    assert(sol_o_.Checkerboard() == Odd);
     setCheckerboard(wout, sol_e_);
-    assert(sol_e_.checkerboard == Even);
+    assert(sol_e_.Checkerboard() == Even);
     setCheckerboard(wout, sol_o_);
-    assert(sol_o_.checkerboard == Odd);
+    assert(sol_o_.Checkerboard() == Odd);
 }
 
 template <typename FImpl>
@@ -217,7 +217,7 @@ void A2AVectorsSchurDiagTwo<FImpl>::makeHighModeV5D(FermionField &vout_4d,
                                                     FermionField &vout_5d, 
                                                     const FermionField &noise)
 {
-    if (noise._grid->Dimensions() == fGrid_->Dimensions() - 1)
+    if (noise.Grid()->Dimensions() == fGrid_->Dimensions() - 1)
     {
         action_.ImportPhysicalFermionSource(noise, tmp5_);
     }
@@ -241,7 +241,7 @@ void A2AVectorsSchurDiagTwo<FImpl>::makeHighModeW5D(FermionField &wout_4d,
                                                     FermionField &wout_5d, 
                                                     const FermionField &noise)
 {
-    if (noise._grid->Dimensions() == fGrid_->Dimensions() - 1)
+    if (noise.Grid()->Dimensions() == fGrid_->Dimensions() - 1)
     {
         action_.ImportUnphysicalFermion(noise, wout_5d);
         wout_4d = noise;
@@ -261,7 +261,7 @@ void A2AVectorsIo::write(const std::string fileStem, std::vector<Field> &vec,
                          const bool multiFile, const int trajectory)
 {
     Record       record;
-    GridBase     *grid = vec[0]._grid;
+    GridBase     *grid = vec[0].Grid();
     ScidacWriter binWriter(grid->IsBoss());
     std::string  filename = vecFilename(fileStem, trajectory, multiFile);
 
diff --git a/Hadrons/Application.cc b/Hadrons/Application.cc
index e4264f84..ad21e199 100644
--- a/Hadrons/Application.cc
+++ b/Hadrons/Application.cc
@@ -31,7 +31,6 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #include <Hadrons/Modules.hpp>
 
 using namespace Grid;
-using namespace QCD;
 using namespace Hadrons;
 
 #define BIG_SEP "================"
diff --git a/Hadrons/Archive/Modules/ScalarVP.cc b/Hadrons/Archive/Modules/ScalarVP.cc
index f2696a19..f4f7e59b 100644
--- a/Hadrons/Archive/Modules/ScalarVP.cc
+++ b/Hadrons/Archive/Modules/ScalarVP.cc
@@ -269,7 +269,7 @@ void TScalarVP::execute(void)
     if (!par().output.empty())
     {
         outputData.projection.resize(par().outputMom.size());
-        outputData.lattice_size = env().getGrid()->_fdimensions;
+        outputData.lattice_size = env().getGrid()->FullDimensions().toVector();
         outputData.mass = static_cast<TChargedProp *>(vm().getModule(par().scalarProp))->par().mass;
         outputData.charge = q;
         for (unsigned int i_p = 0; i_p < par().outputMom.size(); ++i_p)
@@ -476,7 +476,7 @@ void TScalarVP::makeCaches(void)
     {
         LOG(Message) << "Caching phases for momentum projections..."
                      << std::endl;
-        std::vector<int> &l = env().getGrid()->_fdimensions;
+        auto l = env().getGrid()->FullDimensions();
         Complex          ci(0.0,1.0);
 
         // Calculate phase factors
@@ -484,7 +484,7 @@ void TScalarVP::makeCaches(void)
         {
             std::vector<int> mom = strToVec<int>(par().outputMom[i_p]);
             auto &momph_ip = envGet(ScalarField, momPhaseName_[i_p]);
-            momph_ip = zero;
+            momph_ip = Zero();
             for (unsigned int j = 0; j < env().getNd()-1; ++j)
             {
                 Real twoPiL = M_PI*2./l[j];
@@ -541,7 +541,7 @@ void TScalarVP::momD1(ScalarField &s, FFT &fft)
     envGetTmp(ScalarField, result);
     envGetTmp(ScalarField, Amu);
 
-    result = zero;
+    result = Zero();
     for (unsigned int mu = 0; mu < env().getNd(); ++mu)
     {
         Amu = peekLorentz(A, mu);
diff --git a/Hadrons/Archive/Modules/VPCounterTerms.cc b/Hadrons/Archive/Modules/VPCounterTerms.cc
index 4fa90d6a..c324ed5e 100644
--- a/Hadrons/Archive/Modules/VPCounterTerms.cc
+++ b/Hadrons/Archive/Modules/VPCounterTerms.cc
@@ -115,10 +115,10 @@ void TVPCounterTerms::execute(void)
 
     // Phases and hat{p}^2
     auto &phatsq = envGet(ScalarField, phatsqName_);
-    std::vector<int> &l = env().getGrid()->_fdimensions;
+    Coordinate l = env().getGrid()->FullDimensions();
     
     LOG(Message) << "Calculating shift phases..." << std::endl;
-    phatsq = zero;
+    phatsq = Zero();
     for (unsigned int mu = 0; mu < env().getNd(); ++mu)
     {
         Real    twoPiL = M_PI*2./l[mu];
@@ -156,7 +156,7 @@ void TVPCounterTerms::execute(void)
     if (!par().output.empty())
     {
         outputData.projection.resize(par().outputMom.size());
-        outputData.lattice_size = env().getGrid()->_fdimensions;
+        outputData.lattice_size = env().getGrid()->FullDimensions().toVector();
         outputData.mass = par().mass;
         for (unsigned int i_p = 0; i_p < par().outputMom.size(); ++i_p)
         {
@@ -172,7 +172,7 @@ void TVPCounterTerms::execute(void)
             }
             // Calculate phase factors
             auto &momph_ip = envGet(ScalarField, momPhaseName_[i_p]);
-            momph_ip = zero;
+            momph_ip = Zero();
             for (unsigned int j = 0; j < env().getNd()-1; ++j)
             {
                 Real twoPiL = M_PI*2./l[j];
diff --git a/Hadrons/Archive/Modules/WardIdentity.hpp b/Hadrons/Archive/Modules/WardIdentity.hpp
index 34819afc..e7732fc0 100644
--- a/Hadrons/Archive/Modules/WardIdentity.hpp
+++ b/Hadrons/Archive/Modules/WardIdentity.hpp
@@ -146,7 +146,7 @@ void TWardIdentity<FImpl>::execute(void)
     // Compute D_mu V_mu, D here is backward derivative.
     envGetTmp(PropagatorField, tmp);
     envGetTmp(PropagatorField, vector_WI);
-    vector_WI    = zero;
+    vector_WI    = Zero();
     for (unsigned int mu = 0; mu < Nd; ++mu)
     {
         act.ContractConservedCurrent(q, q, tmp, Current::Vector, mu);
@@ -167,7 +167,7 @@ void TWardIdentity<FImpl>::execute(void)
         std::vector<TComplex> axial_buf;
 
         // Compute <P|D_mu A_mu>, D is backwards derivative.
-        axial_defect = zero;
+        axial_defect = Zero();
         for (unsigned int mu = 0; mu < Nd; ++mu)
         {
             act.ContractConservedCurrent(q, q, tmp, Current::Axial, mu);
@@ -175,8 +175,8 @@ void TWardIdentity<FImpl>::execute(void)
             axial_defect += trace(g5*tmp);
         }
 
-        // Get <P|J5q> for 5D (zero for 4D) and <P|P>.
-        PJ5q = zero;
+        // Get <P|J5q> for 5D (Zero(); for 4D) and <P|P>.
+        PJ5q = Zero();
         if (Ls_ > 1)
         {
             // <P|P>
diff --git a/Hadrons/Archive/Modules/WeakHamiltonian.hpp b/Hadrons/Archive/Modules/WeakHamiltonian.hpp
index a8916325..d047a5fb 100644
--- a/Hadrons/Archive/Modules/WeakHamiltonian.hpp
+++ b/Hadrons/Archive/Modules/WeakHamiltonian.hpp
@@ -56,7 +56,7 @@ for (unsigned int t = 0; t < buf.size(); ++t)\
 
 //// Contraction of mu index: use 'mu' variable in exp.
 #define SUM_MU(buf,exp)\
-buf = zero;\
+   buf = Zero();				\
 for (unsigned int mu = 0; mu < ndim; ++mu)\
 {\
     buf += exp;\
diff --git a/Hadrons/DilutedNoise.hpp b/Hadrons/DilutedNoise.hpp
index b91dbc40..cd0c3c60 100644
--- a/Hadrons/DilutedNoise.hpp
+++ b/Hadrons/DilutedNoise.hpp
@@ -165,7 +165,7 @@ TimeDilutedSpinColorDiagonalNoise<FImpl>::
 TimeDilutedSpinColorDiagonalNoise(GridCartesian *g)
 : DilutedNoise<FImpl>(g)
 {
-    nt_ = this->getGrid()->GlobalDimensions().back();
+    nt_ = this->getGrid()->GlobalDimensions().size();
     this->resize(nt_*Ns*FImpl::Dimension);
 }
 
@@ -192,11 +192,11 @@ void TimeDilutedSpinColorDiagonalNoise<FImpl>::generateNoise(GridParallelRNG &rn
         etaCut = where((tLat == t), eta, 0.*eta);
         for (unsigned int s = 0; s < Ns; ++s)
         {
-            etas = zero;
-            pokeSpin(etas, etaCut, s);
+	    etas = Zero();
+	    pokeSpin(etas, etaCut, s);
             for (unsigned int c = 0; c < nc; ++c)
             {
-                noise[i] = zero;
+  	        noise[i] = Zero();
                 pokeColour(noise[i], etas, c);
                 i++;
             }
@@ -233,11 +233,11 @@ void FullVolumeSpinColorDiagonalNoise<FImpl>::generateNoise(GridParallelRNG &rng
     {
         for (unsigned int s = 0; s < Ns; ++s)
         {
-            etas = zero;
+  	    etas = Zero();
             pokeSpin(etas, eta, s);
             for (unsigned int c = 0; c < nc; ++c)
             {
-                noise[i] = zero;
+	        noise[i] = Zero();
                 pokeColour(noise[i], etas, c);
                 i++;
             }
diff --git a/Hadrons/EigenPack.hpp b/Hadrons/EigenPack.hpp
index 1c7497f2..2e7c62c9 100644
--- a/Hadrons/EigenPack.hpp
+++ b/Hadrons/EigenPack.hpp
@@ -180,7 +180,7 @@ namespace EigenPackIo
                           const unsigned int size, bool multiFile, 
                           GridBase *gridIo = nullptr)
     {
-        GridBase             *grid = evec[0]._grid;
+        GridBase             *grid = evec[0].Grid();
         std::unique_ptr<TIo> ioBuf{nullptr}; 
         std::unique_ptr<T>   testBuf{nullptr};
         ScidacWriter         binWriter(grid->IsBoss());
diff --git a/Hadrons/Environment.cc b/Hadrons/Environment.cc
index 540b3620..2287a7fd 100644
--- a/Hadrons/Environment.cc
+++ b/Hadrons/Environment.cc
@@ -31,7 +31,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #include <Hadrons/ModuleFactory.hpp>
 
 using namespace Grid;
-using namespace QCD;
+ 
 using namespace Hadrons;
 
 #define ERROR_NO_ADDRESS(address)\
@@ -43,7 +43,7 @@ HADRONS_ERROR_REF(ObjectDefinition, "no object with address " + std::to_string(a
 // constructor /////////////////////////////////////////////////////////////////
 Environment::Environment(void)
 {
-    dim_ = GridDefaultLatt();
+    dim_ = GridDefaultLatt().toVector();
     nd_  = dim_.size();
     vol_ = 1.;
     for (auto d: dim_)
diff --git a/Hadrons/GeneticScheduler.hpp b/Hadrons/GeneticScheduler.hpp
index f36a0f50..178b8301 100644
--- a/Hadrons/GeneticScheduler.hpp
+++ b/Hadrons/GeneticScheduler.hpp
@@ -133,7 +133,6 @@ void GeneticScheduler<V, T>::nextGeneration(void)
     //LOG(Debug) << "Starting population:\n" << *this << std::endl;
     
     // random mutations
-    //PARALLEL_FOR_LOOP
     for (unsigned int i = 0; i < par_.popSize; ++i)
     {
         doMutation();
@@ -141,7 +140,6 @@ void GeneticScheduler<V, T>::nextGeneration(void)
     //LOG(Debug) << "After mutations:\n" << *this << std::endl;
     
     // mating
-    //PARALLEL_FOR_LOOP
     for (unsigned int i = 0; i < par_.popSize/2; ++i)
     {
         doCrossover();
@@ -177,7 +175,7 @@ void GeneticScheduler<V, T>::doCrossover(void)
     Gene c1, c2;
     
     crossover(c1, c2, p1, p2);
-    PARALLEL_CRITICAL
+    thread_critical
     {
         population_.insert(std::make_pair(func_(c1), c1));
         population_.insert(std::make_pair(func_(c2), c2));
@@ -197,7 +195,7 @@ void GeneticScheduler<V, T>::doMutation(void)
         
         std::advance(it, pdis(gen_));
         mutation(m, it->second);
-        PARALLEL_CRITICAL
+	thread_critical
         {
             population_.insert(std::make_pair(func_(m), m));
         }
diff --git a/Hadrons/Global.cc b/Hadrons/Global.cc
index 7645e956..d74d10d2 100644
--- a/Hadrons/Global.cc
+++ b/Hadrons/Global.cc
@@ -29,7 +29,6 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #include <Hadrons/Global.hpp>
 
 using namespace Grid;
-using namespace QCD;
 using namespace Hadrons;
 
 HadronsLogger Hadrons::HadronsLogError(1,"Error");
@@ -77,7 +76,7 @@ size_t Hadrons::typeHash(const std::type_info *info)
     return info->hash_code();
 }
 
-constexpr unsigned int maxNameSize = 1024u;
+//constexpr unsigned int maxNameSize = 1024u;
 
 std::string Hadrons::typeName(const std::type_info *info)
 {
diff --git a/Hadrons/Global.hpp b/Hadrons/Global.hpp
index a525a30f..9b592721 100644
--- a/Hadrons/Global.hpp
+++ b/Hadrons/Global.hpp
@@ -52,7 +52,6 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 #define BEGIN_HADRONS_NAMESPACE \
 namespace Grid {\
-using namespace QCD;\
 namespace Hadrons {\
 using Grid::operator<<;\
 using Grid::operator>>;
diff --git a/Hadrons/Module.cc b/Hadrons/Module.cc
index c5057f8f..8e332d5a 100644
--- a/Hadrons/Module.cc
+++ b/Hadrons/Module.cc
@@ -29,7 +29,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #include <Hadrons/Module.hpp>
 
 using namespace Grid;
-using namespace QCD;
+ 
 using namespace Hadrons;
 
 /******************************************************************************
diff --git a/Hadrons/Modules/MAction/Wilson.hpp b/Hadrons/Modules/MAction/Wilson.hpp
index 11df89c3..9aec5f79 100644
--- a/Hadrons/Modules/MAction/Wilson.hpp
+++ b/Hadrons/Modules/MAction/Wilson.hpp
@@ -122,10 +122,8 @@ void TWilson<FImpl>::setup(void)
     {
         implParams.twist_n_2pi_L   = strToVec<Real>(par().twist);
     }
-    LOG(Message) << "Fermion boundary conditions: " << implParams.boundary_phases
-                 << std::endl;
-    LOG(Message) << "Twists: " << implParams.twist_n_2pi_L
-                 << std::endl;
+    LOG(Message) << "Fermion boundary conditions: " << implParams.boundary_phases << std::endl;
+    LOG(Message) << "Twists: " << implParams.twist_n_2pi_L << std::endl;
     if (implParams.boundary_phases.size() != env().getNd())
     {
         HADRONS_ERROR(Size, "Wrong number of boundary phase");
diff --git a/Hadrons/Modules/MAction/ZMobiusDWF.hpp b/Hadrons/Modules/MAction/ZMobiusDWF.hpp
index 56efe1a2..62bd1651 100644
--- a/Hadrons/Modules/MAction/ZMobiusDWF.hpp
+++ b/Hadrons/Modules/MAction/ZMobiusDWF.hpp
@@ -49,7 +49,7 @@ public:
                                     double                           , M5,
                                     double                           , b,
                                     double                           , c,
-                                    std::vector<std::complex<double>>, omega,
+                                    std::vector<std::complex<double> >, omega,
                                     std::string                      , boundary,
                                     std::string                      , twist);
 };
@@ -126,6 +126,8 @@ void TZMobiusDWF<FImpl>::setup(void)
     auto &grb5 = *envGetRbGrid(FermionField, par().Ls);
     auto omega = par().omega;
     typename ZMobiusFermion<FImpl>::ImplParams implParams;
+    
+
     if (!par().boundary.empty())
     {
         implParams.boundary_phases = strToVec<Complex>(par().boundary);
@@ -134,10 +136,8 @@ void TZMobiusDWF<FImpl>::setup(void)
     {
         implParams.twist_n_2pi_L   = strToVec<Real>(par().twist);
     }
-    LOG(Message) << "Fermion boundary conditions: " << implParams.boundary_phases
-                 << std::endl;
-    LOG(Message) << "Twists: " << implParams.twist_n_2pi_L
-                 << std::endl;
+    LOG(Message) << "Fermion boundary conditions: " << implParams.boundary_phases << std::endl;
+    LOG(Message) << "Twists: " << implParams.twist_n_2pi_L << std::endl;
     if (implParams.boundary_phases.size() != env().getNd())
     {
         HADRONS_ERROR(Size, "Wrong number of boundary phase");
@@ -146,9 +146,17 @@ void TZMobiusDWF<FImpl>::setup(void)
     {
         HADRONS_ERROR(Size, "Wrong number of twist");
     }
-    envCreateDerived(FMat, ZMobiusFermion<FImpl>, getName(), par().Ls, U, g5,
-                     grb5, g4, grb4, par().mass, par().M5, omega,
-                     par().b, par().c, implParams);
+
+    assert(par().Ls==omega.size());
+    int Ls=par().Ls;
+    std::vector<ComplexD> _omega(Ls);
+    for(int i=0;i<Ls;i++){
+      _omega[i] = omega[i];
+    }
+    envCreateDerived(FMat, ZMobiusFermion<FImpl>, getName(), par().Ls,
+		     U, g5, grb5, g4, grb4, 
+		     par().mass, par().M5, 
+		     _omega, par().b, par().c, implParams);
 }
 
 // execution ///////////////////////////////////////////////////////////////////
diff --git a/Hadrons/Modules/MContraction/A2ALoop.hpp b/Hadrons/Modules/MContraction/A2ALoop.hpp
index 7e7ffac0..316b228d 100644
--- a/Hadrons/Modules/MContraction/A2ALoop.hpp
+++ b/Hadrons/Modules/MContraction/A2ALoop.hpp
@@ -109,7 +109,7 @@ void TA2ALoop<FImpl>::execute(void)
     auto &left  = envGet(std::vector<FermionField>, par().left);
     auto &right = envGet(std::vector<FermionField>, par().right);
 
-    loop = zero;
+    loop = Zero();
     for (unsigned int i = 0; i < left.size(); ++i)
     {
         loop += outerProduct(left[i], right[i]);
diff --git a/Hadrons/Modules/MContraction/A2AMesonField.hpp b/Hadrons/Modules/MContraction/A2AMesonField.hpp
index f01df09f..d3f90959 100644
--- a/Hadrons/Modules/MContraction/A2AMesonField.hpp
+++ b/Hadrons/Modules/MContraction/A2AMesonField.hpp
@@ -258,7 +258,7 @@ void TA2AMesonField<FImpl>::execute(void)
             std::vector<Real> p;
 
             envGetTmp(ComplexField, coor);
-            ph[j] = zero;
+            ph[j] = Zero();
             for(unsigned int mu = 0; mu < mom_[j].size(); mu++)
             {
                 LatticeCoordinate(coor, mu);
diff --git a/Hadrons/Modules/MContraction/WeakMesonDecayKl2.hpp b/Hadrons/Modules/MContraction/WeakMesonDecayKl2.hpp
index d7a45108..7b00ffaa 100644
--- a/Hadrons/Modules/MContraction/WeakMesonDecayKl2.hpp
+++ b/Hadrons/Modules/MContraction/WeakMesonDecayKl2.hpp
@@ -156,7 +156,7 @@ void TWeakMesonDecayKl2<FImpl>::execute(void)
     std::vector<SpinMatrix> res_summed;
     Result                  r;
 
-    auto &res    = envGet(PropagatorField, getName()); res = zero;
+    auto &res    = envGet(PropagatorField, getName()); res = Zero();
     auto &q1     = envGet(PropagatorField, par().q1);
     auto &q2     = envGet(PropagatorField, par().q2);
     auto &lepton = envGet(PropagatorField, par().lepton);
@@ -166,7 +166,7 @@ void TWeakMesonDecayKl2<FImpl>::execute(void)
 
     for (unsigned int mu = 0; mu < 4; ++mu)
     {
-        c = zero;
+        c = Zero();
         //hadronic part: trace(q1*adj(q2)*g5*gL[mu]) 
         c = trace(q1*adj(q2)*g5*GammaL(Gamma::gmu[mu]));
         prop_buf = 1.;
diff --git a/Hadrons/Modules/MFermion/EMLepton.hpp b/Hadrons/Modules/MFermion/EMLepton.hpp
index 3fe544d3..baae7a7e 100644
--- a/Hadrons/Modules/MFermion/EMLepton.hpp
+++ b/Hadrons/Modules/MFermion/EMLepton.hpp
@@ -219,7 +219,7 @@ void TEMLepton<FImpl>::execute(void)
 	    // 5D source if action is 5d
 	    mat.ImportPhysicalFermionSource(tmp, source);
 	}
-        sol = zero;
+        sol = Zero();
 	mat.FreePropagator(source,sol,mass,boundary,twist);
 	if (Ls_ == 1)
 	{
@@ -257,14 +257,14 @@ void TEMLepton<FImpl>::execute(void)
 	proptmp = where( tlat < tl, boundary[Tp]*proptmp, proptmp);
 
         // i*A_mu*gamma_mu
-        sourcetmp = zero;
+        sourcetmp = Zero();
         for(unsigned int mu=0;mu<=3;mu++)
         {
 	    Gamma gmu(Gamma::gmu[mu]);
 	    sourcetmp +=  ci * PeekIndex<LorentzIndex>(stoch_photon, mu) *  (gmu * proptmp );
         }
 
-        proptmp = zero;
+        proptmp = Zero();
 
         //sequential propagator from i*Aslash*S
         LOG(Message) << "Sequential propagator for t= " << tl << std::endl;
@@ -281,7 +281,7 @@ void TEMLepton<FImpl>::execute(void)
 		// 5D source if action is 5d
 		mat.ImportPhysicalFermionSource(tmp, source);
 	    }
-            sol = zero;
+            sol = Zero();
 	    mat.FreePropagator(source,sol,mass,boundary,twist);
 	    if (Ls_ == 1)
 	    {
diff --git a/Hadrons/Modules/MFermion/FreeProp.hpp b/Hadrons/Modules/MFermion/FreeProp.hpp
index baccba6a..dcf77802 100644
--- a/Hadrons/Modules/MFermion/FreeProp.hpp
+++ b/Hadrons/Modules/MFermion/FreeProp.hpp
@@ -167,7 +167,7 @@ void TFreeProp<FImpl>::execute(void)
                 PropToFerm<FImpl>(source, fullSrc, s, c);
             }
         }
-        sol = zero;
+        sol = Zero();
 	std::vector<double> twist = strToVec<double>(par().twist);
 	if(twist.size() != Nd)
 	{
diff --git a/Hadrons/Modules/MFermion/GaugeProp.hpp b/Hadrons/Modules/MFermion/GaugeProp.hpp
index 49b4bd5a..a11274e6 100644
--- a/Hadrons/Modules/MFermion/GaugeProp.hpp
+++ b/Hadrons/Modules/MFermion/GaugeProp.hpp
@@ -174,8 +174,8 @@ void TGaugeProp<FImpl>::execute(void)
                 PropToFerm<FImpl>(source, fullSrc, s, c);
             }
         }
+        sol = Zero();
         LOG(Message) << "Solve" << std::endl;
-        sol = zero;
         solver(sol, source);
         LOG(Message) << "Export solution" << std::endl;
         FermToProp<FImpl>(prop, sol, s, c);
diff --git a/Hadrons/Modules/MIO/LoadEigenPack.hpp b/Hadrons/Modules/MIO/LoadEigenPack.hpp
index 0d6036b8..d036e3d3 100644
--- a/Hadrons/Modules/MIO/LoadEigenPack.hpp
+++ b/Hadrons/Modules/MIO/LoadEigenPack.hpp
@@ -177,7 +177,7 @@ void TLoadEigenPack<Pack, GImpl>::execute(void)
         for (unsigned int i = 0; i < par().size; i++)
         {
             LOG(Message) << "Applying gauge transformation to eigenvector i = " << i << "/" << par().size << std::endl;
-            epack.evec[i].checkerboard = Odd;
+            epack.evec[i].Checkerboard() = Odd;
             epack.evec[i] = tmpXformOdd * epack.evec[i];
         }
         stopTimer("Transform application");
diff --git a/Hadrons/Modules/MNPR/Amputate.hpp b/Hadrons/Modules/MNPR/Amputate.hpp
index 46a148c4..6258e515 100644
--- a/Hadrons/Modules/MNPR/Amputate.hpp
+++ b/Hadrons/Modules/MNPR/Amputate.hpp
@@ -109,8 +109,6 @@ template <typename FImpl1, typename FImpl2>
 std::vector<std::string> TAmputate<FImpl1, FImpl2>::getOutput(void)
 {
     std::vector<std::string> output = {getName()};
-    
-    
     return output;
 }
 
@@ -118,15 +116,16 @@ std::vector<std::string> TAmputate<FImpl1, FImpl2>::getOutput(void)
 template <typename Fimpl1, typename Fimpl2>
 SpinColourMatrix TAmputate<Fimpl1, Fimpl2>::invertspincolmat(SpinColourMatrix &scmat)
 {
-    Eigen::MatrixXcf scmat_2d(Ns*Nc,Ns*Nc);
+    Eigen::MatrixXcd scmat_2d(Ns*Nc,Ns*Nc);
     for(int ic=0; ic<Nc; ic++){
     for(int jc=0; jc<Nc; jc++){
         for(int is=0; is<Ns; is++){
         for(int js=0; js<Ns; js++){
-            scmat_2d(Ns*ic+is,Ns*jc+js) = scmat()(is,js)(ic,jc);
+	  auto z =scmat()(is,js)(ic,jc);;
+	  scmat_2d(Ns*ic+is,Ns*jc+js) = std::complex<double>(real(z),imag(z));
         }}
     }}      
-    Eigen::MatrixXcf scmat_2d_inv = scmat_2d.inverse();
+    Eigen::MatrixXcd scmat_2d_inv = scmat_2d.inverse();
     SpinColourMatrix scmat_inv;
     for(int ic=0; ic<Nc; ic++){
     for(int jc=0; jc<Nc; jc++){
@@ -163,8 +162,8 @@ void TAmputate<FImpl1, FImpl2>::execute(void)
     read(reader,"vertex", vertex);
     LOG(Message) << "vertex read" << std::endl;
 
-    pdotxin=zero;
-    pdotxout=zero;
+    pdotxin=Zero();
+    pdotxout=Zero();
     for (unsigned int mu = 0; mu < 4; ++mu)
     {
         Real TwoPiL =  M_PI * 2.0/ latt_size[mu];
diff --git a/Hadrons/Modules/MNPR/Bilinear.hpp b/Hadrons/Modules/MNPR/Bilinear.hpp
index f9f3320a..444d42a0 100644
--- a/Hadrons/Modules/MNPR/Bilinear.hpp
+++ b/Hadrons/Modules/MNPR/Bilinear.hpp
@@ -129,7 +129,7 @@ LatticeSpinColourMatrix TBilinear<FImpl1, FImpl2>::PhaseProps(LatticeSpinColourM
     LatticeComplex      pdotx(grid),  coor(grid);
     std::vector<int>   latt_size = grid->_fdimensions; 
     Complex             Ci(0.0,1.0);
-    pdotx=zero;
+    pdotx=Zero();
     for (unsigned int mu = 0; mu < 4; ++mu)
     {
         Real TwoPiL =  M_PI * 2.0/ latt_size[mu];
@@ -187,8 +187,8 @@ q = (p1-p2)
 
     //
 
-    pdotxin=zero;
-    pdotxout=zero;
+    pdotxin=Zero();
+    pdotxout=Zero();
     for (unsigned int mu = 0; mu < 4; ++mu)
     {
         Real TwoPiL =  M_PI * 2.0/ latt_size[mu];
diff --git a/Hadrons/Modules/MNPR/FourQuark.hpp b/Hadrons/Modules/MNPR/FourQuark.hpp
index ebf169c5..3fb687ca 100644
--- a/Hadrons/Modules/MNPR/FourQuark.hpp
+++ b/Hadrons/Modules/MNPR/FourQuark.hpp
@@ -116,41 +116,30 @@ std::vector<std::string> TFourQuark<FImpl1, FImpl2>::getOutput(void)
 template <typename FImpl1, typename FImpl2>
 void TFourQuark<FImpl1, FImpl2>::tensorprod(LatticeSpinColourSpinColourMatrix &lret, LatticeSpinColourMatrix a, LatticeSpinColourMatrix b)
 {
-#if 0
-            parallel_for(auto site=lret.begin();site<lret.end();site++) {
-                for (int si; si < 4; ++si){
-                for(int sj; sj <4; ++sj){
-                    for (int ci; ci < 3; ++ci){
-                    for (int cj; cj < 3; ++cj){
-                        for (int sk; sk < 4; ++sk){
-                        for(int sl; sl <4; ++sl){
-                            for (int ck; ck < 3; ++ck){
-                            for (int cl; cl < 3; ++cl){
-                        lret[site]()(si,sj)(ci,cj)(sk,sl)(ck,cl)=a[site]()(si,sj)(ci,cj)*b[site]()(sk,sl)(ck,cl);
-                            }}
-                        }}
-                    }}
-                }}
-        }
-#else 
             // FIXME ; is there a general need for this construct ? In which case we should encapsulate the
             //         below loops in a helper function.
             //LOG(Message) << "sp co mat a is - " << a << std::endl;
             //LOG(Message) << "sp co mat b is - " << b << std::endl;
-            parallel_for(auto site=lret.begin();site<lret.end();site++) {
-            vTComplex left;
+	    auto  lret_v = lret.View();
+	    auto  a_v = a.View();
+	    auto  b_v = b.View();
+#ifdef GRID_NVCC
+#warning "NVCC problem: Removed impossibly slow compile of simple NPR host code in FourQuark.hpp"
+#else
+            thread_foreach( site,lret_v,{
+		vTComplex left;
                 for(int si=0; si < Ns; ++si){
                 for(int sj=0; sj < Ns; ++sj){
                     for (int ci=0; ci < Nc; ++ci){
                     for (int cj=0; cj < Nc; ++cj){
                       //LOG(Message) << "si, sj, ci, cj -  " << si << ", " << sj  << ", "<< ci  << ", "<< cj << std::endl;
-                      left()()() = a[site]()(si,sj)(ci,cj);
+                      left()()() = a_v[site]()(si,sj)(ci,cj);
                       //LOG(Message) << left << std::endl;
-                      lret[site]()(si,sj)(ci,cj)=left()*b[site]();
+                      lret_v[site]()(si,sj)(ci,cj)=left()*b_v[site]();
                     }}
                 }}
-            }
-#endif      
+	      });
+#endif
 }
 
 
@@ -168,6 +157,7 @@ void TFourQuark<FImpl1, FImpl2>::setup(void)
 template <typename FImpl1, typename FImpl2>
 void TFourQuark<FImpl1, FImpl2>::execute(void)
 {
+#ifndef GRID_NVCC
 
 /*********************************************************************************
 
@@ -219,8 +209,8 @@ We have up to 256 of these including the offdiag (G1 != G2).
     //Sout = Grid::QCD::PropUtils::PhaseProps(Sout,pout);
     
     //find p.x for in and out so phase can be accounted for in propagators
-    pdotxin=zero;
-    pdotxout=zero;
+    pdotxin=Zero();
+    pdotxout=Zero();
     for (unsigned int mu = 0; mu < 4; ++mu)
     {
         Real TwoPiL =  M_PI * 2.0/ latt_size[mu];
@@ -231,7 +221,6 @@ We have up to 256 of these including the offdiag (G1 != G2).
     Sin = Sin*exp(-Ci*pdotxin); //phase corrections
     Sout = Sout*exp(-Ci*pdotxout);
 
-
     //Set up Gammas 
     std::vector<Gamma> gammavector;
      for( int i=1; i<Gamma::nGamma; i+=2){
@@ -239,7 +228,7 @@ We have up to 256 of these including the offdiag (G1 != G2).
          gammavector.push_back(Gamma(gam));
        }
     
-    lret = zero;
+    lret = Zero();
     if (fullbasis == true){ // all combinations of mu and nu
         result.fourquark.resize(Gamma::nGamma/2*Gamma::nGamma/2);
         for( int mu=0; mu<Gamma::nGamma/2; mu++){ 
@@ -248,7 +237,7 @@ We have up to 256 of these including the offdiag (G1 != G2).
                 LatticeSpinColourMatrix     bilinear_nu(env().getGrid());
                 bilinear_nu = g5*adj(Sout)*g5*gammavector[nu]*Sin;
                 LOG(Message) << "bilinear_nu for nu = " << nu << " is - " << bilinear_mu << std::endl;
-                result.fourquark[mu*Gamma::nGamma/2 + nu] = zero;
+                result.fourquark[mu*Gamma::nGamma/2 + nu] = Zero();
                 tensorprod(lret,bilinear_mu,bilinear_nu);
                 result.fourquark[mu*Gamma::nGamma/2 + nu] = sum(lret);
             }
@@ -259,12 +248,13 @@ We have up to 256 of these including the offdiag (G1 != G2).
         //for( int mu=0; mu<Gamma::nGamma/2; mu++ ){
             bilinear_mu = g5*adj(Sout)*g5*gammavector[mu]*Sin;
             //LOG(Message) << "bilinear_mu for mu = " << mu << " is - " << bilinear_mu << std::endl;
-            result.fourquark[mu] = zero;
+            result.fourquark[mu] = Zero();
             tensorprod(lret,bilinear_mu,bilinear_mu); //tensor outer product
             result.fourquark[mu] = sum(lret);
         }
     }
     write(writer, "fourquark", result.fourquark);
+#endif
 }
 
 END_MODULE_NAMESPACE
diff --git a/Hadrons/Modules/MScalar/ChargedProp.cc b/Hadrons/Modules/MScalar/ChargedProp.cc
index 7d623a8f..cde5e3ba 100644
--- a/Hadrons/Modules/MScalar/ChargedProp.cc
+++ b/Hadrons/Modules/MScalar/ChargedProp.cc
@@ -149,7 +149,7 @@ void TChargedProp::execute(void)
                      << RESULT_FILE_NAME(par().output, vm().getTrajectory()) << "'..."
                      << std::endl;
         result.projection.resize(par().outputMom.size());
-        result.lattice_size = env().getGrid()->_fdimensions;
+        result.lattice_size = env().getGrid()->FullDimensions().toVector();
         result.mass = par().mass;
         result.charge = q;
         siteCoor.resize(env().getNd());
@@ -160,11 +160,11 @@ void TChargedProp::execute(void)
             LOG(Message) << "Calculating (" << par().outputMom[i_p]
                          << ") momentum projection" << std::endl;
 
-            result.projection[i_p].corr_0.resize(env().getGrid()->_fdimensions[env().getNd()-1]);
-            result.projection[i_p].corr.resize(env().getGrid()->_fdimensions[env().getNd()-1]);
-            result.projection[i_p].corr_Q.resize(env().getGrid()->_fdimensions[env().getNd()-1]);
-            result.projection[i_p].corr_Sun.resize(env().getGrid()->_fdimensions[env().getNd()-1]);
-            result.projection[i_p].corr_Tad.resize(env().getGrid()->_fdimensions[env().getNd()-1]);
+            result.projection[i_p].corr_0.resize(env().getGrid()->FullDimensions()[env().getNd()-1]);
+            result.projection[i_p].corr.resize(env().getGrid()->FullDimensions()[env().getNd()-1]);
+            result.projection[i_p].corr_Q.resize(env().getGrid()->FullDimensions()[env().getNd()-1]);
+            result.projection[i_p].corr_Sun.resize(env().getGrid()->FullDimensions()[env().getNd()-1]);
+            result.projection[i_p].corr_Tad.resize(env().getGrid()->FullDimensions()[env().getNd()-1]);
 
             for (unsigned int j = 0; j < env().getNd()-1; ++j)
             {
@@ -226,7 +226,7 @@ void TChargedProp::makeCaches(void)
 	}
     if (!phasesDone_)
     {
-        std::vector<int> &l = env().getGrid()->_fdimensions;
+        auto l = env().getGrid()->FullDimensions();
         Complex          ci(0.0,1.0);
         
         LOG(Message) << "Caching shift phases..." << std::endl;
@@ -259,7 +259,7 @@ void TChargedProp::momD1(ScalarField &s, FFT &fft)
     envGetTmp(ScalarField, result);
     envGetTmp(ScalarField, Amu);
 
-    result = zero;
+    result = Zero();
     for (unsigned int mu = 0; mu < env().getNd(); ++mu)
     {
         Amu = peekLorentz(A, mu);
@@ -289,7 +289,7 @@ void TChargedProp::momD2(ScalarField &s, FFT &fft)
     envGetTmp(ScalarField, result);
     envGetTmp(ScalarField, Amu);
 
-    result = zero;
+    result = Zero();
     for (unsigned int mu = 0; mu < env().getNd(); ++mu)
     {
         Amu = peekLorentz(A, mu);
diff --git a/Hadrons/Modules/MScalarSUN/Div.hpp b/Hadrons/Modules/MScalarSUN/Div.hpp
index 5d623fa0..af68d696 100644
--- a/Hadrons/Modules/MScalarSUN/Div.hpp
+++ b/Hadrons/Modules/MScalarSUN/Div.hpp
@@ -132,7 +132,7 @@ void TDiv<SImpl>::execute(void)
     std::cout << std::endl;
 
     auto &div = envGet(ComplexField, getName());
-    div = zero;
+    div = Zero();
     for (unsigned int mu = 0; mu < nd; ++mu)
     {
         auto &op = envGet(ComplexField, par().op[mu]);
diff --git a/Hadrons/Modules/MScalarSUN/StochFreeField.hpp b/Hadrons/Modules/MScalarSUN/StochFreeField.hpp
index 2c53a7d3..a66f010c 100644
--- a/Hadrons/Modules/MScalarSUN/StochFreeField.hpp
+++ b/Hadrons/Modules/MScalarSUN/StochFreeField.hpp
@@ -154,7 +154,7 @@ void TStochFreeField<SImpl>::execute(void)
     LOG(Message) << "Generating random momentum-space field" << std::endl;
     envGetTmp(Field, phift);
     envGetTmp(ComplexField, ca);
-    phift = zero;
+    phift = Zero();
     for (int a = 0; a < Nadj; ++a) 
     {
         Site ta;
diff --git a/Hadrons/Modules/MScalarSUN/TrKinetic.hpp b/Hadrons/Modules/MScalarSUN/TrKinetic.hpp
index ff68e314..f26a0f91 100644
--- a/Hadrons/Modules/MScalarSUN/TrKinetic.hpp
+++ b/Hadrons/Modules/MScalarSUN/TrKinetic.hpp
@@ -142,7 +142,7 @@ void TTrKinetic<SImpl>::execute(void)
     auto               &sumkin = envGet(ComplexField, varName(getName(), "sum"));
 
     envGetTmp(std::vector<Field>, der);
-    sumkin = zero;
+    sumkin = Zero();
     if (!par().output.empty())
     {
         result.type = par().type;
diff --git a/Hadrons/Modules/MScalarSUN/TransProj.cc b/Hadrons/Modules/MScalarSUN/TransProj.cc
index 697b72fd..e77f53fb 100644
--- a/Hadrons/Modules/MScalarSUN/TransProj.cc
+++ b/Hadrons/Modules/MScalarSUN/TransProj.cc
@@ -37,3 +37,4 @@ template class Grid::Hadrons::MScalarSUN::TTransProj<ScalarNxNAdjImplR<4>>;
 template class Grid::Hadrons::MScalarSUN::TTransProj<ScalarNxNAdjImplR<5>>;
 template class Grid::Hadrons::MScalarSUN::TTransProj<ScalarNxNAdjImplR<6>>;
 
+    
diff --git a/Hadrons/Modules/MScalarSUN/TransProj.hpp b/Hadrons/Modules/MScalarSUN/TransProj.hpp
index 7be37f33..22ecfffb 100644
--- a/Hadrons/Modules/MScalarSUN/TransProj.hpp
+++ b/Hadrons/Modules/MScalarSUN/TransProj.hpp
@@ -145,7 +145,7 @@ void TTransProj<SImpl>::execute(void)
     envGetTmp(ComplexField, buf1);
     envGetTmp(ComplexField, buf2);
     envGetTmp(ComplexField, lap);
-    lap = zero;
+    lap = Zero();
     if (!par().output.empty())
     {
         result.type = par().type;
diff --git a/Hadrons/Modules/MSink/Point.hpp b/Hadrons/Modules/MSink/Point.hpp
index 2785ba86..a8359eae 100644
--- a/Hadrons/Modules/MSink/Point.hpp
+++ b/Hadrons/Modules/MSink/Point.hpp
@@ -128,7 +128,7 @@ void TPoint<FImpl>::execute(void)
 
         envGetTmp(LatticeComplex, coor);
         p  = strToVec<Real>(par().mom);
-        ph = zero;
+        ph = Zero();
         for(unsigned int mu = 0; mu < p.size(); mu++)
         {
             LatticeCoordinate(coor, mu);
diff --git a/Hadrons/Modules/MSolver/A2AAslashVectors.hpp b/Hadrons/Modules/MSolver/A2AAslashVectors.hpp
index 7c745f10..fa222228 100644
--- a/Hadrons/Modules/MSolver/A2AAslashVectors.hpp
+++ b/Hadrons/Modules/MSolver/A2AAslashVectors.hpp
@@ -156,7 +156,7 @@ void TA2AAslashVectors<FImpl>::execute(void)
                  << par().vector << " and the photon field " << par().emField << std::endl;
     for(unsigned int i=0; i<Nmodes; i++)
     {
-        v4dtmp = zero;
+        v4dtmp = Zero();
         startTimer("Multiply Aslash");
         for(unsigned int mu=0;mu<=3;mu++)
         {
diff --git a/Hadrons/Modules/MSolver/MixedPrecisionRBPrecCG.hpp b/Hadrons/Modules/MSolver/MixedPrecisionRBPrecCG.hpp
index 808e1cb0..5a041c79 100644
--- a/Hadrons/Modules/MSolver/MixedPrecisionRBPrecCG.hpp
+++ b/Hadrons/Modules/MSolver/MixedPrecisionRBPrecCG.hpp
@@ -68,6 +68,7 @@ private:
     class OperatorFunctionWrapper: public OperatorFunction<Field>
     {
     public:
+        using OperatorFunction<Field>::operator();
         OperatorFunctionWrapper(LinearFunction<Field> &fn): fn_(fn) {};
         virtual ~OperatorFunctionWrapper(void) = default;
         virtual void operator()(LinearOperatorBase<Field> &op, 
diff --git a/Hadrons/Modules/MSource/Gauss.hpp b/Hadrons/Modules/MSource/Gauss.hpp
index 80111b75..8957288d 100644
--- a/Hadrons/Modules/MSource/Gauss.hpp
+++ b/Hadrons/Modules/MSource/Gauss.hpp
@@ -117,7 +117,7 @@ void TGauss<FImpl>::execute(void)
     const Real Pi(M_PI);
     const SitePropagator idMat=[](){ SitePropagator s; s=1.; return s; }();
 
-    ScalarRho=zero;
+    ScalarRho=Zero();
     for(int mu=0; mu<dim; mu++) {
         assert(env().getDim(mu)%2==0);
         assert(position_[mu]>=0 && position_[mu]<env().getDim(mu));
diff --git a/Hadrons/Modules/MSource/Momentum.hpp b/Hadrons/Modules/MSource/Momentum.hpp
index d47146e8..87a19fe5 100644
--- a/Hadrons/Modules/MSource/Momentum.hpp
+++ b/Hadrons/Modules/MSource/Momentum.hpp
@@ -126,7 +126,7 @@ void TMomentum<FImpl>::execute(void)
     LOG(Message) << " " << std::endl;
     //get the momentum from parameters
     p  = strToVec<Real>(par().mom);
-    C = zero;
+    C = Zero();
     LOG(Message) << "momentum converted from string - " << std::to_string(p[0]) <<std::to_string(p[1]) <<std::to_string(p[2]) <<   std::to_string(p[3]) << std::endl;
     for(int mu=0;mu<4;mu++){
     Real TwoPiL =  M_PI * 2.0/ latt_size[mu];
diff --git a/Hadrons/Modules/MSource/Point.hpp b/Hadrons/Modules/MSource/Point.hpp
index 93f14030..a33c5308 100644
--- a/Hadrons/Modules/MSource/Point.hpp
+++ b/Hadrons/Modules/MSource/Point.hpp
@@ -132,7 +132,7 @@ void TPoint<FImpl>::execute(void)
                       + " components (must have " + std::to_string(env().getNd()) + ")");
     }
     id  = 1.;
-    src = zero;
+    src = Zero();
     pokeSite(id, src, position);
 }
 
diff --git a/Hadrons/Modules/MSource/SeqAslash.hpp b/Hadrons/Modules/MSource/SeqAslash.hpp
index a31051fd..ccca15d9 100644
--- a/Hadrons/Modules/MSource/SeqAslash.hpp
+++ b/Hadrons/Modules/MSource/SeqAslash.hpp
@@ -148,7 +148,7 @@ void TSeqAslash<FImpl>::execute(void)
                      << par().tA << " <= t <= " << par().tB 
 		     << " using the photon field " << par().emField << std::endl;
     }
-    auto  &src = envGet(PropagatorField, getName()); src=zero;
+    auto  &src = envGet(PropagatorField, getName()); src=Zero();
     auto  &q   = envGet(PropagatorField, par().q);
     auto  &ph  = envGet(LatticeComplex, momphName_);
     auto  &t   = envGet(Lattice<iScalar<vInteger>>, tName_);
@@ -160,7 +160,7 @@ void TSeqAslash<FImpl>::execute(void)
 
         envGetTmp(LatticeComplex, coor);
         p  = strToVec<Real>(par().mom);
-        ph = zero;
+        ph = Zero();
         for(unsigned int mu = 0; mu < env().getNd(); mu++)
         {
             LatticeCoordinate(coor, mu);
diff --git a/Hadrons/Modules/MSource/SeqConserved.hpp b/Hadrons/Modules/MSource/SeqConserved.hpp
index 33a2c9fd..c8d5121d 100644
--- a/Hadrons/Modules/MSource/SeqConserved.hpp
+++ b/Hadrons/Modules/MSource/SeqConserved.hpp
@@ -141,7 +141,7 @@ void TSeqConserved<FImpl>::setup(void)
 {
     auto Ls_ = env().getObjectLs(par().action);
     envCreateLat(PropagatorField, getName(), Ls_);
-    envTmpLat(PropagatorField, "src_tmp");
+    envTmpLat(PropagatorField, "src_tmp",Ls_);
     envCacheLat(LatticeComplex, SeqmomphName_);
     envTmpLat(LatticeComplex, "coor");
     envTmpLat(LatticeComplex, "latt_compl");
@@ -175,14 +175,14 @@ void TSeqConserved<FImpl>::execute(void)
     auto &mat = envGet(FMat, par().action);
     envGetTmp(LatticeComplex, latt_compl);
 
-    src = zero;
+    src = Zero();
 
     //exp(ipx)
     auto &mom_phase = envGet(LatticeComplex, SeqmomphName_);
     if (!SeqhasPhase_)
     {    
         std::vector<Real> mom = strToVec<Real>(par().mom);
-        mom_phase = zero;
+        mom_phase = Zero();
         Complex           i(0.0,1.0);
         envGetTmp(LatticeComplex, coor);
         for(unsigned int mu = 0; mu < env().getNd(); mu++)
diff --git a/Hadrons/Modules/MSource/SeqGamma.hpp b/Hadrons/Modules/MSource/SeqGamma.hpp
index 25f32a75..4d125848 100644
--- a/Hadrons/Modules/MSource/SeqGamma.hpp
+++ b/Hadrons/Modules/MSource/SeqGamma.hpp
@@ -159,7 +159,7 @@ void TSeqGamma<FImpl>::execute(void)
 
         envGetTmp(LatticeComplex, coor);
         p  = strToVec<Real>(par().mom);
-        ph = zero;
+        ph = Zero();
         for(unsigned int mu = 0; mu < env().getNd(); mu++)
         {
             LatticeCoordinate(coor, mu);
diff --git a/Hadrons/Modules/MSource/Wall.hpp b/Hadrons/Modules/MSource/Wall.hpp
index 945e2760..ea26d6c9 100644
--- a/Hadrons/Modules/MSource/Wall.hpp
+++ b/Hadrons/Modules/MSource/Wall.hpp
@@ -142,7 +142,7 @@ void TWall<FImpl>::execute(void)
 
         envGetTmp(LatticeComplex, coor);
         p  = strToVec<Real>(par().mom);
-        ph = zero;
+        ph = Zero();
         for(unsigned int mu = 0; mu < env().getNd(); mu++)
         {
             LatticeCoordinate(coor, mu);
diff --git a/Hadrons/TimerArray.cc b/Hadrons/TimerArray.cc
index 1274f4d5..caf72fc1 100644
--- a/Hadrons/TimerArray.cc
+++ b/Hadrons/TimerArray.cc
@@ -28,7 +28,6 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #include <Hadrons/TimerArray.hpp>
 
 using namespace Grid;
-using namespace QCD;
 using namespace Hadrons;
 
 void TimerArray::startTimer(const std::string &name)
diff --git a/Hadrons/Utilities/Contractor.cc b/Hadrons/Utilities/Contractor.cc
index cfa4492e..7e4b8b26 100644
--- a/Hadrons/Utilities/Contractor.cc
+++ b/Hadrons/Utilities/Contractor.cc
@@ -31,7 +31,6 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #include <Hadrons/TimerArray.hpp>
 
 using namespace Grid;
-using namespace QCD;
 using namespace Hadrons;
 
 #define TIME_MOD(t) (((t) + par.global.nt) % par.global.nt)
@@ -252,18 +251,18 @@ int main(int argc, char* argv[])
 
     // parse parameter file
     ContractorPar par;
-    unsigned int  nMat, nCont;
+    //    unsigned int  nMat,nCont;
     XmlReader     reader(parFilename);
 
     read(reader, "global",    par.global);
     read(reader, "a2aMatrix", par.a2aMatrix);
     read(reader, "product",   par.product);
-    nMat  = par.a2aMatrix.size();
-    nCont = par.product.size();
+    //    nMat  = par.a2aMatrix.size();
+    //    nCont = par.product.size();
 
     // create diskvectors
     std::map<std::string, EigenDiskVector<ComplexD>> a2aMat;
-    unsigned int                                     cacheSize;
+    //    unsigned int                                     cacheSize;
 
     for (auto &p: par.a2aMatrix)
     {
@@ -282,7 +281,8 @@ int main(int argc, char* argv[])
         for (auto &p: par.a2aMatrix)
         {
             std::string filename = p.file;
-            double      t, size;
+            double      t;
+	    //	    double  size;
 
             tokenReplace(filename, "traj", traj);
             std::cout << "======== Loading '" << filename << "'" << std::endl;
@@ -306,7 +306,8 @@ int main(int argc, char* argv[])
             std::vector<A2AMatrixTr<ComplexD>>     lastTerm(par.global.nt);
             A2AMatrix<ComplexD>                    prod, buf, tmp;
             TimerArray                             tAr;
-            double                                 fusec, busec, flops, bytes, tusec;
+            double                                 fusec, busec, flops, bytes;
+	    //	    double  tusec;
             Contractor::CorrelatorResult           result;             
 
             tAr.startTimer("Total");
@@ -352,11 +353,12 @@ int main(int argc, char* argv[])
 
                 tAr.startTimer("Transpose caching");
                 lastTerm[t].resize(ref.rows(), ref.cols());
-                parallel_for (unsigned int j = 0; j < ref.cols(); ++j)
-                for (unsigned int i = 0; i < ref.rows(); ++i)
-                {
-                    lastTerm[t](i, j) = ref(i, j);
-                }
+                thread_for( j,ref.cols(),{
+                  for (unsigned int i = 0; i < ref.rows(); ++i)
+                  {
+                      lastTerm[t](i, j) = ref(i, j);
+                  }
+		});
                 tAr.stopTimer("Transpose caching");
             }
             bytes = par.global.nt*lastTerm[0].rows()*lastTerm[0].cols()*sizeof(ComplexD);
diff --git a/Hadrons/Utilities/ContractorBenchmark.cc b/Hadrons/Utilities/ContractorBenchmark.cc
index 37c410db..d387cfec 100644
--- a/Hadrons/Utilities/ContractorBenchmark.cc
+++ b/Hadrons/Utilities/ContractorBenchmark.cc
@@ -34,16 +34,20 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 using namespace Grid;
 using namespace Hadrons;
+const int RowMajor = Eigen::RowMajor;
+const int ColMajor = Eigen::ColMajor;
 
 #ifdef GRID_COMMS_MPI3
 #define GET_RANK(rank, nMpi) \
 MPI_Comm_size(MPI_COMM_WORLD, &(nMpi));\
-MPI_Comm_rank(MPI_COMM_WORLD, &(rank))
+MPI_Comm_rank(MPI_COMM_WORLD, &(rank));\
+assert(rank<nMpi)
+
 #define BARRIER() MPI_Barrier(MPI_COMM_WORLD)
 #define INIT() MPI_Init(NULL, NULL)
 #define FINALIZE() MPI_Finalize()
 #else
-#define GET_RANK(rank, nMpi) (nMpi) = 1; (rank) = 0
+#define GET_RANK(rank, nMpi) (nMpi) = 1; (rank) = 0 ; assert(rank<nMpi)
 #define BARRIER()
 #define INIT()
 #define FINALIZE()
@@ -74,7 +78,7 @@ inline void trBenchmark(const std::string name, const MatLeft &left,
     if (rank == 0)
     {
         std::cout << std::setw(34) << name << ": diff= "
-                  << std::setw(12) << std::norm(buf-ref)
+                  << std::setw(12) << abs(buf-ref)
                   << std::setw(10) << t/1.0e6 << " sec "
                   << std::setw(10) << flops/t/1.0e3 << " GFlop/s " 
                   << std::setw(10) << bytes/t*1.0e6/1024/1024/1024 << " GB/s "
@@ -88,7 +92,8 @@ inline void mulBenchmark(const std::string name, const MatV &left,
                          const MatV &right, const Mat &ref, Function fn)
 {
     double       t, flops, bytes;
-    double       nr = left[0].rows(), nc = left[0].cols(), n = nr*nc;
+    double       nr = left[0].rows(), nc = left[0].cols();
+    //    double        n = nr*nc;
     unsigned int nMat = left.size();
     int          nMpi, rank;
     Mat          buf(left[0].rows(), left[0].rows());
@@ -126,22 +131,22 @@ static inline void zdotuRow(ComplexD &res, const unsigned int aRow,
     const ComplexD *aPt, *bPt;
     unsigned int   aInc, bInc;
 
-    if (MatLeft::Options == Eigen::RowMajor)
+    if (MatLeft::Options == RowMajor)
     {
         aPt  = a.data() + aRow*a.cols();
         aInc = 1;
     }
-    else if (MatLeft::Options == Eigen::ColMajor)
+    else if (MatLeft::Options == ColMajor)
     {
         aPt  = a.data() + aRow;
         aInc = a.rows();
     }
-    if (MatRight::Options == Eigen::RowMajor)
+    if (MatRight::Options == RowMajor)
     {
         bPt  = b.data() + aRow;
         bInc = b.cols();
     }
-    else if (MatRight::Options == Eigen::ColMajor)
+    else if (MatRight::Options == ColMajor)
     {
         bPt  = b.data() + aRow*b.rows();
         bInc = 1;
@@ -156,22 +161,22 @@ static inline void zdotuCol(ComplexD &res, const unsigned int aCol,
     const ComplexD *aPt, *bPt;
     unsigned int   aInc, bInc;
 
-    if (MatLeft::Options == Eigen::RowMajor)
+    if (MatLeft::Options == RowMajor)
     {
         aPt  = a.data() + aCol;
         aInc = a.cols();
     }
-    else if (MatLeft::Options == Eigen::ColMajor)
+    else if (MatLeft::Options == ColMajor)
     {
         aPt  = a.data() + aCol*a.rows();
         aInc = 1;
     }
-    if (MatRight::Options == Eigen::RowMajor)
+    if (MatRight::Options == RowMajor)
     {
         bPt  = b.data() + aCol*b.cols();
         bInc = 1;
     }
-    else if (MatRight::Options == Eigen::ColMajor)
+    else if (MatRight::Options == ColMajor)
     {
         bPt  = b.data() + aCol;
         bInc = b.rows();
@@ -196,20 +201,20 @@ void fullTrBenchmark(const unsigned int ni, const unsigned int nj, const unsigne
     {
         std::cout << "==== tr(A*B) benchmarks" << std::endl;
         std::cout << "A matrices use ";
-        if (MatLeft::Options == Eigen::RowMajor)
+        if (MatLeft::Options == RowMajor)
         {
             std::cout << "row-major ordering" << std::endl;
         }
-        else if (MatLeft::Options == Eigen::ColMajor)
+        else if (MatLeft::Options == ColMajor)
         {
             std::cout << "col-major ordering" << std::endl;
         }
         std::cout << "B matrices use ";
-        if (MatRight::Options == Eigen::RowMajor)
+        if (MatRight::Options == RowMajor)
         {
             std::cout << "row-major ordering" << std::endl;
         }
-        else if (MatRight::Options == Eigen::ColMajor)
+        else if (MatRight::Options == ColMajor)
         {
             std::cout << "col-major ordering" << std::endl;
         }
@@ -229,7 +234,7 @@ void fullTrBenchmark(const unsigned int ni, const unsigned int nj, const unsigne
         auto nr = a.rows(), nc = a.cols();
         
         res = 0.;
-        parallel_for (unsigned int i = 0; i < nr; ++i)
+        thread_for(i,nr,
         {
             ComplexD tmp = 0.;
 
@@ -237,11 +242,11 @@ void fullTrBenchmark(const unsigned int ni, const unsigned int nj, const unsigne
             {
                 tmp += a(i, j)*b(j, i);
             }
-            parallel_critical
+            thread_critical
             {
                 res += tmp;
             }
-        }
+        });
     });
     trBenchmark("Naive loop cols first", left, right, ref,
     [](ComplexD &res, const MatLeft &a, const MatRight &b)
@@ -249,7 +254,7 @@ void fullTrBenchmark(const unsigned int ni, const unsigned int nj, const unsigne
         auto nr = a.rows(), nc = a.cols();
         
         res = 0.;
-        parallel_for (unsigned int j = 0; j < nc; ++j)
+        thread_for(j,nc,
         {
             ComplexD tmp = 0.;
 
@@ -257,11 +262,11 @@ void fullTrBenchmark(const unsigned int ni, const unsigned int nj, const unsigne
             {
                 tmp += a(i, j)*b(j, i);
             }        
-            parallel_critical
+            thread_critical
             {
                 res += tmp;
             }
-        }
+        });
     });
     trBenchmark("Eigen tr(A*B)", left, right, ref,
     [](ComplexD &res, const MatLeft &a, const MatRight &b)
@@ -272,31 +277,31 @@ void fullTrBenchmark(const unsigned int ni, const unsigned int nj, const unsigne
     [](ComplexD &res, const MatLeft &a, const MatRight &b)
     {
         res = 0.;
-        parallel_for (unsigned int r = 0; r < a.rows(); ++r)
+        thread_for(r,a.rows(),
         {
             ComplexD tmp;
 
             tmp = a.row(r).conjugate().dot(b.col(r));
-            parallel_critical
+            thread_critical
             {
                 res += tmp;
             }
-        }
+        });
     });
     trBenchmark("Eigen col-wise dot", left, right, ref,
     [](ComplexD &res, const MatLeft &a, const MatRight &b)
     {
         res = 0.;
-        parallel_for (unsigned int c = 0; c < a.cols(); ++c)
+        thread_for(c,a.cols(),
         {
             ComplexD tmp;
 
             tmp = a.col(c).conjugate().dot(b.row(c));
-            parallel_critical
+            thread_critical
             {
                 res += tmp;
             }
-        }
+        });
     });
     trBenchmark("Eigen Hadamard", left, right, ref,
     [](ComplexD &res, const MatLeft &a, const MatRight &b)
@@ -308,31 +313,31 @@ void fullTrBenchmark(const unsigned int ni, const unsigned int nj, const unsigne
     [](ComplexD &res, const MatLeft &a, const MatRight &b)
     {
         res = 0.;
-        parallel_for (unsigned int r = 0; r < a.rows(); ++r)
+        thread_for(r,a.rows(),
         {
             ComplexD tmp;
 
             zdotuRow(tmp, r, a, b);
-            parallel_critical
+            thread_critical
             {
                 res += tmp;
             }
-        }
+        });
     });
     trBenchmark("MKL col-wise zdotu", left, right, ref,
     [](ComplexD &res, const MatLeft &a, const MatRight &b)
     {
         res = 0.;
-        parallel_for (unsigned int c = 0; c < a.cols(); ++c)
+        thread_for(c,a.cols(),
         {
             ComplexD tmp;
 
             zdotuCol(tmp, c, a, b);
-            parallel_critical
+            thread_critical
             {
                 res += tmp;
             }
-        }
+        });
     });
 #endif
     BARRIER();
@@ -356,11 +361,11 @@ void fullMulBenchmark(const unsigned int ni, const unsigned int nj, const unsign
     {
         std::cout << "==== A*B benchmarks" << std::endl;
         std::cout << "all matrices use ";
-        if (Mat::Options == Eigen::RowMajor)
+        if (Mat::Options == RowMajor)
         {
             std::cout << "row-major ordering" << std::endl;
         }
-        else if (Mat::Options == Eigen::ColMajor)
+        else if (Mat::Options == ColMajor)
         {
             std::cout << "col-major ordering" << std::endl;
         }
@@ -383,13 +388,13 @@ void fullMulBenchmark(const unsigned int ni, const unsigned int nj, const unsign
     [](Mat &res, const Mat &a, const Mat &b)
     {
         const ComplexD one(1., 0.), zero(0., 0.);
-        if (Mat::Options == Eigen::RowMajor)
+        if (Mat::Options == RowMajor)
         {
             cblas_zgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, a.rows(), b.cols(),
                         a.cols(), &one, a.data(), a.cols(), b.data(), b.cols(), &zero,
                         res.data(), res.cols());
         }
-        else if (Mat::Options == Eigen::ColMajor)
+        else if (Mat::Options == ColMajor)
         {
             cblas_zgemm(CblasColMajor, CblasNoTrans, CblasNoTrans, a.rows(), b.cols(),
                         a.cols(), &one, a.data(), a.rows(), b.data(), b.rows(), &zero,
@@ -430,11 +435,7 @@ int main(int argc, char *argv[])
 
         std::cout << nMpi << " MPI processes" << std::endl;
 #ifdef GRID_OMP
-        #pragma omp parallel
-        {
-            #pragma omp single
-            std::cout << omp_get_num_threads() << " threads\n" << std::endl; 
-        }
+	std::cout << omp_get_num_threads() << " threads\n" << std::endl; 
 #else
         std::cout << "Single-threaded\n" << std::endl; 
 #endif
diff --git a/Hadrons/Utilities/EigenPackCast.cc b/Hadrons/Utilities/EigenPackCast.cc
index d76dc6d2..09a9cc43 100644
--- a/Hadrons/Utilities/EigenPackCast.cc
+++ b/Hadrons/Utilities/EigenPackCast.cc
@@ -29,7 +29,6 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #include <Hadrons/Environment.hpp>
 
 using namespace Grid;
-using namespace QCD;
 using namespace Hadrons;
 
 template <typename FOut, typename FIn>
diff --git a/Hadrons/Utilities/HadronsXmlRun.cc b/Hadrons/Utilities/HadronsXmlRun.cc
index d13dbfe6..a81139af 100644
--- a/Hadrons/Utilities/HadronsXmlRun.cc
+++ b/Hadrons/Utilities/HadronsXmlRun.cc
@@ -29,7 +29,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #include <Hadrons/Application.hpp>
 
 using namespace Grid;
-using namespace QCD;
+ 
 using namespace Hadrons;
 
 int main(int argc, char *argv[])
diff --git a/Hadrons/VirtualMachine.cc b/Hadrons/VirtualMachine.cc
index 8a44ba69..e09dfdee 100644
--- a/Hadrons/VirtualMachine.cc
+++ b/Hadrons/VirtualMachine.cc
@@ -31,7 +31,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #include <Hadrons/ModuleFactory.hpp>
 
 using namespace Grid;
-using namespace QCD;
+ 
 using namespace Hadrons;
 
 /******************************************************************************
@@ -586,7 +586,7 @@ VirtualMachine::Program VirtualMachine::schedule(const GeneticPar &par)
     LOG(Message) << "  max. cst. generation= " << par.maxCstGen << std::endl;
     LOG(Message) << "         mutation rate= " << par.mutationRate << std::endl;
     
-    unsigned int          k = 0, gen, prevPeak, nCstPeak = 0;
+    unsigned int          gen, prevPeak, nCstPeak = 0;
     std::random_device    rd;
     Scheduler::Parameters gpar;
     
diff --git a/Makefile.am b/Makefile.am
index abf38f12..710514bc 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -11,6 +11,6 @@ tests-local: all
 bench-local: all
 check-local: all
 
-AM_CXXFLAGS += -I$(top_builddir)/include
+AM_CXXFLAGS += -I$(top_builddir)
 
 ACLOCAL_AMFLAGS = -I m4
diff --git a/README b/README
deleted file mode 120000
index 42061c01..00000000
--- a/README
+++ /dev/null
@@ -1 +0,0 @@
-README.md
\ No newline at end of file
diff --git a/README b/README
new file mode 100644
index 00000000..86506f52
--- /dev/null
+++ b/README
@@ -0,0 +1,407 @@
+# Grid [![Teamcity status](http://ci.cliath.ph.ed.ac.uk/app/rest/builds/aggregated/strob:(buildType:(affectedProject(id:Grid)),branch:name:develop)/statusIcon.svg)](http://ci.cliath.ph.ed.ac.uk/project.html?projectId=Grid&tab=projectOverview) [![Travis status](https://travis-ci.org/paboyle/Grid.svg?branch=develop)](https://travis-ci.org/paboyle/Grid)
+
+**Data parallel C++ mathematical object library.**
+
+License: GPL v2.
+
+Last update June 2017.
+
+_Please do not send pull requests to the `master` branch which is reserved for releases._
+
+
+
+### Description
+This library provides data parallel C++ container classes with internal memory layout
+that is transformed to map efficiently to SIMD architectures. CSHIFT facilities
+are provided, similar to HPF and cmfortran, and user control is given over the mapping of
+array indices to both MPI tasks and SIMD processing elements.
+
+* Identically shaped arrays then be processed with perfect data parallelisation.
+* Such identically shaped arrays are called conformable arrays.
+
+The transformation is based on the observation that Cartesian array processing involves
+identical processing to be performed on different regions of the Cartesian array.
+
+The library will both geometrically decompose into MPI tasks and across SIMD lanes.
+Local vector loops are parallelised with OpenMP pragmas.
+
+Data parallel array operations can then be specified with a SINGLE data parallel paradigm, but
+optimally use MPI, OpenMP and SIMD parallelism under the hood. This is a significant simplification
+for most programmers.
+
+The layout transformations are parametrised by the SIMD vector length. This adapts according to the architecture.
+Presently SSE4, ARM NEON (128 bits) AVX, AVX2, QPX (256 bits), IMCI and AVX512 (512 bits) targets are supported.
+
+These are presented as `vRealF`, `vRealD`, `vComplexF`, and `vComplexD` internal vector data types. 
+The corresponding scalar types are named `RealF`, `RealD`, `ComplexF` and `ComplexD`.
+
+MPI, OpenMP, and SIMD parallelism are present in the library.
+Please see [this paper](https://arxiv.org/abs/1512.03487) for more detail.
+
+
+### Compilers
+
+Intel ICPC v16.0.3 and later
+
+Clang v3.5 and later (need 3.8 and later for OpenMP)
+
+GCC   v4.9.x (recommended)
+
+GCC   v6.3 and later
+
+### Important: 
+
+Some versions of GCC appear to have a bug under high optimisation (-O2, -O3).
+
+The safety of these compiler versions cannot be guaranteed at this time. Follow Issue 100 for details and updates.
+
+GCC   v5.x
+
+GCC   v6.1, v6.2
+
+### Bug report
+
+_To help us tracking and solving more efficiently issues with Grid, please report problems using the issue system of GitHub rather than sending emails to Grid developers._
+
+When you file an issue, please go though the following checklist:
+
+1. Check that the code is pointing to the `HEAD` of `develop` or any commit in `master` which is tagged with a version number. 
+2. Give a description of the target platform (CPU, network, compiler). Please give the full CPU part description, using for example `cat /proc/cpuinfo | grep 'model name' | uniq` (Linux) or `sysctl machdep.cpu.brand_string` (macOS) and the full output the `--version` option of your compiler.
+3. Give the exact `configure` command used.
+4. Attach `config.log`.
+5. Attach `grid.config.summary`.
+6. Attach the output of `make V=1`.
+7. Describe the issue and any previous attempt to solve it. If relevant, show how to reproduce the issue using a minimal working example.
+
+### Required libraries
+Grid requires:
+
+[GMP](https://gmplib.org/), 
+
+[MPFR](http://www.mpfr.org/) 
+
+Bootstrapping grid downloads and uses for internal dense matrix (non-QCD operations) the Eigen library.
+
+Grid optionally uses:
+
+[HDF5](https://support.hdfgroup.org/HDF5/)  
+
+[LIME](http://usqcd-software.github.io/c-lime/) for ILDG and SciDAC file format support. 
+
+[FFTW](http://www.fftw.org) either generic version or via the Intel MKL library.
+
+LAPACK either generic version or Intel MKL library.
+
+
+### Quick start
+First, start by cloning the repository:
+
+``` bash
+git clone https://github.com/paboyle/Grid.git
+```
+
+Then enter the cloned directory and set up the build system:
+
+``` bash
+cd Grid
+./bootstrap.sh
+```
+
+Now you can execute the `configure` script to generate makefiles (here from a build directory):
+
+``` bash
+mkdir build; cd build
+../configure --enable-precision=double --enable-simd=AVX --enable-comms=mpi-auto --prefix=<path>
+```
+
+where `--enable-precision=` set the default precision,
+`--enable-simd=` set the SIMD type, `--enable-
+comms=`, and `<path>` should be replaced by the prefix path where you want to
+install Grid. Other options are detailed in the next section, you can also use `configure
+--help` to display them. Like with any other program using GNU autotool, the
+`CXX`, `CXXFLAGS`, `LDFLAGS`, ... environment variables can be modified to
+customise the build.
+
+Finally, you can build, check, and install Grid:
+
+``` bash
+make; make check; make install
+```
+
+To minimise the build time, only the tests at the root of the `tests` directory are built by default. If you want to build tests in the sub-directory `<subdir>` you can execute:
+
+``` bash
+make -C tests/<subdir> tests
+```
+If you want to build all the tests at once just use `make tests`.
+
+### Build configuration options
+
+- `--prefix=<path>`: installation prefix for Grid.
+- `--with-gmp=<path>`: look for GMP in the UNIX prefix `<path>`
+- `--with-mpfr=<path>`: look for MPFR in the UNIX prefix `<path>`
+- `--with-fftw=<path>`: look for FFTW in the UNIX prefix `<path>`
+- `--enable-lapack[=<path>]`: enable LAPACK support in Lanczos eigensolver. A UNIX prefix containing the library can be specified (optional).
+- `--enable-mkl[=<path>]`: use Intel MKL for FFT (and LAPACK if enabled) routines. A UNIX prefix containing the library can be specified (optional).
+- `--enable-numa`: enable NUMA first touch optimisation
+- `--enable-simd=<code>`: setup Grid for the SIMD target `<code>` (default: `GEN`). A list of possible SIMD targets is detailed in a section below.
+- `--enable-gen-simd-width=<size>`: select the size (in bytes) of the generic SIMD vector type (default: 32 bytes).
+- `--enable-precision={single|double}`: set the default precision (default: `double`).
+- `--enable-precision=<comm>`: Use `<comm>` for message passing (default: `none`). A list of possible SIMD targets is detailed in a section below.
+- `--enable-rng={sitmo|ranlux48|mt19937}`: choose the RNG (default: `sitmo `).
+- `--disable-timers`: disable system dependent high-resolution timers.
+- `--enable-chroma`: enable Chroma regression tests.
+- `--enable-doxygen-doc`: enable the Doxygen documentation generation (build with `make doxygen-doc`)
+
+### Possible communication interfaces
+
+The following options can be use with the `--enable-comms=` option to target different communication interfaces:
+
+| `<comm>`       | Description                                                   |
+| -------------- | ------------------------------------------------------------- |
+| `none`         | no communications                                             |
+| `mpi[-auto]`   | MPI communications                                            |
+| `mpi3[-auto]`  | MPI communications using MPI 3 shared memory                  |
+| `shmem `       | Cray SHMEM communications                                     |
+
+For the MPI interfaces the optional `-auto` suffix instructs the `configure` scripts to determine all the necessary compilation and linking flags. This is done by extracting the informations from the MPI wrapper specified in the environment variable `MPICXX` (if not specified `configure` will scan though a list of default names). The `-auto` suffix is not supported by the Cray environment wrapper scripts. Use the standard versions instead.  
+
+### Possible SIMD types
+
+The following options can be use with the `--enable-simd=` option to target different SIMD instruction sets:
+
+| `<code>`    | Description                            |
+| ----------- | -------------------------------------- |
+| `GEN`       | generic portable vector code           |
+| `SSE4`      | SSE 4.2 (128 bit)                      |
+| `AVX`       | AVX (256 bit)                          |
+| `AVXFMA`    | AVX (256 bit) + FMA                    |
+| `AVXFMA4`   | AVX (256 bit) + FMA4                   |
+| `AVX2`      | AVX 2 (256 bit)                        |
+| `AVX512`    | AVX 512 bit                            |
+| `NEONv8`    | [ARM NEON](http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.den0024a/ch07s03.html) (128 bit)                     |
+| `QPX`       | IBM QPX (256 bit)                      |
+
+Alternatively, some CPU codenames can be directly used:
+
+| `<code>`    | Description                            |
+| ----------- | -------------------------------------- |
+| `KNL`       | [Intel Xeon Phi codename Knights Landing](http://ark.intel.com/products/codename/48999/Knights-Landing) |
+| `SKL`       | [Intel Skylake with AVX512 extensions](https://ark.intel.com/products/codename/37572/Skylake#@server) |
+| `BGQ`       | Blue Gene/Q                            |
+
+#### Notes:
+- We currently support AVX512 for the Intel compiler and GCC (KNL and SKL target). Support for clang will appear in future versions of Grid when the AVX512 support in the compiler will be more advanced.
+- For BG/Q only [bgclang](http://trac.alcf.anl.gov/projects/llvm-bgq) is supported. We do not presently plan to support more compilers for this platform.
+- BG/Q performances are currently rather poor. This is being investigated for future versions.
+- The vector size for the `GEN` target can be specified with the `configure` script option `--enable-gen-simd-width`.
+
+### Build setup for Intel Knights Landing platform
+
+The following configuration is recommended for the Intel Knights Landing platform:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=KNL        \
+             --enable-comms=mpi-auto  \
+             --enable-mkl             \
+             CXX=icpc MPICXX=mpiicpc
+```
+The MKL flag enables use of BLAS and FFTW from the Intel Math Kernels Library.
+
+If you are working on a Cray machine that does not use the `mpiicpc` wrapper, please use:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=KNL        \
+             --enable-comms=mpi       \
+             --enable-mkl             \
+             CXX=CC CC=cc
+```
+
+If gmp and mpfr are NOT in standard places (/usr/) these flags may be needed:
+``` bash
+               --with-gmp=<path>        \
+               --with-mpfr=<path>       \
+```
+where `<path>` is the UNIX prefix where GMP and MPFR are installed. 
+
+Knight's Landing with Intel Omnipath adapters with two adapters per node 
+presently performs better with use of more than one rank per node, using shared memory 
+for interior communication. This is the mpi3 communications implementation. 
+We recommend four ranks per node for best performance, but optimum is local volume dependent.
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=KNL        \
+             --enable-comms=mpi3-auto \
+             --enable-mkl             \
+             CC=icpc MPICXX=mpiicpc 
+```
+
+### Build setup for Intel Haswell Xeon platform
+
+The following configuration is recommended for the Intel Haswell platform:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX2       \
+             --enable-comms=mpi3-auto \
+             --enable-mkl             \
+             CXX=icpc MPICXX=mpiicpc
+```
+The MKL flag enables use of BLAS and FFTW from the Intel Math Kernels Library.
+
+If gmp and mpfr are NOT in standard places (/usr/) these flags may be needed:
+``` bash
+               --with-gmp=<path>        \
+               --with-mpfr=<path>       \
+```
+where `<path>` is the UNIX prefix where GMP and MPFR are installed. 
+
+If you are working on a Cray machine that does not use the `mpiicpc` wrapper, please use:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX2       \
+             --enable-comms=mpi3      \
+             --enable-mkl             \
+             CXX=CC CC=cc
+```
+Since Dual socket nodes are commonplace, we recommend MPI-3 as the default with the use of 
+one rank per socket. If using the Intel MPI library, threads should be pinned to NUMA domains using
+```
+        export I_MPI_PIN=1
+```
+This is the default.
+
+### Build setup for Intel Skylake Xeon platform
+
+The following configuration is recommended for the Intel Skylake platform:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX512     \
+             --enable-comms=mpi3      \
+             --enable-mkl             \
+             CXX=mpiicpc
+```
+The MKL flag enables use of BLAS and FFTW from the Intel Math Kernels Library.
+
+If gmp and mpfr are NOT in standard places (/usr/) these flags may be needed:
+``` bash
+               --with-gmp=<path>        \
+               --with-mpfr=<path>       \
+```
+where `<path>` is the UNIX prefix where GMP and MPFR are installed. 
+
+If you are working on a Cray machine that does not use the `mpiicpc` wrapper, please use:
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX512     \
+             --enable-comms=mpi3      \
+             --enable-mkl             \
+             CXX=CC CC=cc
+```
+Since Dual socket nodes are commonplace, we recommend MPI-3 as the default with the use of 
+one rank per socket. If using the Intel MPI library, threads should be pinned to NUMA domains using
+``` 
+        export I_MPI_PIN=1
+```
+This is the default. 
+
+#### Expected Skylake Gold 6148 dual socket (single prec, single node 20+20 cores) performance using NUMA MPI mapping): 
+
+mpirun -n 2 benchmarks/Benchmark_dwf --grid 16.16.16.16 --mpi 2.1.1.1 --cacheblocking 2.2.2.2 --dslash-asm --shm 1024 --threads 18 
+
+TBA
+
+
+### Build setup for AMD EPYC / RYZEN
+
+The AMD EPYC is a multichip module comprising 32 cores spread over four distinct chips each with 8 cores.
+So, even with a single socket node there is a quad-chip module. Dual socket nodes with 64 cores total
+are common. Each chip within the module exposes a separate NUMA domain.
+There are four NUMA domains per socket and we recommend one MPI rank per NUMA domain.
+MPI-3 is recommended with the use of four ranks per socket,
+and 8 threads per rank. 
+
+The following configuration is recommended for the AMD EPYC platform.
+
+``` bash
+../configure --enable-precision=double\
+             --enable-simd=AVX2       \
+             --enable-comms=mpi3 \
+             CXX=mpicxx 
+```
+
+If gmp and mpfr are NOT in standard places (/usr/) these flags may be needed:
+``` bash
+               --with-gmp=<path>        \
+               --with-mpfr=<path>       \
+```
+where `<path>` is the UNIX prefix where GMP and MPFR are installed. 
+
+Using MPICH and g++ v4.9.2, best performance can be obtained using explicit GOMP_CPU_AFFINITY flags for each MPI rank.
+This can be done by invoking MPI on a wrapper script omp_bind.sh to handle this. 
+
+It is recommended to run 8 MPI ranks on a single dual socket AMD EPYC, with 8 threads per rank using MPI3 and
+shared memory to communicate within this node:
+
+mpirun -np 8 ./omp_bind.sh ./Benchmark_dwf --mpi 2.2.2.1 --dslash-unroll --threads 8 --grid 16.16.16.16 --cacheblocking 4.4.4.4 
+
+Where omp_bind.sh does the following:
+```
+#!/bin/bash
+
+numanode=` expr $PMI_RANK % 8 `
+basecore=`expr $numanode \* 16`
+core0=`expr $basecore + 0 `
+core1=`expr $basecore + 2 `
+core2=`expr $basecore + 4 `
+core3=`expr $basecore + 6 `
+core4=`expr $basecore + 8 `
+core5=`expr $basecore + 10 `
+core6=`expr $basecore + 12 `
+core7=`expr $basecore + 14 `
+
+export GOMP_CPU_AFFINITY="$core0 $core1 $core2 $core3 $core4 $core5 $core6 $core7"
+echo GOMP_CUP_AFFINITY $GOMP_CPU_AFFINITY
+
+$@
+```
+
+Performance:
+
+#### Expected AMD EPYC 7601 dual socket (single prec, single node 32+32 cores) performance using NUMA MPI mapping): 
+
+mpirun  -np 8 ./omp_bind.sh ./Benchmark_dwf --threads 8 --mpi 2.2.2.1 --dslash-unroll --grid 16.16.16.16 --cacheblocking 4.4.4.4
+
+TBA
+
+### Build setup for BlueGene/Q
+
+To be written...
+
+### Build setup for ARM Neon
+
+To be written...
+
+### Build setup for laptops, other compilers, non-cluster builds
+
+Many versions of g++ and clang++ work with Grid, and involve merely replacing CXX (and MPICXX),
+and omit the enable-mkl flag. 
+
+Single node builds are enabled with 
+```
+            --enable-comms=none
+```
+
+FFTW support that is not in the default search path may then enabled with
+```
+    --with-fftw=<installpath>
+```
+
+BLAS will not be compiled in by default, and Lanczos will default to Eigen diagonalisation.
+
diff --git a/TODO b/TODO
index 746302ca..f1175560 100644
--- a/TODO
+++ b/TODO
@@ -1,8 +1,166 @@
-TODO:
----------------
+=========================================================
+General
+=========================================================
 
-Code item work list
+- Make representations code take Gimpl
+- Simplify the HMCand remove modules
+- Lattice_arith - are the mult, mac etc.. still needed after ET engine?
+- Lattice_rng
+- Lattice_transfer.h
+- accelerate A2Autils -- off critical path for HMC
 
+=========================================================
+GPU branch code item work list
+=========================================================
+
+* sum_cpu promote to double during summation for increased precisoin.
+* Introduce sumD & ReduceD 
+* GPU sum is probably better currently.
+
+* Accelerate the cshift & benchmark
+
+* 0) Single GPU
+- 128 bit integer table load in GPU code.
+  - ImprovedStaggered accelerate & measure perf
+  - Gianluca's changes to Cayley into gpu-port
+  - Mobius kernel fusion.                     -- Gianluca?
+  - Lebesque order reintroduction. StencilView should have pointer to it
+  - Lebesgue reorder in all kernels
+
+* 3) Comms/NVlink
+- OpenMP tasks to run comms threads. Experiment with it 
+- Remove explicit openMP in staggered. 
+- Single parallel region around both the Kernel call and the comms.
+- Fix the halo exchange SIMT loop
+- Stencil gather ??
+- SIMD dirs in stencil
+
+* 4) ET enhancements
+- eval -> scalar ops in ET engine
+- coalescedRead, coalescedWrite in expressions.
+
+* 5) Misc
+- Conserved current clean up.
+- multLinkProp eliminate
+
+8) Merge develop and test HMC
+
+9) Gamma tables on GPU; check this. Appear to work, but no idea why. Are these done on CPU?
+
+10) Audit
+-     pragma once uniformly
+-     Audit NAMESPACE CHANGES
+-     Audit changes
+
+-----
+Gianluca's changes
+- Performance impact of construct in aligned allocator???
+---------
+
+- merge2 where is it used. Audit routines, comment out and check compile.
+
+-----------------------------
+DONE:
+-----------------------------
+=============================================================================================
+AUDIT ContractWWVV with respect to develop    -- DONE
+- GPU accelerate EOFA                                                  -- DONE
+- LinalgUtils  ssp loop not offloaded                                  -- DONE
+- coalescedRead <- threadIdx.x                                         -- DONE
+- Stencil.h : Thread loops in exchange code. Need to offload these     -- DONE ; pending debug
+- Mobius/Domain EOFA cache header implementaiotn has thread_loop       -- DONE ; pending test
+- Differentiate non-temporal coalescedWrite from temporal              -- DONE
+
+- Clean up PRAGMAS, and SIMT_loop                                      -- DONE
+  thread_loop interface revisit.
+  _foreach
+  _for
+
+- Staggered kernels -> GPU coalesced loop, loop in kernels -- DONE
+- Staggered kernels inline for GPU -- DONE
+
+-- Common source GPU and CPU generic kernels???                  ---- DONE
+--   - Uniform coding between GPU kernels and CPU kernels attempt  ---- DONE, got faster !
+
+-- Figure what to do about "multLinkGpu" etc.. in FermionOperatorImpl. -- DONE
+-- Gparity is the awkward one                                          -- DONE
+-- Solve non-Gparity first.                                            -- DONE
+-- Simplify the operator IMPL support                                  -- DONE
+-- 
+--
+-- Investigate why slower than september     --- DONE
+--
+-- Single GPU simd target (VGPU) --- DONE
+--
+-- Reread WilsonKernels and check diffs -- DONE
+--
+
+- AVX512 still broken, lebesgue order missing ?
+
+* Gianluca merger
+  - Cayley coefficients -> GPU retention or prefetch
+  - Make GPU offload reductions deterministic -- Gianluca merge
+  - Lattice_reduction - remnant thread_loops must offload. Audit thread_loop in main code for non-accelerated code  
+  - Inner product compare to Summit inner product optimisation
+  - CayleyFermion5D.cc - flop count line 166 odd. Shouldn't depend on arch
+  -                    - Review Vector use
+  - CayleyFermion5D.h  - DperpGPU unify coding style
+
+- Committed my modifications
+- Accelerate non-dslash elements of Mobius; check accelerator_loop uniformly used in fermion operators
+   - Merged Gianluca modifications
+- Verify HMC one flavour ratio
+- GPU offload reductions: using thrust::reduce?
+- Deprecate JSON.
+- pugixml difficult.
+- Eigen problematic.
+- Audit HMC timestep / traj length size
+- GPU offload reductions; thrust initial ; inclusive_scan vs reduce?
+- Pragmas.h - prune and remove strong_inline (?)
+- GPU offload reductions; thrust initial ; inclusive_scan vs reduce?
+- Remove old parallel_for macros, fix errors
+- - Need (1) omp parallel for     <-- thread_loop
+- -      (2) omp for
+- -      (3) omp for collapse(n)
+- -      (4) omp parallel for collapse(n)
+- - Only (1) has a natural mirror in accelerator_loop
+- - Nested loop macros get cumbersome made a generic interface for N deep
+- - Don't like thread_region and thread_loop_in_region
+- - Could replace with 
+
+    thread_nested(1, 
+      for {
+
+      }
+    );
+    thread_nested(2,
+      for (){
+        for (){
+
+	}
+      }
+    );
+
+    and same "in_region".
+
+
+-----------------------------
+Physics item work list:
+-----------------------------
+
+2)- Consistent linear solver flop count/rate -- PARTIAL, time but no flop/s yet
+4)- Multigrid Wilson and DWF, compare to other Multigrid implementations
+5)- HDCR resume
+
+-----------------------------
+
+
+Nov 2018
+
+1)- BG/Q port and check ; Andrew says ok.
+3)- Physical propagator interface  -- DONE
+
+DONE
 a) namespaces & indentation
  GRID_BEGIN_NAMESPACE();
  GRID_END_NAMESPACE();
@@ -16,14 +174,6 @@ b) GPU branch
 - Start port once Nvidia box is up
 - Cut down volume of code for first port? How?
 
-Physics item work list:
-
-1)- BG/Q port and check ; Andrew says ok.
-2)- Consistent linear solver flop count/rate -- PARTIAL, time but no flop/s yet
-3)- Physical propagator interface
-4)- Multigrid Wilson and DWF, compare to other Multigrid implementations
-5)- HDCR resume
-
 ----------------------------
 Recent DONE 
 -- RNG I/O in ILDG/SciDAC (minor) 
@@ -162,6 +312,8 @@ RECENT
 
 
 DONE:
+
+
 - MultiArray -- MultiRHS done
 - ConjugateGradientMultiShift -- DONE
 - MCR                         -- DONE
diff --git a/benchmarks/Benchmark_IO.cc b/benchmarks/Benchmark_IO.cc
index 58e0340d..3d3b0ce0 100644
--- a/benchmarks/Benchmark_IO.cc
+++ b/benchmarks/Benchmark_IO.cc
@@ -6,7 +6,6 @@
 #endif
 
 using namespace Grid;
-using namespace QCD;
 
 std::string filestem(const int l)
 {
diff --git a/benchmarks/Benchmark_IO.hpp b/benchmarks/Benchmark_IO.hpp
index 9565c928..91fcb61f 100644
--- a/benchmarks/Benchmark_IO.hpp
+++ b/benchmarks/Benchmark_IO.hpp
@@ -18,7 +18,7 @@ template <typename Field>
 void limeWrite(const std::string filestem, Field &vec)
 {
   emptyUserRecord   record;
-  QCD::ScidacWriter binWriter(vec._grid->IsBoss());
+  ScidacWriter binWriter(vec.Grid()->IsBoss());
 
   binWriter.open(filestem + ".bin");
   binWriter.writeScidacFieldRecord(vec, record);
@@ -29,7 +29,7 @@ template <typename Field>
 void limeRead(Field &vec, const std::string filestem)
 {
   emptyUserRecord   record;
-  QCD::ScidacReader binReader;
+  ScidacReader binReader;
 
   binReader.open(filestem + ".bin");
   binReader.readScidacFieldRecord(vec, record);
@@ -44,18 +44,18 @@ inline void makeGrid(std::shared_ptr<GridBase> &gPt,
   {
     if (Ls > 1)
     {
-      gPt.reset(QCD::SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls, gBasePt.get()));
+      gPt.reset(SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls, gBasePt.get()));
     }
     else
     {
-      gPt.reset(QCD::SpaceTimeGrid::makeFourDimRedBlackGrid(gBasePt.get()));
+      gPt.reset(SpaceTimeGrid::makeFourDimRedBlackGrid(gBasePt.get()));
     }
   }
   else
   {
     if (Ls > 1)
     {
-        gPt.reset(QCD::SpaceTimeGrid::makeFiveDimGrid(Ls, gBasePt.get()));
+        gPt.reset(SpaceTimeGrid::makeFiveDimGrid(Ls, gBasePt.get()));
     }
     else
     {
@@ -65,13 +65,13 @@ inline void makeGrid(std::shared_ptr<GridBase> &gPt,
 }
 
 template <typename Field>
-void writeBenchmark(const std::vector<int> &latt, const std::string filename,
+void writeBenchmark(const Coordinate &latt, const std::string filename,
                     const WriterFn<Field> &write, 
                     const unsigned int Ls = 1, const bool rb = false)
 {
   auto                           mpi  = GridDefaultMpi();
   auto                           simd = GridDefaultSimd(latt.size(), Field::vector_type::Nsimd());
-  std::shared_ptr<GridCartesian> gBasePt(QCD::SpaceTimeGrid::makeFourDimGrid(latt, simd, mpi));
+  std::shared_ptr<GridCartesian> gBasePt(SpaceTimeGrid::makeFourDimGrid(latt, simd, mpi));
   std::shared_ptr<GridBase>      gPt;
 
   makeGrid(gPt, gBasePt, Ls, rb);
@@ -85,13 +85,13 @@ void writeBenchmark(const std::vector<int> &latt, const std::string filename,
 }
 
 template <typename Field>
-void readBenchmark(const std::vector<int> &latt, const std::string filename,
+void readBenchmark(const Coordinate &latt, const std::string filename,
                    const ReaderFn<Field> &read, 
                    const unsigned int Ls = 1, const bool rb = false)
 {
   auto                           mpi  = GridDefaultMpi();
   auto                           simd = GridDefaultSimd(latt.size(), Field::vector_type::Nsimd());
-  std::shared_ptr<GridCartesian> gBasePt(QCD::SpaceTimeGrid::makeFourDimGrid(latt, simd, mpi));
+  std::shared_ptr<GridCartesian> gBasePt(SpaceTimeGrid::makeFourDimGrid(latt, simd, mpi));
   std::shared_ptr<GridBase>      gPt;
 
   makeGrid(gPt, gBasePt, Ls, rb);
diff --git a/benchmarks/Benchmark_IO_vs_dir.cc b/benchmarks/Benchmark_IO_vs_dir.cc
index 15f6194f..cb4831ed 100644
--- a/benchmarks/Benchmark_IO_vs_dir.cc
+++ b/benchmarks/Benchmark_IO_vs_dir.cc
@@ -5,7 +5,6 @@
 "============================================================================="
 
 using namespace Grid;
-using namespace QCD;
 
 int main (int argc, char ** argv)
 {
diff --git a/benchmarks/Benchmark_ITT.cc b/benchmarks/Benchmark_ITT.cc
index f811ac32..7ad4a147 100644
--- a/benchmarks/Benchmark_ITT.cc
+++ b/benchmarks/Benchmark_ITT.cc
@@ -28,13 +28,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     /*  END LEGAL */
 #include <Grid/Grid.h>
 
-using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
-
-typedef WilsonFermion5D<DomainWallVec5dImplR> WilsonFermion5DR;
-typedef WilsonFermion5D<DomainWallVec5dImplF> WilsonFermion5DF;
-typedef WilsonFermion5D<DomainWallVec5dImplD> WilsonFermion5DD;
 
 
 std::vector<int> L_list;
@@ -112,8 +106,8 @@ public:
     int nmu=0;
     int maxlat=32;
 
-    std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
-    std::vector<int> mpi_layout  = GridDefaultMpi();
+    Coordinate simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
+    Coordinate mpi_layout  = GridDefaultMpi();
 
     for(int mu=0;mu<Nd;mu++) if (mpi_layout[mu]>1) nmu++;
 
@@ -128,7 +122,7 @@ public:
     for(int lat=4;lat<=maxlat;lat+=4){
       for(int Ls=8;Ls<=8;Ls*=2){
 
-	std::vector<int> latt_size  ({lat*mpi_layout[0],
+	Coordinate latt_size  ({lat*mpi_layout[0],
 	      lat*mpi_layout[1],
 	      lat*mpi_layout[2],
 	      lat*mpi_layout[3]});
@@ -158,11 +152,9 @@ public:
 
 	  dbytes=0;
 	  ncomm=0;
-#ifdef GRID_OMP
-#pragma omp parallel for num_threads(Grid::CartesianCommunicator::nCommThreads)
-#endif
-	  for(int dir=0;dir<8;dir++){
 
+	  thread_for(dir,8,{
+		     
 	    double tbytes;
 	    int mu =dir % 4;
 
@@ -177,26 +169,15 @@ public:
 		int comm_proc = mpi_layout[mu]-1;
 		Grid.ShiftedRanks(mu,comm_proc,xmit_to_rank,recv_from_rank);
 	      }
-#ifdef GRID_OMP
-	int tid = omp_get_thread_num(); 
-#else 
-        int tid = dir;
-#endif
 	      tbytes= Grid.StencilSendToRecvFrom((void *)&xbuf[dir][0], xmit_to_rank,
 						 (void *)&rbuf[dir][0], recv_from_rank,
-						 bytes,tid);
-	  
-#ifdef GRID_OMP
-#pragma omp atomic
-#endif
-	      ncomm++;
-
-#ifdef GRID_OMP
-#pragma omp atomic
-#endif
-	      dbytes+=tbytes;
+						 bytes,dir);
+	      thread_critical {
+		ncomm++;
+		dbytes+=tbytes;
+	      }
 	    }
-	  }
+          });
 	  Grid.Barrier();
 	  double stop=usecond();
 	  t_time[i] = stop-start; // microseconds
@@ -209,7 +190,7 @@ public:
 
 	dbytes=dbytes*ppn;
 	double xbytes    = dbytes*0.5;
-	double rbytes    = dbytes*0.5;
+	//	double rbytes    = dbytes*0.5;
 	double bidibytes = dbytes;
 
 	std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
@@ -233,8 +214,8 @@ public:
     typedef Lattice< iVector< vReal,Nvec> > LatticeVec;
     typedef iVector<vReal,Nvec> Vec;
 
-    std::vector<int> simd_layout = GridDefaultSimd(Nd,vReal::Nsimd());
-    std::vector<int> mpi_layout  = GridDefaultMpi();
+    Coordinate simd_layout = GridDefaultSimd(Nd,vReal::Nsimd());
+    Coordinate mpi_layout  = GridDefaultMpi();
 
     std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
     std::cout<<GridLogMessage << "= Benchmarking a*x + y bandwidth"<<std::endl;
@@ -242,7 +223,7 @@ public:
     std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s"<<"\t\t"<<"Gflop/s"<<"\t\t seconds"<< "\t\tGB/s / node"<<std::endl;
     std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
   
-    uint64_t NP;
+    //    uint64_t NP;
     uint64_t NN;
 
 
@@ -252,11 +233,11 @@ public:
     GridSerialRNG          sRNG;      sRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
     for(int lat=8;lat<=lmax;lat+=4){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
       int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
-      NP= Grid.RankCount();
+      //      NP= Grid.RankCount();
       NN =Grid.NodeCount();
 
       Vec rn ; random(sRNG,rn);
@@ -271,8 +252,11 @@ public:
       double start=usecond();
       for(int i=0;i<Nloop;i++){
 	z=a*x-y;
-        x._odata[0]=z._odata[0]; // force serial dependency to prevent optimise away
-        y._odata[4]=z._odata[4];
+	auto x_v = x.View();
+	auto y_v = y.View();
+	auto z_v = z.View();
+        x_v[0]=z_v[0]; // force serial dependency to prevent optimise away
+        y_v[4]=z_v[4];
       }
       double stop=usecond();
       double time = (stop-start)/Nloop*1000;
@@ -286,9 +270,10 @@ public:
     }
   };
 
+#if 0
   static double DWF5(int Ls,int L)
   {
-    RealD mass=0.1;
+    //    RealD mass=0.1;
     RealD M5  =1.8;
 
     double mflops;
@@ -300,25 +285,25 @@ public:
     // Set/Get the layout & grid size
     ///////////////////////////////////////////////////////
     int threads = GridThread::GetThreads();
-    std::vector<int> mpi = GridDefaultMpi(); assert(mpi.size()==4);
-    std::vector<int> local({L,L,L,L});
+    Coordinate mpi = GridDefaultMpi(); assert(mpi.size()==4);
+    Coordinate local({L,L,L,L});
 
-    GridCartesian         * TmpGrid   = SpaceTimeGrid::makeFourDimGrid(std::vector<int>({64,64,64,64}), 
+    GridCartesian         * TmpGrid   = SpaceTimeGrid::makeFourDimGrid(Coordinate({64,64,64,64}), 
 								       GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
     uint64_t NP = TmpGrid->RankCount();
     uint64_t NN = TmpGrid->NodeCount();
     NN_global=NN;
     uint64_t SHM=NP/NN;
 
-    std::vector<int> internal;
-    if      ( SHM == 1 )   internal = std::vector<int>({1,1,1,1});
-    else if ( SHM == 2 )   internal = std::vector<int>({2,1,1,1});
-    else if ( SHM == 4 )   internal = std::vector<int>({2,2,1,1});
-    else if ( SHM == 8 )   internal = std::vector<int>({2,2,2,1});
+    Coordinate internal;
+    if      ( SHM == 1 )   internal = Coordinate({1,1,1,1});
+    else if ( SHM == 2 )   internal = Coordinate({2,1,1,1});
+    else if ( SHM == 4 )   internal = Coordinate({2,2,1,1});
+    else if ( SHM == 8 )   internal = Coordinate({2,2,2,1});
     else assert(0);
 
-    std::vector<int> nodes({mpi[0]/internal[0],mpi[1]/internal[1],mpi[2]/internal[2],mpi[3]/internal[3]});
-    std::vector<int> latt4({local[0]*nodes[0],local[1]*nodes[1],local[2]*nodes[2],local[3]*nodes[3]});
+    Coordinate nodes({mpi[0]/internal[0],mpi[1]/internal[1],mpi[2]/internal[2],mpi[3]/internal[3]});
+    Coordinate latt4({local[0]*nodes[0],local[1]*nodes[1],local[2]*nodes[2],local[3]*nodes[3]});
 
     ///////// Welcome message ////////////
     std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
@@ -347,8 +332,11 @@ public:
     std::cout << GridLogMessage << "Initialised RNGs" << std::endl;
 
     ///////// Source preparation ////////////
-    LatticeFermion src   (sFGrid); random(RNG5,src);
+    LatticeFermion src   (sFGrid); 
     LatticeFermion tmp   (sFGrid);
+    std::cout << GridLogMessage << "allocated src and tmp" << std::endl;
+    random(RNG5,src);
+    std::cout << GridLogMessage << "intialised random source" << std::endl;
 
     RealD N2 = 1.0/::sqrt(norm2(src));
     src = src*N2;
@@ -376,19 +364,19 @@ public:
 #endif
       controls Cases [] = {
 #ifdef AVX512
-	{ QCD::WilsonKernelsStatic::OptInlineAsm , QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
-	{ QCD::WilsonKernelsStatic::OptInlineAsm , QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{  WilsonKernelsStatic::OptInlineAsm ,  WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{  WilsonKernelsStatic::OptInlineAsm ,  WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
 #endif
-	{ QCD::WilsonKernelsStatic::OptHandUnroll, QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
-	{ QCD::WilsonKernelsStatic::OptHandUnroll, QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
-	{ QCD::WilsonKernelsStatic::OptGeneric   , QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
-	{ QCD::WilsonKernelsStatic::OptGeneric   , QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  }
+	{  WilsonKernelsStatic::OptHandUnroll,  WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{  WilsonKernelsStatic::OptHandUnroll,  WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{  WilsonKernelsStatic::OptGeneric   ,  WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{  WilsonKernelsStatic::OptGeneric   ,  WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  }
       }; 
 
       for(int c=0;c<num_cases;c++) {
 
-	QCD::WilsonKernelsStatic::Comms = Cases[c].CommsOverlap;
-	QCD::WilsonKernelsStatic::Opt   = Cases[c].Opt;
+	 WilsonKernelsStatic::Comms = Cases[c].CommsOverlap;
+	 WilsonKernelsStatic::Opt   = Cases[c].Opt;
 	CartesianCommunicator::SetCommunicatorPolicy(Cases[c].CommsAsynch);
 
 	std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
@@ -464,6 +452,7 @@ public:
     }
     return mflops_best;
   }
+#endif
 
   static double DWF(int Ls,int L, double & robust)
   {
@@ -479,25 +468,25 @@ public:
     // Set/Get the layout & grid size
     ///////////////////////////////////////////////////////
     int threads = GridThread::GetThreads();
-    std::vector<int> mpi = GridDefaultMpi(); assert(mpi.size()==4);
-    std::vector<int> local({L,L,L,L});
+    Coordinate mpi = GridDefaultMpi(); assert(mpi.size()==4);
+    Coordinate local({L,L,L,L});
 
-    GridCartesian         * TmpGrid   = SpaceTimeGrid::makeFourDimGrid(std::vector<int>({64,64,64,64}), 
+    GridCartesian         * TmpGrid   = SpaceTimeGrid::makeFourDimGrid(Coordinate({64,64,64,64}), 
 								       GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
     uint64_t NP = TmpGrid->RankCount();
     uint64_t NN = TmpGrid->NodeCount();
     NN_global=NN;
     uint64_t SHM=NP/NN;
 
-    std::vector<int> internal;
-    if      ( SHM == 1 )   internal = std::vector<int>({1,1,1,1});
-    else if ( SHM == 2 )   internal = std::vector<int>({2,1,1,1});
-    else if ( SHM == 4 )   internal = std::vector<int>({2,2,1,1});
-    else if ( SHM == 8 )   internal = std::vector<int>({2,2,2,1});
+    Coordinate internal;
+    if      ( SHM == 1 )   internal = Coordinate({1,1,1,1});
+    else if ( SHM == 2 )   internal = Coordinate({2,1,1,1});
+    else if ( SHM == 4 )   internal = Coordinate({2,2,1,1});
+    else if ( SHM == 8 )   internal = Coordinate({2,2,2,1});
     else assert(0);
 
-    std::vector<int> nodes({mpi[0]/internal[0],mpi[1]/internal[1],mpi[2]/internal[2],mpi[3]/internal[3]});
-    std::vector<int> latt4({local[0]*nodes[0],local[1]*nodes[1],local[2]*nodes[2],local[3]*nodes[3]});
+    Coordinate nodes({mpi[0]/internal[0],mpi[1]/internal[1],mpi[2]/internal[2],mpi[3]/internal[3]});
+    Coordinate latt4({local[0]*nodes[0],local[1]*nodes[1],local[2]*nodes[2],local[3]*nodes[3]});
 
     ///////// Welcome message ////////////
     std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
@@ -531,9 +520,11 @@ public:
     LatticeFermion tmp   (FGrid);
 
     RealD N2 = 1.0/::sqrt(norm2(src));
+    std::cout<<GridLogMessage << "Normalising src  "<< N2 <<std::endl;
     src = src*N2;
     
     LatticeGaugeField Umu(UGrid);  SU3::HotConfiguration(RNG4,Umu); 
+    
 
     DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
@@ -543,12 +534,14 @@ public:
     {
       LatticeGaugeField Umu5d(FGrid); 
       std::vector<LatticeColourMatrix> U(4,FGrid);
-      for(int ss=0;ss<Umu._grid->oSites();ss++){
+      auto Umu_v = Umu.View();
+      auto Umu5d_v = Umu5d.View();
+      for(int ss=0;ss<Umu.Grid()->oSites();ss++){
 	for(int s=0;s<Ls;s++){
-	  Umu5d._odata[Ls*ss+s] = Umu._odata[ss];
+	  Umu5d_v[Ls*ss+s] = Umu_v[ss];
 	}
       }
-      ref = zero;
+      ref = Zero();
       for(int mu=0;mu<Nd;mu++){
 	U[mu] = PeekIndex<LorentzIndex>(Umu5d,mu);
       }
@@ -584,19 +577,19 @@ public:
 #endif
       controls Cases [] = {
 #ifdef AVX512
-	{ QCD::WilsonKernelsStatic::OptInlineAsm , QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
-	{ QCD::WilsonKernelsStatic::OptInlineAsm , QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{  WilsonKernelsStatic::OptInlineAsm ,  WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{  WilsonKernelsStatic::OptInlineAsm ,  WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
 #endif
-	{ QCD::WilsonKernelsStatic::OptHandUnroll, QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
-	{ QCD::WilsonKernelsStatic::OptHandUnroll, QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
-	{ QCD::WilsonKernelsStatic::OptGeneric   , QCD::WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
-	{ QCD::WilsonKernelsStatic::OptGeneric   , QCD::WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  }
+	{  WilsonKernelsStatic::OptHandUnroll,  WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{  WilsonKernelsStatic::OptHandUnroll,  WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{  WilsonKernelsStatic::OptGeneric   ,  WilsonKernelsStatic::CommsThenCompute ,CartesianCommunicator::CommunicatorPolicySequential  },
+	{  WilsonKernelsStatic::OptGeneric   ,  WilsonKernelsStatic::CommsAndCompute  ,CartesianCommunicator::CommunicatorPolicySequential  }
       }; 
 
       for(int c=0;c<num_cases;c++) {
-
-	QCD::WilsonKernelsStatic::Comms = Cases[c].CommsOverlap;
-	QCD::WilsonKernelsStatic::Opt   = Cases[c].Opt;
+	
+	WilsonKernelsStatic::Comms = Cases[c].CommsOverlap;
+	WilsonKernelsStatic::Opt   = Cases[c].Opt;
 	CartesianCommunicator::SetCommunicatorPolicy(Cases[c].CommsAsynch);
 
 	std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
@@ -663,8 +656,10 @@ public:
 	setCheckerboard(r_eo,r_o);
 	setCheckerboard(r_eo,r_e);
 	err = r_eo-ref; 
-	std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
-	assert((norm2(err)<1.0e-4));
+	RealD absref = norm2(ref);
+	RealD abserr = norm2(err);
+	std::cout<<GridLogMessage << "norm diff   "<< abserr << " / " << absref<<std::endl;
+	assert(abserr<1.0e-4);
 
       }
       robust = mflops_worst/mflops_best;
@@ -710,10 +705,10 @@ int main (int argc, char ** argv)
   }
 #if 1
   int sel=2;
-  std::vector<int> L_list({8,12,16,24});
+  Coordinate L_list({8,12,16,24});
 #else
   int sel=1;
-  std::vector<int> L_list({8,12});
+  Coordinate L_list({8,12});
 #endif
   int selm1=sel-1;
   std::vector<double> robust_list;
@@ -729,7 +724,7 @@ int main (int argc, char ** argv)
     std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
     for(int l=0;l<L_list.size();l++){
       double robust;
-      wilson.push_back(Benchmark::DWF(1,L_list[l],robust));
+      wilson.push_back(Benchmark::DWF(Ls,L_list[l],robust));
     }
   }
 
@@ -746,24 +741,14 @@ int main (int argc, char ** argv)
     }
   }
 
-  if ( do_dwf ) {
-    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
-    std::cout<<GridLogMessage << " Domain wall dslash 4D vectorised" <<std::endl;
-    std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
-    for(int l=0;l<L_list.size();l++){
-      dwf5.push_back(Benchmark::DWF5(Ls,L_list[l]));
-    }
-
-  }
-
   if ( do_dwf ) {
 
   std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
   std::cout<<GridLogMessage << " Summary table Ls="<<Ls <<std::endl;
   std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
-  std::cout<<GridLogMessage << "L \t\t Wilson \t DWF4 \t DWF5 " <<std::endl;
+  std::cout<<GridLogMessage << "L \t\t Wilson \t DWF4 " <<std::endl;
   for(int l=0;l<L_list.size();l++){
-    std::cout<<GridLogMessage << L_list[l] <<" \t\t "<< wilson[l]<<" \t "<<dwf4[l]<<" \t "<<dwf5[l] <<std::endl;
+    std::cout<<GridLogMessage << L_list[l] <<" \t\t "<< wilson[l]<<" \t "<<dwf4[l] <<std::endl;
   }
   std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
   }
@@ -787,9 +772,9 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
   std::cout<<GridLogMessage << " Per Node Summary table Ls="<<Ls <<std::endl;
   std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
-  std::cout<<GridLogMessage << " L \t\t Wilson\t\t DWF4  \t\t DWF5 " <<std::endl;
+  std::cout<<GridLogMessage << " L \t\t Wilson\t\t DWF4  " <<std::endl;
   for(int l=0;l<L_list.size();l++){
-    std::cout<<GridLogMessage << L_list[l] <<" \t\t "<< wilson[l]/NN<<" \t "<<dwf4[l]/NN<<" \t "<<dwf5[l] /NN<<std::endl;
+    std::cout<<GridLogMessage << L_list[l] <<" \t\t "<< wilson[l]/NN<<" \t "<<dwf4[l]/NN<<std::endl;
   }
   std::cout<<GridLogMessage << "=================================================================================="<<std::endl;
 
@@ -802,6 +787,5 @@ int main (int argc, char ** argv)
 
   }
 
-
   Grid_finalize();
 }
diff --git a/benchmarks/Benchmark_comms.cc b/benchmarks/Benchmark_comms.cc
index f87caef0..44ccbd19 100644
--- a/benchmarks/Benchmark_comms.cc
+++ b/benchmarks/Benchmark_comms.cc
@@ -29,7 +29,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 struct time_statistics{
   double mean;
@@ -61,8 +60,8 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
@@ -79,13 +78,13 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "= Benchmarking concurrent halo exchange in "<<nmu<<" dimensions"<<std::endl;
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   header();
-  for(int lat=4;lat<=maxlat;lat+=4){
+  for(int lat=8;lat<=maxlat;lat+=4){
     for(int Ls=8;Ls<=8;Ls*=2){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],
-      				    lat*mpi_layout[1],
-      				    lat*mpi_layout[2],
-      				    lat*mpi_layout[3]});
+      Coordinate latt_size  ({lat*mpi_layout[0],
+	                      lat*mpi_layout[1],
+      			      lat*mpi_layout[2],
+      			      lat*mpi_layout[3]});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
       RealD Nrank = Grid._Nprocessors;
@@ -166,14 +165,13 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   header();
 
-  for(int lat=4;lat<=maxlat;lat+=4){
+  for(int lat=8;lat<=maxlat;lat+=4){
     for(int Ls=8;Ls<=8;Ls*=2){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],
-                                    lat*mpi_layout[1],
-                                    lat*mpi_layout[2],
-                                    lat*mpi_layout[3]});
-
+      Coordinate latt_size  ({lat*mpi_layout[0],
+	                      lat*mpi_layout[1],
+      			      lat*mpi_layout[2],
+      			      lat*mpi_layout[3]});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
       RealD Nrank = Grid._Nprocessors;
@@ -261,13 +259,13 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   header();
 
-  for(int lat=4;lat<=maxlat;lat+=4){
+  for(int lat=8;lat<=maxlat;lat+=4){
     for(int Ls=8;Ls<=8;Ls*=2){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],
-      				    lat*mpi_layout[1],
-      				    lat*mpi_layout[2],
-      				    lat*mpi_layout[3]});
+      Coordinate latt_size  ({lat*mpi_layout[0],
+	                      lat*mpi_layout[1],
+      			      lat*mpi_layout[2],
+      			      lat*mpi_layout[3]});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
       RealD Nrank = Grid._Nprocessors;
@@ -281,8 +279,6 @@ int main (int argc, char ** argv)
       for(int d=0;d<8;d++){
 	xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(bytes);
 	rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(bytes);
-	bzero((void *)xbuf[d],bytes);
-	bzero((void *)rbuf[d],bytes);
       }
 
       int ncomm;
@@ -338,7 +334,7 @@ int main (int argc, char ** argv)
 
       dbytes=dbytes*ppn;
       double xbytes    = dbytes*0.5;
-      double rbytes    = dbytes*0.5;
+      //      double rbytes    = dbytes*0.5;
       double bidibytes = dbytes;
 
       std::cout<<GridLogMessage << std::setw(4) << lat<<"\t"<<Ls<<"\t"
@@ -358,13 +354,13 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   header();
 
-  for(int lat=4;lat<=maxlat;lat+=4){
+  for(int lat=8;lat<=maxlat;lat+=4){
     for(int Ls=8;Ls<=8;Ls*=2){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],
-      				    lat*mpi_layout[1],
-      				    lat*mpi_layout[2],
-      				    lat*mpi_layout[3]});
+      Coordinate latt_size  ({lat*mpi_layout[0],
+      			      lat*mpi_layout[1],
+      			      lat*mpi_layout[2],
+      			      lat*mpi_layout[3]});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
       RealD Nrank = Grid._Nprocessors;
@@ -378,8 +374,6 @@ int main (int argc, char ** argv)
       for(int d=0;d<8;d++){
 	xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(bytes);
 	rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(bytes);
-	bzero((void *)xbuf[d],bytes);
-	bzero((void *)rbuf[d],bytes);
       }
 
       int ncomm;
@@ -435,7 +429,7 @@ int main (int argc, char ** argv)
 
       dbytes=dbytes*ppn;
       double xbytes    = dbytes*0.5;
-      double rbytes    = dbytes*0.5;
+      //      double rbytes    = dbytes*0.5;
       double bidibytes = dbytes;
 
 
@@ -456,13 +450,13 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   header();
 
-  for(int lat=4;lat<=maxlat;lat+=4){
+  for(int lat=8;lat<=maxlat;lat+=4){
     for(int Ls=8;Ls<=8;Ls*=2){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],
-      				    lat*mpi_layout[1],
-      				    lat*mpi_layout[2],
-      				    lat*mpi_layout[3]});
+      Coordinate latt_size  ({lat*mpi_layout[0],
+    	                      lat*mpi_layout[1],
+                              lat*mpi_layout[2],
+	                      lat*mpi_layout[3]});
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
       RealD Nrank = Grid._Nprocessors;
@@ -472,13 +466,10 @@ int main (int argc, char ** argv)
       std::vector<HalfSpinColourVectorD *> xbuf(8);
       std::vector<HalfSpinColourVectorD *> rbuf(8);
       Grid.ShmBufferFreeAll();
-      //      uint64_t bytes = lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
-      uint64_t bytes = 2*1024*1024;
+      uint64_t bytes = lat*lat*lat*Ls*sizeof(HalfSpinColourVectorD);
       for(int d=0;d<8;d++){
 	xbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(bytes);
 	rbuf[d] = (HalfSpinColourVectorD *)Grid.ShmBufferMalloc(bytes);
-	bzero((void *)xbuf[d],bytes);
-	bzero((void *)rbuf[d],bytes);
       }
 
       int ncomm;
@@ -490,8 +481,7 @@ int main (int argc, char ** argv)
 	dbytes=0;
 	ncomm=0;
 
-#pragma omp parallel for num_threads(Grid::CartesianCommunicator::nCommThreads)
-	for(int dir=0;dir<8;dir++){
+	thread_for(dir,8,{
 
 	  double tbytes;
 	  int mu =dir % 4;
@@ -512,10 +502,9 @@ int main (int argc, char ** argv)
 	    tbytes= Grid.StencilSendToRecvFrom((void *)&xbuf[dir][0], xmit_to_rank,
 					       (void *)&rbuf[dir][0], recv_from_rank, bytes,tid);
 
-#pragma omp atomic
-	    dbytes+=tbytes;
+	    thread_critical { dbytes+=tbytes; }
 	  }
-	}
+        });
 	Grid.Barrier();
 	double stop=usecond();
 	t_time[i] = stop-start; // microseconds
@@ -525,7 +514,7 @@ int main (int argc, char ** argv)
 
       dbytes=dbytes*ppn;
       double xbytes    = dbytes*0.5;
-      double rbytes    = dbytes*0.5;
+      //      double rbytes    = dbytes*0.5;
       double bidibytes = dbytes;
 
 
diff --git a/benchmarks/Benchmark_dwf.cc b/benchmarks/Benchmark_dwf.cc
index a7362e35..4d6b026f 100644
--- a/benchmarks/Benchmark_dwf.cc
+++ b/benchmarks/Benchmark_dwf.cc
@@ -21,10 +21,16 @@
     *************************************************************************************/
     /*  END LEGAL */
 #include <Grid/Grid.h>
+#ifdef GRID_NVCC
+#define CUDA_PROFILE
+#endif
+
+#ifdef CUDA_PROFILE
+#include <cuda_profiler_api.h>
+#endif
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 template<class d>
 struct scal {
@@ -38,9 +44,6 @@ struct scal {
     Gamma::Algebra::GammaT
   };
 
-typedef WilsonFermion5D<DomainWallVec5dImplR> WilsonFermion5DR;
-typedef WilsonFermion5D<DomainWallVec5dImplF> WilsonFermion5DF;
-typedef WilsonFermion5D<DomainWallVec5dImplD> WilsonFermion5DD;
 
 int main (int argc, char ** argv)
 {
@@ -49,8 +52,8 @@ int main (int argc, char ** argv)
 
   int threads = GridThread::GetThreads();
 
-  std::vector<int> latt4 = GridDefaultLatt();
-  int Ls=16;
+  Coordinate latt4 = GridDefaultLatt();
+  int Ls=8;
   for(int i=0;i<argc;i++)
     if(std::string(argv[i]) == "-Ls"){
       std::stringstream ss(argv[i+1]); ss >> Ls;
@@ -58,7 +61,7 @@ int main (int argc, char ** argv)
 
   GridLogLayout();
 
-  long unsigned int single_site_flops = 8*QCD::Nc*(7+16*QCD::Nc);
+  long unsigned int single_site_flops = 8*Nc*(7+16*Nc);
 
 
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
@@ -83,11 +86,11 @@ int main (int argc, char ** argv)
 
   LatticeFermion src   (FGrid); random(RNG5,src);
 #if 0
-  src = zero;
+  src = Zero();
   {
-    std::vector<int> origin({0,0,0,latt4[2]-1,0});
+    Coordinate origin({0,0,0,latt4[2]-1,0});
     SpinColourVectorF tmp;
-    tmp=zero;
+    tmp=Zero();
     tmp()(0)(0)=Complex(-2.0,0.0);
     std::cout << " source site 0 " << tmp<<std::endl;
     pokeSite(tmp,src,origin);
@@ -98,8 +101,8 @@ int main (int argc, char ** argv)
 #endif
 
 
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid);    ref=zero;
+  LatticeFermion result(FGrid); result=Zero();
+  LatticeFermion    ref(FGrid);    ref=Zero();
   LatticeFermion    tmp(FGrid);
   LatticeFermion    err(FGrid);
 
@@ -125,9 +128,13 @@ int main (int argc, char ** argv)
   // replicate across fifth dimension
   LatticeGaugeField Umu5d(FGrid); 
   std::vector<LatticeColourMatrix> U(4,FGrid);
-  for(int ss=0;ss<Umu._grid->oSites();ss++){
-    for(int s=0;s<Ls;s++){
-      Umu5d._odata[Ls*ss+s] = Umu._odata[ss];
+  {
+    auto Umu5d_v = Umu5d.View();
+    auto Umu_v = Umu.View();
+    for(int ss=0;ss<Umu.Grid()->oSites();ss++){
+      for(int s=0;s<Ls;s++){
+	Umu5d_v[Ls*ss+s] = Umu_v[ss];
+      }
     }
   }
   for(int mu=0;mu<Nd;mu++){
@@ -137,7 +144,7 @@ int main (int argc, char ** argv)
 
   if (1)
   {
-    ref = zero;
+    ref = Zero();
     for(int mu=0;mu<Nd;mu++){
 
       tmp = U[mu]*Cshift(src,mu+1,1);
@@ -162,6 +169,7 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
   std::cout << GridLogMessage<< "* Benchmarking DomainWallFermionR::Dhop                  "<<std::endl;
   std::cout << GridLogMessage<< "* Vectorising space-time by "<<vComplex::Nsimd()<<std::endl;
+  std::cout << GridLogMessage<< "* VComplex size is "<<sizeof(vComplex)<< " B"<<std::endl;
   if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
   if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
 #ifdef GRID_OMP
@@ -174,7 +182,8 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
 
   DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-  int ncall =500;
+  int ncall =1000;
+
   if (1) {
     FGrid->Barrier();
     Dw.ZeroCounters();
@@ -200,8 +209,8 @@ int main (int argc, char ** argv)
     std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NN<<std::endl;
     err = ref-result; 
     std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
+    //exit(0);
 
-    /*
     if(( norm2(err)>1.0e-4) ) { 
       std::cout << "RESULT\n " << result<<std::endl;
       std::cout << "REF   \n " << ref   <<std::endl;
@@ -209,13 +218,12 @@ int main (int argc, char ** argv)
       FGrid->Barrier();
       exit(-1);
     }
-    */
     assert (norm2(err)< 1.0e-4 );
     Dw.Report();
   }
 
   DomainWallFermionRL DwH(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-  if (1) {
+  if (0) {
     FGrid->Barrier();
     DwH.ZeroCounters();
     DwH.Dhop(src,result,0);
@@ -242,187 +250,29 @@ int main (int argc, char ** argv)
     DwH.Report();
   }
 
-  if (1)
-  {
-
-    std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
-    std::cout << GridLogMessage<< "* Benchmarking WilsonFermion5D<DomainWallVec5dImplR>::Dhop "<<std::endl;
-    std::cout << GridLogMessage<< "* Vectorising fifth dimension by "<<vComplex::Nsimd()<<std::endl;
-    if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
-    if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
-#ifdef GRID_OMP
-  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
-  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
-#endif
-    if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
-    if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
-    if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
-    std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
-
-    typedef WilsonFermion5D<DomainWallVec5dImplR> WilsonFermion5DR;
-    LatticeFermion ssrc(sFGrid);
-    LatticeFermion sref(sFGrid);
-    LatticeFermion sresult(sFGrid);
-
-    WilsonFermion5DR sDw(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,M5);
-
-    localConvert(src,ssrc);
-    std::cout<<GridLogMessage<< "src norms "<< norm2(src)<<" " <<norm2(ssrc)<<std::endl;
-    FGrid->Barrier();
-    sDw.Dhop(ssrc,sresult,0);
-    sDw.ZeroCounters();
-    double t0=usecond();
-    for(int i=0;i<ncall;i++){
-      __SSC_START;
-      sDw.Dhop(ssrc,sresult,0);
-      __SSC_STOP;
-    }
-    double t1=usecond();
-    FGrid->Barrier();
-    double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-    double flops=single_site_flops*volume*ncall;
-
-    std::cout<<GridLogMessage << "Called Dw s_inner "<<ncall<<" times in "<<t1-t0<<" us"<<std::endl;
-    std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t1-t0)<<std::endl;
-    std::cout<<GridLogMessage << "mflop/s per rank =  "<< flops/(t1-t0)/NP<<std::endl;
-    std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NN<<std::endl;
-    //    std::cout<<GridLogMessage<< "res norms "<< norm2(result)<<" " <<norm2(sresult)<<std::endl;
-    sDw.Report();
-    RealD sum=0;
-
-    err=zero;
-    localConvert(sresult,err);
-    err = err - ref;
-    sum = norm2(err);
-    std::cout<<GridLogMessage<<" difference between normal ref and simd is "<<sum<<std::endl;
-    if(sum > 1.0e-4 ){
-      std::cout<< "sD REF\n " <<ref << std::endl;
-      std::cout<< "sD ERR   \n " <<err  <<std::endl;
-    }
-    //    assert(sum < 1.0e-4);
-
-    err=zero;
-    localConvert(sresult,err);
-    err = err - result;
-    sum = norm2(err);
-    std::cout<<GridLogMessage<<" difference between normal result and simd is "<<sum<<std::endl;
-    if(sum > 1.0e-4 ){
-      std::cout<< "sD REF\n " <<result << std::endl;
-      std::cout<< "sD ERR   \n " << err  <<std::endl;
-    }
-    assert(sum < 1.0e-4);
-
-    
-    if(1){
-      std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
-      std::cout << GridLogMessage<< "* Benchmarking WilsonFermion5D<DomainWallVec5dImplR>::DhopEO "<<std::endl;
-      std::cout << GridLogMessage<< "* Vectorising fifth dimension by "<<vComplex::Nsimd()<<std::endl;
-      if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
-      if ( sizeof(Real)==8 )   std::cout << GridLogMessage<< "* DOUBLE precision "<<std::endl;
-#ifdef GRID_OMP
-  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsAndCompute ) std::cout << GridLogMessage<< "* Using Overlapped Comms/Compute" <<std::endl;
-  if ( WilsonKernelsStatic::Comms == WilsonKernelsStatic::CommsThenCompute) std::cout << GridLogMessage<< "* Using sequential comms compute" <<std::endl;
-#endif
-      if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptGeneric   ) 
-	std::cout << GridLogMessage<< "* Using GENERIC Nc WilsonKernels" <<std::endl;
-      if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptHandUnroll) 
-	std::cout << GridLogMessage<< "* Using Nc=3       WilsonKernels" <<std::endl;
-      if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) 
-	std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
-      std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
-
-      LatticeFermion sr_eo(sFGrid);
-      LatticeFermion ssrc_e (sFrbGrid);
-      LatticeFermion ssrc_o (sFrbGrid);
-      LatticeFermion sr_e   (sFrbGrid);
-      LatticeFermion sr_o   (sFrbGrid);
-
-      pickCheckerboard(Even,ssrc_e,ssrc);
-      pickCheckerboard(Odd,ssrc_o,ssrc);
-      //      setCheckerboard(sr_eo,ssrc_o);
-      //      setCheckerboard(sr_eo,ssrc_e);
-
-      sr_e = zero;
-      sr_o = zero;
-
-      FGrid->Barrier();
-      sDw.DhopEO(ssrc_o, sr_e, DaggerNo);
-      sDw.ZeroCounters();
-      //      sDw.stat.init("DhopEO");
-      double t0=usecond();
-      for (int i = 0; i < ncall; i++) {
-        sDw.DhopEO(ssrc_o, sr_e, DaggerNo);
-      }
-      double t1=usecond();
-      FGrid->Barrier();
-      //      sDw.stat.print();
-
-      double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-      double flops=(single_site_flops*volume*ncall)/2.0;
-
-      std::cout<<GridLogMessage << "sDeo mflop/s =   "<< flops/(t1-t0)<<std::endl;
-      std::cout<<GridLogMessage << "sDeo mflop/s per rank   "<< flops/(t1-t0)/NP<<std::endl;
-      std::cout<<GridLogMessage << "sDeo mflop/s per node   "<< flops/(t1-t0)/NN<<std::endl;
-      sDw.Report();
-
-      sDw.DhopEO(ssrc_o,sr_e,DaggerNo);
-      sDw.DhopOE(ssrc_e,sr_o,DaggerNo);
-      sDw.Dhop  (ssrc  ,sresult,DaggerNo);
-
-      pickCheckerboard(Even,ssrc_e,sresult);
-      pickCheckerboard(Odd ,ssrc_o,sresult);
-
-      ssrc_e = ssrc_e - sr_e;
-      RealD error = norm2(ssrc_e);
-      std::cout<<GridLogMessage << "sE norm diff   "<< norm2(ssrc_e)<< "  vec nrm"<<norm2(sr_e) <<std::endl;
-
-      ssrc_o = ssrc_o - sr_o;
-      error+= norm2(ssrc_o);
-      std::cout<<GridLogMessage << "sO norm diff   "<< norm2(ssrc_o)<< "  vec nrm"<<norm2(sr_o) <<std::endl;
-
-      if(( error>1.0e-4) ) { 
-	setCheckerboard(ssrc,ssrc_o);
-	setCheckerboard(ssrc,ssrc_e);
-	std::cout<< "DIFF\n " <<ssrc << std::endl;
-	setCheckerboard(ssrc,sr_o);
-	setCheckerboard(ssrc,sr_e);
-	std::cout<< "CBRESULT\n " <<ssrc << std::endl;
-	std::cout<< "RESULT\n " <<sresult<< std::endl;
-      }
-      assert(error<1.0e-4);
-    }
-
-  if(0){
-    std::cout << "Single cache warm call to sDw.Dhop " <<std::endl;
-    for(int i=0;i< PerformanceCounter::NumTypes(); i++ ){
-      sDw.Dhop(ssrc,sresult,0);
-      PerformanceCounter Counter(i);
-      Counter.Start();
-      sDw.Dhop(ssrc,sresult,0);
-      Counter.Stop();
-      Counter.Report();
-    }
-  }
-
-  }
-
-
-
   if (1)
   { // Naive wilson dag implementation
-    ref = zero;
+    ref = Zero();
     for(int mu=0;mu<Nd;mu++){
 
       //    ref =  src - Gamma(Gamma::Algebra::GammaX)* src ; // 1+gamma_x
       tmp = U[mu]*Cshift(src,mu+1,1);
-      for(int i=0;i<ref._odata.size();i++){
-	ref._odata[i]+= tmp._odata[i] + Gamma(Gmu[mu])*tmp._odata[i]; ;
+      {
+	auto ref_v = ref.View();
+	auto tmp_v = tmp.View();
+	for(int i=0;i<ref_v.size();i++){
+	  ref_v[i]+= tmp_v[i] + Gamma(Gmu[mu])*tmp_v[i]; ;
+	}
       }
 
       tmp =adj(U[mu])*src;
       tmp =Cshift(tmp,mu+1,-1);
-      for(int i=0;i<ref._odata.size();i++){
-	ref._odata[i]+= tmp._odata[i] - Gamma(Gmu[mu])*tmp._odata[i]; ;
+      {
+	auto ref_v = ref.View();
+	auto tmp_v = tmp.View();
+	for(int i=0;i<ref_v.size();i++){
+	  ref_v[i]+= tmp_v[i] - Gamma(Gmu[mu])*tmp_v[i]; ;
+	}
       }
     }
     ref = -0.5*ref;
@@ -475,7 +325,13 @@ int main (int argc, char ** argv)
     Dw.DhopEO(src_o,r_e,DaggerNo);
     double t0=usecond();
     for(int i=0;i<ncall;i++){
+#ifdef CUDA_PROFILE
+      if(i==10) cudaProfilerStart();
+#endif
       Dw.DhopEO(src_o,r_e,DaggerNo);
+#ifdef CUDA_PROFILE
+      if(i==20) cudaProfilerStop();
+#endif
     }
     double t1=usecond();
     FGrid->Barrier();
diff --git a/benchmarks/Benchmark_dwf_sweep.cc b/benchmarks/Benchmark_dwf_sweep.cc
index da8eb044..3f108b16 100644
--- a/benchmarks/Benchmark_dwf_sweep.cc
+++ b/benchmarks/Benchmark_dwf_sweep.cc
@@ -30,7 +30,6 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 template<class d>
 struct scal {
@@ -45,7 +44,6 @@ struct scal {
   };
 
 void benchDw(std::vector<int> & L, int Ls, int threads, int report =0 );
-void benchsDw(std::vector<int> & L, int Ls, int threads, int report=0 );
 
 int main (int argc, char ** argv)
 {
@@ -61,13 +59,13 @@ int main (int argc, char ** argv)
   if ( WilsonKernelsStatic::Opt == WilsonKernelsStatic::OptInlineAsm ) std::cout << GridLogMessage<< "* Using Asm Nc=3   WilsonKernels" <<std::endl;
   std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
 
-  const int Ls=8;
+  const int Ls=12;
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
   std::cout<<GridLogMessage << "=========================================================================="<<std::endl;
   std::cout<<GridLogMessage << "= Benchmarking DWF"<<std::endl;
   std::cout<<GridLogMessage << "=========================================================================="<<std::endl;
-  std::cout<<GridLogMessage << "Volume \t\t\tProcs \t Dw \t eoDw \t sDw \t eosDw (Mflop/s)  "<<std::endl;
+  std::cout<<GridLogMessage << "Volume \t\t\tProcs \t Dw \t eoDw   "<<std::endl;
   std::cout<<GridLogMessage << "=========================================================================="<<std::endl;
 
   int Lmax=16;
@@ -84,19 +82,17 @@ int main (int argc, char ** argv)
       }
       std::cout <<Ls<<"\t" ;
       benchDw (latt4,Ls,threads,0);
-      benchsDw(latt4,Ls,threads,0);
       std::cout<<std::endl;
     }
   }
   std::cout<<GridLogMessage << "=========================================================================="<<std::endl;
+#ifdef PERFCOUNT
   {
     std::vector<int> latt4(4,16);
     std::cout<<GridLogMessage << "16^4 Dw miss rate"<<std::endl;
     benchDw (latt4,Ls,threads,1);
-    std::cout<<GridLogMessage << "16^4 sDw miss rate"<<std::endl;
-    benchsDw(latt4,Ls,threads,1);
   }
-
+#endif
   Grid_finalize();
 }
 
@@ -108,7 +104,7 @@ void benchDw(std::vector<int> & latt4, int Ls, int threads,int report )
   GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
   GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
   GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
-  long unsigned int single_site_flops = 8*QCD::Nc*(7+16*QCD::Nc);
+  long unsigned int single_site_flops = 8*Nc*(7+16*Nc);
 
   std::vector<int> seeds4({1,2,3,4});
   std::vector<int> seeds5({5,6,7,8});
@@ -120,12 +116,12 @@ void benchDw(std::vector<int> & latt4, int Ls, int threads,int report )
   LatticeGaugeField Umu(UGrid); 
   random(RNG4,Umu);
 #else 
-  LatticeFermion src   (FGrid); src=zero;
-  LatticeGaugeField Umu(UGrid); Umu=zero;
+  LatticeFermion src   (FGrid); src=Zero();
+  LatticeGaugeField Umu(UGrid); Umu=Zero();
 #endif
 
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid);    ref=zero;
+  LatticeFermion result(FGrid); result=Zero();
+  LatticeFermion    ref(FGrid);    ref=Zero();
   LatticeFermion    tmp(FGrid);
   LatticeFermion    err(FGrid);
 
@@ -134,9 +130,11 @@ void benchDw(std::vector<int> & latt4, int Ls, int threads,int report )
   LatticeGaugeField Umu5d(FGrid); 
 
   // replicate across fifth dimension
-  for(int ss=0;ss<Umu._grid->oSites();ss++){
+  auto Umu5d_v = Umu5d.View();
+  auto Umu_v   = Umu.View();
+  for(int ss=0;ss<Umu.Grid()->oSites();ss++){
     for(int s=0;s<Ls;s++){
-      Umu5d._odata[Ls*ss+s] = Umu._odata[ss];
+      Umu5d_v[Ls*ss+s] = Umu_v[ss];
     }
   }
 
@@ -151,7 +149,7 @@ void benchDw(std::vector<int> & latt4, int Ls, int threads,int report )
 #ifdef CHECK
   if (1) {
 
-    ref = zero;
+    ref = Zero();
     for(int mu=0;mu<Nd;mu++){
       tmp = U[mu]*Cshift(src,mu+1,1);
       ref=ref + tmp - Gamma(Gmu[mu])*tmp;
@@ -183,19 +181,21 @@ void benchDw(std::vector<int> & latt4, int Ls, int threads,int report )
   if (ncall < 5 ) exit(0);
 
   Dw.Dhop(src,result,0);
-
+#ifdef PERFCOUNT
   PerformanceCounter Counter(8);
   Counter.Start();
+#endif
   t0=usecond();
   for(int i=0;i<ncall;i++){
     Dw.Dhop(src,result,0);
   }
   t1=usecond();
+#ifdef PERFCOUNT
   Counter.Stop();
   if ( report ) {
     Counter.Report();
   }
-  
+#endif  
   if ( ! report ) {
     double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
     double flops=single_site_flops*volume*ncall;
@@ -236,134 +236,5 @@ void benchDw(std::vector<int> & latt4, int Ls, int threads,int report )
   }
 }
 
-#define CHECK_SDW
-void benchsDw(std::vector<int> & latt4, int Ls, int threads, int report )
-{
-  long unsigned int single_site_flops = 8*QCD::Nc*(7+16*QCD::Nc);
-
-  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(latt4, GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
-  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
-  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
-  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
-
-  GridCartesian         * sUGrid   = SpaceTimeGrid::makeFourDimDWFGrid(latt4,GridDefaultMpi());
-  GridRedBlackCartesian * sUrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(sUGrid);
-  GridCartesian         * sFGrid   = SpaceTimeGrid::makeFiveDimDWFGrid(Ls,UGrid);
-  GridRedBlackCartesian * sFrbGrid = SpaceTimeGrid::makeFiveDimDWFRedBlackGrid(Ls,UGrid);
-
-  std::vector<int> seeds4({1,2,3,4});
-  std::vector<int> seeds5({5,6,7,8});
-
-#ifdef CHECK_SDW
-  GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
-  GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
-  LatticeFermion src   (FGrid); random(RNG5,src);
-  LatticeGaugeField Umu(UGrid); 
-  random(RNG4,Umu);
-#else 
-  LatticeFermion src   (FGrid); src=zero;
-  LatticeGaugeField Umu(UGrid); Umu=zero;
-#endif
-
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid);    ref=zero;
-  LatticeFermion    tmp(FGrid);
-  LatticeFermion    err(FGrid);
-
-  ColourMatrix cm = Complex(1.0,0.0);
-
-  LatticeGaugeField Umu5d(FGrid); 
-
-  // replicate across fifth dimension
-  for(int ss=0;ss<Umu._grid->oSites();ss++){
-    for(int s=0;s<Ls;s++){
-      Umu5d._odata[Ls*ss+s] = Umu._odata[ss];
-    }
-  }
-
-  RealD mass=0.1;
-  RealD M5  =1.8;
-
-  typedef WilsonFermion5D<DomainWallVec5dImplR> WilsonFermion5DR;
-  LatticeFermion ssrc(sFGrid);
-  LatticeFermion sref(sFGrid);
-  LatticeFermion sresult(sFGrid);
-  WilsonFermion5DR sDw(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,M5);
-  
-  for(int x=0;x<latt4[0];x++){
-  for(int y=0;y<latt4[1];y++){
-  for(int z=0;z<latt4[2];z++){
-  for(int t=0;t<latt4[3];t++){
-  for(int s=0;s<Ls;s++){
-    std::vector<int> site({s,x,y,z,t});
-    SpinColourVector tmp;
-    peekSite(tmp,src,site);
-    pokeSite(tmp,ssrc,site);
-  }}}}}
-
-  double t0=usecond();
-  sDw.Dhop(ssrc,sresult,0);
-  double t1=usecond();
-
-#ifdef TIMERS_OFF
-  int ncall =10;
-#else 
-  int ncall =1+(int) ((5.0*1000*1000)/(t1-t0));
-#endif
-
-  PerformanceCounter Counter(8);
-  Counter.Start();
-  t0=usecond();
-  for(int i=0;i<ncall;i++){
-    sDw.Dhop(ssrc,sresult,0);
-  }
-  t1=usecond();
-  Counter.Stop();
-  
-  if ( report ) {
-    Counter.Report();
-  } else { 
-    double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-    double flops=single_site_flops*volume*ncall;
-    std::cout<<"\t"<< flops/(t1-t0);
-  }
-
-  LatticeFermion sr_eo(sFGrid);
-  LatticeFermion serr(sFGrid);
-  
-  LatticeFermion ssrc_e (sFrbGrid);
-  LatticeFermion ssrc_o (sFrbGrid);
-  LatticeFermion sr_e   (sFrbGrid);
-  LatticeFermion sr_o   (sFrbGrid);
-      
-  pickCheckerboard(Even,ssrc_e,ssrc);
-  pickCheckerboard(Odd,ssrc_o,ssrc);
-  
-  setCheckerboard(sr_eo,ssrc_o);
-  setCheckerboard(sr_eo,ssrc_e);
-    
-  sr_e = zero;
-  sr_o = zero;
-  
-  sDw.DhopEO(ssrc_o,sr_e,DaggerNo);
-  PerformanceCounter CounterSdw(8);
-  CounterSdw.Start();
-  t0=usecond();
-  for(int i=0;i<ncall;i++){
-    __SSC_START;
-    sDw.DhopEO(ssrc_o,sr_e,DaggerNo);
-    __SSC_STOP;
-  }
-  t1=usecond();
-  CounterSdw.Stop();
-
-  if ( report ) { 
-    CounterSdw.Report();
-  } else {
-    double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-    double flops=(single_site_flops*volume*ncall)/2.0;
-    std::cout<<"\t"<< flops/(t1-t0);
-  }
-}
 
 
diff --git a/benchmarks/Benchmark_gparity.cc b/benchmarks/Benchmark_gparity.cc
index 643d241c..b03e1b63 100644
--- a/benchmarks/Benchmark_gparity.cc
+++ b/benchmarks/Benchmark_gparity.cc
@@ -2,7 +2,7 @@
 #include <sstream>
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -36,7 +36,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
   std::cout<<GridLogMessage << "Ls = " << Ls << std::endl;
 
-  std::vector<int> latt4 = GridDefaultLatt();
+  Coordinate latt4 = GridDefaultLatt();
 
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
   GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
@@ -56,8 +56,8 @@ int main (int argc, char ** argv)
   RealD N2 = 1.0/::sqrt(norm2(src));
   src = src*N2;
 
-  GparityLatticeFermionF result(FGrid); result=zero;
-  GparityLatticeFermionF    ref(FGrid);    ref=zero;
+  GparityLatticeFermionF result(FGrid); result=Zero();
+  GparityLatticeFermionF    ref(FGrid);    ref=Zero();
   GparityLatticeFermionF    tmp(FGrid);
   GparityLatticeFermionF    err(FGrid);
 
diff --git a/benchmarks/Benchmark_memory_asynch.cc b/benchmarks/Benchmark_memory_asynch.cc
index 2124b137..a29d281b 100644
--- a/benchmarks/Benchmark_memory_asynch.cc
+++ b/benchmarks/Benchmark_memory_asynch.cc
@@ -30,7 +30,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
@@ -41,8 +41,8 @@ int main (int argc, char ** argv)
   typedef iVector<vReal,Nvec> Vec;
 
 
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vReal::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vReal::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
@@ -58,13 +58,13 @@ int main (int argc, char ** argv)
 
     int Nloop=lmax*4/lat;
 
-    std::vector<int> latt_size  ({2*mpi_layout[0],2*mpi_layout[1],4*mpi_layout[2],lat*mpi_layout[3]});
+    Coordinate latt_size  ({2*mpi_layout[0],2*mpi_layout[1],4*mpi_layout[2],lat*mpi_layout[3]});
 
     GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
     int vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3]*threads;
 
-    Vec tsum; tsum = zero;
+    Vec tsum; tsum = Zero();
 
     GridParallelRNG          pRNG(&Grid);      
     pRNG.SeedFixedIntegers(std::vector<int>({56,17,89,101}));
@@ -78,16 +78,16 @@ int main (int argc, char ** argv)
     }
 
     double start=usecond();
-    parallel_for(int t=0;t<threads;t++){
-
-      sum[t] = x[t]._odata[0];
+    thread_for(t,threads,{
+      auto x_t = x[t].View();
+      sum[t] = x_t[0];
       for(int i=0;i<Nloop;i++){
-	for(auto ss=x[t].begin();ss<x[t].end();ss++){
-	  sum[t] = sum[t]+x[t]._odata[ss];
+	for(auto ss=x_t.begin();ss<x_t.end();ss++){
+	  sum[t] = sum[t]+x_t[ss];
 	}
       }
       stop[t]=usecond();
-    }
+    });
 
     double max_stop=stop[0];
     double min_stop=stop[0];
diff --git a/benchmarks/Benchmark_memory_bandwidth.cc b/benchmarks/Benchmark_memory_bandwidth.cc
index cc965050..9b921a77 100644
--- a/benchmarks/Benchmark_memory_bandwidth.cc
+++ b/benchmarks/Benchmark_memory_bandwidth.cc
@@ -30,7 +30,6 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 int main (int argc, char ** argv)
 {
@@ -41,10 +40,10 @@ int main (int argc, char ** argv)
   typedef iVector<vReal,Nvec> Vec;
 
 
-  Vec rn = zero;
+  Vec rn = Zero();
 
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vReal::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vReal::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
@@ -55,11 +54,11 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s"<<"\t\t"<<"Gflop/s"<<"\t\t seconds"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
-  uint64_t lmax=64;
-#define NLOOP (10*lmax*lmax*lmax*lmax/vol)
+  uint64_t lmax=48;
+#define NLOOP (100*lmax*lmax*lmax*lmax/vol)
   for(int lat=8;lat<=lmax;lat+=8){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
       int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
@@ -73,11 +72,10 @@ int main (int argc, char ** argv)
       double a=2.0;
 
 
+      axpy(z,a,x,y);
       double start=usecond();
       for(int i=0;i<Nloop;i++){
 	axpy(z,a,x,y);
-        x._odata[0]=z._odata[0]; // serial loop dependence to prevent optimise
-        y._odata[4]=z._odata[4];
       }
       double stop=usecond();
       double time = (stop-start)/Nloop*1000;
@@ -96,7 +94,7 @@ int main (int argc, char ** argv)
   
   for(int lat=8;lat<=lmax;lat+=8){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
       int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
@@ -109,11 +107,10 @@ int main (int argc, char ** argv)
 
       uint64_t Nloop=NLOOP;
 
+      z=a*x-y;
       double start=usecond();
       for(int i=0;i<Nloop;i++){
 	z=a*x-y;
-        x._odata[0]=z._odata[0]; // force serial dependency to prevent optimise away
-        y._odata[4]=z._odata[4];
       }
       double stop=usecond();
       double time = (stop-start)/Nloop*1000;
@@ -132,7 +129,7 @@ int main (int argc, char ** argv)
   for(int lat=8;lat<=lmax;lat+=8){
 
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
       int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       uint64_t Nloop=NLOOP;
 
@@ -145,11 +142,10 @@ int main (int argc, char ** argv)
       LatticeVec y(&Grid);// random(pRNG,y);
       RealD a=2.0;
 
-
+      z=a*x;
       double start=usecond();
       for(int i=0;i<Nloop;i++){
 	z=a*x;
-        x._odata[0]=z._odata[0]*2.0;
       }
       double stop=usecond();
       double time = (stop-start)/Nloop*1000;
@@ -168,7 +164,7 @@ int main (int argc, char ** argv)
 
   for(int lat=8;lat<=lmax;lat+=8){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
       int64_t vol= latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       uint64_t Nloop=NLOOP;
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
@@ -177,12 +173,13 @@ int main (int argc, char ** argv)
       LatticeVec z(&Grid);// random(pRNG,z);
       LatticeVec x(&Grid);// random(pRNG,x);
       LatticeVec y(&Grid);// random(pRNG,y);
-      RealD a=2.0;
+      //      RealD a=2.0;
       Real nn;      
       double start=usecond();
       for(int i=0;i<Nloop;i++){
+	auto x_v = x.View();
 	nn=norm2(x);
-	vsplat(x._odata[0]._internal[0],nn);
+	vsplat(x_v[0]._internal[0],nn);
       }
       double stop=usecond();
       double time = (stop-start)/Nloop*1000;
diff --git a/benchmarks/Benchmark_meson_field.cc b/benchmarks/Benchmark_meson_field.cc
index d007fc7b..d37acb7f 100644
--- a/benchmarks/Benchmark_meson_field.cc
+++ b/benchmarks/Benchmark_meson_field.cc
@@ -30,8 +30,6 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
-
 
 #include "Grid/util/Profiling.h"
 
@@ -48,7 +46,7 @@ void sliceInnerProductMesonField(std::vector< std::vector<ComplexD> > &mat,
   int Lblock = lhs.size();
   int Rblock = rhs.size();
 
-  GridBase *grid = lhs[0]._grid;
+  GridBase *grid = lhs[0].Grid();
   
   const int    Nd = grid->_ndimension;
   const int Nsimd = grid->Nsimd();
@@ -67,9 +65,9 @@ void sliceInnerProductMesonField(std::vector< std::vector<ComplexD> > &mat,
   // sum across these down to scalars
   // splitting the SIMD
   std::vector<vector_type,alignedAllocator<vector_type> > lvSum(rd*Lblock*Rblock);
-  parallel_for (int r = 0; r < rd * Lblock * Rblock; r++){
-    lvSum[r] = zero;
-  }
+  thread_for(r, rd * Lblock * Rblock,{
+    lvSum[r] = Zero();
+  });
 
   std::vector<scalar_type > lsSum(ld*Lblock*Rblock,scalar_type(0.0));             
 
@@ -79,7 +77,7 @@ void sliceInnerProductMesonField(std::vector< std::vector<ComplexD> > &mat,
   
   std::cout << GridLogMessage << " Entering first parallel loop "<<std::endl;
   // Parallelise over t-direction doesn't expose as much parallelism as needed for KNL
-  parallel_for(int r=0;r<rd;r++){
+  thread_for(r,rd,{
 
     int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
 
@@ -87,10 +85,12 @@ void sliceInnerProductMesonField(std::vector< std::vector<ComplexD> > &mat,
       for(int b=0;b<e2;b++){
 	int ss= so+n*stride+b;
 	for(int i=0;i<Lblock;i++){
-	  auto left = conjugate(lhs[i]._odata[ss]);
+	  auto lhs_v = lhs[i].View();
+	  auto left = conjugate(lhs_v[ss]);
 	  for(int j=0;j<Rblock;j++){
 	    int idx = i+Lblock*j+Lblock*Rblock*r;
-	    auto right = rhs[j]._odata[ss];
+	    auto rhs_v = rhs[j].View();
+	    auto right = rhs_v[ss];
 	    vector_type vv = left()(0)(0) * right()(0)(0)
 	      +              left()(0)(1) * right()(0)(1)
 	      +              left()(0)(2) * right()(0)(2)
@@ -108,19 +108,19 @@ void sliceInnerProductMesonField(std::vector< std::vector<ComplexD> > &mat,
 	}
       }
     }
-  }
+  });
 
   std::cout << GridLogMessage << " Entering second parallel loop "<<std::endl;
   // Sum across simd lanes in the plane, breaking out orthog dir.
-  parallel_for(int rt=0;rt<rd;rt++){
-
-    std::vector<int> icoor(Nd);
+  thread_for(rt,rd,{
+    
+    Coordinate icoor(Nd);
 
     for(int i=0;i<Lblock;i++){
     for(int j=0;j<Rblock;j++){
 
       iScalar<vector_type> temp; 
-      std::vector<iScalar<scalar_type> > extracted(Nsimd);               
+      ExtractBuffer<iScalar<scalar_type> > extracted(Nsimd);               
 
       temp._internal = lvSum[i+Lblock*j+Lblock*Rblock*rt];
 
@@ -137,7 +137,7 @@ void sliceInnerProductMesonField(std::vector< std::vector<ComplexD> > &mat,
 
       }
     }}
-  }
+  });
 
   std::cout << GridLogMessage << " Entering non parallel loop "<<std::endl;
   for(int t=0;t<fd;t++)
@@ -174,7 +174,7 @@ void sliceInnerProductMesonFieldGamma(std::vector< std::vector<ComplexD> > &mat,
   int Lblock = lhs.size();
   int Rblock = rhs.size();
 
-  GridBase *grid = lhs[0]._grid;
+  GridBase *grid = lhs[0].Grid();
   
   const int    Nd = grid->_ndimension;
   const int Nsimd = grid->Nsimd();
@@ -197,14 +197,14 @@ void sliceInnerProductMesonFieldGamma(std::vector< std::vector<ComplexD> > &mat,
   int MFlvol = ld*Lblock*Rblock*Ngamma;
 
   std::vector<vector_type,alignedAllocator<vector_type> > lvSum(MFrvol);
-  parallel_for (int r = 0; r < MFrvol; r++){
-    lvSum[r] = zero;
-  }
+  thread_for(r,MFrvol,{
+    lvSum[r] = Zero();
+  });
 
   std::vector<scalar_type > lsSum(MFlvol);             
-  parallel_for (int r = 0; r < MFlvol; r++){
+  thread_for( r,MFlvol,{
     lsSum[r]=scalar_type(0.0);
-  }
+  });
 
   int e1=    grid->_slice_nblock[orthogdim];
   int e2=    grid->_slice_block [orthogdim];
@@ -213,7 +213,7 @@ void sliceInnerProductMesonFieldGamma(std::vector< std::vector<ComplexD> > &mat,
   std::cout << GridLogMessage << " Entering first parallel loop "<<std::endl;
 
   // Parallelise over t-direction doesn't expose as much parallelism as needed for KNL
-  parallel_for(int r=0;r<rd;r++){
+  thread_for(r,rd,{
 
     int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
 
@@ -221,11 +221,13 @@ void sliceInnerProductMesonFieldGamma(std::vector< std::vector<ComplexD> > &mat,
       for(int b=0;b<e2;b++){
 	int ss= so+n*stride+b;
 	for(int i=0;i<Lblock;i++){
-	    auto left = conjugate(lhs[i]._odata[ss]);
-	    for(int j=0;j<Rblock;j++){
+ 	  auto lhs_v=lhs[i].View();
+	  auto left = conjugate(lhs_v[ss]);
+	  for(int j=0;j<Rblock;j++){
 	  for(int mu=0;mu<Ngamma;mu++){
-
-        auto right = Gamma(gammas[mu])*rhs[j]._odata[ss];
+	    
+	    auto rhs_v = rhs[j].View();
+	    auto right = Gamma(gammas[mu])*rhs_v[ss];
 
 	      vector_type vv = left()(0)(0) * right()(0)(0)
 		+              left()(0)(1) * right()(0)(1)
@@ -248,15 +250,15 @@ void sliceInnerProductMesonFieldGamma(std::vector< std::vector<ComplexD> > &mat,
 	}
       }
     }
-  }
+  });
 
   std::cout << GridLogMessage << " Entering second parallel loop "<<std::endl;
   // Sum across simd lanes in the plane, breaking out orthog dir.
-  parallel_for(int rt=0;rt<rd;rt++){
+  thread_for(rt,rd,{
 
     iScalar<vector_type> temp; 
-    std::vector<int> icoor(Nd);
-    std::vector<iScalar<scalar_type> > extracted(Nsimd);               
+    Coordinate icoor(Nd);
+    ExtractBuffer<iScalar<scalar_type> > extracted(Nsimd);               
 
     for(int i=0;i<Lblock;i++){
     for(int j=0;j<Rblock;j++){
@@ -278,7 +280,7 @@ void sliceInnerProductMesonFieldGamma(std::vector< std::vector<ComplexD> > &mat,
 
       }
     }}}
-  }
+  });
 
   std::cout << GridLogMessage << " Entering non parallel loop "<<std::endl;
   for(int t=0;t<fd;t++)
@@ -320,7 +322,7 @@ void sliceInnerProductMesonFieldGamma1(std::vector< std::vector<ComplexD> > &mat
   int Lblock = lhs.size();
   int Rblock = rhs.size();
 
-  GridBase *grid = lhs[0]._grid;
+  GridBase *grid = lhs[0].Grid();
   
   const int    Nd = grid->_ndimension;
   const int Nsimd = grid->Nsimd();
@@ -343,14 +345,14 @@ void sliceInnerProductMesonFieldGamma1(std::vector< std::vector<ComplexD> > &mat
   int MFlvol = ld*Lblock*Rblock;
 
   Vector<SpinMatrix_v > lvSum(MFrvol);
-  parallel_for (int r = 0; r < MFrvol; r++){
-    lvSum[r] = zero;
-  }
+  thread_for(r,MFrvol,{
+    lvSum[r] = Zero();
+  });
 
   Vector<SpinMatrix_s > lsSum(MFlvol);             
-  parallel_for (int r = 0; r < MFlvol; r++){
+  thread_for(r,MFlvol,{
     lsSum[r]=scalar_type(0.0);
-  }
+  });
 
   int e1=    grid->_slice_nblock[orthogdim];
   int e2=    grid->_slice_block [orthogdim];
@@ -359,7 +361,7 @@ void sliceInnerProductMesonFieldGamma1(std::vector< std::vector<ComplexD> > &mat
   std::cout << GridLogMessage << " Entering first parallel loop "<<std::endl;
 
   // Parallelise over t-direction doesn't expose as much parallelism as needed for KNL
-  parallel_for(int r=0;r<rd;r++){
+  thread_for(r,rd,{
 
     int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
 
@@ -367,12 +369,14 @@ void sliceInnerProductMesonFieldGamma1(std::vector< std::vector<ComplexD> > &mat
       for(int b=0;b<e2;b++){
 	int ss= so+n*stride+b;
 	for(int i=0;i<Lblock;i++){
-
-	  auto left = conjugate(lhs[i]._odata[ss]);
+	  
+	  auto lhs_v=lhs[i].View();
+	  auto left = conjugate(lhs_v[ss]);
 	  for(int j=0;j<Rblock;j++){
 
 	    SpinMatrix_v vv;
-	    auto right = rhs[j]._odata[ss];
+	    auto rhs_v = rhs[j].View();
+	    auto right = rhs_v[ss];
 	    for(int s1=0;s1<Ns;s1++){
 	    for(int s2=0;s2<Ns;s2++){
 	     vv()(s2,s1)() = left()(s1)(0) * right()(s2)(0)
@@ -388,14 +392,14 @@ void sliceInnerProductMesonFieldGamma1(std::vector< std::vector<ComplexD> > &mat
 	  }
 	}
       }
-    }
+  });
 
   std::cout << GridLogMessage << " Entering second parallel loop "<<std::endl;
   // Sum across simd lanes in the plane, breaking out orthog dir.
-  parallel_for(int rt=0;rt<rd;rt++){
+  thread_for(rt,rd,{
 
-    std::vector<int> icoor(Nd);
-    std::vector<SpinMatrix_s> extracted(Nsimd);               
+    Coordinate icoor(Nd);
+    ExtractBuffer<SpinMatrix_s> extracted(Nsimd);               
 
     for(int i=0;i<Lblock;i++){
     for(int j=0;j<Rblock;j++){
@@ -416,10 +420,10 @@ void sliceInnerProductMesonFieldGamma1(std::vector< std::vector<ComplexD> > &mat
 
       }
     }}
-  }
+  });
 
   std::cout << GridLogMessage << " Entering third parallel loop "<<std::endl;
-  parallel_for(int t=0;t<fd;t++)
+  thread_for(t,fd,
   {
     int pt = t / ld; // processor plane
     int lt = t % ld;
@@ -428,7 +432,7 @@ void sliceInnerProductMesonFieldGamma1(std::vector< std::vector<ComplexD> > &mat
       if (pt == grid->_processor_coor[orthogdim]){
         int ij_dx = i + Lblock * j + Lblock * Rblock * lt;
     	for(int mu=0;mu<Ngamma;mu++){
-	  mat[mu+i*Ngamma+j*Lblock*Ngamma][t] = trace(lsSum[ij_dx]*Gamma(gammas[mu]));
+	  mat[mu+i*Ngamma+j*Lblock*Ngamma][t] = TensorRemove(trace(lsSum[ij_dx]*Gamma(gammas[mu])));
 	}
       }
       else{
@@ -437,7 +441,7 @@ void sliceInnerProductMesonFieldGamma1(std::vector< std::vector<ComplexD> > &mat
 	}
       }
     }}
-  }
+  });
   std::cout << GridLogMessage << " Done "<<std::endl;
   // defer sum over nodes.
   return;
@@ -460,7 +464,7 @@ void sliceInnerProductMesonFieldGammaMom(std::vector< std::vector<ComplexD> > &m
   int Lblock = lhs.size();
   int Rblock = rhs.size();
 
-  GridBase *grid = lhs[0]._grid;
+  GridBase *grid = lhs[0].Grid();
   
   const int    Nd = grid->_ndimension;
   const int Nsimd = grid->Nsimd();
@@ -484,14 +488,16 @@ void sliceInnerProductMesonFieldGammaMom(std::vector< std::vector<ComplexD> > &m
   int MFlvol = ld*Lblock*Rblock*Nmom;
 
   Vector<SpinMatrix_v > lvSum(MFrvol);
-  parallel_for (int r = 0; r < MFrvol; r++){
-    lvSum[r] = zero;
-  }
+  thread_for(r,MFrvol,
+  {
+    lvSum[r] = Zero();
+  });
 
   Vector<SpinMatrix_s > lsSum(MFlvol);             
-  parallel_for (int r = 0; r < MFlvol; r++){
+  thread_for(r,MFlvol,
+  {
     lsSum[r]=scalar_type(0.0);
-  }
+  });
 
   int e1=    grid->_slice_nblock[orthogdim];
   int e2=    grid->_slice_block [orthogdim];
@@ -500,7 +506,8 @@ void sliceInnerProductMesonFieldGammaMom(std::vector< std::vector<ComplexD> > &m
   std::cout << GridLogMessage << " Entering first parallel loop "<<std::endl;
 
   // Parallelise over t-direction doesn't expose as much parallelism as needed for KNL
-  parallel_for(int r=0;r<rd;r++){
+  thread_for(r,rd,
+  {
 
     int so=r*grid->_ostride[orthogdim]; // base offset for start of plane 
 
@@ -511,11 +518,13 @@ void sliceInnerProductMesonFieldGammaMom(std::vector< std::vector<ComplexD> > &m
 
 	for(int i=0;i<Lblock;i++){
 
-	  auto left = conjugate(lhs[i]._odata[ss]);
+	  auto lhs_v = lhs[i].View();
+	  auto left = conjugate(lhs_v[ss]);
 	  for(int j=0;j<Rblock;j++){
 
 	    SpinMatrix_v vv;
-	    auto right = rhs[j]._odata[ss];
+	    auto rhs_v = rhs[j].View();
+	    auto right = rhs_v[ss];
 	    for(int s1=0;s1<Ns;s1++){
 	    for(int s2=0;s2<Ns;s2++){
 	      vv()(s1,s2)() = left()(s1)(0) * right()(s2)(0)
@@ -528,7 +537,8 @@ void sliceInnerProductMesonFieldGammaMom(std::vector< std::vector<ComplexD> > &m
 	    // Trigger unroll
 	    for ( int m=0;m<Nmom;m++){
 	      int idx = m+base;
-	      auto phase = mom[m]._odata[ss];
+	      auto mom_v = mom[m].View();
+	      auto phase = mom_v[ss];
 	      mac(&lvSum[idx],&vv,&phase);
 	    }
 	  
@@ -536,14 +546,15 @@ void sliceInnerProductMesonFieldGammaMom(std::vector< std::vector<ComplexD> > &m
 	}
       }
     }
-  }
+  });
 
   std::cout << GridLogMessage << " Entering second parallel loop "<<std::endl;
   // Sum across simd lanes in the plane, breaking out orthog dir.
-  parallel_for(int rt=0;rt<rd;rt++){
+  thread_for(rt,rd,
+  {
 
-    std::vector<int> icoor(Nd);
-    std::vector<SpinMatrix_s> extracted(Nsimd);               
+    Coordinate icoor(Nd);
+    ExtractBuffer<SpinMatrix_s> extracted(Nsimd);               
 
 
     for(int i=0;i<Lblock;i++){
@@ -566,10 +577,10 @@ void sliceInnerProductMesonFieldGammaMom(std::vector< std::vector<ComplexD> > &m
 
       }
     }}}
-  }
+  });
 
   std::cout << GridLogMessage << " Entering third parallel loop "<<std::endl;
-  parallel_for(int t=0;t<fd;t++)
+  thread_for(t,fd,
   {
     int pt = t / ld; // processor plane
     int lt = t % ld;
@@ -582,7 +593,7 @@ void sliceInnerProductMesonFieldGammaMom(std::vector< std::vector<ComplexD> > &m
 	    mat[ mu
 		+m*Ngamma
 		+i*Nmom*Ngamma
-		+j*Nmom*Ngamma*Lblock][t] = trace(lsSum[ij_dx]*Gamma(gammas[mu]));
+		 +j*Nmom*Ngamma*Lblock][t] = TensorRemove(trace(lsSum[ij_dx]*Gamma(gammas[mu])));
 	  }
 	}
       }
@@ -593,7 +604,7 @@ void sliceInnerProductMesonFieldGammaMom(std::vector< std::vector<ComplexD> > &m
 	}}
       }
     }}
-  }
+  });
   std::cout << GridLogMessage << " Done "<<std::endl;
   // defer sum over nodes.
   return;
@@ -637,9 +648,9 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  auto latt_size   = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   
   const int Nmom=7;
diff --git a/benchmarks/Benchmark_mooee.cc b/benchmarks/Benchmark_mooee.cc
index 1e51c9d2..ef16c908 100644
--- a/benchmarks/Benchmark_mooee.cc
+++ b/benchmarks/Benchmark_mooee.cc
@@ -30,7 +30,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 
 int main (int argc, char ** argv)
@@ -40,19 +40,13 @@ int main (int argc, char ** argv)
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
-  std::vector<int> latt4 = GridDefaultLatt();
+  Coordinate latt4 = GridDefaultLatt();
   const int Ls=16;
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
   GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
   GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
   GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
 
-  std::cout << GridLogMessage << "Making Vec5d innermost grids"<<std::endl;
-  GridCartesian         * sUGrid   = SpaceTimeGrid::makeFourDimDWFGrid(GridDefaultLatt(),GridDefaultMpi());
-  GridRedBlackCartesian * sUrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(sUGrid);
-  GridCartesian         * sFGrid   = SpaceTimeGrid::makeFiveDimDWFGrid(Ls,UGrid);
-  GridRedBlackCartesian * sFrbGrid = SpaceTimeGrid::makeFiveDimDWFRedBlackGrid(Ls,UGrid);
-
   std::vector<int> seeds4({1,2,3,4});
   std::vector<int> seeds5({5,6,7,8});
 
@@ -65,7 +59,7 @@ int main (int argc, char ** argv)
 
   RealD mass=0.1;
   RealD M5  =1.8;
-  RealD NP = UGrid->_Nprocessors;
+  //  RealD NP = UGrid->_Nprocessors;
 
 
   if (1)
@@ -76,13 +70,20 @@ int main (int argc, char ** argv)
     std::cout << GridLogMessage<< "* Benchmarking DomainWallFermionR::Dhop "<<std::endl;
     std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
 
-    GridParallelRNG RNG5(FGrid);
+    GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
     LatticeFermion src(FGrid); random(RNG5,src);
     LatticeFermion result(FGrid);
 
     DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
     double t0,t1;
-
+    
+    typedef typename DomainWallFermionR::Coeff_t Coeff_t;
+    Vector<Coeff_t> diag = Dw.bs;
+    Vector<Coeff_t> upper= Dw.cs;
+    Vector<Coeff_t> lower= Dw.cs;
+    upper[Ls-1]=-Dw.mass*upper[Ls-1];
+    lower[0]   =-Dw.mass*lower[0];
+    
     LatticeFermion r_eo(FGrid);
     LatticeFermion src_e (FrbGrid);
     LatticeFermion src_o (FrbGrid);
@@ -95,17 +96,17 @@ int main (int argc, char ** argv)
     setCheckerboard(r_eo,src_o);
     setCheckerboard(r_eo,src_e);
     
-    r_e = zero;
-    r_o = zero;
+    r_e = Zero();
+    r_o = Zero();
 
 
-#define BENCH_DW(A,in,out)			\
-    Dw.CayleyZeroCounters();			\
-    Dw. A (in,out);				\
+#define BENCH_DW(A,...)			\
+    Dw. A (__VA_ARGS__);				\
     FGrid->Barrier();				\
+    Dw.CayleyZeroCounters();      \
     t0=usecond();				\
     for(int i=0;i<ncall;i++){			\
-      Dw. A (in,out);				\
+      Dw. A (__VA_ARGS__);				\
     }						\
     t1=usecond();				\
     FGrid->Barrier();				\
@@ -114,9 +115,9 @@ int main (int argc, char ** argv)
     std::cout<<GridLogMessage << "******************"<<std::endl;
 
 #define BENCH_ZDW(A,in,out)			\
-    zDw.CayleyZeroCounters();			\
     zDw. A (in,out);				\
     FGrid->Barrier();				\
+    zDw.CayleyZeroCounters();      \
     t0=usecond();				\
     for(int i=0;i<ncall;i++){			\
       zDw. A (in,out);				\
@@ -128,9 +129,9 @@ int main (int argc, char ** argv)
     std::cout<<GridLogMessage << "******************"<<std::endl;
 
 #define BENCH_DW_SSC(A,in,out)			\
-    Dw.CayleyZeroCounters();			\
     Dw. A (in,out);				\
     FGrid->Barrier();				\
+    Dw.CayleyZeroCounters();      \
     t0=usecond();				\
     for(int i=0;i<ncall;i++){			\
       __SSC_START ;				\
@@ -143,80 +144,14 @@ int main (int argc, char ** argv)
     std::cout<<GridLogMessage << "Called " #A " "<< (t1-t0)/ncall<<" us"<<std::endl;\
     std::cout<<GridLogMessage << "******************"<<std::endl;
 
-#define BENCH_DW_MEO(A,in,out)			\
-    Dw.CayleyZeroCounters();			\
-    Dw. A (in,out,0);				\
-    FGrid->Barrier();				\
-    t0=usecond();				\
-    for(int i=0;i<ncall;i++){			\
-      Dw. A (in,out,0);				\
-    }						\
-    t1=usecond();				\
-    FGrid->Barrier();				\
-    Dw.CayleyReport();					\
-    std::cout<<GridLogMessage << "Called " #A " "<< (t1-t0)/ncall<<" us"<<std::endl;\
-    std::cout<<GridLogMessage << "******************"<<std::endl;
-
-    BENCH_DW_MEO(Dhop    ,src,result);
-    BENCH_DW_MEO(DhopEO  ,src_o,r_e);
+    BENCH_DW(Dhop    ,src,result,0);
+    BENCH_DW(DhopEO  ,src_o,r_e,0);
     BENCH_DW(Meooe   ,src_o,r_e);
+    BENCH_DW(M5D     ,src_o,src_o,r_e,lower,diag,upper);
     BENCH_DW(Mooee   ,src_o,r_o);
     BENCH_DW(MooeeInv,src_o,r_o);
 
   }
 
-  if (1)
-  {
-    const int ncall=1000;
-
-    std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
-    std::cout << GridLogMessage<< "* Benchmarking DomainWallFermionVec5dR::Dhop "<<std::endl;
-    std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
-
-    GridParallelRNG RNG5(sFGrid);
-    LatticeFermion src(sFGrid); random(RNG5,src);
-    LatticeFermion sref(sFGrid);
-    LatticeFermion result(sFGrid);
-
-
-    std::cout<<GridLogMessage << "Constructing Vec5D Dw "<<std::endl;
-    DomainWallFermionVec5dR Dw(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5);
-
-    RealD b=1.5;// Scale factor b+c=2, b-c=1
-    RealD c=0.5;
-    std::vector<ComplexD> gamma(Ls,std::complex<double>(1.0,0.0));
-    ZMobiusFermionVec5dR zDw(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,gamma,b,c);
-
-    std::cout<<GridLogMessage << "Calling Dhop "<<std::endl;
-    FGrid->Barrier();
-
-    double t0,t1;
-
-    LatticeFermion r_eo(sFGrid);
-    LatticeFermion src_e (sFrbGrid);
-    LatticeFermion src_o (sFrbGrid);
-    LatticeFermion r_e   (sFrbGrid);
-    LatticeFermion r_o   (sFrbGrid);
-    
-    pickCheckerboard(Even,src_e,src);
-    pickCheckerboard(Odd,src_o,src);
-    
-    setCheckerboard(r_eo,src_o);
-    setCheckerboard(r_eo,src_e);
-    
-    r_e = zero;
-    r_o = zero;
-
-    BENCH_DW_MEO(Dhop    ,src,result);
-    BENCH_DW_MEO(DhopEO  ,src_o,r_e);
-    BENCH_DW_SSC(Meooe   ,src_o,r_e);
-    BENCH_DW(Mooee   ,src_o,r_o);
-    BENCH_DW(MooeeInv,src_o,r_o);
-
-    BENCH_ZDW(Mooee   ,src_o,r_o);
-    BENCH_ZDW(MooeeInv,src_o,r_o);
-
-  }
-
   Grid_finalize();
 }
diff --git a/benchmarks/Benchmark_staggered.cc b/benchmarks/Benchmark_staggered.cc
index f5325b28..93086927 100644
--- a/benchmarks/Benchmark_staggered.cc
+++ b/benchmarks/Benchmark_staggered.cc
@@ -30,15 +30,15 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -57,10 +57,10 @@ int main (int argc, char ** argv)
   typename ImprovedStaggeredFermionR::ImplParams params; 
 
   FermionField src   (&Grid); random(pRNG,src);
-  FermionField result(&Grid); result=zero;
-  FermionField    ref(&Grid);    ref=zero;
-  FermionField    tmp(&Grid);    tmp=zero;
-  FermionField    err(&Grid);    tmp=zero;
+  FermionField result(&Grid); result=Zero();
+  FermionField    ref(&Grid);    ref=Zero();
+  FermionField    tmp(&Grid);    tmp=Zero();
+  FermionField    err(&Grid);    tmp=Zero();
   LatticeGaugeField Umu(&Grid); random(pRNG,Umu);
   std::vector<LatticeColourMatrix> U(4,&Grid);
 
@@ -71,12 +71,12 @@ int main (int argc, char ** argv)
 
   // Only one non-zero (y)
 #if 0
-  Umu=zero;
+  Umu=Zero();
   Complex cone(1.0,0.0);
   for(int nn=0;nn<Nd;nn++){
     random(pRNG,U[nn]);
     if(1) {
-      if (nn!=2) { U[nn]=zero; std::cout<<GridLogMessage << "zeroing gauge field in dir "<<nn<<std::endl; }
+      if (nn!=2) { U[nn]=Zero(); std::cout<<GridLogMessage << "zeroing gauge field in dir "<<nn<<std::endl; }
       //      else       { U[nn]= cone;std::cout<<GridLogMessage << "unit gauge field in dir "<<nn<<std::endl; }
       else       { std::cout<<GridLogMessage << "random gauge field in dir "<<nn<<std::endl; }
     }
@@ -87,21 +87,21 @@ int main (int argc, char ** argv)
   for(int mu=0;mu<Nd;mu++){
     U[mu] = PeekIndex<LorentzIndex>(Umu,mu);
   }
-  ref = zero;
+  ref = Zero();
   /*  
   { // Naive wilson implementation
-    ref = zero;
+    ref = Zero();
     for(int mu=0;mu<Nd;mu++){
       //    ref =  src + Gamma(Gamma::GammaX)* src ; // 1-gamma_x
       tmp = U[mu]*Cshift(src,mu,1);
       for(int i=0;i<ref._odata.size();i++){
-	ref._odata[i]+= tmp._odata[i] - Gamma(Gmu[mu])*tmp._odata[i]; ;
+	ref[i]+= tmp[i] - Gamma(Gmu[mu])*tmp[i]; ;
       }
 
       tmp =adj(U[mu])*src;
       tmp =Cshift(tmp,mu,-1);
       for(int i=0;i<ref._odata.size();i++){
-	ref._odata[i]+= tmp._odata[i] + Gamma(Gmu[mu])*tmp._odata[i]; ;
+	ref[i]+= tmp[i] + Gamma(Gmu[mu])*tmp[i]; ;
       }
     }
   }
diff --git a/benchmarks/Benchmark_su3.cc b/benchmarks/Benchmark_su3.cc
index 7b1b2c1a..d24a3e25 100644
--- a/benchmarks/Benchmark_su3.cc
+++ b/benchmarks/Benchmark_su3.cc
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -30,19 +30,21 @@ Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
-#define LMAX (32)
-#define LMIN (16)
-#define LINC (4)
 
-  int64_t Nloop=2000;
+#define LMAX (48)
+#define LMIN (8)
+#define LADD (8)
 
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  int64_t Nwarm=50;
+  int64_t Nloop=500;
+
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   int64_t threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
@@ -53,9 +55,9 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=LMIN;lat<=LMAX;lat+=LINC){
+  for(int lat=LMIN;lat<=LMAX;lat+=LADD){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
       int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
       GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
@@ -64,6 +66,9 @@ int main (int argc, char ** argv)
       LatticeColourMatrix x(&Grid); random(pRNG,x);
       LatticeColourMatrix y(&Grid); random(pRNG,y);
 
+      for(int64_t i=0;i<Nwarm;i++){
+	x=x*y;
+      }
       double start=usecond();
       for(int64_t i=0;i<Nloop;i++){
 	x=x*y;
@@ -85,9 +90,9 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=LMIN;lat<=LMAX;lat+=LINC){
+  for(int lat=LMIN;lat<=LMAX;lat+=LADD){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
       int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
@@ -97,6 +102,9 @@ int main (int argc, char ** argv)
       LatticeColourMatrix x(&Grid); random(pRNG,x);
       LatticeColourMatrix y(&Grid); random(pRNG,y);
 
+      for(int64_t i=0;i<Nwarm;i++){
+	z=x*y;
+      }
       double start=usecond();
       for(int64_t i=0;i<Nloop;i++){
 	z=x*y;
@@ -116,9 +124,10 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=LMIN;lat<=LMAX;lat+=LINC){
 
-      std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+  for(int lat=LMIN;lat<=LMAX;lat+=LADD){
+
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
       int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
 
       GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
@@ -128,9 +137,13 @@ int main (int argc, char ** argv)
       LatticeColourMatrix x(&Grid); random(pRNG,x);
       LatticeColourMatrix y(&Grid); random(pRNG,y);
 
+      for(int64_t i=0;i<Nwarm;i++){
+	mult(z,x,y);
+      }
       double start=usecond();
       for(int64_t i=0;i<Nloop;i++){
 	mult(z,x,y);
+	//	mac(z,x,y);
       }
       double stop=usecond();
       double time = (stop-start)/Nloop*1000.0;
@@ -147,30 +160,33 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=LMIN;lat<=LMAX;lat+=LINC){
-    
-    std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
-    int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
-    
-    GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
-    GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
-    
-    LatticeColourMatrix z(&Grid); random(pRNG,z);
-    LatticeColourMatrix x(&Grid); random(pRNG,x);
-    LatticeColourMatrix y(&Grid); random(pRNG,y);
-    
-    double start=usecond();
-    for(int64_t i=0;i<Nloop;i++){
-      mac(z,x,y);
+  for(int lat=LMIN;lat<=LMAX;lat+=LADD){
+
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+
+      LatticeColourMatrix z(&Grid); random(pRNG,z);
+      LatticeColourMatrix x(&Grid); random(pRNG,x);
+      LatticeColourMatrix y(&Grid); random(pRNG,y);
+
+      for(int64_t i=0;i<Nwarm;i++){
+	mac(z,x,y);
+      }
+      double start=usecond();
+      for(int64_t i=0;i<Nloop;i++){
+	mac(z,x,y);
+      }
+      double stop=usecond();
+      double time = (stop-start)/Nloop*1000.0;
+      
+      double bytes=4*vol*Nc*Nc*sizeof(Complex);
+      double flops=Nc*Nc*(8+8+8)*vol;
+      std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
+
     }
-    double stop=usecond();
-    double time = (stop-start)/Nloop*1000.0;
-    
-    double bytes=3*vol*Nc*Nc*sizeof(Complex);
-    double flops=Nc*Nc*(6+8+8)*vol;
-    std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
-    
-  }
 
   std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
   std::cout<<GridLogMessage << "= Benchmarking SU3xSU3  CovShiftForward(z,x,y)"<<std::endl;
@@ -178,7 +194,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=LMIN;lat<=LMAX;lat+=LINC){
+  for(int lat=LMIN;lat<=LMAX;lat+=LADD){
 
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
       int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
@@ -211,7 +227,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
   std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
 
-  for(int lat=LMIN;lat<=LMAX;lat+=LINC){
+  for(int lat=LMIN;lat<=LMAX;lat+=LADD){
       std::vector<int> latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
       int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
 
diff --git a/benchmarks/Benchmark_su3_gpu.cc b/benchmarks/Benchmark_su3_gpu.cc
new file mode 100644
index 00000000..c87f501e
--- /dev/null
+++ b/benchmarks/Benchmark_su3_gpu.cc
@@ -0,0 +1,400 @@
+/*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./benchmarks/Benchmark_su3.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+
+using namespace std;
+using namespace Grid;
+
+
+int main (int argc, char ** argv)
+{
+  Grid_init(&argc,&argv);
+
+#define LMIN (16)
+#define LMAX (48)
+#define LADD (8)
+
+  int64_t Nwarm=20;
+  int64_t Nloop=500;
+
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
+
+  int64_t threads = GridThread::GetThreads();
+  std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
+
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking SU3xSU3  ext/ins z = x*y"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
+  std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
+
+  for(int lat=LMIN;lat<=LMAX;lat+=LADD){
+
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+
+      LatticeColourMatrix z(&Grid);// random(pRNG,z);
+      LatticeColourMatrix x(&Grid);// random(pRNG,x);
+      LatticeColourMatrix y(&Grid);// random(pRNG,y);
+
+      auto x_v = x.View();
+      auto y_v = y.View();
+      auto z_v = z.View();
+      const uint64_t Nsite = x_v.size();
+      const uint64_t nsimd = vComplex::Nsimd();
+      for(int64_t i=0;i<Nwarm;i++){
+	accelerator_for( ss, Nsite, nsimd ,{
+	  auto xx = coalescedRead(x_v[ss]);
+	  auto yy = coalescedRead(y_v[ss]);
+	  auto zz = xx*yy;
+	  coalescedWrite(z_v[ss],zz);
+        });
+      }
+      double start=usecond();
+      for(int64_t i=0;i<Nloop;i++){
+	accelerator_for( ss, Nsite, nsimd ,{
+	  auto xx = coalescedRead(x_v[ss]);
+	  auto yy = coalescedRead(y_v[ss]);
+	  auto zz = xx*yy;
+	  coalescedWrite(z_v[ss],zz);
+        });
+      }
+      double stop=usecond();
+      double time = (stop-start)/Nloop*1000.0;
+      
+      double bytes=3.0*vol*Nc*Nc*sizeof(Complex);
+      double footprint=3.0*vol*Nc*Nc*sizeof(Complex);
+      double flops=Nc*Nc*(6.0+8.0+8.0)*vol;
+      std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<footprint<<"    \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
+
+    }
+
+
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking SU3xSU3  ext/ins x = x*y"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
+  std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
+
+  for(int lat=LMIN;lat<=LMAX;lat+=LADD){
+
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+
+      LatticeColourMatrix z(&Grid);// random(pRNG,z);
+      LatticeColourMatrix x(&Grid);// random(pRNG,x);
+      LatticeColourMatrix y(&Grid);// random(pRNG,y);
+
+      auto x_v = x.View();
+      auto y_v = y.View();
+      auto z_v = z.View();
+      const uint64_t Nsite = x_v.size();
+      const uint64_t nsimd = vComplex::Nsimd();
+      for(int64_t i=0;i<Nwarm;i++){
+	accelerator_for( ss, Nsite, nsimd ,{
+	  auto xx = coalescedRead(x_v[ss]);
+	  auto yy = coalescedRead(y_v[ss]);
+	  auto zz = xx*yy;
+	  coalescedWrite(x_v[ss],zz);
+        });
+      }
+      double start=usecond();
+      for(int64_t i=0;i<Nloop;i++){
+	accelerator_for( ss, Nsite, nsimd ,{
+	  auto xx = coalescedRead(x_v[ss]);
+	  auto yy = coalescedRead(y_v[ss]);
+	  auto zz = xx*yy;
+	  coalescedWrite(x_v[ss],zz);
+        });
+      }
+      double stop=usecond();
+      double time = (stop-start)/Nloop*1000.0;
+      
+      double bytes=3.0*vol*Nc*Nc*sizeof(Complex);
+      double footprint=2.0*vol*Nc*Nc*sizeof(Complex);
+      double flops=Nc*Nc*(6.0+8.0+8.0)*vol;
+      std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<footprint<<"    \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
+
+    }
+
+
+
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking SU3xSU3  ext/ins z = z + x*y"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
+  std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
+
+  for(int lat=LMIN;lat<=LMAX;lat+=LADD){
+
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+
+      LatticeColourMatrix z(&Grid);// random(pRNG,z);
+      LatticeColourMatrix x(&Grid);// random(pRNG,x);
+      LatticeColourMatrix y(&Grid);// random(pRNG,y);
+
+      auto x_v = x.View();
+      auto y_v = y.View();
+      auto z_v = z.View();
+      const uint64_t Nsite = x_v.size();
+      const uint64_t nsimd = vComplex::Nsimd();
+      for(int64_t i=0;i<Nwarm;i++){
+	accelerator_for( ss, Nsite, nsimd ,{
+	  auto xx = coalescedRead(x_v[ss]);
+	  auto yy = coalescedRead(y_v[ss]);
+	  auto zz = coalescedRead(z_v[ss]);
+	  zz = zz+xx*yy;
+	  coalescedWrite(z_v[ss],zz);
+        });
+      }
+      double start=usecond();
+      for(int64_t i=0;i<Nloop;i++){
+	accelerator_for( ss, Nsite, nsimd ,{
+	  auto xx = coalescedRead(x_v[ss]);
+	  auto yy = coalescedRead(y_v[ss]);
+	  auto zz = coalescedRead(z_v[ss]);
+	  zz = zz+xx*yy;
+	  coalescedWrite(z_v[ss],zz);
+        });
+      }
+      double stop=usecond();
+      double time = (stop-start)/Nloop*1000.0;
+      
+      double bytes=4.0*vol*Nc*Nc*sizeof(Complex);
+      double footprint=3.0*vol*Nc*Nc*sizeof(Complex);
+      double flops=Nc*Nc*(6.0+8.0+8.0)*vol;
+      std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<footprint<<"    \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
+
+    }
+
+
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking SU3xSU3  ext/ins w = z + x*y"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
+  std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
+
+  for(int lat=LMIN;lat<=LMAX;lat+=LADD){
+
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+
+      LatticeColourMatrix z(&Grid);// random(pRNG,z);
+      LatticeColourMatrix x(&Grid);// random(pRNG,x);
+      LatticeColourMatrix y(&Grid);// random(pRNG,y);
+      LatticeColourMatrix w(&Grid);// random(pRNG,y);
+
+      auto x_v = x.View();
+      auto y_v = y.View();
+      auto z_v = z.View();
+      auto w_v = z.View();
+      const uint64_t Nsite = x_v.size();
+      const uint64_t nsimd = vComplex::Nsimd();
+      for(int64_t i=0;i<Nwarm;i++){
+	accelerator_for( ss, Nsite, nsimd ,{
+	  auto xx = coalescedRead(x_v[ss]);
+	  auto yy = coalescedRead(y_v[ss]);
+	  auto zz = coalescedRead(z_v[ss]);
+	  auto ww = zz+xx*yy;
+	  coalescedWrite(w_v[ss],ww);
+        });
+      }
+      double start=usecond();
+      for(int64_t i=0;i<Nloop;i++){
+	accelerator_for( ss, Nsite, nsimd ,{
+	  auto xx = coalescedRead(x_v[ss]);
+	  auto yy = coalescedRead(y_v[ss]);
+	  auto zz = coalescedRead(z_v[ss]);
+	  auto ww = zz+xx*yy;
+	  coalescedWrite(w_v[ss],ww);
+        });
+      }
+      double stop=usecond();
+      double time = (stop-start)/Nloop*1000.0;
+      
+      double bytes=4.0*vol*Nc*Nc*sizeof(Complex);
+      double footprint=3.0*vol*Nc*Nc*sizeof(Complex);
+      double flops=Nc*Nc*(6.0+8.0+8.0)*vol;
+      std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<footprint<<"    \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
+
+    }
+
+
+
+#if 1
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking SU3xSU3  x= x*y"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
+  std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
+
+  for(int lat=LMIN;lat<=LMAX;lat+=LADD){
+
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+
+      LatticeColourMatrix z(&Grid); random(pRNG,z);
+      LatticeColourMatrix x(&Grid); random(pRNG,x);
+      LatticeColourMatrix y(&Grid); random(pRNG,y);
+
+      for(int64_t i=0;i<Nwarm;i++){
+	x=x*y;
+      }
+      double start=usecond();
+      for(int64_t i=0;i<Nloop;i++){
+	x=x*y;
+      }
+      double stop=usecond();
+      double time = (stop-start)/Nloop*1000.0;
+      
+      double bytes=3.0*vol*Nc*Nc*sizeof(Complex);
+      double footprint=2.0*vol*Nc*Nc*sizeof(Complex);
+      double flops=Nc*Nc*(6.0+8.0+8.0)*vol;
+      std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<footprint<<"    \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
+
+    }
+
+
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking SU3xSU3  z= x*y"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
+  std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
+
+  for(int lat=LMIN;lat<=LMAX;lat+=LADD){
+
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+
+      LatticeColourMatrix z(&Grid); random(pRNG,z);
+      LatticeColourMatrix x(&Grid); random(pRNG,x);
+      LatticeColourMatrix y(&Grid); random(pRNG,y);
+
+      for(int64_t i=0;i<Nwarm;i++){
+	z=x*y;
+      }
+      double start=usecond();
+      for(int64_t i=0;i<Nloop;i++){
+	z=x*y;
+      }
+      double stop=usecond();
+      double time = (stop-start)/Nloop*1000.0;
+      
+      double bytes=3*vol*Nc*Nc*sizeof(Complex);
+      double flops=Nc*Nc*(6+8+8)*vol;
+      std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"    \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
+
+    }
+
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking SU3xSU3  mult(z,x,y)"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
+  std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
+
+  for(int lat=LMIN;lat<=LMAX;lat+=LADD){
+
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+
+      LatticeColourMatrix z(&Grid); random(pRNG,z);
+      LatticeColourMatrix x(&Grid); random(pRNG,x);
+      LatticeColourMatrix y(&Grid); random(pRNG,y);
+
+      for(int64_t i=0;i<Nwarm;i++){
+	mult(z,x,y);
+      }
+      double start=usecond();
+      for(int64_t i=0;i<Nloop;i++){
+	mult(z,x,y);
+	//	mac(z,x,y);
+      }
+      double stop=usecond();
+      double time = (stop-start)/Nloop*1000.0;
+      
+      double bytes=3*vol*Nc*Nc*sizeof(Complex);
+      double flops=Nc*Nc*(6+8+8)*vol;
+      std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"    \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
+
+    }
+
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "= Benchmarking SU3xSU3  mac(z,x,y)"<<std::endl;
+  std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
+  std::cout<<GridLogMessage << "  L  "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
+  std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
+
+  for(int lat=LMIN;lat<=LMAX;lat+=LADD){
+
+      Coordinate latt_size  ({lat*mpi_layout[0],lat*mpi_layout[1],lat*mpi_layout[2],lat*mpi_layout[3]});
+      int64_t vol = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+
+      GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
+      GridParallelRNG          pRNG(&Grid);      pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9}));
+
+      LatticeColourMatrix z(&Grid); random(pRNG,z);
+      LatticeColourMatrix x(&Grid); random(pRNG,x);
+      LatticeColourMatrix y(&Grid); random(pRNG,y);
+
+      for(int64_t i=0;i<Nwarm;i++){
+	mac(z,x,y);
+      }
+      double start=usecond();
+      for(int64_t i=0;i<Nloop;i++){
+	mac(z,x,y);
+      }
+      double stop=usecond();
+      double time = (stop-start)/Nloop*1000.0;
+      
+      double bytes=4*vol*Nc*Nc*sizeof(Complex);
+      double flops=Nc*Nc*(8+8+8)*vol;
+      std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<"   \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
+
+    }
+#endif
+  Grid_finalize();
+}
diff --git a/benchmarks/Benchmark_wilson.cc b/benchmarks/Benchmark_wilson.cc
index 754051f0..49939620 100644
--- a/benchmarks/Benchmark_wilson.cc
+++ b/benchmarks/Benchmark_wilson.cc
@@ -30,7 +30,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 
 #include "Grid/util/Profiling.h"
@@ -61,12 +61,13 @@ int main (int argc, char ** argv)
     perfProfiling = true;
   }
 
-  long unsigned int single_site_flops = 8*QCD::Nc*(7+16*QCD::Nc);
+  long unsigned int single_site_flops = 8*Nc*(7+16*Nc);
 
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  auto latt_size   = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
+
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -77,7 +78,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "Grid floating point word size is REALF"<< sizeof(RealF)<<std::endl;
   std::cout<<GridLogMessage << "Grid floating point word size is REALD"<< sizeof(RealD)<<std::endl;
   std::cout<<GridLogMessage << "Grid floating point word size is REAL"<< sizeof(Real)<<std::endl;
-  std::cout<<GridLogMessage << "Grid number of colours : "<< QCD::Nc <<std::endl;
+  std::cout<<GridLogMessage << "Grid number of colours : "<< Nc <<std::endl;
   std::cout<<GridLogMessage << "Benchmarking Wilson operator in the fundamental representation" << std::endl;
 
 
@@ -87,10 +88,10 @@ int main (int argc, char ** argv)
   //  pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9});
 
   LatticeFermion src   (&Grid); random(pRNG,src);
-  LatticeFermion result(&Grid); result=zero;
-  LatticeFermion    ref(&Grid);    ref=zero;
-  LatticeFermion    tmp(&Grid);    tmp=zero;
-  LatticeFermion    err(&Grid);    tmp=zero;
+  LatticeFermion result(&Grid); result=Zero();
+  LatticeFermion    ref(&Grid);    ref=Zero();
+  LatticeFermion    tmp(&Grid);    tmp=Zero();
+  LatticeFermion    err(&Grid);    tmp=Zero();
   LatticeGaugeField Umu(&Grid); random(pRNG,Umu);
   std::vector<LatticeColourMatrix> U(4,&Grid);
 
@@ -101,12 +102,12 @@ int main (int argc, char ** argv)
 
   // Only one non-zero (y)
 #if 0
-  Umu=zero;
+  Umu=Zero();
   Complex cone(1.0,0.0);
   for(int nn=0;nn<Nd;nn++){
     random(pRNG,U[nn]);
     if(1) {
-      if (nn!=2) { U[nn]=zero; std::cout<<GridLogMessage << "zeroing gauge field in dir "<<nn<<std::endl; }
+      if (nn!=2) { U[nn]=Zero(); std::cout<<GridLogMessage << "zeroing gauge field in dir "<<nn<<std::endl; }
       //      else       { U[nn]= cone;std::cout<<GridLogMessage << "unit gauge field in dir "<<nn<<std::endl; }
       else       { std::cout<<GridLogMessage << "random gauge field in dir "<<nn<<std::endl; }
     }
@@ -119,18 +120,26 @@ int main (int argc, char ** argv)
   }
   
   { // Naive wilson implementation
-    ref = zero;
+    ref = Zero();
     for(int mu=0;mu<Nd;mu++){
       //    ref =  src + Gamma(Gamma::Algebra::GammaX)* src ; // 1-gamma_x
       tmp = U[mu]*Cshift(src,mu,1);
-      for(int i=0;i<ref._odata.size();i++){
-	ref._odata[i]+= tmp._odata[i] - Gamma(Gmu[mu])*tmp._odata[i]; ;
+      {
+	auto ref_v = ref.View();
+	auto tmp_v = tmp.View();
+	for(int i=0;i<ref_v.size();i++){
+	  ref_v[i]+= tmp_v[i] - Gamma(Gmu[mu])*tmp_v[i]; ;
+	}
       }
 
       tmp =adj(U[mu])*src;
       tmp =Cshift(tmp,mu,-1);
-      for(int i=0;i<ref._odata.size();i++){
-	ref._odata[i]+= tmp._odata[i] + Gamma(Gmu[mu])*tmp._odata[i]; ;
+      {
+	auto ref_v = ref.View();
+	auto tmp_v = tmp.View();
+	for(int i=0;i<ref_v.size();i++){
+	  ref_v[i]+= tmp_v[i] + Gamma(Gmu[mu])*tmp_v[i]; ;
+	}
       }
     }
   }
@@ -138,7 +147,6 @@ int main (int argc, char ** argv)
   RealD mass=0.1;
 
   typename WilsonFermionR::ImplParams params; 
-  params.overlapCommsCompute = overlapComms;
 
   WilsonFermionR Dw(Umu,Grid,RBGrid,mass,params);
   
@@ -165,7 +173,7 @@ int main (int argc, char ** argv)
 
   }
 
-
+  
   std::cout<<GridLogMessage << "Called Dw"<<std::endl;
   std::cout<<GridLogMessage << "flops per site " << single_site_flops << std::endl;
   std::cout<<GridLogMessage << "norm result "<< norm2(result)<<std::endl;
@@ -179,27 +187,38 @@ int main (int argc, char ** argv)
   for(int ss=0;ss<0;ss++ ){
     for(int i=0;i<Ns;i++){
       for(int j=0;j<Nc;j++){
-	ComplexF * ref_p = (ComplexF *)&ref._odata[ss]()(i)(j);
-	ComplexF * res_p = (ComplexF *)&result._odata[ss]()(i)(j);
+	auto ref_v = ref.View();
+	auto result_v = result.View();
+	ComplexF * ref_p = (ComplexF *)&ref_v[ss]()(i)(j);
+	ComplexF * res_p = (ComplexF *)&result_v[ss]()(i)(j);
 	std::cout<<GridLogMessage << ss<< " "<<i<<" "<<j<<" "<< (*ref_p)<<" " <<(*res_p)<<std::endl;
       }
     }
   }
 
   { // Naive wilson dag implementation
-    ref = zero;
+    ref = Zero();
     for(int mu=0;mu<Nd;mu++){
 
+
       //    ref =  src - Gamma(Gamma::Algebra::GammaX)* src ; // 1+gamma_x
       tmp = U[mu]*Cshift(src,mu,1);
-      for(int i=0;i<ref._odata.size();i++){
-	ref._odata[i]+= tmp._odata[i] + Gamma(Gmu[mu])*tmp._odata[i]; ;
+      {
+	auto ref_v = ref.View();
+	auto tmp_v = tmp.View();
+	for(int i=0;i<ref_v.size();i++){
+	  ref_v[i]+= tmp_v[i] + Gamma(Gmu[mu])*tmp_v[i]; ;
+	}
       }
 
       tmp =adj(U[mu])*src;
       tmp =Cshift(tmp,mu,-1);
-      for(int i=0;i<ref._odata.size();i++){
-	ref._odata[i]+= tmp._odata[i] - Gamma(Gmu[mu])*tmp._odata[i]; ;
+      {
+	auto ref_v = ref.View();
+	auto tmp_v = tmp.View();
+	for(int i=0;i<ref_v.size();i++){
+	  ref_v[i]+= tmp_v[i] - Gamma(Gmu[mu])*tmp_v[i]; ;
+	}
       }
     }
   }
diff --git a/benchmarks/Benchmark_wilson_sweep.cc b/benchmarks/Benchmark_wilson_sweep.cc
index f80cde28..a01a9e1f 100644
--- a/benchmarks/Benchmark_wilson_sweep.cc
+++ b/benchmarks/Benchmark_wilson_sweep.cc
@@ -23,7 +23,7 @@ Author: Richard Rollins <rprollins@users.noreply.github.com>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -37,8 +37,6 @@ Gamma::Algebra Gmu [] = {
     Gamma::Algebra::GammaT
 };
 
-bool overlapComms = false;
-
 void bench_wilson (
 		   LatticeFermion &    src,
 		   LatticeFermion & result,
@@ -49,12 +47,10 @@ void bench_wilson (
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
-  if( GridCmdOptionExists(argv,argv+argc,"--asynch") ){ overlapComms = true; }
   typename WilsonFermionR::ImplParams params;
-  params.overlapCommsCompute = overlapComms;
 
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   std::vector<int> seeds({1,2,3,4});
   RealD mass = 0.1;
 
@@ -62,7 +58,7 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage<< "* Kernel options --dslash-generic, --dslash-unroll, --dslash-asm" <<std::endl;
   std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
   std::cout << GridLogMessage<< "*****************************************************************" <<std::endl;
-  std::cout << GridLogMessage<< "* Number of colours "<< QCD::Nc <<std::endl;
+  std::cout << GridLogMessage<< "* Number of colours "<< Nc <<std::endl;
   std::cout << GridLogMessage<< "* Benchmarking WilsonFermionR::Dhop                  "<<std::endl;
   std::cout << GridLogMessage<< "* Vectorising space-time by "<<vComplex::Nsimd()<<std::endl;
   if ( sizeof(Real)==4 )   std::cout << GridLogMessage<< "* SINGLE precision "<<std::endl;
@@ -86,33 +82,32 @@ int main (int argc, char ** argv)
   if ( getenv("DMIN") ) dmin=atoi(getenv("DMIN"));
   for (int L=8; L<=Lmax; L*=2)
     {
-      std::vector<int> latt_size = std::vector<int>(4,L);
+      Coordinate latt_size = Coordinate(4,L);
       for(int d=4; d>dmin; d--)
 	{
 	  if ( d<=3 ) { latt_size[d] *= 2; }
 
 	  std::cout << GridLogMessage;
-	  std::copy( latt_size.begin(), --latt_size.end(), std::ostream_iterator<int>( std::cout, std::string("x").c_str() ) );
-	  std::cout << latt_size.back() << "\t\t";
+	  std::cout << latt_size;
 
 	  GridCartesian           Grid(latt_size,simd_layout,mpi_layout);
 	  GridRedBlackCartesian RBGrid(&Grid);
 
 	  GridParallelRNG  pRNG(&Grid); pRNG.SeedFixedIntegers(seeds);
 	  LatticeGaugeField Umu(&Grid); random(pRNG,Umu);
-	  LatticeFermion        src(&Grid); random(pRNG,src);
+	  LatticeFermion    src(&Grid); random(pRNG,src);
 	  LatticeFermion    src_o(&RBGrid); pickCheckerboard(Odd,src_o,src);
-	  LatticeFermion     result(&Grid); result=zero;
-	  LatticeFermion result_e(&RBGrid); result_e=zero;
+	  LatticeFermion     result(&Grid); result=Zero();
+	  LatticeFermion result_e(&RBGrid); result_e=Zero();
 
 	  double volume = std::accumulate(latt_size.begin(),latt_size.end(),1,std::multiplies<int>());
 
 	  WilsonFermionR Dw(Umu,Grid,RBGrid,mass,params);
-
+      
     // Full operator      
 	  bench_wilson(src,result,Dw,volume,DaggerNo);
 	  bench_wilson(src,result,Dw,volume,DaggerYes);
-    std::cout << "\t";
+	  std::cout << "\t";
     // EO
 	  bench_wilson(src,result,Dw,volume,DaggerNo);
 	  bench_wilson(src,result,Dw,volume,DaggerYes);
@@ -132,7 +127,7 @@ void bench_wilson (
 		   int const           dag )
 {
   int ncall    = 1000;
-  long unsigned int single_site_flops = 8*QCD::Nc*(7+16*QCD::Nc);
+  long unsigned int single_site_flops = 8*Nc*(7+16*Nc);
   double t0    = usecond();
   for(int i=0; i<ncall; i++) { Dw.Dhop(src,result,dag); }
   double t1    = usecond();
@@ -148,7 +143,7 @@ void bench_wilson_eo (
 		   int const           dag )
 {
   int ncall    = 1000;
-  long unsigned int single_site_flops = 8*QCD::Nc*(7+16*QCD::Nc);
+  long unsigned int single_site_flops = 8*Nc*(7+16*Nc);
   double t0    = usecond();
   for(int i=0; i<ncall; i++) { Dw.DhopEO(src,result,dag); }
   double t1    = usecond();
diff --git a/benchmarks/simple_su3_expr.cc b/benchmarks/simple_su3_expr.cc
index 0c6092f6..7e53cce0 100644
--- a/benchmarks/simple_su3_expr.cc
+++ b/benchmarks/simple_su3_expr.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 void su3_test_mult_expression(LatticeColourMatrix &z, LatticeColourMatrix &x,LatticeColourMatrix &y)
 {
diff --git a/benchmarks/simple_su3_test.cc b/benchmarks/simple_su3_test.cc
index 79437c23..899b49d0 100644
--- a/benchmarks/simple_su3_test.cc
+++ b/benchmarks/simple_su3_test.cc
@@ -29,8 +29,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
-
 
 void su3_test_mult_routine(LatticeColourMatrix &z, LatticeColourMatrix &x,LatticeColourMatrix &y)
 {
diff --git a/bootstrap.sh b/bootstrap.sh
index e7495096..8ebd35ea 100755
--- a/bootstrap.sh
+++ b/bootstrap.sh
@@ -1,6 +1,6 @@
 #!/usr/bin/env bash
 
-EIGEN_URL='http://bitbucket.org/eigen/eigen/get/3.3.5.tar.bz2'
+EIGEN_URL='http://bitbucket.org/eigen/eigen/get/3.3.7.tar.bz2'
 
 echo "-- deploying Eigen source..."
 ARC=`basename ${EIGEN_URL}`
diff --git a/configure.ac b/configure.ac
index eb816c07..ed7b9a17 100644
--- a/configure.ac
+++ b/configure.ac
@@ -15,7 +15,20 @@ m4_ifdef([AM_SILENT_RULES], [AM_SILENT_RULES([yes])])
 
 ################ Set flags
 # do not move!
-CXXFLAGS="-O3 $CXXFLAGS"
+AC_ARG_ENABLE([debug],[AC_HELP_STRING([--enable-debug=yes|no], [enable debug compilation ])],
+    [ac_DEBUG=${enable_debug}], [ac_DEBUG=no])
+case ${ac_DEBUG} in
+    yes)
+    CXXFLAGS="-g $CXXFLAGS "
+        ;;
+    no)
+    CXXFLAGS="-O3 $CXXFLAGS "
+        ;;
+    *)
+    CXXFLAGS="-O3 $CXXFLAGS "
+        ;;
+esac
+
 
 ############### Checks for programs
 AC_PROG_CXX
@@ -32,7 +45,6 @@ AC_DEFINE_UNQUOTED([GXX_VERSION],["$GXX_VERSION"],
       [version of g++ that will compile the code])
 
 
-
 ############### Checks for typedefs, structures, and compiler characteristics
 AC_TYPE_SIZE_T
 AC_TYPE_UINT32_T
@@ -47,6 +59,7 @@ if test "${ac_cv_prog_cxx_openmp}X" != "noX"; then
   AM_LDFLAGS="$OPENMP_CXXFLAGS $AM_LDFLAGS"
 fi
 
+
 ############### Checks for header files
 AC_CHECK_HEADERS(stdint.h)
 AC_CHECK_HEADERS(mm_malloc.h)
@@ -252,9 +265,44 @@ AC_ARG_ENABLE([gen-simd-width],
             [ac_gen_simd_width=$enable_gen_simd_width],
             [ac_gen_simd_width=32])
 
+AC_ARG_ENABLE([gen-scalar],
+            [AS_HELP_STRING([--enable-gen-scalar=yes|no],
+            [enable generic scalar implementation])],
+            [ac_gen_scalar=$enable_gen_scalar],
+            [ac_gen_scalar=no])
+
+case ${ac_gen_scalar} in
+  yes)
+    AC_DEFINE([GENERIC_SCALAR],[1],[Use scalar data parallel loops])
+  ;;
+  *)
+  ;;
+esac
+
+##################### Compiler dependent choices
+case ${CXX} in 
+  nvcc) 
+#    CXX="nvcc -keep -v -x cu "
+#    CXXLD="nvcc -v -link"
+    CXX="nvcc -x cu "
+    CXXLD="nvcc -link"
+    CXXFLAGS="$CXXFLAGS -Xcompiler -fno-strict-aliasing -Xcompiler -Wno-unusable-partial-specialization --expt-extended-lambda --expt-relaxed-constexpr"
+    ;;
+  *)
+    CXXLD=${CXX}
+    CXXFLAGS="$CXXFLAGS -fno-strict-aliasing"
+    ;;
+esac
+
 case ${ax_cv_cxx_compiler_vendor} in
   clang|gnu)
     case ${ac_SIMD} in
+      GPU)
+        AC_DEFINE([GPU_VEC],[1],[GPU vectorised 512bit])
+        AC_DEFINE_UNQUOTED([GEN_SIMD_WIDTH],[$ac_gen_simd_width],
+                           [generic SIMD vector width (in bytes)])
+        SIMD_GEN_WIDTH_MSG=" (width= $ac_gen_simd_width)"
+	SIMD_FLAGS='';;
       SSE4)
         AC_DEFINE([SSE4],[1],[SSE4 intrinsics])
 	case ${ac_SFW_FP16} in
@@ -307,6 +355,9 @@ case ${ax_cv_cxx_compiler_vendor} in
     esac;;
   intel)
     case ${ac_SIMD} in
+      GPU)
+        AC_DEFINE([GPU_VEC],[1],[GPU vectorised ])
+	SIMD_FLAGS='';;
       SSE4)
         AC_DEFINE([SSE4],[1],[SSE4 intrinsics])
         SIMD_FLAGS='-msse4.2 -xsse4.2';;
@@ -345,13 +396,6 @@ esac
 AM_CXXFLAGS="$SIMD_FLAGS $AM_CXXFLAGS"
 AM_CFLAGS="$SIMD_FLAGS $AM_CFLAGS"
 
-case ${ac_SIMD} in
-  AVX512|KNL)
-    AC_DEFINE([TEST_ZMM],[1],[compile ZMM test]);;
-  *)
-	;;
-esac
-
 ############### Precision selection
 AC_ARG_ENABLE([precision],
               [AC_HELP_STRING([--enable-precision=single|double],
@@ -424,7 +468,9 @@ esac
 case ${ac_COMMS} in
     *-auto)
         LX_FIND_MPI
-        if test "x$have_CXX_mpi" = 'xno'; then AC_MSG_ERROR(["The configure could not find the MPI compilation flags. N.B. The -auto mode is not supported by Cray wrappers. Use the non -auto version in this case."]); fi
+	echo MPI_CXXFLAGS $MPI_CXXFLAGS
+	echo MPI_CXXLDFLAGS $MPI_CXXLDFLAGS
+	echo MPI_CFLAGS $MPI_CFLAGS
         AM_CXXFLAGS="$MPI_CXXFLAGS $AM_CXXFLAGS"
         AM_CFLAGS="$MPI_CFLAGS $AM_CFLAGS"
         AM_LDFLAGS="`echo $MPI_CXXLDFLAGS | sed -E 's/-l@<:@^ @:>@+//g'` $AM_LDFLAGS"
@@ -433,10 +479,7 @@ case ${ac_COMMS} in
         ;;
 esac
 
-AM_CONDITIONAL(BUILD_COMMS_SHMEM, [ test "${comms_type}X" == "shmemX" ])
-AM_CONDITIONAL(BUILD_COMMS_MPI,   [ test "${comms_type}X" == "mpiX" ])
 AM_CONDITIONAL(BUILD_COMMS_MPI3,  [ test "${comms_type}X" == "mpi3X" ] )
-AM_CONDITIONAL(BUILD_COMMS_MPIT,  [ test "${comms_type}X" == "mpitX" ] )
 AM_CONDITIONAL(BUILD_COMMS_NONE,  [ test "${comms_type}X" == "noneX" ])
 
 ############### RNG selection
@@ -515,6 +558,7 @@ GRID_BRANCH=`git rev-parse --abbrev-ref HEAD`
 AM_CXXFLAGS="-I${abs_srcdir} $AM_CXXFLAGS"
 AM_CFLAGS="-I${abs_srcdir} $AM_CFLAGS"
 AM_LDFLAGS="-L${cwd}/Grid $AM_LDFLAGS"
+AC_SUBST([CXXLD])
 AC_SUBST([AM_CFLAGS])
 AC_SUBST([AM_CXXFLAGS])
 AC_SUBST([AM_LDFLAGS])
diff --git a/include/Grid b/include/Grid
deleted file mode 120000
index dc598c56..00000000
--- a/include/Grid
+++ /dev/null
@@ -1 +0,0 @@
-../lib
\ No newline at end of file
diff --git a/scripts/filelist b/scripts/filelist
index fafa733b..78747315 100755
--- a/scripts/filelist
+++ b/scripts/filelist
@@ -6,7 +6,7 @@ home=`pwd`
 cd $home/Grid
 HFILES=`find . -type f -name '*.h' -not -name '*Hdf5*' -not -path '*/gamma-gen/*' -not -path '*/Old/*' -not -path '*/Eigen/*'`
 HFILES="$HFILES"
-CCFILES=`find . -type f -name '*.cc' -not -path '*/gamma-gen/*' -not -name '*Communicator*.cc' -not -name '*SharedMemory*.cc' -not -name '*Hdf5*'`
+CCFILES=`find . -name '*.cc' -not -path '*/gamma-gen/*' -not -name '*Communicator*.cc' -not -name '*SharedMemory*.cc' -not -name '*Hdf5*'`
 HPPFILES=`find . -type f -name '*.hpp'`
 echo HFILES=$HFILES $HPPFILES > Make.inc
 echo >> Make.inc
diff --git a/scripts/format b/scripts/format
new file mode 100755
index 00000000..df1b1f91
--- /dev/null
+++ b/scripts/format
@@ -0,0 +1,6 @@
+#!/bin/bash
+for f in $@ 
+do
+    cp $f $f.bak
+    emacs -batch -l $HOME/.emacs $f --eval "(progn (c-set-offset 'innamespace 0) ( c++-mode ) ( indent-region (point-min) (point-max) nil ) (save-buffer) ) " 
+done
\ No newline at end of file
diff --git a/scripts/loop.log b/scripts/loop.log
deleted file mode 100644
index 21b00a0d..00000000
--- a/scripts/loop.log
+++ /dev/null
@@ -1,15432 +0,0 @@
-Grid       : Message        :  Stencil 12.6786 GB/s per node
-Grid       : Message        :  Stencil 17.2339 GB/s per node
-Grid       : Message        :  Stencil 17.5982 GB/s per node
-Grid       : Message        :  Stencil 12.79 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671730
-Grid       : Message        : Average mflops/s per call per node       : 802660
-Grid       : Message        : Average mflops/s per call per node       : 827085
-Grid       : Message        : Average mflops/s per call per node       : 665055
-Grid       : Message        : Average mflops/s per call per node (full): 314188
-Grid       : Message        : Average mflops/s per call per node (full): 437622
-Grid       : Message        : Average mflops/s per call per node (full): 446214
-Grid       : Message        : Average mflops/s per call per node (full): 304778
-Grid       : Message        :  Stencil 12.3348 GB/s per node
-Grid       : Message        :  Stencil 16.7072 GB/s per node
-Grid       : Message        :  Stencil 17.5215 GB/s per node
-Grid       : Message        :  Stencil 13.0792 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 673845
-Grid       : Message        : Average mflops/s per call per node       : 803432
-Grid       : Message        : Average mflops/s per call per node       : 827130
-Grid       : Message        : Average mflops/s per call per node       : 661863
-Grid       : Message        : Average mflops/s per call per node (full): 312916
-Grid       : Message        : Average mflops/s per call per node (full): 437409
-Grid       : Message        : Average mflops/s per call per node (full): 446117
-Grid       : Message        : Average mflops/s per call per node (full): 305031
-Grid       : Message        :  Stencil 13.251 GB/s per node
-Grid       : Message        :  Stencil 17.0389 GB/s per node
-Grid       : Message        :  Stencil 17.9444 GB/s per node
-Grid       : Message        :  Stencil 12.8909 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670180
-Grid       : Message        : Average mflops/s per call per node       : 803789
-Grid       : Message        : Average mflops/s per call per node       : 820549
-Grid       : Message        : Average mflops/s per call per node       : 665372
-Grid       : Message        : Average mflops/s per call per node (full): 316277
-Grid       : Message        : Average mflops/s per call per node (full): 435976
-Grid       : Message        : Average mflops/s per call per node (full): 446442
-Grid       : Message        : Average mflops/s per call per node (full): 305106
-Grid       : Message        :  Stencil 13.6097 GB/s per node
-Grid       : Message        :  Stencil 17.7981 GB/s per node
-Grid       : Message        :  Stencil 17.5185 GB/s per node
-Grid       : Message        :  Stencil 14.4014 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666791
-Grid       : Message        : Average mflops/s per call per node       : 799898
-Grid       : Message        : Average mflops/s per call per node       : 826498
-Grid       : Message        : Average mflops/s per call per node       : 663479
-Grid       : Message        : Average mflops/s per call per node (full): 316681
-Grid       : Message        : Average mflops/s per call per node (full): 437896
-Grid       : Message        : Average mflops/s per call per node (full): 445597
-Grid       : Message        : Average mflops/s per call per node (full): 307770
-Grid       : Message        :  Stencil 14.2804 GB/s per node
-Grid       : Message        :  Stencil 16.6792 GB/s per node
-Grid       : Message        :  Stencil 18.2657 GB/s per node
-Grid       : Message        :  Stencil 12.4275 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661518
-Grid       : Message        : Average mflops/s per call per node       : 804204
-Grid       : Message        : Average mflops/s per call per node       : 821252
-Grid       : Message        : Average mflops/s per call per node       : 669218
-Grid       : Message        : Average mflops/s per call per node (full): 315153
-Grid       : Message        : Average mflops/s per call per node (full): 436294
-Grid       : Message        : Average mflops/s per call per node (full): 448032
-Grid       : Message        : Average mflops/s per call per node (full): 305047
-Grid       : Message        :  Stencil 13.1218 GB/s per node
-Grid       : Message        :  Stencil 16.6241 GB/s per node
-Grid       : Message        :  Stencil 17.8377 GB/s per node
-Grid       : Message        :  Stencil 13.1379 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667798
-Grid       : Message        : Average mflops/s per call per node       : 804417
-Grid       : Message        : Average mflops/s per call per node       : 816011
-Grid       : Message        : Average mflops/s per call per node       : 658982
-Grid       : Message        : Average mflops/s per call per node (full): 314849
-Grid       : Message        : Average mflops/s per call per node (full): 436163
-Grid       : Message        : Average mflops/s per call per node (full): 438805
-Grid       : Message        : Average mflops/s per call per node (full): 304163
-Grid       : Message        :  Stencil 13.5449 GB/s per node
-Grid       : Message        :  Stencil 17.0727 GB/s per node
-Grid       : Message        :  Stencil 18.1857 GB/s per node
-Grid       : Message        :  Stencil 13.3749 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667016
-Grid       : Message        : Average mflops/s per call per node       : 805999
-Grid       : Message        : Average mflops/s per call per node       : 825220
-Grid       : Message        : Average mflops/s per call per node       : 664181
-Grid       : Message        : Average mflops/s per call per node (full): 315059
-Grid       : Message        : Average mflops/s per call per node (full): 438031
-Grid       : Message        : Average mflops/s per call per node (full): 447100
-Grid       : Message        : Average mflops/s per call per node (full): 306033
-Grid       : Message        :  Stencil 12.7492 GB/s per node
-Grid       : Message        :  Stencil 16.8398 GB/s per node
-Grid       : Message        :  Stencil 18.5169 GB/s per node
-Grid       : Message        :  Stencil 14.1175 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667433
-Grid       : Message        : Average mflops/s per call per node       : 805205
-Grid       : Message        : Average mflops/s per call per node       : 818690
-Grid       : Message        : Average mflops/s per call per node       : 662018
-Grid       : Message        : Average mflops/s per call per node (full): 312571
-Grid       : Message        : Average mflops/s per call per node (full): 436413
-Grid       : Message        : Average mflops/s per call per node (full): 447176
-Grid       : Message        : Average mflops/s per call per node (full): 307166
-Grid       : Message        :  Stencil 14.073 GB/s per node
-Grid       : Message        :  Stencil 14.0583 GB/s per node
-Grid       : Message        :  Stencil 17.3888 GB/s per node
-Grid       : Message        :  Stencil 12.4105 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662617
-Grid       : Message        : Average mflops/s per call per node       : 809873
-Grid       : Message        : Average mflops/s per call per node       : 823381
-Grid       : Message        : Average mflops/s per call per node       : 664561
-Grid       : Message        : Average mflops/s per call per node (full): 315727
-Grid       : Message        : Average mflops/s per call per node (full): 400767
-Grid       : Message        : Average mflops/s per call per node (full): 444114
-Grid       : Message        : Average mflops/s per call per node (full): 304126
-Grid       : Message        :  Stencil 14.5548 GB/s per node
-Grid       : Message        :  Stencil 16.9469 GB/s per node
-Grid       : Message        :  Stencil 17.4579 GB/s per node
-Grid       : Message        :  Stencil 14.3184 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662519
-Grid       : Message        : Average mflops/s per call per node       : 799987
-Grid       : Message        : Average mflops/s per call per node       : 822293
-Grid       : Message        : Average mflops/s per call per node       : 662139
-Grid       : Message        : Average mflops/s per call per node (full): 316617
-Grid       : Message        : Average mflops/s per call per node (full): 433846
-Grid       : Message        : Average mflops/s per call per node (full): 444509
-Grid       : Message        : Average mflops/s per call per node (full): 305318
-Grid       : Message        :  Stencil 13.7009 GB/s per node
-Grid       : Message        :  Stencil 16.6146 GB/s per node
-Grid       : Message        :  Stencil 17.494 GB/s per node
-Grid       : Message        :  Stencil 12.1012 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664936
-Grid       : Message        : Average mflops/s per call per node       : 808221
-Grid       : Message        : Average mflops/s per call per node       : 822622
-Grid       : Message        : Average mflops/s per call per node       : 670400
-Grid       : Message        : Average mflops/s per call per node (full): 316165
-Grid       : Message        : Average mflops/s per call per node (full): 436412
-Grid       : Message        : Average mflops/s per call per node (full): 444939
-Grid       : Message        : Average mflops/s per call per node (full): 303228
-Grid       : Message        :  Stencil 12.6849 GB/s per node
-Grid       : Message        :  Stencil 16.6473 GB/s per node
-Grid       : Message        :  Stencil 17.4004 GB/s per node
-Grid       : Message        :  Stencil 13.1606 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671546
-Grid       : Message        : Average mflops/s per call per node       : 807297
-Grid       : Message        : Average mflops/s per call per node       : 821422
-Grid       : Message        : Average mflops/s per call per node       : 667455
-Grid       : Message        : Average mflops/s per call per node (full): 314965
-Grid       : Message        : Average mflops/s per call per node (full): 436228
-Grid       : Message        : Average mflops/s per call per node (full): 444879
-Grid       : Message        : Average mflops/s per call per node (full): 305982
-Grid       : Message        :  Stencil 13.809 GB/s per node
-Grid       : Message        :  Stencil 16.3086 GB/s per node
-Grid       : Message        :  Stencil 17.3784 GB/s per node
-Grid       : Message        :  Stencil 12.5934 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667096
-Grid       : Message        : Average mflops/s per call per node       : 803680
-Grid       : Message        : Average mflops/s per call per node       : 823904
-Grid       : Message        : Average mflops/s per call per node       : 667368
-Grid       : Message        : Average mflops/s per call per node (full): 317077
-Grid       : Message        : Average mflops/s per call per node (full): 432108
-Grid       : Message        : Average mflops/s per call per node (full): 445114
-Grid       : Message        : Average mflops/s per call per node (full): 304958
-Grid       : Message        :  Stencil 14.3258 GB/s per node
-Grid       : Message        :  Stencil 16.9233 GB/s per node
-Grid       : Message        :  Stencil 17.6914 GB/s per node
-Grid       : Message        :  Stencil 12.1857 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665305
-Grid       : Message        : Average mflops/s per call per node       : 804694
-Grid       : Message        : Average mflops/s per call per node       : 824508
-Grid       : Message        : Average mflops/s per call per node       : 667104
-Grid       : Message        : Average mflops/s per call per node (full): 316651
-Grid       : Message        : Average mflops/s per call per node (full): 428082
-Grid       : Message        : Average mflops/s per call per node (full): 446379
-Grid       : Message        : Average mflops/s per call per node (full): 303779
-Grid       : Message        :  Stencil 13.2527 GB/s per node
-Grid       : Message        :  Stencil 16.6723 GB/s per node
-Grid       : Message        :  Stencil 17.581 GB/s per node
-Grid       : Message        :  Stencil 13.4152 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669149
-Grid       : Message        : Average mflops/s per call per node       : 799407
-Grid       : Message        : Average mflops/s per call per node       : 820946
-Grid       : Message        : Average mflops/s per call per node       : 658257
-Grid       : Message        : Average mflops/s per call per node (full): 315691
-Grid       : Message        : Average mflops/s per call per node (full): 435436
-Grid       : Message        : Average mflops/s per call per node (full): 445192
-Grid       : Message        : Average mflops/s per call per node (full): 305074
-Grid       : Message        :  Stencil 13.0576 GB/s per node
-Grid       : Message        :  Stencil 16.4259 GB/s per node
-Grid       : Message        :  Stencil 17.3695 GB/s per node
-Grid       : Message        :  Stencil 13.224 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671756
-Grid       : Message        : Average mflops/s per call per node       : 805129
-Grid       : Message        : Average mflops/s per call per node       : 826323
-Grid       : Message        : Average mflops/s per call per node       : 663117
-Grid       : Message        : Average mflops/s per call per node (full): 316939
-Grid       : Message        : Average mflops/s per call per node (full): 434386
-Grid       : Message        : Average mflops/s per call per node (full): 445016
-Grid       : Message        : Average mflops/s per call per node (full): 306607
-Grid       : Message        :  Stencil 12.735 GB/s per node
-Grid       : Message        :  Stencil 18.7133 GB/s per node
-Grid       : Message        :  Stencil 17.5949 GB/s per node
-Grid       : Message        :  Stencil 13.6578 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 673814
-Grid       : Message        : Average mflops/s per call per node       : 803124
-Grid       : Message        : Average mflops/s per call per node       : 822958
-Grid       : Message        : Average mflops/s per call per node       : 662935
-Grid       : Message        : Average mflops/s per call per node (full): 315459
-Grid       : Message        : Average mflops/s per call per node (full): 440286
-Grid       : Message        : Average mflops/s per call per node (full): 446144
-Grid       : Message        : Average mflops/s per call per node (full): 306944
-Grid       : Message        :  Stencil 14.5299 GB/s per node
-Grid       : Message        :  Stencil 16.827 GB/s per node
-Grid       : Message        :  Stencil 18.227 GB/s per node
-Grid       : Message        :  Stencil 12.5667 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664966
-Grid       : Message        : Average mflops/s per call per node       : 805649
-Grid       : Message        : Average mflops/s per call per node       : 820323
-Grid       : Message        : Average mflops/s per call per node       : 662468
-Grid       : Message        : Average mflops/s per call per node (full): 316234
-Grid       : Message        : Average mflops/s per call per node (full): 436979
-Grid       : Message        : Average mflops/s per call per node (full): 445870
-Grid       : Message        : Average mflops/s per call per node (full): 304734
-Grid       : Message        :  Stencil 14.0947 GB/s per node
-Grid       : Message        :  Stencil 17.2047 GB/s per node
-Grid       : Message        :  Stencil 16.3778 GB/s per node
-Grid       : Message        :  Stencil 11.9717 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667253
-Grid       : Message        : Average mflops/s per call per node       : 801164
-Grid       : Message        : Average mflops/s per call per node       : 823547
-Grid       : Message        : Average mflops/s per call per node       : 671607
-Grid       : Message        : Average mflops/s per call per node (full): 317280
-Grid       : Message        : Average mflops/s per call per node (full): 438057
-Grid       : Message        : Average mflops/s per call per node (full): 429727
-Grid       : Message        : Average mflops/s per call per node (full): 302300
-Grid       : Message        :  Stencil 14.4044 GB/s per node
-Grid       : Message        :  Stencil 16.4973 GB/s per node
-Grid       : Message        :  Stencil 16.8518 GB/s per node
-Grid       : Message        :  Stencil 13.1009 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664077
-Grid       : Message        : Average mflops/s per call per node       : 802258
-Grid       : Message        : Average mflops/s per call per node       : 823349
-Grid       : Message        : Average mflops/s per call per node       : 667635
-Grid       : Message        : Average mflops/s per call per node (full): 316900
-Grid       : Message        : Average mflops/s per call per node (full): 433661
-Grid       : Message        : Average mflops/s per call per node (full): 436108
-Grid       : Message        : Average mflops/s per call per node (full): 305754
-Grid       : Message        :  Stencil 13.228 GB/s per node
-Grid       : Message        :  Stencil 16.8887 GB/s per node
-Grid       : Message        :  Stencil 17.6471 GB/s per node
-Grid       : Message        :  Stencil 12.7062 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669067
-Grid       : Message        : Average mflops/s per call per node       : 804099
-Grid       : Message        : Average mflops/s per call per node       : 821604
-Grid       : Message        : Average mflops/s per call per node       : 658039
-Grid       : Message        : Average mflops/s per call per node (full): 316770
-Grid       : Message        : Average mflops/s per call per node (full): 437420
-Grid       : Message        : Average mflops/s per call per node (full): 445975
-Grid       : Message        : Average mflops/s per call per node (full): 298715
-Grid       : Message        :  Stencil 13.5671 GB/s per node
-Grid       : Message        :  Stencil 16.5942 GB/s per node
-Grid       : Message        :  Stencil 18.1271 GB/s per node
-Grid       : Message        :  Stencil 11.9482 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667462
-Grid       : Message        : Average mflops/s per call per node       : 807110
-Grid       : Message        : Average mflops/s per call per node       : 823079
-Grid       : Message        : Average mflops/s per call per node       : 667472
-Grid       : Message        : Average mflops/s per call per node (full): 314605
-Grid       : Message        : Average mflops/s per call per node (full): 436264
-Grid       : Message        : Average mflops/s per call per node (full): 446541
-Grid       : Message        : Average mflops/s per call per node (full): 301224
-Grid       : Message        :  Stencil 12.3664 GB/s per node
-Grid       : Message        :  Stencil 13.5017 GB/s per node
-Grid       : Message        :  Stencil 17.7799 GB/s per node
-Grid       : Message        :  Stencil 12.3154 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 672086
-Grid       : Message        : Average mflops/s per call per node       : 808002
-Grid       : Message        : Average mflops/s per call per node       : 822256
-Grid       : Message        : Average mflops/s per call per node       : 666533
-Grid       : Message        : Average mflops/s per call per node (full): 313568
-Grid       : Message        : Average mflops/s per call per node (full): 391027
-Grid       : Message        : Average mflops/s per call per node (full): 441218
-Grid       : Message        : Average mflops/s per call per node (full): 303786
-Grid       : Message        :  Stencil 13.5781 GB/s per node
-Grid       : Message        :  Stencil 16.4869 GB/s per node
-Grid       : Message        :  Stencil 17.8246 GB/s per node
-Grid       : Message        :  Stencil 13.8167 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667731
-Grid       : Message        : Average mflops/s per call per node       : 803031
-Grid       : Message        : Average mflops/s per call per node       : 826184
-Grid       : Message        : Average mflops/s per call per node       : 662380
-Grid       : Message        : Average mflops/s per call per node (full): 316580
-Grid       : Message        : Average mflops/s per call per node (full): 434474
-Grid       : Message        : Average mflops/s per call per node (full): 447463
-Grid       : Message        : Average mflops/s per call per node (full): 305866
-Grid       : Message        :  Stencil 14.1182 GB/s per node
-Grid       : Message        :  Stencil 16.975 GB/s per node
-Grid       : Message        :  Stencil 18.8744 GB/s per node
-Grid       : Message        :  Stencil 12.1288 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661970
-Grid       : Message        : Average mflops/s per call per node       : 803461
-Grid       : Message        : Average mflops/s per call per node       : 820977
-Grid       : Message        : Average mflops/s per call per node       : 668043
-Grid       : Message        : Average mflops/s per call per node (full): 316254
-Grid       : Message        : Average mflops/s per call per node (full): 437180
-Grid       : Message        : Average mflops/s per call per node (full): 448645
-Grid       : Message        : Average mflops/s per call per node (full): 302267
-Grid       : Message        :  Stencil 13.4813 GB/s per node
-Grid       : Message        :  Stencil 16.4433 GB/s per node
-Grid       : Message        :  Stencil 17.8945 GB/s per node
-Grid       : Message        :  Stencil 11.9791 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667678
-Grid       : Message        : Average mflops/s per call per node       : 802481
-Grid       : Message        : Average mflops/s per call per node       : 823512
-Grid       : Message        : Average mflops/s per call per node       : 662641
-Grid       : Message        : Average mflops/s per call per node (full): 316494
-Grid       : Message        : Average mflops/s per call per node (full): 433789
-Grid       : Message        : Average mflops/s per call per node (full): 447496
-Grid       : Message        : Average mflops/s per call per node (full): 300465
-Grid       : Message        :  Stencil 13.123 GB/s per node
-Grid       : Message        :  Stencil 11.7805 GB/s per node
-Grid       : Message        :  Stencil 17.9281 GB/s per node
-Grid       : Message        :  Stencil 12.2311 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665079
-Grid       : Message        : Average mflops/s per call per node       : 806976
-Grid       : Message        : Average mflops/s per call per node       : 824353
-Grid       : Message        : Average mflops/s per call per node       : 669184
-Grid       : Message        : Average mflops/s per call per node (full): 315687
-Grid       : Message        : Average mflops/s per call per node (full): 356593
-Grid       : Message        : Average mflops/s per call per node (full): 441731
-Grid       : Message        : Average mflops/s per call per node (full): 304029
-Grid       : Message        :  Stencil 14.1549 GB/s per node
-Grid       : Message        :  Stencil 18.8328 GB/s per node
-Grid       : Message        :  Stencil 16.9621 GB/s per node
-Grid       : Message        :  Stencil 12.2749 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664087
-Grid       : Message        : Average mflops/s per call per node       : 798788
-Grid       : Message        : Average mflops/s per call per node       : 817414
-Grid       : Message        : Average mflops/s per call per node       : 667448
-Grid       : Message        : Average mflops/s per call per node (full): 316819
-Grid       : Message        : Average mflops/s per call per node (full): 440185
-Grid       : Message        : Average mflops/s per call per node (full): 439012
-Grid       : Message        : Average mflops/s per call per node (full): 304156
-Grid       : Message        :  Stencil 13.5542 GB/s per node
-Grid       : Message        :  Stencil 18.0781 GB/s per node
-Grid       : Message        :  Stencil 18.4257 GB/s per node
-Grid       : Message        :  Stencil 12.1906 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665956
-Grid       : Message        : Average mflops/s per call per node       : 801445
-Grid       : Message        : Average mflops/s per call per node       : 816661
-Grid       : Message        : Average mflops/s per call per node       : 663424
-Grid       : Message        : Average mflops/s per call per node (full): 316319
-Grid       : Message        : Average mflops/s per call per node (full): 438901
-Grid       : Message        : Average mflops/s per call per node (full): 445947
-Grid       : Message        : Average mflops/s per call per node (full): 301634
-Grid       : Message        :  Stencil 12.6518 GB/s per node
-Grid       : Message        :  Stencil 17.1647 GB/s per node
-Grid       : Message        :  Stencil 17.1054 GB/s per node
-Grid       : Message        :  Stencil 14.0803 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669876
-Grid       : Message        : Average mflops/s per call per node       : 800051
-Grid       : Message        : Average mflops/s per call per node       : 817636
-Grid       : Message        : Average mflops/s per call per node       : 661101
-Grid       : Message        : Average mflops/s per call per node (full): 314390
-Grid       : Message        : Average mflops/s per call per node (full): 436913
-Grid       : Message        : Average mflops/s per call per node (full): 439852
-Grid       : Message        : Average mflops/s per call per node (full): 305943
-Grid       : Message        :  Stencil 12.5875 GB/s per node
-Grid       : Message        :  Stencil 16.6244 GB/s per node
-Grid       : Message        :  Stencil 17.697 GB/s per node
-Grid       : Message        :  Stencil 13.9751 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668687
-Grid       : Message        : Average mflops/s per call per node       : 804376
-Grid       : Message        : Average mflops/s per call per node       : 820248
-Grid       : Message        : Average mflops/s per call per node       : 657872
-Grid       : Message        : Average mflops/s per call per node (full): 312425
-Grid       : Message        : Average mflops/s per call per node (full): 432575
-Grid       : Message        : Average mflops/s per call per node (full): 445039
-Grid       : Message        : Average mflops/s per call per node (full): 306199
-Grid       : Message        :  Stencil 12.9612 GB/s per node
-Grid       : Message        :  Stencil 16.9251 GB/s per node
-Grid       : Message        :  Stencil 18.2427 GB/s per node
-Grid       : Message        :  Stencil 13.9809 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670116
-Grid       : Message        : Average mflops/s per call per node       : 802462
-Grid       : Message        : Average mflops/s per call per node       : 815357
-Grid       : Message        : Average mflops/s per call per node       : 663631
-Grid       : Message        : Average mflops/s per call per node (full): 315610
-Grid       : Message        : Average mflops/s per call per node (full): 437106
-Grid       : Message        : Average mflops/s per call per node (full): 447389
-Grid       : Message        : Average mflops/s per call per node (full): 307173
-Grid       : Message        :  Stencil 14.9913 GB/s per node
-Grid       : Message        :  Stencil 16.6469 GB/s per node
-Grid       : Message        :  Stencil 19.1252 GB/s per node
-Grid       : Message        :  Stencil 12.2903 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662529
-Grid       : Message        : Average mflops/s per call per node       : 808307
-Grid       : Message        : Average mflops/s per call per node       : 819406
-Grid       : Message        : Average mflops/s per call per node       : 671267
-Grid       : Message        : Average mflops/s per call per node (full): 316715
-Grid       : Message        : Average mflops/s per call per node (full): 436353
-Grid       : Message        : Average mflops/s per call per node (full): 448492
-Grid       : Message        : Average mflops/s per call per node (full): 304655
-Grid       : Message        :  Stencil 13.9691 GB/s per node
-Grid       : Message        :  Stencil 17.0655 GB/s per node
-Grid       : Message        :  Stencil 17.2496 GB/s per node
-Grid       : Message        :  Stencil 12.9983 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665233
-Grid       : Message        : Average mflops/s per call per node       : 802284
-Grid       : Message        : Average mflops/s per call per node       : 826984
-Grid       : Message        : Average mflops/s per call per node       : 668970
-Grid       : Message        : Average mflops/s per call per node (full): 315863
-Grid       : Message        : Average mflops/s per call per node (full): 437702
-Grid       : Message        : Average mflops/s per call per node (full): 442069
-Grid       : Message        : Average mflops/s per call per node (full): 306745
-Grid       : Message        :  Stencil 15.661 GB/s per node
-Grid       : Message        :  Stencil 16.8648 GB/s per node
-Grid       : Message        :  Stencil 17.2765 GB/s per node
-Grid       : Message        :  Stencil 12.083 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659457
-Grid       : Message        : Average mflops/s per call per node       : 806816
-Grid       : Message        : Average mflops/s per call per node       : 824213
-Grid       : Message        : Average mflops/s per call per node       : 667384
-Grid       : Message        : Average mflops/s per call per node (full): 316720
-Grid       : Message        : Average mflops/s per call per node (full): 437656
-Grid       : Message        : Average mflops/s per call per node (full): 444228
-Grid       : Message        : Average mflops/s per call per node (full): 302337
-Grid       : Message        :  Stencil 13.4679 GB/s per node
-Grid       : Message        :  Stencil 16.4826 GB/s per node
-Grid       : Message        :  Stencil 17.4329 GB/s per node
-Grid       : Message        :  Stencil 12.1631 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663334
-Grid       : Message        : Average mflops/s per call per node       : 801140
-Grid       : Message        : Average mflops/s per call per node       : 827083
-Grid       : Message        : Average mflops/s per call per node       : 668463
-Grid       : Message        : Average mflops/s per call per node (full): 313267
-Grid       : Message        : Average mflops/s per call per node (full): 432812
-Grid       : Message        : Average mflops/s per call per node (full): 443554
-Grid       : Message        : Average mflops/s per call per node (full): 303006
-Grid       : Message        :  Stencil 14.1353 GB/s per node
-Grid       : Message        :  Stencil 16.8278 GB/s per node
-Grid       : Message        :  Stencil 17.6039 GB/s per node
-Grid       : Message        :  Stencil 12.1976 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664989
-Grid       : Message        : Average mflops/s per call per node       : 800166
-Grid       : Message        : Average mflops/s per call per node       : 826481
-Grid       : Message        : Average mflops/s per call per node       : 666778
-Grid       : Message        : Average mflops/s per call per node (full): 316360
-Grid       : Message        : Average mflops/s per call per node (full): 436968
-Grid       : Message        : Average mflops/s per call per node (full): 446051
-Grid       : Message        : Average mflops/s per call per node (full): 303706
-Grid       : Message        :  Stencil 14.9322 GB/s per node
-Grid       : Message        :  Stencil 17.0553 GB/s per node
-Grid       : Message        :  Stencil 17.8388 GB/s per node
-Grid       : Message        :  Stencil 11.9921 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663931
-Grid       : Message        : Average mflops/s per call per node       : 797226
-Grid       : Message        : Average mflops/s per call per node       : 817099
-Grid       : Message        : Average mflops/s per call per node       : 664999
-Grid       : Message        : Average mflops/s per call per node (full): 317371
-Grid       : Message        : Average mflops/s per call per node (full): 437367
-Grid       : Message        : Average mflops/s per call per node (full): 442754
-Grid       : Message        : Average mflops/s per call per node (full): 301680
-Grid       : Message        :  Stencil 13.8221 GB/s per node
-Grid       : Message        :  Stencil 16.6444 GB/s per node
-Grid       : Message        :  Stencil 18.01 GB/s per node
-Grid       : Message        :  Stencil 12.6676 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664187
-Grid       : Message        : Average mflops/s per call per node       : 805007
-Grid       : Message        : Average mflops/s per call per node       : 816560
-Grid       : Message        : Average mflops/s per call per node       : 662956
-Grid       : Message        : Average mflops/s per call per node (full): 313418
-Grid       : Message        : Average mflops/s per call per node (full): 436533
-Grid       : Message        : Average mflops/s per call per node (full): 441327
-Grid       : Message        : Average mflops/s per call per node (full): 304681
-Grid       : Message        :  Stencil 12.9318 GB/s per node
-Grid       : Message        :  Stencil 16.4364 GB/s per node
-Grid       : Message        :  Stencil 17.768 GB/s per node
-Grid       : Message        :  Stencil 13.492 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669173
-Grid       : Message        : Average mflops/s per call per node       : 806886
-Grid       : Message        : Average mflops/s per call per node       : 829390
-Grid       : Message        : Average mflops/s per call per node       : 663708
-Grid       : Message        : Average mflops/s per call per node (full): 314081
-Grid       : Message        : Average mflops/s per call per node (full): 434090
-Grid       : Message        : Average mflops/s per call per node (full): 443265
-Grid       : Message        : Average mflops/s per call per node (full): 303994
-Grid       : Message        :  Stencil 12.0068 GB/s per node
-Grid       : Message        :  Stencil 18.4427 GB/s per node
-Grid       : Message        :  Stencil 17.3472 GB/s per node
-Grid       : Message        :  Stencil 12.1353 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669657
-Grid       : Message        : Average mflops/s per call per node       : 806409
-Grid       : Message        : Average mflops/s per call per node       : 821373
-Grid       : Message        : Average mflops/s per call per node       : 664660
-Grid       : Message        : Average mflops/s per call per node (full): 308790
-Grid       : Message        : Average mflops/s per call per node (full): 443314
-Grid       : Message        : Average mflops/s per call per node (full): 443511
-Grid       : Message        : Average mflops/s per call per node (full): 302802
-Grid       : Message        :  Stencil 11.8595 GB/s per node
-Grid       : Message        :  Stencil 13.2191 GB/s per node
-Grid       : Message        :  Stencil 17.0265 GB/s per node
-Grid       : Message        :  Stencil 11.9838 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670429
-Grid       : Message        : Average mflops/s per call per node       : 805409
-Grid       : Message        : Average mflops/s per call per node       : 821689
-Grid       : Message        : Average mflops/s per call per node       : 663832
-Grid       : Message        : Average mflops/s per call per node (full): 306883
-Grid       : Message        : Average mflops/s per call per node (full): 385481
-Grid       : Message        : Average mflops/s per call per node (full): 440786
-Grid       : Message        : Average mflops/s per call per node (full): 301769
-Grid       : Message        :  Stencil 12.2979 GB/s per node
-Grid       : Message        :  Stencil 17.2923 GB/s per node
-Grid       : Message        :  Stencil 17.2221 GB/s per node
-Grid       : Message        :  Stencil 13.9567 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671006
-Grid       : Message        : Average mflops/s per call per node       : 803697
-Grid       : Message        : Average mflops/s per call per node       : 826999
-Grid       : Message        : Average mflops/s per call per node       : 663643
-Grid       : Message        : Average mflops/s per call per node (full): 311482
-Grid       : Message        : Average mflops/s per call per node (full): 437719
-Grid       : Message        : Average mflops/s per call per node (full): 443215
-Grid       : Message        : Average mflops/s per call per node (full): 307982
-Grid       : Message        :  Stencil 12.3281 GB/s per node
-Grid       : Message        :  Stencil 17.6035 GB/s per node
-Grid       : Message        :  Stencil 17.9606 GB/s per node
-Grid       : Message        :  Stencil 13.341 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668979
-Grid       : Message        : Average mflops/s per call per node       : 801647
-Grid       : Message        : Average mflops/s per call per node       : 820156
-Grid       : Message        : Average mflops/s per call per node       : 662391
-Grid       : Message        : Average mflops/s per call per node (full): 310984
-Grid       : Message        : Average mflops/s per call per node (full): 438264
-Grid       : Message        : Average mflops/s per call per node (full): 440264
-Grid       : Message        : Average mflops/s per call per node (full): 306365
-Grid       : Message        :  Stencil 12.1279 GB/s per node
-Grid       : Message        :  Stencil 16.409 GB/s per node
-Grid       : Message        :  Stencil 17.7244 GB/s per node
-Grid       : Message        :  Stencil 12.4881 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666087
-Grid       : Message        : Average mflops/s per call per node       : 800875
-Grid       : Message        : Average mflops/s per call per node       : 815854
-Grid       : Message        : Average mflops/s per call per node       : 662393
-Grid       : Message        : Average mflops/s per call per node (full): 310224
-Grid       : Message        : Average mflops/s per call per node (full): 433186
-Grid       : Message        : Average mflops/s per call per node (full): 444792
-Grid       : Message        : Average mflops/s per call per node (full): 303496
-Grid       : Message        :  Stencil 13.0349 GB/s per node
-Grid       : Message        :  Stencil 14.3654 GB/s per node
-Grid       : Message        :  Stencil 17.6391 GB/s per node
-Grid       : Message        :  Stencil 13.7551 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668233
-Grid       : Message        : Average mflops/s per call per node       : 804659
-Grid       : Message        : Average mflops/s per call per node       : 814795
-Grid       : Message        : Average mflops/s per call per node       : 664072
-Grid       : Message        : Average mflops/s per call per node (full): 314520
-Grid       : Message        : Average mflops/s per call per node (full): 406269
-Grid       : Message        : Average mflops/s per call per node (full): 444639
-Grid       : Message        : Average mflops/s per call per node (full): 306410
-Grid       : Message        :  Stencil 14.403 GB/s per node
-Grid       : Message        :  Stencil 16.4274 GB/s per node
-Grid       : Message        :  Stencil 17.764 GB/s per node
-Grid       : Message        :  Stencil 13.506 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662911
-Grid       : Message        : Average mflops/s per call per node       : 804622
-Grid       : Message        : Average mflops/s per call per node       : 820148
-Grid       : Message        : Average mflops/s per call per node       : 661708
-Grid       : Message        : Average mflops/s per call per node (full): 316671
-Grid       : Message        : Average mflops/s per call per node (full): 433131
-Grid       : Message        : Average mflops/s per call per node (full): 445038
-Grid       : Message        : Average mflops/s per call per node (full): 304646
-Grid       : Message        :  Stencil 13.1869 GB/s per node
-Grid       : Message        :  Stencil 16.7596 GB/s per node
-Grid       : Message        :  Stencil 17.5764 GB/s per node
-Grid       : Message        :  Stencil 12.3926 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666711
-Grid       : Message        : Average mflops/s per call per node       : 806839
-Grid       : Message        : Average mflops/s per call per node       : 826096
-Grid       : Message        : Average mflops/s per call per node       : 667397
-Grid       : Message        : Average mflops/s per call per node (full): 315643
-Grid       : Message        : Average mflops/s per call per node (full): 436755
-Grid       : Message        : Average mflops/s per call per node (full): 445928
-Grid       : Message        : Average mflops/s per call per node (full): 304457
-Grid       : Message        :  Stencil 13.263 GB/s per node
-Grid       : Message        :  Stencil 17.5252 GB/s per node
-Grid       : Message        :  Stencil 17.8626 GB/s per node
-Grid       : Message        :  Stencil 13.3588 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668043
-Grid       : Message        : Average mflops/s per call per node       : 799596
-Grid       : Message        : Average mflops/s per call per node       : 822349
-Grid       : Message        : Average mflops/s per call per node       : 663654
-Grid       : Message        : Average mflops/s per call per node (full): 316238
-Grid       : Message        : Average mflops/s per call per node (full): 438539
-Grid       : Message        : Average mflops/s per call per node (full): 445313
-Grid       : Message        : Average mflops/s per call per node (full): 305676
-Grid       : Message        :  Stencil 13.1748 GB/s per node
-Grid       : Message        :  Stencil 16.4387 GB/s per node
-Grid       : Message        :  Stencil 16.8156 GB/s per node
-Grid       : Message        :  Stencil 11.9933 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665377
-Grid       : Message        : Average mflops/s per call per node       : 801811
-Grid       : Message        : Average mflops/s per call per node       : 824709
-Grid       : Message        : Average mflops/s per call per node       : 666862
-Grid       : Message        : Average mflops/s per call per node (full): 314939
-Grid       : Message        : Average mflops/s per call per node (full): 433577
-Grid       : Message        : Average mflops/s per call per node (full): 438749
-Grid       : Message        : Average mflops/s per call per node (full): 302165
-Grid       : Message        :  Stencil 13.6378 GB/s per node
-Grid       : Message        :  Stencil 16.6342 GB/s per node
-Grid       : Message        :  Stencil 17.8761 GB/s per node
-Grid       : Message        :  Stencil 13.0271 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665072
-Grid       : Message        : Average mflops/s per call per node       : 803115
-Grid       : Message        : Average mflops/s per call per node       : 811562
-Grid       : Message        : Average mflops/s per call per node       : 665020
-Grid       : Message        : Average mflops/s per call per node (full): 315482
-Grid       : Message        : Average mflops/s per call per node (full): 435792
-Grid       : Message        : Average mflops/s per call per node (full): 443680
-Grid       : Message        : Average mflops/s per call per node (full): 306156
-Grid       : Message        :  Stencil 14.0895 GB/s per node
-Grid       : Message        :  Stencil 16.4099 GB/s per node
-Grid       : Message        :  Stencil 16.8037 GB/s per node
-Grid       : Message        :  Stencil 12.5877 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665096
-Grid       : Message        : Average mflops/s per call per node       : 804712
-Grid       : Message        : Average mflops/s per call per node       : 825464
-Grid       : Message        : Average mflops/s per call per node       : 665110
-Grid       : Message        : Average mflops/s per call per node (full): 316653
-Grid       : Message        : Average mflops/s per call per node (full): 433047
-Grid       : Message        : Average mflops/s per call per node (full): 438423
-Grid       : Message        : Average mflops/s per call per node (full): 303523
-Grid       : Message        :  Stencil 14.4215 GB/s per node
-Grid       : Message        :  Stencil 18.1001 GB/s per node
-Grid       : Message        :  Stencil 18.8503 GB/s per node
-Grid       : Message        :  Stencil 12.8565 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663644
-Grid       : Message        : Average mflops/s per call per node       : 803815
-Grid       : Message        : Average mflops/s per call per node       : 818369
-Grid       : Message        : Average mflops/s per call per node       : 660859
-Grid       : Message        : Average mflops/s per call per node (full): 316221
-Grid       : Message        : Average mflops/s per call per node (full): 440122
-Grid       : Message        : Average mflops/s per call per node (full): 447962
-Grid       : Message        : Average mflops/s per call per node (full): 303555
-Grid       : Message        :  Stencil 13.2125 GB/s per node
-Grid       : Message        :  Stencil 17.2994 GB/s per node
-Grid       : Message        :  Stencil 17.782 GB/s per node
-Grid       : Message        :  Stencil 12.6348 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666511
-Grid       : Message        : Average mflops/s per call per node       : 803513
-Grid       : Message        : Average mflops/s per call per node       : 820944
-Grid       : Message        : Average mflops/s per call per node       : 666189
-Grid       : Message        : Average mflops/s per call per node (full): 314097
-Grid       : Message        : Average mflops/s per call per node (full): 437299
-Grid       : Message        : Average mflops/s per call per node (full): 443860
-Grid       : Message        : Average mflops/s per call per node (full): 304287
-Grid       : Message        :  Stencil 12.5 GB/s per node
-Grid       : Message        :  Stencil 18.2757 GB/s per node
-Grid       : Message        :  Stencil 18.1259 GB/s per node
-Grid       : Message        :  Stencil 12.4919 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667520
-Grid       : Message        : Average mflops/s per call per node       : 799292
-Grid       : Message        : Average mflops/s per call per node       : 818353
-Grid       : Message        : Average mflops/s per call per node       : 665097
-Grid       : Message        : Average mflops/s per call per node (full): 313158
-Grid       : Message        : Average mflops/s per call per node (full): 438549
-Grid       : Message        : Average mflops/s per call per node (full): 447002
-Grid       : Message        : Average mflops/s per call per node (full): 304029
-Grid       : Message        :  Stencil 15.4003 GB/s per node
-Grid       : Message        :  Stencil 16.7543 GB/s per node
-Grid       : Message        :  Stencil 18.0461 GB/s per node
-Grid       : Message        :  Stencil 12.4582 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659704
-Grid       : Message        : Average mflops/s per call per node       : 802566
-Grid       : Message        : Average mflops/s per call per node       : 827087
-Grid       : Message        : Average mflops/s per call per node       : 667201
-Grid       : Message        : Average mflops/s per call per node (full): 316473
-Grid       : Message        : Average mflops/s per call per node (full): 435987
-Grid       : Message        : Average mflops/s per call per node (full): 448107
-Grid       : Message        : Average mflops/s per call per node (full): 304209
-Grid       : Message        :  Stencil 15.2555 GB/s per node
-Grid       : Message        :  Stencil 17.2014 GB/s per node
-Grid       : Message        :  Stencil 17.4156 GB/s per node
-Grid       : Message        :  Stencil 12.7329 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662821
-Grid       : Message        : Average mflops/s per call per node       : 801309
-Grid       : Message        : Average mflops/s per call per node       : 819615
-Grid       : Message        : Average mflops/s per call per node       : 670265
-Grid       : Message        : Average mflops/s per call per node (full): 317148
-Grid       : Message        : Average mflops/s per call per node (full): 436550
-Grid       : Message        : Average mflops/s per call per node (full): 440511
-Grid       : Message        : Average mflops/s per call per node (full): 306119
-Grid       : Message        :  Stencil 12.5579 GB/s per node
-Grid       : Message        :  Stencil 17.8074 GB/s per node
-Grid       : Message        :  Stencil 18.2134 GB/s per node
-Grid       : Message        :  Stencil 13.6269 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666563
-Grid       : Message        : Average mflops/s per call per node       : 802986
-Grid       : Message        : Average mflops/s per call per node       : 822776
-Grid       : Message        : Average mflops/s per call per node       : 661145
-Grid       : Message        : Average mflops/s per call per node (full): 313962
-Grid       : Message        : Average mflops/s per call per node (full): 439128
-Grid       : Message        : Average mflops/s per call per node (full): 447571
-Grid       : Message        : Average mflops/s per call per node (full): 306357
-Grid       : Message        :  Stencil 13.181 GB/s per node
-Grid       : Message        :  Stencil 17.6128 GB/s per node
-Grid       : Message        :  Stencil 17.5374 GB/s per node
-Grid       : Message        :  Stencil 13.5804 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664015
-Grid       : Message        : Average mflops/s per call per node       : 803507
-Grid       : Message        : Average mflops/s per call per node       : 818642
-Grid       : Message        : Average mflops/s per call per node       : 663962
-Grid       : Message        : Average mflops/s per call per node (full): 314661
-Grid       : Message        : Average mflops/s per call per node (full): 439069
-Grid       : Message        : Average mflops/s per call per node (full): 444028
-Grid       : Message        : Average mflops/s per call per node (full): 307019
-Grid       : Message        :  Stencil 13.6153 GB/s per node
-Grid       : Message        :  Stencil 17.0466 GB/s per node
-Grid       : Message        :  Stencil 17.4903 GB/s per node
-Grid       : Message        :  Stencil 12.5444 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666360
-Grid       : Message        : Average mflops/s per call per node       : 800808
-Grid       : Message        : Average mflops/s per call per node       : 824012
-Grid       : Message        : Average mflops/s per call per node       : 663114
-Grid       : Message        : Average mflops/s per call per node (full): 315889
-Grid       : Message        : Average mflops/s per call per node (full): 437328
-Grid       : Message        : Average mflops/s per call per node (full): 444113
-Grid       : Message        : Average mflops/s per call per node (full): 303517
-Grid       : Message        :  Stencil 14.3039 GB/s per node
-Grid       : Message        :  Stencil 17.4336 GB/s per node
-Grid       : Message        :  Stencil 17.3709 GB/s per node
-Grid       : Message        :  Stencil 13.4663 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661343
-Grid       : Message        : Average mflops/s per call per node       : 801882
-Grid       : Message        : Average mflops/s per call per node       : 825649
-Grid       : Message        : Average mflops/s per call per node       : 661617
-Grid       : Message        : Average mflops/s per call per node (full): 316563
-Grid       : Message        : Average mflops/s per call per node (full): 439155
-Grid       : Message        : Average mflops/s per call per node (full): 443762
-Grid       : Message        : Average mflops/s per call per node (full): 306622
-Grid       : Message        :  Stencil 13.4482 GB/s per node
-Grid       : Message        :  Stencil 16.7404 GB/s per node
-Grid       : Message        :  Stencil 18.4052 GB/s per node
-Grid       : Message        :  Stencil 12.698 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667886
-Grid       : Message        : Average mflops/s per call per node       : 798467
-Grid       : Message        : Average mflops/s per call per node       : 819316
-Grid       : Message        : Average mflops/s per call per node       : 663127
-Grid       : Message        : Average mflops/s per call per node (full): 316365
-Grid       : Message        : Average mflops/s per call per node (full): 435445
-Grid       : Message        : Average mflops/s per call per node (full): 447972
-Grid       : Message        : Average mflops/s per call per node (full): 304094
-Grid       : Message        :  Stencil 13.7252 GB/s per node
-Grid       : Message        :  Stencil 16.6097 GB/s per node
-Grid       : Message        :  Stencil 17.8379 GB/s per node
-Grid       : Message        :  Stencil 13.1634 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665025
-Grid       : Message        : Average mflops/s per call per node       : 804014
-Grid       : Message        : Average mflops/s per call per node       : 813861
-Grid       : Message        : Average mflops/s per call per node       : 660909
-Grid       : Message        : Average mflops/s per call per node (full): 316061
-Grid       : Message        : Average mflops/s per call per node (full): 436117
-Grid       : Message        : Average mflops/s per call per node (full): 444560
-Grid       : Message        : Average mflops/s per call per node (full): 304968
-Grid       : Message        :  Stencil 13.8481 GB/s per node
-Grid       : Message        :  Stencil 14.2236 GB/s per node
-Grid       : Message        :  Stencil 19.1255 GB/s per node
-Grid       : Message        :  Stencil 12.4868 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665570
-Grid       : Message        : Average mflops/s per call per node       : 802607
-Grid       : Message        : Average mflops/s per call per node       : 823274
-Grid       : Message        : Average mflops/s per call per node       : 666035
-Grid       : Message        : Average mflops/s per call per node (full): 316665
-Grid       : Message        : Average mflops/s per call per node (full): 404232
-Grid       : Message        : Average mflops/s per call per node (full): 450107
-Grid       : Message        : Average mflops/s per call per node (full): 303665
-Grid       : Message        :  Stencil 12.5744 GB/s per node
-Grid       : Message        :  Stencil 17.3745 GB/s per node
-Grid       : Message        :  Stencil 18.5195 GB/s per node
-Grid       : Message        :  Stencil 12.7849 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670859
-Grid       : Message        : Average mflops/s per call per node       : 806203
-Grid       : Message        : Average mflops/s per call per node       : 820805
-Grid       : Message        : Average mflops/s per call per node       : 663512
-Grid       : Message        : Average mflops/s per call per node (full): 313422
-Grid       : Message        : Average mflops/s per call per node (full): 439284
-Grid       : Message        : Average mflops/s per call per node (full): 447298
-Grid       : Message        : Average mflops/s per call per node (full): 304256
-Grid       : Message        :  Stencil 13.6758 GB/s per node
-Grid       : Message        :  Stencil 17.4313 GB/s per node
-Grid       : Message        :  Stencil 17.4837 GB/s per node
-Grid       : Message        :  Stencil 12.4561 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664223
-Grid       : Message        : Average mflops/s per call per node       : 803113
-Grid       : Message        : Average mflops/s per call per node       : 823434
-Grid       : Message        : Average mflops/s per call per node       : 665009
-Grid       : Message        : Average mflops/s per call per node (full): 316348
-Grid       : Message        : Average mflops/s per call per node (full): 438938
-Grid       : Message        : Average mflops/s per call per node (full): 444205
-Grid       : Message        : Average mflops/s per call per node (full): 302728
-Grid       : Message        :  Stencil 14.1599 GB/s per node
-Grid       : Message        :  Stencil 18.1794 GB/s per node
-Grid       : Message        :  Stencil 17.5413 GB/s per node
-Grid       : Message        :  Stencil 13.6046 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660446
-Grid       : Message        : Average mflops/s per call per node       : 803789
-Grid       : Message        : Average mflops/s per call per node       : 820256
-Grid       : Message        : Average mflops/s per call per node       : 661423
-Grid       : Message        : Average mflops/s per call per node (full): 316034
-Grid       : Message        : Average mflops/s per call per node (full): 436357
-Grid       : Message        : Average mflops/s per call per node (full): 439043
-Grid       : Message        : Average mflops/s per call per node (full): 305437
-Grid       : Message        :  Stencil 15.3867 GB/s per node
-Grid       : Message        :  Stencil 16.7997 GB/s per node
-Grid       : Message        :  Stencil 17.7 GB/s per node
-Grid       : Message        :  Stencil 12.2701 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660885
-Grid       : Message        : Average mflops/s per call per node       : 802864
-Grid       : Message        : Average mflops/s per call per node       : 822424
-Grid       : Message        : Average mflops/s per call per node       : 670941
-Grid       : Message        : Average mflops/s per call per node (full): 316719
-Grid       : Message        : Average mflops/s per call per node (full): 430612
-Grid       : Message        : Average mflops/s per call per node (full): 445515
-Grid       : Message        : Average mflops/s per call per node (full): 302982
-Grid       : Message        :  Stencil 12.8568 GB/s per node
-Grid       : Message        :  Stencil 16.7052 GB/s per node
-Grid       : Message        :  Stencil 16.6873 GB/s per node
-Grid       : Message        :  Stencil 13.3301 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663559
-Grid       : Message        : Average mflops/s per call per node       : 804200
-Grid       : Message        : Average mflops/s per call per node       : 820259
-Grid       : Message        : Average mflops/s per call per node       : 659533
-Grid       : Message        : Average mflops/s per call per node (full): 313045
-Grid       : Message        : Average mflops/s per call per node (full): 437324
-Grid       : Message        : Average mflops/s per call per node (full): 437247
-Grid       : Message        : Average mflops/s per call per node (full): 305163
-Grid       : Message        :  Stencil 13.1161 GB/s per node
-Grid       : Message        :  Stencil 16.5145 GB/s per node
-Grid       : Message        :  Stencil 17.5881 GB/s per node
-Grid       : Message        :  Stencil 12.1828 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661494
-Grid       : Message        : Average mflops/s per call per node       : 806467
-Grid       : Message        : Average mflops/s per call per node       : 820002
-Grid       : Message        : Average mflops/s per call per node       : 670374
-Grid       : Message        : Average mflops/s per call per node (full): 312961
-Grid       : Message        : Average mflops/s per call per node (full): 435157
-Grid       : Message        : Average mflops/s per call per node (full): 434648
-Grid       : Message        : Average mflops/s per call per node (full): 302641
-Grid       : Message        :  Stencil 13.4692 GB/s per node
-Grid       : Message        :  Stencil 16.3541 GB/s per node
-Grid       : Message        :  Stencil 17.6659 GB/s per node
-Grid       : Message        :  Stencil 12.8814 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663625
-Grid       : Message        : Average mflops/s per call per node       : 804987
-Grid       : Message        : Average mflops/s per call per node       : 822099
-Grid       : Message        : Average mflops/s per call per node       : 662334
-Grid       : Message        : Average mflops/s per call per node (full): 312334
-Grid       : Message        : Average mflops/s per call per node (full): 432101
-Grid       : Message        : Average mflops/s per call per node (full): 446640
-Grid       : Message        : Average mflops/s per call per node (full): 305115
-Grid       : Message        :  Stencil 12.8942 GB/s per node
-Grid       : Message        :  Stencil 16.7898 GB/s per node
-Grid       : Message        :  Stencil 17.5971 GB/s per node
-Grid       : Message        :  Stencil 12.2456 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668954
-Grid       : Message        : Average mflops/s per call per node       : 798117
-Grid       : Message        : Average mflops/s per call per node       : 824941
-Grid       : Message        : Average mflops/s per call per node       : 669293
-Grid       : Message        : Average mflops/s per call per node (full): 316080
-Grid       : Message        : Average mflops/s per call per node (full): 433248
-Grid       : Message        : Average mflops/s per call per node (full): 446401
-Grid       : Message        : Average mflops/s per call per node (full): 303309
-Grid       : Message        :  Stencil 12.6616 GB/s per node
-Grid       : Message        :  Stencil 16.7577 GB/s per node
-Grid       : Message        :  Stencil 17.7365 GB/s per node
-Grid       : Message        :  Stencil 12.4545 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670534
-Grid       : Message        : Average mflops/s per call per node       : 804258
-Grid       : Message        : Average mflops/s per call per node       : 819481
-Grid       : Message        : Average mflops/s per call per node       : 664434
-Grid       : Message        : Average mflops/s per call per node (full): 314697
-Grid       : Message        : Average mflops/s per call per node (full): 436280
-Grid       : Message        : Average mflops/s per call per node (full): 445594
-Grid       : Message        : Average mflops/s per call per node (full): 303729
-Grid       : Message        :  Stencil 13.4683 GB/s per node
-Grid       : Message        :  Stencil 16.6307 GB/s per node
-Grid       : Message        :  Stencil 18.91 GB/s per node
-Grid       : Message        :  Stencil 14.0652 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668763
-Grid       : Message        : Average mflops/s per call per node       : 805773
-Grid       : Message        : Average mflops/s per call per node       : 817079
-Grid       : Message        : Average mflops/s per call per node       : 661217
-Grid       : Message        : Average mflops/s per call per node (full): 316613
-Grid       : Message        : Average mflops/s per call per node (full): 437106
-Grid       : Message        : Average mflops/s per call per node (full): 447723
-Grid       : Message        : Average mflops/s per call per node (full): 305805
-Grid       : Message        :  Stencil 12.8557 GB/s per node
-Grid       : Message        :  Stencil 15.3086 GB/s per node
-Grid       : Message        :  Stencil 17.0187 GB/s per node
-Grid       : Message        :  Stencil 12.391 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671402
-Grid       : Message        : Average mflops/s per call per node       : 804054
-Grid       : Message        : Average mflops/s per call per node       : 829054
-Grid       : Message        : Average mflops/s per call per node       : 663265
-Grid       : Message        : Average mflops/s per call per node (full): 316413
-Grid       : Message        : Average mflops/s per call per node (full): 420778
-Grid       : Message        : Average mflops/s per call per node (full): 441034
-Grid       : Message        : Average mflops/s per call per node (full): 302925
-Grid       : Message        :  Stencil 12.5937 GB/s per node
-Grid       : Message        :  Stencil 14.2749 GB/s per node
-Grid       : Message        :  Stencil 17.3705 GB/s per node
-Grid       : Message        :  Stencil 12.9952 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 673301
-Grid       : Message        : Average mflops/s per call per node       : 811708
-Grid       : Message        : Average mflops/s per call per node       : 825225
-Grid       : Message        : Average mflops/s per call per node       : 666381
-Grid       : Message        : Average mflops/s per call per node (full): 314965
-Grid       : Message        : Average mflops/s per call per node (full): 404682
-Grid       : Message        : Average mflops/s per call per node (full): 444928
-Grid       : Message        : Average mflops/s per call per node (full): 306184
-Grid       : Message        :  Stencil 12.4751 GB/s per node
-Grid       : Message        :  Stencil 16.734 GB/s per node
-Grid       : Message        :  Stencil 17.7405 GB/s per node
-Grid       : Message        :  Stencil 13.0793 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670239
-Grid       : Message        : Average mflops/s per call per node       : 803610
-Grid       : Message        : Average mflops/s per call per node       : 822767
-Grid       : Message        : Average mflops/s per call per node       : 661298
-Grid       : Message        : Average mflops/s per call per node (full): 314063
-Grid       : Message        : Average mflops/s per call per node (full): 436778
-Grid       : Message        : Average mflops/s per call per node (full): 446006
-Grid       : Message        : Average mflops/s per call per node (full): 305186
-Grid       : Message        :  Stencil 13.1137 GB/s per node
-Grid       : Message        :  Stencil 16.3732 GB/s per node
-Grid       : Message        :  Stencil 18.3038 GB/s per node
-Grid       : Message        :  Stencil 12.7076 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666051
-Grid       : Message        : Average mflops/s per call per node       : 803733
-Grid       : Message        : Average mflops/s per call per node       : 825737
-Grid       : Message        : Average mflops/s per call per node       : 667496
-Grid       : Message        : Average mflops/s per call per node (full): 314399
-Grid       : Message        : Average mflops/s per call per node (full): 432762
-Grid       : Message        : Average mflops/s per call per node (full): 449063
-Grid       : Message        : Average mflops/s per call per node (full): 305673
-Grid       : Message        :  Stencil 12.5174 GB/s per node
-Grid       : Message        :  Stencil 16.1717 GB/s per node
-Grid       : Message        :  Stencil 17.2254 GB/s per node
-Grid       : Message        :  Stencil 12.9635 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666187
-Grid       : Message        : Average mflops/s per call per node       : 803564
-Grid       : Message        : Average mflops/s per call per node       : 822850
-Grid       : Message        : Average mflops/s per call per node       : 664979
-Grid       : Message        : Average mflops/s per call per node (full): 313719
-Grid       : Message        : Average mflops/s per call per node (full): 430242
-Grid       : Message        : Average mflops/s per call per node (full): 442125
-Grid       : Message        : Average mflops/s per call per node (full): 304770
-Grid       : Message        :  Stencil 12.8431 GB/s per node
-Grid       : Message        :  Stencil 17.4906 GB/s per node
-Grid       : Message        :  Stencil 17.2259 GB/s per node
-Grid       : Message        :  Stencil 13.003 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666116
-Grid       : Message        : Average mflops/s per call per node       : 801907
-Grid       : Message        : Average mflops/s per call per node       : 818616
-Grid       : Message        : Average mflops/s per call per node       : 661256
-Grid       : Message        : Average mflops/s per call per node (full): 314504
-Grid       : Message        : Average mflops/s per call per node (full): 437613
-Grid       : Message        : Average mflops/s per call per node (full): 443053
-Grid       : Message        : Average mflops/s per call per node (full): 305154
-Grid       : Message        :  Stencil 14.0723 GB/s per node
-Grid       : Message        :  Stencil 17.498 GB/s per node
-Grid       : Message        :  Stencil 17.6098 GB/s per node
-Grid       : Message        :  Stencil 12.2343 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664473
-Grid       : Message        : Average mflops/s per call per node       : 802242
-Grid       : Message        : Average mflops/s per call per node       : 820944
-Grid       : Message        : Average mflops/s per call per node       : 669711
-Grid       : Message        : Average mflops/s per call per node (full): 316594
-Grid       : Message        : Average mflops/s per call per node (full): 439951
-Grid       : Message        : Average mflops/s per call per node (full): 443698
-Grid       : Message        : Average mflops/s per call per node (full): 304216
-Grid       : Message        :  Stencil 13.0752 GB/s per node
-Grid       : Message        :  Stencil 18.2619 GB/s per node
-Grid       : Message        :  Stencil 18.5119 GB/s per node
-Grid       : Message        :  Stencil 12.4328 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664692
-Grid       : Message        : Average mflops/s per call per node       : 802193
-Grid       : Message        : Average mflops/s per call per node       : 824987
-Grid       : Message        : Average mflops/s per call per node       : 664304
-Grid       : Message        : Average mflops/s per call per node (full): 314319
-Grid       : Message        : Average mflops/s per call per node (full): 441373
-Grid       : Message        : Average mflops/s per call per node (full): 448471
-Grid       : Message        : Average mflops/s per call per node (full): 303780
-Grid       : Message        :  Stencil 12.4062 GB/s per node
-Grid       : Message        :  Stencil 16.5179 GB/s per node
-Grid       : Message        :  Stencil 17.8355 GB/s per node
-Grid       : Message        :  Stencil 14.5781 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668287
-Grid       : Message        : Average mflops/s per call per node       : 804701
-Grid       : Message        : Average mflops/s per call per node       : 823599
-Grid       : Message        : Average mflops/s per call per node       : 658889
-Grid       : Message        : Average mflops/s per call per node (full): 312743
-Grid       : Message        : Average mflops/s per call per node (full): 434922
-Grid       : Message        : Average mflops/s per call per node (full): 438520
-Grid       : Message        : Average mflops/s per call per node (full): 306746
-Grid       : Message        :  Stencil 13.3246 GB/s per node
-Grid       : Message        :  Stencil 17.0093 GB/s per node
-Grid       : Message        :  Stencil 18.994 GB/s per node
-Grid       : Message        :  Stencil 13.4837 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664741
-Grid       : Message        : Average mflops/s per call per node       : 800517
-Grid       : Message        : Average mflops/s per call per node       : 820825
-Grid       : Message        : Average mflops/s per call per node       : 660615
-Grid       : Message        : Average mflops/s per call per node (full): 314883
-Grid       : Message        : Average mflops/s per call per node (full): 437339
-Grid       : Message        : Average mflops/s per call per node (full): 449436
-Grid       : Message        : Average mflops/s per call per node (full): 303865
-Grid       : Message        :  Stencil 12.2983 GB/s per node
-Grid       : Message        :  Stencil 16.8586 GB/s per node
-Grid       : Message        :  Stencil 18.3762 GB/s per node
-Grid       : Message        :  Stencil 12.0417 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665390
-Grid       : Message        : Average mflops/s per call per node       : 807029
-Grid       : Message        : Average mflops/s per call per node       : 823583
-Grid       : Message        : Average mflops/s per call per node       : 671107
-Grid       : Message        : Average mflops/s per call per node (full): 311919
-Grid       : Message        : Average mflops/s per call per node (full): 438107
-Grid       : Message        : Average mflops/s per call per node (full): 448653
-Grid       : Message        : Average mflops/s per call per node (full): 302841
-Grid       : Message        :  Stencil 13.0546 GB/s per node
-Grid       : Message        :  Stencil 16.4252 GB/s per node
-Grid       : Message        :  Stencil 17.4846 GB/s per node
-Grid       : Message        :  Stencil 12.7131 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667790
-Grid       : Message        : Average mflops/s per call per node       : 804420
-Grid       : Message        : Average mflops/s per call per node       : 822428
-Grid       : Message        : Average mflops/s per call per node       : 662207
-Grid       : Message        : Average mflops/s per call per node (full): 314403
-Grid       : Message        : Average mflops/s per call per node (full): 433325
-Grid       : Message        : Average mflops/s per call per node (full): 444422
-Grid       : Message        : Average mflops/s per call per node (full): 304579
-Grid       : Message        :  Stencil 13.2941 GB/s per node
-Grid       : Message        :  Stencil 9.77585 GB/s per node
-Grid       : Message        :  Stencil 17.002 GB/s per node
-Grid       : Message        :  Stencil 12.5608 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665332
-Grid       : Message        : Average mflops/s per call per node       : 809864
-Grid       : Message        : Average mflops/s per call per node       : 821372
-Grid       : Message        : Average mflops/s per call per node       : 667238
-Grid       : Message        : Average mflops/s per call per node (full): 315191
-Grid       : Message        : Average mflops/s per call per node (full): 311246
-Grid       : Message        : Average mflops/s per call per node (full): 440765
-Grid       : Message        : Average mflops/s per call per node (full): 305183
-Grid       : Message        :  Stencil 13.4714 GB/s per node
-Grid       : Message        :  Stencil 16.6391 GB/s per node
-Grid       : Message        :  Stencil 17.3293 GB/s per node
-Grid       : Message        :  Stencil 12.9076 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662206
-Grid       : Message        : Average mflops/s per call per node       : 803920
-Grid       : Message        : Average mflops/s per call per node       : 818714
-Grid       : Message        : Average mflops/s per call per node       : 662318
-Grid       : Message        : Average mflops/s per call per node (full): 314274
-Grid       : Message        : Average mflops/s per call per node (full): 436457
-Grid       : Message        : Average mflops/s per call per node (full): 443368
-Grid       : Message        : Average mflops/s per call per node (full): 304443
-Grid       : Message        :  Stencil 13.1367 GB/s per node
-Grid       : Message        :  Stencil 17.4188 GB/s per node
-Grid       : Message        :  Stencil 18.3514 GB/s per node
-Grid       : Message        :  Stencil 13.8068 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668658
-Grid       : Message        : Average mflops/s per call per node       : 805311
-Grid       : Message        : Average mflops/s per call per node       : 819600
-Grid       : Message        : Average mflops/s per call per node       : 662668
-Grid       : Message        : Average mflops/s per call per node (full): 315365
-Grid       : Message        : Average mflops/s per call per node (full): 439100
-Grid       : Message        : Average mflops/s per call per node (full): 445623
-Grid       : Message        : Average mflops/s per call per node (full): 306193
-Grid       : Message        :  Stencil 13.9619 GB/s per node
-Grid       : Message        :  Stencil 14.1776 GB/s per node
-Grid       : Message        :  Stencil 17.4347 GB/s per node
-Grid       : Message        :  Stencil 12.8607 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662908
-Grid       : Message        : Average mflops/s per call per node       : 810045
-Grid       : Message        : Average mflops/s per call per node       : 829036
-Grid       : Message        : Average mflops/s per call per node       : 664978
-Grid       : Message        : Average mflops/s per call per node (full): 316261
-Grid       : Message        : Average mflops/s per call per node (full): 402716
-Grid       : Message        : Average mflops/s per call per node (full): 445448
-Grid       : Message        : Average mflops/s per call per node (full): 304756
-Grid       : Message        :  Stencil 13.2573 GB/s per node
-Grid       : Message        :  Stencil 17.1684 GB/s per node
-Grid       : Message        :  Stencil 17.4563 GB/s per node
-Grid       : Message        :  Stencil 12.9146 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666694
-Grid       : Message        : Average mflops/s per call per node       : 800308
-Grid       : Message        : Average mflops/s per call per node       : 826364
-Grid       : Message        : Average mflops/s per call per node       : 667065
-Grid       : Message        : Average mflops/s per call per node (full): 315370
-Grid       : Message        : Average mflops/s per call per node (full): 435363
-Grid       : Message        : Average mflops/s per call per node (full): 444945
-Grid       : Message        : Average mflops/s per call per node (full): 305344
-Grid       : Message        :  Stencil 12.5017 GB/s per node
-Grid       : Message        :  Stencil 17.702 GB/s per node
-Grid       : Message        :  Stencil 17.1299 GB/s per node
-Grid       : Message        :  Stencil 12.8331 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669806
-Grid       : Message        : Average mflops/s per call per node       : 800532
-Grid       : Message        : Average mflops/s per call per node       : 827147
-Grid       : Message        : Average mflops/s per call per node       : 659226
-Grid       : Message        : Average mflops/s per call per node (full): 313519
-Grid       : Message        : Average mflops/s per call per node (full): 439183
-Grid       : Message        : Average mflops/s per call per node (full): 440451
-Grid       : Message        : Average mflops/s per call per node (full): 302614
-Grid       : Message        :  Stencil 13.0484 GB/s per node
-Grid       : Message        :  Stencil 16.8314 GB/s per node
-Grid       : Message        :  Stencil 18.3745 GB/s per node
-Grid       : Message        :  Stencil 12.5254 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666511
-Grid       : Message        : Average mflops/s per call per node       : 803076
-Grid       : Message        : Average mflops/s per call per node       : 819195
-Grid       : Message        : Average mflops/s per call per node       : 664296
-Grid       : Message        : Average mflops/s per call per node (full): 314529
-Grid       : Message        : Average mflops/s per call per node (full): 435843
-Grid       : Message        : Average mflops/s per call per node (full): 446847
-Grid       : Message        : Average mflops/s per call per node (full): 303151
-Grid       : Message        :  Stencil 12.3858 GB/s per node
-Grid       : Message        :  Stencil 17.2709 GB/s per node
-Grid       : Message        :  Stencil 17.303 GB/s per node
-Grid       : Message        :  Stencil 13.8556 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667433
-Grid       : Message        : Average mflops/s per call per node       : 799882
-Grid       : Message        : Average mflops/s per call per node       : 825126
-Grid       : Message        : Average mflops/s per call per node       : 661650
-Grid       : Message        : Average mflops/s per call per node (full): 312345
-Grid       : Message        : Average mflops/s per call per node (full): 435733
-Grid       : Message        : Average mflops/s per call per node (full): 444059
-Grid       : Message        : Average mflops/s per call per node (full): 306157
-Grid       : Message        :  Stencil 13.2423 GB/s per node
-Grid       : Message        :  Stencil 17.1759 GB/s per node
-Grid       : Message        :  Stencil 17.6773 GB/s per node
-Grid       : Message        :  Stencil 13.8838 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666065
-Grid       : Message        : Average mflops/s per call per node       : 803042
-Grid       : Message        : Average mflops/s per call per node       : 822598
-Grid       : Message        : Average mflops/s per call per node       : 666285
-Grid       : Message        : Average mflops/s per call per node (full): 315382
-Grid       : Message        : Average mflops/s per call per node (full): 437049
-Grid       : Message        : Average mflops/s per call per node (full): 445820
-Grid       : Message        : Average mflops/s per call per node (full): 306615
-Grid       : Message        :  Stencil 13.2119 GB/s per node
-Grid       : Message        :  Stencil 17.4023 GB/s per node
-Grid       : Message        :  Stencil 18.1174 GB/s per node
-Grid       : Message        :  Stencil 12.4868 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666930
-Grid       : Message        : Average mflops/s per call per node       : 805994
-Grid       : Message        : Average mflops/s per call per node       : 821803
-Grid       : Message        : Average mflops/s per call per node       : 665081
-Grid       : Message        : Average mflops/s per call per node (full): 315690
-Grid       : Message        : Average mflops/s per call per node (full): 437432
-Grid       : Message        : Average mflops/s per call per node (full): 447243
-Grid       : Message        : Average mflops/s per call per node (full): 305000
-Grid       : Message        :  Stencil 13.9222 GB/s per node
-Grid       : Message        :  Stencil 16.6438 GB/s per node
-Grid       : Message        :  Stencil 16.9078 GB/s per node
-Grid       : Message        :  Stencil 13.3717 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664940
-Grid       : Message        : Average mflops/s per call per node       : 802706
-Grid       : Message        : Average mflops/s per call per node       : 821356
-Grid       : Message        : Average mflops/s per call per node       : 664479
-Grid       : Message        : Average mflops/s per call per node (full): 316591
-Grid       : Message        : Average mflops/s per call per node (full): 436145
-Grid       : Message        : Average mflops/s per call per node (full): 439766
-Grid       : Message        : Average mflops/s per call per node (full): 305996
-Grid       : Message        :  Stencil 13.0773 GB/s per node
-Grid       : Message        :  Stencil 17.1815 GB/s per node
-Grid       : Message        :  Stencil 17.4962 GB/s per node
-Grid       : Message        :  Stencil 14.5529 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667989
-Grid       : Message        : Average mflops/s per call per node       : 806203
-Grid       : Message        : Average mflops/s per call per node       : 821061
-Grid       : Message        : Average mflops/s per call per node       : 660524
-Grid       : Message        : Average mflops/s per call per node (full): 315502
-Grid       : Message        : Average mflops/s per call per node (full): 436257
-Grid       : Message        : Average mflops/s per call per node (full): 444052
-Grid       : Message        : Average mflops/s per call per node (full): 307003
-Grid       : Message        :  Stencil 14.2909 GB/s per node
-Grid       : Message        :  Stencil 16.5371 GB/s per node
-Grid       : Message        :  Stencil 17.1227 GB/s per node
-Grid       : Message        :  Stencil 13.4769 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665438
-Grid       : Message        : Average mflops/s per call per node       : 804734
-Grid       : Message        : Average mflops/s per call per node       : 825460
-Grid       : Message        : Average mflops/s per call per node       : 667013
-Grid       : Message        : Average mflops/s per call per node (full): 315943
-Grid       : Message        : Average mflops/s per call per node (full): 435302
-Grid       : Message        : Average mflops/s per call per node (full): 436536
-Grid       : Message        : Average mflops/s per call per node (full): 305996
-Grid       : Message        :  Stencil 13.9086 GB/s per node
-Grid       : Message        :  Stencil 17.3296 GB/s per node
-Grid       : Message        :  Stencil 17.3873 GB/s per node
-Grid       : Message        :  Stencil 13.8987 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663859
-Grid       : Message        : Average mflops/s per call per node       : 806287
-Grid       : Message        : Average mflops/s per call per node       : 826110
-Grid       : Message        : Average mflops/s per call per node       : 663951
-Grid       : Message        : Average mflops/s per call per node (full): 314912
-Grid       : Message        : Average mflops/s per call per node (full): 440008
-Grid       : Message        : Average mflops/s per call per node (full): 442199
-Grid       : Message        : Average mflops/s per call per node (full): 307155
-Grid       : Message        :  Stencil 13.5995 GB/s per node
-Grid       : Message        :  Stencil 17.1634 GB/s per node
-Grid       : Message        :  Stencil 17.526 GB/s per node
-Grid       : Message        :  Stencil 13.5101 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665903
-Grid       : Message        : Average mflops/s per call per node       : 804316
-Grid       : Message        : Average mflops/s per call per node       : 823921
-Grid       : Message        : Average mflops/s per call per node       : 654776
-Grid       : Message        : Average mflops/s per call per node (full): 317058
-Grid       : Message        : Average mflops/s per call per node (full): 438237
-Grid       : Message        : Average mflops/s per call per node (full): 445829
-Grid       : Message        : Average mflops/s per call per node (full): 301512
-Grid       : Message        :  Stencil 14.1811 GB/s per node
-Grid       : Message        :  Stencil 17.4119 GB/s per node
-Grid       : Message        :  Stencil 17.4588 GB/s per node
-Grid       : Message        :  Stencil 12.7528 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664117
-Grid       : Message        : Average mflops/s per call per node       : 804561
-Grid       : Message        : Average mflops/s per call per node       : 823380
-Grid       : Message        : Average mflops/s per call per node       : 666616
-Grid       : Message        : Average mflops/s per call per node (full): 316154
-Grid       : Message        : Average mflops/s per call per node (full): 437275
-Grid       : Message        : Average mflops/s per call per node (full): 444732
-Grid       : Message        : Average mflops/s per call per node (full): 306355
-Grid       : Message        :  Stencil 12.5427 GB/s per node
-Grid       : Message        :  Stencil 16.8808 GB/s per node
-Grid       : Message        :  Stencil 17.2794 GB/s per node
-Grid       : Message        :  Stencil 12.1337 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666693
-Grid       : Message        : Average mflops/s per call per node       : 805130
-Grid       : Message        : Average mflops/s per call per node       : 822421
-Grid       : Message        : Average mflops/s per call per node       : 667030
-Grid       : Message        : Average mflops/s per call per node (full): 313444
-Grid       : Message        : Average mflops/s per call per node (full): 438107
-Grid       : Message        : Average mflops/s per call per node (full): 443039
-Grid       : Message        : Average mflops/s per call per node (full): 303071
-Grid       : Message        :  Stencil 13.5952 GB/s per node
-Grid       : Message        :  Stencil 17.2235 GB/s per node
-Grid       : Message        :  Stencil 17.5301 GB/s per node
-Grid       : Message        :  Stencil 12.2712 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666773
-Grid       : Message        : Average mflops/s per call per node       : 798745
-Grid       : Message        : Average mflops/s per call per node       : 825565
-Grid       : Message        : Average mflops/s per call per node       : 663631
-Grid       : Message        : Average mflops/s per call per node (full): 315302
-Grid       : Message        : Average mflops/s per call per node (full): 435436
-Grid       : Message        : Average mflops/s per call per node (full): 445069
-Grid       : Message        : Average mflops/s per call per node (full): 304156
-Grid       : Message        :  Stencil 13.3931 GB/s per node
-Grid       : Message        :  Stencil 16.741 GB/s per node
-Grid       : Message        :  Stencil 18.0163 GB/s per node
-Grid       : Message        :  Stencil 13.0197 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666664
-Grid       : Message        : Average mflops/s per call per node       : 806233
-Grid       : Message        : Average mflops/s per call per node       : 822672
-Grid       : Message        : Average mflops/s per call per node       : 664351
-Grid       : Message        : Average mflops/s per call per node (full): 316088
-Grid       : Message        : Average mflops/s per call per node (full): 436794
-Grid       : Message        : Average mflops/s per call per node (full): 448626
-Grid       : Message        : Average mflops/s per call per node (full): 305654
-Grid       : Message        :  Stencil 13.343 GB/s per node
-Grid       : Message        :  Stencil 16.877 GB/s per node
-Grid       : Message        :  Stencil 18.055 GB/s per node
-Grid       : Message        :  Stencil 13.0708 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666103
-Grid       : Message        : Average mflops/s per call per node       : 804067
-Grid       : Message        : Average mflops/s per call per node       : 821323
-Grid       : Message        : Average mflops/s per call per node       : 661550
-Grid       : Message        : Average mflops/s per call per node (full): 315256
-Grid       : Message        : Average mflops/s per call per node (full): 436469
-Grid       : Message        : Average mflops/s per call per node (full): 445568
-Grid       : Message        : Average mflops/s per call per node (full): 304861
-Grid       : Message        :  Stencil 13.3767 GB/s per node
-Grid       : Message        :  Stencil 17.4307 GB/s per node
-Grid       : Message        :  Stencil 17.8732 GB/s per node
-Grid       : Message        :  Stencil 14.365 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665950
-Grid       : Message        : Average mflops/s per call per node       : 804442
-Grid       : Message        : Average mflops/s per call per node       : 820539
-Grid       : Message        : Average mflops/s per call per node       : 665047
-Grid       : Message        : Average mflops/s per call per node (full): 316375
-Grid       : Message        : Average mflops/s per call per node (full): 439167
-Grid       : Message        : Average mflops/s per call per node (full): 447065
-Grid       : Message        : Average mflops/s per call per node (full): 307428
-Grid       : Message        :  Stencil 12.821 GB/s per node
-Grid       : Message        :  Stencil 13.5966 GB/s per node
-Grid       : Message        :  Stencil 17.6488 GB/s per node
-Grid       : Message        :  Stencil 13.8616 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670359
-Grid       : Message        : Average mflops/s per call per node       : 812443
-Grid       : Message        : Average mflops/s per call per node       : 821539
-Grid       : Message        : Average mflops/s per call per node       : 662747
-Grid       : Message        : Average mflops/s per call per node (full): 315688
-Grid       : Message        : Average mflops/s per call per node (full): 392718
-Grid       : Message        : Average mflops/s per call per node (full): 445675
-Grid       : Message        : Average mflops/s per call per node (full): 306518
-Grid       : Message        :  Stencil 12.7894 GB/s per node
-Grid       : Message        :  Stencil 17.3747 GB/s per node
-Grid       : Message        :  Stencil 18.4541 GB/s per node
-Grid       : Message        :  Stencil 14.2313 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668881
-Grid       : Message        : Average mflops/s per call per node       : 807244
-Grid       : Message        : Average mflops/s per call per node       : 820681
-Grid       : Message        : Average mflops/s per call per node       : 660460
-Grid       : Message        : Average mflops/s per call per node (full): 315852
-Grid       : Message        : Average mflops/s per call per node (full): 433116
-Grid       : Message        : Average mflops/s per call per node (full): 448640
-Grid       : Message        : Average mflops/s per call per node (full): 306582
-Grid       : Message        :  Stencil 12.9534 GB/s per node
-Grid       : Message        :  Stencil 17.086 GB/s per node
-Grid       : Message        :  Stencil 17.926 GB/s per node
-Grid       : Message        :  Stencil 12.4938 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666505
-Grid       : Message        : Average mflops/s per call per node       : 804431
-Grid       : Message        : Average mflops/s per call per node       : 820428
-Grid       : Message        : Average mflops/s per call per node       : 669124
-Grid       : Message        : Average mflops/s per call per node (full): 315643
-Grid       : Message        : Average mflops/s per call per node (full): 439109
-Grid       : Message        : Average mflops/s per call per node (full): 446081
-Grid       : Message        : Average mflops/s per call per node (full): 305512
-Grid       : Message        :  Stencil 13.1323 GB/s per node
-Grid       : Message        :  Stencil 16.6478 GB/s per node
-Grid       : Message        :  Stencil 17.6423 GB/s per node
-Grid       : Message        :  Stencil 12.3352 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665775
-Grid       : Message        : Average mflops/s per call per node       : 807142
-Grid       : Message        : Average mflops/s per call per node       : 819368
-Grid       : Message        : Average mflops/s per call per node       : 666051
-Grid       : Message        : Average mflops/s per call per node (full): 315122
-Grid       : Message        : Average mflops/s per call per node (full): 431694
-Grid       : Message        : Average mflops/s per call per node (full): 444808
-Grid       : Message        : Average mflops/s per call per node (full): 302136
-Grid       : Message        :  Stencil 14.1863 GB/s per node
-Grid       : Message        :  Stencil 16.6056 GB/s per node
-Grid       : Message        :  Stencil 18.3201 GB/s per node
-Grid       : Message        :  Stencil 12.6007 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663137
-Grid       : Message        : Average mflops/s per call per node       : 802466
-Grid       : Message        : Average mflops/s per call per node       : 821600
-Grid       : Message        : Average mflops/s per call per node       : 661086
-Grid       : Message        : Average mflops/s per call per node (full): 315558
-Grid       : Message        : Average mflops/s per call per node (full): 435761
-Grid       : Message        : Average mflops/s per call per node (full): 448283
-Grid       : Message        : Average mflops/s per call per node (full): 302765
-Grid       : Message        :  Stencil 13.3338 GB/s per node
-Grid       : Message        :  Stencil 17.1289 GB/s per node
-Grid       : Message        :  Stencil 17.5447 GB/s per node
-Grid       : Message        :  Stencil 12.1588 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661532
-Grid       : Message        : Average mflops/s per call per node       : 806950
-Grid       : Message        : Average mflops/s per call per node       : 826036
-Grid       : Message        : Average mflops/s per call per node       : 668188
-Grid       : Message        : Average mflops/s per call per node (full): 312926
-Grid       : Message        : Average mflops/s per call per node (full): 437556
-Grid       : Message        : Average mflops/s per call per node (full): 445663
-Grid       : Message        : Average mflops/s per call per node (full): 303389
-Grid       : Message        :  Stencil 13.5784 GB/s per node
-Grid       : Message        :  Stencil 14.0755 GB/s per node
-Grid       : Message        :  Stencil 17.46 GB/s per node
-Grid       : Message        :  Stencil 13.1343 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662222
-Grid       : Message        : Average mflops/s per call per node       : 799662
-Grid       : Message        : Average mflops/s per call per node       : 820866
-Grid       : Message        : Average mflops/s per call per node       : 667038
-Grid       : Message        : Average mflops/s per call per node (full): 314441
-Grid       : Message        : Average mflops/s per call per node (full): 400252
-Grid       : Message        : Average mflops/s per call per node (full): 442573
-Grid       : Message        : Average mflops/s per call per node (full): 306094
-Grid       : Message        :  Stencil 13.7759 GB/s per node
-Grid       : Message        :  Stencil 17.173 GB/s per node
-Grid       : Message        :  Stencil 19.3367 GB/s per node
-Grid       : Message        :  Stencil 14.1301 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664650
-Grid       : Message        : Average mflops/s per call per node       : 806907
-Grid       : Message        : Average mflops/s per call per node       : 820560
-Grid       : Message        : Average mflops/s per call per node       : 659010
-Grid       : Message        : Average mflops/s per call per node (full): 315181
-Grid       : Message        : Average mflops/s per call per node (full): 435972
-Grid       : Message        : Average mflops/s per call per node (full): 449939
-Grid       : Message        : Average mflops/s per call per node (full): 303438
-Grid       : Message        :  Stencil 13.6664 GB/s per node
-Grid       : Message        :  Stencil 17.1214 GB/s per node
-Grid       : Message        :  Stencil 18.6147 GB/s per node
-Grid       : Message        :  Stencil 12.3372 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663164
-Grid       : Message        : Average mflops/s per call per node       : 804815
-Grid       : Message        : Average mflops/s per call per node       : 813523
-Grid       : Message        : Average mflops/s per call per node       : 664938
-Grid       : Message        : Average mflops/s per call per node (full): 315596
-Grid       : Message        : Average mflops/s per call per node (full): 437513
-Grid       : Message        : Average mflops/s per call per node (full): 447273
-Grid       : Message        : Average mflops/s per call per node (full): 304029
-Grid       : Message        :  Stencil 14.2618 GB/s per node
-Grid       : Message        :  Stencil 10.0889 GB/s per node
-Grid       : Message        :  Stencil 17.4008 GB/s per node
-Grid       : Message        :  Stencil 12.3218 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661479
-Grid       : Message        : Average mflops/s per call per node       : 807279
-Grid       : Message        : Average mflops/s per call per node       : 829648
-Grid       : Message        : Average mflops/s per call per node       : 667827
-Grid       : Message        : Average mflops/s per call per node (full): 316019
-Grid       : Message        : Average mflops/s per call per node (full): 318759
-Grid       : Message        : Average mflops/s per call per node (full): 445239
-Grid       : Message        : Average mflops/s per call per node (full): 304443
-Grid       : Message        :  Stencil 13.7394 GB/s per node
-Grid       : Message        :  Stencil 17.5089 GB/s per node
-Grid       : Message        :  Stencil 17.5503 GB/s per node
-Grid       : Message        :  Stencil 13.057 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665386
-Grid       : Message        : Average mflops/s per call per node       : 801394
-Grid       : Message        : Average mflops/s per call per node       : 823385
-Grid       : Message        : Average mflops/s per call per node       : 662710
-Grid       : Message        : Average mflops/s per call per node (full): 316635
-Grid       : Message        : Average mflops/s per call per node (full): 439026
-Grid       : Message        : Average mflops/s per call per node (full): 438906
-Grid       : Message        : Average mflops/s per call per node (full): 304242
-Grid       : Message        :  Stencil 13.4454 GB/s per node
-Grid       : Message        :  Stencil 16.8154 GB/s per node
-Grid       : Message        :  Stencil 19.4219 GB/s per node
-Grid       : Message        :  Stencil 13.3221 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664908
-Grid       : Message        : Average mflops/s per call per node       : 807446
-Grid       : Message        : Average mflops/s per call per node       : 819760
-Grid       : Message        : Average mflops/s per call per node       : 658599
-Grid       : Message        : Average mflops/s per call per node (full): 315954
-Grid       : Message        : Average mflops/s per call per node (full): 437560
-Grid       : Message        : Average mflops/s per call per node (full): 449689
-Grid       : Message        : Average mflops/s per call per node (full): 305350
-Grid       : Message        :  Stencil 14.9133 GB/s per node
-Grid       : Message        :  Stencil 16.8471 GB/s per node
-Grid       : Message        :  Stencil 17.4022 GB/s per node
-Grid       : Message        :  Stencil 13.4123 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664239
-Grid       : Message        : Average mflops/s per call per node       : 806584
-Grid       : Message        : Average mflops/s per call per node       : 823548
-Grid       : Message        : Average mflops/s per call per node       : 661416
-Grid       : Message        : Average mflops/s per call per node (full): 316375
-Grid       : Message        : Average mflops/s per call per node (full): 438739
-Grid       : Message        : Average mflops/s per call per node (full): 443985
-Grid       : Message        : Average mflops/s per call per node (full): 305805
-Grid       : Message        :  Stencil 13.5212 GB/s per node
-Grid       : Message        :  Stencil 17.0275 GB/s per node
-Grid       : Message        :  Stencil 18.5525 GB/s per node
-Grid       : Message        :  Stencil 14.6028 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665280
-Grid       : Message        : Average mflops/s per call per node       : 804417
-Grid       : Message        : Average mflops/s per call per node       : 819650
-Grid       : Message        : Average mflops/s per call per node       : 662591
-Grid       : Message        : Average mflops/s per call per node (full): 315578
-Grid       : Message        : Average mflops/s per call per node (full): 435106
-Grid       : Message        : Average mflops/s per call per node (full): 445912
-Grid       : Message        : Average mflops/s per call per node (full): 307890
-Grid       : Message        :  Stencil 12.5063 GB/s per node
-Grid       : Message        :  Stencil 17.0401 GB/s per node
-Grid       : Message        :  Stencil 17.909 GB/s per node
-Grid       : Message        :  Stencil 12.4003 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670017
-Grid       : Message        : Average mflops/s per call per node       : 797635
-Grid       : Message        : Average mflops/s per call per node       : 822123
-Grid       : Message        : Average mflops/s per call per node       : 669380
-Grid       : Message        : Average mflops/s per call per node (full): 313835
-Grid       : Message        : Average mflops/s per call per node (full): 432405
-Grid       : Message        : Average mflops/s per call per node (full): 445624
-Grid       : Message        : Average mflops/s per call per node (full): 304862
-Grid       : Message        :  Stencil 13.4388 GB/s per node
-Grid       : Message        :  Stencil 10.13 GB/s per node
-Grid       : Message        :  Stencil 18.4098 GB/s per node
-Grid       : Message        :  Stencil 12.6245 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664449
-Grid       : Message        : Average mflops/s per call per node       : 813357
-Grid       : Message        : Average mflops/s per call per node       : 820461
-Grid       : Message        : Average mflops/s per call per node       : 666593
-Grid       : Message        : Average mflops/s per call per node (full): 315191
-Grid       : Message        : Average mflops/s per call per node (full): 319670
-Grid       : Message        : Average mflops/s per call per node (full): 448415
-Grid       : Message        : Average mflops/s per call per node (full): 304444
-Grid       : Message        :  Stencil 12.8495 GB/s per node
-Grid       : Message        :  Stencil 16.3414 GB/s per node
-Grid       : Message        :  Stencil 17.4925 GB/s per node
-Grid       : Message        :  Stencil 12.4912 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667422
-Grid       : Message        : Average mflops/s per call per node       : 806560
-Grid       : Message        : Average mflops/s per call per node       : 825592
-Grid       : Message        : Average mflops/s per call per node       : 667284
-Grid       : Message        : Average mflops/s per call per node (full): 314544
-Grid       : Message        : Average mflops/s per call per node (full): 433583
-Grid       : Message        : Average mflops/s per call per node (full): 444810
-Grid       : Message        : Average mflops/s per call per node (full): 304577
-Grid       : Message        :  Stencil 12.3674 GB/s per node
-Grid       : Message        :  Stencil 14.5964 GB/s per node
-Grid       : Message        :  Stencil 17.6038 GB/s per node
-Grid       : Message        :  Stencil 11.969 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669368
-Grid       : Message        : Average mflops/s per call per node       : 805939
-Grid       : Message        : Average mflops/s per call per node       : 827919
-Grid       : Message        : Average mflops/s per call per node       : 659033
-Grid       : Message        : Average mflops/s per call per node (full): 312956
-Grid       : Message        : Average mflops/s per call per node (full): 409279
-Grid       : Message        : Average mflops/s per call per node (full): 446149
-Grid       : Message        : Average mflops/s per call per node (full): 301303
-Grid       : Message        :  Stencil 12.7816 GB/s per node
-Grid       : Message        :  Stencil 17.2484 GB/s per node
-Grid       : Message        :  Stencil 17.5063 GB/s per node
-Grid       : Message        :  Stencil 14.1445 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666224
-Grid       : Message        : Average mflops/s per call per node       : 800794
-Grid       : Message        : Average mflops/s per call per node       : 824797
-Grid       : Message        : Average mflops/s per call per node       : 663407
-Grid       : Message        : Average mflops/s per call per node (full): 313092
-Grid       : Message        : Average mflops/s per call per node (full): 438551
-Grid       : Message        : Average mflops/s per call per node (full): 445019
-Grid       : Message        : Average mflops/s per call per node (full): 306984
-Grid       : Message        :  Stencil 12.5945 GB/s per node
-Grid       : Message        :  Stencil 17.3642 GB/s per node
-Grid       : Message        :  Stencil 16.9934 GB/s per node
-Grid       : Message        :  Stencil 11.7458 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669025
-Grid       : Message        : Average mflops/s per call per node       : 804992
-Grid       : Message        : Average mflops/s per call per node       : 823272
-Grid       : Message        : Average mflops/s per call per node       : 661218
-Grid       : Message        : Average mflops/s per call per node (full): 314220
-Grid       : Message        : Average mflops/s per call per node (full): 438383
-Grid       : Message        : Average mflops/s per call per node (full): 440220
-Grid       : Message        : Average mflops/s per call per node (full): 297043
-Grid       : Message        :  Stencil 13.5802 GB/s per node
-Grid       : Message        :  Stencil 16.4426 GB/s per node
-Grid       : Message        :  Stencil 18.345 GB/s per node
-Grid       : Message        :  Stencil 12.5696 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665125
-Grid       : Message        : Average mflops/s per call per node       : 803719
-Grid       : Message        : Average mflops/s per call per node       : 817192
-Grid       : Message        : Average mflops/s per call per node       : 663354
-Grid       : Message        : Average mflops/s per call per node (full): 315413
-Grid       : Message        : Average mflops/s per call per node (full): 433859
-Grid       : Message        : Average mflops/s per call per node (full): 446174
-Grid       : Message        : Average mflops/s per call per node (full): 303291
-Grid       : Message        :  Stencil 12.7315 GB/s per node
-Grid       : Message        :  Stencil 15.383 GB/s per node
-Grid       : Message        :  Stencil 18.1797 GB/s per node
-Grid       : Message        :  Stencil 13.1161 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666533
-Grid       : Message        : Average mflops/s per call per node       : 806042
-Grid       : Message        : Average mflops/s per call per node       : 821914
-Grid       : Message        : Average mflops/s per call per node       : 664713
-Grid       : Message        : Average mflops/s per call per node (full): 313225
-Grid       : Message        : Average mflops/s per call per node (full): 421118
-Grid       : Message        : Average mflops/s per call per node (full): 447010
-Grid       : Message        : Average mflops/s per call per node (full): 305894
-Grid       : Message        :  Stencil 12.81 GB/s per node
-Grid       : Message        :  Stencil 16.7324 GB/s per node
-Grid       : Message        :  Stencil 17.5909 GB/s per node
-Grid       : Message        :  Stencil 12.8213 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667673
-Grid       : Message        : Average mflops/s per call per node       : 806181
-Grid       : Message        : Average mflops/s per call per node       : 821357
-Grid       : Message        : Average mflops/s per call per node       : 667377
-Grid       : Message        : Average mflops/s per call per node (full): 315296
-Grid       : Message        : Average mflops/s per call per node (full): 436948
-Grid       : Message        : Average mflops/s per call per node (full): 445420
-Grid       : Message        : Average mflops/s per call per node (full): 305824
-Grid       : Message        :  Stencil 14.4771 GB/s per node
-Grid       : Message        :  Stencil 18.0131 GB/s per node
-Grid       : Message        :  Stencil 17.9633 GB/s per node
-Grid       : Message        :  Stencil 12.8198 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662482
-Grid       : Message        : Average mflops/s per call per node       : 800550
-Grid       : Message        : Average mflops/s per call per node       : 823363
-Grid       : Message        : Average mflops/s per call per node       : 655407
-Grid       : Message        : Average mflops/s per call per node (full): 314434
-Grid       : Message        : Average mflops/s per call per node (full): 438833
-Grid       : Message        : Average mflops/s per call per node (full): 438964
-Grid       : Message        : Average mflops/s per call per node (full): 303720
-Grid       : Message        :  Stencil 12.8066 GB/s per node
-Grid       : Message        :  Stencil 12.4512 GB/s per node
-Grid       : Message        :  Stencil 16.9279 GB/s per node
-Grid       : Message        :  Stencil 13.2672 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669351
-Grid       : Message        : Average mflops/s per call per node       : 808688
-Grid       : Message        : Average mflops/s per call per node       : 824319
-Grid       : Message        : Average mflops/s per call per node       : 662097
-Grid       : Message        : Average mflops/s per call per node (full): 314366
-Grid       : Message        : Average mflops/s per call per node (full): 370483
-Grid       : Message        : Average mflops/s per call per node (full): 439565
-Grid       : Message        : Average mflops/s per call per node (full): 305894
-Grid       : Message        :  Stencil 12.7226 GB/s per node
-Grid       : Message        :  Stencil 17.5536 GB/s per node
-Grid       : Message        :  Stencil 18.3302 GB/s per node
-Grid       : Message        :  Stencil 12.0157 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 673634
-Grid       : Message        : Average mflops/s per call per node       : 799870
-Grid       : Message        : Average mflops/s per call per node       : 825012
-Grid       : Message        : Average mflops/s per call per node       : 665898
-Grid       : Message        : Average mflops/s per call per node (full): 315377
-Grid       : Message        : Average mflops/s per call per node (full): 439524
-Grid       : Message        : Average mflops/s per call per node (full): 448642
-Grid       : Message        : Average mflops/s per call per node (full): 301330
-Grid       : Message        :  Stencil 15.665 GB/s per node
-Grid       : Message        :  Stencil 16.7653 GB/s per node
-Grid       : Message        :  Stencil 17.6713 GB/s per node
-Grid       : Message        :  Stencil 12.1773 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662461
-Grid       : Message        : Average mflops/s per call per node       : 803518
-Grid       : Message        : Average mflops/s per call per node       : 830364
-Grid       : Message        : Average mflops/s per call per node       : 665534
-Grid       : Message        : Average mflops/s per call per node (full): 317567
-Grid       : Message        : Average mflops/s per call per node (full): 436654
-Grid       : Message        : Average mflops/s per call per node (full): 445703
-Grid       : Message        : Average mflops/s per call per node (full): 303153
-Grid       : Message        :  Stencil 14.7891 GB/s per node
-Grid       : Message        :  Stencil 16.8005 GB/s per node
-Grid       : Message        :  Stencil 17.7006 GB/s per node
-Grid       : Message        :  Stencil 12.0703 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660523
-Grid       : Message        : Average mflops/s per call per node       : 801552
-Grid       : Message        : Average mflops/s per call per node       : 819377
-Grid       : Message        : Average mflops/s per call per node       : 668934
-Grid       : Message        : Average mflops/s per call per node (full): 315578
-Grid       : Message        : Average mflops/s per call per node (full): 432654
-Grid       : Message        : Average mflops/s per call per node (full): 445646
-Grid       : Message        : Average mflops/s per call per node (full): 302725
-Grid       : Message        :  Stencil 12.4778 GB/s per node
-Grid       : Message        :  Stencil 17.3503 GB/s per node
-Grid       : Message        :  Stencil 17.3418 GB/s per node
-Grid       : Message        :  Stencil 12.9612 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669193
-Grid       : Message        : Average mflops/s per call per node       : 803618
-Grid       : Message        : Average mflops/s per call per node       : 818478
-Grid       : Message        : Average mflops/s per call per node       : 659136
-Grid       : Message        : Average mflops/s per call per node (full): 313641
-Grid       : Message        : Average mflops/s per call per node (full): 437523
-Grid       : Message        : Average mflops/s per call per node (full): 442583
-Grid       : Message        : Average mflops/s per call per node (full): 305080
-Grid       : Message        :  Stencil 12.5965 GB/s per node
-Grid       : Message        :  Stencil 16.2825 GB/s per node
-Grid       : Message        :  Stencil 17.1698 GB/s per node
-Grid       : Message        :  Stencil 12.9088 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668568
-Grid       : Message        : Average mflops/s per call per node       : 809293
-Grid       : Message        : Average mflops/s per call per node       : 820792
-Grid       : Message        : Average mflops/s per call per node       : 658146
-Grid       : Message        : Average mflops/s per call per node (full): 313932
-Grid       : Message        : Average mflops/s per call per node (full): 432644
-Grid       : Message        : Average mflops/s per call per node (full): 442291
-Grid       : Message        : Average mflops/s per call per node (full): 302658
-Grid       : Message        :  Stencil 12.7875 GB/s per node
-Grid       : Message        :  Stencil 17.3888 GB/s per node
-Grid       : Message        :  Stencil 18.2178 GB/s per node
-Grid       : Message        :  Stencil 15.7381 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666964
-Grid       : Message        : Average mflops/s per call per node       : 795974
-Grid       : Message        : Average mflops/s per call per node       : 822389
-Grid       : Message        : Average mflops/s per call per node       : 658799
-Grid       : Message        : Average mflops/s per call per node (full): 314510
-Grid       : Message        : Average mflops/s per call per node (full): 436541
-Grid       : Message        : Average mflops/s per call per node (full): 448418
-Grid       : Message        : Average mflops/s per call per node (full): 307398
-Grid       : Message        :  Stencil 13.0569 GB/s per node
-Grid       : Message        :  Stencil 18.0567 GB/s per node
-Grid       : Message        :  Stencil 18.5102 GB/s per node
-Grid       : Message        :  Stencil 12.0611 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666304
-Grid       : Message        : Average mflops/s per call per node       : 798361
-Grid       : Message        : Average mflops/s per call per node       : 823575
-Grid       : Message        : Average mflops/s per call per node       : 667601
-Grid       : Message        : Average mflops/s per call per node (full): 315659
-Grid       : Message        : Average mflops/s per call per node (full): 438070
-Grid       : Message        : Average mflops/s per call per node (full): 445969
-Grid       : Message        : Average mflops/s per call per node (full): 302695
-Grid       : Message        :  Stencil 13.1052 GB/s per node
-Grid       : Message        :  Stencil 16.5789 GB/s per node
-Grid       : Message        :  Stencil 16.905 GB/s per node
-Grid       : Message        :  Stencil 12.5662 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666859
-Grid       : Message        : Average mflops/s per call per node       : 803782
-Grid       : Message        : Average mflops/s per call per node       : 821719
-Grid       : Message        : Average mflops/s per call per node       : 667919
-Grid       : Message        : Average mflops/s per call per node (full): 315018
-Grid       : Message        : Average mflops/s per call per node (full): 435319
-Grid       : Message        : Average mflops/s per call per node (full): 439161
-Grid       : Message        : Average mflops/s per call per node (full): 305586
-Grid       : Message        :  Stencil 12.6053 GB/s per node
-Grid       : Message        :  Stencil 16.9889 GB/s per node
-Grid       : Message        :  Stencil 17.299 GB/s per node
-Grid       : Message        :  Stencil 13.3575 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668441
-Grid       : Message        : Average mflops/s per call per node       : 800242
-Grid       : Message        : Average mflops/s per call per node       : 813284
-Grid       : Message        : Average mflops/s per call per node       : 663141
-Grid       : Message        : Average mflops/s per call per node (full): 314179
-Grid       : Message        : Average mflops/s per call per node (full): 436227
-Grid       : Message        : Average mflops/s per call per node (full): 442472
-Grid       : Message        : Average mflops/s per call per node (full): 305967
-Grid       : Message        :  Stencil 12.9178 GB/s per node
-Grid       : Message        :  Stencil 16.6625 GB/s per node
-Grid       : Message        :  Stencil 17.7502 GB/s per node
-Grid       : Message        :  Stencil 13.6215 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669848
-Grid       : Message        : Average mflops/s per call per node       : 805128
-Grid       : Message        : Average mflops/s per call per node       : 819001
-Grid       : Message        : Average mflops/s per call per node       : 664476
-Grid       : Message        : Average mflops/s per call per node (full): 316078
-Grid       : Message        : Average mflops/s per call per node (full): 435852
-Grid       : Message        : Average mflops/s per call per node (full): 445856
-Grid       : Message        : Average mflops/s per call per node (full): 305997
-Grid       : Message        :  Stencil 12.7647 GB/s per node
-Grid       : Message        :  Stencil 16.037 GB/s per node
-Grid       : Message        :  Stencil 17.6307 GB/s per node
-Grid       : Message        :  Stencil 12.981 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668508
-Grid       : Message        : Average mflops/s per call per node       : 807700
-Grid       : Message        : Average mflops/s per call per node       : 823565
-Grid       : Message        : Average mflops/s per call per node       : 659680
-Grid       : Message        : Average mflops/s per call per node (full): 314931
-Grid       : Message        : Average mflops/s per call per node (full): 428598
-Grid       : Message        : Average mflops/s per call per node (full): 445601
-Grid       : Message        : Average mflops/s per call per node (full): 305267
-Grid       : Message        :  Stencil 12.7174 GB/s per node
-Grid       : Message        :  Stencil 15.3893 GB/s per node
-Grid       : Message        :  Stencil 17.7699 GB/s per node
-Grid       : Message        :  Stencil 14.1334 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668861
-Grid       : Message        : Average mflops/s per call per node       : 808645
-Grid       : Message        : Average mflops/s per call per node       : 818086
-Grid       : Message        : Average mflops/s per call per node       : 665203
-Grid       : Message        : Average mflops/s per call per node (full): 314695
-Grid       : Message        : Average mflops/s per call per node (full): 421627
-Grid       : Message        : Average mflops/s per call per node (full): 438315
-Grid       : Message        : Average mflops/s per call per node (full): 307073
-Grid       : Message        :  Stencil 13.5612 GB/s per node
-Grid       : Message        :  Stencil 14.3094 GB/s per node
-Grid       : Message        :  Stencil 17.7979 GB/s per node
-Grid       : Message        :  Stencil 13.7627 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664046
-Grid       : Message        : Average mflops/s per call per node       : 804608
-Grid       : Message        : Average mflops/s per call per node       : 822177
-Grid       : Message        : Average mflops/s per call per node       : 661169
-Grid       : Message        : Average mflops/s per call per node (full): 314385
-Grid       : Message        : Average mflops/s per call per node (full): 401651
-Grid       : Message        : Average mflops/s per call per node (full): 443230
-Grid       : Message        : Average mflops/s per call per node (full): 304789
-Grid       : Message        :  Stencil 13.6262 GB/s per node
-Grid       : Message        :  Stencil 16.5029 GB/s per node
-Grid       : Message        :  Stencil 18.9915 GB/s per node
-Grid       : Message        :  Stencil 11.9887 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669539
-Grid       : Message        : Average mflops/s per call per node       : 802650
-Grid       : Message        : Average mflops/s per call per node       : 819077
-Grid       : Message        : Average mflops/s per call per node       : 666765
-Grid       : Message        : Average mflops/s per call per node (full): 317347
-Grid       : Message        : Average mflops/s per call per node (full): 434704
-Grid       : Message        : Average mflops/s per call per node (full): 449241
-Grid       : Message        : Average mflops/s per call per node (full): 302354
-Grid       : Message        :  Stencil 13.9588 GB/s per node
-Grid       : Message        :  Stencil 12.2464 GB/s per node
-Grid       : Message        :  Stencil 18.1376 GB/s per node
-Grid       : Message        :  Stencil 12.7738 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666887
-Grid       : Message        : Average mflops/s per call per node       : 815235
-Grid       : Message        : Average mflops/s per call per node       : 820800
-Grid       : Message        : Average mflops/s per call per node       : 672403
-Grid       : Message        : Average mflops/s per call per node (full): 316395
-Grid       : Message        : Average mflops/s per call per node (full): 366349
-Grid       : Message        : Average mflops/s per call per node (full): 447028
-Grid       : Message        : Average mflops/s per call per node (full): 305605
-Grid       : Message        :  Stencil 13.3593 GB/s per node
-Grid       : Message        :  Stencil 16.3731 GB/s per node
-Grid       : Message        :  Stencil 16.966 GB/s per node
-Grid       : Message        :  Stencil 12.9026 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667752
-Grid       : Message        : Average mflops/s per call per node       : 805011
-Grid       : Message        : Average mflops/s per call per node       : 824621
-Grid       : Message        : Average mflops/s per call per node       : 664981
-Grid       : Message        : Average mflops/s per call per node (full): 315591
-Grid       : Message        : Average mflops/s per call per node (full): 429557
-Grid       : Message        : Average mflops/s per call per node (full): 441140
-Grid       : Message        : Average mflops/s per call per node (full): 304634
-Grid       : Message        :  Stencil 14.6674 GB/s per node
-Grid       : Message        :  Stencil 16.6037 GB/s per node
-Grid       : Message        :  Stencil 18.8243 GB/s per node
-Grid       : Message        :  Stencil 13.0788 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662330
-Grid       : Message        : Average mflops/s per call per node       : 800743
-Grid       : Message        : Average mflops/s per call per node       : 822556
-Grid       : Message        : Average mflops/s per call per node       : 662263
-Grid       : Message        : Average mflops/s per call per node (full): 316637
-Grid       : Message        : Average mflops/s per call per node (full): 435866
-Grid       : Message        : Average mflops/s per call per node (full): 449299
-Grid       : Message        : Average mflops/s per call per node (full): 305056
-Grid       : Message        :  Stencil 14.3719 GB/s per node
-Grid       : Message        :  Stencil 16.6864 GB/s per node
-Grid       : Message        :  Stencil 16.5486 GB/s per node
-Grid       : Message        :  Stencil 12.9351 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661735
-Grid       : Message        : Average mflops/s per call per node       : 804615
-Grid       : Message        : Average mflops/s per call per node       : 816724
-Grid       : Message        : Average mflops/s per call per node       : 666066
-Grid       : Message        : Average mflops/s per call per node (full): 316431
-Grid       : Message        : Average mflops/s per call per node (full): 435436
-Grid       : Message        : Average mflops/s per call per node (full): 429896
-Grid       : Message        : Average mflops/s per call per node (full): 305633
-Grid       : Message        :  Stencil 13.8614 GB/s per node
-Grid       : Message        :  Stencil 16.7359 GB/s per node
-Grid       : Message        :  Stencil 17.6222 GB/s per node
-Grid       : Message        :  Stencil 12.787 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663609
-Grid       : Message        : Average mflops/s per call per node       : 802481
-Grid       : Message        : Average mflops/s per call per node       : 822284
-Grid       : Message        : Average mflops/s per call per node       : 663024
-Grid       : Message        : Average mflops/s per call per node (full): 316361
-Grid       : Message        : Average mflops/s per call per node (full): 436671
-Grid       : Message        : Average mflops/s per call per node (full): 445878
-Grid       : Message        : Average mflops/s per call per node (full): 304683
-Grid       : Message        :  Stencil 12.3693 GB/s per node
-Grid       : Message        :  Stencil 17.9277 GB/s per node
-Grid       : Message        :  Stencil 17.857 GB/s per node
-Grid       : Message        :  Stencil 12.4024 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666807
-Grid       : Message        : Average mflops/s per call per node       : 799770
-Grid       : Message        : Average mflops/s per call per node       : 814899
-Grid       : Message        : Average mflops/s per call per node       : 664095
-Grid       : Message        : Average mflops/s per call per node (full): 310994
-Grid       : Message        : Average mflops/s per call per node (full): 440054
-Grid       : Message        : Average mflops/s per call per node (full): 444183
-Grid       : Message        : Average mflops/s per call per node (full): 303609
-Grid       : Message        :  Stencil 12.9655 GB/s per node
-Grid       : Message        :  Stencil 16.8382 GB/s per node
-Grid       : Message        :  Stencil 17.7169 GB/s per node
-Grid       : Message        :  Stencil 12.5815 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667126
-Grid       : Message        : Average mflops/s per call per node       : 804406
-Grid       : Message        : Average mflops/s per call per node       : 824484
-Grid       : Message        : Average mflops/s per call per node       : 668240
-Grid       : Message        : Average mflops/s per call per node (full): 315412
-Grid       : Message        : Average mflops/s per call per node (full): 437094
-Grid       : Message        : Average mflops/s per call per node (full): 445690
-Grid       : Message        : Average mflops/s per call per node (full): 304519
-Grid       : Message        :  Stencil 12.4799 GB/s per node
-Grid       : Message        :  Stencil 17.4822 GB/s per node
-Grid       : Message        :  Stencil 17.1568 GB/s per node
-Grid       : Message        :  Stencil 13.2932 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 672305
-Grid       : Message        : Average mflops/s per call per node       : 800954
-Grid       : Message        : Average mflops/s per call per node       : 822328
-Grid       : Message        : Average mflops/s per call per node       : 669005
-Grid       : Message        : Average mflops/s per call per node (full): 313890
-Grid       : Message        : Average mflops/s per call per node (full): 439128
-Grid       : Message        : Average mflops/s per call per node (full): 442636
-Grid       : Message        : Average mflops/s per call per node (full): 306782
-Grid       : Message        :  Stencil 12.5206 GB/s per node
-Grid       : Message        :  Stencil 16.6807 GB/s per node
-Grid       : Message        :  Stencil 18.4892 GB/s per node
-Grid       : Message        :  Stencil 12.5474 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670051
-Grid       : Message        : Average mflops/s per call per node       : 803701
-Grid       : Message        : Average mflops/s per call per node       : 814013
-Grid       : Message        : Average mflops/s per call per node       : 662089
-Grid       : Message        : Average mflops/s per call per node (full): 313988
-Grid       : Message        : Average mflops/s per call per node (full): 434021
-Grid       : Message        : Average mflops/s per call per node (full): 446562
-Grid       : Message        : Average mflops/s per call per node (full): 300483
-Grid       : Message        :  Stencil 12.6668 GB/s per node
-Grid       : Message        :  Stencil 13.4979 GB/s per node
-Grid       : Message        :  Stencil 18.2685 GB/s per node
-Grid       : Message        :  Stencil 13.2324 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670057
-Grid       : Message        : Average mflops/s per call per node       : 810582
-Grid       : Message        : Average mflops/s per call per node       : 819467
-Grid       : Message        : Average mflops/s per call per node       : 664962
-Grid       : Message        : Average mflops/s per call per node (full): 314212
-Grid       : Message        : Average mflops/s per call per node (full): 390848
-Grid       : Message        : Average mflops/s per call per node (full): 446832
-Grid       : Message        : Average mflops/s per call per node (full): 306389
-Grid       : Message        :  Stencil 13.4024 GB/s per node
-Grid       : Message        :  Stencil 17.8795 GB/s per node
-Grid       : Message        :  Stencil 17.4326 GB/s per node
-Grid       : Message        :  Stencil 14.4151 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666151
-Grid       : Message        : Average mflops/s per call per node       : 804497
-Grid       : Message        : Average mflops/s per call per node       : 826075
-Grid       : Message        : Average mflops/s per call per node       : 661044
-Grid       : Message        : Average mflops/s per call per node (full): 315120
-Grid       : Message        : Average mflops/s per call per node (full): 439813
-Grid       : Message        : Average mflops/s per call per node (full): 445667
-Grid       : Message        : Average mflops/s per call per node (full): 307218
-Grid       : Message        :  Stencil 13.5361 GB/s per node
-Grid       : Message        :  Stencil 17.1346 GB/s per node
-Grid       : Message        :  Stencil 18.1014 GB/s per node
-Grid       : Message        :  Stencil 12.8291 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669046
-Grid       : Message        : Average mflops/s per call per node       : 808504
-Grid       : Message        : Average mflops/s per call per node       : 818995
-Grid       : Message        : Average mflops/s per call per node       : 662119
-Grid       : Message        : Average mflops/s per call per node (full): 315850
-Grid       : Message        : Average mflops/s per call per node (full): 438213
-Grid       : Message        : Average mflops/s per call per node (full): 446359
-Grid       : Message        : Average mflops/s per call per node (full): 306122
-Grid       : Message        :  Stencil 13.3801 GB/s per node
-Grid       : Message        :  Stencil 16.8594 GB/s per node
-Grid       : Message        :  Stencil 17.7869 GB/s per node
-Grid       : Message        :  Stencil 12.5915 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668371
-Grid       : Message        : Average mflops/s per call per node       : 804653
-Grid       : Message        : Average mflops/s per call per node       : 824526
-Grid       : Message        : Average mflops/s per call per node       : 667430
-Grid       : Message        : Average mflops/s per call per node (full): 315682
-Grid       : Message        : Average mflops/s per call per node (full): 437676
-Grid       : Message        : Average mflops/s per call per node (full): 446980
-Grid       : Message        : Average mflops/s per call per node (full): 304992
-Grid       : Message        :  Stencil 13.7549 GB/s per node
-Grid       : Message        :  Stencil 14.9992 GB/s per node
-Grid       : Message        :  Stencil 17.5302 GB/s per node
-Grid       : Message        :  Stencil 12.2735 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666096
-Grid       : Message        : Average mflops/s per call per node       : 811499
-Grid       : Message        : Average mflops/s per call per node       : 824694
-Grid       : Message        : Average mflops/s per call per node       : 668519
-Grid       : Message        : Average mflops/s per call per node (full): 316858
-Grid       : Message        : Average mflops/s per call per node (full): 414059
-Grid       : Message        : Average mflops/s per call per node (full): 445936
-Grid       : Message        : Average mflops/s per call per node (full): 303533
-Grid       : Message        :  Stencil 13.6632 GB/s per node
-Grid       : Message        :  Stencil 17.0763 GB/s per node
-Grid       : Message        :  Stencil 17.3609 GB/s per node
-Grid       : Message        :  Stencil 11.946 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664469
-Grid       : Message        : Average mflops/s per call per node       : 804234
-Grid       : Message        : Average mflops/s per call per node       : 823660
-Grid       : Message        : Average mflops/s per call per node       : 665986
-Grid       : Message        : Average mflops/s per call per node (full): 313506
-Grid       : Message        : Average mflops/s per call per node (full): 436407
-Grid       : Message        : Average mflops/s per call per node (full): 445089
-Grid       : Message        : Average mflops/s per call per node (full): 301375
-Grid       : Message        :  Stencil 13.127 GB/s per node
-Grid       : Message        :  Stencil 16.4242 GB/s per node
-Grid       : Message        :  Stencil 17.4365 GB/s per node
-Grid       : Message        :  Stencil 13.6282 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670281
-Grid       : Message        : Average mflops/s per call per node       : 805435
-Grid       : Message        : Average mflops/s per call per node       : 822530
-Grid       : Message        : Average mflops/s per call per node       : 659966
-Grid       : Message        : Average mflops/s per call per node (full): 315917
-Grid       : Message        : Average mflops/s per call per node (full): 434543
-Grid       : Message        : Average mflops/s per call per node (full): 441755
-Grid       : Message        : Average mflops/s per call per node (full): 306428
-Grid       : Message        :  Stencil 13.2015 GB/s per node
-Grid       : Message        :  Stencil 17.5169 GB/s per node
-Grid       : Message        :  Stencil 18 GB/s per node
-Grid       : Message        :  Stencil 12.8812 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666946
-Grid       : Message        : Average mflops/s per call per node       : 804926
-Grid       : Message        : Average mflops/s per call per node       : 821508
-Grid       : Message        : Average mflops/s per call per node       : 665034
-Grid       : Message        : Average mflops/s per call per node (full): 313490
-Grid       : Message        : Average mflops/s per call per node (full): 438264
-Grid       : Message        : Average mflops/s per call per node (full): 446763
-Grid       : Message        : Average mflops/s per call per node (full): 302935
-Grid       : Message        :  Stencil 13.3755 GB/s per node
-Grid       : Message        :  Stencil 17.2901 GB/s per node
-Grid       : Message        :  Stencil 18.1233 GB/s per node
-Grid       : Message        :  Stencil 13.465 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668938
-Grid       : Message        : Average mflops/s per call per node       : 804972
-Grid       : Message        : Average mflops/s per call per node       : 822896
-Grid       : Message        : Average mflops/s per call per node       : 663745
-Grid       : Message        : Average mflops/s per call per node (full): 316147
-Grid       : Message        : Average mflops/s per call per node (full): 438113
-Grid       : Message        : Average mflops/s per call per node (full): 446562
-Grid       : Message        : Average mflops/s per call per node (full): 306549
-Grid       : Message        :  Stencil 14.3151 GB/s per node
-Grid       : Message        :  Stencil 18.0712 GB/s per node
-Grid       : Message        :  Stencil 19.3206 GB/s per node
-Grid       : Message        :  Stencil 13.0481 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662888
-Grid       : Message        : Average mflops/s per call per node       : 803274
-Grid       : Message        : Average mflops/s per call per node       : 821731
-Grid       : Message        : Average mflops/s per call per node       : 666168
-Grid       : Message        : Average mflops/s per call per node (full): 315353
-Grid       : Message        : Average mflops/s per call per node (full): 442293
-Grid       : Message        : Average mflops/s per call per node (full): 449390
-Grid       : Message        : Average mflops/s per call per node (full): 305695
-Grid       : Message        :  Stencil 13.6391 GB/s per node
-Grid       : Message        :  Stencil 16.6396 GB/s per node
-Grid       : Message        :  Stencil 18.4333 GB/s per node
-Grid       : Message        :  Stencil 13.2398 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663417
-Grid       : Message        : Average mflops/s per call per node       : 808102
-Grid       : Message        : Average mflops/s per call per node       : 820793
-Grid       : Message        : Average mflops/s per call per node       : 665197
-Grid       : Message        : Average mflops/s per call per node (full): 314956
-Grid       : Message        : Average mflops/s per call per node (full): 436651
-Grid       : Message        : Average mflops/s per call per node (full): 448931
-Grid       : Message        : Average mflops/s per call per node (full): 303869
-Grid       : Message        :  Stencil 12.8134 GB/s per node
-Grid       : Message        :  Stencil 16.688 GB/s per node
-Grid       : Message        :  Stencil 17.268 GB/s per node
-Grid       : Message        :  Stencil 13.5131 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667084
-Grid       : Message        : Average mflops/s per call per node       : 801652
-Grid       : Message        : Average mflops/s per call per node       : 818306
-Grid       : Message        : Average mflops/s per call per node       : 664899
-Grid       : Message        : Average mflops/s per call per node (full): 313861
-Grid       : Message        : Average mflops/s per call per node (full): 435206
-Grid       : Message        : Average mflops/s per call per node (full): 443181
-Grid       : Message        : Average mflops/s per call per node (full): 306460
-Grid       : Message        :  Stencil 13.1928 GB/s per node
-Grid       : Message        :  Stencil 16.9779 GB/s per node
-Grid       : Message        :  Stencil 18.3528 GB/s per node
-Grid       : Message        :  Stencil 13.6895 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665268
-Grid       : Message        : Average mflops/s per call per node       : 801351
-Grid       : Message        : Average mflops/s per call per node       : 820825
-Grid       : Message        : Average mflops/s per call per node       : 663334
-Grid       : Message        : Average mflops/s per call per node (full): 314589
-Grid       : Message        : Average mflops/s per call per node (full): 437686
-Grid       : Message        : Average mflops/s per call per node (full): 447820
-Grid       : Message        : Average mflops/s per call per node (full): 306741
-Grid       : Message        :  Stencil 14.5065 GB/s per node
-Grid       : Message        :  Stencil 16.8067 GB/s per node
-Grid       : Message        :  Stencil 18.635 GB/s per node
-Grid       : Message        :  Stencil 12.8957 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662023
-Grid       : Message        : Average mflops/s per call per node       : 802875
-Grid       : Message        : Average mflops/s per call per node       : 817871
-Grid       : Message        : Average mflops/s per call per node       : 664987
-Grid       : Message        : Average mflops/s per call per node (full): 316621
-Grid       : Message        : Average mflops/s per call per node (full): 435212
-Grid       : Message        : Average mflops/s per call per node (full): 446336
-Grid       : Message        : Average mflops/s per call per node (full): 304657
-Grid       : Message        :  Stencil 12.4429 GB/s per node
-Grid       : Message        :  Stencil 16.5042 GB/s per node
-Grid       : Message        :  Stencil 17.7928 GB/s per node
-Grid       : Message        :  Stencil 12.6474 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669273
-Grid       : Message        : Average mflops/s per call per node       : 801563
-Grid       : Message        : Average mflops/s per call per node       : 821917
-Grid       : Message        : Average mflops/s per call per node       : 667029
-Grid       : Message        : Average mflops/s per call per node (full): 312870
-Grid       : Message        : Average mflops/s per call per node (full): 433644
-Grid       : Message        : Average mflops/s per call per node (full): 447043
-Grid       : Message        : Average mflops/s per call per node (full): 305892
-Grid       : Message        :  Stencil 11.3677 GB/s per node
-Grid       : Message        :  Stencil 16.7617 GB/s per node
-Grid       : Message        :  Stencil 19.4241 GB/s per node
-Grid       : Message        :  Stencil 12.6271 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662449
-Grid       : Message        : Average mflops/s per call per node       : 800926
-Grid       : Message        : Average mflops/s per call per node       : 820710
-Grid       : Message        : Average mflops/s per call per node       : 663453
-Grid       : Message        : Average mflops/s per call per node (full): 303543
-Grid       : Message        : Average mflops/s per call per node (full): 435451
-Grid       : Message        : Average mflops/s per call per node (full): 449223
-Grid       : Message        : Average mflops/s per call per node (full): 303758
-Grid       : Message        :  Stencil 12.2592 GB/s per node
-Grid       : Message        :  Stencil 17.931 GB/s per node
-Grid       : Message        :  Stencil 18.5877 GB/s per node
-Grid       : Message        :  Stencil 13.3924 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668599
-Grid       : Message        : Average mflops/s per call per node       : 798325
-Grid       : Message        : Average mflops/s per call per node       : 821788
-Grid       : Message        : Average mflops/s per call per node       : 665107
-Grid       : Message        : Average mflops/s per call per node (full): 311819
-Grid       : Message        : Average mflops/s per call per node (full): 439804
-Grid       : Message        : Average mflops/s per call per node (full): 449096
-Grid       : Message        : Average mflops/s per call per node (full): 306771
-Grid       : Message        :  Stencil 13.4397 GB/s per node
-Grid       : Message        :  Stencil 18.4553 GB/s per node
-Grid       : Message        :  Stencil 17.2144 GB/s per node
-Grid       : Message        :  Stencil 13.4523 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663828
-Grid       : Message        : Average mflops/s per call per node       : 803533
-Grid       : Message        : Average mflops/s per call per node       : 829133
-Grid       : Message        : Average mflops/s per call per node       : 667575
-Grid       : Message        : Average mflops/s per call per node (full): 315880
-Grid       : Message        : Average mflops/s per call per node (full): 442364
-Grid       : Message        : Average mflops/s per call per node (full): 444101
-Grid       : Message        : Average mflops/s per call per node (full): 306812
-Grid       : Message        :  Stencil 13.4004 GB/s per node
-Grid       : Message        :  Stencil 16.9872 GB/s per node
-Grid       : Message        :  Stencil 19.2425 GB/s per node
-Grid       : Message        :  Stencil 12.3307 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666260
-Grid       : Message        : Average mflops/s per call per node       : 805756
-Grid       : Message        : Average mflops/s per call per node       : 819795
-Grid       : Message        : Average mflops/s per call per node       : 665190
-Grid       : Message        : Average mflops/s per call per node (full): 314488
-Grid       : Message        : Average mflops/s per call per node (full): 439238
-Grid       : Message        : Average mflops/s per call per node (full): 449947
-Grid       : Message        : Average mflops/s per call per node (full): 304314
-Grid       : Message        :  Stencil 13.2794 GB/s per node
-Grid       : Message        :  Stencil 16.3363 GB/s per node
-Grid       : Message        :  Stencil 17.4506 GB/s per node
-Grid       : Message        :  Stencil 11.8973 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661869
-Grid       : Message        : Average mflops/s per call per node       : 802011
-Grid       : Message        : Average mflops/s per call per node       : 821374
-Grid       : Message        : Average mflops/s per call per node       : 667753
-Grid       : Message        : Average mflops/s per call per node (full): 314151
-Grid       : Message        : Average mflops/s per call per node (full): 432113
-Grid       : Message        : Average mflops/s per call per node (full): 444162
-Grid       : Message        : Average mflops/s per call per node (full): 300812
-Grid       : Message        :  Stencil 13.1259 GB/s per node
-Grid       : Message        :  Stencil 16.3224 GB/s per node
-Grid       : Message        :  Stencil 16.8754 GB/s per node
-Grid       : Message        :  Stencil 12.6539 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667421
-Grid       : Message        : Average mflops/s per call per node       : 800472
-Grid       : Message        : Average mflops/s per call per node       : 823593
-Grid       : Message        : Average mflops/s per call per node       : 662618
-Grid       : Message        : Average mflops/s per call per node (full): 314562
-Grid       : Message        : Average mflops/s per call per node (full): 426645
-Grid       : Message        : Average mflops/s per call per node (full): 434872
-Grid       : Message        : Average mflops/s per call per node (full): 303988
-Grid       : Message        :  Stencil 12.3184 GB/s per node
-Grid       : Message        :  Stencil 17.8475 GB/s per node
-Grid       : Message        :  Stencil 17.5751 GB/s per node
-Grid       : Message        :  Stencil 12.3613 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664840
-Grid       : Message        : Average mflops/s per call per node       : 804370
-Grid       : Message        : Average mflops/s per call per node       : 815424
-Grid       : Message        : Average mflops/s per call per node       : 664303
-Grid       : Message        : Average mflops/s per call per node (full): 311918
-Grid       : Message        : Average mflops/s per call per node (full): 438836
-Grid       : Message        : Average mflops/s per call per node (full): 443442
-Grid       : Message        : Average mflops/s per call per node (full): 303934
-Grid       : Message        :  Stencil 13.7473 GB/s per node
-Grid       : Message        :  Stencil 11.9221 GB/s per node
-Grid       : Message        :  Stencil 17.4339 GB/s per node
-Grid       : Message        :  Stencil 13.3121 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662807
-Grid       : Message        : Average mflops/s per call per node       : 806827
-Grid       : Message        : Average mflops/s per call per node       : 819802
-Grid       : Message        : Average mflops/s per call per node       : 662039
-Grid       : Message        : Average mflops/s per call per node (full): 315933
-Grid       : Message        : Average mflops/s per call per node (full): 359834
-Grid       : Message        : Average mflops/s per call per node (full): 443439
-Grid       : Message        : Average mflops/s per call per node (full): 305833
-Grid       : Message        :  Stencil 12.6701 GB/s per node
-Grid       : Message        :  Stencil 16.6586 GB/s per node
-Grid       : Message        :  Stencil 16.9694 GB/s per node
-Grid       : Message        :  Stencil 12.4529 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666476
-Grid       : Message        : Average mflops/s per call per node       : 804596
-Grid       : Message        : Average mflops/s per call per node       : 824614
-Grid       : Message        : Average mflops/s per call per node       : 666724
-Grid       : Message        : Average mflops/s per call per node (full): 314720
-Grid       : Message        : Average mflops/s per call per node (full): 431010
-Grid       : Message        : Average mflops/s per call per node (full): 430781
-Grid       : Message        : Average mflops/s per call per node (full): 304274
-Grid       : Message        :  Stencil 13.8334 GB/s per node
-Grid       : Message        :  Stencil 17.5333 GB/s per node
-Grid       : Message        :  Stencil 17.2497 GB/s per node
-Grid       : Message        :  Stencil 14.9046 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664783
-Grid       : Message        : Average mflops/s per call per node       : 805252
-Grid       : Message        : Average mflops/s per call per node       : 820404
-Grid       : Message        : Average mflops/s per call per node       : 655310
-Grid       : Message        : Average mflops/s per call per node (full): 315375
-Grid       : Message        : Average mflops/s per call per node (full): 440397
-Grid       : Message        : Average mflops/s per call per node (full): 442864
-Grid       : Message        : Average mflops/s per call per node (full): 305982
-Grid       : Message        :  Stencil 12.6778 GB/s per node
-Grid       : Message        :  Stencil 16.9202 GB/s per node
-Grid       : Message        :  Stencil 17.2772 GB/s per node
-Grid       : Message        :  Stencil 13.485 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666072
-Grid       : Message        : Average mflops/s per call per node       : 806807
-Grid       : Message        : Average mflops/s per call per node       : 825291
-Grid       : Message        : Average mflops/s per call per node       : 664838
-Grid       : Message        : Average mflops/s per call per node (full): 314502
-Grid       : Message        : Average mflops/s per call per node (full): 437343
-Grid       : Message        : Average mflops/s per call per node (full): 443630
-Grid       : Message        : Average mflops/s per call per node (full): 306984
-Grid       : Message        :  Stencil 13.5258 GB/s per node
-Grid       : Message        :  Stencil 17.4691 GB/s per node
-Grid       : Message        :  Stencil 17.4799 GB/s per node
-Grid       : Message        :  Stencil 12.612 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663070
-Grid       : Message        : Average mflops/s per call per node       : 796856
-Grid       : Message        : Average mflops/s per call per node       : 825192
-Grid       : Message        : Average mflops/s per call per node       : 667299
-Grid       : Message        : Average mflops/s per call per node (full): 316114
-Grid       : Message        : Average mflops/s per call per node (full): 438528
-Grid       : Message        : Average mflops/s per call per node (full): 444614
-Grid       : Message        : Average mflops/s per call per node (full): 305545
-Grid       : Message        :  Stencil 12.8092 GB/s per node
-Grid       : Message        :  Stencil 10.023 GB/s per node
-Grid       : Message        :  Stencil 17.95 GB/s per node
-Grid       : Message        :  Stencil 12.5901 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666850
-Grid       : Message        : Average mflops/s per call per node       : 808491
-Grid       : Message        : Average mflops/s per call per node       : 821557
-Grid       : Message        : Average mflops/s per call per node       : 666166
-Grid       : Message        : Average mflops/s per call per node (full): 315060
-Grid       : Message        : Average mflops/s per call per node (full): 316803
-Grid       : Message        : Average mflops/s per call per node (full): 444537
-Grid       : Message        : Average mflops/s per call per node (full): 304435
-Grid       : Message        :  Stencil 13.6691 GB/s per node
-Grid       : Message        :  Stencil 17.0066 GB/s per node
-Grid       : Message        :  Stencil 17.4481 GB/s per node
-Grid       : Message        :  Stencil 13.2227 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665356
-Grid       : Message        : Average mflops/s per call per node       : 799926
-Grid       : Message        : Average mflops/s per call per node       : 820032
-Grid       : Message        : Average mflops/s per call per node       : 664401
-Grid       : Message        : Average mflops/s per call per node (full): 316276
-Grid       : Message        : Average mflops/s per call per node (full): 434571
-Grid       : Message        : Average mflops/s per call per node (full): 444560
-Grid       : Message        : Average mflops/s per call per node (full): 306388
-Grid       : Message        :  Stencil 13.0078 GB/s per node
-Grid       : Message        :  Stencil 14.0913 GB/s per node
-Grid       : Message        :  Stencil 18.2835 GB/s per node
-Grid       : Message        :  Stencil 13.4152 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664674
-Grid       : Message        : Average mflops/s per call per node       : 808509
-Grid       : Message        : Average mflops/s per call per node       : 819343
-Grid       : Message        : Average mflops/s per call per node       : 665803
-Grid       : Message        : Average mflops/s per call per node (full): 314690
-Grid       : Message        : Average mflops/s per call per node (full): 401452
-Grid       : Message        : Average mflops/s per call per node (full): 447251
-Grid       : Message        : Average mflops/s per call per node (full): 306926
-Grid       : Message        :  Stencil 14.6221 GB/s per node
-Grid       : Message        :  Stencil 16.6677 GB/s per node
-Grid       : Message        :  Stencil 17.582 GB/s per node
-Grid       : Message        :  Stencil 12.2818 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659245
-Grid       : Message        : Average mflops/s per call per node       : 804855
-Grid       : Message        : Average mflops/s per call per node       : 819504
-Grid       : Message        : Average mflops/s per call per node       : 667461
-Grid       : Message        : Average mflops/s per call per node (full): 315871
-Grid       : Message        : Average mflops/s per call per node (full): 434870
-Grid       : Message        : Average mflops/s per call per node (full): 443764
-Grid       : Message        : Average mflops/s per call per node (full): 304046
-Grid       : Message        :  Stencil 13.0725 GB/s per node
-Grid       : Message        :  Stencil 16.8008 GB/s per node
-Grid       : Message        :  Stencil 17.7133 GB/s per node
-Grid       : Message        :  Stencil 12.5526 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662510
-Grid       : Message        : Average mflops/s per call per node       : 801213
-Grid       : Message        : Average mflops/s per call per node       : 818551
-Grid       : Message        : Average mflops/s per call per node       : 666014
-Grid       : Message        : Average mflops/s per call per node (full): 313216
-Grid       : Message        : Average mflops/s per call per node (full): 436557
-Grid       : Message        : Average mflops/s per call per node (full): 445124
-Grid       : Message        : Average mflops/s per call per node (full): 305612
-Grid       : Message        :  Stencil 13.0189 GB/s per node
-Grid       : Message        :  Stencil 16.4422 GB/s per node
-Grid       : Message        :  Stencil 18.3631 GB/s per node
-Grid       : Message        :  Stencil 12.9549 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668479
-Grid       : Message        : Average mflops/s per call per node       : 804731
-Grid       : Message        : Average mflops/s per call per node       : 823586
-Grid       : Message        : Average mflops/s per call per node       : 663128
-Grid       : Message        : Average mflops/s per call per node (full): 315350
-Grid       : Message        : Average mflops/s per call per node (full): 433633
-Grid       : Message        : Average mflops/s per call per node (full): 448125
-Grid       : Message        : Average mflops/s per call per node (full): 305041
-Grid       : Message        :  Stencil 13.7261 GB/s per node
-Grid       : Message        :  Stencil 16.5987 GB/s per node
-Grid       : Message        :  Stencil 18.1536 GB/s per node
-Grid       : Message        :  Stencil 13.5897 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662310
-Grid       : Message        : Average mflops/s per call per node       : 797417
-Grid       : Message        : Average mflops/s per call per node       : 821599
-Grid       : Message        : Average mflops/s per call per node       : 659934
-Grid       : Message        : Average mflops/s per call per node (full): 315396
-Grid       : Message        : Average mflops/s per call per node (full): 433694
-Grid       : Message        : Average mflops/s per call per node (full): 446757
-Grid       : Message        : Average mflops/s per call per node (full): 305987
-Grid       : Message        :  Stencil 15.2071 GB/s per node
-Grid       : Message        :  Stencil 17.2064 GB/s per node
-Grid       : Message        :  Stencil 17.2153 GB/s per node
-Grid       : Message        :  Stencil 14.5624 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661738
-Grid       : Message        : Average mflops/s per call per node       : 804050
-Grid       : Message        : Average mflops/s per call per node       : 823835
-Grid       : Message        : Average mflops/s per call per node       : 657306
-Grid       : Message        : Average mflops/s per call per node (full): 316739
-Grid       : Message        : Average mflops/s per call per node (full): 438758
-Grid       : Message        : Average mflops/s per call per node (full): 443199
-Grid       : Message        : Average mflops/s per call per node (full): 306799
-Grid       : Message        :  Stencil 14.368 GB/s per node
-Grid       : Message        :  Stencil 12.6867 GB/s per node
-Grid       : Message        :  Stencil 17.667 GB/s per node
-Grid       : Message        :  Stencil 12.7552 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664257
-Grid       : Message        : Average mflops/s per call per node       : 806847
-Grid       : Message        : Average mflops/s per call per node       : 825078
-Grid       : Message        : Average mflops/s per call per node       : 668175
-Grid       : Message        : Average mflops/s per call per node (full): 316318
-Grid       : Message        : Average mflops/s per call per node (full): 375326
-Grid       : Message        : Average mflops/s per call per node (full): 446654
-Grid       : Message        : Average mflops/s per call per node (full): 305203
-Grid       : Message        :  Stencil 14.4694 GB/s per node
-Grid       : Message        :  Stencil 16.4098 GB/s per node
-Grid       : Message        :  Stencil 17.5862 GB/s per node
-Grid       : Message        :  Stencil 11.8687 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664554
-Grid       : Message        : Average mflops/s per call per node       : 802905
-Grid       : Message        : Average mflops/s per call per node       : 820834
-Grid       : Message        : Average mflops/s per call per node       : 666175
-Grid       : Message        : Average mflops/s per call per node (full): 314881
-Grid       : Message        : Average mflops/s per call per node (full): 431615
-Grid       : Message        : Average mflops/s per call per node (full): 432655
-Grid       : Message        : Average mflops/s per call per node (full): 298226
-Grid       : Message        :  Stencil 13.5055 GB/s per node
-Grid       : Message        :  Stencil 16.7847 GB/s per node
-Grid       : Message        :  Stencil 18.1815 GB/s per node
-Grid       : Message        :  Stencil 13.5363 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662343
-Grid       : Message        : Average mflops/s per call per node       : 802398
-Grid       : Message        : Average mflops/s per call per node       : 822336
-Grid       : Message        : Average mflops/s per call per node       : 667545
-Grid       : Message        : Average mflops/s per call per node (full): 315990
-Grid       : Message        : Average mflops/s per call per node (full): 437402
-Grid       : Message        : Average mflops/s per call per node (full): 443588
-Grid       : Message        : Average mflops/s per call per node (full): 306124
-Grid       : Message        :  Stencil 14.5623 GB/s per node
-Grid       : Message        :  Stencil 16.7777 GB/s per node
-Grid       : Message        :  Stencil 17.9493 GB/s per node
-Grid       : Message        :  Stencil 13.6665 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661588
-Grid       : Message        : Average mflops/s per call per node       : 804256
-Grid       : Message        : Average mflops/s per call per node       : 823914
-Grid       : Message        : Average mflops/s per call per node       : 665377
-Grid       : Message        : Average mflops/s per call per node (full): 315257
-Grid       : Message        : Average mflops/s per call per node (full): 435848
-Grid       : Message        : Average mflops/s per call per node (full): 445922
-Grid       : Message        : Average mflops/s per call per node (full): 306689
-Grid       : Message        :  Stencil 12.7961 GB/s per node
-Grid       : Message        :  Stencil 16.6998 GB/s per node
-Grid       : Message        :  Stencil 17.8133 GB/s per node
-Grid       : Message        :  Stencil 12.2565 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663028
-Grid       : Message        : Average mflops/s per call per node       : 804059
-Grid       : Message        : Average mflops/s per call per node       : 823390
-Grid       : Message        : Average mflops/s per call per node       : 669287
-Grid       : Message        : Average mflops/s per call per node (full): 312929
-Grid       : Message        : Average mflops/s per call per node (full): 431848
-Grid       : Message        : Average mflops/s per call per node (full): 446219
-Grid       : Message        : Average mflops/s per call per node (full): 304734
-Grid       : Message        :  Stencil 12.7893 GB/s per node
-Grid       : Message        :  Stencil 16.221 GB/s per node
-Grid       : Message        :  Stencil 17.2351 GB/s per node
-Grid       : Message        :  Stencil 13.5079 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665857
-Grid       : Message        : Average mflops/s per call per node       : 802147
-Grid       : Message        : Average mflops/s per call per node       : 820083
-Grid       : Message        : Average mflops/s per call per node       : 665001
-Grid       : Message        : Average mflops/s per call per node (full): 310159
-Grid       : Message        : Average mflops/s per call per node (full): 428779
-Grid       : Message        : Average mflops/s per call per node (full): 442738
-Grid       : Message        : Average mflops/s per call per node (full): 306802
-Grid       : Message        :  Stencil 13.1754 GB/s per node
-Grid       : Message        :  Stencil 16.4533 GB/s per node
-Grid       : Message        :  Stencil 17.429 GB/s per node
-Grid       : Message        :  Stencil 13.1193 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668829
-Grid       : Message        : Average mflops/s per call per node       : 800926
-Grid       : Message        : Average mflops/s per call per node       : 820687
-Grid       : Message        : Average mflops/s per call per node       : 661271
-Grid       : Message        : Average mflops/s per call per node (full): 316558
-Grid       : Message        : Average mflops/s per call per node (full): 433815
-Grid       : Message        : Average mflops/s per call per node (full): 444707
-Grid       : Message        : Average mflops/s per call per node (full): 305770
-Grid       : Message        :  Stencil 12.9593 GB/s per node
-Grid       : Message        :  Stencil 17.1329 GB/s per node
-Grid       : Message        :  Stencil 17.0942 GB/s per node
-Grid       : Message        :  Stencil 12.9989 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667452
-Grid       : Message        : Average mflops/s per call per node       : 806554
-Grid       : Message        : Average mflops/s per call per node       : 822115
-Grid       : Message        : Average mflops/s per call per node       : 661052
-Grid       : Message        : Average mflops/s per call per node (full): 315427
-Grid       : Message        : Average mflops/s per call per node (full): 440584
-Grid       : Message        : Average mflops/s per call per node (full): 441843
-Grid       : Message        : Average mflops/s per call per node (full): 304562
-Grid       : Message        :  Stencil 13.518 GB/s per node
-Grid       : Message        :  Stencil 17.0027 GB/s per node
-Grid       : Message        :  Stencil 18.5458 GB/s per node
-Grid       : Message        :  Stencil 12.7962 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667844
-Grid       : Message        : Average mflops/s per call per node       : 810592
-Grid       : Message        : Average mflops/s per call per node       : 808363
-Grid       : Message        : Average mflops/s per call per node       : 663379
-Grid       : Message        : Average mflops/s per call per node (full): 315060
-Grid       : Message        : Average mflops/s per call per node (full): 439231
-Grid       : Message        : Average mflops/s per call per node (full): 445731
-Grid       : Message        : Average mflops/s per call per node (full): 304239
-Grid       : Message        :  Stencil 12.5581 GB/s per node
-Grid       : Message        :  Stencil 16.7383 GB/s per node
-Grid       : Message        :  Stencil 17.4816 GB/s per node
-Grid       : Message        :  Stencil 12.9929 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665838
-Grid       : Message        : Average mflops/s per call per node       : 806940
-Grid       : Message        : Average mflops/s per call per node       : 820026
-Grid       : Message        : Average mflops/s per call per node       : 667036
-Grid       : Message        : Average mflops/s per call per node (full): 313606
-Grid       : Message        : Average mflops/s per call per node (full): 436480
-Grid       : Message        : Average mflops/s per call per node (full): 444863
-Grid       : Message        : Average mflops/s per call per node (full): 306720
-Grid       : Message        :  Stencil 15.0397 GB/s per node
-Grid       : Message        :  Stencil 17.1911 GB/s per node
-Grid       : Message        :  Stencil 17.6592 GB/s per node
-Grid       : Message        :  Stencil 12.6569 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662483
-Grid       : Message        : Average mflops/s per call per node       : 809326
-Grid       : Message        : Average mflops/s per call per node       : 821916
-Grid       : Message        : Average mflops/s per call per node       : 667193
-Grid       : Message        : Average mflops/s per call per node (full): 316709
-Grid       : Message        : Average mflops/s per call per node (full): 440519
-Grid       : Message        : Average mflops/s per call per node (full): 444642
-Grid       : Message        : Average mflops/s per call per node (full): 304478
-Grid       : Message        :  Stencil 14.1893 GB/s per node
-Grid       : Message        :  Stencil 16.7491 GB/s per node
-Grid       : Message        :  Stencil 17.7025 GB/s per node
-Grid       : Message        :  Stencil 13.0426 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663481
-Grid       : Message        : Average mflops/s per call per node       : 803216
-Grid       : Message        : Average mflops/s per call per node       : 824558
-Grid       : Message        : Average mflops/s per call per node       : 664887
-Grid       : Message        : Average mflops/s per call per node (full): 316694
-Grid       : Message        : Average mflops/s per call per node (full): 436122
-Grid       : Message        : Average mflops/s per call per node (full): 446367
-Grid       : Message        : Average mflops/s per call per node (full): 305147
-Grid       : Message        :  Stencil 12.402 GB/s per node
-Grid       : Message        :  Stencil 16.8793 GB/s per node
-Grid       : Message        :  Stencil 17.5857 GB/s per node
-Grid       : Message        :  Stencil 13.9042 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668553
-Grid       : Message        : Average mflops/s per call per node       : 799272
-Grid       : Message        : Average mflops/s per call per node       : 824044
-Grid       : Message        : Average mflops/s per call per node       : 654693
-Grid       : Message        : Average mflops/s per call per node (full): 311949
-Grid       : Message        : Average mflops/s per call per node (full): 436881
-Grid       : Message        : Average mflops/s per call per node (full): 446205
-Grid       : Message        : Average mflops/s per call per node (full): 305577
-Grid       : Message        :  Stencil 13.1478 GB/s per node
-Grid       : Message        :  Stencil 16.31 GB/s per node
-Grid       : Message        :  Stencil 18.0335 GB/s per node
-Grid       : Message        :  Stencil 12.8276 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668461
-Grid       : Message        : Average mflops/s per call per node       : 802629
-Grid       : Message        : Average mflops/s per call per node       : 821788
-Grid       : Message        : Average mflops/s per call per node       : 663470
-Grid       : Message        : Average mflops/s per call per node (full): 315510
-Grid       : Message        : Average mflops/s per call per node (full): 432585
-Grid       : Message        : Average mflops/s per call per node (full): 446386
-Grid       : Message        : Average mflops/s per call per node (full): 305276
-Grid       : Message        :  Stencil 12.7427 GB/s per node
-Grid       : Message        :  Stencil 16.4928 GB/s per node
-Grid       : Message        :  Stencil 17.973 GB/s per node
-Grid       : Message        :  Stencil 13.3166 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 673103
-Grid       : Message        : Average mflops/s per call per node       : 804558
-Grid       : Message        : Average mflops/s per call per node       : 823988
-Grid       : Message        : Average mflops/s per call per node       : 663175
-Grid       : Message        : Average mflops/s per call per node (full): 315899
-Grid       : Message        : Average mflops/s per call per node (full): 433461
-Grid       : Message        : Average mflops/s per call per node (full): 442470
-Grid       : Message        : Average mflops/s per call per node (full): 306588
-Grid       : Message        :  Stencil 13.2121 GB/s per node
-Grid       : Message        :  Stencil 17.1273 GB/s per node
-Grid       : Message        :  Stencil 17.6244 GB/s per node
-Grid       : Message        :  Stencil 13.2802 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668004
-Grid       : Message        : Average mflops/s per call per node       : 799380
-Grid       : Message        : Average mflops/s per call per node       : 824882
-Grid       : Message        : Average mflops/s per call per node       : 662423
-Grid       : Message        : Average mflops/s per call per node (full): 315807
-Grid       : Message        : Average mflops/s per call per node (full): 437571
-Grid       : Message        : Average mflops/s per call per node (full): 445182
-Grid       : Message        : Average mflops/s per call per node (full): 305479
-Grid       : Message        :  Stencil 13.1344 GB/s per node
-Grid       : Message        :  Stencil 12.7333 GB/s per node
-Grid       : Message        :  Stencil 18.0397 GB/s per node
-Grid       : Message        :  Stencil 12.4452 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667610
-Grid       : Message        : Average mflops/s per call per node       : 809721
-Grid       : Message        : Average mflops/s per call per node       : 820494
-Grid       : Message        : Average mflops/s per call per node       : 667970
-Grid       : Message        : Average mflops/s per call per node (full): 315073
-Grid       : Message        : Average mflops/s per call per node (full): 376912
-Grid       : Message        : Average mflops/s per call per node (full): 447594
-Grid       : Message        : Average mflops/s per call per node (full): 302941
-Grid       : Message        :  Stencil 12.9916 GB/s per node
-Grid       : Message        :  Stencil 16.4847 GB/s per node
-Grid       : Message        :  Stencil 18.4717 GB/s per node
-Grid       : Message        :  Stencil 13.8467 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666602
-Grid       : Message        : Average mflops/s per call per node       : 806090
-Grid       : Message        : Average mflops/s per call per node       : 817726
-Grid       : Message        : Average mflops/s per call per node       : 661847
-Grid       : Message        : Average mflops/s per call per node (full): 315127
-Grid       : Message        : Average mflops/s per call per node (full): 432934
-Grid       : Message        : Average mflops/s per call per node (full): 448390
-Grid       : Message        : Average mflops/s per call per node (full): 306390
-Grid       : Message        :  Stencil 12.3533 GB/s per node
-Grid       : Message        :  Stencil 17.2274 GB/s per node
-Grid       : Message        :  Stencil 17.7197 GB/s per node
-Grid       : Message        :  Stencil 13.6455 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665051
-Grid       : Message        : Average mflops/s per call per node       : 806937
-Grid       : Message        : Average mflops/s per call per node       : 822513
-Grid       : Message        : Average mflops/s per call per node       : 660128
-Grid       : Message        : Average mflops/s per call per node (full): 307102
-Grid       : Message        : Average mflops/s per call per node (full): 438441
-Grid       : Message        : Average mflops/s per call per node (full): 446803
-Grid       : Message        : Average mflops/s per call per node (full): 305302
-Grid       : Message        :  Stencil 12.3064 GB/s per node
-Grid       : Message        :  Stencil 16.2822 GB/s per node
-Grid       : Message        :  Stencil 17.2384 GB/s per node
-Grid       : Message        :  Stencil 14.1965 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665604
-Grid       : Message        : Average mflops/s per call per node       : 801811
-Grid       : Message        : Average mflops/s per call per node       : 826139
-Grid       : Message        : Average mflops/s per call per node       : 658406
-Grid       : Message        : Average mflops/s per call per node (full): 311986
-Grid       : Message        : Average mflops/s per call per node (full): 431562
-Grid       : Message        : Average mflops/s per call per node (full): 443145
-Grid       : Message        : Average mflops/s per call per node (full): 304433
-Grid       : Message        :  Stencil 12.532 GB/s per node
-Grid       : Message        :  Stencil 17.5079 GB/s per node
-Grid       : Message        :  Stencil 18.6678 GB/s per node
-Grid       : Message        :  Stencil 12.3361 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667572
-Grid       : Message        : Average mflops/s per call per node       : 801710
-Grid       : Message        : Average mflops/s per call per node       : 818537
-Grid       : Message        : Average mflops/s per call per node       : 668800
-Grid       : Message        : Average mflops/s per call per node (full): 313979
-Grid       : Message        : Average mflops/s per call per node (full): 439231
-Grid       : Message        : Average mflops/s per call per node (full): 446644
-Grid       : Message        : Average mflops/s per call per node (full): 304258
-Grid       : Message        :  Stencil 13.1495 GB/s per node
-Grid       : Message        :  Stencil 16.1774 GB/s per node
-Grid       : Message        :  Stencil 17.0706 GB/s per node
-Grid       : Message        :  Stencil 12.2607 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669214
-Grid       : Message        : Average mflops/s per call per node       : 802115
-Grid       : Message        : Average mflops/s per call per node       : 824918
-Grid       : Message        : Average mflops/s per call per node       : 668359
-Grid       : Message        : Average mflops/s per call per node (full): 315314
-Grid       : Message        : Average mflops/s per call per node (full): 430106
-Grid       : Message        : Average mflops/s per call per node (full): 441147
-Grid       : Message        : Average mflops/s per call per node (full): 303811
-Grid       : Message        :  Stencil 13.1248 GB/s per node
-Grid       : Message        :  Stencil 9.80877 GB/s per node
-Grid       : Message        :  Stencil 17.9584 GB/s per node
-Grid       : Message        :  Stencil 12.2776 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667706
-Grid       : Message        : Average mflops/s per call per node       : 804502
-Grid       : Message        : Average mflops/s per call per node       : 828082
-Grid       : Message        : Average mflops/s per call per node       : 670349
-Grid       : Message        : Average mflops/s per call per node (full): 315359
-Grid       : Message        : Average mflops/s per call per node (full): 310713
-Grid       : Message        : Average mflops/s per call per node (full): 446836
-Grid       : Message        : Average mflops/s per call per node (full): 304280
-Grid       : Message        :  Stencil 13.5257 GB/s per node
-Grid       : Message        :  Stencil 17.2223 GB/s per node
-Grid       : Message        :  Stencil 17.3992 GB/s per node
-Grid       : Message        :  Stencil 14.4036 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668025
-Grid       : Message        : Average mflops/s per call per node       : 803956
-Grid       : Message        : Average mflops/s per call per node       : 821781
-Grid       : Message        : Average mflops/s per call per node       : 656816
-Grid       : Message        : Average mflops/s per call per node (full): 315395
-Grid       : Message        : Average mflops/s per call per node (full): 438030
-Grid       : Message        : Average mflops/s per call per node (full): 443848
-Grid       : Message        : Average mflops/s per call per node (full): 306141
-Grid       : Message        :  Stencil 13.7108 GB/s per node
-Grid       : Message        :  Stencil 17.0664 GB/s per node
-Grid       : Message        :  Stencil 17.4254 GB/s per node
-Grid       : Message        :  Stencil 12.3972 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666651
-Grid       : Message        : Average mflops/s per call per node       : 804934
-Grid       : Message        : Average mflops/s per call per node       : 829044
-Grid       : Message        : Average mflops/s per call per node       : 667967
-Grid       : Message        : Average mflops/s per call per node (full): 316229
-Grid       : Message        : Average mflops/s per call per node (full): 437367
-Grid       : Message        : Average mflops/s per call per node (full): 445282
-Grid       : Message        : Average mflops/s per call per node (full): 304927
-Grid       : Message        :  Stencil 13.6529 GB/s per node
-Grid       : Message        :  Stencil 17.3621 GB/s per node
-Grid       : Message        :  Stencil 17.7085 GB/s per node
-Grid       : Message        :  Stencil 12.1508 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665599
-Grid       : Message        : Average mflops/s per call per node       : 802139
-Grid       : Message        : Average mflops/s per call per node       : 819221
-Grid       : Message        : Average mflops/s per call per node       : 666278
-Grid       : Message        : Average mflops/s per call per node (full): 315715
-Grid       : Message        : Average mflops/s per call per node (full): 437760
-Grid       : Message        : Average mflops/s per call per node (full): 444470
-Grid       : Message        : Average mflops/s per call per node (full): 303334
-Grid       : Message        :  Stencil 13.7906 GB/s per node
-Grid       : Message        :  Stencil 6.55631 GB/s per node
-Grid       : Message        :  Stencil 18.0631 GB/s per node
-Grid       : Message        :  Stencil 12.4737 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663705
-Grid       : Message        : Average mflops/s per call per node       : 807329
-Grid       : Message        : Average mflops/s per call per node       : 822044
-Grid       : Message        : Average mflops/s per call per node       : 667803
-Grid       : Message        : Average mflops/s per call per node (full): 316153
-Grid       : Message        : Average mflops/s per call per node (full): 226870
-Grid       : Message        : Average mflops/s per call per node (full): 445791
-Grid       : Message        : Average mflops/s per call per node (full): 304399
-Grid       : Message        :  Stencil 13.9422 GB/s per node
-Grid       : Message        :  Stencil 18.0489 GB/s per node
-Grid       : Message        :  Stencil 18.202 GB/s per node
-Grid       : Message        :  Stencil 12.7549 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664665
-Grid       : Message        : Average mflops/s per call per node       : 805682
-Grid       : Message        : Average mflops/s per call per node       : 823134
-Grid       : Message        : Average mflops/s per call per node       : 660925
-Grid       : Message        : Average mflops/s per call per node (full): 316494
-Grid       : Message        : Average mflops/s per call per node (full): 441297
-Grid       : Message        : Average mflops/s per call per node (full): 447213
-Grid       : Message        : Average mflops/s per call per node (full): 303280
-Grid       : Message        :  Stencil 14.8899 GB/s per node
-Grid       : Message        :  Stencil 17.1059 GB/s per node
-Grid       : Message        :  Stencil 17.6417 GB/s per node
-Grid       : Message        :  Stencil 12.4758 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663410
-Grid       : Message        : Average mflops/s per call per node       : 805254
-Grid       : Message        : Average mflops/s per call per node       : 825254
-Grid       : Message        : Average mflops/s per call per node       : 667450
-Grid       : Message        : Average mflops/s per call per node (full): 317238
-Grid       : Message        : Average mflops/s per call per node (full): 437415
-Grid       : Message        : Average mflops/s per call per node (full): 437589
-Grid       : Message        : Average mflops/s per call per node (full): 303595
-Grid       : Message        :  Stencil 12.6778 GB/s per node
-Grid       : Message        :  Stencil 17.786 GB/s per node
-Grid       : Message        :  Stencil 17.91 GB/s per node
-Grid       : Message        :  Stencil 13.2347 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671875
-Grid       : Message        : Average mflops/s per call per node       : 802320
-Grid       : Message        : Average mflops/s per call per node       : 826003
-Grid       : Message        : Average mflops/s per call per node       : 664528
-Grid       : Message        : Average mflops/s per call per node (full): 315102
-Grid       : Message        : Average mflops/s per call per node (full): 438671
-Grid       : Message        : Average mflops/s per call per node (full): 448007
-Grid       : Message        : Average mflops/s per call per node (full): 304178
-Grid       : Message        :  Stencil 12.8066 GB/s per node
-Grid       : Message        :  Stencil 9.47955 GB/s per node
-Grid       : Message        :  Stencil 17.1296 GB/s per node
-Grid       : Message        :  Stencil 13.3126 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667534
-Grid       : Message        : Average mflops/s per call per node       : 810356
-Grid       : Message        : Average mflops/s per call per node       : 820584
-Grid       : Message        : Average mflops/s per call per node       : 665002
-Grid       : Message        : Average mflops/s per call per node (full): 315031
-Grid       : Message        : Average mflops/s per call per node (full): 303250
-Grid       : Message        : Average mflops/s per call per node (full): 440421
-Grid       : Message        : Average mflops/s per call per node (full): 306009
-Grid       : Message        :  Stencil 12.8412 GB/s per node
-Grid       : Message        :  Stencil 17.5009 GB/s per node
-Grid       : Message        :  Stencil 19.7497 GB/s per node
-Grid       : Message        :  Stencil 12.4201 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666739
-Grid       : Message        : Average mflops/s per call per node       : 798613
-Grid       : Message        : Average mflops/s per call per node       : 818810
-Grid       : Message        : Average mflops/s per call per node       : 663304
-Grid       : Message        : Average mflops/s per call per node (full): 312930
-Grid       : Message        : Average mflops/s per call per node (full): 438826
-Grid       : Message        : Average mflops/s per call per node (full): 450335
-Grid       : Message        : Average mflops/s per call per node (full): 303353
-Grid       : Message        :  Stencil 12.7016 GB/s per node
-Grid       : Message        :  Stencil 16.2536 GB/s per node
-Grid       : Message        :  Stencil 18.2265 GB/s per node
-Grid       : Message        :  Stencil 12.3248 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670398
-Grid       : Message        : Average mflops/s per call per node       : 804082
-Grid       : Message        : Average mflops/s per call per node       : 821654
-Grid       : Message        : Average mflops/s per call per node       : 665580
-Grid       : Message        : Average mflops/s per call per node (full): 314434
-Grid       : Message        : Average mflops/s per call per node (full): 429458
-Grid       : Message        : Average mflops/s per call per node (full): 446120
-Grid       : Message        : Average mflops/s per call per node (full): 303792
-Grid       : Message        :  Stencil 12.8835 GB/s per node
-Grid       : Message        :  Stencil 16.4583 GB/s per node
-Grid       : Message        :  Stencil 17.5232 GB/s per node
-Grid       : Message        :  Stencil 12.9388 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668522
-Grid       : Message        : Average mflops/s per call per node       : 808236
-Grid       : Message        : Average mflops/s per call per node       : 821330
-Grid       : Message        : Average mflops/s per call per node       : 660228
-Grid       : Message        : Average mflops/s per call per node (full): 313520
-Grid       : Message        : Average mflops/s per call per node (full): 430168
-Grid       : Message        : Average mflops/s per call per node (full): 439663
-Grid       : Message        : Average mflops/s per call per node (full): 304068
-Grid       : Message        :  Stencil 13.021 GB/s per node
-Grid       : Message        :  Stencil 16.8159 GB/s per node
-Grid       : Message        :  Stencil 17.3261 GB/s per node
-Grid       : Message        :  Stencil 12.7807 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667768
-Grid       : Message        : Average mflops/s per call per node       : 807405
-Grid       : Message        : Average mflops/s per call per node       : 825368
-Grid       : Message        : Average mflops/s per call per node       : 667833
-Grid       : Message        : Average mflops/s per call per node (full): 313354
-Grid       : Message        : Average mflops/s per call per node (full): 433510
-Grid       : Message        : Average mflops/s per call per node (full): 443765
-Grid       : Message        : Average mflops/s per call per node (full): 306532
-Grid       : Message        :  Stencil 13.04 GB/s per node
-Grid       : Message        :  Stencil 17.9266 GB/s per node
-Grid       : Message        :  Stencil 16.9641 GB/s per node
-Grid       : Message        :  Stencil 12.1914 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667822
-Grid       : Message        : Average mflops/s per call per node       : 796507
-Grid       : Message        : Average mflops/s per call per node       : 826307
-Grid       : Message        : Average mflops/s per call per node       : 667710
-Grid       : Message        : Average mflops/s per call per node (full): 314874
-Grid       : Message        : Average mflops/s per call per node (full): 438934
-Grid       : Message        : Average mflops/s per call per node (full): 440721
-Grid       : Message        : Average mflops/s per call per node (full): 301811
-Grid       : Message        :  Stencil 14.0713 GB/s per node
-Grid       : Message        :  Stencil 9.84629 GB/s per node
-Grid       : Message        :  Stencil 17.975 GB/s per node
-Grid       : Message        :  Stencil 12.6292 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662677
-Grid       : Message        : Average mflops/s per call per node       : 807628
-Grid       : Message        : Average mflops/s per call per node       : 819486
-Grid       : Message        : Average mflops/s per call per node       : 665414
-Grid       : Message        : Average mflops/s per call per node (full): 314320
-Grid       : Message        : Average mflops/s per call per node (full): 311792
-Grid       : Message        : Average mflops/s per call per node (full): 446145
-Grid       : Message        : Average mflops/s per call per node (full): 305001
-Grid       : Message        :  Stencil 15.2395 GB/s per node
-Grid       : Message        :  Stencil 16.6914 GB/s per node
-Grid       : Message        :  Stencil 16.7894 GB/s per node
-Grid       : Message        :  Stencil 13.1997 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661209
-Grid       : Message        : Average mflops/s per call per node       : 809394
-Grid       : Message        : Average mflops/s per call per node       : 825684
-Grid       : Message        : Average mflops/s per call per node       : 662842
-Grid       : Message        : Average mflops/s per call per node (full): 316996
-Grid       : Message        : Average mflops/s per call per node (full): 437153
-Grid       : Message        : Average mflops/s per call per node (full): 438272
-Grid       : Message        : Average mflops/s per call per node (full): 304307
-Grid       : Message        :  Stencil 13.3791 GB/s per node
-Grid       : Message        :  Stencil 16.8369 GB/s per node
-Grid       : Message        :  Stencil 17.224 GB/s per node
-Grid       : Message        :  Stencil 14.0024 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665689
-Grid       : Message        : Average mflops/s per call per node       : 802725
-Grid       : Message        : Average mflops/s per call per node       : 827751
-Grid       : Message        : Average mflops/s per call per node       : 654313
-Grid       : Message        : Average mflops/s per call per node (full): 316236
-Grid       : Message        : Average mflops/s per call per node (full): 436545
-Grid       : Message        : Average mflops/s per call per node (full): 443947
-Grid       : Message        : Average mflops/s per call per node (full): 304434
-Grid       : Message        :  Stencil 12.5735 GB/s per node
-Grid       : Message        :  Stencil 11.2593 GB/s per node
-Grid       : Message        :  Stencil 17.4278 GB/s per node
-Grid       : Message        :  Stencil 12.9417 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 672094
-Grid       : Message        : Average mflops/s per call per node       : 807479
-Grid       : Message        : Average mflops/s per call per node       : 827536
-Grid       : Message        : Average mflops/s per call per node       : 663187
-Grid       : Message        : Average mflops/s per call per node (full): 314938
-Grid       : Message        : Average mflops/s per call per node (full): 345570
-Grid       : Message        : Average mflops/s per call per node (full): 444522
-Grid       : Message        : Average mflops/s per call per node (full): 304011
-Grid       : Message        :  Stencil 13.6533 GB/s per node
-Grid       : Message        :  Stencil 17.3081 GB/s per node
-Grid       : Message        :  Stencil 17.4874 GB/s per node
-Grid       : Message        :  Stencil 12.6095 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668633
-Grid       : Message        : Average mflops/s per call per node       : 802535
-Grid       : Message        : Average mflops/s per call per node       : 823084
-Grid       : Message        : Average mflops/s per call per node       : 668702
-Grid       : Message        : Average mflops/s per call per node (full): 317040
-Grid       : Message        : Average mflops/s per call per node (full): 438029
-Grid       : Message        : Average mflops/s per call per node (full): 445463
-Grid       : Message        : Average mflops/s per call per node (full): 306130
-Grid       : Message        :  Stencil 12.867 GB/s per node
-Grid       : Message        :  Stencil 17.2766 GB/s per node
-Grid       : Message        :  Stencil 16.1951 GB/s per node
-Grid       : Message        :  Stencil 11.682 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667313
-Grid       : Message        : Average mflops/s per call per node       : 802375
-Grid       : Message        : Average mflops/s per call per node       : 827107
-Grid       : Message        : Average mflops/s per call per node       : 664678
-Grid       : Message        : Average mflops/s per call per node (full): 313388
-Grid       : Message        : Average mflops/s per call per node (full): 438214
-Grid       : Message        : Average mflops/s per call per node (full): 424986
-Grid       : Message        : Average mflops/s per call per node (full): 298506
-Grid       : Message        :  Stencil 13.0763 GB/s per node
-Grid       : Message        :  Stencil 16.5815 GB/s per node
-Grid       : Message        :  Stencil 17.1128 GB/s per node
-Grid       : Message        :  Stencil 12.3072 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666096
-Grid       : Message        : Average mflops/s per call per node       : 805076
-Grid       : Message        : Average mflops/s per call per node       : 820869
-Grid       : Message        : Average mflops/s per call per node       : 664399
-Grid       : Message        : Average mflops/s per call per node (full): 314237
-Grid       : Message        : Average mflops/s per call per node (full): 436072
-Grid       : Message        : Average mflops/s per call per node (full): 440936
-Grid       : Message        : Average mflops/s per call per node (full): 303585
-Grid       : Message        :  Stencil 13.4101 GB/s per node
-Grid       : Message        :  Stencil 16.7424 GB/s per node
-Grid       : Message        :  Stencil 17.8502 GB/s per node
-Grid       : Message        :  Stencil 14.3448 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666799
-Grid       : Message        : Average mflops/s per call per node       : 809391
-Grid       : Message        : Average mflops/s per call per node       : 823629
-Grid       : Message        : Average mflops/s per call per node       : 664725
-Grid       : Message        : Average mflops/s per call per node (full): 315379
-Grid       : Message        : Average mflops/s per call per node (full): 437879
-Grid       : Message        : Average mflops/s per call per node (full): 447103
-Grid       : Message        : Average mflops/s per call per node (full): 307257
-Grid       : Message        :  Stencil 13.568 GB/s per node
-Grid       : Message        :  Stencil 17.6779 GB/s per node
-Grid       : Message        :  Stencil 16.8741 GB/s per node
-Grid       : Message        :  Stencil 12.4839 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668025
-Grid       : Message        : Average mflops/s per call per node       : 804674
-Grid       : Message        : Average mflops/s per call per node       : 821256
-Grid       : Message        : Average mflops/s per call per node       : 664966
-Grid       : Message        : Average mflops/s per call per node (full): 316063
-Grid       : Message        : Average mflops/s per call per node (full): 440264
-Grid       : Message        : Average mflops/s per call per node (full): 439220
-Grid       : Message        : Average mflops/s per call per node (full): 303367
-Grid       : Message        :  Stencil 13.8637 GB/s per node
-Grid       : Message        :  Stencil 16.318 GB/s per node
-Grid       : Message        :  Stencil 17.3829 GB/s per node
-Grid       : Message        :  Stencil 12.7476 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667436
-Grid       : Message        : Average mflops/s per call per node       : 803547
-Grid       : Message        : Average mflops/s per call per node       : 825521
-Grid       : Message        : Average mflops/s per call per node       : 662615
-Grid       : Message        : Average mflops/s per call per node (full): 316257
-Grid       : Message        : Average mflops/s per call per node (full): 429325
-Grid       : Message        : Average mflops/s per call per node (full): 441772
-Grid       : Message        : Average mflops/s per call per node (full): 304181
-Grid       : Message        :  Stencil 14.1388 GB/s per node
-Grid       : Message        :  Stencil 16.9508 GB/s per node
-Grid       : Message        :  Stencil 17.6921 GB/s per node
-Grid       : Message        :  Stencil 12.6858 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666942
-Grid       : Message        : Average mflops/s per call per node       : 802603
-Grid       : Message        : Average mflops/s per call per node       : 824945
-Grid       : Message        : Average mflops/s per call per node       : 668276
-Grid       : Message        : Average mflops/s per call per node (full): 317268
-Grid       : Message        : Average mflops/s per call per node (full): 435966
-Grid       : Message        : Average mflops/s per call per node (full): 445270
-Grid       : Message        : Average mflops/s per call per node (full): 305709
-Grid       : Message        :  Stencil 13.2752 GB/s per node
-Grid       : Message        :  Stencil 14.3609 GB/s per node
-Grid       : Message        :  Stencil 19.2179 GB/s per node
-Grid       : Message        :  Stencil 13.5641 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666144
-Grid       : Message        : Average mflops/s per call per node       : 803392
-Grid       : Message        : Average mflops/s per call per node       : 822152
-Grid       : Message        : Average mflops/s per call per node       : 658488
-Grid       : Message        : Average mflops/s per call per node (full): 315994
-Grid       : Message        : Average mflops/s per call per node (full): 402828
-Grid       : Message        : Average mflops/s per call per node (full): 449873
-Grid       : Message        : Average mflops/s per call per node (full): 304869
-Grid       : Message        :  Stencil 13.6586 GB/s per node
-Grid       : Message        :  Stencil 16.6687 GB/s per node
-Grid       : Message        :  Stencil 17.8118 GB/s per node
-Grid       : Message        :  Stencil 12.9281 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662645
-Grid       : Message        : Average mflops/s per call per node       : 805011
-Grid       : Message        : Average mflops/s per call per node       : 823772
-Grid       : Message        : Average mflops/s per call per node       : 659767
-Grid       : Message        : Average mflops/s per call per node (full): 315917
-Grid       : Message        : Average mflops/s per call per node (full): 434045
-Grid       : Message        : Average mflops/s per call per node (full): 447067
-Grid       : Message        : Average mflops/s per call per node (full): 303836
-Grid       : Message        :  Stencil 12.783 GB/s per node
-Grid       : Message        :  Stencil 16.5007 GB/s per node
-Grid       : Message        :  Stencil 18.1978 GB/s per node
-Grid       : Message        :  Stencil 12.5762 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668439
-Grid       : Message        : Average mflops/s per call per node       : 808408
-Grid       : Message        : Average mflops/s per call per node       : 818862
-Grid       : Message        : Average mflops/s per call per node       : 666769
-Grid       : Message        : Average mflops/s per call per node (full): 313955
-Grid       : Message        : Average mflops/s per call per node (full): 435371
-Grid       : Message        : Average mflops/s per call per node (full): 444736
-Grid       : Message        : Average mflops/s per call per node (full): 302601
-Grid       : Message        :  Stencil 12.3588 GB/s per node
-Grid       : Message        :  Stencil 16.6906 GB/s per node
-Grid       : Message        :  Stencil 17.5897 GB/s per node
-Grid       : Message        :  Stencil 12.731 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 672717
-Grid       : Message        : Average mflops/s per call per node       : 809225
-Grid       : Message        : Average mflops/s per call per node       : 824966
-Grid       : Message        : Average mflops/s per call per node       : 664892
-Grid       : Message        : Average mflops/s per call per node (full): 312967
-Grid       : Message        : Average mflops/s per call per node (full): 437334
-Grid       : Message        : Average mflops/s per call per node (full): 446624
-Grid       : Message        : Average mflops/s per call per node (full): 305067
-Grid       : Message        :  Stencil 13.5086 GB/s per node
-Grid       : Message        :  Stencil 16.7682 GB/s per node
-Grid       : Message        :  Stencil 17.36 GB/s per node
-Grid       : Message        :  Stencil 12.4457 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665369
-Grid       : Message        : Average mflops/s per call per node       : 803451
-Grid       : Message        : Average mflops/s per call per node       : 819366
-Grid       : Message        : Average mflops/s per call per node       : 664067
-Grid       : Message        : Average mflops/s per call per node (full): 314363
-Grid       : Message        : Average mflops/s per call per node (full): 436058
-Grid       : Message        : Average mflops/s per call per node (full): 443030
-Grid       : Message        : Average mflops/s per call per node (full): 303937
-Grid       : Message        :  Stencil 13.3165 GB/s per node
-Grid       : Message        :  Stencil 12.5025 GB/s per node
-Grid       : Message        :  Stencil 17.7703 GB/s per node
-Grid       : Message        :  Stencil 14.2332 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665657
-Grid       : Message        : Average mflops/s per call per node       : 812801
-Grid       : Message        : Average mflops/s per call per node       : 823140
-Grid       : Message        : Average mflops/s per call per node       : 663945
-Grid       : Message        : Average mflops/s per call per node (full): 315523
-Grid       : Message        : Average mflops/s per call per node (full): 372095
-Grid       : Message        : Average mflops/s per call per node (full): 446653
-Grid       : Message        : Average mflops/s per call per node (full): 307241
-Grid       : Message        :  Stencil 14.9576 GB/s per node
-Grid       : Message        :  Stencil 8.89052 GB/s per node
-Grid       : Message        :  Stencil 17.2728 GB/s per node
-Grid       : Message        :  Stencil 12.6808 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662859
-Grid       : Message        : Average mflops/s per call per node       : 812123
-Grid       : Message        : Average mflops/s per call per node       : 830044
-Grid       : Message        : Average mflops/s per call per node       : 665640
-Grid       : Message        : Average mflops/s per call per node (full): 317215
-Grid       : Message        : Average mflops/s per call per node (full): 289792
-Grid       : Message        : Average mflops/s per call per node (full): 444953
-Grid       : Message        : Average mflops/s per call per node (full): 305590
-Grid       : Message        :  Stencil 13.2073 GB/s per node
-Grid       : Message        :  Stencil 16.5902 GB/s per node
-Grid       : Message        :  Stencil 17.9064 GB/s per node
-Grid       : Message        :  Stencil 12.3071 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665944
-Grid       : Message        : Average mflops/s per call per node       : 805445
-Grid       : Message        : Average mflops/s per call per node       : 821582
-Grid       : Message        : Average mflops/s per call per node       : 659905
-Grid       : Message        : Average mflops/s per call per node (full): 315108
-Grid       : Message        : Average mflops/s per call per node (full): 434920
-Grid       : Message        : Average mflops/s per call per node (full): 440533
-Grid       : Message        : Average mflops/s per call per node (full): 300261
-Grid       : Message        :  Stencil 13.6207 GB/s per node
-Grid       : Message        :  Stencil 16.963 GB/s per node
-Grid       : Message        :  Stencil 18.8015 GB/s per node
-Grid       : Message        :  Stencil 12.6053 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662003
-Grid       : Message        : Average mflops/s per call per node       : 802034
-Grid       : Message        : Average mflops/s per call per node       : 823994
-Grid       : Message        : Average mflops/s per call per node       : 664597
-Grid       : Message        : Average mflops/s per call per node (full): 315494
-Grid       : Message        : Average mflops/s per call per node (full): 436003
-Grid       : Message        : Average mflops/s per call per node (full): 449112
-Grid       : Message        : Average mflops/s per call per node (full): 304923
-Grid       : Message        :  Stencil 13.6174 GB/s per node
-Grid       : Message        :  Stencil 12.2346 GB/s per node
-Grid       : Message        :  Stencil 17.3975 GB/s per node
-Grid       : Message        :  Stencil 13.829 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661357
-Grid       : Message        : Average mflops/s per call per node       : 801615
-Grid       : Message        : Average mflops/s per call per node       : 822680
-Grid       : Message        : Average mflops/s per call per node       : 662306
-Grid       : Message        : Average mflops/s per call per node (full): 314288
-Grid       : Message        : Average mflops/s per call per node (full): 366449
-Grid       : Message        : Average mflops/s per call per node (full): 444360
-Grid       : Message        : Average mflops/s per call per node (full): 307243
-Grid       : Message        :  Stencil 13.2055 GB/s per node
-Grid       : Message        :  Stencil 15.5344 GB/s per node
-Grid       : Message        :  Stencil 18.6795 GB/s per node
-Grid       : Message        :  Stencil 13.3403 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667854
-Grid       : Message        : Average mflops/s per call per node       : 805612
-Grid       : Message        : Average mflops/s per call per node       : 825962
-Grid       : Message        : Average mflops/s per call per node       : 661435
-Grid       : Message        : Average mflops/s per call per node (full): 315217
-Grid       : Message        : Average mflops/s per call per node (full): 423496
-Grid       : Message        : Average mflops/s per call per node (full): 449684
-Grid       : Message        : Average mflops/s per call per node (full): 305117
-Grid       : Message        :  Stencil 14.0589 GB/s per node
-Grid       : Message        :  Stencil 17.53 GB/s per node
-Grid       : Message        :  Stencil 17.3587 GB/s per node
-Grid       : Message        :  Stencil 12.8013 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663290
-Grid       : Message        : Average mflops/s per call per node       : 802056
-Grid       : Message        : Average mflops/s per call per node       : 820628
-Grid       : Message        : Average mflops/s per call per node       : 667458
-Grid       : Message        : Average mflops/s per call per node (full): 316086
-Grid       : Message        : Average mflops/s per call per node (full): 433471
-Grid       : Message        : Average mflops/s per call per node (full): 443865
-Grid       : Message        : Average mflops/s per call per node (full): 305638
-Grid       : Message        :  Stencil 12.4114 GB/s per node
-Grid       : Message        :  Stencil 16.5811 GB/s per node
-Grid       : Message        :  Stencil 17.246 GB/s per node
-Grid       : Message        :  Stencil 12.9512 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670801
-Grid       : Message        : Average mflops/s per call per node       : 803141
-Grid       : Message        : Average mflops/s per call per node       : 818087
-Grid       : Message        : Average mflops/s per call per node       : 662536
-Grid       : Message        : Average mflops/s per call per node (full): 313026
-Grid       : Message        : Average mflops/s per call per node (full): 435532
-Grid       : Message        : Average mflops/s per call per node (full): 440716
-Grid       : Message        : Average mflops/s per call per node (full): 304931
-Grid       : Message        :  Stencil 12.3893 GB/s per node
-Grid       : Message        :  Stencil 16.275 GB/s per node
-Grid       : Message        :  Stencil 17.6015 GB/s per node
-Grid       : Message        :  Stencil 14.2541 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670981
-Grid       : Message        : Average mflops/s per call per node       : 808883
-Grid       : Message        : Average mflops/s per call per node       : 824212
-Grid       : Message        : Average mflops/s per call per node       : 662167
-Grid       : Message        : Average mflops/s per call per node (full): 313634
-Grid       : Message        : Average mflops/s per call per node (full): 433070
-Grid       : Message        : Average mflops/s per call per node (full): 445960
-Grid       : Message        : Average mflops/s per call per node (full): 307540
-Grid       : Message        :  Stencil 12.5686 GB/s per node
-Grid       : Message        :  Stencil 17.4453 GB/s per node
-Grid       : Message        :  Stencil 18.5536 GB/s per node
-Grid       : Message        :  Stencil 14.4062 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671233
-Grid       : Message        : Average mflops/s per call per node       : 803489
-Grid       : Message        : Average mflops/s per call per node       : 822710
-Grid       : Message        : Average mflops/s per call per node       : 659322
-Grid       : Message        : Average mflops/s per call per node (full): 314646
-Grid       : Message        : Average mflops/s per call per node (full): 438822
-Grid       : Message        : Average mflops/s per call per node (full): 448936
-Grid       : Message        : Average mflops/s per call per node (full): 307099
-Grid       : Message        :  Stencil 12.17 GB/s per node
-Grid       : Message        :  Stencil 17.2394 GB/s per node
-Grid       : Message        :  Stencil 17.1924 GB/s per node
-Grid       : Message        :  Stencil 14.3786 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670419
-Grid       : Message        : Average mflops/s per call per node       : 803856
-Grid       : Message        : Average mflops/s per call per node       : 819163
-Grid       : Message        : Average mflops/s per call per node       : 660090
-Grid       : Message        : Average mflops/s per call per node (full): 309861
-Grid       : Message        : Average mflops/s per call per node (full): 439206
-Grid       : Message        : Average mflops/s per call per node (full): 443010
-Grid       : Message        : Average mflops/s per call per node (full): 306431
-Grid       : Message        :  Stencil 13.4474 GB/s per node
-Grid       : Message        :  Stencil 16.5257 GB/s per node
-Grid       : Message        :  Stencil 16.5948 GB/s per node
-Grid       : Message        :  Stencil 12.4461 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664070
-Grid       : Message        : Average mflops/s per call per node       : 805521
-Grid       : Message        : Average mflops/s per call per node       : 825787
-Grid       : Message        : Average mflops/s per call per node       : 666217
-Grid       : Message        : Average mflops/s per call per node (full): 314056
-Grid       : Message        : Average mflops/s per call per node (full): 435679
-Grid       : Message        : Average mflops/s per call per node (full): 431690
-Grid       : Message        : Average mflops/s per call per node (full): 305067
-Grid       : Message        :  Stencil 13.2165 GB/s per node
-Grid       : Message        :  Stencil 11.7598 GB/s per node
-Grid       : Message        :  Stencil 17.8103 GB/s per node
-Grid       : Message        :  Stencil 11.9989 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667313
-Grid       : Message        : Average mflops/s per call per node       : 811988
-Grid       : Message        : Average mflops/s per call per node       : 822549
-Grid       : Message        : Average mflops/s per call per node       : 667800
-Grid       : Message        : Average mflops/s per call per node (full): 314792
-Grid       : Message        : Average mflops/s per call per node (full): 356582
-Grid       : Message        : Average mflops/s per call per node (full): 445769
-Grid       : Message        : Average mflops/s per call per node (full): 302482
-Grid       : Message        :  Stencil 13.2337 GB/s per node
-Grid       : Message        :  Stencil 16.9512 GB/s per node
-Grid       : Message        :  Stencil 17.8045 GB/s per node
-Grid       : Message        :  Stencil 13.8251 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664088
-Grid       : Message        : Average mflops/s per call per node       : 803417
-Grid       : Message        : Average mflops/s per call per node       : 818634
-Grid       : Message        : Average mflops/s per call per node       : 663948
-Grid       : Message        : Average mflops/s per call per node (full): 314716
-Grid       : Message        : Average mflops/s per call per node (full): 437263
-Grid       : Message        : Average mflops/s per call per node (full): 445292
-Grid       : Message        : Average mflops/s per call per node (full): 306747
-Grid       : Message        :  Stencil 12.4742 GB/s per node
-Grid       : Message        :  Stencil 17.4951 GB/s per node
-Grid       : Message        :  Stencil 17.6713 GB/s per node
-Grid       : Message        :  Stencil 13.3499 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665881
-Grid       : Message        : Average mflops/s per call per node       : 801179
-Grid       : Message        : Average mflops/s per call per node       : 824577
-Grid       : Message        : Average mflops/s per call per node       : 667595
-Grid       : Message        : Average mflops/s per call per node (full): 311674
-Grid       : Message        : Average mflops/s per call per node (full): 438499
-Grid       : Message        : Average mflops/s per call per node (full): 445682
-Grid       : Message        : Average mflops/s per call per node (full): 306601
-Grid       : Message        :  Stencil 12.5619 GB/s per node
-Grid       : Message        :  Stencil 11.8294 GB/s per node
-Grid       : Message        :  Stencil 16.3323 GB/s per node
-Grid       : Message        :  Stencil 11.958 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667462
-Grid       : Message        : Average mflops/s per call per node       : 803788
-Grid       : Message        : Average mflops/s per call per node       : 827608
-Grid       : Message        : Average mflops/s per call per node       : 670717
-Grid       : Message        : Average mflops/s per call per node (full): 313219
-Grid       : Message        : Average mflops/s per call per node (full): 357704
-Grid       : Message        : Average mflops/s per call per node (full): 420800
-Grid       : Message        : Average mflops/s per call per node (full): 301473
-Grid       : Message        :  Stencil 13.1724 GB/s per node
-Grid       : Message        :  Stencil 17.0143 GB/s per node
-Grid       : Message        :  Stencil 17.4939 GB/s per node
-Grid       : Message        :  Stencil 13.4218 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669132
-Grid       : Message        : Average mflops/s per call per node       : 805514
-Grid       : Message        : Average mflops/s per call per node       : 823415
-Grid       : Message        : Average mflops/s per call per node       : 666192
-Grid       : Message        : Average mflops/s per call per node (full): 315321
-Grid       : Message        : Average mflops/s per call per node (full): 437871
-Grid       : Message        : Average mflops/s per call per node (full): 445335
-Grid       : Message        : Average mflops/s per call per node (full): 306812
-Grid       : Message        :  Stencil 13.7464 GB/s per node
-Grid       : Message        :  Stencil 16.6745 GB/s per node
-Grid       : Message        :  Stencil 17.1911 GB/s per node
-Grid       : Message        :  Stencil 15.0666 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664781
-Grid       : Message        : Average mflops/s per call per node       : 803928
-Grid       : Message        : Average mflops/s per call per node       : 822713
-Grid       : Message        : Average mflops/s per call per node       : 658377
-Grid       : Message        : Average mflops/s per call per node (full): 316466
-Grid       : Message        : Average mflops/s per call per node (full): 436232
-Grid       : Message        : Average mflops/s per call per node (full): 442786
-Grid       : Message        : Average mflops/s per call per node (full): 306735
-Grid       : Message        :  Stencil 14.3112 GB/s per node
-Grid       : Message        :  Stencil 16.9462 GB/s per node
-Grid       : Message        :  Stencil 17.9614 GB/s per node
-Grid       : Message        :  Stencil 13.8318 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665989
-Grid       : Message        : Average mflops/s per call per node       : 803710
-Grid       : Message        : Average mflops/s per call per node       : 819066
-Grid       : Message        : Average mflops/s per call per node       : 664430
-Grid       : Message        : Average mflops/s per call per node (full): 316879
-Grid       : Message        : Average mflops/s per call per node (full): 438108
-Grid       : Message        : Average mflops/s per call per node (full): 446138
-Grid       : Message        : Average mflops/s per call per node (full): 307528
-Grid       : Message        :  Stencil 13.2229 GB/s per node
-Grid       : Message        :  Stencil 17.2677 GB/s per node
-Grid       : Message        :  Stencil 17.7745 GB/s per node
-Grid       : Message        :  Stencil 12.5946 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668613
-Grid       : Message        : Average mflops/s per call per node       : 805485
-Grid       : Message        : Average mflops/s per call per node       : 825436
-Grid       : Message        : Average mflops/s per call per node       : 665651
-Grid       : Message        : Average mflops/s per call per node (full): 316275
-Grid       : Message        : Average mflops/s per call per node (full): 432677
-Grid       : Message        : Average mflops/s per call per node (full): 446369
-Grid       : Message        : Average mflops/s per call per node (full): 303448
-Grid       : Message        :  Stencil 13.5782 GB/s per node
-Grid       : Message        :  Stencil 17.1672 GB/s per node
-Grid       : Message        :  Stencil 17.4806 GB/s per node
-Grid       : Message        :  Stencil 13.6005 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665271
-Grid       : Message        : Average mflops/s per call per node       : 803369
-Grid       : Message        : Average mflops/s per call per node       : 826869
-Grid       : Message        : Average mflops/s per call per node       : 661994
-Grid       : Message        : Average mflops/s per call per node (full): 315703
-Grid       : Message        : Average mflops/s per call per node (full): 438789
-Grid       : Message        : Average mflops/s per call per node (full): 445241
-Grid       : Message        : Average mflops/s per call per node (full): 306011
-Grid       : Message        :  Stencil 13.0431 GB/s per node
-Grid       : Message        :  Stencil 16.4806 GB/s per node
-Grid       : Message        :  Stencil 18.8665 GB/s per node
-Grid       : Message        :  Stencil 13.3775 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667692
-Grid       : Message        : Average mflops/s per call per node       : 805333
-Grid       : Message        : Average mflops/s per call per node       : 829122
-Grid       : Message        : Average mflops/s per call per node       : 663530
-Grid       : Message        : Average mflops/s per call per node (full): 315212
-Grid       : Message        : Average mflops/s per call per node (full): 434699
-Grid       : Message        : Average mflops/s per call per node (full): 450563
-Grid       : Message        : Average mflops/s per call per node (full): 306393
-Grid       : Message        :  Stencil 12.7172 GB/s per node
-Grid       : Message        :  Stencil 16.4615 GB/s per node
-Grid       : Message        :  Stencil 19.0358 GB/s per node
-Grid       : Message        :  Stencil 13.4485 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669490
-Grid       : Message        : Average mflops/s per call per node       : 800479
-Grid       : Message        : Average mflops/s per call per node       : 822324
-Grid       : Message        : Average mflops/s per call per node       : 660896
-Grid       : Message        : Average mflops/s per call per node (full): 314962
-Grid       : Message        : Average mflops/s per call per node (full): 431262
-Grid       : Message        : Average mflops/s per call per node (full): 450903
-Grid       : Message        : Average mflops/s per call per node (full): 306073
-Grid       : Message        :  Stencil 13.6954 GB/s per node
-Grid       : Message        :  Stencil 16.7403 GB/s per node
-Grid       : Message        :  Stencil 17.4408 GB/s per node
-Grid       : Message        :  Stencil 12.6016 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663190
-Grid       : Message        : Average mflops/s per call per node       : 800581
-Grid       : Message        : Average mflops/s per call per node       : 826168
-Grid       : Message        : Average mflops/s per call per node       : 665855
-Grid       : Message        : Average mflops/s per call per node (full): 313723
-Grid       : Message        : Average mflops/s per call per node (full): 437454
-Grid       : Message        : Average mflops/s per call per node (full): 445763
-Grid       : Message        : Average mflops/s per call per node (full): 304748
-Grid       : Message        :  Stencil 14.0301 GB/s per node
-Grid       : Message        :  Stencil 16.8619 GB/s per node
-Grid       : Message        :  Stencil 17.9328 GB/s per node
-Grid       : Message        :  Stencil 14.2081 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665022
-Grid       : Message        : Average mflops/s per call per node       : 808321
-Grid       : Message        : Average mflops/s per call per node       : 826120
-Grid       : Message        : Average mflops/s per call per node       : 662543
-Grid       : Message        : Average mflops/s per call per node (full): 316402
-Grid       : Message        : Average mflops/s per call per node (full): 438387
-Grid       : Message        : Average mflops/s per call per node (full): 447424
-Grid       : Message        : Average mflops/s per call per node (full): 307661
-Grid       : Message        :  Stencil 14.4306 GB/s per node
-Grid       : Message        :  Stencil 16.4094 GB/s per node
-Grid       : Message        :  Stencil 17.3191 GB/s per node
-Grid       : Message        :  Stencil 13.8437 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664494
-Grid       : Message        : Average mflops/s per call per node       : 806160
-Grid       : Message        : Average mflops/s per call per node       : 820601
-Grid       : Message        : Average mflops/s per call per node       : 661708
-Grid       : Message        : Average mflops/s per call per node (full): 315384
-Grid       : Message        : Average mflops/s per call per node (full): 434299
-Grid       : Message        : Average mflops/s per call per node (full): 443111
-Grid       : Message        : Average mflops/s per call per node (full): 306571
-Grid       : Message        :  Stencil 13.5568 GB/s per node
-Grid       : Message        :  Stencil 18.3458 GB/s per node
-Grid       : Message        :  Stencil 18.3744 GB/s per node
-Grid       : Message        :  Stencil 12.6788 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664391
-Grid       : Message        : Average mflops/s per call per node       : 798768
-Grid       : Message        : Average mflops/s per call per node       : 826634
-Grid       : Message        : Average mflops/s per call per node       : 663137
-Grid       : Message        : Average mflops/s per call per node (full): 315463
-Grid       : Message        : Average mflops/s per call per node (full): 440577
-Grid       : Message        : Average mflops/s per call per node (full): 449206
-Grid       : Message        : Average mflops/s per call per node (full): 305236
-Grid       : Message        :  Stencil 12.6774 GB/s per node
-Grid       : Message        :  Stencil 17.6112 GB/s per node
-Grid       : Message        :  Stencil 17.2765 GB/s per node
-Grid       : Message        :  Stencil 13.9594 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670856
-Grid       : Message        : Average mflops/s per call per node       : 804658
-Grid       : Message        : Average mflops/s per call per node       : 818594
-Grid       : Message        : Average mflops/s per call per node       : 654510
-Grid       : Message        : Average mflops/s per call per node (full): 313844
-Grid       : Message        : Average mflops/s per call per node (full): 438540
-Grid       : Message        : Average mflops/s per call per node (full): 443813
-Grid       : Message        : Average mflops/s per call per node (full): 305913
-Grid       : Message        :  Stencil 14.0668 GB/s per node
-Grid       : Message        :  Stencil 16.7665 GB/s per node
-Grid       : Message        :  Stencil 17.9828 GB/s per node
-Grid       : Message        :  Stencil 13.3468 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665086
-Grid       : Message        : Average mflops/s per call per node       : 805773
-Grid       : Message        : Average mflops/s per call per node       : 826598
-Grid       : Message        : Average mflops/s per call per node       : 665131
-Grid       : Message        : Average mflops/s per call per node (full): 314886
-Grid       : Message        : Average mflops/s per call per node (full): 437424
-Grid       : Message        : Average mflops/s per call per node (full): 445201
-Grid       : Message        : Average mflops/s per call per node (full): 303112
-Grid       : Message        :  Stencil 12.9395 GB/s per node
-Grid       : Message        :  Stencil 16.6062 GB/s per node
-Grid       : Message        :  Stencil 17.4802 GB/s per node
-Grid       : Message        :  Stencil 13.1207 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667340
-Grid       : Message        : Average mflops/s per call per node       : 802719
-Grid       : Message        : Average mflops/s per call per node       : 821553
-Grid       : Message        : Average mflops/s per call per node       : 660221
-Grid       : Message        : Average mflops/s per call per node (full): 313758
-Grid       : Message        : Average mflops/s per call per node (full): 435392
-Grid       : Message        : Average mflops/s per call per node (full): 443062
-Grid       : Message        : Average mflops/s per call per node (full): 304683
-Grid       : Message        :  Stencil 13.5273 GB/s per node
-Grid       : Message        :  Stencil 16.7218 GB/s per node
-Grid       : Message        :  Stencil 17.3427 GB/s per node
-Grid       : Message        :  Stencil 12.3055 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666951
-Grid       : Message        : Average mflops/s per call per node       : 808340
-Grid       : Message        : Average mflops/s per call per node       : 825856
-Grid       : Message        : Average mflops/s per call per node       : 669653
-Grid       : Message        : Average mflops/s per call per node (full): 316393
-Grid       : Message        : Average mflops/s per call per node (full): 437208
-Grid       : Message        : Average mflops/s per call per node (full): 444151
-Grid       : Message        : Average mflops/s per call per node (full): 304506
-Grid       : Message        :  Stencil 12.9419 GB/s per node
-Grid       : Message        :  Stencil 17.7185 GB/s per node
-Grid       : Message        :  Stencil 17.6023 GB/s per node
-Grid       : Message        :  Stencil 13.1646 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671486
-Grid       : Message        : Average mflops/s per call per node       : 803835
-Grid       : Message        : Average mflops/s per call per node       : 812919
-Grid       : Message        : Average mflops/s per call per node       : 661156
-Grid       : Message        : Average mflops/s per call per node (full): 316181
-Grid       : Message        : Average mflops/s per call per node (full): 439019
-Grid       : Message        : Average mflops/s per call per node (full): 443163
-Grid       : Message        : Average mflops/s per call per node (full): 305543
-Grid       : Message        :  Stencil 12.8064 GB/s per node
-Grid       : Message        :  Stencil 17.3831 GB/s per node
-Grid       : Message        :  Stencil 19.2456 GB/s per node
-Grid       : Message        :  Stencil 13.7529 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671019
-Grid       : Message        : Average mflops/s per call per node       : 804072
-Grid       : Message        : Average mflops/s per call per node       : 818083
-Grid       : Message        : Average mflops/s per call per node       : 657458
-Grid       : Message        : Average mflops/s per call per node (full): 315531
-Grid       : Message        : Average mflops/s per call per node (full): 437530
-Grid       : Message        : Average mflops/s per call per node (full): 447228
-Grid       : Message        : Average mflops/s per call per node (full): 305933
-Grid       : Message        :  Stencil 13.0671 GB/s per node
-Grid       : Message        :  Stencil 16.3692 GB/s per node
-Grid       : Message        :  Stencil 17.5399 GB/s per node
-Grid       : Message        :  Stencil 12.765 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668696
-Grid       : Message        : Average mflops/s per call per node       : 806903
-Grid       : Message        : Average mflops/s per call per node       : 815258
-Grid       : Message        : Average mflops/s per call per node       : 662416
-Grid       : Message        : Average mflops/s per call per node (full): 314894
-Grid       : Message        : Average mflops/s per call per node (full): 433481
-Grid       : Message        : Average mflops/s per call per node (full): 443106
-Grid       : Message        : Average mflops/s per call per node (full): 304658
-Grid       : Message        :  Stencil 13.2608 GB/s per node
-Grid       : Message        :  Stencil 16.5387 GB/s per node
-Grid       : Message        :  Stencil 17.9303 GB/s per node
-Grid       : Message        :  Stencil 12.9177 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665759
-Grid       : Message        : Average mflops/s per call per node       : 804452
-Grid       : Message        : Average mflops/s per call per node       : 822492
-Grid       : Message        : Average mflops/s per call per node       : 664038
-Grid       : Message        : Average mflops/s per call per node (full): 315706
-Grid       : Message        : Average mflops/s per call per node (full): 435423
-Grid       : Message        : Average mflops/s per call per node (full): 446707
-Grid       : Message        : Average mflops/s per call per node (full): 303776
-Grid       : Message        :  Stencil 12.7231 GB/s per node
-Grid       : Message        :  Stencil 17.212 GB/s per node
-Grid       : Message        :  Stencil 17.2632 GB/s per node
-Grid       : Message        :  Stencil 15.0456 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668041
-Grid       : Message        : Average mflops/s per call per node       : 804260
-Grid       : Message        : Average mflops/s per call per node       : 828111
-Grid       : Message        : Average mflops/s per call per node       : 659790
-Grid       : Message        : Average mflops/s per call per node (full): 315169
-Grid       : Message        : Average mflops/s per call per node (full): 436572
-Grid       : Message        : Average mflops/s per call per node (full): 445038
-Grid       : Message        : Average mflops/s per call per node (full): 307447
-Grid       : Message        :  Stencil 13.9317 GB/s per node
-Grid       : Message        :  Stencil 17.6126 GB/s per node
-Grid       : Message        :  Stencil 17.9129 GB/s per node
-Grid       : Message        :  Stencil 12.1942 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663571
-Grid       : Message        : Average mflops/s per call per node       : 806005
-Grid       : Message        : Average mflops/s per call per node       : 818928
-Grid       : Message        : Average mflops/s per call per node       : 667109
-Grid       : Message        : Average mflops/s per call per node (full): 315229
-Grid       : Message        : Average mflops/s per call per node (full): 437729
-Grid       : Message        : Average mflops/s per call per node (full): 443696
-Grid       : Message        : Average mflops/s per call per node (full): 303522
-Grid       : Message        :  Stencil 13.8441 GB/s per node
-Grid       : Message        :  Stencil 16.8207 GB/s per node
-Grid       : Message        :  Stencil 17.3388 GB/s per node
-Grid       : Message        :  Stencil 12.6899 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667025
-Grid       : Message        : Average mflops/s per call per node       : 803488
-Grid       : Message        : Average mflops/s per call per node       : 826681
-Grid       : Message        : Average mflops/s per call per node       : 668904
-Grid       : Message        : Average mflops/s per call per node (full): 316439
-Grid       : Message        : Average mflops/s per call per node (full): 438327
-Grid       : Message        : Average mflops/s per call per node (full): 443946
-Grid       : Message        : Average mflops/s per call per node (full): 304893
-Grid       : Message        :  Stencil 14.6365 GB/s per node
-Grid       : Message        :  Stencil 16.8923 GB/s per node
-Grid       : Message        :  Stencil 18.0901 GB/s per node
-Grid       : Message        :  Stencil 14.1822 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666147
-Grid       : Message        : Average mflops/s per call per node       : 804740
-Grid       : Message        : Average mflops/s per call per node       : 822198
-Grid       : Message        : Average mflops/s per call per node       : 661873
-Grid       : Message        : Average mflops/s per call per node (full): 316893
-Grid       : Message        : Average mflops/s per call per node (full): 437004
-Grid       : Message        : Average mflops/s per call per node (full): 446809
-Grid       : Message        : Average mflops/s per call per node (full): 307507
-Grid       : Message        :  Stencil 13.0344 GB/s per node
-Grid       : Message        :  Stencil 16.2898 GB/s per node
-Grid       : Message        :  Stencil 18.1261 GB/s per node
-Grid       : Message        :  Stencil 13.4165 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669087
-Grid       : Message        : Average mflops/s per call per node       : 804366
-Grid       : Message        : Average mflops/s per call per node       : 822323
-Grid       : Message        : Average mflops/s per call per node       : 659677
-Grid       : Message        : Average mflops/s per call per node (full): 315904
-Grid       : Message        : Average mflops/s per call per node (full): 432455
-Grid       : Message        : Average mflops/s per call per node (full): 445337
-Grid       : Message        : Average mflops/s per call per node (full): 304386
-Grid       : Message        :  Stencil 13.7753 GB/s per node
-Grid       : Message        :  Stencil 11.5873 GB/s per node
-Grid       : Message        :  Stencil 17.7056 GB/s per node
-Grid       : Message        :  Stencil 12.6323 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666533
-Grid       : Message        : Average mflops/s per call per node       : 808259
-Grid       : Message        : Average mflops/s per call per node       : 821688
-Grid       : Message        : Average mflops/s per call per node       : 669442
-Grid       : Message        : Average mflops/s per call per node (full): 315905
-Grid       : Message        : Average mflops/s per call per node (full): 352761
-Grid       : Message        : Average mflops/s per call per node (full): 444870
-Grid       : Message        : Average mflops/s per call per node (full): 305567
-Grid       : Message        :  Stencil 13.3282 GB/s per node
-Grid       : Message        :  Stencil 16.8213 GB/s per node
-Grid       : Message        :  Stencil 17.8142 GB/s per node
-Grid       : Message        :  Stencil 12.2119 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670208
-Grid       : Message        : Average mflops/s per call per node       : 805861
-Grid       : Message        : Average mflops/s per call per node       : 822623
-Grid       : Message        : Average mflops/s per call per node       : 666740
-Grid       : Message        : Average mflops/s per call per node (full): 316655
-Grid       : Message        : Average mflops/s per call per node (full): 432955
-Grid       : Message        : Average mflops/s per call per node (full): 446403
-Grid       : Message        : Average mflops/s per call per node (full): 303583
-Grid       : Message        :  Stencil 12.71 GB/s per node
-Grid       : Message        :  Stencil 17.2789 GB/s per node
-Grid       : Message        :  Stencil 17.3154 GB/s per node
-Grid       : Message        :  Stencil 12.9706 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668272
-Grid       : Message        : Average mflops/s per call per node       : 803593
-Grid       : Message        : Average mflops/s per call per node       : 821190
-Grid       : Message        : Average mflops/s per call per node       : 665746
-Grid       : Message        : Average mflops/s per call per node (full): 315074
-Grid       : Message        : Average mflops/s per call per node (full): 436341
-Grid       : Message        : Average mflops/s per call per node (full): 438150
-Grid       : Message        : Average mflops/s per call per node (full): 304502
-Grid       : Message        :  Stencil 12.5427 GB/s per node
-Grid       : Message        :  Stencil 17.1143 GB/s per node
-Grid       : Message        :  Stencil 17.4695 GB/s per node
-Grid       : Message        :  Stencil 14.2233 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668353
-Grid       : Message        : Average mflops/s per call per node       : 807108
-Grid       : Message        : Average mflops/s per call per node       : 826014
-Grid       : Message        : Average mflops/s per call per node       : 661985
-Grid       : Message        : Average mflops/s per call per node (full): 313652
-Grid       : Message        : Average mflops/s per call per node (full): 435772
-Grid       : Message        : Average mflops/s per call per node (full): 444672
-Grid       : Message        : Average mflops/s per call per node (full): 306525
-Grid       : Message        :  Stencil 12.3856 GB/s per node
-Grid       : Message        :  Stencil 16.6937 GB/s per node
-Grid       : Message        :  Stencil 18.4506 GB/s per node
-Grid       : Message        :  Stencil 13.9017 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671646
-Grid       : Message        : Average mflops/s per call per node       : 799233
-Grid       : Message        : Average mflops/s per call per node       : 825952
-Grid       : Message        : Average mflops/s per call per node       : 666904
-Grid       : Message        : Average mflops/s per call per node (full): 313151
-Grid       : Message        : Average mflops/s per call per node (full): 435437
-Grid       : Message        : Average mflops/s per call per node (full): 449338
-Grid       : Message        : Average mflops/s per call per node (full): 307579
-Grid       : Message        :  Stencil 13.5081 GB/s per node
-Grid       : Message        :  Stencil 16.904 GB/s per node
-Grid       : Message        :  Stencil 17.8221 GB/s per node
-Grid       : Message        :  Stencil 11.9864 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667372
-Grid       : Message        : Average mflops/s per call per node       : 806954
-Grid       : Message        : Average mflops/s per call per node       : 823843
-Grid       : Message        : Average mflops/s per call per node       : 668278
-Grid       : Message        : Average mflops/s per call per node (full): 315974
-Grid       : Message        : Average mflops/s per call per node (full): 436821
-Grid       : Message        : Average mflops/s per call per node (full): 437610
-Grid       : Message        : Average mflops/s per call per node (full): 301928
-Grid       : Message        :  Stencil 13.2842 GB/s per node
-Grid       : Message        :  Stencil 16.9778 GB/s per node
-Grid       : Message        :  Stencil 17.1985 GB/s per node
-Grid       : Message        :  Stencil 12.6819 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663084
-Grid       : Message        : Average mflops/s per call per node       : 806820
-Grid       : Message        : Average mflops/s per call per node       : 819492
-Grid       : Message        : Average mflops/s per call per node       : 663353
-Grid       : Message        : Average mflops/s per call per node (full): 315298
-Grid       : Message        : Average mflops/s per call per node (full): 438561
-Grid       : Message        : Average mflops/s per call per node (full): 441997
-Grid       : Message        : Average mflops/s per call per node (full): 304187
-Grid       : Message        :  Stencil 12.9111 GB/s per node
-Grid       : Message        :  Stencil 17.6963 GB/s per node
-Grid       : Message        :  Stencil 17.6975 GB/s per node
-Grid       : Message        :  Stencil 13.016 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664176
-Grid       : Message        : Average mflops/s per call per node       : 800210
-Grid       : Message        : Average mflops/s per call per node       : 821890
-Grid       : Message        : Average mflops/s per call per node       : 662729
-Grid       : Message        : Average mflops/s per call per node (full): 314576
-Grid       : Message        : Average mflops/s per call per node (full): 437851
-Grid       : Message        : Average mflops/s per call per node (full): 446601
-Grid       : Message        : Average mflops/s per call per node (full): 305381
-Grid       : Message        :  Stencil 13.1248 GB/s per node
-Grid       : Message        :  Stencil 11.7779 GB/s per node
-Grid       : Message        :  Stencil 17.7619 GB/s per node
-Grid       : Message        :  Stencil 13.4022 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664419
-Grid       : Message        : Average mflops/s per call per node       : 810492
-Grid       : Message        : Average mflops/s per call per node       : 823930
-Grid       : Message        : Average mflops/s per call per node       : 665522
-Grid       : Message        : Average mflops/s per call per node (full): 314881
-Grid       : Message        : Average mflops/s per call per node (full): 356528
-Grid       : Message        : Average mflops/s per call per node (full): 445937
-Grid       : Message        : Average mflops/s per call per node (full): 306116
-Grid       : Message        :  Stencil 14.3878 GB/s per node
-Grid       : Message        :  Stencil 16.7257 GB/s per node
-Grid       : Message        :  Stencil 17.8578 GB/s per node
-Grid       : Message        :  Stencil 13.9036 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660208
-Grid       : Message        : Average mflops/s per call per node       : 807016
-Grid       : Message        : Average mflops/s per call per node       : 823662
-Grid       : Message        : Average mflops/s per call per node       : 664817
-Grid       : Message        : Average mflops/s per call per node (full): 315583
-Grid       : Message        : Average mflops/s per call per node (full): 436472
-Grid       : Message        : Average mflops/s per call per node (full): 446283
-Grid       : Message        : Average mflops/s per call per node (full): 306769
-Grid       : Message        :  Stencil 13.7372 GB/s per node
-Grid       : Message        :  Stencil 14.8812 GB/s per node
-Grid       : Message        :  Stencil 17.5616 GB/s per node
-Grid       : Message        :  Stencil 11.8767 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660390
-Grid       : Message        : Average mflops/s per call per node       : 804542
-Grid       : Message        : Average mflops/s per call per node       : 819512
-Grid       : Message        : Average mflops/s per call per node       : 664682
-Grid       : Message        : Average mflops/s per call per node (full): 315855
-Grid       : Message        : Average mflops/s per call per node (full): 414288
-Grid       : Message        : Average mflops/s per call per node (full): 442000
-Grid       : Message        : Average mflops/s per call per node (full): 300282
-Grid       : Message        :  Stencil 14.2464 GB/s per node
-Grid       : Message        :  Stencil 17.0363 GB/s per node
-Grid       : Message        :  Stencil 17.9915 GB/s per node
-Grid       : Message        :  Stencil 13.461 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661537
-Grid       : Message        : Average mflops/s per call per node       : 805651
-Grid       : Message        : Average mflops/s per call per node       : 824045
-Grid       : Message        : Average mflops/s per call per node       : 667438
-Grid       : Message        : Average mflops/s per call per node (full): 316224
-Grid       : Message        : Average mflops/s per call per node (full): 435645
-Grid       : Message        : Average mflops/s per call per node (full): 446580
-Grid       : Message        : Average mflops/s per call per node (full): 307108
-Grid       : Message        :  Stencil 14.2497 GB/s per node
-Grid       : Message        :  Stencil 16.2639 GB/s per node
-Grid       : Message        :  Stencil 17.1763 GB/s per node
-Grid       : Message        :  Stencil 12.0851 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659856
-Grid       : Message        : Average mflops/s per call per node       : 808352
-Grid       : Message        : Average mflops/s per call per node       : 823527
-Grid       : Message        : Average mflops/s per call per node       : 662137
-Grid       : Message        : Average mflops/s per call per node (full): 315624
-Grid       : Message        : Average mflops/s per call per node (full): 431227
-Grid       : Message        : Average mflops/s per call per node (full): 441847
-Grid       : Message        : Average mflops/s per call per node (full): 302014
-Grid       : Message        :  Stencil 12.6238 GB/s per node
-Grid       : Message        :  Stencil 16.3639 GB/s per node
-Grid       : Message        :  Stencil 17.937 GB/s per node
-Grid       : Message        :  Stencil 12.8845 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664350
-Grid       : Message        : Average mflops/s per call per node       : 803314
-Grid       : Message        : Average mflops/s per call per node       : 819809
-Grid       : Message        : Average mflops/s per call per node       : 660241
-Grid       : Message        : Average mflops/s per call per node (full): 312411
-Grid       : Message        : Average mflops/s per call per node (full): 431811
-Grid       : Message        : Average mflops/s per call per node (full): 445748
-Grid       : Message        : Average mflops/s per call per node (full): 301276
-Grid       : Message        :  Stencil 12.5847 GB/s per node
-Grid       : Message        :  Stencil 17.6308 GB/s per node
-Grid       : Message        :  Stencil 18.0077 GB/s per node
-Grid       : Message        :  Stencil 12.2625 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662560
-Grid       : Message        : Average mflops/s per call per node       : 800141
-Grid       : Message        : Average mflops/s per call per node       : 819948
-Grid       : Message        : Average mflops/s per call per node       : 665869
-Grid       : Message        : Average mflops/s per call per node (full): 313352
-Grid       : Message        : Average mflops/s per call per node (full): 439427
-Grid       : Message        : Average mflops/s per call per node (full): 444876
-Grid       : Message        : Average mflops/s per call per node (full): 304118
-Grid       : Message        :  Stencil 13.4228 GB/s per node
-Grid       : Message        :  Stencil 16.4704 GB/s per node
-Grid       : Message        :  Stencil 17.0233 GB/s per node
-Grid       : Message        :  Stencil 12.4666 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659557
-Grid       : Message        : Average mflops/s per call per node       : 805825
-Grid       : Message        : Average mflops/s per call per node       : 822933
-Grid       : Message        : Average mflops/s per call per node       : 663659
-Grid       : Message        : Average mflops/s per call per node (full): 314515
-Grid       : Message        : Average mflops/s per call per node (full): 435573
-Grid       : Message        : Average mflops/s per call per node (full): 440776
-Grid       : Message        : Average mflops/s per call per node (full): 304630
-Grid       : Message        :  Stencil 12.4644 GB/s per node
-Grid       : Message        :  Stencil 8.22093 GB/s per node
-Grid       : Message        :  Stencil 17.7152 GB/s per node
-Grid       : Message        :  Stencil 14.6792 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665642
-Grid       : Message        : Average mflops/s per call per node       : 808130
-Grid       : Message        : Average mflops/s per call per node       : 823904
-Grid       : Message        : Average mflops/s per call per node       : 661846
-Grid       : Message        : Average mflops/s per call per node (full): 312503
-Grid       : Message        : Average mflops/s per call per node (full): 272466
-Grid       : Message        : Average mflops/s per call per node (full): 445081
-Grid       : Message        : Average mflops/s per call per node (full): 306767
-Grid       : Message        :  Stencil 12.6647 GB/s per node
-Grid       : Message        :  Stencil 17.5637 GB/s per node
-Grid       : Message        :  Stencil 16.5685 GB/s per node
-Grid       : Message        :  Stencil 12.8074 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664429
-Grid       : Message        : Average mflops/s per call per node       : 804657
-Grid       : Message        : Average mflops/s per call per node       : 825316
-Grid       : Message        : Average mflops/s per call per node       : 662416
-Grid       : Message        : Average mflops/s per call per node (full): 314387
-Grid       : Message        : Average mflops/s per call per node (full): 439744
-Grid       : Message        : Average mflops/s per call per node (full): 429863
-Grid       : Message        : Average mflops/s per call per node (full): 304961
-Grid       : Message        :  Stencil 13.6648 GB/s per node
-Grid       : Message        :  Stencil 17.817 GB/s per node
-Grid       : Message        :  Stencil 16.2935 GB/s per node
-Grid       : Message        :  Stencil 12.3396 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665153
-Grid       : Message        : Average mflops/s per call per node       : 799606
-Grid       : Message        : Average mflops/s per call per node       : 823484
-Grid       : Message        : Average mflops/s per call per node       : 664285
-Grid       : Message        : Average mflops/s per call per node (full): 315681
-Grid       : Message        : Average mflops/s per call per node (full): 438394
-Grid       : Message        : Average mflops/s per call per node (full): 421651
-Grid       : Message        : Average mflops/s per call per node (full): 302749
-Grid       : Message        :  Stencil 12.9152 GB/s per node
-Grid       : Message        :  Stencil 17.2826 GB/s per node
-Grid       : Message        :  Stencil 17.2844 GB/s per node
-Grid       : Message        :  Stencil 14.1308 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671951
-Grid       : Message        : Average mflops/s per call per node       : 805099
-Grid       : Message        : Average mflops/s per call per node       : 820406
-Grid       : Message        : Average mflops/s per call per node       : 666892
-Grid       : Message        : Average mflops/s per call per node (full): 315966
-Grid       : Message        : Average mflops/s per call per node (full): 440134
-Grid       : Message        : Average mflops/s per call per node (full): 442765
-Grid       : Message        : Average mflops/s per call per node (full): 305421
-Grid       : Message        :  Stencil 14.0159 GB/s per node
-Grid       : Message        :  Stencil 16.8766 GB/s per node
-Grid       : Message        :  Stencil 17.9298 GB/s per node
-Grid       : Message        :  Stencil 12.887 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663984
-Grid       : Message        : Average mflops/s per call per node       : 804339
-Grid       : Message        : Average mflops/s per call per node       : 825334
-Grid       : Message        : Average mflops/s per call per node       : 662330
-Grid       : Message        : Average mflops/s per call per node (full): 315302
-Grid       : Message        : Average mflops/s per call per node (full): 435971
-Grid       : Message        : Average mflops/s per call per node (full): 446246
-Grid       : Message        : Average mflops/s per call per node (full): 304300
-Grid       : Message        :  Stencil 14.5695 GB/s per node
-Grid       : Message        :  Stencil 16.8476 GB/s per node
-Grid       : Message        :  Stencil 17.7522 GB/s per node
-Grid       : Message        :  Stencil 14.0278 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663605
-Grid       : Message        : Average mflops/s per call per node       : 803894
-Grid       : Message        : Average mflops/s per call per node       : 828041
-Grid       : Message        : Average mflops/s per call per node       : 665135
-Grid       : Message        : Average mflops/s per call per node (full): 317216
-Grid       : Message        : Average mflops/s per call per node (full): 435113
-Grid       : Message        : Average mflops/s per call per node (full): 446534
-Grid       : Message        : Average mflops/s per call per node (full): 307800
-Grid       : Message        :  Stencil 12.3099 GB/s per node
-Grid       : Message        :  Stencil 17.0746 GB/s per node
-Grid       : Message        :  Stencil 17.5469 GB/s per node
-Grid       : Message        :  Stencil 13.6681 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669123
-Grid       : Message        : Average mflops/s per call per node       : 801928
-Grid       : Message        : Average mflops/s per call per node       : 824345
-Grid       : Message        : Average mflops/s per call per node       : 659179
-Grid       : Message        : Average mflops/s per call per node (full): 312030
-Grid       : Message        : Average mflops/s per call per node (full): 432529
-Grid       : Message        : Average mflops/s per call per node (full): 446152
-Grid       : Message        : Average mflops/s per call per node (full): 306748
-Grid       : Message        :  Stencil 12.8134 GB/s per node
-Grid       : Message        :  Stencil 18.2851 GB/s per node
-Grid       : Message        :  Stencil 17.6986 GB/s per node
-Grid       : Message        :  Stencil 12.7626 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666400
-Grid       : Message        : Average mflops/s per call per node       : 804575
-Grid       : Message        : Average mflops/s per call per node       : 828300
-Grid       : Message        : Average mflops/s per call per node       : 663369
-Grid       : Message        : Average mflops/s per call per node (full): 313964
-Grid       : Message        : Average mflops/s per call per node (full): 442054
-Grid       : Message        : Average mflops/s per call per node (full): 447876
-Grid       : Message        : Average mflops/s per call per node (full): 305509
-Grid       : Message        :  Stencil 13.7164 GB/s per node
-Grid       : Message        :  Stencil 14.5666 GB/s per node
-Grid       : Message        :  Stencil 17.8571 GB/s per node
-Grid       : Message        :  Stencil 12.8831 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662667
-Grid       : Message        : Average mflops/s per call per node       : 808341
-Grid       : Message        : Average mflops/s per call per node       : 823866
-Grid       : Message        : Average mflops/s per call per node       : 662070
-Grid       : Message        : Average mflops/s per call per node (full): 315443
-Grid       : Message        : Average mflops/s per call per node (full): 409315
-Grid       : Message        : Average mflops/s per call per node (full): 446895
-Grid       : Message        : Average mflops/s per call per node (full): 304843
-Grid       : Message        :  Stencil 13.3634 GB/s per node
-Grid       : Message        :  Stencil 16.8785 GB/s per node
-Grid       : Message        :  Stencil 17.6547 GB/s per node
-Grid       : Message        :  Stencil 13.929 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665176
-Grid       : Message        : Average mflops/s per call per node       : 804424
-Grid       : Message        : Average mflops/s per call per node       : 821991
-Grid       : Message        : Average mflops/s per call per node       : 659931
-Grid       : Message        : Average mflops/s per call per node (full): 314784
-Grid       : Message        : Average mflops/s per call per node (full): 437025
-Grid       : Message        : Average mflops/s per call per node (full): 446321
-Grid       : Message        : Average mflops/s per call per node (full): 306278
-Grid       : Message        :  Stencil 12.355 GB/s per node
-Grid       : Message        :  Stencil 16.895 GB/s per node
-Grid       : Message        :  Stencil 18.166 GB/s per node
-Grid       : Message        :  Stencil 12.8473 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671496
-Grid       : Message        : Average mflops/s per call per node       : 808884
-Grid       : Message        : Average mflops/s per call per node       : 820349
-Grid       : Message        : Average mflops/s per call per node       : 666351
-Grid       : Message        : Average mflops/s per call per node (full): 312992
-Grid       : Message        : Average mflops/s per call per node (full): 439455
-Grid       : Message        : Average mflops/s per call per node (full): 446030
-Grid       : Message        : Average mflops/s per call per node (full): 306316
-Grid       : Message        :  Stencil 12.6097 GB/s per node
-Grid       : Message        :  Stencil 16.6952 GB/s per node
-Grid       : Message        :  Stencil 17.7702 GB/s per node
-Grid       : Message        :  Stencil 12.9123 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667832
-Grid       : Message        : Average mflops/s per call per node       : 807358
-Grid       : Message        : Average mflops/s per call per node       : 825113
-Grid       : Message        : Average mflops/s per call per node       : 664207
-Grid       : Message        : Average mflops/s per call per node (full): 314251
-Grid       : Message        : Average mflops/s per call per node (full): 436472
-Grid       : Message        : Average mflops/s per call per node (full): 446405
-Grid       : Message        : Average mflops/s per call per node (full): 303965
-Grid       : Message        :  Stencil 14.0821 GB/s per node
-Grid       : Message        :  Stencil 15.1892 GB/s per node
-Grid       : Message        :  Stencil 17.7668 GB/s per node
-Grid       : Message        :  Stencil 12.6889 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664142
-Grid       : Message        : Average mflops/s per call per node       : 806184
-Grid       : Message        : Average mflops/s per call per node       : 826703
-Grid       : Message        : Average mflops/s per call per node       : 659089
-Grid       : Message        : Average mflops/s per call per node (full): 316433
-Grid       : Message        : Average mflops/s per call per node (full): 417763
-Grid       : Message        : Average mflops/s per call per node (full): 445587
-Grid       : Message        : Average mflops/s per call per node (full): 303256
-Grid       : Message        :  Stencil 13.5766 GB/s per node
-Grid       : Message        :  Stencil 17.2951 GB/s per node
-Grid       : Message        :  Stencil 17.4029 GB/s per node
-Grid       : Message        :  Stencil 12.7326 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666642
-Grid       : Message        : Average mflops/s per call per node       : 802032
-Grid       : Message        : Average mflops/s per call per node       : 819241
-Grid       : Message        : Average mflops/s per call per node       : 666999
-Grid       : Message        : Average mflops/s per call per node (full): 316199
-Grid       : Message        : Average mflops/s per call per node (full): 438555
-Grid       : Message        : Average mflops/s per call per node (full): 443653
-Grid       : Message        : Average mflops/s per call per node (full): 305053
-Grid       : Message        :  Stencil 13.2271 GB/s per node
-Grid       : Message        :  Stencil 16.6175 GB/s per node
-Grid       : Message        :  Stencil 18.169 GB/s per node
-Grid       : Message        :  Stencil 12.1386 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668187
-Grid       : Message        : Average mflops/s per call per node       : 804086
-Grid       : Message        : Average mflops/s per call per node       : 826263
-Grid       : Message        : Average mflops/s per call per node       : 664120
-Grid       : Message        : Average mflops/s per call per node (full): 315783
-Grid       : Message        : Average mflops/s per call per node (full): 435193
-Grid       : Message        : Average mflops/s per call per node (full): 446811
-Grid       : Message        : Average mflops/s per call per node (full): 302426
-Grid       : Message        :  Stencil 12.5939 GB/s per node
-Grid       : Message        :  Stencil 16.4433 GB/s per node
-Grid       : Message        :  Stencil 17.8313 GB/s per node
-Grid       : Message        :  Stencil 11.7482 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668082
-Grid       : Message        : Average mflops/s per call per node       : 807633
-Grid       : Message        : Average mflops/s per call per node       : 822405
-Grid       : Message        : Average mflops/s per call per node       : 668530
-Grid       : Message        : Average mflops/s per call per node (full): 314384
-Grid       : Message        : Average mflops/s per call per node (full): 433890
-Grid       : Message        : Average mflops/s per call per node (full): 436885
-Grid       : Message        : Average mflops/s per call per node (full): 294829
-Grid       : Message        :  Stencil 12.9189 GB/s per node
-Grid       : Message        :  Stencil 16.7443 GB/s per node
-Grid       : Message        :  Stencil 18.5545 GB/s per node
-Grid       : Message        :  Stencil 13.4276 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666286
-Grid       : Message        : Average mflops/s per call per node       : 804392
-Grid       : Message        : Average mflops/s per call per node       : 818844
-Grid       : Message        : Average mflops/s per call per node       : 664687
-Grid       : Message        : Average mflops/s per call per node (full): 315228
-Grid       : Message        : Average mflops/s per call per node (full): 436326
-Grid       : Message        : Average mflops/s per call per node (full): 447264
-Grid       : Message        : Average mflops/s per call per node (full): 307040
-Grid       : Message        :  Stencil 13.3248 GB/s per node
-Grid       : Message        :  Stencil 12.1998 GB/s per node
-Grid       : Message        :  Stencil 17.5079 GB/s per node
-Grid       : Message        :  Stencil 13.3229 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667839
-Grid       : Message        : Average mflops/s per call per node       : 813175
-Grid       : Message        : Average mflops/s per call per node       : 814133
-Grid       : Message        : Average mflops/s per call per node       : 665674
-Grid       : Message        : Average mflops/s per call per node (full): 315289
-Grid       : Message        : Average mflops/s per call per node (full): 365548
-Grid       : Message        : Average mflops/s per call per node (full): 443603
-Grid       : Message        : Average mflops/s per call per node (full): 305937
-Grid       : Message        :  Stencil 13.1377 GB/s per node
-Grid       : Message        :  Stencil 16.4798 GB/s per node
-Grid       : Message        :  Stencil 17.8989 GB/s per node
-Grid       : Message        :  Stencil 13.3023 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669090
-Grid       : Message        : Average mflops/s per call per node       : 804535
-Grid       : Message        : Average mflops/s per call per node       : 825306
-Grid       : Message        : Average mflops/s per call per node       : 666189
-Grid       : Message        : Average mflops/s per call per node (full): 315289
-Grid       : Message        : Average mflops/s per call per node (full): 434236
-Grid       : Message        : Average mflops/s per call per node (full): 445521
-Grid       : Message        : Average mflops/s per call per node (full): 303984
-Grid       : Message        :  Stencil 14.0527 GB/s per node
-Grid       : Message        :  Stencil 16.9076 GB/s per node
-Grid       : Message        :  Stencil 17.3601 GB/s per node
-Grid       : Message        :  Stencil 12.4109 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663052
-Grid       : Message        : Average mflops/s per call per node       : 803103
-Grid       : Message        : Average mflops/s per call per node       : 828502
-Grid       : Message        : Average mflops/s per call per node       : 663911
-Grid       : Message        : Average mflops/s per call per node (full): 315560
-Grid       : Message        : Average mflops/s per call per node (full): 436378
-Grid       : Message        : Average mflops/s per call per node (full): 444933
-Grid       : Message        : Average mflops/s per call per node (full): 303394
-Grid       : Message        :  Stencil 13.2076 GB/s per node
-Grid       : Message        :  Stencil 17.8873 GB/s per node
-Grid       : Message        :  Stencil 17.3344 GB/s per node
-Grid       : Message        :  Stencil 13.4283 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666130
-Grid       : Message        : Average mflops/s per call per node       : 802572
-Grid       : Message        : Average mflops/s per call per node       : 824862
-Grid       : Message        : Average mflops/s per call per node       : 665551
-Grid       : Message        : Average mflops/s per call per node (full): 315252
-Grid       : Message        : Average mflops/s per call per node (full): 440432
-Grid       : Message        : Average mflops/s per call per node (full): 443691
-Grid       : Message        : Average mflops/s per call per node (full): 306628
-Grid       : Message        :  Stencil 12.6615 GB/s per node
-Grid       : Message        :  Stencil 17.1256 GB/s per node
-Grid       : Message        :  Stencil 17.3373 GB/s per node
-Grid       : Message        :  Stencil 12.4662 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671242
-Grid       : Message        : Average mflops/s per call per node       : 804096
-Grid       : Message        : Average mflops/s per call per node       : 825872
-Grid       : Message        : Average mflops/s per call per node       : 663829
-Grid       : Message        : Average mflops/s per call per node (full): 314762
-Grid       : Message        : Average mflops/s per call per node (full): 438760
-Grid       : Message        : Average mflops/s per call per node (full): 442389
-Grid       : Message        : Average mflops/s per call per node (full): 304135
-Grid       : Message        :  Stencil 13.9984 GB/s per node
-Grid       : Message        :  Stencil 16.7166 GB/s per node
-Grid       : Message        :  Stencil 17.9318 GB/s per node
-Grid       : Message        :  Stencil 12.5099 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665745
-Grid       : Message        : Average mflops/s per call per node       : 804408
-Grid       : Message        : Average mflops/s per call per node       : 825315
-Grid       : Message        : Average mflops/s per call per node       : 662627
-Grid       : Message        : Average mflops/s per call per node (full): 316107
-Grid       : Message        : Average mflops/s per call per node (full): 436169
-Grid       : Message        : Average mflops/s per call per node (full): 447081
-Grid       : Message        : Average mflops/s per call per node (full): 304059
-Grid       : Message        :  Stencil 14.1307 GB/s per node
-Grid       : Message        :  Stencil 16.955 GB/s per node
-Grid       : Message        :  Stencil 17.413 GB/s per node
-Grid       : Message        :  Stencil 13.2242 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665882
-Grid       : Message        : Average mflops/s per call per node       : 805427
-Grid       : Message        : Average mflops/s per call per node       : 819318
-Grid       : Message        : Average mflops/s per call per node       : 664097
-Grid       : Message        : Average mflops/s per call per node (full): 315293
-Grid       : Message        : Average mflops/s per call per node (full): 433667
-Grid       : Message        : Average mflops/s per call per node (full): 444729
-Grid       : Message        : Average mflops/s per call per node (full): 305803
-Grid       : Message        :  Stencil 14.9627 GB/s per node
-Grid       : Message        :  Stencil 16.6868 GB/s per node
-Grid       : Message        :  Stencil 18.1597 GB/s per node
-Grid       : Message        :  Stencil 13.8765 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663128
-Grid       : Message        : Average mflops/s per call per node       : 805706
-Grid       : Message        : Average mflops/s per call per node       : 816270
-Grid       : Message        : Average mflops/s per call per node       : 662347
-Grid       : Message        : Average mflops/s per call per node (full): 316989
-Grid       : Message        : Average mflops/s per call per node (full): 437238
-Grid       : Message        : Average mflops/s per call per node (full): 446091
-Grid       : Message        : Average mflops/s per call per node (full): 306802
-Grid       : Message        :  Stencil 13.1026 GB/s per node
-Grid       : Message        :  Stencil 16.688 GB/s per node
-Grid       : Message        :  Stencil 17.4324 GB/s per node
-Grid       : Message        :  Stencil 13.3778 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668583
-Grid       : Message        : Average mflops/s per call per node       : 805193
-Grid       : Message        : Average mflops/s per call per node       : 824697
-Grid       : Message        : Average mflops/s per call per node       : 661618
-Grid       : Message        : Average mflops/s per call per node (full): 316151
-Grid       : Message        : Average mflops/s per call per node (full): 437204
-Grid       : Message        : Average mflops/s per call per node (full): 444908
-Grid       : Message        : Average mflops/s per call per node (full): 306332
-Grid       : Message        :  Stencil 13.12 GB/s per node
-Grid       : Message        :  Stencil 17.2708 GB/s per node
-Grid       : Message        :  Stencil 17.37 GB/s per node
-Grid       : Message        :  Stencil 12.1212 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667632
-Grid       : Message        : Average mflops/s per call per node       : 800607
-Grid       : Message        : Average mflops/s per call per node       : 822822
-Grid       : Message        : Average mflops/s per call per node       : 665908
-Grid       : Message        : Average mflops/s per call per node (full): 314548
-Grid       : Message        : Average mflops/s per call per node (full): 437853
-Grid       : Message        : Average mflops/s per call per node (full): 443769
-Grid       : Message        : Average mflops/s per call per node (full): 301555
-Grid       : Message        :  Stencil 12.4694 GB/s per node
-Grid       : Message        :  Stencil 16.5969 GB/s per node
-Grid       : Message        :  Stencil 17.213 GB/s per node
-Grid       : Message        :  Stencil 13.3999 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669733
-Grid       : Message        : Average mflops/s per call per node       : 806943
-Grid       : Message        : Average mflops/s per call per node       : 822933
-Grid       : Message        : Average mflops/s per call per node       : 662354
-Grid       : Message        : Average mflops/s per call per node (full): 313965
-Grid       : Message        : Average mflops/s per call per node (full): 435489
-Grid       : Message        : Average mflops/s per call per node (full): 440837
-Grid       : Message        : Average mflops/s per call per node (full): 306316
-Grid       : Message        :  Stencil 12.5374 GB/s per node
-Grid       : Message        :  Stencil 17.5318 GB/s per node
-Grid       : Message        :  Stencil 16.8417 GB/s per node
-Grid       : Message        :  Stencil 12.4993 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667866
-Grid       : Message        : Average mflops/s per call per node       : 796795
-Grid       : Message        : Average mflops/s per call per node       : 821464
-Grid       : Message        : Average mflops/s per call per node       : 661191
-Grid       : Message        : Average mflops/s per call per node (full): 313613
-Grid       : Message        : Average mflops/s per call per node (full): 435727
-Grid       : Message        : Average mflops/s per call per node (full): 427083
-Grid       : Message        : Average mflops/s per call per node (full): 303886
-Grid       : Message        :  Stencil 12.6277 GB/s per node
-Grid       : Message        :  Stencil 17.4118 GB/s per node
-Grid       : Message        :  Stencil 18.5909 GB/s per node
-Grid       : Message        :  Stencil 12.4698 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666965
-Grid       : Message        : Average mflops/s per call per node       : 798077
-Grid       : Message        : Average mflops/s per call per node       : 817594
-Grid       : Message        : Average mflops/s per call per node       : 669968
-Grid       : Message        : Average mflops/s per call per node (full): 314027
-Grid       : Message        : Average mflops/s per call per node (full): 437054
-Grid       : Message        : Average mflops/s per call per node (full): 447328
-Grid       : Message        : Average mflops/s per call per node (full): 304924
-Grid       : Message        :  Stencil 12.7725 GB/s per node
-Grid       : Message        :  Stencil 16.6166 GB/s per node
-Grid       : Message        :  Stencil 17.397 GB/s per node
-Grid       : Message        :  Stencil 12.4983 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665732
-Grid       : Message        : Average mflops/s per call per node       : 807913
-Grid       : Message        : Average mflops/s per call per node       : 821210
-Grid       : Message        : Average mflops/s per call per node       : 661150
-Grid       : Message        : Average mflops/s per call per node (full): 313604
-Grid       : Message        : Average mflops/s per call per node (full): 435521
-Grid       : Message        : Average mflops/s per call per node (full): 443043
-Grid       : Message        : Average mflops/s per call per node (full): 303752
-Grid       : Message        :  Stencil 13.6786 GB/s per node
-Grid       : Message        :  Stencil 17.4595 GB/s per node
-Grid       : Message        :  Stencil 17.5525 GB/s per node
-Grid       : Message        :  Stencil 13.3556 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661439
-Grid       : Message        : Average mflops/s per call per node       : 805238
-Grid       : Message        : Average mflops/s per call per node       : 815276
-Grid       : Message        : Average mflops/s per call per node       : 662636
-Grid       : Message        : Average mflops/s per call per node (full): 315806
-Grid       : Message        : Average mflops/s per call per node (full): 440289
-Grid       : Message        : Average mflops/s per call per node (full): 443045
-Grid       : Message        : Average mflops/s per call per node (full): 306035
-Grid       : Message        :  Stencil 12.614 GB/s per node
-Grid       : Message        :  Stencil 17.0698 GB/s per node
-Grid       : Message        :  Stencil 17.8556 GB/s per node
-Grid       : Message        :  Stencil 12.4782 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665447
-Grid       : Message        : Average mflops/s per call per node       : 803670
-Grid       : Message        : Average mflops/s per call per node       : 817131
-Grid       : Message        : Average mflops/s per call per node       : 668088
-Grid       : Message        : Average mflops/s per call per node (full): 310869
-Grid       : Message        : Average mflops/s per call per node (full): 439056
-Grid       : Message        : Average mflops/s per call per node (full): 446346
-Grid       : Message        : Average mflops/s per call per node (full): 304479
-Grid       : Message        :  Stencil 13.6138 GB/s per node
-Grid       : Message        :  Stencil 18.4778 GB/s per node
-Grid       : Message        :  Stencil 17.1873 GB/s per node
-Grid       : Message        :  Stencil 13.0416 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666452
-Grid       : Message        : Average mflops/s per call per node       : 795137
-Grid       : Message        : Average mflops/s per call per node       : 828115
-Grid       : Message        : Average mflops/s per call per node       : 662954
-Grid       : Message        : Average mflops/s per call per node (full): 314428
-Grid       : Message        : Average mflops/s per call per node (full): 437022
-Grid       : Message        : Average mflops/s per call per node (full): 442798
-Grid       : Message        : Average mflops/s per call per node (full): 305421
-Grid       : Message        :  Stencil 14.7139 GB/s per node
-Grid       : Message        :  Stencil 16.6484 GB/s per node
-Grid       : Message        :  Stencil 17.9508 GB/s per node
-Grid       : Message        :  Stencil 13.5629 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665832
-Grid       : Message        : Average mflops/s per call per node       : 803880
-Grid       : Message        : Average mflops/s per call per node       : 821362
-Grid       : Message        : Average mflops/s per call per node       : 663756
-Grid       : Message        : Average mflops/s per call per node (full): 316832
-Grid       : Message        : Average mflops/s per call per node (full): 436166
-Grid       : Message        : Average mflops/s per call per node (full): 445924
-Grid       : Message        : Average mflops/s per call per node (full): 305439
-Grid       : Message        :  Stencil 13.0084 GB/s per node
-Grid       : Message        :  Stencil 16.4882 GB/s per node
-Grid       : Message        :  Stencil 17.9387 GB/s per node
-Grid       : Message        :  Stencil 14.2227 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671512
-Grid       : Message        : Average mflops/s per call per node       : 807289
-Grid       : Message        : Average mflops/s per call per node       : 821044
-Grid       : Message        : Average mflops/s per call per node       : 663509
-Grid       : Message        : Average mflops/s per call per node (full): 315038
-Grid       : Message        : Average mflops/s per call per node (full): 434524
-Grid       : Message        : Average mflops/s per call per node (full): 447069
-Grid       : Message        : Average mflops/s per call per node (full): 307107
-Grid       : Message        :  Stencil 13.0866 GB/s per node
-Grid       : Message        :  Stencil 18.7245 GB/s per node
-Grid       : Message        :  Stencil 17.3379 GB/s per node
-Grid       : Message        :  Stencil 12.6833 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667359
-Grid       : Message        : Average mflops/s per call per node       : 800618
-Grid       : Message        : Average mflops/s per call per node       : 824683
-Grid       : Message        : Average mflops/s per call per node       : 665308
-Grid       : Message        : Average mflops/s per call per node (full): 314225
-Grid       : Message        : Average mflops/s per call per node (full): 440026
-Grid       : Message        : Average mflops/s per call per node (full): 443438
-Grid       : Message        : Average mflops/s per call per node (full): 304101
-Grid       : Message        :  Stencil 14.1214 GB/s per node
-Grid       : Message        :  Stencil 11.1327 GB/s per node
-Grid       : Message        :  Stencil 18.438 GB/s per node
-Grid       : Message        :  Stencil 12.6113 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664777
-Grid       : Message        : Average mflops/s per call per node       : 804016
-Grid       : Message        : Average mflops/s per call per node       : 821008
-Grid       : Message        : Average mflops/s per call per node       : 668356
-Grid       : Message        : Average mflops/s per call per node (full): 317217
-Grid       : Message        : Average mflops/s per call per node (full): 342710
-Grid       : Message        : Average mflops/s per call per node (full): 447021
-Grid       : Message        : Average mflops/s per call per node (full): 297876
-Grid       : Message        :  Stencil 13.408 GB/s per node
-Grid       : Message        :  Stencil 17.4022 GB/s per node
-Grid       : Message        :  Stencil 16.8897 GB/s per node
-Grid       : Message        :  Stencil 13.0928 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667130
-Grid       : Message        : Average mflops/s per call per node       : 807309
-Grid       : Message        : Average mflops/s per call per node       : 821878
-Grid       : Message        : Average mflops/s per call per node       : 666926
-Grid       : Message        : Average mflops/s per call per node (full): 315943
-Grid       : Message        : Average mflops/s per call per node (full): 440890
-Grid       : Message        : Average mflops/s per call per node (full): 438122
-Grid       : Message        : Average mflops/s per call per node (full): 306182
-Grid       : Message        :  Stencil 12.573 GB/s per node
-Grid       : Message        :  Stencil 16.7063 GB/s per node
-Grid       : Message        :  Stencil 17.1319 GB/s per node
-Grid       : Message        :  Stencil 13.0951 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668167
-Grid       : Message        : Average mflops/s per call per node       : 802663
-Grid       : Message        : Average mflops/s per call per node       : 825467
-Grid       : Message        : Average mflops/s per call per node       : 663485
-Grid       : Message        : Average mflops/s per call per node (full): 314510
-Grid       : Message        : Average mflops/s per call per node (full): 436425
-Grid       : Message        : Average mflops/s per call per node (full): 441905
-Grid       : Message        : Average mflops/s per call per node (full): 304733
-Grid       : Message        :  Stencil 12.5289 GB/s per node
-Grid       : Message        :  Stencil 17.3595 GB/s per node
-Grid       : Message        :  Stencil 18.2739 GB/s per node
-Grid       : Message        :  Stencil 12.4844 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669621
-Grid       : Message        : Average mflops/s per call per node       : 799998
-Grid       : Message        : Average mflops/s per call per node       : 816993
-Grid       : Message        : Average mflops/s per call per node       : 658972
-Grid       : Message        : Average mflops/s per call per node (full): 314123
-Grid       : Message        : Average mflops/s per call per node (full): 439790
-Grid       : Message        : Average mflops/s per call per node (full): 445963
-Grid       : Message        : Average mflops/s per call per node (full): 303737
-Grid       : Message        :  Stencil 13.4448 GB/s per node
-Grid       : Message        :  Stencil 16.9993 GB/s per node
-Grid       : Message        :  Stencil 17.8453 GB/s per node
-Grid       : Message        :  Stencil 15.0281 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663621
-Grid       : Message        : Average mflops/s per call per node       : 805340
-Grid       : Message        : Average mflops/s per call per node       : 821926
-Grid       : Message        : Average mflops/s per call per node       : 658941
-Grid       : Message        : Average mflops/s per call per node (full): 315829
-Grid       : Message        : Average mflops/s per call per node (full): 436984
-Grid       : Message        : Average mflops/s per call per node (full): 439082
-Grid       : Message        : Average mflops/s per call per node (full): 306427
-Grid       : Message        :  Stencil 13.0982 GB/s per node
-Grid       : Message        :  Stencil 16.6596 GB/s per node
-Grid       : Message        :  Stencil 18.4421 GB/s per node
-Grid       : Message        :  Stencil 13.3526 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665223
-Grid       : Message        : Average mflops/s per call per node       : 803700
-Grid       : Message        : Average mflops/s per call per node       : 822412
-Grid       : Message        : Average mflops/s per call per node       : 668318
-Grid       : Message        : Average mflops/s per call per node (full): 315138
-Grid       : Message        : Average mflops/s per call per node (full): 436240
-Grid       : Message        : Average mflops/s per call per node (full): 449103
-Grid       : Message        : Average mflops/s per call per node (full): 307087
-Grid       : Message        :  Stencil 13.1714 GB/s per node
-Grid       : Message        :  Stencil 16.8825 GB/s per node
-Grid       : Message        :  Stencil 17.5464 GB/s per node
-Grid       : Message        :  Stencil 13.121 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667784
-Grid       : Message        : Average mflops/s per call per node       : 807189
-Grid       : Message        : Average mflops/s per call per node       : 820395
-Grid       : Message        : Average mflops/s per call per node       : 667089
-Grid       : Message        : Average mflops/s per call per node (full): 314869
-Grid       : Message        : Average mflops/s per call per node (full): 438273
-Grid       : Message        : Average mflops/s per call per node (full): 445656
-Grid       : Message        : Average mflops/s per call per node (full): 307350
-Grid       : Message        :  Stencil 13.72 GB/s per node
-Grid       : Message        :  Stencil 12.6906 GB/s per node
-Grid       : Message        :  Stencil 18.2964 GB/s per node
-Grid       : Message        :  Stencil 13.3584 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666867
-Grid       : Message        : Average mflops/s per call per node       : 809633
-Grid       : Message        : Average mflops/s per call per node       : 828112
-Grid       : Message        : Average mflops/s per call per node       : 667237
-Grid       : Message        : Average mflops/s per call per node (full): 315368
-Grid       : Message        : Average mflops/s per call per node (full): 375897
-Grid       : Message        : Average mflops/s per call per node (full): 448476
-Grid       : Message        : Average mflops/s per call per node (full): 306870
-Grid       : Message        :  Stencil 12.6439 GB/s per node
-Grid       : Message        :  Stencil 16.0623 GB/s per node
-Grid       : Message        :  Stencil 18.2302 GB/s per node
-Grid       : Message        :  Stencil 14.5926 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669424
-Grid       : Message        : Average mflops/s per call per node       : 800242
-Grid       : Message        : Average mflops/s per call per node       : 822926
-Grid       : Message        : Average mflops/s per call per node       : 655791
-Grid       : Message        : Average mflops/s per call per node (full): 313407
-Grid       : Message        : Average mflops/s per call per node (full): 428941
-Grid       : Message        : Average mflops/s per call per node (full): 447806
-Grid       : Message        : Average mflops/s per call per node (full): 306464
-Grid       : Message        :  Stencil 14.4725 GB/s per node
-Grid       : Message        :  Stencil 14.2302 GB/s per node
-Grid       : Message        :  Stencil 16.734 GB/s per node
-Grid       : Message        :  Stencil 11.6196 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664432
-Grid       : Message        : Average mflops/s per call per node       : 803748
-Grid       : Message        : Average mflops/s per call per node       : 823327
-Grid       : Message        : Average mflops/s per call per node       : 669175
-Grid       : Message        : Average mflops/s per call per node (full): 316590
-Grid       : Message        : Average mflops/s per call per node (full): 402428
-Grid       : Message        : Average mflops/s per call per node (full): 437468
-Grid       : Message        : Average mflops/s per call per node (full): 298073
-Grid       : Message        :  Stencil 13.7889 GB/s per node
-Grid       : Message        :  Stencil 16.9001 GB/s per node
-Grid       : Message        :  Stencil 17.9268 GB/s per node
-Grid       : Message        :  Stencil 12.4647 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665976
-Grid       : Message        : Average mflops/s per call per node       : 803838
-Grid       : Message        : Average mflops/s per call per node       : 819200
-Grid       : Message        : Average mflops/s per call per node       : 666380
-Grid       : Message        : Average mflops/s per call per node (full): 316274
-Grid       : Message        : Average mflops/s per call per node (full): 437659
-Grid       : Message        : Average mflops/s per call per node (full): 446601
-Grid       : Message        : Average mflops/s per call per node (full): 305255
-Grid       : Message        :  Stencil 12.9962 GB/s per node
-Grid       : Message        :  Stencil 17.5956 GB/s per node
-Grid       : Message        :  Stencil 17.5064 GB/s per node
-Grid       : Message        :  Stencil 12.5874 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668479
-Grid       : Message        : Average mflops/s per call per node       : 799041
-Grid       : Message        : Average mflops/s per call per node       : 818390
-Grid       : Message        : Average mflops/s per call per node       : 664712
-Grid       : Message        : Average mflops/s per call per node (full): 315237
-Grid       : Message        : Average mflops/s per call per node (full): 440281
-Grid       : Message        : Average mflops/s per call per node (full): 443676
-Grid       : Message        : Average mflops/s per call per node (full): 304852
-Grid       : Message        :  Stencil 12.6041 GB/s per node
-Grid       : Message        :  Stencil 17.3707 GB/s per node
-Grid       : Message        :  Stencil 17.659 GB/s per node
-Grid       : Message        :  Stencil 12.3764 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669297
-Grid       : Message        : Average mflops/s per call per node       : 803887
-Grid       : Message        : Average mflops/s per call per node       : 823498
-Grid       : Message        : Average mflops/s per call per node       : 665741
-Grid       : Message        : Average mflops/s per call per node (full): 315614
-Grid       : Message        : Average mflops/s per call per node (full): 436539
-Grid       : Message        : Average mflops/s per call per node (full): 440952
-Grid       : Message        : Average mflops/s per call per node (full): 304036
-Grid       : Message        :  Stencil 13.5348 GB/s per node
-Grid       : Message        :  Stencil 17.038 GB/s per node
-Grid       : Message        :  Stencil 18.1659 GB/s per node
-Grid       : Message        :  Stencil 13.1206 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664512
-Grid       : Message        : Average mflops/s per call per node       : 805318
-Grid       : Message        : Average mflops/s per call per node       : 825232
-Grid       : Message        : Average mflops/s per call per node       : 666321
-Grid       : Message        : Average mflops/s per call per node (full): 314166
-Grid       : Message        : Average mflops/s per call per node (full): 437732
-Grid       : Message        : Average mflops/s per call per node (full): 448411
-Grid       : Message        : Average mflops/s per call per node (full): 306062
-Grid       : Message        :  Stencil 13.7061 GB/s per node
-Grid       : Message        :  Stencil 16.3278 GB/s per node
-Grid       : Message        :  Stencil 18.5878 GB/s per node
-Grid       : Message        :  Stencil 12.4536 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661594
-Grid       : Message        : Average mflops/s per call per node       : 808149
-Grid       : Message        : Average mflops/s per call per node       : 821405
-Grid       : Message        : Average mflops/s per call per node       : 665914
-Grid       : Message        : Average mflops/s per call per node (full): 315010
-Grid       : Message        : Average mflops/s per call per node (full): 432917
-Grid       : Message        : Average mflops/s per call per node (full): 448662
-Grid       : Message        : Average mflops/s per call per node (full): 302919
-Grid       : Message        :  Stencil 12.7146 GB/s per node
-Grid       : Message        :  Stencil 16.0773 GB/s per node
-Grid       : Message        :  Stencil 17.1469 GB/s per node
-Grid       : Message        :  Stencil 12.8855 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666923
-Grid       : Message        : Average mflops/s per call per node       : 805953
-Grid       : Message        : Average mflops/s per call per node       : 828904
-Grid       : Message        : Average mflops/s per call per node       : 665301
-Grid       : Message        : Average mflops/s per call per node (full): 314385
-Grid       : Message        : Average mflops/s per call per node (full): 428018
-Grid       : Message        : Average mflops/s per call per node (full): 440554
-Grid       : Message        : Average mflops/s per call per node (full): 305729
-Grid       : Message        :  Stencil 12.9741 GB/s per node
-Grid       : Message        :  Stencil 18.8804 GB/s per node
-Grid       : Message        :  Stencil 17.7919 GB/s per node
-Grid       : Message        :  Stencil 13.4678 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666032
-Grid       : Message        : Average mflops/s per call per node       : 803449
-Grid       : Message        : Average mflops/s per call per node       : 822860
-Grid       : Message        : Average mflops/s per call per node       : 662973
-Grid       : Message        : Average mflops/s per call per node (full): 313551
-Grid       : Message        : Average mflops/s per call per node (full): 441640
-Grid       : Message        : Average mflops/s per call per node (full): 446690
-Grid       : Message        : Average mflops/s per call per node (full): 306235
-Grid       : Message        :  Stencil 12.521 GB/s per node
-Grid       : Message        :  Stencil 16.3717 GB/s per node
-Grid       : Message        :  Stencil 16.7679 GB/s per node
-Grid       : Message        :  Stencil 12.8942 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665824
-Grid       : Message        : Average mflops/s per call per node       : 806997
-Grid       : Message        : Average mflops/s per call per node       : 825090
-Grid       : Message        : Average mflops/s per call per node       : 664015
-Grid       : Message        : Average mflops/s per call per node (full): 312669
-Grid       : Message        : Average mflops/s per call per node (full): 432103
-Grid       : Message        : Average mflops/s per call per node (full): 428071
-Grid       : Message        : Average mflops/s per call per node (full): 305300
-Grid       : Message        :  Stencil 13.3708 GB/s per node
-Grid       : Message        :  Stencil 16.5066 GB/s per node
-Grid       : Message        :  Stencil 16.9276 GB/s per node
-Grid       : Message        :  Stencil 12.0497 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666178
-Grid       : Message        : Average mflops/s per call per node       : 804912
-Grid       : Message        : Average mflops/s per call per node       : 822589
-Grid       : Message        : Average mflops/s per call per node       : 668246
-Grid       : Message        : Average mflops/s per call per node (full): 316301
-Grid       : Message        : Average mflops/s per call per node (full): 435095
-Grid       : Message        : Average mflops/s per call per node (full): 436080
-Grid       : Message        : Average mflops/s per call per node (full): 302118
-Grid       : Message        :  Stencil 12.7027 GB/s per node
-Grid       : Message        :  Stencil 16.5099 GB/s per node
-Grid       : Message        :  Stencil 17.1585 GB/s per node
-Grid       : Message        :  Stencil 13.2181 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668274
-Grid       : Message        : Average mflops/s per call per node       : 802058
-Grid       : Message        : Average mflops/s per call per node       : 823686
-Grid       : Message        : Average mflops/s per call per node       : 665332
-Grid       : Message        : Average mflops/s per call per node (full): 314731
-Grid       : Message        : Average mflops/s per call per node (full): 433988
-Grid       : Message        : Average mflops/s per call per node (full): 441863
-Grid       : Message        : Average mflops/s per call per node (full): 306171
-Grid       : Message        :  Stencil 13.5096 GB/s per node
-Grid       : Message        :  Stencil 16.6291 GB/s per node
-Grid       : Message        :  Stencil 17.6593 GB/s per node
-Grid       : Message        :  Stencil 12.738 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667716
-Grid       : Message        : Average mflops/s per call per node       : 804795
-Grid       : Message        : Average mflops/s per call per node       : 823643
-Grid       : Message        : Average mflops/s per call per node       : 661819
-Grid       : Message        : Average mflops/s per call per node (full): 315837
-Grid       : Message        : Average mflops/s per call per node (full): 436896
-Grid       : Message        : Average mflops/s per call per node (full): 443696
-Grid       : Message        : Average mflops/s per call per node (full): 303158
-Grid       : Message        :  Stencil 13.764 GB/s per node
-Grid       : Message        :  Stencil 16.4805 GB/s per node
-Grid       : Message        :  Stencil 16.806 GB/s per node
-Grid       : Message        :  Stencil 12.2637 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666394
-Grid       : Message        : Average mflops/s per call per node       : 808315
-Grid       : Message        : Average mflops/s per call per node       : 830523
-Grid       : Message        : Average mflops/s per call per node       : 667099
-Grid       : Message        : Average mflops/s per call per node (full): 316598
-Grid       : Message        : Average mflops/s per call per node (full): 431115
-Grid       : Message        : Average mflops/s per call per node (full): 438804
-Grid       : Message        : Average mflops/s per call per node (full): 303771
-Grid       : Message        :  Stencil 14.0315 GB/s per node
-Grid       : Message        :  Stencil 16.6177 GB/s per node
-Grid       : Message        :  Stencil 17.5158 GB/s per node
-Grid       : Message        :  Stencil 12.8021 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664946
-Grid       : Message        : Average mflops/s per call per node       : 801175
-Grid       : Message        : Average mflops/s per call per node       : 824969
-Grid       : Message        : Average mflops/s per call per node       : 664651
-Grid       : Message        : Average mflops/s per call per node (full): 316485
-Grid       : Message        : Average mflops/s per call per node (full): 434346
-Grid       : Message        : Average mflops/s per call per node (full): 445188
-Grid       : Message        : Average mflops/s per call per node (full): 304439
-Grid       : Message        :  Stencil 13.3968 GB/s per node
-Grid       : Message        :  Stencil 16.7046 GB/s per node
-Grid       : Message        :  Stencil 19.0282 GB/s per node
-Grid       : Message        :  Stencil 13.6061 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665767
-Grid       : Message        : Average mflops/s per call per node       : 806549
-Grid       : Message        : Average mflops/s per call per node       : 824649
-Grid       : Message        : Average mflops/s per call per node       : 659771
-Grid       : Message        : Average mflops/s per call per node (full): 316021
-Grid       : Message        : Average mflops/s per call per node (full): 437159
-Grid       : Message        : Average mflops/s per call per node (full): 448624
-Grid       : Message        : Average mflops/s per call per node (full): 305509
-Grid       : Message        :  Stencil 13.3675 GB/s per node
-Grid       : Message        :  Stencil 16.382 GB/s per node
-Grid       : Message        :  Stencil 18.511 GB/s per node
-Grid       : Message        :  Stencil 13.2205 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665497
-Grid       : Message        : Average mflops/s per call per node       : 803878
-Grid       : Message        : Average mflops/s per call per node       : 827619
-Grid       : Message        : Average mflops/s per call per node       : 665752
-Grid       : Message        : Average mflops/s per call per node (full): 316256
-Grid       : Message        : Average mflops/s per call per node (full): 433503
-Grid       : Message        : Average mflops/s per call per node (full): 448486
-Grid       : Message        : Average mflops/s per call per node (full): 306068
-Grid       : Message        :  Stencil 13.0375 GB/s per node
-Grid       : Message        :  Stencil 16.6764 GB/s per node
-Grid       : Message        :  Stencil 17.3085 GB/s per node
-Grid       : Message        :  Stencil 12.0552 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668003
-Grid       : Message        : Average mflops/s per call per node       : 806652
-Grid       : Message        : Average mflops/s per call per node       : 825741
-Grid       : Message        : Average mflops/s per call per node       : 665785
-Grid       : Message        : Average mflops/s per call per node (full): 313382
-Grid       : Message        : Average mflops/s per call per node (full): 436713
-Grid       : Message        : Average mflops/s per call per node (full): 445120
-Grid       : Message        : Average mflops/s per call per node (full): 302148
-Grid       : Message        :  Stencil 14.8512 GB/s per node
-Grid       : Message        :  Stencil 16.5633 GB/s per node
-Grid       : Message        :  Stencil 17.388 GB/s per node
-Grid       : Message        :  Stencil 12.0063 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663881
-Grid       : Message        : Average mflops/s per call per node       : 804976
-Grid       : Message        : Average mflops/s per call per node       : 820549
-Grid       : Message        : Average mflops/s per call per node       : 668148
-Grid       : Message        : Average mflops/s per call per node (full): 316076
-Grid       : Message        : Average mflops/s per call per node (full): 435694
-Grid       : Message        : Average mflops/s per call per node (full): 443735
-Grid       : Message        : Average mflops/s per call per node (full): 302773
-Grid       : Message        :  Stencil 13.3267 GB/s per node
-Grid       : Message        :  Stencil 17.0588 GB/s per node
-Grid       : Message        :  Stencil 16.5925 GB/s per node
-Grid       : Message        :  Stencil 13.05 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665763
-Grid       : Message        : Average mflops/s per call per node       : 803088
-Grid       : Message        : Average mflops/s per call per node       : 826037
-Grid       : Message        : Average mflops/s per call per node       : 662986
-Grid       : Message        : Average mflops/s per call per node (full): 315659
-Grid       : Message        : Average mflops/s per call per node (full): 435380
-Grid       : Message        : Average mflops/s per call per node (full): 433958
-Grid       : Message        : Average mflops/s per call per node (full): 306295
-Grid       : Message        :  Stencil 12.9447 GB/s per node
-Grid       : Message        :  Stencil 17.6472 GB/s per node
-Grid       : Message        :  Stencil 18.4279 GB/s per node
-Grid       : Message        :  Stencil 13.0117 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667222
-Grid       : Message        : Average mflops/s per call per node       : 801881
-Grid       : Message        : Average mflops/s per call per node       : 821229
-Grid       : Message        : Average mflops/s per call per node       : 660378
-Grid       : Message        : Average mflops/s per call per node (full): 313981
-Grid       : Message        : Average mflops/s per call per node (full): 438879
-Grid       : Message        : Average mflops/s per call per node (full): 447844
-Grid       : Message        : Average mflops/s per call per node (full): 303894
-Grid       : Message        :  Stencil 13.1517 GB/s per node
-Grid       : Message        :  Stencil 16.6235 GB/s per node
-Grid       : Message        :  Stencil 17.8076 GB/s per node
-Grid       : Message        :  Stencil 12.154 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669391
-Grid       : Message        : Average mflops/s per call per node       : 800786
-Grid       : Message        : Average mflops/s per call per node       : 818034
-Grid       : Message        : Average mflops/s per call per node       : 664339
-Grid       : Message        : Average mflops/s per call per node (full): 315474
-Grid       : Message        : Average mflops/s per call per node (full): 434236
-Grid       : Message        : Average mflops/s per call per node (full): 444109
-Grid       : Message        : Average mflops/s per call per node (full): 302887
-Grid       : Message        :  Stencil 13.2738 GB/s per node
-Grid       : Message        :  Stencil 16.4548 GB/s per node
-Grid       : Message        :  Stencil 18.0116 GB/s per node
-Grid       : Message        :  Stencil 14.688 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667392
-Grid       : Message        : Average mflops/s per call per node       : 802106
-Grid       : Message        : Average mflops/s per call per node       : 819500
-Grid       : Message        : Average mflops/s per call per node       : 658955
-Grid       : Message        : Average mflops/s per call per node (full): 316091
-Grid       : Message        : Average mflops/s per call per node (full): 433778
-Grid       : Message        : Average mflops/s per call per node (full): 446855
-Grid       : Message        : Average mflops/s per call per node (full): 305394
-Grid       : Message        :  Stencil 12.7989 GB/s per node
-Grid       : Message        :  Stencil 17.0567 GB/s per node
-Grid       : Message        :  Stencil 17.9343 GB/s per node
-Grid       : Message        :  Stencil 12.7083 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671386
-Grid       : Message        : Average mflops/s per call per node       : 802837
-Grid       : Message        : Average mflops/s per call per node       : 820421
-Grid       : Message        : Average mflops/s per call per node       : 663040
-Grid       : Message        : Average mflops/s per call per node (full): 315605
-Grid       : Message        : Average mflops/s per call per node (full): 437290
-Grid       : Message        : Average mflops/s per call per node (full): 447574
-Grid       : Message        : Average mflops/s per call per node (full): 304870
-Grid       : Message        :  Stencil 12.2852 GB/s per node
-Grid       : Message        :  Stencil 17.3791 GB/s per node
-Grid       : Message        :  Stencil 17.1474 GB/s per node
-Grid       : Message        :  Stencil 13.436 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670569
-Grid       : Message        : Average mflops/s per call per node       : 805222
-Grid       : Message        : Average mflops/s per call per node       : 825606
-Grid       : Message        : Average mflops/s per call per node       : 663347
-Grid       : Message        : Average mflops/s per call per node (full): 312029
-Grid       : Message        : Average mflops/s per call per node (full): 440050
-Grid       : Message        : Average mflops/s per call per node (full): 441858
-Grid       : Message        : Average mflops/s per call per node (full): 305846
-Grid       : Message        :  Stencil 12.2379 GB/s per node
-Grid       : Message        :  Stencil 14.0604 GB/s per node
-Grid       : Message        :  Stencil 18.6758 GB/s per node
-Grid       : Message        :  Stencil 13.2438 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671841
-Grid       : Message        : Average mflops/s per call per node       : 807685
-Grid       : Message        : Average mflops/s per call per node       : 820105
-Grid       : Message        : Average mflops/s per call per node       : 656744
-Grid       : Message        : Average mflops/s per call per node (full): 311636
-Grid       : Message        : Average mflops/s per call per node (full): 401570
-Grid       : Message        : Average mflops/s per call per node (full): 447034
-Grid       : Message        : Average mflops/s per call per node (full): 304702
-Grid       : Message        :  Stencil 13.5392 GB/s per node
-Grid       : Message        :  Stencil 17.0725 GB/s per node
-Grid       : Message        :  Stencil 17.1784 GB/s per node
-Grid       : Message        :  Stencil 12.2392 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664917
-Grid       : Message        : Average mflops/s per call per node       : 801687
-Grid       : Message        : Average mflops/s per call per node       : 815332
-Grid       : Message        : Average mflops/s per call per node       : 656729
-Grid       : Message        : Average mflops/s per call per node (full): 315726
-Grid       : Message        : Average mflops/s per call per node (full): 434728
-Grid       : Message        : Average mflops/s per call per node (full): 431152
-Grid       : Message        : Average mflops/s per call per node (full): 302133
-Grid       : Message        :  Stencil 13.5374 GB/s per node
-Grid       : Message        :  Stencil 16.5367 GB/s per node
-Grid       : Message        :  Stencil 17.4212 GB/s per node
-Grid       : Message        :  Stencil 12.8501 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668077
-Grid       : Message        : Average mflops/s per call per node       : 803680
-Grid       : Message        : Average mflops/s per call per node       : 826720
-Grid       : Message        : Average mflops/s per call per node       : 665432
-Grid       : Message        : Average mflops/s per call per node (full): 316753
-Grid       : Message        : Average mflops/s per call per node (full): 434204
-Grid       : Message        : Average mflops/s per call per node (full): 444228
-Grid       : Message        : Average mflops/s per call per node (full): 305541
-Grid       : Message        :  Stencil 12.8187 GB/s per node
-Grid       : Message        :  Stencil 16.7074 GB/s per node
-Grid       : Message        :  Stencil 18.4389 GB/s per node
-Grid       : Message        :  Stencil 12.8444 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670256
-Grid       : Message        : Average mflops/s per call per node       : 804534
-Grid       : Message        : Average mflops/s per call per node       : 820132
-Grid       : Message        : Average mflops/s per call per node       : 666422
-Grid       : Message        : Average mflops/s per call per node (full): 315529
-Grid       : Message        : Average mflops/s per call per node (full): 435032
-Grid       : Message        : Average mflops/s per call per node (full): 447477
-Grid       : Message        : Average mflops/s per call per node (full): 304948
-Grid       : Message        :  Stencil 13.379 GB/s per node
-Grid       : Message        :  Stencil 17.7912 GB/s per node
-Grid       : Message        :  Stencil 17.7083 GB/s per node
-Grid       : Message        :  Stencil 14.0831 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670713
-Grid       : Message        : Average mflops/s per call per node       : 801535
-Grid       : Message        : Average mflops/s per call per node       : 820180
-Grid       : Message        : Average mflops/s per call per node       : 660857
-Grid       : Message        : Average mflops/s per call per node (full): 316493
-Grid       : Message        : Average mflops/s per call per node (full): 439429
-Grid       : Message        : Average mflops/s per call per node (full): 442506
-Grid       : Message        : Average mflops/s per call per node (full): 306858
-Grid       : Message        :  Stencil 13.1634 GB/s per node
-Grid       : Message        :  Stencil 16.8688 GB/s per node
-Grid       : Message        :  Stencil 17.749 GB/s per node
-Grid       : Message        :  Stencil 13.1366 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669289
-Grid       : Message        : Average mflops/s per call per node       : 802523
-Grid       : Message        : Average mflops/s per call per node       : 820049
-Grid       : Message        : Average mflops/s per call per node       : 663095
-Grid       : Message        : Average mflops/s per call per node (full): 314878
-Grid       : Message        : Average mflops/s per call per node (full): 436156
-Grid       : Message        : Average mflops/s per call per node (full): 444995
-Grid       : Message        : Average mflops/s per call per node (full): 304995
-Grid       : Message        :  Stencil 12.3941 GB/s per node
-Grid       : Message        :  Stencil 16.754 GB/s per node
-Grid       : Message        :  Stencil 18.4191 GB/s per node
-Grid       : Message        :  Stencil 14.0827 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669824
-Grid       : Message        : Average mflops/s per call per node       : 803334
-Grid       : Message        : Average mflops/s per call per node       : 820453
-Grid       : Message        : Average mflops/s per call per node       : 661164
-Grid       : Message        : Average mflops/s per call per node (full): 313443
-Grid       : Message        : Average mflops/s per call per node (full): 437198
-Grid       : Message        : Average mflops/s per call per node (full): 448787
-Grid       : Message        : Average mflops/s per call per node (full): 306816
-Grid       : Message        :  Stencil 13.5058 GB/s per node
-Grid       : Message        :  Stencil 16.1241 GB/s per node
-Grid       : Message        :  Stencil 17.2608 GB/s per node
-Grid       : Message        :  Stencil 12.0562 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664161
-Grid       : Message        : Average mflops/s per call per node       : 805610
-Grid       : Message        : Average mflops/s per call per node       : 825873
-Grid       : Message        : Average mflops/s per call per node       : 668450
-Grid       : Message        : Average mflops/s per call per node (full): 315219
-Grid       : Message        : Average mflops/s per call per node (full): 430622
-Grid       : Message        : Average mflops/s per call per node (full): 443767
-Grid       : Message        : Average mflops/s per call per node (full): 302160
-Grid       : Message        :  Stencil 12.4599 GB/s per node
-Grid       : Message        :  Stencil 17.2007 GB/s per node
-Grid       : Message        :  Stencil 18.89 GB/s per node
-Grid       : Message        :  Stencil 12.934 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668564
-Grid       : Message        : Average mflops/s per call per node       : 807248
-Grid       : Message        : Average mflops/s per call per node       : 818786
-Grid       : Message        : Average mflops/s per call per node       : 657532
-Grid       : Message        : Average mflops/s per call per node (full): 313528
-Grid       : Message        : Average mflops/s per call per node (full): 440095
-Grid       : Message        : Average mflops/s per call per node (full): 448704
-Grid       : Message        : Average mflops/s per call per node (full): 304130
-Grid       : Message        :  Stencil 12.78 GB/s per node
-Grid       : Message        :  Stencil 16.6666 GB/s per node
-Grid       : Message        :  Stencil 18.0844 GB/s per node
-Grid       : Message        :  Stencil 13.0648 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667875
-Grid       : Message        : Average mflops/s per call per node       : 803602
-Grid       : Message        : Average mflops/s per call per node       : 823689
-Grid       : Message        : Average mflops/s per call per node       : 667685
-Grid       : Message        : Average mflops/s per call per node (full): 315202
-Grid       : Message        : Average mflops/s per call per node (full): 435594
-Grid       : Message        : Average mflops/s per call per node (full): 445154
-Grid       : Message        : Average mflops/s per call per node (full): 305952
-Grid       : Message        :  Stencil 13.243 GB/s per node
-Grid       : Message        :  Stencil 18.4039 GB/s per node
-Grid       : Message        :  Stencil 18.6476 GB/s per node
-Grid       : Message        :  Stencil 12.592 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668126
-Grid       : Message        : Average mflops/s per call per node       : 809386
-Grid       : Message        : Average mflops/s per call per node       : 821717
-Grid       : Message        : Average mflops/s per call per node       : 667399
-Grid       : Message        : Average mflops/s per call per node (full): 315372
-Grid       : Message        : Average mflops/s per call per node (full): 443816
-Grid       : Message        : Average mflops/s per call per node (full): 444890
-Grid       : Message        : Average mflops/s per call per node (full): 302195
-Grid       : Message        :  Stencil 12.46 GB/s per node
-Grid       : Message        :  Stencil 18.2155 GB/s per node
-Grid       : Message        :  Stencil 19.4584 GB/s per node
-Grid       : Message        :  Stencil 12.0904 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669848
-Grid       : Message        : Average mflops/s per call per node       : 803086
-Grid       : Message        : Average mflops/s per call per node       : 822755
-Grid       : Message        : Average mflops/s per call per node       : 664231
-Grid       : Message        : Average mflops/s per call per node (full): 313075
-Grid       : Message        : Average mflops/s per call per node (full): 440188
-Grid       : Message        : Average mflops/s per call per node (full): 451379
-Grid       : Message        : Average mflops/s per call per node (full): 301640
-Grid       : Message        :  Stencil 12.8262 GB/s per node
-Grid       : Message        :  Stencil 17.1159 GB/s per node
-Grid       : Message        :  Stencil 19.3861 GB/s per node
-Grid       : Message        :  Stencil 12.3982 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668377
-Grid       : Message        : Average mflops/s per call per node       : 798894
-Grid       : Message        : Average mflops/s per call per node       : 824796
-Grid       : Message        : Average mflops/s per call per node       : 664084
-Grid       : Message        : Average mflops/s per call per node (full): 315225
-Grid       : Message        : Average mflops/s per call per node (full): 432619
-Grid       : Message        : Average mflops/s per call per node (full): 450634
-Grid       : Message        : Average mflops/s per call per node (full): 304496
-Grid       : Message        :  Stencil 13.0438 GB/s per node
-Grid       : Message        :  Stencil 16.5429 GB/s per node
-Grid       : Message        :  Stencil 18.9669 GB/s per node
-Grid       : Message        :  Stencil 13.4612 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666991
-Grid       : Message        : Average mflops/s per call per node       : 806302
-Grid       : Message        : Average mflops/s per call per node       : 821472
-Grid       : Message        : Average mflops/s per call per node       : 659024
-Grid       : Message        : Average mflops/s per call per node (full): 315137
-Grid       : Message        : Average mflops/s per call per node (full): 434857
-Grid       : Message        : Average mflops/s per call per node (full): 449237
-Grid       : Message        : Average mflops/s per call per node (full): 303623
-Grid       : Message        :  Stencil 12.9249 GB/s per node
-Grid       : Message        :  Stencil 16.6397 GB/s per node
-Grid       : Message        :  Stencil 18.3127 GB/s per node
-Grid       : Message        :  Stencil 12.2587 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667634
-Grid       : Message        : Average mflops/s per call per node       : 805112
-Grid       : Message        : Average mflops/s per call per node       : 816695
-Grid       : Message        : Average mflops/s per call per node       : 671686
-Grid       : Message        : Average mflops/s per call per node (full): 315028
-Grid       : Message        : Average mflops/s per call per node (full): 426131
-Grid       : Message        : Average mflops/s per call per node (full): 445520
-Grid       : Message        : Average mflops/s per call per node (full): 305326
-Grid       : Message        :  Stencil 12.682 GB/s per node
-Grid       : Message        :  Stencil 16.5346 GB/s per node
-Grid       : Message        :  Stencil 17.7569 GB/s per node
-Grid       : Message        :  Stencil 12.9644 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671225
-Grid       : Message        : Average mflops/s per call per node       : 807645
-Grid       : Message        : Average mflops/s per call per node       : 820011
-Grid       : Message        : Average mflops/s per call per node       : 665084
-Grid       : Message        : Average mflops/s per call per node (full): 315592
-Grid       : Message        : Average mflops/s per call per node (full): 434958
-Grid       : Message        : Average mflops/s per call per node (full): 445033
-Grid       : Message        : Average mflops/s per call per node (full): 306041
-Grid       : Message        :  Stencil 12.6496 GB/s per node
-Grid       : Message        :  Stencil 18.4247 GB/s per node
-Grid       : Message        :  Stencil 17.1874 GB/s per node
-Grid       : Message        :  Stencil 12.4295 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 672262
-Grid       : Message        : Average mflops/s per call per node       : 804437
-Grid       : Message        : Average mflops/s per call per node       : 821138
-Grid       : Message        : Average mflops/s per call per node       : 668501
-Grid       : Message        : Average mflops/s per call per node (full): 314993
-Grid       : Message        : Average mflops/s per call per node (full): 443670
-Grid       : Message        : Average mflops/s per call per node (full): 443111
-Grid       : Message        : Average mflops/s per call per node (full): 305376
-Grid       : Message        :  Stencil 12.462 GB/s per node
-Grid       : Message        :  Stencil 10.9507 GB/s per node
-Grid       : Message        :  Stencil 17.384 GB/s per node
-Grid       : Message        :  Stencil 15.5983 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669546
-Grid       : Message        : Average mflops/s per call per node       : 814575
-Grid       : Message        : Average mflops/s per call per node       : 817613
-Grid       : Message        : Average mflops/s per call per node       : 658649
-Grid       : Message        : Average mflops/s per call per node (full): 312235
-Grid       : Message        : Average mflops/s per call per node (full): 338314
-Grid       : Message        : Average mflops/s per call per node (full): 443080
-Grid       : Message        : Average mflops/s per call per node (full): 307255
-Grid       : Message        :  Stencil 12.7405 GB/s per node
-Grid       : Message        :  Stencil 8.93222 GB/s per node
-Grid       : Message        :  Stencil 19.2037 GB/s per node
-Grid       : Message        :  Stencil 12.2945 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667278
-Grid       : Message        : Average mflops/s per call per node       : 807521
-Grid       : Message        : Average mflops/s per call per node       : 819983
-Grid       : Message        : Average mflops/s per call per node       : 665896
-Grid       : Message        : Average mflops/s per call per node (full): 313870
-Grid       : Message        : Average mflops/s per call per node (full): 290570
-Grid       : Message        : Average mflops/s per call per node (full): 448551
-Grid       : Message        : Average mflops/s per call per node (full): 303654
-Grid       : Message        :  Stencil 13.1072 GB/s per node
-Grid       : Message        :  Stencil 17.1792 GB/s per node
-Grid       : Message        :  Stencil 17.8879 GB/s per node
-Grid       : Message        :  Stencil 12.8571 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669444
-Grid       : Message        : Average mflops/s per call per node       : 809635
-Grid       : Message        : Average mflops/s per call per node       : 815615
-Grid       : Message        : Average mflops/s per call per node       : 662025
-Grid       : Message        : Average mflops/s per call per node (full): 316555
-Grid       : Message        : Average mflops/s per call per node (full): 439307
-Grid       : Message        : Average mflops/s per call per node (full): 444712
-Grid       : Message        : Average mflops/s per call per node (full): 304171
-Grid       : Message        :  Stencil 13.9813 GB/s per node
-Grid       : Message        :  Stencil 17.1032 GB/s per node
-Grid       : Message        :  Stencil 18.3618 GB/s per node
-Grid       : Message        :  Stencil 12.8147 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663352
-Grid       : Message        : Average mflops/s per call per node       : 802854
-Grid       : Message        : Average mflops/s per call per node       : 819638
-Grid       : Message        : Average mflops/s per call per node       : 666209
-Grid       : Message        : Average mflops/s per call per node (full): 315948
-Grid       : Message        : Average mflops/s per call per node (full): 435988
-Grid       : Message        : Average mflops/s per call per node (full): 445941
-Grid       : Message        : Average mflops/s per call per node (full): 305988
-Grid       : Message        :  Stencil 12.4969 GB/s per node
-Grid       : Message        :  Stencil 16.6988 GB/s per node
-Grid       : Message        :  Stencil 17.843 GB/s per node
-Grid       : Message        :  Stencil 12.6231 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670303
-Grid       : Message        : Average mflops/s per call per node       : 803093
-Grid       : Message        : Average mflops/s per call per node       : 820375
-Grid       : Message        : Average mflops/s per call per node       : 666858
-Grid       : Message        : Average mflops/s per call per node (full): 314096
-Grid       : Message        : Average mflops/s per call per node (full): 436713
-Grid       : Message        : Average mflops/s per call per node (full): 445877
-Grid       : Message        : Average mflops/s per call per node (full): 304598
-Grid       : Message        :  Stencil 13.397 GB/s per node
-Grid       : Message        :  Stencil 7.74249 GB/s per node
-Grid       : Message        :  Stencil 17.764 GB/s per node
-Grid       : Message        :  Stencil 13.4336 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667487
-Grid       : Message        : Average mflops/s per call per node       : 810620
-Grid       : Message        : Average mflops/s per call per node       : 826152
-Grid       : Message        : Average mflops/s per call per node       : 658813
-Grid       : Message        : Average mflops/s per call per node (full): 316835
-Grid       : Message        : Average mflops/s per call per node (full): 259695
-Grid       : Message        : Average mflops/s per call per node (full): 444902
-Grid       : Message        : Average mflops/s per call per node (full): 305398
-Grid       : Message        :  Stencil 12.8336 GB/s per node
-Grid       : Message        :  Stencil 16.7161 GB/s per node
-Grid       : Message        :  Stencil 18.6455 GB/s per node
-Grid       : Message        :  Stencil 12.5245 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670738
-Grid       : Message        : Average mflops/s per call per node       : 804355
-Grid       : Message        : Average mflops/s per call per node       : 819528
-Grid       : Message        : Average mflops/s per call per node       : 666568
-Grid       : Message        : Average mflops/s per call per node (full): 315705
-Grid       : Message        : Average mflops/s per call per node (full): 429704
-Grid       : Message        : Average mflops/s per call per node (full): 447242
-Grid       : Message        : Average mflops/s per call per node (full): 304879
-Grid       : Message        :  Stencil 12.9128 GB/s per node
-Grid       : Message        :  Stencil 16.2871 GB/s per node
-Grid       : Message        :  Stencil 18.1038 GB/s per node
-Grid       : Message        :  Stencil 12.6756 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670906
-Grid       : Message        : Average mflops/s per call per node       : 804414
-Grid       : Message        : Average mflops/s per call per node       : 823795
-Grid       : Message        : Average mflops/s per call per node       : 667857
-Grid       : Message        : Average mflops/s per call per node (full): 315623
-Grid       : Message        : Average mflops/s per call per node (full): 432420
-Grid       : Message        : Average mflops/s per call per node (full): 445608
-Grid       : Message        : Average mflops/s per call per node (full): 305259
-Grid       : Message        :  Stencil 13.1449 GB/s per node
-Grid       : Message        :  Stencil 17.4201 GB/s per node
-Grid       : Message        :  Stencil 18.0655 GB/s per node
-Grid       : Message        :  Stencil 13.6378 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667533
-Grid       : Message        : Average mflops/s per call per node       : 806456
-Grid       : Message        : Average mflops/s per call per node       : 823065
-Grid       : Message        : Average mflops/s per call per node       : 662906
-Grid       : Message        : Average mflops/s per call per node (full): 315087
-Grid       : Message        : Average mflops/s per call per node (full): 439271
-Grid       : Message        : Average mflops/s per call per node (full): 446828
-Grid       : Message        : Average mflops/s per call per node (full): 304904
-Grid       : Message        :  Stencil 12.5603 GB/s per node
-Grid       : Message        :  Stencil 16.4946 GB/s per node
-Grid       : Message        :  Stencil 17.2015 GB/s per node
-Grid       : Message        :  Stencil 12.6719 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667440
-Grid       : Message        : Average mflops/s per call per node       : 804005
-Grid       : Message        : Average mflops/s per call per node       : 819082
-Grid       : Message        : Average mflops/s per call per node       : 659045
-Grid       : Message        : Average mflops/s per call per node (full): 313637
-Grid       : Message        : Average mflops/s per call per node (full): 434386
-Grid       : Message        : Average mflops/s per call per node (full): 441996
-Grid       : Message        : Average mflops/s per call per node (full): 302145
-Grid       : Message        :  Stencil 12.5421 GB/s per node
-Grid       : Message        :  Stencil 16.6179 GB/s per node
-Grid       : Message        :  Stencil 17.576 GB/s per node
-Grid       : Message        :  Stencil 12.7209 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 673301
-Grid       : Message        : Average mflops/s per call per node       : 802954
-Grid       : Message        : Average mflops/s per call per node       : 822063
-Grid       : Message        : Average mflops/s per call per node       : 665198
-Grid       : Message        : Average mflops/s per call per node (full): 314161
-Grid       : Message        : Average mflops/s per call per node (full): 433576
-Grid       : Message        : Average mflops/s per call per node (full): 445628
-Grid       : Message        : Average mflops/s per call per node (full): 305214
-Grid       : Message        :  Stencil 13.2862 GB/s per node
-Grid       : Message        :  Stencil 10.5698 GB/s per node
-Grid       : Message        :  Stencil 17.4027 GB/s per node
-Grid       : Message        :  Stencil 12.7517 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669822
-Grid       : Message        : Average mflops/s per call per node       : 804922
-Grid       : Message        : Average mflops/s per call per node       : 822865
-Grid       : Message        : Average mflops/s per call per node       : 663862
-Grid       : Message        : Average mflops/s per call per node (full): 316809
-Grid       : Message        : Average mflops/s per call per node (full): 329286
-Grid       : Message        : Average mflops/s per call per node (full): 442725
-Grid       : Message        : Average mflops/s per call per node (full): 303149
-Grid       : Message        :  Stencil 12.852 GB/s per node
-Grid       : Message        :  Stencil 16.5856 GB/s per node
-Grid       : Message        :  Stencil 16.8703 GB/s per node
-Grid       : Message        :  Stencil 12.3381 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 672030
-Grid       : Message        : Average mflops/s per call per node       : 806056
-Grid       : Message        : Average mflops/s per call per node       : 824011
-Grid       : Message        : Average mflops/s per call per node       : 663725
-Grid       : Message        : Average mflops/s per call per node (full): 313016
-Grid       : Message        : Average mflops/s per call per node (full): 433934
-Grid       : Message        : Average mflops/s per call per node (full): 437887
-Grid       : Message        : Average mflops/s per call per node (full): 304202
-Grid       : Message        :  Stencil 13.1136 GB/s per node
-Grid       : Message        :  Stencil 17.4085 GB/s per node
-Grid       : Message        :  Stencil 17.6594 GB/s per node
-Grid       : Message        :  Stencil 12.5101 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667205
-Grid       : Message        : Average mflops/s per call per node       : 801624
-Grid       : Message        : Average mflops/s per call per node       : 821358
-Grid       : Message        : Average mflops/s per call per node       : 666993
-Grid       : Message        : Average mflops/s per call per node (full): 315372
-Grid       : Message        : Average mflops/s per call per node (full): 438220
-Grid       : Message        : Average mflops/s per call per node (full): 446266
-Grid       : Message        : Average mflops/s per call per node (full): 305044
-Grid       : Message        :  Stencil 12.615 GB/s per node
-Grid       : Message        :  Stencil 16.9959 GB/s per node
-Grid       : Message        :  Stencil 17.4396 GB/s per node
-Grid       : Message        :  Stencil 12.6798 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668788
-Grid       : Message        : Average mflops/s per call per node       : 800461
-Grid       : Message        : Average mflops/s per call per node       : 826309
-Grid       : Message        : Average mflops/s per call per node       : 664766
-Grid       : Message        : Average mflops/s per call per node (full): 314281
-Grid       : Message        : Average mflops/s per call per node (full): 435637
-Grid       : Message        : Average mflops/s per call per node (full): 445094
-Grid       : Message        : Average mflops/s per call per node (full): 303595
-Grid       : Message        :  Stencil 12.6878 GB/s per node
-Grid       : Message        :  Stencil 16.6081 GB/s per node
-Grid       : Message        :  Stencil 16.9322 GB/s per node
-Grid       : Message        :  Stencil 13.615 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667055
-Grid       : Message        : Average mflops/s per call per node       : 802579
-Grid       : Message        : Average mflops/s per call per node       : 829844
-Grid       : Message        : Average mflops/s per call per node       : 662433
-Grid       : Message        : Average mflops/s per call per node (full): 314107
-Grid       : Message        : Average mflops/s per call per node (full): 434352
-Grid       : Message        : Average mflops/s per call per node (full): 429138
-Grid       : Message        : Average mflops/s per call per node (full): 305501
-Grid       : Message        :  Stencil 13.0521 GB/s per node
-Grid       : Message        :  Stencil 14.0309 GB/s per node
-Grid       : Message        :  Stencil 16.8677 GB/s per node
-Grid       : Message        :  Stencil 12.5594 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668824
-Grid       : Message        : Average mflops/s per call per node       : 810876
-Grid       : Message        : Average mflops/s per call per node       : 821318
-Grid       : Message        : Average mflops/s per call per node       : 664964
-Grid       : Message        : Average mflops/s per call per node (full): 316377
-Grid       : Message        : Average mflops/s per call per node (full): 399792
-Grid       : Message        : Average mflops/s per call per node (full): 438584
-Grid       : Message        : Average mflops/s per call per node (full): 302806
-Grid       : Message        :  Stencil 15.8763 GB/s per node
-Grid       : Message        :  Stencil 17.0735 GB/s per node
-Grid       : Message        :  Stencil 18.3697 GB/s per node
-Grid       : Message        :  Stencil 12.4011 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 658812
-Grid       : Message        : Average mflops/s per call per node       : 807395
-Grid       : Message        : Average mflops/s per call per node       : 825888
-Grid       : Message        : Average mflops/s per call per node       : 663160
-Grid       : Message        : Average mflops/s per call per node (full): 316458
-Grid       : Message        : Average mflops/s per call per node (full): 439902
-Grid       : Message        : Average mflops/s per call per node (full): 450768
-Grid       : Message        : Average mflops/s per call per node (full): 304482
-Grid       : Message        :  Stencil 13.3067 GB/s per node
-Grid       : Message        :  Stencil 10.7314 GB/s per node
-Grid       : Message        :  Stencil 17.4575 GB/s per node
-Grid       : Message        :  Stencil 12.6334 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663778
-Grid       : Message        : Average mflops/s per call per node       : 804718
-Grid       : Message        : Average mflops/s per call per node       : 827613
-Grid       : Message        : Average mflops/s per call per node       : 668973
-Grid       : Message        : Average mflops/s per call per node (full): 315624
-Grid       : Message        : Average mflops/s per call per node (full): 332576
-Grid       : Message        : Average mflops/s per call per node (full): 446449
-Grid       : Message        : Average mflops/s per call per node (full): 306130
-Grid       : Message        :  Stencil 12.8944 GB/s per node
-Grid       : Message        :  Stencil 16.6851 GB/s per node
-Grid       : Message        :  Stencil 17.6033 GB/s per node
-Grid       : Message        :  Stencil 12.1455 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667064
-Grid       : Message        : Average mflops/s per call per node       : 804489
-Grid       : Message        : Average mflops/s per call per node       : 814207
-Grid       : Message        : Average mflops/s per call per node       : 662144
-Grid       : Message        : Average mflops/s per call per node (full): 312704
-Grid       : Message        : Average mflops/s per call per node (full): 436769
-Grid       : Message        : Average mflops/s per call per node (full): 444143
-Grid       : Message        : Average mflops/s per call per node (full): 300723
-Grid       : Message        :  Stencil 13.4574 GB/s per node
-Grid       : Message        :  Stencil 16.9887 GB/s per node
-Grid       : Message        :  Stencil 17.1818 GB/s per node
-Grid       : Message        :  Stencil 12.4573 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666017
-Grid       : Message        : Average mflops/s per call per node       : 801217
-Grid       : Message        : Average mflops/s per call per node       : 828869
-Grid       : Message        : Average mflops/s per call per node       : 669250
-Grid       : Message        : Average mflops/s per call per node (full): 316097
-Grid       : Message        : Average mflops/s per call per node (full): 438066
-Grid       : Message        : Average mflops/s per call per node (full): 443785
-Grid       : Message        : Average mflops/s per call per node (full): 305111
-Grid       : Message        :  Stencil 14.3726 GB/s per node
-Grid       : Message        :  Stencil 16.5001 GB/s per node
-Grid       : Message        :  Stencil 17.4919 GB/s per node
-Grid       : Message        :  Stencil 11.8296 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661241
-Grid       : Message        : Average mflops/s per call per node       : 805278
-Grid       : Message        : Average mflops/s per call per node       : 819257
-Grid       : Message        : Average mflops/s per call per node       : 662070
-Grid       : Message        : Average mflops/s per call per node (full): 316571
-Grid       : Message        : Average mflops/s per call per node (full): 434901
-Grid       : Message        : Average mflops/s per call per node (full): 443915
-Grid       : Message        : Average mflops/s per call per node (full): 299770
-Grid       : Message        :  Stencil 13.9432 GB/s per node
-Grid       : Message        :  Stencil 17.1709 GB/s per node
-Grid       : Message        :  Stencil 18.5397 GB/s per node
-Grid       : Message        :  Stencil 12.0828 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662025
-Grid       : Message        : Average mflops/s per call per node       : 798735
-Grid       : Message        : Average mflops/s per call per node       : 827157
-Grid       : Message        : Average mflops/s per call per node       : 669224
-Grid       : Message        : Average mflops/s per call per node (full): 315170
-Grid       : Message        : Average mflops/s per call per node (full): 436317
-Grid       : Message        : Average mflops/s per call per node (full): 450815
-Grid       : Message        : Average mflops/s per call per node (full): 303414
-Grid       : Message        :  Stencil 14.8256 GB/s per node
-Grid       : Message        :  Stencil 16.8343 GB/s per node
-Grid       : Message        :  Stencil 17.9506 GB/s per node
-Grid       : Message        :  Stencil 12.1935 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 658702
-Grid       : Message        : Average mflops/s per call per node       : 805008
-Grid       : Message        : Average mflops/s per call per node       : 821807
-Grid       : Message        : Average mflops/s per call per node       : 669907
-Grid       : Message        : Average mflops/s per call per node (full): 315640
-Grid       : Message        : Average mflops/s per call per node (full): 436405
-Grid       : Message        : Average mflops/s per call per node (full): 447011
-Grid       : Message        : Average mflops/s per call per node (full): 303721
-Grid       : Message        :  Stencil 12.8896 GB/s per node
-Grid       : Message        :  Stencil 17.0187 GB/s per node
-Grid       : Message        :  Stencil 18.0889 GB/s per node
-Grid       : Message        :  Stencil 12.6104 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668724
-Grid       : Message        : Average mflops/s per call per node       : 804606
-Grid       : Message        : Average mflops/s per call per node       : 816875
-Grid       : Message        : Average mflops/s per call per node       : 669186
-Grid       : Message        : Average mflops/s per call per node (full): 315034
-Grid       : Message        : Average mflops/s per call per node (full): 440288
-Grid       : Message        : Average mflops/s per call per node (full): 445646
-Grid       : Message        : Average mflops/s per call per node (full): 304050
-Grid       : Message        :  Stencil 13.7508 GB/s per node
-Grid       : Message        :  Stencil 16.7089 GB/s per node
-Grid       : Message        :  Stencil 18.2857 GB/s per node
-Grid       : Message        :  Stencil 13.2618 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663381
-Grid       : Message        : Average mflops/s per call per node       : 811779
-Grid       : Message        : Average mflops/s per call per node       : 819120
-Grid       : Message        : Average mflops/s per call per node       : 663294
-Grid       : Message        : Average mflops/s per call per node (full): 316059
-Grid       : Message        : Average mflops/s per call per node (full): 437388
-Grid       : Message        : Average mflops/s per call per node (full): 447847
-Grid       : Message        : Average mflops/s per call per node (full): 305901
-Grid       : Message        :  Stencil 12.8254 GB/s per node
-Grid       : Message        :  Stencil 17.0104 GB/s per node
-Grid       : Message        :  Stencil 19.176 GB/s per node
-Grid       : Message        :  Stencil 12.5891 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663632
-Grid       : Message        : Average mflops/s per call per node       : 805038
-Grid       : Message        : Average mflops/s per call per node       : 820317
-Grid       : Message        : Average mflops/s per call per node       : 661826
-Grid       : Message        : Average mflops/s per call per node (full): 313980
-Grid       : Message        : Average mflops/s per call per node (full): 438791
-Grid       : Message        : Average mflops/s per call per node (full): 448800
-Grid       : Message        : Average mflops/s per call per node (full): 304529
-Grid       : Message        :  Stencil 15.0151 GB/s per node
-Grid       : Message        :  Stencil 16.4325 GB/s per node
-Grid       : Message        :  Stencil 17.3295 GB/s per node
-Grid       : Message        :  Stencil 11.941 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659911
-Grid       : Message        : Average mflops/s per call per node       : 803247
-Grid       : Message        : Average mflops/s per call per node       : 820411
-Grid       : Message        : Average mflops/s per call per node       : 671308
-Grid       : Message        : Average mflops/s per call per node (full): 315850
-Grid       : Message        : Average mflops/s per call per node (full): 433571
-Grid       : Message        : Average mflops/s per call per node (full): 441916
-Grid       : Message        : Average mflops/s per call per node (full): 301998
-Grid       : Message        :  Stencil 14.2255 GB/s per node
-Grid       : Message        :  Stencil 17.661 GB/s per node
-Grid       : Message        :  Stencil 18.0753 GB/s per node
-Grid       : Message        :  Stencil 13.4795 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662586
-Grid       : Message        : Average mflops/s per call per node       : 805452
-Grid       : Message        : Average mflops/s per call per node       : 827893
-Grid       : Message        : Average mflops/s per call per node       : 664931
-Grid       : Message        : Average mflops/s per call per node (full): 315941
-Grid       : Message        : Average mflops/s per call per node (full): 440056
-Grid       : Message        : Average mflops/s per call per node (full): 438518
-Grid       : Message        : Average mflops/s per call per node (full): 305840
-Grid       : Message        :  Stencil 12.5694 GB/s per node
-Grid       : Message        :  Stencil 16.384 GB/s per node
-Grid       : Message        :  Stencil 17.0648 GB/s per node
-Grid       : Message        :  Stencil 12.5373 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667746
-Grid       : Message        : Average mflops/s per call per node       : 805444
-Grid       : Message        : Average mflops/s per call per node       : 820356
-Grid       : Message        : Average mflops/s per call per node       : 664620
-Grid       : Message        : Average mflops/s per call per node (full): 313742
-Grid       : Message        : Average mflops/s per call per node (full): 433099
-Grid       : Message        : Average mflops/s per call per node (full): 440895
-Grid       : Message        : Average mflops/s per call per node (full): 304415
-Grid       : Message        :  Stencil 9.82643 GB/s per node
-Grid       : Message        :  Stencil 17.1429 GB/s per node
-Grid       : Message        :  Stencil 18.2343 GB/s per node
-Grid       : Message        :  Stencil 12.7365 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667814
-Grid       : Message        : Average mflops/s per call per node       : 806928
-Grid       : Message        : Average mflops/s per call per node       : 821145
-Grid       : Message        : Average mflops/s per call per node       : 666313
-Grid       : Message        : Average mflops/s per call per node (full): 278891
-Grid       : Message        : Average mflops/s per call per node (full): 438675
-Grid       : Message        : Average mflops/s per call per node (full): 446588
-Grid       : Message        : Average mflops/s per call per node (full): 305229
-Grid       : Message        :  Stencil 12.7522 GB/s per node
-Grid       : Message        :  Stencil 17.7047 GB/s per node
-Grid       : Message        :  Stencil 18.7484 GB/s per node
-Grid       : Message        :  Stencil 13.3864 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667383
-Grid       : Message        : Average mflops/s per call per node       : 804993
-Grid       : Message        : Average mflops/s per call per node       : 822793
-Grid       : Message        : Average mflops/s per call per node       : 667884
-Grid       : Message        : Average mflops/s per call per node (full): 314158
-Grid       : Message        : Average mflops/s per call per node (full): 439202
-Grid       : Message        : Average mflops/s per call per node (full): 449309
-Grid       : Message        : Average mflops/s per call per node (full): 306541
-Grid       : Message        :  Stencil 12.586 GB/s per node
-Grid       : Message        :  Stencil 16.2104 GB/s per node
-Grid       : Message        :  Stencil 17.4591 GB/s per node
-Grid       : Message        :  Stencil 12.9073 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669139
-Grid       : Message        : Average mflops/s per call per node       : 801610
-Grid       : Message        : Average mflops/s per call per node       : 824310
-Grid       : Message        : Average mflops/s per call per node       : 666021
-Grid       : Message        : Average mflops/s per call per node (full): 314005
-Grid       : Message        : Average mflops/s per call per node (full): 422426
-Grid       : Message        : Average mflops/s per call per node (full): 443867
-Grid       : Message        : Average mflops/s per call per node (full): 304793
-Grid       : Message        :  Stencil 12.5338 GB/s per node
-Grid       : Message        :  Stencil 17.3649 GB/s per node
-Grid       : Message        :  Stencil 17.3414 GB/s per node
-Grid       : Message        :  Stencil 12.9893 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668380
-Grid       : Message        : Average mflops/s per call per node       : 801173
-Grid       : Message        : Average mflops/s per call per node       : 821810
-Grid       : Message        : Average mflops/s per call per node       : 668572
-Grid       : Message        : Average mflops/s per call per node (full): 313675
-Grid       : Message        : Average mflops/s per call per node (full): 437900
-Grid       : Message        : Average mflops/s per call per node (full): 442784
-Grid       : Message        : Average mflops/s per call per node (full): 305719
-Grid       : Message        :  Stencil 12.9458 GB/s per node
-Grid       : Message        :  Stencil 16.6798 GB/s per node
-Grid       : Message        :  Stencil 17.6924 GB/s per node
-Grid       : Message        :  Stencil 13.7427 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669642
-Grid       : Message        : Average mflops/s per call per node       : 804386
-Grid       : Message        : Average mflops/s per call per node       : 825651
-Grid       : Message        : Average mflops/s per call per node       : 665360
-Grid       : Message        : Average mflops/s per call per node (full): 315893
-Grid       : Message        : Average mflops/s per call per node (full): 437608
-Grid       : Message        : Average mflops/s per call per node (full): 446511
-Grid       : Message        : Average mflops/s per call per node (full): 307120
-Grid       : Message        :  Stencil 12.6993 GB/s per node
-Grid       : Message        :  Stencil 17.4135 GB/s per node
-Grid       : Message        :  Stencil 17.7356 GB/s per node
-Grid       : Message        :  Stencil 13.9442 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669407
-Grid       : Message        : Average mflops/s per call per node       : 802858
-Grid       : Message        : Average mflops/s per call per node       : 818423
-Grid       : Message        : Average mflops/s per call per node       : 661573
-Grid       : Message        : Average mflops/s per call per node (full): 314725
-Grid       : Message        : Average mflops/s per call per node (full): 438259
-Grid       : Message        : Average mflops/s per call per node (full): 445175
-Grid       : Message        : Average mflops/s per call per node (full): 306416
-Grid       : Message        :  Stencil 13.4012 GB/s per node
-Grid       : Message        :  Stencil 16.8096 GB/s per node
-Grid       : Message        :  Stencil 18.5014 GB/s per node
-Grid       : Message        :  Stencil 12.5647 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666205
-Grid       : Message        : Average mflops/s per call per node       : 804380
-Grid       : Message        : Average mflops/s per call per node       : 817652
-Grid       : Message        : Average mflops/s per call per node       : 667548
-Grid       : Message        : Average mflops/s per call per node (full): 316299
-Grid       : Message        : Average mflops/s per call per node (full): 436949
-Grid       : Message        : Average mflops/s per call per node (full): 446093
-Grid       : Message        : Average mflops/s per call per node (full): 304572
-Grid       : Message        :  Stencil 12.857 GB/s per node
-Grid       : Message        :  Stencil 18.2254 GB/s per node
-Grid       : Message        :  Stencil 16.6485 GB/s per node
-Grid       : Message        :  Stencil 12.3756 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668871
-Grid       : Message        : Average mflops/s per call per node       : 802096
-Grid       : Message        : Average mflops/s per call per node       : 824877
-Grid       : Message        : Average mflops/s per call per node       : 660461
-Grid       : Message        : Average mflops/s per call per node (full): 315638
-Grid       : Message        : Average mflops/s per call per node (full): 440771
-Grid       : Message        : Average mflops/s per call per node (full): 424600
-Grid       : Message        : Average mflops/s per call per node (full): 299492
-Grid       : Message        :  Stencil 14.3042 GB/s per node
-Grid       : Message        :  Stencil 18.4101 GB/s per node
-Grid       : Message        :  Stencil 17.7716 GB/s per node
-Grid       : Message        :  Stencil 13.5163 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665246
-Grid       : Message        : Average mflops/s per call per node       : 802891
-Grid       : Message        : Average mflops/s per call per node       : 821382
-Grid       : Message        : Average mflops/s per call per node       : 661226
-Grid       : Message        : Average mflops/s per call per node (full): 316755
-Grid       : Message        : Average mflops/s per call per node (full): 442426
-Grid       : Message        : Average mflops/s per call per node (full): 446803
-Grid       : Message        : Average mflops/s per call per node (full): 305937
-Grid       : Message        :  Stencil 13.6618 GB/s per node
-Grid       : Message        :  Stencil 16.3922 GB/s per node
-Grid       : Message        :  Stencil 17.3079 GB/s per node
-Grid       : Message        :  Stencil 13.1425 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668153
-Grid       : Message        : Average mflops/s per call per node       : 806500
-Grid       : Message        : Average mflops/s per call per node       : 823490
-Grid       : Message        : Average mflops/s per call per node       : 663790
-Grid       : Message        : Average mflops/s per call per node (full): 316957
-Grid       : Message        : Average mflops/s per call per node (full): 434594
-Grid       : Message        : Average mflops/s per call per node (full): 442187
-Grid       : Message        : Average mflops/s per call per node (full): 304909
-Grid       : Message        :  Stencil 13.9324 GB/s per node
-Grid       : Message        :  Stencil 17.2523 GB/s per node
-Grid       : Message        :  Stencil 17.4606 GB/s per node
-Grid       : Message        :  Stencil 12.2759 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664155
-Grid       : Message        : Average mflops/s per call per node       : 802125
-Grid       : Message        : Average mflops/s per call per node       : 819229
-Grid       : Message        : Average mflops/s per call per node       : 664597
-Grid       : Message        : Average mflops/s per call per node (full): 316763
-Grid       : Message        : Average mflops/s per call per node (full): 438223
-Grid       : Message        : Average mflops/s per call per node (full): 444372
-Grid       : Message        : Average mflops/s per call per node (full): 303570
-Grid       : Message        :  Stencil 13.0636 GB/s per node
-Grid       : Message        :  Stencil 15.9842 GB/s per node
-Grid       : Message        :  Stencil 17.529 GB/s per node
-Grid       : Message        :  Stencil 14.4215 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664767
-Grid       : Message        : Average mflops/s per call per node       : 807391
-Grid       : Message        : Average mflops/s per call per node       : 821570
-Grid       : Message        : Average mflops/s per call per node       : 661350
-Grid       : Message        : Average mflops/s per call per node (full): 315462
-Grid       : Message        : Average mflops/s per call per node (full): 424480
-Grid       : Message        : Average mflops/s per call per node (full): 444477
-Grid       : Message        : Average mflops/s per call per node (full): 307316
-Grid       : Message        :  Stencil 12.6748 GB/s per node
-Grid       : Message        :  Stencil 16.9539 GB/s per node
-Grid       : Message        :  Stencil 17.3641 GB/s per node
-Grid       : Message        :  Stencil 13.0659 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667960
-Grid       : Message        : Average mflops/s per call per node       : 807442
-Grid       : Message        : Average mflops/s per call per node       : 818275
-Grid       : Message        : Average mflops/s per call per node       : 661018
-Grid       : Message        : Average mflops/s per call per node (full): 314619
-Grid       : Message        : Average mflops/s per call per node (full): 435237
-Grid       : Message        : Average mflops/s per call per node (full): 440508
-Grid       : Message        : Average mflops/s per call per node (full): 305566
-Grid       : Message        :  Stencil 14.8347 GB/s per node
-Grid       : Message        :  Stencil 17.385 GB/s per node
-Grid       : Message        :  Stencil 17.593 GB/s per node
-Grid       : Message        :  Stencil 13.8755 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662295
-Grid       : Message        : Average mflops/s per call per node       : 805309
-Grid       : Message        : Average mflops/s per call per node       : 814833
-Grid       : Message        : Average mflops/s per call per node       : 662705
-Grid       : Message        : Average mflops/s per call per node (full): 315703
-Grid       : Message        : Average mflops/s per call per node (full): 438796
-Grid       : Message        : Average mflops/s per call per node (full): 444603
-Grid       : Message        : Average mflops/s per call per node (full): 304304
-Grid       : Message        :  Stencil 12.7963 GB/s per node
-Grid       : Message        :  Stencil 16.472 GB/s per node
-Grid       : Message        :  Stencil 17.7234 GB/s per node
-Grid       : Message        :  Stencil 14.8846 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666891
-Grid       : Message        : Average mflops/s per call per node       : 805228
-Grid       : Message        : Average mflops/s per call per node       : 826469
-Grid       : Message        : Average mflops/s per call per node       : 666819
-Grid       : Message        : Average mflops/s per call per node (full): 314699
-Grid       : Message        : Average mflops/s per call per node (full): 434810
-Grid       : Message        : Average mflops/s per call per node (full): 447368
-Grid       : Message        : Average mflops/s per call per node (full): 308042
-Grid       : Message        :  Stencil 14.6722 GB/s per node
-Grid       : Message        :  Stencil 17.809 GB/s per node
-Grid       : Message        :  Stencil 17.9151 GB/s per node
-Grid       : Message        :  Stencil 12.8243 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661154
-Grid       : Message        : Average mflops/s per call per node       : 798660
-Grid       : Message        : Average mflops/s per call per node       : 822729
-Grid       : Message        : Average mflops/s per call per node       : 672214
-Grid       : Message        : Average mflops/s per call per node (full): 316280
-Grid       : Message        : Average mflops/s per call per node (full): 438026
-Grid       : Message        : Average mflops/s per call per node (full): 445698
-Grid       : Message        : Average mflops/s per call per node (full): 307144
-Grid       : Message        :  Stencil 14.2181 GB/s per node
-Grid       : Message        :  Stencil 16.8218 GB/s per node
-Grid       : Message        :  Stencil 16.8766 GB/s per node
-Grid       : Message        :  Stencil 12.6323 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663563
-Grid       : Message        : Average mflops/s per call per node       : 806464
-Grid       : Message        : Average mflops/s per call per node       : 824904
-Grid       : Message        : Average mflops/s per call per node       : 664000
-Grid       : Message        : Average mflops/s per call per node (full): 316627
-Grid       : Message        : Average mflops/s per call per node (full): 438328
-Grid       : Message        : Average mflops/s per call per node (full): 439408
-Grid       : Message        : Average mflops/s per call per node (full): 303348
-Grid       : Message        :  Stencil 13.1026 GB/s per node
-Grid       : Message        :  Stencil 17.6235 GB/s per node
-Grid       : Message        :  Stencil 18.184 GB/s per node
-Grid       : Message        :  Stencil 12.6188 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668665
-Grid       : Message        : Average mflops/s per call per node       : 799103
-Grid       : Message        : Average mflops/s per call per node       : 823048
-Grid       : Message        : Average mflops/s per call per node       : 665913
-Grid       : Message        : Average mflops/s per call per node (full): 315766
-Grid       : Message        : Average mflops/s per call per node (full): 439133
-Grid       : Message        : Average mflops/s per call per node (full): 447554
-Grid       : Message        : Average mflops/s per call per node (full): 304968
-Grid       : Message        :  Stencil 13.8485 GB/s per node
-Grid       : Message        :  Stencil 18.4952 GB/s per node
-Grid       : Message        :  Stencil 17.5541 GB/s per node
-Grid       : Message        :  Stencil 12.1178 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663295
-Grid       : Message        : Average mflops/s per call per node       : 805011
-Grid       : Message        : Average mflops/s per call per node       : 825778
-Grid       : Message        : Average mflops/s per call per node       : 671942
-Grid       : Message        : Average mflops/s per call per node (full): 315425
-Grid       : Message        : Average mflops/s per call per node (full): 440979
-Grid       : Message        : Average mflops/s per call per node (full): 446281
-Grid       : Message        : Average mflops/s per call per node (full): 303669
-Grid       : Message        :  Stencil 15.4956 GB/s per node
-Grid       : Message        :  Stencil 16.4981 GB/s per node
-Grid       : Message        :  Stencil 18.3299 GB/s per node
-Grid       : Message        :  Stencil 12.8365 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663849
-Grid       : Message        : Average mflops/s per call per node       : 806486
-Grid       : Message        : Average mflops/s per call per node       : 820322
-Grid       : Message        : Average mflops/s per call per node       : 666420
-Grid       : Message        : Average mflops/s per call per node (full): 317964
-Grid       : Message        : Average mflops/s per call per node (full): 434864
-Grid       : Message        : Average mflops/s per call per node (full): 448068
-Grid       : Message        : Average mflops/s per call per node (full): 305884
-Grid       : Message        :  Stencil 12.9709 GB/s per node
-Grid       : Message        :  Stencil 17.3497 GB/s per node
-Grid       : Message        :  Stencil 17.8545 GB/s per node
-Grid       : Message        :  Stencil 14.3185 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666199
-Grid       : Message        : Average mflops/s per call per node       : 805234
-Grid       : Message        : Average mflops/s per call per node       : 821724
-Grid       : Message        : Average mflops/s per call per node       : 661092
-Grid       : Message        : Average mflops/s per call per node (full): 314666
-Grid       : Message        : Average mflops/s per call per node (full): 439669
-Grid       : Message        : Average mflops/s per call per node (full): 443151
-Grid       : Message        : Average mflops/s per call per node (full): 305498
-Grid       : Message        :  Stencil 14.0359 GB/s per node
-Grid       : Message        :  Stencil 16.959 GB/s per node
-Grid       : Message        :  Stencil 17.5696 GB/s per node
-Grid       : Message        :  Stencil 14.5427 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663099
-Grid       : Message        : Average mflops/s per call per node       : 801907
-Grid       : Message        : Average mflops/s per call per node       : 825632
-Grid       : Message        : Average mflops/s per call per node       : 656212
-Grid       : Message        : Average mflops/s per call per node (full): 316620
-Grid       : Message        : Average mflops/s per call per node (full): 437226
-Grid       : Message        : Average mflops/s per call per node (full): 446519
-Grid       : Message        : Average mflops/s per call per node (full): 307200
-Grid       : Message        :  Stencil 13.1194 GB/s per node
-Grid       : Message        :  Stencil 16.5614 GB/s per node
-Grid       : Message        :  Stencil 17.534 GB/s per node
-Grid       : Message        :  Stencil 12.7429 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667941
-Grid       : Message        : Average mflops/s per call per node       : 798824
-Grid       : Message        : Average mflops/s per call per node       : 827515
-Grid       : Message        : Average mflops/s per call per node       : 667869
-Grid       : Message        : Average mflops/s per call per node (full): 316068
-Grid       : Message        : Average mflops/s per call per node (full): 433793
-Grid       : Message        : Average mflops/s per call per node (full): 446594
-Grid       : Message        : Average mflops/s per call per node (full): 305905
-Grid       : Message        :  Stencil 14.0792 GB/s per node
-Grid       : Message        :  Stencil 16.1996 GB/s per node
-Grid       : Message        :  Stencil 17.2694 GB/s per node
-Grid       : Message        :  Stencil 12.9557 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663768
-Grid       : Message        : Average mflops/s per call per node       : 807580
-Grid       : Message        : Average mflops/s per call per node       : 820271
-Grid       : Message        : Average mflops/s per call per node       : 662811
-Grid       : Message        : Average mflops/s per call per node (full): 316384
-Grid       : Message        : Average mflops/s per call per node (full): 431281
-Grid       : Message        : Average mflops/s per call per node (full): 442801
-Grid       : Message        : Average mflops/s per call per node (full): 305739
-Grid       : Message        :  Stencil 12.9774 GB/s per node
-Grid       : Message        :  Stencil 17.2278 GB/s per node
-Grid       : Message        :  Stencil 17.2784 GB/s per node
-Grid       : Message        :  Stencil 13.1069 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664316
-Grid       : Message        : Average mflops/s per call per node       : 805309
-Grid       : Message        : Average mflops/s per call per node       : 822463
-Grid       : Message        : Average mflops/s per call per node       : 666367
-Grid       : Message        : Average mflops/s per call per node (full): 314621
-Grid       : Message        : Average mflops/s per call per node (full): 431987
-Grid       : Message        : Average mflops/s per call per node (full): 442618
-Grid       : Message        : Average mflops/s per call per node (full): 305410
-Grid       : Message        :  Stencil 13.6128 GB/s per node
-Grid       : Message        :  Stencil 16.5456 GB/s per node
-Grid       : Message        :  Stencil 17.9411 GB/s per node
-Grid       : Message        :  Stencil 11.9459 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660818
-Grid       : Message        : Average mflops/s per call per node       : 799034
-Grid       : Message        : Average mflops/s per call per node       : 821010
-Grid       : Message        : Average mflops/s per call per node       : 663987
-Grid       : Message        : Average mflops/s per call per node (full): 312564
-Grid       : Message        : Average mflops/s per call per node (full): 430588
-Grid       : Message        : Average mflops/s per call per node (full): 445427
-Grid       : Message        : Average mflops/s per call per node (full): 299941
-Grid       : Message        :  Stencil 13.9729 GB/s per node
-Grid       : Message        :  Stencil 17.9032 GB/s per node
-Grid       : Message        :  Stencil 17.6193 GB/s per node
-Grid       : Message        :  Stencil 13.7253 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663104
-Grid       : Message        : Average mflops/s per call per node       : 801495
-Grid       : Message        : Average mflops/s per call per node       : 825126
-Grid       : Message        : Average mflops/s per call per node       : 663916
-Grid       : Message        : Average mflops/s per call per node (full): 316378
-Grid       : Message        : Average mflops/s per call per node (full): 439695
-Grid       : Message        : Average mflops/s per call per node (full): 446413
-Grid       : Message        : Average mflops/s per call per node (full): 305940
-Grid       : Message        :  Stencil 13.0165 GB/s per node
-Grid       : Message        :  Stencil 17.1341 GB/s per node
-Grid       : Message        :  Stencil 18.0119 GB/s per node
-Grid       : Message        :  Stencil 12.7273 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667567
-Grid       : Message        : Average mflops/s per call per node       : 803152
-Grid       : Message        : Average mflops/s per call per node       : 822604
-Grid       : Message        : Average mflops/s per call per node       : 668782
-Grid       : Message        : Average mflops/s per call per node (full): 314125
-Grid       : Message        : Average mflops/s per call per node (full): 436297
-Grid       : Message        : Average mflops/s per call per node (full): 447752
-Grid       : Message        : Average mflops/s per call per node (full): 305420
-Grid       : Message        :  Stencil 13.8381 GB/s per node
-Grid       : Message        :  Stencil 17.2971 GB/s per node
-Grid       : Message        :  Stencil 17.7722 GB/s per node
-Grid       : Message        :  Stencil 12.8672 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662547
-Grid       : Message        : Average mflops/s per call per node       : 802852
-Grid       : Message        : Average mflops/s per call per node       : 821718
-Grid       : Message        : Average mflops/s per call per node       : 665407
-Grid       : Message        : Average mflops/s per call per node (full): 312763
-Grid       : Message        : Average mflops/s per call per node (full): 438642
-Grid       : Message        : Average mflops/s per call per node (full): 443881
-Grid       : Message        : Average mflops/s per call per node (full): 305626
-Grid       : Message        :  Stencil 12.7343 GB/s per node
-Grid       : Message        :  Stencil 17.479 GB/s per node
-Grid       : Message        :  Stencil 17.3539 GB/s per node
-Grid       : Message        :  Stencil 12.7125 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669782
-Grid       : Message        : Average mflops/s per call per node       : 801289
-Grid       : Message        : Average mflops/s per call per node       : 823651
-Grid       : Message        : Average mflops/s per call per node       : 669702
-Grid       : Message        : Average mflops/s per call per node (full): 314697
-Grid       : Message        : Average mflops/s per call per node (full): 439325
-Grid       : Message        : Average mflops/s per call per node (full): 442798
-Grid       : Message        : Average mflops/s per call per node (full): 305791
-Grid       : Message        :  Stencil 12.907 GB/s per node
-Grid       : Message        :  Stencil 17.4736 GB/s per node
-Grid       : Message        :  Stencil 17.468 GB/s per node
-Grid       : Message        :  Stencil 12.2998 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667492
-Grid       : Message        : Average mflops/s per call per node       : 806048
-Grid       : Message        : Average mflops/s per call per node       : 820264
-Grid       : Message        : Average mflops/s per call per node       : 667962
-Grid       : Message        : Average mflops/s per call per node (full): 314788
-Grid       : Message        : Average mflops/s per call per node (full): 440490
-Grid       : Message        : Average mflops/s per call per node (full): 442776
-Grid       : Message        : Average mflops/s per call per node (full): 303726
-Grid       : Message        :  Stencil 12.0599 GB/s per node
-Grid       : Message        :  Stencil 16.8494 GB/s per node
-Grid       : Message        :  Stencil 17.1225 GB/s per node
-Grid       : Message        :  Stencil 14.4116 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667139
-Grid       : Message        : Average mflops/s per call per node       : 801106
-Grid       : Message        : Average mflops/s per call per node       : 820180
-Grid       : Message        : Average mflops/s per call per node       : 661676
-Grid       : Message        : Average mflops/s per call per node (full): 309309
-Grid       : Message        : Average mflops/s per call per node (full): 437179
-Grid       : Message        : Average mflops/s per call per node (full): 441473
-Grid       : Message        : Average mflops/s per call per node (full): 307348
-Grid       : Message        :  Stencil 13.1343 GB/s per node
-Grid       : Message        :  Stencil 16.8507 GB/s per node
-Grid       : Message        :  Stencil 18.1028 GB/s per node
-Grid       : Message        :  Stencil 13.8568 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664792
-Grid       : Message        : Average mflops/s per call per node       : 802847
-Grid       : Message        : Average mflops/s per call per node       : 821274
-Grid       : Message        : Average mflops/s per call per node       : 661640
-Grid       : Message        : Average mflops/s per call per node (full): 315142
-Grid       : Message        : Average mflops/s per call per node (full): 434607
-Grid       : Message        : Average mflops/s per call per node (full): 445560
-Grid       : Message        : Average mflops/s per call per node (full): 305731
-Grid       : Message        :  Stencil 14.1103 GB/s per node
-Grid       : Message        :  Stencil 18.1799 GB/s per node
-Grid       : Message        :  Stencil 16.1787 GB/s per node
-Grid       : Message        :  Stencil 13.0525 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664712
-Grid       : Message        : Average mflops/s per call per node       : 801524
-Grid       : Message        : Average mflops/s per call per node       : 826986
-Grid       : Message        : Average mflops/s per call per node       : 663457
-Grid       : Message        : Average mflops/s per call per node (full): 316484
-Grid       : Message        : Average mflops/s per call per node (full): 440438
-Grid       : Message        : Average mflops/s per call per node (full): 419089
-Grid       : Message        : Average mflops/s per call per node (full): 304400
-Grid       : Message        :  Stencil 14.4843 GB/s per node
-Grid       : Message        :  Stencil 16.718 GB/s per node
-Grid       : Message        :  Stencil 17.5117 GB/s per node
-Grid       : Message        :  Stencil 12.4229 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659603
-Grid       : Message        : Average mflops/s per call per node       : 806019
-Grid       : Message        : Average mflops/s per call per node       : 823493
-Grid       : Message        : Average mflops/s per call per node       : 670796
-Grid       : Message        : Average mflops/s per call per node (full): 316105
-Grid       : Message        : Average mflops/s per call per node (full): 437210
-Grid       : Message        : Average mflops/s per call per node (full): 444135
-Grid       : Message        : Average mflops/s per call per node (full): 302971
-Grid       : Message        :  Stencil 13.5064 GB/s per node
-Grid       : Message        :  Stencil 10.846 GB/s per node
-Grid       : Message        :  Stencil 18.2372 GB/s per node
-Grid       : Message        :  Stencil 12.0147 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664622
-Grid       : Message        : Average mflops/s per call per node       : 811024
-Grid       : Message        : Average mflops/s per call per node       : 826650
-Grid       : Message        : Average mflops/s per call per node       : 664303
-Grid       : Message        : Average mflops/s per call per node (full): 314965
-Grid       : Message        : Average mflops/s per call per node (full): 336084
-Grid       : Message        : Average mflops/s per call per node (full): 449306
-Grid       : Message        : Average mflops/s per call per node (full): 301577
-Grid       : Message        :  Stencil 12.3366 GB/s per node
-Grid       : Message        :  Stencil 17.1414 GB/s per node
-Grid       : Message        :  Stencil 18.1672 GB/s per node
-Grid       : Message        :  Stencil 12.544 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668773
-Grid       : Message        : Average mflops/s per call per node       : 803488
-Grid       : Message        : Average mflops/s per call per node       : 823559
-Grid       : Message        : Average mflops/s per call per node       : 666131
-Grid       : Message        : Average mflops/s per call per node (full): 312170
-Grid       : Message        : Average mflops/s per call per node (full): 435861
-Grid       : Message        : Average mflops/s per call per node (full): 448144
-Grid       : Message        : Average mflops/s per call per node (full): 304729
-Grid       : Message        :  Stencil 13.962 GB/s per node
-Grid       : Message        :  Stencil 16.6905 GB/s per node
-Grid       : Message        :  Stencil 17.603 GB/s per node
-Grid       : Message        :  Stencil 12.6235 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664443
-Grid       : Message        : Average mflops/s per call per node       : 800430
-Grid       : Message        : Average mflops/s per call per node       : 821436
-Grid       : Message        : Average mflops/s per call per node       : 660055
-Grid       : Message        : Average mflops/s per call per node (full): 315811
-Grid       : Message        : Average mflops/s per call per node (full): 435238
-Grid       : Message        : Average mflops/s per call per node (full): 445246
-Grid       : Message        : Average mflops/s per call per node (full): 302370
-Grid       : Message        :  Stencil 12.8179 GB/s per node
-Grid       : Message        :  Stencil 16.5557 GB/s per node
-Grid       : Message        :  Stencil 17.7032 GB/s per node
-Grid       : Message        :  Stencil 11.7236 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669912
-Grid       : Message        : Average mflops/s per call per node       : 805298
-Grid       : Message        : Average mflops/s per call per node       : 823263
-Grid       : Message        : Average mflops/s per call per node       : 665005
-Grid       : Message        : Average mflops/s per call per node (full): 315095
-Grid       : Message        : Average mflops/s per call per node (full): 432900
-Grid       : Message        : Average mflops/s per call per node (full): 445229
-Grid       : Message        : Average mflops/s per call per node (full): 298923
-Grid       : Message        :  Stencil 15.3521 GB/s per node
-Grid       : Message        :  Stencil 17.6488 GB/s per node
-Grid       : Message        :  Stencil 17.7158 GB/s per node
-Grid       : Message        :  Stencil 14.2351 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662351
-Grid       : Message        : Average mflops/s per call per node       : 806679
-Grid       : Message        : Average mflops/s per call per node       : 825100
-Grid       : Message        : Average mflops/s per call per node       : 658530
-Grid       : Message        : Average mflops/s per call per node (full): 317173
-Grid       : Message        : Average mflops/s per call per node (full): 440280
-Grid       : Message        : Average mflops/s per call per node (full): 447566
-Grid       : Message        : Average mflops/s per call per node (full): 306379
-Grid       : Message        :  Stencil 13.817 GB/s per node
-Grid       : Message        :  Stencil 17.4135 GB/s per node
-Grid       : Message        :  Stencil 17.6026 GB/s per node
-Grid       : Message        :  Stencil 14.0141 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662014
-Grid       : Message        : Average mflops/s per call per node       : 801619
-Grid       : Message        : Average mflops/s per call per node       : 827237
-Grid       : Message        : Average mflops/s per call per node       : 656741
-Grid       : Message        : Average mflops/s per call per node (full): 315937
-Grid       : Message        : Average mflops/s per call per node (full): 438308
-Grid       : Message        : Average mflops/s per call per node (full): 445880
-Grid       : Message        : Average mflops/s per call per node (full): 306123
-Grid       : Message        :  Stencil 12.7159 GB/s per node
-Grid       : Message        :  Stencil 17.4458 GB/s per node
-Grid       : Message        :  Stencil 17.2803 GB/s per node
-Grid       : Message        :  Stencil 12.1528 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668929
-Grid       : Message        : Average mflops/s per call per node       : 805120
-Grid       : Message        : Average mflops/s per call per node       : 827176
-Grid       : Message        : Average mflops/s per call per node       : 667771
-Grid       : Message        : Average mflops/s per call per node (full): 315238
-Grid       : Message        : Average mflops/s per call per node (full): 438501
-Grid       : Message        : Average mflops/s per call per node (full): 443475
-Grid       : Message        : Average mflops/s per call per node (full): 302603
-Grid       : Message        :  Stencil 13.6026 GB/s per node
-Grid       : Message        :  Stencil 17.4989 GB/s per node
-Grid       : Message        :  Stencil 17.1213 GB/s per node
-Grid       : Message        :  Stencil 11.9238 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662910
-Grid       : Message        : Average mflops/s per call per node       : 798563
-Grid       : Message        : Average mflops/s per call per node       : 825342
-Grid       : Message        : Average mflops/s per call per node       : 665980
-Grid       : Message        : Average mflops/s per call per node (full): 315142
-Grid       : Message        : Average mflops/s per call per node (full): 437721
-Grid       : Message        : Average mflops/s per call per node (full): 439211
-Grid       : Message        : Average mflops/s per call per node (full): 300811
-Grid       : Message        :  Stencil 13.5228 GB/s per node
-Grid       : Message        :  Stencil 16.4377 GB/s per node
-Grid       : Message        :  Stencil 17.4979 GB/s per node
-Grid       : Message        :  Stencil 12.0545 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664701
-Grid       : Message        : Average mflops/s per call per node       : 806089
-Grid       : Message        : Average mflops/s per call per node       : 822489
-Grid       : Message        : Average mflops/s per call per node       : 664298
-Grid       : Message        : Average mflops/s per call per node (full): 312607
-Grid       : Message        : Average mflops/s per call per node (full): 433771
-Grid       : Message        : Average mflops/s per call per node (full): 444755
-Grid       : Message        : Average mflops/s per call per node (full): 299660
-Grid       : Message        :  Stencil 12.5371 GB/s per node
-Grid       : Message        :  Stencil 17.1109 GB/s per node
-Grid       : Message        :  Stencil 17.6479 GB/s per node
-Grid       : Message        :  Stencil 12.1976 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669989
-Grid       : Message        : Average mflops/s per call per node       : 807323
-Grid       : Message        : Average mflops/s per call per node       : 820493
-Grid       : Message        : Average mflops/s per call per node       : 658760
-Grid       : Message        : Average mflops/s per call per node (full): 311514
-Grid       : Message        : Average mflops/s per call per node (full): 440217
-Grid       : Message        : Average mflops/s per call per node (full): 444441
-Grid       : Message        : Average mflops/s per call per node (full): 302401
-Grid       : Message        :  Stencil 12.8845 GB/s per node
-Grid       : Message        :  Stencil 17.2483 GB/s per node
-Grid       : Message        :  Stencil 18.8034 GB/s per node
-Grid       : Message        :  Stencil 12.3223 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669446
-Grid       : Message        : Average mflops/s per call per node       : 800302
-Grid       : Message        : Average mflops/s per call per node       : 819148
-Grid       : Message        : Average mflops/s per call per node       : 661661
-Grid       : Message        : Average mflops/s per call per node (full): 314380
-Grid       : Message        : Average mflops/s per call per node (full): 437620
-Grid       : Message        : Average mflops/s per call per node (full): 447853
-Grid       : Message        : Average mflops/s per call per node (full): 303941
-Grid       : Message        :  Stencil 12.9305 GB/s per node
-Grid       : Message        :  Stencil 17.4124 GB/s per node
-Grid       : Message        :  Stencil 17.5552 GB/s per node
-Grid       : Message        :  Stencil 14.0837 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666221
-Grid       : Message        : Average mflops/s per call per node       : 799268
-Grid       : Message        : Average mflops/s per call per node       : 822312
-Grid       : Message        : Average mflops/s per call per node       : 658214
-Grid       : Message        : Average mflops/s per call per node (full): 315500
-Grid       : Message        : Average mflops/s per call per node (full): 437563
-Grid       : Message        : Average mflops/s per call per node (full): 434967
-Grid       : Message        : Average mflops/s per call per node (full): 305814
-Grid       : Message        :  Stencil 13.6345 GB/s per node
-Grid       : Message        :  Stencil 18.8579 GB/s per node
-Grid       : Message        :  Stencil 17.4464 GB/s per node
-Grid       : Message        :  Stencil 12.8563 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665744
-Grid       : Message        : Average mflops/s per call per node       : 800913
-Grid       : Message        : Average mflops/s per call per node       : 819724
-Grid       : Message        : Average mflops/s per call per node       : 664268
-Grid       : Message        : Average mflops/s per call per node (full): 315457
-Grid       : Message        : Average mflops/s per call per node (full): 442373
-Grid       : Message        : Average mflops/s per call per node (full): 443078
-Grid       : Message        : Average mflops/s per call per node (full): 304337
-Grid       : Message        :  Stencil 13.2966 GB/s per node
-Grid       : Message        :  Stencil 16.5044 GB/s per node
-Grid       : Message        :  Stencil 17.98 GB/s per node
-Grid       : Message        :  Stencil 13.0868 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667082
-Grid       : Message        : Average mflops/s per call per node       : 804780
-Grid       : Message        : Average mflops/s per call per node       : 824113
-Grid       : Message        : Average mflops/s per call per node       : 662976
-Grid       : Message        : Average mflops/s per call per node (full): 315618
-Grid       : Message        : Average mflops/s per call per node (full): 434837
-Grid       : Message        : Average mflops/s per call per node (full): 444083
-Grid       : Message        : Average mflops/s per call per node (full): 304187
-Grid       : Message        :  Stencil 13.6028 GB/s per node
-Grid       : Message        :  Stencil 11.7606 GB/s per node
-Grid       : Message        :  Stencil 18.1364 GB/s per node
-Grid       : Message        :  Stencil 12.9121 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664563
-Grid       : Message        : Average mflops/s per call per node       : 810802
-Grid       : Message        : Average mflops/s per call per node       : 816970
-Grid       : Message        : Average mflops/s per call per node       : 664115
-Grid       : Message        : Average mflops/s per call per node (full): 315142
-Grid       : Message        : Average mflops/s per call per node (full): 356461
-Grid       : Message        : Average mflops/s per call per node (full): 445728
-Grid       : Message        : Average mflops/s per call per node (full): 306152
-Grid       : Message        :  Stencil 13.0699 GB/s per node
-Grid       : Message        :  Stencil 16.2134 GB/s per node
-Grid       : Message        :  Stencil 17.513 GB/s per node
-Grid       : Message        :  Stencil 13.8265 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667083
-Grid       : Message        : Average mflops/s per call per node       : 805456
-Grid       : Message        : Average mflops/s per call per node       : 830206
-Grid       : Message        : Average mflops/s per call per node       : 661448
-Grid       : Message        : Average mflops/s per call per node (full): 313897
-Grid       : Message        : Average mflops/s per call per node (full): 425836
-Grid       : Message        : Average mflops/s per call per node (full): 444927
-Grid       : Message        : Average mflops/s per call per node (full): 306378
-Grid       : Message        :  Stencil 13.472 GB/s per node
-Grid       : Message        :  Stencil 16.6024 GB/s per node
-Grid       : Message        :  Stencil 17.4968 GB/s per node
-Grid       : Message        :  Stencil 13.2805 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664531
-Grid       : Message        : Average mflops/s per call per node       : 806542
-Grid       : Message        : Average mflops/s per call per node       : 819825
-Grid       : Message        : Average mflops/s per call per node       : 666230
-Grid       : Message        : Average mflops/s per call per node (full): 315455
-Grid       : Message        : Average mflops/s per call per node (full): 436165
-Grid       : Message        : Average mflops/s per call per node (full): 445133
-Grid       : Message        : Average mflops/s per call per node (full): 306197
-Grid       : Message        :  Stencil 12.518 GB/s per node
-Grid       : Message        :  Stencil 17.3645 GB/s per node
-Grid       : Message        :  Stencil 17.1035 GB/s per node
-Grid       : Message        :  Stencil 12.106 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666808
-Grid       : Message        : Average mflops/s per call per node       : 802389
-Grid       : Message        : Average mflops/s per call per node       : 819214
-Grid       : Message        : Average mflops/s per call per node       : 667164
-Grid       : Message        : Average mflops/s per call per node (full): 313082
-Grid       : Message        : Average mflops/s per call per node (full): 439283
-Grid       : Message        : Average mflops/s per call per node (full): 440588
-Grid       : Message        : Average mflops/s per call per node (full): 304126
-Grid       : Message        :  Stencil 12.9822 GB/s per node
-Grid       : Message        :  Stencil 15.7679 GB/s per node
-Grid       : Message        :  Stencil 17.8964 GB/s per node
-Grid       : Message        :  Stencil 12.1026 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667679
-Grid       : Message        : Average mflops/s per call per node       : 802767
-Grid       : Message        : Average mflops/s per call per node       : 820210
-Grid       : Message        : Average mflops/s per call per node       : 666777
-Grid       : Message        : Average mflops/s per call per node (full): 314809
-Grid       : Message        : Average mflops/s per call per node (full): 425015
-Grid       : Message        : Average mflops/s per call per node (full): 444200
-Grid       : Message        : Average mflops/s per call per node (full): 302785
-Grid       : Message        :  Stencil 13.074 GB/s per node
-Grid       : Message        :  Stencil 16.3824 GB/s per node
-Grid       : Message        :  Stencil 17.4478 GB/s per node
-Grid       : Message        :  Stencil 12.6308 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664206
-Grid       : Message        : Average mflops/s per call per node       : 804750
-Grid       : Message        : Average mflops/s per call per node       : 825824
-Grid       : Message        : Average mflops/s per call per node       : 668402
-Grid       : Message        : Average mflops/s per call per node (full): 313047
-Grid       : Message        : Average mflops/s per call per node (full): 433801
-Grid       : Message        : Average mflops/s per call per node (full): 446196
-Grid       : Message        : Average mflops/s per call per node (full): 305143
-Grid       : Message        :  Stencil 14.5859 GB/s per node
-Grid       : Message        :  Stencil 8.49655 GB/s per node
-Grid       : Message        :  Stencil 17.462 GB/s per node
-Grid       : Message        :  Stencil 13.5331 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659024
-Grid       : Message        : Average mflops/s per call per node       : 814575
-Grid       : Message        : Average mflops/s per call per node       : 816364
-Grid       : Message        : Average mflops/s per call per node       : 665033
-Grid       : Message        : Average mflops/s per call per node (full): 316545
-Grid       : Message        : Average mflops/s per call per node (full): 278554
-Grid       : Message        : Average mflops/s per call per node (full): 442461
-Grid       : Message        : Average mflops/s per call per node (full): 304570
-Grid       : Message        :  Stencil 13.853 GB/s per node
-Grid       : Message        :  Stencil 17.2408 GB/s per node
-Grid       : Message        :  Stencil 17.7049 GB/s per node
-Grid       : Message        :  Stencil 13.8443 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661716
-Grid       : Message        : Average mflops/s per call per node       : 803346
-Grid       : Message        : Average mflops/s per call per node       : 816573
-Grid       : Message        : Average mflops/s per call per node       : 662450
-Grid       : Message        : Average mflops/s per call per node (full): 316011
-Grid       : Message        : Average mflops/s per call per node (full): 436728
-Grid       : Message        : Average mflops/s per call per node (full): 441098
-Grid       : Message        : Average mflops/s per call per node (full): 304371
-Grid       : Message        :  Stencil 13.3244 GB/s per node
-Grid       : Message        :  Stencil 16.8162 GB/s per node
-Grid       : Message        :  Stencil 17.7584 GB/s per node
-Grid       : Message        :  Stencil 11.9128 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667918
-Grid       : Message        : Average mflops/s per call per node       : 804741
-Grid       : Message        : Average mflops/s per call per node       : 820470
-Grid       : Message        : Average mflops/s per call per node       : 666901
-Grid       : Message        : Average mflops/s per call per node (full): 314445
-Grid       : Message        : Average mflops/s per call per node (full): 436457
-Grid       : Message        : Average mflops/s per call per node (full): 443473
-Grid       : Message        : Average mflops/s per call per node (full): 300946
-Grid       : Message        :  Stencil 13.4446 GB/s per node
-Grid       : Message        :  Stencil 16.8478 GB/s per node
-Grid       : Message        :  Stencil 17.7755 GB/s per node
-Grid       : Message        :  Stencil 13.8389 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667766
-Grid       : Message        : Average mflops/s per call per node       : 802491
-Grid       : Message        : Average mflops/s per call per node       : 821947
-Grid       : Message        : Average mflops/s per call per node       : 665397
-Grid       : Message        : Average mflops/s per call per node (full): 316293
-Grid       : Message        : Average mflops/s per call per node (full): 437154
-Grid       : Message        : Average mflops/s per call per node (full): 446942
-Grid       : Message        : Average mflops/s per call per node (full): 307415
-Grid       : Message        :  Stencil 14.0143 GB/s per node
-Grid       : Message        :  Stencil 16.6462 GB/s per node
-Grid       : Message        :  Stencil 18.0273 GB/s per node
-Grid       : Message        :  Stencil 12.2425 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666333
-Grid       : Message        : Average mflops/s per call per node       : 805973
-Grid       : Message        : Average mflops/s per call per node       : 818586
-Grid       : Message        : Average mflops/s per call per node       : 667479
-Grid       : Message        : Average mflops/s per call per node (full): 316949
-Grid       : Message        : Average mflops/s per call per node (full): 434593
-Grid       : Message        : Average mflops/s per call per node (full): 439484
-Grid       : Message        : Average mflops/s per call per node (full): 303910
-Grid       : Message        :  Stencil 13.7419 GB/s per node
-Grid       : Message        :  Stencil 16.4797 GB/s per node
-Grid       : Message        :  Stencil 16.9528 GB/s per node
-Grid       : Message        :  Stencil 12.4341 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669863
-Grid       : Message        : Average mflops/s per call per node       : 804020
-Grid       : Message        : Average mflops/s per call per node       : 818802
-Grid       : Message        : Average mflops/s per call per node       : 663249
-Grid       : Message        : Average mflops/s per call per node (full): 316526
-Grid       : Message        : Average mflops/s per call per node (full): 431400
-Grid       : Message        : Average mflops/s per call per node (full): 439791
-Grid       : Message        : Average mflops/s per call per node (full): 303398
-Grid       : Message        :  Stencil 12.9811 GB/s per node
-Grid       : Message        :  Stencil 16.9562 GB/s per node
-Grid       : Message        :  Stencil 18.895 GB/s per node
-Grid       : Message        :  Stencil 14.0543 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669785
-Grid       : Message        : Average mflops/s per call per node       : 805240
-Grid       : Message        : Average mflops/s per call per node       : 820750
-Grid       : Message        : Average mflops/s per call per node       : 657780
-Grid       : Message        : Average mflops/s per call per node (full): 315425
-Grid       : Message        : Average mflops/s per call per node (full): 437260
-Grid       : Message        : Average mflops/s per call per node (full): 448428
-Grid       : Message        : Average mflops/s per call per node (full): 305976
-Grid       : Message        :  Stencil 12.9075 GB/s per node
-Grid       : Message        :  Stencil 16.5077 GB/s per node
-Grid       : Message        :  Stencil 18.1803 GB/s per node
-Grid       : Message        :  Stencil 13.2324 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669413
-Grid       : Message        : Average mflops/s per call per node       : 806530
-Grid       : Message        : Average mflops/s per call per node       : 820634
-Grid       : Message        : Average mflops/s per call per node       : 663184
-Grid       : Message        : Average mflops/s per call per node (full): 314776
-Grid       : Message        : Average mflops/s per call per node (full): 434158
-Grid       : Message        : Average mflops/s per call per node (full): 448497
-Grid       : Message        : Average mflops/s per call per node (full): 304859
-Grid       : Message        :  Stencil 13.287 GB/s per node
-Grid       : Message        :  Stencil 16.4753 GB/s per node
-Grid       : Message        :  Stencil 17.1715 GB/s per node
-Grid       : Message        :  Stencil 15.9104 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668487
-Grid       : Message        : Average mflops/s per call per node       : 806878
-Grid       : Message        : Average mflops/s per call per node       : 820880
-Grid       : Message        : Average mflops/s per call per node       : 661266
-Grid       : Message        : Average mflops/s per call per node (full): 315909
-Grid       : Message        : Average mflops/s per call per node (full): 431333
-Grid       : Message        : Average mflops/s per call per node (full): 442874
-Grid       : Message        : Average mflops/s per call per node (full): 308064
-Grid       : Message        :  Stencil 13.4721 GB/s per node
-Grid       : Message        :  Stencil 17.1158 GB/s per node
-Grid       : Message        :  Stencil 17.2517 GB/s per node
-Grid       : Message        :  Stencil 11.9345 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666143
-Grid       : Message        : Average mflops/s per call per node       : 801154
-Grid       : Message        : Average mflops/s per call per node       : 821732
-Grid       : Message        : Average mflops/s per call per node       : 665116
-Grid       : Message        : Average mflops/s per call per node (full): 316290
-Grid       : Message        : Average mflops/s per call per node (full): 438552
-Grid       : Message        : Average mflops/s per call per node (full): 443037
-Grid       : Message        : Average mflops/s per call per node (full): 301309
-Grid       : Message        :  Stencil 13.3112 GB/s per node
-Grid       : Message        :  Stencil 16.9799 GB/s per node
-Grid       : Message        :  Stencil 17.3504 GB/s per node
-Grid       : Message        :  Stencil 12.3472 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666107
-Grid       : Message        : Average mflops/s per call per node       : 803853
-Grid       : Message        : Average mflops/s per call per node       : 826777
-Grid       : Message        : Average mflops/s per call per node       : 663021
-Grid       : Message        : Average mflops/s per call per node (full): 315759
-Grid       : Message        : Average mflops/s per call per node (full): 437814
-Grid       : Message        : Average mflops/s per call per node (full): 444277
-Grid       : Message        : Average mflops/s per call per node (full): 302103
-Grid       : Message        :  Stencil 14.5476 GB/s per node
-Grid       : Message        :  Stencil 16.744 GB/s per node
-Grid       : Message        :  Stencil 18.094 GB/s per node
-Grid       : Message        :  Stencil 12.4832 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661807
-Grid       : Message        : Average mflops/s per call per node       : 805374
-Grid       : Message        : Average mflops/s per call per node       : 815879
-Grid       : Message        : Average mflops/s per call per node       : 669430
-Grid       : Message        : Average mflops/s per call per node (full): 317179
-Grid       : Message        : Average mflops/s per call per node (full): 436523
-Grid       : Message        : Average mflops/s per call per node (full): 445499
-Grid       : Message        : Average mflops/s per call per node (full): 304739
-Grid       : Message        :  Stencil 13.5476 GB/s per node
-Grid       : Message        :  Stencil 17.0627 GB/s per node
-Grid       : Message        :  Stencil 17.9587 GB/s per node
-Grid       : Message        :  Stencil 12.7293 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664909
-Grid       : Message        : Average mflops/s per call per node       : 802387
-Grid       : Message        : Average mflops/s per call per node       : 819178
-Grid       : Message        : Average mflops/s per call per node       : 662260
-Grid       : Message        : Average mflops/s per call per node (full): 315249
-Grid       : Message        : Average mflops/s per call per node (full): 435294
-Grid       : Message        : Average mflops/s per call per node (full): 446118
-Grid       : Message        : Average mflops/s per call per node (full): 304420
-Grid       : Message        :  Stencil 14.9442 GB/s per node
-Grid       : Message        :  Stencil 17.2568 GB/s per node
-Grid       : Message        :  Stencil 17.3896 GB/s per node
-Grid       : Message        :  Stencil 12.517 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663158
-Grid       : Message        : Average mflops/s per call per node       : 803797
-Grid       : Message        : Average mflops/s per call per node       : 820222
-Grid       : Message        : Average mflops/s per call per node       : 664988
-Grid       : Message        : Average mflops/s per call per node (full): 317171
-Grid       : Message        : Average mflops/s per call per node (full): 437926
-Grid       : Message        : Average mflops/s per call per node (full): 443898
-Grid       : Message        : Average mflops/s per call per node (full): 303122
-Grid       : Message        :  Stencil 13.7926 GB/s per node
-Grid       : Message        :  Stencil 10.182 GB/s per node
-Grid       : Message        :  Stencil 17.1182 GB/s per node
-Grid       : Message        :  Stencil 12.6113 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666118
-Grid       : Message        : Average mflops/s per call per node       : 811413
-Grid       : Message        : Average mflops/s per call per node       : 827386
-Grid       : Message        : Average mflops/s per call per node       : 662404
-Grid       : Message        : Average mflops/s per call per node (full): 316161
-Grid       : Message        : Average mflops/s per call per node (full): 320980
-Grid       : Message        : Average mflops/s per call per node (full): 441691
-Grid       : Message        : Average mflops/s per call per node (full): 304311
-Grid       : Message        :  Stencil 15.0103 GB/s per node
-Grid       : Message        :  Stencil 17.974 GB/s per node
-Grid       : Message        :  Stencil 16.8732 GB/s per node
-Grid       : Message        :  Stencil 11.9428 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659495
-Grid       : Message        : Average mflops/s per call per node       : 798743
-Grid       : Message        : Average mflops/s per call per node       : 828324
-Grid       : Message        : Average mflops/s per call per node       : 669846
-Grid       : Message        : Average mflops/s per call per node (full): 315714
-Grid       : Message        : Average mflops/s per call per node (full): 439913
-Grid       : Message        : Average mflops/s per call per node (full): 439295
-Grid       : Message        : Average mflops/s per call per node (full): 301726
-Grid       : Message        :  Stencil 13.4555 GB/s per node
-Grid       : Message        :  Stencil 16.2566 GB/s per node
-Grid       : Message        :  Stencil 17.5285 GB/s per node
-Grid       : Message        :  Stencil 13.3142 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664309
-Grid       : Message        : Average mflops/s per call per node       : 801698
-Grid       : Message        : Average mflops/s per call per node       : 823972
-Grid       : Message        : Average mflops/s per call per node       : 663338
-Grid       : Message        : Average mflops/s per call per node (full): 314162
-Grid       : Message        : Average mflops/s per call per node (full): 431179
-Grid       : Message        : Average mflops/s per call per node (full): 445222
-Grid       : Message        : Average mflops/s per call per node (full): 305774
-Grid       : Message        :  Stencil 12.698 GB/s per node
-Grid       : Message        :  Stencil 16.5092 GB/s per node
-Grid       : Message        :  Stencil 17.4474 GB/s per node
-Grid       : Message        :  Stencil 12.3678 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665460
-Grid       : Message        : Average mflops/s per call per node       : 803692
-Grid       : Message        : Average mflops/s per call per node       : 814742
-Grid       : Message        : Average mflops/s per call per node       : 665287
-Grid       : Message        : Average mflops/s per call per node (full): 314678
-Grid       : Message        : Average mflops/s per call per node (full): 433743
-Grid       : Message        : Average mflops/s per call per node (full): 432566
-Grid       : Message        : Average mflops/s per call per node (full): 304588
-Grid       : Message        :  Stencil 13.7497 GB/s per node
-Grid       : Message        :  Stencil 16.4933 GB/s per node
-Grid       : Message        :  Stencil 17.969 GB/s per node
-Grid       : Message        :  Stencil 15.9667 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663973
-Grid       : Message        : Average mflops/s per call per node       : 809146
-Grid       : Message        : Average mflops/s per call per node       : 824922
-Grid       : Message        : Average mflops/s per call per node       : 656900
-Grid       : Message        : Average mflops/s per call per node (full): 314409
-Grid       : Message        : Average mflops/s per call per node (full): 434490
-Grid       : Message        : Average mflops/s per call per node (full): 446536
-Grid       : Message        : Average mflops/s per call per node (full): 307389
-Grid       : Message        :  Stencil 12.961 GB/s per node
-Grid       : Message        :  Stencil 12.5404 GB/s per node
-Grid       : Message        :  Stencil 16.8662 GB/s per node
-Grid       : Message        :  Stencil 12.6023 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666260
-Grid       : Message        : Average mflops/s per call per node       : 809160
-Grid       : Message        : Average mflops/s per call per node       : 821739
-Grid       : Message        : Average mflops/s per call per node       : 666750
-Grid       : Message        : Average mflops/s per call per node (full): 313902
-Grid       : Message        : Average mflops/s per call per node (full): 371663
-Grid       : Message        : Average mflops/s per call per node (full): 437243
-Grid       : Message        : Average mflops/s per call per node (full): 300832
-Grid       : Message        :  Stencil 13.1174 GB/s per node
-Grid       : Message        :  Stencil 16.4419 GB/s per node
-Grid       : Message        :  Stencil 17.6359 GB/s per node
-Grid       : Message        :  Stencil 12.9856 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666836
-Grid       : Message        : Average mflops/s per call per node       : 803190
-Grid       : Message        : Average mflops/s per call per node       : 831600
-Grid       : Message        : Average mflops/s per call per node       : 663531
-Grid       : Message        : Average mflops/s per call per node (full): 315346
-Grid       : Message        : Average mflops/s per call per node (full): 434213
-Grid       : Message        : Average mflops/s per call per node (full): 446042
-Grid       : Message        : Average mflops/s per call per node (full): 305281
-Grid       : Message        :  Stencil 13.8596 GB/s per node
-Grid       : Message        :  Stencil 16.4839 GB/s per node
-Grid       : Message        :  Stencil 16.9857 GB/s per node
-Grid       : Message        :  Stencil 11.9013 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665002
-Grid       : Message        : Average mflops/s per call per node       : 803603
-Grid       : Message        : Average mflops/s per call per node       : 826298
-Grid       : Message        : Average mflops/s per call per node       : 666970
-Grid       : Message        : Average mflops/s per call per node (full): 312641
-Grid       : Message        : Average mflops/s per call per node (full): 435404
-Grid       : Message        : Average mflops/s per call per node (full): 439788
-Grid       : Message        : Average mflops/s per call per node (full): 300583
-Grid       : Message        :  Stencil 13.126 GB/s per node
-Grid       : Message        :  Stencil 7.92625 GB/s per node
-Grid       : Message        :  Stencil 17.2697 GB/s per node
-Grid       : Message        :  Stencil 12.5352 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660996
-Grid       : Message        : Average mflops/s per call per node       : 812284
-Grid       : Message        : Average mflops/s per call per node       : 825278
-Grid       : Message        : Average mflops/s per call per node       : 668621
-Grid       : Message        : Average mflops/s per call per node (full): 313449
-Grid       : Message        : Average mflops/s per call per node (full): 264451
-Grid       : Message        : Average mflops/s per call per node (full): 442895
-Grid       : Message        : Average mflops/s per call per node (full): 305272
-Grid       : Message        :  Stencil 14.0345 GB/s per node
-Grid       : Message        :  Stencil 17.1177 GB/s per node
-Grid       : Message        :  Stencil 17.9579 GB/s per node
-Grid       : Message        :  Stencil 13.7344 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664502
-Grid       : Message        : Average mflops/s per call per node       : 805953
-Grid       : Message        : Average mflops/s per call per node       : 825593
-Grid       : Message        : Average mflops/s per call per node       : 660114
-Grid       : Message        : Average mflops/s per call per node (full): 315873
-Grid       : Message        : Average mflops/s per call per node (full): 437394
-Grid       : Message        : Average mflops/s per call per node (full): 447291
-Grid       : Message        : Average mflops/s per call per node (full): 304741
-Grid       : Message        :  Stencil 13.4645 GB/s per node
-Grid       : Message        :  Stencil 17.4987 GB/s per node
-Grid       : Message        :  Stencil 17.4917 GB/s per node
-Grid       : Message        :  Stencil 13.0298 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666735
-Grid       : Message        : Average mflops/s per call per node       : 803387
-Grid       : Message        : Average mflops/s per call per node       : 815749
-Grid       : Message        : Average mflops/s per call per node       : 664987
-Grid       : Message        : Average mflops/s per call per node (full): 316438
-Grid       : Message        : Average mflops/s per call per node (full): 439568
-Grid       : Message        : Average mflops/s per call per node (full): 443471
-Grid       : Message        : Average mflops/s per call per node (full): 305477
-Grid       : Message        :  Stencil 13.3655 GB/s per node
-Grid       : Message        :  Stencil 17.6846 GB/s per node
-Grid       : Message        :  Stencil 18.5753 GB/s per node
-Grid       : Message        :  Stencil 13.1193 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664948
-Grid       : Message        : Average mflops/s per call per node       : 806213
-Grid       : Message        : Average mflops/s per call per node       : 823464
-Grid       : Message        : Average mflops/s per call per node       : 662498
-Grid       : Message        : Average mflops/s per call per node (full): 315014
-Grid       : Message        : Average mflops/s per call per node (full): 441551
-Grid       : Message        : Average mflops/s per call per node (full): 447925
-Grid       : Message        : Average mflops/s per call per node (full): 305503
-Grid       : Message        :  Stencil 14.1704 GB/s per node
-Grid       : Message        :  Stencil 16.4379 GB/s per node
-Grid       : Message        :  Stencil 17.1236 GB/s per node
-Grid       : Message        :  Stencil 12.5188 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665015
-Grid       : Message        : Average mflops/s per call per node       : 804900
-Grid       : Message        : Average mflops/s per call per node       : 821353
-Grid       : Message        : Average mflops/s per call per node       : 666641
-Grid       : Message        : Average mflops/s per call per node (full): 316231
-Grid       : Message        : Average mflops/s per call per node (full): 434665
-Grid       : Message        : Average mflops/s per call per node (full): 435790
-Grid       : Message        : Average mflops/s per call per node (full): 297563
-Grid       : Message        :  Stencil 13.258 GB/s per node
-Grid       : Message        :  Stencil 16.6796 GB/s per node
-Grid       : Message        :  Stencil 17.6714 GB/s per node
-Grid       : Message        :  Stencil 13.6617 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668216
-Grid       : Message        : Average mflops/s per call per node       : 804154
-Grid       : Message        : Average mflops/s per call per node       : 819799
-Grid       : Message        : Average mflops/s per call per node       : 662504
-Grid       : Message        : Average mflops/s per call per node (full): 315646
-Grid       : Message        : Average mflops/s per call per node (full): 436768
-Grid       : Message        : Average mflops/s per call per node (full): 445854
-Grid       : Message        : Average mflops/s per call per node (full): 306523
-Grid       : Message        :  Stencil 12.5776 GB/s per node
-Grid       : Message        :  Stencil 19.1094 GB/s per node
-Grid       : Message        :  Stencil 17.8382 GB/s per node
-Grid       : Message        :  Stencil 12.3613 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668374
-Grid       : Message        : Average mflops/s per call per node       : 803874
-Grid       : Message        : Average mflops/s per call per node       : 822898
-Grid       : Message        : Average mflops/s per call per node       : 663293
-Grid       : Message        : Average mflops/s per call per node (full): 314337
-Grid       : Message        : Average mflops/s per call per node (full): 442282
-Grid       : Message        : Average mflops/s per call per node (full): 445664
-Grid       : Message        : Average mflops/s per call per node (full): 302925
-Grid       : Message        :  Stencil 13.4442 GB/s per node
-Grid       : Message        :  Stencil 16.8888 GB/s per node
-Grid       : Message        :  Stencil 17.1822 GB/s per node
-Grid       : Message        :  Stencil 13.2556 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664317
-Grid       : Message        : Average mflops/s per call per node       : 801116
-Grid       : Message        : Average mflops/s per call per node       : 818191
-Grid       : Message        : Average mflops/s per call per node       : 661072
-Grid       : Message        : Average mflops/s per call per node (full): 314244
-Grid       : Message        : Average mflops/s per call per node (full): 437558
-Grid       : Message        : Average mflops/s per call per node (full): 443380
-Grid       : Message        : Average mflops/s per call per node (full): 305380
-Grid       : Message        :  Stencil 13.4673 GB/s per node
-Grid       : Message        :  Stencil 17.4665 GB/s per node
-Grid       : Message        :  Stencil 18.2401 GB/s per node
-Grid       : Message        :  Stencil 13.2764 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664984
-Grid       : Message        : Average mflops/s per call per node       : 805208
-Grid       : Message        : Average mflops/s per call per node       : 827513
-Grid       : Message        : Average mflops/s per call per node       : 664600
-Grid       : Message        : Average mflops/s per call per node (full): 314235
-Grid       : Message        : Average mflops/s per call per node (full): 438965
-Grid       : Message        : Average mflops/s per call per node (full): 448809
-Grid       : Message        : Average mflops/s per call per node (full): 306326
-Grid       : Message        :  Stencil 14.6654 GB/s per node
-Grid       : Message        :  Stencil 7.58732 GB/s per node
-Grid       : Message        :  Stencil 18.1682 GB/s per node
-Grid       : Message        :  Stencil 11.7598 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661720
-Grid       : Message        : Average mflops/s per call per node       : 806420
-Grid       : Message        : Average mflops/s per call per node       : 818420
-Grid       : Message        : Average mflops/s per call per node       : 668057
-Grid       : Message        : Average mflops/s per call per node (full): 316368
-Grid       : Message        : Average mflops/s per call per node (full): 255634
-Grid       : Message        : Average mflops/s per call per node (full): 446359
-Grid       : Message        : Average mflops/s per call per node (full): 298658
-Grid       : Message        :  Stencil 14.6144 GB/s per node
-Grid       : Message        :  Stencil 16.5195 GB/s per node
-Grid       : Message        :  Stencil 17.374 GB/s per node
-Grid       : Message        :  Stencil 13.3331 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663454
-Grid       : Message        : Average mflops/s per call per node       : 803748
-Grid       : Message        : Average mflops/s per call per node       : 827999
-Grid       : Message        : Average mflops/s per call per node       : 663467
-Grid       : Message        : Average mflops/s per call per node (full): 316129
-Grid       : Message        : Average mflops/s per call per node (full): 434864
-Grid       : Message        : Average mflops/s per call per node (full): 443165
-Grid       : Message        : Average mflops/s per call per node (full): 304171
-Grid       : Message        :  Stencil 14.3637 GB/s per node
-Grid       : Message        :  Stencil 17.712 GB/s per node
-Grid       : Message        :  Stencil 17.178 GB/s per node
-Grid       : Message        :  Stencil 12.8432 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665494
-Grid       : Message        : Average mflops/s per call per node       : 802527
-Grid       : Message        : Average mflops/s per call per node       : 821649
-Grid       : Message        : Average mflops/s per call per node       : 660891
-Grid       : Message        : Average mflops/s per call per node (full): 316872
-Grid       : Message        : Average mflops/s per call per node (full): 436560
-Grid       : Message        : Average mflops/s per call per node (full): 441310
-Grid       : Message        : Average mflops/s per call per node (full): 305030
-Grid       : Message        :  Stencil 13.6686 GB/s per node
-Grid       : Message        :  Stencil 17.4995 GB/s per node
-Grid       : Message        :  Stencil 17.6195 GB/s per node
-Grid       : Message        :  Stencil 12.6683 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663927
-Grid       : Message        : Average mflops/s per call per node       : 804827
-Grid       : Message        : Average mflops/s per call per node       : 828559
-Grid       : Message        : Average mflops/s per call per node       : 667145
-Grid       : Message        : Average mflops/s per call per node (full): 313344
-Grid       : Message        : Average mflops/s per call per node (full): 435859
-Grid       : Message        : Average mflops/s per call per node (full): 445948
-Grid       : Message        : Average mflops/s per call per node (full): 303649
-Grid       : Message        :  Stencil 13.2974 GB/s per node
-Grid       : Message        :  Stencil 16.9957 GB/s per node
-Grid       : Message        :  Stencil 17.9603 GB/s per node
-Grid       : Message        :  Stencil 13.7693 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660695
-Grid       : Message        : Average mflops/s per call per node       : 799165
-Grid       : Message        : Average mflops/s per call per node       : 820450
-Grid       : Message        : Average mflops/s per call per node       : 661012
-Grid       : Message        : Average mflops/s per call per node (full): 312753
-Grid       : Message        : Average mflops/s per call per node (full): 436415
-Grid       : Message        : Average mflops/s per call per node (full): 447591
-Grid       : Message        : Average mflops/s per call per node (full): 306742
-Grid       : Message        :  Stencil 14.316 GB/s per node
-Grid       : Message        :  Stencil 17.3739 GB/s per node
-Grid       : Message        :  Stencil 17.5313 GB/s per node
-Grid       : Message        :  Stencil 12.8304 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 658790
-Grid       : Message        : Average mflops/s per call per node       : 799323
-Grid       : Message        : Average mflops/s per call per node       : 824848
-Grid       : Message        : Average mflops/s per call per node       : 665160
-Grid       : Message        : Average mflops/s per call per node (full): 316336
-Grid       : Message        : Average mflops/s per call per node (full): 435574
-Grid       : Message        : Average mflops/s per call per node (full): 440652
-Grid       : Message        : Average mflops/s per call per node (full): 304537
-Grid       : Message        :  Stencil 14.3048 GB/s per node
-Grid       : Message        :  Stencil 17.5397 GB/s per node
-Grid       : Message        :  Stencil 17.0633 GB/s per node
-Grid       : Message        :  Stencil 13.5409 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659717
-Grid       : Message        : Average mflops/s per call per node       : 804485
-Grid       : Message        : Average mflops/s per call per node       : 825617
-Grid       : Message        : Average mflops/s per call per node       : 661768
-Grid       : Message        : Average mflops/s per call per node (full): 314706
-Grid       : Message        : Average mflops/s per call per node (full): 439069
-Grid       : Message        : Average mflops/s per call per node (full): 440527
-Grid       : Message        : Average mflops/s per call per node (full): 306173
-Grid       : Message        :  Stencil 12.8025 GB/s per node
-Grid       : Message        :  Stencil 17.2434 GB/s per node
-Grid       : Message        :  Stencil 17.9128 GB/s per node
-Grid       : Message        :  Stencil 13.279 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668256
-Grid       : Message        : Average mflops/s per call per node       : 803676
-Grid       : Message        : Average mflops/s per call per node       : 820491
-Grid       : Message        : Average mflops/s per call per node       : 662418
-Grid       : Message        : Average mflops/s per call per node (full): 315205
-Grid       : Message        : Average mflops/s per call per node (full): 437749
-Grid       : Message        : Average mflops/s per call per node (full): 446306
-Grid       : Message        : Average mflops/s per call per node (full): 305289
-Grid       : Message        :  Stencil 12.5601 GB/s per node
-Grid       : Message        :  Stencil 16.8043 GB/s per node
-Grid       : Message        :  Stencil 16.5785 GB/s per node
-Grid       : Message        :  Stencil 12.8368 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668628
-Grid       : Message        : Average mflops/s per call per node       : 803293
-Grid       : Message        : Average mflops/s per call per node       : 821762
-Grid       : Message        : Average mflops/s per call per node       : 660631
-Grid       : Message        : Average mflops/s per call per node (full): 313970
-Grid       : Message        : Average mflops/s per call per node (full): 437030
-Grid       : Message        : Average mflops/s per call per node (full): 426193
-Grid       : Message        : Average mflops/s per call per node (full): 303849
-Grid       : Message        :  Stencil 12.2139 GB/s per node
-Grid       : Message        :  Stencil 16.8379 GB/s per node
-Grid       : Message        :  Stencil 17.767 GB/s per node
-Grid       : Message        :  Stencil 13.2638 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671858
-Grid       : Message        : Average mflops/s per call per node       : 803407
-Grid       : Message        : Average mflops/s per call per node       : 822154
-Grid       : Message        : Average mflops/s per call per node       : 665694
-Grid       : Message        : Average mflops/s per call per node (full): 311775
-Grid       : Message        : Average mflops/s per call per node (full): 438106
-Grid       : Message        : Average mflops/s per call per node (full): 446340
-Grid       : Message        : Average mflops/s per call per node (full): 305394
-Grid       : Message        :  Stencil 13.1395 GB/s per node
-Grid       : Message        :  Stencil 17.0386 GB/s per node
-Grid       : Message        :  Stencil 17.3741 GB/s per node
-Grid       : Message        :  Stencil 12.463 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667151
-Grid       : Message        : Average mflops/s per call per node       : 802197
-Grid       : Message        : Average mflops/s per call per node       : 822477
-Grid       : Message        : Average mflops/s per call per node       : 664371
-Grid       : Message        : Average mflops/s per call per node (full): 314042
-Grid       : Message        : Average mflops/s per call per node (full): 437082
-Grid       : Message        : Average mflops/s per call per node (full): 443449
-Grid       : Message        : Average mflops/s per call per node (full): 303713
-Grid       : Message        :  Stencil 12.8417 GB/s per node
-Grid       : Message        :  Stencil 16.6673 GB/s per node
-Grid       : Message        :  Stencil 19.5679 GB/s per node
-Grid       : Message        :  Stencil 12.7913 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669236
-Grid       : Message        : Average mflops/s per call per node       : 803589
-Grid       : Message        : Average mflops/s per call per node       : 821074
-Grid       : Message        : Average mflops/s per call per node       : 663866
-Grid       : Message        : Average mflops/s per call per node (full): 315084
-Grid       : Message        : Average mflops/s per call per node (full): 436973
-Grid       : Message        : Average mflops/s per call per node (full): 449685
-Grid       : Message        : Average mflops/s per call per node (full): 306224
-Grid       : Message        :  Stencil 12.5789 GB/s per node
-Grid       : Message        :  Stencil 15.6077 GB/s per node
-Grid       : Message        :  Stencil 17.4916 GB/s per node
-Grid       : Message        :  Stencil 13.0994 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666648
-Grid       : Message        : Average mflops/s per call per node       : 802272
-Grid       : Message        : Average mflops/s per call per node       : 822978
-Grid       : Message        : Average mflops/s per call per node       : 661676
-Grid       : Message        : Average mflops/s per call per node (full): 314172
-Grid       : Message        : Average mflops/s per call per node (full): 413146
-Grid       : Message        : Average mflops/s per call per node (full): 444026
-Grid       : Message        : Average mflops/s per call per node (full): 305552
-Grid       : Message        :  Stencil 12.7795 GB/s per node
-Grid       : Message        :  Stencil 16.6517 GB/s per node
-Grid       : Message        :  Stencil 17.9527 GB/s per node
-Grid       : Message        :  Stencil 12.8314 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671120
-Grid       : Message        : Average mflops/s per call per node       : 805540
-Grid       : Message        : Average mflops/s per call per node       : 826565
-Grid       : Message        : Average mflops/s per call per node       : 664837
-Grid       : Message        : Average mflops/s per call per node (full): 315358
-Grid       : Message        : Average mflops/s per call per node (full): 435827
-Grid       : Message        : Average mflops/s per call per node (full): 447974
-Grid       : Message        : Average mflops/s per call per node (full): 304610
-Grid       : Message        :  Stencil 13.9753 GB/s per node
-Grid       : Message        :  Stencil 12.3165 GB/s per node
-Grid       : Message        :  Stencil 18.4297 GB/s per node
-Grid       : Message        :  Stencil 12.4519 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662207
-Grid       : Message        : Average mflops/s per call per node       : 810547
-Grid       : Message        : Average mflops/s per call per node       : 820474
-Grid       : Message        : Average mflops/s per call per node       : 666930
-Grid       : Message        : Average mflops/s per call per node (full): 315541
-Grid       : Message        : Average mflops/s per call per node (full): 367835
-Grid       : Message        : Average mflops/s per call per node (full): 447424
-Grid       : Message        : Average mflops/s per call per node (full): 304845
-Grid       : Message        :  Stencil 12.2834 GB/s per node
-Grid       : Message        :  Stencil 17.2565 GB/s per node
-Grid       : Message        :  Stencil 18.1191 GB/s per node
-Grid       : Message        :  Stencil 11.8336 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670755
-Grid       : Message        : Average mflops/s per call per node       : 802889
-Grid       : Message        : Average mflops/s per call per node       : 820477
-Grid       : Message        : Average mflops/s per call per node       : 667293
-Grid       : Message        : Average mflops/s per call per node (full): 312060
-Grid       : Message        : Average mflops/s per call per node (full): 436713
-Grid       : Message        : Average mflops/s per call per node (full): 445364
-Grid       : Message        : Average mflops/s per call per node (full): 300346
-Grid       : Message        :  Stencil 12.2895 GB/s per node
-Grid       : Message        :  Stencil 17.4713 GB/s per node
-Grid       : Message        :  Stencil 17.4528 GB/s per node
-Grid       : Message        :  Stencil 12.3124 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668955
-Grid       : Message        : Average mflops/s per call per node       : 808029
-Grid       : Message        : Average mflops/s per call per node       : 820266
-Grid       : Message        : Average mflops/s per call per node       : 666742
-Grid       : Message        : Average mflops/s per call per node (full): 310810
-Grid       : Message        : Average mflops/s per call per node (full): 439407
-Grid       : Message        : Average mflops/s per call per node (full): 444057
-Grid       : Message        : Average mflops/s per call per node (full): 303354
-Grid       : Message        :  Stencil 13.0029 GB/s per node
-Grid       : Message        :  Stencil 16.9907 GB/s per node
-Grid       : Message        :  Stencil 18.4251 GB/s per node
-Grid       : Message        :  Stencil 14.6304 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667661
-Grid       : Message        : Average mflops/s per call per node       : 804566
-Grid       : Message        : Average mflops/s per call per node       : 822622
-Grid       : Message        : Average mflops/s per call per node       : 659795
-Grid       : Message        : Average mflops/s per call per node (full): 315440
-Grid       : Message        : Average mflops/s per call per node (full): 437435
-Grid       : Message        : Average mflops/s per call per node (full): 448746
-Grid       : Message        : Average mflops/s per call per node (full): 305813
-Grid       : Message        :  Stencil 12.4992 GB/s per node
-Grid       : Message        :  Stencil 16.8081 GB/s per node
-Grid       : Message        :  Stencil 16.9932 GB/s per node
-Grid       : Message        :  Stencil 11.8852 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668265
-Grid       : Message        : Average mflops/s per call per node       : 805726
-Grid       : Message        : Average mflops/s per call per node       : 825066
-Grid       : Message        : Average mflops/s per call per node       : 667803
-Grid       : Message        : Average mflops/s per call per node (full): 311253
-Grid       : Message        : Average mflops/s per call per node (full): 437038
-Grid       : Message        : Average mflops/s per call per node (full): 440641
-Grid       : Message        : Average mflops/s per call per node (full): 300243
-Grid       : Message        :  Stencil 12.858 GB/s per node
-Grid       : Message        :  Stencil 16.6191 GB/s per node
-Grid       : Message        :  Stencil 17.4867 GB/s per node
-Grid       : Message        :  Stencil 12.0766 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664701
-Grid       : Message        : Average mflops/s per call per node       : 802362
-Grid       : Message        : Average mflops/s per call per node       : 825296
-Grid       : Message        : Average mflops/s per call per node       : 671195
-Grid       : Message        : Average mflops/s per call per node (full): 313291
-Grid       : Message        : Average mflops/s per call per node (full): 435415
-Grid       : Message        : Average mflops/s per call per node (full): 444173
-Grid       : Message        : Average mflops/s per call per node (full): 302506
-Grid       : Message        :  Stencil 12.819 GB/s per node
-Grid       : Message        :  Stencil 16.7382 GB/s per node
-Grid       : Message        :  Stencil 17.8396 GB/s per node
-Grid       : Message        :  Stencil 12.7021 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661640
-Grid       : Message        : Average mflops/s per call per node       : 800746
-Grid       : Message        : Average mflops/s per call per node       : 825440
-Grid       : Message        : Average mflops/s per call per node       : 664101
-Grid       : Message        : Average mflops/s per call per node (full): 312336
-Grid       : Message        : Average mflops/s per call per node (full): 435376
-Grid       : Message        : Average mflops/s per call per node (full): 447927
-Grid       : Message        : Average mflops/s per call per node (full): 303009
-Grid       : Message        :  Stencil 13.2615 GB/s per node
-Grid       : Message        :  Stencil 16.7029 GB/s per node
-Grid       : Message        :  Stencil 17.164 GB/s per node
-Grid       : Message        :  Stencil 15.1159 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663828
-Grid       : Message        : Average mflops/s per call per node       : 803900
-Grid       : Message        : Average mflops/s per call per node       : 823424
-Grid       : Message        : Average mflops/s per call per node       : 661500
-Grid       : Message        : Average mflops/s per call per node (full): 315033
-Grid       : Message        : Average mflops/s per call per node (full): 434653
-Grid       : Message        : Average mflops/s per call per node (full): 427996
-Grid       : Message        : Average mflops/s per call per node (full): 308046
-Grid       : Message        :  Stencil 12.6615 GB/s per node
-Grid       : Message        :  Stencil 16.4168 GB/s per node
-Grid       : Message        :  Stencil 17.8024 GB/s per node
-Grid       : Message        :  Stencil 11.8658 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667893
-Grid       : Message        : Average mflops/s per call per node       : 805218
-Grid       : Message        : Average mflops/s per call per node       : 820360
-Grid       : Message        : Average mflops/s per call per node       : 660027
-Grid       : Message        : Average mflops/s per call per node (full): 314693
-Grid       : Message        : Average mflops/s per call per node (full): 433820
-Grid       : Message        : Average mflops/s per call per node (full): 444477
-Grid       : Message        : Average mflops/s per call per node (full): 299996
-Grid       : Message        :  Stencil 12.4537 GB/s per node
-Grid       : Message        :  Stencil 16.6316 GB/s per node
-Grid       : Message        :  Stencil 17.255 GB/s per node
-Grid       : Message        :  Stencil 13.0949 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667214
-Grid       : Message        : Average mflops/s per call per node       : 807821
-Grid       : Message        : Average mflops/s per call per node       : 821258
-Grid       : Message        : Average mflops/s per call per node       : 665148
-Grid       : Message        : Average mflops/s per call per node (full): 313047
-Grid       : Message        : Average mflops/s per call per node (full): 435113
-Grid       : Message        : Average mflops/s per call per node (full): 441888
-Grid       : Message        : Average mflops/s per call per node (full): 305770
-Grid       : Message        :  Stencil 13.026 GB/s per node
-Grid       : Message        :  Stencil 16.8508 GB/s per node
-Grid       : Message        :  Stencil 18.4139 GB/s per node
-Grid       : Message        :  Stencil 16.757 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664610
-Grid       : Message        : Average mflops/s per call per node       : 796445
-Grid       : Message        : Average mflops/s per call per node       : 819592
-Grid       : Message        : Average mflops/s per call per node       : 657022
-Grid       : Message        : Average mflops/s per call per node (full): 313254
-Grid       : Message        : Average mflops/s per call per node (full): 435785
-Grid       : Message        : Average mflops/s per call per node (full): 447783
-Grid       : Message        : Average mflops/s per call per node (full): 307976
-Grid       : Message        :  Stencil 13.6629 GB/s per node
-Grid       : Message        :  Stencil 14.7722 GB/s per node
-Grid       : Message        :  Stencil 17.0123 GB/s per node
-Grid       : Message        :  Stencil 14.6703 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666158
-Grid       : Message        : Average mflops/s per call per node       : 808269
-Grid       : Message        : Average mflops/s per call per node       : 823241
-Grid       : Message        : Average mflops/s per call per node       : 659966
-Grid       : Message        : Average mflops/s per call per node (full): 315972
-Grid       : Message        : Average mflops/s per call per node (full): 408625
-Grid       : Message        : Average mflops/s per call per node (full): 441349
-Grid       : Message        : Average mflops/s per call per node (full): 307814
-Grid       : Message        :  Stencil 13.8222 GB/s per node
-Grid       : Message        :  Stencil 16.5421 GB/s per node
-Grid       : Message        :  Stencil 17.9573 GB/s per node
-Grid       : Message        :  Stencil 13.5068 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662776
-Grid       : Message        : Average mflops/s per call per node       : 802749
-Grid       : Message        : Average mflops/s per call per node       : 819045
-Grid       : Message        : Average mflops/s per call per node       : 667423
-Grid       : Message        : Average mflops/s per call per node (full): 315685
-Grid       : Message        : Average mflops/s per call per node (full): 434846
-Grid       : Message        : Average mflops/s per call per node (full): 445561
-Grid       : Message        : Average mflops/s per call per node (full): 306973
-Grid       : Message        :  Stencil 12.6947 GB/s per node
-Grid       : Message        :  Stencil 16.6596 GB/s per node
-Grid       : Message        :  Stencil 17.3481 GB/s per node
-Grid       : Message        :  Stencil 12.2762 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664797
-Grid       : Message        : Average mflops/s per call per node       : 803254
-Grid       : Message        : Average mflops/s per call per node       : 817960
-Grid       : Message        : Average mflops/s per call per node       : 663602
-Grid       : Message        : Average mflops/s per call per node (full): 313921
-Grid       : Message        : Average mflops/s per call per node (full): 434428
-Grid       : Message        : Average mflops/s per call per node (full): 443691
-Grid       : Message        : Average mflops/s per call per node (full): 303453
-Grid       : Message        :  Stencil 13.0304 GB/s per node
-Grid       : Message        :  Stencil 16.5727 GB/s per node
-Grid       : Message        :  Stencil 17.5342 GB/s per node
-Grid       : Message        :  Stencil 13.6637 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661446
-Grid       : Message        : Average mflops/s per call per node       : 800983
-Grid       : Message        : Average mflops/s per call per node       : 825268
-Grid       : Message        : Average mflops/s per call per node       : 662414
-Grid       : Message        : Average mflops/s per call per node (full): 312931
-Grid       : Message        : Average mflops/s per call per node (full): 434956
-Grid       : Message        : Average mflops/s per call per node (full): 446353
-Grid       : Message        : Average mflops/s per call per node (full): 306809
-Grid       : Message        :  Stencil 13.4499 GB/s per node
-Grid       : Message        :  Stencil 17.8274 GB/s per node
-Grid       : Message        :  Stencil 17.3187 GB/s per node
-Grid       : Message        :  Stencil 12.2928 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663668
-Grid       : Message        : Average mflops/s per call per node       : 802349
-Grid       : Message        : Average mflops/s per call per node       : 824225
-Grid       : Message        : Average mflops/s per call per node       : 664527
-Grid       : Message        : Average mflops/s per call per node (full): 313580
-Grid       : Message        : Average mflops/s per call per node (full): 437042
-Grid       : Message        : Average mflops/s per call per node (full): 444399
-Grid       : Message        : Average mflops/s per call per node (full): 303315
-Grid       : Message        :  Stencil 13.4204 GB/s per node
-Grid       : Message        :  Stencil 14.6526 GB/s per node
-Grid       : Message        :  Stencil 17.0765 GB/s per node
-Grid       : Message        :  Stencil 13.3068 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665269
-Grid       : Message        : Average mflops/s per call per node       : 801119
-Grid       : Message        : Average mflops/s per call per node       : 821479
-Grid       : Message        : Average mflops/s per call per node       : 660903
-Grid       : Message        : Average mflops/s per call per node (full): 315610
-Grid       : Message        : Average mflops/s per call per node (full): 410340
-Grid       : Message        : Average mflops/s per call per node (full): 440724
-Grid       : Message        : Average mflops/s per call per node (full): 304392
-Grid       : Message        :  Stencil 12.9226 GB/s per node
-Grid       : Message        :  Stencil 16.9886 GB/s per node
-Grid       : Message        :  Stencil 18.0785 GB/s per node
-Grid       : Message        :  Stencil 12.6001 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667969
-Grid       : Message        : Average mflops/s per call per node       : 805671
-Grid       : Message        : Average mflops/s per call per node       : 825970
-Grid       : Message        : Average mflops/s per call per node       : 668389
-Grid       : Message        : Average mflops/s per call per node (full): 315630
-Grid       : Message        : Average mflops/s per call per node (full): 437228
-Grid       : Message        : Average mflops/s per call per node (full): 447175
-Grid       : Message        : Average mflops/s per call per node (full): 305637
-Grid       : Message        :  Stencil 13.0071 GB/s per node
-Grid       : Message        :  Stencil 16.4769 GB/s per node
-Grid       : Message        :  Stencil 17.5628 GB/s per node
-Grid       : Message        :  Stencil 12.6108 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664811
-Grid       : Message        : Average mflops/s per call per node       : 807014
-Grid       : Message        : Average mflops/s per call per node       : 825491
-Grid       : Message        : Average mflops/s per call per node       : 664695
-Grid       : Message        : Average mflops/s per call per node (full): 314531
-Grid       : Message        : Average mflops/s per call per node (full): 434962
-Grid       : Message        : Average mflops/s per call per node (full): 445453
-Grid       : Message        : Average mflops/s per call per node (full): 305524
-Grid       : Message        :  Stencil 13.0581 GB/s per node
-Grid       : Message        :  Stencil 16.6896 GB/s per node
-Grid       : Message        :  Stencil 17.6596 GB/s per node
-Grid       : Message        :  Stencil 12.3528 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666024
-Grid       : Message        : Average mflops/s per call per node       : 803409
-Grid       : Message        : Average mflops/s per call per node       : 820641
-Grid       : Message        : Average mflops/s per call per node       : 668749
-Grid       : Message        : Average mflops/s per call per node (full): 314320
-Grid       : Message        : Average mflops/s per call per node (full): 435101
-Grid       : Message        : Average mflops/s per call per node (full): 444807
-Grid       : Message        : Average mflops/s per call per node (full): 305022
-Grid       : Message        :  Stencil 13.858 GB/s per node
-Grid       : Message        :  Stencil 16.4654 GB/s per node
-Grid       : Message        :  Stencil 16.3535 GB/s per node
-Grid       : Message        :  Stencil 12.7462 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663278
-Grid       : Message        : Average mflops/s per call per node       : 799008
-Grid       : Message        : Average mflops/s per call per node       : 828356
-Grid       : Message        : Average mflops/s per call per node       : 665600
-Grid       : Message        : Average mflops/s per call per node (full): 315862
-Grid       : Message        : Average mflops/s per call per node (full): 424663
-Grid       : Message        : Average mflops/s per call per node (full): 429391
-Grid       : Message        : Average mflops/s per call per node (full): 305352
-Grid       : Message        :  Stencil 13.395 GB/s per node
-Grid       : Message        :  Stencil 16.9958 GB/s per node
-Grid       : Message        :  Stencil 17.0761 GB/s per node
-Grid       : Message        :  Stencil 11.9704 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663609
-Grid       : Message        : Average mflops/s per call per node       : 798959
-Grid       : Message        : Average mflops/s per call per node       : 825384
-Grid       : Message        : Average mflops/s per call per node       : 668680
-Grid       : Message        : Average mflops/s per call per node (full): 315709
-Grid       : Message        : Average mflops/s per call per node (full): 434231
-Grid       : Message        : Average mflops/s per call per node (full): 442366
-Grid       : Message        : Average mflops/s per call per node (full): 301876
-Grid       : Message        :  Stencil 14.9144 GB/s per node
-Grid       : Message        :  Stencil 17.035 GB/s per node
-Grid       : Message        :  Stencil 16.9675 GB/s per node
-Grid       : Message        :  Stencil 12.5384 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660331
-Grid       : Message        : Average mflops/s per call per node       : 799565
-Grid       : Message        : Average mflops/s per call per node       : 820775
-Grid       : Message        : Average mflops/s per call per node       : 664950
-Grid       : Message        : Average mflops/s per call per node (full): 316087
-Grid       : Message        : Average mflops/s per call per node (full): 435532
-Grid       : Message        : Average mflops/s per call per node (full): 439632
-Grid       : Message        : Average mflops/s per call per node (full): 304705
-Grid       : Message        :  Stencil 13.2079 GB/s per node
-Grid       : Message        :  Stencil 9.38049 GB/s per node
-Grid       : Message        :  Stencil 17.5985 GB/s per node
-Grid       : Message        :  Stencil 12.9394 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666286
-Grid       : Message        : Average mflops/s per call per node       : 807472
-Grid       : Message        : Average mflops/s per call per node       : 822532
-Grid       : Message        : Average mflops/s per call per node       : 666927
-Grid       : Message        : Average mflops/s per call per node (full): 315426
-Grid       : Message        : Average mflops/s per call per node (full): 301528
-Grid       : Message        : Average mflops/s per call per node (full): 444270
-Grid       : Message        : Average mflops/s per call per node (full): 305483
-Grid       : Message        :  Stencil 12.9366 GB/s per node
-Grid       : Message        :  Stencil 8.86962 GB/s per node
-Grid       : Message        :  Stencil 18.1196 GB/s per node
-Grid       : Message        :  Stencil 14.2661 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664986
-Grid       : Message        : Average mflops/s per call per node       : 810537
-Grid       : Message        : Average mflops/s per call per node       : 825000
-Grid       : Message        : Average mflops/s per call per node       : 662504
-Grid       : Message        : Average mflops/s per call per node (full): 311648
-Grid       : Message        : Average mflops/s per call per node (full): 288921
-Grid       : Message        : Average mflops/s per call per node (full): 448842
-Grid       : Message        : Average mflops/s per call per node (full): 307355
-Grid       : Message        :  Stencil 12.8223 GB/s per node
-Grid       : Message        :  Stencil 16.8345 GB/s per node
-Grid       : Message        :  Stencil 17.715 GB/s per node
-Grid       : Message        :  Stencil 13.3817 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666438
-Grid       : Message        : Average mflops/s per call per node       : 806506
-Grid       : Message        : Average mflops/s per call per node       : 823873
-Grid       : Message        : Average mflops/s per call per node       : 664154
-Grid       : Message        : Average mflops/s per call per node (full): 314939
-Grid       : Message        : Average mflops/s per call per node (full): 438382
-Grid       : Message        : Average mflops/s per call per node (full): 445696
-Grid       : Message        : Average mflops/s per call per node (full): 304392
-Grid       : Message        :  Stencil 13.5753 GB/s per node
-Grid       : Message        :  Stencil 16.8199 GB/s per node
-Grid       : Message        :  Stencil 17.6639 GB/s per node
-Grid       : Message        :  Stencil 12.508 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661288
-Grid       : Message        : Average mflops/s per call per node       : 801734
-Grid       : Message        : Average mflops/s per call per node       : 826895
-Grid       : Message        : Average mflops/s per call per node       : 666866
-Grid       : Message        : Average mflops/s per call per node (full): 315113
-Grid       : Message        : Average mflops/s per call per node (full): 436627
-Grid       : Message        : Average mflops/s per call per node (full): 445885
-Grid       : Message        : Average mflops/s per call per node (full): 304389
-Grid       : Message        :  Stencil 13.2609 GB/s per node
-Grid       : Message        :  Stencil 16.5876 GB/s per node
-Grid       : Message        :  Stencil 17.518 GB/s per node
-Grid       : Message        :  Stencil 12.622 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666661
-Grid       : Message        : Average mflops/s per call per node       : 806564
-Grid       : Message        : Average mflops/s per call per node       : 829801
-Grid       : Message        : Average mflops/s per call per node       : 666031
-Grid       : Message        : Average mflops/s per call per node (full): 315103
-Grid       : Message        : Average mflops/s per call per node (full): 434902
-Grid       : Message        : Average mflops/s per call per node (full): 446852
-Grid       : Message        : Average mflops/s per call per node (full): 303927
-Grid       : Message        :  Stencil 12.7534 GB/s per node
-Grid       : Message        :  Stencil 18.1301 GB/s per node
-Grid       : Message        :  Stencil 17.676 GB/s per node
-Grid       : Message        :  Stencil 12.6733 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666771
-Grid       : Message        : Average mflops/s per call per node       : 795816
-Grid       : Message        : Average mflops/s per call per node       : 819261
-Grid       : Message        : Average mflops/s per call per node       : 661509
-Grid       : Message        : Average mflops/s per call per node (full): 313882
-Grid       : Message        : Average mflops/s per call per node (full): 436371
-Grid       : Message        : Average mflops/s per call per node (full): 445274
-Grid       : Message        : Average mflops/s per call per node (full): 304463
-Grid       : Message        :  Stencil 13.021 GB/s per node
-Grid       : Message        :  Stencil 17.343 GB/s per node
-Grid       : Message        :  Stencil 17.7043 GB/s per node
-Grid       : Message        :  Stencil 12.9421 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667115
-Grid       : Message        : Average mflops/s per call per node       : 801682
-Grid       : Message        : Average mflops/s per call per node       : 821272
-Grid       : Message        : Average mflops/s per call per node       : 667456
-Grid       : Message        : Average mflops/s per call per node (full): 314729
-Grid       : Message        : Average mflops/s per call per node (full): 437239
-Grid       : Message        : Average mflops/s per call per node (full): 440060
-Grid       : Message        : Average mflops/s per call per node (full): 305517
-Grid       : Message        :  Stencil 14.6619 GB/s per node
-Grid       : Message        :  Stencil 14.2718 GB/s per node
-Grid       : Message        :  Stencil 17.7885 GB/s per node
-Grid       : Message        :  Stencil 12.7771 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663197
-Grid       : Message        : Average mflops/s per call per node       : 804402
-Grid       : Message        : Average mflops/s per call per node       : 818351
-Grid       : Message        : Average mflops/s per call per node       : 659551
-Grid       : Message        : Average mflops/s per call per node (full): 316195
-Grid       : Message        : Average mflops/s per call per node (full): 404190
-Grid       : Message        : Average mflops/s per call per node (full): 445743
-Grid       : Message        : Average mflops/s per call per node (full): 303828
-Grid       : Message        :  Stencil 14.1316 GB/s per node
-Grid       : Message        :  Stencil 16.6678 GB/s per node
-Grid       : Message        :  Stencil 18.9153 GB/s per node
-Grid       : Message        :  Stencil 12.2411 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664796
-Grid       : Message        : Average mflops/s per call per node       : 801994
-Grid       : Message        : Average mflops/s per call per node       : 817793
-Grid       : Message        : Average mflops/s per call per node       : 671268
-Grid       : Message        : Average mflops/s per call per node (full): 316474
-Grid       : Message        : Average mflops/s per call per node (full): 434130
-Grid       : Message        : Average mflops/s per call per node (full): 446851
-Grid       : Message        : Average mflops/s per call per node (full): 304576
-Grid       : Message        :  Stencil 15.3967 GB/s per node
-Grid       : Message        :  Stencil 17.3259 GB/s per node
-Grid       : Message        :  Stencil 17.8009 GB/s per node
-Grid       : Message        :  Stencil 12.0176 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662867
-Grid       : Message        : Average mflops/s per call per node       : 802190
-Grid       : Message        : Average mflops/s per call per node       : 818363
-Grid       : Message        : Average mflops/s per call per node       : 672577
-Grid       : Message        : Average mflops/s per call per node (full): 317209
-Grid       : Message        : Average mflops/s per call per node (full): 437671
-Grid       : Message        : Average mflops/s per call per node (full): 443132
-Grid       : Message        : Average mflops/s per call per node (full): 302658
-Grid       : Message        :  Stencil 13.7304 GB/s per node
-Grid       : Message        :  Stencil 16.5138 GB/s per node
-Grid       : Message        :  Stencil 17.2066 GB/s per node
-Grid       : Message        :  Stencil 12.6344 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667165
-Grid       : Message        : Average mflops/s per call per node       : 801862
-Grid       : Message        : Average mflops/s per call per node       : 821386
-Grid       : Message        : Average mflops/s per call per node       : 668298
-Grid       : Message        : Average mflops/s per call per node (full): 313850
-Grid       : Message        : Average mflops/s per call per node (full): 434568
-Grid       : Message        : Average mflops/s per call per node (full): 441871
-Grid       : Message        : Average mflops/s per call per node (full): 305778
-Grid       : Message        :  Stencil 14.1902 GB/s per node
-Grid       : Message        :  Stencil 12.5243 GB/s per node
-Grid       : Message        :  Stencil 17.6716 GB/s per node
-Grid       : Message        :  Stencil 12.2327 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667550
-Grid       : Message        : Average mflops/s per call per node       : 808486
-Grid       : Message        : Average mflops/s per call per node       : 821750
-Grid       : Message        : Average mflops/s per call per node       : 664300
-Grid       : Message        : Average mflops/s per call per node (full): 317058
-Grid       : Message        : Average mflops/s per call per node (full): 371507
-Grid       : Message        : Average mflops/s per call per node (full): 445703
-Grid       : Message        : Average mflops/s per call per node (full): 302001
-Grid       : Message        :  Stencil 14.7126 GB/s per node
-Grid       : Message        :  Stencil 17.3296 GB/s per node
-Grid       : Message        :  Stencil 16.9344 GB/s per node
-Grid       : Message        :  Stencil 13.7377 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663247
-Grid       : Message        : Average mflops/s per call per node       : 801530
-Grid       : Message        : Average mflops/s per call per node       : 823572
-Grid       : Message        : Average mflops/s per call per node       : 660202
-Grid       : Message        : Average mflops/s per call per node (full): 317715
-Grid       : Message        : Average mflops/s per call per node (full): 438797
-Grid       : Message        : Average mflops/s per call per node (full): 440266
-Grid       : Message        : Average mflops/s per call per node (full): 305603
-Grid       : Message        :  Stencil 14.2874 GB/s per node
-Grid       : Message        :  Stencil 16.3823 GB/s per node
-Grid       : Message        :  Stencil 18.2821 GB/s per node
-Grid       : Message        :  Stencil 12.2695 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664440
-Grid       : Message        : Average mflops/s per call per node       : 802782
-Grid       : Message        : Average mflops/s per call per node       : 817892
-Grid       : Message        : Average mflops/s per call per node       : 664293
-Grid       : Message        : Average mflops/s per call per node (full): 314310
-Grid       : Message        : Average mflops/s per call per node (full): 431204
-Grid       : Message        : Average mflops/s per call per node (full): 445353
-Grid       : Message        : Average mflops/s per call per node (full): 304266
-Grid       : Message        :  Stencil 12.2287 GB/s per node
-Grid       : Message        :  Stencil 9.84358 GB/s per node
-Grid       : Message        :  Stencil 18.2287 GB/s per node
-Grid       : Message        :  Stencil 13.4763 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669877
-Grid       : Message        : Average mflops/s per call per node       : 812425
-Grid       : Message        : Average mflops/s per call per node       : 815365
-Grid       : Message        : Average mflops/s per call per node       : 662986
-Grid       : Message        : Average mflops/s per call per node (full): 310980
-Grid       : Message        : Average mflops/s per call per node (full): 312378
-Grid       : Message        : Average mflops/s per call per node (full): 445350
-Grid       : Message        : Average mflops/s per call per node (full): 304896
-Grid       : Message        :  Stencil 12.4854 GB/s per node
-Grid       : Message        :  Stencil 16.9744 GB/s per node
-Grid       : Message        :  Stencil 17.2508 GB/s per node
-Grid       : Message        :  Stencil 14.1054 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666189
-Grid       : Message        : Average mflops/s per call per node       : 800186
-Grid       : Message        : Average mflops/s per call per node       : 824927
-Grid       : Message        : Average mflops/s per call per node       : 660589
-Grid       : Message        : Average mflops/s per call per node (full): 313277
-Grid       : Message        : Average mflops/s per call per node (full): 437682
-Grid       : Message        : Average mflops/s per call per node (full): 443906
-Grid       : Message        : Average mflops/s per call per node (full): 304831
-Grid       : Message        :  Stencil 13.8368 GB/s per node
-Grid       : Message        :  Stencil 16.3028 GB/s per node
-Grid       : Message        :  Stencil 17.5656 GB/s per node
-Grid       : Message        :  Stencil 12.244 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664243
-Grid       : Message        : Average mflops/s per call per node       : 800247
-Grid       : Message        : Average mflops/s per call per node       : 825207
-Grid       : Message        : Average mflops/s per call per node       : 669564
-Grid       : Message        : Average mflops/s per call per node (full): 316090
-Grid       : Message        : Average mflops/s per call per node (full): 431846
-Grid       : Message        : Average mflops/s per call per node (full): 445461
-Grid       : Message        : Average mflops/s per call per node (full): 303726
-Grid       : Message        :  Stencil 13.2986 GB/s per node
-Grid       : Message        :  Stencil 16.8133 GB/s per node
-Grid       : Message        :  Stencil 17.0501 GB/s per node
-Grid       : Message        :  Stencil 12.7636 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663854
-Grid       : Message        : Average mflops/s per call per node       : 802586
-Grid       : Message        : Average mflops/s per call per node       : 815724
-Grid       : Message        : Average mflops/s per call per node       : 670749
-Grid       : Message        : Average mflops/s per call per node (full): 315614
-Grid       : Message        : Average mflops/s per call per node (full): 437869
-Grid       : Message        : Average mflops/s per call per node (full): 440744
-Grid       : Message        : Average mflops/s per call per node (full): 307138
-Grid       : Message        :  Stencil 12.9122 GB/s per node
-Grid       : Message        :  Stencil 15.6169 GB/s per node
-Grid       : Message        :  Stencil 17.9642 GB/s per node
-Grid       : Message        :  Stencil 13.2988 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666449
-Grid       : Message        : Average mflops/s per call per node       : 810116
-Grid       : Message        : Average mflops/s per call per node       : 822794
-Grid       : Message        : Average mflops/s per call per node       : 666926
-Grid       : Message        : Average mflops/s per call per node (full): 313636
-Grid       : Message        : Average mflops/s per call per node (full): 425292
-Grid       : Message        : Average mflops/s per call per node (full): 447099
-Grid       : Message        : Average mflops/s per call per node (full): 306732
-Grid       : Message        :  Stencil 13.0659 GB/s per node
-Grid       : Message        :  Stencil 17.3714 GB/s per node
-Grid       : Message        :  Stencil 18.1176 GB/s per node
-Grid       : Message        :  Stencil 12.6298 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665254
-Grid       : Message        : Average mflops/s per call per node       : 800461
-Grid       : Message        : Average mflops/s per call per node       : 825509
-Grid       : Message        : Average mflops/s per call per node       : 664509
-Grid       : Message        : Average mflops/s per call per node (full): 314606
-Grid       : Message        : Average mflops/s per call per node (full): 437387
-Grid       : Message        : Average mflops/s per call per node (full): 442048
-Grid       : Message        : Average mflops/s per call per node (full): 304503
-Grid       : Message        :  Stencil 13.1864 GB/s per node
-Grid       : Message        :  Stencil 16.961 GB/s per node
-Grid       : Message        :  Stencil 18.1721 GB/s per node
-Grid       : Message        :  Stencil 13.3021 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668091
-Grid       : Message        : Average mflops/s per call per node       : 797834
-Grid       : Message        : Average mflops/s per call per node       : 822533
-Grid       : Message        : Average mflops/s per call per node       : 665343
-Grid       : Message        : Average mflops/s per call per node (full): 315989
-Grid       : Message        : Average mflops/s per call per node (full): 437285
-Grid       : Message        : Average mflops/s per call per node (full): 448604
-Grid       : Message        : Average mflops/s per call per node (full): 305937
-Grid       : Message        :  Stencil 12.7171 GB/s per node
-Grid       : Message        :  Stencil 17.0547 GB/s per node
-Grid       : Message        :  Stencil 17.2289 GB/s per node
-Grid       : Message        :  Stencil 13.0249 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669958
-Grid       : Message        : Average mflops/s per call per node       : 802599
-Grid       : Message        : Average mflops/s per call per node       : 823252
-Grid       : Message        : Average mflops/s per call per node       : 663794
-Grid       : Message        : Average mflops/s per call per node (full): 315310
-Grid       : Message        : Average mflops/s per call per node (full): 434842
-Grid       : Message        : Average mflops/s per call per node (full): 443271
-Grid       : Message        : Average mflops/s per call per node (full): 305779
-Grid       : Message        :  Stencil 12.7706 GB/s per node
-Grid       : Message        :  Stencil 17.9404 GB/s per node
-Grid       : Message        :  Stencil 17.5594 GB/s per node
-Grid       : Message        :  Stencil 11.8661 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670656
-Grid       : Message        : Average mflops/s per call per node       : 798841
-Grid       : Message        : Average mflops/s per call per node       : 816038
-Grid       : Message        : Average mflops/s per call per node       : 667840
-Grid       : Message        : Average mflops/s per call per node (full): 315570
-Grid       : Message        : Average mflops/s per call per node (full): 438820
-Grid       : Message        : Average mflops/s per call per node (full): 444007
-Grid       : Message        : Average mflops/s per call per node (full): 301392
-Grid       : Message        :  Stencil 12.6387 GB/s per node
-Grid       : Message        :  Stencil 17.87 GB/s per node
-Grid       : Message        :  Stencil 17.3612 GB/s per node
-Grid       : Message        :  Stencil 12.1397 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670097
-Grid       : Message        : Average mflops/s per call per node       : 801686
-Grid       : Message        : Average mflops/s per call per node       : 822358
-Grid       : Message        : Average mflops/s per call per node       : 673319
-Grid       : Message        : Average mflops/s per call per node (full): 314702
-Grid       : Message        : Average mflops/s per call per node (full): 433081
-Grid       : Message        : Average mflops/s per call per node (full): 443404
-Grid       : Message        : Average mflops/s per call per node (full): 303616
-Grid       : Message        :  Stencil 12.6213 GB/s per node
-Grid       : Message        :  Stencil 17.3179 GB/s per node
-Grid       : Message        :  Stencil 17.3871 GB/s per node
-Grid       : Message        :  Stencil 14.4538 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669104
-Grid       : Message        : Average mflops/s per call per node       : 808904
-Grid       : Message        : Average mflops/s per call per node       : 816456
-Grid       : Message        : Average mflops/s per call per node       : 664116
-Grid       : Message        : Average mflops/s per call per node (full): 314805
-Grid       : Message        : Average mflops/s per call per node (full): 438129
-Grid       : Message        : Average mflops/s per call per node (full): 443576
-Grid       : Message        : Average mflops/s per call per node (full): 307045
-Grid       : Message        :  Stencil 13.1342 GB/s per node
-Grid       : Message        :  Stencil 17.2667 GB/s per node
-Grid       : Message        :  Stencil 18.9676 GB/s per node
-Grid       : Message        :  Stencil 13.1063 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666448
-Grid       : Message        : Average mflops/s per call per node       : 803193
-Grid       : Message        : Average mflops/s per call per node       : 820740
-Grid       : Message        : Average mflops/s per call per node       : 661598
-Grid       : Message        : Average mflops/s per call per node (full): 315157
-Grid       : Message        : Average mflops/s per call per node (full): 437486
-Grid       : Message        : Average mflops/s per call per node (full): 447975
-Grid       : Message        : Average mflops/s per call per node (full): 305551
-Grid       : Message        :  Stencil 12.8885 GB/s per node
-Grid       : Message        :  Stencil 16.7165 GB/s per node
-Grid       : Message        :  Stencil 17.4135 GB/s per node
-Grid       : Message        :  Stencil 12.4162 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669610
-Grid       : Message        : Average mflops/s per call per node       : 798878
-Grid       : Message        : Average mflops/s per call per node       : 828761
-Grid       : Message        : Average mflops/s per call per node       : 665508
-Grid       : Message        : Average mflops/s per call per node (full): 315714
-Grid       : Message        : Average mflops/s per call per node (full): 433076
-Grid       : Message        : Average mflops/s per call per node (full): 445197
-Grid       : Message        : Average mflops/s per call per node (full): 304990
-Grid       : Message        :  Stencil 13.4829 GB/s per node
-Grid       : Message        :  Stencil 17.3411 GB/s per node
-Grid       : Message        :  Stencil 19.5368 GB/s per node
-Grid       : Message        :  Stencil 12.2651 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663229
-Grid       : Message        : Average mflops/s per call per node       : 801443
-Grid       : Message        : Average mflops/s per call per node       : 822028
-Grid       : Message        : Average mflops/s per call per node       : 666187
-Grid       : Message        : Average mflops/s per call per node (full): 313415
-Grid       : Message        : Average mflops/s per call per node (full): 437696
-Grid       : Message        : Average mflops/s per call per node (full): 451046
-Grid       : Message        : Average mflops/s per call per node (full): 303781
-Grid       : Message        :  Stencil 12.9546 GB/s per node
-Grid       : Message        :  Stencil 16.7871 GB/s per node
-Grid       : Message        :  Stencil 17.4806 GB/s per node
-Grid       : Message        :  Stencil 12.5693 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667984
-Grid       : Message        : Average mflops/s per call per node       : 804416
-Grid       : Message        : Average mflops/s per call per node       : 820681
-Grid       : Message        : Average mflops/s per call per node       : 662491
-Grid       : Message        : Average mflops/s per call per node (full): 315209
-Grid       : Message        : Average mflops/s per call per node (full): 437613
-Grid       : Message        : Average mflops/s per call per node (full): 444061
-Grid       : Message        : Average mflops/s per call per node (full): 304878
-Grid       : Message        :  Stencil 13.2857 GB/s per node
-Grid       : Message        :  Stencil 16.9461 GB/s per node
-Grid       : Message        :  Stencil 17.5121 GB/s per node
-Grid       : Message        :  Stencil 13.1988 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666158
-Grid       : Message        : Average mflops/s per call per node       : 806146
-Grid       : Message        : Average mflops/s per call per node       : 822367
-Grid       : Message        : Average mflops/s per call per node       : 667978
-Grid       : Message        : Average mflops/s per call per node (full): 315279
-Grid       : Message        : Average mflops/s per call per node (full): 437540
-Grid       : Message        : Average mflops/s per call per node (full): 445660
-Grid       : Message        : Average mflops/s per call per node (full): 306304
-Grid       : Message        :  Stencil 13.1804 GB/s per node
-Grid       : Message        :  Stencil 17.6715 GB/s per node
-Grid       : Message        :  Stencil 17.9868 GB/s per node
-Grid       : Message        :  Stencil 13.724 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667593
-Grid       : Message        : Average mflops/s per call per node       : 801475
-Grid       : Message        : Average mflops/s per call per node       : 823711
-Grid       : Message        : Average mflops/s per call per node       : 668033
-Grid       : Message        : Average mflops/s per call per node (full): 315247
-Grid       : Message        : Average mflops/s per call per node (full): 435588
-Grid       : Message        : Average mflops/s per call per node (full): 446829
-Grid       : Message        : Average mflops/s per call per node (full): 307368
-Grid       : Message        :  Stencil 12.851 GB/s per node
-Grid       : Message        :  Stencil 17.4314 GB/s per node
-Grid       : Message        :  Stencil 17.2006 GB/s per node
-Grid       : Message        :  Stencil 13.0596 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667473
-Grid       : Message        : Average mflops/s per call per node       : 802011
-Grid       : Message        : Average mflops/s per call per node       : 827037
-Grid       : Message        : Average mflops/s per call per node       : 663835
-Grid       : Message        : Average mflops/s per call per node (full): 313637
-Grid       : Message        : Average mflops/s per call per node (full): 440010
-Grid       : Message        : Average mflops/s per call per node (full): 443524
-Grid       : Message        : Average mflops/s per call per node (full): 304671
-Grid       : Message        :  Stencil 13.5352 GB/s per node
-Grid       : Message        :  Stencil 13.4914 GB/s per node
-Grid       : Message        :  Stencil 18.1952 GB/s per node
-Grid       : Message        :  Stencil 13.6803 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666383
-Grid       : Message        : Average mflops/s per call per node       : 810543
-Grid       : Message        : Average mflops/s per call per node       : 819120
-Grid       : Message        : Average mflops/s per call per node       : 661499
-Grid       : Message        : Average mflops/s per call per node (full): 315285
-Grid       : Message        : Average mflops/s per call per node (full): 390981
-Grid       : Message        : Average mflops/s per call per node (full): 437005
-Grid       : Message        : Average mflops/s per call per node (full): 305682
-Grid       : Message        :  Stencil 13.9432 GB/s per node
-Grid       : Message        :  Stencil 16.9757 GB/s per node
-Grid       : Message        :  Stencil 18.1128 GB/s per node
-Grid       : Message        :  Stencil 14.5547 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664137
-Grid       : Message        : Average mflops/s per call per node       : 806558
-Grid       : Message        : Average mflops/s per call per node       : 819594
-Grid       : Message        : Average mflops/s per call per node       : 660482
-Grid       : Message        : Average mflops/s per call per node (full): 315526
-Grid       : Message        : Average mflops/s per call per node (full): 437547
-Grid       : Message        : Average mflops/s per call per node (full): 445861
-Grid       : Message        : Average mflops/s per call per node (full): 306961
-Grid       : Message        :  Stencil 13.4324 GB/s per node
-Grid       : Message        :  Stencil 18.4747 GB/s per node
-Grid       : Message        :  Stencil 17.1373 GB/s per node
-Grid       : Message        :  Stencil 13.291 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666766
-Grid       : Message        : Average mflops/s per call per node       : 798827
-Grid       : Message        : Average mflops/s per call per node       : 824186
-Grid       : Message        : Average mflops/s per call per node       : 665952
-Grid       : Message        : Average mflops/s per call per node (full): 315836
-Grid       : Message        : Average mflops/s per call per node (full): 440620
-Grid       : Message        : Average mflops/s per call per node (full): 442485
-Grid       : Message        : Average mflops/s per call per node (full): 306693
-Grid       : Message        :  Stencil 14.2648 GB/s per node
-Grid       : Message        :  Stencil 17.4152 GB/s per node
-Grid       : Message        :  Stencil 17.9912 GB/s per node
-Grid       : Message        :  Stencil 12.5767 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665482
-Grid       : Message        : Average mflops/s per call per node       : 806993
-Grid       : Message        : Average mflops/s per call per node       : 816942
-Grid       : Message        : Average mflops/s per call per node       : 661803
-Grid       : Message        : Average mflops/s per call per node (full): 317197
-Grid       : Message        : Average mflops/s per call per node (full): 438764
-Grid       : Message        : Average mflops/s per call per node (full): 445909
-Grid       : Message        : Average mflops/s per call per node (full): 304387
-Grid       : Message        :  Stencil 14.4825 GB/s per node
-Grid       : Message        :  Stencil 16.3628 GB/s per node
-Grid       : Message        :  Stencil 17.3344 GB/s per node
-Grid       : Message        :  Stencil 12.5575 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661238
-Grid       : Message        : Average mflops/s per call per node       : 803942
-Grid       : Message        : Average mflops/s per call per node       : 818555
-Grid       : Message        : Average mflops/s per call per node       : 667193
-Grid       : Message        : Average mflops/s per call per node (full): 316596
-Grid       : Message        : Average mflops/s per call per node (full): 426580
-Grid       : Message        : Average mflops/s per call per node (full): 442295
-Grid       : Message        : Average mflops/s per call per node (full): 305003
-Grid       : Message        :  Stencil 13.2119 GB/s per node
-Grid       : Message        :  Stencil 14.9454 GB/s per node
-Grid       : Message        :  Stencil 18.2205 GB/s per node
-Grid       : Message        :  Stencil 13.2841 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666447
-Grid       : Message        : Average mflops/s per call per node       : 805976
-Grid       : Message        : Average mflops/s per call per node       : 825025
-Grid       : Message        : Average mflops/s per call per node       : 662542
-Grid       : Message        : Average mflops/s per call per node (full): 315221
-Grid       : Message        : Average mflops/s per call per node (full): 414701
-Grid       : Message        : Average mflops/s per call per node (full): 449093
-Grid       : Message        : Average mflops/s per call per node (full): 305750
-Grid       : Message        :  Stencil 13.3864 GB/s per node
-Grid       : Message        :  Stencil 16.3266 GB/s per node
-Grid       : Message        :  Stencil 18.0105 GB/s per node
-Grid       : Message        :  Stencil 13.2748 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666430
-Grid       : Message        : Average mflops/s per call per node       : 801581
-Grid       : Message        : Average mflops/s per call per node       : 822402
-Grid       : Message        : Average mflops/s per call per node       : 664154
-Grid       : Message        : Average mflops/s per call per node (full): 316332
-Grid       : Message        : Average mflops/s per call per node (full): 432812
-Grid       : Message        : Average mflops/s per call per node (full): 447225
-Grid       : Message        : Average mflops/s per call per node (full): 306327
-Grid       : Message        :  Stencil 14.0436 GB/s per node
-Grid       : Message        :  Stencil 17.5313 GB/s per node
-Grid       : Message        :  Stencil 17.9904 GB/s per node
-Grid       : Message        :  Stencil 14.0196 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661958
-Grid       : Message        : Average mflops/s per call per node       : 802925
-Grid       : Message        : Average mflops/s per call per node       : 822580
-Grid       : Message        : Average mflops/s per call per node       : 655474
-Grid       : Message        : Average mflops/s per call per node (full): 315807
-Grid       : Message        : Average mflops/s per call per node (full): 440084
-Grid       : Message        : Average mflops/s per call per node (full): 446506
-Grid       : Message        : Average mflops/s per call per node (full): 305478
-Grid       : Message        :  Stencil 13.8095 GB/s per node
-Grid       : Message        :  Stencil 17.219 GB/s per node
-Grid       : Message        :  Stencil 17.8972 GB/s per node
-Grid       : Message        :  Stencil 12.5523 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662486
-Grid       : Message        : Average mflops/s per call per node       : 805245
-Grid       : Message        : Average mflops/s per call per node       : 822277
-Grid       : Message        : Average mflops/s per call per node       : 667800
-Grid       : Message        : Average mflops/s per call per node (full): 315251
-Grid       : Message        : Average mflops/s per call per node (full): 437134
-Grid       : Message        : Average mflops/s per call per node (full): 445001
-Grid       : Message        : Average mflops/s per call per node (full): 303441
-Grid       : Message        :  Stencil 13.4219 GB/s per node
-Grid       : Message        :  Stencil 16.4628 GB/s per node
-Grid       : Message        :  Stencil 17.7018 GB/s per node
-Grid       : Message        :  Stencil 12.3819 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665299
-Grid       : Message        : Average mflops/s per call per node       : 805631
-Grid       : Message        : Average mflops/s per call per node       : 819042
-Grid       : Message        : Average mflops/s per call per node       : 670935
-Grid       : Message        : Average mflops/s per call per node (full): 315640
-Grid       : Message        : Average mflops/s per call per node (full): 434757
-Grid       : Message        : Average mflops/s per call per node (full): 444690
-Grid       : Message        : Average mflops/s per call per node (full): 305656
-Grid       : Message        :  Stencil 12.4988 GB/s per node
-Grid       : Message        :  Stencil 16.9083 GB/s per node
-Grid       : Message        :  Stencil 17.7146 GB/s per node
-Grid       : Message        :  Stencil 14.3923 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665591
-Grid       : Message        : Average mflops/s per call per node       : 803923
-Grid       : Message        : Average mflops/s per call per node       : 828013
-Grid       : Message        : Average mflops/s per call per node       : 663066
-Grid       : Message        : Average mflops/s per call per node (full): 313228
-Grid       : Message        : Average mflops/s per call per node (full): 435404
-Grid       : Message        : Average mflops/s per call per node (full): 442913
-Grid       : Message        : Average mflops/s per call per node (full): 306970
-Grid       : Message        :  Stencil 13.3269 GB/s per node
-Grid       : Message        :  Stencil 16.7782 GB/s per node
-Grid       : Message        :  Stencil 18.0916 GB/s per node
-Grid       : Message        :  Stencil 13.0793 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666367
-Grid       : Message        : Average mflops/s per call per node       : 806076
-Grid       : Message        : Average mflops/s per call per node       : 819398
-Grid       : Message        : Average mflops/s per call per node       : 663443
-Grid       : Message        : Average mflops/s per call per node (full): 315773
-Grid       : Message        : Average mflops/s per call per node (full): 436873
-Grid       : Message        : Average mflops/s per call per node (full): 445575
-Grid       : Message        : Average mflops/s per call per node (full): 305899
-Grid       : Message        :  Stencil 12.4975 GB/s per node
-Grid       : Message        :  Stencil 16.6148 GB/s per node
-Grid       : Message        :  Stencil 16.7767 GB/s per node
-Grid       : Message        :  Stencil 14.3518 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666796
-Grid       : Message        : Average mflops/s per call per node       : 806592
-Grid       : Message        : Average mflops/s per call per node       : 823710
-Grid       : Message        : Average mflops/s per call per node       : 664034
-Grid       : Message        : Average mflops/s per call per node (full): 312829
-Grid       : Message        : Average mflops/s per call per node (full): 435589
-Grid       : Message        : Average mflops/s per call per node (full): 431787
-Grid       : Message        : Average mflops/s per call per node (full): 307315
-Grid       : Message        :  Stencil 12.7171 GB/s per node
-Grid       : Message        :  Stencil 17.2914 GB/s per node
-Grid       : Message        :  Stencil 17.1318 GB/s per node
-Grid       : Message        :  Stencil 14.4639 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664303
-Grid       : Message        : Average mflops/s per call per node       : 805861
-Grid       : Message        : Average mflops/s per call per node       : 827627
-Grid       : Message        : Average mflops/s per call per node       : 657617
-Grid       : Message        : Average mflops/s per call per node (full): 311895
-Grid       : Message        : Average mflops/s per call per node (full): 440178
-Grid       : Message        : Average mflops/s per call per node (full): 438531
-Grid       : Message        : Average mflops/s per call per node (full): 305532
-Grid       : Message        :  Stencil 13.8382 GB/s per node
-Grid       : Message        :  Stencil 16.6842 GB/s per node
-Grid       : Message        :  Stencil 17.5183 GB/s per node
-Grid       : Message        :  Stencil 12.7917 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663577
-Grid       : Message        : Average mflops/s per call per node       : 805710
-Grid       : Message        : Average mflops/s per call per node       : 820054
-Grid       : Message        : Average mflops/s per call per node       : 661076
-Grid       : Message        : Average mflops/s per call per node (full): 315325
-Grid       : Message        : Average mflops/s per call per node (full): 437400
-Grid       : Message        : Average mflops/s per call per node (full): 443599
-Grid       : Message        : Average mflops/s per call per node (full): 304844
-Grid       : Message        :  Stencil 13.2596 GB/s per node
-Grid       : Message        :  Stencil 16.6761 GB/s per node
-Grid       : Message        :  Stencil 17.0363 GB/s per node
-Grid       : Message        :  Stencil 14.1222 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665035
-Grid       : Message        : Average mflops/s per call per node       : 806591
-Grid       : Message        : Average mflops/s per call per node       : 819211
-Grid       : Message        : Average mflops/s per call per node       : 660628
-Grid       : Message        : Average mflops/s per call per node (full): 314882
-Grid       : Message        : Average mflops/s per call per node (full): 436507
-Grid       : Message        : Average mflops/s per call per node (full): 440248
-Grid       : Message        : Average mflops/s per call per node (full): 306543
-Grid       : Message        :  Stencil 13.9848 GB/s per node
-Grid       : Message        :  Stencil 18.3976 GB/s per node
-Grid       : Message        :  Stencil 19.3124 GB/s per node
-Grid       : Message        :  Stencil 12.9166 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664656
-Grid       : Message        : Average mflops/s per call per node       : 799203
-Grid       : Message        : Average mflops/s per call per node       : 819769
-Grid       : Message        : Average mflops/s per call per node       : 669625
-Grid       : Message        : Average mflops/s per call per node (full): 316861
-Grid       : Message        : Average mflops/s per call per node (full): 439961
-Grid       : Message        : Average mflops/s per call per node (full): 448614
-Grid       : Message        : Average mflops/s per call per node (full): 307114
-Grid       : Message        :  Stencil 12.4935 GB/s per node
-Grid       : Message        :  Stencil 11.2988 GB/s per node
-Grid       : Message        :  Stencil 17.3402 GB/s per node
-Grid       : Message        :  Stencil 12.5342 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668122
-Grid       : Message        : Average mflops/s per call per node       : 813150
-Grid       : Message        : Average mflops/s per call per node       : 823168
-Grid       : Message        : Average mflops/s per call per node       : 665416
-Grid       : Message        : Average mflops/s per call per node (full): 313288
-Grid       : Message        : Average mflops/s per call per node (full): 346150
-Grid       : Message        : Average mflops/s per call per node (full): 437381
-Grid       : Message        : Average mflops/s per call per node (full): 304229
-Grid       : Message        :  Stencil 14.216 GB/s per node
-Grid       : Message        :  Stencil 17.0074 GB/s per node
-Grid       : Message        :  Stencil 17.2103 GB/s per node
-Grid       : Message        :  Stencil 13.9827 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663028
-Grid       : Message        : Average mflops/s per call per node       : 805888
-Grid       : Message        : Average mflops/s per call per node       : 820719
-Grid       : Message        : Average mflops/s per call per node       : 663627
-Grid       : Message        : Average mflops/s per call per node (full): 316786
-Grid       : Message        : Average mflops/s per call per node (full): 438756
-Grid       : Message        : Average mflops/s per call per node (full): 441710
-Grid       : Message        : Average mflops/s per call per node (full): 307103
-Grid       : Message        :  Stencil 13.1908 GB/s per node
-Grid       : Message        :  Stencil 14.8085 GB/s per node
-Grid       : Message        :  Stencil 17.4607 GB/s per node
-Grid       : Message        :  Stencil 12.3831 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665860
-Grid       : Message        : Average mflops/s per call per node       : 807115
-Grid       : Message        : Average mflops/s per call per node       : 824929
-Grid       : Message        : Average mflops/s per call per node       : 670725
-Grid       : Message        : Average mflops/s per call per node (full): 314340
-Grid       : Message        : Average mflops/s per call per node (full): 414385
-Grid       : Message        : Average mflops/s per call per node (full): 445376
-Grid       : Message        : Average mflops/s per call per node (full): 304375
-Grid       : Message        :  Stencil 13.5194 GB/s per node
-Grid       : Message        :  Stencil 13.942 GB/s per node
-Grid       : Message        :  Stencil 17.6336 GB/s per node
-Grid       : Message        :  Stencil 12.367 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669064
-Grid       : Message        : Average mflops/s per call per node       : 804949
-Grid       : Message        : Average mflops/s per call per node       : 817478
-Grid       : Message        : Average mflops/s per call per node       : 665460
-Grid       : Message        : Average mflops/s per call per node (full): 316926
-Grid       : Message        : Average mflops/s per call per node (full): 399455
-Grid       : Message        : Average mflops/s per call per node (full): 446074
-Grid       : Message        : Average mflops/s per call per node (full): 304541
-Grid       : Message        :  Stencil 13.6126 GB/s per node
-Grid       : Message        :  Stencil 15.9337 GB/s per node
-Grid       : Message        :  Stencil 17.3187 GB/s per node
-Grid       : Message        :  Stencil 12.4468 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665578
-Grid       : Message        : Average mflops/s per call per node       : 812716
-Grid       : Message        : Average mflops/s per call per node       : 821287
-Grid       : Message        : Average mflops/s per call per node       : 665420
-Grid       : Message        : Average mflops/s per call per node (full): 316025
-Grid       : Message        : Average mflops/s per call per node (full): 426919
-Grid       : Message        : Average mflops/s per call per node (full): 443410
-Grid       : Message        : Average mflops/s per call per node (full): 304065
-Grid       : Message        :  Stencil 12.7614 GB/s per node
-Grid       : Message        :  Stencil 17.1654 GB/s per node
-Grid       : Message        :  Stencil 18.0923 GB/s per node
-Grid       : Message        :  Stencil 13.4006 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671025
-Grid       : Message        : Average mflops/s per call per node       : 797503
-Grid       : Message        : Average mflops/s per call per node       : 823647
-Grid       : Message        : Average mflops/s per call per node       : 659844
-Grid       : Message        : Average mflops/s per call per node (full): 316108
-Grid       : Message        : Average mflops/s per call per node (full): 438833
-Grid       : Message        : Average mflops/s per call per node (full): 447618
-Grid       : Message        : Average mflops/s per call per node (full): 305267
-Grid       : Message        :  Stencil 12.9773 GB/s per node
-Grid       : Message        :  Stencil 17.1714 GB/s per node
-Grid       : Message        :  Stencil 17.282 GB/s per node
-Grid       : Message        :  Stencil 11.9263 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671307
-Grid       : Message        : Average mflops/s per call per node       : 804090
-Grid       : Message        : Average mflops/s per call per node       : 827698
-Grid       : Message        : Average mflops/s per call per node       : 672546
-Grid       : Message        : Average mflops/s per call per node (full): 316159
-Grid       : Message        : Average mflops/s per call per node (full): 438851
-Grid       : Message        : Average mflops/s per call per node (full): 445167
-Grid       : Message        : Average mflops/s per call per node (full): 300966
-Grid       : Message        :  Stencil 13.6184 GB/s per node
-Grid       : Message        :  Stencil 17.0852 GB/s per node
-Grid       : Message        :  Stencil 17.8702 GB/s per node
-Grid       : Message        :  Stencil 12.7869 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665162
-Grid       : Message        : Average mflops/s per call per node       : 803811
-Grid       : Message        : Average mflops/s per call per node       : 828009
-Grid       : Message        : Average mflops/s per call per node       : 662307
-Grid       : Message        : Average mflops/s per call per node (full): 316505
-Grid       : Message        : Average mflops/s per call per node (full): 438134
-Grid       : Message        : Average mflops/s per call per node (full): 447694
-Grid       : Message        : Average mflops/s per call per node (full): 304436
-Grid       : Message        :  Stencil 12.6024 GB/s per node
-Grid       : Message        :  Stencil 17.1863 GB/s per node
-Grid       : Message        :  Stencil 18.0182 GB/s per node
-Grid       : Message        :  Stencil 13.271 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668267
-Grid       : Message        : Average mflops/s per call per node       : 800026
-Grid       : Message        : Average mflops/s per call per node       : 822868
-Grid       : Message        : Average mflops/s per call per node       : 661952
-Grid       : Message        : Average mflops/s per call per node (full): 310228
-Grid       : Message        : Average mflops/s per call per node (full): 435863
-Grid       : Message        : Average mflops/s per call per node (full): 446221
-Grid       : Message        : Average mflops/s per call per node (full): 305856
-Grid       : Message        :  Stencil 13.1809 GB/s per node
-Grid       : Message        :  Stencil 13.4887 GB/s per node
-Grid       : Message        :  Stencil 17.3289 GB/s per node
-Grid       : Message        :  Stencil 13.1904 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664234
-Grid       : Message        : Average mflops/s per call per node       : 803314
-Grid       : Message        : Average mflops/s per call per node       : 815947
-Grid       : Message        : Average mflops/s per call per node       : 661962
-Grid       : Message        : Average mflops/s per call per node (full): 314075
-Grid       : Message        : Average mflops/s per call per node (full): 389819
-Grid       : Message        : Average mflops/s per call per node (full): 441479
-Grid       : Message        : Average mflops/s per call per node (full): 305914
-Grid       : Message        :  Stencil 13.1528 GB/s per node
-Grid       : Message        :  Stencil 16.7188 GB/s per node
-Grid       : Message        :  Stencil 16.5866 GB/s per node
-Grid       : Message        :  Stencil 12.213 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663048
-Grid       : Message        : Average mflops/s per call per node       : 804374
-Grid       : Message        : Average mflops/s per call per node       : 819659
-Grid       : Message        : Average mflops/s per call per node       : 667866
-Grid       : Message        : Average mflops/s per call per node (full): 312955
-Grid       : Message        : Average mflops/s per call per node (full): 436178
-Grid       : Message        : Average mflops/s per call per node (full): 435129
-Grid       : Message        : Average mflops/s per call per node (full): 299583
-Grid       : Message        :  Stencil 12.839 GB/s per node
-Grid       : Message        :  Stencil 17.4587 GB/s per node
-Grid       : Message        :  Stencil 17.6344 GB/s per node
-Grid       : Message        :  Stencil 13.0173 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660284
-Grid       : Message        : Average mflops/s per call per node       : 804717
-Grid       : Message        : Average mflops/s per call per node       : 827121
-Grid       : Message        : Average mflops/s per call per node       : 663707
-Grid       : Message        : Average mflops/s per call per node (full): 312460
-Grid       : Message        : Average mflops/s per call per node (full): 438847
-Grid       : Message        : Average mflops/s per call per node (full): 446329
-Grid       : Message        : Average mflops/s per call per node (full): 305415
-Grid       : Message        :  Stencil 13.2555 GB/s per node
-Grid       : Message        :  Stencil 17.6703 GB/s per node
-Grid       : Message        :  Stencil 18.5159 GB/s per node
-Grid       : Message        :  Stencil 13.467 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663868
-Grid       : Message        : Average mflops/s per call per node       : 800212
-Grid       : Message        : Average mflops/s per call per node       : 824806
-Grid       : Message        : Average mflops/s per call per node       : 664923
-Grid       : Message        : Average mflops/s per call per node (full): 315689
-Grid       : Message        : Average mflops/s per call per node (full): 439311
-Grid       : Message        : Average mflops/s per call per node (full): 449243
-Grid       : Message        : Average mflops/s per call per node (full): 305734
-Grid       : Message        :  Stencil 12.3396 GB/s per node
-Grid       : Message        :  Stencil 16.991 GB/s per node
-Grid       : Message        :  Stencil 17.5986 GB/s per node
-Grid       : Message        :  Stencil 11.9851 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665254
-Grid       : Message        : Average mflops/s per call per node       : 807953
-Grid       : Message        : Average mflops/s per call per node       : 821190
-Grid       : Message        : Average mflops/s per call per node       : 668761
-Grid       : Message        : Average mflops/s per call per node (full): 312356
-Grid       : Message        : Average mflops/s per call per node (full): 432148
-Grid       : Message        : Average mflops/s per call per node (full): 435364
-Grid       : Message        : Average mflops/s per call per node (full): 302434
-Grid       : Message        :  Stencil 12.6533 GB/s per node
-Grid       : Message        :  Stencil 16.6566 GB/s per node
-Grid       : Message        :  Stencil 17.421 GB/s per node
-Grid       : Message        :  Stencil 13.6647 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664238
-Grid       : Message        : Average mflops/s per call per node       : 807496
-Grid       : Message        : Average mflops/s per call per node       : 826445
-Grid       : Message        : Average mflops/s per call per node       : 657369
-Grid       : Message        : Average mflops/s per call per node (full): 313584
-Grid       : Message        : Average mflops/s per call per node (full): 435134
-Grid       : Message        : Average mflops/s per call per node (full): 441673
-Grid       : Message        : Average mflops/s per call per node (full): 305415
-Grid       : Message        :  Stencil 12.4773 GB/s per node
-Grid       : Message        :  Stencil 18.1009 GB/s per node
-Grid       : Message        :  Stencil 17.9373 GB/s per node
-Grid       : Message        :  Stencil 12.6955 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664810
-Grid       : Message        : Average mflops/s per call per node       : 798276
-Grid       : Message        : Average mflops/s per call per node       : 820774
-Grid       : Message        : Average mflops/s per call per node       : 667568
-Grid       : Message        : Average mflops/s per call per node (full): 313601
-Grid       : Message        : Average mflops/s per call per node (full): 439527
-Grid       : Message        : Average mflops/s per call per node (full): 446031
-Grid       : Message        : Average mflops/s per call per node (full): 304211
-Grid       : Message        :  Stencil 12.8544 GB/s per node
-Grid       : Message        :  Stencil 18.2949 GB/s per node
-Grid       : Message        :  Stencil 16.4737 GB/s per node
-Grid       : Message        :  Stencil 12.2715 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670412
-Grid       : Message        : Average mflops/s per call per node       : 805830
-Grid       : Message        : Average mflops/s per call per node       : 826551
-Grid       : Message        : Average mflops/s per call per node       : 667492
-Grid       : Message        : Average mflops/s per call per node (full): 315155
-Grid       : Message        : Average mflops/s per call per node (full): 441838
-Grid       : Message        : Average mflops/s per call per node (full): 432139
-Grid       : Message        : Average mflops/s per call per node (full): 304623
-Grid       : Message        :  Stencil 14.1742 GB/s per node
-Grid       : Message        :  Stencil 16.563 GB/s per node
-Grid       : Message        :  Stencil 17.5296 GB/s per node
-Grid       : Message        :  Stencil 13.3106 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666231
-Grid       : Message        : Average mflops/s per call per node       : 809848
-Grid       : Message        : Average mflops/s per call per node       : 815874
-Grid       : Message        : Average mflops/s per call per node       : 662812
-Grid       : Message        : Average mflops/s per call per node (full): 315886
-Grid       : Message        : Average mflops/s per call per node (full): 436379
-Grid       : Message        : Average mflops/s per call per node (full): 444229
-Grid       : Message        : Average mflops/s per call per node (full): 306706
-Grid       : Message        :  Stencil 13.297 GB/s per node
-Grid       : Message        :  Stencil 10.6758 GB/s per node
-Grid       : Message        :  Stencil 18.0975 GB/s per node
-Grid       : Message        :  Stencil 12.9191 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664986
-Grid       : Message        : Average mflops/s per call per node       : 809930
-Grid       : Message        : Average mflops/s per call per node       : 824570
-Grid       : Message        : Average mflops/s per call per node       : 662503
-Grid       : Message        : Average mflops/s per call per node (full): 314510
-Grid       : Message        : Average mflops/s per call per node (full): 332417
-Grid       : Message        : Average mflops/s per call per node (full): 447999
-Grid       : Message        : Average mflops/s per call per node (full): 304748
-Grid       : Message        :  Stencil 13.2278 GB/s per node
-Grid       : Message        :  Stencil 10.026 GB/s per node
-Grid       : Message        :  Stencil 18.1989 GB/s per node
-Grid       : Message        :  Stencil 13.4784 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668048
-Grid       : Message        : Average mflops/s per call per node       : 807632
-Grid       : Message        : Average mflops/s per call per node       : 822427
-Grid       : Message        : Average mflops/s per call per node       : 663594
-Grid       : Message        : Average mflops/s per call per node (full): 315399
-Grid       : Message        : Average mflops/s per call per node (full): 317245
-Grid       : Message        : Average mflops/s per call per node (full): 446650
-Grid       : Message        : Average mflops/s per call per node (full): 306189
-Grid       : Message        :  Stencil 13.2856 GB/s per node
-Grid       : Message        :  Stencil 16.8058 GB/s per node
-Grid       : Message        :  Stencil 18.3342 GB/s per node
-Grid       : Message        :  Stencil 13.7579 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666376
-Grid       : Message        : Average mflops/s per call per node       : 807190
-Grid       : Message        : Average mflops/s per call per node       : 820858
-Grid       : Message        : Average mflops/s per call per node       : 658153
-Grid       : Message        : Average mflops/s per call per node (full): 314857
-Grid       : Message        : Average mflops/s per call per node (full): 435853
-Grid       : Message        : Average mflops/s per call per node (full): 447087
-Grid       : Message        : Average mflops/s per call per node (full): 305513
-Grid       : Message        :  Stencil 15.1605 GB/s per node
-Grid       : Message        :  Stencil 17.0136 GB/s per node
-Grid       : Message        :  Stencil 18.373 GB/s per node
-Grid       : Message        :  Stencil 12.3423 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662966
-Grid       : Message        : Average mflops/s per call per node       : 808304
-Grid       : Message        : Average mflops/s per call per node       : 818004
-Grid       : Message        : Average mflops/s per call per node       : 666981
-Grid       : Message        : Average mflops/s per call per node (full): 316955
-Grid       : Message        : Average mflops/s per call per node (full): 439448
-Grid       : Message        : Average mflops/s per call per node (full): 447067
-Grid       : Message        : Average mflops/s per call per node (full): 302182
-Grid       : Message        :  Stencil 12.5394 GB/s per node
-Grid       : Message        :  Stencil 16.3613 GB/s per node
-Grid       : Message        :  Stencil 17.7478 GB/s per node
-Grid       : Message        :  Stencil 12.9737 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666957
-Grid       : Message        : Average mflops/s per call per node       : 803510
-Grid       : Message        : Average mflops/s per call per node       : 822896
-Grid       : Message        : Average mflops/s per call per node       : 661411
-Grid       : Message        : Average mflops/s per call per node (full): 314184
-Grid       : Message        : Average mflops/s per call per node (full): 432010
-Grid       : Message        : Average mflops/s per call per node (full): 441724
-Grid       : Message        : Average mflops/s per call per node (full): 301640
-Grid       : Message        :  Stencil 13.5336 GB/s per node
-Grid       : Message        :  Stencil 16.7272 GB/s per node
-Grid       : Message        :  Stencil 16.7701 GB/s per node
-Grid       : Message        :  Stencil 11.9536 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663344
-Grid       : Message        : Average mflops/s per call per node       : 807277
-Grid       : Message        : Average mflops/s per call per node       : 828164
-Grid       : Message        : Average mflops/s per call per node       : 666565
-Grid       : Message        : Average mflops/s per call per node (full): 314791
-Grid       : Message        : Average mflops/s per call per node (full): 436726
-Grid       : Message        : Average mflops/s per call per node (full): 438177
-Grid       : Message        : Average mflops/s per call per node (full): 300950
-Grid       : Message        :  Stencil 13.0412 GB/s per node
-Grid       : Message        :  Stencil 18.249 GB/s per node
-Grid       : Message        :  Stencil 17.3893 GB/s per node
-Grid       : Message        :  Stencil 12.5904 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667289
-Grid       : Message        : Average mflops/s per call per node       : 800480
-Grid       : Message        : Average mflops/s per call per node       : 826123
-Grid       : Message        : Average mflops/s per call per node       : 668870
-Grid       : Message        : Average mflops/s per call per node (full): 314766
-Grid       : Message        : Average mflops/s per call per node (full): 440731
-Grid       : Message        : Average mflops/s per call per node (full): 444114
-Grid       : Message        : Average mflops/s per call per node (full): 305542
-Grid       : Message        :  Stencil 14.1855 GB/s per node
-Grid       : Message        :  Stencil 17.3159 GB/s per node
-Grid       : Message        :  Stencil 17.3951 GB/s per node
-Grid       : Message        :  Stencil 14.7946 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664318
-Grid       : Message        : Average mflops/s per call per node       : 802147
-Grid       : Message        : Average mflops/s per call per node       : 823625
-Grid       : Message        : Average mflops/s per call per node       : 660868
-Grid       : Message        : Average mflops/s per call per node (full): 315826
-Grid       : Message        : Average mflops/s per call per node (full): 438869
-Grid       : Message        : Average mflops/s per call per node (full): 445153
-Grid       : Message        : Average mflops/s per call per node (full): 307083
-Grid       : Message        :  Stencil 14.3068 GB/s per node
-Grid       : Message        :  Stencil 16.111 GB/s per node
-Grid       : Message        :  Stencil 17.2469 GB/s per node
-Grid       : Message        :  Stencil 12.5397 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662917
-Grid       : Message        : Average mflops/s per call per node       : 809051
-Grid       : Message        : Average mflops/s per call per node       : 821687
-Grid       : Message        : Average mflops/s per call per node       : 666637
-Grid       : Message        : Average mflops/s per call per node (full): 316955
-Grid       : Message        : Average mflops/s per call per node (full): 422775
-Grid       : Message        : Average mflops/s per call per node (full): 442596
-Grid       : Message        : Average mflops/s per call per node (full): 304989
-Grid       : Message        :  Stencil 13.1324 GB/s per node
-Grid       : Message        :  Stencil 15.9543 GB/s per node
-Grid       : Message        :  Stencil 17.6188 GB/s per node
-Grid       : Message        :  Stencil 13.5612 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666396
-Grid       : Message        : Average mflops/s per call per node       : 809715
-Grid       : Message        : Average mflops/s per call per node       : 820205
-Grid       : Message        : Average mflops/s per call per node       : 664769
-Grid       : Message        : Average mflops/s per call per node (full): 315161
-Grid       : Message        : Average mflops/s per call per node (full): 426166
-Grid       : Message        : Average mflops/s per call per node (full): 444006
-Grid       : Message        : Average mflops/s per call per node (full): 305982
-Grid       : Message        :  Stencil 12.978 GB/s per node
-Grid       : Message        :  Stencil 17.042 GB/s per node
-Grid       : Message        :  Stencil 17.937 GB/s per node
-Grid       : Message        :  Stencil 11.8709 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666662
-Grid       : Message        : Average mflops/s per call per node       : 799634
-Grid       : Message        : Average mflops/s per call per node       : 819919
-Grid       : Message        : Average mflops/s per call per node       : 663800
-Grid       : Message        : Average mflops/s per call per node (full): 315038
-Grid       : Message        : Average mflops/s per call per node (full): 436877
-Grid       : Message        : Average mflops/s per call per node (full): 445358
-Grid       : Message        : Average mflops/s per call per node (full): 300508
-Grid       : Message        :  Stencil 12.4893 GB/s per node
-Grid       : Message        :  Stencil 17.7135 GB/s per node
-Grid       : Message        :  Stencil 17.2763 GB/s per node
-Grid       : Message        :  Stencil 12.7222 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664912
-Grid       : Message        : Average mflops/s per call per node       : 806802
-Grid       : Message        : Average mflops/s per call per node       : 826341
-Grid       : Message        : Average mflops/s per call per node       : 659176
-Grid       : Message        : Average mflops/s per call per node (full): 312454
-Grid       : Message        : Average mflops/s per call per node (full): 441070
-Grid       : Message        : Average mflops/s per call per node (full): 442784
-Grid       : Message        : Average mflops/s per call per node (full): 304059
-Grid       : Message        :  Stencil 12.7942 GB/s per node
-Grid       : Message        :  Stencil 17.2664 GB/s per node
-Grid       : Message        :  Stencil 17.7487 GB/s per node
-Grid       : Message        :  Stencil 12.6147 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663989
-Grid       : Message        : Average mflops/s per call per node       : 802431
-Grid       : Message        : Average mflops/s per call per node       : 825882
-Grid       : Message        : Average mflops/s per call per node       : 669656
-Grid       : Message        : Average mflops/s per call per node (full): 303988
-Grid       : Message        : Average mflops/s per call per node (full): 437662
-Grid       : Message        : Average mflops/s per call per node (full): 444731
-Grid       : Message        : Average mflops/s per call per node (full): 305627
-Grid       : Message        :  Stencil 13.6631 GB/s per node
-Grid       : Message        :  Stencil 16.8758 GB/s per node
-Grid       : Message        :  Stencil 17.9437 GB/s per node
-Grid       : Message        :  Stencil 12.5825 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664333
-Grid       : Message        : Average mflops/s per call per node       : 806678
-Grid       : Message        : Average mflops/s per call per node       : 818733
-Grid       : Message        : Average mflops/s per call per node       : 661848
-Grid       : Message        : Average mflops/s per call per node (full): 315757
-Grid       : Message        : Average mflops/s per call per node (full): 438414
-Grid       : Message        : Average mflops/s per call per node (full): 445361
-Grid       : Message        : Average mflops/s per call per node (full): 302864
-Grid       : Message        :  Stencil 12.1964 GB/s per node
-Grid       : Message        :  Stencil 14.0574 GB/s per node
-Grid       : Message        :  Stencil 17.7224 GB/s per node
-Grid       : Message        :  Stencil 12.4657 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667607
-Grid       : Message        : Average mflops/s per call per node       : 807422
-Grid       : Message        : Average mflops/s per call per node       : 825382
-Grid       : Message        : Average mflops/s per call per node       : 670496
-Grid       : Message        : Average mflops/s per call per node (full): 310813
-Grid       : Message        : Average mflops/s per call per node (full): 400623
-Grid       : Message        : Average mflops/s per call per node (full): 446110
-Grid       : Message        : Average mflops/s per call per node (full): 305631
-Grid       : Message        :  Stencil 12.9772 GB/s per node
-Grid       : Message        :  Stencil 16.5692 GB/s per node
-Grid       : Message        :  Stencil 17.082 GB/s per node
-Grid       : Message        :  Stencil 13.1432 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663156
-Grid       : Message        : Average mflops/s per call per node       : 800475
-Grid       : Message        : Average mflops/s per call per node       : 827650
-Grid       : Message        : Average mflops/s per call per node       : 663837
-Grid       : Message        : Average mflops/s per call per node (full): 313312
-Grid       : Message        : Average mflops/s per call per node (full): 433722
-Grid       : Message        : Average mflops/s per call per node (full): 440128
-Grid       : Message        : Average mflops/s per call per node (full): 305750
-Grid       : Message        :  Stencil 12.6959 GB/s per node
-Grid       : Message        :  Stencil 9.75687 GB/s per node
-Grid       : Message        :  Stencil 17.5084 GB/s per node
-Grid       : Message        :  Stencil 12.0511 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670135
-Grid       : Message        : Average mflops/s per call per node       : 813094
-Grid       : Message        : Average mflops/s per call per node       : 828302
-Grid       : Message        : Average mflops/s per call per node       : 668614
-Grid       : Message        : Average mflops/s per call per node (full): 314985
-Grid       : Message        : Average mflops/s per call per node (full): 310972
-Grid       : Message        : Average mflops/s per call per node (full): 445489
-Grid       : Message        : Average mflops/s per call per node (full): 302304
-Grid       : Message        :  Stencil 15.1776 GB/s per node
-Grid       : Message        :  Stencil 16.8651 GB/s per node
-Grid       : Message        :  Stencil 16.6968 GB/s per node
-Grid       : Message        :  Stencil 12.2462 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664631
-Grid       : Message        : Average mflops/s per call per node       : 804873
-Grid       : Message        : Average mflops/s per call per node       : 816983
-Grid       : Message        : Average mflops/s per call per node       : 671057
-Grid       : Message        : Average mflops/s per call per node (full): 317617
-Grid       : Message        : Average mflops/s per call per node (full): 436935
-Grid       : Message        : Average mflops/s per call per node (full): 426973
-Grid       : Message        : Average mflops/s per call per node (full): 304192
-Grid       : Message        :  Stencil 13.7144 GB/s per node
-Grid       : Message        :  Stencil 18.7334 GB/s per node
-Grid       : Message        :  Stencil 17.9377 GB/s per node
-Grid       : Message        :  Stencil 13.5519 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665825
-Grid       : Message        : Average mflops/s per call per node       : 807739
-Grid       : Message        : Average mflops/s per call per node       : 821023
-Grid       : Message        : Average mflops/s per call per node       : 665112
-Grid       : Message        : Average mflops/s per call per node (full): 317255
-Grid       : Message        : Average mflops/s per call per node (full): 441000
-Grid       : Message        : Average mflops/s per call per node (full): 446026
-Grid       : Message        : Average mflops/s per call per node (full): 306568
-Grid       : Message        :  Stencil 14.4005 GB/s per node
-Grid       : Message        :  Stencil 14.9304 GB/s per node
-Grid       : Message        :  Stencil 17.8885 GB/s per node
-Grid       : Message        :  Stencil 12.5922 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666094
-Grid       : Message        : Average mflops/s per call per node       : 805924
-Grid       : Message        : Average mflops/s per call per node       : 821798
-Grid       : Message        : Average mflops/s per call per node       : 670074
-Grid       : Message        : Average mflops/s per call per node (full): 316933
-Grid       : Message        : Average mflops/s per call per node (full): 415751
-Grid       : Message        : Average mflops/s per call per node (full): 446741
-Grid       : Message        : Average mflops/s per call per node (full): 305588
-Grid       : Message        :  Stencil 13.4556 GB/s per node
-Grid       : Message        :  Stencil 17.7989 GB/s per node
-Grid       : Message        :  Stencil 17.8123 GB/s per node
-Grid       : Message        :  Stencil 13.2153 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665540
-Grid       : Message        : Average mflops/s per call per node       : 798428
-Grid       : Message        : Average mflops/s per call per node       : 819450
-Grid       : Message        : Average mflops/s per call per node       : 662678
-Grid       : Message        : Average mflops/s per call per node (full): 314840
-Grid       : Message        : Average mflops/s per call per node (full): 438204
-Grid       : Message        : Average mflops/s per call per node (full): 444908
-Grid       : Message        : Average mflops/s per call per node (full): 305172
-Grid       : Message        :  Stencil 14.147 GB/s per node
-Grid       : Message        :  Stencil 16.5951 GB/s per node
-Grid       : Message        :  Stencil 17.045 GB/s per node
-Grid       : Message        :  Stencil 12.3705 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667115
-Grid       : Message        : Average mflops/s per call per node       : 802583
-Grid       : Message        : Average mflops/s per call per node       : 820262
-Grid       : Message        : Average mflops/s per call per node       : 661130
-Grid       : Message        : Average mflops/s per call per node (full): 316760
-Grid       : Message        : Average mflops/s per call per node (full): 425907
-Grid       : Message        : Average mflops/s per call per node (full): 440866
-Grid       : Message        : Average mflops/s per call per node (full): 302945
-Grid       : Message        :  Stencil 12.4367 GB/s per node
-Grid       : Message        :  Stencil 16.6064 GB/s per node
-Grid       : Message        :  Stencil 17.5276 GB/s per node
-Grid       : Message        :  Stencil 12.5734 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666864
-Grid       : Message        : Average mflops/s per call per node       : 798125
-Grid       : Message        : Average mflops/s per call per node       : 820718
-Grid       : Message        : Average mflops/s per call per node       : 665262
-Grid       : Message        : Average mflops/s per call per node (full): 310215
-Grid       : Message        : Average mflops/s per call per node (full): 429816
-Grid       : Message        : Average mflops/s per call per node (full): 442081
-Grid       : Message        : Average mflops/s per call per node (full): 304755
-Grid       : Message        :  Stencil 13.0628 GB/s per node
-Grid       : Message        :  Stencil 16.6893 GB/s per node
-Grid       : Message        :  Stencil 17.1925 GB/s per node
-Grid       : Message        :  Stencil 12.3717 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665550
-Grid       : Message        : Average mflops/s per call per node       : 803240
-Grid       : Message        : Average mflops/s per call per node       : 817650
-Grid       : Message        : Average mflops/s per call per node       : 661483
-Grid       : Message        : Average mflops/s per call per node (full): 313860
-Grid       : Message        : Average mflops/s per call per node (full): 435647
-Grid       : Message        : Average mflops/s per call per node (full): 442649
-Grid       : Message        : Average mflops/s per call per node (full): 303165
-Grid       : Message        :  Stencil 12.7405 GB/s per node
-Grid       : Message        :  Stencil 17.0502 GB/s per node
-Grid       : Message        :  Stencil 17.5352 GB/s per node
-Grid       : Message        :  Stencil 11.8669 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667635
-Grid       : Message        : Average mflops/s per call per node       : 797258
-Grid       : Message        : Average mflops/s per call per node       : 815596
-Grid       : Message        : Average mflops/s per call per node       : 663218
-Grid       : Message        : Average mflops/s per call per node (full): 314792
-Grid       : Message        : Average mflops/s per call per node (full): 436536
-Grid       : Message        : Average mflops/s per call per node (full): 443681
-Grid       : Message        : Average mflops/s per call per node (full): 300390
-Grid       : Message        :  Stencil 12.4671 GB/s per node
-Grid       : Message        :  Stencil 16.5639 GB/s per node
-Grid       : Message        :  Stencil 18.9733 GB/s per node
-Grid       : Message        :  Stencil 11.9817 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664843
-Grid       : Message        : Average mflops/s per call per node       : 800012
-Grid       : Message        : Average mflops/s per call per node       : 819876
-Grid       : Message        : Average mflops/s per call per node       : 663148
-Grid       : Message        : Average mflops/s per call per node (full): 311164
-Grid       : Message        : Average mflops/s per call per node (full): 435468
-Grid       : Message        : Average mflops/s per call per node (full): 448897
-Grid       : Message        : Average mflops/s per call per node (full): 301392
-Grid       : Message        :  Stencil 12.9655 GB/s per node
-Grid       : Message        :  Stencil 16.8094 GB/s per node
-Grid       : Message        :  Stencil 17.5092 GB/s per node
-Grid       : Message        :  Stencil 12.7611 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661145
-Grid       : Message        : Average mflops/s per call per node       : 802228
-Grid       : Message        : Average mflops/s per call per node       : 826971
-Grid       : Message        : Average mflops/s per call per node       : 660795
-Grid       : Message        : Average mflops/s per call per node (full): 314337
-Grid       : Message        : Average mflops/s per call per node (full): 435938
-Grid       : Message        : Average mflops/s per call per node (full): 445923
-Grid       : Message        : Average mflops/s per call per node (full): 304664
-Grid       : Message        :  Stencil 13.1316 GB/s per node
-Grid       : Message        :  Stencil 13.4993 GB/s per node
-Grid       : Message        :  Stencil 17.7713 GB/s per node
-Grid       : Message        :  Stencil 12.6489 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663786
-Grid       : Message        : Average mflops/s per call per node       : 800213
-Grid       : Message        : Average mflops/s per call per node       : 822439
-Grid       : Message        : Average mflops/s per call per node       : 662126
-Grid       : Message        : Average mflops/s per call per node (full): 315340
-Grid       : Message        : Average mflops/s per call per node (full): 388301
-Grid       : Message        : Average mflops/s per call per node (full): 446386
-Grid       : Message        : Average mflops/s per call per node (full): 304489
-Grid       : Message        :  Stencil 12.78 GB/s per node
-Grid       : Message        :  Stencil 16.1777 GB/s per node
-Grid       : Message        :  Stencil 17.0707 GB/s per node
-Grid       : Message        :  Stencil 13.3816 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666441
-Grid       : Message        : Average mflops/s per call per node       : 803715
-Grid       : Message        : Average mflops/s per call per node       : 819076
-Grid       : Message        : Average mflops/s per call per node       : 661900
-Grid       : Message        : Average mflops/s per call per node (full): 314891
-Grid       : Message        : Average mflops/s per call per node (full): 430371
-Grid       : Message        : Average mflops/s per call per node (full): 441525
-Grid       : Message        : Average mflops/s per call per node (full): 304762
-Grid       : Message        :  Stencil 14.2821 GB/s per node
-Grid       : Message        :  Stencil 17.3567 GB/s per node
-Grid       : Message        :  Stencil 17.0311 GB/s per node
-Grid       : Message        :  Stencil 12.3513 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660644
-Grid       : Message        : Average mflops/s per call per node       : 807218
-Grid       : Message        : Average mflops/s per call per node       : 829333
-Grid       : Message        : Average mflops/s per call per node       : 671991
-Grid       : Message        : Average mflops/s per call per node (full): 315644
-Grid       : Message        : Average mflops/s per call per node (full): 438153
-Grid       : Message        : Average mflops/s per call per node (full): 441341
-Grid       : Message        : Average mflops/s per call per node (full): 304531
-Grid       : Message        :  Stencil 14.2211 GB/s per node
-Grid       : Message        :  Stencil 16.8738 GB/s per node
-Grid       : Message        :  Stencil 17.8739 GB/s per node
-Grid       : Message        :  Stencil 12.481 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 658591
-Grid       : Message        : Average mflops/s per call per node       : 803936
-Grid       : Message        : Average mflops/s per call per node       : 819286
-Grid       : Message        : Average mflops/s per call per node       : 666410
-Grid       : Message        : Average mflops/s per call per node (full): 314857
-Grid       : Message        : Average mflops/s per call per node (full): 438458
-Grid       : Message        : Average mflops/s per call per node (full): 436476
-Grid       : Message        : Average mflops/s per call per node (full): 301718
-Grid       : Message        :  Stencil 12.858 GB/s per node
-Grid       : Message        :  Stencil 17.1302 GB/s per node
-Grid       : Message        :  Stencil 17.3094 GB/s per node
-Grid       : Message        :  Stencil 12.7569 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664558
-Grid       : Message        : Average mflops/s per call per node       : 801165
-Grid       : Message        : Average mflops/s per call per node       : 821151
-Grid       : Message        : Average mflops/s per call per node       : 662358
-Grid       : Message        : Average mflops/s per call per node (full): 314254
-Grid       : Message        : Average mflops/s per call per node (full): 435368
-Grid       : Message        : Average mflops/s per call per node (full): 444693
-Grid       : Message        : Average mflops/s per call per node (full): 304294
-Grid       : Message        :  Stencil 15.4771 GB/s per node
-Grid       : Message        :  Stencil 14.5214 GB/s per node
-Grid       : Message        :  Stencil 18.2442 GB/s per node
-Grid       : Message        :  Stencil 13.1201 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659965
-Grid       : Message        : Average mflops/s per call per node       : 807033
-Grid       : Message        : Average mflops/s per call per node       : 820979
-Grid       : Message        : Average mflops/s per call per node       : 662589
-Grid       : Message        : Average mflops/s per call per node (full): 316696
-Grid       : Message        : Average mflops/s per call per node (full): 409084
-Grid       : Message        : Average mflops/s per call per node (full): 448128
-Grid       : Message        : Average mflops/s per call per node (full): 304887
-Grid       : Message        :  Stencil 12.5206 GB/s per node
-Grid       : Message        :  Stencil 16.7692 GB/s per node
-Grid       : Message        :  Stencil 17.6077 GB/s per node
-Grid       : Message        :  Stencil 12.2452 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664987
-Grid       : Message        : Average mflops/s per call per node       : 804050
-Grid       : Message        : Average mflops/s per call per node       : 822386
-Grid       : Message        : Average mflops/s per call per node       : 665008
-Grid       : Message        : Average mflops/s per call per node (full): 312935
-Grid       : Message        : Average mflops/s per call per node (full): 435278
-Grid       : Message        : Average mflops/s per call per node (full): 445270
-Grid       : Message        : Average mflops/s per call per node (full): 302602
-Grid       : Message        :  Stencil 12.8442 GB/s per node
-Grid       : Message        :  Stencil 17.2721 GB/s per node
-Grid       : Message        :  Stencil 18.2259 GB/s per node
-Grid       : Message        :  Stencil 12.1537 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665418
-Grid       : Message        : Average mflops/s per call per node       : 800007
-Grid       : Message        : Average mflops/s per call per node       : 815405
-Grid       : Message        : Average mflops/s per call per node       : 670901
-Grid       : Message        : Average mflops/s per call per node (full): 312220
-Grid       : Message        : Average mflops/s per call per node (full): 432543
-Grid       : Message        : Average mflops/s per call per node (full): 445549
-Grid       : Message        : Average mflops/s per call per node (full): 303527
-Grid       : Message        :  Stencil 12.5181 GB/s per node
-Grid       : Message        :  Stencil 17.2748 GB/s per node
-Grid       : Message        :  Stencil 18.1267 GB/s per node
-Grid       : Message        :  Stencil 13.6044 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667650
-Grid       : Message        : Average mflops/s per call per node       : 800526
-Grid       : Message        : Average mflops/s per call per node       : 818541
-Grid       : Message        : Average mflops/s per call per node       : 660433
-Grid       : Message        : Average mflops/s per call per node (full): 313808
-Grid       : Message        : Average mflops/s per call per node (full): 437457
-Grid       : Message        : Average mflops/s per call per node (full): 445161
-Grid       : Message        : Average mflops/s per call per node (full): 305531
-Grid       : Message        :  Stencil 13.9883 GB/s per node
-Grid       : Message        :  Stencil 17.3733 GB/s per node
-Grid       : Message        :  Stencil 18.4282 GB/s per node
-Grid       : Message        :  Stencil 12.2481 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665282
-Grid       : Message        : Average mflops/s per call per node       : 803784
-Grid       : Message        : Average mflops/s per call per node       : 821363
-Grid       : Message        : Average mflops/s per call per node       : 673318
-Grid       : Message        : Average mflops/s per call per node (full): 316685
-Grid       : Message        : Average mflops/s per call per node (full): 438703
-Grid       : Message        : Average mflops/s per call per node (full): 446737
-Grid       : Message        : Average mflops/s per call per node (full): 304041
-Grid       : Message        :  Stencil 13.2789 GB/s per node
-Grid       : Message        :  Stencil 16.4879 GB/s per node
-Grid       : Message        :  Stencil 17.6746 GB/s per node
-Grid       : Message        :  Stencil 13.917 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665807
-Grid       : Message        : Average mflops/s per call per node       : 803510
-Grid       : Message        : Average mflops/s per call per node       : 816688
-Grid       : Message        : Average mflops/s per call per node       : 661748
-Grid       : Message        : Average mflops/s per call per node (full): 313889
-Grid       : Message        : Average mflops/s per call per node (full): 434971
-Grid       : Message        : Average mflops/s per call per node (full): 445035
-Grid       : Message        : Average mflops/s per call per node (full): 307077
-Grid       : Message        :  Stencil 12.2734 GB/s per node
-Grid       : Message        :  Stencil 12.66 GB/s per node
-Grid       : Message        :  Stencil 17.0617 GB/s per node
-Grid       : Message        :  Stencil 13.5623 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667015
-Grid       : Message        : Average mflops/s per call per node       : 807794
-Grid       : Message        : Average mflops/s per call per node       : 822143
-Grid       : Message        : Average mflops/s per call per node       : 661033
-Grid       : Message        : Average mflops/s per call per node (full): 311024
-Grid       : Message        : Average mflops/s per call per node (full): 374728
-Grid       : Message        : Average mflops/s per call per node (full): 441706
-Grid       : Message        : Average mflops/s per call per node (full): 305472
-Grid       : Message        :  Stencil 12.1509 GB/s per node
-Grid       : Message        :  Stencil 17.0251 GB/s per node
-Grid       : Message        :  Stencil 16.1424 GB/s per node
-Grid       : Message        :  Stencil 12.3466 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668234
-Grid       : Message        : Average mflops/s per call per node       : 806025
-Grid       : Message        : Average mflops/s per call per node       : 821907
-Grid       : Message        : Average mflops/s per call per node       : 665139
-Grid       : Message        : Average mflops/s per call per node (full): 310571
-Grid       : Message        : Average mflops/s per call per node (full): 438555
-Grid       : Message        : Average mflops/s per call per node (full): 417720
-Grid       : Message        : Average mflops/s per call per node (full): 301979
-Grid       : Message        :  Stencil 12.3345 GB/s per node
-Grid       : Message        :  Stencil 16.5728 GB/s per node
-Grid       : Message        :  Stencil 17.1912 GB/s per node
-Grid       : Message        :  Stencil 12.7909 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667776
-Grid       : Message        : Average mflops/s per call per node       : 809703
-Grid       : Message        : Average mflops/s per call per node       : 827346
-Grid       : Message        : Average mflops/s per call per node       : 665721
-Grid       : Message        : Average mflops/s per call per node (full): 310230
-Grid       : Message        : Average mflops/s per call per node (full): 435928
-Grid       : Message        : Average mflops/s per call per node (full): 443387
-Grid       : Message        : Average mflops/s per call per node (full): 305302
-Grid       : Message        :  Stencil 14.1641 GB/s per node
-Grid       : Message        :  Stencil 16.2782 GB/s per node
-Grid       : Message        :  Stencil 17.5187 GB/s per node
-Grid       : Message        :  Stencil 12.2126 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665588
-Grid       : Message        : Average mflops/s per call per node       : 803008
-Grid       : Message        : Average mflops/s per call per node       : 823725
-Grid       : Message        : Average mflops/s per call per node       : 670796
-Grid       : Message        : Average mflops/s per call per node (full): 316399
-Grid       : Message        : Average mflops/s per call per node (full): 432108
-Grid       : Message        : Average mflops/s per call per node (full): 445275
-Grid       : Message        : Average mflops/s per call per node (full): 304455
-Grid       : Message        :  Stencil 12.3719 GB/s per node
-Grid       : Message        :  Stencil 9.44157 GB/s per node
-Grid       : Message        :  Stencil 18.2184 GB/s per node
-Grid       : Message        :  Stencil 13.0357 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 673287
-Grid       : Message        : Average mflops/s per call per node       : 810733
-Grid       : Message        : Average mflops/s per call per node       : 826413
-Grid       : Message        : Average mflops/s per call per node       : 664360
-Grid       : Message        : Average mflops/s per call per node (full): 313111
-Grid       : Message        : Average mflops/s per call per node (full): 303110
-Grid       : Message        : Average mflops/s per call per node (full): 447822
-Grid       : Message        : Average mflops/s per call per node (full): 305094
-Grid       : Message        :  Stencil 14.4216 GB/s per node
-Grid       : Message        :  Stencil 16.8166 GB/s per node
-Grid       : Message        :  Stencil 17.0233 GB/s per node
-Grid       : Message        :  Stencil 12.4243 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664592
-Grid       : Message        : Average mflops/s per call per node       : 799741
-Grid       : Message        : Average mflops/s per call per node       : 817291
-Grid       : Message        : Average mflops/s per call per node       : 660091
-Grid       : Message        : Average mflops/s per call per node (full): 316638
-Grid       : Message        : Average mflops/s per call per node (full): 436341
-Grid       : Message        : Average mflops/s per call per node (full): 440702
-Grid       : Message        : Average mflops/s per call per node (full): 302313
-Grid       : Message        :  Stencil 14.0866 GB/s per node
-Grid       : Message        :  Stencil 15.121 GB/s per node
-Grid       : Message        :  Stencil 17.1513 GB/s per node
-Grid       : Message        :  Stencil 14.4614 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662020
-Grid       : Message        : Average mflops/s per call per node       : 805280
-Grid       : Message        : Average mflops/s per call per node       : 825135
-Grid       : Message        : Average mflops/s per call per node       : 663528
-Grid       : Message        : Average mflops/s per call per node (full): 315369
-Grid       : Message        : Average mflops/s per call per node (full): 416997
-Grid       : Message        : Average mflops/s per call per node (full): 442667
-Grid       : Message        : Average mflops/s per call per node (full): 307982
-Grid       : Message        :  Stencil 13.0421 GB/s per node
-Grid       : Message        :  Stencil 13.475 GB/s per node
-Grid       : Message        :  Stencil 17.8514 GB/s per node
-Grid       : Message        :  Stencil 12.3485 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668948
-Grid       : Message        : Average mflops/s per call per node       : 806437
-Grid       : Message        : Average mflops/s per call per node       : 816250
-Grid       : Message        : Average mflops/s per call per node       : 667938
-Grid       : Message        : Average mflops/s per call per node (full): 315978
-Grid       : Message        : Average mflops/s per call per node (full): 390547
-Grid       : Message        : Average mflops/s per call per node (full): 446984
-Grid       : Message        : Average mflops/s per call per node (full): 304395
-Grid       : Message        :  Stencil 12.5389 GB/s per node
-Grid       : Message        :  Stencil 16.7972 GB/s per node
-Grid       : Message        :  Stencil 17.4921 GB/s per node
-Grid       : Message        :  Stencil 12.8593 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668172
-Grid       : Message        : Average mflops/s per call per node       : 801761
-Grid       : Message        : Average mflops/s per call per node       : 818191
-Grid       : Message        : Average mflops/s per call per node       : 667537
-Grid       : Message        : Average mflops/s per call per node (full): 313733
-Grid       : Message        : Average mflops/s per call per node (full): 435641
-Grid       : Message        : Average mflops/s per call per node (full): 443637
-Grid       : Message        : Average mflops/s per call per node (full): 305751
-Grid       : Message        :  Stencil 12.9159 GB/s per node
-Grid       : Message        :  Stencil 17.2392 GB/s per node
-Grid       : Message        :  Stencil 17.8093 GB/s per node
-Grid       : Message        :  Stencil 12.1072 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665873
-Grid       : Message        : Average mflops/s per call per node       : 801432
-Grid       : Message        : Average mflops/s per call per node       : 819327
-Grid       : Message        : Average mflops/s per call per node       : 666187
-Grid       : Message        : Average mflops/s per call per node (full): 315585
-Grid       : Message        : Average mflops/s per call per node (full): 439621
-Grid       : Message        : Average mflops/s per call per node (full): 445822
-Grid       : Message        : Average mflops/s per call per node (full): 302671
-Grid       : Message        :  Stencil 15.4449 GB/s per node
-Grid       : Message        :  Stencil 10.9898 GB/s per node
-Grid       : Message        :  Stencil 16.9168 GB/s per node
-Grid       : Message        :  Stencil 12.7072 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659037
-Grid       : Message        : Average mflops/s per call per node       : 807417
-Grid       : Message        : Average mflops/s per call per node       : 823099
-Grid       : Message        : Average mflops/s per call per node       : 667598
-Grid       : Message        : Average mflops/s per call per node (full): 316591
-Grid       : Message        : Average mflops/s per call per node (full): 339479
-Grid       : Message        : Average mflops/s per call per node (full): 430694
-Grid       : Message        : Average mflops/s per call per node (full): 303012
-Grid       : Message        :  Stencil 14.7637 GB/s per node
-Grid       : Message        :  Stencil 16.2874 GB/s per node
-Grid       : Message        :  Stencil 18.6323 GB/s per node
-Grid       : Message        :  Stencil 12.9178 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660695
-Grid       : Message        : Average mflops/s per call per node       : 801611
-Grid       : Message        : Average mflops/s per call per node       : 815351
-Grid       : Message        : Average mflops/s per call per node       : 669428
-Grid       : Message        : Average mflops/s per call per node (full): 316066
-Grid       : Message        : Average mflops/s per call per node (full): 431920
-Grid       : Message        : Average mflops/s per call per node (full): 448644
-Grid       : Message        : Average mflops/s per call per node (full): 306518
-Grid       : Message        :  Stencil 14.3132 GB/s per node
-Grid       : Message        :  Stencil 17.1266 GB/s per node
-Grid       : Message        :  Stencil 17.0616 GB/s per node
-Grid       : Message        :  Stencil 13.8447 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 658407
-Grid       : Message        : Average mflops/s per call per node       : 803101
-Grid       : Message        : Average mflops/s per call per node       : 819603
-Grid       : Message        : Average mflops/s per call per node       : 659596
-Grid       : Message        : Average mflops/s per call per node (full): 315588
-Grid       : Message        : Average mflops/s per call per node (full): 438032
-Grid       : Message        : Average mflops/s per call per node (full): 440639
-Grid       : Message        : Average mflops/s per call per node (full): 305823
-Grid       : Message        :  Stencil 13.4098 GB/s per node
-Grid       : Message        :  Stencil 16.5838 GB/s per node
-Grid       : Message        :  Stencil 17.4944 GB/s per node
-Grid       : Message        :  Stencil 12.611 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661273
-Grid       : Message        : Average mflops/s per call per node       : 800292
-Grid       : Message        : Average mflops/s per call per node       : 821205
-Grid       : Message        : Average mflops/s per call per node       : 666224
-Grid       : Message        : Average mflops/s per call per node (full): 314171
-Grid       : Message        : Average mflops/s per call per node (full): 435713
-Grid       : Message        : Average mflops/s per call per node (full): 445784
-Grid       : Message        : Average mflops/s per call per node (full): 305169
-Grid       : Message        :  Stencil 13.1361 GB/s per node
-Grid       : Message        :  Stencil 17.236 GB/s per node
-Grid       : Message        :  Stencil 17.6766 GB/s per node
-Grid       : Message        :  Stencil 14.132 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663250
-Grid       : Message        : Average mflops/s per call per node       : 807288
-Grid       : Message        : Average mflops/s per call per node       : 824157
-Grid       : Message        : Average mflops/s per call per node       : 660656
-Grid       : Message        : Average mflops/s per call per node (full): 314434
-Grid       : Message        : Average mflops/s per call per node (full): 437413
-Grid       : Message        : Average mflops/s per call per node (full): 446797
-Grid       : Message        : Average mflops/s per call per node (full): 307108
-Grid       : Message        :  Stencil 12.9989 GB/s per node
-Grid       : Message        :  Stencil 17.8977 GB/s per node
-Grid       : Message        :  Stencil 17.4466 GB/s per node
-Grid       : Message        :  Stencil 12.2233 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662489
-Grid       : Message        : Average mflops/s per call per node       : 801498
-Grid       : Message        : Average mflops/s per call per node       : 827275
-Grid       : Message        : Average mflops/s per call per node       : 667737
-Grid       : Message        : Average mflops/s per call per node (full): 311317
-Grid       : Message        : Average mflops/s per call per node (full): 435001
-Grid       : Message        : Average mflops/s per call per node (full): 441079
-Grid       : Message        : Average mflops/s per call per node (full): 303713
-Grid       : Message        :  Stencil 13.7653 GB/s per node
-Grid       : Message        :  Stencil 17.0009 GB/s per node
-Grid       : Message        :  Stencil 17.617 GB/s per node
-Grid       : Message        :  Stencil 12.0234 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664655
-Grid       : Message        : Average mflops/s per call per node       : 802383
-Grid       : Message        : Average mflops/s per call per node       : 821660
-Grid       : Message        : Average mflops/s per call per node       : 664596
-Grid       : Message        : Average mflops/s per call per node (full): 315360
-Grid       : Message        : Average mflops/s per call per node (full): 437320
-Grid       : Message        : Average mflops/s per call per node (full): 444749
-Grid       : Message        : Average mflops/s per call per node (full): 302217
-Grid       : Message        :  Stencil 13.2126 GB/s per node
-Grid       : Message        :  Stencil 17.8005 GB/s per node
-Grid       : Message        :  Stencil 17.0058 GB/s per node
-Grid       : Message        :  Stencil 12.0881 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664981
-Grid       : Message        : Average mflops/s per call per node       : 806023
-Grid       : Message        : Average mflops/s per call per node       : 822818
-Grid       : Message        : Average mflops/s per call per node       : 667398
-Grid       : Message        : Average mflops/s per call per node (full): 315465
-Grid       : Message        : Average mflops/s per call per node (full): 441619
-Grid       : Message        : Average mflops/s per call per node (full): 440684
-Grid       : Message        : Average mflops/s per call per node (full): 302959
-Grid       : Message        :  Stencil 13.5388 GB/s per node
-Grid       : Message        :  Stencil 18.0366 GB/s per node
-Grid       : Message        :  Stencil 17.8471 GB/s per node
-Grid       : Message        :  Stencil 12.4445 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663666
-Grid       : Message        : Average mflops/s per call per node       : 798323
-Grid       : Message        : Average mflops/s per call per node       : 822953
-Grid       : Message        : Average mflops/s per call per node       : 666327
-Grid       : Message        : Average mflops/s per call per node (full): 311731
-Grid       : Message        : Average mflops/s per call per node (full): 439801
-Grid       : Message        : Average mflops/s per call per node (full): 446810
-Grid       : Message        : Average mflops/s per call per node (full): 304155
-Grid       : Message        :  Stencil 13.4909 GB/s per node
-Grid       : Message        :  Stencil 16.82 GB/s per node
-Grid       : Message        :  Stencil 17.578 GB/s per node
-Grid       : Message        :  Stencil 12.1492 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663940
-Grid       : Message        : Average mflops/s per call per node       : 804399
-Grid       : Message        : Average mflops/s per call per node       : 826470
-Grid       : Message        : Average mflops/s per call per node       : 667211
-Grid       : Message        : Average mflops/s per call per node (full): 314458
-Grid       : Message        : Average mflops/s per call per node (full): 433138
-Grid       : Message        : Average mflops/s per call per node (full): 445496
-Grid       : Message        : Average mflops/s per call per node (full): 303364
-Grid       : Message        :  Stencil 13.2232 GB/s per node
-Grid       : Message        :  Stencil 16.7092 GB/s per node
-Grid       : Message        :  Stencil 18.4317 GB/s per node
-Grid       : Message        :  Stencil 13.5936 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667945
-Grid       : Message        : Average mflops/s per call per node       : 808507
-Grid       : Message        : Average mflops/s per call per node       : 819600
-Grid       : Message        : Average mflops/s per call per node       : 664098
-Grid       : Message        : Average mflops/s per call per node (full): 315045
-Grid       : Message        : Average mflops/s per call per node (full): 437537
-Grid       : Message        : Average mflops/s per call per node (full): 446572
-Grid       : Message        : Average mflops/s per call per node (full): 307152
-Grid       : Message        :  Stencil 12.7131 GB/s per node
-Grid       : Message        :  Stencil 18.2008 GB/s per node
-Grid       : Message        :  Stencil 17.2221 GB/s per node
-Grid       : Message        :  Stencil 12.6037 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666033
-Grid       : Message        : Average mflops/s per call per node       : 807661
-Grid       : Message        : Average mflops/s per call per node       : 826467
-Grid       : Message        : Average mflops/s per call per node       : 663763
-Grid       : Message        : Average mflops/s per call per node (full): 314663
-Grid       : Message        : Average mflops/s per call per node (full): 442038
-Grid       : Message        : Average mflops/s per call per node (full): 443390
-Grid       : Message        : Average mflops/s per call per node (full): 303637
-Grid       : Message        :  Stencil 13.5181 GB/s per node
-Grid       : Message        :  Stencil 10.8977 GB/s per node
-Grid       : Message        :  Stencil 17.9444 GB/s per node
-Grid       : Message        :  Stencil 12.0613 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661244
-Grid       : Message        : Average mflops/s per call per node       : 810256
-Grid       : Message        : Average mflops/s per call per node       : 827175
-Grid       : Message        : Average mflops/s per call per node       : 663484
-Grid       : Message        : Average mflops/s per call per node (full): 314360
-Grid       : Message        : Average mflops/s per call per node (full): 337415
-Grid       : Message        : Average mflops/s per call per node (full): 448417
-Grid       : Message        : Average mflops/s per call per node (full): 302725
-Grid       : Message        :  Stencil 13.5593 GB/s per node
-Grid       : Message        :  Stencil 16.7705 GB/s per node
-Grid       : Message        :  Stencil 17.8027 GB/s per node
-Grid       : Message        :  Stencil 15.5264 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666063
-Grid       : Message        : Average mflops/s per call per node       : 804006
-Grid       : Message        : Average mflops/s per call per node       : 823762
-Grid       : Message        : Average mflops/s per call per node       : 658362
-Grid       : Message        : Average mflops/s per call per node (full): 316752
-Grid       : Message        : Average mflops/s per call per node (full): 425272
-Grid       : Message        : Average mflops/s per call per node (full): 446540
-Grid       : Message        : Average mflops/s per call per node (full): 307569
-Grid       : Message        :  Stencil 14.9456 GB/s per node
-Grid       : Message        :  Stencil 17.4885 GB/s per node
-Grid       : Message        :  Stencil 17.3422 GB/s per node
-Grid       : Message        :  Stencil 12.3453 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663467
-Grid       : Message        : Average mflops/s per call per node       : 801409
-Grid       : Message        : Average mflops/s per call per node       : 816290
-Grid       : Message        : Average mflops/s per call per node       : 669270
-Grid       : Message        : Average mflops/s per call per node (full): 317182
-Grid       : Message        : Average mflops/s per call per node (full): 438211
-Grid       : Message        : Average mflops/s per call per node (full): 442788
-Grid       : Message        : Average mflops/s per call per node (full): 304823
-Grid       : Message        :  Stencil 13.9943 GB/s per node
-Grid       : Message        :  Stencil 17.1094 GB/s per node
-Grid       : Message        :  Stencil 18.2979 GB/s per node
-Grid       : Message        :  Stencil 12.9114 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664422
-Grid       : Message        : Average mflops/s per call per node       : 807973
-Grid       : Message        : Average mflops/s per call per node       : 818457
-Grid       : Message        : Average mflops/s per call per node       : 668307
-Grid       : Message        : Average mflops/s per call per node (full): 316225
-Grid       : Message        : Average mflops/s per call per node (full): 438810
-Grid       : Message        : Average mflops/s per call per node (full): 447538
-Grid       : Message        : Average mflops/s per call per node (full): 306457
-Grid       : Message        :  Stencil 12.4988 GB/s per node
-Grid       : Message        :  Stencil 16.2525 GB/s per node
-Grid       : Message        :  Stencil 17.7376 GB/s per node
-Grid       : Message        :  Stencil 11.9912 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666295
-Grid       : Message        : Average mflops/s per call per node       : 803164
-Grid       : Message        : Average mflops/s per call per node       : 824180
-Grid       : Message        : Average mflops/s per call per node       : 670430
-Grid       : Message        : Average mflops/s per call per node (full): 313251
-Grid       : Message        : Average mflops/s per call per node (full): 432096
-Grid       : Message        : Average mflops/s per call per node (full): 446445
-Grid       : Message        : Average mflops/s per call per node (full): 302679
-Grid       : Message        :  Stencil 14.3437 GB/s per node
-Grid       : Message        :  Stencil 17.4697 GB/s per node
-Grid       : Message        :  Stencil 16.9784 GB/s per node
-Grid       : Message        :  Stencil 13.9946 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661222
-Grid       : Message        : Average mflops/s per call per node       : 802776
-Grid       : Message        : Average mflops/s per call per node       : 824277
-Grid       : Message        : Average mflops/s per call per node       : 662738
-Grid       : Message        : Average mflops/s per call per node (full): 315779
-Grid       : Message        : Average mflops/s per call per node (full): 439965
-Grid       : Message        : Average mflops/s per call per node (full): 440859
-Grid       : Message        : Average mflops/s per call per node (full): 307038
-Grid       : Message        :  Stencil 14.8629 GB/s per node
-Grid       : Message        :  Stencil 16.1011 GB/s per node
-Grid       : Message        :  Stencil 18.7162 GB/s per node
-Grid       : Message        :  Stencil 12.7471 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662883
-Grid       : Message        : Average mflops/s per call per node       : 804761
-Grid       : Message        : Average mflops/s per call per node       : 821957
-Grid       : Message        : Average mflops/s per call per node       : 666030
-Grid       : Message        : Average mflops/s per call per node (full): 316443
-Grid       : Message        : Average mflops/s per call per node (full): 429850
-Grid       : Message        : Average mflops/s per call per node (full): 448767
-Grid       : Message        : Average mflops/s per call per node (full): 304830
-Grid       : Message        :  Stencil 14.3938 GB/s per node
-Grid       : Message        :  Stencil 16.8236 GB/s per node
-Grid       : Message        :  Stencil 18.0328 GB/s per node
-Grid       : Message        :  Stencil 11.6708 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665694
-Grid       : Message        : Average mflops/s per call per node       : 803772
-Grid       : Message        : Average mflops/s per call per node       : 820641
-Grid       : Message        : Average mflops/s per call per node       : 668786
-Grid       : Message        : Average mflops/s per call per node (full): 316913
-Grid       : Message        : Average mflops/s per call per node (full): 437280
-Grid       : Message        : Average mflops/s per call per node (full): 446599
-Grid       : Message        : Average mflops/s per call per node (full): 298644
-Grid       : Message        :  Stencil 13.6226 GB/s per node
-Grid       : Message        :  Stencil 10.1949 GB/s per node
-Grid       : Message        :  Stencil 18.0608 GB/s per node
-Grid       : Message        :  Stencil 11.8098 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663352
-Grid       : Message        : Average mflops/s per call per node       : 806740
-Grid       : Message        : Average mflops/s per call per node       : 823623
-Grid       : Message        : Average mflops/s per call per node       : 669411
-Grid       : Message        : Average mflops/s per call per node (full): 315926
-Grid       : Message        : Average mflops/s per call per node (full): 321727
-Grid       : Message        : Average mflops/s per call per node (full): 447028
-Grid       : Message        : Average mflops/s per call per node (full): 300846
-Grid       : Message        :  Stencil 13.8048 GB/s per node
-Grid       : Message        :  Stencil 17.1501 GB/s per node
-Grid       : Message        :  Stencil 17.915 GB/s per node
-Grid       : Message        :  Stencil 12.8037 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665595
-Grid       : Message        : Average mflops/s per call per node       : 800369
-Grid       : Message        : Average mflops/s per call per node       : 821968
-Grid       : Message        : Average mflops/s per call per node       : 663864
-Grid       : Message        : Average mflops/s per call per node (full): 316305
-Grid       : Message        : Average mflops/s per call per node (full): 435911
-Grid       : Message        : Average mflops/s per call per node (full): 445401
-Grid       : Message        : Average mflops/s per call per node (full): 304844
-Grid       : Message        :  Stencil 13.3276 GB/s per node
-Grid       : Message        :  Stencil 16.5479 GB/s per node
-Grid       : Message        :  Stencil 15.9232 GB/s per node
-Grid       : Message        :  Stencil 12.6292 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664191
-Grid       : Message        : Average mflops/s per call per node       : 805749
-Grid       : Message        : Average mflops/s per call per node       : 824182
-Grid       : Message        : Average mflops/s per call per node       : 662231
-Grid       : Message        : Average mflops/s per call per node (full): 316444
-Grid       : Message        : Average mflops/s per call per node (full): 433993
-Grid       : Message        : Average mflops/s per call per node (full): 418336
-Grid       : Message        : Average mflops/s per call per node (full): 304125
-Grid       : Message        :  Stencil 12.9556 GB/s per node
-Grid       : Message        :  Stencil 16.5463 GB/s per node
-Grid       : Message        :  Stencil 17.379 GB/s per node
-Grid       : Message        :  Stencil 12.8479 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667084
-Grid       : Message        : Average mflops/s per call per node       : 801127
-Grid       : Message        : Average mflops/s per call per node       : 824920
-Grid       : Message        : Average mflops/s per call per node       : 668330
-Grid       : Message        : Average mflops/s per call per node (full): 315162
-Grid       : Message        : Average mflops/s per call per node (full): 435167
-Grid       : Message        : Average mflops/s per call per node (full): 444288
-Grid       : Message        : Average mflops/s per call per node (full): 306298
-Grid       : Message        :  Stencil 13.9152 GB/s per node
-Grid       : Message        :  Stencil 17.8934 GB/s per node
-Grid       : Message        :  Stencil 17.9279 GB/s per node
-Grid       : Message        :  Stencil 12.7926 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662878
-Grid       : Message        : Average mflops/s per call per node       : 798961
-Grid       : Message        : Average mflops/s per call per node       : 824106
-Grid       : Message        : Average mflops/s per call per node       : 663412
-Grid       : Message        : Average mflops/s per call per node (full): 315995
-Grid       : Message        : Average mflops/s per call per node (full): 437028
-Grid       : Message        : Average mflops/s per call per node (full): 446450
-Grid       : Message        : Average mflops/s per call per node (full): 305560
-Grid       : Message        :  Stencil 14.802 GB/s per node
-Grid       : Message        :  Stencil 17.5445 GB/s per node
-Grid       : Message        :  Stencil 17.7725 GB/s per node
-Grid       : Message        :  Stencil 12.817 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660844
-Grid       : Message        : Average mflops/s per call per node       : 797512
-Grid       : Message        : Average mflops/s per call per node       : 816649
-Grid       : Message        : Average mflops/s per call per node       : 661087
-Grid       : Message        : Average mflops/s per call per node (full): 316362
-Grid       : Message        : Average mflops/s per call per node (full): 436374
-Grid       : Message        : Average mflops/s per call per node (full): 446109
-Grid       : Message        : Average mflops/s per call per node (full): 304547
-Grid       : Message        :  Stencil 13.819 GB/s per node
-Grid       : Message        :  Stencil 17.337 GB/s per node
-Grid       : Message        :  Stencil 17.8609 GB/s per node
-Grid       : Message        :  Stencil 12.4582 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661106
-Grid       : Message        : Average mflops/s per call per node       : 801815
-Grid       : Message        : Average mflops/s per call per node       : 824608
-Grid       : Message        : Average mflops/s per call per node       : 666840
-Grid       : Message        : Average mflops/s per call per node (full): 315404
-Grid       : Message        : Average mflops/s per call per node (full): 436362
-Grid       : Message        : Average mflops/s per call per node (full): 446372
-Grid       : Message        : Average mflops/s per call per node (full): 304462
-Grid       : Message        :  Stencil 13.3501 GB/s per node
-Grid       : Message        :  Stencil 16.2943 GB/s per node
-Grid       : Message        :  Stencil 17.9905 GB/s per node
-Grid       : Message        :  Stencil 13.1651 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667754
-Grid       : Message        : Average mflops/s per call per node       : 805087
-Grid       : Message        : Average mflops/s per call per node       : 815001
-Grid       : Message        : Average mflops/s per call per node       : 663820
-Grid       : Message        : Average mflops/s per call per node (full): 315198
-Grid       : Message        : Average mflops/s per call per node (full): 430787
-Grid       : Message        : Average mflops/s per call per node (full): 444664
-Grid       : Message        : Average mflops/s per call per node (full): 304994
-Grid       : Message        :  Stencil 12.5859 GB/s per node
-Grid       : Message        :  Stencil 16.4739 GB/s per node
-Grid       : Message        :  Stencil 17.3418 GB/s per node
-Grid       : Message        :  Stencil 12.8162 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671092
-Grid       : Message        : Average mflops/s per call per node       : 803434
-Grid       : Message        : Average mflops/s per call per node       : 819660
-Grid       : Message        : Average mflops/s per call per node       : 667983
-Grid       : Message        : Average mflops/s per call per node (full): 314765
-Grid       : Message        : Average mflops/s per call per node (full): 434197
-Grid       : Message        : Average mflops/s per call per node (full): 442487
-Grid       : Message        : Average mflops/s per call per node (full): 306098
-Grid       : Message        :  Stencil 13.4637 GB/s per node
-Grid       : Message        :  Stencil 7.73841 GB/s per node
-Grid       : Message        :  Stencil 17.9445 GB/s per node
-Grid       : Message        :  Stencil 13.2531 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666729
-Grid       : Message        : Average mflops/s per call per node       : 809090
-Grid       : Message        : Average mflops/s per call per node       : 823783
-Grid       : Message        : Average mflops/s per call per node       : 661230
-Grid       : Message        : Average mflops/s per call per node (full): 315758
-Grid       : Message        : Average mflops/s per call per node (full): 259334
-Grid       : Message        : Average mflops/s per call per node (full): 447738
-Grid       : Message        : Average mflops/s per call per node (full): 304430
-Grid       : Message        :  Stencil 12.637 GB/s per node
-Grid       : Message        :  Stencil 16.3219 GB/s per node
-Grid       : Message        :  Stencil 17.178 GB/s per node
-Grid       : Message        :  Stencil 12.8779 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665199
-Grid       : Message        : Average mflops/s per call per node       : 808316
-Grid       : Message        : Average mflops/s per call per node       : 825699
-Grid       : Message        : Average mflops/s per call per node       : 667714
-Grid       : Message        : Average mflops/s per call per node (full): 313642
-Grid       : Message        : Average mflops/s per call per node (full): 432798
-Grid       : Message        : Average mflops/s per call per node (full): 442997
-Grid       : Message        : Average mflops/s per call per node (full): 305697
-Grid       : Message        :  Stencil 13.8312 GB/s per node
-Grid       : Message        :  Stencil 17.2629 GB/s per node
-Grid       : Message        :  Stencil 17.0892 GB/s per node
-Grid       : Message        :  Stencil 12.7407 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660395
-Grid       : Message        : Average mflops/s per call per node       : 798639
-Grid       : Message        : Average mflops/s per call per node       : 825638
-Grid       : Message        : Average mflops/s per call per node       : 661845
-Grid       : Message        : Average mflops/s per call per node (full): 315821
-Grid       : Message        : Average mflops/s per call per node (full): 436751
-Grid       : Message        : Average mflops/s per call per node (full): 436490
-Grid       : Message        : Average mflops/s per call per node (full): 303363
-Grid       : Message        :  Stencil 14.028 GB/s per node
-Grid       : Message        :  Stencil 16.6012 GB/s per node
-Grid       : Message        :  Stencil 17.5076 GB/s per node
-Grid       : Message        :  Stencil 12.9486 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662658
-Grid       : Message        : Average mflops/s per call per node       : 803401
-Grid       : Message        : Average mflops/s per call per node       : 820992
-Grid       : Message        : Average mflops/s per call per node       : 662085
-Grid       : Message        : Average mflops/s per call per node (full): 316026
-Grid       : Message        : Average mflops/s per call per node (full): 436247
-Grid       : Message        : Average mflops/s per call per node (full): 443812
-Grid       : Message        : Average mflops/s per call per node (full): 304274
-Grid       : Message        :  Stencil 13.5494 GB/s per node
-Grid       : Message        :  Stencil 10.9216 GB/s per node
-Grid       : Message        :  Stencil 17.3544 GB/s per node
-Grid       : Message        :  Stencil 12.5985 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665674
-Grid       : Message        : Average mflops/s per call per node       : 805542
-Grid       : Message        : Average mflops/s per call per node       : 830375
-Grid       : Message        : Average mflops/s per call per node       : 670696
-Grid       : Message        : Average mflops/s per call per node (full): 315974
-Grid       : Message        : Average mflops/s per call per node (full): 337951
-Grid       : Message        : Average mflops/s per call per node (full): 444669
-Grid       : Message        : Average mflops/s per call per node (full): 306025
-Grid       : Message        :  Stencil 14.5241 GB/s per node
-Grid       : Message        :  Stencil 18.0398 GB/s per node
-Grid       : Message        :  Stencil 17.3894 GB/s per node
-Grid       : Message        :  Stencil 13.0366 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662707
-Grid       : Message        : Average mflops/s per call per node       : 802766
-Grid       : Message        : Average mflops/s per call per node       : 816841
-Grid       : Message        : Average mflops/s per call per node       : 665898
-Grid       : Message        : Average mflops/s per call per node (full): 317317
-Grid       : Message        : Average mflops/s per call per node (full): 439686
-Grid       : Message        : Average mflops/s per call per node (full): 441888
-Grid       : Message        : Average mflops/s per call per node (full): 305500
-Grid       : Message        :  Stencil 12.7563 GB/s per node
-Grid       : Message        :  Stencil 16.6327 GB/s per node
-Grid       : Message        :  Stencil 17.4013 GB/s per node
-Grid       : Message        :  Stencil 14.5206 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670543
-Grid       : Message        : Average mflops/s per call per node       : 807913
-Grid       : Message        : Average mflops/s per call per node       : 824241
-Grid       : Message        : Average mflops/s per call per node       : 664469
-Grid       : Message        : Average mflops/s per call per node (full): 313226
-Grid       : Message        : Average mflops/s per call per node (full): 435507
-Grid       : Message        : Average mflops/s per call per node (full): 444819
-Grid       : Message        : Average mflops/s per call per node (full): 308239
-Grid       : Message        :  Stencil 13.8565 GB/s per node
-Grid       : Message        :  Stencil 16.9393 GB/s per node
-Grid       : Message        :  Stencil 17.3037 GB/s per node
-Grid       : Message        :  Stencil 12.6889 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668007
-Grid       : Message        : Average mflops/s per call per node       : 809567
-Grid       : Message        : Average mflops/s per call per node       : 819754
-Grid       : Message        : Average mflops/s per call per node       : 661690
-Grid       : Message        : Average mflops/s per call per node (full): 314114
-Grid       : Message        : Average mflops/s per call per node (full): 438418
-Grid       : Message        : Average mflops/s per call per node (full): 442341
-Grid       : Message        : Average mflops/s per call per node (full): 303115
-Grid       : Message        :  Stencil 14.6176 GB/s per node
-Grid       : Message        :  Stencil 16.4156 GB/s per node
-Grid       : Message        :  Stencil 17.9656 GB/s per node
-Grid       : Message        :  Stencil 12.1492 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663949
-Grid       : Message        : Average mflops/s per call per node       : 811066
-Grid       : Message        : Average mflops/s per call per node       : 818656
-Grid       : Message        : Average mflops/s per call per node       : 660849
-Grid       : Message        : Average mflops/s per call per node (full): 316079
-Grid       : Message        : Average mflops/s per call per node (full): 432736
-Grid       : Message        : Average mflops/s per call per node (full): 445519
-Grid       : Message        : Average mflops/s per call per node (full): 301745
-Grid       : Message        :  Stencil 13.1905 GB/s per node
-Grid       : Message        :  Stencil 17.9314 GB/s per node
-Grid       : Message        :  Stencil 18.1212 GB/s per node
-Grid       : Message        :  Stencil 13.4665 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667469
-Grid       : Message        : Average mflops/s per call per node       : 799710
-Grid       : Message        : Average mflops/s per call per node       : 823445
-Grid       : Message        : Average mflops/s per call per node       : 662119
-Grid       : Message        : Average mflops/s per call per node (full): 315512
-Grid       : Message        : Average mflops/s per call per node (full): 440524
-Grid       : Message        : Average mflops/s per call per node (full): 446878
-Grid       : Message        : Average mflops/s per call per node (full): 306150
-Grid       : Message        :  Stencil 12.7929 GB/s per node
-Grid       : Message        :  Stencil 9.08547 GB/s per node
-Grid       : Message        :  Stencil 18.3256 GB/s per node
-Grid       : Message        :  Stencil 14.0092 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669543
-Grid       : Message        : Average mflops/s per call per node       : 807014
-Grid       : Message        : Average mflops/s per call per node       : 818948
-Grid       : Message        : Average mflops/s per call per node       : 658341
-Grid       : Message        : Average mflops/s per call per node (full): 314512
-Grid       : Message        : Average mflops/s per call per node (full): 294179
-Grid       : Message        : Average mflops/s per call per node (full): 446968
-Grid       : Message        : Average mflops/s per call per node (full): 306159
-Grid       : Message        :  Stencil 13.6148 GB/s per node
-Grid       : Message        :  Stencil 16.5713 GB/s per node
-Grid       : Message        :  Stencil 17.8983 GB/s per node
-Grid       : Message        :  Stencil 12.3309 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666637
-Grid       : Message        : Average mflops/s per call per node       : 806221
-Grid       : Message        : Average mflops/s per call per node       : 819564
-Grid       : Message        : Average mflops/s per call per node       : 668249
-Grid       : Message        : Average mflops/s per call per node (full): 313595
-Grid       : Message        : Average mflops/s per call per node (full): 436378
-Grid       : Message        : Average mflops/s per call per node (full): 440557
-Grid       : Message        : Average mflops/s per call per node (full): 301593
-Grid       : Message        :  Stencil 12.452 GB/s per node
-Grid       : Message        :  Stencil 17.205 GB/s per node
-Grid       : Message        :  Stencil 17.6185 GB/s per node
-Grid       : Message        :  Stencil 12.9366 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 673064
-Grid       : Message        : Average mflops/s per call per node       : 800476
-Grid       : Message        : Average mflops/s per call per node       : 821564
-Grid       : Message        : Average mflops/s per call per node       : 661854
-Grid       : Message        : Average mflops/s per call per node (full): 313548
-Grid       : Message        : Average mflops/s per call per node (full): 434248
-Grid       : Message        : Average mflops/s per call per node (full): 445327
-Grid       : Message        : Average mflops/s per call per node (full): 304364
-Grid       : Message        :  Stencil 13.0521 GB/s per node
-Grid       : Message        :  Stencil 16.7585 GB/s per node
-Grid       : Message        :  Stencil 17.2553 GB/s per node
-Grid       : Message        :  Stencil 13.3428 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663655
-Grid       : Message        : Average mflops/s per call per node       : 806512
-Grid       : Message        : Average mflops/s per call per node       : 819044
-Grid       : Message        : Average mflops/s per call per node       : 661744
-Grid       : Message        : Average mflops/s per call per node (full): 314477
-Grid       : Message        : Average mflops/s per call per node (full): 436994
-Grid       : Message        : Average mflops/s per call per node (full): 443491
-Grid       : Message        : Average mflops/s per call per node (full): 305860
-Grid       : Message        :  Stencil 14.1847 GB/s per node
-Grid       : Message        :  Stencil 12.4488 GB/s per node
-Grid       : Message        :  Stencil 17.0334 GB/s per node
-Grid       : Message        :  Stencil 12.6692 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666023
-Grid       : Message        : Average mflops/s per call per node       : 808708
-Grid       : Message        : Average mflops/s per call per node       : 819004
-Grid       : Message        : Average mflops/s per call per node       : 666128
-Grid       : Message        : Average mflops/s per call per node (full): 317309
-Grid       : Message        : Average mflops/s per call per node (full): 370478
-Grid       : Message        : Average mflops/s per call per node (full): 438950
-Grid       : Message        : Average mflops/s per call per node (full): 304976
-Grid       : Message        :  Stencil 13.2247 GB/s per node
-Grid       : Message        :  Stencil 9.3349 GB/s per node
-Grid       : Message        :  Stencil 17.4661 GB/s per node
-Grid       : Message        :  Stencil 12.2981 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664947
-Grid       : Message        : Average mflops/s per call per node       : 812408
-Grid       : Message        : Average mflops/s per call per node       : 816737
-Grid       : Message        : Average mflops/s per call per node       : 668435
-Grid       : Message        : Average mflops/s per call per node (full): 315500
-Grid       : Message        : Average mflops/s per call per node (full): 300518
-Grid       : Message        : Average mflops/s per call per node (full): 443404
-Grid       : Message        : Average mflops/s per call per node (full): 303863
-Grid       : Message        :  Stencil 13.7203 GB/s per node
-Grid       : Message        :  Stencil 17.2069 GB/s per node
-Grid       : Message        :  Stencil 17.7459 GB/s per node
-Grid       : Message        :  Stencil 12.5509 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663900
-Grid       : Message        : Average mflops/s per call per node       : 799419
-Grid       : Message        : Average mflops/s per call per node       : 821869
-Grid       : Message        : Average mflops/s per call per node       : 666586
-Grid       : Message        : Average mflops/s per call per node (full): 315915
-Grid       : Message        : Average mflops/s per call per node (full): 437670
-Grid       : Message        : Average mflops/s per call per node (full): 445017
-Grid       : Message        : Average mflops/s per call per node (full): 302862
-Grid       : Message        :  Stencil 13.4093 GB/s per node
-Grid       : Message        :  Stencil 16.9177 GB/s per node
-Grid       : Message        :  Stencil 17.0182 GB/s per node
-Grid       : Message        :  Stencil 12.9735 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667387
-Grid       : Message        : Average mflops/s per call per node       : 805937
-Grid       : Message        : Average mflops/s per call per node       : 820998
-Grid       : Message        : Average mflops/s per call per node       : 663590
-Grid       : Message        : Average mflops/s per call per node (full): 316559
-Grid       : Message        : Average mflops/s per call per node (full): 437874
-Grid       : Message        : Average mflops/s per call per node (full): 441596
-Grid       : Message        : Average mflops/s per call per node (full): 305372
-Grid       : Message        :  Stencil 12.5414 GB/s per node
-Grid       : Message        :  Stencil 13.4095 GB/s per node
-Grid       : Message        :  Stencil 17.9259 GB/s per node
-Grid       : Message        :  Stencil 14.453 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671866
-Grid       : Message        : Average mflops/s per call per node       : 808558
-Grid       : Message        : Average mflops/s per call per node       : 816009
-Grid       : Message        : Average mflops/s per call per node       : 662280
-Grid       : Message        : Average mflops/s per call per node (full): 313778
-Grid       : Message        : Average mflops/s per call per node (full): 390067
-Grid       : Message        : Average mflops/s per call per node (full): 445621
-Grid       : Message        : Average mflops/s per call per node (full): 307685
-Grid       : Message        :  Stencil 13.0586 GB/s per node
-Grid       : Message        :  Stencil 17.8407 GB/s per node
-Grid       : Message        :  Stencil 17.6383 GB/s per node
-Grid       : Message        :  Stencil 12.424 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665049
-Grid       : Message        : Average mflops/s per call per node       : 802252
-Grid       : Message        : Average mflops/s per call per node       : 818440
-Grid       : Message        : Average mflops/s per call per node       : 664308
-Grid       : Message        : Average mflops/s per call per node (full): 314421
-Grid       : Message        : Average mflops/s per call per node (full): 439421
-Grid       : Message        : Average mflops/s per call per node (full): 444839
-Grid       : Message        : Average mflops/s per call per node (full): 301055
-Grid       : Message        :  Stencil 13.8105 GB/s per node
-Grid       : Message        :  Stencil 16.9644 GB/s per node
-Grid       : Message        :  Stencil 17.9834 GB/s per node
-Grid       : Message        :  Stencil 12.6671 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664235
-Grid       : Message        : Average mflops/s per call per node       : 807501
-Grid       : Message        : Average mflops/s per call per node       : 817287
-Grid       : Message        : Average mflops/s per call per node       : 662850
-Grid       : Message        : Average mflops/s per call per node (full): 315857
-Grid       : Message        : Average mflops/s per call per node (full): 438824
-Grid       : Message        : Average mflops/s per call per node (full): 446259
-Grid       : Message        : Average mflops/s per call per node (full): 303882
-Grid       : Message        :  Stencil 14.374 GB/s per node
-Grid       : Message        :  Stencil 16.8686 GB/s per node
-Grid       : Message        :  Stencil 18.0065 GB/s per node
-Grid       : Message        :  Stencil 12.4734 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664446
-Grid       : Message        : Average mflops/s per call per node       : 804749
-Grid       : Message        : Average mflops/s per call per node       : 815605
-Grid       : Message        : Average mflops/s per call per node       : 669340
-Grid       : Message        : Average mflops/s per call per node (full): 316594
-Grid       : Message        : Average mflops/s per call per node (full): 436341
-Grid       : Message        : Average mflops/s per call per node (full): 446244
-Grid       : Message        : Average mflops/s per call per node (full): 304947
-Grid       : Message        :  Stencil 14.0086 GB/s per node
-Grid       : Message        :  Stencil 14.369 GB/s per node
-Grid       : Message        :  Stencil 17.0426 GB/s per node
-Grid       : Message        :  Stencil 14.3542 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666933
-Grid       : Message        : Average mflops/s per call per node       : 807542
-Grid       : Message        : Average mflops/s per call per node       : 821032
-Grid       : Message        : Average mflops/s per call per node       : 660200
-Grid       : Message        : Average mflops/s per call per node (full): 317049
-Grid       : Message        : Average mflops/s per call per node (full): 405986
-Grid       : Message        : Average mflops/s per call per node (full): 440205
-Grid       : Message        : Average mflops/s per call per node (full): 306721
-Grid       : Message        :  Stencil 13.74 GB/s per node
-Grid       : Message        :  Stencil 16.4136 GB/s per node
-Grid       : Message        :  Stencil 17.747 GB/s per node
-Grid       : Message        :  Stencil 13.0566 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667117
-Grid       : Message        : Average mflops/s per call per node       : 799704
-Grid       : Message        : Average mflops/s per call per node       : 825934
-Grid       : Message        : Average mflops/s per call per node       : 666386
-Grid       : Message        : Average mflops/s per call per node (full): 316527
-Grid       : Message        : Average mflops/s per call per node (full): 431300
-Grid       : Message        : Average mflops/s per call per node (full): 445056
-Grid       : Message        : Average mflops/s per call per node (full): 305967
-Grid       : Message        :  Stencil 15.0414 GB/s per node
-Grid       : Message        :  Stencil 16.3805 GB/s per node
-Grid       : Message        :  Stencil 17.7157 GB/s per node
-Grid       : Message        :  Stencil 11.8524 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662854
-Grid       : Message        : Average mflops/s per call per node       : 803063
-Grid       : Message        : Average mflops/s per call per node       : 819147
-Grid       : Message        : Average mflops/s per call per node       : 668262
-Grid       : Message        : Average mflops/s per call per node (full): 316698
-Grid       : Message        : Average mflops/s per call per node (full): 425852
-Grid       : Message        : Average mflops/s per call per node (full): 446307
-Grid       : Message        : Average mflops/s per call per node (full): 299278
-Grid       : Message        :  Stencil 14.2677 GB/s per node
-Grid       : Message        :  Stencil 17.8884 GB/s per node
-Grid       : Message        :  Stencil 18.4766 GB/s per node
-Grid       : Message        :  Stencil 12.2529 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661618
-Grid       : Message        : Average mflops/s per call per node       : 803210
-Grid       : Message        : Average mflops/s per call per node       : 819503
-Grid       : Message        : Average mflops/s per call per node       : 665029
-Grid       : Message        : Average mflops/s per call per node (full): 314492
-Grid       : Message        : Average mflops/s per call per node (full): 439913
-Grid       : Message        : Average mflops/s per call per node (full): 447901
-Grid       : Message        : Average mflops/s per call per node (full): 303560
-Grid       : Message        :  Stencil 15.4062 GB/s per node
-Grid       : Message        :  Stencil 14.2115 GB/s per node
-Grid       : Message        :  Stencil 17.3118 GB/s per node
-Grid       : Message        :  Stencil 12.8169 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660690
-Grid       : Message        : Average mflops/s per call per node       : 806604
-Grid       : Message        : Average mflops/s per call per node       : 825682
-Grid       : Message        : Average mflops/s per call per node       : 663405
-Grid       : Message        : Average mflops/s per call per node (full): 317080
-Grid       : Message        : Average mflops/s per call per node (full): 403412
-Grid       : Message        : Average mflops/s per call per node (full): 444843
-Grid       : Message        : Average mflops/s per call per node (full): 305059
-Grid       : Message        :  Stencil 13.2666 GB/s per node
-Grid       : Message        :  Stencil 18.5442 GB/s per node
-Grid       : Message        :  Stencil 18.4985 GB/s per node
-Grid       : Message        :  Stencil 14.6172 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666265
-Grid       : Message        : Average mflops/s per call per node       : 799181
-Grid       : Message        : Average mflops/s per call per node       : 815491
-Grid       : Message        : Average mflops/s per call per node       : 657783
-Grid       : Message        : Average mflops/s per call per node (full): 315400
-Grid       : Message        : Average mflops/s per call per node (full): 440243
-Grid       : Message        : Average mflops/s per call per node (full): 444845
-Grid       : Message        : Average mflops/s per call per node (full): 307117
-Grid       : Message        :  Stencil 13.0131 GB/s per node
-Grid       : Message        :  Stencil 17.2194 GB/s per node
-Grid       : Message        :  Stencil 17.1459 GB/s per node
-Grid       : Message        :  Stencil 12.3687 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670289
-Grid       : Message        : Average mflops/s per call per node       : 808556
-Grid       : Message        : Average mflops/s per call per node       : 823513
-Grid       : Message        : Average mflops/s per call per node       : 665368
-Grid       : Message        : Average mflops/s per call per node (full): 314349
-Grid       : Message        : Average mflops/s per call per node (full): 440568
-Grid       : Message        : Average mflops/s per call per node (full): 441278
-Grid       : Message        : Average mflops/s per call per node (full): 303793
-Grid       : Message        :  Stencil 12.5975 GB/s per node
-Grid       : Message        :  Stencil 17.4383 GB/s per node
-Grid       : Message        :  Stencil 17.8292 GB/s per node
-Grid       : Message        :  Stencil 12.6179 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671699
-Grid       : Message        : Average mflops/s per call per node       : 808745
-Grid       : Message        : Average mflops/s per call per node       : 825012
-Grid       : Message        : Average mflops/s per call per node       : 667837
-Grid       : Message        : Average mflops/s per call per node (full): 314785
-Grid       : Message        : Average mflops/s per call per node (full): 441635
-Grid       : Message        : Average mflops/s per call per node (full): 445869
-Grid       : Message        : Average mflops/s per call per node (full): 305078
-Grid       : Message        :  Stencil 13.4828 GB/s per node
-Grid       : Message        :  Stencil 17.391 GB/s per node
-Grid       : Message        :  Stencil 17.5985 GB/s per node
-Grid       : Message        :  Stencil 13.0579 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670305
-Grid       : Message        : Average mflops/s per call per node       : 800124
-Grid       : Message        : Average mflops/s per call per node       : 821717
-Grid       : Message        : Average mflops/s per call per node       : 665973
-Grid       : Message        : Average mflops/s per call per node (full): 316858
-Grid       : Message        : Average mflops/s per call per node (full): 436514
-Grid       : Message        : Average mflops/s per call per node (full): 445506
-Grid       : Message        : Average mflops/s per call per node (full): 305669
-Grid       : Message        :  Stencil 13.8375 GB/s per node
-Grid       : Message        :  Stencil 16.8669 GB/s per node
-Grid       : Message        :  Stencil 17.1294 GB/s per node
-Grid       : Message        :  Stencil 13.6601 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670173
-Grid       : Message        : Average mflops/s per call per node       : 802703
-Grid       : Message        : Average mflops/s per call per node       : 826415
-Grid       : Message        : Average mflops/s per call per node       : 665591
-Grid       : Message        : Average mflops/s per call per node (full): 316417
-Grid       : Message        : Average mflops/s per call per node (full): 436306
-Grid       : Message        : Average mflops/s per call per node (full): 432316
-Grid       : Message        : Average mflops/s per call per node (full): 306859
-Grid       : Message        :  Stencil 13.1054 GB/s per node
-Grid       : Message        :  Stencil 16.5859 GB/s per node
-Grid       : Message        :  Stencil 18.0068 GB/s per node
-Grid       : Message        :  Stencil 13.3095 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669786
-Grid       : Message        : Average mflops/s per call per node       : 804012
-Grid       : Message        : Average mflops/s per call per node       : 815813
-Grid       : Message        : Average mflops/s per call per node       : 664266
-Grid       : Message        : Average mflops/s per call per node (full): 315101
-Grid       : Message        : Average mflops/s per call per node (full): 434858
-Grid       : Message        : Average mflops/s per call per node (full): 444859
-Grid       : Message        : Average mflops/s per call per node (full): 306065
-Grid       : Message        :  Stencil 12.876 GB/s per node
-Grid       : Message        :  Stencil 17.3527 GB/s per node
-Grid       : Message        :  Stencil 16.3377 GB/s per node
-Grid       : Message        :  Stencil 12.7555 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670670
-Grid       : Message        : Average mflops/s per call per node       : 803765
-Grid       : Message        : Average mflops/s per call per node       : 823899
-Grid       : Message        : Average mflops/s per call per node       : 660135
-Grid       : Message        : Average mflops/s per call per node (full): 315159
-Grid       : Message        : Average mflops/s per call per node (full): 438142
-Grid       : Message        : Average mflops/s per call per node (full): 422732
-Grid       : Message        : Average mflops/s per call per node (full): 303690
-Grid       : Message        :  Stencil 14.5704 GB/s per node
-Grid       : Message        :  Stencil 12.0734 GB/s per node
-Grid       : Message        :  Stencil 18.0296 GB/s per node
-Grid       : Message        :  Stencil 12.6776 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666129
-Grid       : Message        : Average mflops/s per call per node       : 803061
-Grid       : Message        : Average mflops/s per call per node       : 814283
-Grid       : Message        : Average mflops/s per call per node       : 657234
-Grid       : Message        : Average mflops/s per call per node (full): 316864
-Grid       : Message        : Average mflops/s per call per node (full): 362979
-Grid       : Message        : Average mflops/s per call per node (full): 444805
-Grid       : Message        : Average mflops/s per call per node (full): 302650
-Grid       : Message        :  Stencil 13.5439 GB/s per node
-Grid       : Message        :  Stencil 16.9154 GB/s per node
-Grid       : Message        :  Stencil 20.0573 GB/s per node
-Grid       : Message        :  Stencil 13.6491 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666737
-Grid       : Message        : Average mflops/s per call per node       : 796338
-Grid       : Message        : Average mflops/s per call per node       : 816756
-Grid       : Message        : Average mflops/s per call per node       : 663693
-Grid       : Message        : Average mflops/s per call per node (full): 316049
-Grid       : Message        : Average mflops/s per call per node (full): 434567
-Grid       : Message        : Average mflops/s per call per node (full): 449696
-Grid       : Message        : Average mflops/s per call per node (full): 306704
-Grid       : Message        :  Stencil 12.5183 GB/s per node
-Grid       : Message        :  Stencil 16.5827 GB/s per node
-Grid       : Message        :  Stencil 18.559 GB/s per node
-Grid       : Message        :  Stencil 14.6143 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669246
-Grid       : Message        : Average mflops/s per call per node       : 804736
-Grid       : Message        : Average mflops/s per call per node       : 813387
-Grid       : Message        : Average mflops/s per call per node       : 663006
-Grid       : Message        : Average mflops/s per call per node (full): 313893
-Grid       : Message        : Average mflops/s per call per node (full): 435101
-Grid       : Message        : Average mflops/s per call per node (full): 445623
-Grid       : Message        : Average mflops/s per call per node (full): 307351
-Grid       : Message        :  Stencil 12.4063 GB/s per node
-Grid       : Message        :  Stencil 16.7085 GB/s per node
-Grid       : Message        :  Stencil 17.7427 GB/s per node
-Grid       : Message        :  Stencil 14.278 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667017
-Grid       : Message        : Average mflops/s per call per node       : 806776
-Grid       : Message        : Average mflops/s per call per node       : 818848
-Grid       : Message        : Average mflops/s per call per node       : 663672
-Grid       : Message        : Average mflops/s per call per node (full): 309893
-Grid       : Message        : Average mflops/s per call per node (full): 437230
-Grid       : Message        : Average mflops/s per call per node (full): 443248
-Grid       : Message        : Average mflops/s per call per node (full): 307578
-Grid       : Message        :  Stencil 12.4697 GB/s per node
-Grid       : Message        :  Stencil 17.7348 GB/s per node
-Grid       : Message        :  Stencil 17.8787 GB/s per node
-Grid       : Message        :  Stencil 12.386 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668605
-Grid       : Message        : Average mflops/s per call per node       : 803182
-Grid       : Message        : Average mflops/s per call per node       : 821161
-Grid       : Message        : Average mflops/s per call per node       : 666723
-Grid       : Message        : Average mflops/s per call per node (full): 312663
-Grid       : Message        : Average mflops/s per call per node (full): 438756
-Grid       : Message        : Average mflops/s per call per node (full): 445401
-Grid       : Message        : Average mflops/s per call per node (full): 303558
-Grid       : Message        :  Stencil 13.4515 GB/s per node
-Grid       : Message        :  Stencil 11.3553 GB/s per node
-Grid       : Message        :  Stencil 18.5098 GB/s per node
-Grid       : Message        :  Stencil 13.0622 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663087
-Grid       : Message        : Average mflops/s per call per node       : 812456
-Grid       : Message        : Average mflops/s per call per node       : 814407
-Grid       : Message        : Average mflops/s per call per node       : 665146
-Grid       : Message        : Average mflops/s per call per node (full): 316157
-Grid       : Message        : Average mflops/s per call per node (full): 347578
-Grid       : Message        : Average mflops/s per call per node (full): 444978
-Grid       : Message        : Average mflops/s per call per node (full): 305370
-Grid       : Message        :  Stencil 13.2173 GB/s per node
-Grid       : Message        :  Stencil 16.2429 GB/s per node
-Grid       : Message        :  Stencil 18.5115 GB/s per node
-Grid       : Message        :  Stencil 12.5288 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668092
-Grid       : Message        : Average mflops/s per call per node       : 805973
-Grid       : Message        : Average mflops/s per call per node       : 824247
-Grid       : Message        : Average mflops/s per call per node       : 669384
-Grid       : Message        : Average mflops/s per call per node (full): 315328
-Grid       : Message        : Average mflops/s per call per node (full): 430969
-Grid       : Message        : Average mflops/s per call per node (full): 443273
-Grid       : Message        : Average mflops/s per call per node (full): 304709
-Grid       : Message        :  Stencil 13.1158 GB/s per node
-Grid       : Message        :  Stencil 16.8212 GB/s per node
-Grid       : Message        :  Stencil 17.6751 GB/s per node
-Grid       : Message        :  Stencil 12.5535 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664131
-Grid       : Message        : Average mflops/s per call per node       : 804083
-Grid       : Message        : Average mflops/s per call per node       : 824802
-Grid       : Message        : Average mflops/s per call per node       : 664059
-Grid       : Message        : Average mflops/s per call per node (full): 315125
-Grid       : Message        : Average mflops/s per call per node (full): 431866
-Grid       : Message        : Average mflops/s per call per node (full): 445963
-Grid       : Message        : Average mflops/s per call per node (full): 304673
-Grid       : Message        :  Stencil 14.1054 GB/s per node
-Grid       : Message        :  Stencil 17.0876 GB/s per node
-Grid       : Message        :  Stencil 16.244 GB/s per node
-Grid       : Message        :  Stencil 12.5805 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663063
-Grid       : Message        : Average mflops/s per call per node       : 800649
-Grid       : Message        : Average mflops/s per call per node       : 832698
-Grid       : Message        : Average mflops/s per call per node       : 668734
-Grid       : Message        : Average mflops/s per call per node (full): 316797
-Grid       : Message        : Average mflops/s per call per node (full): 438163
-Grid       : Message        : Average mflops/s per call per node (full): 421935
-Grid       : Message        : Average mflops/s per call per node (full): 303610
-Grid       : Message        :  Stencil 12.8483 GB/s per node
-Grid       : Message        :  Stencil 17.087 GB/s per node
-Grid       : Message        :  Stencil 17.8021 GB/s per node
-Grid       : Message        :  Stencil 14.3434 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666039
-Grid       : Message        : Average mflops/s per call per node       : 800967
-Grid       : Message        : Average mflops/s per call per node       : 826722
-Grid       : Message        : Average mflops/s per call per node       : 660834
-Grid       : Message        : Average mflops/s per call per node (full): 315401
-Grid       : Message        : Average mflops/s per call per node (full): 436993
-Grid       : Message        : Average mflops/s per call per node (full): 446784
-Grid       : Message        : Average mflops/s per call per node (full): 306784
-Grid       : Message        :  Stencil 12.7988 GB/s per node
-Grid       : Message        :  Stencil 16.4221 GB/s per node
-Grid       : Message        :  Stencil 17.3057 GB/s per node
-Grid       : Message        :  Stencil 14.4481 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668754
-Grid       : Message        : Average mflops/s per call per node       : 799936
-Grid       : Message        : Average mflops/s per call per node       : 824533
-Grid       : Message        : Average mflops/s per call per node       : 661878
-Grid       : Message        : Average mflops/s per call per node (full): 315340
-Grid       : Message        : Average mflops/s per call per node (full): 433837
-Grid       : Message        : Average mflops/s per call per node (full): 445578
-Grid       : Message        : Average mflops/s per call per node (full): 306850
-Grid       : Message        :  Stencil 14.4045 GB/s per node
-Grid       : Message        :  Stencil 16.6417 GB/s per node
-Grid       : Message        :  Stencil 17.9008 GB/s per node
-Grid       : Message        :  Stencil 13.8014 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662777
-Grid       : Message        : Average mflops/s per call per node       : 802338
-Grid       : Message        : Average mflops/s per call per node       : 820027
-Grid       : Message        : Average mflops/s per call per node       : 664195
-Grid       : Message        : Average mflops/s per call per node (full): 316459
-Grid       : Message        : Average mflops/s per call per node (full): 437409
-Grid       : Message        : Average mflops/s per call per node (full): 445966
-Grid       : Message        : Average mflops/s per call per node (full): 306817
-Grid       : Message        :  Stencil 13.1521 GB/s per node
-Grid       : Message        :  Stencil 17.1275 GB/s per node
-Grid       : Message        :  Stencil 17.9507 GB/s per node
-Grid       : Message        :  Stencil 12.6391 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668265
-Grid       : Message        : Average mflops/s per call per node       : 808937
-Grid       : Message        : Average mflops/s per call per node       : 820979
-Grid       : Message        : Average mflops/s per call per node       : 662241
-Grid       : Message        : Average mflops/s per call per node (full): 316102
-Grid       : Message        : Average mflops/s per call per node (full): 435257
-Grid       : Message        : Average mflops/s per call per node (full): 445397
-Grid       : Message        : Average mflops/s per call per node (full): 303644
-Grid       : Message        :  Stencil 12.8526 GB/s per node
-Grid       : Message        :  Stencil 16.7457 GB/s per node
-Grid       : Message        :  Stencil 17.2491 GB/s per node
-Grid       : Message        :  Stencil 13.7785 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666915
-Grid       : Message        : Average mflops/s per call per node       : 807411
-Grid       : Message        : Average mflops/s per call per node       : 822023
-Grid       : Message        : Average mflops/s per call per node       : 656054
-Grid       : Message        : Average mflops/s per call per node (full): 314974
-Grid       : Message        : Average mflops/s per call per node (full): 437865
-Grid       : Message        : Average mflops/s per call per node (full): 443123
-Grid       : Message        : Average mflops/s per call per node (full): 305622
-Grid       : Message        :  Stencil 13.0399 GB/s per node
-Grid       : Message        :  Stencil 17.3983 GB/s per node
-Grid       : Message        :  Stencil 17.8769 GB/s per node
-Grid       : Message        :  Stencil 12.7115 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669243
-Grid       : Message        : Average mflops/s per call per node       : 804714
-Grid       : Message        : Average mflops/s per call per node       : 821109
-Grid       : Message        : Average mflops/s per call per node       : 665715
-Grid       : Message        : Average mflops/s per call per node (full): 315930
-Grid       : Message        : Average mflops/s per call per node (full): 439006
-Grid       : Message        : Average mflops/s per call per node (full): 445645
-Grid       : Message        : Average mflops/s per call per node (full): 304465
-Grid       : Message        :  Stencil 13.0748 GB/s per node
-Grid       : Message        :  Stencil 17.0022 GB/s per node
-Grid       : Message        :  Stencil 17.9339 GB/s per node
-Grid       : Message        :  Stencil 13.2873 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670384
-Grid       : Message        : Average mflops/s per call per node       : 803062
-Grid       : Message        : Average mflops/s per call per node       : 821719
-Grid       : Message        : Average mflops/s per call per node       : 664236
-Grid       : Message        : Average mflops/s per call per node (full): 315475
-Grid       : Message        : Average mflops/s per call per node (full): 437462
-Grid       : Message        : Average mflops/s per call per node (full): 446076
-Grid       : Message        : Average mflops/s per call per node (full): 304854
-Grid       : Message        :  Stencil 13.3451 GB/s per node
-Grid       : Message        :  Stencil 16.987 GB/s per node
-Grid       : Message        :  Stencil 17.4433 GB/s per node
-Grid       : Message        :  Stencil 13.2438 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668057
-Grid       : Message        : Average mflops/s per call per node       : 807339
-Grid       : Message        : Average mflops/s per call per node       : 823540
-Grid       : Message        : Average mflops/s per call per node       : 662882
-Grid       : Message        : Average mflops/s per call per node (full): 313029
-Grid       : Message        : Average mflops/s per call per node (full): 437000
-Grid       : Message        : Average mflops/s per call per node (full): 444704
-Grid       : Message        : Average mflops/s per call per node (full): 305305
-Grid       : Message        :  Stencil 13.985 GB/s per node
-Grid       : Message        :  Stencil 10.8433 GB/s per node
-Grid       : Message        :  Stencil 18.4694 GB/s per node
-Grid       : Message        :  Stencil 12.2394 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660780
-Grid       : Message        : Average mflops/s per call per node       : 811878
-Grid       : Message        : Average mflops/s per call per node       : 819535
-Grid       : Message        : Average mflops/s per call per node       : 663962
-Grid       : Message        : Average mflops/s per call per node (full): 315398
-Grid       : Message        : Average mflops/s per call per node (full): 335978
-Grid       : Message        : Average mflops/s per call per node (full): 447079
-Grid       : Message        : Average mflops/s per call per node (full): 299551
-Grid       : Message        :  Stencil 12.7344 GB/s per node
-Grid       : Message        :  Stencil 16.7795 GB/s per node
-Grid       : Message        :  Stencil 17.9974 GB/s per node
-Grid       : Message        :  Stencil 12.3923 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666265
-Grid       : Message        : Average mflops/s per call per node       : 806199
-Grid       : Message        : Average mflops/s per call per node       : 821853
-Grid       : Message        : Average mflops/s per call per node       : 665136
-Grid       : Message        : Average mflops/s per call per node (full): 314563
-Grid       : Message        : Average mflops/s per call per node (full): 437706
-Grid       : Message        : Average mflops/s per call per node (full): 445625
-Grid       : Message        : Average mflops/s per call per node (full): 304797
-Grid       : Message        :  Stencil 13.5227 GB/s per node
-Grid       : Message        :  Stencil 9.9904 GB/s per node
-Grid       : Message        :  Stencil 18.1087 GB/s per node
-Grid       : Message        :  Stencil 13.3208 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665052
-Grid       : Message        : Average mflops/s per call per node       : 806549
-Grid       : Message        : Average mflops/s per call per node       : 820800
-Grid       : Message        : Average mflops/s per call per node       : 668015
-Grid       : Message        : Average mflops/s per call per node (full): 314606
-Grid       : Message        : Average mflops/s per call per node (full): 316705
-Grid       : Message        : Average mflops/s per call per node (full): 446817
-Grid       : Message        : Average mflops/s per call per node (full): 306683
-Grid       : Message        :  Stencil 14.0079 GB/s per node
-Grid       : Message        :  Stencil 10.2811 GB/s per node
-Grid       : Message        :  Stencil 17.4514 GB/s per node
-Grid       : Message        :  Stencil 12.1589 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664551
-Grid       : Message        : Average mflops/s per call per node       : 807886
-Grid       : Message        : Average mflops/s per call per node       : 823665
-Grid       : Message        : Average mflops/s per call per node       : 668830
-Grid       : Message        : Average mflops/s per call per node (full): 315811
-Grid       : Message        : Average mflops/s per call per node (full): 322821
-Grid       : Message        : Average mflops/s per call per node (full): 444862
-Grid       : Message        : Average mflops/s per call per node (full): 303960
-Grid       : Message        :  Stencil 12.9543 GB/s per node
-Grid       : Message        :  Stencil 17.2676 GB/s per node
-Grid       : Message        :  Stencil 17.5014 GB/s per node
-Grid       : Message        :  Stencil 12.4183 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666039
-Grid       : Message        : Average mflops/s per call per node       : 801726
-Grid       : Message        : Average mflops/s per call per node       : 818973
-Grid       : Message        : Average mflops/s per call per node       : 668015
-Grid       : Message        : Average mflops/s per call per node (full): 312869
-Grid       : Message        : Average mflops/s per call per node (full): 434229
-Grid       : Message        : Average mflops/s per call per node (full): 444635
-Grid       : Message        : Average mflops/s per call per node (full): 304935
-Grid       : Message        :  Stencil 12.5383 GB/s per node
-Grid       : Message        :  Stencil 14.5256 GB/s per node
-Grid       : Message        :  Stencil 18.1822 GB/s per node
-Grid       : Message        :  Stencil 12.5818 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669210
-Grid       : Message        : Average mflops/s per call per node       : 807821
-Grid       : Message        : Average mflops/s per call per node       : 822054
-Grid       : Message        : Average mflops/s per call per node       : 664072
-Grid       : Message        : Average mflops/s per call per node (full): 313947
-Grid       : Message        : Average mflops/s per call per node (full): 408152
-Grid       : Message        : Average mflops/s per call per node (full): 446600
-Grid       : Message        : Average mflops/s per call per node (full): 304352
-Grid       : Message        :  Stencil 13.144 GB/s per node
-Grid       : Message        :  Stencil 17.0589 GB/s per node
-Grid       : Message        :  Stencil 17.9078 GB/s per node
-Grid       : Message        :  Stencil 12.6028 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663790
-Grid       : Message        : Average mflops/s per call per node       : 801031
-Grid       : Message        : Average mflops/s per call per node       : 826611
-Grid       : Message        : Average mflops/s per call per node       : 668085
-Grid       : Message        : Average mflops/s per call per node (full): 314625
-Grid       : Message        : Average mflops/s per call per node (full): 435639
-Grid       : Message        : Average mflops/s per call per node (full): 447655
-Grid       : Message        : Average mflops/s per call per node (full): 304941
-Grid       : Message        :  Stencil 13.2036 GB/s per node
-Grid       : Message        :  Stencil 16.4167 GB/s per node
-Grid       : Message        :  Stencil 18.1195 GB/s per node
-Grid       : Message        :  Stencil 12.1919 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668091
-Grid       : Message        : Average mflops/s per call per node       : 804708
-Grid       : Message        : Average mflops/s per call per node       : 826018
-Grid       : Message        : Average mflops/s per call per node       : 665139
-Grid       : Message        : Average mflops/s per call per node (full): 316459
-Grid       : Message        : Average mflops/s per call per node (full): 432613
-Grid       : Message        : Average mflops/s per call per node (full): 447761
-Grid       : Message        : Average mflops/s per call per node (full): 303228
-Grid       : Message        :  Stencil 12.9496 GB/s per node
-Grid       : Message        :  Stencil 8.04132 GB/s per node
-Grid       : Message        :  Stencil 18.2398 GB/s per node
-Grid       : Message        :  Stencil 12.7744 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666903
-Grid       : Message        : Average mflops/s per call per node       : 812769
-Grid       : Message        : Average mflops/s per call per node       : 812983
-Grid       : Message        : Average mflops/s per call per node       : 666973
-Grid       : Message        : Average mflops/s per call per node (full): 312722
-Grid       : Message        : Average mflops/s per call per node (full): 267572
-Grid       : Message        : Average mflops/s per call per node (full): 445639
-Grid       : Message        : Average mflops/s per call per node (full): 305841
-Grid       : Message        :  Stencil 12.3602 GB/s per node
-Grid       : Message        :  Stencil 16.9159 GB/s per node
-Grid       : Message        :  Stencil 18.6838 GB/s per node
-Grid       : Message        :  Stencil 12.9963 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669015
-Grid       : Message        : Average mflops/s per call per node       : 807804
-Grid       : Message        : Average mflops/s per call per node       : 824645
-Grid       : Message        : Average mflops/s per call per node       : 665724
-Grid       : Message        : Average mflops/s per call per node (full): 312875
-Grid       : Message        : Average mflops/s per call per node (full): 432958
-Grid       : Message        : Average mflops/s per call per node (full): 450242
-Grid       : Message        : Average mflops/s per call per node (full): 305161
-Grid       : Message        :  Stencil 13.4647 GB/s per node
-Grid       : Message        :  Stencil 16.3383 GB/s per node
-Grid       : Message        :  Stencil 17.0735 GB/s per node
-Grid       : Message        :  Stencil 12.7368 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668122
-Grid       : Message        : Average mflops/s per call per node       : 803149
-Grid       : Message        : Average mflops/s per call per node       : 825487
-Grid       : Message        : Average mflops/s per call per node       : 667634
-Grid       : Message        : Average mflops/s per call per node (full): 315650
-Grid       : Message        : Average mflops/s per call per node (full): 432558
-Grid       : Message        : Average mflops/s per call per node (full): 441983
-Grid       : Message        : Average mflops/s per call per node (full): 305838
-Grid       : Message        :  Stencil 13.4652 GB/s per node
-Grid       : Message        :  Stencil 16.738 GB/s per node
-Grid       : Message        :  Stencil 17.9649 GB/s per node
-Grid       : Message        :  Stencil 12.6534 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664658
-Grid       : Message        : Average mflops/s per call per node       : 800925
-Grid       : Message        : Average mflops/s per call per node       : 821484
-Grid       : Message        : Average mflops/s per call per node       : 665831
-Grid       : Message        : Average mflops/s per call per node (full): 315984
-Grid       : Message        : Average mflops/s per call per node (full): 435073
-Grid       : Message        : Average mflops/s per call per node (full): 445590
-Grid       : Message        : Average mflops/s per call per node (full): 305339
-Grid       : Message        :  Stencil 13.2929 GB/s per node
-Grid       : Message        :  Stencil 17.0294 GB/s per node
-Grid       : Message        :  Stencil 17.992 GB/s per node
-Grid       : Message        :  Stencil 12.0384 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665322
-Grid       : Message        : Average mflops/s per call per node       : 806397
-Grid       : Message        : Average mflops/s per call per node       : 817346
-Grid       : Message        : Average mflops/s per call per node       : 671984
-Grid       : Message        : Average mflops/s per call per node (full): 316333
-Grid       : Message        : Average mflops/s per call per node (full): 438125
-Grid       : Message        : Average mflops/s per call per node (full): 444428
-Grid       : Message        : Average mflops/s per call per node (full): 302906
-Grid       : Message        :  Stencil 12.8307 GB/s per node
-Grid       : Message        :  Stencil 17.5075 GB/s per node
-Grid       : Message        :  Stencil 17.4408 GB/s per node
-Grid       : Message        :  Stencil 13.0046 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 672324
-Grid       : Message        : Average mflops/s per call per node       : 798334
-Grid       : Message        : Average mflops/s per call per node       : 817300
-Grid       : Message        : Average mflops/s per call per node       : 664807
-Grid       : Message        : Average mflops/s per call per node (full): 316035
-Grid       : Message        : Average mflops/s per call per node (full): 435968
-Grid       : Message        : Average mflops/s per call per node (full): 439149
-Grid       : Message        : Average mflops/s per call per node (full): 306540
-Grid       : Message        :  Stencil 12.664 GB/s per node
-Grid       : Message        :  Stencil 16.3831 GB/s per node
-Grid       : Message        :  Stencil 18.4677 GB/s per node
-Grid       : Message        :  Stencil 12.9777 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 672242
-Grid       : Message        : Average mflops/s per call per node       : 806043
-Grid       : Message        : Average mflops/s per call per node       : 823586
-Grid       : Message        : Average mflops/s per call per node       : 668304
-Grid       : Message        : Average mflops/s per call per node (full): 315131
-Grid       : Message        : Average mflops/s per call per node (full): 433011
-Grid       : Message        : Average mflops/s per call per node (full): 447910
-Grid       : Message        : Average mflops/s per call per node (full): 304572
-Grid       : Message        :  Stencil 12.647 GB/s per node
-Grid       : Message        :  Stencil 17.7233 GB/s per node
-Grid       : Message        :  Stencil 19.045 GB/s per node
-Grid       : Message        :  Stencil 14.4043 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670068
-Grid       : Message        : Average mflops/s per call per node       : 802508
-Grid       : Message        : Average mflops/s per call per node       : 822255
-Grid       : Message        : Average mflops/s per call per node       : 660874
-Grid       : Message        : Average mflops/s per call per node (full): 314382
-Grid       : Message        : Average mflops/s per call per node (full): 440019
-Grid       : Message        : Average mflops/s per call per node (full): 449501
-Grid       : Message        : Average mflops/s per call per node (full): 307181
-Grid       : Message        :  Stencil 12.933 GB/s per node
-Grid       : Message        :  Stencil 16.7577 GB/s per node
-Grid       : Message        :  Stencil 17.9727 GB/s per node
-Grid       : Message        :  Stencil 13.1741 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667285
-Grid       : Message        : Average mflops/s per call per node       : 807065
-Grid       : Message        : Average mflops/s per call per node       : 816317
-Grid       : Message        : Average mflops/s per call per node       : 660677
-Grid       : Message        : Average mflops/s per call per node (full): 313925
-Grid       : Message        : Average mflops/s per call per node (full): 437370
-Grid       : Message        : Average mflops/s per call per node (full): 445682
-Grid       : Message        : Average mflops/s per call per node (full): 304500
-Grid       : Message        :  Stencil 12.8929 GB/s per node
-Grid       : Message        :  Stencil 16.5216 GB/s per node
-Grid       : Message        :  Stencil 17.6462 GB/s per node
-Grid       : Message        :  Stencil 11.6618 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670583
-Grid       : Message        : Average mflops/s per call per node       : 809937
-Grid       : Message        : Average mflops/s per call per node       : 818639
-Grid       : Message        : Average mflops/s per call per node       : 669559
-Grid       : Message        : Average mflops/s per call per node (full): 314614
-Grid       : Message        : Average mflops/s per call per node (full): 434299
-Grid       : Message        : Average mflops/s per call per node (full): 444539
-Grid       : Message        : Average mflops/s per call per node (full): 298526
-Grid       : Message        :  Stencil 13.0396 GB/s per node
-Grid       : Message        :  Stencil 16.8357 GB/s per node
-Grid       : Message        :  Stencil 17.251 GB/s per node
-Grid       : Message        :  Stencil 14.0986 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667395
-Grid       : Message        : Average mflops/s per call per node       : 807567
-Grid       : Message        : Average mflops/s per call per node       : 825726
-Grid       : Message        : Average mflops/s per call per node       : 661632
-Grid       : Message        : Average mflops/s per call per node (full): 314509
-Grid       : Message        : Average mflops/s per call per node (full): 437961
-Grid       : Message        : Average mflops/s per call per node (full): 443090
-Grid       : Message        : Average mflops/s per call per node (full): 303319
-Grid       : Message        :  Stencil 14.2244 GB/s per node
-Grid       : Message        :  Stencil 16.71 GB/s per node
-Grid       : Message        :  Stencil 17.8663 GB/s per node
-Grid       : Message        :  Stencil 13.1549 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663781
-Grid       : Message        : Average mflops/s per call per node       : 802430
-Grid       : Message        : Average mflops/s per call per node       : 822847
-Grid       : Message        : Average mflops/s per call per node       : 662736
-Grid       : Message        : Average mflops/s per call per node (full): 315901
-Grid       : Message        : Average mflops/s per call per node (full): 434370
-Grid       : Message        : Average mflops/s per call per node (full): 447766
-Grid       : Message        : Average mflops/s per call per node (full): 305592
-Grid       : Message        :  Stencil 13.8167 GB/s per node
-Grid       : Message        :  Stencil 16.4358 GB/s per node
-Grid       : Message        :  Stencil 17.2993 GB/s per node
-Grid       : Message        :  Stencil 12.7085 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661245
-Grid       : Message        : Average mflops/s per call per node       : 803105
-Grid       : Message        : Average mflops/s per call per node       : 823692
-Grid       : Message        : Average mflops/s per call per node       : 664362
-Grid       : Message        : Average mflops/s per call per node (full): 314367
-Grid       : Message        : Average mflops/s per call per node (full): 433581
-Grid       : Message        : Average mflops/s per call per node (full): 442903
-Grid       : Message        : Average mflops/s per call per node (full): 305447
-Grid       : Message        :  Stencil 13.1055 GB/s per node
-Grid       : Message        :  Stencil 9.4695 GB/s per node
-Grid       : Message        :  Stencil 17.4881 GB/s per node
-Grid       : Message        :  Stencil 13.3955 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664820
-Grid       : Message        : Average mflops/s per call per node       : 809060
-Grid       : Message        : Average mflops/s per call per node       : 822009
-Grid       : Message        : Average mflops/s per call per node       : 663366
-Grid       : Message        : Average mflops/s per call per node (full): 313768
-Grid       : Message        : Average mflops/s per call per node (full): 303820
-Grid       : Message        : Average mflops/s per call per node (full): 443817
-Grid       : Message        : Average mflops/s per call per node (full): 304245
-Grid       : Message        :  Stencil 13.5332 GB/s per node
-Grid       : Message        :  Stencil 14.8741 GB/s per node
-Grid       : Message        :  Stencil 17.4688 GB/s per node
-Grid       : Message        :  Stencil 12.3018 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663171
-Grid       : Message        : Average mflops/s per call per node       : 807899
-Grid       : Message        : Average mflops/s per call per node       : 816213
-Grid       : Message        : Average mflops/s per call per node       : 665794
-Grid       : Message        : Average mflops/s per call per node (full): 314666
-Grid       : Message        : Average mflops/s per call per node (full): 414230
-Grid       : Message        : Average mflops/s per call per node (full): 436437
-Grid       : Message        : Average mflops/s per call per node (full): 300328
-Grid       : Message        :  Stencil 13.4418 GB/s per node
-Grid       : Message        :  Stencil 16.7514 GB/s per node
-Grid       : Message        :  Stencil 18.5606 GB/s per node
-Grid       : Message        :  Stencil 13.6725 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663475
-Grid       : Message        : Average mflops/s per call per node       : 802921
-Grid       : Message        : Average mflops/s per call per node       : 820846
-Grid       : Message        : Average mflops/s per call per node       : 660621
-Grid       : Message        : Average mflops/s per call per node (full): 315321
-Grid       : Message        : Average mflops/s per call per node (full): 436917
-Grid       : Message        : Average mflops/s per call per node (full): 447510
-Grid       : Message        : Average mflops/s per call per node (full): 306476
-Grid       : Message        :  Stencil 13.2975 GB/s per node
-Grid       : Message        :  Stencil 16.294 GB/s per node
-Grid       : Message        :  Stencil 16.9049 GB/s per node
-Grid       : Message        :  Stencil 12.8291 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666696
-Grid       : Message        : Average mflops/s per call per node       : 804316
-Grid       : Message        : Average mflops/s per call per node       : 820510
-Grid       : Message        : Average mflops/s per call per node       : 664933
-Grid       : Message        : Average mflops/s per call per node (full): 315006
-Grid       : Message        : Average mflops/s per call per node (full): 429726
-Grid       : Message        : Average mflops/s per call per node (full): 439125
-Grid       : Message        : Average mflops/s per call per node (full): 305590
-Grid       : Message        :  Stencil 14.0116 GB/s per node
-Grid       : Message        :  Stencil 17.0097 GB/s per node
-Grid       : Message        :  Stencil 17.7337 GB/s per node
-Grid       : Message        :  Stencil 12.7172 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664280
-Grid       : Message        : Average mflops/s per call per node       : 803742
-Grid       : Message        : Average mflops/s per call per node       : 827551
-Grid       : Message        : Average mflops/s per call per node       : 665486
-Grid       : Message        : Average mflops/s per call per node (full): 316235
-Grid       : Message        : Average mflops/s per call per node (full): 438807
-Grid       : Message        : Average mflops/s per call per node (full): 447663
-Grid       : Message        : Average mflops/s per call per node (full): 304516
-Grid       : Message        :  Stencil 13.3899 GB/s per node
-Grid       : Message        :  Stencil 18.1307 GB/s per node
-Grid       : Message        :  Stencil 17.2902 GB/s per node
-Grid       : Message        :  Stencil 13.0768 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670032
-Grid       : Message        : Average mflops/s per call per node       : 801219
-Grid       : Message        : Average mflops/s per call per node       : 829903
-Grid       : Message        : Average mflops/s per call per node       : 655400
-Grid       : Message        : Average mflops/s per call per node (full): 316179
-Grid       : Message        : Average mflops/s per call per node (full): 438143
-Grid       : Message        : Average mflops/s per call per node (full): 444077
-Grid       : Message        : Average mflops/s per call per node (full): 304791
-Grid       : Message        :  Stencil 13.0342 GB/s per node
-Grid       : Message        :  Stencil 17.5561 GB/s per node
-Grid       : Message        :  Stencil 18.0282 GB/s per node
-Grid       : Message        :  Stencil 13.7715 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666334
-Grid       : Message        : Average mflops/s per call per node       : 799060
-Grid       : Message        : Average mflops/s per call per node       : 821650
-Grid       : Message        : Average mflops/s per call per node       : 662056
-Grid       : Message        : Average mflops/s per call per node (full): 315412
-Grid       : Message        : Average mflops/s per call per node (full): 437403
-Grid       : Message        : Average mflops/s per call per node (full): 446257
-Grid       : Message        : Average mflops/s per call per node (full): 307217
-Grid       : Message        :  Stencil 13.5037 GB/s per node
-Grid       : Message        :  Stencil 18.5172 GB/s per node
-Grid       : Message        :  Stencil 17.1817 GB/s per node
-Grid       : Message        :  Stencil 12.5464 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665393
-Grid       : Message        : Average mflops/s per call per node       : 799830
-Grid       : Message        : Average mflops/s per call per node       : 824258
-Grid       : Message        : Average mflops/s per call per node       : 671050
-Grid       : Message        : Average mflops/s per call per node (full): 315641
-Grid       : Message        : Average mflops/s per call per node (full): 441107
-Grid       : Message        : Average mflops/s per call per node (full): 441294
-Grid       : Message        : Average mflops/s per call per node (full): 305549
-Grid       : Message        :  Stencil 13.3831 GB/s per node
-Grid       : Message        :  Stencil 14.2564 GB/s per node
-Grid       : Message        :  Stencil 17.669 GB/s per node
-Grid       : Message        :  Stencil 13.9748 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668898
-Grid       : Message        : Average mflops/s per call per node       : 807392
-Grid       : Message        : Average mflops/s per call per node       : 823886
-Grid       : Message        : Average mflops/s per call per node       : 659980
-Grid       : Message        : Average mflops/s per call per node (full): 316380
-Grid       : Message        : Average mflops/s per call per node (full): 403824
-Grid       : Message        : Average mflops/s per call per node (full): 445627
-Grid       : Message        : Average mflops/s per call per node (full): 305922
-Grid       : Message        :  Stencil 13.0011 GB/s per node
-Grid       : Message        :  Stencil 16.4418 GB/s per node
-Grid       : Message        :  Stencil 17.5423 GB/s per node
-Grid       : Message        :  Stencil 12.2651 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667833
-Grid       : Message        : Average mflops/s per call per node       : 799772
-Grid       : Message        : Average mflops/s per call per node       : 825062
-Grid       : Message        : Average mflops/s per call per node       : 662430
-Grid       : Message        : Average mflops/s per call per node (full): 312990
-Grid       : Message        : Average mflops/s per call per node (full): 432380
-Grid       : Message        : Average mflops/s per call per node (full): 446061
-Grid       : Message        : Average mflops/s per call per node (full): 301703
-Grid       : Message        :  Stencil 13.122 GB/s per node
-Grid       : Message        :  Stencil 14.4751 GB/s per node
-Grid       : Message        :  Stencil 17.8534 GB/s per node
-Grid       : Message        :  Stencil 14.0202 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668032
-Grid       : Message        : Average mflops/s per call per node       : 801144
-Grid       : Message        : Average mflops/s per call per node       : 825762
-Grid       : Message        : Average mflops/s per call per node       : 665085
-Grid       : Message        : Average mflops/s per call per node (full): 315796
-Grid       : Message        : Average mflops/s per call per node (full): 405994
-Grid       : Message        : Average mflops/s per call per node (full): 446302
-Grid       : Message        : Average mflops/s per call per node (full): 306393
-Grid       : Message        :  Stencil 12.8932 GB/s per node
-Grid       : Message        :  Stencil 16.6554 GB/s per node
-Grid       : Message        :  Stencil 17.8851 GB/s per node
-Grid       : Message        :  Stencil 12.5967 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665908
-Grid       : Message        : Average mflops/s per call per node       : 808734
-Grid       : Message        : Average mflops/s per call per node       : 822333
-Grid       : Message        : Average mflops/s per call per node       : 662094
-Grid       : Message        : Average mflops/s per call per node (full): 314977
-Grid       : Message        : Average mflops/s per call per node (full): 434727
-Grid       : Message        : Average mflops/s per call per node (full): 446631
-Grid       : Message        : Average mflops/s per call per node (full): 304237
-Grid       : Message        :  Stencil 12.4235 GB/s per node
-Grid       : Message        :  Stencil 16.2135 GB/s per node
-Grid       : Message        :  Stencil 17.6276 GB/s per node
-Grid       : Message        :  Stencil 14.727 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666897
-Grid       : Message        : Average mflops/s per call per node       : 806787
-Grid       : Message        : Average mflops/s per call per node       : 825036
-Grid       : Message        : Average mflops/s per call per node       : 662896
-Grid       : Message        : Average mflops/s per call per node (full): 311217
-Grid       : Message        : Average mflops/s per call per node (full): 431113
-Grid       : Message        : Average mflops/s per call per node (full): 445192
-Grid       : Message        : Average mflops/s per call per node (full): 307938
-Grid       : Message        :  Stencil 12.8052 GB/s per node
-Grid       : Message        :  Stencil 16.6575 GB/s per node
-Grid       : Message        :  Stencil 17.5731 GB/s per node
-Grid       : Message        :  Stencil 13.5217 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665083
-Grid       : Message        : Average mflops/s per call per node       : 802066
-Grid       : Message        : Average mflops/s per call per node       : 822189
-Grid       : Message        : Average mflops/s per call per node       : 664236
-Grid       : Message        : Average mflops/s per call per node (full): 312239
-Grid       : Message        : Average mflops/s per call per node (full): 436940
-Grid       : Message        : Average mflops/s per call per node (full): 446146
-Grid       : Message        : Average mflops/s per call per node (full): 305529
-Grid       : Message        :  Stencil 12.0483 GB/s per node
-Grid       : Message        :  Stencil 17.8025 GB/s per node
-Grid       : Message        :  Stencil 17.3049 GB/s per node
-Grid       : Message        :  Stencil 11.5328 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668534
-Grid       : Message        : Average mflops/s per call per node       : 803686
-Grid       : Message        : Average mflops/s per call per node       : 815868
-Grid       : Message        : Average mflops/s per call per node       : 663886
-Grid       : Message        : Average mflops/s per call per node (full): 308896
-Grid       : Message        : Average mflops/s per call per node (full): 439722
-Grid       : Message        : Average mflops/s per call per node (full): 432427
-Grid       : Message        : Average mflops/s per call per node (full): 287890
-Grid       : Message        :  Stencil 12.676 GB/s per node
-Grid       : Message        :  Stencil 17.23 GB/s per node
-Grid       : Message        :  Stencil 17.5567 GB/s per node
-Grid       : Message        :  Stencil 12.3602 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662447
-Grid       : Message        : Average mflops/s per call per node       : 804064
-Grid       : Message        : Average mflops/s per call per node       : 822752
-Grid       : Message        : Average mflops/s per call per node       : 660888
-Grid       : Message        : Average mflops/s per call per node (full): 310141
-Grid       : Message        : Average mflops/s per call per node (full): 439871
-Grid       : Message        : Average mflops/s per call per node (full): 445413
-Grid       : Message        : Average mflops/s per call per node (full): 300370
-Grid       : Message        :  Stencil 12.3903 GB/s per node
-Grid       : Message        :  Stencil 13.9581 GB/s per node
-Grid       : Message        :  Stencil 17.4805 GB/s per node
-Grid       : Message        :  Stencil 13.8736 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666960
-Grid       : Message        : Average mflops/s per call per node       : 811450
-Grid       : Message        : Average mflops/s per call per node       : 818355
-Grid       : Message        : Average mflops/s per call per node       : 666749
-Grid       : Message        : Average mflops/s per call per node (full): 312948
-Grid       : Message        : Average mflops/s per call per node (full): 399590
-Grid       : Message        : Average mflops/s per call per node (full): 444163
-Grid       : Message        : Average mflops/s per call per node (full): 307748
-Grid       : Message        :  Stencil 12.2576 GB/s per node
-Grid       : Message        :  Stencil 16.9103 GB/s per node
-Grid       : Message        :  Stencil 17.6251 GB/s per node
-Grid       : Message        :  Stencil 13.1704 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669561
-Grid       : Message        : Average mflops/s per call per node       : 804700
-Grid       : Message        : Average mflops/s per call per node       : 822080
-Grid       : Message        : Average mflops/s per call per node       : 663517
-Grid       : Message        : Average mflops/s per call per node (full): 311869
-Grid       : Message        : Average mflops/s per call per node (full): 438410
-Grid       : Message        : Average mflops/s per call per node (full): 445679
-Grid       : Message        : Average mflops/s per call per node (full): 306253
-Grid       : Message        :  Stencil 13.1253 GB/s per node
-Grid       : Message        :  Stencil 17.0406 GB/s per node
-Grid       : Message        :  Stencil 17.4521 GB/s per node
-Grid       : Message        :  Stencil 12.5503 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663884
-Grid       : Message        : Average mflops/s per call per node       : 805130
-Grid       : Message        : Average mflops/s per call per node       : 822047
-Grid       : Message        : Average mflops/s per call per node       : 666019
-Grid       : Message        : Average mflops/s per call per node (full): 313695
-Grid       : Message        : Average mflops/s per call per node (full): 433054
-Grid       : Message        : Average mflops/s per call per node (full): 444033
-Grid       : Message        : Average mflops/s per call per node (full): 304935
-Grid       : Message        :  Stencil 14.2193 GB/s per node
-Grid       : Message        :  Stencil 16.7231 GB/s per node
-Grid       : Message        :  Stencil 17.4057 GB/s per node
-Grid       : Message        :  Stencil 12.8035 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659679
-Grid       : Message        : Average mflops/s per call per node       : 802868
-Grid       : Message        : Average mflops/s per call per node       : 822679
-Grid       : Message        : Average mflops/s per call per node       : 668637
-Grid       : Message        : Average mflops/s per call per node (full): 316201
-Grid       : Message        : Average mflops/s per call per node (full): 436200
-Grid       : Message        : Average mflops/s per call per node (full): 443477
-Grid       : Message        : Average mflops/s per call per node (full): 305573
-Grid       : Message        :  Stencil 12.5265 GB/s per node
-Grid       : Message        :  Stencil 11.3738 GB/s per node
-Grid       : Message        :  Stencil 17.8532 GB/s per node
-Grid       : Message        :  Stencil 12.9027 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664605
-Grid       : Message        : Average mflops/s per call per node       : 806182
-Grid       : Message        : Average mflops/s per call per node       : 817960
-Grid       : Message        : Average mflops/s per call per node       : 665454
-Grid       : Message        : Average mflops/s per call per node (full): 312821
-Grid       : Message        : Average mflops/s per call per node (full): 347998
-Grid       : Message        : Average mflops/s per call per node (full): 445111
-Grid       : Message        : Average mflops/s per call per node (full): 305415
-Grid       : Message        :  Stencil 12.8852 GB/s per node
-Grid       : Message        :  Stencil 15.8886 GB/s per node
-Grid       : Message        :  Stencil 17.8262 GB/s per node
-Grid       : Message        :  Stencil 14.1356 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666314
-Grid       : Message        : Average mflops/s per call per node       : 807479
-Grid       : Message        : Average mflops/s per call per node       : 819779
-Grid       : Message        : Average mflops/s per call per node       : 661020
-Grid       : Message        : Average mflops/s per call per node (full): 314064
-Grid       : Message        : Average mflops/s per call per node (full): 426703
-Grid       : Message        : Average mflops/s per call per node (full): 443793
-Grid       : Message        : Average mflops/s per call per node (full): 307501
-Grid       : Message        :  Stencil 13.3414 GB/s per node
-Grid       : Message        :  Stencil 18.0181 GB/s per node
-Grid       : Message        :  Stencil 17.9612 GB/s per node
-Grid       : Message        :  Stencil 13.7132 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662865
-Grid       : Message        : Average mflops/s per call per node       : 800521
-Grid       : Message        : Average mflops/s per call per node       : 831440
-Grid       : Message        : Average mflops/s per call per node       : 663001
-Grid       : Message        : Average mflops/s per call per node (full): 312104
-Grid       : Message        : Average mflops/s per call per node (full): 440473
-Grid       : Message        : Average mflops/s per call per node (full): 449193
-Grid       : Message        : Average mflops/s per call per node (full): 305778
-Grid       : Message        :  Stencil 12.4744 GB/s per node
-Grid       : Message        :  Stencil 12.4069 GB/s per node
-Grid       : Message        :  Stencil 18.6512 GB/s per node
-Grid       : Message        :  Stencil 13.9883 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668317
-Grid       : Message        : Average mflops/s per call per node       : 809376
-Grid       : Message        : Average mflops/s per call per node       : 824709
-Grid       : Message        : Average mflops/s per call per node       : 659530
-Grid       : Message        : Average mflops/s per call per node (full): 313873
-Grid       : Message        : Average mflops/s per call per node (full): 370099
-Grid       : Message        : Average mflops/s per call per node (full): 447544
-Grid       : Message        : Average mflops/s per call per node (full): 306571
-Grid       : Message        :  Stencil 13.0316 GB/s per node
-Grid       : Message        :  Stencil 16.6968 GB/s per node
-Grid       : Message        :  Stencil 17.7661 GB/s per node
-Grid       : Message        :  Stencil 14.0046 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666400
-Grid       : Message        : Average mflops/s per call per node       : 802491
-Grid       : Message        : Average mflops/s per call per node       : 821831
-Grid       : Message        : Average mflops/s per call per node       : 661370
-Grid       : Message        : Average mflops/s per call per node (full): 314349
-Grid       : Message        : Average mflops/s per call per node (full): 435465
-Grid       : Message        : Average mflops/s per call per node (full): 446423
-Grid       : Message        : Average mflops/s per call per node (full): 306860
-Grid       : Message        :  Stencil 12.8639 GB/s per node
-Grid       : Message        :  Stencil 16.4533 GB/s per node
-Grid       : Message        :  Stencil 17.201 GB/s per node
-Grid       : Message        :  Stencil 12.2991 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665471
-Grid       : Message        : Average mflops/s per call per node       : 801379
-Grid       : Message        : Average mflops/s per call per node       : 817508
-Grid       : Message        : Average mflops/s per call per node       : 663412
-Grid       : Message        : Average mflops/s per call per node (full): 314238
-Grid       : Message        : Average mflops/s per call per node (full): 433165
-Grid       : Message        : Average mflops/s per call per node (full): 441811
-Grid       : Message        : Average mflops/s per call per node (full): 303330
-Grid       : Message        :  Stencil 12.5565 GB/s per node
-Grid       : Message        :  Stencil 15.1478 GB/s per node
-Grid       : Message        :  Stencil 18.294 GB/s per node
-Grid       : Message        :  Stencil 14.0358 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668709
-Grid       : Message        : Average mflops/s per call per node       : 807656
-Grid       : Message        : Average mflops/s per call per node       : 821970
-Grid       : Message        : Average mflops/s per call per node       : 660987
-Grid       : Message        : Average mflops/s per call per node (full): 314130
-Grid       : Message        : Average mflops/s per call per node (full): 417627
-Grid       : Message        : Average mflops/s per call per node (full): 447522
-Grid       : Message        : Average mflops/s per call per node (full): 306286
-Grid       : Message        :  Stencil 13.4528 GB/s per node
-Grid       : Message        :  Stencil 16.5353 GB/s per node
-Grid       : Message        :  Stencil 16.4358 GB/s per node
-Grid       : Message        :  Stencil 14.2175 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667247
-Grid       : Message        : Average mflops/s per call per node       : 803885
-Grid       : Message        : Average mflops/s per call per node       : 824360
-Grid       : Message        : Average mflops/s per call per node       : 663661
-Grid       : Message        : Average mflops/s per call per node (full): 316220
-Grid       : Message        : Average mflops/s per call per node (full): 434687
-Grid       : Message        : Average mflops/s per call per node (full): 424565
-Grid       : Message        : Average mflops/s per call per node (full): 306818
-Grid       : Message        :  Stencil 12.9009 GB/s per node
-Grid       : Message        :  Stencil 16.6621 GB/s per node
-Grid       : Message        :  Stencil 17.2524 GB/s per node
-Grid       : Message        :  Stencil 12.7505 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665737
-Grid       : Message        : Average mflops/s per call per node       : 802449
-Grid       : Message        : Average mflops/s per call per node       : 824897
-Grid       : Message        : Average mflops/s per call per node       : 663450
-Grid       : Message        : Average mflops/s per call per node (full): 312619
-Grid       : Message        : Average mflops/s per call per node (full): 434975
-Grid       : Message        : Average mflops/s per call per node (full): 441741
-Grid       : Message        : Average mflops/s per call per node (full): 303787
-Grid       : Message        :  Stencil 13.8603 GB/s per node
-Grid       : Message        :  Stencil 17.3898 GB/s per node
-Grid       : Message        :  Stencil 18.0875 GB/s per node
-Grid       : Message        :  Stencil 12.4416 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661775
-Grid       : Message        : Average mflops/s per call per node       : 806436
-Grid       : Message        : Average mflops/s per call per node       : 824965
-Grid       : Message        : Average mflops/s per call per node       : 666546
-Grid       : Message        : Average mflops/s per call per node (full): 315060
-Grid       : Message        : Average mflops/s per call per node (full): 436662
-Grid       : Message        : Average mflops/s per call per node (full): 443945
-Grid       : Message        : Average mflops/s per call per node (full): 304778
-Grid       : Message        :  Stencil 13.7038 GB/s per node
-Grid       : Message        :  Stencil 17.6172 GB/s per node
-Grid       : Message        :  Stencil 18.4053 GB/s per node
-Grid       : Message        :  Stencil 12.6816 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663760
-Grid       : Message        : Average mflops/s per call per node       : 801627
-Grid       : Message        : Average mflops/s per call per node       : 825152
-Grid       : Message        : Average mflops/s per call per node       : 667139
-Grid       : Message        : Average mflops/s per call per node (full): 315050
-Grid       : Message        : Average mflops/s per call per node (full): 438868
-Grid       : Message        : Average mflops/s per call per node (full): 448242
-Grid       : Message        : Average mflops/s per call per node (full): 305717
-Grid       : Message        :  Stencil 12.9353 GB/s per node
-Grid       : Message        :  Stencil 16.0852 GB/s per node
-Grid       : Message        :  Stencil 17.3692 GB/s per node
-Grid       : Message        :  Stencil 12.9265 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669251
-Grid       : Message        : Average mflops/s per call per node       : 805532
-Grid       : Message        : Average mflops/s per call per node       : 826799
-Grid       : Message        : Average mflops/s per call per node       : 663717
-Grid       : Message        : Average mflops/s per call per node (full): 313709
-Grid       : Message        : Average mflops/s per call per node (full): 424664
-Grid       : Message        : Average mflops/s per call per node (full): 445421
-Grid       : Message        : Average mflops/s per call per node (full): 305257
-Grid       : Message        :  Stencil 13.379 GB/s per node
-Grid       : Message        :  Stencil 16.9515 GB/s per node
-Grid       : Message        :  Stencil 18.3665 GB/s per node
-Grid       : Message        :  Stencil 12.7137 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665154
-Grid       : Message        : Average mflops/s per call per node       : 802761
-Grid       : Message        : Average mflops/s per call per node       : 818096
-Grid       : Message        : Average mflops/s per call per node       : 664144
-Grid       : Message        : Average mflops/s per call per node (full): 315941
-Grid       : Message        : Average mflops/s per call per node (full): 438033
-Grid       : Message        : Average mflops/s per call per node (full): 446798
-Grid       : Message        : Average mflops/s per call per node (full): 304773
-Grid       : Message        :  Stencil 13.3168 GB/s per node
-Grid       : Message        :  Stencil 10.2857 GB/s per node
-Grid       : Message        :  Stencil 17.1783 GB/s per node
-Grid       : Message        :  Stencil 12.9949 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665785
-Grid       : Message        : Average mflops/s per call per node       : 810825
-Grid       : Message        : Average mflops/s per call per node       : 823353
-Grid       : Message        : Average mflops/s per call per node       : 665362
-Grid       : Message        : Average mflops/s per call per node (full): 315764
-Grid       : Message        : Average mflops/s per call per node (full): 323579
-Grid       : Message        : Average mflops/s per call per node (full): 443309
-Grid       : Message        : Average mflops/s per call per node (full): 305253
-Grid       : Message        :  Stencil 13.9634 GB/s per node
-Grid       : Message        :  Stencil 17.0929 GB/s per node
-Grid       : Message        :  Stencil 18.7741 GB/s per node
-Grid       : Message        :  Stencil 12.8815 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666241
-Grid       : Message        : Average mflops/s per call per node       : 809735
-Grid       : Message        : Average mflops/s per call per node       : 813012
-Grid       : Message        : Average mflops/s per call per node       : 661935
-Grid       : Message        : Average mflops/s per call per node (full): 316880
-Grid       : Message        : Average mflops/s per call per node (full): 438139
-Grid       : Message        : Average mflops/s per call per node (full): 445881
-Grid       : Message        : Average mflops/s per call per node (full): 305467
-Grid       : Message        :  Stencil 13.1792 GB/s per node
-Grid       : Message        :  Stencil 17.8518 GB/s per node
-Grid       : Message        :  Stencil 17.7641 GB/s per node
-Grid       : Message        :  Stencil 13.0227 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667995
-Grid       : Message        : Average mflops/s per call per node       : 806047
-Grid       : Message        : Average mflops/s per call per node       : 825949
-Grid       : Message        : Average mflops/s per call per node       : 663069
-Grid       : Message        : Average mflops/s per call per node (full): 314623
-Grid       : Message        : Average mflops/s per call per node (full): 440785
-Grid       : Message        : Average mflops/s per call per node (full): 446951
-Grid       : Message        : Average mflops/s per call per node (full): 305984
-Grid       : Message        :  Stencil 12.8287 GB/s per node
-Grid       : Message        :  Stencil 16.6164 GB/s per node
-Grid       : Message        :  Stencil 18.9068 GB/s per node
-Grid       : Message        :  Stencil 12.0297 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668841
-Grid       : Message        : Average mflops/s per call per node       : 803523
-Grid       : Message        : Average mflops/s per call per node       : 818370
-Grid       : Message        : Average mflops/s per call per node       : 666093
-Grid       : Message        : Average mflops/s per call per node (full): 315297
-Grid       : Message        : Average mflops/s per call per node (full): 435944
-Grid       : Message        : Average mflops/s per call per node (full): 447388
-Grid       : Message        : Average mflops/s per call per node (full): 302011
-Grid       : Message        :  Stencil 13.2067 GB/s per node
-Grid       : Message        :  Stencil 17.3298 GB/s per node
-Grid       : Message        :  Stencil 17.6517 GB/s per node
-Grid       : Message        :  Stencil 12.6095 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666902
-Grid       : Message        : Average mflops/s per call per node       : 800780
-Grid       : Message        : Average mflops/s per call per node       : 827430
-Grid       : Message        : Average mflops/s per call per node       : 670098
-Grid       : Message        : Average mflops/s per call per node (full): 315707
-Grid       : Message        : Average mflops/s per call per node (full): 438855
-Grid       : Message        : Average mflops/s per call per node (full): 446502
-Grid       : Message        : Average mflops/s per call per node (full): 305800
-Grid       : Message        :  Stencil 12.8391 GB/s per node
-Grid       : Message        :  Stencil 18.1332 GB/s per node
-Grid       : Message        :  Stencil 18.4131 GB/s per node
-Grid       : Message        :  Stencil 13.6612 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667111
-Grid       : Message        : Average mflops/s per call per node       : 800736
-Grid       : Message        : Average mflops/s per call per node       : 825266
-Grid       : Message        : Average mflops/s per call per node       : 659026
-Grid       : Message        : Average mflops/s per call per node (full): 315263
-Grid       : Message        : Average mflops/s per call per node (full): 440530
-Grid       : Message        : Average mflops/s per call per node (full): 448816
-Grid       : Message        : Average mflops/s per call per node (full): 306118
-Grid       : Message        :  Stencil 13.2343 GB/s per node
-Grid       : Message        :  Stencil 16.9053 GB/s per node
-Grid       : Message        :  Stencil 18.1019 GB/s per node
-Grid       : Message        :  Stencil 12.6898 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666497
-Grid       : Message        : Average mflops/s per call per node       : 808899
-Grid       : Message        : Average mflops/s per call per node       : 821718
-Grid       : Message        : Average mflops/s per call per node       : 663943
-Grid       : Message        : Average mflops/s per call per node (full): 316154
-Grid       : Message        : Average mflops/s per call per node (full): 437821
-Grid       : Message        : Average mflops/s per call per node (full): 447299
-Grid       : Message        : Average mflops/s per call per node (full): 302565
-Grid       : Message        :  Stencil 13.6119 GB/s per node
-Grid       : Message        :  Stencil 16.4649 GB/s per node
-Grid       : Message        :  Stencil 16.6632 GB/s per node
-Grid       : Message        :  Stencil 13.1453 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665911
-Grid       : Message        : Average mflops/s per call per node       : 802839
-Grid       : Message        : Average mflops/s per call per node       : 822157
-Grid       : Message        : Average mflops/s per call per node       : 659798
-Grid       : Message        : Average mflops/s per call per node (full): 317175
-Grid       : Message        : Average mflops/s per call per node (full): 434421
-Grid       : Message        : Average mflops/s per call per node (full): 426729
-Grid       : Message        : Average mflops/s per call per node (full): 301588
-Grid       : Message        :  Stencil 13.6702 GB/s per node
-Grid       : Message        :  Stencil 16.2963 GB/s per node
-Grid       : Message        :  Stencil 18.2252 GB/s per node
-Grid       : Message        :  Stencil 12.2345 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666119
-Grid       : Message        : Average mflops/s per call per node       : 803426
-Grid       : Message        : Average mflops/s per call per node       : 823702
-Grid       : Message        : Average mflops/s per call per node       : 668199
-Grid       : Message        : Average mflops/s per call per node (full): 317097
-Grid       : Message        : Average mflops/s per call per node (full): 431819
-Grid       : Message        : Average mflops/s per call per node (full): 447193
-Grid       : Message        : Average mflops/s per call per node (full): 303845
-Grid       : Message        :  Stencil 13.3411 GB/s per node
-Grid       : Message        :  Stencil 17.1795 GB/s per node
-Grid       : Message        :  Stencil 18.0849 GB/s per node
-Grid       : Message        :  Stencil 12.496 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668590
-Grid       : Message        : Average mflops/s per call per node       : 800821
-Grid       : Message        : Average mflops/s per call per node       : 829262
-Grid       : Message        : Average mflops/s per call per node       : 664689
-Grid       : Message        : Average mflops/s per call per node (full): 316188
-Grid       : Message        : Average mflops/s per call per node (full): 438769
-Grid       : Message        : Average mflops/s per call per node (full): 448396
-Grid       : Message        : Average mflops/s per call per node (full): 304490
-Grid       : Message        :  Stencil 14.7099 GB/s per node
-Grid       : Message        :  Stencil 17.1644 GB/s per node
-Grid       : Message        :  Stencil 17.1071 GB/s per node
-Grid       : Message        :  Stencil 12.4515 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663658
-Grid       : Message        : Average mflops/s per call per node       : 804636
-Grid       : Message        : Average mflops/s per call per node       : 821849
-Grid       : Message        : Average mflops/s per call per node       : 659873
-Grid       : Message        : Average mflops/s per call per node (full): 317243
-Grid       : Message        : Average mflops/s per call per node (full): 435777
-Grid       : Message        : Average mflops/s per call per node (full): 438993
-Grid       : Message        : Average mflops/s per call per node (full): 304428
-Grid       : Message        :  Stencil 14.5401 GB/s per node
-Grid       : Message        :  Stencil 17.653 GB/s per node
-Grid       : Message        :  Stencil 16.8231 GB/s per node
-Grid       : Message        :  Stencil 12.4649 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663737
-Grid       : Message        : Average mflops/s per call per node       : 799291
-Grid       : Message        : Average mflops/s per call per node       : 820476
-Grid       : Message        : Average mflops/s per call per node       : 659852
-Grid       : Message        : Average mflops/s per call per node (full): 316420
-Grid       : Message        : Average mflops/s per call per node (full): 437281
-Grid       : Message        : Average mflops/s per call per node (full): 437689
-Grid       : Message        : Average mflops/s per call per node (full): 302147
-Grid       : Message        :  Stencil 12.8809 GB/s per node
-Grid       : Message        :  Stencil 17.1122 GB/s per node
-Grid       : Message        :  Stencil 17.9883 GB/s per node
-Grid       : Message        :  Stencil 12.297 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666660
-Grid       : Message        : Average mflops/s per call per node       : 804044
-Grid       : Message        : Average mflops/s per call per node       : 821177
-Grid       : Message        : Average mflops/s per call per node       : 665370
-Grid       : Message        : Average mflops/s per call per node (full): 314404
-Grid       : Message        : Average mflops/s per call per node (full): 435744
-Grid       : Message        : Average mflops/s per call per node (full): 445622
-Grid       : Message        : Average mflops/s per call per node (full): 303978
-Grid       : Message        :  Stencil 13.4031 GB/s per node
-Grid       : Message        :  Stencil 16.5617 GB/s per node
-Grid       : Message        :  Stencil 17.8289 GB/s per node
-Grid       : Message        :  Stencil 12.8881 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663272
-Grid       : Message        : Average mflops/s per call per node       : 802300
-Grid       : Message        : Average mflops/s per call per node       : 819337
-Grid       : Message        : Average mflops/s per call per node       : 663035
-Grid       : Message        : Average mflops/s per call per node (full): 315588
-Grid       : Message        : Average mflops/s per call per node (full): 435178
-Grid       : Message        : Average mflops/s per call per node (full): 445795
-Grid       : Message        : Average mflops/s per call per node (full): 301042
-Grid       : Message        :  Stencil 14.0041 GB/s per node
-Grid       : Message        :  Stencil 16.5044 GB/s per node
-Grid       : Message        :  Stencil 16.8942 GB/s per node
-Grid       : Message        :  Stencil 12.0931 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659529
-Grid       : Message        : Average mflops/s per call per node       : 800857
-Grid       : Message        : Average mflops/s per call per node       : 821965
-Grid       : Message        : Average mflops/s per call per node       : 668337
-Grid       : Message        : Average mflops/s per call per node (full): 313966
-Grid       : Message        : Average mflops/s per call per node (full): 434530
-Grid       : Message        : Average mflops/s per call per node (full): 438095
-Grid       : Message        : Average mflops/s per call per node (full): 303502
-Grid       : Message        :  Stencil 13.632 GB/s per node
-Grid       : Message        :  Stencil 16.2975 GB/s per node
-Grid       : Message        :  Stencil 17.9547 GB/s per node
-Grid       : Message        :  Stencil 13.4796 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663517
-Grid       : Message        : Average mflops/s per call per node       : 803902
-Grid       : Message        : Average mflops/s per call per node       : 816525
-Grid       : Message        : Average mflops/s per call per node       : 664545
-Grid       : Message        : Average mflops/s per call per node (full): 314656
-Grid       : Message        : Average mflops/s per call per node (full): 432209
-Grid       : Message        : Average mflops/s per call per node (full): 445822
-Grid       : Message        : Average mflops/s per call per node (full): 305938
-Grid       : Message        :  Stencil 13.1287 GB/s per node
-Grid       : Message        :  Stencil 16.983 GB/s per node
-Grid       : Message        :  Stencil 17.7084 GB/s per node
-Grid       : Message        :  Stencil 12.3646 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666280
-Grid       : Message        : Average mflops/s per call per node       : 802273
-Grid       : Message        : Average mflops/s per call per node       : 818273
-Grid       : Message        : Average mflops/s per call per node       : 669441
-Grid       : Message        : Average mflops/s per call per node (full): 315151
-Grid       : Message        : Average mflops/s per call per node (full): 433794
-Grid       : Message        : Average mflops/s per call per node (full): 444069
-Grid       : Message        : Average mflops/s per call per node (full): 304140
-Grid       : Message        :  Stencil 12.8293 GB/s per node
-Grid       : Message        :  Stencil 16.3603 GB/s per node
-Grid       : Message        :  Stencil 17.4713 GB/s per node
-Grid       : Message        :  Stencil 12.433 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666807
-Grid       : Message        : Average mflops/s per call per node       : 803321
-Grid       : Message        : Average mflops/s per call per node       : 827302
-Grid       : Message        : Average mflops/s per call per node       : 668460
-Grid       : Message        : Average mflops/s per call per node (full): 314888
-Grid       : Message        : Average mflops/s per call per node (full): 433450
-Grid       : Message        : Average mflops/s per call per node (full): 446392
-Grid       : Message        : Average mflops/s per call per node (full): 305275
-Grid       : Message        :  Stencil 13.0512 GB/s per node
-Grid       : Message        :  Stencil 17.9731 GB/s per node
-Grid       : Message        :  Stencil 17.1651 GB/s per node
-Grid       : Message        :  Stencil 11.6682 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664991
-Grid       : Message        : Average mflops/s per call per node       : 805192
-Grid       : Message        : Average mflops/s per call per node       : 817573
-Grid       : Message        : Average mflops/s per call per node       : 662967
-Grid       : Message        : Average mflops/s per call per node (full): 315623
-Grid       : Message        : Average mflops/s per call per node (full): 441507
-Grid       : Message        : Average mflops/s per call per node (full): 441601
-Grid       : Message        : Average mflops/s per call per node (full): 298035
-Grid       : Message        :  Stencil 13.253 GB/s per node
-Grid       : Message        :  Stencil 16.3774 GB/s per node
-Grid       : Message        :  Stencil 18.6875 GB/s per node
-Grid       : Message        :  Stencil 12.2165 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667799
-Grid       : Message        : Average mflops/s per call per node       : 803575
-Grid       : Message        : Average mflops/s per call per node       : 818975
-Grid       : Message        : Average mflops/s per call per node       : 670247
-Grid       : Message        : Average mflops/s per call per node (full): 316219
-Grid       : Message        : Average mflops/s per call per node (full): 433350
-Grid       : Message        : Average mflops/s per call per node (full): 448136
-Grid       : Message        : Average mflops/s per call per node (full): 304675
-Grid       : Message        :  Stencil 14.3041 GB/s per node
-Grid       : Message        :  Stencil 17.0718 GB/s per node
-Grid       : Message        :  Stencil 17.9263 GB/s per node
-Grid       : Message        :  Stencil 12.7822 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665593
-Grid       : Message        : Average mflops/s per call per node       : 805329
-Grid       : Message        : Average mflops/s per call per node       : 826751
-Grid       : Message        : Average mflops/s per call per node       : 664358
-Grid       : Message        : Average mflops/s per call per node (full): 316931
-Grid       : Message        : Average mflops/s per call per node (full): 437511
-Grid       : Message        : Average mflops/s per call per node (full): 442404
-Grid       : Message        : Average mflops/s per call per node (full): 303260
-Grid       : Message        :  Stencil 12.5125 GB/s per node
-Grid       : Message        :  Stencil 16.6135 GB/s per node
-Grid       : Message        :  Stencil 17.3698 GB/s per node
-Grid       : Message        :  Stencil 13.3542 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669120
-Grid       : Message        : Average mflops/s per call per node       : 801584
-Grid       : Message        : Average mflops/s per call per node       : 823063
-Grid       : Message        : Average mflops/s per call per node       : 659636
-Grid       : Message        : Average mflops/s per call per node (full): 314085
-Grid       : Message        : Average mflops/s per call per node (full): 434002
-Grid       : Message        : Average mflops/s per call per node (full): 443555
-Grid       : Message        : Average mflops/s per call per node (full): 305938
-Grid       : Message        :  Stencil 12.6212 GB/s per node
-Grid       : Message        :  Stencil 16.2682 GB/s per node
-Grid       : Message        :  Stencil 19.0768 GB/s per node
-Grid       : Message        :  Stencil 12.0519 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669969
-Grid       : Message        : Average mflops/s per call per node       : 806422
-Grid       : Message        : Average mflops/s per call per node       : 824098
-Grid       : Message        : Average mflops/s per call per node       : 668735
-Grid       : Message        : Average mflops/s per call per node (full): 315226
-Grid       : Message        : Average mflops/s per call per node (full): 432388
-Grid       : Message        : Average mflops/s per call per node (full): 450312
-Grid       : Message        : Average mflops/s per call per node (full): 301944
-Grid       : Message        :  Stencil 13.7274 GB/s per node
-Grid       : Message        :  Stencil 17.1047 GB/s per node
-Grid       : Message        :  Stencil 17.2058 GB/s per node
-Grid       : Message        :  Stencil 12.2715 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664176
-Grid       : Message        : Average mflops/s per call per node       : 801792
-Grid       : Message        : Average mflops/s per call per node       : 823566
-Grid       : Message        : Average mflops/s per call per node       : 667132
-Grid       : Message        : Average mflops/s per call per node (full): 315721
-Grid       : Message        : Average mflops/s per call per node (full): 437657
-Grid       : Message        : Average mflops/s per call per node (full): 442652
-Grid       : Message        : Average mflops/s per call per node (full): 304294
-Grid       : Message        :  Stencil 12.5236 GB/s per node
-Grid       : Message        :  Stencil 16.2429 GB/s per node
-Grid       : Message        :  Stencil 17.4594 GB/s per node
-Grid       : Message        :  Stencil 12.6165 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669461
-Grid       : Message        : Average mflops/s per call per node       : 808103
-Grid       : Message        : Average mflops/s per call per node       : 820303
-Grid       : Message        : Average mflops/s per call per node       : 665400
-Grid       : Message        : Average mflops/s per call per node (full): 311139
-Grid       : Message        : Average mflops/s per call per node (full): 428508
-Grid       : Message        : Average mflops/s per call per node (full): 444642
-Grid       : Message        : Average mflops/s per call per node (full): 303657
-Grid       : Message        :  Stencil 13.1776 GB/s per node
-Grid       : Message        :  Stencil 14.7654 GB/s per node
-Grid       : Message        :  Stencil 17.2279 GB/s per node
-Grid       : Message        :  Stencil 13.697 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667935
-Grid       : Message        : Average mflops/s per call per node       : 801285
-Grid       : Message        : Average mflops/s per call per node       : 823621
-Grid       : Message        : Average mflops/s per call per node       : 665857
-Grid       : Message        : Average mflops/s per call per node (full): 315134
-Grid       : Message        : Average mflops/s per call per node (full): 411898
-Grid       : Message        : Average mflops/s per call per node (full): 442485
-Grid       : Message        : Average mflops/s per call per node (full): 306877
-Grid       : Message        :  Stencil 13.3373 GB/s per node
-Grid       : Message        :  Stencil 17.5889 GB/s per node
-Grid       : Message        :  Stencil 18.0192 GB/s per node
-Grid       : Message        :  Stencil 13.6967 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669391
-Grid       : Message        : Average mflops/s per call per node       : 807592
-Grid       : Message        : Average mflops/s per call per node       : 820767
-Grid       : Message        : Average mflops/s per call per node       : 662345
-Grid       : Message        : Average mflops/s per call per node (full): 315870
-Grid       : Message        : Average mflops/s per call per node (full): 439749
-Grid       : Message        : Average mflops/s per call per node (full): 446580
-Grid       : Message        : Average mflops/s per call per node (full): 304171
-Grid       : Message        :  Stencil 13.801 GB/s per node
-Grid       : Message        :  Stencil 16.7288 GB/s per node
-Grid       : Message        :  Stencil 17.6663 GB/s per node
-Grid       : Message        :  Stencil 12.7834 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664427
-Grid       : Message        : Average mflops/s per call per node       : 804456
-Grid       : Message        : Average mflops/s per call per node       : 822575
-Grid       : Message        : Average mflops/s per call per node       : 667446
-Grid       : Message        : Average mflops/s per call per node (full): 315898
-Grid       : Message        : Average mflops/s per call per node (full): 437236
-Grid       : Message        : Average mflops/s per call per node (full): 445612
-Grid       : Message        : Average mflops/s per call per node (full): 304453
-Grid       : Message        :  Stencil 14.3359 GB/s per node
-Grid       : Message        :  Stencil 16.4792 GB/s per node
-Grid       : Message        :  Stencil 16.8484 GB/s per node
-Grid       : Message        :  Stencil 11.8385 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665692
-Grid       : Message        : Average mflops/s per call per node       : 802731
-Grid       : Message        : Average mflops/s per call per node       : 820411
-Grid       : Message        : Average mflops/s per call per node       : 668575
-Grid       : Message        : Average mflops/s per call per node (full): 315999
-Grid       : Message        : Average mflops/s per call per node (full): 434040
-Grid       : Message        : Average mflops/s per call per node (full): 435236
-Grid       : Message        : Average mflops/s per call per node (full): 300225
-Grid       : Message        :  Stencil 13.8964 GB/s per node
-Grid       : Message        :  Stencil 16.9391 GB/s per node
-Grid       : Message        :  Stencil 17.6979 GB/s per node
-Grid       : Message        :  Stencil 12.8515 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662069
-Grid       : Message        : Average mflops/s per call per node       : 803656
-Grid       : Message        : Average mflops/s per call per node       : 815813
-Grid       : Message        : Average mflops/s per call per node       : 665169
-Grid       : Message        : Average mflops/s per call per node (full): 314933
-Grid       : Message        : Average mflops/s per call per node (full): 437751
-Grid       : Message        : Average mflops/s per call per node (full): 445789
-Grid       : Message        : Average mflops/s per call per node (full): 304065
-Grid       : Message        :  Stencil 13.561 GB/s per node
-Grid       : Message        :  Stencil 17.9284 GB/s per node
-Grid       : Message        :  Stencil 17.2071 GB/s per node
-Grid       : Message        :  Stencil 13.2587 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662725
-Grid       : Message        : Average mflops/s per call per node       : 801438
-Grid       : Message        : Average mflops/s per call per node       : 814631
-Grid       : Message        : Average mflops/s per call per node       : 664032
-Grid       : Message        : Average mflops/s per call per node (full): 314547
-Grid       : Message        : Average mflops/s per call per node (full): 438516
-Grid       : Message        : Average mflops/s per call per node (full): 439489
-Grid       : Message        : Average mflops/s per call per node (full): 306046
-Grid       : Message        :  Stencil 12.3496 GB/s per node
-Grid       : Message        :  Stencil 18.8328 GB/s per node
-Grid       : Message        :  Stencil 18.8839 GB/s per node
-Grid       : Message        :  Stencil 12.5511 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665142
-Grid       : Message        : Average mflops/s per call per node       : 803796
-Grid       : Message        : Average mflops/s per call per node       : 813808
-Grid       : Message        : Average mflops/s per call per node       : 664365
-Grid       : Message        : Average mflops/s per call per node (full): 311260
-Grid       : Message        : Average mflops/s per call per node (full): 440747
-Grid       : Message        : Average mflops/s per call per node (full): 445725
-Grid       : Message        : Average mflops/s per call per node (full): 301717
-Grid       : Message        :  Stencil 13.3218 GB/s per node
-Grid       : Message        :  Stencil 16.7001 GB/s per node
-Grid       : Message        :  Stencil 17.7836 GB/s per node
-Grid       : Message        :  Stencil 13.2898 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667594
-Grid       : Message        : Average mflops/s per call per node       : 802337
-Grid       : Message        : Average mflops/s per call per node       : 821749
-Grid       : Message        : Average mflops/s per call per node       : 659994
-Grid       : Message        : Average mflops/s per call per node (full): 315952
-Grid       : Message        : Average mflops/s per call per node (full): 436347
-Grid       : Message        : Average mflops/s per call per node (full): 445341
-Grid       : Message        : Average mflops/s per call per node (full): 305925
-Grid       : Message        :  Stencil 13.3547 GB/s per node
-Grid       : Message        :  Stencil 16.3847 GB/s per node
-Grid       : Message        :  Stencil 18.0773 GB/s per node
-Grid       : Message        :  Stencil 12.2147 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663244
-Grid       : Message        : Average mflops/s per call per node       : 803493
-Grid       : Message        : Average mflops/s per call per node       : 820519
-Grid       : Message        : Average mflops/s per call per node       : 665691
-Grid       : Message        : Average mflops/s per call per node (full): 315608
-Grid       : Message        : Average mflops/s per call per node (full): 433807
-Grid       : Message        : Average mflops/s per call per node (full): 442851
-Grid       : Message        : Average mflops/s per call per node (full): 299870
-Grid       : Message        :  Stencil 13.4964 GB/s per node
-Grid       : Message        :  Stencil 17.4179 GB/s per node
-Grid       : Message        :  Stencil 17.2663 GB/s per node
-Grid       : Message        :  Stencil 12.9006 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665195
-Grid       : Message        : Average mflops/s per call per node       : 806421
-Grid       : Message        : Average mflops/s per call per node       : 821652
-Grid       : Message        : Average mflops/s per call per node       : 664788
-Grid       : Message        : Average mflops/s per call per node (full): 316373
-Grid       : Message        : Average mflops/s per call per node (full): 438593
-Grid       : Message        : Average mflops/s per call per node (full): 442859
-Grid       : Message        : Average mflops/s per call per node (full): 303968
-Grid       : Message        :  Stencil 13.7037 GB/s per node
-Grid       : Message        :  Stencil 16.746 GB/s per node
-Grid       : Message        :  Stencil 18.4778 GB/s per node
-Grid       : Message        :  Stencil 12.3692 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669746
-Grid       : Message        : Average mflops/s per call per node       : 801249
-Grid       : Message        : Average mflops/s per call per node       : 819217
-Grid       : Message        : Average mflops/s per call per node       : 671226
-Grid       : Message        : Average mflops/s per call per node (full): 316275
-Grid       : Message        : Average mflops/s per call per node (full): 434504
-Grid       : Message        : Average mflops/s per call per node (full): 447449
-Grid       : Message        : Average mflops/s per call per node (full): 303978
-Grid       : Message        :  Stencil 13.6722 GB/s per node
-Grid       : Message        :  Stencil 17.1809 GB/s per node
-Grid       : Message        :  Stencil 17.5852 GB/s per node
-Grid       : Message        :  Stencil 12.8785 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667196
-Grid       : Message        : Average mflops/s per call per node       : 808159
-Grid       : Message        : Average mflops/s per call per node       : 822944
-Grid       : Message        : Average mflops/s per call per node       : 660357
-Grid       : Message        : Average mflops/s per call per node (full): 315620
-Grid       : Message        : Average mflops/s per call per node (full): 437790
-Grid       : Message        : Average mflops/s per call per node (full): 445648
-Grid       : Message        : Average mflops/s per call per node (full): 304434
-Grid       : Message        :  Stencil 12.5679 GB/s per node
-Grid       : Message        :  Stencil 17.4199 GB/s per node
-Grid       : Message        :  Stencil 17.2132 GB/s per node
-Grid       : Message        :  Stencil 14.2392 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 672005
-Grid       : Message        : Average mflops/s per call per node       : 802189
-Grid       : Message        : Average mflops/s per call per node       : 828593
-Grid       : Message        : Average mflops/s per call per node       : 662159
-Grid       : Message        : Average mflops/s per call per node (full): 314643
-Grid       : Message        : Average mflops/s per call per node (full): 437588
-Grid       : Message        : Average mflops/s per call per node (full): 443841
-Grid       : Message        : Average mflops/s per call per node (full): 307739
-Grid       : Message        :  Stencil 13.6386 GB/s per node
-Grid       : Message        :  Stencil 17.1214 GB/s per node
-Grid       : Message        :  Stencil 18.3172 GB/s per node
-Grid       : Message        :  Stencil 13.1393 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666778
-Grid       : Message        : Average mflops/s per call per node       : 803811
-Grid       : Message        : Average mflops/s per call per node       : 817101
-Grid       : Message        : Average mflops/s per call per node       : 668798
-Grid       : Message        : Average mflops/s per call per node (full): 316127
-Grid       : Message        : Average mflops/s per call per node (full): 439144
-Grid       : Message        : Average mflops/s per call per node (full): 445765
-Grid       : Message        : Average mflops/s per call per node (full): 306233
-Grid       : Message        :  Stencil 12.5875 GB/s per node
-Grid       : Message        :  Stencil 16.5928 GB/s per node
-Grid       : Message        :  Stencil 17.1334 GB/s per node
-Grid       : Message        :  Stencil 12.7421 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668509
-Grid       : Message        : Average mflops/s per call per node       : 802915
-Grid       : Message        : Average mflops/s per call per node       : 826550
-Grid       : Message        : Average mflops/s per call per node       : 663648
-Grid       : Message        : Average mflops/s per call per node (full): 313506
-Grid       : Message        : Average mflops/s per call per node (full): 435660
-Grid       : Message        : Average mflops/s per call per node (full): 442517
-Grid       : Message        : Average mflops/s per call per node (full): 304384
-Grid       : Message        :  Stencil 12.7432 GB/s per node
-Grid       : Message        :  Stencil 17.4419 GB/s per node
-Grid       : Message        :  Stencil 18.3907 GB/s per node
-Grid       : Message        :  Stencil 12.7773 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665510
-Grid       : Message        : Average mflops/s per call per node       : 803177
-Grid       : Message        : Average mflops/s per call per node       : 825020
-Grid       : Message        : Average mflops/s per call per node       : 665546
-Grid       : Message        : Average mflops/s per call per node (full): 313556
-Grid       : Message        : Average mflops/s per call per node (full): 438560
-Grid       : Message        : Average mflops/s per call per node (full): 447051
-Grid       : Message        : Average mflops/s per call per node (full): 305798
-Grid       : Message        :  Stencil 13.6475 GB/s per node
-Grid       : Message        :  Stencil 14.9065 GB/s per node
-Grid       : Message        :  Stencil 17.237 GB/s per node
-Grid       : Message        :  Stencil 12.7836 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666778
-Grid       : Message        : Average mflops/s per call per node       : 807523
-Grid       : Message        : Average mflops/s per call per node       : 824240
-Grid       : Message        : Average mflops/s per call per node       : 662212
-Grid       : Message        : Average mflops/s per call per node (full): 315898
-Grid       : Message        : Average mflops/s per call per node (full): 413795
-Grid       : Message        : Average mflops/s per call per node (full): 442913
-Grid       : Message        : Average mflops/s per call per node (full): 304489
-Grid       : Message        :  Stencil 13.3365 GB/s per node
-Grid       : Message        :  Stencil 17.3594 GB/s per node
-Grid       : Message        :  Stencil 17.4579 GB/s per node
-Grid       : Message        :  Stencil 12.355 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667231
-Grid       : Message        : Average mflops/s per call per node       : 802404
-Grid       : Message        : Average mflops/s per call per node       : 828621
-Grid       : Message        : Average mflops/s per call per node       : 665864
-Grid       : Message        : Average mflops/s per call per node (full): 314036
-Grid       : Message        : Average mflops/s per call per node (full): 435686
-Grid       : Message        : Average mflops/s per call per node (full): 445797
-Grid       : Message        : Average mflops/s per call per node (full): 303018
-Grid       : Message        :  Stencil 13.3896 GB/s per node
-Grid       : Message        :  Stencil 16.8074 GB/s per node
-Grid       : Message        :  Stencil 17.3581 GB/s per node
-Grid       : Message        :  Stencil 13.3522 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668693
-Grid       : Message        : Average mflops/s per call per node       : 804947
-Grid       : Message        : Average mflops/s per call per node       : 823825
-Grid       : Message        : Average mflops/s per call per node       : 663289
-Grid       : Message        : Average mflops/s per call per node (full): 315872
-Grid       : Message        : Average mflops/s per call per node (full): 437838
-Grid       : Message        : Average mflops/s per call per node (full): 444007
-Grid       : Message        : Average mflops/s per call per node (full): 306131
-Grid       : Message        :  Stencil 12.2837 GB/s per node
-Grid       : Message        :  Stencil 16.4696 GB/s per node
-Grid       : Message        :  Stencil 18.1402 GB/s per node
-Grid       : Message        :  Stencil 13.8135 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667352
-Grid       : Message        : Average mflops/s per call per node       : 803747
-Grid       : Message        : Average mflops/s per call per node       : 824454
-Grid       : Message        : Average mflops/s per call per node       : 661954
-Grid       : Message        : Average mflops/s per call per node (full): 310871
-Grid       : Message        : Average mflops/s per call per node (full): 434036
-Grid       : Message        : Average mflops/s per call per node (full): 449157
-Grid       : Message        : Average mflops/s per call per node (full): 304027
-Grid       : Message        :  Stencil 10.6491 GB/s per node
-Grid       : Message        :  Stencil 16.7318 GB/s per node
-Grid       : Message        :  Stencil 17.8447 GB/s per node
-Grid       : Message        :  Stencil 12.1491 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671516
-Grid       : Message        : Average mflops/s per call per node       : 804093
-Grid       : Message        : Average mflops/s per call per node       : 820766
-Grid       : Message        : Average mflops/s per call per node       : 669280
-Grid       : Message        : Average mflops/s per call per node (full): 293317
-Grid       : Message        : Average mflops/s per call per node (full): 436082
-Grid       : Message        : Average mflops/s per call per node (full): 446245
-Grid       : Message        : Average mflops/s per call per node (full): 303620
-Grid       : Message        :  Stencil 13.9134 GB/s per node
-Grid       : Message        :  Stencil 13.2131 GB/s per node
-Grid       : Message        :  Stencil 18.4007 GB/s per node
-Grid       : Message        :  Stencil 11.608 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665008
-Grid       : Message        : Average mflops/s per call per node       : 804174
-Grid       : Message        : Average mflops/s per call per node       : 820197
-Grid       : Message        : Average mflops/s per call per node       : 665283
-Grid       : Message        : Average mflops/s per call per node (full): 316562
-Grid       : Message        : Average mflops/s per call per node (full): 385144
-Grid       : Message        : Average mflops/s per call per node (full): 446507
-Grid       : Message        : Average mflops/s per call per node (full): 294405
-Grid       : Message        :  Stencil 14.206 GB/s per node
-Grid       : Message        :  Stencil 17.3668 GB/s per node
-Grid       : Message        :  Stencil 17.3285 GB/s per node
-Grid       : Message        :  Stencil 12.5141 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666293
-Grid       : Message        : Average mflops/s per call per node       : 799737
-Grid       : Message        : Average mflops/s per call per node       : 825270
-Grid       : Message        : Average mflops/s per call per node       : 665729
-Grid       : Message        : Average mflops/s per call per node (full): 316554
-Grid       : Message        : Average mflops/s per call per node (full): 437565
-Grid       : Message        : Average mflops/s per call per node (full): 442063
-Grid       : Message        : Average mflops/s per call per node (full): 304515
-Grid       : Message        :  Stencil 13.162 GB/s per node
-Grid       : Message        :  Stencil 16.3511 GB/s per node
-Grid       : Message        :  Stencil 17.517 GB/s per node
-Grid       : Message        :  Stencil 12.0928 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666213
-Grid       : Message        : Average mflops/s per call per node       : 804341
-Grid       : Message        : Average mflops/s per call per node       : 818709
-Grid       : Message        : Average mflops/s per call per node       : 662936
-Grid       : Message        : Average mflops/s per call per node (full): 315494
-Grid       : Message        : Average mflops/s per call per node (full): 433146
-Grid       : Message        : Average mflops/s per call per node (full): 444798
-Grid       : Message        : Average mflops/s per call per node (full): 302146
-Grid       : Message        :  Stencil 13.6063 GB/s per node
-Grid       : Message        :  Stencil 16.6806 GB/s per node
-Grid       : Message        :  Stencil 17.1707 GB/s per node
-Grid       : Message        :  Stencil 12.1258 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662785
-Grid       : Message        : Average mflops/s per call per node       : 810464
-Grid       : Message        : Average mflops/s per call per node       : 820353
-Grid       : Message        : Average mflops/s per call per node       : 670523
-Grid       : Message        : Average mflops/s per call per node (full): 315716
-Grid       : Message        : Average mflops/s per call per node (full): 436969
-Grid       : Message        : Average mflops/s per call per node (full): 440335
-Grid       : Message        : Average mflops/s per call per node (full): 302790
-Grid       : Message        :  Stencil 14.1364 GB/s per node
-Grid       : Message        :  Stencil 16.9377 GB/s per node
-Grid       : Message        :  Stencil 18.1098 GB/s per node
-Grid       : Message        :  Stencil 12.7042 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664049
-Grid       : Message        : Average mflops/s per call per node       : 803685
-Grid       : Message        : Average mflops/s per call per node       : 825457
-Grid       : Message        : Average mflops/s per call per node       : 667548
-Grid       : Message        : Average mflops/s per call per node (full): 315970
-Grid       : Message        : Average mflops/s per call per node (full): 434618
-Grid       : Message        : Average mflops/s per call per node (full): 448142
-Grid       : Message        : Average mflops/s per call per node (full): 305991
-Grid       : Message        :  Stencil 12.8699 GB/s per node
-Grid       : Message        :  Stencil 16.6379 GB/s per node
-Grid       : Message        :  Stencil 18.522 GB/s per node
-Grid       : Message        :  Stencil 13.0003 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668760
-Grid       : Message        : Average mflops/s per call per node       : 807728
-Grid       : Message        : Average mflops/s per call per node       : 824873
-Grid       : Message        : Average mflops/s per call per node       : 661669
-Grid       : Message        : Average mflops/s per call per node (full): 315270
-Grid       : Message        : Average mflops/s per call per node (full): 436781
-Grid       : Message        : Average mflops/s per call per node (full): 449931
-Grid       : Message        : Average mflops/s per call per node (full): 304919
-Grid       : Message        :  Stencil 14.9178 GB/s per node
-Grid       : Message        :  Stencil 16.4403 GB/s per node
-Grid       : Message        :  Stencil 17.7768 GB/s per node
-Grid       : Message        :  Stencil 14.7444 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666216
-Grid       : Message        : Average mflops/s per call per node       : 803335
-Grid       : Message        : Average mflops/s per call per node       : 823697
-Grid       : Message        : Average mflops/s per call per node       : 658365
-Grid       : Message        : Average mflops/s per call per node (full): 316364
-Grid       : Message        : Average mflops/s per call per node (full): 432841
-Grid       : Message        : Average mflops/s per call per node (full): 445773
-Grid       : Message        : Average mflops/s per call per node (full): 307202
-Grid       : Message        :  Stencil 13.4562 GB/s per node
-Grid       : Message        :  Stencil 17.3964 GB/s per node
-Grid       : Message        :  Stencil 18.6967 GB/s per node
-Grid       : Message        :  Stencil 13.176 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668300
-Grid       : Message        : Average mflops/s per call per node       : 796984
-Grid       : Message        : Average mflops/s per call per node       : 824084
-Grid       : Message        : Average mflops/s per call per node       : 662759
-Grid       : Message        : Average mflops/s per call per node (full): 316025
-Grid       : Message        : Average mflops/s per call per node (full): 436907
-Grid       : Message        : Average mflops/s per call per node (full): 447205
-Grid       : Message        : Average mflops/s per call per node (full): 305831
-Grid       : Message        :  Stencil 13.7019 GB/s per node
-Grid       : Message        :  Stencil 17.0464 GB/s per node
-Grid       : Message        :  Stencil 17.5725 GB/s per node
-Grid       : Message        :  Stencil 13.1021 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668842
-Grid       : Message        : Average mflops/s per call per node       : 806323
-Grid       : Message        : Average mflops/s per call per node       : 817625
-Grid       : Message        : Average mflops/s per call per node       : 662144
-Grid       : Message        : Average mflops/s per call per node (full): 316260
-Grid       : Message        : Average mflops/s per call per node (full): 439138
-Grid       : Message        : Average mflops/s per call per node (full): 444101
-Grid       : Message        : Average mflops/s per call per node (full): 305157
-Grid       : Message        :  Stencil 13.3656 GB/s per node
-Grid       : Message        :  Stencil 18.3142 GB/s per node
-Grid       : Message        :  Stencil 17.2804 GB/s per node
-Grid       : Message        :  Stencil 12.0448 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669417
-Grid       : Message        : Average mflops/s per call per node       : 804771
-Grid       : Message        : Average mflops/s per call per node       : 821989
-Grid       : Message        : Average mflops/s per call per node       : 670615
-Grid       : Message        : Average mflops/s per call per node (full): 316023
-Grid       : Message        : Average mflops/s per call per node (full): 441714
-Grid       : Message        : Average mflops/s per call per node (full): 442340
-Grid       : Message        : Average mflops/s per call per node (full): 302689
-Grid       : Message        :  Stencil 13.4482 GB/s per node
-Grid       : Message        :  Stencil 17.1081 GB/s per node
-Grid       : Message        :  Stencil 17.2833 GB/s per node
-Grid       : Message        :  Stencil 12.2318 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666082
-Grid       : Message        : Average mflops/s per call per node       : 803669
-Grid       : Message        : Average mflops/s per call per node       : 822570
-Grid       : Message        : Average mflops/s per call per node       : 671442
-Grid       : Message        : Average mflops/s per call per node (full): 315574
-Grid       : Message        : Average mflops/s per call per node (full): 438385
-Grid       : Message        : Average mflops/s per call per node (full): 442237
-Grid       : Message        : Average mflops/s per call per node (full): 304457
-Grid       : Message        :  Stencil 12.9332 GB/s per node
-Grid       : Message        :  Stencil 19.7532 GB/s per node
-Grid       : Message        :  Stencil 18.0717 GB/s per node
-Grid       : Message        :  Stencil 12.6331 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666753
-Grid       : Message        : Average mflops/s per call per node       : 804898
-Grid       : Message        : Average mflops/s per call per node       : 818580
-Grid       : Message        : Average mflops/s per call per node       : 671362
-Grid       : Message        : Average mflops/s per call per node (full): 314607
-Grid       : Message        : Average mflops/s per call per node (full): 444353
-Grid       : Message        : Average mflops/s per call per node (full): 445560
-Grid       : Message        : Average mflops/s per call per node (full): 305843
-Grid       : Message        :  Stencil 12.5512 GB/s per node
-Grid       : Message        :  Stencil 17.5269 GB/s per node
-Grid       : Message        :  Stencil 17.2834 GB/s per node
-Grid       : Message        :  Stencil 13.4437 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 672546
-Grid       : Message        : Average mflops/s per call per node       : 803988
-Grid       : Message        : Average mflops/s per call per node       : 823710
-Grid       : Message        : Average mflops/s per call per node       : 661994
-Grid       : Message        : Average mflops/s per call per node (full): 314246
-Grid       : Message        : Average mflops/s per call per node (full): 441246
-Grid       : Message        : Average mflops/s per call per node (full): 442659
-Grid       : Message        : Average mflops/s per call per node (full): 306323
-Grid       : Message        :  Stencil 13.7881 GB/s per node
-Grid       : Message        :  Stencil 17.1041 GB/s per node
-Grid       : Message        :  Stencil 17.8683 GB/s per node
-Grid       : Message        :  Stencil 14.7355 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662517
-Grid       : Message        : Average mflops/s per call per node       : 802137
-Grid       : Message        : Average mflops/s per call per node       : 820673
-Grid       : Message        : Average mflops/s per call per node       : 660132
-Grid       : Message        : Average mflops/s per call per node (full): 315677
-Grid       : Message        : Average mflops/s per call per node (full): 435985
-Grid       : Message        : Average mflops/s per call per node (full): 446418
-Grid       : Message        : Average mflops/s per call per node (full): 306190
-Grid       : Message        :  Stencil 14.4227 GB/s per node
-Grid       : Message        :  Stencil 16.9638 GB/s per node
-Grid       : Message        :  Stencil 17.2421 GB/s per node
-Grid       : Message        :  Stencil 13.146 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662917
-Grid       : Message        : Average mflops/s per call per node       : 805807
-Grid       : Message        : Average mflops/s per call per node       : 817935
-Grid       : Message        : Average mflops/s per call per node       : 664956
-Grid       : Message        : Average mflops/s per call per node (full): 316235
-Grid       : Message        : Average mflops/s per call per node (full): 436484
-Grid       : Message        : Average mflops/s per call per node (full): 440255
-Grid       : Message        : Average mflops/s per call per node (full): 305928
-Grid       : Message        :  Stencil 12.6977 GB/s per node
-Grid       : Message        :  Stencil 10.1535 GB/s per node
-Grid       : Message        :  Stencil 17.3576 GB/s per node
-Grid       : Message        :  Stencil 12.7675 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669493
-Grid       : Message        : Average mflops/s per call per node       : 808986
-Grid       : Message        : Average mflops/s per call per node       : 823811
-Grid       : Message        : Average mflops/s per call per node       : 668299
-Grid       : Message        : Average mflops/s per call per node (full): 314590
-Grid       : Message        : Average mflops/s per call per node (full): 320442
-Grid       : Message        : Average mflops/s per call per node (full): 444171
-Grid       : Message        : Average mflops/s per call per node (full): 303688
-Grid       : Message        :  Stencil 13.195 GB/s per node
-Grid       : Message        :  Stencil 16.5774 GB/s per node
-Grid       : Message        :  Stencil 17.064 GB/s per node
-Grid       : Message        :  Stencil 12.4249 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668161
-Grid       : Message        : Average mflops/s per call per node       : 805483
-Grid       : Message        : Average mflops/s per call per node       : 819279
-Grid       : Message        : Average mflops/s per call per node       : 668078
-Grid       : Message        : Average mflops/s per call per node (full): 315648
-Grid       : Message        : Average mflops/s per call per node (full): 435404
-Grid       : Message        : Average mflops/s per call per node (full): 440242
-Grid       : Message        : Average mflops/s per call per node (full): 304294
-Grid       : Message        :  Stencil 13.1611 GB/s per node
-Grid       : Message        :  Stencil 10.293 GB/s per node
-Grid       : Message        :  Stencil 18.4021 GB/s per node
-Grid       : Message        :  Stencil 12.2993 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664944
-Grid       : Message        : Average mflops/s per call per node       : 810756
-Grid       : Message        : Average mflops/s per call per node       : 823706
-Grid       : Message        : Average mflops/s per call per node       : 665806
-Grid       : Message        : Average mflops/s per call per node (full): 315544
-Grid       : Message        : Average mflops/s per call per node (full): 322690
-Grid       : Message        : Average mflops/s per call per node (full): 446038
-Grid       : Message        : Average mflops/s per call per node (full): 303429
-Grid       : Message        :  Stencil 14.224 GB/s per node
-Grid       : Message        :  Stencil 16.9568 GB/s per node
-Grid       : Message        :  Stencil 17.3814 GB/s per node
-Grid       : Message        :  Stencil 12.3383 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663420
-Grid       : Message        : Average mflops/s per call per node       : 803452
-Grid       : Message        : Average mflops/s per call per node       : 822770
-Grid       : Message        : Average mflops/s per call per node       : 664749
-Grid       : Message        : Average mflops/s per call per node (full): 316778
-Grid       : Message        : Average mflops/s per call per node (full): 436976
-Grid       : Message        : Average mflops/s per call per node (full): 443927
-Grid       : Message        : Average mflops/s per call per node (full): 303698
-Grid       : Message        :  Stencil 14.2383 GB/s per node
-Grid       : Message        :  Stencil 16.4187 GB/s per node
-Grid       : Message        :  Stencil 18.1444 GB/s per node
-Grid       : Message        :  Stencil 14.1145 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662589
-Grid       : Message        : Average mflops/s per call per node       : 803662
-Grid       : Message        : Average mflops/s per call per node       : 824074
-Grid       : Message        : Average mflops/s per call per node       : 655333
-Grid       : Message        : Average mflops/s per call per node (full): 316927
-Grid       : Message        : Average mflops/s per call per node (full): 433461
-Grid       : Message        : Average mflops/s per call per node (full): 447367
-Grid       : Message        : Average mflops/s per call per node (full): 305873
-Grid       : Message        :  Stencil 14.1274 GB/s per node
-Grid       : Message        :  Stencil 16.8365 GB/s per node
-Grid       : Message        :  Stencil 18.1599 GB/s per node
-Grid       : Message        :  Stencil 13.2011 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662459
-Grid       : Message        : Average mflops/s per call per node       : 804676
-Grid       : Message        : Average mflops/s per call per node       : 824916
-Grid       : Message        : Average mflops/s per call per node       : 663064
-Grid       : Message        : Average mflops/s per call per node (full): 315764
-Grid       : Message        : Average mflops/s per call per node (full): 437445
-Grid       : Message        : Average mflops/s per call per node (full): 447714
-Grid       : Message        : Average mflops/s per call per node (full): 305780
-Grid       : Message        :  Stencil 14.4141 GB/s per node
-Grid       : Message        :  Stencil 17.0002 GB/s per node
-Grid       : Message        :  Stencil 17.4644 GB/s per node
-Grid       : Message        :  Stencil 12.4324 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666248
-Grid       : Message        : Average mflops/s per call per node       : 807964
-Grid       : Message        : Average mflops/s per call per node       : 817203
-Grid       : Message        : Average mflops/s per call per node       : 668184
-Grid       : Message        : Average mflops/s per call per node (full): 315953
-Grid       : Message        : Average mflops/s per call per node (full): 436210
-Grid       : Message        : Average mflops/s per call per node (full): 443869
-Grid       : Message        : Average mflops/s per call per node (full): 304638
-Grid       : Message        :  Stencil 13.9462 GB/s per node
-Grid       : Message        :  Stencil 9.16624 GB/s per node
-Grid       : Message        :  Stencil 17.3436 GB/s per node
-Grid       : Message        :  Stencil 12.7577 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663975
-Grid       : Message        : Average mflops/s per call per node       : 815086
-Grid       : Message        : Average mflops/s per call per node       : 819869
-Grid       : Message        : Average mflops/s per call per node       : 661803
-Grid       : Message        : Average mflops/s per call per node (full): 316413
-Grid       : Message        : Average mflops/s per call per node (full): 296466
-Grid       : Message        : Average mflops/s per call per node (full): 443320
-Grid       : Message        : Average mflops/s per call per node (full): 305294
-Grid       : Message        :  Stencil 13.1198 GB/s per node
-Grid       : Message        :  Stencil 16.541 GB/s per node
-Grid       : Message        :  Stencil 17.5719 GB/s per node
-Grid       : Message        :  Stencil 12.426 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666706
-Grid       : Message        : Average mflops/s per call per node       : 806562
-Grid       : Message        : Average mflops/s per call per node       : 827599
-Grid       : Message        : Average mflops/s per call per node       : 671380
-Grid       : Message        : Average mflops/s per call per node (full): 314551
-Grid       : Message        : Average mflops/s per call per node (full): 435958
-Grid       : Message        : Average mflops/s per call per node (full): 447131
-Grid       : Message        : Average mflops/s per call per node (full): 305332
-Grid       : Message        :  Stencil 14.0187 GB/s per node
-Grid       : Message        :  Stencil 16.7282 GB/s per node
-Grid       : Message        :  Stencil 17.5016 GB/s per node
-Grid       : Message        :  Stencil 12.3897 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665303
-Grid       : Message        : Average mflops/s per call per node       : 803576
-Grid       : Message        : Average mflops/s per call per node       : 820190
-Grid       : Message        : Average mflops/s per call per node       : 662707
-Grid       : Message        : Average mflops/s per call per node (full): 316797
-Grid       : Message        : Average mflops/s per call per node (full): 436659
-Grid       : Message        : Average mflops/s per call per node (full): 444002
-Grid       : Message        : Average mflops/s per call per node (full): 303495
-Grid       : Message        :  Stencil 12.9872 GB/s per node
-Grid       : Message        :  Stencil 10.7783 GB/s per node
-Grid       : Message        :  Stencil 17.6598 GB/s per node
-Grid       : Message        :  Stencil 12.1232 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666356
-Grid       : Message        : Average mflops/s per call per node       : 808219
-Grid       : Message        : Average mflops/s per call per node       : 821321
-Grid       : Message        : Average mflops/s per call per node       : 667910
-Grid       : Message        : Average mflops/s per call per node (full): 315669
-Grid       : Message        : Average mflops/s per call per node (full): 332681
-Grid       : Message        : Average mflops/s per call per node (full): 442435
-Grid       : Message        : Average mflops/s per call per node (full): 302498
-Grid       : Message        :  Stencil 12.3811 GB/s per node
-Grid       : Message        :  Stencil 17.0083 GB/s per node
-Grid       : Message        :  Stencil 17.8352 GB/s per node
-Grid       : Message        :  Stencil 13.3962 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669156
-Grid       : Message        : Average mflops/s per call per node       : 805285
-Grid       : Message        : Average mflops/s per call per node       : 818262
-Grid       : Message        : Average mflops/s per call per node       : 657968
-Grid       : Message        : Average mflops/s per call per node (full): 312932
-Grid       : Message        : Average mflops/s per call per node (full): 438176
-Grid       : Message        : Average mflops/s per call per node (full): 437432
-Grid       : Message        : Average mflops/s per call per node (full): 304967
-Grid       : Message        :  Stencil 12.6208 GB/s per node
-Grid       : Message        :  Stencil 16.5345 GB/s per node
-Grid       : Message        :  Stencil 17.303 GB/s per node
-Grid       : Message        :  Stencil 13.6195 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667075
-Grid       : Message        : Average mflops/s per call per node       : 802960
-Grid       : Message        : Average mflops/s per call per node       : 826809
-Grid       : Message        : Average mflops/s per call per node       : 658842
-Grid       : Message        : Average mflops/s per call per node (full): 311157
-Grid       : Message        : Average mflops/s per call per node (full): 433094
-Grid       : Message        : Average mflops/s per call per node (full): 441505
-Grid       : Message        : Average mflops/s per call per node (full): 304660
-Grid       : Message        :  Stencil 13.5426 GB/s per node
-Grid       : Message        :  Stencil 16.4742 GB/s per node
-Grid       : Message        :  Stencil 18.4648 GB/s per node
-Grid       : Message        :  Stencil 14.1341 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665545
-Grid       : Message        : Average mflops/s per call per node       : 804158
-Grid       : Message        : Average mflops/s per call per node       : 819437
-Grid       : Message        : Average mflops/s per call per node       : 664279
-Grid       : Message        : Average mflops/s per call per node (full): 315607
-Grid       : Message        : Average mflops/s per call per node (full): 432159
-Grid       : Message        : Average mflops/s per call per node (full): 446260
-Grid       : Message        : Average mflops/s per call per node (full): 305848
-Grid       : Message        :  Stencil 13.2587 GB/s per node
-Grid       : Message        :  Stencil 16.5012 GB/s per node
-Grid       : Message        :  Stencil 17.3578 GB/s per node
-Grid       : Message        :  Stencil 11.7879 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668955
-Grid       : Message        : Average mflops/s per call per node       : 804300
-Grid       : Message        : Average mflops/s per call per node       : 821209
-Grid       : Message        : Average mflops/s per call per node       : 669854
-Grid       : Message        : Average mflops/s per call per node (full): 315721
-Grid       : Message        : Average mflops/s per call per node (full): 431274
-Grid       : Message        : Average mflops/s per call per node (full): 444321
-Grid       : Message        : Average mflops/s per call per node (full): 300279
-Grid       : Message        :  Stencil 12.8008 GB/s per node
-Grid       : Message        :  Stencil 16.3146 GB/s per node
-Grid       : Message        :  Stencil 17.1493 GB/s per node
-Grid       : Message        :  Stencil 12.0482 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669211
-Grid       : Message        : Average mflops/s per call per node       : 807401
-Grid       : Message        : Average mflops/s per call per node       : 822916
-Grid       : Message        : Average mflops/s per call per node       : 665423
-Grid       : Message        : Average mflops/s per call per node (full): 313918
-Grid       : Message        : Average mflops/s per call per node (full): 428195
-Grid       : Message        : Average mflops/s per call per node (full): 442564
-Grid       : Message        : Average mflops/s per call per node (full): 301389
-Grid       : Message        :  Stencil 12.6949 GB/s per node
-Grid       : Message        :  Stencil 16.4159 GB/s per node
-Grid       : Message        :  Stencil 18.1188 GB/s per node
-Grid       : Message        :  Stencil 13.1025 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666446
-Grid       : Message        : Average mflops/s per call per node       : 804369
-Grid       : Message        : Average mflops/s per call per node       : 824741
-Grid       : Message        : Average mflops/s per call per node       : 663836
-Grid       : Message        : Average mflops/s per call per node (full): 312030
-Grid       : Message        : Average mflops/s per call per node (full): 433563
-Grid       : Message        : Average mflops/s per call per node (full): 448410
-Grid       : Message        : Average mflops/s per call per node (full): 305915
-Grid       : Message        :  Stencil 15.1128 GB/s per node
-Grid       : Message        :  Stencil 17.6454 GB/s per node
-Grid       : Message        :  Stencil 17.5161 GB/s per node
-Grid       : Message        :  Stencil 12.8753 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663917
-Grid       : Message        : Average mflops/s per call per node       : 801411
-Grid       : Message        : Average mflops/s per call per node       : 828285
-Grid       : Message        : Average mflops/s per call per node       : 661367
-Grid       : Message        : Average mflops/s per call per node (full): 316998
-Grid       : Message        : Average mflops/s per call per node (full): 437959
-Grid       : Message        : Average mflops/s per call per node (full): 444458
-Grid       : Message        : Average mflops/s per call per node (full): 304657
-Grid       : Message        :  Stencil 13.4841 GB/s per node
-Grid       : Message        :  Stencil 16.5592 GB/s per node
-Grid       : Message        :  Stencil 17.373 GB/s per node
-Grid       : Message        :  Stencil 13.1126 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667237
-Grid       : Message        : Average mflops/s per call per node       : 800749
-Grid       : Message        : Average mflops/s per call per node       : 822447
-Grid       : Message        : Average mflops/s per call per node       : 664609
-Grid       : Message        : Average mflops/s per call per node (full): 315971
-Grid       : Message        : Average mflops/s per call per node (full): 434444
-Grid       : Message        : Average mflops/s per call per node (full): 444853
-Grid       : Message        : Average mflops/s per call per node (full): 305671
-Grid       : Message        :  Stencil 12.8365 GB/s per node
-Grid       : Message        :  Stencil 16.8574 GB/s per node
-Grid       : Message        :  Stencil 18.38 GB/s per node
-Grid       : Message        :  Stencil 13.3748 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667470
-Grid       : Message        : Average mflops/s per call per node       : 808901
-Grid       : Message        : Average mflops/s per call per node       : 818028
-Grid       : Message        : Average mflops/s per call per node       : 668664
-Grid       : Message        : Average mflops/s per call per node (full): 312274
-Grid       : Message        : Average mflops/s per call per node (full): 437422
-Grid       : Message        : Average mflops/s per call per node (full): 445399
-Grid       : Message        : Average mflops/s per call per node (full): 307343
-Grid       : Message        :  Stencil 13.3121 GB/s per node
-Grid       : Message        :  Stencil 17.5213 GB/s per node
-Grid       : Message        :  Stencil 18.3939 GB/s per node
-Grid       : Message        :  Stencil 13.0664 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667802
-Grid       : Message        : Average mflops/s per call per node       : 801730
-Grid       : Message        : Average mflops/s per call per node       : 827544
-Grid       : Message        : Average mflops/s per call per node       : 665724
-Grid       : Message        : Average mflops/s per call per node (full): 315980
-Grid       : Message        : Average mflops/s per call per node (full): 438621
-Grid       : Message        : Average mflops/s per call per node (full): 448983
-Grid       : Message        : Average mflops/s per call per node (full): 306193
-Grid       : Message        :  Stencil 13.4233 GB/s per node
-Grid       : Message        :  Stencil 17.1774 GB/s per node
-Grid       : Message        :  Stencil 17.7923 GB/s per node
-Grid       : Message        :  Stencil 11.9634 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667088
-Grid       : Message        : Average mflops/s per call per node       : 802362
-Grid       : Message        : Average mflops/s per call per node       : 820102
-Grid       : Message        : Average mflops/s per call per node       : 667356
-Grid       : Message        : Average mflops/s per call per node (full): 315180
-Grid       : Message        : Average mflops/s per call per node (full): 436717
-Grid       : Message        : Average mflops/s per call per node (full): 444688
-Grid       : Message        : Average mflops/s per call per node (full): 301726
-Grid       : Message        :  Stencil 12.8978 GB/s per node
-Grid       : Message        :  Stencil 16.6616 GB/s per node
-Grid       : Message        :  Stencil 17.9218 GB/s per node
-Grid       : Message        :  Stencil 13.2227 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667878
-Grid       : Message        : Average mflops/s per call per node       : 802312
-Grid       : Message        : Average mflops/s per call per node       : 815521
-Grid       : Message        : Average mflops/s per call per node       : 660750
-Grid       : Message        : Average mflops/s per call per node (full): 314861
-Grid       : Message        : Average mflops/s per call per node (full): 434376
-Grid       : Message        : Average mflops/s per call per node (full): 445104
-Grid       : Message        : Average mflops/s per call per node (full): 305488
-Grid       : Message        :  Stencil 12.7428 GB/s per node
-Grid       : Message        :  Stencil 16.777 GB/s per node
-Grid       : Message        :  Stencil 16.8466 GB/s per node
-Grid       : Message        :  Stencil 12.3905 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668370
-Grid       : Message        : Average mflops/s per call per node       : 804917
-Grid       : Message        : Average mflops/s per call per node       : 825243
-Grid       : Message        : Average mflops/s per call per node       : 666776
-Grid       : Message        : Average mflops/s per call per node (full): 314673
-Grid       : Message        : Average mflops/s per call per node (full): 435603
-Grid       : Message        : Average mflops/s per call per node (full): 438303
-Grid       : Message        : Average mflops/s per call per node (full): 303690
-Grid       : Message        :  Stencil 13.3228 GB/s per node
-Grid       : Message        :  Stencil 17.4223 GB/s per node
-Grid       : Message        :  Stencil 16.754 GB/s per node
-Grid       : Message        :  Stencil 12.6843 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667883
-Grid       : Message        : Average mflops/s per call per node       : 805010
-Grid       : Message        : Average mflops/s per call per node       : 820299
-Grid       : Message        : Average mflops/s per call per node       : 663059
-Grid       : Message        : Average mflops/s per call per node (full): 315375
-Grid       : Message        : Average mflops/s per call per node (full): 440281
-Grid       : Message        : Average mflops/s per call per node (full): 430543
-Grid       : Message        : Average mflops/s per call per node (full): 301193
-Grid       : Message        :  Stencil 13.3313 GB/s per node
-Grid       : Message        :  Stencil 9.28908 GB/s per node
-Grid       : Message        :  Stencil 18.1805 GB/s per node
-Grid       : Message        :  Stencil 13.6462 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667383
-Grid       : Message        : Average mflops/s per call per node       : 815148
-Grid       : Message        : Average mflops/s per call per node       : 817603
-Grid       : Message        : Average mflops/s per call per node       : 660300
-Grid       : Message        : Average mflops/s per call per node (full): 315710
-Grid       : Message        : Average mflops/s per call per node (full): 299406
-Grid       : Message        : Average mflops/s per call per node (full): 446043
-Grid       : Message        : Average mflops/s per call per node (full): 306193
-Grid       : Message        :  Stencil 12.3255 GB/s per node
-Grid       : Message        :  Stencil 10.828 GB/s per node
-Grid       : Message        :  Stencil 17.9218 GB/s per node
-Grid       : Message        :  Stencil 12.2895 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668942
-Grid       : Message        : Average mflops/s per call per node       : 808118
-Grid       : Message        : Average mflops/s per call per node       : 815742
-Grid       : Message        : Average mflops/s per call per node       : 666199
-Grid       : Message        : Average mflops/s per call per node (full): 312475
-Grid       : Message        : Average mflops/s per call per node (full): 335871
-Grid       : Message        : Average mflops/s per call per node (full): 444799
-Grid       : Message        : Average mflops/s per call per node (full): 303522
-Grid       : Message        :  Stencil 12.3314 GB/s per node
-Grid       : Message        :  Stencil 16.9833 GB/s per node
-Grid       : Message        :  Stencil 17.2071 GB/s per node
-Grid       : Message        :  Stencil 13.1734 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669271
-Grid       : Message        : Average mflops/s per call per node       : 802289
-Grid       : Message        : Average mflops/s per call per node       : 831405
-Grid       : Message        : Average mflops/s per call per node       : 661655
-Grid       : Message        : Average mflops/s per call per node (full): 312489
-Grid       : Message        : Average mflops/s per call per node (full): 434880
-Grid       : Message        : Average mflops/s per call per node (full): 443932
-Grid       : Message        : Average mflops/s per call per node (full): 305986
-Grid       : Message        :  Stencil 12.5033 GB/s per node
-Grid       : Message        :  Stencil 18.164 GB/s per node
-Grid       : Message        :  Stencil 17.9494 GB/s per node
-Grid       : Message        :  Stencil 11.9873 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668282
-Grid       : Message        : Average mflops/s per call per node       : 798868
-Grid       : Message        : Average mflops/s per call per node       : 821773
-Grid       : Message        : Average mflops/s per call per node       : 668449
-Grid       : Message        : Average mflops/s per call per node (full): 312460
-Grid       : Message        : Average mflops/s per call per node (full): 440317
-Grid       : Message        : Average mflops/s per call per node (full): 445596
-Grid       : Message        : Average mflops/s per call per node (full): 302268
-Grid       : Message        :  Stencil 13.6313 GB/s per node
-Grid       : Message        :  Stencil 9.02077 GB/s per node
-Grid       : Message        :  Stencil 17.3999 GB/s per node
-Grid       : Message        :  Stencil 13.8185 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667175
-Grid       : Message        : Average mflops/s per call per node       : 811629
-Grid       : Message        : Average mflops/s per call per node       : 821079
-Grid       : Message        : Average mflops/s per call per node       : 660880
-Grid       : Message        : Average mflops/s per call per node (full): 315879
-Grid       : Message        : Average mflops/s per call per node (full): 293111
-Grid       : Message        : Average mflops/s per call per node (full): 444471
-Grid       : Message        : Average mflops/s per call per node (full): 305386
-Grid       : Message        :  Stencil 13.1158 GB/s per node
-Grid       : Message        :  Stencil 17.6661 GB/s per node
-Grid       : Message        :  Stencil 17.4208 GB/s per node
-Grid       : Message        :  Stencil 12.3309 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667019
-Grid       : Message        : Average mflops/s per call per node       : 804662
-Grid       : Message        : Average mflops/s per call per node       : 812437
-Grid       : Message        : Average mflops/s per call per node       : 666289
-Grid       : Message        : Average mflops/s per call per node (full): 314758
-Grid       : Message        : Average mflops/s per call per node (full): 441791
-Grid       : Message        : Average mflops/s per call per node (full): 437368
-Grid       : Message        : Average mflops/s per call per node (full): 303335
-Grid       : Message        :  Stencil 13.6911 GB/s per node
-Grid       : Message        :  Stencil 14.1241 GB/s per node
-Grid       : Message        :  Stencil 16.9226 GB/s per node
-Grid       : Message        :  Stencil 12.4308 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665603
-Grid       : Message        : Average mflops/s per call per node       : 807566
-Grid       : Message        : Average mflops/s per call per node       : 824624
-Grid       : Message        : Average mflops/s per call per node       : 668689
-Grid       : Message        : Average mflops/s per call per node (full): 315008
-Grid       : Message        : Average mflops/s per call per node (full): 401068
-Grid       : Message        : Average mflops/s per call per node (full): 439576
-Grid       : Message        : Average mflops/s per call per node (full): 305056
-Grid       : Message        :  Stencil 13.4017 GB/s per node
-Grid       : Message        :  Stencil 17.9305 GB/s per node
-Grid       : Message        :  Stencil 18.0716 GB/s per node
-Grid       : Message        :  Stencil 13.3354 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663292
-Grid       : Message        : Average mflops/s per call per node       : 801563
-Grid       : Message        : Average mflops/s per call per node       : 818955
-Grid       : Message        : Average mflops/s per call per node       : 669092
-Grid       : Message        : Average mflops/s per call per node (full): 315241
-Grid       : Message        : Average mflops/s per call per node (full): 438778
-Grid       : Message        : Average mflops/s per call per node (full): 445765
-Grid       : Message        : Average mflops/s per call per node (full): 306511
-Grid       : Message        :  Stencil 13.4992 GB/s per node
-Grid       : Message        :  Stencil 16.8372 GB/s per node
-Grid       : Message        :  Stencil 18.0878 GB/s per node
-Grid       : Message        :  Stencil 13.7108 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664701
-Grid       : Message        : Average mflops/s per call per node       : 805242
-Grid       : Message        : Average mflops/s per call per node       : 816353
-Grid       : Message        : Average mflops/s per call per node       : 660762
-Grid       : Message        : Average mflops/s per call per node (full): 316150
-Grid       : Message        : Average mflops/s per call per node (full): 436528
-Grid       : Message        : Average mflops/s per call per node (full): 444918
-Grid       : Message        : Average mflops/s per call per node (full): 304280
-Grid       : Message        :  Stencil 13.2318 GB/s per node
-Grid       : Message        :  Stencil 9.11562 GB/s per node
-Grid       : Message        :  Stencil 17.3461 GB/s per node
-Grid       : Message        :  Stencil 12.4613 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664415
-Grid       : Message        : Average mflops/s per call per node       : 812585
-Grid       : Message        : Average mflops/s per call per node       : 823017
-Grid       : Message        : Average mflops/s per call per node       : 664295
-Grid       : Message        : Average mflops/s per call per node (full): 314799
-Grid       : Message        : Average mflops/s per call per node (full): 294282
-Grid       : Message        : Average mflops/s per call per node (full): 443996
-Grid       : Message        : Average mflops/s per call per node (full): 302613
-Grid       : Message        :  Stencil 12.8501 GB/s per node
-Grid       : Message        :  Stencil 18.1365 GB/s per node
-Grid       : Message        :  Stencil 17.5319 GB/s per node
-Grid       : Message        :  Stencil 13.6167 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667116
-Grid       : Message        : Average mflops/s per call per node       : 805284
-Grid       : Message        : Average mflops/s per call per node       : 824300
-Grid       : Message        : Average mflops/s per call per node       : 663720
-Grid       : Message        : Average mflops/s per call per node (full): 314912
-Grid       : Message        : Average mflops/s per call per node (full): 441021
-Grid       : Message        : Average mflops/s per call per node (full): 446060
-Grid       : Message        : Average mflops/s per call per node (full): 306658
-Grid       : Message        :  Stencil 12.9248 GB/s per node
-Grid       : Message        :  Stencil 17.9967 GB/s per node
-Grid       : Message        :  Stencil 16.997 GB/s per node
-Grid       : Message        :  Stencil 13.6847 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664023
-Grid       : Message        : Average mflops/s per call per node       : 804162
-Grid       : Message        : Average mflops/s per call per node       : 824007
-Grid       : Message        : Average mflops/s per call per node       : 664748
-Grid       : Message        : Average mflops/s per call per node (full): 313748
-Grid       : Message        : Average mflops/s per call per node (full): 440774
-Grid       : Message        : Average mflops/s per call per node (full): 440280
-Grid       : Message        : Average mflops/s per call per node (full): 306476
-Grid       : Message        :  Stencil 13.4693 GB/s per node
-Grid       : Message        :  Stencil 10.9112 GB/s per node
-Grid       : Message        :  Stencil 18.7818 GB/s per node
-Grid       : Message        :  Stencil 13.5111 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665246
-Grid       : Message        : Average mflops/s per call per node       : 814175
-Grid       : Message        : Average mflops/s per call per node       : 821192
-Grid       : Message        : Average mflops/s per call per node       : 665535
-Grid       : Message        : Average mflops/s per call per node (full): 316162
-Grid       : Message        : Average mflops/s per call per node (full): 337525
-Grid       : Message        : Average mflops/s per call per node (full): 447987
-Grid       : Message        : Average mflops/s per call per node (full): 306323
-Grid       : Message        :  Stencil 13.0777 GB/s per node
-Grid       : Message        :  Stencil 16.9577 GB/s per node
-Grid       : Message        :  Stencil 18.3557 GB/s per node
-Grid       : Message        :  Stencil 14.4107 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665813
-Grid       : Message        : Average mflops/s per call per node       : 805514
-Grid       : Message        : Average mflops/s per call per node       : 819381
-Grid       : Message        : Average mflops/s per call per node       : 658705
-Grid       : Message        : Average mflops/s per call per node (full): 314850
-Grid       : Message        : Average mflops/s per call per node (full): 439586
-Grid       : Message        : Average mflops/s per call per node (full): 447271
-Grid       : Message        : Average mflops/s per call per node (full): 306939
-Grid       : Message        :  Stencil 13.3862 GB/s per node
-Grid       : Message        :  Stencil 17.1971 GB/s per node
-Grid       : Message        :  Stencil 17.6139 GB/s per node
-Grid       : Message        :  Stencil 12.9573 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666671
-Grid       : Message        : Average mflops/s per call per node       : 803421
-Grid       : Message        : Average mflops/s per call per node       : 825401
-Grid       : Message        : Average mflops/s per call per node       : 659545
-Grid       : Message        : Average mflops/s per call per node (full): 315837
-Grid       : Message        : Average mflops/s per call per node (full): 438396
-Grid       : Message        : Average mflops/s per call per node (full): 445524
-Grid       : Message        : Average mflops/s per call per node (full): 304195
-Grid       : Message        :  Stencil 13.5591 GB/s per node
-Grid       : Message        :  Stencil 17.7183 GB/s per node
-Grid       : Message        :  Stencil 17.2019 GB/s per node
-Grid       : Message        :  Stencil 12.3189 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665892
-Grid       : Message        : Average mflops/s per call per node       : 806264
-Grid       : Message        : Average mflops/s per call per node       : 824604
-Grid       : Message        : Average mflops/s per call per node       : 667157
-Grid       : Message        : Average mflops/s per call per node (full): 315081
-Grid       : Message        : Average mflops/s per call per node (full): 440782
-Grid       : Message        : Average mflops/s per call per node (full): 440963
-Grid       : Message        : Average mflops/s per call per node (full): 304250
-Grid       : Message        :  Stencil 13.248 GB/s per node
-Grid       : Message        :  Stencil 14.8018 GB/s per node
-Grid       : Message        :  Stencil 17.461 GB/s per node
-Grid       : Message        :  Stencil 13.2692 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665956
-Grid       : Message        : Average mflops/s per call per node       : 807889
-Grid       : Message        : Average mflops/s per call per node       : 825939
-Grid       : Message        : Average mflops/s per call per node       : 665659
-Grid       : Message        : Average mflops/s per call per node (full): 313670
-Grid       : Message        : Average mflops/s per call per node (full): 412631
-Grid       : Message        : Average mflops/s per call per node (full): 445575
-Grid       : Message        : Average mflops/s per call per node (full): 306363
-Grid       : Message        :  Stencil 13.7145 GB/s per node
-Grid       : Message        :  Stencil 17.3324 GB/s per node
-Grid       : Message        :  Stencil 18.1848 GB/s per node
-Grid       : Message        :  Stencil 12.9419 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665029
-Grid       : Message        : Average mflops/s per call per node       : 808743
-Grid       : Message        : Average mflops/s per call per node       : 821243
-Grid       : Message        : Average mflops/s per call per node       : 662212
-Grid       : Message        : Average mflops/s per call per node (full): 314960
-Grid       : Message        : Average mflops/s per call per node (full): 439231
-Grid       : Message        : Average mflops/s per call per node (full): 447120
-Grid       : Message        : Average mflops/s per call per node (full): 305205
-Grid       : Message        :  Stencil 13.5834 GB/s per node
-Grid       : Message        :  Stencil 13.9187 GB/s per node
-Grid       : Message        :  Stencil 17.0826 GB/s per node
-Grid       : Message        :  Stencil 13.5903 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668611
-Grid       : Message        : Average mflops/s per call per node       : 810477
-Grid       : Message        : Average mflops/s per call per node       : 820414
-Grid       : Message        : Average mflops/s per call per node       : 662241
-Grid       : Message        : Average mflops/s per call per node (full): 316736
-Grid       : Message        : Average mflops/s per call per node (full): 397739
-Grid       : Message        : Average mflops/s per call per node (full): 434490
-Grid       : Message        : Average mflops/s per call per node (full): 306342
-Grid       : Message        :  Stencil 12.8785 GB/s per node
-Grid       : Message        :  Stencil 18.0336 GB/s per node
-Grid       : Message        :  Stencil 17.5106 GB/s per node
-Grid       : Message        :  Stencil 12.2151 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668606
-Grid       : Message        : Average mflops/s per call per node       : 805111
-Grid       : Message        : Average mflops/s per call per node       : 822796
-Grid       : Message        : Average mflops/s per call per node       : 664684
-Grid       : Message        : Average mflops/s per call per node (full): 313610
-Grid       : Message        : Average mflops/s per call per node (full): 440398
-Grid       : Message        : Average mflops/s per call per node (full): 444645
-Grid       : Message        : Average mflops/s per call per node (full): 303118
-Grid       : Message        :  Stencil 13.3419 GB/s per node
-Grid       : Message        :  Stencil 17.7552 GB/s per node
-Grid       : Message        :  Stencil 18.0255 GB/s per node
-Grid       : Message        :  Stencil 12.1451 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666457
-Grid       : Message        : Average mflops/s per call per node       : 803015
-Grid       : Message        : Average mflops/s per call per node       : 819714
-Grid       : Message        : Average mflops/s per call per node       : 666522
-Grid       : Message        : Average mflops/s per call per node (full): 314487
-Grid       : Message        : Average mflops/s per call per node (full): 440563
-Grid       : Message        : Average mflops/s per call per node (full): 446709
-Grid       : Message        : Average mflops/s per call per node (full): 302842
-Grid       : Message        :  Stencil 13.0752 GB/s per node
-Grid       : Message        :  Stencil 16.7879 GB/s per node
-Grid       : Message        :  Stencil 17.1563 GB/s per node
-Grid       : Message        :  Stencil 12.2405 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661678
-Grid       : Message        : Average mflops/s per call per node       : 806525
-Grid       : Message        : Average mflops/s per call per node       : 822295
-Grid       : Message        : Average mflops/s per call per node       : 662354
-Grid       : Message        : Average mflops/s per call per node (full): 314978
-Grid       : Message        : Average mflops/s per call per node (full): 437910
-Grid       : Message        : Average mflops/s per call per node (full): 432688
-Grid       : Message        : Average mflops/s per call per node (full): 300123
-Grid       : Message        :  Stencil 12.8164 GB/s per node
-Grid       : Message        :  Stencil 10.8645 GB/s per node
-Grid       : Message        :  Stencil 17.5877 GB/s per node
-Grid       : Message        :  Stencil 12.2955 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663617
-Grid       : Message        : Average mflops/s per call per node       : 815350
-Grid       : Message        : Average mflops/s per call per node       : 822622
-Grid       : Message        : Average mflops/s per call per node       : 670769
-Grid       : Message        : Average mflops/s per call per node (full): 311572
-Grid       : Message        : Average mflops/s per call per node (full): 336451
-Grid       : Message        : Average mflops/s per call per node (full): 445744
-Grid       : Message        : Average mflops/s per call per node (full): 305249
-Grid       : Message        :  Stencil 13.4844 GB/s per node
-Grid       : Message        :  Stencil 9.41537 GB/s per node
-Grid       : Message        :  Stencil 17.3923 GB/s per node
-Grid       : Message        :  Stencil 12.6071 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664692
-Grid       : Message        : Average mflops/s per call per node       : 804293
-Grid       : Message        : Average mflops/s per call per node       : 820679
-Grid       : Message        : Average mflops/s per call per node       : 666786
-Grid       : Message        : Average mflops/s per call per node (full): 315823
-Grid       : Message        : Average mflops/s per call per node (full): 302137
-Grid       : Message        : Average mflops/s per call per node (full): 444841
-Grid       : Message        : Average mflops/s per call per node (full): 304938
-Grid       : Message        :  Stencil 13.9875 GB/s per node
-Grid       : Message        :  Stencil 17.01 GB/s per node
-Grid       : Message        :  Stencil 18.6205 GB/s per node
-Grid       : Message        :  Stencil 13.0562 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664500
-Grid       : Message        : Average mflops/s per call per node       : 802832
-Grid       : Message        : Average mflops/s per call per node       : 818612
-Grid       : Message        : Average mflops/s per call per node       : 665705
-Grid       : Message        : Average mflops/s per call per node (full): 316672
-Grid       : Message        : Average mflops/s per call per node (full): 438180
-Grid       : Message        : Average mflops/s per call per node (full): 447127
-Grid       : Message        : Average mflops/s per call per node (full): 306157
-Grid       : Message        :  Stencil 14.0333 GB/s per node
-Grid       : Message        :  Stencil 16.5672 GB/s per node
-Grid       : Message        :  Stencil 17.9764 GB/s per node
-Grid       : Message        :  Stencil 12.9231 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663626
-Grid       : Message        : Average mflops/s per call per node       : 804403
-Grid       : Message        : Average mflops/s per call per node       : 819891
-Grid       : Message        : Average mflops/s per call per node       : 666901
-Grid       : Message        : Average mflops/s per call per node (full): 315999
-Grid       : Message        : Average mflops/s per call per node (full): 435532
-Grid       : Message        : Average mflops/s per call per node (full): 446810
-Grid       : Message        : Average mflops/s per call per node (full): 306073
-Grid       : Message        :  Stencil 13.809 GB/s per node
-Grid       : Message        :  Stencil 18.2537 GB/s per node
-Grid       : Message        :  Stencil 17.7281 GB/s per node
-Grid       : Message        :  Stencil 14.1155 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665162
-Grid       : Message        : Average mflops/s per call per node       : 799502
-Grid       : Message        : Average mflops/s per call per node       : 826679
-Grid       : Message        : Average mflops/s per call per node       : 660384
-Grid       : Message        : Average mflops/s per call per node (full): 316186
-Grid       : Message        : Average mflops/s per call per node (full): 439300
-Grid       : Message        : Average mflops/s per call per node (full): 447229
-Grid       : Message        : Average mflops/s per call per node (full): 307210
-Grid       : Message        :  Stencil 13.7922 GB/s per node
-Grid       : Message        :  Stencil 17.225 GB/s per node
-Grid       : Message        :  Stencil 18.2881 GB/s per node
-Grid       : Message        :  Stencil 13.7297 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665724
-Grid       : Message        : Average mflops/s per call per node       : 799898
-Grid       : Message        : Average mflops/s per call per node       : 825497
-Grid       : Message        : Average mflops/s per call per node       : 660536
-Grid       : Message        : Average mflops/s per call per node (full): 315508
-Grid       : Message        : Average mflops/s per call per node (full): 437931
-Grid       : Message        : Average mflops/s per call per node (full): 448941
-Grid       : Message        : Average mflops/s per call per node (full): 306707
-Grid       : Message        :  Stencil 13.9262 GB/s per node
-Grid       : Message        :  Stencil 16.7104 GB/s per node
-Grid       : Message        :  Stencil 18.1467 GB/s per node
-Grid       : Message        :  Stencil 13.2761 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665930
-Grid       : Message        : Average mflops/s per call per node       : 805745
-Grid       : Message        : Average mflops/s per call per node       : 819049
-Grid       : Message        : Average mflops/s per call per node       : 658184
-Grid       : Message        : Average mflops/s per call per node (full): 316284
-Grid       : Message        : Average mflops/s per call per node (full): 436808
-Grid       : Message        : Average mflops/s per call per node (full): 445900
-Grid       : Message        : Average mflops/s per call per node (full): 304988
-Grid       : Message        :  Stencil 13.6979 GB/s per node
-Grid       : Message        :  Stencil 16.5942 GB/s per node
-Grid       : Message        :  Stencil 17.6454 GB/s per node
-Grid       : Message        :  Stencil 13.6655 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667274
-Grid       : Message        : Average mflops/s per call per node       : 804315
-Grid       : Message        : Average mflops/s per call per node       : 818993
-Grid       : Message        : Average mflops/s per call per node       : 661050
-Grid       : Message        : Average mflops/s per call per node (full): 315545
-Grid       : Message        : Average mflops/s per call per node (full): 435802
-Grid       : Message        : Average mflops/s per call per node (full): 444969
-Grid       : Message        : Average mflops/s per call per node (full): 305752
-Grid       : Message        :  Stencil 14.8011 GB/s per node
-Grid       : Message        :  Stencil 17.4577 GB/s per node
-Grid       : Message        :  Stencil 16.0438 GB/s per node
-Grid       : Message        :  Stencil 11.8119 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666647
-Grid       : Message        : Average mflops/s per call per node       : 803512
-Grid       : Message        : Average mflops/s per call per node       : 825357
-Grid       : Message        : Average mflops/s per call per node       : 669817
-Grid       : Message        : Average mflops/s per call per node (full): 318122
-Grid       : Message        : Average mflops/s per call per node (full): 438539
-Grid       : Message        : Average mflops/s per call per node (full): 420134
-Grid       : Message        : Average mflops/s per call per node (full): 299258
-Grid       : Message        :  Stencil 14.2815 GB/s per node
-Grid       : Message        :  Stencil 16.5677 GB/s per node
-Grid       : Message        :  Stencil 16.7552 GB/s per node
-Grid       : Message        :  Stencil 12.4322 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665525
-Grid       : Message        : Average mflops/s per call per node       : 802985
-Grid       : Message        : Average mflops/s per call per node       : 820117
-Grid       : Message        : Average mflops/s per call per node       : 668353
-Grid       : Message        : Average mflops/s per call per node (full): 317822
-Grid       : Message        : Average mflops/s per call per node (full): 435373
-Grid       : Message        : Average mflops/s per call per node (full): 437842
-Grid       : Message        : Average mflops/s per call per node (full): 304972
-Grid       : Message        :  Stencil 12.445 GB/s per node
-Grid       : Message        :  Stencil 13.4312 GB/s per node
-Grid       : Message        :  Stencil 17.1194 GB/s per node
-Grid       : Message        :  Stencil 13.338 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668202
-Grid       : Message        : Average mflops/s per call per node       : 807277
-Grid       : Message        : Average mflops/s per call per node       : 817844
-Grid       : Message        : Average mflops/s per call per node       : 661735
-Grid       : Message        : Average mflops/s per call per node (full): 313315
-Grid       : Message        : Average mflops/s per call per node (full): 390731
-Grid       : Message        : Average mflops/s per call per node (full): 440923
-Grid       : Message        : Average mflops/s per call per node (full): 306331
-Grid       : Message        :  Stencil 12.7171 GB/s per node
-Grid       : Message        :  Stencil 16.314 GB/s per node
-Grid       : Message        :  Stencil 18.5352 GB/s per node
-Grid       : Message        :  Stencil 13.2695 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666693
-Grid       : Message        : Average mflops/s per call per node       : 800267
-Grid       : Message        : Average mflops/s per call per node       : 826845
-Grid       : Message        : Average mflops/s per call per node       : 660822
-Grid       : Message        : Average mflops/s per call per node (full): 314168
-Grid       : Message        : Average mflops/s per call per node (full): 431113
-Grid       : Message        : Average mflops/s per call per node (full): 448873
-Grid       : Message        : Average mflops/s per call per node (full): 305170
-Grid       : Message        :  Stencil 12.9387 GB/s per node
-Grid       : Message        :  Stencil 16.4085 GB/s per node
-Grid       : Message        :  Stencil 16.8739 GB/s per node
-Grid       : Message        :  Stencil 13.6264 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670156
-Grid       : Message        : Average mflops/s per call per node       : 805849
-Grid       : Message        : Average mflops/s per call per node       : 818217
-Grid       : Message        : Average mflops/s per call per node       : 660247
-Grid       : Message        : Average mflops/s per call per node (full): 315820
-Grid       : Message        : Average mflops/s per call per node (full): 429584
-Grid       : Message        : Average mflops/s per call per node (full): 440164
-Grid       : Message        : Average mflops/s per call per node (full): 305487
-Grid       : Message        :  Stencil 12.917 GB/s per node
-Grid       : Message        :  Stencil 16.6841 GB/s per node
-Grid       : Message        :  Stencil 17.6266 GB/s per node
-Grid       : Message        :  Stencil 12.6883 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671234
-Grid       : Message        : Average mflops/s per call per node       : 806459
-Grid       : Message        : Average mflops/s per call per node       : 823954
-Grid       : Message        : Average mflops/s per call per node       : 662970
-Grid       : Message        : Average mflops/s per call per node (full): 314276
-Grid       : Message        : Average mflops/s per call per node (full): 433660
-Grid       : Message        : Average mflops/s per call per node (full): 445980
-Grid       : Message        : Average mflops/s per call per node (full): 303927
-Grid       : Message        :  Stencil 12.3094 GB/s per node
-Grid       : Message        :  Stencil 16.6002 GB/s per node
-Grid       : Message        :  Stencil 17.6975 GB/s per node
-Grid       : Message        :  Stencil 12.2183 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671314
-Grid       : Message        : Average mflops/s per call per node       : 805711
-Grid       : Message        : Average mflops/s per call per node       : 819593
-Grid       : Message        : Average mflops/s per call per node       : 664132
-Grid       : Message        : Average mflops/s per call per node (full): 312627
-Grid       : Message        : Average mflops/s per call per node (full): 435765
-Grid       : Message        : Average mflops/s per call per node (full): 445720
-Grid       : Message        : Average mflops/s per call per node (full): 303954
-Grid       : Message        :  Stencil 12.8645 GB/s per node
-Grid       : Message        :  Stencil 15.0431 GB/s per node
-Grid       : Message        :  Stencil 18.2157 GB/s per node
-Grid       : Message        :  Stencil 12.3631 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666465
-Grid       : Message        : Average mflops/s per call per node       : 800239
-Grid       : Message        : Average mflops/s per call per node       : 819023
-Grid       : Message        : Average mflops/s per call per node       : 665368
-Grid       : Message        : Average mflops/s per call per node (full): 314038
-Grid       : Message        : Average mflops/s per call per node (full): 414968
-Grid       : Message        : Average mflops/s per call per node (full): 447939
-Grid       : Message        : Average mflops/s per call per node (full): 301079
-Grid       : Message        :  Stencil 14.8432 GB/s per node
-Grid       : Message        :  Stencil 9.11744 GB/s per node
-Grid       : Message        :  Stencil 17.9773 GB/s per node
-Grid       : Message        :  Stencil 13.6707 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662854
-Grid       : Message        : Average mflops/s per call per node       : 809082
-Grid       : Message        : Average mflops/s per call per node       : 820687
-Grid       : Message        : Average mflops/s per call per node       : 662361
-Grid       : Message        : Average mflops/s per call per node (full): 315454
-Grid       : Message        : Average mflops/s per call per node (full): 295354
-Grid       : Message        : Average mflops/s per call per node (full): 446626
-Grid       : Message        : Average mflops/s per call per node (full): 305813
-Grid       : Message        :  Stencil 12.7809 GB/s per node
-Grid       : Message        :  Stencil 16.2911 GB/s per node
-Grid       : Message        :  Stencil 17.3266 GB/s per node
-Grid       : Message        :  Stencil 13.7354 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668695
-Grid       : Message        : Average mflops/s per call per node       : 803473
-Grid       : Message        : Average mflops/s per call per node       : 820087
-Grid       : Message        : Average mflops/s per call per node       : 662076
-Grid       : Message        : Average mflops/s per call per node (full): 315302
-Grid       : Message        : Average mflops/s per call per node (full): 432491
-Grid       : Message        : Average mflops/s per call per node (full): 443546
-Grid       : Message        : Average mflops/s per call per node (full): 305753
-Grid       : Message        :  Stencil 15.392 GB/s per node
-Grid       : Message        :  Stencil 16.6395 GB/s per node
-Grid       : Message        :  Stencil 18.4007 GB/s per node
-Grid       : Message        :  Stencil 11.7254 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661087
-Grid       : Message        : Average mflops/s per call per node       : 800899
-Grid       : Message        : Average mflops/s per call per node       : 823878
-Grid       : Message        : Average mflops/s per call per node       : 665867
-Grid       : Message        : Average mflops/s per call per node (full): 316630
-Grid       : Message        : Average mflops/s per call per node (full): 435665
-Grid       : Message        : Average mflops/s per call per node (full): 447474
-Grid       : Message        : Average mflops/s per call per node (full): 299133
-Grid       : Message        :  Stencil 12.7109 GB/s per node
-Grid       : Message        :  Stencil 16.6862 GB/s per node
-Grid       : Message        :  Stencil 17.2417 GB/s per node
-Grid       : Message        :  Stencil 12.1287 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668518
-Grid       : Message        : Average mflops/s per call per node       : 805298
-Grid       : Message        : Average mflops/s per call per node       : 817691
-Grid       : Message        : Average mflops/s per call per node       : 669558
-Grid       : Message        : Average mflops/s per call per node (full): 315247
-Grid       : Message        : Average mflops/s per call per node (full): 436881
-Grid       : Message        : Average mflops/s per call per node (full): 442923
-Grid       : Message        : Average mflops/s per call per node (full): 303777
-Grid       : Message        :  Stencil 14.6304 GB/s per node
-Grid       : Message        :  Stencil 17.0628 GB/s per node
-Grid       : Message        :  Stencil 17.9849 GB/s per node
-Grid       : Message        :  Stencil 12.5003 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661927
-Grid       : Message        : Average mflops/s per call per node       : 800619
-Grid       : Message        : Average mflops/s per call per node       : 824018
-Grid       : Message        : Average mflops/s per call per node       : 669003
-Grid       : Message        : Average mflops/s per call per node (full): 316799
-Grid       : Message        : Average mflops/s per call per node (full): 435870
-Grid       : Message        : Average mflops/s per call per node (full): 448908
-Grid       : Message        : Average mflops/s per call per node (full): 305761
-Grid       : Message        :  Stencil 14.2657 GB/s per node
-Grid       : Message        :  Stencil 17.7772 GB/s per node
-Grid       : Message        :  Stencil 16.4253 GB/s per node
-Grid       : Message        :  Stencil 12.0975 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662242
-Grid       : Message        : Average mflops/s per call per node       : 803620
-Grid       : Message        : Average mflops/s per call per node       : 822478
-Grid       : Message        : Average mflops/s per call per node       : 666255
-Grid       : Message        : Average mflops/s per call per node (full): 316413
-Grid       : Message        : Average mflops/s per call per node (full): 440755
-Grid       : Message        : Average mflops/s per call per node (full): 425590
-Grid       : Message        : Average mflops/s per call per node (full): 301935
-Grid       : Message        :  Stencil 15.2164 GB/s per node
-Grid       : Message        :  Stencil 9.49935 GB/s per node
-Grid       : Message        :  Stencil 17.2102 GB/s per node
-Grid       : Message        :  Stencil 13.7943 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659534
-Grid       : Message        : Average mflops/s per call per node       : 809661
-Grid       : Message        : Average mflops/s per call per node       : 823187
-Grid       : Message        : Average mflops/s per call per node       : 661130
-Grid       : Message        : Average mflops/s per call per node (full): 316258
-Grid       : Message        : Average mflops/s per call per node (full): 304426
-Grid       : Message        : Average mflops/s per call per node (full): 442628
-Grid       : Message        : Average mflops/s per call per node (full): 306161
-Grid       : Message        :  Stencil 14.6314 GB/s per node
-Grid       : Message        :  Stencil 16.992 GB/s per node
-Grid       : Message        :  Stencil 17.8637 GB/s per node
-Grid       : Message        :  Stencil 12.3308 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 658515
-Grid       : Message        : Average mflops/s per call per node       : 801185
-Grid       : Message        : Average mflops/s per call per node       : 823355
-Grid       : Message        : Average mflops/s per call per node       : 670100
-Grid       : Message        : Average mflops/s per call per node (full): 315176
-Grid       : Message        : Average mflops/s per call per node (full): 437692
-Grid       : Message        : Average mflops/s per call per node (full): 444062
-Grid       : Message        : Average mflops/s per call per node (full): 304858
-Grid       : Message        :  Stencil 12.3942 GB/s per node
-Grid       : Message        :  Stencil 17.204 GB/s per node
-Grid       : Message        :  Stencil 17.2984 GB/s per node
-Grid       : Message        :  Stencil 12.1147 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663799
-Grid       : Message        : Average mflops/s per call per node       : 800854
-Grid       : Message        : Average mflops/s per call per node       : 823783
-Grid       : Message        : Average mflops/s per call per node       : 667952
-Grid       : Message        : Average mflops/s per call per node (full): 312919
-Grid       : Message        : Average mflops/s per call per node (full): 439336
-Grid       : Message        : Average mflops/s per call per node (full): 442674
-Grid       : Message        : Average mflops/s per call per node (full): 302111
-Grid       : Message        :  Stencil 13.5046 GB/s per node
-Grid       : Message        :  Stencil 8.65121 GB/s per node
-Grid       : Message        :  Stencil 16.8146 GB/s per node
-Grid       : Message        :  Stencil 12.6679 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664548
-Grid       : Message        : Average mflops/s per call per node       : 810316
-Grid       : Message        : Average mflops/s per call per node       : 824111
-Grid       : Message        : Average mflops/s per call per node       : 668561
-Grid       : Message        : Average mflops/s per call per node (full): 315068
-Grid       : Message        : Average mflops/s per call per node (full): 283443
-Grid       : Message        : Average mflops/s per call per node (full): 438671
-Grid       : Message        : Average mflops/s per call per node (full): 305982
-Grid       : Message        :  Stencil 13.1484 GB/s per node
-Grid       : Message        :  Stencil 17.4256 GB/s per node
-Grid       : Message        :  Stencil 18.1358 GB/s per node
-Grid       : Message        :  Stencil 12.7422 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663668
-Grid       : Message        : Average mflops/s per call per node       : 800415
-Grid       : Message        : Average mflops/s per call per node       : 808348
-Grid       : Message        : Average mflops/s per call per node       : 666082
-Grid       : Message        : Average mflops/s per call per node (full): 313980
-Grid       : Message        : Average mflops/s per call per node (full): 437498
-Grid       : Message        : Average mflops/s per call per node (full): 443415
-Grid       : Message        : Average mflops/s per call per node (full): 305528
-Grid       : Message        :  Stencil 13.3168 GB/s per node
-Grid       : Message        :  Stencil 16.8823 GB/s per node
-Grid       : Message        :  Stencil 17.3926 GB/s per node
-Grid       : Message        :  Stencil 14.2876 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667078
-Grid       : Message        : Average mflops/s per call per node       : 801243
-Grid       : Message        : Average mflops/s per call per node       : 823023
-Grid       : Message        : Average mflops/s per call per node       : 661535
-Grid       : Message        : Average mflops/s per call per node (full): 315190
-Grid       : Message        : Average mflops/s per call per node (full): 432799
-Grid       : Message        : Average mflops/s per call per node (full): 442973
-Grid       : Message        : Average mflops/s per call per node (full): 303404
-Grid       : Message        :  Stencil 12.6679 GB/s per node
-Grid       : Message        :  Stencil 16.6478 GB/s per node
-Grid       : Message        :  Stencil 17.3866 GB/s per node
-Grid       : Message        :  Stencil 13.3003 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670601
-Grid       : Message        : Average mflops/s per call per node       : 802308
-Grid       : Message        : Average mflops/s per call per node       : 818386
-Grid       : Message        : Average mflops/s per call per node       : 661566
-Grid       : Message        : Average mflops/s per call per node (full): 315534
-Grid       : Message        : Average mflops/s per call per node (full): 435309
-Grid       : Message        : Average mflops/s per call per node (full): 443217
-Grid       : Message        : Average mflops/s per call per node (full): 305060
-Grid       : Message        :  Stencil 13.1465 GB/s per node
-Grid       : Message        :  Stencil 11.5329 GB/s per node
-Grid       : Message        :  Stencil 17.4103 GB/s per node
-Grid       : Message        :  Stencil 13.828 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664492
-Grid       : Message        : Average mflops/s per call per node       : 807782
-Grid       : Message        : Average mflops/s per call per node       : 829479
-Grid       : Message        : Average mflops/s per call per node       : 657773
-Grid       : Message        : Average mflops/s per call per node (full): 315037
-Grid       : Message        : Average mflops/s per call per node (full): 351401
-Grid       : Message        : Average mflops/s per call per node (full): 445392
-Grid       : Message        : Average mflops/s per call per node (full): 305751
-Grid       : Message        :  Stencil 13.773 GB/s per node
-Grid       : Message        :  Stencil 17.1506 GB/s per node
-Grid       : Message        :  Stencil 17.6133 GB/s per node
-Grid       : Message        :  Stencil 13.3226 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664042
-Grid       : Message        : Average mflops/s per call per node       : 804459
-Grid       : Message        : Average mflops/s per call per node       : 822795
-Grid       : Message        : Average mflops/s per call per node       : 664842
-Grid       : Message        : Average mflops/s per call per node (full): 316073
-Grid       : Message        : Average mflops/s per call per node (full): 437764
-Grid       : Message        : Average mflops/s per call per node (full): 446153
-Grid       : Message        : Average mflops/s per call per node (full): 299502
-Grid       : Message        :  Stencil 15.0845 GB/s per node
-Grid       : Message        :  Stencil 17.7414 GB/s per node
-Grid       : Message        :  Stencil 16.8645 GB/s per node
-Grid       : Message        :  Stencil 12.3923 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661352
-Grid       : Message        : Average mflops/s per call per node       : 800592
-Grid       : Message        : Average mflops/s per call per node       : 817293
-Grid       : Message        : Average mflops/s per call per node       : 664372
-Grid       : Message        : Average mflops/s per call per node (full): 316852
-Grid       : Message        : Average mflops/s per call per node (full): 439206
-Grid       : Message        : Average mflops/s per call per node (full): 437628
-Grid       : Message        : Average mflops/s per call per node (full): 304183
-Grid       : Message        :  Stencil 14.7196 GB/s per node
-Grid       : Message        :  Stencil 16.5359 GB/s per node
-Grid       : Message        :  Stencil 18.1543 GB/s per node
-Grid       : Message        :  Stencil 13.2842 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662539
-Grid       : Message        : Average mflops/s per call per node       : 802369
-Grid       : Message        : Average mflops/s per call per node       : 824979
-Grid       : Message        : Average mflops/s per call per node       : 661577
-Grid       : Message        : Average mflops/s per call per node (full): 316824
-Grid       : Message        : Average mflops/s per call per node (full): 434027
-Grid       : Message        : Average mflops/s per call per node (full): 446898
-Grid       : Message        : Average mflops/s per call per node (full): 305991
-Grid       : Message        :  Stencil 15.5375 GB/s per node
-Grid       : Message        :  Stencil 16.5096 GB/s per node
-Grid       : Message        :  Stencil 18.3063 GB/s per node
-Grid       : Message        :  Stencil 13.1609 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664450
-Grid       : Message        : Average mflops/s per call per node       : 802683
-Grid       : Message        : Average mflops/s per call per node       : 823743
-Grid       : Message        : Average mflops/s per call per node       : 662443
-Grid       : Message        : Average mflops/s per call per node (full): 317778
-Grid       : Message        : Average mflops/s per call per node (full): 435735
-Grid       : Message        : Average mflops/s per call per node (full): 448569
-Grid       : Message        : Average mflops/s per call per node (full): 305253
-Grid       : Message        :  Stencil 12.8296 GB/s per node
-Grid       : Message        :  Stencil 9.10781 GB/s per node
-Grid       : Message        :  Stencil 17.7547 GB/s per node
-Grid       : Message        :  Stencil 12.4499 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669156
-Grid       : Message        : Average mflops/s per call per node       : 807513
-Grid       : Message        : Average mflops/s per call per node       : 825799
-Grid       : Message        : Average mflops/s per call per node       : 670206
-Grid       : Message        : Average mflops/s per call per node (full): 314775
-Grid       : Message        : Average mflops/s per call per node (full): 294729
-Grid       : Message        : Average mflops/s per call per node (full): 446972
-Grid       : Message        : Average mflops/s per call per node (full): 305790
-Grid       : Message        :  Stencil 12.6955 GB/s per node
-Grid       : Message        :  Stencil 16.6487 GB/s per node
-Grid       : Message        :  Stencil 17.6868 GB/s per node
-Grid       : Message        :  Stencil 12.7983 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 673509
-Grid       : Message        : Average mflops/s per call per node       : 802936
-Grid       : Message        : Average mflops/s per call per node       : 827030
-Grid       : Message        : Average mflops/s per call per node       : 664901
-Grid       : Message        : Average mflops/s per call per node (full): 315828
-Grid       : Message        : Average mflops/s per call per node (full): 435932
-Grid       : Message        : Average mflops/s per call per node (full): 446750
-Grid       : Message        : Average mflops/s per call per node (full): 305108
-Grid       : Message        :  Stencil 12.6757 GB/s per node
-Grid       : Message        :  Stencil 16.1476 GB/s per node
-Grid       : Message        :  Stencil 18.0383 GB/s per node
-Grid       : Message        :  Stencil 14.2022 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 672735
-Grid       : Message        : Average mflops/s per call per node       : 802948
-Grid       : Message        : Average mflops/s per call per node       : 822948
-Grid       : Message        : Average mflops/s per call per node       : 659746
-Grid       : Message        : Average mflops/s per call per node (full): 315118
-Grid       : Message        : Average mflops/s per call per node (full): 429520
-Grid       : Message        : Average mflops/s per call per node (full): 446367
-Grid       : Message        : Average mflops/s per call per node (full): 306267
-Grid       : Message        :  Stencil 13.1888 GB/s per node
-Grid       : Message        :  Stencil 17.1689 GB/s per node
-Grid       : Message        :  Stencil 17.964 GB/s per node
-Grid       : Message        :  Stencil 12.4117 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666876
-Grid       : Message        : Average mflops/s per call per node       : 806357
-Grid       : Message        : Average mflops/s per call per node       : 820983
-Grid       : Message        : Average mflops/s per call per node       : 665899
-Grid       : Message        : Average mflops/s per call per node (full): 316192
-Grid       : Message        : Average mflops/s per call per node (full): 440065
-Grid       : Message        : Average mflops/s per call per node (full): 445936
-Grid       : Message        : Average mflops/s per call per node (full): 304403
-Grid       : Message        :  Stencil 13.9266 GB/s per node
-Grid       : Message        :  Stencil 9.79511 GB/s per node
-Grid       : Message        :  Stencil 17.4312 GB/s per node
-Grid       : Message        :  Stencil 12.2498 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665834
-Grid       : Message        : Average mflops/s per call per node       : 810976
-Grid       : Message        : Average mflops/s per call per node       : 821093
-Grid       : Message        : Average mflops/s per call per node       : 667009
-Grid       : Message        : Average mflops/s per call per node (full): 312307
-Grid       : Message        : Average mflops/s per call per node (full): 311635
-Grid       : Message        : Average mflops/s per call per node (full): 445404
-Grid       : Message        : Average mflops/s per call per node (full): 303006
-Grid       : Message        :  Stencil 12.81 GB/s per node
-Grid       : Message        :  Stencil 16.5723 GB/s per node
-Grid       : Message        :  Stencil 18.4032 GB/s per node
-Grid       : Message        :  Stencil 13.6021 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671280
-Grid       : Message        : Average mflops/s per call per node       : 806415
-Grid       : Message        : Average mflops/s per call per node       : 818215
-Grid       : Message        : Average mflops/s per call per node       : 664436
-Grid       : Message        : Average mflops/s per call per node (full): 315501
-Grid       : Message        : Average mflops/s per call per node (full): 435810
-Grid       : Message        : Average mflops/s per call per node (full): 444505
-Grid       : Message        : Average mflops/s per call per node (full): 302928
-Grid       : Message        :  Stencil 12.8577 GB/s per node
-Grid       : Message        :  Stencil 17.5319 GB/s per node
-Grid       : Message        :  Stencil 18.6292 GB/s per node
-Grid       : Message        :  Stencil 12.6488 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668092
-Grid       : Message        : Average mflops/s per call per node       : 803661
-Grid       : Message        : Average mflops/s per call per node       : 820460
-Grid       : Message        : Average mflops/s per call per node       : 670432
-Grid       : Message        : Average mflops/s per call per node (full): 315044
-Grid       : Message        : Average mflops/s per call per node (full): 439046
-Grid       : Message        : Average mflops/s per call per node (full): 447930
-Grid       : Message        : Average mflops/s per call per node (full): 304999
-Grid       : Message        :  Stencil 13.0255 GB/s per node
-Grid       : Message        :  Stencil 17.3176 GB/s per node
-Grid       : Message        :  Stencil 17.2674 GB/s per node
-Grid       : Message        :  Stencil 13.168 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668654
-Grid       : Message        : Average mflops/s per call per node       : 802245
-Grid       : Message        : Average mflops/s per call per node       : 822532
-Grid       : Message        : Average mflops/s per call per node       : 660223
-Grid       : Message        : Average mflops/s per call per node (full): 315458
-Grid       : Message        : Average mflops/s per call per node (full): 439511
-Grid       : Message        : Average mflops/s per call per node (full): 443161
-Grid       : Message        : Average mflops/s per call per node (full): 304948
-Grid       : Message        :  Stencil 13.5487 GB/s per node
-Grid       : Message        :  Stencil 17.5157 GB/s per node
-Grid       : Message        :  Stencil 17.6107 GB/s per node
-Grid       : Message        :  Stencil 13.0801 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662485
-Grid       : Message        : Average mflops/s per call per node       : 801571
-Grid       : Message        : Average mflops/s per call per node       : 825590
-Grid       : Message        : Average mflops/s per call per node       : 667309
-Grid       : Message        : Average mflops/s per call per node (full): 315559
-Grid       : Message        : Average mflops/s per call per node (full): 439667
-Grid       : Message        : Average mflops/s per call per node (full): 446588
-Grid       : Message        : Average mflops/s per call per node (full): 304921
-Grid       : Message        :  Stencil 13.1937 GB/s per node
-Grid       : Message        :  Stencil 17.687 GB/s per node
-Grid       : Message        :  Stencil 17.401 GB/s per node
-Grid       : Message        :  Stencil 13.7869 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662902
-Grid       : Message        : Average mflops/s per call per node       : 805299
-Grid       : Message        : Average mflops/s per call per node       : 822967
-Grid       : Message        : Average mflops/s per call per node       : 662737
-Grid       : Message        : Average mflops/s per call per node (full): 313476
-Grid       : Message        : Average mflops/s per call per node (full): 439484
-Grid       : Message        : Average mflops/s per call per node (full): 445173
-Grid       : Message        : Average mflops/s per call per node (full): 306650
-Grid       : Message        :  Stencil 13.5789 GB/s per node
-Grid       : Message        :  Stencil 16.1773 GB/s per node
-Grid       : Message        :  Stencil 17.6389 GB/s per node
-Grid       : Message        :  Stencil 13.2715 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664326
-Grid       : Message        : Average mflops/s per call per node       : 804113
-Grid       : Message        : Average mflops/s per call per node       : 822547
-Grid       : Message        : Average mflops/s per call per node       : 663436
-Grid       : Message        : Average mflops/s per call per node (full): 314307
-Grid       : Message        : Average mflops/s per call per node (full): 431593
-Grid       : Message        : Average mflops/s per call per node (full): 445804
-Grid       : Message        : Average mflops/s per call per node (full): 306092
-Grid       : Message        :  Stencil 13.654 GB/s per node
-Grid       : Message        :  Stencil 17.5621 GB/s per node
-Grid       : Message        :  Stencil 16.9713 GB/s per node
-Grid       : Message        :  Stencil 12.6528 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665081
-Grid       : Message        : Average mflops/s per call per node       : 805621
-Grid       : Message        : Average mflops/s per call per node       : 826388
-Grid       : Message        : Average mflops/s per call per node       : 669706
-Grid       : Message        : Average mflops/s per call per node (full): 315746
-Grid       : Message        : Average mflops/s per call per node (full): 441183
-Grid       : Message        : Average mflops/s per call per node (full): 440868
-Grid       : Message        : Average mflops/s per call per node (full): 305130
-Grid       : Message        :  Stencil 12.7877 GB/s per node
-Grid       : Message        :  Stencil 16.8454 GB/s per node
-Grid       : Message        :  Stencil 17.1362 GB/s per node
-Grid       : Message        :  Stencil 12.2521 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667733
-Grid       : Message        : Average mflops/s per call per node       : 805130
-Grid       : Message        : Average mflops/s per call per node       : 818806
-Grid       : Message        : Average mflops/s per call per node       : 667100
-Grid       : Message        : Average mflops/s per call per node (full): 314589
-Grid       : Message        : Average mflops/s per call per node (full): 436885
-Grid       : Message        : Average mflops/s per call per node (full): 441502
-Grid       : Message        : Average mflops/s per call per node (full): 303569
-Grid       : Message        :  Stencil 13.6816 GB/s per node
-Grid       : Message        :  Stencil 16.5988 GB/s per node
-Grid       : Message        :  Stencil 17.4377 GB/s per node
-Grid       : Message        :  Stencil 13.7525 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664947
-Grid       : Message        : Average mflops/s per call per node       : 807086
-Grid       : Message        : Average mflops/s per call per node       : 822228
-Grid       : Message        : Average mflops/s per call per node       : 664514
-Grid       : Message        : Average mflops/s per call per node (full): 315168
-Grid       : Message        : Average mflops/s per call per node (full): 436002
-Grid       : Message        : Average mflops/s per call per node (full): 443007
-Grid       : Message        : Average mflops/s per call per node (full): 306737
-Grid       : Message        :  Stencil 13.0057 GB/s per node
-Grid       : Message        :  Stencil 16.9441 GB/s per node
-Grid       : Message        :  Stencil 17.679 GB/s per node
-Grid       : Message        :  Stencil 12.2985 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668090
-Grid       : Message        : Average mflops/s per call per node       : 806060
-Grid       : Message        : Average mflops/s per call per node       : 826981
-Grid       : Message        : Average mflops/s per call per node       : 670588
-Grid       : Message        : Average mflops/s per call per node (full): 314896
-Grid       : Message        : Average mflops/s per call per node (full): 438351
-Grid       : Message        : Average mflops/s per call per node (full): 446426
-Grid       : Message        : Average mflops/s per call per node (full): 304257
-Grid       : Message        :  Stencil 13.1404 GB/s per node
-Grid       : Message        :  Stencil 12.6783 GB/s per node
-Grid       : Message        :  Stencil 18.8603 GB/s per node
-Grid       : Message        :  Stencil 12.6124 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665879
-Grid       : Message        : Average mflops/s per call per node       : 806963
-Grid       : Message        : Average mflops/s per call per node       : 814888
-Grid       : Message        : Average mflops/s per call per node       : 669744
-Grid       : Message        : Average mflops/s per call per node (full): 313332
-Grid       : Message        : Average mflops/s per call per node (full): 374495
-Grid       : Message        : Average mflops/s per call per node (full): 447120
-Grid       : Message        : Average mflops/s per call per node (full): 305294
-Grid       : Message        :  Stencil 12.582 GB/s per node
-Grid       : Message        :  Stencil 15.9827 GB/s per node
-Grid       : Message        :  Stencil 18.3784 GB/s per node
-Grid       : Message        :  Stencil 12.6071 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666960
-Grid       : Message        : Average mflops/s per call per node       : 801866
-Grid       : Message        : Average mflops/s per call per node       : 815942
-Grid       : Message        : Average mflops/s per call per node       : 667845
-Grid       : Message        : Average mflops/s per call per node (full): 314410
-Grid       : Message        : Average mflops/s per call per node (full): 428552
-Grid       : Message        : Average mflops/s per call per node (full): 447166
-Grid       : Message        : Average mflops/s per call per node (full): 304418
-Grid       : Message        :  Stencil 13.5239 GB/s per node
-Grid       : Message        :  Stencil 17.4 GB/s per node
-Grid       : Message        :  Stencil 16.7494 GB/s per node
-Grid       : Message        :  Stencil 12.2387 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665104
-Grid       : Message        : Average mflops/s per call per node       : 804421
-Grid       : Message        : Average mflops/s per call per node       : 821566
-Grid       : Message        : Average mflops/s per call per node       : 668199
-Grid       : Message        : Average mflops/s per call per node (full): 316575
-Grid       : Message        : Average mflops/s per call per node (full): 437744
-Grid       : Message        : Average mflops/s per call per node (full): 438409
-Grid       : Message        : Average mflops/s per call per node (full): 304113
-Grid       : Message        :  Stencil 12.9829 GB/s per node
-Grid       : Message        :  Stencil 10.9168 GB/s per node
-Grid       : Message        :  Stencil 18.3933 GB/s per node
-Grid       : Message        :  Stencil 13.774 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666011
-Grid       : Message        : Average mflops/s per call per node       : 807342
-Grid       : Message        : Average mflops/s per call per node       : 821382
-Grid       : Message        : Average mflops/s per call per node       : 660536
-Grid       : Message        : Average mflops/s per call per node (full): 314102
-Grid       : Message        : Average mflops/s per call per node (full): 337750
-Grid       : Message        : Average mflops/s per call per node (full): 447207
-Grid       : Message        : Average mflops/s per call per node (full): 303080
-Grid       : Message        :  Stencil 13.3301 GB/s per node
-Grid       : Message        :  Stencil 17.2677 GB/s per node
-Grid       : Message        :  Stencil 17.264 GB/s per node
-Grid       : Message        :  Stencil 12.5656 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663154
-Grid       : Message        : Average mflops/s per call per node       : 803852
-Grid       : Message        : Average mflops/s per call per node       : 823789
-Grid       : Message        : Average mflops/s per call per node       : 665791
-Grid       : Message        : Average mflops/s per call per node (full): 314605
-Grid       : Message        : Average mflops/s per call per node (full): 439696
-Grid       : Message        : Average mflops/s per call per node (full): 442832
-Grid       : Message        : Average mflops/s per call per node (full): 303476
-Grid       : Message        :  Stencil 13.0953 GB/s per node
-Grid       : Message        :  Stencil 16.4118 GB/s per node
-Grid       : Message        :  Stencil 17.0283 GB/s per node
-Grid       : Message        :  Stencil 12.9251 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663595
-Grid       : Message        : Average mflops/s per call per node       : 799392
-Grid       : Message        : Average mflops/s per call per node       : 821889
-Grid       : Message        : Average mflops/s per call per node       : 663291
-Grid       : Message        : Average mflops/s per call per node (full): 314648
-Grid       : Message        : Average mflops/s per call per node (full): 433149
-Grid       : Message        : Average mflops/s per call per node (full): 439817
-Grid       : Message        : Average mflops/s per call per node (full): 305093
-Grid       : Message        :  Stencil 14.1367 GB/s per node
-Grid       : Message        :  Stencil 16.7558 GB/s per node
-Grid       : Message        :  Stencil 17.9835 GB/s per node
-Grid       : Message        :  Stencil 13.1096 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665854
-Grid       : Message        : Average mflops/s per call per node       : 803889
-Grid       : Message        : Average mflops/s per call per node       : 822656
-Grid       : Message        : Average mflops/s per call per node       : 662789
-Grid       : Message        : Average mflops/s per call per node (full): 316383
-Grid       : Message        : Average mflops/s per call per node (full): 437067
-Grid       : Message        : Average mflops/s per call per node (full): 445712
-Grid       : Message        : Average mflops/s per call per node (full): 304553
-Grid       : Message        :  Stencil 14.3347 GB/s per node
-Grid       : Message        :  Stencil 15.9031 GB/s per node
-Grid       : Message        :  Stencil 17.775 GB/s per node
-Grid       : Message        :  Stencil 12.7522 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666072
-Grid       : Message        : Average mflops/s per call per node       : 804652
-Grid       : Message        : Average mflops/s per call per node       : 819328
-Grid       : Message        : Average mflops/s per call per node       : 670436
-Grid       : Message        : Average mflops/s per call per node (full): 316941
-Grid       : Message        : Average mflops/s per call per node (full): 423228
-Grid       : Message        : Average mflops/s per call per node (full): 444860
-Grid       : Message        : Average mflops/s per call per node (full): 305769
-Grid       : Message        :  Stencil 14.0882 GB/s per node
-Grid       : Message        :  Stencil 16.6305 GB/s per node
-Grid       : Message        :  Stencil 17.847 GB/s per node
-Grid       : Message        :  Stencil 13.1699 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669135
-Grid       : Message        : Average mflops/s per call per node       : 804670
-Grid       : Message        : Average mflops/s per call per node       : 821916
-Grid       : Message        : Average mflops/s per call per node       : 660372
-Grid       : Message        : Average mflops/s per call per node (full): 317075
-Grid       : Message        : Average mflops/s per call per node (full): 436490
-Grid       : Message        : Average mflops/s per call per node (full): 446030
-Grid       : Message        : Average mflops/s per call per node (full): 305999
-Grid       : Message        :  Stencil 12.3987 GB/s per node
-Grid       : Message        :  Stencil 17.0523 GB/s per node
-Grid       : Message        :  Stencil 16.8339 GB/s per node
-Grid       : Message        :  Stencil 13.2098 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667431
-Grid       : Message        : Average mflops/s per call per node       : 806676
-Grid       : Message        : Average mflops/s per call per node       : 819751
-Grid       : Message        : Average mflops/s per call per node       : 660084
-Grid       : Message        : Average mflops/s per call per node (full): 312865
-Grid       : Message        : Average mflops/s per call per node (full): 439743
-Grid       : Message        : Average mflops/s per call per node (full): 438997
-Grid       : Message        : Average mflops/s per call per node (full): 303877
-Grid       : Message        :  Stencil 13.8262 GB/s per node
-Grid       : Message        :  Stencil 16.6954 GB/s per node
-Grid       : Message        :  Stencil 18.417 GB/s per node
-Grid       : Message        :  Stencil 13.1479 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668787
-Grid       : Message        : Average mflops/s per call per node       : 798666
-Grid       : Message        : Average mflops/s per call per node       : 831464
-Grid       : Message        : Average mflops/s per call per node       : 668642
-Grid       : Message        : Average mflops/s per call per node (full): 317648
-Grid       : Message        : Average mflops/s per call per node (full): 436717
-Grid       : Message        : Average mflops/s per call per node (full): 450757
-Grid       : Message        : Average mflops/s per call per node (full): 306965
-Grid       : Message        :  Stencil 13.7838 GB/s per node
-Grid       : Message        :  Stencil 9.57078 GB/s per node
-Grid       : Message        :  Stencil 16.9114 GB/s per node
-Grid       : Message        :  Stencil 12.9244 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667864
-Grid       : Message        : Average mflops/s per call per node       : 810482
-Grid       : Message        : Average mflops/s per call per node       : 825563
-Grid       : Message        : Average mflops/s per call per node       : 666999
-Grid       : Message        : Average mflops/s per call per node (full): 316673
-Grid       : Message        : Average mflops/s per call per node (full): 305942
-Grid       : Message        : Average mflops/s per call per node (full): 440459
-Grid       : Message        : Average mflops/s per call per node (full): 303764
-Grid       : Message        :  Stencil 14.3677 GB/s per node
-Grid       : Message        :  Stencil 15.2322 GB/s per node
-Grid       : Message        :  Stencil 17.9007 GB/s per node
-Grid       : Message        :  Stencil 13.0004 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664683
-Grid       : Message        : Average mflops/s per call per node       : 805400
-Grid       : Message        : Average mflops/s per call per node       : 817421
-Grid       : Message        : Average mflops/s per call per node       : 665777
-Grid       : Message        : Average mflops/s per call per node (full): 316767
-Grid       : Message        : Average mflops/s per call per node (full): 417760
-Grid       : Message        : Average mflops/s per call per node (full): 446308
-Grid       : Message        : Average mflops/s per call per node (full): 303530
-Grid       : Message        :  Stencil 13.3195 GB/s per node
-Grid       : Message        :  Stencil 16.2863 GB/s per node
-Grid       : Message        :  Stencil 18.3161 GB/s per node
-Grid       : Message        :  Stencil 13.6078 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667666
-Grid       : Message        : Average mflops/s per call per node       : 803458
-Grid       : Message        : Average mflops/s per call per node       : 820298
-Grid       : Message        : Average mflops/s per call per node       : 661331
-Grid       : Message        : Average mflops/s per call per node (full): 316283
-Grid       : Message        : Average mflops/s per call per node (full): 431512
-Grid       : Message        : Average mflops/s per call per node (full): 448171
-Grid       : Message        : Average mflops/s per call per node (full): 306466
-Grid       : Message        :  Stencil 12.7435 GB/s per node
-Grid       : Message        :  Stencil 17.6106 GB/s per node
-Grid       : Message        :  Stencil 18.4497 GB/s per node
-Grid       : Message        :  Stencil 12.8518 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667716
-Grid       : Message        : Average mflops/s per call per node       : 802090
-Grid       : Message        : Average mflops/s per call per node       : 819461
-Grid       : Message        : Average mflops/s per call per node       : 659626
-Grid       : Message        : Average mflops/s per call per node (full): 315239
-Grid       : Message        : Average mflops/s per call per node (full): 439084
-Grid       : Message        : Average mflops/s per call per node (full): 446296
-Grid       : Message        : Average mflops/s per call per node (full): 304276
-Grid       : Message        :  Stencil 13.4268 GB/s per node
-Grid       : Message        :  Stencil 16.7193 GB/s per node
-Grid       : Message        :  Stencil 17.3289 GB/s per node
-Grid       : Message        :  Stencil 12.5761 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665303
-Grid       : Message        : Average mflops/s per call per node       : 806654
-Grid       : Message        : Average mflops/s per call per node       : 822986
-Grid       : Message        : Average mflops/s per call per node       : 669838
-Grid       : Message        : Average mflops/s per call per node (full): 315531
-Grid       : Message        : Average mflops/s per call per node (full): 437072
-Grid       : Message        : Average mflops/s per call per node (full): 443806
-Grid       : Message        : Average mflops/s per call per node (full): 305796
-Grid       : Message        :  Stencil 14.2954 GB/s per node
-Grid       : Message        :  Stencil 16.7181 GB/s per node
-Grid       : Message        :  Stencil 17.119 GB/s per node
-Grid       : Message        :  Stencil 12.2928 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660850
-Grid       : Message        : Average mflops/s per call per node       : 805612
-Grid       : Message        : Average mflops/s per call per node       : 822817
-Grid       : Message        : Average mflops/s per call per node       : 662549
-Grid       : Message        : Average mflops/s per call per node (full): 315957
-Grid       : Message        : Average mflops/s per call per node (full): 436318
-Grid       : Message        : Average mflops/s per call per node (full): 439364
-Grid       : Message        : Average mflops/s per call per node (full): 297783
-Grid       : Message        :  Stencil 12.7076 GB/s per node
-Grid       : Message        :  Stencil 8.72101 GB/s per node
-Grid       : Message        :  Stencil 18.2561 GB/s per node
-Grid       : Message        :  Stencil 13.1808 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668366
-Grid       : Message        : Average mflops/s per call per node       : 806517
-Grid       : Message        : Average mflops/s per call per node       : 821816
-Grid       : Message        : Average mflops/s per call per node       : 662894
-Grid       : Message        : Average mflops/s per call per node (full): 314523
-Grid       : Message        : Average mflops/s per call per node (full): 285263
-Grid       : Message        : Average mflops/s per call per node (full): 448258
-Grid       : Message        : Average mflops/s per call per node (full): 306647
-Grid       : Message        :  Stencil 12.9634 GB/s per node
-Grid       : Message        :  Stencil 17.0186 GB/s per node
-Grid       : Message        :  Stencil 17.4637 GB/s per node
-Grid       : Message        :  Stencil 12.5254 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666297
-Grid       : Message        : Average mflops/s per call per node       : 804426
-Grid       : Message        : Average mflops/s per call per node       : 829078
-Grid       : Message        : Average mflops/s per call per node       : 666390
-Grid       : Message        : Average mflops/s per call per node (full): 313370
-Grid       : Message        : Average mflops/s per call per node (full): 437900
-Grid       : Message        : Average mflops/s per call per node (full): 444086
-Grid       : Message        : Average mflops/s per call per node (full): 304885
-Grid       : Message        :  Stencil 12.691 GB/s per node
-Grid       : Message        :  Stencil 14.2967 GB/s per node
-Grid       : Message        :  Stencil 18.1624 GB/s per node
-Grid       : Message        :  Stencil 14.6255 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667652
-Grid       : Message        : Average mflops/s per call per node       : 802604
-Grid       : Message        : Average mflops/s per call per node       : 821366
-Grid       : Message        : Average mflops/s per call per node       : 662911
-Grid       : Message        : Average mflops/s per call per node (full): 312773
-Grid       : Message        : Average mflops/s per call per node (full): 403134
-Grid       : Message        : Average mflops/s per call per node (full): 445599
-Grid       : Message        : Average mflops/s per call per node (full): 307073
-Grid       : Message        :  Stencil 13.343 GB/s per node
-Grid       : Message        :  Stencil 13.1466 GB/s per node
-Grid       : Message        :  Stencil 17.5352 GB/s per node
-Grid       : Message        :  Stencil 13.6102 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665664
-Grid       : Message        : Average mflops/s per call per node       : 806207
-Grid       : Message        : Average mflops/s per call per node       : 818724
-Grid       : Message        : Average mflops/s per call per node       : 662443
-Grid       : Message        : Average mflops/s per call per node (full): 314916
-Grid       : Message        : Average mflops/s per call per node (full): 377419
-Grid       : Message        : Average mflops/s per call per node (full): 445019
-Grid       : Message        : Average mflops/s per call per node (full): 305836
-Grid       : Message        :  Stencil 13.2107 GB/s per node
-Grid       : Message        :  Stencil 17.44 GB/s per node
-Grid       : Message        :  Stencil 16.9745 GB/s per node
-Grid       : Message        :  Stencil 11.7252 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666774
-Grid       : Message        : Average mflops/s per call per node       : 805953
-Grid       : Message        : Average mflops/s per call per node       : 814111
-Grid       : Message        : Average mflops/s per call per node       : 662423
-Grid       : Message        : Average mflops/s per call per node (full): 313687
-Grid       : Message        : Average mflops/s per call per node (full): 442093
-Grid       : Message        : Average mflops/s per call per node (full): 440174
-Grid       : Message        : Average mflops/s per call per node (full): 298957
-Grid       : Message        :  Stencil 12.3945 GB/s per node
-Grid       : Message        :  Stencil 8.02693 GB/s per node
-Grid       : Message        :  Stencil 17.5639 GB/s per node
-Grid       : Message        :  Stencil 13.2315 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667677
-Grid       : Message        : Average mflops/s per call per node       : 811328
-Grid       : Message        : Average mflops/s per call per node       : 821375
-Grid       : Message        : Average mflops/s per call per node       : 666391
-Grid       : Message        : Average mflops/s per call per node (full): 312447
-Grid       : Message        : Average mflops/s per call per node (full): 267131
-Grid       : Message        : Average mflops/s per call per node (full): 444425
-Grid       : Message        : Average mflops/s per call per node (full): 306431
-Grid       : Message        :  Stencil 12.422 GB/s per node
-Grid       : Message        :  Stencil 16.6346 GB/s per node
-Grid       : Message        :  Stencil 18.3315 GB/s per node
-Grid       : Message        :  Stencil 12.0226 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668795
-Grid       : Message        : Average mflops/s per call per node       : 805787
-Grid       : Message        : Average mflops/s per call per node       : 822035
-Grid       : Message        : Average mflops/s per call per node       : 670737
-Grid       : Message        : Average mflops/s per call per node (full): 312813
-Grid       : Message        : Average mflops/s per call per node (full): 435939
-Grid       : Message        : Average mflops/s per call per node (full): 446949
-Grid       : Message        : Average mflops/s per call per node (full): 302645
-Grid       : Message        :  Stencil 13.6331 GB/s per node
-Grid       : Message        :  Stencil 16.8444 GB/s per node
-Grid       : Message        :  Stencil 17.3844 GB/s per node
-Grid       : Message        :  Stencil 12.7145 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662530
-Grid       : Message        : Average mflops/s per call per node       : 802929
-Grid       : Message        : Average mflops/s per call per node       : 823807
-Grid       : Message        : Average mflops/s per call per node       : 660727
-Grid       : Message        : Average mflops/s per call per node (full): 310885
-Grid       : Message        : Average mflops/s per call per node (full): 432574
-Grid       : Message        : Average mflops/s per call per node (full): 444738
-Grid       : Message        : Average mflops/s per call per node (full): 303776
-Grid       : Message        :  Stencil 13.319 GB/s per node
-Grid       : Message        :  Stencil 16.6364 GB/s per node
-Grid       : Message        :  Stencil 17.5618 GB/s per node
-Grid       : Message        :  Stencil 12.1875 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666624
-Grid       : Message        : Average mflops/s per call per node       : 802022
-Grid       : Message        : Average mflops/s per call per node       : 820670
-Grid       : Message        : Average mflops/s per call per node       : 662307
-Grid       : Message        : Average mflops/s per call per node (full): 315583
-Grid       : Message        : Average mflops/s per call per node (full): 433871
-Grid       : Message        : Average mflops/s per call per node (full): 445169
-Grid       : Message        : Average mflops/s per call per node (full): 295793
-Grid       : Message        :  Stencil 13.5188 GB/s per node
-Grid       : Message        :  Stencil 17.4021 GB/s per node
-Grid       : Message        :  Stencil 18.2654 GB/s per node
-Grid       : Message        :  Stencil 14.2935 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664121
-Grid       : Message        : Average mflops/s per call per node       : 802832
-Grid       : Message        : Average mflops/s per call per node       : 820333
-Grid       : Message        : Average mflops/s per call per node       : 662363
-Grid       : Message        : Average mflops/s per call per node (full): 315530
-Grid       : Message        : Average mflops/s per call per node (full): 437807
-Grid       : Message        : Average mflops/s per call per node (full): 446585
-Grid       : Message        : Average mflops/s per call per node (full): 306735
-Grid       : Message        :  Stencil 13.6374 GB/s per node
-Grid       : Message        :  Stencil 16.7984 GB/s per node
-Grid       : Message        :  Stencil 17.4589 GB/s per node
-Grid       : Message        :  Stencil 13.3826 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659833
-Grid       : Message        : Average mflops/s per call per node       : 807520
-Grid       : Message        : Average mflops/s per call per node       : 823853
-Grid       : Message        : Average mflops/s per call per node       : 665819
-Grid       : Message        : Average mflops/s per call per node (full): 314759
-Grid       : Message        : Average mflops/s per call per node (full): 436571
-Grid       : Message        : Average mflops/s per call per node (full): 445851
-Grid       : Message        : Average mflops/s per call per node (full): 306885
-Grid       : Message        :  Stencil 13.1851 GB/s per node
-Grid       : Message        :  Stencil 9.70481 GB/s per node
-Grid       : Message        :  Stencil 17.4339 GB/s per node
-Grid       : Message        :  Stencil 12.6306 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661627
-Grid       : Message        : Average mflops/s per call per node       : 805451
-Grid       : Message        : Average mflops/s per call per node       : 817025
-Grid       : Message        : Average mflops/s per call per node       : 664777
-Grid       : Message        : Average mflops/s per call per node (full): 312092
-Grid       : Message        : Average mflops/s per call per node (full): 309023
-Grid       : Message        : Average mflops/s per call per node (full): 443869
-Grid       : Message        : Average mflops/s per call per node (full): 304307
-Grid       : Message        :  Stencil 13.8383 GB/s per node
-Grid       : Message        :  Stencil 17.767 GB/s per node
-Grid       : Message        :  Stencil 17.4802 GB/s per node
-Grid       : Message        :  Stencil 13.0814 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662545
-Grid       : Message        : Average mflops/s per call per node       : 804357
-Grid       : Message        : Average mflops/s per call per node       : 823795
-Grid       : Message        : Average mflops/s per call per node       : 668634
-Grid       : Message        : Average mflops/s per call per node (full): 314985
-Grid       : Message        : Average mflops/s per call per node (full): 437156
-Grid       : Message        : Average mflops/s per call per node (full): 445449
-Grid       : Message        : Average mflops/s per call per node (full): 306406
-Grid       : Message        :  Stencil 12.8009 GB/s per node
-Grid       : Message        :  Stencil 17.7658 GB/s per node
-Grid       : Message        :  Stencil 16.9127 GB/s per node
-Grid       : Message        :  Stencil 13.0729 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664961
-Grid       : Message        : Average mflops/s per call per node       : 798250
-Grid       : Message        : Average mflops/s per call per node       : 817036
-Grid       : Message        : Average mflops/s per call per node       : 666474
-Grid       : Message        : Average mflops/s per call per node (full): 313921
-Grid       : Message        : Average mflops/s per call per node (full): 437420
-Grid       : Message        : Average mflops/s per call per node (full): 438986
-Grid       : Message        : Average mflops/s per call per node (full): 305255
-Grid       : Message        :  Stencil 12.5554 GB/s per node
-Grid       : Message        :  Stencil 15.1018 GB/s per node
-Grid       : Message        :  Stencil 17.9067 GB/s per node
-Grid       : Message        :  Stencil 11.922 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663244
-Grid       : Message        : Average mflops/s per call per node       : 808394
-Grid       : Message        : Average mflops/s per call per node       : 816336
-Grid       : Message        : Average mflops/s per call per node       : 669089
-Grid       : Message        : Average mflops/s per call per node (full): 312840
-Grid       : Message        : Average mflops/s per call per node (full): 417492
-Grid       : Message        : Average mflops/s per call per node (full): 446369
-Grid       : Message        : Average mflops/s per call per node (full): 300965
-Grid       : Message        :  Stencil 14.4247 GB/s per node
-Grid       : Message        :  Stencil 16.806 GB/s per node
-Grid       : Message        :  Stencil 17.3214 GB/s per node
-Grid       : Message        :  Stencil 13.4102 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662793
-Grid       : Message        : Average mflops/s per call per node       : 805114
-Grid       : Message        : Average mflops/s per call per node       : 820028
-Grid       : Message        : Average mflops/s per call per node       : 663970
-Grid       : Message        : Average mflops/s per call per node (full): 316120
-Grid       : Message        : Average mflops/s per call per node (full): 436892
-Grid       : Message        : Average mflops/s per call per node (full): 444044
-Grid       : Message        : Average mflops/s per call per node (full): 306598
-Grid       : Message        :  Stencil 13.1868 GB/s per node
-Grid       : Message        :  Stencil 16.79 GB/s per node
-Grid       : Message        :  Stencil 17.9053 GB/s per node
-Grid       : Message        :  Stencil 12.5273 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665115
-Grid       : Message        : Average mflops/s per call per node       : 800161
-Grid       : Message        : Average mflops/s per call per node       : 821672
-Grid       : Message        : Average mflops/s per call per node       : 668719
-Grid       : Message        : Average mflops/s per call per node (full): 315324
-Grid       : Message        : Average mflops/s per call per node (full): 429540
-Grid       : Message        : Average mflops/s per call per node (full): 446378
-Grid       : Message        : Average mflops/s per call per node (full): 306237
-Grid       : Message        :  Stencil 13.2253 GB/s per node
-Grid       : Message        :  Stencil 16.8698 GB/s per node
-Grid       : Message        :  Stencil 18.2046 GB/s per node
-Grid       : Message        :  Stencil 12.7647 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667962
-Grid       : Message        : Average mflops/s per call per node       : 802042
-Grid       : Message        : Average mflops/s per call per node       : 820920
-Grid       : Message        : Average mflops/s per call per node       : 668023
-Grid       : Message        : Average mflops/s per call per node (full): 314959
-Grid       : Message        : Average mflops/s per call per node (full): 438077
-Grid       : Message        : Average mflops/s per call per node (full): 445637
-Grid       : Message        : Average mflops/s per call per node (full): 305345
-Grid       : Message        :  Stencil 13.2723 GB/s per node
-Grid       : Message        :  Stencil 17.0483 GB/s per node
-Grid       : Message        :  Stencil 17.3302 GB/s per node
-Grid       : Message        :  Stencil 12.4456 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664764
-Grid       : Message        : Average mflops/s per call per node       : 803607
-Grid       : Message        : Average mflops/s per call per node       : 820416
-Grid       : Message        : Average mflops/s per call per node       : 665533
-Grid       : Message        : Average mflops/s per call per node (full): 315139
-Grid       : Message        : Average mflops/s per call per node (full): 439429
-Grid       : Message        : Average mflops/s per call per node (full): 444492
-Grid       : Message        : Average mflops/s per call per node (full): 304859
-Grid       : Message        :  Stencil 14.0213 GB/s per node
-Grid       : Message        :  Stencil 13.8357 GB/s per node
-Grid       : Message        :  Stencil 17.182 GB/s per node
-Grid       : Message        :  Stencil 12.6582 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664981
-Grid       : Message        : Average mflops/s per call per node       : 806282
-Grid       : Message        : Average mflops/s per call per node       : 824095
-Grid       : Message        : Average mflops/s per call per node       : 665473
-Grid       : Message        : Average mflops/s per call per node (full): 316013
-Grid       : Message        : Average mflops/s per call per node (full): 396530
-Grid       : Message        : Average mflops/s per call per node (full): 442563
-Grid       : Message        : Average mflops/s per call per node (full): 304489
-Grid       : Message        :  Stencil 15.2683 GB/s per node
-Grid       : Message        :  Stencil 10.9317 GB/s per node
-Grid       : Message        :  Stencil 17.2713 GB/s per node
-Grid       : Message        :  Stencil 13.4705 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661032
-Grid       : Message        : Average mflops/s per call per node       : 808017
-Grid       : Message        : Average mflops/s per call per node       : 819256
-Grid       : Message        : Average mflops/s per call per node       : 669147
-Grid       : Message        : Average mflops/s per call per node (full): 316887
-Grid       : Message        : Average mflops/s per call per node (full): 337035
-Grid       : Message        : Average mflops/s per call per node (full): 442321
-Grid       : Message        : Average mflops/s per call per node (full): 306892
-Grid       : Message        :  Stencil 13.1559 GB/s per node
-Grid       : Message        :  Stencil 16.449 GB/s per node
-Grid       : Message        :  Stencil 17.9936 GB/s per node
-Grid       : Message        :  Stencil 13.389 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667960
-Grid       : Message        : Average mflops/s per call per node       : 803262
-Grid       : Message        : Average mflops/s per call per node       : 823091
-Grid       : Message        : Average mflops/s per call per node       : 664836
-Grid       : Message        : Average mflops/s per call per node (full): 315214
-Grid       : Message        : Average mflops/s per call per node (full): 433790
-Grid       : Message        : Average mflops/s per call per node (full): 441761
-Grid       : Message        : Average mflops/s per call per node (full): 301416
-Grid       : Message        :  Stencil 13.7506 GB/s per node
-Grid       : Message        :  Stencil 16.4977 GB/s per node
-Grid       : Message        :  Stencil 17.039 GB/s per node
-Grid       : Message        :  Stencil 15.0107 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663396
-Grid       : Message        : Average mflops/s per call per node       : 810470
-Grid       : Message        : Average mflops/s per call per node       : 824869
-Grid       : Message        : Average mflops/s per call per node       : 659004
-Grid       : Message        : Average mflops/s per call per node (full): 316219
-Grid       : Message        : Average mflops/s per call per node (full): 435982
-Grid       : Message        : Average mflops/s per call per node (full): 441402
-Grid       : Message        : Average mflops/s per call per node (full): 306600
-Grid       : Message        :  Stencil 13.6079 GB/s per node
-Grid       : Message        :  Stencil 16.9354 GB/s per node
-Grid       : Message        :  Stencil 16.8547 GB/s per node
-Grid       : Message        :  Stencil 13.1611 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665147
-Grid       : Message        : Average mflops/s per call per node       : 803221
-Grid       : Message        : Average mflops/s per call per node       : 817284
-Grid       : Message        : Average mflops/s per call per node       : 664592
-Grid       : Message        : Average mflops/s per call per node (full): 316696
-Grid       : Message        : Average mflops/s per call per node (full): 438191
-Grid       : Message        : Average mflops/s per call per node (full): 439513
-Grid       : Message        : Average mflops/s per call per node (full): 306013
-Grid       : Message        :  Stencil 12.8309 GB/s per node
-Grid       : Message        :  Stencil 17.2165 GB/s per node
-Grid       : Message        :  Stencil 17.9916 GB/s per node
-Grid       : Message        :  Stencil 13.2875 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665799
-Grid       : Message        : Average mflops/s per call per node       : 803556
-Grid       : Message        : Average mflops/s per call per node       : 824068
-Grid       : Message        : Average mflops/s per call per node       : 666535
-Grid       : Message        : Average mflops/s per call per node (full): 313823
-Grid       : Message        : Average mflops/s per call per node (full): 438727
-Grid       : Message        : Average mflops/s per call per node (full): 446218
-Grid       : Message        : Average mflops/s per call per node (full): 306419
-Grid       : Message        :  Stencil 13.2736 GB/s per node
-Grid       : Message        :  Stencil 17.2401 GB/s per node
-Grid       : Message        :  Stencil 17.6393 GB/s per node
-Grid       : Message        :  Stencil 12.4333 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664483
-Grid       : Message        : Average mflops/s per call per node       : 802986
-Grid       : Message        : Average mflops/s per call per node       : 825353
-Grid       : Message        : Average mflops/s per call per node       : 671556
-Grid       : Message        : Average mflops/s per call per node (full): 315141
-Grid       : Message        : Average mflops/s per call per node (full): 431488
-Grid       : Message        : Average mflops/s per call per node (full): 447340
-Grid       : Message        : Average mflops/s per call per node (full): 305656
-Grid       : Message        :  Stencil 12.2857 GB/s per node
-Grid       : Message        :  Stencil 16.7884 GB/s per node
-Grid       : Message        :  Stencil 17.3337 GB/s per node
-Grid       : Message        :  Stencil 12.6979 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669141
-Grid       : Message        : Average mflops/s per call per node       : 807060
-Grid       : Message        : Average mflops/s per call per node       : 826403
-Grid       : Message        : Average mflops/s per call per node       : 669655
-Grid       : Message        : Average mflops/s per call per node (full): 312097
-Grid       : Message        : Average mflops/s per call per node (full): 437245
-Grid       : Message        : Average mflops/s per call per node (full): 444011
-Grid       : Message        : Average mflops/s per call per node (full): 306295
-Grid       : Message        :  Stencil 13.0719 GB/s per node
-Grid       : Message        :  Stencil 17.8789 GB/s per node
-Grid       : Message        :  Stencil 16.9782 GB/s per node
-Grid       : Message        :  Stencil 12.5719 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665336
-Grid       : Message        : Average mflops/s per call per node       : 806073
-Grid       : Message        : Average mflops/s per call per node       : 820607
-Grid       : Message        : Average mflops/s per call per node       : 660984
-Grid       : Message        : Average mflops/s per call per node (full): 315686
-Grid       : Message        : Average mflops/s per call per node (full): 440215
-Grid       : Message        : Average mflops/s per call per node (full): 439389
-Grid       : Message        : Average mflops/s per call per node (full): 304164
-Grid       : Message        :  Stencil 13.3607 GB/s per node
-Grid       : Message        :  Stencil 16.2373 GB/s per node
-Grid       : Message        :  Stencil 17.749 GB/s per node
-Grid       : Message        :  Stencil 13.0617 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667798
-Grid       : Message        : Average mflops/s per call per node       : 808757
-Grid       : Message        : Average mflops/s per call per node       : 826930
-Grid       : Message        : Average mflops/s per call per node       : 667184
-Grid       : Message        : Average mflops/s per call per node (full): 316106
-Grid       : Message        : Average mflops/s per call per node (full): 431650
-Grid       : Message        : Average mflops/s per call per node (full): 446141
-Grid       : Message        : Average mflops/s per call per node (full): 305764
-Grid       : Message        :  Stencil 12.1402 GB/s per node
-Grid       : Message        :  Stencil 17.6135 GB/s per node
-Grid       : Message        :  Stencil 18.115 GB/s per node
-Grid       : Message        :  Stencil 12.9154 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669426
-Grid       : Message        : Average mflops/s per call per node       : 802030
-Grid       : Message        : Average mflops/s per call per node       : 821277
-Grid       : Message        : Average mflops/s per call per node       : 664313
-Grid       : Message        : Average mflops/s per call per node (full): 310554
-Grid       : Message        : Average mflops/s per call per node (full): 438004
-Grid       : Message        : Average mflops/s per call per node (full): 448393
-Grid       : Message        : Average mflops/s per call per node (full): 303893
-Grid       : Message        :  Stencil 12.8153 GB/s per node
-Grid       : Message        :  Stencil 17.2724 GB/s per node
-Grid       : Message        :  Stencil 16.7937 GB/s per node
-Grid       : Message        :  Stencil 12.4661 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666144
-Grid       : Message        : Average mflops/s per call per node       : 800199
-Grid       : Message        : Average mflops/s per call per node       : 825550
-Grid       : Message        : Average mflops/s per call per node       : 667956
-Grid       : Message        : Average mflops/s per call per node (full): 314030
-Grid       : Message        : Average mflops/s per call per node (full): 436611
-Grid       : Message        : Average mflops/s per call per node (full): 438304
-Grid       : Message        : Average mflops/s per call per node (full): 303741
-Grid       : Message        :  Stencil 13.1943 GB/s per node
-Grid       : Message        :  Stencil 14.5485 GB/s per node
-Grid       : Message        :  Stencil 17.4574 GB/s per node
-Grid       : Message        :  Stencil 13.4498 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667448
-Grid       : Message        : Average mflops/s per call per node       : 806169
-Grid       : Message        : Average mflops/s per call per node       : 824639
-Grid       : Message        : Average mflops/s per call per node       : 666959
-Grid       : Message        : Average mflops/s per call per node (full): 315919
-Grid       : Message        : Average mflops/s per call per node (full): 408858
-Grid       : Message        : Average mflops/s per call per node (full): 445835
-Grid       : Message        : Average mflops/s per call per node (full): 306758
-Grid       : Message        :  Stencil 12.5642 GB/s per node
-Grid       : Message        :  Stencil 16.9379 GB/s per node
-Grid       : Message        :  Stencil 18.1868 GB/s per node
-Grid       : Message        :  Stencil 12.0624 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668822
-Grid       : Message        : Average mflops/s per call per node       : 803972
-Grid       : Message        : Average mflops/s per call per node       : 820042
-Grid       : Message        : Average mflops/s per call per node       : 672336
-Grid       : Message        : Average mflops/s per call per node (full): 314087
-Grid       : Message        : Average mflops/s per call per node (full): 436495
-Grid       : Message        : Average mflops/s per call per node (full): 446760
-Grid       : Message        : Average mflops/s per call per node (full): 302805
-Grid       : Message        :  Stencil 12.4464 GB/s per node
-Grid       : Message        :  Stencil 10.9115 GB/s per node
-Grid       : Message        :  Stencil 17.3986 GB/s per node
-Grid       : Message        :  Stencil 12.7824 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667093
-Grid       : Message        : Average mflops/s per call per node       : 815614
-Grid       : Message        : Average mflops/s per call per node       : 830593
-Grid       : Message        : Average mflops/s per call per node       : 656955
-Grid       : Message        : Average mflops/s per call per node (full): 314114
-Grid       : Message        : Average mflops/s per call per node (full): 337799
-Grid       : Message        : Average mflops/s per call per node (full): 445295
-Grid       : Message        : Average mflops/s per call per node (full): 299650
-Grid       : Message        :  Stencil 12.5693 GB/s per node
-Grid       : Message        :  Stencil 16.035 GB/s per node
-Grid       : Message        :  Stencil 16.6488 GB/s per node
-Grid       : Message        :  Stencil 13.8067 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665625
-Grid       : Message        : Average mflops/s per call per node       : 807346
-Grid       : Message        : Average mflops/s per call per node       : 830478
-Grid       : Message        : Average mflops/s per call per node       : 667173
-Grid       : Message        : Average mflops/s per call per node (full): 313665
-Grid       : Message        : Average mflops/s per call per node (full): 426106
-Grid       : Message        : Average mflops/s per call per node (full): 426730
-Grid       : Message        : Average mflops/s per call per node (full): 305773
-Grid       : Message        :  Stencil 13.5136 GB/s per node
-Grid       : Message        :  Stencil 9.41642 GB/s per node
-Grid       : Message        :  Stencil 19.2267 GB/s per node
-Grid       : Message        :  Stencil 12.8042 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662943
-Grid       : Message        : Average mflops/s per call per node       : 812832
-Grid       : Message        : Average mflops/s per call per node       : 818514
-Grid       : Message        : Average mflops/s per call per node       : 663012
-Grid       : Message        : Average mflops/s per call per node (full): 314230
-Grid       : Message        : Average mflops/s per call per node (full): 302620
-Grid       : Message        : Average mflops/s per call per node (full): 449719
-Grid       : Message        : Average mflops/s per call per node (full): 305924
-Grid       : Message        :  Stencil 12.9002 GB/s per node
-Grid       : Message        :  Stencil 17.2748 GB/s per node
-Grid       : Message        :  Stencil 17.5933 GB/s per node
-Grid       : Message        :  Stencil 13.646 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671705
-Grid       : Message        : Average mflops/s per call per node       : 803725
-Grid       : Message        : Average mflops/s per call per node       : 821077
-Grid       : Message        : Average mflops/s per call per node       : 661167
-Grid       : Message        : Average mflops/s per call per node (full): 315733
-Grid       : Message        : Average mflops/s per call per node (full): 438357
-Grid       : Message        : Average mflops/s per call per node (full): 445776
-Grid       : Message        : Average mflops/s per call per node (full): 305797
-Grid       : Message        :  Stencil 13.0162 GB/s per node
-Grid       : Message        :  Stencil 16.996 GB/s per node
-Grid       : Message        :  Stencil 17.4666 GB/s per node
-Grid       : Message        :  Stencil 13.2392 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664654
-Grid       : Message        : Average mflops/s per call per node       : 802776
-Grid       : Message        : Average mflops/s per call per node       : 820496
-Grid       : Message        : Average mflops/s per call per node       : 661304
-Grid       : Message        : Average mflops/s per call per node (full): 314583
-Grid       : Message        : Average mflops/s per call per node (full): 437784
-Grid       : Message        : Average mflops/s per call per node (full): 444315
-Grid       : Message        : Average mflops/s per call per node (full): 305299
-Grid       : Message        :  Stencil 13.4746 GB/s per node
-Grid       : Message        :  Stencil 16.9801 GB/s per node
-Grid       : Message        :  Stencil 17.6159 GB/s per node
-Grid       : Message        :  Stencil 13.1317 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666274
-Grid       : Message        : Average mflops/s per call per node       : 802737
-Grid       : Message        : Average mflops/s per call per node       : 822278
-Grid       : Message        : Average mflops/s per call per node       : 658715
-Grid       : Message        : Average mflops/s per call per node (full): 314579
-Grid       : Message        : Average mflops/s per call per node (full): 432928
-Grid       : Message        : Average mflops/s per call per node (full): 445152
-Grid       : Message        : Average mflops/s per call per node (full): 303967
-Grid       : Message        :  Stencil 13.2933 GB/s per node
-Grid       : Message        :  Stencil 16.6674 GB/s per node
-Grid       : Message        :  Stencil 17.6578 GB/s per node
-Grid       : Message        :  Stencil 12.655 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663792
-Grid       : Message        : Average mflops/s per call per node       : 805521
-Grid       : Message        : Average mflops/s per call per node       : 817418
-Grid       : Message        : Average mflops/s per call per node       : 667583
-Grid       : Message        : Average mflops/s per call per node (full): 315054
-Grid       : Message        : Average mflops/s per call per node (full): 432370
-Grid       : Message        : Average mflops/s per call per node (full): 445767
-Grid       : Message        : Average mflops/s per call per node (full): 304227
-Grid       : Message        :  Stencil 14.0909 GB/s per node
-Grid       : Message        :  Stencil 18.2789 GB/s per node
-Grid       : Message        :  Stencil 17.553 GB/s per node
-Grid       : Message        :  Stencil 11.7614 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664315
-Grid       : Message        : Average mflops/s per call per node       : 801435
-Grid       : Message        : Average mflops/s per call per node       : 822475
-Grid       : Message        : Average mflops/s per call per node       : 666305
-Grid       : Message        : Average mflops/s per call per node (full): 316926
-Grid       : Message        : Average mflops/s per call per node (full): 439618
-Grid       : Message        : Average mflops/s per call per node (full): 445020
-Grid       : Message        : Average mflops/s per call per node (full): 299389
-Grid       : Message        :  Stencil 13.56 GB/s per node
-Grid       : Message        :  Stencil 16.6586 GB/s per node
-Grid       : Message        :  Stencil 17.5614 GB/s per node
-Grid       : Message        :  Stencil 13.1493 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664674
-Grid       : Message        : Average mflops/s per call per node       : 802079
-Grid       : Message        : Average mflops/s per call per node       : 825166
-Grid       : Message        : Average mflops/s per call per node       : 665438
-Grid       : Message        : Average mflops/s per call per node (full): 314723
-Grid       : Message        : Average mflops/s per call per node (full): 435905
-Grid       : Message        : Average mflops/s per call per node (full): 445889
-Grid       : Message        : Average mflops/s per call per node (full): 306380
-Grid       : Message        :  Stencil 12.9297 GB/s per node
-Grid       : Message        :  Stencil 16.6985 GB/s per node
-Grid       : Message        :  Stencil 18.1772 GB/s per node
-Grid       : Message        :  Stencil 13.673 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666231
-Grid       : Message        : Average mflops/s per call per node       : 800458
-Grid       : Message        : Average mflops/s per call per node       : 821243
-Grid       : Message        : Average mflops/s per call per node       : 660904
-Grid       : Message        : Average mflops/s per call per node (full): 313374
-Grid       : Message        : Average mflops/s per call per node (full): 432784
-Grid       : Message        : Average mflops/s per call per node (full): 448213
-Grid       : Message        : Average mflops/s per call per node (full): 304931
-Grid       : Message        :  Stencil 13.248 GB/s per node
-Grid       : Message        :  Stencil 16.5793 GB/s per node
-Grid       : Message        :  Stencil 17.6583 GB/s per node
-Grid       : Message        :  Stencil 13.0133 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667207
-Grid       : Message        : Average mflops/s per call per node       : 801544
-Grid       : Message        : Average mflops/s per call per node       : 820776
-Grid       : Message        : Average mflops/s per call per node       : 664668
-Grid       : Message        : Average mflops/s per call per node (full): 315702
-Grid       : Message        : Average mflops/s per call per node (full): 434485
-Grid       : Message        : Average mflops/s per call per node (full): 445645
-Grid       : Message        : Average mflops/s per call per node (full): 305435
-Grid       : Message        :  Stencil 13.0376 GB/s per node
-Grid       : Message        :  Stencil 16.7384 GB/s per node
-Grid       : Message        :  Stencil 17.5046 GB/s per node
-Grid       : Message        :  Stencil 12.7411 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665202
-Grid       : Message        : Average mflops/s per call per node       : 801357
-Grid       : Message        : Average mflops/s per call per node       : 822786
-Grid       : Message        : Average mflops/s per call per node       : 661364
-Grid       : Message        : Average mflops/s per call per node (full): 314427
-Grid       : Message        : Average mflops/s per call per node (full): 437039
-Grid       : Message        : Average mflops/s per call per node (full): 444813
-Grid       : Message        : Average mflops/s per call per node (full): 304747
-Grid       : Message        :  Stencil 13.1282 GB/s per node
-Grid       : Message        :  Stencil 17.8233 GB/s per node
-Grid       : Message        :  Stencil 17.638 GB/s per node
-Grid       : Message        :  Stencil 13.3392 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669809
-Grid       : Message        : Average mflops/s per call per node       : 803166
-Grid       : Message        : Average mflops/s per call per node       : 818551
-Grid       : Message        : Average mflops/s per call per node       : 663436
-Grid       : Message        : Average mflops/s per call per node (full): 315817
-Grid       : Message        : Average mflops/s per call per node (full): 439138
-Grid       : Message        : Average mflops/s per call per node (full): 443409
-Grid       : Message        : Average mflops/s per call per node (full): 305966
-Grid       : Message        :  Stencil 13.3358 GB/s per node
-Grid       : Message        :  Stencil 17.2628 GB/s per node
-Grid       : Message        :  Stencil 17.9616 GB/s per node
-Grid       : Message        :  Stencil 13.314 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668368
-Grid       : Message        : Average mflops/s per call per node       : 801644
-Grid       : Message        : Average mflops/s per call per node       : 823157
-Grid       : Message        : Average mflops/s per call per node       : 661210
-Grid       : Message        : Average mflops/s per call per node (full): 315243
-Grid       : Message        : Average mflops/s per call per node (full): 438471
-Grid       : Message        : Average mflops/s per call per node (full): 446383
-Grid       : Message        : Average mflops/s per call per node (full): 305657
-Grid       : Message        :  Stencil 14.2109 GB/s per node
-Grid       : Message        :  Stencil 18.6749 GB/s per node
-Grid       : Message        :  Stencil 17.7836 GB/s per node
-Grid       : Message        :  Stencil 12.6261 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666304
-Grid       : Message        : Average mflops/s per call per node       : 798222
-Grid       : Message        : Average mflops/s per call per node       : 817149
-Grid       : Message        : Average mflops/s per call per node       : 665280
-Grid       : Message        : Average mflops/s per call per node (full): 316455
-Grid       : Message        : Average mflops/s per call per node (full): 440022
-Grid       : Message        : Average mflops/s per call per node (full): 439907
-Grid       : Message        : Average mflops/s per call per node (full): 301760
-Grid       : Message        :  Stencil 15.4239 GB/s per node
-Grid       : Message        :  Stencil 9.59995 GB/s per node
-Grid       : Message        :  Stencil 17.6231 GB/s per node
-Grid       : Message        :  Stencil 12.2973 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662349
-Grid       : Message        : Average mflops/s per call per node       : 810307
-Grid       : Message        : Average mflops/s per call per node       : 818809
-Grid       : Message        : Average mflops/s per call per node       : 665904
-Grid       : Message        : Average mflops/s per call per node (full): 317328
-Grid       : Message        : Average mflops/s per call per node (full): 306879
-Grid       : Message        : Average mflops/s per call per node (full): 442167
-Grid       : Message        : Average mflops/s per call per node (full): 303953
-Grid       : Message        :  Stencil 14.5719 GB/s per node
-Grid       : Message        :  Stencil 17.0462 GB/s per node
-Grid       : Message        :  Stencil 17.0002 GB/s per node
-Grid       : Message        :  Stencil 13.1762 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667267
-Grid       : Message        : Average mflops/s per call per node       : 803553
-Grid       : Message        : Average mflops/s per call per node       : 824600
-Grid       : Message        : Average mflops/s per call per node       : 666051
-Grid       : Message        : Average mflops/s per call per node (full): 317919
-Grid       : Message        : Average mflops/s per call per node (full): 438698
-Grid       : Message        : Average mflops/s per call per node (full): 441306
-Grid       : Message        : Average mflops/s per call per node (full): 306707
-Grid       : Message        :  Stencil 13.3023 GB/s per node
-Grid       : Message        :  Stencil 16.7035 GB/s per node
-Grid       : Message        :  Stencil 17.7461 GB/s per node
-Grid       : Message        :  Stencil 13.8032 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669854
-Grid       : Message        : Average mflops/s per call per node       : 802712
-Grid       : Message        : Average mflops/s per call per node       : 820285
-Grid       : Message        : Average mflops/s per call per node       : 661765
-Grid       : Message        : Average mflops/s per call per node (full): 316185
-Grid       : Message        : Average mflops/s per call per node (full): 436687
-Grid       : Message        : Average mflops/s per call per node (full): 445842
-Grid       : Message        : Average mflops/s per call per node (full): 306302
-Grid       : Message        :  Stencil 13.788 GB/s per node
-Grid       : Message        :  Stencil 16.0696 GB/s per node
-Grid       : Message        :  Stencil 17.3617 GB/s per node
-Grid       : Message        :  Stencil 13.2007 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666466
-Grid       : Message        : Average mflops/s per call per node       : 805791
-Grid       : Message        : Average mflops/s per call per node       : 825870
-Grid       : Message        : Average mflops/s per call per node       : 663506
-Grid       : Message        : Average mflops/s per call per node (full): 315594
-Grid       : Message        : Average mflops/s per call per node (full): 428294
-Grid       : Message        : Average mflops/s per call per node (full): 444063
-Grid       : Message        : Average mflops/s per call per node (full): 305500
-Grid       : Message        :  Stencil 13.5511 GB/s per node
-Grid       : Message        :  Stencil 8.71329 GB/s per node
-Grid       : Message        :  Stencil 17.9582 GB/s per node
-Grid       : Message        :  Stencil 14.0022 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670027
-Grid       : Message        : Average mflops/s per call per node       : 809139
-Grid       : Message        : Average mflops/s per call per node       : 823711
-Grid       : Message        : Average mflops/s per call per node       : 660064
-Grid       : Message        : Average mflops/s per call per node (full): 316759
-Grid       : Message        : Average mflops/s per call per node (full): 283689
-Grid       : Message        : Average mflops/s per call per node (full): 446189
-Grid       : Message        : Average mflops/s per call per node (full): 306759
-Grid       : Message        :  Stencil 12.7662 GB/s per node
-Grid       : Message        :  Stencil 16.7982 GB/s per node
-Grid       : Message        :  Stencil 19.0757 GB/s per node
-Grid       : Message        :  Stencil 13.4446 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670826
-Grid       : Message        : Average mflops/s per call per node       : 808418
-Grid       : Message        : Average mflops/s per call per node       : 822184
-Grid       : Message        : Average mflops/s per call per node       : 664689
-Grid       : Message        : Average mflops/s per call per node (full): 316037
-Grid       : Message        : Average mflops/s per call per node (full): 437563
-Grid       : Message        : Average mflops/s per call per node (full): 450183
-Grid       : Message        : Average mflops/s per call per node (full): 307322
-Grid       : Message        :  Stencil 14.3459 GB/s per node
-Grid       : Message        :  Stencil 16.6088 GB/s per node
-Grid       : Message        :  Stencil 17.5173 GB/s per node
-Grid       : Message        :  Stencil 12.2539 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659901
-Grid       : Message        : Average mflops/s per call per node       : 806660
-Grid       : Message        : Average mflops/s per call per node       : 824506
-Grid       : Message        : Average mflops/s per call per node       : 666331
-Grid       : Message        : Average mflops/s per call per node (full): 315648
-Grid       : Message        : Average mflops/s per call per node (full): 436396
-Grid       : Message        : Average mflops/s per call per node (full): 444001
-Grid       : Message        : Average mflops/s per call per node (full): 303238
-Grid       : Message        :  Stencil 13.6295 GB/s per node
-Grid       : Message        :  Stencil 15.0441 GB/s per node
-Grid       : Message        :  Stencil 16.959 GB/s per node
-Grid       : Message        :  Stencil 14.2892 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663747
-Grid       : Message        : Average mflops/s per call per node       : 803580
-Grid       : Message        : Average mflops/s per call per node       : 821219
-Grid       : Message        : Average mflops/s per call per node       : 653734
-Grid       : Message        : Average mflops/s per call per node (full): 314459
-Grid       : Message        : Average mflops/s per call per node (full): 416182
-Grid       : Message        : Average mflops/s per call per node (full): 437365
-Grid       : Message        : Average mflops/s per call per node (full): 305736
-Grid       : Message        :  Stencil 12.9195 GB/s per node
-Grid       : Message        :  Stencil 11.4968 GB/s per node
-Grid       : Message        :  Stencil 17.197 GB/s per node
-Grid       : Message        :  Stencil 13.6375 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665932
-Grid       : Message        : Average mflops/s per call per node       : 808899
-Grid       : Message        : Average mflops/s per call per node       : 827618
-Grid       : Message        : Average mflops/s per call per node       : 663740
-Grid       : Message        : Average mflops/s per call per node (full): 314343
-Grid       : Message        : Average mflops/s per call per node (full): 350331
-Grid       : Message        : Average mflops/s per call per node (full): 437483
-Grid       : Message        : Average mflops/s per call per node (full): 303981
-Grid       : Message        :  Stencil 13.2593 GB/s per node
-Grid       : Message        :  Stencil 16.3093 GB/s per node
-Grid       : Message        :  Stencil 17.8383 GB/s per node
-Grid       : Message        :  Stencil 12.615 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665928
-Grid       : Message        : Average mflops/s per call per node       : 804355
-Grid       : Message        : Average mflops/s per call per node       : 825301
-Grid       : Message        : Average mflops/s per call per node       : 667432
-Grid       : Message        : Average mflops/s per call per node (full): 315382
-Grid       : Message        : Average mflops/s per call per node (full): 432655
-Grid       : Message        : Average mflops/s per call per node (full): 446212
-Grid       : Message        : Average mflops/s per call per node (full): 305544
-Grid       : Message        :  Stencil 12.7735 GB/s per node
-Grid       : Message        :  Stencil 16.5774 GB/s per node
-Grid       : Message        :  Stencil 17.6756 GB/s per node
-Grid       : Message        :  Stencil 13.2914 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664773
-Grid       : Message        : Average mflops/s per call per node       : 810050
-Grid       : Message        : Average mflops/s per call per node       : 816671
-Grid       : Message        : Average mflops/s per call per node       : 662241
-Grid       : Message        : Average mflops/s per call per node (full): 314265
-Grid       : Message        : Average mflops/s per call per node (full): 436636
-Grid       : Message        : Average mflops/s per call per node (full): 445247
-Grid       : Message        : Average mflops/s per call per node (full): 305656
-Grid       : Message        :  Stencil 12.8344 GB/s per node
-Grid       : Message        :  Stencil 17.1858 GB/s per node
-Grid       : Message        :  Stencil 17.413 GB/s per node
-Grid       : Message        :  Stencil 12.3467 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666922
-Grid       : Message        : Average mflops/s per call per node       : 802121
-Grid       : Message        : Average mflops/s per call per node       : 826953
-Grid       : Message        : Average mflops/s per call per node       : 665500
-Grid       : Message        : Average mflops/s per call per node (full): 313807
-Grid       : Message        : Average mflops/s per call per node (full): 436324
-Grid       : Message        : Average mflops/s per call per node (full): 445232
-Grid       : Message        : Average mflops/s per call per node (full): 302145
-Grid       : Message        :  Stencil 15.1025 GB/s per node
-Grid       : Message        :  Stencil 9.66352 GB/s per node
-Grid       : Message        :  Stencil 17.8292 GB/s per node
-Grid       : Message        :  Stencil 14.5742 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662633
-Grid       : Message        : Average mflops/s per call per node       : 812434
-Grid       : Message        : Average mflops/s per call per node       : 826027
-Grid       : Message        : Average mflops/s per call per node       : 655861
-Grid       : Message        : Average mflops/s per call per node (full): 317147
-Grid       : Message        : Average mflops/s per call per node (full): 308756
-Grid       : Message        : Average mflops/s per call per node (full): 447746
-Grid       : Message        : Average mflops/s per call per node (full): 306200
-Grid       : Message        :  Stencil 13.9373 GB/s per node
-Grid       : Message        :  Stencil 15.1548 GB/s per node
-Grid       : Message        :  Stencil 17.3736 GB/s per node
-Grid       : Message        :  Stencil 13.8492 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664275
-Grid       : Message        : Average mflops/s per call per node       : 805750
-Grid       : Message        : Average mflops/s per call per node       : 823774
-Grid       : Message        : Average mflops/s per call per node       : 661882
-Grid       : Message        : Average mflops/s per call per node (full): 315808
-Grid       : Message        : Average mflops/s per call per node (full): 418754
-Grid       : Message        : Average mflops/s per call per node (full): 443329
-Grid       : Message        : Average mflops/s per call per node (full): 305700
-Grid       : Message        :  Stencil 13.141 GB/s per node
-Grid       : Message        :  Stencil 16.7009 GB/s per node
-Grid       : Message        :  Stencil 17.8167 GB/s per node
-Grid       : Message        :  Stencil 13.6201 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667151
-Grid       : Message        : Average mflops/s per call per node       : 801398
-Grid       : Message        : Average mflops/s per call per node       : 824131
-Grid       : Message        : Average mflops/s per call per node       : 661978
-Grid       : Message        : Average mflops/s per call per node (full): 314659
-Grid       : Message        : Average mflops/s per call per node (full): 435937
-Grid       : Message        : Average mflops/s per call per node (full): 445330
-Grid       : Message        : Average mflops/s per call per node (full): 306688
-Grid       : Message        :  Stencil 13.9323 GB/s per node
-Grid       : Message        :  Stencil 16.4841 GB/s per node
-Grid       : Message        :  Stencil 16.7951 GB/s per node
-Grid       : Message        :  Stencil 13.1655 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664108
-Grid       : Message        : Average mflops/s per call per node       : 801113
-Grid       : Message        : Average mflops/s per call per node       : 826510
-Grid       : Message        : Average mflops/s per call per node       : 661766
-Grid       : Message        : Average mflops/s per call per node (full): 316038
-Grid       : Message        : Average mflops/s per call per node (full): 434101
-Grid       : Message        : Average mflops/s per call per node (full): 438453
-Grid       : Message        : Average mflops/s per call per node (full): 305318
-Grid       : Message        :  Stencil 14.1698 GB/s per node
-Grid       : Message        :  Stencil 16.633 GB/s per node
-Grid       : Message        :  Stencil 17.4947 GB/s per node
-Grid       : Message        :  Stencil 13.9853 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662381
-Grid       : Message        : Average mflops/s per call per node       : 804158
-Grid       : Message        : Average mflops/s per call per node       : 825228
-Grid       : Message        : Average mflops/s per call per node       : 659796
-Grid       : Message        : Average mflops/s per call per node (full): 315789
-Grid       : Message        : Average mflops/s per call per node (full): 431645
-Grid       : Message        : Average mflops/s per call per node (full): 444926
-Grid       : Message        : Average mflops/s per call per node (full): 306277
-Grid       : Message        :  Stencil 14.2747 GB/s per node
-Grid       : Message        :  Stencil 16.6401 GB/s per node
-Grid       : Message        :  Stencil 18.2928 GB/s per node
-Grid       : Message        :  Stencil 13.065 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665096
-Grid       : Message        : Average mflops/s per call per node       : 804366
-Grid       : Message        : Average mflops/s per call per node       : 820500
-Grid       : Message        : Average mflops/s per call per node       : 666332
-Grid       : Message        : Average mflops/s per call per node (full): 317067
-Grid       : Message        : Average mflops/s per call per node (full): 436069
-Grid       : Message        : Average mflops/s per call per node (full): 446419
-Grid       : Message        : Average mflops/s per call per node (full): 306089
-Grid       : Message        :  Stencil 14.6541 GB/s per node
-Grid       : Message        :  Stencil 16.9482 GB/s per node
-Grid       : Message        :  Stencil 17.7746 GB/s per node
-Grid       : Message        :  Stencil 12.6571 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664825
-Grid       : Message        : Average mflops/s per call per node       : 805429
-Grid       : Message        : Average mflops/s per call per node       : 816785
-Grid       : Message        : Average mflops/s per call per node       : 664280
-Grid       : Message        : Average mflops/s per call per node (full): 316962
-Grid       : Message        : Average mflops/s per call per node (full): 437622
-Grid       : Message        : Average mflops/s per call per node (full): 445335
-Grid       : Message        : Average mflops/s per call per node (full): 303974
-Grid       : Message        :  Stencil 14.2193 GB/s per node
-Grid       : Message        :  Stencil 16.7243 GB/s per node
-Grid       : Message        :  Stencil 17.3418 GB/s per node
-Grid       : Message        :  Stencil 12.8809 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666393
-Grid       : Message        : Average mflops/s per call per node       : 810603
-Grid       : Message        : Average mflops/s per call per node       : 818960
-Grid       : Message        : Average mflops/s per call per node       : 665596
-Grid       : Message        : Average mflops/s per call per node (full): 316810
-Grid       : Message        : Average mflops/s per call per node (full): 437826
-Grid       : Message        : Average mflops/s per call per node (full): 443884
-Grid       : Message        : Average mflops/s per call per node (full): 305181
-Grid       : Message        :  Stencil 13.7367 GB/s per node
-Grid       : Message        :  Stencil 13.7684 GB/s per node
-Grid       : Message        :  Stencil 17.1511 GB/s per node
-Grid       : Message        :  Stencil 12.483 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665979
-Grid       : Message        : Average mflops/s per call per node       : 805947
-Grid       : Message        : Average mflops/s per call per node       : 821545
-Grid       : Message        : Average mflops/s per call per node       : 663268
-Grid       : Message        : Average mflops/s per call per node (full): 316166
-Grid       : Message        : Average mflops/s per call per node (full): 395981
-Grid       : Message        : Average mflops/s per call per node (full): 442434
-Grid       : Message        : Average mflops/s per call per node (full): 303984
-Grid       : Message        :  Stencil 13.9408 GB/s per node
-Grid       : Message        :  Stencil 17.3043 GB/s per node
-Grid       : Message        :  Stencil 17.7234 GB/s per node
-Grid       : Message        :  Stencil 14.3152 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667601
-Grid       : Message        : Average mflops/s per call per node       : 801981
-Grid       : Message        : Average mflops/s per call per node       : 822114
-Grid       : Message        : Average mflops/s per call per node       : 662198
-Grid       : Message        : Average mflops/s per call per node (full): 316300
-Grid       : Message        : Average mflops/s per call per node (full): 438069
-Grid       : Message        : Average mflops/s per call per node (full): 445210
-Grid       : Message        : Average mflops/s per call per node (full): 306345
-Grid       : Message        :  Stencil 12.5711 GB/s per node
-Grid       : Message        :  Stencil 15.0578 GB/s per node
-Grid       : Message        :  Stencil 17.8285 GB/s per node
-Grid       : Message        :  Stencil 12.6284 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671125
-Grid       : Message        : Average mflops/s per call per node       : 806146
-Grid       : Message        : Average mflops/s per call per node       : 821990
-Grid       : Message        : Average mflops/s per call per node       : 661048
-Grid       : Message        : Average mflops/s per call per node (full): 312972
-Grid       : Message        : Average mflops/s per call per node (full): 416531
-Grid       : Message        : Average mflops/s per call per node (full): 446886
-Grid       : Message        : Average mflops/s per call per node (full): 302882
-Grid       : Message        :  Stencil 12.5281 GB/s per node
-Grid       : Message        :  Stencil 15.8613 GB/s per node
-Grid       : Message        :  Stencil 17.6163 GB/s per node
-Grid       : Message        :  Stencil 13.0894 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670975
-Grid       : Message        : Average mflops/s per call per node       : 801534
-Grid       : Message        : Average mflops/s per call per node       : 823667
-Grid       : Message        : Average mflops/s per call per node       : 663701
-Grid       : Message        : Average mflops/s per call per node (full): 314256
-Grid       : Message        : Average mflops/s per call per node (full): 425895
-Grid       : Message        : Average mflops/s per call per node (full): 444898
-Grid       : Message        : Average mflops/s per call per node (full): 305136
-Grid       : Message        :  Stencil 13.8193 GB/s per node
-Grid       : Message        :  Stencil 16.2551 GB/s per node
-Grid       : Message        :  Stencil 17.1763 GB/s per node
-Grid       : Message        :  Stencil 13.4658 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666654
-Grid       : Message        : Average mflops/s per call per node       : 798824
-Grid       : Message        : Average mflops/s per call per node       : 822433
-Grid       : Message        : Average mflops/s per call per node       : 662772
-Grid       : Message        : Average mflops/s per call per node (full): 316261
-Grid       : Message        : Average mflops/s per call per node (full): 431431
-Grid       : Message        : Average mflops/s per call per node (full): 442904
-Grid       : Message        : Average mflops/s per call per node (full): 306560
-Grid       : Message        :  Stencil 14.0584 GB/s per node
-Grid       : Message        :  Stencil 16.9526 GB/s per node
-Grid       : Message        :  Stencil 16.855 GB/s per node
-Grid       : Message        :  Stencil 12.106 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667419
-Grid       : Message        : Average mflops/s per call per node       : 801288
-Grid       : Message        : Average mflops/s per call per node       : 828394
-Grid       : Message        : Average mflops/s per call per node       : 664532
-Grid       : Message        : Average mflops/s per call per node (full): 315332
-Grid       : Message        : Average mflops/s per call per node (full): 430937
-Grid       : Message        : Average mflops/s per call per node (full): 439427
-Grid       : Message        : Average mflops/s per call per node (full): 302734
-Grid       : Message        :  Stencil 12.7819 GB/s per node
-Grid       : Message        :  Stencil 11.0133 GB/s per node
-Grid       : Message        :  Stencil 17.3597 GB/s per node
-Grid       : Message        :  Stencil 12.6736 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671730
-Grid       : Message        : Average mflops/s per call per node       : 814118
-Grid       : Message        : Average mflops/s per call per node       : 824538
-Grid       : Message        : Average mflops/s per call per node       : 668701
-Grid       : Message        : Average mflops/s per call per node (full): 315767
-Grid       : Message        : Average mflops/s per call per node (full): 340190
-Grid       : Message        : Average mflops/s per call per node (full): 444378
-Grid       : Message        : Average mflops/s per call per node (full): 305611
-Grid       : Message        :  Stencil 13.3355 GB/s per node
-Grid       : Message        :  Stencil 17.0643 GB/s per node
-Grid       : Message        :  Stencil 17.77 GB/s per node
-Grid       : Message        :  Stencil 12.2757 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668068
-Grid       : Message        : Average mflops/s per call per node       : 804577
-Grid       : Message        : Average mflops/s per call per node       : 824356
-Grid       : Message        : Average mflops/s per call per node       : 668566
-Grid       : Message        : Average mflops/s per call per node (full): 315740
-Grid       : Message        : Average mflops/s per call per node (full): 439670
-Grid       : Message        : Average mflops/s per call per node (full): 446852
-Grid       : Message        : Average mflops/s per call per node (full): 304567
-Grid       : Message        :  Stencil 13.6138 GB/s per node
-Grid       : Message        :  Stencil 14.3006 GB/s per node
-Grid       : Message        :  Stencil 18.8566 GB/s per node
-Grid       : Message        :  Stencil 12.6303 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665029
-Grid       : Message        : Average mflops/s per call per node       : 806340
-Grid       : Message        : Average mflops/s per call per node       : 824355
-Grid       : Message        : Average mflops/s per call per node       : 668042
-Grid       : Message        : Average mflops/s per call per node (full): 316228
-Grid       : Message        : Average mflops/s per call per node (full): 404877
-Grid       : Message        : Average mflops/s per call per node (full): 450779
-Grid       : Message        : Average mflops/s per call per node (full): 305727
-Grid       : Message        :  Stencil 12.594 GB/s per node
-Grid       : Message        :  Stencil 16.5838 GB/s per node
-Grid       : Message        :  Stencil 17.822 GB/s per node
-Grid       : Message        :  Stencil 12.2811 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665494
-Grid       : Message        : Average mflops/s per call per node       : 803374
-Grid       : Message        : Average mflops/s per call per node       : 825090
-Grid       : Message        : Average mflops/s per call per node       : 667290
-Grid       : Message        : Average mflops/s per call per node (full): 311572
-Grid       : Message        : Average mflops/s per call per node (full): 435649
-Grid       : Message        : Average mflops/s per call per node (full): 444993
-Grid       : Message        : Average mflops/s per call per node (full): 304229
-Grid       : Message        :  Stencil 12.9542 GB/s per node
-Grid       : Message        :  Stencil 14.1127 GB/s per node
-Grid       : Message        :  Stencil 18.8473 GB/s per node
-Grid       : Message        :  Stencil 14.0909 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663959
-Grid       : Message        : Average mflops/s per call per node       : 807391
-Grid       : Message        : Average mflops/s per call per node       : 822112
-Grid       : Message        : Average mflops/s per call per node       : 662313
-Grid       : Message        : Average mflops/s per call per node (full): 313998
-Grid       : Message        : Average mflops/s per call per node (full): 400634
-Grid       : Message        : Average mflops/s per call per node (full): 450565
-Grid       : Message        : Average mflops/s per call per node (full): 305393
-Grid       : Message        :  Stencil 13.4408 GB/s per node
-Grid       : Message        :  Stencil 17.5697 GB/s per node
-Grid       : Message        :  Stencil 17.3413 GB/s per node
-Grid       : Message        :  Stencil 13.1107 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665884
-Grid       : Message        : Average mflops/s per call per node       : 805244
-Grid       : Message        : Average mflops/s per call per node       : 822770
-Grid       : Message        : Average mflops/s per call per node       : 663204
-Grid       : Message        : Average mflops/s per call per node (full): 316076
-Grid       : Message        : Average mflops/s per call per node (full): 440380
-Grid       : Message        : Average mflops/s per call per node (full): 444103
-Grid       : Message        : Average mflops/s per call per node (full): 304950
-Grid       : Message        :  Stencil 13.5491 GB/s per node
-Grid       : Message        :  Stencil 16.4245 GB/s per node
-Grid       : Message        :  Stencil 17.5451 GB/s per node
-Grid       : Message        :  Stencil 12.533 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666679
-Grid       : Message        : Average mflops/s per call per node       : 801902
-Grid       : Message        : Average mflops/s per call per node       : 822733
-Grid       : Message        : Average mflops/s per call per node       : 667394
-Grid       : Message        : Average mflops/s per call per node (full): 317129
-Grid       : Message        : Average mflops/s per call per node (full): 434095
-Grid       : Message        : Average mflops/s per call per node (full): 443780
-Grid       : Message        : Average mflops/s per call per node (full): 305218
-Grid       : Message        :  Stencil 13.6534 GB/s per node
-Grid       : Message        :  Stencil 13.8195 GB/s per node
-Grid       : Message        :  Stencil 17.2644 GB/s per node
-Grid       : Message        :  Stencil 12.3142 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667011
-Grid       : Message        : Average mflops/s per call per node       : 810067
-Grid       : Message        : Average mflops/s per call per node       : 821524
-Grid       : Message        : Average mflops/s per call per node       : 671276
-Grid       : Message        : Average mflops/s per call per node (full): 317031
-Grid       : Message        : Average mflops/s per call per node (full): 396478
-Grid       : Message        : Average mflops/s per call per node (full): 443169
-Grid       : Message        : Average mflops/s per call per node (full): 303807
-Grid       : Message        :  Stencil 13.3901 GB/s per node
-Grid       : Message        :  Stencil 18.0472 GB/s per node
-Grid       : Message        :  Stencil 17.2049 GB/s per node
-Grid       : Message        :  Stencil 13.2758 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665144
-Grid       : Message        : Average mflops/s per call per node       : 798251
-Grid       : Message        : Average mflops/s per call per node       : 820132
-Grid       : Message        : Average mflops/s per call per node       : 667648
-Grid       : Message        : Average mflops/s per call per node (full): 316090
-Grid       : Message        : Average mflops/s per call per node (full): 437012
-Grid       : Message        : Average mflops/s per call per node (full): 438289
-Grid       : Message        : Average mflops/s per call per node (full): 306613
-Grid       : Message        :  Stencil 13.1945 GB/s per node
-Grid       : Message        :  Stencil 17.1718 GB/s per node
-Grid       : Message        :  Stencil 17.3392 GB/s per node
-Grid       : Message        :  Stencil 11.9572 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667084
-Grid       : Message        : Average mflops/s per call per node       : 796989
-Grid       : Message        : Average mflops/s per call per node       : 824578
-Grid       : Message        : Average mflops/s per call per node       : 666011
-Grid       : Message        : Average mflops/s per call per node (full): 316372
-Grid       : Message        : Average mflops/s per call per node (full): 435100
-Grid       : Message        : Average mflops/s per call per node (full): 443137
-Grid       : Message        : Average mflops/s per call per node (full): 301185
-Grid       : Message        :  Stencil 12.8059 GB/s per node
-Grid       : Message        :  Stencil 16.412 GB/s per node
-Grid       : Message        :  Stencil 17.7064 GB/s per node
-Grid       : Message        :  Stencil 12.3333 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669057
-Grid       : Message        : Average mflops/s per call per node       : 803586
-Grid       : Message        : Average mflops/s per call per node       : 823735
-Grid       : Message        : Average mflops/s per call per node       : 668583
-Grid       : Message        : Average mflops/s per call per node (full): 315115
-Grid       : Message        : Average mflops/s per call per node (full): 434082
-Grid       : Message        : Average mflops/s per call per node (full): 445782
-Grid       : Message        : Average mflops/s per call per node (full): 303832
-Grid       : Message        :  Stencil 12.9454 GB/s per node
-Grid       : Message        :  Stencil 9.35683 GB/s per node
-Grid       : Message        :  Stencil 17.5506 GB/s per node
-Grid       : Message        :  Stencil 12.7689 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663821
-Grid       : Message        : Average mflops/s per call per node       : 811170
-Grid       : Message        : Average mflops/s per call per node       : 818730
-Grid       : Message        : Average mflops/s per call per node       : 662393
-Grid       : Message        : Average mflops/s per call per node (full): 313741
-Grid       : Message        : Average mflops/s per call per node (full): 301128
-Grid       : Message        : Average mflops/s per call per node (full): 444151
-Grid       : Message        : Average mflops/s per call per node (full): 305199
-Grid       : Message        :  Stencil 13.8853 GB/s per node
-Grid       : Message        :  Stencil 17.2209 GB/s per node
-Grid       : Message        :  Stencil 17.2958 GB/s per node
-Grid       : Message        :  Stencil 13.6379 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667527
-Grid       : Message        : Average mflops/s per call per node       : 799978
-Grid       : Message        : Average mflops/s per call per node       : 827915
-Grid       : Message        : Average mflops/s per call per node       : 666053
-Grid       : Message        : Average mflops/s per call per node (full): 313304
-Grid       : Message        : Average mflops/s per call per node (full): 437343
-Grid       : Message        : Average mflops/s per call per node (full): 443923
-Grid       : Message        : Average mflops/s per call per node (full): 307158
-Grid       : Message        :  Stencil 13.6277 GB/s per node
-Grid       : Message        :  Stencil 16.4862 GB/s per node
-Grid       : Message        :  Stencil 18.3482 GB/s per node
-Grid       : Message        :  Stencil 12.4 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664650
-Grid       : Message        : Average mflops/s per call per node       : 803574
-Grid       : Message        : Average mflops/s per call per node       : 818372
-Grid       : Message        : Average mflops/s per call per node       : 661592
-Grid       : Message        : Average mflops/s per call per node (full): 316026
-Grid       : Message        : Average mflops/s per call per node (full): 434754
-Grid       : Message        : Average mflops/s per call per node (full): 445993
-Grid       : Message        : Average mflops/s per call per node (full): 302797
-Grid       : Message        :  Stencil 12.6584 GB/s per node
-Grid       : Message        :  Stencil 14.6628 GB/s per node
-Grid       : Message        :  Stencil 17.4541 GB/s per node
-Grid       : Message        :  Stencil 12.0402 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666614
-Grid       : Message        : Average mflops/s per call per node       : 808074
-Grid       : Message        : Average mflops/s per call per node       : 820717
-Grid       : Message        : Average mflops/s per call per node       : 670324
-Grid       : Message        : Average mflops/s per call per node (full): 312049
-Grid       : Message        : Average mflops/s per call per node (full): 410942
-Grid       : Message        : Average mflops/s per call per node (full): 445355
-Grid       : Message        : Average mflops/s per call per node (full): 303320
-Grid       : Message        :  Stencil 12.9024 GB/s per node
-Grid       : Message        :  Stencil 16.444 GB/s per node
-Grid       : Message        :  Stencil 17.3399 GB/s per node
-Grid       : Message        :  Stencil 14.0399 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665405
-Grid       : Message        : Average mflops/s per call per node       : 808103
-Grid       : Message        : Average mflops/s per call per node       : 821609
-Grid       : Message        : Average mflops/s per call per node       : 665294
-Grid       : Message        : Average mflops/s per call per node (full): 313606
-Grid       : Message        : Average mflops/s per call per node (full): 434873
-Grid       : Message        : Average mflops/s per call per node (full): 444264
-Grid       : Message        : Average mflops/s per call per node (full): 306771
-Grid       : Message        :  Stencil 13.9308 GB/s per node
-Grid       : Message        :  Stencil 17.4158 GB/s per node
-Grid       : Message        :  Stencil 18.4423 GB/s per node
-Grid       : Message        :  Stencil 12.855 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665272
-Grid       : Message        : Average mflops/s per call per node       : 797460
-Grid       : Message        : Average mflops/s per call per node       : 823126
-Grid       : Message        : Average mflops/s per call per node       : 662335
-Grid       : Message        : Average mflops/s per call per node (full): 314975
-Grid       : Message        : Average mflops/s per call per node (full): 435467
-Grid       : Message        : Average mflops/s per call per node (full): 449253
-Grid       : Message        : Average mflops/s per call per node (full): 305424
-Grid       : Message        :  Stencil 12.9925 GB/s per node
-Grid       : Message        :  Stencil 16.3539 GB/s per node
-Grid       : Message        :  Stencil 17.2369 GB/s per node
-Grid       : Message        :  Stencil 13.4937 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669940
-Grid       : Message        : Average mflops/s per call per node       : 803238
-Grid       : Message        : Average mflops/s per call per node       : 825965
-Grid       : Message        : Average mflops/s per call per node       : 663766
-Grid       : Message        : Average mflops/s per call per node (full): 314761
-Grid       : Message        : Average mflops/s per call per node (full): 434155
-Grid       : Message        : Average mflops/s per call per node (full): 444234
-Grid       : Message        : Average mflops/s per call per node (full): 299831
-Grid       : Message        :  Stencil 14.106 GB/s per node
-Grid       : Message        :  Stencil 16.6361 GB/s per node
-Grid       : Message        :  Stencil 17.0601 GB/s per node
-Grid       : Message        :  Stencil 12.1837 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660382
-Grid       : Message        : Average mflops/s per call per node       : 804401
-Grid       : Message        : Average mflops/s per call per node       : 823592
-Grid       : Message        : Average mflops/s per call per node       : 670380
-Grid       : Message        : Average mflops/s per call per node (full): 315078
-Grid       : Message        : Average mflops/s per call per node (full): 432738
-Grid       : Message        : Average mflops/s per call per node (full): 441811
-Grid       : Message        : Average mflops/s per call per node (full): 303707
-Grid       : Message        :  Stencil 12.9924 GB/s per node
-Grid       : Message        :  Stencil 17.4822 GB/s per node
-Grid       : Message        :  Stencil 16.9255 GB/s per node
-Grid       : Message        :  Stencil 12.131 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666019
-Grid       : Message        : Average mflops/s per call per node       : 804316
-Grid       : Message        : Average mflops/s per call per node       : 828336
-Grid       : Message        : Average mflops/s per call per node       : 669150
-Grid       : Message        : Average mflops/s per call per node (full): 314690
-Grid       : Message        : Average mflops/s per call per node (full): 438882
-Grid       : Message        : Average mflops/s per call per node (full): 440236
-Grid       : Message        : Average mflops/s per call per node (full): 303263
-Grid       : Message        :  Stencil 12.9286 GB/s per node
-Grid       : Message        :  Stencil 16.5674 GB/s per node
-Grid       : Message        :  Stencil 17.4028 GB/s per node
-Grid       : Message        :  Stencil 12.2927 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663144
-Grid       : Message        : Average mflops/s per call per node       : 802347
-Grid       : Message        : Average mflops/s per call per node       : 822576
-Grid       : Message        : Average mflops/s per call per node       : 664923
-Grid       : Message        : Average mflops/s per call per node (full): 313191
-Grid       : Message        : Average mflops/s per call per node (full): 434915
-Grid       : Message        : Average mflops/s per call per node (full): 443819
-Grid       : Message        : Average mflops/s per call per node (full): 303684
-Grid       : Message        :  Stencil 12.9843 GB/s per node
-Grid       : Message        :  Stencil 16.5282 GB/s per node
-Grid       : Message        :  Stencil 16.3675 GB/s per node
-Grid       : Message        :  Stencil 12.5887 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663888
-Grid       : Message        : Average mflops/s per call per node       : 806737
-Grid       : Message        : Average mflops/s per call per node       : 820169
-Grid       : Message        : Average mflops/s per call per node       : 668303
-Grid       : Message        : Average mflops/s per call per node (full): 314496
-Grid       : Message        : Average mflops/s per call per node (full): 434932
-Grid       : Message        : Average mflops/s per call per node (full): 424715
-Grid       : Message        : Average mflops/s per call per node (full): 305269
-Grid       : Message        :  Stencil 14.4789 GB/s per node
-Grid       : Message        :  Stencil 9.60314 GB/s per node
-Grid       : Message        :  Stencil 17.4542 GB/s per node
-Grid       : Message        :  Stencil 12.7538 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661383
-Grid       : Message        : Average mflops/s per call per node       : 814741
-Grid       : Message        : Average mflops/s per call per node       : 824469
-Grid       : Message        : Average mflops/s per call per node       : 665327
-Grid       : Message        : Average mflops/s per call per node (full): 316536
-Grid       : Message        : Average mflops/s per call per node (full): 306930
-Grid       : Message        : Average mflops/s per call per node (full): 444745
-Grid       : Message        : Average mflops/s per call per node (full): 304862
-Grid       : Message        :  Stencil 14.1053 GB/s per node
-Grid       : Message        :  Stencil 17.8723 GB/s per node
-Grid       : Message        :  Stencil 17.2534 GB/s per node
-Grid       : Message        :  Stencil 12.2174 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661647
-Grid       : Message        : Average mflops/s per call per node       : 803667
-Grid       : Message        : Average mflops/s per call per node       : 821296
-Grid       : Message        : Average mflops/s per call per node       : 667759
-Grid       : Message        : Average mflops/s per call per node (full): 314634
-Grid       : Message        : Average mflops/s per call per node (full): 438959
-Grid       : Message        : Average mflops/s per call per node (full): 442505
-Grid       : Message        : Average mflops/s per call per node (full): 304194
-Grid       : Message        :  Stencil 13.0328 GB/s per node
-Grid       : Message        :  Stencil 17.0942 GB/s per node
-Grid       : Message        :  Stencil 18.1395 GB/s per node
-Grid       : Message        :  Stencil 12.7188 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666217
-Grid       : Message        : Average mflops/s per call per node       : 804184
-Grid       : Message        : Average mflops/s per call per node       : 821901
-Grid       : Message        : Average mflops/s per call per node       : 665614
-Grid       : Message        : Average mflops/s per call per node (full): 315688
-Grid       : Message        : Average mflops/s per call per node (full): 438573
-Grid       : Message        : Average mflops/s per call per node (full): 443813
-Grid       : Message        : Average mflops/s per call per node (full): 305679
-Grid       : Message        :  Stencil 12.8203 GB/s per node
-Grid       : Message        :  Stencil 16.6636 GB/s per node
-Grid       : Message        :  Stencil 17.7953 GB/s per node
-Grid       : Message        :  Stencil 14.0385 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671915
-Grid       : Message        : Average mflops/s per call per node       : 806958
-Grid       : Message        : Average mflops/s per call per node       : 827955
-Grid       : Message        : Average mflops/s per call per node       : 661803
-Grid       : Message        : Average mflops/s per call per node (full): 315523
-Grid       : Message        : Average mflops/s per call per node (full): 435924
-Grid       : Message        : Average mflops/s per call per node (full): 447476
-Grid       : Message        : Average mflops/s per call per node (full): 306720
-Grid       : Message        :  Stencil 13.0128 GB/s per node
-Grid       : Message        :  Stencil 16.5301 GB/s per node
-Grid       : Message        :  Stencil 17.1159 GB/s per node
-Grid       : Message        :  Stencil 12.2497 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668199
-Grid       : Message        : Average mflops/s per call per node       : 804835
-Grid       : Message        : Average mflops/s per call per node       : 824454
-Grid       : Message        : Average mflops/s per call per node       : 672602
-Grid       : Message        : Average mflops/s per call per node (full): 315815
-Grid       : Message        : Average mflops/s per call per node (full): 435739
-Grid       : Message        : Average mflops/s per call per node (full): 441843
-Grid       : Message        : Average mflops/s per call per node (full): 303662
-Grid       : Message        :  Stencil 12.9599 GB/s per node
-Grid       : Message        :  Stencil 16.5674 GB/s per node
-Grid       : Message        :  Stencil 18.6981 GB/s per node
-Grid       : Message        :  Stencil 12.4395 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668982
-Grid       : Message        : Average mflops/s per call per node       : 805101
-Grid       : Message        : Average mflops/s per call per node       : 819369
-Grid       : Message        : Average mflops/s per call per node       : 662323
-Grid       : Message        : Average mflops/s per call per node (full): 316156
-Grid       : Message        : Average mflops/s per call per node (full): 434381
-Grid       : Message        : Average mflops/s per call per node (full): 447002
-Grid       : Message        : Average mflops/s per call per node (full): 301518
-Grid       : Message        :  Stencil 13.6958 GB/s per node
-Grid       : Message        :  Stencil 17.0557 GB/s per node
-Grid       : Message        :  Stencil 17.5598 GB/s per node
-Grid       : Message        :  Stencil 12.6534 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667324
-Grid       : Message        : Average mflops/s per call per node       : 800973
-Grid       : Message        : Average mflops/s per call per node       : 824928
-Grid       : Message        : Average mflops/s per call per node       : 662691
-Grid       : Message        : Average mflops/s per call per node (full): 316434
-Grid       : Message        : Average mflops/s per call per node (full): 435862
-Grid       : Message        : Average mflops/s per call per node (full): 445696
-Grid       : Message        : Average mflops/s per call per node (full): 303382
-Grid       : Message        :  Stencil 13.1224 GB/s per node
-Grid       : Message        :  Stencil 17.4906 GB/s per node
-Grid       : Message        :  Stencil 17.5314 GB/s per node
-Grid       : Message        :  Stencil 12.9159 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669689
-Grid       : Message        : Average mflops/s per call per node       : 801373
-Grid       : Message        : Average mflops/s per call per node       : 823561
-Grid       : Message        : Average mflops/s per call per node       : 663627
-Grid       : Message        : Average mflops/s per call per node (full): 314853
-Grid       : Message        : Average mflops/s per call per node (full): 438798
-Grid       : Message        : Average mflops/s per call per node (full): 446137
-Grid       : Message        : Average mflops/s per call per node (full): 305165
-Grid       : Message        :  Stencil 13.3894 GB/s per node
-Grid       : Message        :  Stencil 17.0121 GB/s per node
-Grid       : Message        :  Stencil 18.0152 GB/s per node
-Grid       : Message        :  Stencil 13.0537 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668239
-Grid       : Message        : Average mflops/s per call per node       : 803511
-Grid       : Message        : Average mflops/s per call per node       : 820475
-Grid       : Message        : Average mflops/s per call per node       : 668274
-Grid       : Message        : Average mflops/s per call per node (full): 315111
-Grid       : Message        : Average mflops/s per call per node (full): 434183
-Grid       : Message        : Average mflops/s per call per node (full): 441978
-Grid       : Message        : Average mflops/s per call per node (full): 305313
-Grid       : Message        :  Stencil 13.515 GB/s per node
-Grid       : Message        :  Stencil 10.2688 GB/s per node
-Grid       : Message        :  Stencil 17.7445 GB/s per node
-Grid       : Message        :  Stencil 13.5992 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667074
-Grid       : Message        : Average mflops/s per call per node       : 810297
-Grid       : Message        : Average mflops/s per call per node       : 827854
-Grid       : Message        : Average mflops/s per call per node       : 666083
-Grid       : Message        : Average mflops/s per call per node (full): 314808
-Grid       : Message        : Average mflops/s per call per node (full): 323184
-Grid       : Message        : Average mflops/s per call per node (full): 447110
-Grid       : Message        : Average mflops/s per call per node (full): 307365
-Grid       : Message        :  Stencil 12.6776 GB/s per node
-Grid       : Message        :  Stencil 17.9313 GB/s per node
-Grid       : Message        :  Stencil 18.4433 GB/s per node
-Grid       : Message        :  Stencil 11.8563 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668535
-Grid       : Message        : Average mflops/s per call per node       : 806107
-Grid       : Message        : Average mflops/s per call per node       : 824265
-Grid       : Message        : Average mflops/s per call per node       : 668328
-Grid       : Message        : Average mflops/s per call per node (full): 314950
-Grid       : Message        : Average mflops/s per call per node (full): 440526
-Grid       : Message        : Average mflops/s per call per node (full): 449722
-Grid       : Message        : Average mflops/s per call per node (full): 300696
-Grid       : Message        :  Stencil 15.4819 GB/s per node
-Grid       : Message        :  Stencil 16.8911 GB/s per node
-Grid       : Message        :  Stencil 20.2369 GB/s per node
-Grid       : Message        :  Stencil 12.9396 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659746
-Grid       : Message        : Average mflops/s per call per node       : 805149
-Grid       : Message        : Average mflops/s per call per node       : 825933
-Grid       : Message        : Average mflops/s per call per node       : 668331
-Grid       : Message        : Average mflops/s per call per node (full): 316586
-Grid       : Message        : Average mflops/s per call per node (full): 437437
-Grid       : Message        : Average mflops/s per call per node (full): 452041
-Grid       : Message        : Average mflops/s per call per node (full): 306154
-Grid       : Message        :  Stencil 13.272 GB/s per node
-Grid       : Message        :  Stencil 17.3632 GB/s per node
-Grid       : Message        :  Stencil 19.4385 GB/s per node
-Grid       : Message        :  Stencil 12.092 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663519
-Grid       : Message        : Average mflops/s per call per node       : 805282
-Grid       : Message        : Average mflops/s per call per node       : 817363
-Grid       : Message        : Average mflops/s per call per node       : 671292
-Grid       : Message        : Average mflops/s per call per node (full): 314996
-Grid       : Message        : Average mflops/s per call per node (full): 437586
-Grid       : Message        : Average mflops/s per call per node (full): 450000
-Grid       : Message        : Average mflops/s per call per node (full): 302839
-Grid       : Message        :  Stencil 13.126 GB/s per node
-Grid       : Message        :  Stencil 17.6082 GB/s per node
-Grid       : Message        :  Stencil 17.4959 GB/s per node
-Grid       : Message        :  Stencil 12.5887 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666639
-Grid       : Message        : Average mflops/s per call per node       : 804402
-Grid       : Message        : Average mflops/s per call per node       : 824793
-Grid       : Message        : Average mflops/s per call per node       : 663605
-Grid       : Message        : Average mflops/s per call per node (full): 314758
-Grid       : Message        : Average mflops/s per call per node (full): 439559
-Grid       : Message        : Average mflops/s per call per node (full): 445120
-Grid       : Message        : Average mflops/s per call per node (full): 300752
-Grid       : Message        :  Stencil 14.3865 GB/s per node
-Grid       : Message        :  Stencil 13.7631 GB/s per node
-Grid       : Message        :  Stencil 17.914 GB/s per node
-Grid       : Message        :  Stencil 13.5841 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661551
-Grid       : Message        : Average mflops/s per call per node       : 813095
-Grid       : Message        : Average mflops/s per call per node       : 817403
-Grid       : Message        : Average mflops/s per call per node       : 664747
-Grid       : Message        : Average mflops/s per call per node (full): 315777
-Grid       : Message        : Average mflops/s per call per node (full): 395402
-Grid       : Message        : Average mflops/s per call per node (full): 446025
-Grid       : Message        : Average mflops/s per call per node (full): 307213
-Grid       : Message        :  Stencil 13.154 GB/s per node
-Grid       : Message        :  Stencil 16.589 GB/s per node
-Grid       : Message        :  Stencil 17.3772 GB/s per node
-Grid       : Message        :  Stencil 12.6145 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668811
-Grid       : Message        : Average mflops/s per call per node       : 808048
-Grid       : Message        : Average mflops/s per call per node       : 823148
-Grid       : Message        : Average mflops/s per call per node       : 663637
-Grid       : Message        : Average mflops/s per call per node (full): 316526
-Grid       : Message        : Average mflops/s per call per node (full): 436480
-Grid       : Message        : Average mflops/s per call per node (full): 445049
-Grid       : Message        : Average mflops/s per call per node (full): 304377
-Grid       : Message        :  Stencil 13.6708 GB/s per node
-Grid       : Message        :  Stencil 16.7315 GB/s per node
-Grid       : Message        :  Stencil 19.2646 GB/s per node
-Grid       : Message        :  Stencil 12.3974 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664753
-Grid       : Message        : Average mflops/s per call per node       : 804731
-Grid       : Message        : Average mflops/s per call per node       : 825383
-Grid       : Message        : Average mflops/s per call per node       : 670540
-Grid       : Message        : Average mflops/s per call per node (full): 314606
-Grid       : Message        : Average mflops/s per call per node (full): 436933
-Grid       : Message        : Average mflops/s per call per node (full): 450415
-Grid       : Message        : Average mflops/s per call per node (full): 305052
-Grid       : Message        :  Stencil 14.1461 GB/s per node
-Grid       : Message        :  Stencil 17.0066 GB/s per node
-Grid       : Message        :  Stencil 17.0841 GB/s per node
-Grid       : Message        :  Stencil 12.017 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664273
-Grid       : Message        : Average mflops/s per call per node       : 803468
-Grid       : Message        : Average mflops/s per call per node       : 825833
-Grid       : Message        : Average mflops/s per call per node       : 666602
-Grid       : Message        : Average mflops/s per call per node (full): 316771
-Grid       : Message        : Average mflops/s per call per node (full): 436762
-Grid       : Message        : Average mflops/s per call per node (full): 442465
-Grid       : Message        : Average mflops/s per call per node (full): 302696
-Grid       : Message        :  Stencil 13.5987 GB/s per node
-Grid       : Message        :  Stencil 16.8984 GB/s per node
-Grid       : Message        :  Stencil 17.6919 GB/s per node
-Grid       : Message        :  Stencil 12.3808 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667584
-Grid       : Message        : Average mflops/s per call per node       : 801722
-Grid       : Message        : Average mflops/s per call per node       : 823101
-Grid       : Message        : Average mflops/s per call per node       : 669154
-Grid       : Message        : Average mflops/s per call per node (full): 316780
-Grid       : Message        : Average mflops/s per call per node (full): 438047
-Grid       : Message        : Average mflops/s per call per node (full): 445333
-Grid       : Message        : Average mflops/s per call per node (full): 304774
-Grid       : Message        :  Stencil 13.6975 GB/s per node
-Grid       : Message        :  Stencil 18.6487 GB/s per node
-Grid       : Message        :  Stencil 17.5858 GB/s per node
-Grid       : Message        :  Stencil 14.6524 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667923
-Grid       : Message        : Average mflops/s per call per node       : 807162
-Grid       : Message        : Average mflops/s per call per node       : 825609
-Grid       : Message        : Average mflops/s per call per node       : 659829
-Grid       : Message        : Average mflops/s per call per node (full): 316104
-Grid       : Message        : Average mflops/s per call per node (full): 441285
-Grid       : Message        : Average mflops/s per call per node (full): 446761
-Grid       : Message        : Average mflops/s per call per node (full): 306605
-Grid       : Message        :  Stencil 16.0286 GB/s per node
-Grid       : Message        :  Stencil 12.0197 GB/s per node
-Grid       : Message        :  Stencil 17.7871 GB/s per node
-Grid       : Message        :  Stencil 13.0547 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664227
-Grid       : Message        : Average mflops/s per call per node       : 809418
-Grid       : Message        : Average mflops/s per call per node       : 822235
-Grid       : Message        : Average mflops/s per call per node       : 668032
-Grid       : Message        : Average mflops/s per call per node (full): 318100
-Grid       : Message        : Average mflops/s per call per node (full): 361966
-Grid       : Message        : Average mflops/s per call per node (full): 446497
-Grid       : Message        : Average mflops/s per call per node (full): 306591
-Grid       : Message        :  Stencil 13.76 GB/s per node
-Grid       : Message        :  Stencil 17.5266 GB/s per node
-Grid       : Message        :  Stencil 15.9418 GB/s per node
-Grid       : Message        :  Stencil 12.6726 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668512
-Grid       : Message        : Average mflops/s per call per node       : 804067
-Grid       : Message        : Average mflops/s per call per node       : 823981
-Grid       : Message        : Average mflops/s per call per node       : 662906
-Grid       : Message        : Average mflops/s per call per node (full): 315789
-Grid       : Message        : Average mflops/s per call per node (full): 438450
-Grid       : Message        : Average mflops/s per call per node (full): 418927
-Grid       : Message        : Average mflops/s per call per node (full): 302969
-Grid       : Message        :  Stencil 13.7171 GB/s per node
-Grid       : Message        :  Stencil 16.5623 GB/s per node
-Grid       : Message        :  Stencil 17.1249 GB/s per node
-Grid       : Message        :  Stencil 13.0189 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668511
-Grid       : Message        : Average mflops/s per call per node       : 806847
-Grid       : Message        : Average mflops/s per call per node       : 822286
-Grid       : Message        : Average mflops/s per call per node       : 665505
-Grid       : Message        : Average mflops/s per call per node (full): 316746
-Grid       : Message        : Average mflops/s per call per node (full): 435621
-Grid       : Message        : Average mflops/s per call per node (full): 442049
-Grid       : Message        : Average mflops/s per call per node (full): 305990
-Grid       : Message        :  Stencil 13.4105 GB/s per node
-Grid       : Message        :  Stencil 11.3204 GB/s per node
-Grid       : Message        :  Stencil 17.3826 GB/s per node
-Grid       : Message        :  Stencil 12.9017 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668608
-Grid       : Message        : Average mflops/s per call per node       : 809832
-Grid       : Message        : Average mflops/s per call per node       : 824581
-Grid       : Message        : Average mflops/s per call per node       : 668790
-Grid       : Message        : Average mflops/s per call per node (full): 315761
-Grid       : Message        : Average mflops/s per call per node (full): 346191
-Grid       : Message        : Average mflops/s per call per node (full): 444567
-Grid       : Message        : Average mflops/s per call per node (full): 305959
-Grid       : Message        :  Stencil 12.7964 GB/s per node
-Grid       : Message        :  Stencil 16.375 GB/s per node
-Grid       : Message        :  Stencil 18.0128 GB/s per node
-Grid       : Message        :  Stencil 12.5739 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666980
-Grid       : Message        : Average mflops/s per call per node       : 805908
-Grid       : Message        : Average mflops/s per call per node       : 823029
-Grid       : Message        : Average mflops/s per call per node       : 669658
-Grid       : Message        : Average mflops/s per call per node (full): 314223
-Grid       : Message        : Average mflops/s per call per node (full): 432713
-Grid       : Message        : Average mflops/s per call per node (full): 446420
-Grid       : Message        : Average mflops/s per call per node (full): 305992
-Grid       : Message        :  Stencil 12.5729 GB/s per node
-Grid       : Message        :  Stencil 15.7072 GB/s per node
-Grid       : Message        :  Stencil 17.8107 GB/s per node
-Grid       : Message        :  Stencil 12.6056 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667120
-Grid       : Message        : Average mflops/s per call per node       : 808876
-Grid       : Message        : Average mflops/s per call per node       : 821364
-Grid       : Message        : Average mflops/s per call per node       : 667564
-Grid       : Message        : Average mflops/s per call per node (full): 313410
-Grid       : Message        : Average mflops/s per call per node (full): 416363
-Grid       : Message        : Average mflops/s per call per node (full): 446932
-Grid       : Message        : Average mflops/s per call per node (full): 304678
-Grid       : Message        :  Stencil 12.945 GB/s per node
-Grid       : Message        :  Stencil 16.4332 GB/s per node
-Grid       : Message        :  Stencil 18.5724 GB/s per node
-Grid       : Message        :  Stencil 13.4346 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664294
-Grid       : Message        : Average mflops/s per call per node       : 808725
-Grid       : Message        : Average mflops/s per call per node       : 823405
-Grid       : Message        : Average mflops/s per call per node       : 665357
-Grid       : Message        : Average mflops/s per call per node (full): 313774
-Grid       : Message        : Average mflops/s per call per node (full): 434302
-Grid       : Message        : Average mflops/s per call per node (full): 448348
-Grid       : Message        : Average mflops/s per call per node (full): 306846
-Grid       : Message        :  Stencil 12.766 GB/s per node
-Grid       : Message        :  Stencil 10.3715 GB/s per node
-Grid       : Message        :  Stencil 17.2152 GB/s per node
-Grid       : Message        :  Stencil 13.2096 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668238
-Grid       : Message        : Average mflops/s per call per node       : 809538
-Grid       : Message        : Average mflops/s per call per node       : 821750
-Grid       : Message        : Average mflops/s per call per node       : 665493
-Grid       : Message        : Average mflops/s per call per node (full): 314801
-Grid       : Message        : Average mflops/s per call per node (full): 325253
-Grid       : Message        : Average mflops/s per call per node (full): 443233
-Grid       : Message        : Average mflops/s per call per node (full): 305861
-Grid       : Message        :  Stencil 14.2297 GB/s per node
-Grid       : Message        :  Stencil 8.79049 GB/s per node
-Grid       : Message        :  Stencil 17.2774 GB/s per node
-Grid       : Message        :  Stencil 12.9415 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664583
-Grid       : Message        : Average mflops/s per call per node       : 805974
-Grid       : Message        : Average mflops/s per call per node       : 819344
-Grid       : Message        : Average mflops/s per call per node       : 664789
-Grid       : Message        : Average mflops/s per call per node (full): 316928
-Grid       : Message        : Average mflops/s per call per node (full): 286856
-Grid       : Message        : Average mflops/s per call per node (full): 443311
-Grid       : Message        : Average mflops/s per call per node (full): 305485
-Grid       : Message        :  Stencil 12.3881 GB/s per node
-Grid       : Message        :  Stencil 10.9088 GB/s per node
-Grid       : Message        :  Stencil 17.466 GB/s per node
-Grid       : Message        :  Stencil 12.9535 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666538
-Grid       : Message        : Average mflops/s per call per node       : 813628
-Grid       : Message        : Average mflops/s per call per node       : 826126
-Grid       : Message        : Average mflops/s per call per node       : 661329
-Grid       : Message        : Average mflops/s per call per node (full): 311884
-Grid       : Message        : Average mflops/s per call per node (full): 337706
-Grid       : Message        : Average mflops/s per call per node (full): 444964
-Grid       : Message        : Average mflops/s per call per node (full): 305091
-Grid       : Message        :  Stencil 13.237 GB/s per node
-Grid       : Message        :  Stencil 12.3816 GB/s per node
-Grid       : Message        :  Stencil 17.6251 GB/s per node
-Grid       : Message        :  Stencil 14.167 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667943
-Grid       : Message        : Average mflops/s per call per node       : 810737
-Grid       : Message        : Average mflops/s per call per node       : 822289
-Grid       : Message        : Average mflops/s per call per node       : 659425
-Grid       : Message        : Average mflops/s per call per node (full): 315870
-Grid       : Message        : Average mflops/s per call per node (full): 369389
-Grid       : Message        : Average mflops/s per call per node (full): 446280
-Grid       : Message        : Average mflops/s per call per node (full): 305260
-Grid       : Message        :  Stencil 14.7861 GB/s per node
-Grid       : Message        :  Stencil 16.559 GB/s per node
-Grid       : Message        :  Stencil 17.8089 GB/s per node
-Grid       : Message        :  Stencil 13.5257 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663141
-Grid       : Message        : Average mflops/s per call per node       : 805069
-Grid       : Message        : Average mflops/s per call per node       : 815702
-Grid       : Message        : Average mflops/s per call per node       : 664549
-Grid       : Message        : Average mflops/s per call per node (full): 316511
-Grid       : Message        : Average mflops/s per call per node (full): 435230
-Grid       : Message        : Average mflops/s per call per node (full): 444902
-Grid       : Message        : Average mflops/s per call per node (full): 306236
-Grid       : Message        :  Stencil 13.1419 GB/s per node
-Grid       : Message        :  Stencil 18.032 GB/s per node
-Grid       : Message        :  Stencil 17.8433 GB/s per node
-Grid       : Message        :  Stencil 12.9064 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666758
-Grid       : Message        : Average mflops/s per call per node       : 807885
-Grid       : Message        : Average mflops/s per call per node       : 821856
-Grid       : Message        : Average mflops/s per call per node       : 660427
-Grid       : Message        : Average mflops/s per call per node (full): 314625
-Grid       : Message        : Average mflops/s per call per node (full): 440284
-Grid       : Message        : Average mflops/s per call per node (full): 446483
-Grid       : Message        : Average mflops/s per call per node (full): 304466
-Grid       : Message        :  Stencil 12.881 GB/s per node
-Grid       : Message        :  Stencil 17.0216 GB/s per node
-Grid       : Message        :  Stencil 18.7172 GB/s per node
-Grid       : Message        :  Stencil 13.37 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667531
-Grid       : Message        : Average mflops/s per call per node       : 801594
-Grid       : Message        : Average mflops/s per call per node       : 822271
-Grid       : Message        : Average mflops/s per call per node       : 663941
-Grid       : Message        : Average mflops/s per call per node (full): 314968
-Grid       : Message        : Average mflops/s per call per node (full): 435815
-Grid       : Message        : Average mflops/s per call per node (full): 449114
-Grid       : Message        : Average mflops/s per call per node (full): 306456
-Grid       : Message        :  Stencil 12.9034 GB/s per node
-Grid       : Message        :  Stencil 16.9758 GB/s per node
-Grid       : Message        :  Stencil 17.4356 GB/s per node
-Grid       : Message        :  Stencil 12.6256 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669880
-Grid       : Message        : Average mflops/s per call per node       : 804816
-Grid       : Message        : Average mflops/s per call per node       : 830930
-Grid       : Message        : Average mflops/s per call per node       : 658692
-Grid       : Message        : Average mflops/s per call per node (full): 312269
-Grid       : Message        : Average mflops/s per call per node (full): 433171
-Grid       : Message        : Average mflops/s per call per node (full): 445211
-Grid       : Message        : Average mflops/s per call per node (full): 303503
-Grid       : Message        :  Stencil 13.768 GB/s per node
-Grid       : Message        :  Stencil 16.5521 GB/s per node
-Grid       : Message        :  Stencil 17.7681 GB/s per node
-Grid       : Message        :  Stencil 12.1193 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667809
-Grid       : Message        : Average mflops/s per call per node       : 805744
-Grid       : Message        : Average mflops/s per call per node       : 824123
-Grid       : Message        : Average mflops/s per call per node       : 668852
-Grid       : Message        : Average mflops/s per call per node (full): 316518
-Grid       : Message        : Average mflops/s per call per node (full): 431066
-Grid       : Message        : Average mflops/s per call per node (full): 445982
-Grid       : Message        : Average mflops/s per call per node (full): 303187
-Grid       : Message        :  Stencil 13.1464 GB/s per node
-Grid       : Message        :  Stencil 16.9932 GB/s per node
-Grid       : Message        :  Stencil 18.8166 GB/s per node
-Grid       : Message        :  Stencil 13.4326 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669118
-Grid       : Message        : Average mflops/s per call per node       : 806856
-Grid       : Message        : Average mflops/s per call per node       : 823368
-Grid       : Message        : Average mflops/s per call per node       : 666544
-Grid       : Message        : Average mflops/s per call per node (full): 315823
-Grid       : Message        : Average mflops/s per call per node (full): 437027
-Grid       : Message        : Average mflops/s per call per node (full): 443807
-Grid       : Message        : Average mflops/s per call per node (full): 307407
-Grid       : Message        :  Stencil 13.491 GB/s per node
-Grid       : Message        :  Stencil 15.9341 GB/s per node
-Grid       : Message        :  Stencil 17.8917 GB/s per node
-Grid       : Message        :  Stencil 13.607 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668430
-Grid       : Message        : Average mflops/s per call per node       : 804180
-Grid       : Message        : Average mflops/s per call per node       : 822103
-Grid       : Message        : Average mflops/s per call per node       : 661251
-Grid       : Message        : Average mflops/s per call per node (full): 316500
-Grid       : Message        : Average mflops/s per call per node (full): 427481
-Grid       : Message        : Average mflops/s per call per node (full): 443448
-Grid       : Message        : Average mflops/s per call per node (full): 305580
-Grid       : Message        :  Stencil 13.5886 GB/s per node
-Grid       : Message        :  Stencil 17.1736 GB/s per node
-Grid       : Message        :  Stencil 17.3722 GB/s per node
-Grid       : Message        :  Stencil 12.6624 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667919
-Grid       : Message        : Average mflops/s per call per node       : 801088
-Grid       : Message        : Average mflops/s per call per node       : 820868
-Grid       : Message        : Average mflops/s per call per node       : 656734
-Grid       : Message        : Average mflops/s per call per node (full): 313716
-Grid       : Message        : Average mflops/s per call per node (full): 435734
-Grid       : Message        : Average mflops/s per call per node (full): 443237
-Grid       : Message        : Average mflops/s per call per node (full): 303009
-Grid       : Message        :  Stencil 12.5853 GB/s per node
-Grid       : Message        :  Stencil 17.522 GB/s per node
-Grid       : Message        :  Stencil 17.2681 GB/s per node
-Grid       : Message        :  Stencil 12.761 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670653
-Grid       : Message        : Average mflops/s per call per node       : 807256
-Grid       : Message        : Average mflops/s per call per node       : 821593
-Grid       : Message        : Average mflops/s per call per node       : 660975
-Grid       : Message        : Average mflops/s per call per node (full): 314512
-Grid       : Message        : Average mflops/s per call per node (full): 438086
-Grid       : Message        : Average mflops/s per call per node (full): 439179
-Grid       : Message        : Average mflops/s per call per node (full): 303605
-Grid       : Message        :  Stencil 13.697 GB/s per node
-Grid       : Message        :  Stencil 16.6424 GB/s per node
-Grid       : Message        :  Stencil 17.3883 GB/s per node
-Grid       : Message        :  Stencil 13.4778 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670865
-Grid       : Message        : Average mflops/s per call per node       : 803527
-Grid       : Message        : Average mflops/s per call per node       : 823427
-Grid       : Message        : Average mflops/s per call per node       : 659487
-Grid       : Message        : Average mflops/s per call per node (full): 315548
-Grid       : Message        : Average mflops/s per call per node (full): 434599
-Grid       : Message        : Average mflops/s per call per node (full): 444751
-Grid       : Message        : Average mflops/s per call per node (full): 305546
-Grid       : Message        :  Stencil 12.6026 GB/s per node
-Grid       : Message        :  Stencil 16.4971 GB/s per node
-Grid       : Message        :  Stencil 17.1935 GB/s per node
-Grid       : Message        :  Stencil 12.9868 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670945
-Grid       : Message        : Average mflops/s per call per node       : 802586
-Grid       : Message        : Average mflops/s per call per node       : 822190
-Grid       : Message        : Average mflops/s per call per node       : 661417
-Grid       : Message        : Average mflops/s per call per node (full): 314391
-Grid       : Message        : Average mflops/s per call per node (full): 433381
-Grid       : Message        : Average mflops/s per call per node (full): 441169
-Grid       : Message        : Average mflops/s per call per node (full): 304820
-Grid       : Message        :  Stencil 12.2678 GB/s per node
-Grid       : Message        :  Stencil 16.8147 GB/s per node
-Grid       : Message        :  Stencil 19.618 GB/s per node
-Grid       : Message        :  Stencil 13.8669 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669378
-Grid       : Message        : Average mflops/s per call per node       : 800923
-Grid       : Message        : Average mflops/s per call per node       : 818294
-Grid       : Message        : Average mflops/s per call per node       : 661224
-Grid       : Message        : Average mflops/s per call per node (full): 311876
-Grid       : Message        : Average mflops/s per call per node (full): 434853
-Grid       : Message        : Average mflops/s per call per node (full): 449593
-Grid       : Message        : Average mflops/s per call per node (full): 306914
-Grid       : Message        :  Stencil 12.4139 GB/s per node
-Grid       : Message        :  Stencil 17.2397 GB/s per node
-Grid       : Message        :  Stencil 17.7356 GB/s per node
-Grid       : Message        :  Stencil 13.2463 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668627
-Grid       : Message        : Average mflops/s per call per node       : 800456
-Grid       : Message        : Average mflops/s per call per node       : 818989
-Grid       : Message        : Average mflops/s per call per node       : 665886
-Grid       : Message        : Average mflops/s per call per node (full): 313481
-Grid       : Message        : Average mflops/s per call per node (full): 436451
-Grid       : Message        : Average mflops/s per call per node (full): 435769
-Grid       : Message        : Average mflops/s per call per node (full): 306570
-Grid       : Message        :  Stencil 13.5333 GB/s per node
-Grid       : Message        :  Stencil 8.13661 GB/s per node
-Grid       : Message        :  Stencil 16.9329 GB/s per node
-Grid       : Message        :  Stencil 12.8284 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664504
-Grid       : Message        : Average mflops/s per call per node       : 804766
-Grid       : Message        : Average mflops/s per call per node       : 821491
-Grid       : Message        : Average mflops/s per call per node       : 664154
-Grid       : Message        : Average mflops/s per call per node (full): 316740
-Grid       : Message        : Average mflops/s per call per node (full): 270001
-Grid       : Message        : Average mflops/s per call per node (full): 439291
-Grid       : Message        : Average mflops/s per call per node (full): 305179
-Grid       : Message        :  Stencil 14.5014 GB/s per node
-Grid       : Message        :  Stencil 16.4796 GB/s per node
-Grid       : Message        :  Stencil 17.684 GB/s per node
-Grid       : Message        :  Stencil 12.7725 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660744
-Grid       : Message        : Average mflops/s per call per node       : 806758
-Grid       : Message        : Average mflops/s per call per node       : 822278
-Grid       : Message        : Average mflops/s per call per node       : 659832
-Grid       : Message        : Average mflops/s per call per node (full): 316390
-Grid       : Message        : Average mflops/s per call per node (full): 434171
-Grid       : Message        : Average mflops/s per call per node (full): 445325
-Grid       : Message        : Average mflops/s per call per node (full): 303568
-Grid       : Message        :  Stencil 14.5585 GB/s per node
-Grid       : Message        :  Stencil 16.9893 GB/s per node
-Grid       : Message        :  Stencil 17.6812 GB/s per node
-Grid       : Message        :  Stencil 14.2824 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665943
-Grid       : Message        : Average mflops/s per call per node       : 802984
-Grid       : Message        : Average mflops/s per call per node       : 819906
-Grid       : Message        : Average mflops/s per call per node       : 662809
-Grid       : Message        : Average mflops/s per call per node (full): 317714
-Grid       : Message        : Average mflops/s per call per node (full): 438366
-Grid       : Message        : Average mflops/s per call per node (full): 446131
-Grid       : Message        : Average mflops/s per call per node (full): 307211
-Grid       : Message        :  Stencil 15.4553 GB/s per node
-Grid       : Message        :  Stencil 16.9426 GB/s per node
-Grid       : Message        :  Stencil 17.2209 GB/s per node
-Grid       : Message        :  Stencil 12.0195 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664098
-Grid       : Message        : Average mflops/s per call per node       : 802553
-Grid       : Message        : Average mflops/s per call per node       : 824960
-Grid       : Message        : Average mflops/s per call per node       : 664260
-Grid       : Message        : Average mflops/s per call per node (full): 316082
-Grid       : Message        : Average mflops/s per call per node (full): 432776
-Grid       : Message        : Average mflops/s per call per node (full): 443286
-Grid       : Message        : Average mflops/s per call per node (full): 302022
-Grid       : Message        :  Stencil 15.7701 GB/s per node
-Grid       : Message        :  Stencil 16.2108 GB/s per node
-Grid       : Message        :  Stencil 17.2166 GB/s per node
-Grid       : Message        :  Stencil 11.9306 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663469
-Grid       : Message        : Average mflops/s per call per node       : 801281
-Grid       : Message        : Average mflops/s per call per node       : 823253
-Grid       : Message        : Average mflops/s per call per node       : 667932
-Grid       : Message        : Average mflops/s per call per node (full): 318237
-Grid       : Message        : Average mflops/s per call per node (full): 430731
-Grid       : Message        : Average mflops/s per call per node (full): 442433
-Grid       : Message        : Average mflops/s per call per node (full): 300951
-Grid       : Message        :  Stencil 14.2805 GB/s per node
-Grid       : Message        :  Stencil 17.6347 GB/s per node
-Grid       : Message        :  Stencil 18.0806 GB/s per node
-Grid       : Message        :  Stencil 14.1374 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661408
-Grid       : Message        : Average mflops/s per call per node       : 798410
-Grid       : Message        : Average mflops/s per call per node       : 821662
-Grid       : Message        : Average mflops/s per call per node       : 661402
-Grid       : Message        : Average mflops/s per call per node (full): 315896
-Grid       : Message        : Average mflops/s per call per node (full): 436256
-Grid       : Message        : Average mflops/s per call per node (full): 446757
-Grid       : Message        : Average mflops/s per call per node (full): 306241
-Grid       : Message        :  Stencil 12.6773 GB/s per node
-Grid       : Message        :  Stencil 16.234 GB/s per node
-Grid       : Message        :  Stencil 18.0985 GB/s per node
-Grid       : Message        :  Stencil 13.6585 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668940
-Grid       : Message        : Average mflops/s per call per node       : 806973
-Grid       : Message        : Average mflops/s per call per node       : 819816
-Grid       : Message        : Average mflops/s per call per node       : 663629
-Grid       : Message        : Average mflops/s per call per node (full): 315716
-Grid       : Message        : Average mflops/s per call per node (full): 432828
-Grid       : Message        : Average mflops/s per call per node (full): 446896
-Grid       : Message        : Average mflops/s per call per node (full): 307358
-Grid       : Message        :  Stencil 13.1497 GB/s per node
-Grid       : Message        :  Stencil 17.7945 GB/s per node
-Grid       : Message        :  Stencil 17.2219 GB/s per node
-Grid       : Message        :  Stencil 13.6643 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661906
-Grid       : Message        : Average mflops/s per call per node       : 795510
-Grid       : Message        : Average mflops/s per call per node       : 823272
-Grid       : Message        : Average mflops/s per call per node       : 660621
-Grid       : Message        : Average mflops/s per call per node (full): 310264
-Grid       : Message        : Average mflops/s per call per node (full): 438583
-Grid       : Message        : Average mflops/s per call per node (full): 442278
-Grid       : Message        : Average mflops/s per call per node (full): 306590
-Grid       : Message        :  Stencil 13.7303 GB/s per node
-Grid       : Message        :  Stencil 14.0153 GB/s per node
-Grid       : Message        :  Stencil 17.1919 GB/s per node
-Grid       : Message        :  Stencil 12.1508 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665851
-Grid       : Message        : Average mflops/s per call per node       : 806390
-Grid       : Message        : Average mflops/s per call per node       : 826798
-Grid       : Message        : Average mflops/s per call per node       : 668700
-Grid       : Message        : Average mflops/s per call per node (full): 315276
-Grid       : Message        : Average mflops/s per call per node (full): 401084
-Grid       : Message        : Average mflops/s per call per node (full): 440578
-Grid       : Message        : Average mflops/s per call per node (full): 303352
-Grid       : Message        :  Stencil 12.9474 GB/s per node
-Grid       : Message        :  Stencil 16.7575 GB/s per node
-Grid       : Message        :  Stencil 18.5942 GB/s per node
-Grid       : Message        :  Stencil 12.6604 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660774
-Grid       : Message        : Average mflops/s per call per node       : 806489
-Grid       : Message        : Average mflops/s per call per node       : 824226
-Grid       : Message        : Average mflops/s per call per node       : 665717
-Grid       : Message        : Average mflops/s per call per node (full): 311902
-Grid       : Message        : Average mflops/s per call per node (full): 438116
-Grid       : Message        : Average mflops/s per call per node (full): 448158
-Grid       : Message        : Average mflops/s per call per node (full): 304377
-Grid       : Message        :  Stencil 14.3589 GB/s per node
-Grid       : Message        :  Stencil 16.9713 GB/s per node
-Grid       : Message        :  Stencil 17.9925 GB/s per node
-Grid       : Message        :  Stencil 11.8981 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660765
-Grid       : Message        : Average mflops/s per call per node       : 801897
-Grid       : Message        : Average mflops/s per call per node       : 821621
-Grid       : Message        : Average mflops/s per call per node       : 670225
-Grid       : Message        : Average mflops/s per call per node (full): 316036
-Grid       : Message        : Average mflops/s per call per node (full): 436107
-Grid       : Message        : Average mflops/s per call per node (full): 446054
-Grid       : Message        : Average mflops/s per call per node (full): 301453
-Grid       : Message        :  Stencil 13.7643 GB/s per node
-Grid       : Message        :  Stencil 16.6706 GB/s per node
-Grid       : Message        :  Stencil 17.0874 GB/s per node
-Grid       : Message        :  Stencil 12.087 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662129
-Grid       : Message        : Average mflops/s per call per node       : 806221
-Grid       : Message        : Average mflops/s per call per node       : 821227
-Grid       : Message        : Average mflops/s per call per node       : 666438
-Grid       : Message        : Average mflops/s per call per node (full): 315264
-Grid       : Message        : Average mflops/s per call per node (full): 437614
-Grid       : Message        : Average mflops/s per call per node (full): 441200
-Grid       : Message        : Average mflops/s per call per node (full): 302746
-Grid       : Message        :  Stencil 15.2251 GB/s per node
-Grid       : Message        :  Stencil 17.2964 GB/s per node
-Grid       : Message        :  Stencil 18.0299 GB/s per node
-Grid       : Message        :  Stencil 12.6288 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662399
-Grid       : Message        : Average mflops/s per call per node       : 803931
-Grid       : Message        : Average mflops/s per call per node       : 820320
-Grid       : Message        : Average mflops/s per call per node       : 663829
-Grid       : Message        : Average mflops/s per call per node (full): 316822
-Grid       : Message        : Average mflops/s per call per node (full): 437254
-Grid       : Message        : Average mflops/s per call per node (full): 438214
-Grid       : Message        : Average mflops/s per call per node (full): 304733
-Grid       : Message        :  Stencil 13.3113 GB/s per node
-Grid       : Message        :  Stencil 16.7028 GB/s per node
-Grid       : Message        :  Stencil 17.3581 GB/s per node
-Grid       : Message        :  Stencil 12.4669 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664624
-Grid       : Message        : Average mflops/s per call per node       : 805500
-Grid       : Message        : Average mflops/s per call per node       : 830041
-Grid       : Message        : Average mflops/s per call per node       : 662596
-Grid       : Message        : Average mflops/s per call per node (full): 315637
-Grid       : Message        : Average mflops/s per call per node (full): 435982
-Grid       : Message        : Average mflops/s per call per node (full): 445041
-Grid       : Message        : Average mflops/s per call per node (full): 303697
-Grid       : Message        :  Stencil 14.1203 GB/s per node
-Grid       : Message        :  Stencil 17.0781 GB/s per node
-Grid       : Message        :  Stencil 16.4489 GB/s per node
-Grid       : Message        :  Stencil 12.5695 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661332
-Grid       : Message        : Average mflops/s per call per node       : 800493
-Grid       : Message        : Average mflops/s per call per node       : 823953
-Grid       : Message        : Average mflops/s per call per node       : 664719
-Grid       : Message        : Average mflops/s per call per node (full): 315811
-Grid       : Message        : Average mflops/s per call per node (full): 435198
-Grid       : Message        : Average mflops/s per call per node (full): 428685
-Grid       : Message        : Average mflops/s per call per node (full): 304638
-Grid       : Message        :  Stencil 13.7742 GB/s per node
-Grid       : Message        :  Stencil 17.3357 GB/s per node
-Grid       : Message        :  Stencil 18.761 GB/s per node
-Grid       : Message        :  Stencil 12.7716 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663846
-Grid       : Message        : Average mflops/s per call per node       : 801600
-Grid       : Message        : Average mflops/s per call per node       : 819316
-Grid       : Message        : Average mflops/s per call per node       : 669795
-Grid       : Message        : Average mflops/s per call per node (full): 315898
-Grid       : Message        : Average mflops/s per call per node (full): 436843
-Grid       : Message        : Average mflops/s per call per node (full): 447877
-Grid       : Message        : Average mflops/s per call per node (full): 306082
-Grid       : Message        :  Stencil 13.2072 GB/s per node
-Grid       : Message        :  Stencil 16.8379 GB/s per node
-Grid       : Message        :  Stencil 17.4784 GB/s per node
-Grid       : Message        :  Stencil 13.3794 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666336
-Grid       : Message        : Average mflops/s per call per node       : 805100
-Grid       : Message        : Average mflops/s per call per node       : 829079
-Grid       : Message        : Average mflops/s per call per node       : 663901
-Grid       : Message        : Average mflops/s per call per node (full): 315246
-Grid       : Message        : Average mflops/s per call per node (full): 432623
-Grid       : Message        : Average mflops/s per call per node (full): 445247
-Grid       : Message        : Average mflops/s per call per node (full): 306374
-Grid       : Message        :  Stencil 12.3984 GB/s per node
-Grid       : Message        :  Stencil 8.37936 GB/s per node
-Grid       : Message        :  Stencil 17.0412 GB/s per node
-Grid       : Message        :  Stencil 12.0361 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667711
-Grid       : Message        : Average mflops/s per call per node       : 810908
-Grid       : Message        : Average mflops/s per call per node       : 825617
-Grid       : Message        : Average mflops/s per call per node       : 667513
-Grid       : Message        : Average mflops/s per call per node (full): 312778
-Grid       : Message        : Average mflops/s per call per node (full): 276363
-Grid       : Message        : Average mflops/s per call per node (full): 441323
-Grid       : Message        : Average mflops/s per call per node (full): 302656
-Grid       : Message        :  Stencil 13.0633 GB/s per node
-Grid       : Message        :  Stencil 16.331 GB/s per node
-Grid       : Message        :  Stencil 16.8624 GB/s per node
-Grid       : Message        :  Stencil 12.9714 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663183
-Grid       : Message        : Average mflops/s per call per node       : 807105
-Grid       : Message        : Average mflops/s per call per node       : 824210
-Grid       : Message        : Average mflops/s per call per node       : 663522
-Grid       : Message        : Average mflops/s per call per node (full): 314204
-Grid       : Message        : Average mflops/s per call per node (full): 432882
-Grid       : Message        : Average mflops/s per call per node (full): 439099
-Grid       : Message        : Average mflops/s per call per node (full): 304943
-Grid       : Message        :  Stencil 12.3296 GB/s per node
-Grid       : Message        :  Stencil 16.9761 GB/s per node
-Grid       : Message        :  Stencil 17.8134 GB/s per node
-Grid       : Message        :  Stencil 13.9537 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664208
-Grid       : Message        : Average mflops/s per call per node       : 804127
-Grid       : Message        : Average mflops/s per call per node       : 820009
-Grid       : Message        : Average mflops/s per call per node       : 660766
-Grid       : Message        : Average mflops/s per call per node (full): 311581
-Grid       : Message        : Average mflops/s per call per node (full): 438930
-Grid       : Message        : Average mflops/s per call per node (full): 445649
-Grid       : Message        : Average mflops/s per call per node (full): 306281
-Grid       : Message        :  Stencil 12.625 GB/s per node
-Grid       : Message        :  Stencil 17.463 GB/s per node
-Grid       : Message        :  Stencil 17.838 GB/s per node
-Grid       : Message        :  Stencil 14.241 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666891
-Grid       : Message        : Average mflops/s per call per node       : 798629
-Grid       : Message        : Average mflops/s per call per node       : 824752
-Grid       : Message        : Average mflops/s per call per node       : 660324
-Grid       : Message        : Average mflops/s per call per node (full): 313583
-Grid       : Message        : Average mflops/s per call per node (full): 436361
-Grid       : Message        : Average mflops/s per call per node (full): 446309
-Grid       : Message        : Average mflops/s per call per node (full): 306683
-Grid       : Message        :  Stencil 12.6714 GB/s per node
-Grid       : Message        :  Stencil 17.6931 GB/s per node
-Grid       : Message        :  Stencil 17.0934 GB/s per node
-Grid       : Message        :  Stencil 12.7118 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666993
-Grid       : Message        : Average mflops/s per call per node       : 802461
-Grid       : Message        : Average mflops/s per call per node       : 826448
-Grid       : Message        : Average mflops/s per call per node       : 664123
-Grid       : Message        : Average mflops/s per call per node (full): 313724
-Grid       : Message        : Average mflops/s per call per node (full): 438044
-Grid       : Message        : Average mflops/s per call per node (full): 441526
-Grid       : Message        : Average mflops/s per call per node (full): 303764
-Grid       : Message        :  Stencil 12.3299 GB/s per node
-Grid       : Message        :  Stencil 16.8856 GB/s per node
-Grid       : Message        :  Stencil 18.2979 GB/s per node
-Grid       : Message        :  Stencil 12.9924 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666837
-Grid       : Message        : Average mflops/s per call per node       : 804554
-Grid       : Message        : Average mflops/s per call per node       : 822620
-Grid       : Message        : Average mflops/s per call per node       : 667675
-Grid       : Message        : Average mflops/s per call per node (full): 311836
-Grid       : Message        : Average mflops/s per call per node (full): 437328
-Grid       : Message        : Average mflops/s per call per node (full): 448041
-Grid       : Message        : Average mflops/s per call per node (full): 306129
-Grid       : Message        :  Stencil 12.4805 GB/s per node
-Grid       : Message        :  Stencil 16.7107 GB/s per node
-Grid       : Message        :  Stencil 18.7081 GB/s per node
-Grid       : Message        :  Stencil 11.8256 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666222
-Grid       : Message        : Average mflops/s per call per node       : 804660
-Grid       : Message        : Average mflops/s per call per node       : 821869
-Grid       : Message        : Average mflops/s per call per node       : 668984
-Grid       : Message        : Average mflops/s per call per node (full): 312524
-Grid       : Message        : Average mflops/s per call per node (full): 436016
-Grid       : Message        : Average mflops/s per call per node (full): 445127
-Grid       : Message        : Average mflops/s per call per node (full): 299541
-Grid       : Message        :  Stencil 13.1368 GB/s per node
-Grid       : Message        :  Stencil 16.7553 GB/s per node
-Grid       : Message        :  Stencil 17.9153 GB/s per node
-Grid       : Message        :  Stencil 13.4052 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663625
-Grid       : Message        : Average mflops/s per call per node       : 811113
-Grid       : Message        : Average mflops/s per call per node       : 825860
-Grid       : Message        : Average mflops/s per call per node       : 663241
-Grid       : Message        : Average mflops/s per call per node (full): 314396
-Grid       : Message        : Average mflops/s per call per node (full): 436770
-Grid       : Message        : Average mflops/s per call per node (full): 446415
-Grid       : Message        : Average mflops/s per call per node (full): 303725
-Grid       : Message        :  Stencil 13.5957 GB/s per node
-Grid       : Message        :  Stencil 17.1293 GB/s per node
-Grid       : Message        :  Stencil 17.3253 GB/s per node
-Grid       : Message        :  Stencil 12.8351 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668326
-Grid       : Message        : Average mflops/s per call per node       : 804576
-Grid       : Message        : Average mflops/s per call per node       : 817037
-Grid       : Message        : Average mflops/s per call per node       : 665455
-Grid       : Message        : Average mflops/s per call per node (full): 314691
-Grid       : Message        : Average mflops/s per call per node (full): 435872
-Grid       : Message        : Average mflops/s per call per node (full): 443981
-Grid       : Message        : Average mflops/s per call per node (full): 306470
-Grid       : Message        :  Stencil 14.4764 GB/s per node
-Grid       : Message        :  Stencil 17.9404 GB/s per node
-Grid       : Message        :  Stencil 18.3239 GB/s per node
-Grid       : Message        :  Stencil 14.7409 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666658
-Grid       : Message        : Average mflops/s per call per node       : 801177
-Grid       : Message        : Average mflops/s per call per node       : 824112
-Grid       : Message        : Average mflops/s per call per node       : 660454
-Grid       : Message        : Average mflops/s per call per node (full): 316984
-Grid       : Message        : Average mflops/s per call per node (full): 440871
-Grid       : Message        : Average mflops/s per call per node (full): 448721
-Grid       : Message        : Average mflops/s per call per node (full): 307159
-Grid       : Message        :  Stencil 12.4847 GB/s per node
-Grid       : Message        :  Stencil 17.617 GB/s per node
-Grid       : Message        :  Stencil 17.778 GB/s per node
-Grid       : Message        :  Stencil 13.3334 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 673753
-Grid       : Message        : Average mflops/s per call per node       : 800196
-Grid       : Message        : Average mflops/s per call per node       : 821568
-Grid       : Message        : Average mflops/s per call per node       : 664242
-Grid       : Message        : Average mflops/s per call per node (full): 313021
-Grid       : Message        : Average mflops/s per call per node (full): 436903
-Grid       : Message        : Average mflops/s per call per node (full): 443523
-Grid       : Message        : Average mflops/s per call per node (full): 305924
-Grid       : Message        :  Stencil 13.2565 GB/s per node
-Grid       : Message        :  Stencil 17.3985 GB/s per node
-Grid       : Message        :  Stencil 17.83 GB/s per node
-Grid       : Message        :  Stencil 12.8742 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670556
-Grid       : Message        : Average mflops/s per call per node       : 806249
-Grid       : Message        : Average mflops/s per call per node       : 820354
-Grid       : Message        : Average mflops/s per call per node       : 663650
-Grid       : Message        : Average mflops/s per call per node (full): 317296
-Grid       : Message        : Average mflops/s per call per node (full): 440128
-Grid       : Message        : Average mflops/s per call per node (full): 445629
-Grid       : Message        : Average mflops/s per call per node (full): 305392
-Grid       : Message        :  Stencil 13.6076 GB/s per node
-Grid       : Message        :  Stencil 16.6494 GB/s per node
-Grid       : Message        :  Stencil 17.5183 GB/s per node
-Grid       : Message        :  Stencil 14.4296 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666624
-Grid       : Message        : Average mflops/s per call per node       : 804377
-Grid       : Message        : Average mflops/s per call per node       : 817964
-Grid       : Message        : Average mflops/s per call per node       : 665206
-Grid       : Message        : Average mflops/s per call per node (full): 316269
-Grid       : Message        : Average mflops/s per call per node (full): 434950
-Grid       : Message        : Average mflops/s per call per node (full): 443647
-Grid       : Message        : Average mflops/s per call per node (full): 307731
-Grid       : Message        :  Stencil 14.0593 GB/s per node
-Grid       : Message        :  Stencil 17.2712 GB/s per node
-Grid       : Message        :  Stencil 18.0003 GB/s per node
-Grid       : Message        :  Stencil 12.4421 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666307
-Grid       : Message        : Average mflops/s per call per node       : 803766
-Grid       : Message        : Average mflops/s per call per node       : 823949
-Grid       : Message        : Average mflops/s per call per node       : 666058
-Grid       : Message        : Average mflops/s per call per node (full): 317264
-Grid       : Message        : Average mflops/s per call per node (full): 437667
-Grid       : Message        : Average mflops/s per call per node (full): 437707
-Grid       : Message        : Average mflops/s per call per node (full): 299945
-Grid       : Message        :  Stencil 12.4696 GB/s per node
-Grid       : Message        :  Stencil 16.5084 GB/s per node
-Grid       : Message        :  Stencil 17.5922 GB/s per node
-Grid       : Message        :  Stencil 12.8264 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 672217
-Grid       : Message        : Average mflops/s per call per node       : 807160
-Grid       : Message        : Average mflops/s per call per node       : 825964
-Grid       : Message        : Average mflops/s per call per node       : 667014
-Grid       : Message        : Average mflops/s per call per node (full): 313820
-Grid       : Message        : Average mflops/s per call per node (full): 433654
-Grid       : Message        : Average mflops/s per call per node (full): 446353
-Grid       : Message        : Average mflops/s per call per node (full): 305633
-Grid       : Message        :  Stencil 12.6734 GB/s per node
-Grid       : Message        :  Stencil 16.4931 GB/s per node
-Grid       : Message        :  Stencil 17.477 GB/s per node
-Grid       : Message        :  Stencil 14.7119 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671820
-Grid       : Message        : Average mflops/s per call per node       : 804476
-Grid       : Message        : Average mflops/s per call per node       : 817111
-Grid       : Message        : Average mflops/s per call per node       : 658700
-Grid       : Message        : Average mflops/s per call per node (full): 314863
-Grid       : Message        : Average mflops/s per call per node (full): 435448
-Grid       : Message        : Average mflops/s per call per node (full): 443580
-Grid       : Message        : Average mflops/s per call per node (full): 306743
-Grid       : Message        :  Stencil 12.5016 GB/s per node
-Grid       : Message        :  Stencil 9.52792 GB/s per node
-Grid       : Message        :  Stencil 17.9562 GB/s per node
-Grid       : Message        :  Stencil 12.3768 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671064
-Grid       : Message        : Average mflops/s per call per node       : 815749
-Grid       : Message        : Average mflops/s per call per node       : 821584
-Grid       : Message        : Average mflops/s per call per node       : 670523
-Grid       : Message        : Average mflops/s per call per node (full): 313644
-Grid       : Message        : Average mflops/s per call per node (full): 305467
-Grid       : Message        : Average mflops/s per call per node (full): 445200
-Grid       : Message        : Average mflops/s per call per node (full): 305141
-Grid       : Message        :  Stencil 12.1447 GB/s per node
-Grid       : Message        :  Stencil 16.8099 GB/s per node
-Grid       : Message        :  Stencil 17.3329 GB/s per node
-Grid       : Message        :  Stencil 13.932 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669301
-Grid       : Message        : Average mflops/s per call per node       : 806765
-Grid       : Message        : Average mflops/s per call per node       : 821776
-Grid       : Message        : Average mflops/s per call per node       : 657008
-Grid       : Message        : Average mflops/s per call per node (full): 309647
-Grid       : Message        : Average mflops/s per call per node (full): 438769
-Grid       : Message        : Average mflops/s per call per node (full): 444367
-Grid       : Message        : Average mflops/s per call per node (full): 304731
-Grid       : Message        :  Stencil 12.6801 GB/s per node
-Grid       : Message        :  Stencil 16.7348 GB/s per node
-Grid       : Message        :  Stencil 17.7477 GB/s per node
-Grid       : Message        :  Stencil 12.265 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670334
-Grid       : Message        : Average mflops/s per call per node       : 807859
-Grid       : Message        : Average mflops/s per call per node       : 825198
-Grid       : Message        : Average mflops/s per call per node       : 664489
-Grid       : Message        : Average mflops/s per call per node (full): 315221
-Grid       : Message        : Average mflops/s per call per node (full): 436968
-Grid       : Message        : Average mflops/s per call per node (full): 446639
-Grid       : Message        : Average mflops/s per call per node (full): 303895
-Grid       : Message        :  Stencil 12.4594 GB/s per node
-Grid       : Message        :  Stencil 17.1054 GB/s per node
-Grid       : Message        :  Stencil 17.0036 GB/s per node
-Grid       : Message        :  Stencil 13.1651 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667348
-Grid       : Message        : Average mflops/s per call per node       : 804015
-Grid       : Message        : Average mflops/s per call per node       : 826743
-Grid       : Message        : Average mflops/s per call per node       : 661906
-Grid       : Message        : Average mflops/s per call per node (full): 311877
-Grid       : Message        : Average mflops/s per call per node (full): 437038
-Grid       : Message        : Average mflops/s per call per node (full): 441547
-Grid       : Message        : Average mflops/s per call per node (full): 305557
-Grid       : Message        :  Stencil 13.786 GB/s per node
-Grid       : Message        :  Stencil 16.8368 GB/s per node
-Grid       : Message        :  Stencil 17.6723 GB/s per node
-Grid       : Message        :  Stencil 12.2742 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668162
-Grid       : Message        : Average mflops/s per call per node       : 800651
-Grid       : Message        : Average mflops/s per call per node       : 819596
-Grid       : Message        : Average mflops/s per call per node       : 664426
-Grid       : Message        : Average mflops/s per call per node (full): 316998
-Grid       : Message        : Average mflops/s per call per node (full): 436396
-Grid       : Message        : Average mflops/s per call per node (full): 443602
-Grid       : Message        : Average mflops/s per call per node (full): 303972
-Grid       : Message        :  Stencil 13.7633 GB/s per node
-Grid       : Message        :  Stencil 16.2817 GB/s per node
-Grid       : Message        :  Stencil 17.8035 GB/s per node
-Grid       : Message        :  Stencil 12.6944 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664684
-Grid       : Message        : Average mflops/s per call per node       : 808546
-Grid       : Message        : Average mflops/s per call per node       : 822845
-Grid       : Message        : Average mflops/s per call per node       : 664413
-Grid       : Message        : Average mflops/s per call per node (full): 316050
-Grid       : Message        : Average mflops/s per call per node (full): 432889
-Grid       : Message        : Average mflops/s per call per node (full): 446074
-Grid       : Message        : Average mflops/s per call per node (full): 304574
-Grid       : Message        :  Stencil 13.4668 GB/s per node
-Grid       : Message        :  Stencil 16.8483 GB/s per node
-Grid       : Message        :  Stencil 18.8989 GB/s per node
-Grid       : Message        :  Stencil 11.9368 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667145
-Grid       : Message        : Average mflops/s per call per node       : 802864
-Grid       : Message        : Average mflops/s per call per node       : 814890
-Grid       : Message        : Average mflops/s per call per node       : 669128
-Grid       : Message        : Average mflops/s per call per node (full): 315400
-Grid       : Message        : Average mflops/s per call per node (full): 436546
-Grid       : Message        : Average mflops/s per call per node (full): 447561
-Grid       : Message        : Average mflops/s per call per node (full): 301691
-Grid       : Message        :  Stencil 13.2105 GB/s per node
-Grid       : Message        :  Stencil 16.9977 GB/s per node
-Grid       : Message        :  Stencil 17.0903 GB/s per node
-Grid       : Message        :  Stencil 12.057 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663785
-Grid       : Message        : Average mflops/s per call per node       : 803215
-Grid       : Message        : Average mflops/s per call per node       : 823658
-Grid       : Message        : Average mflops/s per call per node       : 663578
-Grid       : Message        : Average mflops/s per call per node (full): 314991
-Grid       : Message        : Average mflops/s per call per node (full): 436633
-Grid       : Message        : Average mflops/s per call per node (full): 441939
-Grid       : Message        : Average mflops/s per call per node (full): 301989
-Grid       : Message        :  Stencil 14.4148 GB/s per node
-Grid       : Message        :  Stencil 17.3747 GB/s per node
-Grid       : Message        :  Stencil 17.3421 GB/s per node
-Grid       : Message        :  Stencil 12.1089 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662050
-Grid       : Message        : Average mflops/s per call per node       : 800697
-Grid       : Message        : Average mflops/s per call per node       : 823547
-Grid       : Message        : Average mflops/s per call per node       : 672339
-Grid       : Message        : Average mflops/s per call per node (full): 315394
-Grid       : Message        : Average mflops/s per call per node (full): 438004
-Grid       : Message        : Average mflops/s per call per node (full): 440362
-Grid       : Message        : Average mflops/s per call per node (full): 303484
-Grid       : Message        :  Stencil 14.5963 GB/s per node
-Grid       : Message        :  Stencil 9.37368 GB/s per node
-Grid       : Message        :  Stencil 17.0918 GB/s per node
-Grid       : Message        :  Stencil 12.5873 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663264
-Grid       : Message        : Average mflops/s per call per node       : 808126
-Grid       : Message        : Average mflops/s per call per node       : 824264
-Grid       : Message        : Average mflops/s per call per node       : 668386
-Grid       : Message        : Average mflops/s per call per node (full): 316801
-Grid       : Message        : Average mflops/s per call per node (full): 300452
-Grid       : Message        : Average mflops/s per call per node (full): 441820
-Grid       : Message        : Average mflops/s per call per node (full): 304562
-Grid       : Message        :  Stencil 12.9733 GB/s per node
-Grid       : Message        :  Stencil 18.1228 GB/s per node
-Grid       : Message        :  Stencil 17.4571 GB/s per node
-Grid       : Message        :  Stencil 13.2088 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665026
-Grid       : Message        : Average mflops/s per call per node       : 803643
-Grid       : Message        : Average mflops/s per call per node       : 820059
-Grid       : Message        : Average mflops/s per call per node       : 662757
-Grid       : Message        : Average mflops/s per call per node (full): 314393
-Grid       : Message        : Average mflops/s per call per node (full): 440506
-Grid       : Message        : Average mflops/s per call per node (full): 435037
-Grid       : Message        : Average mflops/s per call per node (full): 304707
-Grid       : Message        :  Stencil 14.9553 GB/s per node
-Grid       : Message        :  Stencil 17.5984 GB/s per node
-Grid       : Message        :  Stencil 17.2416 GB/s per node
-Grid       : Message        :  Stencil 12.3825 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662569
-Grid       : Message        : Average mflops/s per call per node       : 803959
-Grid       : Message        : Average mflops/s per call per node       : 824464
-Grid       : Message        : Average mflops/s per call per node       : 663959
-Grid       : Message        : Average mflops/s per call per node (full): 316442
-Grid       : Message        : Average mflops/s per call per node (full): 440130
-Grid       : Message        : Average mflops/s per call per node (full): 442776
-Grid       : Message        : Average mflops/s per call per node (full): 303533
-Grid       : Message        :  Stencil 13.6284 GB/s per node
-Grid       : Message        :  Stencil 16.7526 GB/s per node
-Grid       : Message        :  Stencil 17.4366 GB/s per node
-Grid       : Message        :  Stencil 11.8692 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664654
-Grid       : Message        : Average mflops/s per call per node       : 805596
-Grid       : Message        : Average mflops/s per call per node       : 826425
-Grid       : Message        : Average mflops/s per call per node       : 666243
-Grid       : Message        : Average mflops/s per call per node (full): 315764
-Grid       : Message        : Average mflops/s per call per node (full): 437992
-Grid       : Message        : Average mflops/s per call per node (full): 446462
-Grid       : Message        : Average mflops/s per call per node (full): 300689
-Grid       : Message        :  Stencil 13.1773 GB/s per node
-Grid       : Message        :  Stencil 17.1496 GB/s per node
-Grid       : Message        :  Stencil 17.2845 GB/s per node
-Grid       : Message        :  Stencil 12.2689 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662457
-Grid       : Message        : Average mflops/s per call per node       : 799747
-Grid       : Message        : Average mflops/s per call per node       : 823813
-Grid       : Message        : Average mflops/s per call per node       : 668403
-Grid       : Message        : Average mflops/s per call per node (full): 315019
-Grid       : Message        : Average mflops/s per call per node (full): 437159
-Grid       : Message        : Average mflops/s per call per node (full): 443460
-Grid       : Message        : Average mflops/s per call per node (full): 304224
-Grid       : Message        :  Stencil 13.1181 GB/s per node
-Grid       : Message        :  Stencil 17.2829 GB/s per node
-Grid       : Message        :  Stencil 17.6311 GB/s per node
-Grid       : Message        :  Stencil 12.7872 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662005
-Grid       : Message        : Average mflops/s per call per node       : 810197
-Grid       : Message        : Average mflops/s per call per node       : 821979
-Grid       : Message        : Average mflops/s per call per node       : 659287
-Grid       : Message        : Average mflops/s per call per node (full): 311259
-Grid       : Message        : Average mflops/s per call per node (full): 440542
-Grid       : Message        : Average mflops/s per call per node (full): 445679
-Grid       : Message        : Average mflops/s per call per node (full): 304177
-Grid       : Message        :  Stencil 12.2806 GB/s per node
-Grid       : Message        :  Stencil 17.0642 GB/s per node
-Grid       : Message        :  Stencil 17.5733 GB/s per node
-Grid       : Message        :  Stencil 12.6914 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669888
-Grid       : Message        : Average mflops/s per call per node       : 805610
-Grid       : Message        : Average mflops/s per call per node       : 824584
-Grid       : Message        : Average mflops/s per call per node       : 657127
-Grid       : Message        : Average mflops/s per call per node (full): 312348
-Grid       : Message        : Average mflops/s per call per node (full): 439208
-Grid       : Message        : Average mflops/s per call per node (full): 443666
-Grid       : Message        : Average mflops/s per call per node (full): 303205
-Grid       : Message        :  Stencil 12.9661 GB/s per node
-Grid       : Message        :  Stencil 16.5873 GB/s per node
-Grid       : Message        :  Stencil 17.5306 GB/s per node
-Grid       : Message        :  Stencil 12.8656 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667796
-Grid       : Message        : Average mflops/s per call per node       : 804159
-Grid       : Message        : Average mflops/s per call per node       : 822987
-Grid       : Message        : Average mflops/s per call per node       : 667076
-Grid       : Message        : Average mflops/s per call per node (full): 316412
-Grid       : Message        : Average mflops/s per call per node (full): 435030
-Grid       : Message        : Average mflops/s per call per node (full): 445900
-Grid       : Message        : Average mflops/s per call per node (full): 305052
-Grid       : Message        :  Stencil 13.1916 GB/s per node
-Grid       : Message        :  Stencil 10.271 GB/s per node
-Grid       : Message        :  Stencil 18.4837 GB/s per node
-Grid       : Message        :  Stencil 12.6862 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667469
-Grid       : Message        : Average mflops/s per call per node       : 809427
-Grid       : Message        : Average mflops/s per call per node       : 815268
-Grid       : Message        : Average mflops/s per call per node       : 665818
-Grid       : Message        : Average mflops/s per call per node (full): 315679
-Grid       : Message        : Average mflops/s per call per node (full): 323040
-Grid       : Message        : Average mflops/s per call per node (full): 446261
-Grid       : Message        : Average mflops/s per call per node (full): 304701
-Grid       : Message        :  Stencil 13.1754 GB/s per node
-Grid       : Message        :  Stencil 15.6857 GB/s per node
-Grid       : Message        :  Stencil 17.1186 GB/s per node
-Grid       : Message        :  Stencil 12.8238 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668281
-Grid       : Message        : Average mflops/s per call per node       : 804696
-Grid       : Message        : Average mflops/s per call per node       : 824252
-Grid       : Message        : Average mflops/s per call per node       : 661608
-Grid       : Message        : Average mflops/s per call per node (full): 316585
-Grid       : Message        : Average mflops/s per call per node (full): 424725
-Grid       : Message        : Average mflops/s per call per node (full): 442445
-Grid       : Message        : Average mflops/s per call per node (full): 304938
-Grid       : Message        :  Stencil 14.1363 GB/s per node
-Grid       : Message        :  Stencil 17.2076 GB/s per node
-Grid       : Message        :  Stencil 17.2527 GB/s per node
-Grid       : Message        :  Stencil 12.6664 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664212
-Grid       : Message        : Average mflops/s per call per node       : 807256
-Grid       : Message        : Average mflops/s per call per node       : 828470
-Grid       : Message        : Average mflops/s per call per node       : 658701
-Grid       : Message        : Average mflops/s per call per node (full): 316642
-Grid       : Message        : Average mflops/s per call per node (full): 440364
-Grid       : Message        : Average mflops/s per call per node (full): 444454
-Grid       : Message        : Average mflops/s per call per node (full): 300892
-Grid       : Message        :  Stencil 12.4873 GB/s per node
-Grid       : Message        :  Stencil 18.1535 GB/s per node
-Grid       : Message        :  Stencil 17.2971 GB/s per node
-Grid       : Message        :  Stencil 12.6544 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669576
-Grid       : Message        : Average mflops/s per call per node       : 804299
-Grid       : Message        : Average mflops/s per call per node       : 824912
-Grid       : Message        : Average mflops/s per call per node       : 666060
-Grid       : Message        : Average mflops/s per call per node (full): 314263
-Grid       : Message        : Average mflops/s per call per node (full): 442052
-Grid       : Message        : Average mflops/s per call per node (full): 443849
-Grid       : Message        : Average mflops/s per call per node (full): 304899
-Grid       : Message        :  Stencil 12.5467 GB/s per node
-Grid       : Message        :  Stencil 16.4949 GB/s per node
-Grid       : Message        :  Stencil 17.1648 GB/s per node
-Grid       : Message        :  Stencil 12.1105 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668703
-Grid       : Message        : Average mflops/s per call per node       : 801782
-Grid       : Message        : Average mflops/s per call per node       : 818095
-Grid       : Message        : Average mflops/s per call per node       : 667291
-Grid       : Message        : Average mflops/s per call per node (full): 314193
-Grid       : Message        : Average mflops/s per call per node (full): 433869
-Grid       : Message        : Average mflops/s per call per node (full): 441706
-Grid       : Message        : Average mflops/s per call per node (full): 303151
-Grid       : Message        :  Stencil 13.4886 GB/s per node
-Grid       : Message        :  Stencil 16.6148 GB/s per node
-Grid       : Message        :  Stencil 17.7462 GB/s per node
-Grid       : Message        :  Stencil 11.9484 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666020
-Grid       : Message        : Average mflops/s per call per node       : 806462
-Grid       : Message        : Average mflops/s per call per node       : 821792
-Grid       : Message        : Average mflops/s per call per node       : 667299
-Grid       : Message        : Average mflops/s per call per node (full): 316359
-Grid       : Message        : Average mflops/s per call per node (full): 433941
-Grid       : Message        : Average mflops/s per call per node (full): 446264
-Grid       : Message        : Average mflops/s per call per node (full): 301910
-Grid       : Message        :  Stencil 13.7286 GB/s per node
-Grid       : Message        :  Stencil 16.5805 GB/s per node
-Grid       : Message        :  Stencil 17.4372 GB/s per node
-Grid       : Message        :  Stencil 13.3038 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664492
-Grid       : Message        : Average mflops/s per call per node       : 805003
-Grid       : Message        : Average mflops/s per call per node       : 819586
-Grid       : Message        : Average mflops/s per call per node       : 660935
-Grid       : Message        : Average mflops/s per call per node (full): 314319
-Grid       : Message        : Average mflops/s per call per node (full): 434736
-Grid       : Message        : Average mflops/s per call per node (full): 435192
-Grid       : Message        : Average mflops/s per call per node (full): 306127
-Grid       : Message        :  Stencil 13.5175 GB/s per node
-Grid       : Message        :  Stencil 16.9837 GB/s per node
-Grid       : Message        :  Stencil 17.7435 GB/s per node
-Grid       : Message        :  Stencil 13.2658 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667154
-Grid       : Message        : Average mflops/s per call per node       : 803437
-Grid       : Message        : Average mflops/s per call per node       : 822090
-Grid       : Message        : Average mflops/s per call per node       : 660675
-Grid       : Message        : Average mflops/s per call per node (full): 315651
-Grid       : Message        : Average mflops/s per call per node (full): 437381
-Grid       : Message        : Average mflops/s per call per node (full): 444310
-Grid       : Message        : Average mflops/s per call per node (full): 305053
-Grid       : Message        :  Stencil 12.718 GB/s per node
-Grid       : Message        :  Stencil 13.4649 GB/s per node
-Grid       : Message        :  Stencil 17.1457 GB/s per node
-Grid       : Message        :  Stencil 13.5868 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670258
-Grid       : Message        : Average mflops/s per call per node       : 808276
-Grid       : Message        : Average mflops/s per call per node       : 823242
-Grid       : Message        : Average mflops/s per call per node       : 662121
-Grid       : Message        : Average mflops/s per call per node (full): 314845
-Grid       : Message        : Average mflops/s per call per node (full): 389476
-Grid       : Message        : Average mflops/s per call per node (full): 442441
-Grid       : Message        : Average mflops/s per call per node (full): 306231
-Grid       : Message        :  Stencil 13.9675 GB/s per node
-Grid       : Message        :  Stencil 17.0644 GB/s per node
-Grid       : Message        :  Stencil 17.6995 GB/s per node
-Grid       : Message        :  Stencil 12.7459 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663407
-Grid       : Message        : Average mflops/s per call per node       : 808739
-Grid       : Message        : Average mflops/s per call per node       : 819816
-Grid       : Message        : Average mflops/s per call per node       : 667386
-Grid       : Message        : Average mflops/s per call per node (full): 315984
-Grid       : Message        : Average mflops/s per call per node (full): 440156
-Grid       : Message        : Average mflops/s per call per node (full): 445980
-Grid       : Message        : Average mflops/s per call per node (full): 305803
-Grid       : Message        :  Stencil 14.1632 GB/s per node
-Grid       : Message        :  Stencil 15.7343 GB/s per node
-Grid       : Message        :  Stencil 18.4273 GB/s per node
-Grid       : Message        :  Stencil 14.4432 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664937
-Grid       : Message        : Average mflops/s per call per node       : 805253
-Grid       : Message        : Average mflops/s per call per node       : 828233
-Grid       : Message        : Average mflops/s per call per node       : 664513
-Grid       : Message        : Average mflops/s per call per node (full): 315813
-Grid       : Message        : Average mflops/s per call per node (full): 415168
-Grid       : Message        : Average mflops/s per call per node (full): 447412
-Grid       : Message        : Average mflops/s per call per node (full): 307131
-Grid       : Message        :  Stencil 12.4829 GB/s per node
-Grid       : Message        :  Stencil 16.5053 GB/s per node
-Grid       : Message        :  Stencil 17.2726 GB/s per node
-Grid       : Message        :  Stencil 14.047 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670858
-Grid       : Message        : Average mflops/s per call per node       : 801237
-Grid       : Message        : Average mflops/s per call per node       : 822537
-Grid       : Message        : Average mflops/s per call per node       : 662443
-Grid       : Message        : Average mflops/s per call per node (full): 313466
-Grid       : Message        : Average mflops/s per call per node (full): 433622
-Grid       : Message        : Average mflops/s per call per node (full): 443957
-Grid       : Message        : Average mflops/s per call per node (full): 306748
-Grid       : Message        :  Stencil 12.8276 GB/s per node
-Grid       : Message        :  Stencil 17.1233 GB/s per node
-Grid       : Message        :  Stencil 18.0525 GB/s per node
-Grid       : Message        :  Stencil 13.1973 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668220
-Grid       : Message        : Average mflops/s per call per node       : 800530
-Grid       : Message        : Average mflops/s per call per node       : 824932
-Grid       : Message        : Average mflops/s per call per node       : 665198
-Grid       : Message        : Average mflops/s per call per node (full): 315263
-Grid       : Message        : Average mflops/s per call per node (full): 438804
-Grid       : Message        : Average mflops/s per call per node (full): 446733
-Grid       : Message        : Average mflops/s per call per node (full): 306020
-Grid       : Message        :  Stencil 13.3079 GB/s per node
-Grid       : Message        :  Stencil 17.2573 GB/s per node
-Grid       : Message        :  Stencil 17.4127 GB/s per node
-Grid       : Message        :  Stencil 12.1482 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668893
-Grid       : Message        : Average mflops/s per call per node       : 806882
-Grid       : Message        : Average mflops/s per call per node       : 819174
-Grid       : Message        : Average mflops/s per call per node       : 664891
-Grid       : Message        : Average mflops/s per call per node (full): 315958
-Grid       : Message        : Average mflops/s per call per node (full): 438004
-Grid       : Message        : Average mflops/s per call per node (full): 443438
-Grid       : Message        : Average mflops/s per call per node (full): 303070
-Grid       : Message        :  Stencil 13.6325 GB/s per node
-Grid       : Message        :  Stencil 13.3344 GB/s per node
-Grid       : Message        :  Stencil 17.6158 GB/s per node
-Grid       : Message        :  Stencil 13.8209 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664060
-Grid       : Message        : Average mflops/s per call per node       : 806781
-Grid       : Message        : Average mflops/s per call per node       : 819397
-Grid       : Message        : Average mflops/s per call per node       : 659580
-Grid       : Message        : Average mflops/s per call per node (full): 315675
-Grid       : Message        : Average mflops/s per call per node (full): 387874
-Grid       : Message        : Average mflops/s per call per node (full): 443746
-Grid       : Message        : Average mflops/s per call per node (full): 305557
-Grid       : Message        :  Stencil 12.8853 GB/s per node
-Grid       : Message        :  Stencil 16.7132 GB/s per node
-Grid       : Message        :  Stencil 17.1811 GB/s per node
-Grid       : Message        :  Stencil 13.2971 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668706
-Grid       : Message        : Average mflops/s per call per node       : 805007
-Grid       : Message        : Average mflops/s per call per node       : 820607
-Grid       : Message        : Average mflops/s per call per node       : 660532
-Grid       : Message        : Average mflops/s per call per node (full): 315550
-Grid       : Message        : Average mflops/s per call per node (full): 436813
-Grid       : Message        : Average mflops/s per call per node (full): 441014
-Grid       : Message        : Average mflops/s per call per node (full): 303187
-Grid       : Message        :  Stencil 12.3614 GB/s per node
-Grid       : Message        :  Stencil 16.6071 GB/s per node
-Grid       : Message        :  Stencil 19.2775 GB/s per node
-Grid       : Message        :  Stencil 12.0097 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667522
-Grid       : Message        : Average mflops/s per call per node       : 802893
-Grid       : Message        : Average mflops/s per call per node       : 818988
-Grid       : Message        : Average mflops/s per call per node       : 663512
-Grid       : Message        : Average mflops/s per call per node (full): 312664
-Grid       : Message        : Average mflops/s per call per node (full): 435656
-Grid       : Message        : Average mflops/s per call per node (full): 449837
-Grid       : Message        : Average mflops/s per call per node (full): 300482
-Grid       : Message        :  Stencil 13.3121 GB/s per node
-Grid       : Message        :  Stencil 17.2751 GB/s per node
-Grid       : Message        :  Stencil 17.4883 GB/s per node
-Grid       : Message        :  Stencil 13.266 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665762
-Grid       : Message        : Average mflops/s per call per node       : 801452
-Grid       : Message        : Average mflops/s per call per node       : 827915
-Grid       : Message        : Average mflops/s per call per node       : 663806
-Grid       : Message        : Average mflops/s per call per node (full): 314945
-Grid       : Message        : Average mflops/s per call per node (full): 437703
-Grid       : Message        : Average mflops/s per call per node (full): 443208
-Grid       : Message        : Average mflops/s per call per node (full): 305507
-Grid       : Message        :  Stencil 13.3057 GB/s per node
-Grid       : Message        :  Stencil 16.4284 GB/s per node
-Grid       : Message        :  Stencil 17.7674 GB/s per node
-Grid       : Message        :  Stencil 12.8116 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665746
-Grid       : Message        : Average mflops/s per call per node       : 806662
-Grid       : Message        : Average mflops/s per call per node       : 820972
-Grid       : Message        : Average mflops/s per call per node       : 663597
-Grid       : Message        : Average mflops/s per call per node (full): 315385
-Grid       : Message        : Average mflops/s per call per node (full): 434085
-Grid       : Message        : Average mflops/s per call per node (full): 445698
-Grid       : Message        : Average mflops/s per call per node (full): 304993
-Grid       : Message        :  Stencil 13.4849 GB/s per node
-Grid       : Message        :  Stencil 9.38352 GB/s per node
-Grid       : Message        :  Stencil 17.645 GB/s per node
-Grid       : Message        :  Stencil 13.2214 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663124
-Grid       : Message        : Average mflops/s per call per node       : 809265
-Grid       : Message        : Average mflops/s per call per node       : 823398
-Grid       : Message        : Average mflops/s per call per node       : 664345
-Grid       : Message        : Average mflops/s per call per node (full): 313884
-Grid       : Message        : Average mflops/s per call per node (full): 301790
-Grid       : Message        : Average mflops/s per call per node (full): 443023
-Grid       : Message        : Average mflops/s per call per node (full): 304831
-Grid       : Message        :  Stencil 13.7732 GB/s per node
-Grid       : Message        :  Stencil 15.1391 GB/s per node
-Grid       : Message        :  Stencil 18.2708 GB/s per node
-Grid       : Message        :  Stencil 14.2929 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664574
-Grid       : Message        : Average mflops/s per call per node       : 800209
-Grid       : Message        : Average mflops/s per call per node       : 821873
-Grid       : Message        : Average mflops/s per call per node       : 659978
-Grid       : Message        : Average mflops/s per call per node (full): 315542
-Grid       : Message        : Average mflops/s per call per node (full): 417439
-Grid       : Message        : Average mflops/s per call per node (full): 444265
-Grid       : Message        : Average mflops/s per call per node (full): 306522
-Grid       : Message        :  Stencil 12.7582 GB/s per node
-Grid       : Message        :  Stencil 16.5874 GB/s per node
-Grid       : Message        :  Stencil 17.4572 GB/s per node
-Grid       : Message        :  Stencil 12.9153 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665604
-Grid       : Message        : Average mflops/s per call per node       : 803460
-Grid       : Message        : Average mflops/s per call per node       : 822525
-Grid       : Message        : Average mflops/s per call per node       : 664818
-Grid       : Message        : Average mflops/s per call per node (full): 313795
-Grid       : Message        : Average mflops/s per call per node (full): 435096
-Grid       : Message        : Average mflops/s per call per node (full): 444071
-Grid       : Message        : Average mflops/s per call per node (full): 305431
-Grid       : Message        :  Stencil 13.8559 GB/s per node
-Grid       : Message        :  Stencil 17.0497 GB/s per node
-Grid       : Message        :  Stencil 16.8828 GB/s per node
-Grid       : Message        :  Stencil 12.9922 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659253
-Grid       : Message        : Average mflops/s per call per node       : 805404
-Grid       : Message        : Average mflops/s per call per node       : 819715
-Grid       : Message        : Average mflops/s per call per node       : 664332
-Grid       : Message        : Average mflops/s per call per node (full): 313997
-Grid       : Message        : Average mflops/s per call per node (full): 435225
-Grid       : Message        : Average mflops/s per call per node (full): 439067
-Grid       : Message        : Average mflops/s per call per node (full): 305231
-Grid       : Message        :  Stencil 13.0007 GB/s per node
-Grid       : Message        :  Stencil 8.69655 GB/s per node
-Grid       : Message        :  Stencil 17.5308 GB/s per node
-Grid       : Message        :  Stencil 12.2284 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665359
-Grid       : Message        : Average mflops/s per call per node       : 813330
-Grid       : Message        : Average mflops/s per call per node       : 824392
-Grid       : Message        : Average mflops/s per call per node       : 666731
-Grid       : Message        : Average mflops/s per call per node (full): 314743
-Grid       : Message        : Average mflops/s per call per node (full): 284313
-Grid       : Message        : Average mflops/s per call per node (full): 444231
-Grid       : Message        : Average mflops/s per call per node (full): 303602
-Grid       : Message        :  Stencil 14.0179 GB/s per node
-Grid       : Message        :  Stencil 16.4361 GB/s per node
-Grid       : Message        :  Stencil 17.3788 GB/s per node
-Grid       : Message        :  Stencil 13.9105 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661910
-Grid       : Message        : Average mflops/s per call per node       : 801215
-Grid       : Message        : Average mflops/s per call per node       : 825305
-Grid       : Message        : Average mflops/s per call per node       : 662391
-Grid       : Message        : Average mflops/s per call per node (full): 315920
-Grid       : Message        : Average mflops/s per call per node (full): 432965
-Grid       : Message        : Average mflops/s per call per node (full): 440216
-Grid       : Message        : Average mflops/s per call per node (full): 306253
-Grid       : Message        :  Stencil 14.7332 GB/s per node
-Grid       : Message        :  Stencil 17.0097 GB/s per node
-Grid       : Message        :  Stencil 18.3556 GB/s per node
-Grid       : Message        :  Stencil 12.3901 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660267
-Grid       : Message        : Average mflops/s per call per node       : 804795
-Grid       : Message        : Average mflops/s per call per node       : 822348
-Grid       : Message        : Average mflops/s per call per node       : 665592
-Grid       : Message        : Average mflops/s per call per node (full): 315863
-Grid       : Message        : Average mflops/s per call per node (full): 439178
-Grid       : Message        : Average mflops/s per call per node (full): 448396
-Grid       : Message        : Average mflops/s per call per node (full): 303397
-Grid       : Message        :  Stencil 13.3785 GB/s per node
-Grid       : Message        :  Stencil 17.2514 GB/s per node
-Grid       : Message        :  Stencil 17.3377 GB/s per node
-Grid       : Message        :  Stencil 14.8653 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668976
-Grid       : Message        : Average mflops/s per call per node       : 804297
-Grid       : Message        : Average mflops/s per call per node       : 823445
-Grid       : Message        : Average mflops/s per call per node       : 659754
-Grid       : Message        : Average mflops/s per call per node (full): 316369
-Grid       : Message        : Average mflops/s per call per node (full): 439002
-Grid       : Message        : Average mflops/s per call per node (full): 443709
-Grid       : Message        : Average mflops/s per call per node (full): 306861
-Grid       : Message        :  Stencil 14.4473 GB/s per node
-Grid       : Message        :  Stencil 18.4221 GB/s per node
-Grid       : Message        :  Stencil 18.7098 GB/s per node
-Grid       : Message        :  Stencil 12.6884 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665908
-Grid       : Message        : Average mflops/s per call per node       : 797772
-Grid       : Message        : Average mflops/s per call per node       : 821779
-Grid       : Message        : Average mflops/s per call per node       : 665933
-Grid       : Message        : Average mflops/s per call per node (full): 316513
-Grid       : Message        : Average mflops/s per call per node (full): 439698
-Grid       : Message        : Average mflops/s per call per node (full): 446877
-Grid       : Message        : Average mflops/s per call per node (full): 304760
-Grid       : Message        :  Stencil 12.9669 GB/s per node
-Grid       : Message        :  Stencil 16.3339 GB/s per node
-Grid       : Message        :  Stencil 17.6991 GB/s per node
-Grid       : Message        :  Stencil 13.9045 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671505
-Grid       : Message        : Average mflops/s per call per node       : 807941
-Grid       : Message        : Average mflops/s per call per node       : 819920
-Grid       : Message        : Average mflops/s per call per node       : 661226
-Grid       : Message        : Average mflops/s per call per node (full): 315894
-Grid       : Message        : Average mflops/s per call per node (full): 433325
-Grid       : Message        : Average mflops/s per call per node (full): 445189
-Grid       : Message        : Average mflops/s per call per node (full): 305056
-Grid       : Message        :  Stencil 13.389 GB/s per node
-Grid       : Message        :  Stencil 18.4311 GB/s per node
-Grid       : Message        :  Stencil 17.3176 GB/s per node
-Grid       : Message        :  Stencil 12.428 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669226
-Grid       : Message        : Average mflops/s per call per node       : 802742
-Grid       : Message        : Average mflops/s per call per node       : 821110
-Grid       : Message        : Average mflops/s per call per node       : 667706
-Grid       : Message        : Average mflops/s per call per node (full): 316891
-Grid       : Message        : Average mflops/s per call per node (full): 441526
-Grid       : Message        : Average mflops/s per call per node (full): 442371
-Grid       : Message        : Average mflops/s per call per node (full): 303524
-Grid       : Message        :  Stencil 13.2963 GB/s per node
-Grid       : Message        :  Stencil 16.6181 GB/s per node
-Grid       : Message        :  Stencil 17.976 GB/s per node
-Grid       : Message        :  Stencil 12.7132 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668142
-Grid       : Message        : Average mflops/s per call per node       : 803065
-Grid       : Message        : Average mflops/s per call per node       : 826397
-Grid       : Message        : Average mflops/s per call per node       : 662717
-Grid       : Message        : Average mflops/s per call per node (full): 316414
-Grid       : Message        : Average mflops/s per call per node (full): 433704
-Grid       : Message        : Average mflops/s per call per node (full): 449195
-Grid       : Message        : Average mflops/s per call per node (full): 303450
-Grid       : Message        :  Stencil 13.407 GB/s per node
-Grid       : Message        :  Stencil 16.42 GB/s per node
-Grid       : Message        :  Stencil 18.5532 GB/s per node
-Grid       : Message        :  Stencil 12.0239 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667511
-Grid       : Message        : Average mflops/s per call per node       : 808754
-Grid       : Message        : Average mflops/s per call per node       : 817579
-Grid       : Message        : Average mflops/s per call per node       : 664542
-Grid       : Message        : Average mflops/s per call per node (full): 316072
-Grid       : Message        : Average mflops/s per call per node (full): 434448
-Grid       : Message        : Average mflops/s per call per node (full): 448228
-Grid       : Message        : Average mflops/s per call per node (full): 302366
-Grid       : Message        :  Stencil 12.6577 GB/s per node
-Grid       : Message        :  Stencil 16.6525 GB/s per node
-Grid       : Message        :  Stencil 18.0754 GB/s per node
-Grid       : Message        :  Stencil 13.328 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670269
-Grid       : Message        : Average mflops/s per call per node       : 806147
-Grid       : Message        : Average mflops/s per call per node       : 822395
-Grid       : Message        : Average mflops/s per call per node       : 663421
-Grid       : Message        : Average mflops/s per call per node (full): 315162
-Grid       : Message        : Average mflops/s per call per node (full): 434256
-Grid       : Message        : Average mflops/s per call per node (full): 443098
-Grid       : Message        : Average mflops/s per call per node (full): 306801
-Grid       : Message        :  Stencil 12.5144 GB/s per node
-Grid       : Message        :  Stencil 16.1335 GB/s per node
-Grid       : Message        :  Stencil 17.0158 GB/s per node
-Grid       : Message        :  Stencil 11.8636 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671085
-Grid       : Message        : Average mflops/s per call per node       : 800540
-Grid       : Message        : Average mflops/s per call per node       : 821557
-Grid       : Message        : Average mflops/s per call per node       : 668438
-Grid       : Message        : Average mflops/s per call per node (full): 314542
-Grid       : Message        : Average mflops/s per call per node (full): 429368
-Grid       : Message        : Average mflops/s per call per node (full): 441488
-Grid       : Message        : Average mflops/s per call per node (full): 300842
-Grid       : Message        :  Stencil 12.9445 GB/s per node
-Grid       : Message        :  Stencil 16.5649 GB/s per node
-Grid       : Message        :  Stencil 17.6354 GB/s per node
-Grid       : Message        :  Stencil 12.6334 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670258
-Grid       : Message        : Average mflops/s per call per node       : 807901
-Grid       : Message        : Average mflops/s per call per node       : 820599
-Grid       : Message        : Average mflops/s per call per node       : 664701
-Grid       : Message        : Average mflops/s per call per node (full): 316240
-Grid       : Message        : Average mflops/s per call per node (full): 436680
-Grid       : Message        : Average mflops/s per call per node (full): 444438
-Grid       : Message        : Average mflops/s per call per node (full): 304377
-Grid       : Message        :  Stencil 12.9229 GB/s per node
-Grid       : Message        :  Stencil 16.476 GB/s per node
-Grid       : Message        :  Stencil 17.3273 GB/s per node
-Grid       : Message        :  Stencil 12.1586 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671176
-Grid       : Message        : Average mflops/s per call per node       : 804588
-Grid       : Message        : Average mflops/s per call per node       : 824591
-Grid       : Message        : Average mflops/s per call per node       : 668922
-Grid       : Message        : Average mflops/s per call per node (full): 316452
-Grid       : Message        : Average mflops/s per call per node (full): 431674
-Grid       : Message        : Average mflops/s per call per node (full): 443769
-Grid       : Message        : Average mflops/s per call per node (full): 304071
-Grid       : Message        :  Stencil 13.7492 GB/s per node
-Grid       : Message        :  Stencil 16.0205 GB/s per node
-Grid       : Message        :  Stencil 17.7682 GB/s per node
-Grid       : Message        :  Stencil 12.5497 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667267
-Grid       : Message        : Average mflops/s per call per node       : 804527
-Grid       : Message        : Average mflops/s per call per node       : 822570
-Grid       : Message        : Average mflops/s per call per node       : 668442
-Grid       : Message        : Average mflops/s per call per node (full): 315015
-Grid       : Message        : Average mflops/s per call per node (full): 428099
-Grid       : Message        : Average mflops/s per call per node (full): 446695
-Grid       : Message        : Average mflops/s per call per node (full): 304877
-Grid       : Message        :  Stencil 16.0749 GB/s per node
-Grid       : Message        :  Stencil 16.7101 GB/s per node
-Grid       : Message        :  Stencil 17.5895 GB/s per node
-Grid       : Message        :  Stencil 12.6814 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664654
-Grid       : Message        : Average mflops/s per call per node       : 804560
-Grid       : Message        : Average mflops/s per call per node       : 820827
-Grid       : Message        : Average mflops/s per call per node       : 663960
-Grid       : Message        : Average mflops/s per call per node (full): 317931
-Grid       : Message        : Average mflops/s per call per node (full): 436764
-Grid       : Message        : Average mflops/s per call per node (full): 444919
-Grid       : Message        : Average mflops/s per call per node (full): 304502
-Grid       : Message        :  Stencil 13.8452 GB/s per node
-Grid       : Message        :  Stencil 16.5895 GB/s per node
-Grid       : Message        :  Stencil 17.6105 GB/s per node
-Grid       : Message        :  Stencil 12.7434 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662680
-Grid       : Message        : Average mflops/s per call per node       : 804209
-Grid       : Message        : Average mflops/s per call per node       : 826764
-Grid       : Message        : Average mflops/s per call per node       : 665513
-Grid       : Message        : Average mflops/s per call per node (full): 315795
-Grid       : Message        : Average mflops/s per call per node (full): 435470
-Grid       : Message        : Average mflops/s per call per node (full): 446375
-Grid       : Message        : Average mflops/s per call per node (full): 306667
-Grid       : Message        :  Stencil 12.696 GB/s per node
-Grid       : Message        :  Stencil 17.8462 GB/s per node
-Grid       : Message        :  Stencil 17.4581 GB/s per node
-Grid       : Message        :  Stencil 12.4621 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668704
-Grid       : Message        : Average mflops/s per call per node       : 801632
-Grid       : Message        : Average mflops/s per call per node       : 821837
-Grid       : Message        : Average mflops/s per call per node       : 665805
-Grid       : Message        : Average mflops/s per call per node (full): 314958
-Grid       : Message        : Average mflops/s per call per node (full): 439770
-Grid       : Message        : Average mflops/s per call per node (full): 444800
-Grid       : Message        : Average mflops/s per call per node (full): 304450
-Grid       : Message        :  Stencil 13.195 GB/s per node
-Grid       : Message        :  Stencil 16.3625 GB/s per node
-Grid       : Message        :  Stencil 17.9368 GB/s per node
-Grid       : Message        :  Stencil 12.7961 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664900
-Grid       : Message        : Average mflops/s per call per node       : 802314
-Grid       : Message        : Average mflops/s per call per node       : 822453
-Grid       : Message        : Average mflops/s per call per node       : 662885
-Grid       : Message        : Average mflops/s per call per node (full): 312141
-Grid       : Message        : Average mflops/s per call per node (full): 433682
-Grid       : Message        : Average mflops/s per call per node (full): 445677
-Grid       : Message        : Average mflops/s per call per node (full): 304522
-Grid       : Message        :  Stencil 13.1627 GB/s per node
-Grid       : Message        :  Stencil 16.5828 GB/s per node
-Grid       : Message        :  Stencil 18.8868 GB/s per node
-Grid       : Message        :  Stencil 12.1598 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667526
-Grid       : Message        : Average mflops/s per call per node       : 804126
-Grid       : Message        : Average mflops/s per call per node       : 826797
-Grid       : Message        : Average mflops/s per call per node       : 666150
-Grid       : Message        : Average mflops/s per call per node (full): 315161
-Grid       : Message        : Average mflops/s per call per node (full): 436237
-Grid       : Message        : Average mflops/s per call per node (full): 450383
-Grid       : Message        : Average mflops/s per call per node (full): 303367
-Grid       : Message        :  Stencil 14.3233 GB/s per node
-Grid       : Message        :  Stencil 17.0524 GB/s per node
-Grid       : Message        :  Stencil 17.7682 GB/s per node
-Grid       : Message        :  Stencil 13.6733 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662040
-Grid       : Message        : Average mflops/s per call per node       : 802978
-Grid       : Message        : Average mflops/s per call per node       : 827704
-Grid       : Message        : Average mflops/s per call per node       : 663840
-Grid       : Message        : Average mflops/s per call per node (full): 316403
-Grid       : Message        : Average mflops/s per call per node (full): 438560
-Grid       : Message        : Average mflops/s per call per node (full): 447686
-Grid       : Message        : Average mflops/s per call per node (full): 306479
-Grid       : Message        :  Stencil 14.0166 GB/s per node
-Grid       : Message        :  Stencil 17.7287 GB/s per node
-Grid       : Message        :  Stencil 17.3918 GB/s per node
-Grid       : Message        :  Stencil 13.5967 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662621
-Grid       : Message        : Average mflops/s per call per node       : 803027
-Grid       : Message        : Average mflops/s per call per node       : 828930
-Grid       : Message        : Average mflops/s per call per node       : 666495
-Grid       : Message        : Average mflops/s per call per node (full): 315597
-Grid       : Message        : Average mflops/s per call per node (full): 439267
-Grid       : Message        : Average mflops/s per call per node (full): 444967
-Grid       : Message        : Average mflops/s per call per node (full): 306897
-Grid       : Message        :  Stencil 12.7254 GB/s per node
-Grid       : Message        :  Stencil 16.7791 GB/s per node
-Grid       : Message        :  Stencil 17.8821 GB/s per node
-Grid       : Message        :  Stencil 12.1478 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667639
-Grid       : Message        : Average mflops/s per call per node       : 808509
-Grid       : Message        : Average mflops/s per call per node       : 822811
-Grid       : Message        : Average mflops/s per call per node       : 669762
-Grid       : Message        : Average mflops/s per call per node (full): 314173
-Grid       : Message        : Average mflops/s per call per node (full): 437440
-Grid       : Message        : Average mflops/s per call per node (full): 445486
-Grid       : Message        : Average mflops/s per call per node (full): 303879
-Grid       : Message        :  Stencil 12.7592 GB/s per node
-Grid       : Message        :  Stencil 17.3833 GB/s per node
-Grid       : Message        :  Stencil 18.0241 GB/s per node
-Grid       : Message        :  Stencil 11.543 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666434
-Grid       : Message        : Average mflops/s per call per node       : 800539
-Grid       : Message        : Average mflops/s per call per node       : 824422
-Grid       : Message        : Average mflops/s per call per node       : 664539
-Grid       : Message        : Average mflops/s per call per node (full): 314702
-Grid       : Message        : Average mflops/s per call per node (full): 436754
-Grid       : Message        : Average mflops/s per call per node (full): 438523
-Grid       : Message        : Average mflops/s per call per node (full): 295096
-Grid       : Message        :  Stencil 13.6429 GB/s per node
-Grid       : Message        :  Stencil 17.1495 GB/s per node
-Grid       : Message        :  Stencil 17.9739 GB/s per node
-Grid       : Message        :  Stencil 12.6514 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666162
-Grid       : Message        : Average mflops/s per call per node       : 803120
-Grid       : Message        : Average mflops/s per call per node       : 820330
-Grid       : Message        : Average mflops/s per call per node       : 665553
-Grid       : Message        : Average mflops/s per call per node (full): 315273
-Grid       : Message        : Average mflops/s per call per node (full): 436051
-Grid       : Message        : Average mflops/s per call per node (full): 445863
-Grid       : Message        : Average mflops/s per call per node (full): 305580
-Grid       : Message        :  Stencil 13.2379 GB/s per node
-Grid       : Message        :  Stencil 17.1006 GB/s per node
-Grid       : Message        :  Stencil 17.7767 GB/s per node
-Grid       : Message        :  Stencil 13.0215 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664785
-Grid       : Message        : Average mflops/s per call per node       : 800435
-Grid       : Message        : Average mflops/s per call per node       : 823378
-Grid       : Message        : Average mflops/s per call per node       : 668449
-Grid       : Message        : Average mflops/s per call per node (full): 314326
-Grid       : Message        : Average mflops/s per call per node (full): 438018
-Grid       : Message        : Average mflops/s per call per node (full): 446735
-Grid       : Message        : Average mflops/s per call per node (full): 305534
-Grid       : Message        :  Stencil 13.5944 GB/s per node
-Grid       : Message        :  Stencil 17.0677 GB/s per node
-Grid       : Message        :  Stencil 17.4291 GB/s per node
-Grid       : Message        :  Stencil 12.6877 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666724
-Grid       : Message        : Average mflops/s per call per node       : 802857
-Grid       : Message        : Average mflops/s per call per node       : 822555
-Grid       : Message        : Average mflops/s per call per node       : 666741
-Grid       : Message        : Average mflops/s per call per node (full): 315873
-Grid       : Message        : Average mflops/s per call per node (full): 438425
-Grid       : Message        : Average mflops/s per call per node (full): 445125
-Grid       : Message        : Average mflops/s per call per node (full): 305291
-Grid       : Message        :  Stencil 13.3329 GB/s per node
-Grid       : Message        :  Stencil 13.2059 GB/s per node
-Grid       : Message        :  Stencil 17.6162 GB/s per node
-Grid       : Message        :  Stencil 13.8388 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665180
-Grid       : Message        : Average mflops/s per call per node       : 810879
-Grid       : Message        : Average mflops/s per call per node       : 821220
-Grid       : Message        : Average mflops/s per call per node       : 658602
-Grid       : Message        : Average mflops/s per call per node (full): 315310
-Grid       : Message        : Average mflops/s per call per node (full): 385291
-Grid       : Message        : Average mflops/s per call per node (full): 445718
-Grid       : Message        : Average mflops/s per call per node (full): 306487
-Grid       : Message        :  Stencil 13.7408 GB/s per node
-Grid       : Message        :  Stencil 14.8181 GB/s per node
-Grid       : Message        :  Stencil 17.5145 GB/s per node
-Grid       : Message        :  Stencil 12.4369 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663746
-Grid       : Message        : Average mflops/s per call per node       : 802457
-Grid       : Message        : Average mflops/s per call per node       : 818522
-Grid       : Message        : Average mflops/s per call per node       : 663289
-Grid       : Message        : Average mflops/s per call per node (full): 315225
-Grid       : Message        : Average mflops/s per call per node (full): 412019
-Grid       : Message        : Average mflops/s per call per node (full): 445551
-Grid       : Message        : Average mflops/s per call per node (full): 303850
-Grid       : Message        :  Stencil 13.3968 GB/s per node
-Grid       : Message        :  Stencil 16.7506 GB/s per node
-Grid       : Message        :  Stencil 16.9505 GB/s per node
-Grid       : Message        :  Stencil 13.0396 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663215
-Grid       : Message        : Average mflops/s per call per node       : 803452
-Grid       : Message        : Average mflops/s per call per node       : 816860
-Grid       : Message        : Average mflops/s per call per node       : 662333
-Grid       : Message        : Average mflops/s per call per node (full): 314571
-Grid       : Message        : Average mflops/s per call per node (full): 436462
-Grid       : Message        : Average mflops/s per call per node (full): 438271
-Grid       : Message        : Average mflops/s per call per node (full): 304220
-Grid       : Message        :  Stencil 13.2516 GB/s per node
-Grid       : Message        :  Stencil 18.4683 GB/s per node
-Grid       : Message        :  Stencil 20.1626 GB/s per node
-Grid       : Message        :  Stencil 12.3178 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665033
-Grid       : Message        : Average mflops/s per call per node       : 805690
-Grid       : Message        : Average mflops/s per call per node       : 821338
-Grid       : Message        : Average mflops/s per call per node       : 667382
-Grid       : Message        : Average mflops/s per call per node (full): 313801
-Grid       : Message        : Average mflops/s per call per node (full): 442080
-Grid       : Message        : Average mflops/s per call per node (full): 451370
-Grid       : Message        : Average mflops/s per call per node (full): 304378
-Grid       : Message        :  Stencil 12.4761 GB/s per node
-Grid       : Message        :  Stencil 17.063 GB/s per node
-Grid       : Message        :  Stencil 17.2998 GB/s per node
-Grid       : Message        :  Stencil 13.0949 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669469
-Grid       : Message        : Average mflops/s per call per node       : 807207
-Grid       : Message        : Average mflops/s per call per node       : 818619
-Grid       : Message        : Average mflops/s per call per node       : 662204
-Grid       : Message        : Average mflops/s per call per node (full): 313152
-Grid       : Message        : Average mflops/s per call per node (full): 438548
-Grid       : Message        : Average mflops/s per call per node (full): 438583
-Grid       : Message        : Average mflops/s per call per node (full): 304705
-Grid       : Message        :  Stencil 13.9202 GB/s per node
-Grid       : Message        :  Stencil 16.7782 GB/s per node
-Grid       : Message        :  Stencil 18.2511 GB/s per node
-Grid       : Message        :  Stencil 12.3493 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665295
-Grid       : Message        : Average mflops/s per call per node       : 800954
-Grid       : Message        : Average mflops/s per call per node       : 820727
-Grid       : Message        : Average mflops/s per call per node       : 667125
-Grid       : Message        : Average mflops/s per call per node (full): 315384
-Grid       : Message        : Average mflops/s per call per node (full): 433980
-Grid       : Message        : Average mflops/s per call per node (full): 447066
-Grid       : Message        : Average mflops/s per call per node (full): 304322
-Grid       : Message        :  Stencil 12.2009 GB/s per node
-Grid       : Message        :  Stencil 17.8894 GB/s per node
-Grid       : Message        :  Stencil 17.733 GB/s per node
-Grid       : Message        :  Stencil 12.1817 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669787
-Grid       : Message        : Average mflops/s per call per node       : 799609
-Grid       : Message        : Average mflops/s per call per node       : 818548
-Grid       : Message        : Average mflops/s per call per node       : 667138
-Grid       : Message        : Average mflops/s per call per node (full): 311653
-Grid       : Message        : Average mflops/s per call per node (full): 438106
-Grid       : Message        : Average mflops/s per call per node (full): 444209
-Grid       : Message        : Average mflops/s per call per node (full): 304070
-Grid       : Message        :  Stencil 14.1701 GB/s per node
-Grid       : Message        :  Stencil 17.9506 GB/s per node
-Grid       : Message        :  Stencil 18.1714 GB/s per node
-Grid       : Message        :  Stencil 12.1593 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664334
-Grid       : Message        : Average mflops/s per call per node       : 804515
-Grid       : Message        : Average mflops/s per call per node       : 814830
-Grid       : Message        : Average mflops/s per call per node       : 662621
-Grid       : Message        : Average mflops/s per call per node (full): 316632
-Grid       : Message        : Average mflops/s per call per node (full): 441857
-Grid       : Message        : Average mflops/s per call per node (full): 444490
-Grid       : Message        : Average mflops/s per call per node (full): 302082
-Grid       : Message        :  Stencil 14.012 GB/s per node
-Grid       : Message        :  Stencil 16.5864 GB/s per node
-Grid       : Message        :  Stencil 17.2479 GB/s per node
-Grid       : Message        :  Stencil 12.3873 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666742
-Grid       : Message        : Average mflops/s per call per node       : 799540
-Grid       : Message        : Average mflops/s per call per node       : 815256
-Grid       : Message        : Average mflops/s per call per node       : 666099
-Grid       : Message        : Average mflops/s per call per node (full): 317252
-Grid       : Message        : Average mflops/s per call per node (full): 435178
-Grid       : Message        : Average mflops/s per call per node (full): 442294
-Grid       : Message        : Average mflops/s per call per node (full): 304068
-Grid       : Message        :  Stencil 13.1105 GB/s per node
-Grid       : Message        :  Stencil 17.0605 GB/s per node
-Grid       : Message        :  Stencil 17.4726 GB/s per node
-Grid       : Message        :  Stencil 13.9293 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665709
-Grid       : Message        : Average mflops/s per call per node       : 802462
-Grid       : Message        : Average mflops/s per call per node       : 821252
-Grid       : Message        : Average mflops/s per call per node       : 660691
-Grid       : Message        : Average mflops/s per call per node (full): 314770
-Grid       : Message        : Average mflops/s per call per node (full): 438294
-Grid       : Message        : Average mflops/s per call per node (full): 445239
-Grid       : Message        : Average mflops/s per call per node (full): 304619
-Grid       : Message        :  Stencil 12.7894 GB/s per node
-Grid       : Message        :  Stencil 13.614 GB/s per node
-Grid       : Message        :  Stencil 18.0553 GB/s per node
-Grid       : Message        :  Stencil 14.1673 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670212
-Grid       : Message        : Average mflops/s per call per node       : 810108
-Grid       : Message        : Average mflops/s per call per node       : 822900
-Grid       : Message        : Average mflops/s per call per node       : 663050
-Grid       : Message        : Average mflops/s per call per node (full): 315724
-Grid       : Message        : Average mflops/s per call per node (full): 392709
-Grid       : Message        : Average mflops/s per call per node (full): 447040
-Grid       : Message        : Average mflops/s per call per node (full): 307350
-Grid       : Message        :  Stencil 12.6603 GB/s per node
-Grid       : Message        :  Stencil 17.4165 GB/s per node
-Grid       : Message        :  Stencil 18.1449 GB/s per node
-Grid       : Message        :  Stencil 11.9735 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668180
-Grid       : Message        : Average mflops/s per call per node       : 803492
-Grid       : Message        : Average mflops/s per call per node       : 821395
-Grid       : Message        : Average mflops/s per call per node       : 666526
-Grid       : Message        : Average mflops/s per call per node (full): 314481
-Grid       : Message        : Average mflops/s per call per node (full): 438394
-Grid       : Message        : Average mflops/s per call per node (full): 446746
-Grid       : Message        : Average mflops/s per call per node (full): 301260
-Grid       : Message        :  Stencil 14.5401 GB/s per node
-Grid       : Message        :  Stencil 18.8236 GB/s per node
-Grid       : Message        :  Stencil 17.4676 GB/s per node
-Grid       : Message        :  Stencil 12.6769 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662498
-Grid       : Message        : Average mflops/s per call per node       : 806168
-Grid       : Message        : Average mflops/s per call per node       : 820787
-Grid       : Message        : Average mflops/s per call per node       : 667998
-Grid       : Message        : Average mflops/s per call per node (full): 316850
-Grid       : Message        : Average mflops/s per call per node (full): 442363
-Grid       : Message        : Average mflops/s per call per node (full): 444850
-Grid       : Message        : Average mflops/s per call per node (full): 304213
-Grid       : Message        :  Stencil 14.4069 GB/s per node
-Grid       : Message        :  Stencil 17.4845 GB/s per node
-Grid       : Message        :  Stencil 18.0741 GB/s per node
-Grid       : Message        :  Stencil 11.8651 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667022
-Grid       : Message        : Average mflops/s per call per node       : 806637
-Grid       : Message        : Average mflops/s per call per node       : 822004
-Grid       : Message        : Average mflops/s per call per node       : 669972
-Grid       : Message        : Average mflops/s per call per node (full): 317121
-Grid       : Message        : Average mflops/s per call per node (full): 439636
-Grid       : Message        : Average mflops/s per call per node (full): 439085
-Grid       : Message        : Average mflops/s per call per node (full): 300811
-Grid       : Message        :  Stencil 13.3774 GB/s per node
-Grid       : Message        :  Stencil 11.3078 GB/s per node
-Grid       : Message        :  Stencil 17.8454 GB/s per node
-Grid       : Message        :  Stencil 12.793 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668798
-Grid       : Message        : Average mflops/s per call per node       : 810858
-Grid       : Message        : Average mflops/s per call per node       : 818351
-Grid       : Message        : Average mflops/s per call per node       : 666968
-Grid       : Message        : Average mflops/s per call per node (full): 316429
-Grid       : Message        : Average mflops/s per call per node (full): 346497
-Grid       : Message        : Average mflops/s per call per node (full): 445644
-Grid       : Message        : Average mflops/s per call per node (full): 305367
-Grid       : Message        :  Stencil 13.1528 GB/s per node
-Grid       : Message        :  Stencil 17.1553 GB/s per node
-Grid       : Message        :  Stencil 18.4991 GB/s per node
-Grid       : Message        :  Stencil 12.9203 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666186
-Grid       : Message        : Average mflops/s per call per node       : 807320
-Grid       : Message        : Average mflops/s per call per node       : 825001
-Grid       : Message        : Average mflops/s per call per node       : 663941
-Grid       : Message        : Average mflops/s per call per node (full): 315095
-Grid       : Message        : Average mflops/s per call per node (full): 436414
-Grid       : Message        : Average mflops/s per call per node (full): 446638
-Grid       : Message        : Average mflops/s per call per node (full): 305184
-Grid       : Message        :  Stencil 14.3512 GB/s per node
-Grid       : Message        :  Stencil 8.08582 GB/s per node
-Grid       : Message        :  Stencil 17.2098 GB/s per node
-Grid       : Message        :  Stencil 12.9761 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664921
-Grid       : Message        : Average mflops/s per call per node       : 811951
-Grid       : Message        : Average mflops/s per call per node       : 820638
-Grid       : Message        : Average mflops/s per call per node       : 662247
-Grid       : Message        : Average mflops/s per call per node (full): 317454
-Grid       : Message        : Average mflops/s per call per node (full): 269204
-Grid       : Message        : Average mflops/s per call per node (full): 442623
-Grid       : Message        : Average mflops/s per call per node (full): 302247
-Grid       : Message        :  Stencil 13.6523 GB/s per node
-Grid       : Message        :  Stencil 12.1863 GB/s per node
-Grid       : Message        :  Stencil 18.2226 GB/s per node
-Grid       : Message        :  Stencil 14.5407 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668120
-Grid       : Message        : Average mflops/s per call per node       : 810323
-Grid       : Message        : Average mflops/s per call per node       : 829405
-Grid       : Message        : Average mflops/s per call per node       : 660045
-Grid       : Message        : Average mflops/s per call per node (full): 316597
-Grid       : Message        : Average mflops/s per call per node (full): 365500
-Grid       : Message        : Average mflops/s per call per node (full): 449915
-Grid       : Message        : Average mflops/s per call per node (full): 306907
-Grid       : Message        :  Stencil 13.3278 GB/s per node
-Grid       : Message        :  Stencil 16.5323 GB/s per node
-Grid       : Message        :  Stencil 17.9966 GB/s per node
-Grid       : Message        :  Stencil 12.5225 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668482
-Grid       : Message        : Average mflops/s per call per node       : 806322
-Grid       : Message        : Average mflops/s per call per node       : 817287
-Grid       : Message        : Average mflops/s per call per node       : 667975
-Grid       : Message        : Average mflops/s per call per node (full): 315330
-Grid       : Message        : Average mflops/s per call per node (full): 435216
-Grid       : Message        : Average mflops/s per call per node (full): 444432
-Grid       : Message        : Average mflops/s per call per node (full): 304925
-Grid       : Message        :  Stencil 12.8598 GB/s per node
-Grid       : Message        :  Stencil 17.2892 GB/s per node
-Grid       : Message        :  Stencil 17.7201 GB/s per node
-Grid       : Message        :  Stencil 12.3424 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670194
-Grid       : Message        : Average mflops/s per call per node       : 805940
-Grid       : Message        : Average mflops/s per call per node       : 816860
-Grid       : Message        : Average mflops/s per call per node       : 670715
-Grid       : Message        : Average mflops/s per call per node (full): 316157
-Grid       : Message        : Average mflops/s per call per node (full): 440925
-Grid       : Message        : Average mflops/s per call per node (full): 445275
-Grid       : Message        : Average mflops/s per call per node (full): 304670
-Grid       : Message        :  Stencil 12.7423 GB/s per node
-Grid       : Message        :  Stencil 17.3792 GB/s per node
-Grid       : Message        :  Stencil 17.3747 GB/s per node
-Grid       : Message        :  Stencil 12.4431 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670648
-Grid       : Message        : Average mflops/s per call per node       : 801905
-Grid       : Message        : Average mflops/s per call per node       : 824759
-Grid       : Message        : Average mflops/s per call per node       : 664315
-Grid       : Message        : Average mflops/s per call per node (full): 315426
-Grid       : Message        : Average mflops/s per call per node (full): 435399
-Grid       : Message        : Average mflops/s per call per node (full): 444680
-Grid       : Message        : Average mflops/s per call per node (full): 303482
-Grid       : Message        :  Stencil 12.3573 GB/s per node
-Grid       : Message        :  Stencil 17.5181 GB/s per node
-Grid       : Message        :  Stencil 17.2293 GB/s per node
-Grid       : Message        :  Stencil 12.7231 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 670000
-Grid       : Message        : Average mflops/s per call per node       : 800657
-Grid       : Message        : Average mflops/s per call per node       : 822182
-Grid       : Message        : Average mflops/s per call per node       : 661835
-Grid       : Message        : Average mflops/s per call per node (full): 313166
-Grid       : Message        : Average mflops/s per call per node (full): 439074
-Grid       : Message        : Average mflops/s per call per node (full): 442154
-Grid       : Message        : Average mflops/s per call per node (full): 301674
-Grid       : Message        :  Stencil 13.6493 GB/s per node
-Grid       : Message        :  Stencil 10.9089 GB/s per node
-Grid       : Message        :  Stencil 16.1128 GB/s per node
-Grid       : Message        :  Stencil 12.0041 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668159
-Grid       : Message        : Average mflops/s per call per node       : 814123
-Grid       : Message        : Average mflops/s per call per node       : 826384
-Grid       : Message        : Average mflops/s per call per node       : 668706
-Grid       : Message        : Average mflops/s per call per node (full): 315568
-Grid       : Message        : Average mflops/s per call per node (full): 337505
-Grid       : Message        : Average mflops/s per call per node (full): 417961
-Grid       : Message        : Average mflops/s per call per node (full): 302061
-Grid       : Message        :  Stencil 13.8407 GB/s per node
-Grid       : Message        :  Stencil 12.1634 GB/s per node
-Grid       : Message        :  Stencil 17.1616 GB/s per node
-Grid       : Message        :  Stencil 13.0852 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666780
-Grid       : Message        : Average mflops/s per call per node       : 804701
-Grid       : Message        : Average mflops/s per call per node       : 820652
-Grid       : Message        : Average mflops/s per call per node       : 663281
-Grid       : Message        : Average mflops/s per call per node (full): 317003
-Grid       : Message        : Average mflops/s per call per node (full): 364820
-Grid       : Message        : Average mflops/s per call per node (full): 443175
-Grid       : Message        : Average mflops/s per call per node (full): 305023
-Grid       : Message        :  Stencil 14.5455 GB/s per node
-Grid       : Message        :  Stencil 16.8268 GB/s per node
-Grid       : Message        :  Stencil 17.682 GB/s per node
-Grid       : Message        :  Stencil 12.4407 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665431
-Grid       : Message        : Average mflops/s per call per node       : 805665
-Grid       : Message        : Average mflops/s per call per node       : 828909
-Grid       : Message        : Average mflops/s per call per node       : 668099
-Grid       : Message        : Average mflops/s per call per node (full): 317273
-Grid       : Message        : Average mflops/s per call per node (full): 437080
-Grid       : Message        : Average mflops/s per call per node (full): 446707
-Grid       : Message        : Average mflops/s per call per node (full): 305324
-Grid       : Message        :  Stencil 12.7316 GB/s per node
-Grid       : Message        :  Stencil 17.028 GB/s per node
-Grid       : Message        :  Stencil 17.9791 GB/s per node
-Grid       : Message        :  Stencil 13.2229 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668528
-Grid       : Message        : Average mflops/s per call per node       : 803356
-Grid       : Message        : Average mflops/s per call per node       : 820340
-Grid       : Message        : Average mflops/s per call per node       : 659327
-Grid       : Message        : Average mflops/s per call per node (full): 314400
-Grid       : Message        : Average mflops/s per call per node (full): 439496
-Grid       : Message        : Average mflops/s per call per node (full): 445632
-Grid       : Message        : Average mflops/s per call per node (full): 305293
-Grid       : Message        :  Stencil 12.5028 GB/s per node
-Grid       : Message        :  Stencil 15.9845 GB/s per node
-Grid       : Message        :  Stencil 18.9131 GB/s per node
-Grid       : Message        :  Stencil 12.9499 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668710
-Grid       : Message        : Average mflops/s per call per node       : 807360
-Grid       : Message        : Average mflops/s per call per node       : 821502
-Grid       : Message        : Average mflops/s per call per node       : 668162
-Grid       : Message        : Average mflops/s per call per node (full): 313671
-Grid       : Message        : Average mflops/s per call per node (full): 426907
-Grid       : Message        : Average mflops/s per call per node (full): 449433
-Grid       : Message        : Average mflops/s per call per node (full): 306237
-Grid       : Message        :  Stencil 12.4208 GB/s per node
-Grid       : Message        :  Stencil 17.8274 GB/s per node
-Grid       : Message        :  Stencil 18.3711 GB/s per node
-Grid       : Message        :  Stencil 12.2259 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666707
-Grid       : Message        : Average mflops/s per call per node       : 800849
-Grid       : Message        : Average mflops/s per call per node       : 818819
-Grid       : Message        : Average mflops/s per call per node       : 668127
-Grid       : Message        : Average mflops/s per call per node (full): 312720
-Grid       : Message        : Average mflops/s per call per node (full): 439674
-Grid       : Message        : Average mflops/s per call per node (full): 446184
-Grid       : Message        : Average mflops/s per call per node (full): 303664
-Grid       : Message        :  Stencil 12.5386 GB/s per node
-Grid       : Message        :  Stencil 16.7961 GB/s per node
-Grid       : Message        :  Stencil 17.123 GB/s per node
-Grid       : Message        :  Stencil 12.187 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669033
-Grid       : Message        : Average mflops/s per call per node       : 806477
-Grid       : Message        : Average mflops/s per call per node       : 824230
-Grid       : Message        : Average mflops/s per call per node       : 665084
-Grid       : Message        : Average mflops/s per call per node (full): 313818
-Grid       : Message        : Average mflops/s per call per node (full): 437820
-Grid       : Message        : Average mflops/s per call per node (full): 442023
-Grid       : Message        : Average mflops/s per call per node (full): 303308
-Grid       : Message        :  Stencil 13.9051 GB/s per node
-Grid       : Message        :  Stencil 14.1793 GB/s per node
-Grid       : Message        :  Stencil 17.3172 GB/s per node
-Grid       : Message        :  Stencil 13.5939 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662731
-Grid       : Message        : Average mflops/s per call per node       : 806139
-Grid       : Message        : Average mflops/s per call per node       : 826644
-Grid       : Message        : Average mflops/s per call per node       : 662587
-Grid       : Message        : Average mflops/s per call per node (full): 316606
-Grid       : Message        : Average mflops/s per call per node (full): 403966
-Grid       : Message        : Average mflops/s per call per node (full): 444894
-Grid       : Message        : Average mflops/s per call per node (full): 306687
-Grid       : Message        :  Stencil 12.713 GB/s per node
-Grid       : Message        :  Stencil 17.2275 GB/s per node
-Grid       : Message        :  Stencil 17.3628 GB/s per node
-Grid       : Message        :  Stencil 12.5522 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667285
-Grid       : Message        : Average mflops/s per call per node       : 799222
-Grid       : Message        : Average mflops/s per call per node       : 825492
-Grid       : Message        : Average mflops/s per call per node       : 669083
-Grid       : Message        : Average mflops/s per call per node (full): 313531
-Grid       : Message        : Average mflops/s per call per node (full): 437657
-Grid       : Message        : Average mflops/s per call per node (full): 444462
-Grid       : Message        : Average mflops/s per call per node (full): 305101
-Grid       : Message        :  Stencil 13.1261 GB/s per node
-Grid       : Message        :  Stencil 17.3873 GB/s per node
-Grid       : Message        :  Stencil 18.1324 GB/s per node
-Grid       : Message        :  Stencil 13.4224 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662768
-Grid       : Message        : Average mflops/s per call per node       : 805359
-Grid       : Message        : Average mflops/s per call per node       : 818840
-Grid       : Message        : Average mflops/s per call per node       : 664498
-Grid       : Message        : Average mflops/s per call per node (full): 314186
-Grid       : Message        : Average mflops/s per call per node (full): 437617
-Grid       : Message        : Average mflops/s per call per node (full): 446608
-Grid       : Message        : Average mflops/s per call per node (full): 306661
-Grid       : Message        :  Stencil 13.5709 GB/s per node
-Grid       : Message        :  Stencil 16.8654 GB/s per node
-Grid       : Message        :  Stencil 17.8845 GB/s per node
-Grid       : Message        :  Stencil 13.5503 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659042
-Grid       : Message        : Average mflops/s per call per node       : 808434
-Grid       : Message        : Average mflops/s per call per node       : 819807
-Grid       : Message        : Average mflops/s per call per node       : 661740
-Grid       : Message        : Average mflops/s per call per node (full): 314042
-Grid       : Message        : Average mflops/s per call per node (full): 439197
-Grid       : Message        : Average mflops/s per call per node (full): 445970
-Grid       : Message        : Average mflops/s per call per node (full): 305878
-Grid       : Message        :  Stencil 14.4503 GB/s per node
-Grid       : Message        :  Stencil 16.376 GB/s per node
-Grid       : Message        :  Stencil 17.6563 GB/s per node
-Grid       : Message        :  Stencil 12.3184 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661059
-Grid       : Message        : Average mflops/s per call per node       : 810627
-Grid       : Message        : Average mflops/s per call per node       : 824265
-Grid       : Message        : Average mflops/s per call per node       : 666180
-Grid       : Message        : Average mflops/s per call per node (full): 316759
-Grid       : Message        : Average mflops/s per call per node (full): 433890
-Grid       : Message        : Average mflops/s per call per node (full): 446667
-Grid       : Message        : Average mflops/s per call per node (full): 303785
-Grid       : Message        :  Stencil 12.8266 GB/s per node
-Grid       : Message        :  Stencil 16.7177 GB/s per node
-Grid       : Message        :  Stencil 17.8434 GB/s per node
-Grid       : Message        :  Stencil 12.9931 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666756
-Grid       : Message        : Average mflops/s per call per node       : 801430
-Grid       : Message        : Average mflops/s per call per node       : 823620
-Grid       : Message        : Average mflops/s per call per node       : 664019
-Grid       : Message        : Average mflops/s per call per node (full): 314420
-Grid       : Message        : Average mflops/s per call per node (full): 436518
-Grid       : Message        : Average mflops/s per call per node (full): 445382
-Grid       : Message        : Average mflops/s per call per node (full): 305582
-Grid       : Message        :  Stencil 14.6546 GB/s per node
-Grid       : Message        :  Stencil 17.191 GB/s per node
-Grid       : Message        :  Stencil 17.9299 GB/s per node
-Grid       : Message        :  Stencil 12.0428 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664627
-Grid       : Message        : Average mflops/s per call per node       : 796301
-Grid       : Message        : Average mflops/s per call per node       : 822949
-Grid       : Message        : Average mflops/s per call per node       : 668229
-Grid       : Message        : Average mflops/s per call per node (full): 316538
-Grid       : Message        : Average mflops/s per call per node (full): 435682
-Grid       : Message        : Average mflops/s per call per node (full): 436898
-Grid       : Message        : Average mflops/s per call per node (full): 301776
-Grid       : Message        :  Stencil 13.7608 GB/s per node
-Grid       : Message        :  Stencil 12.5123 GB/s per node
-Grid       : Message        :  Stencil 17.7201 GB/s per node
-Grid       : Message        :  Stencil 12.7085 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666215
-Grid       : Message        : Average mflops/s per call per node       : 810734
-Grid       : Message        : Average mflops/s per call per node       : 820886
-Grid       : Message        : Average mflops/s per call per node       : 664598
-Grid       : Message        : Average mflops/s per call per node (full): 315574
-Grid       : Message        : Average mflops/s per call per node (full): 371951
-Grid       : Message        : Average mflops/s per call per node (full): 445993
-Grid       : Message        : Average mflops/s per call per node (full): 304257
-Grid       : Message        :  Stencil 13.5357 GB/s per node
-Grid       : Message        :  Stencil 17.5152 GB/s per node
-Grid       : Message        :  Stencil 17.4573 GB/s per node
-Grid       : Message        :  Stencil 13.2572 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663986
-Grid       : Message        : Average mflops/s per call per node       : 803379
-Grid       : Message        : Average mflops/s per call per node       : 820745
-Grid       : Message        : Average mflops/s per call per node       : 664000
-Grid       : Message        : Average mflops/s per call per node (full): 315566
-Grid       : Message        : Average mflops/s per call per node (full): 438003
-Grid       : Message        : Average mflops/s per call per node (full): 440111
-Grid       : Message        : Average mflops/s per call per node (full): 306596
-Grid       : Message        :  Stencil 14.4558 GB/s per node
-Grid       : Message        :  Stencil 10.5721 GB/s per node
-Grid       : Message        :  Stencil 17.7932 GB/s per node
-Grid       : Message        :  Stencil 14.146 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665281
-Grid       : Message        : Average mflops/s per call per node       : 810654
-Grid       : Message        : Average mflops/s per call per node       : 818240
-Grid       : Message        : Average mflops/s per call per node       : 661650
-Grid       : Message        : Average mflops/s per call per node (full): 317179
-Grid       : Message        : Average mflops/s per call per node (full): 329822
-Grid       : Message        : Average mflops/s per call per node (full): 446330
-Grid       : Message        : Average mflops/s per call per node (full): 306634
-Grid       : Message        :  Stencil 13.8995 GB/s per node
-Grid       : Message        :  Stencil 17.8563 GB/s per node
-Grid       : Message        :  Stencil 17.7542 GB/s per node
-Grid       : Message        :  Stencil 13.4172 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664829
-Grid       : Message        : Average mflops/s per call per node       : 802048
-Grid       : Message        : Average mflops/s per call per node       : 819825
-Grid       : Message        : Average mflops/s per call per node       : 657813
-Grid       : Message        : Average mflops/s per call per node (full): 316146
-Grid       : Message        : Average mflops/s per call per node (full): 439566
-Grid       : Message        : Average mflops/s per call per node (full): 443082
-Grid       : Message        : Average mflops/s per call per node (full): 304548
-Grid       : Message        :  Stencil 13.312 GB/s per node
-Grid       : Message        :  Stencil 16.6707 GB/s per node
-Grid       : Message        :  Stencil 18.6503 GB/s per node
-Grid       : Message        :  Stencil 13.0266 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667661
-Grid       : Message        : Average mflops/s per call per node       : 800789
-Grid       : Message        : Average mflops/s per call per node       : 823530
-Grid       : Message        : Average mflops/s per call per node       : 662616
-Grid       : Message        : Average mflops/s per call per node (full): 315744
-Grid       : Message        : Average mflops/s per call per node (full): 435038
-Grid       : Message        : Average mflops/s per call per node (full): 447909
-Grid       : Message        : Average mflops/s per call per node (full): 303186
-Grid       : Message        :  Stencil 12.6566 GB/s per node
-Grid       : Message        :  Stencil 17.8213 GB/s per node
-Grid       : Message        :  Stencil 17.9463 GB/s per node
-Grid       : Message        :  Stencil 13.4799 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666857
-Grid       : Message        : Average mflops/s per call per node       : 803446
-Grid       : Message        : Average mflops/s per call per node       : 814357
-Grid       : Message        : Average mflops/s per call per node       : 658835
-Grid       : Message        : Average mflops/s per call per node (full): 314154
-Grid       : Message        : Average mflops/s per call per node (full): 440308
-Grid       : Message        : Average mflops/s per call per node (full): 444496
-Grid       : Message        : Average mflops/s per call per node (full): 305207
-Grid       : Message        :  Stencil 12.8388 GB/s per node
-Grid       : Message        :  Stencil 17.0231 GB/s per node
-Grid       : Message        :  Stencil 18.4924 GB/s per node
-Grid       : Message        :  Stencil 12.3361 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669083
-Grid       : Message        : Average mflops/s per call per node       : 804015
-Grid       : Message        : Average mflops/s per call per node       : 821390
-Grid       : Message        : Average mflops/s per call per node       : 667814
-Grid       : Message        : Average mflops/s per call per node (full): 315517
-Grid       : Message        : Average mflops/s per call per node (full): 434609
-Grid       : Message        : Average mflops/s per call per node (full): 447579
-Grid       : Message        : Average mflops/s per call per node (full): 303505
-Grid       : Message        :  Stencil 13.7879 GB/s per node
-Grid       : Message        :  Stencil 16.5433 GB/s per node
-Grid       : Message        :  Stencil 17.908 GB/s per node
-Grid       : Message        :  Stencil 13.8103 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662509
-Grid       : Message        : Average mflops/s per call per node       : 806916
-Grid       : Message        : Average mflops/s per call per node       : 818770
-Grid       : Message        : Average mflops/s per call per node       : 662643
-Grid       : Message        : Average mflops/s per call per node (full): 315036
-Grid       : Message        : Average mflops/s per call per node (full): 435654
-Grid       : Message        : Average mflops/s per call per node (full): 445632
-Grid       : Message        : Average mflops/s per call per node (full): 306895
-Grid       : Message        :  Stencil 13.4876 GB/s per node
-Grid       : Message        :  Stencil 16.8819 GB/s per node
-Grid       : Message        :  Stencil 17.6484 GB/s per node
-Grid       : Message        :  Stencil 12.5259 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663824
-Grid       : Message        : Average mflops/s per call per node       : 808188
-Grid       : Message        : Average mflops/s per call per node       : 823638
-Grid       : Message        : Average mflops/s per call per node       : 662030
-Grid       : Message        : Average mflops/s per call per node (full): 314759
-Grid       : Message        : Average mflops/s per call per node (full): 438549
-Grid       : Message        : Average mflops/s per call per node (full): 445451
-Grid       : Message        : Average mflops/s per call per node (full): 303669
-Grid       : Message        :  Stencil 12.3675 GB/s per node
-Grid       : Message        :  Stencil 16.4276 GB/s per node
-Grid       : Message        :  Stencil 17.5417 GB/s per node
-Grid       : Message        :  Stencil 12.1676 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 669071
-Grid       : Message        : Average mflops/s per call per node       : 806133
-Grid       : Message        : Average mflops/s per call per node       : 820619
-Grid       : Message        : Average mflops/s per call per node       : 667351
-Grid       : Message        : Average mflops/s per call per node (full): 312679
-Grid       : Message        : Average mflops/s per call per node (full): 434983
-Grid       : Message        : Average mflops/s per call per node (full): 444984
-Grid       : Message        : Average mflops/s per call per node (full): 303637
-Grid       : Message        :  Stencil 13.0709 GB/s per node
-Grid       : Message        :  Stencil 17.5349 GB/s per node
-Grid       : Message        :  Stencil 17.7934 GB/s per node
-Grid       : Message        :  Stencil 12.1751 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668344
-Grid       : Message        : Average mflops/s per call per node       : 807623
-Grid       : Message        : Average mflops/s per call per node       : 824665
-Grid       : Message        : Average mflops/s per call per node       : 666928
-Grid       : Message        : Average mflops/s per call per node (full): 315271
-Grid       : Message        : Average mflops/s per call per node (full): 438487
-Grid       : Message        : Average mflops/s per call per node (full): 445571
-Grid       : Message        : Average mflops/s per call per node (full): 303691
-Grid       : Message        :  Stencil 12.9941 GB/s per node
-Grid       : Message        :  Stencil 14.9089 GB/s per node
-Grid       : Message        :  Stencil 17.9194 GB/s per node
-Grid       : Message        :  Stencil 12.0044 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664444
-Grid       : Message        : Average mflops/s per call per node       : 808597
-Grid       : Message        : Average mflops/s per call per node       : 823230
-Grid       : Message        : Average mflops/s per call per node       : 665374
-Grid       : Message        : Average mflops/s per call per node (full): 314230
-Grid       : Message        : Average mflops/s per call per node (full): 414717
-Grid       : Message        : Average mflops/s per call per node (full): 447203
-Grid       : Message        : Average mflops/s per call per node (full): 302432
-Grid       : Message        :  Stencil 12.563 GB/s per node
-Grid       : Message        :  Stencil 17.4419 GB/s per node
-Grid       : Message        :  Stencil 16.3727 GB/s per node
-Grid       : Message        :  Stencil 12.8744 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667188
-Grid       : Message        : Average mflops/s per call per node       : 803660
-Grid       : Message        : Average mflops/s per call per node       : 828617
-Grid       : Message        : Average mflops/s per call per node       : 665791
-Grid       : Message        : Average mflops/s per call per node (full): 313854
-Grid       : Message        : Average mflops/s per call per node (full): 439680
-Grid       : Message        : Average mflops/s per call per node (full): 432089
-Grid       : Message        : Average mflops/s per call per node (full): 305079
-Grid       : Message        :  Stencil 13.1201 GB/s per node
-Grid       : Message        :  Stencil 16.4825 GB/s per node
-Grid       : Message        :  Stencil 17.2141 GB/s per node
-Grid       : Message        :  Stencil 12.8028 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662617
-Grid       : Message        : Average mflops/s per call per node       : 805707
-Grid       : Message        : Average mflops/s per call per node       : 819975
-Grid       : Message        : Average mflops/s per call per node       : 663049
-Grid       : Message        : Average mflops/s per call per node (full): 314151
-Grid       : Message        : Average mflops/s per call per node (full): 432487
-Grid       : Message        : Average mflops/s per call per node (full): 441567
-Grid       : Message        : Average mflops/s per call per node (full): 300225
-Grid       : Message        :  Stencil 13.494 GB/s per node
-Grid       : Message        :  Stencil 17.4044 GB/s per node
-Grid       : Message        :  Stencil 17.391 GB/s per node
-Grid       : Message        :  Stencil 13.3663 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 659710
-Grid       : Message        : Average mflops/s per call per node       : 804423
-Grid       : Message        : Average mflops/s per call per node       : 819191
-Grid       : Message        : Average mflops/s per call per node       : 659182
-Grid       : Message        : Average mflops/s per call per node (full): 313886
-Grid       : Message        : Average mflops/s per call per node (full): 439049
-Grid       : Message        : Average mflops/s per call per node (full): 443820
-Grid       : Message        : Average mflops/s per call per node (full): 304886
-Grid       : Message        :  Stencil 13.3257 GB/s per node
-Grid       : Message        :  Stencil 16.5307 GB/s per node
-Grid       : Message        :  Stencil 17.5747 GB/s per node
-Grid       : Message        :  Stencil 14.698 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660623
-Grid       : Message        : Average mflops/s per call per node       : 807067
-Grid       : Message        : Average mflops/s per call per node       : 818778
-Grid       : Message        : Average mflops/s per call per node       : 662805
-Grid       : Message        : Average mflops/s per call per node (full): 314423
-Grid       : Message        : Average mflops/s per call per node (full): 432619
-Grid       : Message        : Average mflops/s per call per node (full): 444622
-Grid       : Message        : Average mflops/s per call per node (full): 306808
-Grid       : Message        :  Stencil 13.282 GB/s per node
-Grid       : Message        :  Stencil 16.7499 GB/s per node
-Grid       : Message        :  Stencil 18.5887 GB/s per node
-Grid       : Message        :  Stencil 13.7764 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663404
-Grid       : Message        : Average mflops/s per call per node       : 804737
-Grid       : Message        : Average mflops/s per call per node       : 820664
-Grid       : Message        : Average mflops/s per call per node       : 666191
-Grid       : Message        : Average mflops/s per call per node (full): 314406
-Grid       : Message        : Average mflops/s per call per node (full): 435830
-Grid       : Message        : Average mflops/s per call per node (full): 448424
-Grid       : Message        : Average mflops/s per call per node (full): 306801
-Grid       : Message        :  Stencil 13.2265 GB/s per node
-Grid       : Message        :  Stencil 17.4913 GB/s per node
-Grid       : Message        :  Stencil 17.7804 GB/s per node
-Grid       : Message        :  Stencil 12.8921 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661827
-Grid       : Message        : Average mflops/s per call per node       : 804946
-Grid       : Message        : Average mflops/s per call per node       : 823488
-Grid       : Message        : Average mflops/s per call per node       : 667613
-Grid       : Message        : Average mflops/s per call per node (full): 314051
-Grid       : Message        : Average mflops/s per call per node (full): 438066
-Grid       : Message        : Average mflops/s per call per node (full): 446464
-Grid       : Message        : Average mflops/s per call per node (full): 305594
-Grid       : Message        :  Stencil 13.2206 GB/s per node
-Grid       : Message        :  Stencil 17.0128 GB/s per node
-Grid       : Message        :  Stencil 18.401 GB/s per node
-Grid       : Message        :  Stencil 13.7508 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663232
-Grid       : Message        : Average mflops/s per call per node       : 798296
-Grid       : Message        : Average mflops/s per call per node       : 822587
-Grid       : Message        : Average mflops/s per call per node       : 660829
-Grid       : Message        : Average mflops/s per call per node (full): 314733
-Grid       : Message        : Average mflops/s per call per node (full): 437104
-Grid       : Message        : Average mflops/s per call per node (full): 447399
-Grid       : Message        : Average mflops/s per call per node (full): 306421
-Grid       : Message        :  Stencil 12.5284 GB/s per node
-Grid       : Message        :  Stencil 16.9445 GB/s per node
-Grid       : Message        :  Stencil 17.8376 GB/s per node
-Grid       : Message        :  Stencil 12.8427 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668494
-Grid       : Message        : Average mflops/s per call per node       : 807032
-Grid       : Message        : Average mflops/s per call per node       : 823168
-Grid       : Message        : Average mflops/s per call per node       : 663439
-Grid       : Message        : Average mflops/s per call per node (full): 313697
-Grid       : Message        : Average mflops/s per call per node (full): 437154
-Grid       : Message        : Average mflops/s per call per node (full): 446625
-Grid       : Message        : Average mflops/s per call per node (full): 303679
-Grid       : Message        :  Stencil 13.193 GB/s per node
-Grid       : Message        :  Stencil 16.6136 GB/s per node
-Grid       : Message        :  Stencil 17.4497 GB/s per node
-Grid       : Message        :  Stencil 12.2637 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666149
-Grid       : Message        : Average mflops/s per call per node       : 805048
-Grid       : Message        : Average mflops/s per call per node       : 825446
-Grid       : Message        : Average mflops/s per call per node       : 663333
-Grid       : Message        : Average mflops/s per call per node (full): 315673
-Grid       : Message        : Average mflops/s per call per node (full): 436055
-Grid       : Message        : Average mflops/s per call per node (full): 444547
-Grid       : Message        : Average mflops/s per call per node (full): 303758
-Grid       : Message        :  Stencil 13.4605 GB/s per node
-Grid       : Message        :  Stencil 16.4581 GB/s per node
-Grid       : Message        :  Stencil 18.7226 GB/s per node
-Grid       : Message        :  Stencil 12.3411 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667912
-Grid       : Message        : Average mflops/s per call per node       : 807641
-Grid       : Message        : Average mflops/s per call per node       : 827192
-Grid       : Message        : Average mflops/s per call per node       : 663584
-Grid       : Message        : Average mflops/s per call per node (full): 313717
-Grid       : Message        : Average mflops/s per call per node (full): 434972
-Grid       : Message        : Average mflops/s per call per node (full): 450158
-Grid       : Message        : Average mflops/s per call per node (full): 303868
-Grid       : Message        :  Stencil 12.5401 GB/s per node
-Grid       : Message        :  Stencil 11.6789 GB/s per node
-Grid       : Message        :  Stencil 17.4623 GB/s per node
-Grid       : Message        :  Stencil 14.3254 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671632
-Grid       : Message        : Average mflops/s per call per node       : 813861
-Grid       : Message        : Average mflops/s per call per node       : 829044
-Grid       : Message        : Average mflops/s per call per node       : 657813
-Grid       : Message        : Average mflops/s per call per node (full): 314374
-Grid       : Message        : Average mflops/s per call per node (full): 354116
-Grid       : Message        : Average mflops/s per call per node (full): 446129
-Grid       : Message        : Average mflops/s per call per node (full): 306680
-Grid       : Message        :  Stencil 12.7752 GB/s per node
-Grid       : Message        :  Stencil 16.5636 GB/s per node
-Grid       : Message        :  Stencil 17.2492 GB/s per node
-Grid       : Message        :  Stencil 12.8813 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667719
-Grid       : Message        : Average mflops/s per call per node       : 803560
-Grid       : Message        : Average mflops/s per call per node       : 827933
-Grid       : Message        : Average mflops/s per call per node       : 661443
-Grid       : Message        : Average mflops/s per call per node (full): 314865
-Grid       : Message        : Average mflops/s per call per node (full): 434684
-Grid       : Message        : Average mflops/s per call per node (full): 443851
-Grid       : Message        : Average mflops/s per call per node (full): 304589
-Grid       : Message        :  Stencil 14.3801 GB/s per node
-Grid       : Message        :  Stencil 9.78086 GB/s per node
-Grid       : Message        :  Stencil 17.8346 GB/s per node
-Grid       : Message        :  Stencil 12.1078 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662537
-Grid       : Message        : Average mflops/s per call per node       : 813351
-Grid       : Message        : Average mflops/s per call per node       : 821884
-Grid       : Message        : Average mflops/s per call per node       : 664214
-Grid       : Message        : Average mflops/s per call per node (full): 315769
-Grid       : Message        : Average mflops/s per call per node (full): 311062
-Grid       : Message        : Average mflops/s per call per node (full): 443457
-Grid       : Message        : Average mflops/s per call per node (full): 302417
-Grid       : Message        :  Stencil 13.2075 GB/s per node
-Grid       : Message        :  Stencil 12.8025 GB/s per node
-Grid       : Message        :  Stencil 17.5761 GB/s per node
-Grid       : Message        :  Stencil 15.2306 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662740
-Grid       : Message        : Average mflops/s per call per node       : 807134
-Grid       : Message        : Average mflops/s per call per node       : 820897
-Grid       : Message        : Average mflops/s per call per node       : 658357
-Grid       : Message        : Average mflops/s per call per node (full): 315204
-Grid       : Message        : Average mflops/s per call per node (full): 377502
-Grid       : Message        : Average mflops/s per call per node (full): 445586
-Grid       : Message        : Average mflops/s per call per node (full): 307638
-Grid       : Message        :  Stencil 12.9764 GB/s per node
-Grid       : Message        :  Stencil 17.1323 GB/s per node
-Grid       : Message        :  Stencil 18.2949 GB/s per node
-Grid       : Message        :  Stencil 12.4703 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665907
-Grid       : Message        : Average mflops/s per call per node       : 800038
-Grid       : Message        : Average mflops/s per call per node       : 822861
-Grid       : Message        : Average mflops/s per call per node       : 662657
-Grid       : Message        : Average mflops/s per call per node (full): 313891
-Grid       : Message        : Average mflops/s per call per node (full): 436920
-Grid       : Message        : Average mflops/s per call per node (full): 444330
-Grid       : Message        : Average mflops/s per call per node (full): 298764
-Grid       : Message        :  Stencil 15.4686 GB/s per node
-Grid       : Message        :  Stencil 16.8296 GB/s per node
-Grid       : Message        :  Stencil 17.0712 GB/s per node
-Grid       : Message        :  Stencil 12.3795 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662064
-Grid       : Message        : Average mflops/s per call per node       : 804013
-Grid       : Message        : Average mflops/s per call per node       : 824463
-Grid       : Message        : Average mflops/s per call per node       : 670516
-Grid       : Message        : Average mflops/s per call per node (full): 317322
-Grid       : Message        : Average mflops/s per call per node (full): 437408
-Grid       : Message        : Average mflops/s per call per node (full): 441839
-Grid       : Message        : Average mflops/s per call per node (full): 305327
-Grid       : Message        :  Stencil 13.1037 GB/s per node
-Grid       : Message        :  Stencil 16.9373 GB/s per node
-Grid       : Message        :  Stencil 17.7799 GB/s per node
-Grid       : Message        :  Stencil 13.197 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665706
-Grid       : Message        : Average mflops/s per call per node       : 806766
-Grid       : Message        : Average mflops/s per call per node       : 819907
-Grid       : Message        : Average mflops/s per call per node       : 662447
-Grid       : Message        : Average mflops/s per call per node (full): 315174
-Grid       : Message        : Average mflops/s per call per node (full): 438405
-Grid       : Message        : Average mflops/s per call per node (full): 445012
-Grid       : Message        : Average mflops/s per call per node (full): 304715
-Grid       : Message        :  Stencil 13.7707 GB/s per node
-Grid       : Message        :  Stencil 16.8391 GB/s per node
-Grid       : Message        :  Stencil 17.6309 GB/s per node
-Grid       : Message        :  Stencil 12.15 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661970
-Grid       : Message        : Average mflops/s per call per node       : 807154
-Grid       : Message        : Average mflops/s per call per node       : 819629
-Grid       : Message        : Average mflops/s per call per node       : 664143
-Grid       : Message        : Average mflops/s per call per node (full): 314464
-Grid       : Message        : Average mflops/s per call per node (full): 435569
-Grid       : Message        : Average mflops/s per call per node (full): 441555
-Grid       : Message        : Average mflops/s per call per node (full): 302821
-Grid       : Message        :  Stencil 14.2817 GB/s per node
-Grid       : Message        :  Stencil 17.0494 GB/s per node
-Grid       : Message        :  Stencil 16.4168 GB/s per node
-Grid       : Message        :  Stencil 12.3744 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661496
-Grid       : Message        : Average mflops/s per call per node       : 803949
-Grid       : Message        : Average mflops/s per call per node       : 825886
-Grid       : Message        : Average mflops/s per call per node       : 663562
-Grid       : Message        : Average mflops/s per call per node (full): 316101
-Grid       : Message        : Average mflops/s per call per node (full): 436399
-Grid       : Message        : Average mflops/s per call per node (full): 428930
-Grid       : Message        : Average mflops/s per call per node (full): 301876
-Grid       : Message        :  Stencil 13.2501 GB/s per node
-Grid       : Message        :  Stencil 17.3485 GB/s per node
-Grid       : Message        :  Stencil 17.4801 GB/s per node
-Grid       : Message        :  Stencil 12.7829 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663072
-Grid       : Message        : Average mflops/s per call per node       : 802940
-Grid       : Message        : Average mflops/s per call per node       : 822540
-Grid       : Message        : Average mflops/s per call per node       : 662228
-Grid       : Message        : Average mflops/s per call per node (full): 314512
-Grid       : Message        : Average mflops/s per call per node (full): 438926
-Grid       : Message        : Average mflops/s per call per node (full): 442773
-Grid       : Message        : Average mflops/s per call per node (full): 305542
-Grid       : Message        :  Stencil 14.7135 GB/s per node
-Grid       : Message        :  Stencil 16.5017 GB/s per node
-Grid       : Message        :  Stencil 17.6741 GB/s per node
-Grid       : Message        :  Stencil 12.5544 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 663528
-Grid       : Message        : Average mflops/s per call per node       : 803113
-Grid       : Message        : Average mflops/s per call per node       : 824588
-Grid       : Message        : Average mflops/s per call per node       : 664828
-Grid       : Message        : Average mflops/s per call per node (full): 316604
-Grid       : Message        : Average mflops/s per call per node (full): 435590
-Grid       : Message        : Average mflops/s per call per node (full): 446359
-Grid       : Message        : Average mflops/s per call per node (full): 305030
-Grid       : Message        :  Stencil 13.0648 GB/s per node
-Grid       : Message        :  Stencil 17.6308 GB/s per node
-Grid       : Message        :  Stencil 18.0907 GB/s per node
-Grid       : Message        :  Stencil 12.67 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664297
-Grid       : Message        : Average mflops/s per call per node       : 803366
-Grid       : Message        : Average mflops/s per call per node       : 818922
-Grid       : Message        : Average mflops/s per call per node       : 662458
-Grid       : Message        : Average mflops/s per call per node (full): 314283
-Grid       : Message        : Average mflops/s per call per node (full): 437036
-Grid       : Message        : Average mflops/s per call per node (full): 445853
-Grid       : Message        : Average mflops/s per call per node (full): 303738
-Grid       : Message        :  Stencil 12.5967 GB/s per node
-Grid       : Message        :  Stencil 16.0996 GB/s per node
-Grid       : Message        :  Stencil 17.1523 GB/s per node
-Grid       : Message        :  Stencil 11.9602 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 671367
-Grid       : Message        : Average mflops/s per call per node       : 802117
-Grid       : Message        : Average mflops/s per call per node       : 822388
-Grid       : Message        : Average mflops/s per call per node       : 670004
-Grid       : Message        : Average mflops/s per call per node (full): 314223
-Grid       : Message        : Average mflops/s per call per node (full): 429436
-Grid       : Message        : Average mflops/s per call per node (full): 442282
-Grid       : Message        : Average mflops/s per call per node (full): 301663
-Grid       : Message        :  Stencil 13.3652 GB/s per node
-Grid       : Message        :  Stencil 14.7891 GB/s per node
-Grid       : Message        :  Stencil 17.3862 GB/s per node
-Grid       : Message        :  Stencil 13.2581 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 666124
-Grid       : Message        : Average mflops/s per call per node       : 805548
-Grid       : Message        : Average mflops/s per call per node       : 822880
-Grid       : Message        : Average mflops/s per call per node       : 661956
-Grid       : Message        : Average mflops/s per call per node (full): 315515
-Grid       : Message        : Average mflops/s per call per node (full): 411718
-Grid       : Message        : Average mflops/s per call per node (full): 444732
-Grid       : Message        : Average mflops/s per call per node (full): 303817
-Grid       : Message        :  Stencil 13.6342 GB/s per node
-Grid       : Message        :  Stencil 16.4294 GB/s per node
-Grid       : Message        :  Stencil 17.1953 GB/s per node
-Grid       : Message        :  Stencil 13.369 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665092
-Grid       : Message        : Average mflops/s per call per node       : 804618
-Grid       : Message        : Average mflops/s per call per node       : 824222
-Grid       : Message        : Average mflops/s per call per node       : 657667
-Grid       : Message        : Average mflops/s per call per node (full): 315976
-Grid       : Message        : Average mflops/s per call per node (full): 433965
-Grid       : Message        : Average mflops/s per call per node (full): 441241
-Grid       : Message        : Average mflops/s per call per node (full): 304885
-Grid       : Message        :  Stencil 14.6131 GB/s per node
-Grid       : Message        :  Stencil 16.4093 GB/s per node
-Grid       : Message        :  Stencil 18.1192 GB/s per node
-Grid       : Message        :  Stencil 13.5067 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665242
-Grid       : Message        : Average mflops/s per call per node       : 807371
-Grid       : Message        : Average mflops/s per call per node       : 822732
-Grid       : Message        : Average mflops/s per call per node       : 661930
-Grid       : Message        : Average mflops/s per call per node (full): 317716
-Grid       : Message        : Average mflops/s per call per node (full): 431823
-Grid       : Message        : Average mflops/s per call per node (full): 447833
-Grid       : Message        : Average mflops/s per call per node (full): 306907
-Grid       : Message        :  Stencil 14.0475 GB/s per node
-Grid       : Message        :  Stencil 14.4059 GB/s per node
-Grid       : Message        :  Stencil 18.8707 GB/s per node
-Grid       : Message        :  Stencil 12.4078 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667247
-Grid       : Message        : Average mflops/s per call per node       : 807574
-Grid       : Message        : Average mflops/s per call per node       : 819580
-Grid       : Message        : Average mflops/s per call per node       : 670418
-Grid       : Message        : Average mflops/s per call per node (full): 316787
-Grid       : Message        : Average mflops/s per call per node (full): 407665
-Grid       : Message        : Average mflops/s per call per node (full): 447793
-Grid       : Message        : Average mflops/s per call per node (full): 304722
-Grid       : Message        :  Stencil 12.8318 GB/s per node
-Grid       : Message        :  Stencil 16.8843 GB/s per node
-Grid       : Message        :  Stencil 18.1065 GB/s per node
-Grid       : Message        :  Stencil 11.7948 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667312
-Grid       : Message        : Average mflops/s per call per node       : 799072
-Grid       : Message        : Average mflops/s per call per node       : 826647
-Grid       : Message        : Average mflops/s per call per node       : 667891
-Grid       : Message        : Average mflops/s per call per node (full): 314263
-Grid       : Message        : Average mflops/s per call per node (full): 436737
-Grid       : Message        : Average mflops/s per call per node (full): 438770
-Grid       : Message        : Average mflops/s per call per node (full): 297490
-Grid       : Message        :  Stencil 13.3866 GB/s per node
-Grid       : Message        :  Stencil 12.1769 GB/s per node
-Grid       : Message        :  Stencil 17.4186 GB/s per node
-Grid       : Message        :  Stencil 11.9079 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664332
-Grid       : Message        : Average mflops/s per call per node       : 810631
-Grid       : Message        : Average mflops/s per call per node       : 821828
-Grid       : Message        : Average mflops/s per call per node       : 664980
-Grid       : Message        : Average mflops/s per call per node (full): 315441
-Grid       : Message        : Average mflops/s per call per node (full): 364776
-Grid       : Message        : Average mflops/s per call per node (full): 444564
-Grid       : Message        : Average mflops/s per call per node (full): 301166
-Grid       : Message        :  Stencil 13.3728 GB/s per node
-Grid       : Message        :  Stencil 17.5614 GB/s per node
-Grid       : Message        :  Stencil 18.0237 GB/s per node
-Grid       : Message        :  Stencil 13.8207 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667161
-Grid       : Message        : Average mflops/s per call per node       : 800879
-Grid       : Message        : Average mflops/s per call per node       : 815462
-Grid       : Message        : Average mflops/s per call per node       : 661988
-Grid       : Message        : Average mflops/s per call per node (full): 315692
-Grid       : Message        : Average mflops/s per call per node (full): 437055
-Grid       : Message        : Average mflops/s per call per node (full): 444586
-Grid       : Message        : Average mflops/s per call per node (full): 305587
-Grid       : Message        :  Stencil 12.6136 GB/s per node
-Grid       : Message        :  Stencil 17.8142 GB/s per node
-Grid       : Message        :  Stencil 18.7351 GB/s per node
-Grid       : Message        :  Stencil 12.0376 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667354
-Grid       : Message        : Average mflops/s per call per node       : 803405
-Grid       : Message        : Average mflops/s per call per node       : 823025
-Grid       : Message        : Average mflops/s per call per node       : 665668
-Grid       : Message        : Average mflops/s per call per node (full): 313391
-Grid       : Message        : Average mflops/s per call per node (full): 440352
-Grid       : Message        : Average mflops/s per call per node (full): 448308
-Grid       : Message        : Average mflops/s per call per node (full): 302534
-Grid       : Message        :  Stencil 13.6781 GB/s per node
-Grid       : Message        :  Stencil 16.4091 GB/s per node
-Grid       : Message        :  Stencil 17.8156 GB/s per node
-Grid       : Message        :  Stencil 12.0971 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 661133
-Grid       : Message        : Average mflops/s per call per node       : 801506
-Grid       : Message        : Average mflops/s per call per node       : 820190
-Grid       : Message        : Average mflops/s per call per node       : 669406
-Grid       : Message        : Average mflops/s per call per node (full): 314497
-Grid       : Message        : Average mflops/s per call per node (full): 433770
-Grid       : Message        : Average mflops/s per call per node (full): 445647
-Grid       : Message        : Average mflops/s per call per node (full): 303000
-Grid       : Message        :  Stencil 14.0661 GB/s per node
-Grid       : Message        :  Stencil 16.698 GB/s per node
-Grid       : Message        :  Stencil 16.8629 GB/s per node
-Grid       : Message        :  Stencil 13.4859 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 660056
-Grid       : Message        : Average mflops/s per call per node       : 800700
-Grid       : Message        : Average mflops/s per call per node       : 826733
-Grid       : Message        : Average mflops/s per call per node       : 660054
-Grid       : Message        : Average mflops/s per call per node (full): 314176
-Grid       : Message        : Average mflops/s per call per node (full): 431452
-Grid       : Message        : Average mflops/s per call per node (full): 438813
-Grid       : Message        : Average mflops/s per call per node (full): 306706
-Grid       : Message        :  Stencil 15.2755 GB/s per node
-Grid       : Message        :  Stencil 16.3466 GB/s per node
-Grid       : Message        :  Stencil 16.9219 GB/s per node
-Grid       : Message        :  Stencil 12.5556 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 662693
-Grid       : Message        : Average mflops/s per call per node       : 802976
-Grid       : Message        : Average mflops/s per call per node       : 819842
-Grid       : Message        : Average mflops/s per call per node       : 661755
-Grid       : Message        : Average mflops/s per call per node (full): 317332
-Grid       : Message        : Average mflops/s per call per node (full): 433726
-Grid       : Message        : Average mflops/s per call per node (full): 439478
-Grid       : Message        : Average mflops/s per call per node (full): 303366
-Grid       : Message        :  Stencil 12.7011 GB/s per node
-Grid       : Message        :  Stencil 9.36114 GB/s per node
-Grid       : Message        :  Stencil 18.2933 GB/s per node
-Grid       : Message        :  Stencil 13.0964 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 665822
-Grid       : Message        : Average mflops/s per call per node       : 812056
-Grid       : Message        : Average mflops/s per call per node       : 823525
-Grid       : Message        : Average mflops/s per call per node       : 660552
-Grid       : Message        : Average mflops/s per call per node (full): 314345
-Grid       : Message        : Average mflops/s per call per node (full): 301000
-Grid       : Message        : Average mflops/s per call per node (full): 447518
-Grid       : Message        : Average mflops/s per call per node (full): 303723
-Grid       : Message        :  Stencil 13.1927 GB/s per node
-Grid       : Message        :  Stencil 16.8595 GB/s per node
-Grid       : Message        :  Stencil 17.886 GB/s per node
-Grid       : Message        :  Stencil 12.3675 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 667425
-Grid       : Message        : Average mflops/s per call per node       : 807531
-Grid       : Message        : Average mflops/s per call per node       : 820923
-Grid       : Message        : Average mflops/s per call per node       : 672801
-Grid       : Message        : Average mflops/s per call per node (full): 315755
-Grid       : Message        : Average mflops/s per call per node (full): 438818
-Grid       : Message        : Average mflops/s per call per node (full): 444555
-Grid       : Message        : Average mflops/s per call per node (full): 304944
-Grid       : Message        :  Stencil 12.9532 GB/s per node
-Grid       : Message        :  Stencil 10.464 GB/s per node
-Grid       : Message        :  Stencil 16.678 GB/s per node
-Grid       : Message        :  Stencil 13.437 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 668065
-Grid       : Message        : Average mflops/s per call per node       : 811810
-Grid       : Message        : Average mflops/s per call per node       : 821025
-Grid       : Message        : Average mflops/s per call per node       : 665160
-Grid       : Message        : Average mflops/s per call per node (full): 313479
-Grid       : Message        : Average mflops/s per call per node (full): 327432
-Grid       : Message        : Average mflops/s per call per node (full): 428518
-Grid       : Message        : Average mflops/s per call per node (full): 307295
-Grid       : Message        :  Stencil 13.5958 GB/s per node
-Grid       : Message        :  Stencil 10.7364 GB/s per node
-Grid       : Message        :  Stencil 17.6069 GB/s per node
-Grid       : Message        :  Stencil 13.6256 GB/s per node
-Grid       : Message        : Average mflops/s per call per node       : 664957
-Grid       : Message        : Average mflops/s per call per node       : 812429
-Grid       : Message        : Average mflops/s per call per node       : 819163
-Grid       : Message        : Average mflops/s per call per node       : 662325
-Grid       : Message        : Average mflops/s per call per node (full): 315613
-Grid       : Message        : Average mflops/s per call per node (full): 333839
-Grid       : Message        : Average mflops/s per call per node (full): 444411
-Grid       : Message        : Average mflops/s per call per node (full): 304561
diff --git a/scripts/update_eigen.sh b/scripts/update_eigen.sh
index b1834175..ab7b8f90 100755
--- a/scripts/update_eigen.sh
+++ b/scripts/update_eigen.sh
@@ -7,6 +7,7 @@ fi
 ARC=$1
 
 INITDIR=`pwd`
+
 rm -f Grid/Eigen
 rm -rf Eigen
 
@@ -33,3 +34,4 @@ find -L Eigen -type f -print | sed 's/^/  /;$q;s/$/ \\/' >> ${INITDIR}/Grid/Eige
 # back to home
 ###################################
 cd ${INITDIR}
+
diff --git a/tests/IO/Test_ildg_io.cc b/tests/IO/Test_ildg_io.cc
index cb5efed2..369acfd1 100644
--- a/tests/IO/Test_ildg_io.cc
+++ b/tests/IO/Test_ildg_io.cc
@@ -31,7 +31,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 
 int main (int argc, char ** argv)
@@ -41,12 +41,12 @@ int main (int argc, char ** argv)
 
   std::cout <<GridLogMessage<< " main "<<std::endl;
 
-  std::vector<int> simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  auto simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
   //std::vector<int> latt_size  ({48,48,48,96});
   //std::vector<int> latt_size  ({32,32,32,32});
-  std::vector<int> latt_size  ({16,16,16,32});
-  std::vector<int> clatt_size  ({4,4,4,8});
+  Coordinate latt_size  ({16,16,16,32});
+  Coordinate clatt_size  ({4,4,4,8});
   int orthodir=3;
   int orthosz =latt_size[orthodir];
     
diff --git a/tests/IO/Test_ildg_read.cc b/tests/IO/Test_ildg_read.cc
index 958fbe4e..d846453d 100644
--- a/tests/IO/Test_ildg_read.cc
+++ b/tests/IO/Test_ildg_read.cc
@@ -31,7 +31,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 
 int main (int argc, char ** argv)
@@ -40,9 +40,9 @@ int main (int argc, char ** argv)
   Grid_init(&argc,&argv);
 
 
-  std::vector<int> simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
-  std::vector<int> latt_size = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size = GridDefaultLatt();
   int orthodir=3;
   int orthosz =latt_size[orthodir];
     
@@ -64,7 +64,7 @@ int main (int argc, char ** argv)
 
   // Painful ; fix syntactical niceness
   LatticeComplex LinkTrace(&Fine);
-  LinkTrace=zero;
+  LinkTrace=Zero();
   for(int mu=0;mu<Nd;mu++){
     LinkTrace = LinkTrace + trace(U[mu]);
   }
@@ -72,7 +72,7 @@ int main (int argc, char ** argv)
   // (1+2+3)=6 = N(N-1)/2 terms
   LatticeComplex Plaq(&Fine);
 
-  Plaq = zero;
+  Plaq = Zero();
 
   for(int mu=1;mu<Nd;mu++){
     for(int nu=0;nu<mu;nu++){
@@ -89,7 +89,7 @@ int main (int argc, char ** argv)
   int Nt = Plaq_T.size();
 
   TComplex Plaq_T_sum; 
-  Plaq_T_sum=zero;
+  Plaq_T_sum=Zero();
   for(int t=0;t<Nt;t++){
     Plaq_T_sum = Plaq_T_sum+Plaq_T[t];
     Complex Pt=TensorRemove(Plaq_T[t]);
diff --git a/tests/IO/Test_nersc_io.cc b/tests/IO/Test_nersc_io.cc
index ca04e623..f5413e3b 100644
--- a/tests/IO/Test_nersc_io.cc
+++ b/tests/IO/Test_nersc_io.cc
@@ -31,7 +31,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 
 int main (int argc, char ** argv)
@@ -40,19 +40,18 @@ int main (int argc, char ** argv)
 
   std::cout <<GridLogMessage<< " main "<<std::endl;
 
-  std::vector<int> simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   //std::vector<int> latt_size  ({48,48,48,96});
   //std::vector<int> latt_size  ({32,32,32,32});
-  std::vector<int> latt_size  ({16,16,16,32});
-  std::vector<int> clatt_size  ({4,4,4,8});
+  Coordinate latt_size  ({16,16,16,32});
+  Coordinate clatt_size  ({4,4,4,8});
   int orthodir=3;
   int orthosz =latt_size[orthodir];
     
   GridCartesian     Fine(latt_size,simd_layout,mpi_layout);
   GridCartesian     Coarse(clatt_size,simd_layout,mpi_layout);
 
-
   GridParallelRNG   pRNGa(&Fine);
   GridParallelRNG   pRNGb(&Fine);
   GridSerialRNG     sRNGa;
@@ -106,7 +105,7 @@ int main (int argc, char ** argv)
 
   // Painful ; fix syntactical niceness
   LatticeComplex LinkTrace(&Fine);
-  LinkTrace=zero;
+  LinkTrace=Zero();
   for(int mu=0;mu<Nd;mu++){
     LinkTrace = LinkTrace + trace(U[mu]);
   }
@@ -115,7 +114,7 @@ int main (int argc, char ** argv)
   LatticeComplex Plaq(&Fine);
   LatticeComplex cPlaq(&Coarse);
 
-  Plaq = zero;
+  Plaq = Zero();
 #if 1
   for(int mu=1;mu<Nd;mu++){
     for(int nu=0;nu<mu;nu++){
@@ -131,7 +130,7 @@ int main (int argc, char ** argv)
   int Nt = Plaq_T.size();
 
   TComplex Plaq_T_sum; 
-  Plaq_T_sum=zero;
+  Plaq_T_sum=Zero();
   for(int t=0;t<Nt;t++){
     Plaq_T_sum = Plaq_T_sum+Plaq_T[t];
     Complex Pt=TensorRemove(Plaq_T[t]);
diff --git a/tests/IO/Test_nersc_read.cc b/tests/IO/Test_nersc_read.cc
index 66b2d62e..353bf02d 100644
--- a/tests/IO/Test_nersc_read.cc
+++ b/tests/IO/Test_nersc_read.cc
@@ -31,17 +31,14 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
-
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-
-  std::vector<int> simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
-  std::vector<int> latt_size = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size = GridDefaultLatt();
   int orthodir=3;
   int orthosz =latt_size[orthodir];
     
@@ -60,7 +57,7 @@ int main (int argc, char ** argv)
 
   // Painful ; fix syntactical niceness
   LatticeComplex LinkTrace(&Fine);
-  LinkTrace=zero;
+  LinkTrace=Zero();
   for(int mu=0;mu<Nd;mu++){
     LinkTrace = LinkTrace + trace(U[mu]);
   }
@@ -68,7 +65,7 @@ int main (int argc, char ** argv)
   // (1+2+3)=6 = N(N-1)/2 terms
   LatticeComplex Plaq(&Fine);
 
-  Plaq = zero;
+  Plaq = Zero();
 #if 1
   for(int mu=1;mu<Nd;mu++){
     for(int nu=0;nu<mu;nu++){
@@ -85,7 +82,7 @@ int main (int argc, char ** argv)
   int Nt = Plaq_T.size();
 
   TComplex Plaq_T_sum; 
-  Plaq_T_sum=zero;
+  Plaq_T_sum=Zero();
   for(int t=0;t<Nt;t++){
     Plaq_T_sum = Plaq_T_sum+Plaq_T[t];
     Complex Pt=TensorRemove(Plaq_T[t]);
diff --git a/tests/IO/Test_serialisation.cc b/tests/IO/Test_serialisation.cc
index 522e0cb5..7f3e5bca 100644
--- a/tests/IO/Test_serialisation.cc
+++ b/tests/IO/Test_serialisation.cc
@@ -32,12 +32,12 @@ Author: Michael Marshall <michael.marshall@ed.ac.uk>
 #include <typeinfo>
 
 using namespace Grid;
-using namespace Grid::QCD;
 
 GRID_SERIALIZABLE_ENUM(myenum, undef, red, 1, blue, 2, green, 3);
 
 class myclass: Serializable {
 public:
+
   GRID_SERIALIZABLE_CLASS_MEMBERS(myclass,
                           myenum, e,
                           std::vector<myenum>, ve,
@@ -47,8 +47,8 @@ public:
                           bool , b,
                           std::vector<double>, array,
                           std::vector<std::vector<double> >, twodimarray,
-                          std::vector<std::vector<std::vector<Complex> > >, cmplx3darray,
-                          SpinColourMatrix, scm
+			  std::vector<std::vector<std::vector<Complex> > >, cmplx3darray,
+			  SpinColourMatrix, scm
                           );
   myclass() {}
   myclass(int i)
@@ -311,8 +311,8 @@ int main(int argc,char **argv)
   ioTest<TextWriter, TextReader>("iotest.dat", obj, "text   (object)           ");
   ioTest<TextWriter, TextReader>("iotest.dat", vec, "text   (vector of objects)");
   //// text
-  ioTest<JSONWriter, JSONReader>("iotest.json", obj,  "JSON   (object)           ");
-  ioTest<JSONWriter, JSONReader>("iotest.json", vec,  "JSON   (vector of objects)");
+  //  ioTest<JSONWriter, JSONReader>("iotest.json", obj,  "JSON   (object)           ");
+  //  ioTest<JSONWriter, JSONReader>("iotest.json", vec,  "JSON   (vector of objects)");
 
   //// HDF5
 #ifdef HAVE_HDF5
diff --git a/tests/Test_cayley_even_odd_vec.cc b/tests/Test_cayley_even_odd_vec.cc
index fc271001..0e71d910 100644
--- a/tests/Test_cayley_even_odd_vec.cc
+++ b/tests/Test_cayley_even_odd_vec.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -43,15 +43,6 @@ struct scal {
         Gamma::Algebra::GammaT
     };
 
-typedef DomainWallFermion<DomainWallVec5dImplR>                      DomainWallVecFermionR;
-typedef ZMobiusFermion<ZDomainWallVec5dImplR>                        ZMobiusVecFermionR;
-typedef MobiusFermion<DomainWallVec5dImplR>                          MobiusVecFermionR;
-typedef MobiusZolotarevFermion<DomainWallVec5dImplR>                 MobiusZolotarevVecFermionR;
-typedef ScaledShamirFermion<DomainWallVec5dImplR>                    ScaledShamirVecFermionR;
-typedef ShamirZolotarevFermion<DomainWallVec5dImplR>                 ShamirZolotarevVecFermionR;
-typedef OverlapWilsonCayleyTanhFermion<DomainWallVec5dImplR>         OverlapWilsonCayleyTanhVecFermionR;
-typedef OverlapWilsonCayleyZolotarevFermion<DomainWallVec5dImplR>    OverlapWilsonCayleyZolotarevVecFermionR;
-
 template<class What> 
 void  TestWhat(What & Ddwf,
 	       GridCartesian         * FGrid,	       GridRedBlackCartesian * FrbGrid,
@@ -59,9 +50,6 @@ void  TestWhat(What & Ddwf,
 	       RealD mass, RealD M5,
 	       GridParallelRNG *RNG4,   GridParallelRNG *RNG5);
 
-template<class This,class That> 
-void  TestMoo(This & Dw, That &sDw);
-
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
@@ -71,7 +59,7 @@ int main (int argc, char ** argv)
 
   const int Ls=16;
 
-  std::vector<int> latt4  =GridDefaultLatt();
+  auto latt4  =GridDefaultLatt();
 
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(latt4, GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
   GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
@@ -101,11 +89,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage <<"DomainWallFermion vectorised test"<<std::endl;
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
   DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-  DomainWallVecFermionR sDdwf(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5);
-
-  TestMoo(Ddwf,sDdwf);
   TestWhat<DomainWallFermionR>(Ddwf,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
-  TestWhat<DomainWallVecFermionR>(sDdwf,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
 
   RealD b=1.5;// Scale factor b+c=2, b-c=1
   RealD c=0.5;
@@ -114,75 +98,53 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage <<"MobiusFermion test"<<std::endl;
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
   MobiusFermionR     Dmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c);
-  MobiusVecFermionR sDmob(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,b,c);
-  TestMoo(Dmob,sDmob);
   TestWhat<MobiusFermionR>(Dmob,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
-  TestWhat<MobiusVecFermionR>(sDmob,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
-
 
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
   std::cout<<GridLogMessage <<"Z-MobiusFermion test"<<std::endl;
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
   std::vector<ComplexD> gamma(Ls,std::complex<double>(1.0,0.0));
   ZMobiusFermionR     zDmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,gamma,b,c);
-  ZMobiusVecFermionR szDmob(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,gamma,b,c);
-  TestMoo(zDmob,szDmob);
   TestWhat<ZMobiusFermionR>(zDmob,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
-  TestWhat<ZMobiusVecFermionR>(szDmob,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
 
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
   std::cout<<GridLogMessage <<"MobiusZolotarevFermion test"<<std::endl;
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
 
   MobiusZolotarevFermionR Dzolo(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,0.1,2.0);
-  MobiusZolotarevVecFermionR sDzolo(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,b,c,0.1,2.0);
 
-  TestMoo(Dzolo,sDzolo);
   TestWhat<MobiusZolotarevFermionR>(Dzolo,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
-  TestWhat<MobiusZolotarevVecFermionR>(sDzolo,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
 
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
   std::cout<<GridLogMessage <<"ScaledShamirFermion test"<<std::endl;
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
 
   ScaledShamirFermionR Dsham(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0);
-  ScaledShamirVecFermionR sDsham(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,2.0);
 
-  TestMoo(Dsham,sDsham);
   TestWhat<ScaledShamirFermionR>(Dsham,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
-  TestWhat<ScaledShamirVecFermionR>(sDsham,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
 
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
   std::cout<<GridLogMessage <<"ShamirZolotarevFermion test"<<std::endl;
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
 
   ShamirZolotarevFermionR Dshamz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
-  ShamirZolotarevVecFermionR sDshamz(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,0.1,2.0);
 
-  TestMoo(Dshamz,sDshamz);
   TestWhat<ShamirZolotarevFermionR>(Dshamz,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
-  TestWhat<ShamirZolotarevVecFermionR>(sDshamz,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
 
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
   std::cout<<GridLogMessage <<"OverlapWilsonCayleyTanhFermion test"<<std::endl;
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
   OverlapWilsonCayleyTanhFermionR Dov(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
-  OverlapWilsonCayleyTanhVecFermionR sDov(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,1.0);
 
-  TestMoo(Dov,sDov);
   TestWhat<OverlapWilsonCayleyTanhFermionR>(Dov,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
-  TestWhat<OverlapWilsonCayleyTanhVecFermionR>(sDov,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
 
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
   std::cout<<GridLogMessage <<"OverlapWilsonCayleyZolotarevFermion test"<<std::endl;
   std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
 
   OverlapWilsonCayleyZolotarevFermionR Dovz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
-  OverlapWilsonCayleyZolotarevVecFermionR sDovz(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,0.1,2.0);
 
-  TestMoo(Dovz,sDovz);
   TestWhat<OverlapWilsonCayleyZolotarevFermionR>(Dovz,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
-  TestWhat<OverlapWilsonCayleyZolotarevVecFermionR>(sDovz,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
 
   std::cout<<GridLogMessage<<"=============================================================="<<std::endl;
 
@@ -201,10 +163,10 @@ void  TestWhat(What & Ddwf,
   LatticeFermion src   (FGrid); random(*RNG5,src);
   LatticeFermion phi   (FGrid); random(*RNG5,phi);
   LatticeFermion chi   (FGrid); random(*RNG5,chi);
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid);    ref=zero;
-  LatticeFermion    tmp(FGrid);    tmp=zero;
-  LatticeFermion    err(FGrid);    tmp=zero;
+  LatticeFermion result(FGrid); result=Zero();
+  LatticeFermion    ref(FGrid);    ref=Zero();
+  LatticeFermion    tmp(FGrid);    tmp=Zero();
+  LatticeFermion    err(FGrid);    tmp=Zero();
 
   LatticeFermion src_e (FrbGrid);
   LatticeFermion src_o (FrbGrid);
@@ -324,150 +286,3 @@ void  TestWhat(What & Ddwf,
 }
 
 
-template<class This,class That> 
-void  TestMoo(This & Dw, That &sDw)
-{
-  GridBase *sgrid= sDw.FermionGrid();
-  GridBase *ngrid=  Dw.FermionGrid();
-
-  int Ls = Dw.Ls;
-
-  LatticeFermion ssrc(sgrid);
-  LatticeFermion nsrc(ngrid);
-  LatticeFermion zz(ngrid); zz=zero;
-  LatticeFermion sres(sgrid);
-  LatticeFermion nres(ngrid);
-  LatticeFermion ndiff(ngrid);
-  LatticeFermion sdiff(sgrid);
-
-    Gamma g5( Gamma::Algebra::Gamma5 );
-
-  std::vector<int> seeds({1,2,3,4,5,7,8});
-  GridParallelRNG    RNG5(ngrid);  
-  RNG5.SeedFixedIntegers(seeds);
-  random(RNG5,nsrc);
-  //  nsrc = nsrc + g5*nsrc;
-
-  //  Lattice<iScalar<vInteger> > coor(ngrid);
-  //  LatticeCoordinate(coor,0);//scoor
-  //  nsrc = where(coor==(Integer)0,zz,nsrc);
-
-  std::vector<int> latt4(4);
-  for(int d=0;d<4;d++){
-    latt4[d] = ngrid->_fdimensions[d+1];
-  }
-
-  for(int x=0;x<latt4[0];x++){
-  for(int y=0;y<latt4[1];y++){
-  for(int z=0;z<latt4[2];z++){
-  for(int t=0;t<latt4[3];t++){
-  for(int s=0;s<Ls;s++){
-    std::vector<int> site({s,x,y,z,t});
-    SpinColourVector tmp;
-    peekSite(tmp,nsrc,site);
-    pokeSite(tmp,ssrc,site);
-  }}}}}
-
-  sDw.Mooee(ssrc,sres);
-   Dw.Mooee(nsrc,nres);
-
-  sDw.MooeeInternal(ssrc,sdiff,DaggerNo,InverseNo);
-
-
-  for(int x=0;x<latt4[0];x++){
-  for(int y=0;y<latt4[1];y++){
-  for(int z=0;z<latt4[2];z++){
-  for(int t=0;t<latt4[3];t++){
-  for(int s=0;s<Ls;s++){
-
-    std::vector<int> site({s,x,y,z,t});
-    SpinColourVector stmp;
-    SpinColourVector itmp;
-    SpinColourVector dtmp;
-    peekSite(stmp,sres,site);
-    peekSite(itmp,sdiff,site);
-
-    dtmp=itmp-stmp;
-    if ( norm2(dtmp)>1.0e-6) { 
-      std::cout<<GridLogMessage<<"=============================================================="<<std::endl;
-      std:: cout << x<<" "<<y<<" "<< z<< " "<<t<<"; s= "<<s<<std::endl;
-      std:: cout << "stmp "<< stmp <<std::endl;
-      std:: cout << "itmp "<< itmp <<std::endl;
-      std::cout<<GridLogMessage<<"=============================================================="<<std::endl;
-    }
-  }}}}}
-  sdiff = sdiff -sres;
-  std::cout<<GridLogMessage<<" norm MooInternal diff "<<norm2(sdiff)<<std::endl;
-
-  for(int x=0;x<latt4[0];x++){
-  for(int y=0;y<latt4[1];y++){
-  for(int z=0;z<latt4[2];z++){
-  for(int t=0;t<latt4[3];t++){
-  for(int s=0;s<Ls;s++){
-    std::vector<int> site({s,x,y,z,t});
-
-    SpinColourVector tmp;
-    peekSite(tmp,sres,site);
-    pokeSite(tmp,ndiff,site);
-  }}}}}
-  ndiff=ndiff-nres;
-  std::cout<<GridLogMessage<<" norm Moo diff "<<norm2(ndiff)<<std::endl;
-
-  sDw.MooeeDag(ssrc,sres);
-   Dw.MooeeDag(nsrc,nres);
-  sDw.MooeeInternal(ssrc,sdiff,DaggerYes,InverseNo);
-
-  sdiff = sdiff -sres;
-  std::cout<<GridLogMessage<<" norm MooInternalDag diff "<<norm2(sdiff)<<std::endl;
-
-  for(int x=0;x<latt4[0];x++){
-  for(int y=0;y<latt4[1];y++){
-  for(int z=0;z<latt4[2];z++){
-  for(int t=0;t<latt4[3];t++){
-  for(int s=0;s<Ls;s++){
-    std::vector<int> site({s,x,y,z,t});
-
-    SpinColourVector tmp;
-    peekSite(tmp,sres,site);
-    pokeSite(tmp,ndiff,site);
-  }}}}}
-  ndiff=ndiff-nres;
-  std::cout<<GridLogMessage<<" norm MooeeDag diff "<<norm2(ndiff)<<std::endl;
-
-  sDw.MooeeInv(ssrc,sres);
-   Dw.MooeeInv(nsrc,nres);
-
-  for(int x=0;x<latt4[0];x++){
-  for(int y=0;y<latt4[1];y++){
-  for(int z=0;z<latt4[2];z++){
-  for(int t=0;t<latt4[3];t++){
-  for(int s=0;s<Ls;s++){
-    std::vector<int> site({s,x,y,z,t});
-
-    SpinColourVector tmp;
-    peekSite(tmp,sres,site);
-    pokeSite(tmp,ndiff,site);
-  }}}}}
-  ndiff=ndiff-nres;
-  std::cout<<GridLogMessage<<" norm MooeeInv diff "<<norm2(ndiff)<<std::endl;
-
-  sDw.MooeeInvDag(ssrc,sres);
-   Dw.MooeeInvDag(nsrc,nres);
-
-  for(int x=0;x<latt4[0];x++){
-  for(int y=0;y<latt4[1];y++){
-  for(int z=0;z<latt4[2];z++){
-  for(int t=0;t<latt4[3];t++){
-  for(int s=0;s<Ls;s++){
-    std::vector<int> site({s,x,y,z,t});
-    SpinColourVector tmp;
-    peekSite(tmp,sres,site);
-    pokeSite(tmp,ndiff,site);
-  }}}}}
-  ndiff=ndiff-nres;
-  std::cout<<GridLogMessage<<" norm MooeeInvDag diff "<<norm2(ndiff)<<std::endl;
-
-  std::cout<<GridLogMessage<<"=============================================================="<<std::endl;
-
-
-}
diff --git a/tests/Test_compressed_lanczos_hot_start.cc b/tests/Test_compressed_lanczos_hot_start.cc
index 65f69c54..8eb7a921 100644
--- a/tests/Test_compressed_lanczos_hot_start.cc
+++ b/tests/Test_compressed_lanczos_hot_start.cc
@@ -36,7 +36,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 template<class Fobj,class CComplex,int nbasis>
 class LocalCoherenceLanczosScidac : public LocalCoherenceLanczos<Fobj,CComplex,nbasis>
@@ -59,7 +58,7 @@ public:
 #ifdef HAVE_LIME
     assert(this->subspace.size()==nbasis);
     emptyUserRecord record;
-    Grid::QCD::ScidacWriter WR(this->_FineGrid->IsBoss());
+    Grid::ScidacWriter WR(this->_FineGrid->IsBoss());
     WR.open(evecs_file);
     for(int k=0;k<nbasis;k++) {
       WR.writeScidacFieldRecord(this->subspace[k],record);
@@ -87,7 +86,7 @@ public:
     
     std::cout << GridLogIRL<< "checkpointFineRestore:  Reading evecs from "<<evecs_file<<std::endl;
     emptyUserRecord record;
-    Grid::QCD::ScidacReader RD ;
+    Grid::ScidacReader RD ;
     RD.open(evecs_file);
     for(int k=0;k<nbasis;k++) {
       this->subspace[k].checkerboard=this->_checkerboard;
@@ -105,7 +104,7 @@ public:
 #ifdef HAVE_LIME
     int n = this->evec_coarse.size();
     emptyUserRecord record;
-    Grid::QCD::ScidacWriter WR(this->_CoarseGrid->IsBoss());
+    Grid::ScidacWriter WR(this->_CoarseGrid->IsBoss());
     WR.open(evecs_file);
     for(int k=0;k<n;k++) {
       WR.writeScidacFieldRecord(this->evec_coarse[k],record);
@@ -132,7 +131,7 @@ public:
     assert(this->evals_coarse.size()==nvec);
     emptyUserRecord record;
     std::cout << GridLogIRL<< "checkpointCoarseRestore:  Reading evecs from "<<evecs_file<<std::endl;
-    Grid::QCD::ScidacReader RD ;
+    Grid::ScidacReader RD ;
     RD.open(evecs_file);
     for(int k=0;k<nvec;k++) {
       RD.readScidacFieldRecord(this->evec_coarse[k],record);
@@ -169,7 +168,7 @@ int main (int argc, char ** argv) {
   std::vector<int> blockSize = Params.blockSize;
   std::vector<int> latt({32,32,32,32});
   uint64_t     vol = Ls*latt[0]*latt[1]*latt[2]*latt[3];
-  double   mat_flop= 2.0*1320.0*vol;    
+  //  double   mat_flop= 2.0*1320.0*vol;    
   // Grids
   GridCartesian         * UGrid     = SpaceTimeGrid::makeFourDimGrid(latt,
 								     GridDefaultSimd(Nd,vComplex::Nsimd()),
@@ -217,8 +216,8 @@ int main (int argc, char ** argv) {
   LanczosParams fine  =Params.FineParams;  
   LanczosParams coarse=Params.CoarseParams;  
 
-  const int Ns1 = fine.Nstop;   const int Ns2 = coarse.Nstop;
-  const int Nk1 = fine.Nk;      const int Nk2 = coarse.Nk;
+  const int Ns1 = fine.Nstop;   //const int Ns2 = coarse.Nstop;
+  const int Nk1 = fine.Nk;      //const int Nk2 = coarse.Nk;
   const int Nm1 = fine.Nm;      const int Nm2 = coarse.Nm;
 
   std::cout << GridLogMessage << "Keep " << fine.Nstop   << " fine   vectors" << std::endl;
diff --git a/tests/Test_cshift.cc b/tests/Test_cshift.cc
index f9559a83..ad28aa54 100644
--- a/tests/Test_cshift.cc
+++ b/tests/Test_cshift.cc
@@ -29,15 +29,15 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  auto latt_size   = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
 
   GridCartesian        Fine(latt_size,simd_layout,mpi_layout);
 
@@ -47,10 +47,9 @@ int main (int argc, char ** argv)
   LatticeComplex ShiftU(&Fine);
 
   LatticeComplex lex(&Fine);
-  lex=zero;
+  lex=Zero();
   Integer stride =1;
   {
-    double nrm;
     LatticeComplex coor(&Fine);
 
     for(int d=0;d<4;d++){
@@ -79,7 +78,7 @@ int main (int argc, char ** argv)
 
 	ShiftU  = Cshift(U,dir,shift);    // Shift everything
 
-	std::vector<int> coor(4);
+	Coordinate coor(4);
 
 	for(coor[3]=0;coor[3]<latt_size[3];coor[3]++){
 	for(coor[2]=0;coor[2]<latt_size[2];coor[2]++){
@@ -90,7 +89,7 @@ int main (int argc, char ** argv)
 
 	  double nrm=norm2(U);
 
-	  std::vector<int> scoor(coor);
+	  Coordinate scoor(coor);
 	  scoor[dir] = (scoor[dir]+shift)%latt_size[dir];
 	  
 	  Integer slex = scoor[0]
@@ -101,7 +100,7 @@ int main (int argc, char ** argv)
 	  Complex scm(slex);
 	  
 	  nrm = abs(scm-cm()()());
-	  std::vector<int> peer(4);
+	  Coordinate peer(4);
 	  Complex tmp  =cm;
 	  Integer index=real(tmp);
 	  Lexicographic::CoorFromIndex(peer,index,latt_size);
diff --git a/tests/Test_dwf_mixedcg_prec.cc b/tests/Test_dwf_mixedcg_prec.cc
index 92567b6f..0b8463d9 100644
--- a/tests/Test_dwf_mixedcg_prec.cc
+++ b/tests/Test_dwf_mixedcg_prec.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -67,7 +67,7 @@ int main (int argc, char ** argv)
   GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
 
   LatticeFermionD    src(FGrid); random(RNG5,src);
-  LatticeFermionD result(FGrid); result=zero;
+  LatticeFermionD result(FGrid); result=Zero();
   LatticeGaugeFieldD Umu(UGrid);
   LatticeGaugeFieldF Umu_f(UGrid_f); 
   
@@ -84,10 +84,10 @@ int main (int argc, char ** argv)
   LatticeFermionD result_o(FrbGrid);
   LatticeFermionD result_o_2(FrbGrid);
   pickCheckerboard(Odd,src_o,src);
-  result_o.checkerboard = Odd;
-  result_o = zero;
-  result_o_2.checkerboard = Odd;
-  result_o_2 = zero;
+  result_o.Checkerboard() = Odd;
+  result_o = Zero();
+  result_o_2.Checkerboard() = Odd;
+  result_o_2 = Zero();
 
   SchurDiagMooeeOperator<DomainWallFermionD,LatticeFermionD> HermOpEO(Ddwf);
   SchurDiagMooeeOperator<DomainWallFermionF,LatticeFermionF> HermOpEO_f(Ddwf_f);
@@ -131,6 +131,6 @@ int main (int argc, char ** argv)
   }
   #endif
 
-
+  
   Grid_finalize();
 }
diff --git a/tests/Test_dwf_mixedcg_prec_halfcomms.cc b/tests/Test_dwf_mixedcg_prec_halfcomms.cc
index d6aaa21e..4d94632c 100644
--- a/tests/Test_dwf_mixedcg_prec_halfcomms.cc
+++ b/tests/Test_dwf_mixedcg_prec_halfcomms.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -65,7 +65,7 @@ int main (int argc, char ** argv)
   GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
 
   LatticeFermionD    src(FGrid); random(RNG5,src);
-  LatticeFermionD result(FGrid); result=zero;
+  LatticeFermionD result(FGrid); result=Zero();
   LatticeGaugeFieldD Umu(UGrid);
   LatticeGaugeFieldF Umu_f(UGrid_f); 
   
@@ -82,8 +82,8 @@ int main (int argc, char ** argv)
   LatticeFermionD    src_o(FrbGrid);
   LatticeFermionD result_cg(FrbGrid);
   pickCheckerboard(Odd,src_o,src);
-  result_cg.checkerboard = Odd;
-  result_cg = zero;
+  result_cg.Checkerboard() = Odd;
+  result_cg = Zero();
   LatticeFermionD result_mcg(result_cg);
   LatticeFermionD result_rlcg(result_cg);
 
diff --git a/tests/Test_general_stencil.cc b/tests/Test_general_stencil.cc
new file mode 100644
index 00000000..f03677d0
--- /dev/null
+++ b/tests/Test_general_stencil.cc
@@ -0,0 +1,162 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid
+
+    Source file: ./tests/Test_stencil.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+#include <Grid/stencil/GeneralLocalStencil.h>
+
+using namespace std;
+using namespace Grid;
+
+int main(int argc, char ** argv) 
+{
+  Grid_init(&argc, &argv);
+
+  //  typedef LatticeColourMatrix Field;
+  typedef LatticeComplex Field;
+  typedef typename Field::vector_object vobj;
+  typedef typename vobj::scalar_object sobj;
+
+  auto latt_size   = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
+
+  double volume = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
+
+  GridCartesian Fine(latt_size,simd_layout,mpi_layout);
+  GridRedBlackCartesian rbFine(&Fine);
+  GridParallelRNG       fRNG(&Fine);
+
+  //  fRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9});
+  std::vector<int> seeds({1,2,3,4});
+  fRNG.SeedFixedIntegers(seeds);
+
+  Field Foo(&Fine);
+  Field Bar(&Fine);
+  Field Check(&Fine);
+  Field Diff(&Fine);
+  LatticeComplex lex(&Fine);
+
+  lex = Zero();
+  random(fRNG,Foo);
+  gaussian(fRNG,Bar);
+
+  for (int i=0;i<simd_layout.size();i++){
+    std::cout <<" simd layout "<<i<<" = "<<simd_layout[i]<<std::endl;
+  }
+  Integer stride =1000;
+  {
+    LatticeComplex coor(&Fine);
+
+    for(int d=0;d<Nd;d++){
+      LatticeCoordinate(coor,d);
+      lex = lex + coor*stride;
+      stride=stride/10;
+    }
+    Foo=lex;
+  }
+
+  typedef GeneralLocalStencil        GeneralStencil;
+
+    for(int dir1=0;dir1<4;dir1++){
+    for(int dir2=0;dir2<4;dir2++){
+      if ( dir2!= dir1){
+      for(int disp1=0;disp1<Fine._fdimensions[dir1];disp1++){
+      for(int disp2=0;disp2<Fine._fdimensions[dir2];disp2++){
+
+	std::cout<< std::fixed <<GridLogMessage << "Using stencil to shift dim "<<dir1<< " by "<<disp1<<std::endl;
+	std::cout<< std::fixed <<GridLogMessage << "                   and dim "<<dir2<< " by "<<disp2<<std::endl;
+	// start to test the Cartesian npoint stencil infrastructure
+	int npoint=1;
+
+	Coordinate shift(Nd); 
+	for(int d=0;d<Nd;d++) shift[d]=0;
+	shift[dir1]=disp1;
+	shift[dir2]=disp2;
+	std::vector<Coordinate> shifts(npoint,shift);
+	GeneralLocalStencil gStencil(&Fine,shifts);
+
+	Bar = Cshift(Foo,dir1,disp1);
+	Bar = Cshift(Bar,dir2,disp2);
+
+	// Implement a stencil code that should agree with cshift!
+	for(int i=0;i<Check.Grid()->oSites();i++){
+	  auto SE = gStencil.GetEntry(0,i);
+	  auto check = Check.View();
+	  auto foo   = Foo.View();
+
+	  // Encapsulate in a general wrapper
+	  check[i] = foo[SE->_offset];                   auto tmp=check[i];
+	  if (SE->_permute & 0x1 ) { permute(check[i],tmp,0); tmp=check[i];}
+	  if (SE->_permute & 0x2 ) { permute(check[i],tmp,1); tmp=check[i];}
+	  if (SE->_permute & 0x4 ) { permute(check[i],tmp,2); tmp=check[i];}
+	  if (SE->_permute & 0x8 ) { permute(check[i],tmp,3); tmp=check[i];}
+	}
+
+	Real nrmC = norm2(Check);
+	Real nrmB = norm2(Bar);
+	Diff = Check-Bar;
+	Real nrm  = norm2(Diff);
+	std::cout<<GridLogMessage<<"N2diff ="<<nrm<<" "<<nrmC<<" " <<nrmB<<std::endl;
+
+	Coordinate coor(4);
+	for(coor[3]=0;coor[3]<latt_size[3]/mpi_layout[3];coor[3]++){
+	for(coor[2]=0;coor[2]<latt_size[2]/mpi_layout[2];coor[2]++){
+	for(coor[1]=0;coor[1]<latt_size[1]/mpi_layout[1];coor[1]++){
+	for(coor[0]=0;coor[0]<latt_size[0]/mpi_layout[0];coor[0]++){
+
+	  RealD diff;
+	  sobj check,bar;
+	  peekSite(check,Check,coor);
+	  peekSite(bar,Bar,coor);
+
+	  sobj ddiff;
+	  ddiff = check -bar;
+	  diff =norm2(ddiff);
+	  if ( diff > 0){
+	    std::cout <<"Coor (" << coor[0]<<","<<coor[1]<<","<<coor[2]<<","<<coor[3]
+		      <<") " <<check<<" vs "<<bar<<std::endl;
+	  }
+
+
+	}}}}
+
+	if (nrm > 1.0e-4) {
+	  auto check = Check.View();
+	  auto bar   = Bar.View();
+	  for(int i=0;i<check.size();i++){
+	    std::cout << i<<" Check "<<check[i]<< "\n"<<i<<" Bar "<<bar[i]<<std::endl;
+	  }
+	}
+	if (nrm > 1.0e-4) exit(-1);
+
+      }}
+      }
+    }}
+ Grid_finalize();
+}
diff --git a/tests/Test_simd.cc b/tests/Test_simd.cc
index b2e8d68e..b5799b7c 100644
--- a/tests/Test_simd.cc
+++ b/tests/Test_simd.cc
@@ -28,9 +28,7 @@ with this program; if not, write to the Free Software Foundation, Inc.,
     /*  END LEGAL */
 #include <Grid/Grid.h>
 
-using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 class funcPlus {
 public:
@@ -106,7 +104,7 @@ public:
 //  norm2, 
 //  Reduce,
 //
-//  mac,mult,sub,add, vone,vzero,vcomplex_i, =zero,
+//  mac,mult,sub,add, vone,vzero,vcomplex_i, =Zero(),
 //  vset,vsplat,vstore,vstream,vload, scalar*vec, vec*scalar
 //  unary -,
 //  *= , -=, +=
@@ -129,10 +127,10 @@ void Tester(const functor &func)
   
   int Nsimd = vec::Nsimd();
 
-  std::vector<scal> input1(Nsimd);
-  std::vector<scal> input2(Nsimd);
-  std::vector<scal> result(Nsimd);
-  std::vector<scal> reference(Nsimd);
+  ExtractBuffer<scal> input1(Nsimd);
+  ExtractBuffer<scal> input2(Nsimd);
+  ExtractBuffer<scal> result(Nsimd);
+  ExtractBuffer<scal> reference(Nsimd);
 
   std::vector<vec,alignedAllocator<vec> > buf(3);
   vec & v_input1 = buf[0];
@@ -186,10 +184,10 @@ void IntTester(const functor &func)
 
   int Nsimd = vec::Nsimd();
 
-  std::vector<scal> input1(Nsimd);
-  std::vector<scal> input2(Nsimd);
-  std::vector<scal> result(Nsimd);
-  std::vector<scal> reference(Nsimd);
+  ExtractBuffer<scal> input1(Nsimd);
+  ExtractBuffer<scal> input2(Nsimd);
+  ExtractBuffer<scal> result(Nsimd);
+  ExtractBuffer<scal> reference(Nsimd);
 
   std::vector<vec,alignedAllocator<vec> > buf(3);
   vec & v_input1 = buf[0];
@@ -244,8 +242,8 @@ void ReductionTester(const functor &func)
   
   int Nsimd = vec::Nsimd();
 
-  std::vector<scal> input1(Nsimd);
-  std::vector<scal> input2(Nsimd);
+  ExtractBuffer<scal> input1(Nsimd);
+  ExtractBuffer<scal> input2(Nsimd);
   reduced result(0);
   reduced reference(0);
   reduced tmp;
@@ -290,8 +288,8 @@ void IntReductionTester(const functor &func)
 {
   int Nsimd = vec::Nsimd();
 
-  std::vector<scal> input1(Nsimd);
-  std::vector<scal> input2(Nsimd);
+  ExtractBuffer<scal> input1(Nsimd);
+  ExtractBuffer<scal> input2(Nsimd);
   reduced result(0);
   reduced reference(0);
   reduced tmp;
@@ -335,7 +333,7 @@ public:
   int n;
   funcPermute(int _n) { n=_n;};
   template<class vec>    void operator()(vec &rr,vec &i1,vec &i2) const { permute(rr,i1,n);}
-  template<class scal>   void apply(std::vector<scal> &rr,std::vector<scal> &in)  const { 
+  template<class scal>   void apply(ExtractBuffer<scal> &rr,ExtractBuffer<scal> &in)  const { 
     int sz=in.size();
     int msk = sz>>(n+1);
     for(int i=0;i<sz;i++){
@@ -350,10 +348,10 @@ public:
   int n;
   funcExchange(int _n) { n=_n;};
   template<class vec>    void operator()(vec &r1,vec &r2,vec &i1,vec &i2) const { exchange(r1,r2,i1,i2,n);}
-  template<class scal>   void apply(std::vector<scal> &r1,
-				    std::vector<scal> &r2,
-				    std::vector<scal> &in1,
-				    std::vector<scal> &in2)  const 
+  template<class scal>   void apply(ExtractBuffer<scal> &r1,
+				    ExtractBuffer<scal> &r2,
+				    ExtractBuffer<scal> &in1,
+				    ExtractBuffer<scal> &in2)  const 
   { 
     int sz=in1.size();
     int msk = sz>>(n+1);
@@ -376,7 +374,7 @@ public:
   int n;
   funcRotate(int _n) { n=_n;};
   template<class vec>    void operator()(vec &rr,vec &i1,vec &i2) const { rr=rotate(i1,n);}
-  template<class scal>   void apply(std::vector<scal> &rr,std::vector<scal> &in)  const { 
+  template<class scal>   void apply(ExtractBuffer<scal> &rr,ExtractBuffer<scal> &in)  const { 
     int sz = in.size();
     for(int i=0;i<sz;i++){
       rr[i] = in[(i+n)%sz];
@@ -394,10 +392,10 @@ void PermTester(const functor &func)
   
   int Nsimd = vec::Nsimd();
 
-  std::vector<scal> input1(Nsimd);
-  std::vector<scal> input2(Nsimd);
-  std::vector<scal> result(Nsimd);
-  std::vector<scal> reference(Nsimd);
+  ExtractBuffer<scal> input1(Nsimd);
+  ExtractBuffer<scal> input2(Nsimd);
+  ExtractBuffer<scal> result(Nsimd);
+  ExtractBuffer<scal> reference(Nsimd);
 
   std::vector<vec,alignedAllocator<vec> > buf(3);
   vec & v_input1 = buf[0];
@@ -460,14 +458,14 @@ void ExchangeTester(const functor &func)
   
   int Nsimd = vec::Nsimd();
 
-  std::vector<scal> input1(Nsimd);
-  std::vector<scal> input2(Nsimd);
-  std::vector<scal> result1(Nsimd);
-  std::vector<scal> result2(Nsimd);
-  std::vector<scal> reference1(Nsimd);
-  std::vector<scal> reference2(Nsimd);
-  std::vector<scal> test1(Nsimd);
-  std::vector<scal> test2(Nsimd);
+  ExtractBuffer<scal> input1(Nsimd);
+  ExtractBuffer<scal> input2(Nsimd);
+  ExtractBuffer<scal> result1(Nsimd);
+  ExtractBuffer<scal> result2(Nsimd);
+  ExtractBuffer<scal> reference1(Nsimd);
+  ExtractBuffer<scal> reference2(Nsimd);
+  ExtractBuffer<scal> test1(Nsimd);
+  ExtractBuffer<scal> test2(Nsimd);
 
   std::vector<vec,alignedAllocator<vec> > buf(6);
   vec & v_input1 = buf[0];
@@ -513,6 +511,7 @@ void ExchangeTester(const functor &func)
       }
     }
     //    assert(found==1);
+    assert(found==1||found==0);
   }
   for(int i=0;i<Nsimd;i++){
     int found=0;
@@ -523,6 +522,7 @@ void ExchangeTester(const functor &func)
       }
     }
     //    assert(found==1);
+    assert(found==1||found==0);
   }
 
   /*
@@ -547,9 +547,25 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  auto latt_size   = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
+
+  {
+    std::cout << " Constructing Test({1,2,3,4,5,6}) " << std::endl;
+    Coordinate Test({1,2,3,4,5,6});
+    std::cout << " Test({1,2,3,4,5,6}) = " << Test <<std::endl;
+  }
+  /*
+  {
+    Coordinate Test =  {1,2,3,4} ;
+    std::cout << " Test = {1,2,3,4} " << Test <<std::endl;
+  }
+  {
+    Coordinate Test {1,2,3,4};
+    std::cout << " Test {1,2,3,4} " << Test <<std::endl;
+  }
+  */
     
   GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
   std::vector<int> seeds({1,2,3,4});
diff --git a/tests/Test_stencil.cc b/tests/Test_stencil.cc
index e9517446..0bd97fea 100644
--- a/tests/Test_stencil.cc
+++ b/tests/Test_stencil.cc
@@ -31,7 +31,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main(int argc, char ** argv) {
   Grid_init(&argc, &argv);
@@ -41,9 +41,9 @@ int main(int argc, char ** argv) {
   typedef typename Field::vector_object vobj;
   typedef typename vobj::scalar_object sobj;
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  auto latt_size   = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
 
   double volume = latt_size[0]*latt_size[1]*latt_size[2]*latt_size[3];
 
@@ -61,7 +61,7 @@ int main(int argc, char ** argv) {
   Field Diff(&Fine);
   LatticeComplex lex(&Fine);
 
-  lex = zero;
+  lex = Zero();
   random(fRNG,Foo);
   gaussian(fRNG,Bar);
 
@@ -70,7 +70,6 @@ int main(int argc, char ** argv) {
   }
   Integer stride =1000;
   {
-    double nrm;
     LatticeComplex coor(&Fine);
 
     for(int d=0;d<Nd;d++){
@@ -81,7 +80,7 @@ int main(int argc, char ** argv) {
     Foo=lex;
   }
 
-  typedef CartesianStencil<vobj,vobj> Stencil;
+  typedef CartesianStencil<vobj,vobj,int> Stencil;
     for(int dir=0;dir<4;dir++){
       for(int disp=0;disp<Fine._fdimensions[dir];disp++){
 
@@ -91,8 +90,8 @@ int main(int argc, char ** argv) {
 	std::vector<int> directions(npoint,dir);
 	std::vector<int> displacements(npoint,disp);
 
-	Stencil myStencil(&Fine,npoint,0,directions,displacements);
-	std::vector<int> ocoor(4);
+	Stencil myStencil(&Fine,npoint,0,directions,displacements,0);
+	Coordinate ocoor(4);
 	for(int o=0;o<Fine.oSites();o++){
 	  Fine.oCoorFromOindex(ocoor,o);
 	  ocoor[dir]=(ocoor[dir]+disp)%Fine._rdimensions[dir];
@@ -104,20 +103,22 @@ int main(int argc, char ** argv) {
 	Bar = Cshift(Foo,dir,disp);
 
 	// Implement a stencil code that should agree with cshift!
-	for(int i=0;i<Check._grid->oSites();i++){
+	for(int i=0;i<Check.Grid()->oSites();i++){
 	  
 	  int permute_type;
 	  StencilEntry *SE;
 	  SE = myStencil.GetEntry(permute_type,0,i);
 	  
+	  auto check = Check.View();
+	  auto foo   = Foo.View();
 	  if ( SE->_is_local && SE->_permute )
-	    permute(Check._odata[i],Foo._odata[SE->_offset],permute_type);
+	    permute(check[i],foo[SE->_offset],permute_type);
 	  else if (SE->_is_local)
-	    Check._odata[i] = Foo._odata[SE->_offset];
+	    check[i] = foo[SE->_offset];
 	  else { 
-	    Check._odata[i] = myStencil.CommBuf()[SE->_offset];
-	    //	    std::cout << " receive "<<i<<" " << Check._odata[i]<<std::endl;
-	    //	    std::cout << " Foo     "<<i<<" " <<   Foo._odata[i]<<std::endl;
+	    check[i] = myStencil.CommBuf()[SE->_offset];
+	    //	    std::cout << " receive "<<i<<" " << Check[i]<<std::endl;
+	    //	    std::cout << " Foo     "<<i<<" " <<   Foo[i]<<std::endl;
 	  }
 	}
 
@@ -127,7 +128,7 @@ int main(int argc, char ** argv) {
 	Real nrm  = norm2(Diff);
 	std::cout<<GridLogMessage<<"N2diff ="<<nrm<<" "<<nrmC<<" " <<nrmB<<std::endl;
 
-	std::vector<int> coor(4);
+	Coordinate coor(4);
 	for(coor[3]=0;coor[3]<latt_size[3]/mpi_layout[3];coor[3]++){
 	for(coor[2]=0;coor[2]<latt_size[2]/mpi_layout[2];coor[2]++){
 	for(coor[1]=0;coor[1]<latt_size[1]/mpi_layout[1];coor[1]++){
@@ -150,8 +151,10 @@ int main(int argc, char ** argv) {
 	}}}}
 
 	if (nrm > 1.0e-4) {
-	  for(int i=0;i<Check._odata.size();i++){
-	    std::cout << i<<" Check.odata "<<Check._odata[i]<< "\n"<<i<<" Bar.odata "<<Bar._odata[i]<<std::endl;
+	  auto check = Check.View();
+	  auto bar   = Bar.View();
+	  for(int i=0;i<check.size();i++){
+	    std::cout << i<<" Check "<<check[i]<< "\n"<<i<<" Bar "<<bar[i]<<std::endl;
 	  }
 	}
 	if (nrm > 1.0e-4) exit(-1);
@@ -178,10 +181,10 @@ int main(int argc, char ** argv) {
 	std::vector<int> directions(npoint,dir);
 	std::vector<int> displacements(npoint,disp);
 
-	Stencil EStencil(&rbFine,npoint,Even,directions,displacements);
-	Stencil OStencil(&rbFine,npoint,Odd,directions,displacements);
+	Stencil EStencil(&rbFine,npoint,Even,directions,displacements,0);
+	Stencil OStencil(&rbFine,npoint,Odd,directions,displacements,0);
 
-	std::vector<int> ocoor(4);
+	Coordinate ocoor(4);
 	for(int o=0;o<Fine.oSites();o++){
 	  Fine.oCoorFromOindex(ocoor,o);
 	  ocoor[dir]=(ocoor[dir]+disp)%Fine._rdimensions[dir];
@@ -192,41 +195,45 @@ int main(int argc, char ** argv) {
 	Bar = Cshift(Foo,dir,disp);
 
 	if ( disp & 0x1 ) {
-	  ECheck.checkerboard = Even;
-	  OCheck.checkerboard = Odd;
+	  ECheck.Checkerboard() = Even;
+	  OCheck.Checkerboard() = Odd;
 	} else {
-	  ECheck.checkerboard = Odd;
-	  OCheck.checkerboard = Even;
+	  ECheck.Checkerboard() = Odd;
+	  OCheck.Checkerboard() = Even;
 	}
 
 	// Implement a stencil code that should agree with that darn cshift!
 	EStencil.HaloExchange(EFoo,compress);
-	for(int i=0;i<OCheck._grid->oSites();i++){
+	for(int i=0;i<OCheck.Grid()->oSites();i++){
 	  int permute_type;
 	  StencilEntry *SE;
 	  SE = EStencil.GetEntry(permute_type,0,i);
 	  //	  std::cout << "Even source "<< i<<" -> " <<SE->_offset << " "<< SE->_is_local<<std::endl;
 
+	  auto ocheck = OCheck.View();
+	  auto efoo   = EFoo.View();
 	  if ( SE->_is_local && SE->_permute )
-	    permute(OCheck._odata[i],EFoo._odata[SE->_offset],permute_type);
+	    permute(ocheck[i],efoo[SE->_offset],permute_type);
 	  else if (SE->_is_local)
-	    OCheck._odata[i] = EFoo._odata[SE->_offset];
+	    ocheck[i] = efoo[SE->_offset];
 	  else
-	    OCheck._odata[i] = EStencil.CommBuf()[SE->_offset];
+	    ocheck[i] = EStencil.CommBuf()[SE->_offset];
 	}
 	OStencil.HaloExchange(OFoo,compress);
-	for(int i=0;i<ECheck._grid->oSites();i++){
+	for(int i=0;i<ECheck.Grid()->oSites();i++){
 	  int permute_type;
 	  StencilEntry *SE;
 	  SE = OStencil.GetEntry(permute_type,0,i);
 	  //	  std::cout << "ODD source "<< i<<" -> " <<SE->_offset << " "<< SE->_is_local<<std::endl;
 
+	  auto echeck = ECheck.View();
+	  auto ofoo   = OFoo.View();
 	  if ( SE->_is_local && SE->_permute )
-	    permute(ECheck._odata[i],OFoo._odata[SE->_offset],permute_type);
+	    permute(echeck[i],ofoo[SE->_offset],permute_type);
 	  else if (SE->_is_local)
-	    ECheck._odata[i] = OFoo._odata[SE->_offset];
+	    echeck[i] = ofoo[SE->_offset];
 	  else
-	    ECheck._odata[i] = OStencil.CommBuf()[SE->_offset];
+	    echeck[i] = OStencil.CommBuf()[SE->_offset];
 	}
 
 	setCheckerboard(Check,ECheck);
@@ -238,7 +245,7 @@ int main(int argc, char ** argv) {
 	Real nrm  = norm2(Diff);
 	std::cout<<GridLogMessage<<"RB N2diff ="<<nrm<<" "<<nrmC<<" " <<nrmB<<std::endl;
 
-	std::vector<int> coor(4);
+	Coordinate coor(4);
 	for(coor[3]=0;coor[3]<latt_size[3]/mpi_layout[3];coor[3]++){
 	for(coor[2]=0;coor[2]<latt_size[2]/mpi_layout[2];coor[2]++){
 	for(coor[1]=0;coor[1]<latt_size[1]/mpi_layout[1];coor[1]++){
diff --git a/tests/core/Test_GaugeAction.cc b/tests/core/Test_GaugeAction.cc
index 572f19fb..fcac6935 100644
--- a/tests/core/Test_GaugeAction.cc
+++ b/tests/core/Test_GaugeAction.cc
@@ -31,7 +31,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 
 /* For Metropolis */
@@ -59,10 +59,10 @@ int main (int argc, char ** argv)
   Grid_init(&argc,&argv);
 
 
-  std::vector<int> simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
-  std::vector<int> latt_size  ({16,16,16,32});
-  std::vector<int> clatt_size  ({4,4,4,8});
+  Coordinate simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size  ({16,16,16,32});
+  Coordinate clatt_size  ({4,4,4,8});
   int orthodir=3;
   int orthosz =latt_size[orthodir];
     
@@ -84,7 +84,7 @@ int main (int argc, char ** argv)
 
   // Painful ; fix syntactical niceness : to check reader
   LatticeComplex LinkTrace(&Fine);
-  LinkTrace=zero;
+  LinkTrace=Zero();
   for(int mu=0;mu<Nd;mu++){
     LinkTrace = LinkTrace + trace(U[mu]);
   }
@@ -110,7 +110,7 @@ int main (int argc, char ** argv)
 
   LatticeComplex Plaq(&Fine);
   LatticeComplex cPlaq(&Coarse);
-  Plaq = zero;
+  Plaq = Zero();
   for(int mu=1;mu<Nd;mu++){
     for(int nu=0;nu<mu;nu++){
       Plaq = Plaq + trace(PeriodicBC::CovShiftForward(U[mu],mu,U[nu])*adj(PeriodicBC::CovShiftForward(U[nu],nu,U[mu])));
@@ -127,7 +127,7 @@ int main (int argc, char ** argv)
 
 
   TComplex Plaq_T_sum; 
-  Plaq_T_sum=zero;
+  Plaq_T_sum=Zero();
   for(int t=0;t<Nt;t++){
     Plaq_T_sum = Plaq_T_sum+Plaq_T[t];
     Complex Pt=TensorRemove(Plaq_T[t]);
diff --git a/tests/core/Test_RectPlaq.cc b/tests/core/Test_RectPlaq.cc
index 2e9cc832..12d36795 100644
--- a/tests/core/Test_RectPlaq.cc
+++ b/tests/core/Test_RectPlaq.cc
@@ -30,7 +30,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 
 /* For Metropolis */
@@ -54,7 +54,7 @@ public:
 
 void RectPlaq(const std::vector<LatticeColourMatrix> &U, LatticeComplex &RectPlaqValue )
 {
-  RectPlaqValue=zero;
+  RectPlaqValue=Zero();
   // 12 * vol loops
   for(int mu=1;mu<Nd;mu++){
     for(int nu=0;nu<mu;nu++){
@@ -76,10 +76,10 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
-  std::vector<int> latt_size  ({16,16,16,32});
-  std::vector<int> clatt_size  ({4,4,4,8});
+  Coordinate simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size  ({16,16,16,32});
+  Coordinate clatt_size  ({4,4,4,8});
   int orthodir=3;
   int orthosz =latt_size[orthodir];
     
@@ -101,7 +101,7 @@ int main (int argc, char ** argv)
 
   // Painful ; fix syntactical niceness : to check reader
   LatticeComplex LinkTrace(&Fine);
-  LinkTrace=zero;
+  LinkTrace=Zero();
   for(int mu=0;mu<Nd;mu++){
     LinkTrace = LinkTrace + trace(U[mu]);
   }
@@ -109,7 +109,7 @@ int main (int argc, char ** argv)
 
   LatticeComplex Plaq(&Fine);
   LatticeComplex cPlaq(&Coarse);
-  Plaq = zero;
+  Plaq = Zero();
   for(int mu=1;mu<Nd;mu++){
     for(int nu=0;nu<mu;nu++){
       Plaq = Plaq + trace(PeriodicBC::CovShiftForward(U[mu],mu,U[nu])*adj(PeriodicBC::CovShiftForward(U[nu],nu,U[mu])));
@@ -134,7 +134,7 @@ int main (int argc, char ** argv)
  // Rect Plaq Calc Deriv
 
   LatticeComplex RectPlaq_d(&Fine);
-  RectPlaq_d = zero;
+  RectPlaq_d = Zero();
   LatticeColourMatrix ds_U(&Fine);
   LatticeColourMatrix left_2(&Fine);
   LatticeColourMatrix upper_l(&Fine);
@@ -151,7 +151,7 @@ int main (int argc, char ** argv)
   //     // 
   for(int mu=0;mu<Nd;mu++){
 
-    ds_U=zero; // dS / dUmu
+    ds_U=Zero(); // dS / dUmu
 
     for(int nu=0;nu<Nd;nu++){
 
@@ -275,7 +275,7 @@ int main (int argc, char ** argv)
 
 
   TComplex Plaq_T_sum; 
-  Plaq_T_sum=zero;
+  Plaq_T_sum=Zero();
   for(int t=0;t<Nt;t++){
     Plaq_T_sum = Plaq_T_sum+Plaq_T[t];
     Complex Pt=TensorRemove(Plaq_T[t]);
diff --git a/tests/core/Test_cf_coarsen_support.cc b/tests/core/Test_cf_coarsen_support.cc
index 35a8a764..e787905e 100644
--- a/tests/core/Test_cf_coarsen_support.cc
+++ b/tests/core/Test_cf_coarsen_support.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -60,8 +60,8 @@ int main (int argc, char ** argv)
   GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
 
   LatticeFermion    src(FGrid); random(RNG5,src);
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid); ref=zero;
+  LatticeFermion result(FGrid); result=Zero();
+  LatticeFermion    ref(FGrid); ref=Zero();
   LatticeFermion    tmp(FGrid);
   LatticeFermion    err(FGrid);
   LatticeGaugeField Umu(UGrid); SU3::HotConfiguration(RNG4,Umu);
diff --git a/tests/core/Test_checker.cc b/tests/core/Test_checker.cc
index 56b0eb60..9d15f2cb 100644
--- a/tests/core/Test_checker.cc
+++ b/tests/core/Test_checker.cc
@@ -30,7 +30,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -56,7 +56,7 @@ int main (int argc, char ** argv)
 
   assert(argc >= 5);
   
-  std::vector<int> latt(4,0);
+  Coordinate latt(4,0);
   latt[0] = toint(argv[1]);
   latt[1] = toint(argv[2]);
   latt[2] = toint(argv[3]);
@@ -65,7 +65,7 @@ int main (int argc, char ** argv)
   const int Ls= toint(argv[5]);
   
   std::cout << "Lattice size (" << latt[0] << "," << latt[1] << "," << latt[2] << "," << latt[3] << ") Ls=" << Ls << std::endl;
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
   std::cout << "SIMD layout (" << simd_layout[0] << "," << simd_layout[1] << "," << simd_layout[2] << "," << simd_layout[3] << ")" << std::endl;
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(latt, simd_layout,GridDefaultMpi());
   GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
@@ -85,8 +85,8 @@ int main (int argc, char ** argv)
   LatticeType src_o(FrbGrid);
   pickCheckerboard(Odd,src_o,src);
 
-  std::vector<int> site(5);
-  std::vector<int> cbsite(5);
+  Coordinate site(5);
+  Coordinate cbsite(5);
   typedef typename GridTypeMapper<LatticeType::vector_object>::scalar_object sobj;
 
   // std::cout << "sizeof(vobj) " << sizeof(LatticeType::vector_object) << std::endl;
@@ -98,7 +98,7 @@ int main (int argc, char ** argv)
       for(site[3]=0;site[3]<latt[2];site[3]++){
 	for(site[2]=0;site[2]<latt[1];site[2]++){
 	  for(site[1]=0;site[1]<latt[0];site[1]++){
-	    if(src_o._grid->CheckerBoard(site) != src_o.checkerboard)
+	    if(src_o.Grid()->CheckerBoard(site) != src_o.Checkerboard())
 	      continue;
 
 	    std::cout << "Site (" << site[0] << "," << site[1] << "," << site[2] << "," << site[3] << "," << site[4] << ")" << std::endl;
@@ -128,7 +128,7 @@ int main (int argc, char ** argv)
 
 
   
-  // LatticeFermion result(FGrid); result=zero;
+  // LatticeFermion result(FGrid); result=Zero();
   // LatticeGaugeField Umu(UGrid); 
 
   // SU3::HotConfiguration(RNG4,Umu);
@@ -145,7 +145,7 @@ int main (int argc, char ** argv)
   // LatticeFermion    src_o(FrbGrid);
   // LatticeFermion result_o(FrbGrid);
   // pickCheckerboard(Odd,src_o,src);
-  // result_o=zero;
+  // result_o=Zero();
 
   // SchurDiagMooeeOperator<DomainWallFermionR,LatticeFermion> HermOpEO(Ddwf);
   // ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
diff --git a/tests/core/Test_contfrac_even_odd.cc b/tests/core/Test_contfrac_even_odd.cc
index 9ad0be5a..25affd00 100644
--- a/tests/core/Test_contfrac_even_odd.cc
+++ b/tests/core/Test_contfrac_even_odd.cc
@@ -29,7 +29,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 template<class d>
 struct scal {
@@ -107,10 +106,10 @@ void  TestWhat(What & Ddwf,
   LatticeFermion src   (FGrid); random(*RNG5,src);
   LatticeFermion phi   (FGrid); random(*RNG5,phi);
   LatticeFermion chi   (FGrid); random(*RNG5,chi);
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid);    ref=zero;
-  LatticeFermion    tmp(FGrid);    tmp=zero;
-  LatticeFermion    err(FGrid);    tmp=zero;
+  LatticeFermion result(FGrid); result=Zero();
+  LatticeFermion    ref(FGrid);    ref=Zero();
+  LatticeFermion    tmp(FGrid);    tmp=Zero();
+  LatticeFermion    err(FGrid);    tmp=Zero();
 
   LatticeFermion src_e (FrbGrid);
   LatticeFermion src_o (FrbGrid);
diff --git a/tests/core/Test_cshift_red_black.cc b/tests/core/Test_cshift_red_black.cc
index c7b0c2f1..34325072 100644
--- a/tests/core/Test_cshift_red_black.cc
+++ b/tests/core/Test_cshift_red_black.cc
@@ -30,7 +30,6 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
 
 #define POWER10
 
@@ -38,12 +37,12 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
+  Coordinate latt_size   = GridDefaultLatt();
   int Nd = latt_size.size();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
-  std::vector<int> mask(Nd,1);
+  Coordinate mask(Nd,1);
   mask[0]=0;
 
   GridCartesian         Fine  (latt_size,simd_layout,mpi_layout);
@@ -60,10 +59,9 @@ int main (int argc, char ** argv)
   LatticeComplex ShiftUe(&RBFine);
   LatticeComplex ShiftUo(&RBFine);
   LatticeComplex lex(&Fine);
-  lex=zero;
+  lex=Zero();
   Integer stride =1;
   {
-    double nrm;
     LatticeComplex coor(&Fine);
 
     for(int d=0;d<Nd;d++){
@@ -84,10 +82,11 @@ int main (int argc, char ** argv)
   pickCheckerboard(Odd,Uo,U);
 
   //  std::cout<<GridLogMessage << U<<std::endl;
+  std::cout<<GridLogMessage<< U <<std::endl;
+  std::cout<<GridLogMessage << "U " <<norm2(U)<<std::endl;
   std::cout<<GridLogMessage << "Ue " <<norm2(Ue)<<std::endl;
   std::cout<<GridLogMessage << "Uo " <<norm2(Uo)<<std::endl;
 
-
   TComplex cm;
   TComplex cmeo;
   for(int dir=0;dir<Nd;dir++){
@@ -116,7 +115,7 @@ int main (int argc, char ** argv)
 	err = ShiftU - rbShiftU;
 	std::cout<< "\terror " <<norm2(err)<<std::endl;
 
-	std::vector<int> coor(4);
+	Coordinate coor(4);
 
 	std::cout<<GridLogMessage << "  Checking the non-checkerboard shift "<<shift <<" dir "<<dir <<" ... ";
 	for(coor[3]=0;coor[3]<latt_size[3];coor[3]++){
@@ -128,17 +127,17 @@ int main (int argc, char ** argv)
 
 	  /////////	  double nrm=norm2(U);
 
-	  std::vector<int> scoor(coor);
+	  Coordinate scoor(coor);
 	  scoor[dir] = (scoor[dir]+shift)%latt_size[dir];
 	  
 #ifndef POWER10
-	  std::vector<int> powers=latt_size;
+	  Coordinate powers=latt_size;
 	  Integer slex = scoor[0]
 	    + latt_size[0]*scoor[1]
 	    + latt_size[0]*latt_size[1]*scoor[2]
 	    + latt_size[0]*latt_size[1]*latt_size[2]*scoor[3];
 #else
-	  std::vector<int> powers({1,10,100,1000});
+	  Coordinate powers({1,10,100,1000});
 	  Integer slex = scoor[0]
 	    + 10        *scoor[1]
 	    + 100       *scoor[2]
@@ -147,7 +146,7 @@ int main (int argc, char ** argv)
 	  Complex scm(slex);
 	  
 	  double nrm = abs(scm-cm()()());
-	  std::vector<int> peer(4);
+	  Coordinate peer(4);
 	  Complex ctmp = cm;
 	  Integer index=real(ctmp);
 	  Lexicographic::CoorFromIndex(peer,index,powers);
@@ -176,23 +175,23 @@ int main (int argc, char ** argv)
 
 	  Integer checkerboard = RBFine.CheckerBoard(coor);
 
-	  if ( checkerboard == ShiftUo.checkerboard ) {
+	  if ( checkerboard == ShiftUo.Checkerboard() ) {
 	    peekSite(cmeo,ShiftUo,coor);
 	  } else { 
 	    peekSite(cmeo,ShiftUe,coor);
 	  }
 
-	  std::vector<int> scoor(coor);
+	  Coordinate scoor(coor);
 	  scoor[dir] = (scoor[dir]+shift)%latt_size[dir];
 	  
 #ifndef POWER10
-	  std::vector<int> powers=latt_size;
+	  Coordinate powers=latt_size;
 	  Integer slex = scoor[0]
 	    + latt_size[0]*scoor[1]
 	    + latt_size[0]*latt_size[1]*scoor[2]
 	    + latt_size[0]*latt_size[1]*latt_size[2]*scoor[3];
 #else 
-	  std::vector<int> powers({1,10,100,1000});
+	  Coordinate powers({1,10,100,1000});
 	  Integer slex = scoor[0]
 	    + 10        *scoor[1]
 	    + 100       *scoor[2]
@@ -200,7 +199,7 @@ int main (int argc, char ** argv)
 #endif
 	  Complex scm(slex);
 
-	  std::vector<int> peer(4);
+	  Coordinate peer(4);
 	  Complex ctmp=cmeo;
 	  Integer index=real(ctmp);
 	  Lexicographic::CoorFromIndex(peer,index,powers);
@@ -210,7 +209,7 @@ int main (int argc, char ** argv)
 
 	    std::cout << " coor "<<" ["<<coor[0]<<","<<coor[1]<<","<<coor[2]<<","<<coor[3]<<"] \n ";
 	    std::cout << "shift "<< shift <<" dir "<<dir<< " checker board "<< checkerboard << " ";
-	    std::cout << "Uo cb = "   << ShiftUo.checkerboard << " Ue cb= "<<ShiftUe.checkerboard<<std::endl;
+	    std::cout << "Uo cb = "   << ShiftUo.Checkerboard() << " Ue cb= "<<ShiftUe.Checkerboard()<<std::endl;
 
 	    std::cout<<"EOFAIL shift "<< shift<<" in dir "<< dir
 		     <<" ["<<coor[0]<<","<<coor[1]<<","<<coor[2]<<","<<coor[3]<<"] = "
diff --git a/tests/core/Test_cshift_red_black_rotate.cc b/tests/core/Test_cshift_red_black_rotate.cc
index aa9e6104..ce81999b 100644
--- a/tests/core/Test_cshift_red_black_rotate.cc
+++ b/tests/core/Test_cshift_red_black_rotate.cc
@@ -30,7 +30,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 #define POWER10
 
@@ -38,12 +38,12 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
+  Coordinate latt_size   = GridDefaultLatt();
   int Nd = latt_size.size();
-  std::vector<int> simd_layout( { vComplex::Nsimd(),1,1,1});
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate simd_layout( { vComplex::Nsimd(),1,1,1});
+  Coordinate mpi_layout  = GridDefaultMpi();
 
-  std::vector<int> mask(Nd,1);
+  Coordinate mask(Nd,1);
   mask[0]=0;
 
   GridCartesian         Fine  (latt_size,simd_layout,mpi_layout);
@@ -60,10 +60,9 @@ int main (int argc, char ** argv)
   LatticeComplex ShiftUe(&RBFine);
   LatticeComplex ShiftUo(&RBFine);
   LatticeComplex lex(&Fine);
-  lex=zero;
+  lex=Zero();
   Integer stride =1;
   {
-    double nrm;
     LatticeComplex coor(&Fine);
 
     for(int d=0;d<Nd;d++){
@@ -117,7 +116,7 @@ int main (int argc, char ** argv)
 	err = ShiftU - rbShiftU;
 	std::cout<< "\terror " <<norm2(err)<<std::endl;
 
-	std::vector<int> coor(4);
+	Coordinate coor(4);
 
 	std::cout<<GridLogMessage << "  Checking the non-checkerboard shift "<< shift << " dir "<<dir  <<"... ";
 	for(coor[3]=0;coor[3]<latt_size[3];coor[3]++){
@@ -129,17 +128,17 @@ int main (int argc, char ** argv)
 
 	  /////////	  double nrm=norm2(U);
 
-	  std::vector<int> scoor(coor);
+	  Coordinate scoor(coor);
 	  scoor[dir] = (scoor[dir]+shift)%latt_size[dir];
 
 #ifdef POWER10
-	  std::vector<int> powers({1,10,100,1000});
+	  Coordinate powers({1,10,100,1000});
 	  Integer slex = scoor[3]
 	    + 10        *scoor[2]
 	    + 100       *scoor[1]
 	    + 1000      *scoor[0];
 #else
-	  std::vector<int> powers=latt_size;
+	  Coordinate powers=latt_size;
 	  Integer slex = scoor[0]
 	    + latt_size[0]*scoor[1]
 	    + latt_size[0]*latt_size[1]*scoor[2]
@@ -149,7 +148,7 @@ int main (int argc, char ** argv)
 	  Complex scm(slex);
 	  
 	  double nrm = abs(scm-cm()()());
-	  std::vector<int> peer(4);
+	  Coordinate peer(4);
 	  Complex ctmp = cm;
 	  Integer index=real(ctmp);
 	  Lexicographic::CoorFromIndex(peer,index,powers);
@@ -181,25 +180,25 @@ int main (int argc, char ** argv)
 
 	  //	  std::cout << " coor "<<" ["<<coor[0]<<","<<coor[1]<<","<<coor[2]<<","<<coor[3]<<"] \n ";
 	  //	  std::cout << "shift "<< shift <<" dir "<<dir<< " checker board "<< checkerboard << " ";
-	  //	  std::cout << "Uo "   << ShiftUo.checkerboard << " Ue "<<ShiftUe.checkerboard<<std::endl;
-	  if ( checkerboard == ShiftUo.checkerboard ) {
+	  //	  std::cout << "Uo "   << ShiftUo.Checkerboard() << " Ue "<<ShiftUe.Checkerboard()<<std::endl;
+	  if ( checkerboard == ShiftUo.Checkerboard() ) {
 	    peekSite(cmeo,ShiftUo,coor);
 	  } else { 
 	    peekSite(cmeo,ShiftUe,coor);
 	  }
 
 
-	  std::vector<int> scoor(coor);
+	  Coordinate scoor(coor);
 	  scoor[dir] = (scoor[dir]+shift)%latt_size[dir];
 
 #ifdef POWER10	  
-	  std::vector<int> powers({1,10,100,1000});
+	  Coordinate powers({1,10,100,1000});
 	  Integer slex = scoor[3]
 	    + 10        *scoor[2]
 	    + 100       *scoor[1]
 	    + 1000      *scoor[0];
 #else
-	  std::vector<int> powers = latt_size;
+	  Coordinate powers = latt_size;
 	  Integer slex = scoor[0]
 	    + latt_size[0]*scoor[1]
 	    + latt_size[0]*latt_size[1]*scoor[2]
@@ -207,7 +206,7 @@ int main (int argc, char ** argv)
 #endif
 	  Complex scm(slex);
 
-	  std::vector<int> peer(4);
+	  Coordinate peer(4);
 	  Complex ctmp=cmeo;
 	  Integer index=real(ctmp);
 	  Lexicographic::CoorFromIndex(peer,index,powers);
diff --git a/tests/core/Test_cshift_rotate.cc b/tests/core/Test_cshift_rotate.cc
index 64c08892..9e572b70 100644
--- a/tests/core/Test_cshift_rotate.cc
+++ b/tests/core/Test_cshift_rotate.cc
@@ -29,15 +29,15 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout( { vComplex::Nsimd(),1,1,1});
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout( { vComplex::Nsimd(),1,1,1});
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   GridCartesian        Fine(latt_size,simd_layout,mpi_layout);
 
@@ -46,10 +46,9 @@ int main (int argc, char ** argv)
   LatticeComplex U(&Fine);
   LatticeComplex ShiftU(&Fine);
   LatticeComplex lex(&Fine);
-  lex=zero;
+  lex=Zero();
   Integer stride =1;
   {
-    double nrm;
     LatticeComplex coor(&Fine);
 
     for(int d=0;d<4;d++){
@@ -76,7 +75,7 @@ int main (int argc, char ** argv)
 	std::cout << "ShiftU[0]" << ShiftU[0]<<std::endl;
 	std::cout << "ShiftU[1]" << ShiftU[1]<<std::endl;
 	*/
-	std::vector<int> coor(4);
+	Coordinate coor(4);
 
 	for(coor[3]=0;coor[3]<latt_size[3];coor[3]++){
 	for(coor[2]=0;coor[2]<latt_size[2];coor[2]++){
@@ -87,7 +86,7 @@ int main (int argc, char ** argv)
 
 	  double nrm=norm2(U);
 
-	  std::vector<int> scoor(coor);
+	  Coordinate scoor(coor);
 	  scoor[dir] = (scoor[dir]+shift)%latt_size[dir];
 	  
 	  Integer slex = scoor[0]
@@ -98,7 +97,7 @@ int main (int argc, char ** argv)
 	  Complex scm(slex);
 	  
 	  nrm = abs(scm-cm()()());
-	  std::vector<int> peer(4);
+	  Coordinate peer(4);
 	  Complex tmp  =cm;
 	  Integer index=real(tmp);
 	  Lexicographic::CoorFromIndex(peer,index,latt_size);
diff --git a/tests/core/Test_dwf_eofa_even_odd.cc b/tests/core/Test_dwf_eofa_even_odd.cc
index 5fe0f653..1d2f2909 100644
--- a/tests/core/Test_dwf_eofa_even_odd.cc
+++ b/tests/core/Test_dwf_eofa_even_odd.cc
@@ -32,7 +32,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -69,18 +69,18 @@ int main (int argc, char ** argv)
     LatticeFermion    src   (FGrid); random(RNG5, src);
     LatticeFermion    phi   (FGrid); random(RNG5, phi);
     LatticeFermion    chi   (FGrid); random(RNG5, chi);
-    LatticeFermion    result(FGrid); result = zero;
-    LatticeFermion    ref   (FGrid); ref = zero;
-    LatticeFermion    tmp   (FGrid); tmp = zero;
-    LatticeFermion    err   (FGrid); err = zero;
+    LatticeFermion    result(FGrid); result = Zero();
+    LatticeFermion    ref   (FGrid); ref = Zero();
+    LatticeFermion    tmp   (FGrid); tmp = Zero();
+    LatticeFermion    err   (FGrid); err = Zero();
     LatticeGaugeField Umu   (UGrid); SU3::HotConfiguration(RNG4, Umu);
     std::vector<LatticeColourMatrix> U(4,UGrid);
 
     // Only one non-zero (y)
-    Umu = zero;
+    Umu = Zero();
     for(int nn=0; nn<Nd; nn++){
         random(RNG4, U[nn]);
-        if(nn>0){ U[nn] = zero; }
+        if(nn>0){ U[nn] = Zero(); }
         PokeIndex<LorentzIndex>(Umu, U[nn], nn);
     }
 
diff --git a/tests/core/Test_dwf_even_odd.cc b/tests/core/Test_dwf_even_odd.cc
index 0424301c..d654e588 100644
--- a/tests/core/Test_dwf_even_odd.cc
+++ b/tests/core/Test_dwf_even_odd.cc
@@ -30,7 +30,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -68,19 +68,19 @@ int main (int argc, char ** argv)
   LatticeFermion src   (FGrid); random(RNG5,src);
   LatticeFermion phi   (FGrid); random(RNG5,phi);
   LatticeFermion chi   (FGrid); random(RNG5,chi);
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid);    ref=zero;
-  LatticeFermion    tmp(FGrid);    tmp=zero;
-  LatticeFermion    err(FGrid);    tmp=zero;
+  LatticeFermion result(FGrid); result=Zero();
+  LatticeFermion    ref(FGrid);    ref=Zero();
+  LatticeFermion    tmp(FGrid);    tmp=Zero();
+  LatticeFermion    err(FGrid);    tmp=Zero();
   LatticeGaugeField Umu(UGrid); SU3::HotConfiguration(RNG4,Umu);
   std::vector<LatticeColourMatrix> U(4,UGrid);
 
   // Only one non-zero (y)
-  Umu=zero;
+  Umu=Zero();
   for(int nn=0;nn<Nd;nn++){
     random(RNG4,U[nn]);
     if ( nn>0 ) 
-      U[nn]=zero;
+      U[nn]=Zero();
     PokeIndex<LorentzIndex>(Umu,U[nn],nn);
   }
 
diff --git a/tests/core/Test_dwf_rb5d.cc b/tests/core/Test_dwf_rb5d.cc
deleted file mode 100644
index 68c1755e..00000000
--- a/tests/core/Test_dwf_rb5d.cc
+++ /dev/null
@@ -1,136 +0,0 @@
-    /*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid 
-
-    Source file: ./tests/Test_dwf_even_odd.cc
-
-    Copyright (C) 2015
-
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: paboyle <paboyle@ph.ed.ac.uk>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-    /*  END LEGAL */
-#include <Grid/Grid.h>
-
-using namespace std;
-using namespace Grid;
-using namespace Grid::QCD;
-
-template<class d>
-struct scal {
-  d internal;
-};
-
-  Gamma::Algebra Gmu [] = {
-    Gamma::Algebra::GammaX,
-    Gamma::Algebra::GammaY,
-    Gamma::Algebra::GammaZ,
-    Gamma::Algebra::GammaT
-  };
-
-typedef WilsonFermion5D<DomainWallVec5dImplR> WilsonFermion5DR;
-typedef WilsonFermion5D<DomainWallVec5dImplF> WilsonFermion5DF;
-typedef WilsonFermion5D<DomainWallVec5dImplD> WilsonFermion5DD;
-
-typedef WilsonFermion5D<WilsonImplR> WilsonFermion5D_OKR;
-
-int main (int argc, char ** argv)
-{
-  Grid_init(&argc,&argv);
-
-  int threads = GridThread::GetThreads();
-  std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
-
-
-  const int Ls=32;
-
-  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
-  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
-
-  GridCartesian         * sUGrid   = SpaceTimeGrid::makeFourDimDWFGrid(GridDefaultLatt(),GridDefaultMpi());
-  GridRedBlackCartesian * sUrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(sUGrid);
-
-  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimDWFGrid(Ls,UGrid);
-  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimDWFRedBlackGrid(Ls,UGrid);
-
-  std::vector<int> seeds4({1,2,3,4});
-  std::vector<int> seeds5({5,6,7,8});
-
-  GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
-  GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
-
-  LatticeFermion src   (FGrid); random(RNG5,src);
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid);    ref=zero;
-  LatticeFermion    tmp(FGrid);
-  LatticeFermion    err(FGrid);
-
-  LatticeGaugeField Umu(UGrid); SU3::HotConfiguration(RNG4,Umu);
-  std::vector<LatticeColourMatrix> U(4,UGrid);
-
-  // Only one non-zero (y)
-  /*
-  Umu=zero;
-  for(int nn=0;nn<Nd;nn++){
-    random(RNG4,U[nn]);
-    PokeIndex<LorentzIndex>(Umu,U[nn],nn);
-  }
-  */
-
-  RealD mass=0.1;
-  RealD M5  =1.8;
-  typename WilsonFermion5DR::ImplParams params;
-
-  WilsonFermion5DR Dw(Umu,*FGrid,*FrbGrid,*sUGrid,*sUrbGrid,M5,params);
-
-  Dw.Dhop(src,result,0);
-
-  std::cout << "Norm src = "<<norm2(src)<<" Norm res = "<<norm2(result) << std::endl;
-
-  GridCartesian         * FokGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
-  GridRedBlackCartesian * FokrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
-  WilsonFermion5D_OKR Dok(Umu,*FokGrid,*FokrbGrid,*UGrid,*UrbGrid,M5,params);
-  
-  LatticeFermion src_ok   (FokGrid);
-  LatticeFermion ref_ok(FokGrid); 
-  LatticeFermion result_ok(FokGrid);
-  
-  
-  for(int lidx=0;lidx<FGrid->lSites();lidx++){
-    std::vector<int> lcoor;
-    FGrid->LocalIndexToLocalCoor(lidx,lcoor);
-    
-    SpinColourVector siteSrc;
-
-    peekLocalSite(siteSrc,src,lcoor);
-    pokeLocalSite(siteSrc,src_ok,lcoor);
-
-    peekLocalSite(siteSrc,result,lcoor);
-    pokeLocalSite(siteSrc,result_ok,lcoor);
-  }
-  
-  Dok.Dhop(src_ok,ref_ok,0);
-  
-  std::cout << "Reference = "<<norm2(src_ok)<<" res = "<<norm2(ref_ok) << std::endl;
-  ref_ok = ref_ok - result_ok;
-  std::cout << "Result = "<<norm2(result_ok)<< std::endl;
-  std::cout << "Reference diff = "<<norm2(ref_ok)<< std::endl;
-
-  Grid_finalize();
-}
diff --git a/tests/core/Test_fft.cc b/tests/core/Test_fft.cc
index b7f9f000..2ba3752b 100644
--- a/tests/core/Test_fft.cc
+++ b/tests/core/Test_fft.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
@@ -38,9 +38,9 @@ int main (int argc, char ** argv)
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout( { vComplexD::Nsimd(),1,1,1});
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout( { vComplexD::Nsimd(),1,1,1});
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   int vol = 1;
   for(int d=0;d<latt_size.size();d++){
@@ -60,7 +60,7 @@ int main (int argc, char ** argv)
   LatticeSpinMatrixD    S(&GRID);
   LatticeSpinMatrixD    Stilde(&GRID);
   
-  std::vector<int> p({1,3,2,3});
+  Coordinate p({1,3,2,3});
 
   one = ComplexD(1.0,0.0);
   zz  = ComplexD(0.0,0.0);
@@ -71,7 +71,7 @@ int main (int argc, char ** argv)
   std::cout<<"*************************************************"<<std::endl;
   std::cout<<"Testing Fourier from of known plane wave         "<<std::endl;
   std::cout<<"*************************************************"<<std::endl;
-  C=zero;
+  C=Zero();
   for(int mu=0;mu<4;mu++){
     RealD TwoPiL =  M_PI * 2.0/ latt_size[mu];
     LatticeCoordinate(coor,mu);
@@ -80,7 +80,7 @@ int main (int argc, char ** argv)
 
   C = exp(C*ci);
   Csav = C;
-  S=zero;
+  S=Zero();
   S = S+C;
 
   FFT theFFT(&GRID);
@@ -92,12 +92,12 @@ int main (int argc, char ** argv)
   theFFT.FFT_dim(Ctilde,Ctilde,2,FFT::forward); std::cout << theFFT.MFlops()<<" Mflops "<<std::endl;
   theFFT.FFT_dim(Ctilde,Ctilde,3,FFT::forward); std::cout << theFFT.MFlops()<<" Mflops "<<std::endl;
 
-  //  C=zero;
+  //  C=Zero();
   //  Ctilde = where(abs(Ctilde)<1.0e-10,C,Ctilde);
   TComplexD cVol;
   cVol()()() = vol;
 
-  Cref=zero;
+  Cref=Zero();
   pokeSite(cVol,Cref,p);
   //  std::cout <<"Ctilde "<< Ctilde <<std::endl;
   //  std::cout <<"Cref   "<< Cref <<std::endl;
@@ -121,9 +121,9 @@ int main (int argc, char ** argv)
   theFFT.FFT_dim(Stilde,S,3,FFT::forward); std::cout << theFFT.MFlops()<<" mflops "<<std::endl;
 
   SpinMatrixD Sp; 
-  Sp = zero; Sp = Sp+cVol;
+  Sp = Zero(); Sp = Sp+cVol;
 
-  S=zero;
+  S=Zero();
   pokeSite(Sp,S,p);
 
   S= S-Stilde;
@@ -139,7 +139,7 @@ int main (int argc, char ** argv)
   LatticeGaugeFieldD Umu(&GRID);
 
   SU3::ColdConfiguration(pRNG,Umu); // Unit gauge
-  //  Umu=zero;
+  //  Umu=Zero();
   ////////////////////////////////////////////////////
   // Wilson test
   ////////////////////////////////////////////////////
@@ -225,13 +225,13 @@ int main (int argc, char ** argv)
       Gamma::Algebra::GammaT,
       Gamma::Algebra::Gamma5
     };
-    LatticeFermionD    Kinetic(FGrid); Kinetic = zero;
+    LatticeFermionD    Kinetic(FGrid); Kinetic = Zero();
     LatticeComplexD    kmu(FGrid); 
     LatticeInteger     scoor(FGrid); 
-    LatticeComplexD    sk (FGrid); sk = zero;
-    LatticeComplexD    sk2(FGrid); sk2= zero;
-    LatticeComplexD    W(FGrid); W= zero;
-    //      LatticeComplexD    a(FGrid); a= zero;
+    LatticeComplexD    sk (FGrid); sk = Zero();
+    LatticeComplexD    sk2(FGrid); sk2= Zero();
+    LatticeComplexD    W(FGrid); W= Zero();
+    //      LatticeComplexD    a(FGrid); a= Zero();
     LatticeComplexD    one(FGrid); one =ComplexD(1.0,0.0);
     ComplexD ci(0.0,1.0);
     
@@ -294,8 +294,8 @@ int main (int argc, char ** argv)
     LatticeFermionD    ref(&GRID);
     LatticeFermionD    diff(&GRID);
 
-    std::vector<int> point(4,0);
-    src=zero;
+    Coordinate point(4,0);
+    src=Zero();
     SpinColourVectorD ferm; gaussian(sRNG,ferm);
     pokeSite(ferm,src,point);
 
@@ -314,9 +314,9 @@ int main (int argc, char ** argv)
 
     Gamma G5(Gamma::Algebra::Gamma5);
 
-    LatticeFermionD    src5(FGrid); src5=zero;
+    LatticeFermionD    src5(FGrid); src5=Zero();
     LatticeFermionD    tmp5(FGrid); 
-    LatticeFermionD    result5(FGrid); result5=zero;
+    LatticeFermionD    result5(FGrid); result5=Zero();
     LatticeFermionD    result4(&GRID); 
     const int sdir=0;
 
@@ -374,8 +374,8 @@ int main (int argc, char ** argv)
     LatticeFermionD    ref(&GRID);
     LatticeFermionD    diff(&GRID);
 
-    std::vector<int> point(4,0);
-    src=zero;
+    Coordinate point(4,0);
+    src=Zero();
     SpinColourVectorD ferm; gaussian(sRNG,ferm);
     pokeSite(ferm,src,point);
 
@@ -394,9 +394,9 @@ int main (int argc, char ** argv)
 
     Gamma G5(Gamma::Algebra::Gamma5);
 
-    LatticeFermionD    src5(FGrid); src5=zero;
+    LatticeFermionD    src5(FGrid); src5=Zero();
     LatticeFermionD    tmp5(FGrid); 
-    LatticeFermionD    result5(FGrid); result5=zero;
+    LatticeFermionD    result5(FGrid); result5=Zero();
     LatticeFermionD    result4(&GRID); 
     const int sdir=0;
 
@@ -449,8 +449,8 @@ int main (int argc, char ** argv)
     typedef Photon<QEDGimplTypesD>       QEDGaction;
 
     QEDGaction Maxwell(QEDGaction::FEYNMAN_L);
-    QEDGaction::GaugeField Prop(&GRID);Prop=zero;
-    QEDGaction::GaugeField Source(&GRID);Source=zero;
+    QEDGaction::GaugeField Prop(&GRID);Prop=Zero();
+    QEDGaction::GaugeField Source(&GRID);Source=Zero();
 
     Maxwell.FreePropagator (Source,Prop);
     std::cout << " MaxwellFree propagator\n";
diff --git a/tests/core/Test_fft_gfix.cc b/tests/core/Test_fft_gfix.cc
index 4eb6887d..228770a8 100644
--- a/tests/core/Test_fft_gfix.cc
+++ b/tests/core/Test_fft_gfix.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
@@ -39,9 +39,9 @@ int main (int argc, char ** argv)
 
   int threads = GridThread::GetThreads();
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout( { vComplex::Nsimd(),1,1,1});
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout( { vComplex::Nsimd(),1,1,1});
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   int vol = 1;
   for(int d=0;d<latt_size.size();d++){
diff --git a/tests/core/Test_fftf.cc b/tests/core/Test_fftf.cc
index 22838f7b..e5b6f75b 100644
--- a/tests/core/Test_fftf.cc
+++ b/tests/core/Test_fftf.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
@@ -38,9 +38,9 @@ int main (int argc, char ** argv)
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout( { vComplexF::Nsimd(),1,1,1});
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout( { vComplexF::Nsimd(),1,1,1});
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   int vol = 1;
   for(int d=0;d<latt_size.size();d++){
@@ -57,14 +57,14 @@ int main (int argc, char ** argv)
   LatticeSpinMatrixF    S(&Fine);
   LatticeSpinMatrixF    Stilde(&Fine);
   
-  std::vector<int> p({1,2,3,2});
+  Coordinate p({1,2,3,2});
 
   one = ComplexF(1.0,0.0);
   zz  = ComplexF(0.0,0.0);
 
   ComplexF ci(0.0,1.0);
 
-  C=zero;
+  C=Zero();
   for(int mu=0;mu<4;mu++){
     RealD TwoPiL =  M_PI * 2.0/ latt_size[mu];
     LatticeCoordinate(coor,mu);
@@ -73,7 +73,7 @@ int main (int argc, char ** argv)
 
   C = exp(C*ci);
 
-  S=zero;
+  S=Zero();
   S = S+C;
 
   FFT theFFT(&Fine);
@@ -84,12 +84,12 @@ int main (int argc, char ** argv)
   theFFT.FFT_dim(Ctilde,Ctilde,2,FFT::forward); std::cout << theFFT.MFlops()<<std::endl;
   theFFT.FFT_dim(Ctilde,Ctilde,3,FFT::forward); std::cout << theFFT.MFlops()<<std::endl;
 
-  //  C=zero;
+  //  C=Zero();
   //  Ctilde = where(abs(Ctilde)<1.0e-10,C,Ctilde);
   TComplexF cVol;
   cVol()()() = vol;
 
-  C=zero;
+  C=Zero();
   pokeSite(cVol,C,p);
   C=C-Ctilde;
   std::cout << "diff scalar "<<norm2(C) << std::endl;
@@ -101,9 +101,9 @@ int main (int argc, char ** argv)
   theFFT.FFT_dim(Stilde,Stilde,3,FFT::forward); std::cout << theFFT.MFlops()<<" "<<theFFT.USec() <<std::endl;
 
   SpinMatrixF Sp; 
-  Sp = zero; Sp = Sp+cVol;
+  Sp = Zero(); Sp = Sp+cVol;
 
-  S=zero;
+  S=Zero();
   pokeSite(Sp,S,p);
 
   S= S-Stilde;
diff --git a/tests/core/Test_gamma.cc b/tests/core/Test_gamma.cc
index f3805bdb..2658f2c1 100644
--- a/tests/core/Test_gamma.cc
+++ b/tests/core/Test_gamma.cc
@@ -29,9 +29,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
     /*  END LEGAL */
 #include <Grid/Grid.h>
 
-using namespace std;
 using namespace Grid;
-using namespace QCD;
 
 static constexpr double                      tolerance = 1.0e-6;
 static std::array<SpinMatrix, Gamma::nGamma> testAlgebra;
@@ -66,28 +64,28 @@ void createTestAlgebra(void)
   SpinMatrix                testg5;
   const Complex             I(0., 1.), mI(0., -1.);
   
-  testg[0] = zero;
+  testg[0] = Zero();
   testg[0]()(0, 3) = I;
   testg[0]()(1, 2) = I;
   testg[0]()(2, 1) = mI;
   testg[0]()(3, 0) = mI;
   std::cout << GridLogMessage << "test GammaX= " << std::endl;
   print(testg[0]);
-  testg[1] = zero;
+  testg[1] = Zero();
   testg[1]()(0, 3) = -1.;
   testg[1]()(1, 2) = 1.;
   testg[1]()(2, 1) = 1.;
   testg[1]()(3, 0) = -1.;
   std::cout << GridLogMessage << "test GammaY= " << std::endl;
   print(testg[1]);
-  testg[2] = zero;
+  testg[2] = Zero();
   testg[2]()(0, 2) = I;
   testg[2]()(1, 3) = mI;
   testg[2]()(2, 0) = mI;
   testg[2]()(3, 1) = I;
   std::cout << GridLogMessage << "test GammaZ= " << std::endl;
   print(testg[2]);
-  testg[3] = zero;
+  testg[3] = Zero();
   testg[3]()(0, 2) = 1.;
   testg[3]()(1, 3) = 1.;
   testg[3]()(2, 0) = 1.;
@@ -142,7 +140,6 @@ void checkGamma(const Gamma::Algebra a, GridSerialRNG &rng)
   SpinVector v;
   SpinMatrix m, &testg = testAlgebra[a];
   Gamma      g(a);
-  bool       pass = true;
   
   random(rng, v);
   random(rng, m);
@@ -161,7 +158,6 @@ void checkProd(const Gamma::Algebra a, const Gamma::Algebra b)
 {
   SpinMatrix gm, testg = testAlgebra[a]*testAlgebra[b];
   Gamma      g = Gamma(a)*Gamma(b);
-  bool       pass = true;
   
   std::cout << GridLogMessage << "Checking " << Gamma::name[a] << " * "
             << Gamma::name[b] << ": ";
@@ -175,7 +171,6 @@ void checkAdj(const Gamma::Algebra a)
 {
   SpinMatrix gm, testg = adj(testAlgebra[a]);
   Gamma      g(adj(Gamma(a)));
-  bool       pass = true;
   
   std::cout << GridLogMessage << "Checking adj(" << Gamma::name[a] << "): ";
   gm = 1.0;
@@ -217,7 +212,6 @@ void checkGammaL(const Gamma::Algebra a, GridSerialRNG &rng)
   SpinVector v;
   SpinMatrix m, &testg = testAlgebra[a], pl;
   GammaL     gl(a);
-  bool       pass = true;
   
   random(rng, v);
   random(rng, m);
@@ -238,9 +232,9 @@ int main(int argc, char *argv[])
 {
   Grid_init(&argc,&argv);
   
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   
   GridCartesian Grid(latt_size,simd_layout,mpi_layout);
   GridSerialRNG sRNG;
diff --git a/tests/core/Test_gparity.cc b/tests/core/Test_gparity.cc
index 81091e9e..32cf9c5f 100644
--- a/tests/core/Test_gparity.cc
+++ b/tests/core/Test_gparity.cc
@@ -31,7 +31,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 //typedef GparityDomainWallFermionD GparityDiracOp;
 //typedef DomainWallFermionD StandardDiracOp;
@@ -84,18 +84,18 @@ int main (int argc, char ** argv)
   //const int L =4;
   //std::vector<int> latt_2f(Nd,L);
 
-  std::vector<int> latt_2f = GridDefaultLatt();
-  std::vector<int> latt_1f(latt_2f); latt_1f[nu] = 2*latt_2f[nu];
+  Coordinate latt_2f = GridDefaultLatt();
+  Coordinate latt_1f(latt_2f); latt_1f[nu] = 2*latt_2f[nu];
   int L = latt_2f[nu];
 
 
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplexType::Nsimd());
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplexType::Nsimd());
 
   std::cout << GridLogMessage << "SIMD layout: ";
   for(int i=0;i<simd_layout.size();i++) std::cout << simd_layout[i] << " ";
   std::cout << std::endl;
   
-  std::vector<int> mpi_layout  = GridDefaultMpi(); //node layout
+  Coordinate mpi_layout  = GridDefaultMpi(); //node layout
 
   GridCartesian         * UGrid_1f   = SpaceTimeGrid::makeFourDimGrid(latt_1f, simd_layout, mpi_layout);
   GridRedBlackCartesian * UrbGrid_1f = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid_1f);
@@ -127,7 +127,7 @@ int main (int argc, char ** argv)
   tmpsrc=src*2.0;
   PokeIndex<0>(src_2f,tmpsrc,1);
 
-  StandardFermionField result_1f(FGrid_1f); result_1f=zero;
+  StandardFermionField result_1f(FGrid_1f); result_1f=Zero();
   StandardGaugeField Umu_1f(UGrid_1f); 
   Replicate(Umu_2f,Umu_1f);
 
@@ -166,7 +166,7 @@ int main (int argc, char ** argv)
   StandardFermionField    src_o_1f(FrbGrid_1f);
   StandardFermionField result_o_1f(FrbGrid_1f);
   pickCheckerboard(Odd,src_o_1f,src_1f);
-  result_o_1f=zero;
+  result_o_1f=Zero();
 
   SchurDiagMooeeOperator<StandardDiracOp,StandardFermionField> HermOpEO(Ddwf);
   ConjugateGradient<StandardFermionField> CG(1.0e-8,10000);
@@ -258,18 +258,18 @@ int main (int argc, char ** argv)
 
   }
 
-  GparityFermionField result_2f(FGrid_2f); result_2f=zero;
+  GparityFermionField result_2f(FGrid_2f); result_2f=Zero();
   GparityFermionField    src_o_2f(FrbGrid_2f);
   GparityFermionField result_o_2f(FrbGrid_2f);
   pickCheckerboard(Odd,src_o_2f,src_2f);
-  result_o_2f=zero;
+  result_o_2f=Zero();
 
   ConjugateGradient<GparityFermionField> CG2f(1.0e-8,10000);
   SchurDiagMooeeOperator<GparityDiracOp,GparityFermionField> HermOpEO2f(GPDdwf);
   CG2f(HermOpEO2f,src_o_2f,result_o_2f);
 
-  std::cout << "2f cb "<<result_o_2f.checkerboard<<std::endl;
-  std::cout << "1f cb "<<result_o_1f.checkerboard<<std::endl;
+  std::cout << "2f cb "<<result_o_2f.Checkerboard()<<std::endl;
+  std::cout << "1f cb "<<result_o_1f.Checkerboard()<<std::endl;
 
   std::cout << " result norms " <<norm2(result_o_2f)<<" " <<norm2(result_o_1f)<<std::endl;
 
@@ -278,14 +278,14 @@ int main (int argc, char ** argv)
   StandardFermionField    res0  (FGrid_2f); 
   StandardFermionField    res1  (FGrid_2f); 
 
-  res0=zero;
-  res1=zero;
+  res0=Zero();
+  res1=Zero();
 
   res0o = PeekIndex<0>(result_o_2f,0);
   res1o = PeekIndex<0>(result_o_2f,1);
 
-  std::cout << "res cb "<<res0o.checkerboard<<std::endl;
-  std::cout << "res cb "<<res1o.checkerboard<<std::endl;
+  std::cout << "res cb "<<res0o.Checkerboard()<<std::endl;
+  std::cout << "res cb "<<res1o.Checkerboard()<<std::endl;
 
   setCheckerboard(res0,res0o);
   setCheckerboard(res1,res1o);
@@ -298,7 +298,7 @@ int main (int argc, char ** argv)
 
   replica = where( xcoor_1f5 >= Integer(L), replica1,replica0 );
 
-  replica0 = zero;
+  replica0 = Zero();
   setCheckerboard(replica0,result_o_1f);
 
   std::cout << "Norm2 solutions is " <<norm2(replica)<<" "<< norm2(replica0)<<std::endl;
diff --git a/tests/core/Test_gpwilson_even_odd.cc b/tests/core/Test_gpwilson_even_odd.cc
index 2069eb40..ac4cde99 100644
--- a/tests/core/Test_gpwilson_even_odd.cc
+++ b/tests/core/Test_gpwilson_even_odd.cc
@@ -30,15 +30,15 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -57,10 +57,10 @@ int main (int argc, char ** argv)
   FermionField src   (&Grid); random(pRNG,src);
   FermionField phi   (&Grid); random(pRNG,phi);
   FermionField chi   (&Grid); random(pRNG,chi);
-  FermionField result(&Grid); result=zero;
-  FermionField    ref(&Grid);    ref=zero;
-  FermionField    tmp(&Grid);    tmp=zero;
-  FermionField    err(&Grid);    tmp=zero;
+  FermionField result(&Grid); result=Zero();
+  FermionField    ref(&Grid);    ref=Zero();
+  FermionField    tmp(&Grid);    tmp=Zero();
+  FermionField    err(&Grid);    tmp=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
   std::vector<LatticeColourMatrix> U(4,&Grid);
 
@@ -70,7 +70,7 @@ int main (int argc, char ** argv)
   }  
 
   // Only one non-zero (y)
-  Umu=zero;
+  Umu=Zero();
   for(int nn=0;nn<Nd;nn++){
     random(pRNG,U[nn]);
     std::cout<<GridLogMessage<<"U[nn]"<<norm2(U[nn])<<std::endl;
diff --git a/tests/core/Test_lie_generators.cc b/tests/core/Test_lie_generators.cc
index 114b49f7..471cea25 100644
--- a/tests/core/Test_lie_generators.cc
+++ b/tests/core/Test_lie_generators.cc
@@ -42,7 +42,7 @@ directory
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main(int argc, char** argv) {
   Grid_init(&argc, &argv);
@@ -110,7 +110,7 @@ int main(int argc, char** argv) {
   AdjointRep<Nc> AdjRep(grid);
 
   // AdjointRepresentation has the predefined number of colours Nc
-  Representations<FundamentalRepresentation, AdjointRepresentation, TwoIndexSymmetricRepresentation> RepresentationTypes(grid);  
+  //  Representations<FundamentalRepresentation, AdjointRepresentation, TwoIndexSymmetricRepresentation> RepresentationTypes(grid);  
 
   
   LatticeGaugeField U(grid), V(grid);
@@ -121,7 +121,7 @@ int main(int argc, char** argv) {
   // Test group structure
   // (U_f * V_f)_r = U_r * V_r
   LatticeGaugeField UV(grid);
-  UV = zero;
+  UV = Zero();
   for (int mu = 0; mu < Nd; mu++) {
     SU<Nc>::LatticeMatrix Umu = peekLorentz(U,mu);
     SU<Nc>::LatticeMatrix Vmu = peekLorentz(V,mu);
@@ -139,7 +139,7 @@ int main(int argc, char** argv) {
   typename AdjointRep<Nc>::LatticeField Vr = AdjRep.U;  // V_r
 
   typename AdjointRep<Nc>::LatticeField UrVr(grid);
-  UrVr = zero;
+  UrVr = Zero();
   for (int mu = 0; mu < Nd; mu++) {
     typename AdjointRep<Nc>::LatticeMatrix Urmu = peekLorentz(Ur,mu);
     typename AdjointRep<Nc>::LatticeMatrix Vrmu = peekLorentz(Vr,mu);
@@ -200,7 +200,7 @@ int main(int argc, char** argv) {
 
 
   // Tranform to the adjoint representation
-  U = zero; // fill this with only one direction
+  U = Zero(); // fill this with only one direction
   pokeLorentz(U,Ufund,0); // the representation transf acts on full gauge fields
 
   AdjRep.update_representation(U);
@@ -315,7 +315,7 @@ int main(int argc, char** argv) {
   SU<Nc>::HotConfiguration<LatticeGaugeField>(gridRNG, V2);
   
   LatticeGaugeField UV2(grid);
-  UV2 = zero;
+  UV2 = Zero();
   for (int mu = 0; mu < Nd; mu++) {
     SU<Nc>::LatticeMatrix Umu2 = peekLorentz(U2,mu);
     SU<Nc>::LatticeMatrix Vmu2 = peekLorentz(V2,mu);
@@ -332,7 +332,7 @@ int main(int argc, char** argv) {
   typename TwoIndexRep< Nc, Symmetric >::LatticeField Vr2 = TIndexRep.U;  // V_r
   
   typename TwoIndexRep< Nc, Symmetric >::LatticeField Ur2Vr2(grid);
-  Ur2Vr2 = zero;
+  Ur2Vr2 = Zero();
   for (int mu = 0; mu < Nd; mu++) {
     typename TwoIndexRep< Nc, Symmetric >::LatticeMatrix Urmu2 = peekLorentz(Ur2,mu);
     typename TwoIndexRep< Nc, Symmetric >::LatticeMatrix Vrmu2 = peekLorentz(Vr2,mu);
@@ -393,7 +393,7 @@ int main(int argc, char** argv) {
   
 
   // Tranform to the 2Index Sym representation
-  U = zero; // fill this with only one direction
+  U = Zero(); // fill this with only one direction
   pokeLorentz(U,Ufund2,0); // the representation transf acts on full gauge fields
   
   TIndexRep.update_representation(U);
@@ -425,7 +425,7 @@ int main(int argc, char** argv) {
   SU<Nc>::HotConfiguration<LatticeGaugeField>(gridRNG, V2A);
   
   LatticeGaugeField UV2A(grid);
-  UV2A = zero;
+  UV2A = Zero();
   for (int mu = 0; mu < Nd; mu++) {
     SU<Nc>::LatticeMatrix Umu2A = peekLorentz(U2,mu);
     SU<Nc>::LatticeMatrix Vmu2A = peekLorentz(V2,mu);
@@ -442,7 +442,7 @@ int main(int argc, char** argv) {
   typename TwoIndexRep< Nc, AntiSymmetric >::LatticeField Vr2A = TIndexRepA.U;  // V_r
   
   typename TwoIndexRep< Nc, AntiSymmetric >::LatticeField Ur2Vr2A(grid);
-  Ur2Vr2A = zero;
+  Ur2Vr2A = Zero();
   for (int mu = 0; mu < Nd; mu++) {
     typename TwoIndexRep< Nc, AntiSymmetric >::LatticeMatrix Urmu2A = peekLorentz(Ur2A,mu);
     typename TwoIndexRep< Nc, AntiSymmetric >::LatticeMatrix Vrmu2A = peekLorentz(Vr2A,mu);
@@ -503,7 +503,7 @@ int main(int argc, char** argv) {
   
 
   // Tranform to the 2Index Sym representation
-  U = zero; // fill this with only one direction
+  U = Zero(); // fill this with only one direction
   pokeLorentz(U,Ufund2A,0); // the representation transf acts on full gauge fields
   
   TIndexRepA.update_representation(U);
diff --git a/tests/core/Test_main.cc b/tests/core/Test_main.cc
index 1fdc572f..cd2373ce 100644
--- a/tests/core/Test_main.cc
+++ b/tests/core/Test_main.cc
@@ -32,7 +32,6 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 /*
  Grid_main.cc(232): error: no suitable user-defined conversion from
@@ -58,9 +57,9 @@ auto peekDumKopf(const vobj &rhs, int i) -> decltype(peekIndex<3>(rhs, 0)) {
 int main(int argc, char **argv) {
   Grid_init(&argc, &argv);
 
-  std::vector<int> latt_size = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(4, vComplex::Nsimd());
-  std::vector<int> mpi_layout = GridDefaultMpi();
+  Coordinate latt_size = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(4, vComplex::Nsimd());
+  Coordinate mpi_layout = GridDefaultMpi();
 
   latt_size.resize(4);
 
@@ -180,7 +179,7 @@ int main(int argc, char **argv) {
       std::cout << "Norm2 LatticeReal : "<< norm2(BarReal) << std::endl;
       std::cout << "Norm2 LatticeComplex : "<< norm2(BarComplex) << std::endl;
 
-      exit(0);
+      //      exit(0);
 
       TComplex tr = trace(cmat);
 
@@ -388,7 +387,7 @@ int main(int argc, char **argv) {
         std::vector<int> coor(4);
         for(int d=0;d<4;d++) coor[d] = 0;
         peekSite(cmat,Foo,coor);
-        Foo = zero;
+        Foo = Zero();
         pokeSite(cmat,Foo,coor);
       }
       random(Foo);
@@ -402,14 +401,14 @@ int main(int argc, char **argv) {
           Fine._ldimensions[0] * Fine._ldimensions[1] * Fine._ldimensions[2];
 
       LatticeInteger lex(&Fine);
-      lex = zero;
+      lex = Zero();
       for (int d = 0; d < 4; d++) {
         LatticeInteger coor(&Fine);
         LatticeCoordinate(coor, d);
         lex = lex + coor * mm[d];
       }
 
-      //    Bar = zero;
+      //    Bar = Zero();
       //    Bar = where(lex<Integer(10),Foo,Bar);
 
       cout << "peeking sites..\n";
@@ -449,7 +448,7 @@ int main(int argc, char **argv) {
       double t0, t1, flops;
       double bytes;
       int ncall = 5000;
-      int Nc = Grid::QCD::Nc;
+      int Nc = Grid::Nc;
 
       LatticeGaugeField U(&Fine);
       //    LatticeColourMatrix Uy = peekLorentz(U,1);
@@ -515,7 +514,6 @@ int main(int argc, char **argv) {
       double nrm = 0;
 
       LatticeColourMatrix deriv(&Fine);
-      double half = 0.5;
       deriv = 0.5 * Cshift(Foo, 0, 1) - 0.5 * Cshift(Foo, 0, -1);
 
       for (int dir = 0; dir < 4; dir++) {
@@ -533,7 +531,7 @@ int main(int argc, char **argv) {
           bShifted = Cshift(rFoo, dir, shift);  // Shift red->black
           rShifted = Cshift(bFoo, dir, shift);  // Shift black->red
 
-          ShiftedCheck = zero;
+          ShiftedCheck = Zero();
           setCheckerboard(ShiftedCheck, bShifted);  // Put them all together
           setCheckerboard(ShiftedCheck,
                           rShifted);  // and check the results (later)
@@ -547,7 +545,7 @@ int main(int argc, char **argv) {
                    coor[1]++) {
                 for (coor[0] = 0; coor[0] < latt_size[0] / mpi_layout[0];
                      coor[0]++) {
-                  std::complex<Grid::Real> diff;
+                  Complex diff;
 
                   std::vector<int> shiftcoor = coor;
                   shiftcoor[dir] = (shiftcoor[dir] + shift + latt_size[dir]) %
diff --git a/tests/core/Test_mobius_eofa_even_odd.cc b/tests/core/Test_mobius_eofa_even_odd.cc
index fceb58f6..bfd53c72 100644
--- a/tests/core/Test_mobius_eofa_even_odd.cc
+++ b/tests/core/Test_mobius_eofa_even_odd.cc
@@ -32,7 +32,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -69,18 +69,18 @@ int main (int argc, char ** argv)
     LatticeFermion    src   (FGrid); random(RNG5, src);
     LatticeFermion    phi   (FGrid); random(RNG5, phi);
     LatticeFermion    chi   (FGrid); random(RNG5, chi);
-    LatticeFermion    result(FGrid); result = zero;
-    LatticeFermion    ref   (FGrid); ref = zero;
-    LatticeFermion    tmp   (FGrid); tmp = zero;
-    LatticeFermion    err   (FGrid); err = zero;
+    LatticeFermion    result(FGrid); result = Zero();
+    LatticeFermion    ref   (FGrid); ref = Zero();
+    LatticeFermion    tmp   (FGrid); tmp = Zero();
+    LatticeFermion    err   (FGrid); err = Zero();
     LatticeGaugeField Umu   (UGrid); SU3::HotConfiguration(RNG4, Umu);
     std::vector<LatticeColourMatrix> U(4,UGrid);
 
     // Only one non-zero (y)
-    Umu = zero;
+    Umu = Zero();
     for(int nn=0; nn<Nd; nn++){
         random(RNG4, U[nn]);
-        if(nn>0){ U[nn] = zero; }
+        if(nn>0){ U[nn] = Zero(); }
         PokeIndex<LorentzIndex>(Umu, U[nn], nn);
     }
 
diff --git a/tests/core/Test_mobius_even_odd.cc b/tests/core/Test_mobius_even_odd.cc
index 3c9ac520..0a035dc8 100644
--- a/tests/core/Test_mobius_even_odd.cc
+++ b/tests/core/Test_mobius_even_odd.cc
@@ -30,7 +30,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -68,19 +68,19 @@ int main (int argc, char ** argv)
   LatticeFermion src   (FGrid); random(RNG5,src);
   LatticeFermion phi   (FGrid); random(RNG5,phi);
   LatticeFermion chi   (FGrid); random(RNG5,chi);
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid);    ref=zero;
-  LatticeFermion    tmp(FGrid);    tmp=zero;
-  LatticeFermion    err(FGrid);    tmp=zero;
+  LatticeFermion result(FGrid); result=Zero();
+  LatticeFermion    ref(FGrid);    ref=Zero();
+  LatticeFermion    tmp(FGrid);    tmp=Zero();
+  LatticeFermion    err(FGrid);    tmp=Zero();
   LatticeGaugeField Umu(UGrid); random(RNG4,Umu);
   std::vector<LatticeColourMatrix> U(4,UGrid);
 
   // Only one non-zero (y)
-  Umu=zero;
+  Umu=Zero();
   for(int nn=0;nn<Nd;nn++){
     random(RNG4,U[nn]);
     if ( nn>0 ) 
-      U[nn]=zero;
+      U[nn]=Zero();
     PokeIndex<LorentzIndex>(Umu,U[nn],nn);
   }
 
diff --git a/tests/core/Test_poisson_fft.cc b/tests/core/Test_poisson_fft.cc
index 2252ec7a..ad93523e 100644
--- a/tests/core/Test_poisson_fft.cc
+++ b/tests/core/Test_poisson_fft.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
@@ -67,7 +67,7 @@ int main (int argc, char ** argv)
   neg = -pos;
   zz  = ComplexD(0.0,0.0);
 
-  Charge=zero;
+  Charge=Zero();
   
   // Parallel plate capacitor
   {
@@ -90,7 +90,7 @@ int main (int argc, char ** argv)
   std::vector<LatticeComplexD> k(4,&GRID);
   LatticeComplexD ksq(&GRID);
 
-  ksq=zero;
+  ksq=Zero();
   for(int mu=0;mu<nd;mu++) {
     
     Integer L=latt_size[mu];
diff --git a/tests/core/Test_qed.cc b/tests/core/Test_qed.cc
index f3b33454..775d6307 100644
--- a/tests/core/Test_qed.cc
+++ b/tests/core/Test_qed.cc
@@ -29,7 +29,6 @@ See the full license in the file "LICENSE" in the top level distribution directo
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace QCD;
 
 typedef PeriodicGaugeImpl<QedGImplR>  QedPeriodicGImplR;
 typedef PhotonR::GaugeField           EmField;
@@ -45,9 +44,9 @@ int main(int argc, char *argv[])
   std::cout << GridLogMessage << "Grid initialized" << std::endl;
   
   // QED stuff
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(4, vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(4, vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian    grid(latt_size,simd_layout,mpi_layout);
   GridParallelRNG  pRNG(&grid);
   PhotonR          photon(&grid, PhotonR::Gauge::coulomb, PhotonR::ZmScheme::qedL);
@@ -77,7 +76,7 @@ int main(int argc, char *argv[])
 
       std::cout << GridLogMessage << "Spatial zero-mode norm 2" << std::endl;
       sliceSum(a, zm, grid.Nd() - 1);
-      for (unsigned int t = 0; t < latt_size.back(); ++t)
+      for (unsigned int t = 0; t < latt_size.size(); ++t)
       {
         std::cout << GridLogMessage << "t = " << t << " " << std::sqrt(norm2(zm[t])) << std::endl;
       }
@@ -85,7 +84,7 @@ int main(int argc, char *argv[])
       // Calculate divergence
       EmComp diva(&grid), amu(&grid);
 
-      diva = zero;
+      diva = Zero();
       for (unsigned int mu = 0; mu < grid.Nd(); ++mu)
       {
         amu   = peekLorentz(a, mu);
diff --git a/tests/core/Test_quenched_update.cc b/tests/core/Test_quenched_update.cc
index 66d65309..ef428d1b 100644
--- a/tests/core/Test_quenched_update.cc
+++ b/tests/core/Test_quenched_update.cc
@@ -30,7 +30,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
@@ -58,7 +58,6 @@ int main (int argc, char ** argv)
   // SU3 colour operatoions
   LatticeColourMatrix link(grid);
   LatticeColourMatrix staple(grid);
-  int mu=0;
 
   // Apply heatbath to the link
   RealD beta=6.0;
@@ -75,8 +74,8 @@ int main (int argc, char ** argv)
 
     for( int cb=0;cb<2;cb++ ) {
 
-      one.checkerboard=subsets[cb];
-      mask= zero;
+      one.Checkerboard()=subsets[cb];
+      mask= Zero();
       setCheckerboard(mask,one);
 
       //      std::cout<<GridLogMessage<<mask<<std::endl;
diff --git a/tests/core/Test_rng.cc b/tests/core/Test_rng.cc
index b5d27c29..f6830cf9 100644
--- a/tests/core/Test_rng.cc
+++ b/tests/core/Test_rng.cc
@@ -29,15 +29,15 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
      
   GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
diff --git a/tests/core/Test_rng_fixed.cc b/tests/core/Test_rng_fixed.cc
index 3c4a27e1..1f935497 100644
--- a/tests/core/Test_rng_fixed.cc
+++ b/tests/core/Test_rng_fixed.cc
@@ -29,15 +29,15 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(4,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
      
   GridCartesian     Grid(latt_size,simd_layout,mpi_layout);
 
diff --git a/tests/core/Test_staggered.cc b/tests/core/Test_staggered.cc
index f8f7035a..c85d4090 100644
--- a/tests/core/Test_staggered.cc
+++ b/tests/core/Test_staggered.cc
@@ -30,15 +30,15 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -58,10 +58,10 @@ int main (int argc, char ** argv)
   typename ImprovedStaggeredFermionR::ImplParams params; 
 
   FermionField src   (&Grid); random(pRNG,src);
-  FermionField result(&Grid); result=zero;
-  FermionField    ref(&Grid);    ref=zero;
-  FermionField    tmp(&Grid);    tmp=zero;
-  FermionField    err(&Grid);    tmp=zero;
+  FermionField result(&Grid); result=Zero();
+  FermionField    ref(&Grid);    ref=Zero();
+  FermionField    tmp(&Grid);    tmp=Zero();
+  FermionField    err(&Grid);    tmp=Zero();
   FermionField phi   (&Grid); random(pRNG,phi);
   FermionField chi   (&Grid); random(pRNG,chi);
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
@@ -82,7 +82,7 @@ int main (int argc, char ** argv)
   */
   }
 
-  ref = zero;
+  ref = Zero();
 
   RealD mass=0.1;
   RealD c1=9.0/8.0;
@@ -90,7 +90,7 @@ int main (int argc, char ** argv)
   RealD u0=1.0;
 
   { // Simple improved staggered implementation
-    ref = zero;
+    ref = Zero();
     RealD c1tad = 0.5*c1/u0;
     RealD c2tad = 0.5*c2/u0/u0/u0;
 
diff --git a/tests/core/Test_staggered5D.cc b/tests/core/Test_staggered5D.cc
index be31c438..7055d183 100644
--- a/tests/core/Test_staggered5D.cc
+++ b/tests/core/Test_staggered5D.cc
@@ -30,15 +30,14 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   std::cout << GridLogMessage << "Making s innermost grids"<<std::endl;
 
@@ -69,10 +68,10 @@ int main (int argc, char ** argv)
 
   random(pRNG5,src);
 
-  FermionField result(FGrid); result=zero;
-  FermionField    ref(FGrid);    ref=zero;
-  FermionField    tmp(FGrid);    tmp=zero;
-  FermionField    err(FGrid);    tmp=zero;
+  FermionField result(FGrid); result=Zero();
+  FermionField    ref(FGrid);    ref=Zero();
+  FermionField    tmp(FGrid);    tmp=Zero();
+  FermionField    err(FGrid);    tmp=Zero();
   FermionField phi   (FGrid); random(pRNG5,phi);
   FermionField chi   (FGrid); random(pRNG5,chi);
 
@@ -89,9 +88,11 @@ int main (int argc, char ** argv)
   // replicate across fifth dimension
   ////////////////////////////////////
   LatticeGaugeField Umu5d(FGrid); 
-  for(int ss=0;ss<Umu._grid->oSites();ss++){
+  auto umu5d = Umu5d.View();
+  auto umu   = Umu.View();
+  for(int ss=0;ss<Umu.Grid()->oSites();ss++){
     for(int s=0;s<Ls;s++){
-      Umu5d._odata[Ls*ss+s] = Umu._odata[ss];
+      umu5d[Ls*ss+s] = umu[ss];
     }
   }
 
@@ -107,7 +108,7 @@ int main (int argc, char ** argv)
   RealD u0=1.0;
 
   { // Simple improved staggered implementation
-    ref = zero;
+    ref = Zero();
     RealD c1tad = 0.5*c1/u0;
     RealD c2tad = 0.5*c2/u0/u0/u0;
 
diff --git a/tests/core/Test_staggered5Dvec.cc b/tests/core/Test_staggered5Dvec.cc
index 2720c24c..73241276 100644
--- a/tests/core/Test_staggered5Dvec.cc
+++ b/tests/core/Test_staggered5Dvec.cc
@@ -30,15 +30,16 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+
 
 int main (int argc, char ** argv)
 {
+#if 0
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   const int Ls=16;
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
@@ -70,15 +71,15 @@ int main (int argc, char ** argv)
   FermionField src   (FGrid);
   random(pRNG5,src);
   /*
-  std::vector<int> site({0,1,2,0,0});
-  ColourVector cv = zero;
+  Coordinate site({0,1,2,0,0});
+  ColourVector cv = Zero();
   cv()()(0)=1.0;
-  src = zero;
+  src = Zero();
   pokeSite(cv,src,site);
   */
-  FermionField result(FGrid); result=zero;
-  FermionField    tmp(FGrid);    tmp=zero;
-  FermionField    err(FGrid);    tmp=zero;
+  FermionField result(FGrid); result=Zero();
+  FermionField    tmp(FGrid);    tmp=Zero();
+  FermionField    err(FGrid);    tmp=Zero();
   FermionField phi   (FGrid); random(pRNG5,phi);
   FermionField chi   (FGrid); random(pRNG5,chi);
 
@@ -88,7 +89,7 @@ int main (int argc, char ** argv)
   /*
   for(int mu=1;mu<4;mu++){
     auto tmp = PeekIndex<LorentzIndex>(Umu,mu);
-        tmp = zero;
+        tmp = Zero();
     PokeIndex<LorentzIndex>(Umu,tmp,mu);
   }
   */
@@ -129,9 +130,9 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "Calling vectorised staggered operator"<<std::endl;
 
 #ifdef AVX512
-  QCD::StaggeredKernelsStatic::Opt=QCD::StaggeredKernelsStatic::OptInlineAsm;
+  StaggeredKernelsStatic::Opt=StaggeredKernelsStatic::OptInlineAsm;
 #else
-  QCD::StaggeredKernelsStatic::Opt=QCD::StaggeredKernelsStatic::OptGeneric;
+  StaggeredKernelsStatic::Opt=StaggeredKernelsStatic::OptGeneric;
 #endif
 
   t0=usecond();
@@ -150,9 +151,9 @@ int main (int argc, char ** argv)
 
   
   FermionField ssrc  (sFGrid);  localConvert(src,ssrc);
-  FermionField sresult(sFGrid); sresult=zero;
+  FermionField sresult(sFGrid); sresult=Zero();
 
-  QCD::StaggeredKernelsStatic::Opt=QCD::StaggeredKernelsStatic::OptHandUnroll;
+  StaggeredKernelsStatic::Opt=StaggeredKernelsStatic::OptHandUnroll;
   t0=usecond();
   for(int i=0;i<ncall1;i++){
     sDs.Dhop(ssrc,sresult,0);
@@ -167,9 +168,9 @@ int main (int argc, char ** argv)
 
 
 #ifdef AVX512
-  QCD::StaggeredKernelsStatic::Opt=QCD::StaggeredKernelsStatic::OptInlineAsm;
+  StaggeredKernelsStatic::Opt=StaggeredKernelsStatic::OptInlineAsm;
 #else
-  QCD::StaggeredKernelsStatic::Opt=QCD::StaggeredKernelsStatic::OptGeneric;
+  StaggeredKernelsStatic::Opt=StaggeredKernelsStatic::OptGeneric;
 #endif
 
   err = tmp-result; 
@@ -193,4 +194,5 @@ int main (int argc, char ** argv)
 
 
   Grid_finalize();
+#endif
 }
diff --git a/tests/core/Test_staggered5DvecF.cc b/tests/core/Test_staggered5DvecF.cc
index 5d421673..2386d054 100644
--- a/tests/core/Test_staggered5DvecF.cc
+++ b/tests/core/Test_staggered5DvecF.cc
@@ -30,15 +30,15 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 int main (int argc, char ** argv)
 {
+#if 0
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  auto latt_size   = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
 
   const int Ls=16;
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
@@ -76,9 +76,9 @@ int main (int argc, char ** argv)
   src = zero;
   pokeSite(cv,src,site);
   */
-  FermionField result(FGrid); result=zero;
-  FermionField    tmp(FGrid);    tmp=zero;
-  FermionField    err(FGrid);    tmp=zero;
+  FermionField result(FGrid); result=Zero();
+  FermionField    tmp(FGrid);    tmp=Zero();
+  FermionField    err(FGrid);    tmp=Zero();
   FermionField phi   (FGrid); random(pRNG5,phi);
   FermionField chi   (FGrid); random(pRNG5,chi);
 
@@ -88,7 +88,7 @@ int main (int argc, char ** argv)
   /*
   for(int mu=1;mu<4;mu++){
     auto tmp = PeekIndex<LorentzIndex>(Umu,mu);
-        tmp = zero;
+        tmp = Zero();
     PokeIndex<LorentzIndex>(Umu,tmp,mu);
   }
   */
@@ -129,9 +129,9 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "Calling vectorised staggered operator"<<std::endl;
 
 #ifdef AVX512
-  QCD::StaggeredKernelsStatic::Opt=QCD::StaggeredKernelsStatic::OptInlineAsm;
+  StaggeredKernelsStatic::Opt=StaggeredKernelsStatic::OptInlineAsm;
 #else
-  QCD::StaggeredKernelsStatic::Opt=QCD::StaggeredKernelsStatic::OptGeneric;
+  StaggeredKernelsStatic::Opt=StaggeredKernelsStatic::OptGeneric;
 #endif
 
   t0=usecond();
@@ -150,9 +150,9 @@ int main (int argc, char ** argv)
 
   
   FermionField ssrc  (sFGrid);  localConvert(src,ssrc);
-  FermionField sresult(sFGrid); sresult=zero;
+  FermionField sresult(sFGrid); sresult=Zero();
 
-  QCD::StaggeredKernelsStatic::Opt=QCD::StaggeredKernelsStatic::OptHandUnroll;
+  StaggeredKernelsStatic::Opt=StaggeredKernelsStatic::OptHandUnroll;
   t0=usecond();
   for(int i=0;i<ncall1;i++){
     sDs.Dhop(ssrc,sresult,0);
@@ -167,9 +167,9 @@ int main (int argc, char ** argv)
 
 
 #ifdef AVX512
-  QCD::StaggeredKernelsStatic::Opt=QCD::StaggeredKernelsStatic::OptInlineAsm;
+  StaggeredKernelsStatic::Opt=StaggeredKernelsStatic::OptInlineAsm;
 #else
-  QCD::StaggeredKernelsStatic::Opt=QCD::StaggeredKernelsStatic::OptGeneric;
+  StaggeredKernelsStatic::Opt=StaggeredKernelsStatic::OptGeneric;
 #endif
 
   err = tmp-result; 
@@ -193,4 +193,5 @@ int main (int argc, char ** argv)
 
 
   Grid_finalize();
+#endif
 }
diff --git a/tests/core/Test_wilson_clover.cc b/tests/core/Test_wilson_clover.cc
index 9281e298..3e31f7f6 100644
--- a/tests/core/Test_wilson_clover.cc
+++ b/tests/core/Test_wilson_clover.cc
@@ -29,15 +29,14 @@
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 int main(int argc, char **argv)
 {
   Grid_init(&argc, &argv);
 
-  std::vector<int> latt_size = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd, vComplex::Nsimd());
-  std::vector<int> mpi_layout = GridDefaultMpi();
+  auto latt_size = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(Nd, vComplex::Nsimd());
+  auto mpi_layout = GridDefaultMpi();
   GridCartesian Grid(latt_size, simd_layout, mpi_layout);
   GridRedBlackCartesian RBGrid(&Grid);
 
@@ -59,17 +58,17 @@ int main(int argc, char **argv)
   FermionField src(&Grid);
   random(pRNG, src);
   FermionField result(&Grid);
-  result = zero;
+  result = Zero();
   FermionField result2(&Grid);
-  result2 = zero;
+  result2 = Zero();
   FermionField ref(&Grid);
-  ref = zero;
+  ref = Zero();
   FermionField tmp(&Grid);
-  tmp = zero;
+  tmp = Zero();
   FermionField err(&Grid);
-  err = zero;
+  err = Zero();
   FermionField err2(&Grid);
-  err2 = zero;
+  err2 = Zero();
   FermionField phi(&Grid);
   random(pRNG, phi);
   FermionField chi(&Grid);
@@ -214,9 +213,9 @@ int main(int argc, char **argv)
   std::cout << GridLogMessage << "= Testing gauge covariance Clover term with EO preconditioning  " << std::endl;
   std::cout << GridLogMessage << "================================================================" << std::endl;
 
-  chi = zero;
-  phi = zero;
-  tmp = zero;
+  chi = Zero();
+  phi = Zero();
+  tmp = Zero();
   pickCheckerboard(Even, chi_e, chi);
   pickCheckerboard(Odd, chi_o, chi);
   pickCheckerboard(Even, phi_e, phi);
@@ -236,9 +235,9 @@ int main(int argc, char **argv)
   LatticeColourMatrix Omega(&Grid);
   LatticeColourMatrix ShiftedOmega(&Grid);
   LatticeGaugeField U_prime(&Grid);
-  U_prime = zero;
+  U_prime = Zero();
   LatticeColourMatrix U_prime_mu(&Grid);
-  U_prime_mu = zero;
+  U_prime_mu = Zero();
   SU<Nc>::LieRandomize(pRNG2, Omega, 1.0);
   for (int mu = 0; mu < Nd; mu++)
   {
@@ -269,8 +268,8 @@ int main(int argc, char **argv)
   std::cout << GridLogMessage << "= Testing gauge covariance Clover term w/o EO preconditioning  " << std::endl;
   std::cout << GridLogMessage << "================================================================" << std::endl;
 
-  chi = zero;
-  phi = zero;
+  chi = Zero();
+  phi = Zero();
 
   WilsonFermionR Dw(Umu, Grid, RBGrid, mass, params);
   Dw.ImportGauge(Umu);
@@ -293,9 +292,9 @@ int main(int argc, char **argv)
   std::cout << GridLogMessage << "= Testing Mooee(csw=0) Clover to reproduce Mooee Wilson   " << std::endl;
   std::cout << GridLogMessage << "==========================================================" << std::endl;
 
-  chi = zero;
-  phi = zero;
-  err = zero;
+  chi = Zero();
+  phi = Zero();
+  err = Zero();
   WilsonCloverFermionR Dwc_csw0(Umu, Grid, RBGrid, mass, 0.0, 0.0, anis, params); //  <-- Notice: csw=0
   Dwc_csw0.ImportGauge(Umu);
 
@@ -323,9 +322,9 @@ int main(int argc, char **argv)
   std::cout << GridLogMessage << "= Testing EO operator is equal to the unprec              " << std::endl;
   std::cout << GridLogMessage << "==========================================================" << std::endl;
 
-  chi = zero;
-  phi = zero;
-  err = zero;
+  chi = Zero();
+  phi = Zero();
+  err = Zero();
 
   pickCheckerboard(Even, phi_e, phi);
   pickCheckerboard(Odd, phi_o, phi);
diff --git a/tests/core/Test_wilson_even_odd.cc b/tests/core/Test_wilson_even_odd.cc
index 909a2509..8f325f1c 100644
--- a/tests/core/Test_wilson_even_odd.cc
+++ b/tests/core/Test_wilson_even_odd.cc
@@ -29,7 +29,6 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 template<class d>
 struct scal {
@@ -47,9 +46,9 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -66,11 +65,12 @@ int main (int argc, char ** argv)
   LatticeFermion src   (&Grid); random(pRNG,src);
   LatticeFermion phi   (&Grid); random(pRNG,phi);
   LatticeFermion chi   (&Grid); random(pRNG,chi);
-  LatticeFermion result(&Grid); result=zero;
-  LatticeFermion    ref(&Grid);    ref=zero;
-  LatticeFermion    tmp(&Grid);    tmp=zero;
-  LatticeFermion    err(&Grid);    tmp=zero;
-  LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
+  LatticeFermion result(&Grid); result=Zero();
+  LatticeFermion    ref(&Grid);    ref=Zero();
+  LatticeFermion    tmp(&Grid);    tmp=Zero();
+  LatticeFermion    err(&Grid);    tmp=Zero();
+  LatticeGaugeField Umu(&Grid); 
+  SU3::HotConfiguration(pRNG,Umu);
   std::vector<LatticeColourMatrix> U(4,&Grid);
 
   double volume=1;
@@ -79,7 +79,7 @@ int main (int argc, char ** argv)
   }  
 
   // Only one non-zero (y)
-  Umu=zero;
+  Umu=Zero();
   for(int nn=0;nn<Nd;nn++){
     random(pRNG,U[nn]);
     std::cout<<GridLogMessage<<"U[nn]"<<norm2(U[nn])<<std::endl;
diff --git a/tests/core/Test_wilson_twisted_mass_even_odd.cc b/tests/core/Test_wilson_twisted_mass_even_odd.cc
index adc5c9c9..58b0b60f 100644
--- a/tests/core/Test_wilson_twisted_mass_even_odd.cc
+++ b/tests/core/Test_wilson_twisted_mass_even_odd.cc
@@ -30,7 +30,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -48,9 +48,9 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -67,10 +67,10 @@ int main (int argc, char ** argv)
   LatticeFermion src   (&Grid); random(pRNG,src);
   LatticeFermion phi   (&Grid); random(pRNG,phi);
   LatticeFermion chi   (&Grid); random(pRNG,chi);
-  LatticeFermion result(&Grid); result=zero;
-  LatticeFermion    ref(&Grid);    ref=zero;
-  LatticeFermion    tmp(&Grid);    tmp=zero;
-  LatticeFermion    err(&Grid);    tmp=zero;
+  LatticeFermion result(&Grid); result=Zero();
+  LatticeFermion    ref(&Grid);    ref=Zero();
+  LatticeFermion    tmp(&Grid);    tmp=Zero();
+  LatticeFermion    err(&Grid);    tmp=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
   std::vector<LatticeColourMatrix> U(4,&Grid);
 
@@ -80,7 +80,7 @@ int main (int argc, char ** argv)
   }  
 
   // Only one non-zero (y)
-  Umu=zero;
+  Umu=Zero();
   for(int nn=0;nn<Nd;nn++){
     random(pRNG,U[nn]);
     std::cout<<GridLogMessage<<"U[nn]"<<norm2(U[nn])<<std::endl;
diff --git a/tests/core/Test_zmobius_even_odd.cc b/tests/core/Test_zmobius_even_odd.cc
index 867c3359..1150930b 100644
--- a/tests/core/Test_zmobius_even_odd.cc
+++ b/tests/core/Test_zmobius_even_odd.cc
@@ -30,7 +30,6 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 template<class d>
 struct scal {
@@ -52,7 +51,6 @@ int main (int argc, char ** argv)
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 
-
   const int Ls=10;
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
   GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
@@ -68,25 +66,25 @@ int main (int argc, char ** argv)
   LatticeFermion src   (FGrid); random(RNG5,src);
   LatticeFermion phi   (FGrid); random(RNG5,phi);
   LatticeFermion chi   (FGrid); random(RNG5,chi);
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid);    ref=zero;
-  LatticeFermion    tmp(FGrid);    tmp=zero;
-  LatticeFermion    err(FGrid);    tmp=zero;
+  LatticeFermion result(FGrid); result=Zero();
+  LatticeFermion    ref(FGrid);    ref=Zero();
+  LatticeFermion    tmp(FGrid);    tmp=Zero();
+  LatticeFermion    err(FGrid);    tmp=Zero();
   LatticeGaugeField Umu(UGrid); random(RNG4,Umu);
   std::vector<LatticeColourMatrix> U(4,UGrid);
 
   // Only one non-zero (y)
-  Umu=zero;
+  Umu=Zero();
   for(int nn=0;nn<Nd;nn++){
     random(RNG4,U[nn]);
     if ( nn>0 ) 
-      U[nn]=zero;
+      U[nn]=Zero();
     PokeIndex<LorentzIndex>(Umu,U[nn],nn);
   }
 
   RealD mass=0.1;
   RealD M5  =1.8;
-  std::vector < std::complex<double>  > omegas;
+  std::vector < ComplexD  > omegas;
 #if 0
   for(int i=0;i<Ls;i++){
     double imag = 0.;
@@ -107,9 +105,25 @@ int main (int argc, char ** argv)
   omegas.push_back( std::complex<double>(0.0686324988446592,0.0550658530827402) );
   omegas.push_back( std::complex<double>(0.0686324988446592,-0.0550658530827402) );
 #endif
-
-  ZMobiusFermionR Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, omegas,1.,0.);
-//  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+  /*
+  argument types are: (Grid::LatticeGaugeField, 
+		       Grid::GridCartesian, 
+		       Grid::GridRedBlackCartesian, 
+		       Grid::GridCartesian, 
+		       Grid::GridRedBlackCartesian, 
+		       Grid::RealD, 
+		       Grid::RealD, 
+		       std::__1::vector<std::__1::complex<double>, 
+		       std::__1::allocator<std::__1::complex<double>>>, double, double)
+  ZMobiusFermion(GaugeField &_Umu,
+		 GridCartesian         &FiveDimGrid,
+		 GridRedBlackCartesian &FiveDimRedBlackGrid,
+		 GridCartesian         &FourDimGrid,
+		 GridRedBlackCartesian &FourDimRedBlackGrid,
+		 RealD _mass,RealD _M5,
+		 std::vector<ComplexD> &gamma, RealD b,RealD c,const ImplParams &p= ImplParams()) : 
+  */
+  ZMobiusFermionR Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, omegas,RealD(1.),RealD(0.));
 
   LatticeFermion src_e (FrbGrid);
   LatticeFermion src_o (FrbGrid);
diff --git a/tests/debug/Test_cayley_cg.cc b/tests/debug/Test_cayley_cg.cc
index 5150268f..5a9c696f 100644
--- a/tests/debug/Test_cayley_cg.cc
+++ b/tests/debug/Test_cayley_cg.cc
@@ -30,19 +30,9 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
-template<class d>
-struct scal {
-  d internal;
-};
 
-  Gamma::Algebra Gmu [] = {
-    Gamma::Algebra::GammaX,
-    Gamma::Algebra::GammaY,
-    Gamma::Algebra::GammaZ,
-    Gamma::Algebra::GammaT
-  };
+
 
 template<class What> 
 void  TestCGinversions(What & Ddwf, 
@@ -184,7 +174,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage <<"======================"<<std::endl;
   std::cout<<GridLogMessage <<"OverlapWilsonCayleyTanhFermion test"<<std::endl;
   std::cout<<GridLogMessage <<"======================"<<std::endl;
-  OverlapWilsonCayleyTanhFermionR Dov (Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+  OverlapWilsonCayleyTanhFermionR Dov(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
   OverlapWilsonCayleyTanhFermionF DovF(UmuF,*FGridF,*FrbGridF,*UGridF,*UrbGridF,mass,M5,1.0);
   TestCGinversions<OverlapWilsonCayleyTanhFermionR>(Dov,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
   TestReconstruct5DFA<OverlapWilsonCayleyTanhFermionR,OverlapWilsonCayleyTanhFermionF>(Dov,DovF,Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
@@ -225,7 +215,7 @@ void  TestCGunprec(What & Ddwf,
 		   GridParallelRNG *RNG5)
 {
   LatticeFermion src   (FGrid); random(*RNG5,src);
-  LatticeFermion result(FGrid); result=zero;
+  LatticeFermion result(FGrid); result=Zero();
 
   MdagMLinearOperator<What,LatticeFermion> HermOp(Ddwf);
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
@@ -244,7 +234,7 @@ void  TestCGprec(What & Ddwf,
   LatticeFermion    src_o(FrbGrid);
   LatticeFermion result_o(FrbGrid);
   pickCheckerboard(Odd,src_o,src);
-  result_o=zero;
+  result_o=Zero();
 
   SchurDiagMooeeOperator<What,LatticeFermion> HermOpEO(Ddwf);
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
@@ -261,7 +251,7 @@ void  TestReconstruct5D(What & Ddwf,
 			GridParallelRNG *RNG5)
 {
   LatticeFermion src4   (UGrid); random(*RNG4,src4);
-  LatticeFermion res4   (UGrid); res4 = zero;
+  LatticeFermion res4   (UGrid); res4 = Zero();
 
   LatticeFermion src   (FGrid);
   LatticeFermion src_NE(FGrid);
@@ -332,7 +322,7 @@ void  TestReconstruct5DFA(What & Ddwf,
 			  GridParallelRNG *RNG5)
 {
   LatticeFermion src4   (UGrid); random(*RNG4,src4);
-  LatticeFermion res4   (UGrid); res4 = zero;
+  LatticeFermion res4   (UGrid); res4 = Zero();
 
   LatticeFermion src   (FGrid);
   LatticeFermion src_NE(FGrid);
@@ -402,7 +392,7 @@ void  TestCGschur(What & Ddwf,
 		   GridParallelRNG *RNG5)
 {
   LatticeFermion src   (FGrid); random(*RNG5,src);
-  LatticeFermion result(FGrid); result=zero;
+  LatticeFermion result(FGrid); result=Zero();
 
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
   SchurRedBlackDiagMooeeSolve<LatticeFermion> SchurSolver(CG);
diff --git a/tests/debug/Test_cayley_coarsen_support.cc b/tests/debug/Test_cayley_coarsen_support.cc
index f57823e5..e6e18250 100644
--- a/tests/debug/Test_cayley_coarsen_support.cc
+++ b/tests/debug/Test_cayley_coarsen_support.cc
@@ -30,7 +30,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 
 template<class d>
@@ -58,7 +58,7 @@ int main (int argc, char ** argv)
   GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
 
   // Construct a coarsened grid
-  std::vector<int> clatt = GridDefaultLatt();
+  Coordinate clatt = GridDefaultLatt();
   for(int d=0;d<clatt.size();d++){
     clatt[d] = clatt[d]/2;
   }
@@ -73,19 +73,19 @@ int main (int argc, char ** argv)
   GridParallelRNG          CRNG(Coarse5d);CRNG.SeedFixedIntegers(cseeds);
 
   LatticeFermion    src(FGrid); random(RNG5,src);
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid); ref=zero;
+  LatticeFermion result(FGrid); result=Zero();
+  LatticeFermion    ref(FGrid); ref=Zero();
   LatticeFermion    tmp(FGrid);
   LatticeFermion    err(FGrid);
   LatticeGaugeField Umu(UGrid); SU3::HotConfiguration(RNG4,Umu);
 
 #if 0
   std::vector<LatticeColourMatrix> U(4,UGrid);
-  Umu=zero;
+  Umu=Zero();
   Complex cone(1.0,0.0);
   for(int nn=0;nn<Nd;nn++){
     if(1) {
-      if (nn>2) { U[nn]=zero; std::cout<<GridLogMessage << "zeroing gauge field in dir "<<nn<<std::endl; }
+      if (nn>2) { U[nn]=Zero(); std::cout<<GridLogMessage << "zeroing gauge field in dir "<<nn<<std::endl; }
       else      { U[nn]=cone; std::cout<<GridLogMessage << "unit gauge field in dir "<<nn<<std::endl; }
     }
     pokeIndex<LorentzIndex>(Umu,U[nn],nn);
@@ -120,6 +120,7 @@ int main (int argc, char ** argv)
 
   MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermDefOp(Ddwf);
   typedef Aggregation<vSpinColourVector,vTComplex,nbasis> Subspace;
+
   Subspace Aggregates(Coarse5d,FGrid,cb);
   Aggregates.CreateSubspaceRandom(RNG5);
 
@@ -143,7 +144,7 @@ int main (int argc, char ** argv)
   
   blockPromote(c_src,err,subspace);
 
-  prom=zero;
+  prom=Zero();
   for(int b=0;b<nbasis;b++){
     prom=prom+subspace[b];
   }
diff --git a/tests/debug/Test_cayley_even_odd.cc b/tests/debug/Test_cayley_even_odd.cc
index e226ac4a..433f0722 100644
--- a/tests/debug/Test_cayley_even_odd.cc
+++ b/tests/debug/Test_cayley_even_odd.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -127,10 +127,10 @@ void  TestWhat(What & Ddwf,
   LatticeFermion src   (FGrid); random(*RNG5,src);
   LatticeFermion phi   (FGrid); random(*RNG5,phi);
   LatticeFermion chi   (FGrid); random(*RNG5,chi);
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid);    ref=zero;
-  LatticeFermion    tmp(FGrid);    tmp=zero;
-  LatticeFermion    err(FGrid);    tmp=zero;
+  LatticeFermion result(FGrid); result=Zero();
+  LatticeFermion    ref(FGrid);    ref=Zero();
+  LatticeFermion    tmp(FGrid);    tmp=Zero();
+  LatticeFermion    err(FGrid);    tmp=Zero();
 
   LatticeFermion src_e (FrbGrid);
   LatticeFermion src_o (FrbGrid);
diff --git a/tests/debug/Test_cayley_ldop_cr.cc b/tests/debug/Test_cayley_ldop_cr.cc
index c6005fd0..82f388ab 100644
--- a/tests/debug/Test_cayley_ldop_cr.cc
+++ b/tests/debug/Test_cayley_ldop_cr.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
@@ -46,7 +46,7 @@ int main (int argc, char ** argv)
   ///////////////////////////////////////////////////
   // Construct a coarsened grid; utility for this?
   ///////////////////////////////////////////////////
-  std::vector<int> clatt = GridDefaultLatt();
+  Coordinate clatt = GridDefaultLatt();
   for(int d=0;d<clatt.size();d++){
     clatt[d] = clatt[d]/2;
   }
@@ -61,8 +61,8 @@ int main (int argc, char ** argv)
   GridParallelRNG          CRNG(Coarse5d);CRNG.SeedFixedIntegers(cseeds);
 
   LatticeFermion    src(FGrid); gaussian(RNG5,src);
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid); ref=zero;
+  LatticeFermion result(FGrid); result=Zero();
+  LatticeFermion    ref(FGrid); ref=Zero();
   LatticeFermion    tmp(FGrid);
   LatticeFermion    err(FGrid);
   LatticeGaugeField Umu(UGrid); 
@@ -74,7 +74,7 @@ int main (int argc, char ** argv)
   //  SU3::ColdConfiguration(RNG4,Umu);
   //  SU3::TepidConfiguration(RNG4,Umu);
   //  SU3::HotConfiguration(RNG4,Umu);
-  //  Umu=zero;
+  //  Umu=Zero();
 
   RealD mass=0.1;
   RealD M5=1.5;
@@ -99,14 +99,13 @@ int main (int argc, char ** argv)
   Subspace Aggregates(Coarse5d,FGrid,cb);
   Aggregates.CreateSubspace(RNG5,HermDefOp);
 
-
   LittleDiracOperator LittleDiracOp(*Coarse5d);
   LittleDiracOp.CoarsenOperator(FGrid,HermIndefOp,Aggregates);
   
   CoarseVector c_src (Coarse5d);
   CoarseVector c_res (Coarse5d);
   gaussian(CRNG,c_src);
-  c_res=zero;
+  c_res=Zero();
 
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Solving mdagm-CG on coarse space "<< std::endl;
diff --git a/tests/debug/Test_cheby.cc b/tests/debug/Test_cheby.cc
index 72d07885..ad5022ec 100644
--- a/tests/debug/Test_cheby.cc
+++ b/tests/debug/Test_cheby.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 RealD InverseApproximation(RealD x){
   return 1.0/x;
diff --git a/tests/debug/Test_heatbath_dwf_eofa.cc b/tests/debug/Test_heatbath_dwf_eofa.cc
index b77bc982..1e64a568 100644
--- a/tests/debug/Test_heatbath_dwf_eofa.cc
+++ b/tests/debug/Test_heatbath_dwf_eofa.cc
@@ -38,7 +38,7 @@ Author: David Murphy <dmurphy@phys.columbia.edu>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 // Parameters for test
 const std::vector<int> grid_dim = { 8, 8, 8, 8 };
diff --git a/tests/debug/Test_heatbath_dwf_eofa_gparity.cc b/tests/debug/Test_heatbath_dwf_eofa_gparity.cc
index 5c9d4923..cc118d1d 100644
--- a/tests/debug/Test_heatbath_dwf_eofa_gparity.cc
+++ b/tests/debug/Test_heatbath_dwf_eofa_gparity.cc
@@ -38,7 +38,7 @@ Author: David Murphy <dmurphy@phys.columbia.edu>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 typedef GparityWilsonImplR FermionImplPolicy;
 typedef GparityDomainWallEOFAFermionR FermionAction;
diff --git a/tests/debug/Test_heatbath_mobius_eofa.cc b/tests/debug/Test_heatbath_mobius_eofa.cc
index a952873d..95ab935e 100644
--- a/tests/debug/Test_heatbath_mobius_eofa.cc
+++ b/tests/debug/Test_heatbath_mobius_eofa.cc
@@ -38,7 +38,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 // Parameters for test
 const std::vector<int> grid_dim = { 8, 8, 8, 8 };
diff --git a/tests/debug/Test_heatbath_mobius_eofa_gparity.cc b/tests/debug/Test_heatbath_mobius_eofa_gparity.cc
index 08c6d566..7ed3a308 100644
--- a/tests/debug/Test_heatbath_mobius_eofa_gparity.cc
+++ b/tests/debug/Test_heatbath_mobius_eofa_gparity.cc
@@ -38,7 +38,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 typedef GparityWilsonImplR FermionImplPolicy;
 typedef GparityMobiusEOFAFermionR FermionAction;
diff --git a/tests/debug/Test_reweight_dwf_eofa.cc b/tests/debug/Test_reweight_dwf_eofa.cc
index 98a17e2f..728fbf78 100644
--- a/tests/debug/Test_reweight_dwf_eofa.cc
+++ b/tests/debug/Test_reweight_dwf_eofa.cc
@@ -31,7 +31,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 // parameters for test
 const std::vector<int> grid_dim = { 8, 8, 8, 8 };
@@ -178,7 +178,7 @@ int main(int argc, char **argv)
     for(int s=0; s<Ls; ++s){ axpby_ssp_pminus(spProj_Phi, 0.0, Phi, 1.0, Phi, s, s); }
     Deofa_L.Omega(spProj_Phi, tmp[0], -1, 0);
     G5R5(tmp[1], tmp[0]);
-    tmp[0] = zero;
+    tmp[0] = Zero();
     SchurSolver(Deofa_L, tmp[1], tmp[0]);
     Deofa_L.Omega(tmp[0], tmp[1], -1, 1);
     rw_eofa[hit] = -k*innerProduct(spProj_Phi,tmp[1]).real();
@@ -187,7 +187,7 @@ int main(int argc, char **argv)
     for(int s=0; s<Ls; ++s){ axpby_ssp_pplus(spProj_Phi, 0.0, Phi, 1.0, Phi, s, s); }
     Deofa_R.Omega(spProj_Phi, tmp[0], 1, 0);
     G5R5(tmp[1], tmp[0]);
-    tmp[0] = zero;
+    tmp[0] = Zero();
     SchurSolver(Deofa_R, tmp[1], tmp[0]);
     Deofa_R.Omega(tmp[0], tmp[1], 1, 1);
     rw_eofa[hit] += k*innerProduct(spProj_Phi,tmp[1]).real();
diff --git a/tests/debug/Test_reweight_dwf_eofa_gparity.cc b/tests/debug/Test_reweight_dwf_eofa_gparity.cc
index bb0fd98e..fcc01b8d 100644
--- a/tests/debug/Test_reweight_dwf_eofa_gparity.cc
+++ b/tests/debug/Test_reweight_dwf_eofa_gparity.cc
@@ -31,7 +31,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 typedef typename GparityDomainWallFermionR::FermionField FermionField;
 
@@ -181,7 +181,7 @@ int main(int argc, char **argv)
     for(int s=0; s<Ls; ++s){ axpby_ssp_pminus(spProj_Phi, 0.0, Phi, 1.0, Phi, s, s); }
     Deofa_L.Omega(spProj_Phi, tmp[0], -1, 0);
     G5R5(tmp[1], tmp[0]);
-    tmp[0] = zero;
+    tmp[0] = Zero();
     SchurSolver(Deofa_L, tmp[1], tmp[0]);
     Deofa_L.Omega(tmp[0], tmp[1], -1, 1);
     rw_eofa[hit] = -k*innerProduct(spProj_Phi,tmp[1]).real();
@@ -190,7 +190,7 @@ int main(int argc, char **argv)
     for(int s=0; s<Ls; ++s){ axpby_ssp_pplus(spProj_Phi, 0.0, Phi, 1.0, Phi, s, s); }
     Deofa_R.Omega(spProj_Phi, tmp[0], 1, 0);
     G5R5(tmp[1], tmp[0]);
-    tmp[0] = zero;
+    tmp[0] = Zero();
     SchurSolver(Deofa_R, tmp[1], tmp[0]);
     Deofa_R.Omega(tmp[0], tmp[1], 1, 1);
     rw_eofa[hit] += k*innerProduct(spProj_Phi,tmp[1]).real();
diff --git a/tests/debug/Test_reweight_mobius_eofa.cc b/tests/debug/Test_reweight_mobius_eofa.cc
index c4fa78d0..c5e46bcf 100644
--- a/tests/debug/Test_reweight_mobius_eofa.cc
+++ b/tests/debug/Test_reweight_mobius_eofa.cc
@@ -31,7 +31,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 // parameters for test
 const std::vector<int> grid_dim = { 8, 8, 8, 8 };
@@ -185,7 +185,7 @@ int main(int argc, char **argv)
     for(int s=0; s<Ls; ++s){ axpby_ssp_pminus(spProj_Phi, 0.0, Phi, 1.0, Phi, s, s); }
     Deofa_L.Omega(spProj_Phi, tmp[0], -1, 0);
     G5R5(tmp[1], tmp[0]);
-    tmp[0] = zero;
+    tmp[0] = Zero();
     SchurSolver(Deofa_L, tmp[1], tmp[0]);
     Deofa_L.Dtilde(tmp[0], tmp[1]);
     Deofa_L.Omega(tmp[1], tmp[0], -1, 1);
@@ -195,7 +195,7 @@ int main(int argc, char **argv)
     for(int s=0; s<Ls; ++s){ axpby_ssp_pplus(spProj_Phi, 0.0, Phi, 1.0, Phi, s, s); }
     Deofa_R.Omega(spProj_Phi, tmp[0], 1, 0);
     G5R5(tmp[1], tmp[0]);
-    tmp[0] = zero;
+    tmp[0] = Zero();
     SchurSolver(Deofa_R, tmp[1], tmp[0]);
     Deofa_R.Dtilde(tmp[0], tmp[1]);
     Deofa_R.Omega(tmp[1], tmp[0], 1, 1);
diff --git a/tests/debug/Test_reweight_mobius_eofa_gparity.cc b/tests/debug/Test_reweight_mobius_eofa_gparity.cc
index 11a242d2..bfc7543a 100644
--- a/tests/debug/Test_reweight_mobius_eofa_gparity.cc
+++ b/tests/debug/Test_reweight_mobius_eofa_gparity.cc
@@ -31,7 +31,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 typedef typename GparityDomainWallFermionR::FermionField FermionField;
 
@@ -188,7 +188,7 @@ int main(int argc, char **argv)
     for(int s=0; s<Ls; ++s){ axpby_ssp_pminus(spProj_Phi, 0.0, Phi, 1.0, Phi, s, s); }
     Deofa_L.Omega(spProj_Phi, tmp[0], -1, 0);
     G5R5(tmp[1], tmp[0]);
-    tmp[0] = zero;
+    tmp[0] = Zero();
     SchurSolver(Deofa_L, tmp[1], tmp[0]);
     Deofa_L.Dtilde(tmp[0], tmp[1]);
     Deofa_L.Omega(tmp[1], tmp[0], -1, 1);
@@ -198,7 +198,7 @@ int main(int argc, char **argv)
     for(int s=0; s<Ls; ++s){ axpby_ssp_pplus(spProj_Phi, 0.0, Phi, 1.0, Phi, s, s); }
     Deofa_R.Omega(spProj_Phi, tmp[0], 1, 0);
     G5R5(tmp[1], tmp[0]);
-    tmp[0] = zero;
+    tmp[0] = Zero();
     SchurSolver(Deofa_R, tmp[1], tmp[0]);
     Deofa_R.Dtilde(tmp[0], tmp[1]);
     Deofa_R.Omega(tmp[1], tmp[0], 1, 1);
diff --git a/tests/debug/Test_split_laplacian.cc b/tests/debug/Test_split_laplacian.cc
index 174dc1b1..1e41e112 100644
--- a/tests/debug/Test_split_laplacian.cc
+++ b/tests/debug/Test_split_laplacian.cc
@@ -29,7 +29,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 
 int main (int argc, char ** argv)
@@ -38,13 +37,13 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
+  Coordinate latt_size   = GridDefaultLatt();
   int nd   = latt_size.size();
   int ndm1 = nd-1;
 
-  std::vector<int> simd_layout = GridDefaultSimd(nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
-  std::vector<int> mpi_split (mpi_layout.size(),1);
+  Coordinate simd_layout = GridDefaultSimd(nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
+  Coordinate mpi_split (mpi_layout.size(),1);
 
   std::cout << " Full " << GridCmdVectorIntToString(latt_size)  << " subgrid"         <<std::endl;
   std::cout << " Full " << GridCmdVectorIntToString(mpi_layout) << " sub communicator"<<std::endl;
@@ -54,9 +53,9 @@ int main (int argc, char ** argv)
 						    simd_layout,
 						    mpi_layout);
 
-  std::vector<int> latt_m  = latt_size;   latt_m[nd-1] = 1;
-  std::vector<int> mpi_m   = mpi_layout;  mpi_m [nd-1] = 1;
-  std::vector<int> simd_m  = GridDefaultSimd(ndm1,vComplex::Nsimd()); simd_m.push_back(1);
+  Coordinate latt_m  = latt_size;   latt_m[nd-1] = 1;
+  Coordinate mpi_m   = mpi_layout;  mpi_m [nd-1] = 1;
+  Coordinate simd_m  = GridDefaultSimd(ndm1,vComplex::Nsimd()); simd_m.push_back(1);
 
 
   std::cout << " Requesting " << GridCmdVectorIntToString(latt_m)<< " subgrid"         <<std::endl;
@@ -88,7 +87,7 @@ int main (int argc, char ** argv)
 
   auto local_latt = GridN->LocalDimensions();
 
-  Full_cpy = zero;
+  Full_cpy = Zero();
   std::vector<int> seeds({1,2,3,4});
   GridParallelRNG          RNG(GridN);  RNG.SeedFixedIntegers(seeds);
 
diff --git a/tests/debug/Test_zmm.cc b/tests/debug/Test_zmm.cc
deleted file mode 100644
index 92bae882..00000000
--- a/tests/debug/Test_zmm.cc
+++ /dev/null
@@ -1,494 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./tests/Test_zmm.cc
-
-Copyright (C) 2015
-
-Author: paboyle <paboyle@ph.ed.ac.uk>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-#include <Grid/Grid.h>
-
-#ifdef TEST_ZMM
-
-int main(int argc,char **argv)
-{
-  return 0;
-}
-
-#if 0
-#include <simd/Intel512wilson.h>
-using namespace Grid;
-using namespace Grid::QCD;
-
-void ZmulF(void *ptr1,void *ptr2,void *ptr3);
-void Zmul(void *ptr1,void *ptr2,void *ptr3);
-void WilsonDslashAvx512(void *ptr1,void *ptr2,void *ptr3);
-void WilsonDslashAvx512F(void *ptr1,void *ptr2,void *ptr3);
-void TimesIAvx512F(void *ptr1,void *ptr3);
-void TimesIAvx512(void *ptr1,void *ptr3);
-void TimesMinusIAvx512F(void *ptr1,void *ptr3);
-void TimesMinusIAvx512(void *ptr1,void *ptr3);
-
-
-
-int main(int argc,char **argv)
-{
-  Grid_init(&argc,&argv);
-
-  
-  std::vector<int> latt4 = GridDefaultLatt();
-  const int Ls=16;
-
-  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(latt4, GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
-  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
-  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
-  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
-
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
-  int threads = GridThread::GetThreads();
-
-  std::vector<int> seeds4({1,2,3,4});
-  std::vector<int> seeds5({5,6,7,8});
-
-  GridSerialRNG sRNG; sRNG.SeedFixedIntegers(seeds4);
-
-  vColourMatrixD mat;
-  vHalfSpinColourVectorD vec;
-  vHalfSpinColourVectorD vec1;
-  vHalfSpinColourVectorD vec2;
-  vHalfSpinColourVectorD vec3;
- 
-  vHalfSpinColourVectorD matvec;
-  vHalfSpinColourVectorD ref;
-  vComplexD err;
-
-  random(sRNG,vec1); 
-  vec1 = std::complex<double>(0.1,3.0);
-  random(sRNG,vec2);
-  vec2=2.0;
-  random(sRNG,vec3);
-
-  //std::cout << "Zmul  vec1"<<vec1<<" &vec1 "<<& vec1<<std::endl;
-  //std::cout << "Zmul  vec2"<<vec2<<" &vec2 "<<& vec2<<std::endl;
-  //std::cout << "Zmul  vec3"<<vec3<<" &vec3 "<<& vec3<<std::endl;
-  for(int sp=0;sp<2;sp++){
-  for(int co=0;co<3;co++){
-  ref()(sp)(co) = vec1()(sp)(co)*vec2()(sp)(co);
-  }}
-
-  Zmul((void *)&vec1,(void *)&vec2,(void *)&vec3);
-  //std::cout << "Zmul  vec3"<<vec3<<" &vec3 "<<& vec3<<std::endl;
-  //std::cout << "Zmul \n\t ref "<<ref<<"\n\t vec3"<<vec3 <<std::endl;
-  ref = ref - vec3;
-  err = TensorRemove(innerProduct(ref,ref));
-  std::cout <<"Zmul diff   "<< Reduce(err)<<std::endl;
-
-  random(sRNG,mat);
-  mat = zero;
-  mat()()(0,0) = 1.0;
-  random(sRNG,vec);
-
-  ref = mat*vec;
-  
-  WilsonDslashAvx512((void *)&vec, (void *)&mat,(void *)&matvec);
-
-  //std::cout << ref   <<std::endl;
-  //std::cout << matvec<<std::endl;
-  ref = ref - matvec;
-  err = TensorRemove(innerProduct(ref,ref));
-  std::cout <<"Double SU3 x 2spin diff   "<< Reduce(err)<<std::endl;
-  vColourMatrixF matF;
-  vHalfSpinColourVectorF vec1F;
-  vHalfSpinColourVectorF vec2F;
-  vHalfSpinColourVectorF vec3F;
-  vHalfSpinColourVectorF vecF;
-  vHalfSpinColourVectorF matvecF;
-  vHalfSpinColourVectorF refF;
-  vComplexF errF;
-
-  random(sRNG,matF);
-  matF = zero;
-  matF()()(0,0)=1.0;
-  random(sRNG,vecF);
-
-  refF = matF*vecF;
-
-  WilsonDslashAvx512F((void *)&vecF, (void *)&matF,(void *)&matvecF);
-  //std::cout << refF   <<std::endl;
-  //std::cout << matvecF<<std::endl;
- 
-  refF = refF-matvecF;
-  errF = TensorRemove(innerProduct(refF,refF));
-  std::cout <<"Single SU3 x 2spin diff   "<< Reduce(errF)<<std::endl;
-
-  TimesIAvx512F((void *)&vecF,(void *)&matvecF);
-  //std::cout << timesI(vecF)<<std::endl;
-  //std::cout << matvecF<<std::endl;
-  refF = timesI(vecF)-matvecF;
-  errF = TensorRemove(innerProduct(refF,refF));
-  std::cout <<" timesI single diff  "<< Reduce(errF)<<std::endl;
-
-  TimesIAvx512((void *)&vec,(void *)&matvec);
-  //std::cout << timesI(vec)<<std::endl;
-  //std::cout << matvec<<std::endl;
- 
-  ref = timesI(vec)-matvec;
-  err = TensorRemove(innerProduct(ref,ref));
-  std::cout <<" timesI double diff  "<< Reduce(err)<<std::endl;
-
-  TimesMinusIAvx512F((void *)&vecF,(void *)&matvecF);
-  //std::cout << timesMinusI(vecF)<<std::endl;
-  //std::cout << matvecF<<std::endl;
-  refF = timesMinusI(vecF)-matvecF;
-  errF = TensorRemove(innerProduct(refF,refF));
-  std::cout <<" timesMinusI single diff  "<< Reduce(errF)<<std::endl;
-
-  TimesMinusIAvx512((void *)&vec,(void *)&matvec);
-  //std::cout << timesMinusI(vec)<<std::endl;
-  //std::cout << matvec<<std::endl;
-
-  ref = timesMinusI(vec)-matvec;
-  err = TensorRemove(innerProduct(ref,ref));
-  std::cout <<" timesMinusI double diff  "<< Reduce(err)<<std::endl;
-
-
-  LatticeFermion src (FGrid);
-  LatticeFermion tmp (FGrid);
-  LatticeFermion srce(FrbGrid);
-
-  LatticeFermion resulto(FrbGrid); resulto=zero;
-  LatticeFermion resulta(FrbGrid); resulta=zero;
-  LatticeFermion diff(FrbGrid); 
-  LatticeGaugeField Umu(UGrid);
-
-
-  GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
-  GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
-  random(RNG5,src);
-#if 1
-  SU3::HotConfiguration(RNG4,Umu);
-#else
-  int mmu=2;
-  std::vector<LatticeColourMatrix> U(4,UGrid);
-  for(int mu=0;mu<Nd;mu++){
-    U[mu] = PeekIndex<LorentzIndex>(Umu,mu);
-    if ( mu!=mmu ) U[mu] = zero;
-    if ( mu==mmu ) U[mu] = 1.0;
-    PokeIndex<LorentzIndex>(Umu,U[mu],mu);
-  }
-#endif
- pickCheckerboard(Even,srce,src);
-
-  RealD mass=0.1;
-  RealD M5  =1.8;
-  DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
-
-  std::cout<<GridLogMessage << "Calling Dw"<<std::endl;
-  int ncall=50;
-  double t0=usecond();
-  for(int i=0;i<ncall;i++){
-    Dw.DhopOE(srce,resulto,0);
-  }
-  double t1=usecond();
-
-  double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
-  double flops=1344*volume/2;
-  
-  std::cout<<GridLogMessage << "Called Dw"<<std::endl;
-  std::cout<<GridLogMessage << "norm result "<< norm2(resulto)<<std::endl;
-  std::cout<<GridLogMessage << "mflop/s =   "<< flops*ncall/(t1-t0)<<std::endl;
-
-  QCD::WilsonFermion5DStatic::AsmOptDslash=1;
-  t0=usecond();
-  for(int i=0;i<ncall;i++){
-    Dw.DhopOE(srce,resulta,0);
-  }
-  t1=usecond();
-
-#if 1
-  for(int i=0;i< PerformanceCounter::NumTypes(); i++ ){
-    Dw.DhopOE(srce,resulta,0);
-    PerformanceCounter Counter(i);
-    Counter.Start();
-    Dw.DhopOE(srce,resulta,0);
-    Counter.Stop();
-    Counter.Report();
-  }
-#endif
-  //resulta = (-0.5) * resulta;
-
-  std::cout<<GridLogMessage << "Called Asm Dw"<<std::endl;
-  std::cout<<GridLogMessage << "norm result "<< norm2(resulta)<<std::endl;
-  std::cout<<GridLogMessage << "mflop/s =   "<< flops*ncall/(t1-t0)<<std::endl;
-  diff = resulto-resulta;
-  std::cout<<GridLogMessage << "diff "<< norm2(diff)<<std::endl;
-  std::cout<<std::endl;
-#if 0
-  std::cout<<"=========== result Grid ============="<<std::endl;
-  std::cout<<std::endl;
-  tmp = zero;
-  setCheckerboard(tmp,resulto);
-  std::cout<<tmp<<std::endl;
-  std::cout<<std::endl;
-  std::cout<<"=========== result ASM ============="<<std::endl;
-  std::cout<<std::endl;
-  tmp = zero;
-  setCheckerboard(tmp,resulta);
-  std::cout<<tmp<<std::endl;
-#endif
-}
-
-#include <simd/Intel512double.h>
-
-#define zz Z0
-
-
-void Zmul(void *ptr1,void *ptr2,void *ptr3)
-{
-  __asm__ ("mov     $0xAAAA, %%eax "  : : :"%eax");
-  __asm__ ("kmovw    %%eax, %%k6 " : : :);
-  __asm__ ("mov     $0x5555, %%eax "  : : :"%eax");
-  __asm__ ("kmovw    %%eax, %%k7 " : : :);
-
-#define CC result_00
-  LOAD64(%r9,ptr1);
-  LOAD64(%r8,ptr2);
-  LOAD64(%r10,ptr3)
-  __asm__ (
-  VLOAD(0,%r8,CC)
-  ZLOAD(0,%r9,Chi_00,Z0) 
-  ZMUL(Chi_00,Z0,CC,UChi_00,Z1)
-  //VSTORE(0,%r10,UChi_00)
-  //VSTORE(1,%r10,Z1)
-  ZEND1(UChi_00,Z1,Z0)
-  //VSTORE(2,%r10,UChi_00)
-  ZEND2(UChi_00,Z1,Z0)
-  //VSTORE(3,%r10,UChi_00)
-  VSTORE(0,%r10,UChi_00)
-  VLOAD(1,%r8,CC)
-  ZLOAD(1,%r9,Chi_01,Z0) 
-  ZMUL(Chi_01,Z0,CC,UChi_01,Z1)
-  ZEND1(UChi_01,Z1,Z0)
-  ZEND2(UChi_01,Z1,Z0)
-  VSTORE(1,%r10,UChi_01)
-  VLOAD(2,%r8,CC)
-  ZLOAD(2,%r9,Chi_02,Z0) 
-  ZMUL(Chi_02,Z0,CC,UChi_02,Z1)
-  ZEND1(UChi_02,Z1,Z0)
-  ZEND2(UChi_02,Z1,Z0)
-  VSTORE(2,%r10,UChi_02)
-  VLOAD(3,%r8,CC)
-  ZLOAD(3,%r9,Chi_10,Z0) 
-  ZMUL(Chi_10,Z0,CC,UChi_10,Z1)
-  ZEND1(UChi_10,Z1,Z0)
-  ZEND2(UChi_10,Z1,Z0)
-  VSTORE(3,%r10,UChi_10)
-  VLOAD(4,%r8,CC)
-  ZLOAD(4,%r9,Chi_11,Z0) 
-  ZMUL(Chi_11,Z0,CC,UChi_11,Z1)
-  ZEND1(UChi_11,Z1,Z0)
-  ZEND2(UChi_11,Z1,Z0)
-  VSTORE(4,%r10,UChi_11)
-  VLOAD(5,%r8,CC)
-  ZLOAD(5,%r9,Chi_12,Z0) 
-  ZMUL(Chi_12,Z0,CC,UChi_12,Z1)
-  ZEND1(UChi_12,Z1,Z0)
-  ZEND2(UChi_12,Z1,Z0)
-  VSTORE(5,%r10,UChi_12)
-  );
-}
-void TimesMinusIAvx512(void *ptr1,void *ptr3)
-{
-  __asm__ ("mov     $0xAAAA, %%eax "  : : :"%eax");
-  __asm__ ("kmovw    %%eax, %%k6 " : : :);
-  __asm__ ("mov     $0x5555, %%eax "  : : :"%eax");
-  __asm__ ("kmovw    %%eax, %%k7 " : : :);
-
-  MASK_REGS;
-
-  LOAD_CHI(ptr1);
-
-  __asm__ (
-  VZERO(zz)
-  VTIMESMINUSI(Chi_00,UChi_00,zz)
-  VTIMESMINUSI(Chi_01,UChi_01,zz)
-  VTIMESMINUSI(Chi_02,UChi_02,zz)
-  VTIMESMINUSI(Chi_10,UChi_10,zz)
-  VTIMESMINUSI(Chi_11,UChi_11,zz)
-  VTIMESMINUSI(Chi_12,UChi_12,zz)
-  );
-
-  SAVE_UCHI(ptr3);
-}
-
-void TimesIAvx512(void *ptr1,void *ptr3)
-{
-  __asm__ ("mov     $0xAAAA, %%eax "  : : :"%eax");
-  __asm__ ("kmovw    %%eax, %%k6 " : : :);
-  __asm__ ("mov     $0x5555, %%eax "  : : :"%eax");
-  __asm__ ("kmovw    %%eax, %%k7 " : : :);
-
-  MASK_REGS;
-  
-  LOAD_CHI(ptr1);
-  
-  __asm__ (
-  VZERO(zz)
-  VTIMESI(Chi_00,UChi_00,zz)
-  VTIMESI(Chi_01,UChi_01,zz)
-  VTIMESI(Chi_02,UChi_02,zz)
-  VTIMESI(Chi_10,UChi_10,zz)
-  VTIMESI(Chi_11,UChi_11,zz)
-  VTIMESI(Chi_12,UChi_12,zz)
-  );
-
-  SAVE_UCHI(ptr3);
-}
-
-void WilsonDslashAvx512(void *ptr1,void *ptr2,void *ptr3)
-{
-  int return_address;
-  // prototype computed goto to eliminate ABI save restore on call/return in
-  // generated assembly.
-  static void * table[] = { &&save, &&mult };
-
-  MASK_REGS;
-
-  LOAD_CHI(ptr1);
-
-  return_address = 0;
-  goto mult;
-
- save:
-  SAVE_UCHI(ptr3);
-  return;
-
- mult:
-  MULT_2SPIN(ptr2);
-  goto *table[return_address];
-
-}
-
-#include <simd/Intel512single.h>
-
-void ZmulF(void *ptr1,void *ptr2,void *ptr3)
-{
-  __asm__ ("mov     $0xAAAA, %%eax "  : : :"%eax");
-  __asm__ ("kmovw    %%eax, %%k6 " : : :);
-  __asm__ ("mov     $0x5555, %%eax "  : : :"%eax");
-  __asm__ ("kmovw    %%eax, %%k7 " : : :);
-  MASK_REGS;
-  ZLOAD(0,ptr1,Chi_00,Z0);
-  ZLOAD(1,ptr1,Chi_01,Z1);
-  ZLOAD(2,ptr1,Chi_02,Z2);
-  ZLOAD(3,ptr1,Chi_10,Z3);
-  ZLOAD(4,ptr1,Chi_11,Z4);
-  ZLOAD(5,ptr1,Chi_12,Z5);
-
-  VLOAD(0,ptr2,Chi_20);
-  VLOAD(1,ptr2,Chi_21);
-  VLOAD(2,ptr2,Chi_22);
-  VLOAD(3,ptr2,Chi_30);  
-  VLOAD(4,ptr2,Chi_31);  
-  VLOAD(5,ptr2,Chi_32);  
-
-  ZMUL(Chi_00,Z0,Chi_20,UChi_00,UChi_20);
-  ZMUL(Chi_01,Z1,Chi_21,UChi_01,UChi_21);
-  ZMUL(Chi_02,Z2,Chi_22,UChi_02,UChi_22);
-  ZMUL(Chi_10,Z3,Chi_23,UChi_10,UChi_30);
-  ZMUL(Chi_11,Z4,Chi_24,UChi_11,UChi_31);
-  ZMUL(Chi_12,Z5,Chi_25,UChi_12,UChi_32);
-  
-  ZEND1(UChi_00,UChi_20,Z0);
-  ZEND1(UChi_01,UChi_21,Z1);
-  ZEND1(UChi_02,UChi_22,Z2);
-  ZEND1(UChi_10,UChi_30,Z3);
-  ZEND1(UChi_11,UChi_31,Z4);
-  ZEND1(UChi_12,UChi_32,Z5);
-
-  ZEND2(UChi_00,UChi_20,Z0);
-  ZEND2(UChi_01,UChi_21,Z1);
-  ZEND2(UChi_02,UChi_22,Z2);
-  ZEND2(UChi_10,UChi_30,Z3);
-  ZEND2(UChi_11,UChi_31,Z4);
-  ZEND2(UChi_12,UChi_32,Z5);
-
-  SAVE_UCHI(ptr3); 
-}
-
-void TimesMinusIAvx512F(void *ptr1,void *ptr3)
-{
-  MASK_REGS;
-
-  LOAD_CHI(ptr1);
-  __asm__ (
-  VZERO(zz)
-  VTIMESMINUSI(Chi_00,UChi_00,zz)
-  VTIMESMINUSI(Chi_01,UChi_01,zz)
-  VTIMESMINUSI(Chi_02,UChi_02,zz)
-  VTIMESMINUSI(Chi_10,UChi_10,zz)
-  VTIMESMINUSI(Chi_11,UChi_11,zz)
-  VTIMESMINUSI(Chi_12,UChi_12,zz)
-           );
-  SAVE_UCHI(ptr3);
-}
-
-void TimesIAvx512F(void *ptr1,void *ptr3)
-{
-  MASK_REGS;
-  
-  LOAD_CHI(ptr1);
-  __asm__ (
-  VZERO(zz)
-  VTIMESI(Chi_00,UChi_00,zz)
-  VTIMESI(Chi_01,UChi_01,zz)
-  VTIMESI(Chi_02,UChi_02,zz)
-  VTIMESI(Chi_10,UChi_10,zz)
-  VTIMESI(Chi_11,UChi_11,zz)
-  VTIMESI(Chi_12,UChi_12,zz)
-	   );
-  SAVE_UCHI(ptr3);
-}
-
-void WilsonDslashAvx512F(void *ptr1,void *ptr2,void *ptr3)
-{
-  MASK_REGS;
-
-  LOAD_CHI(ptr1);
-
-  MULT_ADDSUB_2SPIN(ptr2);
-  //MULT_2SPIN(ptr2);
-
-  SAVE_UCHI(ptr3);
-
-  return;
-}
-#endif
-#else
-int main(int argc, char **argv)
-{
-  std::cerr << "error: no ZMM test for the selected architecture" << std::endl;
-
-  return 1;
-}
-#endif
diff --git a/tests/debug/test_Grid_jacobi.cc b/tests/debug/test_Grid_jacobi.cc
index 8561f686..5d4c89b7 100644
--- a/tests/debug/test_Grid_jacobi.cc
+++ b/tests/debug/test_Grid_jacobi.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class vobj>
 class LinearOperator {
@@ -81,10 +81,10 @@ class LinearOperatorJacobi : public LinearOperator<vobj>
 	
 	vobj *nbr;
 	if ( local && perm ){
-	  permute(tmp,src._odata[offset],ptype);
+	  permute(tmp,src[offset],ptype);
 	  nbr = &tmp;
 	} else if (local) {
-	  nbr = &src._odata[offset];
+	  nbr = &src[offset];
 	} else  {
 	  nbr = &comm_buf[offset];
 	}
@@ -182,18 +182,18 @@ int main (int argc, char ** argv)
 	Bar = Cshift(Foo,dir,disp);
 
 	// Implement a stencil code that should agree with cshift!
-	for(int i=0;i<Check._grid->oSites();i++){
+	for(int i=0;i<Check.Grid()->oSites();i++){
 
 	  int offset = myStencil._offsets [0][i];
 	  int  local = myStencil._is_local[0][i];
 	  int permute_type = myStencil._permute_type[0];
 	  int perm =myStencil._permute[0][i];
 	  if ( local && perm )
-	    permute(Check._odata[i],Foo._odata[offset],permute_type);
+	    permute(Check[i],Foo[offset],permute_type);
 	  else if (local)
-	    Check._odata[i] = Foo._odata[offset];
+	    Check[i] = Foo[offset];
 	  else 
-	    Check._odata[i] = comm_buf[offset];
+	    Check[i] = comm_buf[offset];
 	  
 
 	}
diff --git a/tests/forces/Test_contfrac_force.cc b/tests/forces/Test_contfrac_force.cc
index 2afb4dde..4eeb8c27 100644
--- a/tests/forces/Test_contfrac_force.cc
+++ b/tests/forces/Test_contfrac_force.cc
@@ -29,15 +29,14 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   const int Ls=9;
 
@@ -99,18 +98,20 @@ int main (int argc, char ** argv)
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin();i<mom.end();i++){
-      Uprime[i](mu) =
-	  U[i](mu)
-	+ mom[i](mu)*U[i](mu)*dt 
-	+ mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt/2.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt/6.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt/24.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt/120.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
+    auto mom_v = mom.View();
+    auto Uprime_v = Uprime.View();
+    auto U_v = U.View();
+    thread_foreach( i,mom_v,{
+      Uprime_v[i](mu) =
+	  U_v[i](mu)
+	+ mom_v[i](mu)*U_v[i](mu)*dt 
+	+ mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt/2.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt/6.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt/24.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt/120.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
 	;
-    }
-
+    });
   }
   
   Dcf.ImportGauge(Uprime);
@@ -123,7 +124,7 @@ int main (int argc, char ** argv)
   //////////////////////////////////////////////
 
 
-  LatticeComplex dS(UGrid); dS = zero;
+  LatticeComplex dS(UGrid); dS = Zero();
   for(int mu=0;mu<Nd;mu++){
     mommu   = PeekIndex<LorentzIndex>(UdSdU,mu);
     mommu=Ta(mommu)*2.0;
diff --git a/tests/forces/Test_dwf_force.cc b/tests/forces/Test_dwf_force.cc
index 59aee7b7..009f50b3 100644
--- a/tests/forces/Test_dwf_force.cc
+++ b/tests/forces/Test_dwf_force.cc
@@ -29,15 +29,15 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   const int Ls=8;
 
@@ -46,18 +46,7 @@ int main (int argc, char ** argv)
   GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
   GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
   
-  // Want a different conf at every run
-  // First create an instance of an engine.
-  std::random_device rnd_device;
-  // Specify the engine and distribution.
-  std::mt19937 mersenne_engine(rnd_device());
-  std::uniform_int_distribution<int> dist(1, 100);
-  
-  auto gen = std::bind(dist, mersenne_engine);
-  std::vector<int> seeds4(4);
-  generate(begin(seeds4), end(seeds4), gen);
-  
-  //std::vector<int> seeds4({1,2,3,5});
+  std::vector<int> seeds4({1,2,3,5});
   std::vector<int> seeds5({5,6,7,8});
   GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
   GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
@@ -110,18 +99,21 @@ int main (int argc, char ** argv)
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin();i<mom.end();i++){
-      Uprime[i](mu) =
-	  U[i](mu)
-	+ mom[i](mu)*U[i](mu)*dt 
-	+ mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt/2.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt/6.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt/24.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt/120.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
-	;
-    }
 
+    auto mom_v    = mom.View();
+    auto U_v      = U.View();
+    auto Uprime_v = Uprime.View();
+
+    thread_foreach( i,mom_v,{
+      Uprime_v[i](mu) =	  U_v[i](mu)
+	+ mom_v[i](mu)*U_v[i](mu)*dt 
+	+ mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt/2.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt/6.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt/24.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt/120.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
+	;
+    });
   }
   
   Ddwf.ImportGauge(Uprime);
@@ -134,7 +126,7 @@ int main (int argc, char ** argv)
   //////////////////////////////////////////////
 
 
-  LatticeComplex dS(UGrid); dS = zero;
+  LatticeComplex dS(UGrid); dS = Zero();
   for(int mu=0;mu<Nd;mu++){
     mommu   = PeekIndex<LorentzIndex>(UdSdU,mu);
     mommu=Ta(mommu)*2.0;
@@ -152,6 +144,7 @@ int main (int argc, char ** argv)
 
   ComplexD dSpred    = sum(dS);
 
+  std::cout << std::setprecision(14);
   std::cout << GridLogMessage << " S      "<<S<<std::endl;
   std::cout << GridLogMessage << " Sprime "<<Sprime<<std::endl;
   std::cout << GridLogMessage << "dS      "<<Sprime-S<<std::endl;
diff --git a/tests/forces/Test_dwf_force_eofa.cc b/tests/forces/Test_dwf_force_eofa.cc
index 18e339ab..670e7589 100644
--- a/tests/forces/Test_dwf_force_eofa.cc
+++ b/tests/forces/Test_dwf_force_eofa.cc
@@ -31,15 +31,15 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char** argv)
 {
   Grid_init(&argc, &argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   const int Ls = 8;
 
@@ -110,13 +110,20 @@ int main (int argc, char** argv)
     PokeIndex<LorentzIndex>(mom, mommu, mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin(); i<mom.end(); i++){
-      Uprime[i](mu) = U[i](mu) + mom[i](mu)*U[i](mu)*dt + mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt/2.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt/6.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt/24.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt/120.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt*dt/720.0);
-    }
+    auto mom_v    = mom.View();
+    auto U_v      = U.View();
+    auto Uprime_v = Uprime.View();
+
+    thread_foreach(i,mom_v,{
+      Uprime_v[i](mu) =	  U_v[i](mu)
+	+ mom_v[i](mu)*U_v[i](mu)*dt 
+	+ mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt/2.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt/6.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt/24.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt/120.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
+	;
+    });
   }
 
   /*Ddwf.ImportGauge(Uprime);
@@ -130,7 +137,7 @@ int main (int argc, char** argv)
   //////////////////////////////////////////////
 
   LatticeComplex dS(UGrid);
-  dS = zero;
+  dS = Zero();
   for(int mu=0; mu<Nd; mu++){
     mommu = PeekIndex<LorentzIndex>(UdSdU, mu);
     mommu = Ta(mommu)*2.0;
diff --git a/tests/forces/Test_dwf_gpforce.cc b/tests/forces/Test_dwf_gpforce.cc
index 98453aab..d762e22a 100644
--- a/tests/forces/Test_dwf_gpforce.cc
+++ b/tests/forces/Test_dwf_gpforce.cc
@@ -29,7 +29,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
  
 
@@ -37,9 +37,9 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   const int Ls=8;
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
@@ -105,7 +105,6 @@ int main (int argc, char ** argv)
   RealD dt = 0.0001;
   RealD Hmom = 0.0;
   RealD Hmomprime = 0.0;
-  RealD Hmompp    = 0.0;
   LatticeColourMatrix mommu(UGrid); 
   LatticeColourMatrix forcemu(UGrid); 
   LatticeGaugeField mom(UGrid); 
@@ -120,18 +119,20 @@ int main (int argc, char ** argv)
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin();i<mom.end();i++){
-      Uprime[i](mu) =
-	  U[i](mu)
-	+ mom[i](mu)*U[i](mu)*dt 
-	+ mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt/2.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt/6.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt/24.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt/120.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
-	;
-    }
+    auto mom_v    = mom.View();
+    auto U_v      = U.View();
+    auto Uprime_v = Uprime.View();
 
+    thread_foreach(i,mom_v,{
+      Uprime_v[i](mu) =	  U_v[i](mu)
+	+ mom_v[i](mu)*U_v[i](mu)*dt 
+	+ mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt/2.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt/6.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt/24.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt/120.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
+	;
+    });
   }
 
   std::cout << GridLogMessage <<"Initial mom hamiltonian is "<< Hmom <<std::endl;
@@ -157,9 +158,9 @@ int main (int argc, char ** argv)
     std::cout << GridLogMessage<< " dsdumu + dag  " << norm2(mommu)<<std::endl;
   }
 
-  LatticeComplex dS(UGrid); dS = zero;
-  LatticeComplex dSmom(UGrid); dSmom = zero;
-  LatticeComplex dSmom2(UGrid); dSmom2 = zero;
+  LatticeComplex dS(UGrid); dS = Zero();
+  LatticeComplex dSmom(UGrid); dSmom = Zero();
+  LatticeComplex dSmom2(UGrid); dSmom2 = Zero();
   for(int mu=0;mu<Nd;mu++){
     mommu   = PeekIndex<LorentzIndex>(UdSdU,mu);
     mommu=Ta(mommu)*2.0;
diff --git a/tests/forces/Test_dwf_gpforce_eofa.cc b/tests/forces/Test_dwf_gpforce_eofa.cc
index 32b3ed6b..66ae9dcf 100644
--- a/tests/forces/Test_dwf_gpforce_eofa.cc
+++ b/tests/forces/Test_dwf_gpforce_eofa.cc
@@ -31,7 +31,6 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 typedef GparityWilsonImplR FermionImplPolicy;
 typedef GparityDomainWallEOFAFermionR FermionAction;
@@ -41,9 +40,9 @@ int main (int argc, char** argv)
 {
   Grid_init(&argc, &argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   const int Ls = 8;
 
@@ -115,13 +114,20 @@ int main (int argc, char** argv)
     PokeIndex<LorentzIndex>(mom, mommu, mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin(); i<mom.end(); i++){
-      Uprime[i](mu) = U[i](mu) + mom[i](mu)*U[i](mu)*dt + mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt/2.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt/6.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt/24.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt/120.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt*dt/720.0);
-    }
+    auto mom_v    = mom.View();
+    auto U_v      = U.View();
+    auto Uprime_v = Uprime.View();
+
+    thread_foreach(i,mom_v,{
+      Uprime_v[i](mu) =	  U_v[i](mu)
+	+ mom_v[i](mu)*U_v[i](mu)*dt 
+	+ mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt/2.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt/6.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt/24.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt/120.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
+	;
+    });
   }
 
   /*Ddwf.ImportGauge(Uprime);
@@ -135,7 +141,7 @@ int main (int argc, char** argv)
   //////////////////////////////////////////////
 
   LatticeComplex dS(UGrid);
-  dS = zero;
+  dS = Zero();
   for(int mu=0; mu<Nd; mu++){
     mommu = PeekIndex<LorentzIndex>(UdSdU, mu);
     mommu = Ta(mommu)*2.0;
diff --git a/tests/forces/Test_gp_plaq_force.cc b/tests/forces/Test_gp_plaq_force.cc
index e121f21b..c4e214bb 100644
--- a/tests/forces/Test_gp_plaq_force.cc
+++ b/tests/forces/Test_gp_plaq_force.cc
@@ -29,15 +29,14 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  auto latt_size   = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
 
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
@@ -55,10 +54,11 @@ int main (int argc, char ** argv)
   SU3::HotConfiguration(pRNG,U);
   
   double beta = 1.0;
-  double c1   = 0.331;
-
-  //ConjugatePlaqPlusRectangleActionR Action(beta,c1);
   ConjugateWilsonGaugeActionR Action(beta);
+
+  //  double c1   = 0.331;
+  //ConjugatePlaqPlusRectangleActionR Action(beta,c1);
+
   //WilsonGaugeActionR Action(beta);
 
   ComplexD S    = Action.S(U);
@@ -85,9 +85,12 @@ int main (int argc, char ** argv)
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin();i<mom.end();i++){ // exp(pmu dt) * Umu
-      Uprime[i](mu) = U[i](mu) + mom[i](mu)*U[i](mu)*dt ;
-    }
+    auto Uprime_v = Uprime.View();
+    auto U_v = U.View();
+    auto mom_v = mom.View();
+    thread_foreach(i,mom_v,{ // exp(pmu dt) * Umu
+      Uprime_v[i](mu) = U_v[i](mu) + mom_v[i](mu)*U_v[i](mu)*dt ;
+    });
   }
 
   ComplexD Sprime    = Action.S(Uprime);
@@ -96,7 +99,7 @@ int main (int argc, char ** argv)
   // Use derivative to estimate dS
   //////////////////////////////////////////////
 
-  LatticeComplex dS(&Grid); dS = zero;
+  LatticeComplex dS(&Grid); dS = Zero();
 
   for(int mu=0;mu<Nd;mu++){
 
diff --git a/tests/forces/Test_gp_rect_force.cc b/tests/forces/Test_gp_rect_force.cc
index 6b3349e0..2573af6a 100644
--- a/tests/forces/Test_gp_rect_force.cc
+++ b/tests/forces/Test_gp_rect_force.cc
@@ -29,7 +29,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
  
 
@@ -37,9 +37,9 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
@@ -60,7 +60,7 @@ int main (int argc, char ** argv)
   double c1   = 0.331;
 
   ConjugatePlaqPlusRectangleActionR Action(beta,c1);
-  //  ConjugateWilsonGaugeActionR Action(beta);
+  //ConjugateWilsonGaugeActionR Action(beta);
   //WilsonGaugeActionR Action(beta);
 
   ComplexD S    = Action.S(U);
@@ -73,7 +73,7 @@ int main (int argc, char ** argv)
   ////////////////////////////////////
   // Modify the gauge field a little 
   ////////////////////////////////////
-  RealD dt = 0.0001;
+  RealD dt = 0.01;
 
   LatticeColourMatrix mommu(&Grid); 
   LatticeColourMatrix forcemu(&Grid); 
@@ -87,9 +87,12 @@ int main (int argc, char ** argv)
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin();i<mom.end();i++){ // exp(pmu dt) * Umu
-      Uprime[i](mu) = U[i](mu) + mom[i](mu)*U[i](mu)*dt ;
-    }
+    auto mom_v   = mom.View();
+    auto Uprime_v= Uprime.View();
+    auto U_v     = U.View();
+    thread_foreach(i,mom_v,{ // exp(pmu dt) * Umu
+      Uprime_v[i](mu) = U_v[i](mu) + mom_v[i](mu)*U_v[i](mu)*dt ;
+    });
   }
 
   ComplexD Sprime    = Action.S(Uprime);
@@ -98,7 +101,7 @@ int main (int argc, char ** argv)
   // Use derivative to estimate dS
   //////////////////////////////////////////////
 
-  LatticeComplex dS(&Grid); dS = zero;
+  LatticeComplex dS(&Grid); dS = Zero();
 
   for(int mu=0;mu<Nd;mu++){
 
@@ -119,7 +122,7 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << " Sprime "<<Sprime<<std::endl;
   std::cout << GridLogMessage << "dS      "<<Sprime-S<<std::endl;
   std::cout << GridLogMessage << "pred dS "<< dSpred <<std::endl;
-  assert( fabs(real(Sprime-S-dSpred)) < 1.0e-2 ) ;
+  assert( fabs(real(Sprime-S-dSpred)) < 1.0e-1 ) ;
   std::cout<< GridLogMessage << "Done" <<std::endl;
   Grid_finalize();
 }
diff --git a/tests/forces/Test_gpdwf_force.cc b/tests/forces/Test_gpdwf_force.cc
index ca9b0a2c..09a1dc4b 100644
--- a/tests/forces/Test_gpdwf_force.cc
+++ b/tests/forces/Test_gpdwf_force.cc
@@ -29,7 +29,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
  
 
@@ -37,9 +37,9 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   const int Ls=8;
 
@@ -105,18 +105,19 @@ int main (int argc, char ** argv)
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin();i<mom.end();i++){
-      Uprime[i](mu) =
-	  U[i](mu)
-	+ mom[i](mu)*U[i](mu)*dt 
-	+ mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt/2.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt/6.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt/24.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt/120.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
+    auto U_v = U.View();
+    auto mom_v = mom.View();
+    auto Uprime_v = Uprime.View();
+    thread_foreach(i,mom_v,{
+      Uprime_v[i](mu) = U_v[i](mu)
+	+ mom_v[i](mu)*U_v[i](mu)*dt 
+	+ mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt/2.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt/6.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt/24.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt/120.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
 	;
-    }
-
+    });
   }
   
   Ddwf.ImportGauge(Uprime);
@@ -128,7 +129,7 @@ int main (int argc, char ** argv)
   // Use derivative to estimate dS
   //////////////////////////////////////////////
 
-  LatticeComplex dS(UGrid); dS = zero;
+  LatticeComplex dS(UGrid); dS = Zero();
   for(int mu=0;mu<Nd;mu++){
     mommu   = PeekIndex<LorentzIndex>(UdSdU,mu);
     mommu=Ta(mommu)*2.0;
@@ -165,19 +166,19 @@ int main (int argc, char ** argv)
   //
   // Gparity --- deriv is pc Uc dSdUc + p U dSdU 
   //
-  //	Pmu = zero;
+  //	Pmu = Zero();
   //	for(int mu=0;mu<Nd;mu++){
   //	  SU<Ncol>::GaussianFundamentalLieAlgebraMatrix(pRNG, Pmu);
   //	  PokeIndex<LorentzIndex>(P, Pmu, mu);
   //	}
   //
   //
-  //    GridBase *grid = out._grid;
+  //    GridBase *grid = out.Grid();
   //    LatticeReal ca (grid);
   //    LatticeMatrix  la (grid);
   //    Complex ci(0.0,scale);
   //    Matrix ta;
-  //    out=zero;
+  //    out=Zero();
   //    for(int a=0;a<generators();a++){
   //      gaussian(pRNG,ca); 
   //      generator(a,ta);
diff --git a/tests/forces/Test_gpwilson_force.cc b/tests/forces/Test_gpwilson_force.cc
index e52ed7ee..cd30d898 100644
--- a/tests/forces/Test_gpwilson_force.cc
+++ b/tests/forces/Test_gpwilson_force.cc
@@ -29,26 +29,20 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
-
- 
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
-
-  const int Ls=8;
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), 
 								   GridDefaultSimd(Nd,vComplex::Nsimd()),
 								   GridDefaultMpi());
   GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
   GridCartesian         * FGrid   = UGrid;
-  GridRedBlackCartesian * FrbGrid = UrbGrid;
 
   std::vector<int> seeds4({1,2,3,4});
   GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
@@ -69,7 +63,6 @@ int main (int argc, char ** argv)
   // Unmodified matrix element
   ////////////////////////////////////
   RealD mass=0.01; 
-  RealD M5=1.8; 
 
   const int nu = 3;
   std::vector<int> twists(Nd,0);  twists[nu] = 1;
@@ -106,17 +99,21 @@ int main (int argc, char ** argv)
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin();i<mom.end();i++){
-      Uprime[i](mu) =
-	  U[i](mu)
-	+ mom[i](mu)*U[i](mu)*dt 
-	+ mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt/2.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt/6.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt/24.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt/120.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
+    auto mom_v    = mom.View();
+    auto U_v      = U.View();
+    auto Uprime_v = Uprime.View();
+
+    thread_foreach(i,mom_v,{
+      Uprime_v[i](mu) =	  U_v[i](mu)
+	+ mom_v[i](mu)*U_v[i](mu)*dt 
+	+ mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt/2.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt/6.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt/24.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt/120.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
 	;
-    }
+    });
+
   }
   
   Wil.ImportGauge(Uprime);
@@ -134,7 +131,7 @@ int main (int argc, char ** argv)
     PokeIndex<LorentzIndex>(UdSdU,mommu,mu);
   }
 
-  LatticeComplex dS(UGrid); dS = zero;
+  LatticeComplex dS(UGrid); dS = Zero();
   for(int mu=0;mu<Nd;mu++){
     forcemu = PeekIndex<LorentzIndex>(UdSdU,mu);
     mommu   = PeekIndex<LorentzIndex>(mom,mu);
diff --git a/tests/forces/Test_laplacian_force.cc b/tests/forces/Test_laplacian_force.cc
index bae80c0b..639378dc 100644
--- a/tests/forces/Test_laplacian_force.cc
+++ b/tests/forces/Test_laplacian_force.cc
@@ -29,17 +29,14 @@ Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
-
-#define parallel_for PARALLEL_FOR_LOOP for
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
@@ -114,7 +111,7 @@ int main (int argc, char ** argv)
   // Use derivative to estimate dS
   //////////////////////////////////////////////
 
-  LatticeComplex dS(&Grid); dS = zero;
+  LatticeComplex dS(&Grid); dS = Zero();
 
   for(int mu=0;mu<Nd;mu++){
     auto UdSdUmu = PeekIndex<LorentzIndex>(UdSdU,mu);
@@ -150,7 +147,7 @@ int main (int argc, char ** argv)
 
   // Prediciton
 
-  dS = zero;
+  dS = Zero();
    for(int mu=0;mu<Nd;mu++){
     auto dSdPmu = PeekIndex<LorentzIndex>(UdSdP,mu);
     auto Pmu = PeekIndex<LorentzIndex>(P,mu);
diff --git a/tests/forces/Test_mobius_force.cc b/tests/forces/Test_mobius_force.cc
index 06a17ffe..a1c4e930 100644
--- a/tests/forces/Test_mobius_force.cc
+++ b/tests/forces/Test_mobius_force.cc
@@ -29,15 +29,15 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   const int Ls=8;
 
@@ -101,18 +101,19 @@ int main (int argc, char ** argv)
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin();i<mom.end();i++){
-      Uprime[i](mu) =
-	  U[i](mu)
-	+ mom[i](mu)*U[i](mu)*dt 
-	+ mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt/2.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt/6.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt/24.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt/120.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
+    auto U_v      = U.View();
+    auto mom_v    = mom.View();
+    auto Uprime_v = Uprime.View();
+    thread_foreach(i,mom_v,{
+      Uprime_v[i](mu) =  U_v[i](mu)
+	+ mom_v[i](mu)*U_v[i](mu)*dt 
+	+ mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt/2.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt/6.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt/24.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt/120.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
 	;
-    }
-
+    });
   }
   
   Ddwf.ImportGauge(Uprime);
@@ -125,7 +126,7 @@ int main (int argc, char ** argv)
   //////////////////////////////////////////////
 
 
-  LatticeComplex dS(UGrid); dS = zero;
+  LatticeComplex dS(UGrid); dS = Zero();
   for(int mu=0;mu<Nd;mu++){
     mommu   = PeekIndex<LorentzIndex>(UdSdU,mu);
     mommu=Ta(mommu)*2.0;
diff --git a/tests/forces/Test_mobius_force_eofa.cc b/tests/forces/Test_mobius_force_eofa.cc
index 68a2931f..f71e2d41 100644
--- a/tests/forces/Test_mobius_force_eofa.cc
+++ b/tests/forces/Test_mobius_force_eofa.cc
@@ -31,15 +31,15 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char** argv)
 {
   Grid_init(&argc, &argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   const int Ls = 8;
 
@@ -112,13 +112,20 @@ int main (int argc, char** argv)
     PokeIndex<LorentzIndex>(mom, mommu, mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin(); i<mom.end(); i++){
-      Uprime[i](mu) = U[i](mu) + mom[i](mu)*U[i](mu)*dt + mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt/2.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt/6.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt/24.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt/120.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt*dt/720.0);
-    }
+    auto mom_v    = mom.View();
+    auto U_v      = U.View();
+    auto Uprime_v = Uprime.View();
+
+    thread_foreach(i,mom_v,{
+      Uprime_v[i](mu) =	  U_v[i](mu)
+	+ mom_v[i](mu)*U_v[i](mu)*dt 
+	+ mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt/2.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt/6.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt/24.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt/120.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
+	;
+    });
   }
 
   /*Ddwf.ImportGauge(Uprime);
@@ -132,7 +139,7 @@ int main (int argc, char** argv)
   //////////////////////////////////////////////
 
   LatticeComplex dS(UGrid);
-  dS = zero;
+  dS = Zero();
   for(int mu=0; mu<Nd; mu++){
     mommu = PeekIndex<LorentzIndex>(UdSdU, mu);
     mommu = Ta(mommu)*2.0;
diff --git a/tests/forces/Test_mobius_gpforce_eofa.cc b/tests/forces/Test_mobius_gpforce_eofa.cc
index 577b4477..4975f36d 100644
--- a/tests/forces/Test_mobius_gpforce_eofa.cc
+++ b/tests/forces/Test_mobius_gpforce_eofa.cc
@@ -31,7 +31,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 typedef GparityWilsonImplR FermionImplPolicy;
 typedef GparityMobiusEOFAFermionR FermionAction;
@@ -41,9 +41,9 @@ int main (int argc, char** argv)
 {
   Grid_init(&argc, &argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   const int Ls = 8;
 
@@ -115,15 +115,17 @@ int main (int argc, char** argv)
     SU3::GaussianFundamentalLieAlgebraMatrix(RNG4, mommu); // Traceless antihermitian momentum; gaussian in lie alg
 
     PokeIndex<LorentzIndex>(mom, mommu, mu);
-
+    auto U_v      = U.View();
+    auto mom_v    = mom.View();
+    auto Uprime_v = Uprime.View();
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin(); i<mom.end(); i++){
-      Uprime[i](mu) = U[i](mu) + mom[i](mu)*U[i](mu)*dt + mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt/2.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt/6.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt/24.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt/120.0)
-                        + mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt*dt/720.0);
-    }
+    thread_foreach( i, mom_v,{
+      Uprime_v[i](mu) = U_v[i](mu) + mom_v[i](mu)*U_v[i](mu)*dt + mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt/2.0)
+                        + mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt/6.0)
+                        + mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt/24.0)
+                        + mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt/120.0)
+                        + mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt*dt/720.0);
+    });
   }
 
   /*Ddwf.ImportGauge(Uprime);
@@ -137,7 +139,7 @@ int main (int argc, char** argv)
   //////////////////////////////////////////////
 
   LatticeComplex dS(UGrid);
-  dS = zero;
+  dS = Zero();
   for(int mu=0; mu<Nd; mu++){
     mommu = PeekIndex<LorentzIndex>(UdSdU, mu);
     mommu = Ta(mommu)*2.0;
diff --git a/tests/forces/Test_partfrac_force.cc b/tests/forces/Test_partfrac_force.cc
index 81f9aaa5..3ea2c6aa 100644
--- a/tests/forces/Test_partfrac_force.cc
+++ b/tests/forces/Test_partfrac_force.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
  
 
@@ -37,9 +37,9 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   const int Ls=9;
 
@@ -101,18 +101,19 @@ int main (int argc, char ** argv)
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin();i<mom.end();i++){
-      Uprime[i](mu) =
-	  U[i](mu)
-	+ mom[i](mu)*U[i](mu)*dt 
-	+ mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt/2.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt/6.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt/24.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt/120.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
+    auto U_v = U.View();
+    auto mom_v = mom.View();
+    auto Uprime_v = Uprime.View();
+    thread_foreach(i,mom_v,{
+      Uprime_v[i](mu) =  U_v[i](mu)
+	+ mom_v[i](mu)*U_v[i](mu)*dt 
+	+ mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt/2.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt/6.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt/24.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt/120.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
 	;
-    }
-
+    });
   }
   
   Dpf.ImportGauge(Uprime);
@@ -125,7 +126,7 @@ int main (int argc, char ** argv)
   //////////////////////////////////////////////
 
 
-  LatticeComplex dS(UGrid); dS = zero;
+  LatticeComplex dS(UGrid); dS = Zero();
   for(int mu=0;mu<Nd;mu++){
     mommu   = PeekIndex<LorentzIndex>(UdSdU,mu);
     mommu=Ta(mommu)*2.0;
diff --git a/tests/forces/Test_rect_force.cc b/tests/forces/Test_rect_force.cc
index 259090eb..9a78de24 100644
--- a/tests/forces/Test_rect_force.cc
+++ b/tests/forces/Test_rect_force.cc
@@ -29,7 +29,7 @@ Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
  
 
@@ -37,9 +37,9 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
@@ -87,9 +87,12 @@ int main (int argc, char ** argv)
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin();i<mom.end();i++){ // exp(pmu dt) * Umu
-      Uprime[i](mu) = U[i](mu) + mom[i](mu)*U[i](mu)*dt ;
-    }
+    auto Uprime_v = Uprime.View();
+    auto U_v      = U.View();
+    auto mom_v    = mom.View();
+    thread_foreach(i,mom_v,{ // exp(pmu dt) * Umu
+      Uprime_v[i](mu) = U_v[i](mu) + mom_v[i](mu)*U_v[i](mu)*dt ;
+    });
   }
 
   ComplexD Sprime    = Action.S(Uprime);
@@ -98,7 +101,7 @@ int main (int argc, char ** argv)
   // Use derivative to estimate dS
   //////////////////////////////////////////////
 
-  LatticeComplex dS(&Grid); dS = zero;
+  LatticeComplex dS(&Grid); dS = Zero();
 
   for(int mu=0;mu<Nd;mu++){
 
diff --git a/tests/forces/Test_wilson_force.cc b/tests/forces/Test_wilson_force.cc
index 338bed8c..47f1516a 100644
--- a/tests/forces/Test_wilson_force.cc
+++ b/tests/forces/Test_wilson_force.cc
@@ -29,17 +29,14 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
-
- 
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
@@ -93,7 +90,6 @@ int main (int argc, char ** argv)
   RealD dt = 0.0001;
   RealD Hmom = 0.0;
   RealD Hmomprime = 0.0;
-  RealD Hmompp    = 0.0;
   LatticeColourMatrix mommu(&Grid); 
   LatticeColourMatrix forcemu(&Grid); 
   LatticeGaugeField mom(&Grid); 
@@ -109,16 +105,18 @@ int main (int argc, char ** argv)
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin();i<mom.end();i++) {
-      Uprime[i](mu)  =	  U[i](mu);
-      Uprime[i](mu) += mom[i](mu)*U[i](mu)*dt ;
-      Uprime[i](mu) += mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt/2.0);
-      Uprime[i](mu) += mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt/6.0);
-      Uprime[i](mu) += mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt/24.0);
-      Uprime[i](mu) += mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt/120.0);
-      Uprime[i](mu) += mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt*dt/720.0);
-    }
-
+    auto U_v = U.View();
+    auto mom_v = mom.View();
+    auto Uprime_v = Uprime.View();
+    thread_foreach( i,mom_v,{
+      Uprime_v[i](mu)  =	  U_v[i](mu);
+      Uprime_v[i](mu) += mom_v[i](mu)*U_v[i](mu)*dt ;
+      Uprime_v[i](mu) += mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt/2.0);
+      Uprime_v[i](mu) += mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt/6.0);
+      Uprime_v[i](mu) += mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt/24.0);
+      Uprime_v[i](mu) += mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt/120.0);
+      Uprime_v[i](mu) += mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt*dt/720.0);
+    });
   }
 
   std::cout << GridLogMessage <<"Initial mom hamiltonian is "<< Hmom <<std::endl;
@@ -131,7 +129,6 @@ int main (int argc, char ** argv)
   // Use derivative to estimate dS
   //////////////////////////////////////////////
 
-
   for(int mu=0;mu<Nd;mu++){
     std::cout << "" <<std::endl;
     mommu   = PeekIndex<LorentzIndex>(mom,mu);
@@ -144,9 +141,9 @@ int main (int argc, char ** argv)
     std::cout << GridLogMessage<< " dsdumu + dag  " << norm2(mommu)<<std::endl;
   }
 
-  LatticeComplex dS(&Grid); dS = zero;
-  LatticeComplex dSmom(&Grid); dSmom = zero;
-  LatticeComplex dSmom2(&Grid); dSmom2 = zero;
+  LatticeComplex dS(&Grid); dS = Zero();
+  LatticeComplex dSmom(&Grid); dSmom = Zero();
+  LatticeComplex dSmom2(&Grid); dSmom2 = Zero();
   
   for(int mu=0;mu<Nd;mu++){
     mommu   = PeekIndex<LorentzIndex>(UdSdU,mu);
diff --git a/tests/forces/Test_wilsonclover_force.cc b/tests/forces/Test_wilsonclover_force.cc
index f7090845..d9ace23c 100644
--- a/tests/forces/Test_wilsonclover_force.cc
+++ b/tests/forces/Test_wilsonclover_force.cc
@@ -29,15 +29,14 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 int main(int argc, char **argv)
 {
   Grid_init(&argc, &argv);
 
-  std::vector<int> latt_size = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd, vComplex::Nsimd());
-  std::vector<int> mpi_layout = GridDefaultMpi();
+  auto latt_size = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(Nd, vComplex::Nsimd());
+  auto mpi_layout = GridDefaultMpi();
 
   GridCartesian Grid(latt_size, simd_layout, mpi_layout);
   GridRedBlackCartesian RBGrid(&Grid);
@@ -63,9 +62,8 @@ int main(int argc, char **argv)
 
   LatticeGaugeField U(&Grid);
 
-  std::vector<int> site = {0, 0, 0, 0};
   SU3::HotConfiguration(pRNG, U);
-  //SU3::ColdConfiguration(pRNG, U);// Clover term zero
+  //SU3::ColdConfiguration(pRNG, U);// Clover term Zero()
 
   ////////////////////////////////////
   // Unmodified matrix element
@@ -94,7 +92,7 @@ int main(int argc, char **argv)
   RealD dt = 0.00005;
   RealD Hmom = 0.0;
   RealD Hmomprime = 0.0;
-  RealD Hmompp = 0.0;
+  //  RealD Hmompp = 0.0;
   LatticeColourMatrix mommu(&Grid);
   LatticeColourMatrix forcemu(&Grid);
   LatticeGaugeField mom(&Grid);
@@ -107,10 +105,13 @@ int main(int argc, char **argv)
     Hmom -= real(sum(trace(mommu * mommu)));
     PokeIndex<LorentzIndex>(mom, mommu, mu);
 
-    parallel_for(int ss = 0; ss < mom._grid->oSites(); ss++)
+    auto Uprime_v = Uprime.View();
+    auto U_v      = U.View();
+    auto mom_v    = mom.View();
+    thread_foreach(ss,mom_v,
     {
-      Uprime[ss]._internal[mu] = ProjectOnGroup(Exponentiate(mom[ss]._internal[mu], dt, 12) * U[ss]._internal[mu]);
-    }
+      Uprime_v[ss]._internal[mu] = ProjectOnGroup(Exponentiate(mom_v[ss]._internal[mu], dt, 12) * U_v[ss]._internal[mu]);
+    });
   }
 
   std::cout << GridLogMessage << "Initial mom hamiltonian is " << Hmom << std::endl;
@@ -125,11 +126,11 @@ int main(int argc, char **argv)
   //////////////////////////////////////////////
 
   LatticeComplex dS(&Grid);
-  dS = zero;
+  dS = Zero();
   LatticeComplex dSmom(&Grid);
-  dSmom = zero;
+  dSmom = Zero();
   LatticeComplex dSmom2(&Grid);
-  dSmom2 = zero;
+  dSmom2 = Zero();
 
   for (int mu = 0; mu < Nd; mu++)
   {
diff --git a/tests/forces/Test_zmobius_force.cc b/tests/forces/Test_zmobius_force.cc
index 329d77e2..2730885f 100644
--- a/tests/forces/Test_zmobius_force.cc
+++ b/tests/forces/Test_zmobius_force.cc
@@ -29,15 +29,14 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   const int Ls=8;
 
@@ -70,7 +69,7 @@ int main (int argc, char ** argv)
   RealD b=0.5;
   RealD c=0.5;
 
-  std::vector < std::complex<double>  > omegas;
+  std::vector < ComplexD  > omegas;
   omegas.push_back( std::complex<double>(1.45806438985048,-0) );
   omegas.push_back( std::complex<double>(1.18231318389348,-0) );
   omegas.push_back( std::complex<double>(0.830951166685955,-0) );
@@ -101,7 +100,7 @@ int main (int argc, char ** argv)
   ////////////////////////////////////
   // Modify the gauge field a little 
   ////////////////////////////////////
-  RealD dt = 0.0001;
+  RealD dt = 0.001;
 
   LatticeColourMatrix mommu(UGrid); 
   LatticeColourMatrix forcemu(UGrid); 
@@ -115,18 +114,20 @@ int main (int argc, char ** argv)
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 
     // fourth order exponential approx
-    parallel_for(auto i=mom.begin();i<mom.end();i++){
-      Uprime[i](mu) =
-	  U[i](mu)
-	+ mom[i](mu)*U[i](mu)*dt 
-	+ mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt/2.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt/6.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt/24.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt/120.0)
-	+ mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *mom[i](mu) *U[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
-	;
-    }
+    auto mom_v    = mom.View();
+    auto U_v      = U.View();
+    auto Uprime_v = Uprime.View();
 
+    thread_foreach(i,mom_v,{
+      Uprime_v[i](mu) =	  U_v[i](mu)
+	+ mom_v[i](mu)*U_v[i](mu)*dt 
+	+ mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt/2.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt/6.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt/24.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt/120.0)
+	+ mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *mom_v[i](mu) *U_v[i](mu)*(dt*dt*dt*dt*dt*dt/720.0)
+	;
+    });
   }
   
   Ddwf.ImportGauge(Uprime);
@@ -139,7 +140,7 @@ int main (int argc, char ** argv)
   //////////////////////////////////////////////
 
 
-  LatticeComplex dS(UGrid); dS = zero;
+  LatticeComplex dS(UGrid); dS = Zero();
   for(int mu=0;mu<Nd;mu++){
     mommu   = PeekIndex<LorentzIndex>(UdSdU,mu);
     mommu=Ta(mommu)*2.0;
diff --git a/tests/hmc/Test_hmc_EODWFRatio.cc b/tests/hmc/Test_hmc_EODWFRatio.cc
index 9a755dd5..93469ffe 100644
--- a/tests/hmc/Test_hmc_EODWFRatio.cc
+++ b/tests/hmc/Test_hmc_EODWFRatio.cc
@@ -33,7 +33,7 @@ directory
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_hmc_EODWFRatioLsVectorised.cc b/tests/hmc/Test_hmc_EODWFRatioLsVectorised.cc
deleted file mode 100644
index fb2621a8..00000000
--- a/tests/hmc/Test_hmc_EODWFRatioLsVectorised.cc
+++ /dev/null
@@ -1,153 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./tests/Test_hmc_WilsonFermionGauge.cc
-
-Copyright (C) 2015
-
-Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
-Author: neo <cossu@post.kek.jp>
-Author: Guido Cossu <guido.cossu@ed.ac.uk>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution
-directory
-*************************************************************************************/
-/*  END LEGAL */
-#include <Grid/Grid.h>
-
-
-
-int main(int argc, char **argv) {
-  using namespace Grid;
-  using namespace Grid::QCD;
-
-  Grid_init(&argc, &argv);
-  int threads = GridThread::GetThreads();
-  // here make a routine to print all the relevant information on the run
-  std::cout << GridLogMessage << "Grid is setup to use " << threads << " threads" << std::endl;
-
-   // Typedefs to simplify notation
-  typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
-
-  typedef DomainWallVec5dImplR FermionImplPolicy;
-  typedef DomainWallFermion<FermionImplPolicy> FermionAction;
-  typedef typename FermionAction::FermionField FermionField;
-
-
-  //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
-  HMCWrapper TheHMC;
-
-  // Grid from the command line
-  TheHMC.Resources.AddFourDimGrid("gauge");
-  // Possibile to create the module by hand 
-  // hardcoding parameters or using a Reader
-
-
-  // Checkpointer definition
-  CheckpointerParameters CPparams;  
-  CPparams.config_prefix = "ckpoint_EODWF_lat";
-  CPparams.rng_prefix = "ckpoint_EODWF_rng";
-  CPparams.saveInterval = 5;
-  CPparams.format = "IEEE64BIG";
-  
-  TheHMC.Resources.LoadNerscCheckpointer(CPparams);
-
-  RNGModuleParameters RNGpar;
-  RNGpar.serial_seeds = "1 2 3 4 5";
-  RNGpar.parallel_seeds = "6 7 8 9 10";
-  TheHMC.Resources.SetRNGSeeds(RNGpar);
-
-  // Construct observables
-  // here there is too much indirection 
-  typedef PlaquetteMod<HMCWrapper::ImplPolicy> PlaqObs;
-  TheHMC.Resources.AddObservable<PlaqObs>();
-  //////////////////////////////////////////////
-
-  /////////////////////////////////////////////////////////////
-  // Collect actions, here use more encapsulation
-  // need wrappers of the fermionic classes 
-  // that have a complex construction
-  // standard
-  RealD beta = 5.6 ;
-  WilsonGaugeActionR Waction(beta);
-    
-
-  const int Ls = 8;
-  auto GridPtr   = TheHMC.Resources.GetCartesian();
-  auto GridRBPtr = TheHMC.Resources.GetRBCartesian();
-  auto FGrid     = SpaceTimeGrid::makeFiveDimGrid(Ls,GridPtr);
-  auto FrbGrid   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,GridPtr);
-  auto sUGrid    = SpaceTimeGrid::makeFourDimDWFGrid(GridDefaultLatt(),GridDefaultMpi());
-  auto sUrbGrid  = SpaceTimeGrid::makeFourDimRedBlackGrid(sUGrid);
-
-  // temporarily need a gauge field
-  LatticeGaugeField U(GridPtr);
-
-  Real mass = 0.04;
-  Real pv   = 1.0;
-  RealD M5  = 1.5;
-
-  FermionAction DenOp(U,*FGrid,*FrbGrid,*sUGrid,*sUrbGrid,mass,M5);
-  FermionAction NumOp(U,*FGrid,*FrbGrid,*sUGrid,*sUrbGrid,pv,M5);
-
-
-  double StoppingCondition = 1.0e-8;
-  double MaxCGIterations = 10000;
-  ConjugateGradient<FermionField>  CG(StoppingCondition,MaxCGIterations);
-  TwoFlavourEvenOddRatioPseudoFermionAction<FermionImplPolicy> Nf2(NumOp, DenOp,CG,CG);
-
-    // Set smearing (true/false), default: false
-  Nf2.is_smeared = false;
-
-  // Collect actions
-  ActionLevel<HMCWrapper::Field> Level1(1);
-  Level1.push_back(&Nf2);
-
-  ActionLevel<HMCWrapper::Field> Level2(4);
-  Level2.push_back(&Waction);
-
-  TheHMC.TheAction.push_back(Level1);
-  TheHMC.TheAction.push_back(Level2);
-  /////////////////////////////////////////////////////////////
-  // HMC parameters are serialisable 
-  TheHMC.Parameters.MD.MDsteps = 20;
-  TheHMC.Parameters.MD.trajL   = 1.0;
-
-  TheHMC.ReadCommandLine(argc, argv); // these can be parameters from file
-
-  // Reset performance counters 
-  NumOp.ZeroCounters();
-  DenOp.ZeroCounters();
-  TheHMC.Run();  // no smearing
-
-  std::cout << GridLogMessage << "Numerator report, Pauli-Villars term         : " << std::endl;
-  NumOp.Report();
-  std::cout << GridLogMessage << "Denominator report, Dw(m) term (includes CG) : " << std::endl;
-  DenOp.Report();
-
-
-
-
-  Grid_finalize();
-
-
-
-
-} // main
-
-
diff --git a/tests/hmc/Test_hmc_EODWFRatio_Gparity.cc b/tests/hmc/Test_hmc_EODWFRatio_Gparity.cc
index 16d225be..3434fccc 100644
--- a/tests/hmc/Test_hmc_EODWFRatio_Gparity.cc
+++ b/tests/hmc/Test_hmc_EODWFRatio_Gparity.cc
@@ -31,7 +31,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_hmc_EOMobiusRatio.cc b/tests/hmc/Test_hmc_EOMobiusRatio.cc
index d6ca96db..63b4d4fa 100644
--- a/tests/hmc/Test_hmc_EOMobiusRatio.cc
+++ b/tests/hmc/Test_hmc_EOMobiusRatio.cc
@@ -73,7 +73,7 @@ namespace Grid{
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
@@ -86,8 +86,8 @@ int main(int argc, char **argv) {
   typedef MobiusFermionR FermionAction;
   typedef typename FermionAction::FermionField FermionField;
   // Serialiser
-  //typedef Grid::XmlReader       Serialiser;
-  typedef Grid::JSONReader       Serialiser;
+  typedef Grid::XmlReader       Serialiser;
+  //typedef Grid::JSONReader       Serialiser;
   
   //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
   HMCWrapper TheHMC;
diff --git a/tests/hmc/Test_hmc_EOMobiusRatioManyFlavour.cc b/tests/hmc/Test_hmc_EOMobiusRatioManyFlavour.cc
index d4eac654..790433f2 100644
--- a/tests/hmc/Test_hmc_EOMobiusRatioManyFlavour.cc
+++ b/tests/hmc/Test_hmc_EOMobiusRatioManyFlavour.cc
@@ -79,7 +79,7 @@ namespace Grid{
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
@@ -92,8 +92,8 @@ int main(int argc, char **argv) {
   typedef MobiusFermionR FermionAction;
   typedef typename FermionAction::FermionField FermionField;
   // Serialiser
-  //typedef Grid::XmlReader       Serialiser;
-  typedef Grid::JSONReader       Serialiser;
+  typedef Grid::XmlReader       Serialiser;
+  //typedef Grid::JSONReader       Serialiser;
   
   //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
   HMCWrapper TheHMC;
diff --git a/tests/hmc/Test_hmc_EOWilsonCloverFermionGauge.cc b/tests/hmc/Test_hmc_EOWilsonCloverFermionGauge.cc
index fbb8a4a3..6b9b70b5 100644
--- a/tests/hmc/Test_hmc_EOWilsonCloverFermionGauge.cc
+++ b/tests/hmc/Test_hmc_EOWilsonCloverFermionGauge.cc
@@ -30,7 +30,6 @@ directory
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_hmc_EOWilsonFermionGauge.cc b/tests/hmc/Test_hmc_EOWilsonFermionGauge.cc
index eccb10b7..3643c0ad 100644
--- a/tests/hmc/Test_hmc_EOWilsonFermionGauge.cc
+++ b/tests/hmc/Test_hmc_EOWilsonFermionGauge.cc
@@ -30,7 +30,7 @@ directory
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_hmc_EOWilsonRatio.cc b/tests/hmc/Test_hmc_EOWilsonRatio.cc
index 0344a534..675bc605 100644
--- a/tests/hmc/Test_hmc_EOWilsonRatio.cc
+++ b/tests/hmc/Test_hmc_EOWilsonRatio.cc
@@ -32,7 +32,7 @@
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_hmc_Factories.cc b/tests/hmc/Test_hmc_Factories.cc
index 07174c7b..0a6c2d2d 100644
--- a/tests/hmc/Test_hmc_Factories.cc
+++ b/tests/hmc/Test_hmc_Factories.cc
@@ -32,9 +32,9 @@ namespace Grid{
 
 // Put this section in a separate header
 // ifdefs ?? Local makefile suggestion , policy as make parameter
-typedef QCD::PeriodicGimplR   ImplementationPolicy;
-typedef QCD::WilsonImplR      FermionImplementationPolicy;
-typedef QCD::NoHirep          RepresentationPolicy;
+typedef PeriodicGimplR   ImplementationPolicy;
+typedef WilsonImplR      FermionImplementationPolicy;
+typedef NoHirep          RepresentationPolicy;
 typedef Grid::XmlReader       Serialiser;
 
 // Register all object names
@@ -44,7 +44,7 @@ typedef Grid::XmlReader       Serialiser;
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
 
diff --git a/tests/hmc/Test_hmc_GparityIwasakiGauge.cc b/tests/hmc/Test_hmc_GparityIwasakiGauge.cc
index 6fbb389f..bc47b6c2 100644
--- a/tests/hmc/Test_hmc_GparityIwasakiGauge.cc
+++ b/tests/hmc/Test_hmc_GparityIwasakiGauge.cc
@@ -30,7 +30,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_hmc_GparityWilsonGauge.cc b/tests/hmc/Test_hmc_GparityWilsonGauge.cc
index 4fbb63d0..eb057181 100644
--- a/tests/hmc/Test_hmc_GparityWilsonGauge.cc
+++ b/tests/hmc/Test_hmc_GparityWilsonGauge.cc
@@ -31,7 +31,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_hmc_IwasakiGauge.cc b/tests/hmc/Test_hmc_IwasakiGauge.cc
index c4d96d83..66345bc0 100644
--- a/tests/hmc/Test_hmc_IwasakiGauge.cc
+++ b/tests/hmc/Test_hmc_IwasakiGauge.cc
@@ -30,7 +30,7 @@ Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_hmc_RectGauge.cc b/tests/hmc/Test_hmc_RectGauge.cc
index c0601d2b..9352a35b 100644
--- a/tests/hmc/Test_hmc_RectGauge.cc
+++ b/tests/hmc/Test_hmc_RectGauge.cc
@@ -31,7 +31,7 @@ directory
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_hmc_ScalarAction.cc b/tests/hmc/Test_hmc_ScalarAction.cc
index 65d3b399..5cf8ab35 100644
--- a/tests/hmc/Test_hmc_ScalarAction.cc
+++ b/tests/hmc/Test_hmc_ScalarAction.cc
@@ -29,7 +29,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_hmc_ScalarActionNxN.cc b/tests/hmc/Test_hmc_ScalarActionNxN.cc
index 33e52d82..0c398306 100644
--- a/tests/hmc/Test_hmc_ScalarActionNxN.cc
+++ b/tests/hmc/Test_hmc_ScalarActionNxN.cc
@@ -45,7 +45,7 @@ class ScalarActionParameters : Serializable {
 }
 
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template <class Impl>
 class MagMeas : public HmcObservable<typename Impl::Field> {
@@ -88,7 +88,8 @@ public:
 };
 
 int main(int argc, char **argv) {
-  typedef Grid::JSONReader       Serialiser;
+  //typedef Grid::JSONReader       Serialiser;
+  typedef Grid::XmlReader       Serialiser;
   
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
@@ -126,7 +127,7 @@ int main(int argc, char **argv) {
   ScalarGrid.set_rb(new GridRedBlackCartesian(ScalarGrid.get_full()));
   TheHMC.Resources.AddGrid("scalar", ScalarGrid);
   std::cout << "Lattice size : " << GridDefaultLatt() << std::endl;
-  
+
   ScalarActionParameters SPar(Reader);
 
   // Checkpointer definition
diff --git a/tests/hmc/Test_hmc_WC2ASFG_Production.cc b/tests/hmc/Test_hmc_WC2ASFG_Production.cc
index d255ab5d..0bbf4ece 100644
--- a/tests/hmc/Test_hmc_WC2ASFG_Production.cc
+++ b/tests/hmc/Test_hmc_WC2ASFG_Production.cc
@@ -69,7 +69,6 @@ namespace Grid{
 int main(int argc, char **argv)
 {
   using namespace Grid;
-  using namespace Grid::QCD;
 
   typedef Representations< FundamentalRepresentation, TwoIndexAntiSymmetricRepresentation > TheRepresentations;  
 
@@ -83,7 +82,8 @@ int main(int argc, char **argv)
   typedef WilsonTwoIndexAntiSymmetricImplR FermionImplPolicy; // gauge field implemetation for the pseudofermions
   typedef WilsonCloverTwoIndexAntiSymmetricFermionR FermionAction; // type of lattice fermions (Wilson, DW, ...)
   typedef typename FermionAction::FermionField FermionField;
-  typedef Grid::JSONReader Serialiser;
+  //typedef Grid::JSONReader Serialiser;
+  typedef Grid::XmlReader Serialiser;
 
   //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
   HMCWrapper TheHMC;
diff --git a/tests/hmc/Test_hmc_WC2SFG_Production.cc b/tests/hmc/Test_hmc_WC2SFG_Production.cc
index 8d5fc458..64a3f1cb 100644
--- a/tests/hmc/Test_hmc_WC2SFG_Production.cc
+++ b/tests/hmc/Test_hmc_WC2SFG_Production.cc
@@ -69,7 +69,7 @@ namespace Grid{
 int main(int argc, char **argv)
 {
   using namespace Grid;
-  using namespace Grid::QCD;
+   
 
   typedef Representations< FundamentalRepresentation, TwoIndexSymmetricRepresentation > TheRepresentations;  
 
@@ -83,7 +83,8 @@ int main(int argc, char **argv)
   typedef WilsonTwoIndexSymmetricImplR FermionImplPolicy; // gauge field implemetation for the pseudofermions
   typedef WilsonCloverTwoIndexSymmetricFermionR FermionAction; // type of lattice fermions (Wilson, DW, ...)
   typedef typename FermionAction::FermionField FermionField;
-  typedef Grid::JSONReader Serialiser;
+  //typedef Grid::JSONReader Serialiser;
+  typedef Grid::XmlReader Serialiser;
 
   //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
   HMCWrapper TheHMC;
diff --git a/tests/hmc/Test_hmc_WCFG_Production.cc b/tests/hmc/Test_hmc_WCFG_Production.cc
index 895e4f81..cebe3791 100644
--- a/tests/hmc/Test_hmc_WCFG_Production.cc
+++ b/tests/hmc/Test_hmc_WCFG_Production.cc
@@ -69,7 +69,7 @@ namespace Grid{
 int main(int argc, char **argv)
 {
   using namespace Grid;
-  using namespace Grid::QCD;
+   
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
@@ -81,7 +81,7 @@ int main(int argc, char **argv)
   typedef WilsonImplR FermionImplPolicy;
   typedef WilsonCloverFermionR FermionAction;
   typedef typename FermionAction::FermionField FermionField;
-  typedef Grid::JSONReader Serialiser;
+  typedef Grid::XmlReader Serialiser;
 
   //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
   HMCWrapper TheHMC;
diff --git a/tests/hmc/Test_hmc_WCMixedRepFG_Production.cc b/tests/hmc/Test_hmc_WCMixedRepFG_Production.cc
index aa5cce85..211900be 100644
--- a/tests/hmc/Test_hmc_WCMixedRepFG_Production.cc
+++ b/tests/hmc/Test_hmc_WCMixedRepFG_Production.cc
@@ -70,7 +70,7 @@ namespace Grid{
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   
 
   // Here change the allowed (higher) representations
   typedef Representations< FundamentalRepresentation, TwoIndexAntiSymmetricRepresentation> TheRepresentations;
@@ -91,7 +91,7 @@ int main(int argc, char **argv) {
   typedef WilsonCloverTwoIndexAntiSymmetricFermionR ASymmFermionAction; 
   typedef typename ASymmFermionAction::FermionField ASymmFermionField;
 
-  typedef Grid::JSONReader Serialiser;
+  typedef Grid::XmlReader Serialiser;
   //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
   HMCWrapper TheHMC;
   
diff --git a/tests/hmc/Test_hmc_WCadjFG_Production.cc b/tests/hmc/Test_hmc_WCadjFG_Production.cc
index 48cea756..5cf4bac1 100644
--- a/tests/hmc/Test_hmc_WCadjFG_Production.cc
+++ b/tests/hmc/Test_hmc_WCadjFG_Production.cc
@@ -69,7 +69,7 @@ namespace Grid{
 int main(int argc, char **argv)
 {
   using namespace Grid;
-  using namespace Grid::QCD;
+   
 
   typedef Representations< FundamentalRepresentation, AdjointRepresentation > TheRepresentations;  
 
@@ -83,7 +83,7 @@ int main(int argc, char **argv)
   typedef WilsonAdjImplR FermionImplPolicy; // gauge field implemetation for the pseudofermions
   typedef WilsonCloverAdjFermionR FermionAction; // type of lattice fermions (Wilson, DW, ...)
   typedef typename FermionAction::FermionField FermionField;
-  typedef Grid::JSONReader Serialiser;
+  typedef Grid::XmlReader Serialiser;
 
   //::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
   HMCWrapper TheHMC;
diff --git a/tests/hmc/Test_hmc_WG_Production.cc b/tests/hmc/Test_hmc_WG_Production.cc
index 7f8d8124..b16c073b 100644
--- a/tests/hmc/Test_hmc_WG_Production.cc
+++ b/tests/hmc/Test_hmc_WG_Production.cc
@@ -47,7 +47,7 @@ namespace Grid{
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   
 
   Grid_init(&argc, &argv);
   GridLogLayout();
@@ -55,7 +55,7 @@ int main(int argc, char **argv) {
   // Typedefs to simplify notation
   typedef GenericHMCRunner<MinimumNorm2> HMCWrapper;  // Uses the default minimum norm
   HMCWrapper TheHMC;
-  typedef Grid::JSONReader       Serialiser;
+  typedef Grid::XmlReader       Serialiser;
 
   // Grid from the command line
   TheHMC.Resources.AddFourDimGrid("gauge");
diff --git a/tests/hmc/Test_hmc_WilsonAdjointFermionGauge.cc b/tests/hmc/Test_hmc_WilsonAdjointFermionGauge.cc
index af4fdae1..cc56cae3 100644
--- a/tests/hmc/Test_hmc_WilsonAdjointFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonAdjointFermionGauge.cc
@@ -33,7 +33,7 @@ directory
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   // Here change the allowed (higher) representations
   typedef Representations< FundamentalRepresentation, AdjointRepresentation > TheRepresentations;
diff --git a/tests/hmc/Test_hmc_WilsonCloverFermionGauge.cc b/tests/hmc/Test_hmc_WilsonCloverFermionGauge.cc
index 8d5701b8..9e27e3ec 100644
--- a/tests/hmc/Test_hmc_WilsonCloverFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonCloverFermionGauge.cc
@@ -31,7 +31,7 @@ directory
 int main(int argc, char **argv)
 {
   using namespace Grid;
-  using namespace Grid::QCD;
+   
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_hmc_WilsonFermionGauge.cc b/tests/hmc/Test_hmc_WilsonFermionGauge.cc
index 9b709a1a..cc1f2474 100644
--- a/tests/hmc/Test_hmc_WilsonFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonFermionGauge.cc
@@ -32,7 +32,7 @@ directory
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_hmc_WilsonGauge.cc b/tests/hmc/Test_hmc_WilsonGauge.cc
index 05bf81a2..4ef0e658 100644
--- a/tests/hmc/Test_hmc_WilsonGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonGauge.cc
@@ -30,9 +30,9 @@ directory
 /*  END LEGAL */
 #include <Grid/Grid.h>
 
-int main(int argc, char **argv) {
+int main(int argc, char **argv) 
+{
   using namespace Grid;
-  using namespace Grid::QCD;
 
   Grid_init(&argc, &argv);
   GridLogLayout();
diff --git a/tests/hmc/Test_hmc_WilsonMixedRepresentationsFermionGauge.cc b/tests/hmc/Test_hmc_WilsonMixedRepresentationsFermionGauge.cc
index b54345cf..6fa90f32 100644
--- a/tests/hmc/Test_hmc_WilsonMixedRepresentationsFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonMixedRepresentationsFermionGauge.cc
@@ -35,8 +35,9 @@ directory
 
 
 int main(int argc, char **argv) {
+#ifndef GRID_NVCC
   using namespace Grid;
-  using namespace Grid::QCD;
+
 
   // Here change the allowed (higher) representations
   typedef Representations< FundamentalRepresentation, AdjointRepresentation , TwoIndexSymmetricRepresentation> TheRepresentations;
@@ -133,7 +134,7 @@ int main(int argc, char **argv) {
   TheHMC.Run();  // no smearing
 
   Grid_finalize();
-
+#endif
 } // main
 
 
diff --git a/tests/hmc/Test_hmc_WilsonRatio.cc b/tests/hmc/Test_hmc_WilsonRatio.cc
index 4e6426ca..3e3cac7e 100644
--- a/tests/hmc/Test_hmc_WilsonRatio.cc
+++ b/tests/hmc/Test_hmc_WilsonRatio.cc
@@ -31,7 +31,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_hmc_WilsonTMFermionGauge.cc b/tests/hmc/Test_hmc_WilsonTMFermionGauge.cc
index fbb9bd6f..5928efbe 100644
--- a/tests/hmc/Test_hmc_WilsonTMFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonTMFermionGauge.cc
@@ -32,7 +32,7 @@ directory
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_hmc_WilsonTwoIndexSymmetricFermionGauge.cc b/tests/hmc/Test_hmc_WilsonTwoIndexSymmetricFermionGauge.cc
index 853d8a92..387842f7 100644
--- a/tests/hmc/Test_hmc_WilsonTwoIndexSymmetricFermionGauge.cc
+++ b/tests/hmc/Test_hmc_WilsonTwoIndexSymmetricFermionGauge.cc
@@ -31,7 +31,7 @@ directory
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   // Here change the allowed (higher) representations
   typedef Representations< FundamentalRepresentation, TwoIndexSymmetricRepresentation > TheRepresentations;
diff --git a/tests/hmc/Test_multishift_sqrt.cc b/tests/hmc/Test_multishift_sqrt.cc
index 8f12b3bd..834e6fc8 100644
--- a/tests/hmc/Test_multishift_sqrt.cc
+++ b/tests/hmc/Test_multishift_sqrt.cc
@@ -31,7 +31,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class Field> class DumbOperator  : public LinearOperatorBase<Field> {
 public:
diff --git a/tests/hmc/Test_remez.cc b/tests/hmc/Test_remez.cc
index 5f4b0a25..ff20e7f4 100644
--- a/tests/hmc/Test_remez.cc
+++ b/tests/hmc/Test_remez.cc
@@ -30,7 +30,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
diff --git a/tests/hmc/Test_rhmc_EOWilson1p1.cc b/tests/hmc/Test_rhmc_EOWilson1p1.cc
index 37713766..51a966b1 100644
--- a/tests/hmc/Test_rhmc_EOWilson1p1.cc
+++ b/tests/hmc/Test_rhmc_EOWilson1p1.cc
@@ -31,7 +31,7 @@ See the full license in the file "LICENSE" in the top level distribution directo
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_rhmc_EOWilsonRatio.cc b/tests/hmc/Test_rhmc_EOWilsonRatio.cc
index d7d6f35e..44fb6d47 100644
--- a/tests/hmc/Test_rhmc_EOWilsonRatio.cc
+++ b/tests/hmc/Test_rhmc_EOWilsonRatio.cc
@@ -32,7 +32,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_rhmc_Wilson1p1.cc b/tests/hmc/Test_rhmc_Wilson1p1.cc
index bab86ab9..93b748d2 100644
--- a/tests/hmc/Test_rhmc_Wilson1p1.cc
+++ b/tests/hmc/Test_rhmc_Wilson1p1.cc
@@ -32,7 +32,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/hmc/Test_rhmc_WilsonRatio.cc b/tests/hmc/Test_rhmc_WilsonRatio.cc
index 40bea28e..4896d329 100644
--- a/tests/hmc/Test_rhmc_WilsonRatio.cc
+++ b/tests/hmc/Test_rhmc_WilsonRatio.cc
@@ -31,7 +31,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   int threads = GridThread::GetThreads();
diff --git a/tests/lanczos/BlockProjector.h b/tests/lanczos/BlockProjector.h
deleted file mode 100644
index 6becaa66..00000000
--- a/tests/lanczos/BlockProjector.h
+++ /dev/null
@@ -1,143 +0,0 @@
-namespace Grid { 
-
-/*
-  BlockProjector
-
-  If _HP_BLOCK_PROJECTORS_ is defined, we assume that _evec is a basis that is not
-  fully orthonormalized (to the precision of the coarse field) and we allow for higher-precision
-  coarse field than basis field.
-
-*/
-//#define _HP_BLOCK_PROJECTORS_
-
-template<typename Field>
-class BlockProjector {
-public:
-
-  BasisFieldVector<Field>& _evec;
-  BlockedGrid<Field>& _bgrid;
-
-  BlockProjector(BasisFieldVector<Field>& evec, BlockedGrid<Field>& bgrid) : _evec(evec), _bgrid(bgrid) {
-  }
-
-  void createOrthonormalBasis(RealD thres = 0.0) {
-
-    GridStopWatch sw;
-    sw.Start();
-
-    int cnt = 0;
-
-#pragma omp parallel shared(cnt)
-    {
-      int lcnt = 0;
-
-#pragma omp for
-      for (int b=0;b<_bgrid._o_blocks;b++) {
-	
-	for (int i=0;i<_evec._Nm;i++) {
-	  
-	  auto nrm0 = _bgrid.block_sp(b,_evec._v[i],_evec._v[i]);
-	  
-	  // |i> -= <j|i> |j>
-	  for (int j=0;j<i;j++) {
-	    _bgrid.block_caxpy(b,_evec._v[i],-_bgrid.block_sp(b,_evec._v[j],_evec._v[i]),_evec._v[j],_evec._v[i]);
-	  }
-	  
-	  auto nrm = _bgrid.block_sp(b,_evec._v[i],_evec._v[i]);
-	  
-	  auto eps = nrm/nrm0;
-	  if (Reduce(eps).real() < thres) {
-	    lcnt++;
-	  }
-	  
-	  // TODO: if norm is too small, remove this eigenvector/mark as not needed; in practice: set it to zero norm here and return a mask
-	  // that is then used later to decide not to write certain eigenvectors to disk (add a norm calculation before subtraction step and look at nrm/nrm0 < eps to decide)
-	  _bgrid.block_cscale(b,1.0 / sqrt(nrm),_evec._v[i]);
-	  
-	}
-	
-      }
-
-#pragma omp critical
-      {
-	cnt += lcnt;
-      }
-    }
-    sw.Stop();
-    std::cout << GridLogMessage << "Gram-Schmidt to create blocked basis took " << sw.Elapsed() << " (" << ((RealD)cnt / (RealD)_bgrid._o_blocks / (RealD)_evec._Nm) 
-	      << " below threshold)" << std::endl;
-
-  }
-
-  template<typename CoarseField>
-  void coarseToFine(const CoarseField& in, Field& out) {
-
-    out = zero;
-    out.checkerboard = _evec._v[0].checkerboard;
-
-    int Nbasis = sizeof(in._odata[0]._internal._internal) / sizeof(in._odata[0]._internal._internal[0]);
-    assert(Nbasis == _evec._Nm);
-    
-#pragma omp parallel for
-    for (int b=0;b<_bgrid._o_blocks;b++) {
-      for (int j=0;j<_evec._Nm;j++) {
-	_bgrid.block_caxpy(b,out,in._odata[b]._internal._internal[j],_evec._v[j],out);
-      }
-    }
-
-  }
-
-  template<typename CoarseField>
-  void fineToCoarse(const Field& in, CoarseField& out) {
-
-    out = zero;
-
-    int Nbasis = sizeof(out._odata[0]._internal._internal) / sizeof(out._odata[0]._internal._internal[0]);
-    assert(Nbasis == _evec._Nm);
-
-
-    Field tmp(_bgrid._grid);
-    tmp = in;
-    
-#pragma omp parallel for
-    for (int b=0;b<_bgrid._o_blocks;b++) {
-      for (int j=0;j<_evec._Nm;j++) {
-	// |rhs> -= <j|rhs> |j>
-	auto c = _bgrid.block_sp(b,_evec._v[j],tmp);
-	_bgrid.block_caxpy(b,tmp,-c,_evec._v[j],tmp); // may make this more numerically stable
-	out._odata[b]._internal._internal[j] = c;
-      }
-    }
-
-  }
-
-  template<typename CoarseField>
-    void deflateFine(BasisFieldVector<CoarseField>& _coef,const std::vector<RealD>& eval,int N,const Field& src_orig,Field& result) {
-    result = zero;
-    for (int i=0;i<N;i++) {
-      Field tmp(result._grid);
-      coarseToFine(_coef._v[i],tmp);
-      axpy(result,TensorRemove(innerProduct(tmp,src_orig)) / eval[i],tmp,result);
-    }
-  }
-
-  template<typename CoarseField>
-    void deflateCoarse(BasisFieldVector<CoarseField>& _coef,const std::vector<RealD>& eval,int N,const Field& src_orig,Field& result) {
-    CoarseField src_coarse(_coef._v[0]._grid);
-    CoarseField result_coarse = src_coarse;
-    result_coarse = zero;
-    fineToCoarse(src_orig,src_coarse);
-    for (int i=0;i<N;i++) {
-      axpy(result_coarse,TensorRemove(innerProduct(_coef._v[i],src_coarse)) / eval[i],_coef._v[i],result_coarse);
-    }
-    coarseToFine(result_coarse,result);
-  }
-
-  template<typename CoarseField>
-    void deflate(BasisFieldVector<CoarseField>& _coef,const std::vector<RealD>& eval,int N,const Field& src_orig,Field& result) {
-    // Deflation on coarse Grid is much faster, so use it by default.  Deflation on fine Grid is kept for legacy reasons for now.
-    deflateCoarse(_coef,eval,N,src_orig,result);
-  }
-
-};
-}
diff --git a/tests/lanczos/BlockedGrid.h b/tests/lanczos/BlockedGrid.h
deleted file mode 100644
index 821272de..00000000
--- a/tests/lanczos/BlockedGrid.h
+++ /dev/null
@@ -1,401 +0,0 @@
-namespace Grid {
-
-template<typename Field>
-class BlockedGrid {
-public:
-  GridBase* _grid;
-  typedef typename Field::scalar_type  Coeff_t;
-  typedef typename Field::vector_type vCoeff_t;
-  
-  std::vector<int> _bs; // block size
-  std::vector<int> _nb; // number of blocks
-  std::vector<int> _l;  // local dimensions irrespective of cb
-  std::vector<int> _l_cb;  // local dimensions of checkerboarded vector
-  std::vector<int> _l_cb_o;  // local dimensions of inner checkerboarded vector
-  std::vector<int> _bs_cb; // block size in checkerboarded vector
-  std::vector<int> _nb_o; // number of blocks of simd o-sites
-
-  int _nd, _blocks, _cf_size, _cf_block_size, _cf_o_block_size, _o_blocks, _block_sites;
-  
-  BlockedGrid(GridBase* grid, const std::vector<int>& block_size) :
-    _grid(grid), _bs(block_size), _nd((int)_bs.size()), 
-      _nb(block_size), _l(block_size), _l_cb(block_size), _nb_o(block_size),
-      _l_cb_o(block_size), _bs_cb(block_size) {
-
-    _blocks = 1;
-    _o_blocks = 1;
-    _l = grid->FullDimensions();
-    _l_cb = grid->LocalDimensions();
-    _l_cb_o = grid->_rdimensions;
-
-    _cf_size = 1;
-    _block_sites = 1;
-    for (int i=0;i<_nd;i++) {
-      _l[i] /= grid->_processors[i];
-
-      assert(!(_l[i] % _bs[i])); // lattice must accommodate choice of blocksize
-
-      int r = _l[i] / _l_cb[i];
-      assert(!(_bs[i] % r)); // checkerboarding must accommodate choice of blocksize
-      _bs_cb[i] = _bs[i] / r;
-      _block_sites *= _bs_cb[i];
-      _nb[i] = _l[i] / _bs[i];
-      _nb_o[i] = _nb[i] / _grid->_simd_layout[i];
-      if (_nb[i] % _grid->_simd_layout[i]) { // simd must accommodate choice of blocksize
-	std::cout << GridLogMessage << "Problem: _nb[" << i << "] = " << _nb[i] << " _grid->_simd_layout[" << i << "] = " << _grid->_simd_layout[i] << std::endl;
-	assert(0);
-      }
-      _blocks *= _nb[i];
-      _o_blocks *= _nb_o[i];
-      _cf_size *= _l[i];
-    }
-
-    _cf_size *= 12 / 2;
-    _cf_block_size = _cf_size / _blocks;
-    _cf_o_block_size = _cf_size / _o_blocks;
-
-    std::cout << GridLogMessage << "BlockedGrid:" << std::endl;
-    std::cout << GridLogMessage << " _l     = " << _l << std::endl;
-    std::cout << GridLogMessage << " _l_cb     = " << _l_cb << std::endl;
-    std::cout << GridLogMessage << " _l_cb_o     = " << _l_cb_o << std::endl;
-    std::cout << GridLogMessage << " _bs    = " << _bs << std::endl;
-    std::cout << GridLogMessage << " _bs_cb    = " << _bs_cb << std::endl;
-
-    std::cout << GridLogMessage << " _nb    = " << _nb << std::endl;
-    std::cout << GridLogMessage << " _nb_o    = " << _nb_o << std::endl;
-    std::cout << GridLogMessage << " _blocks = " << _blocks << std::endl;
-    std::cout << GridLogMessage << " _o_blocks = " << _o_blocks << std::endl;
-    std::cout << GridLogMessage << " sizeof(vCoeff_t) = " << sizeof(vCoeff_t) << std::endl;
-    std::cout << GridLogMessage << " _cf_size = " << _cf_size << std::endl;
-    std::cout << GridLogMessage << " _cf_block_size = " << _cf_block_size << std::endl;
-    std::cout << GridLogMessage << " _block_sites = " << _block_sites << std::endl;
-    std::cout << GridLogMessage << " _grid->oSites() = " << _grid->oSites() << std::endl;
-
-    //    _grid->Barrier();
-    //abort();
-  }
-
-    void block_to_coor(int b, std::vector<int>& x0) {
-
-      std::vector<int> bcoor;
-      bcoor.resize(_nd);
-      x0.resize(_nd);
-      assert(b < _o_blocks);
-      Lexicographic::CoorFromIndex(bcoor,b,_nb_o);
-      int i;
-
-      for (i=0;i<_nd;i++) {
-	x0[i] = bcoor[i]*_bs_cb[i];
-      }
-
-      //std::cout << GridLogMessage << "Map block b -> " << x0 << std::endl;
-
-    }
-
-    void block_site_to_o_coor(const std::vector<int>& x0, std::vector<int>& coor, int i) {
-      Lexicographic::CoorFromIndex(coor,i,_bs_cb);
-      for (int j=0;j<_nd;j++)
-	coor[j] += x0[j];
-    }
-
-    int block_site_to_o_site(const std::vector<int>& x0, int i) {
-      std::vector<int> coor;  coor.resize(_nd);
-      block_site_to_o_coor(x0,coor,i);
-      Lexicographic::IndexFromCoor(coor,i,_l_cb_o);
-      return i;
-    }
-
-    vCoeff_t block_sp(int b, const Field& x, const Field& y) {
-
-      std::vector<int> x0;
-      block_to_coor(b,x0);
-
-      vCoeff_t ret = 0.0;
-      for (int i=0;i<_block_sites;i++) { // only odd sites
-	int ss = block_site_to_o_site(x0,i);
-	ret += TensorRemove(innerProduct(x._odata[ss],y._odata[ss]));
-      }
-
-      return ret;
-
-    }
-
-    vCoeff_t block_sp(int b, const Field& x, const std::vector< ComplexD >& y) {
-
-      std::vector<int> x0;
-      block_to_coor(b,x0);
-
-      constexpr int nsimd = sizeof(vCoeff_t) / sizeof(Coeff_t);
-      int lsize = _cf_o_block_size / _block_sites;
-
-      std::vector< ComplexD > ret(nsimd);
-      for (int i=0;i<nsimd;i++)
-	ret[i] = 0.0;
-
-      for (int i=0;i<_block_sites;i++) { // only odd sites
-	int ss = block_site_to_o_site(x0,i);
-
-	int n = lsize / nsimd;
-	for (int l=0;l<n;l++) {
-	  for (int j=0;j<nsimd;j++) {
-	    int t = lsize * i + l*nsimd + j;
-
-	    ret[j] += conjugate(((Coeff_t*)&x._odata[ss]._internal)[l*nsimd + j]) * y[t];
-	  }
-	}
-      }
-
-      vCoeff_t vret;
-      for (int i=0;i<nsimd;i++)
-	((Coeff_t*)&vret)[i] = (Coeff_t)ret[i];
-
-      return vret;
-
-    }
-
-    template<class T>
-      void vcaxpy(iScalar<T>& r,const vCoeff_t& a,const iScalar<T>& x,const iScalar<T>& y) {
-      vcaxpy(r._internal,a,x._internal,y._internal);
-    }
-
-    template<class T,int N>
-      void vcaxpy(iVector<T,N>& r,const vCoeff_t& a,const iVector<T,N>& x,const iVector<T,N>& y) {
-      for (int i=0;i<N;i++)
-	vcaxpy(r._internal[i],a,x._internal[i],y._internal[i]);
-    }
-
-    void vcaxpy(vCoeff_t& r,const vCoeff_t& a,const vCoeff_t& x,const vCoeff_t& y) {
-      r = a*x + y;
-    }
-
-    void block_caxpy(int b, Field& ret, const vCoeff_t& a, const Field& x, const Field& y) {
-
-      std::vector<int> x0;
-      block_to_coor(b,x0);
-
-      for (int i=0;i<_block_sites;i++) { // only odd sites
-	int ss = block_site_to_o_site(x0,i);
-	vcaxpy(ret._odata[ss],a,x._odata[ss],y._odata[ss]);
-      }
-
-    }
-
-    void block_caxpy(int b, std::vector< ComplexD >& ret, const vCoeff_t& a, const Field& x, const std::vector< ComplexD >& y) {
-      std::vector<int> x0;
-      block_to_coor(b,x0);
-
-      constexpr int nsimd = sizeof(vCoeff_t) / sizeof(Coeff_t);
-      int lsize = _cf_o_block_size / _block_sites;
-
-      for (int i=0;i<_block_sites;i++) { // only odd sites
-	int ss = block_site_to_o_site(x0,i);
-
-	int n = lsize / nsimd;
-	for (int l=0;l<n;l++) {
-	  vCoeff_t r = a* ((vCoeff_t*)&x._odata[ss]._internal)[l];
-
-	  for (int j=0;j<nsimd;j++) {
-	    int t = lsize * i + l*nsimd + j;
-	    ret[t] = y[t] + ((Coeff_t*)&r)[j];
-	  }
-	}
-      }
-
-    }
-
-    void block_set(int b, Field& ret, const std::vector< ComplexD >& x) {
-      std::vector<int> x0;
-      block_to_coor(b,x0);
-
-      int lsize = _cf_o_block_size / _block_sites;
-
-      for (int i=0;i<_block_sites;i++) { // only odd sites
-	int ss = block_site_to_o_site(x0,i);
-
-	for (int l=0;l<lsize;l++)
-	  ((Coeff_t*)&ret._odata[ss]._internal)[l] = (Coeff_t)x[lsize * i + l]; // convert precision
-      }
-
-    }
-
-    void block_get(int b, const Field& ret, std::vector< ComplexD >& x) {
-      std::vector<int> x0;
-      block_to_coor(b,x0);
-
-      int lsize = _cf_o_block_size / _block_sites;
-
-      for (int i=0;i<_block_sites;i++) { // only odd sites
-	int ss = block_site_to_o_site(x0,i);
-
-	for (int l=0;l<lsize;l++)
-	  x[lsize * i + l] = (ComplexD)((Coeff_t*)&ret._odata[ss]._internal)[l];
-      }
-
-    }
-
-    template<class T>
-    void vcscale(iScalar<T>& r,const vCoeff_t& a,const iScalar<T>& x) {
-      vcscale(r._internal,a,x._internal);
-    }
-
-    template<class T,int N>
-    void vcscale(iVector<T,N>& r,const vCoeff_t& a,const iVector<T,N>& x) {
-      for (int i=0;i<N;i++)
-	vcscale(r._internal[i],a,x._internal[i]);
-    }
-
-    void vcscale(vCoeff_t& r,const vCoeff_t& a,const vCoeff_t& x) {
-      r = a*x;
-    }
-
-    void block_cscale(int b, const vCoeff_t& a, Field& ret) {
-
-      std::vector<int> x0;
-      block_to_coor(b,x0);
-      
-      for (int i=0;i<_block_sites;i++) { // only odd sites
-	int ss = block_site_to_o_site(x0,i);
-	vcscale(ret._odata[ss],a,ret._odata[ss]);
-      }
-    }
-
-    void getCanonicalBlockOffset(int cb, std::vector<int>& x0) {
-      const int ndim = 5;
-      assert(_nb.size() == ndim);
-      std::vector<int> _nbc = { _nb[1], _nb[2], _nb[3], _nb[4], _nb[0] };
-      std::vector<int> _bsc = { _bs[1], _bs[2], _bs[3], _bs[4], _bs[0] };
-      x0.resize(ndim);
-
-      assert(cb >= 0);
-      assert(cb < _nbc[0]*_nbc[1]*_nbc[2]*_nbc[3]*_nbc[4]);
-
-      Lexicographic::CoorFromIndex(x0,cb,_nbc);
-      int i;
-
-      for (i=0;i<ndim;i++) {
-	x0[i] *= _bsc[i];
-      }
-
-      //if (cb < 2)
-      //	std::cout << GridLogMessage << "Map: " << cb << " To: " << x0 << std::endl;
-    }
-
-    void pokeBlockOfVectorCanonical(int cb,Field& v,const std::vector<float>& buf) {
-      std::vector<int> _bsc = { _bs[1], _bs[2], _bs[3], _bs[4], _bs[0] };
-      std::vector<int> ldim = v._grid->LocalDimensions();
-      std::vector<int> cldim = { ldim[1], ldim[2], ldim[3], ldim[4], ldim[0] };
-      const int _nbsc = _bs_cb[0]*_bs_cb[1]*_bs_cb[2]*_bs_cb[3]*_bs_cb[4];
-      // take canonical block cb of v and put it in canonical ordering in buf
-      std::vector<int> cx0;
-      getCanonicalBlockOffset(cb,cx0);
-
-#pragma omp parallel
-      {
-	std::vector<int> co0,cl0;
-	co0=cx0; cl0=cx0;
-
-#pragma omp for
-	for (int i=0;i<_nbsc;i++) {
-	  Lexicographic::CoorFromIndex(co0,2*i,_bsc); // 2* for eo
-	  for (int j=0;j<(int)_bsc.size();j++)
-	    cl0[j] = cx0[j] + co0[j];
-	  
-	  std::vector<int> l0 = { cl0[4], cl0[0], cl0[1], cl0[2], cl0[3] };
-	  int oi = v._grid->oIndex(l0);
-	  int ii = v._grid->iIndex(l0);
-	  int lti = i;
-
-	  //if (cb < 2 && i<2)
-	  //  std::cout << GridLogMessage << "Map: " << cb << ", " << i << " To: " << cl0 << ", " << cx0 << ", " << oi << ", " << ii << std::endl;
-	  
-	  for (int s=0;s<4;s++)
-	    for (int c=0;c<3;c++) {
-	      Coeff_t& ld = ((Coeff_t*)&v._odata[oi]._internal._internal[s]._internal[c])[ii];
-	      int ti = 12*lti + 3*s + c;
-	      ld = Coeff_t(buf[2*ti+0], buf[2*ti+1]);
-	    }
-	}
-      }
-    }
-
-    void peekBlockOfVectorCanonical(int cb,const Field& v,std::vector<float>& buf) {
-      std::vector<int> _bsc = { _bs[1], _bs[2], _bs[3], _bs[4], _bs[0] };
-      std::vector<int> ldim = v._grid->LocalDimensions();
-      std::vector<int> cldim = { ldim[1], ldim[2], ldim[3], ldim[4], ldim[0] };
-      const int _nbsc = _bs_cb[0]*_bs_cb[1]*_bs_cb[2]*_bs_cb[3]*_bs_cb[4];
-      // take canonical block cb of v and put it in canonical ordering in buf
-      std::vector<int> cx0;
-      getCanonicalBlockOffset(cb,cx0);
-
-      buf.resize(_cf_block_size * 2);
-
-#pragma omp parallel
-      {
-	std::vector<int> co0,cl0;
-	co0=cx0; cl0=cx0;
-
-#pragma omp for
-	for (int i=0;i<_nbsc;i++) {
-	  Lexicographic::CoorFromIndex(co0,2*i,_bsc); // 2* for eo
-	  for (int j=0;j<(int)_bsc.size();j++)
-	    cl0[j] = cx0[j] + co0[j];
-	  
-	  std::vector<int> l0 = { cl0[4], cl0[0], cl0[1], cl0[2], cl0[3] };
-	  int oi = v._grid->oIndex(l0);
-	  int ii = v._grid->iIndex(l0);
-	  int lti = i;
-	  
-	  //if (cb < 2 && i<2)
-	  //  std::cout << GridLogMessage << "Map: " << cb << ", " << i << " To: " << cl0 << ", " << cx0 << ", " << oi << ", " << ii << std::endl;
-
-	  for (int s=0;s<4;s++)
-	    for (int c=0;c<3;c++) {
-	      Coeff_t& ld = ((Coeff_t*)&v._odata[oi]._internal._internal[s]._internal[c])[ii];
-	      int ti = 12*lti + 3*s + c;
-	      buf[2*ti+0] = ld.real();
-	      buf[2*ti+1] = ld.imag();
-	    }
-	}
-      }
-    }
-
-    int globalToLocalCanonicalBlock(int slot,const std::vector<int>& src_nodes,int nb) {
-      // processor coordinate
-      int _nd = (int)src_nodes.size();
-      std::vector<int> _src_nodes = src_nodes;
-      std::vector<int> pco(_nd);
-      Lexicographic::CoorFromIndex(pco,slot,_src_nodes);
-      std::vector<int> cpco = { pco[1], pco[2], pco[3], pco[4], pco[0] };
-
-      // get local block
-      std::vector<int> _nbc = { _nb[1], _nb[2], _nb[3], _nb[4], _nb[0] };
-      assert(_nd == 5);
-      std::vector<int> c_src_local_blocks(_nd);
-      for (int i=0;i<_nd;i++) {
-	assert(_grid->_fdimensions[i] % (src_nodes[i] * _bs[i]) == 0);
-	c_src_local_blocks[(i+4) % 5] = _grid->_fdimensions[i] / src_nodes[i] / _bs[i];
-      }
-      std::vector<int> cbcoor(_nd); // coordinate of block in slot in canonical form
-      Lexicographic::CoorFromIndex(cbcoor,nb,c_src_local_blocks);
-
-      // cpco, cbcoor
-      std::vector<int> clbcoor(_nd);
-      for (int i=0;i<_nd;i++) {
-	int cgcoor = cpco[i] * c_src_local_blocks[i] + cbcoor[i]; // global block coordinate
-	int pcoor = cgcoor / _nbc[i]; // processor coordinate in my Grid
-	int tpcoor = _grid->_processor_coor[(i+1)%5];
-	if (pcoor != tpcoor)
-	  return -1;
-	clbcoor[i] = cgcoor - tpcoor * _nbc[i]; // canonical local block coordinate for canonical dimension i
-      }
-
-      int lnb;
-      Lexicographic::IndexFromCoor(clbcoor,lnb,_nbc);
-      //std::cout << "Mapped slot = " << slot << " nb = " << nb << " to " << lnb << std::endl;
-      return lnb;
-    }
-
-
- };
-
-}
diff --git a/tests/lanczos/FieldBasisVector.h b/tests/lanczos/FieldBasisVector.h
deleted file mode 100644
index 9a21aa46..00000000
--- a/tests/lanczos/FieldBasisVector.h
+++ /dev/null
@@ -1,81 +0,0 @@
-namespace Grid { 
-
-template<class Field>
-class BasisFieldVector {
- public:
-  int _Nm;
-
-  typedef typename Field::scalar_type Coeff_t;
-  typedef typename Field::vector_type vCoeff_t;
-  typedef typename Field::vector_object vobj;
-  typedef typename vobj::scalar_object sobj;
-
-  std::vector<Field> _v; // _Nfull vectors
-
-  void report(int n,GridBase* value) {
-
-    std::cout << GridLogMessage << "BasisFieldVector allocated:\n";
-    std::cout << GridLogMessage << " Delta N = " << n << "\n";
-    std::cout << GridLogMessage << " Size of full vectors (size) = " << 
-      ((double)n*sizeof(vobj)*value->oSites() / 1024./1024./1024.) << " GB\n";
-    std::cout << GridLogMessage << " Size = " << _v.size() << " Capacity = " << _v.capacity() << std::endl;
-
-    value->Barrier();
-
-#ifdef __linux
-    if (value->IsBoss()) {
-      system("cat /proc/meminfo");
-    }
-#endif
-
-    value->Barrier();
-
-  }
-
-  BasisFieldVector(int Nm,GridBase* value) : _Nm(Nm), _v(Nm,value) {
-    report(Nm,value);
-  }
-  
-  ~BasisFieldVector() {
-  }
-
-  Field& operator[](int i) {
-    return _v[i];
-  }
-
-  void orthogonalize(Field& w, int k) {
-    basisOrthogonalize(_v,w,k);
-  }
-
-  void rotate(Eigen::MatrixXd& Qt,int j0, int j1, int k0,int k1,int Nm) {
-    basisRotate(_v,Qt,j0,j1,k0,k1,Nm);
-  }
-
-  size_t size() const {
-    return _Nm;
-  }
-
-  void resize(int n) {
-    if (n > _Nm)
-      _v.reserve(n);
-    
-    _v.resize(n,_v[0]._grid);
-
-    if (n < _Nm)
-      _v.shrink_to_fit();
-
-    report(n - _Nm,_v[0]._grid);
-
-    _Nm = n;
-  }
-
-  void sortInPlace(std::vector<RealD>& sort_vals, bool reverse) {
-    basisSortInPlace(_v,sort_vals,reverse);
-  }
-
-  void deflate(const std::vector<RealD>& eval,const Field& src_orig,Field& result) {
-    basisDeflate(_v,eval,src_orig,result);
-  }
-
- }; 
-}
diff --git a/tests/lanczos/FieldVectorIO.h b/tests/lanczos/FieldVectorIO.h
deleted file mode 100644
index 344e48fe..00000000
--- a/tests/lanczos/FieldVectorIO.h
+++ /dev/null
@@ -1,1085 +0,0 @@
-namespace Grid { 
-
- namespace FieldVectorIO {
-
-   // zlib's crc32 gets 0.4 GB/s on KNL single thread
-   // below gets 4.8 GB/s
-   static uint32_t crc32_threaded(unsigned char* data, int64_t len, uint32_t previousCrc32 = 0) {
-
-     // crc32 of zlib was incorrect for very large sizes, so do it block-wise
-     uint32_t crc = previousCrc32;
-     off_t blk = 0;
-     off_t step = 1024*1024*1024;
-     while (len > step) {
-       crc = crc32(crc,&data[blk],step);
-       blk += step;
-       len -= step;
-     }
-
-     crc = crc32(crc,&data[blk],len);
-     return crc;
-
-   }
- 
-   static int get_bfm_index( int* pos, int co, int* s ) {
-     
-     int ls = s[0];
-     int NtHalf = s[4] / 2;
-     int simd_coor = pos[4] / NtHalf;
-     int regu_coor = (pos[1] + s[1] * (pos[2] + s[2] * ( pos[3] + s[3] * (pos[4] % NtHalf) ) )) / 2;
-     
-     return regu_coor * ls * 48 + pos[0] * 48 + co * 4 + simd_coor * 2;
-   }
-   
-   static void get_read_geometry(const GridBase* _grid,const std::vector<int>& cnodes,
-			  std::map<int, std::vector<int> >& slots, 
-			  std::vector<int>& slot_lvol,
-			  std::vector<int>& lvol,
-			  int64_t& slot_lsites,int& ntotal) {
-
-     int _nd = (int)cnodes.size();
-     std::vector<int> nodes = cnodes;
-
-     slots.clear();
-     slot_lvol.clear();
-     lvol.clear();
-
-     int i;
-     ntotal = 1;
-     int64_t lsites = 1;
-     slot_lsites = 1;
-     for (i=0;i<_nd;i++) {
-       assert(_grid->_fdimensions[i] % nodes[i] == 0);
-       slot_lvol.push_back(_grid->_fdimensions[i] / nodes[i]);
-       lvol.push_back(_grid->_fdimensions[i] / _grid->_processors[i]);
-       lsites *= lvol.back();
-       slot_lsites *= slot_lvol.back();
-       ntotal *= nodes[i];
-     }
-
-     std::vector<int> lcoor, gcoor, scoor;
-     lcoor.resize(_nd); gcoor.resize(_nd);  scoor.resize(_nd);
-     
-     // create mapping of indices to slots
-     for (int lidx = 0; lidx < lsites; lidx++) {
-       Lexicographic::CoorFromIndex(lcoor,lidx,lvol);
-       for (int i=0;i<_nd;i++) {
-	 gcoor[i] = lcoor[i] + _grid->_processor_coor[i]*lvol[i];
-	 scoor[i] = gcoor[i] / slot_lvol[i];
-       }
-       int slot;
-       Lexicographic::IndexFromCoor(scoor,slot,nodes);
-       auto sl = slots.find(slot);
-       if (sl == slots.end())
-	 slots[slot] = std::vector<int>();
-       slots[slot].push_back(lidx);
-     }
-   }
-   
-   static void canonical_block_to_coarse_coordinates(GridBase* _coarsegrid,int nb,int& ii,int& oi) {
-      // canonical nb needs to be mapped in a coordinate on my coarsegrid (ii,io)
-      std::vector<int> _l = _coarsegrid->LocalDimensions();
-      std::vector<int> _cl = { _l[1], _l[2], _l[3], _l[4], _l[0] };
-      std::vector<int> _cc(_l.size());
-      Lexicographic::CoorFromIndex(_cc,nb,_cl);
-      std::vector<int> _c = { _cc[4], _cc[0], _cc[1], _cc[2], _cc[3] };
-      ii = _coarsegrid->iIndex(_c);
-      oi = _coarsegrid->oIndex(_c);
-    }
-
-    template<typename Field>
-     static bool read_argonne(BasisFieldVector<Field>& ret,const char* dir, const std::vector<int>& cnodes) {
-
-     GridBase* _grid = ret._v[0]._grid;
-
-     std::map<int, std::vector<int> > slots;
-     std::vector<int> slot_lvol, lvol;
-     int64_t slot_lsites;
-     int ntotal;
-     get_read_geometry(_grid,cnodes,
-		       slots,slot_lvol,lvol,slot_lsites,
-		       ntotal);
-     int _nd = (int)lvol.size();
-
-     // this is slow code to read the argonne file format for debugging purposes
-     int nperdir = ntotal / 32;
-     if (nperdir < 1)
-       nperdir=1;
-     std::cout << GridLogMessage << " Read " << dir << " nodes = " << cnodes << std::endl;
-     std::cout << GridLogMessage << " lvol = " << lvol << std::endl;
-     
-     // for error messages
-     char hostname[1024];
-     gethostname(hostname, 1024);
-
-     // now load one slot at a time and fill the vector
-     for (auto sl=slots.begin();sl!=slots.end();sl++) {
-       std::vector<int>& idx = sl->second;
-       int slot = sl->first;
-       std::vector<float> rdata;
-       
-       char buf[4096];
-       
-       sprintf(buf,"%s/checksums.txt",dir); printf("read_argonne: Reading from %s\n",buf);
-       FILE* f = fopen(buf,"rt");
-       if (!f) {
-	 fprintf(stderr,"Node %s cannot read %s\n",hostname,buf); fflush(stderr);
-	 return false;
-       }
-       
-       for (int l=0;l<3+slot;l++)
-	 fgets(buf,sizeof(buf),f);
-       uint32_t crc_exp = strtol(buf, NULL, 16);
-       fclose(f);
-       
-       // load one slot vector
-       sprintf(buf,"%s/%2.2d/%10.10d",dir,slot/nperdir,slot);
-       f = fopen(buf,"rb");
-       if (!f) {
-	 fprintf(stderr,"Node %s cannot read %s\n",hostname,buf); fflush(stderr);
-	 return false;
-       }
-       
-       fseeko(f,0,SEEK_END);
-       off_t total_size = ftello(f);
-       fseeko(f,0,SEEK_SET);
-       
-       int64_t size = slot_lsites / 2 * 24*4;
-       rdata.resize(size);
-       
-       assert(total_size % size == 0);
-       
-       int _Nfull = total_size / size;
-       ret._v.resize(_Nfull,ret._v[0]);
-       ret._Nm = _Nfull;
-       
-       uint32_t crc = 0x0;
-       GridStopWatch gsw,gsw2;
-       for (int nev = 0;nev < _Nfull;nev++) {
-	 
-	 gsw.Start();
-	 assert(fread(&rdata[0],size,1,f) == 1);
-	 gsw.Stop();
-	 
-	 gsw2.Start();
-	 crc = crc32_threaded((unsigned char*)&rdata[0],size,crc);
-	 gsw2.Stop();
-	 
-	 for (int i=0;i<size/4;i++) {
-	   char* c = (char*)&rdata[i];
-	   char tmp; int j;
-	   for (j=0;j<2;j++) {
-	     tmp = c[j]; c[j] = c[3-j]; c[3-j] = tmp;
-	   }
-	 }
-	 
-	 // loop
-	 ret._v[nev].checkerboard = Odd;
-#pragma omp parallel 
-	 {
-	   
-	   std::vector<int> lcoor, gcoor, scoor, slcoor;
-	   lcoor.resize(_nd); gcoor.resize(_nd); 
-	   slcoor.resize(_nd); scoor.resize(_nd);
-	   
-#pragma omp for
-	   for (int64_t lidx = 0; lidx < idx.size(); lidx++) {
-	     int llidx = idx[lidx];
-	     Lexicographic::CoorFromIndex(lcoor,llidx,lvol);
-	     for (int i=0;i<_nd;i++) {
-	       gcoor[i] = lcoor[i] + _grid->_processor_coor[i]*lvol[i];
-	       scoor[i] = gcoor[i] / slot_lvol[i];
-	       slcoor[i] = gcoor[i] - scoor[i]*slot_lvol[i];
-	     }
-	     
-	     if ((lcoor[1]+lcoor[2]+lcoor[3]+lcoor[4]) % 2 == 1) {
-	       // poke 
-	       iScalar<iVector<iVector<ComplexF, 3>, 4> > sc;
-	       for (int s=0;s<4;s++)
-		 for (int c=0;c<3;c++)
-		   sc()(s)(c) = *(std::complex<float>*)&rdata[get_bfm_index(&slcoor[0],c+s*3, &slot_lvol[0] )];
-	       
-	       pokeLocalSite(sc,ret._v[nev],lcoor);
-	     }
-	     
-	   }
-	 }
-       }
-       
-       fclose(f);      
-       std::cout << GridLogMessage << "Loading slot " << slot << " with " << idx.size() << " points and " 
-		 << _Nfull << " vectors in "
-		 << gsw.Elapsed() << " at " 
-		 << ( (double)size * _Nfull / 1024./1024./1024. / gsw.useconds()*1000.*1000. )
-		 << " GB/s " << " crc32 = " << std::hex << crc << " crc32_expected = " << crc_exp << std::dec
-		 << " computed at "
-		 << ( (double)size * _Nfull / 1024./1024./1024. / gsw2.useconds()*1000.*1000. )
-		 << " GB/s "
-		 << std::endl;
-       
-       assert(crc == crc_exp);
-     }
-
-     _grid->Barrier();
-     std::cout << GridLogMessage  << "Loading complete" << std::endl;
-     
-     return true;
-   }
-
-   template<typename Field>
-     static bool read_argonne(BasisFieldVector<Field>& ret,const char* dir) {
-
-
-     GridBase* _grid = ret._v[0]._grid;
-
-     char buf[4096];
-     sprintf(buf,"%s/nodes.txt",dir);
-     FILE* f = fopen(buf,"rt");
-     if (!f) {
-       if (_grid->IsBoss()) {
-	 fprintf(stderr,"Attempting to load eigenvectors without secifying node layout failed due to absence of nodes.txt\n");
-	 fflush(stderr);
-       }
-       return false;
-     }
-
-
-     std::vector<int> nodes((int)_grid->_processors.size());
-     for (int i =0;i<(int)_grid->_processors.size();i++)
-       assert(fscanf(f,"%d\n",&nodes[i])==1);
-     fclose(f);
-
-     return read_argonne(ret,dir,nodes);
-   }
-
-   static void flush_bytes(FILE* f, std::vector<char>& fbuf) {
-     if (fbuf.size()) {
-
-       if (fwrite(&fbuf[0],fbuf.size(),1,f) != 1) {
-	 fprintf(stderr,"Write failed of %g GB!\n",(double)fbuf.size() / 1024./1024./1024.);
-	 exit(2);
-       }
-
-       fbuf.resize(0);
-
-     }
-   }
-
-   static void write_bytes(void* buf, int64_t s, FILE* f, std::vector<char>& fbuf, uint32_t& crc) {
-     static double data_counter = 0.0;
-     static GridStopWatch gsw_crc, gsw_flush1,gsw_flush2,gsw_write,gsw_memcpy;
-     if (s == 0)
-       return;
-
-     // checksum
-     gsw_crc.Start();
-     crc = crc32_threaded((unsigned char*)buf,s,crc);
-     gsw_crc.Stop();
-
-     if (s > fbuf.capacity()) {
-       // cannot buffer this, so first flush current buffer contents and then write this directly to file
-       gsw_flush1.Start();
-       flush_bytes(f,fbuf);
-       gsw_flush1.Stop();
-
-       gsw_write.Start();
-       if (fwrite(buf,s,1,f) != 1) {
-	 fprintf(stderr,"Write failed of %g GB!\n",(double)s / 1024./1024./1024.);
-	 exit(2);
-       }
-       gsw_write.Stop();
-
-     }
-
-     // no room left in buffer, flush to disk
-     if (fbuf.size() + s > fbuf.capacity()) {
-       gsw_flush2.Start();
-       flush_bytes(f,fbuf);
-       gsw_flush2.Stop();
-     }
-
-     // then fill buffer again
-     {
-       gsw_memcpy.Start();
-       size_t t = fbuf.size();
-       fbuf.resize(t + s);
-       memcpy(&fbuf[t],buf,s);
-       gsw_memcpy.Stop();
-     }
-
-     data_counter += (double)s;
-     if (data_counter > 1024.*1024.*20.) {
-       std::cout << GridLogMessage << "Writing " << ((double)data_counter / 1024./1024./1024.) << " GB at"
-	 " crc = " << gsw_crc.Elapsed() << " flush1 = " << gsw_flush1.Elapsed() << " flush2 = " << gsw_flush2.Elapsed() <<
-	 " write = " << gsw_write.Elapsed() << " memcpy = " << gsw_memcpy.Elapsed() << std::endl;
-       data_counter = 0.0;
-       gsw_crc.Reset();
-       gsw_write.Reset();
-       gsw_memcpy.Reset();
-       gsw_flush1.Reset();
-       gsw_flush2.Reset();
-     }
-   }
-
-   static void write_floats(FILE* f, std::vector<char>& fbuf, uint32_t& crc, float* buf, int64_t n) {
-     write_bytes(buf,n*sizeof(float),f,fbuf,crc);
-   }
-
-   static void read_floats(char* & ptr, float* out, int64_t n) {
-     float* in = (float*)ptr;
-     ptr += 4*n;
-
-     for (int64_t i=0;i<n;i++)
-       out[i] = in[i];
-   }
-
-   static int fp_map(float in, float min, float max, int N) {
-     // Idea:
-     //
-     // min=-6
-     // max=6
-     //
-     // N=1
-     // [-6,0] -> 0, [0,6] -> 1;  reconstruct 0 -> -3, 1-> 3
-     //
-     // N=2
-     // [-6,-2] -> 0, [-2,2] -> 1, [2,6] -> 2;  reconstruct 0 -> -4, 1->0, 2->4
-     int ret =  (int) ( (float)(N+1) * ( (in - min) / (max - min) ) );
-     if (ret == N+1) {
-       ret = N;
-     }
-     return ret;
-   }
-
-   static float fp_unmap(int val, float min, float max, int N) {
-     return min + (float)(val + 0.5) * (max - min)  / (float)( N + 1 );
-   }
-
-#define SHRT_UMAX 65535
-#define FP16_BASE 1.4142135623730950488
-#define FP16_COEF_EXP_SHARE_FLOATS 10
-   static float unmap_fp16_exp(unsigned short e) {
-     float de = (float)((int)e - SHRT_UMAX / 2);
-     return ::pow( FP16_BASE, de );
-   }
-
-   // can assume that v >=0 and need to guarantee that unmap_fp16_exp(map_fp16_exp(v)) >= v
-   static unsigned short map_fp16_exp(float v) {
-     // float has exponents 10^{-44.85} .. 10^{38.53}
-     int exp = (int)ceil(::log(v) / ::log(FP16_BASE)) + SHRT_UMAX / 2;
-     if (exp < 0 || exp > SHRT_UMAX) {
-       fprintf(stderr,"Error in map_fp16_exp(%g,%d)\n",v,exp);
-       exit(3);
-     }
-
-     return (unsigned short)exp;
-   }
-  
-   template<typename OPT>
-     static void read_floats_fp16(char* & ptr, OPT* out, int64_t n, int nsc) {
-
-     int64_t nsites = n / nsc;
-     if (n % nsc) {
-       fprintf(stderr,"Invalid size in write_floats_fp16\n");
-       exit(4);
-     }
-
-     unsigned short* in = (unsigned short*)ptr;
-     ptr += 2*(n+nsites);
-
-     // do for each site
-     for (int64_t site = 0;site<nsites;site++) {
-
-       OPT* ev = &out[site*nsc];
-
-       unsigned short* bptr = &in[site*(nsc + 1)];
-
-       unsigned short exp = *bptr++;
-       OPT max = unmap_fp16_exp(exp);
-       OPT min = -max;
-
-       for (int i=0;i<nsc;i++) {
-	 ev[i] = fp_unmap( *bptr++, min, max, SHRT_UMAX );
-       }
-
-     }
-
-   }
-
-   template<typename OPT>
-     static void write_floats_fp16(FILE* f, std::vector<char>& fbuf, uint32_t& crc, OPT* in, int64_t n, int nsc) {
-
-     int64_t nsites = n / nsc;
-     if (n % nsc) {
-       fprintf(stderr,"Invalid size in write_floats_fp16\n");
-       exit(4);
-     }
-
-     unsigned short* buf = (unsigned short*)malloc( sizeof(short) * (n + nsites) );
-     if (!buf) {
-       fprintf(stderr,"Out of mem\n");
-       exit(1);
-     }
-
-     // do for each site
-#pragma omp parallel for
-     for (int64_t site = 0;site<nsites;site++) {
-
-       OPT* ev = &in[site*nsc];
-
-       unsigned short* bptr = &buf[site*(nsc + 1)];
-
-       OPT max = fabs(ev[0]);
-       OPT min;
-
-       for (int i=0;i<nsc;i++) {
-	 if (fabs(ev[i]) > max)
-	   max = fabs(ev[i]);
-       }
-
-       unsigned short exp = map_fp16_exp(max);
-       max = unmap_fp16_exp(exp);
-       min = -max;
-
-       *bptr++ = exp;
-
-       for (int i=0;i<nsc;i++) {
-	 int val = fp_map( ev[i], min, max, SHRT_UMAX );
-	 if (val < 0 || val > SHRT_UMAX) {
-	   fprintf(stderr,"Assert failed: val = %d (%d), ev[i] = %.15g, max = %.15g, exp = %d\n",val,SHRT_UMAX,ev[i],max,(int)exp);
-	   exit(48);
-	 }
-	 *bptr++ = (unsigned short)val;
-       }
-
-     }
-
-     write_bytes(buf,sizeof(short)*(n + nsites),f,fbuf,crc);
-
-     free(buf);
-   }
-
-   template<typename Field,typename CoarseField>
-     static bool read_compressed_vectors(const char* dir,BlockProjector<Field>& pr,BasisFieldVector<CoarseField>& coef, int ngroups = 1) {
-
-     const BasisFieldVector<Field>& basis = pr._evec;
-     GridBase* _grid = basis._v[0]._grid;
-
-     // for error messages
-     char hostname[1024];
-     gethostname(hostname, 1024);
-
-     std::cout << GridLogMessage << "Ready on host " << hostname << " with " << ngroups << " reader groups" << std::endl;
-
-     // first read metadata
-     char buf[4096];
-     sprintf(buf,"%s/metadata.txt",dir);
-
-     std::vector<uint32_t> s,b,nb,nn,crc32;
-     s.resize(5); b.resize(5); nb.resize(5); nn.resize(5);
-     uint32_t neig, nkeep, nkeep_single, blocks, _FP16_COEF_EXP_SHARE_FLOATS;
-     uint32_t nprocessors = 1;
-
-     FILE* f = 0;
-     uint32_t status = 0;
-     if (_grid->IsBoss()) {
-       f = fopen(buf,"rb");
-       status=f ? 1 : 0;
-     }
-     _grid->GlobalSum(status);
-     std::cout << GridLogMessage << "Read params status " << status << std::endl;
-
-     if (!status) {
-       return false;
-     }
-
-#define _IRL_READ_INT(buf,p) if (f) { assert(fscanf(f,buf,p)==1); } else { *(p) = 0; } _grid->GlobalSum(*(p));
-
-     for (int i=0;i<5;i++) {
-       sprintf(buf,"s[%d] = %%d\n",i);
-       _IRL_READ_INT(buf,&s[(i+1)%5]);
-     }
-     for (int i=0;i<5;i++) {
-       sprintf(buf,"b[%d] = %%d\n",i);
-       _IRL_READ_INT(buf,&b[(i+1)%5]);
-     }
-     for (int i=0;i<5;i++) {
-       sprintf(buf,"nb[%d] = %%d\n",i);
-       _IRL_READ_INT(buf,&nb[(i+1)%5]);
-     }
-     _IRL_READ_INT("neig = %d\n",&neig);
-     _IRL_READ_INT("nkeep = %d\n",&nkeep);
-     _IRL_READ_INT("nkeep_single = %d\n",&nkeep_single);
-     _IRL_READ_INT("blocks = %d\n",&blocks);
-     _IRL_READ_INT("FP16_COEF_EXP_SHARE_FLOATS = %d\n",&_FP16_COEF_EXP_SHARE_FLOATS);
-
-     for (int i=0;i<5;i++) {
-       assert(_grid->FullDimensions()[i] % s[i] == 0);
-       nn[i] = _grid->FullDimensions()[i] / s[i];
-       nprocessors *= nn[i];
-     }
-
-     std::cout << GridLogMessage << "Reading data that was generated on node-layout " << nn << std::endl;
-
-     crc32.resize(nprocessors);
-     for (int i =0;i<nprocessors;i++) {
-       sprintf(buf,"crc32[%d] = %%X\n",i);
-       _IRL_READ_INT(buf,&crc32[i]);
-     }
-
-#undef _IRL_READ_INT
-
-     if (f)
-       fclose(f);
-
-     // allocate memory
-     assert(std::equal(pr._bgrid._bs.begin(),pr._bgrid._bs.end(),b.begin())); // needs to be the same for now
-     assert(pr._evec.size() == nkeep); // needs to be the same since this is compile-time fixed
-     coef.resize(neig);
-
-
-     // now get read geometry
-     std::map<int, std::vector<int> > slots;
-     std::vector<int> slot_lvol, lvol;
-     int64_t slot_lsites;
-     int ntotal;
-     std::vector<int> _nn(nn.begin(),nn.end());
-     get_read_geometry(_grid,_nn,
-		       slots,slot_lvol,lvol,slot_lsites,
-		       ntotal);
-     int _nd = (int)lvol.size();
-
-     // types
-     typedef typename Field::scalar_type Coeff_t;
-     typedef typename CoarseField::scalar_type CoeffCoarse_t;
-
-     // slot layout
-     int nperdir = ntotal / 32;
-     if (nperdir < 1)
-       nperdir=1;
-
-     // add read groups
-     for (int ngroup=0;ngroup<ngroups;ngroup++) {
-
-       bool action = _grid->ThisRank() % ngroups == ngroup;
-       
-       std::cout << GridLogMessage << "Reading in group " << ngroup << " / " << ngroups << std::endl;
-
-       // load all necessary slots and store them appropriately
-       for (auto sl=slots.begin();sl!=slots.end();sl++) {
-
-	 std::vector<int>& idx = sl->second;
-	 int slot = sl->first;
-	 std::vector<float> rdata;
-       
-	 char buf[4096];
-       
-	 if (action) {
-	   // load one slot vector
-	   sprintf(buf,"%s/%2.2d/%10.10d.compressed",dir,slot/nperdir,slot);
-	   f = fopen(buf,"rb");
-	   if (!f) {
-	     fprintf(stderr,"Node %s cannot read %s\n",hostname,buf); fflush(stderr);
-	     return false;
-	   }
-	 }
-
-	 uint32_t crc = 0x0;
-	 off_t size;
-	   
-	 GridStopWatch gsw;
-	 _grid->Barrier();
-	 gsw.Start();
-	 
-	 std::vector<char> raw_in(0);
-	 if (action) {
-	   fseeko(f,0,SEEK_END);
-	   size = ftello(f);
-	   fseeko(f,0,SEEK_SET);
-
-	   raw_in.resize(size);
-	   assert(fread(&raw_in[0],size,1,f) == 1);
-	 }
-
-	 _grid->Barrier();
-	 gsw.Stop();
-
-	 RealD totalGB = (RealD)size / 1024./1024./1024 * _grid->_Nprocessors;
-	 RealD seconds = gsw.useconds() / 1e6;
-
-	 if (action) {
-	   std::cout << GridLogMessage << "[" << slot << "]  Read " << totalGB << " GB of compressed data at " << totalGB/seconds << " GB/s" << std::endl;
-	   
-	   uint32_t crc_comp = crc32_threaded((unsigned char*)&raw_in[0],size,0);
-	   
-	   if (crc_comp != crc32[slot]) {
-	     std::cout << "Node " << hostname << " found crc mismatch for file " << buf << " (" << std::hex << crc_comp << " vs " << crc32[slot] << std::dec << ")" << std::endl;
-	     std::cout << "Byte size: " << size << std::endl;
-	   }
-
-	   assert(crc_comp == crc32[slot]);
-	 }
-
-	 _grid->Barrier();
-       
-	 if (action) {
-	   fclose(f);
-	 }
-
-	 char* ptr = &raw_in[0];
-	 
-	 GridStopWatch gsw2;
-	 gsw2.Start();
-	 if (action) {
-	   int nsingleCap = nkeep_single;
-	   if (pr._evec.size() < nsingleCap)
-	     nsingleCap = pr._evec.size();
-	   
-	   int _cf_block_size = slot_lsites * 12 / 2 / blocks;
-	   
-#define FP_16_SIZE(a,b)  (( (a) + (a/b) )*2)
-	   
-	   // first read single precision basis vectors
-#pragma omp parallel
-	   {
-	     std::vector<float> buf(_cf_block_size * 2);
-#pragma omp for
-	     for (int nb=0;nb<blocks;nb++) {
-	       for (int i=0;i<nsingleCap;i++) {
-		 char* lptr = ptr + buf.size()*(i + nsingleCap*nb)*4;
-		 read_floats(lptr, &buf[0], buf.size() );
-		 int mnb = pr._bgrid.globalToLocalCanonicalBlock(slot,_nn,nb);
-		 if (mnb != -1)
-		   pr._bgrid.pokeBlockOfVectorCanonical(mnb,pr._evec._v[i],buf);
-	       }
-	     }
-	     
-#pragma omp barrier
-#pragma omp single
-	     {
-	       ptr = ptr + buf.size()*nsingleCap*blocks*4;
-	     }
-	     
-	   }
-	   
-	   // TODO: at this point I should add a checksum test for block_sp(nb,v,v) for all blocks, then I would know that the mapping
-	   // to blocks is OK at this point; after that ...
-	   
-	   // then read fixed precision basis vectors
-#pragma omp parallel
-	   {
-	     std::vector<float> buf(_cf_block_size * 2);
-#pragma omp for
-	     for (int nb=0;nb<blocks;nb++) {
-	       for (int i=nsingleCap;i<(int)pr._evec.size();i++) {
-		 char* lptr = ptr + FP_16_SIZE( buf.size(), 24 )*((i-nsingleCap) + (pr._evec.size() - nsingleCap)*nb);
-		 read_floats_fp16(lptr, &buf[0], buf.size(), 24);
-		 int mnb = pr._bgrid.globalToLocalCanonicalBlock(slot,_nn,nb);
-		 if (mnb != -1)
-		   pr._bgrid.pokeBlockOfVectorCanonical(mnb,pr._evec._v[i],buf);
-	       }
-	     }
-	     
-#pragma omp barrier
-#pragma omp single
-	     {
-	       ptr = ptr + FP_16_SIZE( buf.size()*(pr._evec.size() - nsingleCap)*blocks, 24 );
-	     }
-	   }
-	   
-#pragma omp parallel
-	   {
-	     std::vector<float> buf1(nkeep_single*2);
-	     std::vector<float> buf2((nkeep - nkeep_single)*2);
-	     
-#pragma omp for
-	     for (int j=0;j<(int)coef.size();j++)
-	       for (int nb=0;nb<blocks;nb++) {
-		 // get local coordinate on coarse grid
-		 int ii,oi;
-		 int mnb = pr._bgrid.globalToLocalCanonicalBlock(slot,_nn,nb);
-		 if (mnb != -1)
-		   canonical_block_to_coarse_coordinates(coef._v[0]._grid,mnb,ii,oi);
-		 
-		 char* lptr = ptr + (4*buf1.size() + FP_16_SIZE(buf2.size(), _FP16_COEF_EXP_SHARE_FLOATS))*(nb + j*blocks);
-		 int l;
-		 read_floats(lptr, &buf1[0], buf1.size() );
-		 if (mnb != -1) {
-		   for (l=0;l<nkeep_single;l++) {
-		     ((CoeffCoarse_t*)&coef._v[j]._odata[oi]._internal._internal[l])[ii] = CoeffCoarse_t(buf1[2*l+0],buf1[2*l+1]);
-		   }
-		 }
-		 read_floats_fp16(lptr, &buf2[0], buf2.size(), _FP16_COEF_EXP_SHARE_FLOATS);
-		 if (mnb != -1) {
-		   for (l=nkeep_single;l<nkeep;l++) {
-		     ((CoeffCoarse_t*)&coef._v[j]._odata[oi]._internal._internal[l])[ii] = CoeffCoarse_t(buf2[2*(l-nkeep_single)+0],buf2[2*(l-nkeep_single)+1]);
-		   }
-		 }
-		 
-	       }
-	   }
-	   
-	   // set checkerboard
-	   for (int i=0;i<(int)pr._evec.size();i++)
-	     pr._evec._v[i].checkerboard = Odd;
-	 
-	   gsw2.Stop();
-	   seconds=gsw2.useconds()/1e6;
-	   std::cout << GridLogMessage << "Processed " << totalGB << " GB of compressed data at " << totalGB/seconds << " GB/s" << std::endl;
-	 }
-       }
-     }
-#undef FP_16_SIZE
-     return true;
-
-   }
- 
-   static bool DirectoryExists(const char *path) {
-     struct stat info;
-     return ((stat( path, &info ) == 0) && (info.st_mode & S_IFDIR));
-   }
- 
-   static void conditionalMkDir(const char* path) {
-     if (!DirectoryExists(path))
-       mkdir(path,ACCESSPERMS);
-   }
- 
-   template<typename Field,typename CoarseField>
-   static void write_compressed_vectors(const char* dir,const BlockProjector<Field>& pr,
-					const BasisFieldVector<CoarseField>& coef,
-					int nsingle,int writer_nodes = 0) {
-
-     GridStopWatch gsw;
-     
-     const BasisFieldVector<Field>& basis = pr._evec;
-     GridBase* _grid = basis._v[0]._grid;
-     std::vector<int> _l = _grid->FullDimensions();
-     for (int i=0;i<(int)_l.size();i++)
-       _l[i] /= _grid->_processors[i];
-
-     _grid->Barrier();
-     gsw.Start();
-
-     char buf[4096];
-     
-     // Making the directories is somewhat tricky.
-     // If we run on a joint filesystem we would just 
-     // have the boss create the directories and then
-     // have a barrier.  We also want to be able to run
-     // on local /scratch, so potentially all nodes need
-     // to create their own directories.  So do the following
-     // for now.
-     for (int j=0;j<_grid->_Nprocessors;j++) {
-       if (j == _grid->ThisRank()) {
-	 conditionalMkDir(dir);
-	 for (int i=0;i<32;i++) {
-	   sprintf(buf,"%s/%2.2d",dir,i);
-	   conditionalMkDir(buf);
-	 }       
-	 _grid->Barrier(); // make sure directories are ready
-       }
-     }
-     
-
-     typedef typename Field::scalar_type Coeff_t;
-     typedef typename CoarseField::scalar_type CoeffCoarse_t;
-     
-     int nperdir = _grid->_Nprocessors / 32;
-     if (nperdir < 1)
-       nperdir=1;
-
-     int slot;
-     Lexicographic::IndexFromCoor(_grid->_processor_coor,slot,_grid->_processors);
-
-     int64_t off = 0x0;
-     uint32_t crc = 0x0;
-     if (writer_nodes < 1)
-       writer_nodes = _grid->_Nprocessors;
-     int groups = _grid->_Nprocessors / writer_nodes;
-     if (groups<1)
-       groups = 1;
-
-     std::cout << GridLogMessage << " Write " << dir << " nodes = " << writer_nodes << std::endl;
-
-     for (int group=0;group<groups;group++) {
-       _grid->Barrier();
-       if (_grid->ThisRank() % groups == group) {
-	 
-	 sprintf(buf,"%s/%2.2d/%10.10d.compressed",dir,slot/nperdir,slot);
-	 FILE* f = fopen(buf,"wb");
-	 assert(f);
-
-	 //buffer does not seem to help
-	 //assert(!setvbuf ( f , NULL , _IOFBF , 1024*1024*2 ));
-	 	 
-	 int nsingleCap = nsingle;
-	 if (pr._evec.size() < nsingleCap)
-	   nsingleCap = pr._evec.size();
-	 
-	 GridStopWatch gsw1,gsw2,gsw3,gsw4,gsw5;
-
-	 gsw1.Start();
-
-	 std::vector<char> fbuf;
-	 fbuf.reserve( 1024 * 1024 * 8 );
-
-	 // first write single precision basis vectors
-	 for (int nb=0;nb<pr._bgrid._blocks;nb++) {
-	   for (int i=0;i<nsingleCap;i++) {
-	     std::vector<float> buf;
-	     pr._bgrid.peekBlockOfVectorCanonical(nb,pr._evec._v[i],buf);
-	     
-#if 0
-	     {
-	       RealD nrm = 0.0;
-	       for (int j=0;j<(int)buf.size();j++)
-		 nrm += buf[j]*buf[j];
-	       std::cout << GridLogMessage << "Norm: " << nrm << std::endl;
-	     }
-#endif
-	     write_floats(f,fbuf,crc, &buf[0], buf.size() );
-	   }
-	 }
-
-	 gsw1.Stop();
-	 gsw2.Start();
-
-	 // then write fixed precision basis vectors
-	 for (int nb=0;nb<pr._bgrid._blocks;nb++) {
-	   for (int i=nsingleCap;i<(int)pr._evec.size();i++) {
-	     std::vector<float> buf;
-	     pr._bgrid.peekBlockOfVectorCanonical(nb,pr._evec._v[i],buf);
-	     write_floats_fp16(f,fbuf,crc, &buf[0], buf.size(), 24);
-	   }
-	 }
-
-	 gsw2.Stop();
-	 assert(coef._v[0]._grid->_isites*coef._v[0]._grid->_osites == pr._bgrid._blocks);
-
-	 gsw3.Start();
-	 for (int j=0;j<(int)coef.size();j++) {
-
-	   int64_t size1 = nsingleCap*2;
-	   int64_t size2 = 2*(pr._evec.size()-nsingleCap);
-	   int64_t size = size1;
-	   if (size2>size)
-	     size=size2;
-	   std::vector<float> buf(size);
-
-	   //RealD nrmTest = 0.0;
-	   for (int nb=0;nb<pr._bgrid._blocks;nb++) {
-	     // get local coordinate on coarse grid
-	     int ii, oi;
-	     canonical_block_to_coarse_coordinates(coef._v[0]._grid,nb,ii,oi);
-	     
-	     gsw4.Start();
-	     gsw5.Start();
-	     for (int l=0;l<nsingleCap;l++) {
-	       auto res = ((CoeffCoarse_t*)&coef._v[j]._odata[oi]._internal._internal[l])[ii];
-	       buf[2*l+0] = res.real();
-	       buf[2*l+1] = res.imag();
-	       //nrmTest += res.real() * res.real() + res.imag() * res.imag();
-	     }
-	     gsw5.Stop();
-	     write_floats(f,fbuf,crc, &buf[0], size1 );
-	     gsw4.Stop();
-	     for (int l=nsingleCap;l<(int)pr._evec.size();l++) {
-	       auto res = ((CoeffCoarse_t*)&coef._v[j]._odata[oi]._internal._internal[l])[ii];
-	       buf[2*(l-nsingleCap)+0] = res.real();
-	       buf[2*(l-nsingleCap)+1] = res.imag();
-	       //nrmTest += res.real() * res.real() + res.imag() * res.imag();
-	     }
-	     write_floats_fp16(f,fbuf,crc, &buf[0], size2, FP16_COEF_EXP_SHARE_FLOATS);
-	   }
-
-	   //_grid->GlobalSum(nrmTest);
-	   //std::cout << GridLogMessage << "Test norm: " << nrmTest << std::endl;
-	 }
-	 gsw3.Stop();
-
-	 flush_bytes(f,fbuf);
-
-	 off = ftello(f);	 
-	 fclose(f);
-
-	 std::cout<<GridLogMessage << "Timing: write single basis in " << gsw1.Elapsed() << " and fp16 in " << gsw2.Elapsed() << " and coefficients in " << gsw3.Elapsed() << " (" <<
-	   gsw4.Elapsed() << ", " << gsw5.Elapsed() << ")" << std::endl;
-       }
-     }
-	 
-     _grid->Barrier();
-     gsw.Stop();
-     
-     RealD totalGB = (RealD)off / 1024./1024./1024 * _grid->_Nprocessors;
-     RealD seconds = gsw.useconds() / 1e6;
-     std::cout << GridLogMessage << "Write " << totalGB << " GB of compressed data at " << totalGB/seconds << " GB/s in " << seconds << " s" << std::endl;
-
-     // gather crcs
-     std::vector<uint32_t> crcs(_grid->_Nprocessors);
-     for (int i=0;i<_grid->_Nprocessors;i++) {
-       crcs[i] = 0x0;
-     }
-     crcs[slot] = crc;
-     for (int i=0;i<_grid->_Nprocessors;i++) {
-       _grid->GlobalSum(crcs[i]);
-     }
-     
-     if (_grid->IsBoss()) {
-       sprintf(buf,"%s/metadata.txt",dir);
-       FILE* f = fopen(buf,"wb");
-       assert(f);
-       for (int i=0;i<5;i++)
-	 fprintf(f,"s[%d] = %d\n",i,_grid->FullDimensions()[(i+1)%5] / _grid->_processors[(i+1)%5]);
-       for (int i=0;i<5;i++)
-	 fprintf(f,"b[%d] = %d\n",i,pr._bgrid._bs[(i+1)%5]);
-       for (int i=0;i<5;i++)
-	 fprintf(f,"nb[%d] = %d\n",i,pr._bgrid._nb[(i+1)%5]);
-       fprintf(f,"neig = %d\n",(int)coef.size());
-       fprintf(f,"nkeep = %d\n",(int)pr._evec.size());
-       fprintf(f,"nkeep_single = %d\n",nsingle);
-       fprintf(f,"blocks = %d\n",pr._bgrid._blocks);
-       fprintf(f,"FP16_COEF_EXP_SHARE_FLOATS = %d\n",FP16_COEF_EXP_SHARE_FLOATS);
-       for (int i =0;i<_grid->_Nprocessors;i++)
-	 fprintf(f,"crc32[%d] = %X\n",i,crcs[i]);
-       fclose(f);
-     }
-
-   }
-
-   template<typename Field>
-     static void write_argonne(const BasisFieldVector<Field>& ret,const char* dir) {
-     
-     GridBase* _grid = ret._v[0]._grid;
-     std::vector<int> _l = _grid->FullDimensions();
-     for (int i=0;i<(int)_l.size();i++)
-       _l[i] /= _grid->_processors[i];
-
-     char buf[4096];
-     
-     if (_grid->IsBoss()) {
-       mkdir(dir,ACCESSPERMS);
-       
-       for (int i=0;i<32;i++) {
-	 sprintf(buf,"%s/%2.2d",dir,i);
-	 mkdir(buf,ACCESSPERMS);
-       }
-     }
-     
-     _grid->Barrier(); // make sure directories are ready
-
-     
-     int nperdir = _grid->_Nprocessors / 32;
-     if (nperdir < 1)
-       nperdir=1;
-     std::cout << GridLogMessage << " Write " << dir << " nodes = " << _grid->_Nprocessors << std::endl;
-     
-     int slot;
-     Lexicographic::IndexFromCoor(_grid->_processor_coor,slot,_grid->_processors);
-     //printf("Slot: %d <> %d\n",slot, _grid->ThisRank());
-     
-     sprintf(buf,"%s/%2.2d/%10.10d",dir,slot/nperdir,slot);
-     FILE* f = fopen(buf,"wb");
-     assert(f);
-     
-     int N = (int)ret._v.size();
-     uint32_t crc = 0x0;
-     int64_t cf_size = _grid->oSites()*_grid->iSites()*12;
-     std::vector< float > rdata(cf_size*2);
-     
-     GridStopWatch gsw1,gsw2;
-     
-     for (int i=0;i<N;i++) {
-
-       if (ret._v[i].checkerboard != Odd)
-	 continue;
-
-       // create buffer and put data in argonne format in there
-       std::vector<int> coor(_l.size());
-       for (coor[1] = 0;coor[1]<_l[1];coor[1]++) {
-	 for (coor[2] = 0;coor[2]<_l[2];coor[2]++) {
-	   for (coor[3] = 0;coor[3]<_l[3];coor[3]++) {
-	     for (coor[4] = 0;coor[4]<_l[4];coor[4]++) {
-	       for (coor[0] = 0;coor[0]<_l[0];coor[0]++) {
-		 
-		 if ((coor[1]+coor[2]+coor[3]+coor[4]) % 2 == 1) {
-		   // peek
-		   iScalar<iVector<iVector<ComplexF, 3>, 4> > sc;
-		   peekLocalSite(sc,ret._v[i],coor);
-		   for (int s=0;s<4;s++)
-		     for (int c=0;c<3;c++)
-		       *(std::complex<float>*)&rdata[get_bfm_index(&coor[0],c+s*3, &_l[0] )] = sc()(s)(c);
-		 }
-	       }
-	     }
-	   }
-	 }
-       }
-       
-       // endian flip
-       for (int i=0;i<cf_size*2;i++) {
-	 char* c = (char*)&rdata[i];
-	 char tmp; int j;
-	 for (j=0;j<2;j++) {
-	   tmp = c[j]; c[j] = c[3-j]; c[3-j] = tmp;
-	 }
-       }
-       
-       // create crc of buffer
-       gsw1.Start();
-       crc = crc32_threaded((unsigned char*)&rdata[0],cf_size*2*4,crc);    
-       gsw1.Stop();
-       
-       // write out
-       gsw2.Start();
-       assert(fwrite(&rdata[0],cf_size*2*4,1,f)==1);
-       gsw2.Stop();
-       
-     }
-     
-     fclose(f);
-     
-     
-     // gather crc's and write out
-     std::vector<uint32_t> crcs(_grid->_Nprocessors);
-     for (int i=0;i<_grid->_Nprocessors;i++) {
-       crcs[i] = 0x0;
-     }
-     crcs[slot] = crc;
-     for (int i=0;i<_grid->_Nprocessors;i++) {
-       _grid->GlobalSum(crcs[i]);
-     }
-     
-     if (_grid->IsBoss()) {
-       sprintf(buf,"%s/checksums.txt",dir);
-       FILE* f = fopen(buf,"wt");
-       assert(f);
-       fprintf(f,"00000000\n\n");
-       for (int i =0;i<_grid->_Nprocessors;i++)
-	 fprintf(f,"%X\n",crcs[i]);
-       fclose(f);
-       
-       sprintf(buf,"%s/nodes.txt",dir);
-       f = fopen(buf,"wt");
-       assert(f);
-       for (int i =0;i<(int)_grid->_processors.size();i++)
-	 fprintf(f,"%d\n",_grid->_processors[i]);
-       fclose(f);
-     }
-     
-     
-     std::cout << GridLogMessage << "Writing slot " << slot << " with "
-	       << N << " vectors in "
-	       << gsw2.Elapsed() << " at " 
-	       << ( (double)cf_size*2*4 * N / 1024./1024./1024. / gsw2.useconds()*1000.*1000. )
-	       << " GB/s  with crc computed at "
-	       << ( (double)cf_size*2*4 * N / 1024./1024./1024. / gsw1.useconds()*1000.*1000. )
-	       << " GB/s "
-	       << std::endl;
-     
-     _grid->Barrier();
-     std::cout << GridLogMessage  << "Writing complete" << std::endl;
-     
-   }
- }
-
-}
diff --git a/tests/lanczos/Params.h b/tests/lanczos/Params.h
deleted file mode 100644
index d9a6d3b3..00000000
--- a/tests/lanczos/Params.h
+++ /dev/null
@@ -1,136 +0,0 @@
-/*
-  Params IO
-
-  Author: Christoph Lehner
-  Date:   2017
-*/
-
-#define PADD(p,X) p.get(#X,X);
-
-class Params {
- protected:
-
-  std::string trim(const std::string& sc) {
-    std::string s = sc;
-    s.erase(s.begin(), std::find_if(s.begin(), s.end(),
-				    std::not1(std::ptr_fun<int, int>(std::isspace))));
-    s.erase(std::find_if(s.rbegin(), s.rend(),
-			 std::not1(std::ptr_fun<int, int>(std::isspace))).base(), s.end());
-    return s;
-  }
-
- public:
-
-  std::map< std::string, std::string > lines;
-  std::string _fn;
-
- Params(const char* fn) : _fn(fn) {
-    FILE* f = fopen(fn,"rt");
-    assert(f);
-    while (!feof(f)) {
-      char buf[4096];
-      if (fgets(buf,sizeof(buf),f)) {
-	if (buf[0] != '#' && buf[0] != '\r' && buf[0] != '\n') {
-	  char* sep = strchr(buf,'=');
-	  assert(sep);
-	  *sep = '\0';
-	  lines[trim(buf)] = trim(sep+1);
-	}
-      }
-    }      
-    fclose(f);
-  }
-
-  ~Params() {
-  }
-
-  std::string loghead() {
-    return _fn + ": ";
-  }
-
-  bool has(const char* name) {
-    auto f = lines.find(name);
-    return (f != lines.end());
-  }
-
-  const std::string& get(const char* name) {
-    auto f = lines.find(name);
-    if (f == lines.end()) {
-      std::cout << Grid::GridLogMessage << loghead() << "Could not find value for " << name << std::endl;
-      abort();
-    }
-    return f->second;
-  }
-
-  void parse(std::string& s, const std::string& cval) {
-    std::stringstream trimmer;
-    trimmer << cval;
-    s.clear();
-    trimmer >> s;
-  }
-
-  void parse(int& i, const std::string& cval) {
-    assert(sscanf(cval.c_str(),"%d",&i)==1);
-  }
-
-  void parse(long long& i, const std::string& cval) {
-    assert(sscanf(cval.c_str(),"%lld",&i)==1);
-  }
-
-  void parse(double& f, const std::string& cval) {
-    assert(sscanf(cval.c_str(),"%lf",&f)==1);
-  }
-
-  void parse(float& f, const std::string& cval) {
-    assert(sscanf(cval.c_str(),"%f",&f)==1);
-  }
-
-  void parse(bool& b, const std::string& cval) {
-    std::string lcval = cval;
-    std::transform(lcval.begin(), lcval.end(), lcval.begin(), ::tolower);
-    if (lcval == "true" || lcval == "yes") {
-      b = true;
-    } else if (lcval == "false" || lcval == "no") {
-      b = false;
-    } else {
-      std::cout << "Invalid value for boolean: " << b << std::endl;
-      assert(0);
-    }
-  }
-
-  void parse(std::complex<double>& f, const std::string& cval) {
-    double r,i;
-    assert(sscanf(cval.c_str(),"%lf %lf",&r,&i)==2);
-    f = std::complex<double>(r,i);
-  }
-
-  void parse(std::complex<float>& f, const std::string& cval) {
-    float r,i;
-    assert(sscanf(cval.c_str(),"%f %f",&r,&i)==2);
-    f = std::complex<float>(r,i);
-  }
-
-  template<class T>
-    void get(const char* name, std::vector<T>& v) {
-    int i = 0;
-    v.resize(0);
-    while (true) {
-      char buf[4096];
-      sprintf(buf,"%s[%d]",name,i++);
-      if (!has(buf))
-	break;
-      T val;
-      parse(val,get(buf));
-      std::cout << Grid::GridLogMessage << loghead() << "Set " << buf << " to " << val << std::endl;
-      v.push_back(val);
-    }
-  }
-
-  template<class T>
-    void get(const char* name, T& f) {
-    parse(f,get(name));
-    std::cout << Grid::GridLogMessage << loghead() << "Set " << name << " to " << f << std::endl;
-  }
-
-  
-};
diff --git a/tests/lanczos/Test_WCMultiRep_lanczos.cc b/tests/lanczos/Test_WCMultiRep_lanczos.cc
index b6d69aee..58759c96 100644
--- a/tests/lanczos/Test_WCMultiRep_lanczos.cc
+++ b/tests/lanczos/Test_WCMultiRep_lanczos.cc
@@ -30,7 +30,6 @@ directory
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 
 //typedef WilsonCloverFermionR FermionOp;
@@ -154,11 +153,11 @@ NerscHmcCheckpointer<PeriodicGimplR> Checkpoint(CPparams);
   std::vector<ASymmFermionField>   ASevec(Nm, FGrid);
   
   for (int i = 0; i < 1; i++) {
-    std::cout << i << " / " << Nm << "Fund: grid pointer " << Fundevec[i]._grid
+    std::cout << i << " / " << Nm << "Fund: grid pointer " << Fundevec[i].Grid()
               << std::endl;
   };
   for (int i = 0; i < 1; i++) {
-    std::cout << i << " / " << Nm << "AS: grid pointer " << ASevec[i]._grid
+    std::cout << i << " / " << Nm << "AS: grid pointer " << ASevec[i].Grid()
               << std::endl;
   };
   
diff --git a/tests/lanczos/Test_compressed_lanczos.cc b/tests/lanczos/Test_compressed_lanczos.cc
index 8bce82bb..3fa3c490 100644
--- a/tests/lanczos/Test_compressed_lanczos.cc
+++ b/tests/lanczos/Test_compressed_lanczos.cc
@@ -36,7 +36,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 template<class Fobj,class CComplex,int nbasis>
 class LocalCoherenceLanczosScidac : public LocalCoherenceLanczos<Fobj,CComplex,nbasis>
@@ -58,7 +57,7 @@ public:
   {
     assert(this->subspace.size()==nbasis);
     emptyUserRecord record;
-    Grid::QCD::ScidacWriter WR(this->_FineGrid->IsBoss());
+    Grid::ScidacWriter WR(this->_FineGrid->IsBoss());
     WR.open(evecs_file);
     for(int k=0;k<nbasis;k++) {
       WR.writeScidacFieldRecord(this->subspace[k],record);
@@ -82,10 +81,10 @@ public:
     
     std::cout << GridLogIRL<< "checkpointFineRestore:  Reading evecs from "<<evecs_file<<std::endl;
     emptyUserRecord record;
-    Grid::QCD::ScidacReader RD ;
+    Grid::ScidacReader RD ;
     RD.open(evecs_file);
     for(int k=0;k<nbasis;k++) {
-      this->subspace[k].checkerboard=this->_checkerboard;
+      this->subspace[k].Checkerboard()=this->_checkerboard;
       RD.readScidacFieldRecord(this->subspace[k],record);
       
     }
@@ -96,7 +95,7 @@ public:
   {
     int n = this->evec_coarse.size();
     emptyUserRecord record;
-    Grid::QCD::ScidacWriter WR(this->_CoarseGrid->IsBoss());
+    Grid::ScidacWriter WR(this->_CoarseGrid->IsBoss());
     WR.open(evecs_file);
     for(int k=0;k<n;k++) {
       WR.writeScidacFieldRecord(this->evec_coarse[k],record);
@@ -119,7 +118,7 @@ public:
     assert(this->evals_coarse.size()==nvec);
     emptyUserRecord record;
     std::cout << GridLogIRL<< "checkpointCoarseRestore:  Reading evecs from "<<evecs_file<<std::endl;
-    Grid::QCD::ScidacReader RD ;
+    Grid::ScidacReader RD ;
     RD.open(evecs_file);
     for(int k=0;k<nvec;k++) {
       RD.readScidacFieldRecord(this->evec_coarse[k],record);
@@ -150,7 +149,7 @@ int main (int argc, char ** argv) {
   int     Ls = (int)Params.omega.size();
   RealD mass = Params.mass;
   RealD M5   = Params.M5;
-  std::vector<int> blockSize = Params.blockSize;
+  auto blockSize = Params.blockSize;
 
   // Grids
   GridCartesian         * UGrid     = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(),
@@ -160,11 +159,10 @@ int main (int argc, char ** argv) {
   GridCartesian         * FGrid     = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
   GridRedBlackCartesian * FrbGrid   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
 
-  std::vector<int> fineLatt     = GridDefaultLatt();
+  auto fineLatt     = GridDefaultLatt();
   int dims=fineLatt.size();
   assert(blockSize.size()==dims+1);
-  std::vector<int> coarseLatt(dims);
-  std::vector<int> coarseLatt5d ;
+  Coordinate coarseLatt(dims);
 
   for (int d=0;d<coarseLatt.size();d++){
     coarseLatt[d] = fineLatt[d]/blockSize[d];    assert(coarseLatt[d]*blockSize[d]==fineLatt[d]);
diff --git a/tests/lanczos/Test_dwf_compressed_lanczos.cc b/tests/lanczos/Test_dwf_compressed_lanczos.cc
deleted file mode 100644
index 45690f05..00000000
--- a/tests/lanczos/Test_dwf_compressed_lanczos.cc
+++ /dev/null
@@ -1,712 +0,0 @@
-/*
-  Authors: Christoph Lehner
-  Date: 2017
-
-  Multigrid Lanczos
-
-
-
-  TODO:
-
-  High priority:
-  - Explore filtering of starting vector again, should really work:  If cheby has 4 for low mode region and 1 for high mode, applying 15 iterations has 1e9 suppression
-    of high modes, which should create the desired invariant subspace already?  Missing something here???  Maybe dynamic range dangerous, i.e., could also kill interesting
-    eigenrange if not careful.
-
-    Better: Use all Cheby up to order N in order to approximate a step function; try this!  Problem: width of step function.  Can kill eigenspace > 1e-3 and have < 1e-5 equal
-            to 1
-
-  Low priority:
-  - Given that I seem to need many restarts and high degree poly to create the base and this takes about 1 day, seriously consider a simple method to create a basis
-    (ortho krylov low poly); and then fix up lowest say 200 eigenvalues by 1 run with high-degree poly (600 could be enough)
-*/
-#include <Grid/Grid.h>
-#include <Grid/algorithms/iterative/ImplicitlyRestartedLanczos.h>
-/////////////////////////////////////////////////////////////////////////////
-// The following are now decoupled from the Lanczos and deal with grids.
-// Safe to replace functionality
-/////////////////////////////////////////////////////////////////////////////
-#include "BlockedGrid.h"
-#include "FieldBasisVector.h"
-#include "BlockProjector.h"
-#include "FieldVectorIO.h"
-#include "Params.h"
-
-using namespace std;
-using namespace Grid;
-using namespace Grid::QCD;
-
-bool read_evals(GridBase* _grid, char* fn, std::vector<RealD>& evals) {
-
-  FILE* f = 0;
-  uint32_t status = 0;
-  if (_grid->IsBoss()) {
-    f = fopen(fn,"rt");
-    status = f ? 1 : 0;
-  }
-  _grid->GlobalSum(status);
-
-  if (!status)
-    return false;
-
-  uint32_t N;
-  if (f)
-    assert(fscanf(f,"%d\n",&N)==1);
-  else
-    N = 0;
-  _grid->GlobalSum(N);
-
-  std::cout << "Reading " << N << " eigenvalues" << std::endl;
-
-  evals.resize(N);
-
-  for (int i=0;i<N;i++) {
-    if (f)
-      assert(fscanf(f,"%lf",&evals[i])==1);
-    else
-      evals[i] = 0;
-  }
-
-  _grid->GlobalSumVector(&evals[0],evals.size());
-
-  if (f)
-    fclose(f);
-  return true;
-}
-
-void write_evals(char* fn, std::vector<RealD>& evals) {
-  FILE* f = fopen(fn,"wt");
-  assert(f);
-
-  int N = (int)evals.size();
-  fprintf(f,"%d\n",N);
-
-  for (int i=0;i<N;i++) {
-    fprintf(f,"%.15E\n",evals[i]);
-  }
-
-  fclose(f);
-}
-
-void write_history(char* fn, std::vector<RealD>& hist) {
-  FILE* f = fopen(fn,"wt");
-  assert(f);
-
-  int N = (int)hist.size();
-  for (int i=0;i<N;i++) {
-    fprintf(f,"%d %.15E\n",i,hist[i]);
-  }
-
-  fclose(f);
-}
-
-
-template<typename Field>
-class CheckpointedLinearFunction : public LinearFunction<Field> {
-public:
-  LinearFunction<Field>& _op;
-  std::string _dir;
-  int _max_apply;
-  int _apply, _apply_actual;
-  GridBase* _grid;
-  FILE* _f;
-
-  CheckpointedLinearFunction(GridBase* grid, LinearFunction<Field>& op, const char* dir,int max_apply) : _op(op), _dir(dir), _grid(grid), _f(0),
-													 _max_apply(max_apply), _apply(0), _apply_actual(0) {
-
-    FieldVectorIO::conditionalMkDir(dir);
-
-    char fn[4096];
-    sprintf(fn,"%s/ckpt_op.%4.4d",_dir.c_str(),_grid->ThisRank());
-    printf("CheckpointLinearFunction:: file %s\n",fn);
-    _f = fopen(fn,"r+b");
-    if (!_f)
-      _f = fopen(fn,"w+b");
-    assert(_f);
-    fseek(_f,0,SEEK_CUR);
-
-  }
-
-  ~CheckpointedLinearFunction() {
-    if (_f) {
-      fclose(_f);
-      _f = 0;
-    }
-  }
-
-  bool load_ckpt(const Field& in, Field& out) {
-
-    off_t cur = ftello(_f);
-    fseeko(_f,0,SEEK_END);
-    if (cur == ftello(_f))
-      return false;
-    fseeko(_f,cur,SEEK_SET);
-
-    size_t sz = sizeof(out._odata[0]) * out._odata.size();
-
-    GridStopWatch gsw;
-    gsw.Start();
-    uint32_t crc_exp;
-    assert(fread(&crc_exp,4,1,_f)==1);
-    assert(fread(&out._odata[0],sz,1,_f)==1);
-    assert(FieldVectorIO::crc32_threaded((unsigned char*)&out._odata[0],sz,0x0)==crc_exp);
-    gsw.Stop();
-
-    printf("CheckpointLinearFunction:: reading %lld\n",(long long)sz);
-    std::cout << GridLogMessage << "Loading " << ((RealD)sz/1024./1024./1024.) << " GB in " << gsw.Elapsed() << std::endl;
-    return true;
-  }
-
-  void save_ckpt(const Field& in, Field& out) {
-
-    fseek(_f,0,SEEK_CUR); // switch to write
-
-    size_t sz = sizeof(out._odata[0]) * out._odata.size();
-
-    GridStopWatch gsw;
-    gsw.Start();
-    uint32_t crc = FieldVectorIO::crc32_threaded((unsigned char*)&out._odata[0],sz,0x0);
-    assert(fwrite(&crc,4,1,_f)==1);
-    assert(fwrite(&out._odata[0],sz,1,_f)==1);
-    fflush(_f); // try this on the GPFS to suppress OPA usage for disk during dslash; this is not needed at Lustre/JLAB
-    gsw.Stop();
-
-    printf("CheckpointLinearFunction:: writing %lld\n",(long long)sz);
-    std::cout << GridLogMessage << "Saving " << ((RealD)sz/1024./1024./1024.) << " GB in " << gsw.Elapsed() << std::endl;
-  }
-
-  void operator()(const Field& in, Field& out) {
-
-    _apply++;
-
-    if (load_ckpt(in,out))
-      return;
-
-    _op(in,out);
-    
-    save_ckpt(in,out);
-
-    if (_apply_actual++ >= _max_apply) {
-      std::cout << GridLogMessage << "Maximum application of operator reached, checkpoint and finish in future job" << std::endl;
-      if (_f) { fclose(_f); _f=0; }
-      in._grid->Barrier();
-      Grid_finalize();
-      exit(3);
-    }
-  }
-};
-
-template<typename CoarseField,typename Field>
-class ProjectedFunctionHermOp : public LinearFunction<CoarseField> {
-public:
-  OperatorFunction<Field>   & _poly;
-  LinearOperatorBase<Field> &_Linop;
-  BlockProjector<Field>& _pr;
-
-  ProjectedFunctionHermOp(BlockProjector<Field>& pr,OperatorFunction<Field> & poly,LinearOperatorBase<Field>& linop) : _poly(poly), _Linop(linop), _pr(pr) {
-  }
-
-  void operator()(const CoarseField& in, CoarseField& out) {
-    assert(_pr._bgrid._o_blocks == in._grid->oSites());
-
-    Field fin(_pr._bgrid._grid);
-    Field fout(_pr._bgrid._grid);
-
-    GridStopWatch gsw1,gsw2,gsw3;
-    // fill fin
-    gsw1.Start();
-    _pr.coarseToFine(in,fin);
-    gsw1.Stop();
-
-    // apply poly
-    gsw2.Start();
-    _poly(_Linop,fin,fout);
-    gsw2.Stop();
-
-    // fill out
-    gsw3.Start();
-    _pr.fineToCoarse(fout,out);
-    gsw3.Stop();
-
-    auto eps = innerProduct(in,out);
-    std::cout << GridLogMessage << "Operator timing details: c2f = " << gsw1.Elapsed() << " poly = " << gsw2.Elapsed() << " f2c = " << gsw3.Elapsed() << 
-      "   Complimentary Hermiticity check: " << eps.imag() / std::abs(eps) << std::endl;
-
-  }
-};
-
-template<typename CoarseField,typename Field>
-class ProjectedHermOp : public LinearFunction<CoarseField> {
-public:
-  LinearOperatorBase<Field> &_Linop;
-  BlockProjector<Field>& _pr;
-
-  ProjectedHermOp(BlockProjector<Field>& pr,LinearOperatorBase<Field>& linop) : _Linop(linop), _pr(pr) {
-  }
-
-  void operator()(const CoarseField& in, CoarseField& out) {
-    assert(_pr._bgrid._o_blocks == in._grid->oSites());
-    Field fin(_pr._bgrid._grid);
-    Field fout(_pr._bgrid._grid);
-    _pr.coarseToFine(in,fin);
-    _Linop.HermOp(fin,fout);
-    _pr.fineToCoarse(fout,out);
-
-  }
-};
-
-template<typename vtype, int N > using CoarseSiteFieldGeneral = iScalar< iVector<vtype, N> >;
-template<int N> using CoarseSiteFieldD = CoarseSiteFieldGeneral< vComplexD, N >;
-template<int N> using CoarseSiteFieldF = CoarseSiteFieldGeneral< vComplexF, N >;
-template<int N> using CoarseSiteField  = CoarseSiteFieldGeneral< vComplex,  N >;
-template<int N> using CoarseLatticeFermion  = Lattice< CoarseSiteField<N> >;
-template<int N> using CoarseLatticeFermionD = Lattice< CoarseSiteFieldD<N> >;
-
-template<typename Field,int Nstop1>
-void CoarseGridLanczos(BlockProjector<Field>& pr,RealD alpha2,RealD beta,int Npoly2,
-		       int Nstop2,int Nk2,int Nm2,RealD resid2,RealD betastp2,int MaxIt,int MinRes2,
-		       LinearOperatorBase<Field>& HermOp, std::vector<RealD>& eval1, bool cg_test_enabled, 
-		       int cg_test_maxiter,int nsingle,int SkipTest2, int MaxApply2,bool smoothed_eval_enabled,
-		       int smoothed_eval_inner,int smoothed_eval_outer,int smoothed_eval_begin,
-		       int smoothed_eval_end,RealD smoothed_eval_inner_resid) {
-
-  BlockedGrid<Field>& bgrid = pr._bgrid;
-  BasisFieldVector<Field>& basis = pr._evec;
-
-
-  std::vector<int> coarseFourDimLatt;
-  for (int i=0;i<4;i++)
-    coarseFourDimLatt.push_back(bgrid._nb[1+i] * bgrid._grid->_processors[1+i]);
-  assert(bgrid._grid->_processors[0] == 1);
-
-  std::cout << GridLogMessage << "CoarseGrid = " << coarseFourDimLatt << " with basis = " << Nstop1 << std::endl;
-  GridCartesian         * UCoarseGrid   = SpaceTimeGrid::makeFourDimGrid(coarseFourDimLatt, GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
-  GridCartesian         * FCoarseGrid   = SpaceTimeGrid::makeFiveDimGrid(bgrid._nb[0],UCoarseGrid);
-
-  Chebyshev<Field> Cheb2(alpha2,beta,Npoly2);
-  CoarseLatticeFermion<Nstop1> src_coarse(FCoarseGrid);
-
-  // Second round of Lanczos in blocked space
-  std::vector<RealD>         eval2(Nm2);
-  std::vector<RealD>         eval3(Nm2);
-  BasisFieldVector<CoarseLatticeFermion<Nstop1> > coef(Nm2,FCoarseGrid);
-
-  ProjectedFunctionHermOp<CoarseLatticeFermion<Nstop1>,LatticeFermion> Op2plain(pr,Cheb2,HermOp);
-  CheckpointedLinearFunction<CoarseLatticeFermion<Nstop1> > Op2ckpt(src_coarse._grid,Op2plain,"checkpoint",MaxApply2);
-  LinearFunction< CoarseLatticeFermion<Nstop1> >* Op2;
-  if (MaxApply2) {
-    Op2 = &Op2ckpt;
-  } else {
-    Op2 = &Op2plain;
-  }
-  ProjectedHermOp<CoarseLatticeFermion<Nstop1>,LatticeFermion> Op2nopoly(pr,HermOp);
-  ImplicitlyRestartedLanczos<CoarseLatticeFermion<Nstop1> > IRL2(*Op2,*Op2,Nstop2,Nk2,Nm2,resid2,MaxIt,betastp2,MinRes2);
-
-
-  src_coarse = 1.0;
-  
-  // Precision test
-  {
-    Field tmp(bgrid._grid);
-    CoarseLatticeFermion<Nstop1> tmp2(FCoarseGrid);
-    CoarseLatticeFermion<Nstop1> tmp3(FCoarseGrid);
-    tmp2 = 1.0;
-    tmp3 = 1.0;
-
-    pr.coarseToFine(tmp2,tmp);
-    pr.fineToCoarse(tmp,tmp2);
-
-    tmp2 -= tmp3;
-    std::cout << GridLogMessage << "Precision Test c->f->c: " << norm2(tmp2) / norm2(tmp3) << std::endl;
-
-    //bgrid._grid->Barrier();
-    //return;
-  }
-
-  int Nconv;
-  if (!FieldVectorIO::read_compressed_vectors("lanczos.output",pr,coef) ||
-      !read_evals(UCoarseGrid,(char *)"lanczos.output/eigen-values.txt",eval3) ||
-      !read_evals(UCoarseGrid,(char *)"lanczos.output/eigen-values.txt.linear",eval1) ||
-      !read_evals(UCoarseGrid,(char *)"lanczos.output/eigen-values.txt.poly",eval2)
-      ) {
-    
-
-    IRL2.calc(eval2,coef._v,src_coarse,Nconv,true);
-
-    coef.resize(Nstop2);
-    eval2.resize(Nstop2);
-    eval3.resize(Nstop2);
-
-    std::vector<Field> step3_cache;
-
-    // reconstruct eigenvalues of original operator
-    for (int i=0;i<Nstop2;i++){
-      RealD eval2_linear;
-
-      if (i<Nstop1) {
-	eval2_linear = eval1[i];
-      } else {
-	eval2_linear = eval2[i-1];
-      }
-
-      RealD eval2_poly = eval2[i];
-      RealD eval_reconstruct = Cheb2.approxInv(eval2_poly,eval2_linear,100,1e-10);
-      std::cout << i << " Reconstructed eval = " << eval_reconstruct << " from quess " << eval2_linear << std::endl;
-      eval2[i] = eval_reconstruct;
-    }
-    
-    // as demonstrated in CG test below, best result from mixed determination
-    for (int i=0;i<Nstop2;i++)
-      eval3[i] = (i < Nstop1) ? eval1[i] : eval2[i];
-    
-    for(int i=0;i<Nstop2;i++){
-      std::cout << i<<" / "<< Nstop2<< " eigenvalue "<< eval3[i] <<std::endl;
-    };
-    
-    // write
-    mkdir("lanczos.output",ACCESSPERMS);
-    FieldVectorIO::write_compressed_vectors("lanczos.output",pr,coef,nsingle);
-    if (bgrid._grid->IsBoss()) {
-      write_evals((char *)"lanczos.output/eigen-values.txt",eval3);
-      write_evals((char *)"lanczos.output/eigen-values.txt.linear",eval1);
-      write_evals((char *)"lanczos.output/eigen-values.txt.poly",eval2);
-    }
-
-  }
-
-  // fix up eigenvalues
-  if (!read_evals(UCoarseGrid,(char *)"lanczos.output/eigen-values.txt.smoothed",eval3) && smoothed_eval_enabled) {
-
-    ConjugateGradient<LatticeFermion> CG(smoothed_eval_inner_resid, smoothed_eval_inner, false);
-
-    LatticeFermion v_i(basis[0]._grid);
-    auto tmp = v_i;
-    auto tmp2 = v_i;
-
-    for (int i=smoothed_eval_begin;i<smoothed_eval_end;i++) {
-
-      GridStopWatch gsw;
-
-      gsw.Start();
-
-      pr.coarseToFine(coef[i],v_i);
-      v_i.checkerboard = Odd;
-      
-      for (int j=0;j<smoothed_eval_outer;j++) {
-	tmp=zero;
-	//pr.deflate(coef,eval3,Nstop2,v_i,tmp);
-	CG(HermOp, v_i, tmp);
-
-	v_i = 1.0 / ::sqrt( norm2(tmp) ) * tmp;
-      }
-
-      tmp = v_i;
-
-      HermOp.HermOp(tmp,tmp2);
-
-      RealD ev = innerProduct(tmp,tmp2).real();
-
-      gsw.Stop();
-
-      std::cout << GridLogMessage << "Smoothed eigenvalue " << i << " from " << eval3[i] << " to " << ev << " in " << gsw.Elapsed() << std::endl;
-      //	" with effective smoother precision " << (CG.ResHistory.back() / CG.ResHistory.front() ) << std::endl;
-      //      CG.ResHistory.clear();
-
-      eval3[i] = ev;
-    }
-
-    if (bgrid._grid->IsBoss()) {
-      write_evals((char *)"lanczos.output/eigen-values.txt.smoothed",eval3);
-      write_evals((char *)"lanczos.output/eigen-values.txt",eval3); // also reset this to the best ones we have available
-    }
-  }
-
-  // do CG test with and without deflation
-  if (cg_test_enabled) {
-    ConjugateGradient<LatticeFermion> CG(1.0e-8, cg_test_maxiter, false);
-    LatticeFermion src_orig(bgrid._grid);
-    src_orig.checkerboard = Odd;
-    src_orig = 1.0;
-    src_orig = src_orig * (1.0 / ::sqrt(norm2(src_orig)) );
-    auto result = src_orig; 
-
-    // undeflated solve
-    std::cout << GridLogMessage << " Undeflated solve "<<std::endl;
-    result = zero;
-    CG(HermOp, src_orig, result);
-    //    if (UCoarseGrid->IsBoss())
-    //      write_history("cg_test.undefl",CG.ResHistory);
-    //    CG.ResHistory.clear();
-
-    // deflated solve with all eigenvectors
-    std::cout << GridLogMessage << " Deflated solve with all evectors"<<std::endl;
-    result = zero;
-    pr.deflate(coef,eval2,Nstop2,src_orig,result);
-    CG(HermOp, src_orig, result);
-    //    if (UCoarseGrid->IsBoss())
-    //      write_history("cg_test.defl_all",CG.ResHistory);
-    //    CG.ResHistory.clear();
-
-    // deflated solve with non-blocked eigenvectors
-    std::cout << GridLogMessage << " Deflated solve with non-blocked evectors"<<std::endl;
-    result = zero;
-    pr.deflate(coef,eval1,Nstop1,src_orig,result);
-    CG(HermOp, src_orig, result);
-    //    if (UCoarseGrid->IsBoss())
-    //      write_history("cg_test.defl_full",CG.ResHistory);
-    //    CG.ResHistory.clear();
-
-    // deflated solve with all eigenvectors and original eigenvalues from proj
-    std::cout << GridLogMessage << " Deflated solve with all eigenvectors and original eigenvalues from proj"<<std::endl;
-    result = zero;
-    pr.deflate(coef,eval3,Nstop2,src_orig,result);
-    CG(HermOp, src_orig, result);
-    //    if (UCoarseGrid->IsBoss())
-    //      write_history("cg_test.defl_all_ev3",CG.ResHistory);
-    //    CG.ResHistory.clear();
-
-  }
-  
-}
-
-
-template<typename Field>
-void quick_krylov_basis(BasisFieldVector<Field>& evec,Field& src,LinearFunction<Field>& Op,int Nstop) {
-  Field tmp = src;
-  Field tmp2 = tmp;
-
-  for (int i=0;i<Nstop;i++) {
-    GridStopWatch gsw;
-    gsw.Start();
-    Op(tmp,tmp2);
-    gsw.Stop();
-    evec.orthogonalize(tmp2,i);
-
-    RealD nn = norm2(tmp2);
-    nn = Grid::sqrt(nn);
-    tmp2 = tmp2 * (1.0/nn);
-
-    evec[i] = tmp2;
-    tmp = tmp2;
-    std::cout << GridLogMessage << "Quick_krylov_basis: " << i << "/" << Nstop << " timing of operator=" << gsw.Elapsed() << std::endl;
-  }
-
-}
-
-
-
-int main (int argc, char ** argv) {
-
-  Grid_init(&argc,&argv);
-
-  const int MaxIt = 10000;
-
-  int Ls;
-  RealD mass;
-  RealD M5;
-  std::vector < std::complex<double>  > omega;
-  
-  RealD alpha1, alpha2, beta;
-  int Npoly1, Npoly2;
-  int Nstop1, Nstop2;
-  int Nk1, Nk2;
-  int Np1, Np2;
-  int MinRes1, MinRes2;
-  int SkipTest2, MaxApply2;
-  bool checkpoint_basis;
-  bool cg_test_enabled;
-  bool exit_after_basis_calculation;
-  bool simple_krylov_basis;
-  int cg_test_maxiter;
-  int nsingle; // store in single precision, the rest in FP16
-  int max_cheb_time_ms;
-  bool smoothed_eval_enabled;
-  int smoothed_eval_inner;
-  int smoothed_eval_outer;
-  int smoothed_eval_begin;
-  int smoothed_eval_end;
-  RealD smoothed_eval_inner_resid;
-
-  // vector representation
-  std::vector<int> block_size; // 5d block size
-
-  RealD resid1, resid2, betastp1, betastp2, basis_norm_threshold;
-
-  std::string config;
-  
-  Params jp("params.txt");
-  PADD(jp,Npoly1); PADD(jp,Npoly2);
-  PADD(jp,max_cheb_time_ms);
-  PADD(jp,Nstop1); PADD(jp,Nstop2); PADD(jp,MaxApply2);
-  PADD(jp,Nk1); PADD(jp,Nk2); PADD(jp,betastp1); PADD(jp,betastp2);
-  PADD(jp,Np1); PADD(jp,Np2); basis_norm_threshold = 1e-5; //PADD(jp,basis_norm_threshold);
-  PADD(jp,block_size); PADD(jp,smoothed_eval_enabled); PADD(jp,smoothed_eval_inner);
-  PADD(jp,resid1); PADD(jp,resid2); PADD(jp,smoothed_eval_outer);
-  PADD(jp,alpha1); PADD(jp,alpha2); PADD(jp,smoothed_eval_begin);
-  PADD(jp,MinRes1); PADD(jp,MinRes2); PADD(jp,smoothed_eval_end);
-  PADD(jp,beta); PADD(jp,mass); PADD(jp,smoothed_eval_inner_resid);
-  PADD(jp,omega); PADD(jp,config); 
-  PADD(jp,M5); PADD(jp,cg_test_enabled);
-  PADD(jp,cg_test_maxiter); PADD(jp,checkpoint_basis);
-  PADD(jp,nsingle); PADD(jp,exit_after_basis_calculation);
-  PADD(jp,simple_krylov_basis); PADD(jp,SkipTest2);
-
-  Ls = (int)omega.size();
-
-  // Grids
-  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
-  GridCartesian         * UGridHP = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexD::Nsimd()),GridDefaultMpi());
-  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
-  GridRedBlackCartesian * UrbGridHP = SpaceTimeGrid::makeFourDimRedBlackGrid(UGridHP);
-  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
-  GridCartesian         * FGridHP   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGridHP);
-  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
-  GridRedBlackCartesian * FrbGridHP = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGridHP);
-
-  // Gauge field
-  LatticeGaugeField Umu(UGrid);
-  FieldMetaData header;
-  NerscIO::readConfiguration(Umu,header,config);
-  std::cout << GridLogMessage << "Lattice dimensions: " << GridDefaultLatt()
-            << "   Ls: " << Ls << std::endl;
-
-  // ZMobius EO Operator
-  ZMobiusFermionR Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5, omega,1.,0.);
-  SchurDiagTwoOperator<ZMobiusFermionR,LatticeFermion> HermOp(Ddwf);
-
-  // Eigenvector storage
-  const int Nm1 = Np1 + Nk1;
-  const int Nm2 = Np2 + Nk2; // maximum number of vectors we need to keep
-  std::cout << GridLogMessage << "Keep " << Nm1 << " full vectors" << std::endl;
-  std::cout << GridLogMessage << "Keep " << Nm2 << " total vectors" << std::endl;
-  assert(Nm2 >= Nm1);
-  BasisFieldVector<LatticeFermion> evec(Nm1,FrbGrid); // start off with keeping full vectors
-
-  // First and second cheby
-  Chebyshev<LatticeFermion> Cheb1(alpha1,beta,Npoly1);
-  FunctionHermOp<LatticeFermion> Op1(Cheb1,HermOp);
-  PlainHermOp<LatticeFermion> Op1test(HermOp);
-
-  // Eigenvalue storage
-  std::vector<RealD>          eval1(evec.size());
-
-  // Construct source vector
-  LatticeFermion    src(FrbGrid);
-  {
-    src=1.0;
-    src.checkerboard = Odd;
-
-    // normalize
-    RealD nn = norm2(src);
-    nn = Grid::sqrt(nn);
-    src = src * (1.0/nn);
-  }
-
-  // Do a benchmark and a quick exit if performance is too little (ugly but needed due to performance fluctuations)
-  if (max_cheb_time_ms) {
-    // one round of warmup
-    auto tmp = src;
-    GridStopWatch gsw1,gsw2;
-    gsw1.Start();
-    Cheb1(HermOp,src,tmp);
-    gsw1.Stop();
-    Ddwf.ZeroCounters();
-    gsw2.Start();
-    Cheb1(HermOp,src,tmp);
-    gsw2.Stop();
-    Ddwf.Report();
-    std::cout << GridLogMessage << "Performance check; warmup = " << gsw1.Elapsed() << "  test = " << gsw2.Elapsed() << std::endl;
-    int ms = (int)(gsw2.useconds()/1e3);
-    if (ms > max_cheb_time_ms) {
-      std::cout << GridLogMessage << "Performance too poor: " << ms << " ms, cutoff = " << max_cheb_time_ms << " ms" << std::endl;
-      Grid_finalize();
-      return 2;
-    }
-
-  }
-
-  // First round of Lanczos to get low mode basis
-  ImplicitlyRestartedLanczos<LatticeFermion> IRL1(Op1,Op1test,Nstop1,Nk1,Nm1,resid1,MaxIt,betastp1,MinRes1);
-  int Nconv;
-
-  char tag[1024];
-  if (!FieldVectorIO::read_argonne(evec,(char *)"checkpoint") || !read_evals(UGrid,(char *)"checkpoint/eigen-values.txt",eval1)) {
-
-    if (simple_krylov_basis) {
-      quick_krylov_basis(evec,src,Op1,Nstop1);
-    } else {
-      IRL1.calc(eval1,evec._v,src,Nconv,false);
-    }
-    evec.resize(Nstop1); // and throw away superfluous
-    eval1.resize(Nstop1);
-    if (checkpoint_basis)
-      FieldVectorIO::write_argonne(evec,(char *)"checkpoint");
-    if (UGrid->IsBoss() && checkpoint_basis)
-      write_evals((char *)"checkpoint/eigen-values.txt",eval1);
-
-    Ddwf.Report();
-
-    if (exit_after_basis_calculation) {
-      Grid_finalize();
-      return 0;
-    }
-  }
-
-  // now test eigenvectors
-  if (!simple_krylov_basis) {
-    for (int i=0;i<Nstop1;i++){
-      auto B = evec[i];
-      auto tmp = B;
-      auto v = B;
-      
-      {
-	HermOp.HermOp(B,v);
-	
-	RealD vnum = real(innerProduct(B,v)); // HermOp.
-	RealD vden = norm2(B);
-	RealD vv0 = norm2(v);
-	RealD eval2 = vnum/vden;
-	v -= eval2*B;
-	RealD vv = norm2(v);
-	
-	std::cout << i << " OP eval = " << eval2 << " (" << eval1[i] << ") "
-		  << "res2 = " << vv << " norm2 = " << norm2(B) << std::endl;
-      }
-    }
-  }
-
-  // do second step only if needed
-  if (Nstop1 <= Nstop2) {
-    
-    // Now setup blocking
-    assert(evec.size() == Nstop1);
-    BlockedGrid<LatticeFermion> bgrid(FrbGrid, block_size);
-    BlockProjector<LatticeFermion> pr(evec,bgrid);
-    pr.createOrthonormalBasis(basis_norm_threshold);
-    pr.createOrthonormalBasis(basis_norm_threshold); // another round due to precision issues created by local coherence
-
-    constexpr int common_basis_sizes[] = { 60, 250, 400 };
-    constexpr int n_common_basis_sizes = sizeof(common_basis_sizes) / sizeof(common_basis_sizes[0]);
-    switch (Nstop1) {
-#define BASIS(n) case common_basis_sizes[n]:\
-      CoarseGridLanczos<LatticeFermion,common_basis_sizes[n]>\
-	(pr,alpha2,beta,Npoly2,Nstop2,Nk2,Nm2,resid2,betastp2,MaxIt,MinRes2,HermOp,eval1, \
-	 cg_test_enabled,cg_test_maxiter,nsingle,SkipTest2, \
-	 MaxApply2,smoothed_eval_enabled,smoothed_eval_inner,smoothed_eval_outer, \
-	 smoothed_eval_begin,smoothed_eval_end,smoothed_eval_inner_resid); break;
-      BASIS(0);
-      BASIS(1);
-      BASIS(2);
-    default:
-      std::cout << GridLogMessage << "Basis size " << Nstop1 << " must be added at compile-time" << std::endl;
-      std::cout << GridLogMessage << "Currently available sizes: " << std::endl;
-      for (int i=0;i<n_common_basis_sizes;i++) {
-	std::cout << GridLogMessage << "  " << common_basis_sizes[i] << std::endl;
-      }
-    }
-
-  }
-    
-  Grid_finalize();
-}
-
diff --git a/tests/lanczos/Test_dwf_compressed_lanczos_reorg.cc b/tests/lanczos/Test_dwf_compressed_lanczos_reorg.cc
index 8bce82bb..ecd0e629 100644
--- a/tests/lanczos/Test_dwf_compressed_lanczos_reorg.cc
+++ b/tests/lanczos/Test_dwf_compressed_lanczos_reorg.cc
@@ -36,7 +36,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class Fobj,class CComplex,int nbasis>
 class LocalCoherenceLanczosScidac : public LocalCoherenceLanczos<Fobj,CComplex,nbasis>
@@ -58,7 +58,7 @@ public:
   {
     assert(this->subspace.size()==nbasis);
     emptyUserRecord record;
-    Grid::QCD::ScidacWriter WR(this->_FineGrid->IsBoss());
+    Grid::ScidacWriter WR(this->_FineGrid->IsBoss());
     WR.open(evecs_file);
     for(int k=0;k<nbasis;k++) {
       WR.writeScidacFieldRecord(this->subspace[k],record);
@@ -82,12 +82,11 @@ public:
     
     std::cout << GridLogIRL<< "checkpointFineRestore:  Reading evecs from "<<evecs_file<<std::endl;
     emptyUserRecord record;
-    Grid::QCD::ScidacReader RD ;
+    Grid::ScidacReader RD ;
     RD.open(evecs_file);
     for(int k=0;k<nbasis;k++) {
-      this->subspace[k].checkerboard=this->_checkerboard;
+      this->subspace[k].Checkerboard()=this->_checkerboard;
       RD.readScidacFieldRecord(this->subspace[k],record);
-      
     }
     RD.close();
   }
@@ -96,7 +95,7 @@ public:
   {
     int n = this->evec_coarse.size();
     emptyUserRecord record;
-    Grid::QCD::ScidacWriter WR(this->_CoarseGrid->IsBoss());
+    Grid::ScidacWriter WR(this->_CoarseGrid->IsBoss());
     WR.open(evecs_file);
     for(int k=0;k<n;k++) {
       WR.writeScidacFieldRecord(this->evec_coarse[k],record);
@@ -119,7 +118,7 @@ public:
     assert(this->evals_coarse.size()==nvec);
     emptyUserRecord record;
     std::cout << GridLogIRL<< "checkpointCoarseRestore:  Reading evecs from "<<evecs_file<<std::endl;
-    Grid::QCD::ScidacReader RD ;
+    Grid::ScidacReader RD ;
     RD.open(evecs_file);
     for(int k=0;k<nvec;k++) {
       RD.readScidacFieldRecord(this->evec_coarse[k],record);
@@ -160,11 +159,11 @@ int main (int argc, char ** argv) {
   GridCartesian         * FGrid     = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
   GridRedBlackCartesian * FrbGrid   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
 
-  std::vector<int> fineLatt     = GridDefaultLatt();
+  Coordinate fineLatt     = GridDefaultLatt();
   int dims=fineLatt.size();
   assert(blockSize.size()==dims+1);
-  std::vector<int> coarseLatt(dims);
-  std::vector<int> coarseLatt5d ;
+  Coordinate coarseLatt(dims);
+  Coordinate coarseLatt5d ;
 
   for (int d=0;d<coarseLatt.size();d++){
     coarseLatt[d] = fineLatt[d]/blockSize[d];    assert(coarseLatt[d]*blockSize[d]==fineLatt[d]);
diff --git a/tests/lanczos/Test_dwf_compressed_lanczos_reorg_synthetic.cc b/tests/lanczos/Test_dwf_compressed_lanczos_reorg_synthetic.cc
index 132dff4e..0b1c3a1f 100644
--- a/tests/lanczos/Test_dwf_compressed_lanczos_reorg_synthetic.cc
+++ b/tests/lanczos/Test_dwf_compressed_lanczos_reorg_synthetic.cc
@@ -35,7 +35,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class Fobj,class CComplex,int nbasis>
 class ProjectedHermOp : public LinearFunction<Lattice<iVector<CComplex,nbasis > > > {
@@ -87,8 +87,8 @@ public:
 
     GridBase *FineGrid = _Aggregate.FineGrid;
 
-    FineField fin(FineGrid) ;fin.checkerboard  =_Aggregate.checkerboard;
-    FineField fout(FineGrid);fout.checkerboard =_Aggregate.checkerboard;
+    FineField fin(FineGrid) ;fin.Checkerboard()  =_Aggregate.Checkerboard();
+    FineField fout(FineGrid);fout.Checkerboard() =_Aggregate.Checkerboard();
 
     _Aggregate.PromoteFromSubspace(in,fin);
     _poly(_Linop,fin,fout);
@@ -139,7 +139,7 @@ public:
     int Nk = nbasis;
     _Aggregate.subspace.resize(Nk,_FineGrid);
     _Aggregate.subspace[0]=1.0;
-    _Aggregate.subspace[0].checkerboard=_checkerboard;
+    _Aggregate.subspace[0].Checkerboard()=_checkerboard;
     normalise(_Aggregate.subspace[0]);
     PlainHermOp<FineField>    Op(_FineOp);
     for(int k=1;k<Nk;k++){
@@ -148,7 +148,7 @@ public:
       std::cout << GridLogMessage << "testFine subspace "<<k<<" " <<norm2(_Aggregate.subspace[k])<<std::endl;
     }
     for(int k=0;k<Nk;k++){
-      std::cout << GridLogMessage << "testFine subspace "<<k<<"  cb " <<_Aggregate.subspace[k].checkerboard<<std::endl;
+      std::cout << GridLogMessage << "testFine subspace "<<k<<"  cb " <<_Aggregate.subspace[k].Checkerboard()<<std::endl;
     }
     _Aggregate.Orthogonalise();
   }
@@ -165,14 +165,14 @@ public:
 
     std::vector<RealD>          eval(Nm);
 
-    FineField src(_FineGrid); src=1.0; src.checkerboard = _checkerboard;
+    FineField src(_FineGrid); src=1.0; src.Checkerboard() = _checkerboard;
 
     ImplicitlyRestartedLanczos<FineField> IRL(ChebyOp,Op,Nk,Nk,Nm,resid,MaxIt,betastp,MinRes);
     _Aggregate.subspace.resize(Nm,_FineGrid);
     IRL.calc(eval,_Aggregate.subspace,src,Nk,false);
     _Aggregate.subspace.resize(Nk,_FineGrid);
     for(int k=0;k<Nk;k++){
-      std::cout << GridLogMessage << "testFine subspace "<<k<<"  cb " <<_Aggregate.subspace[k].checkerboard<<std::endl;
+      std::cout << GridLogMessage << "testFine subspace "<<k<<"  cb " <<_Aggregate.subspace[k].Checkerboard()<<std::endl;
     }
     _Aggregate.Orthogonalise();
   }
@@ -229,7 +229,7 @@ struct CompressedLanczosParams : Serializable {
 				  LanczosParams, CoarseParams,
 				  std::vector<int>, blockSize,
 				  std::string, config,
-				  std::vector < std::complex<double>  >, omega,
+				  std::vector < ComplexD  >, omega,
 				  RealD, mass,
 				  RealD, M5
 				  );
@@ -264,11 +264,11 @@ int main (int argc, char ** argv) {
   GridCartesian         * FGrid     = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
   GridRedBlackCartesian * FrbGrid   = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
 
-  std::vector<int> fineLatt     = GridDefaultLatt();
+  Coordinate fineLatt     = GridDefaultLatt();
   int dims=fineLatt.size();
   assert(blockSize.size()==dims+1);
-  std::vector<int> coarseLatt(dims);
-  std::vector<int> coarseLatt5d ;
+  Coordinate coarseLatt(dims);
+  Coordinate coarseLatt5d ;
 
   for (int d=0;d<coarseLatt.size();d++){
     coarseLatt[d] = fineLatt[d]/blockSize[d];    assert(coarseLatt[d]*blockSize[d]==fineLatt[d]);
diff --git a/tests/lanczos/Test_dwf_lanczos.cc b/tests/lanczos/Test_dwf_lanczos.cc
index b1e205cf..12283921 100644
--- a/tests/lanczos/Test_dwf_lanczos.cc
+++ b/tests/lanczos/Test_dwf_lanczos.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 typedef typename GparityDomainWallFermionR::FermionField FermionField;
 
@@ -97,7 +97,7 @@ int main (int argc, char ** argv)
   gaussian(RNG5rb,src);
   std::vector<FermionField> evec(Nm,FrbGrid);
   for(int i=0;i<1;i++){
-    std::cout << GridLogMessage <<i<<" / "<< Nm<< " grid pointer "<<evec[i]._grid<<std::endl;
+    std::cout << GridLogMessage <<i<<" / "<< Nm<< " grid pointer "<<evec[i].Grid()<<std::endl;
   };
 
   int Nconv;
diff --git a/tests/lanczos/Test_synthetic_lanczos.cc b/tests/lanczos/Test_synthetic_lanczos.cc
index 4be9ca31..7a3591dd 100644
--- a/tests/lanczos/Test_synthetic_lanczos.cc
+++ b/tests/lanczos/Test_synthetic_lanczos.cc
@@ -31,8 +31,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
+/*
 static int
 FEenableexcept (unsigned int excepts)
 {
@@ -53,6 +53,7 @@ FEenableexcept (unsigned int excepts)
   return 0;
 #endif
 }
+*/
 
 
 template<class Field> class DumbOperator  : public LinearOperatorBase<Field> {
@@ -102,7 +103,6 @@ public:
 int main (int argc, char ** argv)
 {
 
-  //  FEenableexcept(FE_ALL_EXCEPT & ~FE_INEXACT); 
 
   Grid_init(&argc,&argv);
 
@@ -117,7 +117,6 @@ int main (int argc, char ** argv)
 
   RealD alpha = 1.2;
   RealD beta  = 0.1;
-  RealD mu    = 0.0;
   int order = 11;
   Chebyshev<LatticeComplex> Cheby(alpha,beta,order);
   std::ofstream file("cheby.dat");
diff --git a/tests/lanczos/Test_wilson_lanczos.cc b/tests/lanczos/Test_wilson_lanczos.cc
index eabc86d7..01258ac7 100644
--- a/tests/lanczos/Test_wilson_lanczos.cc
+++ b/tests/lanczos/Test_wilson_lanczos.cc
@@ -30,7 +30,7 @@ directory
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 typedef WilsonFermionR FermionOp;
 typedef typename WilsonFermionR::FermionField FermionField;
@@ -71,8 +71,6 @@ int main(int argc, char** argv) {
 */
 
   RealD mass = -0.1;
-  RealD M5 = 1.8;
-  RealD mob_b = 1.5;
   FermionOp WilsonOperator(Umu,*FGrid,*FrbGrid,mass);
   MdagMLinearOperator<FermionOp,LatticeFermion> HermOp(WilsonOperator); /// <-----
   //SchurDiagTwoOperator<FermionOp,FermionField> HermOp(WilsonOperator);
@@ -98,7 +96,7 @@ int main(int argc, char** argv) {
   gaussian(RNG5, src);
   std::vector<FermionField> evec(Nm, FGrid);
   for (int i = 0; i < 1; i++) {
-    std::cout << i << " / " << Nm << " grid pointer " << evec[i]._grid
+    std::cout << i << " / " << Nm << " grid pointer " << evec[i].Grid()
               << std::endl;
   };
 
diff --git a/tests/qdpxx/Test_qdpxx_loops_staples.cc b/tests/qdpxx/Test_qdpxx_loops_staples.cc
index db384fcc..53202a4d 100644
--- a/tests/qdpxx/Test_qdpxx_loops_staples.cc
+++ b/tests/qdpxx/Test_qdpxx_loops_staples.cc
@@ -63,7 +63,7 @@ public:
 
     std::vector<int> x(4);
     QDP::multi1d<int> cx(4);
-    std::vector<int> gd= gr._grid->GlobalDimensions();
+    std::vector<int> gd= gr.Grid()->GlobalDimensions();
 
     for (x[0]=0;x[0]<gd[0];x[0]++){
     for (x[1]=0;x[1]<gd[1];x[1]++){
@@ -274,7 +274,7 @@ double calc_grid_p(Grid::QCD::LatticeGaugeField & Umu)
 {
   std::vector<int> seeds4({1,2,3,4});
 
-  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu._grid;
+  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu.Grid();
   Grid::GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
 
   Grid::QCD::SU3::HotConfiguration(RNG4,Umu);
@@ -291,7 +291,7 @@ double calc_grid_p(Grid::QCD::LatticeGaugeField & Umu)
 } 
 double calc_grid_r(Grid::QCD::LatticeGaugeField & Umu)
 {
-  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu._grid;
+  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu.Grid();
 
   Grid::QCD::PlaqPlusRectangleActionR Wilson(0.0,c1); // Just take beta = 0.0
   
@@ -299,7 +299,7 @@ double calc_grid_r(Grid::QCD::LatticeGaugeField & Umu)
 } 
 double calc_grid_IW(Grid::QCD::LatticeGaugeField & Umu)
 {
-  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu._grid;
+  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu.Grid();
 
   Grid::QCD::IwasakiGaugeActionR Iwasaki(beta);
   
@@ -307,7 +307,7 @@ double calc_grid_IW(Grid::QCD::LatticeGaugeField & Umu)
 } 
 double calc_grid_r_dir(Grid::QCD::LatticeGaugeField & Umu)
 {
-  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu._grid;
+  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu.Grid();
 
   std::vector<Grid::QCD::LatticeColourMatrix> U(4,UGrid);
   for(int mu=0;mu<Nd;mu++){
@@ -611,7 +611,7 @@ void check_grid_r_staple(Grid::QCD::LatticeGaugeField & Umu)
 
   std::vector<int> seeds4({1,2,3,4});
 
-  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu._grid;
+  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu.Grid();
 
   Grid::QCD::PlaqPlusRectangleActionR Wilson(0.0,c1); // Just take beta = 0.0
 
@@ -650,7 +650,7 @@ void check_grid_p_staple(Grid::QCD::LatticeGaugeField & Umu)
 
   std::vector<int> seeds4({1,2,3,4});
 
-  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu._grid;
+  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu.Grid();
 
   Grid::QCD::PlaqPlusRectangleActionR Wilson(1.0,0.0); // Just take c1 = 0.0
 
diff --git a/tests/qdpxx/Test_qdpxx_munprec.cc b/tests/qdpxx/Test_qdpxx_munprec.cc
index fd60566d..3c80a287 100644
--- a/tests/qdpxx/Test_qdpxx_munprec.cc
+++ b/tests/qdpxx/Test_qdpxx_munprec.cc
@@ -81,7 +81,7 @@ public:
 
     std::vector<int> x(4);
     QDP::multi1d<int> cx(4);
-    std::vector<int> gd= gr._grid->GlobalDimensions();
+    std::vector<int> gd= gr.Grid()->GlobalDimensions();
 
     for (x[0]=0;x[0]<gd[0];x[0]++){
     for (x[1]=0;x[1]<gd[1];x[1]++){
@@ -124,7 +124,7 @@ public:
 
     std::vector<int> x(5);
     QDP::multi1d<int> cx(4);
-    std::vector<int> gd= gr._grid->GlobalDimensions();
+    std::vector<int> gd= gr.Grid()->GlobalDimensions();
 
     for (x[0]=0;x[0]<gd[0];x[0]++){
     for (x[1]=0;x[1]<gd[1];x[1]++){
@@ -166,7 +166,7 @@ public:
 
     std::vector<int> x(5);
     QDP::multi1d<int> cx(4);
-    std::vector<int> gd= gr._grid->GlobalDimensions();
+    std::vector<int> gd= gr.Grid()->GlobalDimensions();
 
     for (x[0]=0;x[0]<gd[0];x[0]++){
     for (x[1]=0;x[1]<gd[1];x[1]++){
@@ -486,13 +486,13 @@ void calc_chroma(ChromaAction action,Grid::QCD::LatticeGaugeField & lat, Grid::Q
 void calc_grid(ChromaAction action,Grid::QCD::LatticeGaugeField & Umu, Grid::QCD::LatticeFermion &src, Grid::QCD::LatticeFermion &res,int dag)
 {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   std::vector<int> seeds4({1,2,3,4});
   std::vector<int> seeds5({5,6,7,8});
 
-  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu._grid;
-  Grid::GridCartesian         * FGrid   = (Grid::GridCartesian *) src._grid;
+  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu.Grid();
+  Grid::GridCartesian         * FGrid   = (Grid::GridCartesian *) src.Grid();
   Grid::GridRedBlackCartesian * UrbGrid = Grid::QCD::SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
   Grid::GridRedBlackCartesian * FrbGrid = Grid::QCD::SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
 
diff --git a/tests/qdpxx/Test_qdpxx_stag.cc b/tests/qdpxx/Test_qdpxx_stag.cc
index a7563924..24ac3c2a 100644
--- a/tests/qdpxx/Test_qdpxx_stag.cc
+++ b/tests/qdpxx/Test_qdpxx_stag.cc
@@ -60,7 +60,7 @@ public:
 
     std::vector<int> x(4);
     QDP::multi1d<int> cx(4);
-    std::vector<int> gd= gr._grid->GlobalDimensions();
+    std::vector<int> gd= gr.Grid()->GlobalDimensions();
 
     for (x[0]=0;x[0]<gd[0];x[0]++){
     for (x[1]=0;x[1]<gd[1];x[1]++){
@@ -95,7 +95,7 @@ public:
 
     std::vector<int> x(4);
     QDP::multi1d<int> cx(4);
-    std::vector<int> gd= gr._grid->GlobalDimensions();
+    std::vector<int> gd= gr.Grid()->GlobalDimensions();
 
     for (x[0]=0;x[0]<gd[0];x[0]++){
     for (x[1]=0;x[1]<gd[1];x[1]++){
@@ -130,7 +130,7 @@ public:
 
     std::vector<int> x(5);
     QDP::multi1d<int> cx(4);
-    std::vector<int> gd= gr._grid->GlobalDimensions();
+    std::vector<int> gd= gr.Grid()->GlobalDimensions();
 
     for (x[0]=0;x[0]<gd[0];x[0]++){
     for (x[1]=0;x[1]<gd[1];x[1]++){
@@ -162,7 +162,7 @@ public:
 
     std::vector<int> x(5);
     QDP::multi1d<int> cx(4);
-    std::vector<int> gd= gr._grid->GlobalDimensions();
+    std::vector<int> gd= gr.Grid()->GlobalDimensions();
 
     for (x[0]=0;x[0]<gd[0];x[0]++){
     for (x[1]=0;x[1]<gd[1];x[1]++){
@@ -265,12 +265,12 @@ int main (int argc,char **argv )
       std::cout << "Norm of Grid Asqtad multiply "<<Grid::norm2(res_grid)<<std::endl;
       
       /*
-      std::cout << " site 0 of Uthin  "<<uthin._odata[0] <<std::endl;
-      std::cout << " site 0 of Utriple"<<utriple._odata[0] <<std::endl;
-      std::cout << " site 0 of Ufat   "<<ufat._odata[0] <<std::endl;
+      std::cout << " site 0 of Uthin  "<<uthin[0] <<std::endl;
+      std::cout << " site 0 of Utriple"<<utriple[0] <<std::endl;
+      std::cout << " site 0 of Ufat   "<<ufat[0] <<std::endl;
 
-      std::cout << " site 0 of Grid   "<<res_grid._odata[0] <<std::endl;
-      std::cout << " site 0 of Chroma "<<res_chroma._odata[0] <<std::endl;
+      std::cout << " site 0 of Grid   "<<res_grid[0] <<std::endl;
+      std::cout << " site 0 of Chroma "<<res_chroma[0] <<std::endl;
       */
 
       res_chroma=res_chroma - res_grid;
@@ -327,11 +327,11 @@ void calc_chroma(GaugeField & lat, GaugeField &uthin, GaugeField &ufat, FermionF
 void make_gauge(GaugeField & Umu,FermionField &src)
 {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   std::vector<int> seeds4({1,2,3,4});
 
-  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu._grid;
+  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Umu.Grid();
   Grid::GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
   Grid::QCD::SU3::HotConfiguration(RNG4,Umu);
   Grid::gaussian(RNG4,src);
@@ -340,9 +340,9 @@ void make_gauge(GaugeField & Umu,FermionField &src)
 void calc_grid(GaugeField & Uthin, GaugeField & Utriple, GaugeField & Ufat, FermionField &src, FermionField &res,int dag)
 {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
-  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Uthin._grid;
+  Grid::GridCartesian         * UGrid   = (Grid::GridCartesian *) Uthin.Grid();
   Grid::GridRedBlackCartesian * UrbGrid = Grid::QCD::SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
 
   Grid::QCD::ImprovedStaggeredFermionR Dstag(Uthin,Utriple,Ufat,*UGrid,*UrbGrid,mq*2.0);
diff --git a/tests/smearing/Test_WilsonFlow.cc b/tests/smearing/Test_WilsonFlow.cc
index 9d706af0..f339959a 100644
--- a/tests/smearing/Test_WilsonFlow.cc
+++ b/tests/smearing/Test_WilsonFlow.cc
@@ -62,14 +62,14 @@ namespace Grid{
 
 int main(int argc, char **argv) {
   using namespace Grid;
-  using namespace Grid::QCD;
+   ;
 
   Grid_init(&argc, &argv);
   GridLogLayout();
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd, vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  auto latt_size   = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(Nd, vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size, simd_layout, mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -81,8 +81,8 @@ int main(int argc, char **argv) {
   LatticeGaugeField Umu(&Grid), Uflow(&Grid);
   SU<Nc>::HotConfiguration(pRNG, Umu);
   
-  typedef Grid::JSONReader       Serialiser;
-  Serialiser Reader("input.json");
+  typedef Grid::XmlReader       Serialiser;
+  Serialiser Reader("input.xml");
   WFParameters WFPar(Reader);
   ConfParameters CPar(Reader);
   CheckpointerParameters CPPar(CPar.conf_prefix, CPar.rng_prefix);
diff --git a/tests/smearing/Test_smearing.cc b/tests/smearing/Test_smearing.cc
index 996671f0..efc336c7 100644
--- a/tests/smearing/Test_smearing.cc
+++ b/tests/smearing/Test_smearing.cc
@@ -37,10 +37,9 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
 
   std::vector<int> seeds({1,2,3,4});
@@ -48,7 +47,7 @@ int main (int argc, char ** argv)
 
   LatticeFermion src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  LatticeFermion result(&Grid); result=zero;
+  LatticeFermion result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid);
   //  SU3::HotConfiguration(pRNG,Umu);
   SU3::ColdConfiguration(Umu);
@@ -58,10 +57,10 @@ int main (int argc, char ** argv)
     U[mu] = PeekIndex<LorentzIndex>(Umu,mu);
   }
 
-  std::vector<int> site({4,4,0,0});
-  src=zero;
+  Coordinate site({4,4,0,0});
+  src=Zero();
   SpinColourVector scv;
-  scv=zero;
+  scv=Zero();
   scv()(0)(0) = 1.0;
   pokeSite(scv,src,site);
 
diff --git a/tests/solver/Test_cf_cr_unprec.cc b/tests/solver/Test_cf_cr_unprec.cc
index 05f31ae6..8b0ce873 100644
--- a/tests/solver/Test_cf_cr_unprec.cc
+++ b/tests/solver/Test_cf_cr_unprec.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -60,7 +60,7 @@ int main (int argc, char ** argv)
   GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
 
   LatticeFermion    src(FGrid); random(RNG5,src);
-  LatticeFermion result(FGrid); result=zero;
+  LatticeFermion result(FGrid); result=Zero();
   LatticeGaugeField Umu(UGrid); SU3::HotConfiguration(RNG4,Umu);
 
   std::vector<LatticeColourMatrix> U(4,UGrid);
diff --git a/tests/solver/Test_contfrac_cg.cc b/tests/solver/Test_contfrac_cg.cc
index 1d7c121f..36614dfa 100644
--- a/tests/solver/Test_contfrac_cg.cc
+++ b/tests/solver/Test_contfrac_cg.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -146,7 +146,7 @@ void  TestCGunprec(What & Ddwf,
 		   GridParallelRNG *RNG5)
 {
   LatticeFermion src   (FGrid); random(*RNG5,src);
-  LatticeFermion result(FGrid); result=zero;
+  LatticeFermion result(FGrid); result=Zero();
 
   MdagMLinearOperator<What,LatticeFermion> HermOp(Ddwf);
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
@@ -165,7 +165,7 @@ void  TestCGprec(What & Ddwf,
   LatticeFermion    src_o(FrbGrid);
   LatticeFermion result_o(FrbGrid);
   pickCheckerboard(Odd,src_o,src);
-  result_o=zero;
+  result_o=Zero();
 
   SchurDiagMooeeOperator<What,LatticeFermion> HermOpEO(Ddwf);
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
@@ -182,7 +182,7 @@ void  TestCGschur(What & Ddwf,
 		   GridParallelRNG *RNG5)
 {
   LatticeFermion src   (FGrid); random(*RNG5,src);
-  LatticeFermion result(FGrid); result=zero;
+  LatticeFermion result(FGrid); result=Zero();
 
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
   SchurRedBlackDiagMooeeSolve<LatticeFermion> SchurSolver(CG);
diff --git a/tests/solver/Test_dwf_cg_prec.cc b/tests/solver/Test_dwf_cg_prec.cc
index 0ede352e..cb53894f 100644
--- a/tests/solver/Test_dwf_cg_prec.cc
+++ b/tests/solver/Test_dwf_cg_prec.cc
@@ -30,7 +30,7 @@ directory
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template <class d>
 struct scal {
@@ -64,7 +64,7 @@ int main(int argc, char** argv) {
   LatticeFermion src(FGrid);
   random(RNG5, src);
   LatticeFermion result(FGrid);
-  result = zero;
+  result = Zero();
   LatticeGaugeField Umu(UGrid);
 
   SU3::HotConfiguration(RNG4, Umu);
@@ -84,7 +84,7 @@ int main(int argc, char** argv) {
   LatticeFermion src_o(FrbGrid);
   LatticeFermion result_o(FrbGrid);
   pickCheckerboard(Odd, src_o, src);
-  result_o = zero;
+  result_o = Zero();
 
   GridStopWatch CGTimer;
 
diff --git a/tests/solver/Test_dwf_cg_prec_LsVec.cc b/tests/solver/Test_dwf_cg_prec_LsVec.cc
deleted file mode 100644
index 27799739..00000000
--- a/tests/solver/Test_dwf_cg_prec_LsVec.cc
+++ /dev/null
@@ -1,110 +0,0 @@
-/*************************************************************************************
-
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./tests/Test_dwf_cg_prec.cc
-
-Copyright (C) 2016
-
-Author: Guido Cossu <guido.cossu@ed.ac.uk>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution
-directory
-*************************************************************************************/
-/*  END LEGAL */
-#include <Grid/Grid.h>
-
-using namespace std;
-using namespace Grid;
-using namespace Grid::QCD;
-
-template <class d>
-struct scal {
-  d internal;
-};
-
-int main(int argc, char** argv) {
-  Grid_init(&argc, &argv);
-
-  const int Ls = 16;
-
-  GridCartesian* UGrid = SpaceTimeGrid::makeFourDimGrid(
-      GridDefaultLatt(), GridDefaultSimd(Nd, vComplex::Nsimd()),
-      GridDefaultMpi());
-  GridRedBlackCartesian* UrbGrid =
-      SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
-
-  GridCartesian* sUGrid = SpaceTimeGrid::makeFourDimDWFGrid(GridDefaultLatt(), GridDefaultMpi());
-  GridRedBlackCartesian* sUrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(sUGrid);
-
-  GridCartesian*           FGrid = SpaceTimeGrid::makeFiveDimDWFGrid(Ls, UGrid);
-  GridRedBlackCartesian* FrbGrid = SpaceTimeGrid::makeFiveDimDWFRedBlackGrid(Ls, UGrid);
-
-  std::cout << GridLogMessage << "Lattice dimensions: " << GridDefaultLatt()
-            << "   Ls: " << Ls << std::endl;
-
-  std::vector<int> seeds4({1, 2, 3, 4});
-  std::vector<int> seeds5({5, 6, 7, 8});
-  GridParallelRNG RNG5(FGrid);
-  RNG5.SeedFixedIntegers(seeds5);
-  GridParallelRNG RNG4(UGrid);
-  RNG4.SeedFixedIntegers(seeds4);
-
-  std::cout << GridLogMessage << "Generating random fermion field" << std::endl;
-  LatticeFermion src(FGrid);
-  random(RNG5, src);
-  LatticeFermion result(FGrid);
-  result = zero;
-  LatticeGaugeField Umu(UGrid);
-
-  std::cout << GridLogMessage << "Generating random gauge field" << std::endl;
-  SU3::HotConfiguration(RNG4, Umu);
-
-
-
-  std::vector<LatticeColourMatrix> U(4, UGrid);
-  for (int mu = 0; mu < Nd; mu++) {
-    U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
-  }
-
-  RealD mass = 0.01;
-  RealD M5 = 1.8;
-  DomainWallFermionVec5dR Ddwf(Umu, *FGrid, *FrbGrid, *sUGrid, *sUrbGrid, mass, M5);
-
-  LatticeFermion src_o(FrbGrid);
-  LatticeFermion result_o(FrbGrid);
-  pickCheckerboard(Odd, src_o, src);
-  result_o = zero;
-
-  GridStopWatch CGTimer;
-
-  SchurDiagMooeeOperator<DomainWallFermionVec5dR, LatticeFermion> HermOpEO(Ddwf);
-  ConjugateGradient<LatticeFermion> CG(1.0e-8, 10000, 0);  // switch off the assert
-
-  Ddwf.ZeroCounters();
-  CGTimer.Start();
-  CG(HermOpEO, src_o, result_o);
-  CGTimer.Stop();
-
-  std::cout << GridLogMessage << "Total CG time : " << CGTimer.Elapsed()
-            << std::endl;
-
-  std::cout << GridLogMessage << "######## Dhop calls summary" << std::endl;
-  Ddwf.Report();
-
-  Grid_finalize();
-}
diff --git a/tests/solver/Test_dwf_cg_schur.cc b/tests/solver/Test_dwf_cg_schur.cc
index 0f75eb3c..6216c366 100644
--- a/tests/solver/Test_dwf_cg_schur.cc
+++ b/tests/solver/Test_dwf_cg_schur.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -60,7 +60,7 @@ int main (int argc, char ** argv)
   GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
 
   LatticeFermion    src(FGrid); random(RNG5,src);
-  LatticeFermion result(FGrid); result=zero;
+  LatticeFermion result(FGrid); result=Zero();
   LatticeGaugeField Umu(UGrid); SU3::HotConfiguration(RNG4,Umu);
 
   std::vector<LatticeColourMatrix> U(4,UGrid);
diff --git a/tests/solver/Test_dwf_cg_unprec.cc b/tests/solver/Test_dwf_cg_unprec.cc
index 38f632ef..0af7290c 100644
--- a/tests/solver/Test_dwf_cg_unprec.cc
+++ b/tests/solver/Test_dwf_cg_unprec.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -60,7 +60,7 @@ int main (int argc, char ** argv)
   GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
 
   LatticeFermion    src(FGrid); random(RNG5,src);
-  LatticeFermion result(FGrid); result=zero;
+  LatticeFermion result(FGrid); result=Zero();
   LatticeGaugeField Umu(UGrid); SU3::HotConfiguration(RNG4,Umu);
 
   std::vector<LatticeColourMatrix> U(4,UGrid);
diff --git a/tests/solver/Test_dwf_cr_unprec.cc b/tests/solver/Test_dwf_cr_unprec.cc
index fbfa2965..087ae376 100644
--- a/tests/solver/Test_dwf_cr_unprec.cc
+++ b/tests/solver/Test_dwf_cr_unprec.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 
 template<class d>
@@ -64,7 +64,7 @@ int main (int argc, char ** argv)
   GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
 
   LatticeFermion    src(FGrid); random(RNG5,src);
-  LatticeFermion result(FGrid); result=zero;
+  LatticeFermion result(FGrid); result=Zero();
   LatticeGaugeField Umu(UGrid); SU3::HotConfiguration(RNG4,Umu);
 
   std::vector<LatticeColourMatrix> U(4,UGrid);
diff --git a/tests/solver/Test_dwf_fpgcr.cc b/tests/solver/Test_dwf_fpgcr.cc
index aaf622b9..226bd933 100644
--- a/tests/solver/Test_dwf_fpgcr.cc
+++ b/tests/solver/Test_dwf_fpgcr.cc
@@ -30,7 +30,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/algorithms/iterative/PrecGeneralisedConjugateResidual.h>
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 
 template<class d>
@@ -65,7 +65,7 @@ int main (int argc, char ** argv)
   GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
 
   LatticeFermion    src(FGrid); random(RNG5,src);
-  LatticeFermion result(FGrid); result=zero;
+  LatticeFermion result(FGrid); result=Zero();
   LatticeGaugeField Umu(UGrid); 
 
   SU3::HotConfiguration(RNG4,Umu);
@@ -86,32 +86,32 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage<<"* Solving with MdagM VPGCR "<<std::endl;
   std::cout<<GridLogMessage<<"*********************************************************"<<std::endl;
   MdagMLinearOperator<DomainWallFermionR,LatticeFermion> HermOp(Ddwf);
-  result=zero;
+  result=Zero();
   PGCR(HermOp,src,result);
 
   std::cout<<GridLogMessage<<"*********************************************************"<<std::endl;
   std::cout<<GridLogMessage<<"* Solving with g5-VPGCR "<<std::endl;
   std::cout<<GridLogMessage<<"*********************************************************"<<std::endl;
   Gamma5R5HermitianLinearOperator<DomainWallFermionR,LatticeFermion> g5HermOp(Ddwf);
-  result=zero;
+  result=Zero();
   PGCR(g5HermOp,src,result);
 
   std::cout<<GridLogMessage<<"*********************************************************"<<std::endl;
   std::cout<<GridLogMessage<<"* Solving with MdagM-CR "<<std::endl;
   std::cout<<GridLogMessage<<"*********************************************************"<<std::endl;
-  result=zero;
+  result=Zero();
   CR(HermOp,src,result);
 
   std::cout<<GridLogMessage<<"*********************************************************"<<std::endl;
   std::cout<<GridLogMessage<<"* Solving with g5-CR "<<std::endl;
   std::cout<<GridLogMessage<<"*********************************************************"<<std::endl;
-  result=zero;
+  result=Zero();
   CR(g5HermOp,src,result);
 
   std::cout<<GridLogMessage<<"*********************************************************"<<std::endl;
   std::cout<<GridLogMessage<<"* Solving with MdagM-CG "<<std::endl;
   std::cout<<GridLogMessage<<"*********************************************************"<<std::endl;
-  result=zero;
+  result=Zero();
   CG(HermOp,src,result);
 
   Grid_finalize();
diff --git a/tests/solver/Test_dwf_hdcr.cc b/tests/solver/Test_dwf_hdcr.cc
index b3373238..58bf9032 100644
--- a/tests/solver/Test_dwf_hdcr.cc
+++ b/tests/solver/Test_dwf_hdcr.cc
@@ -33,7 +33,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 class myclass: Serializable {
 public:
@@ -93,13 +93,12 @@ public:
 
   void PowerMethod(const FineField &in) {
 
-    FineField p1(in._grid);
-    FineField p2(in._grid);
+    FineField p1(in.Grid());
+    FineField p2(in.Grid());
 
     MdagMLinearOperator<Matrix,FineField>   fMdagMOp(_FineMatrix);
 
     p1=in;
-    RealD absp2;
     for(int i=0;i<20;i++){
       RealD absp1=std::sqrt(norm2(p1));
       fMdagMOp.HermOp(p1,p2);// this is the G5 herm bit      
@@ -137,9 +136,9 @@ public:
     MdagMLinearOperator<CoarseOperator,CoarseVector>     MdagMOp(_CoarseOperator);
     MdagMLinearOperator<Matrix,FineField>               fMdagMOp(_FineMatrix);
 
-    FineField tmp(in._grid);
-    FineField res(in._grid);
-    FineField Min(in._grid);
+    FineField tmp(in.Grid());
+    FineField res(in.Grid());
+    FineField Min(in.Grid());
 
     // Monitor completeness of low mode space
     _Aggregates.ProjectToSubspace  (Csrc,in);
@@ -158,7 +157,7 @@ public:
     _FineOperator.Op(Min,tmp);
     tmp = in - tmp;   // in - A Min
 
-    Csol=zero;
+    Csol=Zero();
     _Aggregates.ProjectToSubspace  (Csrc,tmp);
     HermOp.AdjOp(Csrc,Ctmp);// Normal equations
     CG(MdagMOp,Ctmp,Csol);
@@ -193,7 +192,7 @@ public:
 
     CoarseVector Csrc(_CoarseOperator.Grid());
     CoarseVector Ctmp(_CoarseOperator.Grid());
-    CoarseVector Csol(_CoarseOperator.Grid()); Csol=zero;
+    CoarseVector Csol(_CoarseOperator.Grid()); Csol=Zero();
 
     ConjugateGradient<CoarseVector>  CG(1.0e-10,100000);
     ConjugateGradient<FineField>    fCG(3.0e-2,1000);
@@ -202,9 +201,9 @@ public:
     MdagMLinearOperator<CoarseOperator,CoarseVector>     MdagMOp(_CoarseOperator);
     ShiftedMdagMLinearOperator<Matrix,FineField>        fMdagMOp(_FineMatrix,0.1);
 
-    FineField tmp(in._grid);
-    FineField res(in._grid);
-    FineField Qin(in._grid);
+    FineField tmp(in.Grid());
+    FineField res(in.Grid());
+    FineField Qin(in.Grid());
 
     // Monitor completeness of low mode space
     //    _Aggregates.ProjectToSubspace  (Csrc,in);
@@ -235,17 +234,17 @@ public:
 
   void SAP (const FineField & src,FineField & psi){
 
-    Lattice<iScalar<vInteger> > coor(src._grid);
-    Lattice<iScalar<vInteger> > subset(src._grid);
+    Lattice<iScalar<vInteger> > coor(src.Grid());
+    Lattice<iScalar<vInteger> > subset(src.Grid());
     
-    FineField r(src._grid);
-    FineField zz(src._grid); zz=zero;
-    FineField vec1(src._grid);
-    FineField vec2(src._grid);
+    FineField r(src.Grid());
+    FineField zz(src.Grid()); zz=Zero();
+    FineField vec1(src.Grid());
+    FineField vec2(src.Grid());
 
-    const Integer block=params.domainsize;
+    const RealD block=params.domainsize;
 
-    subset=zero;
+    subset=Zero();
     for(int mu=0;mu<Nd;mu++){
       LatticeCoordinate(coor,mu+1);
       coor = div(coor,block);
@@ -295,8 +294,8 @@ public:
 
   void SmootherTest (const FineField & in){
     
-    FineField vec1(in._grid);
-    FineField vec2(in._grid);
+    FineField vec1(in.Grid());
+    FineField vec2(in.Grid());
     RealD lo[3] = { 0.5, 1.0, 2.0};
 
     //    MdagMLinearOperator<Matrix,FineField>        fMdagMOp(_FineMatrix);
@@ -327,7 +326,7 @@ public:
 
     CoarseVector Csrc(_CoarseOperator.Grid());
     CoarseVector Ctmp(_CoarseOperator.Grid());
-    CoarseVector Csol(_CoarseOperator.Grid()); Csol=zero;
+    CoarseVector Csol(_CoarseOperator.Grid()); Csol=Zero();
 
     ConjugateGradient<CoarseVector>  CG(3.0e-3,100000);
     //    ConjugateGradient<FineField>    fCG(3.0e-2,1000);
@@ -337,8 +336,8 @@ public:
     //    MdagMLinearOperator<Matrix,FineField>        fMdagMOp(_FineMatrix);
     ShiftedMdagMLinearOperator<Matrix,FineField> fMdagMOp(_SmootherMatrix,0.0);
 
-    FineField vec1(in._grid);
-    FineField vec2(in._grid);
+    FineField vec1(in.Grid());
+    FineField vec2(in.Grid());
 
     //    Chebyshev<FineField> Cheby    (0.5,70.0,30,InverseApproximation);
     //    Chebyshev<FineField> ChebyAccu(0.5,70.0,30,InverseApproximation);
@@ -407,15 +406,15 @@ public:
 
     CoarseVector Csrc(_CoarseOperator.Grid());
     CoarseVector Ctmp(_CoarseOperator.Grid());
-    CoarseVector Csol(_CoarseOperator.Grid()); Csol=zero;
+    CoarseVector Csol(_CoarseOperator.Grid()); Csol=Zero();
 
     ConjugateGradient<CoarseVector>  CG(1.0e-3,100000);
 
     HermitianLinearOperator<CoarseOperator,CoarseVector>  HermOp(_CoarseOperator);
     MdagMLinearOperator<CoarseOperator,CoarseVector>     MdagMOp(_CoarseOperator);
 
-    FineField vec1(in._grid);
-    FineField vec2(in._grid);
+    FineField vec1(in.Grid());
+    FineField vec2(in.Grid());
 
     _Aggregates.ProjectToSubspace  (Csrc,in);
     _Aggregates.PromoteFromSubspace(Csrc,out);
@@ -490,7 +489,7 @@ int main (int argc, char ** argv)
   std::vector<int> block ({2,2,2,2});
   const int nbasis= 32;
 
-  std::vector<int> clatt = GridDefaultLatt();
+  auto clatt = GridDefaultLatt();
   for(int d=0;d<clatt.size();d++){
     clatt[d] = clatt[d]/block[d];
   }
@@ -507,8 +506,8 @@ int main (int argc, char ** argv)
   Gamma g5(Gamma::Algebra::Gamma5);
 
   LatticeFermion    src(FGrid); gaussian(RNG5,src);// src=src+g5*src;
-  LatticeFermion result(FGrid); result=zero;
-  LatticeFermion    ref(FGrid); ref=zero;
+  LatticeFermion result(FGrid); result=Zero();
+  LatticeFermion    ref(FGrid); ref=Zero();
   LatticeFermion    tmp(FGrid);
   LatticeFermion    err(FGrid);
   LatticeGaugeField Umu(UGrid); 
@@ -523,7 +522,7 @@ int main (int argc, char ** argv)
 
   if ( params.domaindecompose ) { 
     Lattice<iScalar<vInteger> > coor(UGrid);
-    zz=zero;
+    zz=Zero();
     for(int mu=0;mu<Nd;mu++){
       LatticeCoordinate(coor,mu);
       U = PeekIndex<LorentzIndex>(Umu,mu);
@@ -536,7 +535,7 @@ int main (int argc, char ** argv)
   //  SU3::ColdConfiguration(RNG4,Umu);
   //  SU3::TepidConfiguration(RNG4,Umu);
   //  SU3::HotConfiguration(RNG4,Umu);
-  //  Umu=zero;
+  //  Umu=Zero();
 
   RealD mass=params.mq;
   RealD M5=1.8;
@@ -574,7 +573,7 @@ int main (int argc, char ** argv)
 //    result =     Aggregates.subspace[i];
 //    Aggregates.subspace[i]=result+g5*result;
 //  }
-  result=zero;
+  result=Zero();
   
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Building coarse representation of Indef operator" <<std::endl;
@@ -590,7 +589,7 @@ int main (int argc, char ** argv)
   CoarseVector c_src (Coarse5d);
   CoarseVector c_res (Coarse5d);
   gaussian(CRNG,c_src);
-  c_res=zero;
+  c_res=Zero();
 
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   std::cout<<GridLogMessage << "Solving posdef-CG on coarse space "<< std::endl;
@@ -661,7 +660,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "Building a two level DDPGCR "<< std::endl;
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   //  PrecGeneralisedConjugateResidual<LatticeFermion> PGCRDD(1.0e-8,100000,PreconDD,8,128);
-  //  result=zero;
+  //  result=Zero();
   //  std::cout<<GridLogMessage<<"checking norm src "<<norm2(src)<<std::endl;
   //  PGCRDD(HermIndefOp,src,result);
 
@@ -670,7 +669,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
   PrecGeneralisedConjugateResidual<LatticeFermion> PGCR(1.0e-8,100000,Precon,8,8);
   std::cout<<GridLogMessage<<"checking norm src "<<norm2(src)<<std::endl;
-  result=zero;
+  result=Zero();
   PGCR(HermIndefOp,src,result);
 
   std::cout<<GridLogMessage << "**************************************************"<< std::endl;
@@ -682,7 +681,7 @@ int main (int argc, char ** argv)
   LatticeFermion    src_o(FrbGrid);
   LatticeFermion result_o(FrbGrid);
   pickCheckerboard(Odd,src_o,src);
-  result_o=zero;
+  result_o=Zero();
 
   pCG(HermOpEO,src_o,result_o);
 
diff --git a/tests/solver/Test_dwf_mrhs_cg.cc b/tests/solver/Test_dwf_mrhs_cg.cc
index 72ba3d26..982a8247 100644
--- a/tests/solver/Test_dwf_mrhs_cg.cc
+++ b/tests/solver/Test_dwf_mrhs_cg.cc
@@ -30,7 +30,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
@@ -42,10 +42,10 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
-  std::vector<int> mpi_split (mpi_layout.size(),1);
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
+  Coordinate mpi_split (mpi_layout.size(),1);
 
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
   GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
@@ -91,7 +91,7 @@ int main (int argc, char ** argv)
   FermionField tmp(FGrid);
 
   for(int s=0;s<nrhs;s++) random(pRNG5,src[s]);
-  for(int s=0;s<nrhs;s++) result[s]=zero;
+  for(int s=0;s<nrhs;s++) result[s]=Zero();
 
   LatticeGaugeField Umu(UGrid); SU3::HotConfiguration(pRNG,Umu);
 
@@ -203,7 +203,7 @@ int main (int argc, char ** argv)
   MdagMLinearOperator<DomainWallFermionR,FermionField> HermOp(Ddwf);
   MdagMLinearOperator<DomainWallFermionR,FermionField> HermOpCk(Dchk);
   ConjugateGradient<FermionField> CG((1.0e-5/(me+1)),10000);
-  s_res = zero;
+  s_res = Zero();
   CG(HermOp,s_src,s_res);
 
   /////////////////////////////////////////////////////////////
diff --git a/tests/solver/Test_dwf_mrhs_cg_mpi.cc b/tests/solver/Test_dwf_mrhs_cg_mpi.cc
index c1d8ce62..8ace9b43 100644
--- a/tests/solver/Test_dwf_mrhs_cg_mpi.cc
+++ b/tests/solver/Test_dwf_mrhs_cg_mpi.cc
@@ -30,7 +30,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
@@ -43,10 +43,10 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
-  std::vector<int> mpi_split (mpi_layout.size(),1);
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
+  Coordinate mpi_split (mpi_layout.size(),1);
 
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), 
 								   GridDefaultSimd(Nd,vComplex::Nsimd()),
@@ -97,11 +97,11 @@ int main (int argc, char ** argv)
   FermionField tmp(FGrid);
   std::cout << GridLogMessage << "Made the Fermion Fields"<<std::endl;
 
-  for(int s=0;s<nrhs;s++) result[s]=zero;
+  for(int s=0;s<nrhs;s++) result[s]=Zero();
 #undef LEXICO_TEST
 #ifdef LEXICO_TEST
   {
-    LatticeFermion lex(FGrid);  lex = zero;
+    LatticeFermion lex(FGrid);  lex = Zero();
     LatticeFermion ftmp(FGrid);
     Integer stride =10000;
     double nrm;
@@ -138,7 +138,7 @@ int main (int argc, char ** argv)
     std::cout << GridLogMessage << "Intialised 4D RNG "<<std::endl;
     SU3::HotConfiguration(pRNG,Umu);
     std::cout << GridLogMessage << "Intialised the HOT Gauge Field"<<std::endl;
-    //    std::cout << " Site zero "<< Umu._odata[0]   <<std::endl;
+    //    std::cout << " Site zero "<< Umu[0]   <<std::endl;
   } else { 
     SU3::ColdConfiguration(Umu);
     std::cout << GridLogMessage << "Intialised the COLD Gauge Field"<<std::endl;
@@ -163,7 +163,7 @@ int main (int argc, char ** argv)
   FermionField s_src_tmp(SFGrid);
   FermionField s_src_diff(SFGrid);
   {
-    LatticeFermion lex(SFGrid);  lex = zero;
+    LatticeFermion lex(SFGrid);  lex = Zero();
     LatticeFermion ftmp(SFGrid);
     Integer stride =10000;
     double nrm;
@@ -199,7 +199,7 @@ int main (int argc, char ** argv)
   MdagMLinearOperator<DomainWallFermionR,FermionField> HermOp(Ddwf);
   MdagMLinearOperator<DomainWallFermionR,FermionField> HermOpCk(Dchk);
   ConjugateGradient<FermionField> CG((stp),10000);
-  s_res = zero;
+  s_res = Zero();
   CG(HermOp,s_src,s_res);
 
   std::cout << GridLogMessage << " split residual norm "<<norm2(s_res)<<std::endl;
@@ -228,7 +228,7 @@ int main (int argc, char ** argv)
     std::cout << GridLogMessage<<" resid["<<n<<"]  "<< norm2(tmp)/norm2(src[n])<<std::endl;
   }
 
-  for(int s=0;s<nrhs;s++) result[s]=zero;
+  for(int s=0;s<nrhs;s++) result[s]=Zero();
   int blockDim = 0;//not used for BlockCGVec
   BlockConjugateGradient<FermionField>    BCGV  (BlockCGVec,blockDim,stp,10000);
   BCGV.PrintInterval=10;
diff --git a/tests/solver/Test_dwf_mrhs_cg_mpieo.cc b/tests/solver/Test_dwf_mrhs_cg_mpieo.cc
index a6dfcd57..6e5208ea 100644
--- a/tests/solver/Test_dwf_mrhs_cg_mpieo.cc
+++ b/tests/solver/Test_dwf_mrhs_cg_mpieo.cc
@@ -30,7 +30,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
@@ -42,10 +42,10 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
-  std::vector<int> mpi_split (mpi_layout.size(),1);
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
+  Coordinate mpi_split (mpi_layout.size(),1);
 
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), 
 								   GridDefaultSimd(Nd,vComplex::Nsimd()),
@@ -85,7 +85,7 @@ int main (int argc, char ** argv)
   std::vector<FermionField> src_o(nrhs,FrbGrid);
 
   for(int s=0;s<nrhs;s++) random(pRNG5,src[s]);
-  for(int s=0;s<nrhs;s++) result[s]=zero;
+  for(int s=0;s<nrhs;s++) result[s]=Zero();
 
   LatticeGaugeField Umu(UGrid); SU3::HotConfiguration(pRNG,Umu);
 
@@ -134,7 +134,7 @@ int main (int argc, char ** argv)
   MdagMLinearOperator<DomainWallFermionR,FermionField> HermOp(Ddwf);
   MdagMLinearOperator<DomainWallFermionR,FermionField> HermOpCk(Dchk);
   ConjugateGradient<FermionField> CG((1.0e-8/(me+1)),10000);
-  s_res = zero;
+  s_res = Zero();
   CG(HermOp,s_src,s_res);
 
   /////////////////////////////////////////////////////////////
diff --git a/tests/solver/Test_laplacian.cc b/tests/solver/Test_laplacian.cc
index f163fe54..0ea26ff0 100644
--- a/tests/solver/Test_laplacian.cc
+++ b/tests/solver/Test_laplacian.cc
@@ -29,16 +29,16 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -56,11 +56,11 @@ int main (int argc, char ** argv)
   // Source and result in the algebra
   // needed for the second test
   AVector src_vec(&Grid); random(pRNG, src_vec);
-  AVector result_vec(&Grid); result_vec = zero;
+  AVector result_vec(&Grid); result_vec = Zero();
   
   LatticeColourMatrix src(&Grid); 
   SU<Nc>::FundamentalLieAlgebraMatrix(src_vec, src);
-  LatticeColourMatrix result(&Grid); result=zero;
+  LatticeColourMatrix result(&Grid); result=Zero();
 
 
   // Generate a field of adjoint matrices
diff --git a/tests/solver/Test_mobius_bcg.cc b/tests/solver/Test_mobius_bcg.cc
index e59cb7e0..8b34a6a5 100644
--- a/tests/solver/Test_mobius_bcg.cc
+++ b/tests/solver/Test_mobius_bcg.cc
@@ -30,7 +30,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 int main (int argc, char ** argv)
 {
@@ -42,14 +41,14 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
-  std::vector<int> mpi_split (mpi_layout.size(),1);
-  std::vector<int> split_coor (mpi_layout.size(),1);
-  std::vector<int> split_dim (mpi_layout.size(),1);
+  auto latt_size   = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
+  Coordinate mpi_split (mpi_layout.size(),1);
+  Coordinate split_coor (mpi_layout.size(),1);
+  Coordinate split_dim (mpi_layout.size(),1);
 
-  std::vector<ComplexD> boundary_phases(Nd,1.);
+  std::vector<Complex> boundary_phases(Nd,1.);
   boundary_phases[Nd-1]=-1.;
   params.boundary_phases = boundary_phases;
 
@@ -108,7 +107,7 @@ int main (int argc, char ** argv)
   FermionField tmp(FGrid);
   std::cout << GridLogMessage << "Made the Fermion Fields"<<std::endl;
 
-  for(int s=0;s<nrhs;s++) result[s]=zero;
+  for(int s=0;s<nrhs;s++) result[s]=Zero();
   GridParallelRNG pRNG5(FGrid);  pRNG5.SeedFixedIntegers(seeds);
   for(int s=0;s<nrhs;s++) {
     random(pRNG5,src[s]);
@@ -131,7 +130,6 @@ int main (int argc, char ** argv)
     std::cout << GridLogMessage << "Intialised 4D RNG "<<std::endl;
     SU3::HotConfiguration(pRNG,Umu);
     std::cout << GridLogMessage << "Intialised the HOT Gauge Field"<<std::endl;
-    std::cout << " Site zero "<< Umu._odata[0]   <<std::endl;
   } 
 
   /////////////////
@@ -170,7 +168,7 @@ int main (int argc, char ** argv)
   MdagMLinearOperator<MobiusFermionR,FermionField> HermOp(Ddwf);
   MdagMLinearOperator<MobiusFermionR,FermionField> HermOpCk(Dchk);
   ConjugateGradient<FermionField> CG((stp),100000);
-  s_res = zero;
+  s_res = Zero();
 
   CG(HermOp,s_src,s_res);
 
@@ -202,7 +200,7 @@ int main (int argc, char ** argv)
 
 
   for(int s=0;s<nrhs;s++){
-    result[s]=zero;
+    result[s]=Zero();
   }
 
   /////////////////////////////////////////////////////////////
diff --git a/tests/solver/Test_mobius_bcg_nosplit.cc b/tests/solver/Test_mobius_bcg_nosplit.cc
index f3ed621f..3f0b7058 100644
--- a/tests/solver/Test_mobius_bcg_nosplit.cc
+++ b/tests/solver/Test_mobius_bcg_nosplit.cc
@@ -31,7 +31,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/algorithms/iterative/BlockConjugateGradient.h>
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
@@ -43,11 +43,11 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  auto latt_size   = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
 
-  std::vector<ComplexD> boundary_phases(Nd,1.);
+  std::vector<Complex> boundary_phases(Nd,1.);
   boundary_phases[Nd-1]=-1.;
   params.boundary_phases = boundary_phases;
 
@@ -72,7 +72,7 @@ int main (int argc, char ** argv)
   FermionField tmp(FGrid);
   std::cout << GridLogMessage << "Made the Fermion Fields"<<std::endl;
 
-  for(int s=0;s<nrhs;s++) result[s]=zero;
+  for(int s=0;s<nrhs;s++) result[s]=Zero();
   GridParallelRNG pRNG5(FGrid);  pRNG5.SeedFixedIntegers(seeds);
   for(int s=0;s<nrhs;s++) {
     random(pRNG5,src[s]);
@@ -116,7 +116,7 @@ int main (int argc, char ** argv)
   ConjugateGradient<FermionField> CG((stp),100000);
 
   for(int rhs=0;rhs<1;rhs++){
-    result[rhs] = zero;
+    result[rhs] = Zero();
     CG(HermOp,src[rhs],result[rhs]);
   }
 
@@ -129,7 +129,7 @@ int main (int argc, char ** argv)
   /////////////////////////////////////////////////////////////
   int blockDim = 0;//not used for BlockCGVec
   for(int s=0;s<nrhs;s++){
-      result[s]=zero;
+      result[s]=Zero();
    }
   BlockConjugateGradient<FermionField>    BCGV  (BlockCGrQVec,blockDim,stp,100000);
   {
diff --git a/tests/solver/Test_mobius_bcg_phys_nosplit.cc b/tests/solver/Test_mobius_bcg_phys_nosplit.cc
index 15617a05..f929bd30 100644
--- a/tests/solver/Test_mobius_bcg_phys_nosplit.cc
+++ b/tests/solver/Test_mobius_bcg_phys_nosplit.cc
@@ -31,7 +31,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/algorithms/iterative/BlockConjugateGradient.h>
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
@@ -43,11 +43,11 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  auto latt_size   = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
 
-  std::vector<ComplexD> boundary_phases(Nd,1.);
+  std::vector<Complex> boundary_phases(Nd,1.);
   boundary_phases[Nd-1]=-1.;
   params.boundary_phases = boundary_phases;
 
@@ -73,7 +73,7 @@ int main (int argc, char ** argv)
   FermionField tmp(FGrid);
   std::cout << GridLogMessage << "Made the Fermion Fields"<<std::endl;
 
-  for(int s=0;s<nrhs;s++) result[s]=zero;
+  for(int s=0;s<nrhs;s++) result[s]=Zero();
   GridParallelRNG pRNG4(UGrid);  pRNG4.SeedFixedIntegers(seeds);
   for(int s=0;s<nrhs;s++) {
     random(pRNG4,src4[s]);
@@ -120,7 +120,7 @@ int main (int argc, char ** argv)
   ConjugateGradient<FermionField> CG((stp),100000);
 
   for(int rhs=0;rhs<1;rhs++){
-    result[rhs] = zero;
+    result[rhs] = Zero();
     //    CG(HermOp,src[rhs],result[rhs]);
   }
 
@@ -133,7 +133,7 @@ int main (int argc, char ** argv)
   /////////////////////////////////////////////////////////////
   int blockDim = 0;//not used for BlockCGVec
   for(int s=0;s<nrhs;s++){
-    result[s]=zero;
+    result[s]=Zero();
   }
   BlockConjugateGradient<FermionField>    BCGV  (BlockCGrQVec,blockDim,stp,100000);
   {
diff --git a/tests/solver/Test_mobius_bcg_prec_nosplit.cc b/tests/solver/Test_mobius_bcg_prec_nosplit.cc
index 63078613..31120a8e 100644
--- a/tests/solver/Test_mobius_bcg_prec_nosplit.cc
+++ b/tests/solver/Test_mobius_bcg_prec_nosplit.cc
@@ -31,7 +31,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include <Grid/algorithms/iterative/BlockConjugateGradient.h>
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
@@ -43,11 +43,11 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  auto latt_size   = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
 
-  std::vector<ComplexD> boundary_phases(Nd,1.);
+  std::vector<Complex> boundary_phases(Nd,1.);
   boundary_phases[Nd-1]=-1.;
   params.boundary_phases = boundary_phases;
 
@@ -72,7 +72,7 @@ int main (int argc, char ** argv)
   FermionField tmp(FGrid);
   std::cout << GridLogMessage << "Made the Fermion Fields"<<std::endl;
 
-  for(int s=0;s<nrhs;s++) result[s]=zero;
+  for(int s=0;s<nrhs;s++) result[s]=Zero();
   GridParallelRNG pRNG5(FGrid);  pRNG5.SeedFixedIntegers(seeds);
   for(int s=0;s<nrhs;s++) {
     random(pRNG5,src[s]);
@@ -116,7 +116,7 @@ int main (int argc, char ** argv)
   ConjugateGradient<FermionField> CG((stp),100000);
 
   for(int rhs=0;rhs<1;rhs++){
-    result[rhs] = zero;
+    result[rhs] = Zero();
     CG(HermOp,src[rhs],result[rhs]);
   }
 
@@ -129,7 +129,7 @@ int main (int argc, char ** argv)
   /////////////////////////////////////////////////////////////
   int blockDim = 0;//not used for BlockCGVec
   for(int s=0;s<nrhs;s++){
-    result[s]=zero;
+    result[s]=Zero();
   }
 
 
diff --git a/tests/solver/Test_multigrid_common.h b/tests/solver/Test_multigrid_common.h
index add833f2..22743b72 100644
--- a/tests/solver/Test_multigrid_common.h
+++ b/tests/solver/Test_multigrid_common.h
@@ -30,6 +30,8 @@
 
 namespace Grid {
 
+Zero zero;
+
 // TODO: Can think about having one parameter struct per level and then a
 // vector of these structs. How well would that work together with the
 // serialization strategy of Grid?
@@ -124,7 +126,7 @@ public:
         Seeds[level][d] = (level)*Nd + d + 1;
       }
 
-      Grids.push_back(QCD::SpaceTimeGrid::makeFourDimGrid(tmp, Grids[level - 1]->_simd_layout, GridDefaultMpi()));
+      Grids.push_back(SpaceTimeGrid::makeFourDimGrid(tmp, Grids[level - 1]->_simd_layout, GridDefaultMpi()));
       PRNGs.push_back(GridParallelRNG(Grids[level]));
 
       PRNGs[level].SeedFixedIntegers(Seeds[level]);
@@ -227,8 +229,8 @@ public:
     _SetupCreateSubspaceTimer.Stop();
 
     _SetupProjectToChiralitiesTimer.Start();
-    FineVector tmp1(_Aggregates.subspace[0]._grid);
-    FineVector tmp2(_Aggregates.subspace[0]._grid);
+    FineVector tmp1(_Aggregates.subspace[0].Grid());
+    FineVector tmp2(_Aggregates.subspace[0].Grid());
     for(int n = 0; n < nb; n++) {
       auto tmp1 = _Aggregates.subspace[n];
       G5C(tmp2, _Aggregates.subspace[n]);
@@ -253,7 +255,7 @@ public:
 
   virtual void operator()(FineVector const &in, FineVector &out) {
 
-    conformable(_LevelInfo.Grids[_CurrentLevel], in._grid);
+    conformable(_LevelInfo.Grids[_CurrentLevel], in.Grid());
     conformable(in, out);
 
     // TODO: implement a W-cycle
@@ -273,7 +275,7 @@ public:
     CoarseVector coarseSol(_LevelInfo.Grids[_NextCoarserLevel]);
     coarseSol = zero;
 
-    FineVector fineTmp(in._grid);
+    FineVector fineTmp(in.Grid());
 
     auto maxSmootherIter = _MultiGridParams.smootherMaxOuterIter[_CurrentLevel] * _MultiGridParams.smootherMaxInnerIter[_CurrentLevel];
 
@@ -330,7 +332,7 @@ public:
     CoarseVector coarseSol(_LevelInfo.Grids[_NextCoarserLevel]);
     coarseSol = zero;
 
-    FineVector fineTmp(in._grid);
+    FineVector fineTmp(in.Grid());
 
     auto smootherMaxIter = _MultiGridParams.smootherMaxOuterIter[_CurrentLevel] * _MultiGridParams.smootherMaxInnerIter[_CurrentLevel];
     auto kCycleMaxIter   = _MultiGridParams.kCycleMaxOuterIter[_CurrentLevel] * _MultiGridParams.kCycleMaxInnerIter[_CurrentLevel];
@@ -605,7 +607,7 @@ public:
 
     _SolveTotalTimer.Start();
 
-    conformable(_LevelInfo.Grids[_CurrentLevel], in._grid);
+    conformable(_LevelInfo.Grids[_CurrentLevel], in.Grid());
     conformable(in, out);
 
     auto coarseSolverMaxIter = _MultiGridParams.coarseSolverMaxOuterIter * _MultiGridParams.coarseSolverMaxInnerIter;
diff --git a/tests/solver/Test_split_grid.cc b/tests/solver/Test_split_grid.cc
index 2b6a4bf7..74c51a05 100644
--- a/tests/solver/Test_split_grid.cc
+++ b/tests/solver/Test_split_grid.cc
@@ -30,7 +30,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 int main (int argc, char ** argv)
 {
@@ -42,10 +42,10 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
-  std::vector<int> mpi_split (mpi_layout.size(),1);
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
+  Coordinate mpi_split (mpi_layout.size(),1);
 
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
   GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
@@ -92,7 +92,7 @@ int main (int argc, char ** argv)
   FermionField tmp(FGrid);
 
   for(int s=0;s<nrhs;s++) random(pRNG5,src[s]);
-  for(int s=0;s<nrhs;s++) result[s]=zero;
+  for(int s=0;s<nrhs;s++) result[s]=Zero();
 
   LatticeGaugeField Umu(UGrid); SU3::HotConfiguration(pRNG,Umu);
 
@@ -127,7 +127,7 @@ int main (int argc, char ** argv)
   MdagMLinearOperator<DomainWallFermionR,FermionField> HermOp(Ddwf);
   MdagMLinearOperator<DomainWallFermionR,FermionField> HermOpCk(Dchk);
   ConjugateGradient<FermionField> CG((1.0e-8/(me+1)),10000);
-  s_res = zero;
+  s_res = Zero();
   CG(HermOp,s_src,s_res);
 
   /////////////////////////////////////////////////////////////
diff --git a/tests/solver/Test_staggered_block_cg_prec.cc b/tests/solver/Test_staggered_block_cg_prec.cc
index acd90bbf..95156939 100644
--- a/tests/solver/Test_staggered_block_cg_prec.cc
+++ b/tests/solver/Test_staggered_block_cg_prec.cc
@@ -30,7 +30,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -54,9 +54,9 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
   GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
@@ -72,7 +72,7 @@ int main (int argc, char ** argv)
 #if 1
   random(pRNG5,src);
 #else
-  src=zero;
+  src=Zero();
   ComplexField coor(FGrid);
   LatticeCoordinate(coor,0);
   for(int ss=0;ss<FGrid->oSites();ss++){
@@ -84,7 +84,7 @@ int main (int argc, char ** argv)
   }
 #endif
   FermionField src_o(FrbGrid);   pickCheckerboard(Odd,src_o,src);
-  FermionField result_o(FrbGrid); result_o=zero; 
+  FermionField result_o(FrbGrid); result_o=Zero(); 
   RealD nrm = norm2(src);
 
   LatticeGaugeField Umu(UGrid); SU3::HotConfiguration(pRNG,Umu);
@@ -117,8 +117,8 @@ int main (int argc, char ** argv)
   FermionField src4d_o(UrbGrid);   pickCheckerboard(Odd,src4d_o,src4d);
   FermionField result4d_o(UrbGrid); 
 
+  result4d_o=Zero();
   double deodoe_flops=(16*(3*(6+8+8)) + 15*3*2)*volume; // == 66*16 +  == 1146
-  result4d_o=zero;
   {
     double t1=usecond();
     CG(HermOp4d,src4d_o,result4d_o);
@@ -138,7 +138,7 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << " Calling 5d CG for "<<Ls <<" right hand sides" <<std::endl;
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   Ds.ZeroCounters();
-  result_o=zero;
+  result_o=Zero();
   {
     double t1=usecond();
     CG(HermOp,src_o,result_o);
@@ -157,7 +157,7 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << " Calling multiRHS CG for "<<Ls <<" right hand sides" <<std::endl;
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   Ds.ZeroCounters();
-  result_o=zero;
+  result_o=Zero();
   {
     double t1=usecond();
     mCG(HermOp,src_o,result_o);
@@ -177,7 +177,7 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << " Calling Block CGrQ for "<<Ls <<" right hand sides" <<std::endl;
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   Ds.ZeroCounters();
-  result_o=zero;
+  result_o=Zero();
   {
     double t1=usecond();
     BCGrQ(HermOp,src_o,result_o);
@@ -196,7 +196,7 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << " Calling Block CG for "<<Ls <<" right hand sides" <<std::endl;
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   Ds.ZeroCounters();
-  result_o=zero;
+  result_o=Zero();
   {
     double t1=usecond();
     BCG(HermOp,src_o,result_o);
@@ -216,7 +216,7 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << "****************************************************************** "<<std::endl;
   std::vector<FermionField> src_v   (Ls,UrbGrid);
   std::vector<FermionField> result_v(Ls,UrbGrid);
-  for(int s=0;s<Ls;s++) result_v[s] = zero;
+  for(int s=0;s<Ls;s++) result_v[s] = Zero();
   for(int s=0;s<Ls;s++) {
     FermionField src4(UGrid);
     ExtractSlice(src4,src,s,0);
diff --git a/tests/solver/Test_staggered_block_cg_unprec.cc b/tests/solver/Test_staggered_block_cg_unprec.cc
index 748b5332..b1797355 100644
--- a/tests/solver/Test_staggered_block_cg_unprec.cc
+++ b/tests/solver/Test_staggered_block_cg_unprec.cc
@@ -30,7 +30,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -54,9 +54,9 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
 
   GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
   GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
@@ -68,7 +68,7 @@ int main (int argc, char ** argv)
   GridParallelRNG pRNG5(FGrid);  pRNG5.SeedFixedIntegers(seeds);
 
   FermionField src(FGrid); random(pRNG5,src);
-  FermionField result(FGrid); result=zero;
+  FermionField result(FGrid); result=Zero();
   RealD nrm = norm2(src);
 
   LatticeGaugeField Umu(UGrid); SU3::HotConfiguration(pRNG,Umu);
@@ -98,7 +98,7 @@ int main (int argc, char ** argv)
   ImprovedStaggeredFermionR Ds4d(Umu,Umu,*UGrid,*UrbGrid,mass,c1,c2,u0);
   MdagMLinearOperator<ImprovedStaggeredFermionR,FermionField> HermOp4d(Ds4d);
   FermionField src4d(UGrid); random(pRNG,src4d);
-  FermionField result4d(UGrid); result4d=zero;
+  FermionField result4d(UGrid); result4d=Zero();
 
   double deodoe_flops=(16*(3*(6+8+8)) + 15*3*2)*volume; // == 66*16 +  == 1146
   {
@@ -118,7 +118,7 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   std::cout << GridLogMessage << " Calling 5d CG for "<<Ls <<" right hand sides" <<std::endl;
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
-  result=zero;
+  result=Zero();
 {
   Ds.ZeroCounters();
   double t1=usecond();
@@ -136,7 +136,7 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   std::cout << GridLogMessage << " Calling multiRHS CG for "<<Ls <<" right hand sides" <<std::endl;
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
-  result=zero;
+  result=Zero();
   Ds.ZeroCounters();
 {
   double t1=usecond();
@@ -154,7 +154,7 @@ int main (int argc, char ** argv)
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
   std::cout << GridLogMessage << " Calling Block CG for "<<Ls <<" right hand sides" <<std::endl;
   std::cout << GridLogMessage << "************************************************************************ "<<std::endl;
-  result=zero;
+  result=Zero();
   Ds.ZeroCounters();
 {
   double t1=usecond();
diff --git a/tests/solver/Test_staggered_cagmres_unprec.cc b/tests/solver/Test_staggered_cagmres_unprec.cc
index b82ecaeb..8121c90d 100644
--- a/tests/solver/Test_staggered_cagmres_unprec.cc
+++ b/tests/solver/Test_staggered_cagmres_unprec.cc
@@ -29,7 +29,7 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
@@ -39,9 +39,9 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -50,7 +50,7 @@ int main (int argc, char ** argv)
 
   FermionField src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  FermionField result(&Grid); result=zero;
+  FermionField result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_staggered_cg_prec.cc b/tests/solver/Test_staggered_cg_prec.cc
index 42e91a35..94373301 100644
--- a/tests/solver/Test_staggered_cg_prec.cc
+++ b/tests/solver/Test_staggered_cg_prec.cc
@@ -31,7 +31,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -52,9 +52,9 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -79,7 +79,7 @@ int main (int argc, char ** argv)
   FermionField res_o(&RBGrid); 
   FermionField src_o(&RBGrid); 
   pickCheckerboard(Odd,src_o,src);
-  res_o=zero;
+  res_o=Zero();
 
   SchurStaggeredOperator<ImprovedStaggeredFermionR,FermionField> HermOpEO(Ds);
   ConjugateGradient<FermionField> CG(1.0e-8,10000);
diff --git a/tests/solver/Test_staggered_cg_schur.cc b/tests/solver/Test_staggered_cg_schur.cc
index 03a173f9..81a2a0fa 100644
--- a/tests/solver/Test_staggered_cg_schur.cc
+++ b/tests/solver/Test_staggered_cg_schur.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -49,9 +49,9 @@ int main (int argc, char ** argv)
   typename ImprovedStaggeredFermionR::ImplParams params; 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -61,7 +61,7 @@ int main (int argc, char ** argv)
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   FermionField    src(&Grid); random(pRNG,src);
-  FermionField result(&Grid); result=zero;
+  FermionField result(&Grid); result=Zero();
   FermionField  resid(&Grid); 
 
   RealD mass=0.1;
diff --git a/tests/solver/Test_staggered_cg_unprec.cc b/tests/solver/Test_staggered_cg_unprec.cc
index 3fae925a..9625a9c8 100644
--- a/tests/solver/Test_staggered_cg_unprec.cc
+++ b/tests/solver/Test_staggered_cg_unprec.cc
@@ -31,7 +31,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -53,9 +53,9 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -64,7 +64,7 @@ int main (int argc, char ** argv)
 
   FermionField src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  FermionField result(&Grid); result=zero;
+  FermionField result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_staggered_fcagmres_prec.cc b/tests/solver/Test_staggered_fcagmres_prec.cc
index 7685585b..d08073c7 100644
--- a/tests/solver/Test_staggered_fcagmres_prec.cc
+++ b/tests/solver/Test_staggered_fcagmres_prec.cc
@@ -29,7 +29,7 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
@@ -39,9 +39,9 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -50,7 +50,7 @@ int main (int argc, char ** argv)
 
   FermionField src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  FermionField result(&Grid); result=zero;
+  FermionField result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_staggered_fgmres_prec.cc b/tests/solver/Test_staggered_fgmres_prec.cc
index 30905e35..98ef83a6 100644
--- a/tests/solver/Test_staggered_fgmres_prec.cc
+++ b/tests/solver/Test_staggered_fgmres_prec.cc
@@ -29,7 +29,7 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
@@ -39,9 +39,9 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -50,7 +50,7 @@ int main (int argc, char ** argv)
 
   FermionField src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  FermionField result(&Grid); result=zero;
+  FermionField result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_staggered_gmres_unprec.cc b/tests/solver/Test_staggered_gmres_unprec.cc
index d65b0b31..e6bb5fc7 100644
--- a/tests/solver/Test_staggered_gmres_unprec.cc
+++ b/tests/solver/Test_staggered_gmres_unprec.cc
@@ -29,7 +29,7 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
@@ -39,9 +39,9 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -50,7 +50,7 @@ int main (int argc, char ** argv)
 
   FermionField src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  FermionField result(&Grid); result=zero;
+  FermionField result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_staggered_mr_unprec.cc b/tests/solver/Test_staggered_mr_unprec.cc
index ca60edb4..d288c9c2 100644
--- a/tests/solver/Test_staggered_mr_unprec.cc
+++ b/tests/solver/Test_staggered_mr_unprec.cc
@@ -29,7 +29,7 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
@@ -39,9 +39,9 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -50,7 +50,7 @@ int main (int argc, char ** argv)
 
   FermionField src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  FermionField result(&Grid); result=zero;
+  FermionField result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_staggered_multishift.cc b/tests/solver/Test_staggered_multishift.cc
index 04386027..0d455414 100644
--- a/tests/solver/Test_staggered_multishift.cc
+++ b/tests/solver/Test_staggered_multishift.cc
@@ -31,7 +31,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 template<class d>
 struct scal {
@@ -52,9 +51,9 @@ int main (int argc, char ** argv)
 
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  auto latt_size   = GridDefaultLatt();
+  auto simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  auto mpi_layout  = GridDefaultMpi();
 
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
diff --git a/tests/solver/Test_wilson_cagmres_unprec.cc b/tests/solver/Test_wilson_cagmres_unprec.cc
index 46f9e6a6..ad502e05 100644
--- a/tests/solver/Test_wilson_cagmres_unprec.cc
+++ b/tests/solver/Test_wilson_cagmres_unprec.cc
@@ -29,15 +29,15 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -46,7 +46,7 @@ int main (int argc, char ** argv)
 
   LatticeFermion src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  LatticeFermion result(&Grid); result=zero;
+  LatticeFermion result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_wilson_cg_prec.cc b/tests/solver/Test_wilson_cg_prec.cc
index 99ddfceb..5696a24b 100644
--- a/tests/solver/Test_wilson_cg_prec.cc
+++ b/tests/solver/Test_wilson_cg_prec.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -48,9 +48,9 @@ int main (int argc, char ** argv)
   Grid_init(&argc,&argv);
 
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -59,7 +59,7 @@ int main (int argc, char ** argv)
 
   LatticeFermion src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  LatticeFermion result(&Grid); result=zero;
+  LatticeFermion result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   std::vector<LatticeColourMatrix> U(4,&Grid);
@@ -78,7 +78,7 @@ int main (int argc, char ** argv)
   LatticeFermion    src_o(&RBGrid);
   LatticeFermion result_o(&RBGrid);
   pickCheckerboard(Odd,src_o,src);
-  result_o=zero;
+  result_o=Zero();
 
   SchurDiagMooeeOperator<WilsonFermionR,LatticeFermion> HermOpEO(Dw);
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
diff --git a/tests/solver/Test_wilson_cg_schur.cc b/tests/solver/Test_wilson_cg_schur.cc
index 13ac0090..23383032 100644
--- a/tests/solver/Test_wilson_cg_schur.cc
+++ b/tests/solver/Test_wilson_cg_schur.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -48,9 +48,9 @@ int main (int argc, char ** argv)
   Grid_init(&argc,&argv);
 
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -60,7 +60,7 @@ int main (int argc, char ** argv)
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   LatticeFermion    src(&Grid); random(pRNG,src);
-  LatticeFermion result(&Grid); result=zero;
+  LatticeFermion result(&Grid); result=Zero();
   LatticeFermion resid(&Grid); 
 
   RealD mass=0.5;
diff --git a/tests/solver/Test_wilson_cg_unprec.cc b/tests/solver/Test_wilson_cg_unprec.cc
index db227ec8..f3335d45 100644
--- a/tests/solver/Test_wilson_cg_unprec.cc
+++ b/tests/solver/Test_wilson_cg_unprec.cc
@@ -30,7 +30,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -48,9 +48,9 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -59,7 +59,7 @@ int main (int argc, char ** argv)
 
   LatticeFermion src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  LatticeFermion result(&Grid); result=zero;
+  LatticeFermion result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_wilson_cr_unprec.cc b/tests/solver/Test_wilson_cr_unprec.cc
index eccd7e74..1f3ae35d 100644
--- a/tests/solver/Test_wilson_cr_unprec.cc
+++ b/tests/solver/Test_wilson_cr_unprec.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template<class d>
 struct scal {
@@ -48,9 +48,9 @@ int main (int argc, char ** argv)
   Grid_init(&argc,&argv);
 
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -59,7 +59,7 @@ int main (int argc, char ** argv)
 
   LatticeFermion src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  LatticeFermion result(&Grid); result=zero;
+  LatticeFermion result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   std::vector<LatticeColourMatrix> U(4,&Grid);
diff --git a/tests/solver/Test_wilson_fcagmres_prec.cc b/tests/solver/Test_wilson_fcagmres_prec.cc
index f802984f..b821a25f 100644
--- a/tests/solver/Test_wilson_fcagmres_prec.cc
+++ b/tests/solver/Test_wilson_fcagmres_prec.cc
@@ -29,15 +29,15 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -46,7 +46,7 @@ int main (int argc, char ** argv)
 
   LatticeFermion src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  LatticeFermion result(&Grid); result=zero;
+  LatticeFermion result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_wilson_fgmres_prec.cc b/tests/solver/Test_wilson_fgmres_prec.cc
index f55516da..cf1818ba 100644
--- a/tests/solver/Test_wilson_fgmres_prec.cc
+++ b/tests/solver/Test_wilson_fgmres_prec.cc
@@ -29,15 +29,15 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -46,7 +46,7 @@ int main (int argc, char ** argv)
 
   LatticeFermion src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  LatticeFermion result(&Grid); result=zero;
+  LatticeFermion result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_wilson_gmres_unprec.cc b/tests/solver/Test_wilson_gmres_unprec.cc
index 443f7ebc..7d788391 100644
--- a/tests/solver/Test_wilson_gmres_unprec.cc
+++ b/tests/solver/Test_wilson_gmres_unprec.cc
@@ -29,15 +29,15 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -46,7 +46,7 @@ int main (int argc, char ** argv)
 
   LatticeFermion src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  LatticeFermion result(&Grid); result=zero;
+  LatticeFermion result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_wilson_mg.cc b/tests/solver/Test_wilson_mg.cc
index 1609c1fc..c022590c 100644
--- a/tests/solver/Test_wilson_mg.cc
+++ b/tests/solver/Test_wilson_mg.cc
@@ -31,7 +31,6 @@
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
 
 int main(int argc, char **argv) {
 
@@ -46,7 +45,7 @@ int main(int argc, char **argv) {
 
   // clang-format off
   LatticeFermion    src(FGrid); gaussian(fPRNG, src);
-  LatticeFermion result(FGrid); result = zero;
+  LatticeFermion result(FGrid); result = Zero();
   LatticeGaugeField Umu(FGrid); SU3::HotConfiguration(fPRNG, Umu);
   // clang-format on
 
diff --git a/tests/solver/Test_wilson_mg_mp.cc b/tests/solver/Test_wilson_mg_mp.cc
index 0cd51227..e631cd15 100644
--- a/tests/solver/Test_wilson_mg_mp.cc
+++ b/tests/solver/Test_wilson_mg_mp.cc
@@ -31,7 +31,7 @@
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main(int argc, char **argv) {
 
@@ -50,8 +50,8 @@ int main(int argc, char **argv) {
 
   // clang-format off
   LatticeFermionD       src_d(FGrid_d); gaussian(fPRNG, src_d);
-  LatticeFermionD resultMGD_d(FGrid_d); resultMGD_d = zero;
-  LatticeFermionD resultMGF_d(FGrid_d); resultMGF_d = zero;
+  LatticeFermionD resultMGD_d(FGrid_d); resultMGD_d = Zero();
+  LatticeFermionD resultMGF_d(FGrid_d); resultMGF_d = Zero();
   LatticeGaugeFieldD    Umu_d(FGrid_d); SU3::HotConfiguration(fPRNG, Umu_d);
   LatticeGaugeFieldF    Umu_f(FGrid_f); precisionChange(Umu_f, Umu_d);
   // clang-format on
@@ -141,8 +141,8 @@ int main(int argc, char **argv) {
 
     // clang-format off
     LatticeFermionF src_f(FGrid_f);    precisionChange(src_f, src_d);
-    LatticeFermionF resMGF_f(FGrid_f); resMGF_f = zero;
-    LatticeFermionD resMGD_d(FGrid_d); resMGD_d = zero;
+    LatticeFermionF resMGF_f(FGrid_f); resMGF_f = Zero();
+    LatticeFermionD resMGD_d(FGrid_d); resMGD_d = Zero();
     // clang-format on
 
     (*MGPreconDw_f)(src_f, resMGF_f);
diff --git a/tests/solver/Test_wilson_mr_unprec.cc b/tests/solver/Test_wilson_mr_unprec.cc
index 976130d3..1cc1f418 100644
--- a/tests/solver/Test_wilson_mr_unprec.cc
+++ b/tests/solver/Test_wilson_mr_unprec.cc
@@ -29,15 +29,15 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -46,7 +46,7 @@ int main (int argc, char ** argv)
 
   LatticeFermion src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  LatticeFermion result(&Grid); result=zero;
+  LatticeFermion result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_wilsonclover_cagmres_unprec.cc b/tests/solver/Test_wilsonclover_cagmres_unprec.cc
index 3ecdf738..e755dcd2 100644
--- a/tests/solver/Test_wilsonclover_cagmres_unprec.cc
+++ b/tests/solver/Test_wilsonclover_cagmres_unprec.cc
@@ -29,15 +29,15 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -50,7 +50,7 @@ int main (int argc, char ** argv)
 
   FermionField src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  FermionField result(&Grid); result=zero;
+  FermionField result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_wilsonclover_fcagmres_prec.cc b/tests/solver/Test_wilsonclover_fcagmres_prec.cc
index 3cbbfc02..95fe890e 100644
--- a/tests/solver/Test_wilsonclover_fcagmres_prec.cc
+++ b/tests/solver/Test_wilsonclover_fcagmres_prec.cc
@@ -29,15 +29,15 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -50,7 +50,7 @@ int main (int argc, char ** argv)
 
   FermionField src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  FermionField result(&Grid); result=zero;
+  FermionField result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_wilsonclover_fgmres_prec.cc b/tests/solver/Test_wilsonclover_fgmres_prec.cc
index 7ad0fa24..daea527a 100644
--- a/tests/solver/Test_wilsonclover_fgmres_prec.cc
+++ b/tests/solver/Test_wilsonclover_fgmres_prec.cc
@@ -29,15 +29,15 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -50,7 +50,7 @@ int main (int argc, char ** argv)
 
   FermionField src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  FermionField result(&Grid); result=zero;
+  FermionField result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_wilsonclover_gmres_unprec.cc b/tests/solver/Test_wilsonclover_gmres_unprec.cc
index a9fe7181..667d17d9 100644
--- a/tests/solver/Test_wilsonclover_gmres_unprec.cc
+++ b/tests/solver/Test_wilsonclover_gmres_unprec.cc
@@ -29,15 +29,15 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -50,7 +50,7 @@ int main (int argc, char ** argv)
 
   FermionField src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  FermionField result(&Grid); result=zero;
+  FermionField result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_wilsonclover_mg.cc b/tests/solver/Test_wilsonclover_mg.cc
index e749aacb..a5fa907d 100644
--- a/tests/solver/Test_wilsonclover_mg.cc
+++ b/tests/solver/Test_wilsonclover_mg.cc
@@ -31,7 +31,7 @@
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main(int argc, char **argv) {
 
diff --git a/tests/solver/Test_wilsonclover_mg_mp.cc b/tests/solver/Test_wilsonclover_mg_mp.cc
index d9ed1d33..b5178d2e 100644
--- a/tests/solver/Test_wilsonclover_mg_mp.cc
+++ b/tests/solver/Test_wilsonclover_mg_mp.cc
@@ -31,7 +31,7 @@
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main(int argc, char **argv) {
 
diff --git a/tests/solver/Test_wilsonclover_mr_unprec.cc b/tests/solver/Test_wilsonclover_mr_unprec.cc
index e3aa8838..c7b5ecfe 100644
--- a/tests/solver/Test_wilsonclover_mr_unprec.cc
+++ b/tests/solver/Test_wilsonclover_mr_unprec.cc
@@ -29,15 +29,15 @@ directory
 #include <Grid/Grid.h>
 
 using namespace Grid;
-using namespace Grid::QCD;
+ 
 
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  std::vector<int> latt_size   = GridDefaultLatt();
-  std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
-  std::vector<int> mpi_layout  = GridDefaultMpi();
+  Coordinate latt_size   = GridDefaultLatt();
+  Coordinate simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
+  Coordinate mpi_layout  = GridDefaultMpi();
   GridCartesian               Grid(latt_size,simd_layout,mpi_layout);
   GridRedBlackCartesian     RBGrid(&Grid);
 
@@ -50,7 +50,7 @@ int main (int argc, char ** argv)
 
   FermionField src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  FermionField result(&Grid); result=zero;
+  FermionField result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU3::HotConfiguration(pRNG,Umu);
 
   double volume=1;
diff --git a/tests/solver/Test_zmobius_cg_prec.cc b/tests/solver/Test_zmobius_cg_prec.cc
index 922ca8ae..fb57cff1 100644
--- a/tests/solver/Test_zmobius_cg_prec.cc
+++ b/tests/solver/Test_zmobius_cg_prec.cc
@@ -30,7 +30,7 @@ directory
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 template <class d>
 struct scal {
@@ -64,7 +64,7 @@ int main(int argc, char** argv) {
   LatticeFermion src(FGrid);
   random(RNG5, src);
   LatticeFermion result(FGrid);
-  result = zero;
+  result = Zero();
   LatticeGaugeField Umu(UGrid);
 
   SU3::HotConfiguration(RNG4, Umu);
@@ -79,7 +79,7 @@ int main(int argc, char** argv) {
 
   RealD mass = 0.01;
   RealD M5 = 1.8;
-  std::vector < std::complex<double>  > omegas;
+  std::vector < ComplexD  > omegas;
 #if 0
   for(int i=0;i<Ls;i++){
     double imag = 0.;
@@ -106,7 +106,7 @@ int main(int argc, char** argv) {
   LatticeFermion src_o(FrbGrid);
   LatticeFermion result_o(FrbGrid);
   pickCheckerboard(Odd, src_o, src);
-  result_o = zero;
+  result_o = Zero();
 
   GridStopWatch CGTimer;
 
diff --git a/tests/testu01/Test_smallcrush.cc b/tests/testu01/Test_smallcrush.cc
index d09cd577..ca3a9464 100644
--- a/tests/testu01/Test_smallcrush.cc
+++ b/tests/testu01/Test_smallcrush.cc
@@ -29,7 +29,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-using namespace Grid::QCD;
+ ;
 
 // Wrap Grid's parallel RNG for testU01
 #undef BIG_CRUSH             // Big crush enable (long running)