Switching Block lanczos precision to explicitly single

Adding sample run script and input file
Checking block lanczos
2025-10-30 19:44:32 +00:00 · 2022-03-08 10:18:36 -08:00 · 2021-09-05 23:04:41 -04:00 · 2021-09-03 17:38:10 -04:00 · 2020-10-01 15:35:29 -04:00 · 2020-10-01 15:35:03 -04:00
202 changed files with 3528 additions and 5612 deletions
--- a/.gitignore
+++ b/.gitignore
@@ -10,6 +10,8 @@
 *~
 *#
 *.sublime-*
+.ctags
+tags

 # Precompiled Headers #
 #######################
@@ -88,7 +90,6 @@ Thumbs.db
 # build directory #
 ###################
 build*/*
-Documentation/_build

 # IDE related files #
 #####################
--- a/Grid/algorithms/CoarsenedMatrix.h
+++ b/Grid/algorithms/CoarsenedMatrix.h
@@ -358,7 +358,7 @@ public:
    autoView( in_v , in, AcceleratorRead);
    autoView( out_v , out, AcceleratorWrite);
    autoView( Stencil_v  , Stencil, AcceleratorRead);
-    int npoint = geom.npoint;
+    auto& geom_v = geom;
    typedef LatticeView<Cobj> Aview;
      
    Vector<Aview> AcceleratorViewContainer;
@@ -380,7 +380,7 @@ public:
      int ptype;
      StencilEntry *SE;

-      for(int point=0;point<npoint;point++){
+      for(int point=0;point<geom_v.npoint;point++){

 	SE=Stencil_v.GetEntry(ptype,point,ss);
 	  
@@ -424,7 +424,7 @@ public:
    autoView( in_v , in, AcceleratorRead);
    autoView( out_v , out, AcceleratorWrite);
    autoView( Stencil_v  , Stencil, AcceleratorRead);
-    int npoint = geom.npoint;
+    auto& geom_v = geom;
    typedef LatticeView<Cobj> Aview;

    Vector<Aview> AcceleratorViewContainer;
@@ -442,8 +442,6 @@ public:
    for(int p=0; p<geom.npoint; p++)
      points[p] = geom.points_dagger[p];

-    auto points_p = &points[0];
-
    RealD* dag_factor_p = &dag_factor[0];

    accelerator_for(sss, Grid()->oSites()*nbasis, Nsimd, {
@@ -454,8 +452,8 @@ public:
      int ptype;
      StencilEntry *SE;

-      for(int p=0;p<npoint;p++){
-        int point = points_p[p];
+      for(int p=0;p<geom_v.npoint;p++){
+        int point = points[p];

 	SE=Stencil_v.GetEntry(ptype,point,ss);

@@ -710,8 +708,6 @@ public:
    for(int p=0; p<npoint; p++)
      points[p] = (dag && !hermitian) ? geom.points_dagger[p] : p;

-    auto points_p = &points[0];
-
    Vector<Aview> AcceleratorViewContainer;
    for(int p=0;p<npoint;p++) AcceleratorViewContainer.push_back(a[p].View(AcceleratorRead));
    Aview *Aview_p = & AcceleratorViewContainer[0];
@@ -732,7 +728,7 @@ public:
        StencilEntry *SE;

        for(int p=0;p<npoint;p++){
-          int point = points_p[p];
+          int point = points[p];
          SE=st_v.GetEntry(ptype,point,ss);

          if(SE->_is_local) {
@@ -758,7 +754,7 @@ public:
        StencilEntry *SE;

        for(int p=0;p<npoint;p++){
-          int point = points_p[p];
+          int point = points[p];
          SE=st_v.GetEntry(ptype,point,ss);

          if(SE->_is_local) {
--- a/Grid/algorithms/FFT.h
+++ b/Grid/algorithms/FFT.h
@@ -136,7 +136,7 @@ public:
    flops=0;
    usec =0;
    Coordinate layout(Nd,1);
-    sgrid = new GridCartesian(dimensions,layout,processors,*grid);
+    sgrid = new GridCartesian(dimensions,layout,processors);
  };
    
  ~FFT ( void)  {
@@ -182,7 +182,7 @@ public:
    pencil_gd[dim] = G*processors[dim];
      
    // Pencil global vol LxLxGxLxL per node
-    GridCartesian pencil_g(pencil_gd,layout,processors,*vgrid);
+    GridCartesian pencil_g(pencil_gd,layout,processors);
      
    // Construct pencils
    typedef typename vobj::scalar_object sobj;
--- a/Grid/algorithms/LinearOperator.h
+++ b/Grid/algorithms/LinearOperator.h
@@ -52,7 +52,6 @@ public:
  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 ~LinearOperatorBase(){};
 };


@@ -508,7 +507,7 @@ class SchurStaggeredOperator :  public SchurOperatorBase<Field> {
  virtual  void MpcDag   (const Field &in, Field &out){
    Mpc(in,out);
  }
-  virtual void MpcDagMpc(const Field &in, Field &out) {
+  virtual void MpcDagMpc(const Field &in, Field &out,RealD &ni,RealD &no) {
    assert(0);// Never need with staggered
  }
 };
@@ -531,16 +530,6 @@ public:
 template<class Field> class LinearFunction {
 public:
  virtual void operator() (const Field &in, Field &out) = 0;
-
-  virtual void operator() (const std::vector<Field> &in, std::vector<Field> &out)
-  {
-    assert(in.size() == out.size());
-
-    for (unsigned int i = 0; i < in.size(); ++i)
-    {
-      (*this)(in[i], out[i]);
-    }
-  }
 };

 template<class Field> class IdentityLinearFunction : public LinearFunction<Field> {
@@ -586,7 +575,6 @@ class HermOpOperatorFunction : public OperatorFunction<Field> {
 template<typename Field>
 class PlainHermOp : public LinearFunction<Field> {
 public:
-  using LinearFunction<Field>::operator();
  LinearOperatorBase<Field> &_Linop;
      
  PlainHermOp(LinearOperatorBase<Field>& linop) : _Linop(linop) 
@@ -600,7 +588,6 @@ public:
 template<typename Field>
 class FunctionHermOp : public LinearFunction<Field> {
 public:
-  using LinearFunction<Field>::operator(); 
  OperatorFunction<Field>   & _poly;
  LinearOperatorBase<Field> &_Linop;
      
--- a/Grid/algorithms/Preconditioner.h
+++ b/Grid/algorithms/Preconditioner.h
@@ -30,19 +30,13 @@ Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>

 NAMESPACE_BEGIN(Grid);

-template<class Field> using Preconditioner =  LinearFunction<Field> ;
-
-/*
-template<class Field> class Preconditioner :  public LinearFunction<Field> {
-  using LinearFunction<Field>::operator();
+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:
-  using Preconditioner<Field>::operator();
-  virtual void operator()(const Field &src, Field & psi){
+  void operator()(const Field &src, Field & psi){
    psi = src;
  }
  TrivialPrecon(void){};
--- a/Grid/algorithms/SparseMatrix.h
+++ b/Grid/algorithms/SparseMatrix.h
@@ -48,7 +48,6 @@ public:
  virtual  void Mdiag    (const Field &in, Field &out)=0;
  virtual  void Mdir     (const Field &in, Field &out,int dir, int disp)=0;
  virtual  void MdirAll  (const Field &in, std::vector<Field> &out)=0;
-  virtual ~SparseMatrixBase() {};
 };

 /////////////////////////////////////////////////////////////////////////////////////////////
@@ -73,7 +72,7 @@ public:
  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 ~CheckerBoardedSparseMatrixBase() {};
+
 };

 NAMESPACE_END(Grid);
--- a/Grid/algorithms/approx/Chebyshev.h
+++ b/Grid/algorithms/approx/Chebyshev.h
@@ -240,12 +240,14 @@ public:
    Field T0(grid); T0 = in;  
    Field T1(grid); 
    Field T2(grid);
+    Field Tout(grid);
    Field y(grid);
      
    Field *Tnm = &T0;
    Field *Tn  = &T1;
    Field *Tnp = &T2;

+    std::cout << GridLogMessage << "Chebyshev() starts"<<std::endl;
    // Tn=T1 = (xscale M + mscale)in
    RealD xscale = 2.0/(hi-lo);
    RealD mscale = -(hi+lo)/(hi-lo);
@@ -254,7 +256,7 @@ public:

    // sum = .5 c[0] T0 + c[1] T1
    //    out = ()*T0 + Coeffs[1]*T1;
-    axpby(out,0.5*Coeffs[0],Coeffs[1],T0,T1);
+    axpby(Tout,0.5*Coeffs[0],Coeffs[1],T0,T1);
    for(int n=2;n<order;n++){

      Linop.HermOp(*Tn,y);
@@ -275,7 +277,7 @@ public:
      axpby(y,xscale,mscale,y,(*Tn));
      axpby(*Tnp,2.0,-1.0,y,(*Tnm));
      if ( Coeffs[n] != 0.0) {
-	axpy(out,Coeffs[n],*Tnp,out);
+	axpy(Tout,Coeffs[n],*Tnp,Tout);
      }
 #endif
      // Cycle pointers to avoid copies
@@ -285,6 +287,8 @@ public:
      Tnp    =swizzle;
 	  
    }
+    out = Tout;
+    std::cout << GridLogMessage << "Chebyshev() ends"<<std::endl;
  }
 };

@@ -377,24 +381,26 @@ public:
    Field T0(grid); T0 = in;  
    Field T1(grid); 
    Field T2(grid);
+    Field Tout(grid);
    Field  y(grid);
      
    Field *Tnm = &T0;
    Field *Tn  = &T1;
    Field *Tnp = &T2;

+    std::cout << GridLogMessage << "ChebyshevLanczos() starts"<<std::endl;
    // 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;
+    Tout = (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);
+      Tout=Tout+Coeffs[n]* (*Tnp);

      // Cycle pointers to avoid copies
      Field *swizzle = Tnm;
@@ -403,6 +409,8 @@ public:
      Tnp    =swizzle;
 	  
    }
+    out=Tout;
+    std::cout << GridLogMessage << "ChebyshevLanczos() ends"<<std::endl;
  }
 };
 NAMESPACE_END(Grid);
--- a/Grid/algorithms/iterative/BiCGSTABMixedPrec.h
+++ b/Grid/algorithms/iterative/BiCGSTABMixedPrec.h
@@ -36,8 +36,7 @@ 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 MixedPrecisionBiCGSTAB : public LinearFunction<FieldD> 
 {
-  public:
-    using LinearFunction<FieldD>::operator();
+  public:                                                
    RealD   Tolerance;
    RealD   InnerTolerance; // Initial tolerance for inner CG. Defaults to Tolerance but can be changed
    Integer MaxInnerIterations;
--- a/Grid/algorithms/iterative/ConjugateGradientMixedPrec.h
+++ b/Grid/algorithms/iterative/ConjugateGradientMixedPrec.h
@@ -35,8 +35,7 @@ NAMESPACE_BEGIN(Grid);
    typename std::enable_if< getPrecision<FieldD>::value == 2, int>::type = 0,
    typename std::enable_if< getPrecision<FieldF>::value == 1, int>::type = 0> 
  class MixedPrecisionConjugateGradient : public LinearFunction<FieldD> {
-  public:
-    using LinearFunction<FieldD>::operator();
+  public:                                                
    RealD   Tolerance;
    RealD   InnerTolerance; //Initial tolerance for inner CG. Defaults to Tolerance but can be changed
    Integer MaxInnerIterations;
--- a/Grid/algorithms/iterative/Deflation.h
+++ b/Grid/algorithms/iterative/Deflation.h
@@ -33,19 +33,16 @@ namespace Grid {
 template<class Field>
 class ZeroGuesser: public LinearFunction<Field> {
 public:
-  using LinearFunction<Field>::operator();
    virtual void operator()(const Field &src, Field &guess) { guess = Zero(); };
 };
 template<class Field>
 class DoNothingGuesser: public LinearFunction<Field> {
 public:
-  using LinearFunction<Field>::operator();
  virtual void operator()(const Field &src, Field &guess) {  };
 };
 template<class Field>
 class SourceGuesser: public LinearFunction<Field> {
 public:
-  using LinearFunction<Field>::operator();
  virtual void operator()(const Field &src, Field &guess) { guess = src; };
 };

@@ -57,24 +54,15 @@ class DeflatedGuesser: public LinearFunction<Field> {
 private:
  const std::vector<Field> &evec;
  const std::vector<RealD> &eval;
-  const unsigned int       N;

 public:
-  using LinearFunction<Field>::operator();

-  DeflatedGuesser(const std::vector<Field> & _evec,const std::vector<RealD> & _eval)
-  : DeflatedGuesser(_evec, _eval, _evec.size())
-  {}
-
-  DeflatedGuesser(const std::vector<Field> & _evec, const std::vector<RealD> & _eval, const unsigned int _N)
-  : evec(_evec), eval(_eval), N(_N)
-  {
-    assert(evec.size()==eval.size());
-    assert(N <= evec.size());
-  } 
+  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();
+    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);
@@ -91,7 +79,6 @@ private:
  const std::vector<RealD>       &eval_coarse;
 public:
  
-  using LinearFunction<FineField>::operator();
  LocalCoherenceDeflatedGuesser(const std::vector<FineField>   &_subspace,
 				const std::vector<CoarseField> &_evec_coarse,
 				const std::vector<RealD>       &_eval_coarse)
--- a/Grid/algorithms/iterative/ImplicitlyRestartedBlockLanczos.h
+++ b/Grid/algorithms/iterative/ImplicitlyRestartedBlockLanczos.h
--- a/Grid/algorithms/iterative/LocalCoherenceLanczos.h
+++ b/Grid/algorithms/iterative/LocalCoherenceLanczos.h
@@ -67,7 +67,6 @@ public:
 template<class Fobj,class CComplex,int nbasis>
 class ProjectedHermOp : public LinearFunction<Lattice<iVector<CComplex,nbasis > > > {
 public:
-  using LinearFunction<Lattice<iVector<CComplex,nbasis > > >::operator();
  typedef iVector<CComplex,nbasis >           CoarseSiteVector;
  typedef Lattice<CoarseSiteVector>           CoarseField;
  typedef Lattice<CComplex>   CoarseScalar; // used for inner products on fine field
@@ -98,7 +97,6 @@ public:
 template<class Fobj,class CComplex,int nbasis>
 class ProjectedFunctionHermOp : public LinearFunction<Lattice<iVector<CComplex,nbasis > > > {
 public:
-  using LinearFunction<Lattice<iVector<CComplex,nbasis > > >::operator();
  typedef iVector<CComplex,nbasis >           CoarseSiteVector;
  typedef Lattice<CoarseSiteVector>           CoarseField;
  typedef Lattice<CComplex>   CoarseScalar; // used for inner products on fine field
--- a/Grid/algorithms/iterative/PrecGeneralisedConjugateResidual.h
+++ b/Grid/algorithms/iterative/PrecGeneralisedConjugateResidual.h
@@ -43,7 +43,7 @@ NAMESPACE_BEGIN(Grid);
 template<class Field>
 class PrecGeneralisedConjugateResidual : public LinearFunction<Field> {
 public:                                                
-  using LinearFunction<Field>::operator();
+
  RealD   Tolerance;
  Integer MaxIterations;
  int verbose;
--- a/Grid/algorithms/iterative/PrecGeneralisedConjugateResidualNonHermitian.h
+++ b/Grid/algorithms/iterative/PrecGeneralisedConjugateResidualNonHermitian.h
@@ -43,7 +43,7 @@ NAMESPACE_BEGIN(Grid);
 template<class Field>
 class PrecGeneralisedConjugateResidualNonHermitian : public LinearFunction<Field> {
 public:                                                
-  using LinearFunction<Field>::operator();
+
  RealD   Tolerance;
  Integer MaxIterations;
  int verbose;
@@ -119,8 +119,7 @@ public:
  RealD GCRnStep(const Field &src, Field &psi,RealD rsq){

    RealD cp;
-    ComplexD a, b;
-    //    ComplexD zAz;
+    ComplexD a, b, zAz;
    RealD zAAz;
    ComplexD rq;

@@ -147,7 +146,7 @@ public:
    //////////////////////////////////
    MatTimer.Start();
    Linop.Op(psi,Az);
-    //    zAz = innerProduct(Az,psi);
+    zAz = innerProduct(Az,psi);
    zAAz= norm2(Az);
    MatTimer.Stop();
    
@@ -171,7 +170,7 @@ public:

    LinalgTimer.Start();

-    //    zAz = innerProduct(Az,psi);
+    zAz = innerProduct(Az,psi);
    zAAz= norm2(Az);

    //p[0],q[0],qq[0] 
@@ -213,7 +212,7 @@ public:
      MatTimer.Start();
      Linop.Op(z,Az);
      MatTimer.Stop();
-      //      zAz = innerProduct(Az,psi);
+      zAz = innerProduct(Az,psi);
      zAAz= norm2(Az);

      LinalgTimer.Start();
--- a/Grid/algorithms/iterative/SchurRedBlack.h
+++ b/Grid/algorithms/iterative/SchurRedBlack.h
@@ -132,31 +132,6 @@ namespace Grid {
      (*this)(_Matrix,in,out,guess);
    }

-    void RedBlackSource(Matrix &_Matrix, const std::vector<Field> &in, std::vector<Field> &src_o) 
-    {
-      GridBase *grid = _Matrix.RedBlackGrid();
-      Field tmp(grid);
-      int nblock = in.size();
-      for(int b=0;b<nblock;b++){
-	RedBlackSource(_Matrix,in[b],tmp,src_o[b]);
-      }
-    }
-    // James can write his own deflated guesser
-    // with optimised code for the inner products
-    //    RedBlackSolveSplitGrid();
-    //    RedBlackSolve(_Matrix,src_o,sol_o); 
-
-    void RedBlackSolution(Matrix &_Matrix, const std::vector<Field> &in, const std::vector<Field> &sol_o, std::vector<Field> &out)
-    {
-      GridBase *grid = _Matrix.RedBlackGrid();
-      Field tmp(grid);
-      int nblock = in.size();
-      for(int b=0;b<nblock;b++) {
-	pickCheckerboard(Even,tmp,in[b]);
-	RedBlackSolution(_Matrix,sol_o[b],tmp,out[b]);
-      }
-    }
-
    template<class Guesser>
    void operator()(Matrix &_Matrix, const std::vector<Field> &in, std::vector<Field> &out,Guesser &guess) 
    {
@@ -175,29 +150,24 @@ namespace Grid {
      ////////////////////////////////////////////////
      // Prepare RedBlack source
      ////////////////////////////////////////////////
-      RedBlackSource(_Matrix,in,src_o);
-	//      for(int b=0;b<nblock;b++){
-	//	RedBlackSource(_Matrix,in[b],tmp,src_o[b]);
-	//      }
-      
+      for(int b=0;b<nblock;b++){
+	RedBlackSource(_Matrix,in[b],tmp,src_o[b]);
+      }
      ////////////////////////////////////////////////
      // Make the guesses
      ////////////////////////////////////////////////
      if ( subGuess ) guess_save.resize(nblock,grid);

-      
-      if(useSolnAsInitGuess) {
-        for(int b=0;b<nblock;b++){
+      for(int b=0;b<nblock;b++){
+        if(useSolnAsInitGuess) {
          pickCheckerboard(Odd, sol_o[b], out[b]);
+        } else {
+          guess(src_o[b],sol_o[b]); 
        }
-      } else {
-        guess(src_o, sol_o); 
-      }

-	    if ( subGuess ) { 
-        for(int b=0;b<nblock;b++){
-          guess_save[b] = sol_o[b];
-        }
+	if ( subGuess ) { 
+	  guess_save[b] = sol_o[b];
+	}
      }
      //////////////////////////////////////////////////////////////
      // Call the block solver
--- a/Grid/allocator/MemoryManager.cc
+++ b/Grid/allocator/MemoryManager.cc
@@ -9,30 +9,14 @@ NAMESPACE_BEGIN(Grid);
 #define AccSmall (3)
 #define Shared   (4)
 #define SharedSmall (5)
-#undef GRID_MM_VERBOSE 
 uint64_t total_shared;
 uint64_t total_device;
 uint64_t total_host;;
 void MemoryManager::PrintBytes(void)
 {
-  std::cout << " MemoryManager : ------------------------------------ "<<std::endl;
-  std::cout << " MemoryManager : PrintBytes "<<std::endl;
-  std::cout << " MemoryManager : ------------------------------------ "<<std::endl;
-  std::cout << " MemoryManager : "<<(total_shared>>20)<<" shared      Mbytes "<<std::endl;
-  std::cout << " MemoryManager : "<<(total_device>>20)<<" accelerator Mbytes "<<std::endl;
-  std::cout << " MemoryManager : "<<(total_host>>20)  <<" cpu         Mbytes "<<std::endl;
-  uint64_t cacheBytes;
-  cacheBytes = CacheBytes[Cpu];
-  std::cout << " MemoryManager : "<<(cacheBytes>>20) <<" cpu cache Mbytes "<<std::endl;
-  cacheBytes = CacheBytes[Acc];
-  std::cout << " MemoryManager : "<<(cacheBytes>>20) <<" acc cache Mbytes "<<std::endl;
-  cacheBytes = CacheBytes[Shared];
-  std::cout << " MemoryManager : "<<(cacheBytes>>20) <<" shared cache Mbytes "<<std::endl;
-  
-#ifdef GRID_CUDA
-  cuda_mem();
-#endif
-  
+  std::cout << " MemoryManager : "<<total_shared<<" shared      bytes "<<std::endl;
+  std::cout << " MemoryManager : "<<total_device<<" accelerator bytes "<<std::endl;
+  std::cout << " MemoryManager : "<<total_host  <<" cpu         bytes "<<std::endl;
 }

 //////////////////////////////////////////////////////////////////////
@@ -40,114 +24,88 @@ void MemoryManager::PrintBytes(void)
 //////////////////////////////////////////////////////////////////////
 MemoryManager::AllocationCacheEntry MemoryManager::Entries[MemoryManager::NallocType][MemoryManager::NallocCacheMax];
 int MemoryManager::Victim[MemoryManager::NallocType];
-int MemoryManager::Ncache[MemoryManager::NallocType] = { 2, 8, 2, 8, 2, 8 };
-uint64_t MemoryManager::CacheBytes[MemoryManager::NallocType];
+int MemoryManager::Ncache[MemoryManager::NallocType] = { 8, 32, 8, 32, 8, 32 };
+
 //////////////////////////////////////////////////////////////////////
 // Actual allocation and deallocation utils
 //////////////////////////////////////////////////////////////////////
 void *MemoryManager::AcceleratorAllocate(size_t bytes)
 {
-  total_device+=bytes;
  void *ptr = (void *) Lookup(bytes,Acc);
  if ( ptr == (void *) NULL ) {
    ptr = (void *) acceleratorAllocDevice(bytes);
+    total_device+=bytes;
  }
-#ifdef GRID_MM_VERBOSE
-  std::cout <<"AcceleratorAllocate "<<std::endl;
-  PrintBytes();
-#endif
  return ptr;
 }
 void  MemoryManager::AcceleratorFree    (void *ptr,size_t bytes)
 {
-  total_device-=bytes;
  void *__freeme = Insert(ptr,bytes,Acc);
  if ( __freeme ) {
    acceleratorFreeDevice(__freeme);
+    total_device-=bytes;
+//       PrintBytes();
  }
-#ifdef GRID_MM_VERBOSE
-  std::cout <<"AcceleratorFree "<<std::endl;
-  PrintBytes();
-#endif
 }
 void *MemoryManager::SharedAllocate(size_t bytes)
 {
-  total_shared+=bytes;
  void *ptr = (void *) Lookup(bytes,Shared);
  if ( ptr == (void *) NULL ) {
    ptr = (void *) acceleratorAllocShared(bytes);
+    total_shared+=bytes;
+        std::cout <<GridLogMessage<<"AcceleratorAllocate: allocated Shared pointer "<<std::hex<<ptr<<std::dec<<std::endl;
+//        PrintBytes();
  }
-#ifdef GRID_MM_VERBOSE
-  std::cout <<"SharedAllocate "<<std::endl;
-  PrintBytes();
-#endif
  return ptr;
 }
 void  MemoryManager::SharedFree    (void *ptr,size_t bytes)
 {
-  total_shared-=bytes;
  void *__freeme = Insert(ptr,bytes,Shared);
  if ( __freeme ) {
    acceleratorFreeShared(__freeme);
+    total_shared-=bytes;
+    //    PrintBytes();
  }
-#ifdef GRID_MM_VERBOSE
-  std::cout <<"SharedFree "<<std::endl;
-  PrintBytes();
-#endif
 }
 #ifdef GRID_UVM
 void *MemoryManager::CpuAllocate(size_t bytes)
 {
-  total_host+=bytes;
  void *ptr = (void *) Lookup(bytes,Cpu);
  if ( ptr == (void *) NULL ) {
    ptr = (void *) acceleratorAllocShared(bytes);
+    total_host+=bytes;
+//    std::cout << GridLogMessage<< "MemoryManager:: CpuAllocate  total_host= "<<total_host<<" "<< ptr << std::endl;
  }
-#ifdef GRID_MM_VERBOSE
-  std::cout <<"CpuAllocate "<<std::endl;
-  PrintBytes();
-#endif
  return ptr;
 }
 void  MemoryManager::CpuFree    (void *_ptr,size_t bytes)
 {
-  total_host-=bytes;
  NotifyDeletion(_ptr);
  void *__freeme = Insert(_ptr,bytes,Cpu);
  if ( __freeme ) { 
    acceleratorFreeShared(__freeme);
+//    std::cout << GridLogMessage<< "MemoryManager:: CpuFree  total_host= "<<total_host<<" "<< __freeme << std::endl;
+    total_host-=bytes;
  }
-#ifdef GRID_MM_VERBOSE
-  std::cout <<"CpuFree "<<std::endl;
-  PrintBytes();
-#endif
 }
 #else
 void *MemoryManager::CpuAllocate(size_t bytes)
 {
-  total_host+=bytes;
  void *ptr = (void *) Lookup(bytes,Cpu);
  if ( ptr == (void *) NULL ) {
    ptr = (void *) acceleratorAllocCpu(bytes);
+    total_host+=bytes;
  }
-#ifdef GRID_MM_VERBOSE
-  std::cout <<"CpuAllocate "<<std::endl;
-  PrintBytes();
-#endif
  return ptr;
 }
 void  MemoryManager::CpuFree    (void *_ptr,size_t bytes)
 {
-  total_host-=bytes;
  NotifyDeletion(_ptr);
  void *__freeme = Insert(_ptr,bytes,Cpu);
  if ( __freeme ) { 
    acceleratorFreeCpu(__freeme);
+    total_host-=bytes;
  }
-#ifdef GRID_MM_VERBOSE
-  std::cout <<"CpuFree "<<std::endl;
-  PrintBytes();
-#endif
 }
 #endif

@@ -159,6 +117,7 @@ void MemoryManager::Init(void)

  char * str;
  int Nc;
+  int NcS;
  
  str= getenv("GRID_ALLOC_NCACHE_LARGE");
  if ( str ) {
@@ -224,13 +183,13 @@ void *MemoryManager::Insert(void *ptr,size_t bytes,int type)
 #ifdef ALLOCATION_CACHE
  bool small = (bytes < GRID_ALLOC_SMALL_LIMIT);
  int cache = type + small;
-  return Insert(ptr,bytes,Entries[cache],Ncache[cache],Victim[cache],CacheBytes[cache]);  
+  return Insert(ptr,bytes,Entries[cache],Ncache[cache],Victim[cache]);  
 #else
  return ptr;
 #endif
 }

-void *MemoryManager::Insert(void *ptr,size_t bytes,AllocationCacheEntry *entries,int ncache,int &victim, uint64_t &cacheBytes) 
+void *MemoryManager::Insert(void *ptr,size_t bytes,AllocationCacheEntry *entries,int ncache,int &victim) 
 {
  assert(ncache>0);
 #ifdef GRID_OMP
@@ -254,7 +213,6 @@ void *MemoryManager::Insert(void *ptr,size_t bytes,AllocationCacheEntry *entries

  if ( entries[v].valid ) {
    ret = entries[v].address;
-    cacheBytes -= entries[v].bytes;
    entries[v].valid = 0;
    entries[v].address = NULL;
    entries[v].bytes = 0;
@@ -263,7 +221,6 @@ void *MemoryManager::Insert(void *ptr,size_t bytes,AllocationCacheEntry *entries
  entries[v].address=ptr;
  entries[v].bytes  =bytes;
  entries[v].valid  =1;
-  cacheBytes += bytes;

  return ret;
 }
@@ -273,13 +230,13 @@ void *MemoryManager::Lookup(size_t bytes,int type)
 #ifdef ALLOCATION_CACHE
  bool small = (bytes < GRID_ALLOC_SMALL_LIMIT);
  int cache = type+small;
-  return Lookup(bytes,Entries[cache],Ncache[cache],CacheBytes[cache]);
+  return Lookup(bytes,Entries[cache],Ncache[cache]);
 #else
  return NULL;
 #endif
 }

-void *MemoryManager::Lookup(size_t bytes,AllocationCacheEntry *entries,int ncache,uint64_t & cacheBytes) 
+void *MemoryManager::Lookup(size_t bytes,AllocationCacheEntry *entries,int ncache) 
 {
  assert(ncache>0);
 #ifdef GRID_OMP
@@ -288,7 +245,6 @@ void *MemoryManager::Lookup(size_t bytes,AllocationCacheEntry *entries,int ncach
  for(int e=0;e<ncache;e++){
    if ( entries[e].valid && ( entries[e].bytes == bytes ) ) {
      entries[e].valid = 0;
-      cacheBytes -= entries[e].bytes;
      return entries[e].address;
    }
  }
--- a/Grid/allocator/MemoryManager.h
+++ b/Grid/allocator/MemoryManager.h
@@ -82,15 +82,14 @@ private:
  static AllocationCacheEntry Entries[NallocType][NallocCacheMax];
  static int Victim[NallocType];
  static int Ncache[NallocType];
-  static uint64_t CacheBytes[NallocType];

  /////////////////////////////////////////////////
  // Free pool
  /////////////////////////////////////////////////
  static void *Insert(void *ptr,size_t bytes,int type) ;
  static void *Lookup(size_t bytes,int type) ;
-  static void *Insert(void *ptr,size_t bytes,AllocationCacheEntry *entries,int ncache,int &victim,uint64_t &cbytes) ;
-  static void *Lookup(size_t bytes,AllocationCacheEntry *entries,int ncache,uint64_t &cbytes) ;
+  static void *Insert(void *ptr,size_t bytes,AllocationCacheEntry *entries,int ncache,int &victim) ;
+  static void *Lookup(size_t bytes,AllocationCacheEntry *entries,int ncache) ;

  static void PrintBytes(void);
 public:
@@ -113,11 +112,6 @@ private:
  static uint64_t     DeviceToHostBytes;
  static uint64_t     HostToDeviceXfer;
  static uint64_t     DeviceToHostXfer;
-  
-  static uint64_t     DeviceAccesses;
-  static uint64_t     HostAccesses;
-  static uint64_t     DeviceAccessBytes;
-  static uint64_t     HostAccessBytes;
 
 private:
 #ifndef GRID_UVM
@@ -157,7 +151,6 @@ private:

  //  static void  LRUupdate(AcceleratorViewEntry &AccCache);
  static void  LRUinsert(AcceleratorViewEntry &AccCache);
-  static void  LRUinsertback(AcceleratorViewEntry &AccCache);
  static void  LRUremove(AcceleratorViewEntry &AccCache);
  
  // manage entries in the table
@@ -176,7 +169,6 @@ private:

 public:
  static void Print(void);
-  static void PrintState( void* CpuPtr);
  static int   isOpen   (void* CpuPtr);
  static void  ViewClose(void* CpuPtr,ViewMode mode);
  static void *ViewOpen (void* CpuPtr,size_t bytes,ViewMode mode,ViewAdvise hint);
--- a/Grid/allocator/MemoryManagerCache.cc
+++ b/Grid/allocator/MemoryManagerCache.cc
@@ -3,7 +3,7 @@

 #warning "Using explicit device memory copies"
 NAMESPACE_BEGIN(Grid);
-//#define dprintf(...) printf ( __VA_ARGS__ ); fflush(stdout);
+//define dprintf(...) printf ( __VA_ARGS__ ); fflush(stdout);
 #define dprintf(...)


@@ -23,11 +23,6 @@ uint64_t  MemoryManager::HostToDeviceBytes;
 uint64_t  MemoryManager::DeviceToHostBytes;
 uint64_t  MemoryManager::HostToDeviceXfer;
 uint64_t  MemoryManager::DeviceToHostXfer;
-uint64_t  MemoryManager::DeviceAccesses;
-uint64_t  MemoryManager::HostAccesses;
-uint64_t  MemoryManager::DeviceAccessBytes;
-uint64_t  MemoryManager::HostAccessBytes;
- 

 ////////////////////////////////////
 // Priority ordering for unlocked entries
@@ -91,14 +86,6 @@ void  MemoryManager::LRUinsert(AcceleratorViewEntry &AccCache)
  AccCache.LRU_valid = 1;
  DeviceLRUBytes+=AccCache.bytes;
 }
-void  MemoryManager::LRUinsertback(AcceleratorViewEntry &AccCache)
-{
-  assert(AccCache.LRU_valid==0);
-  LRU.push_back(AccCache.CpuPtr);
-  AccCache.LRU_entry = --LRU.end();
-  AccCache.LRU_valid = 1;
-  DeviceLRUBytes+=AccCache.bytes;
-}
 void  MemoryManager::LRUremove(AcceleratorViewEntry &AccCache)
 {
  assert(AccCache.LRU_valid==1);
@@ -142,7 +129,6 @@ void MemoryManager::Evict(AcceleratorViewEntry &AccCache)
  dprintf("MemoryManager: Evict(%llx) %llx\n",(uint64_t)AccCache.CpuPtr,(uint64_t)AccCache.AccPtr); 
  assert(AccCache.accLock==0);
  assert(AccCache.cpuLock==0);
-  
  if(AccCache.state==AccDirty) {
    Flush(AccCache);
  }
@@ -245,9 +231,6 @@ uint64_t MemoryManager::AcceleratorViewOpen(uint64_t CpuPtr,size_t bytes,ViewMod
    EntryCreate(CpuPtr,bytes,mode,hint);
  }

-  DeviceAccesses++;
-  DeviceAccessBytes+=bytes;
-
  auto AccCacheIterator = EntryLookup(CpuPtr);
  auto & AccCache = AccCacheIterator->second;
  if (!AccCache.AccPtr) {
@@ -366,10 +349,6 @@ void MemoryManager::CpuViewClose(uint64_t CpuPtr)
  assert(AccCache.accLock==0);

  AccCache.cpuLock--;
-
-  if(AccCache.cpuLock==0) {
-    LRUinsertback(AccCache);
-  }
 }
 /*
 *  Action  State   StateNext         Flush    Clone
@@ -392,9 +371,6 @@ uint64_t MemoryManager::CpuViewOpen(uint64_t CpuPtr,size_t bytes,ViewMode mode,V
    EntryCreate(CpuPtr,bytes,mode,transient);
  }

-  HostAccesses++;
-  HostAccessBytes+=bytes;
-  
  auto AccCacheIterator = EntryLookup(CpuPtr);
  auto & AccCache = AccCacheIterator->second;

@@ -440,12 +416,6 @@ uint64_t MemoryManager::CpuViewOpen(uint64_t CpuPtr,size_t bytes,ViewMode mode,V

  AccCache.transient= transient? EvictNext : 0;

-  // If view is opened on host remove from LRU
-  // Host close says evict next from device
-  if(AccCache.LRU_valid==1){
-    LRUremove(AccCache);
-  }
-  
  return AccCache.CpuPtr;
 }
 void  MemoryManager::NotifyDeletion(void *_ptr)
@@ -459,7 +429,6 @@ void  MemoryManager::NotifyDeletion(void *_ptr)
 }
 void  MemoryManager::Print(void)
 {
-  PrintBytes();
  std::cout << GridLogDebug << "--------------------------------------------" << std::endl;
  std::cout << GridLogDebug << "Memory Manager                             " << std::endl;
  std::cout << GridLogDebug << "--------------------------------------------" << std::endl;
@@ -504,32 +473,6 @@ int   MemoryManager::isOpen   (void* _CpuPtr)
  }
 }

-void MemoryManager::PrintState(void* _CpuPtr)
-{
-  uint64_t CpuPtr = (uint64_t)_CpuPtr;
-
-  if ( EntryPresent(CpuPtr) ){
-    auto AccCacheIterator = EntryLookup(CpuPtr);
-    auto & AccCache = AccCacheIterator->second;
-    std::string str;
-    if ( AccCache.state==Empty    ) str = std::string("Empty");
-    if ( AccCache.state==CpuDirty ) str = std::string("CpuDirty");
-    if ( AccCache.state==AccDirty ) str = std::string("AccDirty");
-    if ( AccCache.state==Consistent)str = std::string("Consistent");
-    if ( AccCache.state==EvictNext) str = std::string("EvictNext");
-
-    std::cout << GridLogMessage << "CpuAddr\t\tAccAddr\t\tState\t\tcpuLock\taccLock\tLRU_valid "<<std::endl;
-    std::cout << GridLogMessage << "0x"<<std::hex<<AccCache.CpuPtr<<std::dec
-    << "\t0x"<<std::hex<<AccCache.AccPtr<<std::dec<<"\t" <<str
-    << "\t" << AccCache.cpuLock
-    << "\t" << AccCache.accLock
-    << "\t" << AccCache.LRU_valid<<std::endl;
-
-  } else {
-    std::cout << GridLogMessage << "No Entry in AccCache table." << std::endl; 
-  }
-}
-
 NAMESPACE_END(Grid);

 #endif
--- a/Grid/allocator/MemoryManagerShared.cc
+++ b/Grid/allocator/MemoryManagerShared.cc
@@ -12,18 +12,10 @@ uint64_t  MemoryManager::HostToDeviceBytes;
 uint64_t  MemoryManager::DeviceToHostBytes;
 uint64_t  MemoryManager::HostToDeviceXfer;
 uint64_t  MemoryManager::DeviceToHostXfer;
-uint64_t  MemoryManager::DeviceAccesses;
-uint64_t  MemoryManager::HostAccesses;
-uint64_t  MemoryManager::DeviceAccessBytes;
-uint64_t  MemoryManager::HostAccessBytes;

 void  MemoryManager::ViewClose(void* AccPtr,ViewMode mode){};
 void *MemoryManager::ViewOpen(void* CpuPtr,size_t bytes,ViewMode mode,ViewAdvise hint){ return CpuPtr; };
 int   MemoryManager::isOpen   (void* CpuPtr) { return 0;}
-void  MemoryManager::PrintState(void* CpuPtr)
-{
-std::cout << GridLogMessage << "Host<->Device memory movement not currently managed by Grid." << std::endl;
-};
 void  MemoryManager::Print(void){};
 void  MemoryManager::NotifyDeletion(void *ptr){};

--- a/Grid/communicator/Communicator_base.cc
+++ b/Grid/communicator/Communicator_base.cc
@@ -33,8 +33,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>

 NAMESPACE_BEGIN(Grid);

-bool Stencil_force_mpi = true;
-
 ///////////////////////////////////////////////////////////////
 // Info that is setup once and indept of cartesian layout
 ///////////////////////////////////////////////////////////////
--- a/Grid/communicator/Communicator_base.h
+++ b/Grid/communicator/Communicator_base.h
@@ -35,8 +35,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>

 NAMESPACE_BEGIN(Grid);

-extern bool Stencil_force_mpi ;
-
 class CartesianCommunicator : public SharedMemory {

 public:    
--- a/Grid/communicator/Communicator_mpi3.cc
+++ b/Grid/communicator/Communicator_mpi3.cc
@@ -370,7 +370,7 @@ double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsReques
  double off_node_bytes=0.0;
  int tag;

-  if ( (gfrom ==MPI_UNDEFINED) || Stencil_force_mpi ) {
+  if ( gfrom ==MPI_UNDEFINED) {
    tag= dir+from*32;
    ierr=MPI_Irecv(recv, bytes, MPI_CHAR,from,tag,communicator_halo[commdir],&rrq);
    assert(ierr==0);
@@ -378,18 +378,12 @@ double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsReques
    off_node_bytes+=bytes;
  }

-  if ( (gdest == MPI_UNDEFINED) || Stencil_force_mpi ) {
+  if ( gdest == MPI_UNDEFINED ) {
    tag= dir+_processor*32;
    ierr =MPI_Isend(xmit, bytes, MPI_CHAR,dest,tag,communicator_halo[commdir],&xrq);
    assert(ierr==0);
    list.push_back(xrq);
    off_node_bytes+=bytes;
-  } else {
-    // TODO : make a OMP loop on CPU, call threaded bcopy
-    void *shm = (void *) this->ShmBufferTranslate(dest,recv);
-    assert(shm!=NULL);
-    //    std::cout <<"acceleratorCopyDeviceToDeviceAsynch"<< std::endl;
-    acceleratorCopyDeviceToDeviceAsynch(xmit,shm,bytes);
  }

  if ( CommunicatorPolicy == CommunicatorPolicySequential ) {
@@ -400,9 +394,6 @@ double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsReques
 }
 void CartesianCommunicator::StencilSendToRecvFromComplete(std::vector<CommsRequest_t> &list,int dir)
 {
-  //   std::cout << "Copy Synchronised\n"<<std::endl;
-  acceleratorCopySynchronise();
-
  int nreq=list.size();

  if (nreq==0) return;
--- a/Grid/communicator/SharedMemoryMPI.cc
+++ b/Grid/communicator/SharedMemoryMPI.cc
@@ -513,16 +513,26 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
  ///////////////////////////////////////////////////////////////////////////////////////////////////////////
  // Each MPI rank should allocate our own buffer
  ///////////////////////////////////////////////////////////////////////////////////////////////////////////
+#ifdef GRID_SYCL_LEVEL_ZERO_IPC
+  auto zeDevice = cl::sycl::get_native<cl::sycl::backend::level_zero>(theGridAccelerator->get_device());
+  auto zeContext= cl::sycl::get_native<cl::sycl::backend::level_zero>(theGridAccelerator->get_context());
+  ze_device_mem_alloc_desc_t zeDesc = {};
+  zeMemAllocDevice(zeContext,&zeDesc,bytes,2*1024*1024,zeDevice,&ShmCommBuf);
+  std::cout << WorldRank << header " SharedMemoryMPI.cc zeMemAllocDevice "<< bytes 
+	      << "bytes at "<< std::hex<< ShmCommBuf <<std::dec<<" for comms buffers " <<std::endl;
+#else  
  ShmCommBuf = acceleratorAllocDevice(bytes);
+#endif  
  if (ShmCommBuf == (void *)NULL ) {
    std::cerr << " SharedMemoryMPI.cc acceleratorAllocDevice failed NULL pointer for " << bytes<<" bytes " << std::endl;
    exit(EXIT_FAILURE);  
  }
-  if ( WorldRank == 0 ){
+  //  if ( WorldRank == 0 ){
+  if ( 1 ){
    std::cout << WorldRank << header " SharedMemoryMPI.cc acceleratorAllocDevice "<< bytes 
 	      << "bytes at "<< std::hex<< ShmCommBuf <<std::dec<<" for comms buffers " <<std::endl;
  }
-  SharedMemoryZero(ShmCommBuf,bytes);
+  //  SharedMemoryZero(ShmCommBuf,bytes);
  std::cout<< "Setting up IPC"<<std::endl;
  ///////////////////////////////////////////////////////////////////////////////////////////////////////////
  // Loop over ranks/gpu's on our node
@@ -533,27 +543,21 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
    //////////////////////////////////////////////////
    // If it is me, pass around the IPC access key
    //////////////////////////////////////////////////
-    void * thisBuf = ShmCommBuf;
-    if(!Stencil_force_mpi) {
 #ifdef GRID_SYCL_LEVEL_ZERO_IPC
-    typedef struct { int fd; pid_t pid ; } clone_mem_t;
-
-    auto zeDevice    = cl::sycl::get_native<cl::sycl::backend::level_zero>(theGridAccelerator->get_device());
-    auto zeContext   = cl::sycl::get_native<cl::sycl::backend::level_zero>(theGridAccelerator->get_context());
-      
-    ze_ipc_mem_handle_t ihandle;
-    clone_mem_t handle;
-
+    ze_ipc_mem_handle_t handle;
    if ( r==WorldShmRank ) { 
-      auto err = zeMemGetIpcHandle(zeContext,ShmCommBuf,&ihandle);
+      auto err = zeMemGetIpcHandle(zeContext,ShmCommBuf,&handle);
      if ( err != ZE_RESULT_SUCCESS ) {
-	std::cout << "SharedMemoryMPI.cc zeMemGetIpcHandle failed for rank "<<r<<" "<<std::hex<<err<<std::dec<<std::endl;
+	std::cerr << "SharedMemoryMPI.cc zeMemGetIpcHandle failed for rank "<<r<<" "<<std::hex<<err<<std::dec<<std::endl;
 	exit(EXIT_FAILURE);
      } else {
-	std::cout << "SharedMemoryMPI.cc zeMemGetIpcHandle succeeded for rank "<<r<<" "<<std::hex<<err<<std::dec<<std::endl;
+	std::cerr << "SharedMemoryMPI.cc zeMemGetIpcHandle succeeded for rank "<<r<<" "<<std::hex<<err<<std::dec<<std::endl;
      }
-      memcpy((void *)&handle.fd,(void *)&ihandle,sizeof(int));
-      handle.pid = getpid();
+      std::cerr<<"Allocated IpcHandle rank "<<r<<" (hex) ";
+      for(int c=0;c<ZE_MAX_IPC_HANDLE_SIZE;c++){
+	std::cerr<<std::hex<<(uint32_t)((uint8_t)handle.data[c])<<std::dec;
+      }
+      std::cerr<<std::endl;
    }
 #endif
 #ifdef GRID_CUDA
@@ -576,7 +580,6 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
      }
    }
 #endif
-
    //////////////////////////////////////////////////
    // Share this IPC handle across the Shm Comm
    //////////////////////////////////////////////////
@@ -592,31 +595,22 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
    ///////////////////////////////////////////////////////////////
    // If I am not the source, overwrite thisBuf with remote buffer
    ///////////////////////////////////////////////////////////////
-
+    void * thisBuf = ShmCommBuf;
 #ifdef GRID_SYCL_LEVEL_ZERO_IPC
    if ( r!=WorldShmRank ) {
      thisBuf = nullptr;
-      std::cout<<"mapping seeking remote pid/fd "
-	       <<handle.pid<<"/"
-	       <<handle.fd<<std::endl;
-
-      int pidfd = syscall(SYS_pidfd_open,handle.pid,0);
-      std::cout<<"Using IpcHandle pidfd "<<pidfd<<"\n";
-      //      int myfd  = syscall(SYS_pidfd_getfd,pidfd,handle.fd,0);
-      int myfd  = syscall(438,pidfd,handle.fd,0);
-
-      std::cout<<"Using IpcHandle myfd "<<myfd<<"\n";
-      
-      memcpy((void *)&ihandle,(void *)&myfd,sizeof(int));
-
-      auto err = zeMemOpenIpcHandle(zeContext,zeDevice,ihandle,0,&thisBuf);
+      std::cerr<<"Using IpcHandle rank "<<r<<" ";
+      for(int c=0;c<ZE_MAX_IPC_HANDLE_SIZE;c++){
+	std::cerr<<std::hex<<(uint32_t)((uint8_t)handle.data[c])<<std::dec;
+      }
+      std::cerr<<std::endl;
+      auto err = zeMemOpenIpcHandle(zeContext,zeDevice,handle,0,&thisBuf);
      if ( err != ZE_RESULT_SUCCESS ) {
-	std::cout << "SharedMemoryMPI.cc "<<zeContext<<" "<<zeDevice<<std::endl;
-	std::cout << "SharedMemoryMPI.cc zeMemOpenIpcHandle failed for rank "<<r<<" "<<std::hex<<err<<std::dec<<std::endl; 
+	std::cerr << "SharedMemoryMPI.cc "<<zeContext<<" "<<zeDevice<<std::endl;
+	std::cerr << "SharedMemoryMPI.cc zeMemOpenIpcHandle failed for rank "<<r<<" "<<std::hex<<err<<std::dec<<std::endl; 
 	exit(EXIT_FAILURE);
      } else {
-	std::cout << "SharedMemoryMPI.cc zeMemOpenIpcHandle succeeded for rank "<<r<<std::endl;
-	std::cout << "SharedMemoryMPI.cc zeMemOpenIpcHandle pointer is "<<std::hex<<thisBuf<<std::dec<<std::endl;
+	std::cerr << "SharedMemoryMPI.cc zeMemOpenIpcHandle succeeded for rank "<<r<<" "<<std::hex<<err<<std::dec<<std::endl;
      }
      assert(thisBuf!=nullptr);
    }
@@ -642,7 +636,6 @@ void GlobalSharedMemory::SharedMemoryAllocate(uint64_t bytes, int flags)
    ///////////////////////////////////////////////////////////////
    // Save a copy of the device buffers
    ///////////////////////////////////////////////////////////////
-    }
    WorldShmCommBufs[r] = thisBuf;
 #else
    WorldShmCommBufs[r] = ShmCommBuf;
--- a/Grid/lattice/Lattice_arith.h
+++ b/Grid/lattice/Lattice_arith.h
@@ -225,7 +225,7 @@ void axpy(Lattice<vobj> &ret,sobj a,const Lattice<vobj> &x,const Lattice<vobj> &
  autoView( x_v , x, AcceleratorRead);
  autoView( y_v , y, AcceleratorRead);
  accelerator_for(ss,x_v.size(),vobj::Nsimd(),{
-    auto tmp = a*coalescedRead(x_v[ss])+coalescedRead(y_v[ss]);
+    auto tmp = a*x_v(ss)+y_v(ss);
    coalescedWrite(ret_v[ss],tmp);
  });
 }
--- a/Grid/lattice/Lattice_base.h
+++ b/Grid/lattice/Lattice_base.h
@@ -88,13 +88,6 @@ public:
    LatticeView<vobj> accessor(*( (LatticeAccelerator<vobj> *) this),mode);
    accessor.ViewClose();
  }
-
-  // Helper function to print the state of this object in the AccCache
-  void PrintCacheState(void)
-  {
-    MemoryManager::PrintState(this->_odata);
-  }
-
  /////////////////////////////////////////////////////////////////////////////////
  // 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
--- a/Grid/lattice/Lattice_basis.h
+++ b/Grid/lattice/Lattice_basis.h
@@ -125,7 +125,7 @@ void basisRotate(VField &basis,Matrix& Qt,int j0, int j1, int k0,int k1,int Nm)

 	for(int k=k0; k<k1; ++k){
 	  auto tmp = coalescedRead(Bp[ss*nrot+j]);
-	  coalescedWrite(Bp[ss*nrot+j],tmp+ Qt_p[jj*Nm+k] * coalescedRead(basis_vp[k][sss]));
+	  coalescedWrite(Bp[ss*nrot+j],tmp+ Qt_p[jj*Nm+k] * coalescedRead(basis_v[k][sss]));
 	}
      });

@@ -134,7 +134,7 @@ void basisRotate(VField &basis,Matrix& Qt,int j0, int j1, int k0,int k1,int Nm)
 	int jj  =j0+j;
 	int ss =sj/nrot;
 	int sss=ss+s;
-	coalescedWrite(basis_vp[jj][sss],coalescedRead(Bp[ss*nrot+j]));
+	coalescedWrite(basis_v[jj][sss],coalescedRead(Bp[ss*nrot+j]));
      });
  }
 #endif
--- a/Grid/lattice/Lattice_reduction.h
+++ b/Grid/lattice/Lattice_reduction.h
@@ -361,7 +361,6 @@ 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;
-  typedef typename vobj::scalar_object::scalar_type scalar_type;
  GridBase  *grid = Data.Grid();
  assert(grid!=NULL);

@@ -420,19 +419,20 @@ template<class vobj> inline void sliceSum(const Lattice<vobj> &Data,std::vector<
  }
  
  // sum over nodes.
+  sobj gsum;
  for(int t=0;t<fd;t++){
    int pt = t/ld; // processor plane
    int lt = t%ld;
    if ( pt == grid->_processor_coor[orthogdim] ) {
-      result[t]=lsSum[lt];
+      gsum=lsSum[lt];
    } else {
-      result[t]=Zero();
+      gsum=Zero();
    }

+    grid->GlobalSum(gsum);
+
+    result[t]=gsum;
  }
-  scalar_type * ptr = (scalar_type *) &result[0];
-  int words = fd*sizeof(sobj)/sizeof(scalar_type);
-  grid->GlobalSumVector(ptr, words);
 }

 template<class vobj>
--- a/Grid/lattice/Lattice_reduction_gpu.h
+++ b/Grid/lattice/Lattice_reduction_gpu.h
@@ -42,6 +42,7 @@ void getNumBlocksAndThreads(const Iterator n, const size_t sizeofsobj, Iterator
  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) {
@@ -51,10 +52,6 @@ void getNumBlocksAndThreads(const Iterator n, const size_t sizeofsobj, Iterator
  
  // let the number of threads in a block be a multiple of 2, starting from warpSize
  threads = warpSize;
-  if ( threads*sizeofsobj > sharedMemPerBlock ) {
-    std::cout << GridLogError << "The object is too large for the shared memory." << std::endl;
-    exit(EXIT_FAILURE);
-  }
  while( 2*threads*sizeofsobj < sharedMemPerBlock && 2*threads <= maxThreadsPerBlock ) threads *= 2;
  // keep all the streaming multiprocessors busy
  blocks = nextPow2(multiProcessorCount);
--- a/Grid/lattice/Lattice_transfer.h
+++ b/Grid/lattice/Lattice_transfer.h
@@ -85,76 +85,6 @@ template<class vobj> inline void setCheckerboard(Lattice<vobj> &full,const Latti
  });
 }

-template<class vobj> inline void acceleratorPickCheckerboard(int cb,Lattice<vobj> &half,const Lattice<vobj> &full, int checker_dim_half=0)
-{
-  half.Checkerboard() = cb;
-  autoView(half_v, half, AcceleratorWrite);
-  autoView(full_v, full, AcceleratorRead);
-  Coordinate rdim_full             = full.Grid()->_rdimensions;
-  Coordinate rdim_half             = half.Grid()->_rdimensions;
-  unsigned long ndim_half          = half.Grid()->_ndimension;
-  Coordinate checker_dim_mask_half = half.Grid()->_checker_dim_mask;
-  Coordinate ostride_half          = half.Grid()->_ostride;
-  accelerator_for(ss, full.Grid()->oSites(),full.Grid()->Nsimd(),{
-    
-    Coordinate coor;
-    int cbos;
-    int linear=0;
-
-    Lexicographic::CoorFromIndex(coor,ss,rdim_full);
-    assert(coor.size()==ndim_half);
-
-    for(int d=0;d<ndim_half;d++){ 
-      if(checker_dim_mask_half[d]) linear += coor[d];
-    }
-    cbos = (linear&0x1);
-
-    if (cbos==cb) {
-      int ssh=0;
-      for(int d=0;d<ndim_half;d++) {
-        if (d == checker_dim_half) ssh += ostride_half[d] * ((coor[d] / 2) % rdim_half[d]);
-        else ssh += ostride_half[d] * (coor[d] % rdim_half[d]);
-      }
-      coalescedWrite(half_v[ssh],full_v(ss));
-    }
-  });
-}
-template<class vobj> inline void acceleratorSetCheckerboard(Lattice<vobj> &full,const Lattice<vobj> &half, int checker_dim_half=0)
-{
-  int cb = half.Checkerboard();
-  autoView(half_v , half, AcceleratorRead);
-  autoView(full_v , full, AcceleratorWrite);
-  Coordinate rdim_full             = full.Grid()->_rdimensions;
-  Coordinate rdim_half             = half.Grid()->_rdimensions;
-  unsigned long ndim_half          = half.Grid()->_ndimension;
-  Coordinate checker_dim_mask_half = half.Grid()->_checker_dim_mask;
-  Coordinate ostride_half          = half.Grid()->_ostride;
-  accelerator_for(ss,full.Grid()->oSites(),full.Grid()->Nsimd(),{
-
-    Coordinate coor;
-    int cbos;
-    int linear=0;
-  
-    Lexicographic::CoorFromIndex(coor,ss,rdim_full);
-    assert(coor.size()==ndim_half);
-
-    for(int d=0;d<ndim_half;d++){ 
-      if(checker_dim_mask_half[d]) linear += coor[d];
-    }
-    cbos = (linear&0x1);
-
-    if (cbos==cb) {
-      int ssh=0;
-      for(int d=0;d<ndim_half;d++){
-        if (d == checker_dim_half) ssh += ostride_half[d] * ((coor[d] / 2) % rdim_half[d]);
-        else ssh += ostride_half[d] * (coor[d] % rdim_half[d]);
-      }
-      coalescedWrite(full_v[ss],half_v(ssh));
-    }
-
-  });
-}
-
 ////////////////////////////////////////////////////////////////////////////////////////////
 // Flexible Type Conversion for internal promotion to double as well as graceful
 // treatment of scalar-compatible types
@@ -434,22 +364,16 @@ inline void blockSum(Lattice<vobj> &coarseData,const Lattice<vobj> &fineData)
  autoView( coarseData_ , coarseData, AcceleratorWrite);
  autoView( fineData_   , fineData, AcceleratorRead);

-  auto coarseData_p = &coarseData_[0];
-  auto fineData_p = &fineData_[0];
-  
  Coordinate fine_rdimensions = fine->_rdimensions;
  Coordinate coarse_rdimensions = coarse->_rdimensions;
-
-  vobj zz = Zero();
  
  accelerator_for(sc,coarse->oSites(),1,{

      // One thread per sub block
      Coordinate coor_c(_ndimension);
      Lexicographic::CoorFromIndex(coor_c,sc,coarse_rdimensions);  // Block coordinate
+      coarseData_[sc]=Zero();

-      vobj cd = zz;
-      
      for(int sb=0;sb<blockVol;sb++){

 	int sf;
@@ -459,11 +383,9 @@ inline void blockSum(Lattice<vobj> &coarseData,const Lattice<vobj> &fineData)
 	for(int d=0;d<_ndimension;d++) coor_f[d]=coor_c[d]*block_r[d] + coor_b[d];
 	Lexicographic::IndexFromCoor(coor_f,sf,fine_rdimensions);

-	cd=cd+fineData_p[sf];
+	coarseData_[sc]=coarseData_[sc]+fineData_[sf];
      }

-      coarseData_p[sc] = cd;
-
    });
  return;
 }
--- a/Grid/parallelIO/IldgIO.h
+++ b/Grid/parallelIO/IldgIO.h
@@ -576,8 +576,6 @@ class ScidacReader : public GridLimeReader {
    std::string rec_name(ILDG_BINARY_DATA);
    while ( limeReaderNextRecord(LimeR) == LIME_SUCCESS ) { 
      if ( !strncmp(limeReaderType(LimeR), rec_name.c_str(),strlen(rec_name.c_str()) )  ) {
-  // in principle should do the line below, but that breaks backard compatibility with old data
-  // skipPastObjectRecord(std::string(GRID_FIELD_NORM));
 	skipPastObjectRecord(std::string(SCIDAC_CHECKSUM));
 	return;
      }
--- a/Grid/parallelIO/NerscIO.h
+++ b/Grid/parallelIO/NerscIO.h
@@ -198,7 +198,7 @@ public:
      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.plaquette -header.plaquette ) < 1.0e-1 );
    assert(fabs(clone.link_trace-header.link_trace) < 1.0e-6 );
    assert(nersc_csum == header.checksum );
      
--- a/Grid/qcd/action/fermion/Fermion.h
+++ b/Grid/qcd/action/fermion/Fermion.h
@@ -115,9 +115,9 @@ typedef WilsonFermion<WilsonImplR> WilsonFermionR;
 typedef WilsonFermion<WilsonImplF> WilsonFermionF;
 typedef WilsonFermion<WilsonImplD> WilsonFermionD;

-//typedef WilsonFermion<WilsonImplRL> WilsonFermionRL;
-//typedef WilsonFermion<WilsonImplFH> WilsonFermionFH;
-//typedef WilsonFermion<WilsonImplDF> WilsonFermionDF;
+typedef WilsonFermion<WilsonImplRL> WilsonFermionRL;
+typedef WilsonFermion<WilsonImplFH> WilsonFermionFH;
+typedef WilsonFermion<WilsonImplDF> WilsonFermionDF;

 typedef WilsonFermion<WilsonAdjImplR> WilsonAdjFermionR;
 typedef WilsonFermion<WilsonAdjImplF> WilsonAdjFermionF;
@@ -158,41 +158,41 @@ typedef DomainWallFermion<WilsonImplR> DomainWallFermionR;
 typedef DomainWallFermion<WilsonImplF> DomainWallFermionF;
 typedef DomainWallFermion<WilsonImplD> DomainWallFermionD;

-//typedef DomainWallFermion<WilsonImplRL> DomainWallFermionRL;
-//typedef DomainWallFermion<WilsonImplFH> DomainWallFermionFH;
-//typedef DomainWallFermion<WilsonImplDF> DomainWallFermionDF;
+typedef DomainWallFermion<WilsonImplRL> DomainWallFermionRL;
+typedef DomainWallFermion<WilsonImplFH> DomainWallFermionFH;
+typedef DomainWallFermion<WilsonImplDF> DomainWallFermionDF;

 typedef DomainWallEOFAFermion<WilsonImplR> DomainWallEOFAFermionR;
 typedef DomainWallEOFAFermion<WilsonImplF> DomainWallEOFAFermionF;
 typedef DomainWallEOFAFermion<WilsonImplD> DomainWallEOFAFermionD;

-//typedef DomainWallEOFAFermion<WilsonImplRL> DomainWallEOFAFermionRL;
-//typedef DomainWallEOFAFermion<WilsonImplFH> DomainWallEOFAFermionFH;
-//typedef DomainWallEOFAFermion<WilsonImplDF> DomainWallEOFAFermionDF;
+typedef DomainWallEOFAFermion<WilsonImplRL> DomainWallEOFAFermionRL;
+typedef DomainWallEOFAFermion<WilsonImplFH> DomainWallEOFAFermionFH;
+typedef DomainWallEOFAFermion<WilsonImplDF> DomainWallEOFAFermionDF;

 typedef MobiusFermion<WilsonImplR> MobiusFermionR;
 typedef MobiusFermion<WilsonImplF> MobiusFermionF;
 typedef MobiusFermion<WilsonImplD> MobiusFermionD;

-//typedef MobiusFermion<WilsonImplRL> MobiusFermionRL;
-//typedef MobiusFermion<WilsonImplFH> MobiusFermionFH;
-//typedef MobiusFermion<WilsonImplDF> MobiusFermionDF;
+typedef MobiusFermion<WilsonImplRL> MobiusFermionRL;
+typedef MobiusFermion<WilsonImplFH> MobiusFermionFH;
+typedef MobiusFermion<WilsonImplDF> MobiusFermionDF;

 typedef MobiusEOFAFermion<WilsonImplR> MobiusEOFAFermionR;
 typedef MobiusEOFAFermion<WilsonImplF> MobiusEOFAFermionF;
 typedef MobiusEOFAFermion<WilsonImplD> MobiusEOFAFermionD;

-//typedef MobiusEOFAFermion<WilsonImplRL> MobiusEOFAFermionRL;
-//typedef MobiusEOFAFermion<WilsonImplFH> MobiusEOFAFermionFH;
-//typedef MobiusEOFAFermion<WilsonImplDF> MobiusEOFAFermionDF;
+typedef MobiusEOFAFermion<WilsonImplRL> MobiusEOFAFermionRL;
+typedef MobiusEOFAFermion<WilsonImplFH> MobiusEOFAFermionFH;
+typedef MobiusEOFAFermion<WilsonImplDF> MobiusEOFAFermionDF;

 typedef ZMobiusFermion<ZWilsonImplR> ZMobiusFermionR;
 typedef ZMobiusFermion<ZWilsonImplF> ZMobiusFermionF;
 typedef ZMobiusFermion<ZWilsonImplD> ZMobiusFermionD;

-//typedef ZMobiusFermion<ZWilsonImplRL> ZMobiusFermionRL;
-//typedef ZMobiusFermion<ZWilsonImplFH> ZMobiusFermionFH;
-//typedef ZMobiusFermion<ZWilsonImplDF> ZMobiusFermionDF;
+typedef ZMobiusFermion<ZWilsonImplRL> ZMobiusFermionRL;
+typedef ZMobiusFermion<ZWilsonImplFH> ZMobiusFermionFH;
+typedef ZMobiusFermion<ZWilsonImplDF> ZMobiusFermionDF;

 // Ls vectorised
 typedef ScaledShamirFermion<WilsonImplR> ScaledShamirFermionR;
@@ -235,49 +235,49 @@ typedef WilsonFermion<GparityWilsonImplR>     GparityWilsonFermionR;
 typedef WilsonFermion<GparityWilsonImplF>     GparityWilsonFermionF;
 typedef WilsonFermion<GparityWilsonImplD>     GparityWilsonFermionD;

-//typedef WilsonFermion<GparityWilsonImplRL>     GparityWilsonFermionRL;
-//typedef WilsonFermion<GparityWilsonImplFH>     GparityWilsonFermionFH;
-//typedef WilsonFermion<GparityWilsonImplDF>     GparityWilsonFermionDF;
+typedef WilsonFermion<GparityWilsonImplRL>     GparityWilsonFermionRL;
+typedef WilsonFermion<GparityWilsonImplFH>     GparityWilsonFermionFH;
+typedef WilsonFermion<GparityWilsonImplDF>     GparityWilsonFermionDF;

 typedef DomainWallFermion<GparityWilsonImplR> GparityDomainWallFermionR;
 typedef DomainWallFermion<GparityWilsonImplF> GparityDomainWallFermionF;
 typedef DomainWallFermion<GparityWilsonImplD> GparityDomainWallFermionD;

-//typedef DomainWallFermion<GparityWilsonImplRL> GparityDomainWallFermionRL;
-//typedef DomainWallFermion<GparityWilsonImplFH> GparityDomainWallFermionFH;
-//typedef DomainWallFermion<GparityWilsonImplDF> GparityDomainWallFermionDF;
+typedef DomainWallFermion<GparityWilsonImplRL> GparityDomainWallFermionRL;
+typedef DomainWallFermion<GparityWilsonImplFH> GparityDomainWallFermionFH;
+typedef DomainWallFermion<GparityWilsonImplDF> GparityDomainWallFermionDF;

 typedef DomainWallEOFAFermion<GparityWilsonImplR> GparityDomainWallEOFAFermionR;
 typedef DomainWallEOFAFermion<GparityWilsonImplF> GparityDomainWallEOFAFermionF;
 typedef DomainWallEOFAFermion<GparityWilsonImplD> GparityDomainWallEOFAFermionD;

-//typedef DomainWallEOFAFermion<GparityWilsonImplRL> GparityDomainWallEOFAFermionRL;
-//typedef DomainWallEOFAFermion<GparityWilsonImplFH> GparityDomainWallEOFAFermionFH;
-//typedef DomainWallEOFAFermion<GparityWilsonImplDF> GparityDomainWallEOFAFermionDF;
+typedef DomainWallEOFAFermion<GparityWilsonImplRL> GparityDomainWallEOFAFermionRL;
+typedef DomainWallEOFAFermion<GparityWilsonImplFH> GparityDomainWallEOFAFermionFH;
+typedef DomainWallEOFAFermion<GparityWilsonImplDF> GparityDomainWallEOFAFermionDF;

 typedef WilsonTMFermion<GparityWilsonImplR> GparityWilsonTMFermionR;
 typedef WilsonTMFermion<GparityWilsonImplF> GparityWilsonTMFermionF;
 typedef WilsonTMFermion<GparityWilsonImplD> GparityWilsonTMFermionD;

-//typedef WilsonTMFermion<GparityWilsonImplRL> GparityWilsonTMFermionRL;
-//typedef WilsonTMFermion<GparityWilsonImplFH> GparityWilsonTMFermionFH;
-//typedef WilsonTMFermion<GparityWilsonImplDF> GparityWilsonTMFermionDF;
+typedef WilsonTMFermion<GparityWilsonImplRL> GparityWilsonTMFermionRL;
+typedef WilsonTMFermion<GparityWilsonImplFH> GparityWilsonTMFermionFH;
+typedef WilsonTMFermion<GparityWilsonImplDF> GparityWilsonTMFermionDF;

 typedef MobiusFermion<GparityWilsonImplR> GparityMobiusFermionR;
 typedef MobiusFermion<GparityWilsonImplF> GparityMobiusFermionF;
 typedef MobiusFermion<GparityWilsonImplD> GparityMobiusFermionD;

-//typedef MobiusFermion<GparityWilsonImplRL> GparityMobiusFermionRL;
-//typedef MobiusFermion<GparityWilsonImplFH> GparityMobiusFermionFH;
-//typedef MobiusFermion<GparityWilsonImplDF> GparityMobiusFermionDF;
+typedef MobiusFermion<GparityWilsonImplRL> GparityMobiusFermionRL;
+typedef MobiusFermion<GparityWilsonImplFH> GparityMobiusFermionFH;
+typedef MobiusFermion<GparityWilsonImplDF> GparityMobiusFermionDF;

 typedef MobiusEOFAFermion<GparityWilsonImplR> GparityMobiusEOFAFermionR;
 typedef MobiusEOFAFermion<GparityWilsonImplF> GparityMobiusEOFAFermionF;
 typedef MobiusEOFAFermion<GparityWilsonImplD> GparityMobiusEOFAFermionD;

-//typedef MobiusEOFAFermion<GparityWilsonImplRL> GparityMobiusEOFAFermionRL;
-//typedef MobiusEOFAFermion<GparityWilsonImplFH> GparityMobiusEOFAFermionFH;
-//typedef MobiusEOFAFermion<GparityWilsonImplDF> GparityMobiusEOFAFermionDF;
+typedef MobiusEOFAFermion<GparityWilsonImplRL> GparityMobiusEOFAFermionRL;
+typedef MobiusEOFAFermion<GparityWilsonImplFH> GparityMobiusEOFAFermionFH;
+typedef MobiusEOFAFermion<GparityWilsonImplDF> GparityMobiusEOFAFermionDF;

 typedef ImprovedStaggeredFermion<StaggeredImplR> ImprovedStaggeredFermionR;
 typedef ImprovedStaggeredFermion<StaggeredImplF> ImprovedStaggeredFermionF;
--- a/Grid/qcd/action/fermion/GparityWilsonImpl.h
+++ b/Grid/qcd/action/fermion/GparityWilsonImpl.h
@@ -327,8 +327,8 @@ typedef GparityWilsonImpl<vComplex , FundamentalRepresentation,CoeffReal> Gparit
 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 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);
--- a/Grid/qcd/action/fermion/WilsonCompressor.h
+++ b/Grid/qcd/action/fermion/WilsonCompressor.h
@@ -68,12 +68,11 @@ public:
  /*****************************************************/
  /* Compress includes precision change if mpi data is not same */
  /*****************************************************/
-  accelerator_inline void Compress(SiteHalfSpinor &buf,const SiteSpinor &in) const {
-    typedef decltype(coalescedRead(buf)) sobj;
-    sobj sp;
-    auto sin = coalescedRead(in);
-    projector::Proj(sp,sin,mu,dag);
-    coalescedWrite(buf,sp);
+  template<class _SiteHalfSpinor, class _SiteSpinor>
+  accelerator_inline void Compress(_SiteHalfSpinor *buf,Integer o,const _SiteSpinor &in) const {
+    _SiteHalfSpinor tmp;
+    projector::Proj(tmp,in,mu,dag);
+    vstream(buf[o],tmp);
  }

  /*****************************************************/
@@ -83,18 +82,13 @@ public:
 				   const SiteHalfSpinor * __restrict__ vp0,
 				   const SiteHalfSpinor * __restrict__ vp1,
 				   Integer type,Integer o) const {
-#ifdef GRID_SIMT
-    exchangeSIMT(mp[2*o],mp[2*o+1],vp0[o],vp1[o],type);
-#else
    SiteHalfSpinor tmp1;
    SiteHalfSpinor tmp2;
    exchange(tmp1,tmp2,vp0[o],vp1[o],type);
    vstream(mp[2*o  ],tmp1);
    vstream(mp[2*o+1],tmp2);
-#endif
  }

-
  /*****************************************************/
  /* Have a decompression step if mpi data is not same */
  /*****************************************************/
@@ -111,28 +105,6 @@ public:
 					   const SiteSpinor * __restrict__ in,
 					   Integer j,Integer k, Integer m,Integer type) const
  {
-#ifdef GRID_SIMT
-    typedef SiteSpinor vobj;
-    typedef SiteHalfSpinor hvobj;
-    typedef decltype(coalescedRead(*in))    sobj;
-    typedef decltype(coalescedRead(*out0)) hsobj;
-
-    unsigned int Nsimd = vobj::Nsimd();
-    unsigned int mask = Nsimd >> (type + 1);
-    int lane = acceleratorSIMTlane(Nsimd);
-    int j0 = lane &(~mask); // inner coor zero
-    int j1 = lane |(mask) ; // inner coor one
-    const vobj *vp0 = &in[k];
-    const vobj *vp1 = &in[m];
-    const vobj *vp = (lane&mask) ? vp1:vp0;
-    auto sa = coalescedRead(*vp,j0);
-    auto sb = coalescedRead(*vp,j1);
-    hsobj psa, psb;
-    projector::Proj(psa,sa,mu,dag);
-    projector::Proj(psb,sb,mu,dag);
-    coalescedWrite(out0[j],psa);
-    coalescedWrite(out1[j],psb);
-#else
    SiteHalfSpinor temp1, temp2;
    SiteHalfSpinor temp3, temp4;
    projector::Proj(temp1,in[k],mu,dag);
@@ -140,7 +112,6 @@ public:
    exchange(temp3,temp4,temp1,temp2,type);
    vstream(out0[j],temp3);
    vstream(out1[j],temp4);
-#endif
  }

  /*****************************************************/
@@ -150,7 +121,6 @@ public:

 };

-#if 0
 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 >
@@ -179,23 +149,13 @@ public:
  /*****************************************************/
  /* Compress includes precision change if mpi data is not same */
  /*****************************************************/
-  accelerator_inline void Compress(SiteHalfSpinor &buf,const SiteSpinor &in) const {
-    SiteHalfSpinor hsp;
+  template<class _SiteHalfSpinor, class _SiteSpinor>
+  accelerator_inline void Compress(_SiteHalfSpinor *buf,Integer o,const _SiteSpinor &in) const {
+    _SiteHalfSpinor hsp;
    SiteHalfCommSpinor *hbuf = (SiteHalfCommSpinor *)buf;
    projector::Proj(hsp,in,mu,dag);
    precisionChange((vComplexLow *)&hbuf[o],(vComplexHigh *)&hsp,Nw);
  }
-  accelerator_inline void Compress(SiteHalfSpinor &buf,const SiteSpinor &in) const {
-#ifdef GRID_SIMT
-    typedef decltype(coalescedRead(buf)) sobj;
-    sobj sp;
-    auto sin = coalescedRead(in);
-    projector::Proj(sp,sin,mu,dag);
-    coalescedWrite(buf,sp);
-#else
-    projector::Proj(buf,in,mu,dag);
-#endif
-  }

  /*****************************************************/
  /* Exchange includes precision change if mpi data is not same */
@@ -243,7 +203,6 @@ public:
  accelerator_inline bool DecompressionStep(void) const { return true; }

 };
-#endif

 #define DECLARE_PROJ(Projector,Compressor,spProj)			\
  class Projector {							\
@@ -294,8 +253,33 @@ public:
  typedef typename Base::View_type View_type;
  typedef typename Base::StencilVector StencilVector;

-  void ZeroCountersi(void)  {  }
-  void Reporti(int calls)  {  }
+  double timer0;
+  double timer1;
+  double timer2;
+  double timer3;
+  double timer4;
+  double timer5;
+  double timer6;
+  uint64_t callsi;
+  void ZeroCountersi(void)
+  {
+    timer0=0;
+    timer1=0;
+    timer2=0;
+    timer3=0;
+    timer4=0;
+    timer5=0;
+    timer6=0;
+    callsi=0;
+  }
+  void Reporti(int calls)
+  {
+    if ( timer0 ) std::cout << GridLogMessage << " timer0 (HaloGatherOpt) " <<timer0/calls <<std::endl;
+    if ( timer1 ) std::cout << GridLogMessage << " timer1 (Communicate)   " <<timer1/calls <<std::endl;
+    if ( timer2 ) std::cout << GridLogMessage << " timer2 (CommsMerge )   " <<timer2/calls <<std::endl;
+    if ( timer3 ) std::cout << GridLogMessage << " timer3 (commsMergeShm) " <<timer3/calls <<std::endl;
+    if ( timer4 ) std::cout << GridLogMessage << " timer4 " <<timer4 <<std::endl;
+  }

  std::vector<int> surface_list;

@@ -337,18 +321,26 @@ public:
  {
    std::vector<std::vector<CommsRequest_t> > reqs;
    this->HaloExchangeOptGather(source,compress);
+    double t1=usecond();
    // Asynchronous MPI calls multidirectional, Isend etc...
    // Non-overlapped directions within a thread. Asynchronous calls except MPI3, threaded up to comm threads ways.
    this->Communicate();
+    double t2=usecond(); timer1 += t2-t1;
    this->CommsMerge(compress);
+    double t3=usecond(); timer2 += t3-t2;
    this->CommsMergeSHM(compress);
+    double t4=usecond(); timer3 += t4-t3;
  }
  
  template <class compressor>
  void HaloExchangeOptGather(const Lattice<vobj> &source,compressor &compress) 
  {
    this->Prepare();
+    double t0=usecond();
    this->HaloGatherOpt(source,compress);
+    double t1=usecond();
+    timer0 += t1-t0;
+    callsi++;
  }

  template <class compressor>
@@ -360,9 +352,12 @@ public:
    typedef typename compressor::SiteHalfSpinor     SiteHalfSpinor;
    typedef typename compressor::SiteHalfCommSpinor SiteHalfCommSpinor;

+    this->mpi3synctime_g-=usecond();
    this->_grid->StencilBarrier();
+    this->mpi3synctime_g+=usecond();

    assert(source.Grid()==this->_grid);
+    this->halogtime-=usecond();
    
    this->u_comm_offset=0;
      
@@ -398,6 +393,7 @@ public:
    }
    this->face_table_computed=1;
    assert(this->u_comm_offset==this->_unified_buffer_size);
+    this->halogtime+=usecond();
    accelerator_barrier();
  }

--- a/Grid/qcd/action/fermion/WilsonImpl.h
+++ b/Grid/qcd/action/fermion/WilsonImpl.h
@@ -243,17 +243,17 @@ typedef WilsonImpl<vComplex,  FundamentalRepresentation, CoeffReal > WilsonImplR
 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, 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,  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
--- a/Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h
@@ -828,7 +828,6 @@ void CayleyFermion5D<Impl>::SeqConservedCurrent(PropagatorField &q_in,

 #if (!defined(GRID_HIP))
  int tshift = (mu == Nd-1) ? 1 : 0;
-  unsigned int LLt    = GridDefaultLatt()[Tp];
  ////////////////////////////////////////////////
  // GENERAL CAYLEY CASE
  ////////////////////////////////////////////////
@@ -881,7 +880,7 @@ void CayleyFermion5D<Impl>::SeqConservedCurrent(PropagatorField &q_in,
  }

  std::vector<RealD> G_s(Ls,1.0);
-  RealD sign = 1.0; // sign flip for vector/tadpole
+  Integer sign = 1; // sign flip for vector/tadpole
  if ( curr_type == Current::Axial ) {
    for(int s=0;s<Ls/2;s++){
      G_s[s] = -1.0;
@@ -891,7 +890,7 @@ void CayleyFermion5D<Impl>::SeqConservedCurrent(PropagatorField &q_in,
    auto b=this->_b;
    auto c=this->_c;
    if ( b == 1 && c == 0 ) {
-      sign = -1.0;    
+      sign = -1;    
    }
    else {
      std::cerr << "Error: Tadpole implementation currently unavailable for non-Shamir actions." << std::endl;
@@ -902,8 +901,8 @@ void CayleyFermion5D<Impl>::SeqConservedCurrent(PropagatorField &q_in,
  for(int s=0;s<Ls;s++){

    int sp = (s+1)%Ls;
-    //    int sr = Ls-1-s;
-    //    int srp= (sr+1)%Ls;
+    int sr = Ls-1-s;
+    int srp= (sr+1)%Ls;

    // Mobius parameters
    auto b=this->bs[s];
@@ -935,13 +934,7 @@ void CayleyFermion5D<Impl>::SeqConservedCurrent(PropagatorField &q_in,
    tmp    = Cshift(tmp,mu,-1);
    Impl::multLinkField(Utmp,this->Umu,tmp,mu+Nd); // Adjoint link
    tmp = -G_s[s]*( Utmp + gmu*Utmp );
-    // Mask the time
-    if (tmax == LLt - 1 && tshift == 1){ // quick fix to include timeslice 0 if tmax + tshift is over the last timeslice
-      unsigned int t0 = 0;
-      tmp    = where(((lcoor==t0) || (lcoor>=tmin+tshift)),tmp,zz);
-    } else {
-      tmp    = where((lcoor>=tmin+tshift),tmp,zz);
-    }
+    tmp    = where((lcoor>=tmin+tshift),tmp,zz); // Mask the time 
    L_Q   += where((lcoor<=tmax+tshift),tmp,zz); // Position of current complicated

    InsertSlice(L_Q, q_out, s , 0);
--- a/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmA64FX.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmA64FX.h
@@ -73,17 +73,17 @@ WilsonKernels<ZWilsonImplF>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView
 						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
 #include <qcd/action/fermion/implementation/WilsonKernelsAsmBodyA64FX.h>

-//#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-//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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>

-//#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-//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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>


 #undef INTERIOR_AND_EXTERIOR
@@ -102,17 +102,17 @@ WilsonKernels<ZWilsonImplF>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldVi
 						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
 #include <qcd/action/fermion/implementation/WilsonKernelsAsmBodyA64FX.h>

-//#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-//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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>

-//#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-//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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>


 #undef INTERIOR_AND_EXTERIOR
@@ -131,17 +131,17 @@ WilsonKernels<ZWilsonImplF>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldVi
 						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
 #include <qcd/action/fermion/implementation/WilsonKernelsAsmBodyA64FX.h>

-//#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-//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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>

-//#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-//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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>



@@ -165,17 +165,17 @@ WilsonKernels<ZWilsonImplF>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldVi
 						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
 #include <qcd/action/fermion/implementation/WilsonKernelsAsmBodyA64FX.h>

-//#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-//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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>

-//#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-//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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>


 #undef INTERIOR_AND_EXTERIOR
@@ -194,17 +194,17 @@ WilsonKernels<ZWilsonImplF>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFiel
 						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
 #include <qcd/action/fermion/implementation/WilsonKernelsAsmBodyA64FX.h>

-//#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-//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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>

-//#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-//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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>


 #undef INTERIOR_AND_EXTERIOR
@@ -223,17 +223,17 @@ WilsonKernels<ZWilsonImplF>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFiel
 						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
 #include <qcd/action/fermion/implementation/WilsonKernelsAsmBodyA64FX.h>

-//#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-//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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>

-//#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-//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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>



@@ -280,17 +280,17 @@ WilsonKernels<ZWilsonImplD>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView
 						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
 #include <qcd/action/fermion/implementation/WilsonKernelsAsmBodyA64FX.h>

-// #pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-// 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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>

-// #pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-// 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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>


 #undef INTERIOR_AND_EXTERIOR
@@ -309,17 +309,17 @@ WilsonKernels<ZWilsonImplD>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldVi
 						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
 #include <qcd/action/fermion/implementation/WilsonKernelsAsmBodyA64FX.h>

-// #pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-// 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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>

-// #pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-// 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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>


 #undef INTERIOR_AND_EXTERIOR
@@ -338,17 +338,17 @@ WilsonKernels<ZWilsonImplD>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldVi
 						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
 #include <qcd/action/fermion/implementation/WilsonKernelsAsmBodyA64FX.h>

-// #pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-// 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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>

-// #pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-// 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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>


 /////////////////////////////////////////////////////////////////
@@ -371,17 +371,17 @@ WilsonKernels<ZWilsonImplD>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldVi
 						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
 #include <qcd/action/fermion/implementation/WilsonKernelsAsmBodyA64FX.h>

-// #pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-// 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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>

-// #pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-// 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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>


 #undef INTERIOR_AND_EXTERIOR
@@ -400,17 +400,17 @@ WilsonKernels<ZWilsonImplD>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFiel
 						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
 #include <qcd/action/fermion/implementation/WilsonKernelsAsmBodyA64FX.h>

-// #pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-// 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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>

-// #pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-// 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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>


 #undef INTERIOR_AND_EXTERIOR
@@ -429,17 +429,17 @@ WilsonKernels<ZWilsonImplD>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFiel
 						int ss,int ssU,int Ls,int Ns,const FermionFieldView &in, FermionFieldView &out)
 #include <qcd/action/fermion/implementation/WilsonKernelsAsmBodyA64FX.h>

-// #pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-// 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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>

-// #pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
-// 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/WilsonKernelsAsmBodyA64FX.h>
+#pragma GCC optimize ("-O3", "-fno-schedule-insns", "-fno-schedule-insns2")
+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/WilsonKernelsAsmBodyA64FX.h>



--- a/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmAvx512.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsAsmAvx512.h
@@ -74,15 +74,15 @@ WilsonKernels<ZWilsonImplF>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView
 						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>
+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
@@ -97,15 +97,15 @@ WilsonKernels<ZWilsonImplF>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldVi
 						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>
+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
@@ -121,15 +121,15 @@ WilsonKernels<ZWilsonImplF>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldVi
 						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>
+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
@@ -148,15 +148,15 @@ WilsonKernels<ZWilsonImplF>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldVi
 						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>
+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
@@ -171,15 +171,15 @@ WilsonKernels<ZWilsonImplF>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFiel
 						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>
+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
@@ -194,15 +194,15 @@ WilsonKernels<ZWilsonImplF>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFiel
 						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>
+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
@@ -228,14 +228,14 @@ WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSite(StencilView &st, DoubledGaugeF
 							 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>
+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
@@ -249,14 +249,14 @@ WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteInt(StencilView &st, DoubledGau
 							 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>
+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
@@ -273,15 +273,15 @@ WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteExt(StencilView &st, DoubledGau
 							 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>
+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
@@ -299,14 +299,14 @@ WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteDag(StencilView &st, DoubledGau
 							    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>
+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
@@ -320,14 +320,14 @@ WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteDagInt(StencilView &st, Doubled
 							    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>
+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
@@ -341,14 +341,14 @@ WilsonKernels<ZDomainWallVec5dImplF>::AsmDhopSiteDagExt(StencilView &st, Doubled
 							    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>
+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

@@ -392,14 +392,14 @@ WilsonKernels<ZWilsonImplD>::AsmDhopSite(StencilView &st, DoubledGaugeFieldView
 						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>
+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
@@ -413,14 +413,14 @@ WilsonKernels<ZWilsonImplD>::AsmDhopSiteInt(StencilView &st, DoubledGaugeFieldVi
 						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>
+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
@@ -434,14 +434,14 @@ WilsonKernels<ZWilsonImplD>::AsmDhopSiteExt(StencilView &st, DoubledGaugeFieldVi
 						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>
+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
@@ -459,14 +459,14 @@ WilsonKernels<ZWilsonImplD>::AsmDhopSiteDag(StencilView &st, DoubledGaugeFieldVi
 						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>
+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
@@ -480,14 +480,14 @@ WilsonKernels<ZWilsonImplD>::AsmDhopSiteDagInt(StencilView &st, DoubledGaugeFiel
 						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>
+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
@@ -501,14 +501,14 @@ WilsonKernels<ZWilsonImplD>::AsmDhopSiteDagExt(StencilView &st, DoubledGaugeFiel
 						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>
+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
@@ -533,14 +533,14 @@ WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSite(StencilView &st, DoubledGaugeF
 							 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>
+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
@@ -554,14 +554,14 @@ WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteInt(StencilView &st, DoubledGau
 							 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>
+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
@@ -577,14 +577,14 @@ WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteExt(StencilView &st, DoubledGau
 							 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>
+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
@@ -602,14 +602,14 @@ WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteDag(StencilView &st, DoubledGau
 							    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>
+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
@@ -623,14 +623,14 @@ WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteDagInt(StencilView &st, Doubled
 							    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>
+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
@@ -645,14 +645,14 @@ WilsonKernels<ZDomainWallVec5dImplD>::AsmDhopSiteDagExt(StencilView &st, Doubled
 							    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>
+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

--- a/Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h
@@ -77,23 +77,23 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define REGISTER

 #ifdef GRID_SIMT
-#define LOAD_CHIMU(Ptype)		\
+#define LOAD_CHIMU(ptype)		\
  {const SiteSpinor & ref (in[offset]);	\
-    Chimu_00=coalescedReadPermute<Ptype>(ref()(0)(0),perm,lane);	\
-    Chimu_01=coalescedReadPermute<Ptype>(ref()(0)(1),perm,lane);		\
-    Chimu_02=coalescedReadPermute<Ptype>(ref()(0)(2),perm,lane);		\
-    Chimu_10=coalescedReadPermute<Ptype>(ref()(1)(0),perm,lane);		\
-    Chimu_11=coalescedReadPermute<Ptype>(ref()(1)(1),perm,lane);		\
-    Chimu_12=coalescedReadPermute<Ptype>(ref()(1)(2),perm,lane);		\
-    Chimu_20=coalescedReadPermute<Ptype>(ref()(2)(0),perm,lane);		\
-    Chimu_21=coalescedReadPermute<Ptype>(ref()(2)(1),perm,lane);		\
-    Chimu_22=coalescedReadPermute<Ptype>(ref()(2)(2),perm,lane);		\
-    Chimu_30=coalescedReadPermute<Ptype>(ref()(3)(0),perm,lane);		\
-    Chimu_31=coalescedReadPermute<Ptype>(ref()(3)(1),perm,lane);		\
-    Chimu_32=coalescedReadPermute<Ptype>(ref()(3)(2),perm,lane);	}
+    Chimu_00=coalescedReadPermute<ptype>(ref()(0)(0),perm,lane);	\
+    Chimu_01=coalescedReadPermute<ptype>(ref()(0)(1),perm,lane);		\
+    Chimu_02=coalescedReadPermute<ptype>(ref()(0)(2),perm,lane);		\
+    Chimu_10=coalescedReadPermute<ptype>(ref()(1)(0),perm,lane);		\
+    Chimu_11=coalescedReadPermute<ptype>(ref()(1)(1),perm,lane);		\
+    Chimu_12=coalescedReadPermute<ptype>(ref()(1)(2),perm,lane);		\
+    Chimu_20=coalescedReadPermute<ptype>(ref()(2)(0),perm,lane);		\
+    Chimu_21=coalescedReadPermute<ptype>(ref()(2)(1),perm,lane);		\
+    Chimu_22=coalescedReadPermute<ptype>(ref()(2)(2),perm,lane);		\
+    Chimu_30=coalescedReadPermute<ptype>(ref()(3)(0),perm,lane);		\
+    Chimu_31=coalescedReadPermute<ptype>(ref()(3)(1),perm,lane);		\
+    Chimu_32=coalescedReadPermute<ptype>(ref()(3)(2),perm,lane);	}
 #define PERMUTE_DIR(dir) ;
 #else
-#define LOAD_CHIMU(Ptype)		\
+#define LOAD_CHIMU(ptype)		\
  {const SiteSpinor & ref (in[offset]);	\
    Chimu_00=ref()(0)(0);\
    Chimu_01=ref()(0)(1);\
@@ -109,12 +109,12 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
    Chimu_32=ref()(3)(2);}

 #define PERMUTE_DIR(dir)			\
-  permute##dir(Chi_00,Chi_00);			\
-  permute##dir(Chi_01,Chi_01);			\
-  permute##dir(Chi_02,Chi_02);			\
-  permute##dir(Chi_10,Chi_10);			\
-  permute##dir(Chi_11,Chi_11);			\
-  permute##dir(Chi_12,Chi_12);
+  permute##dir(Chi_00,Chi_00);	\
+      permute##dir(Chi_01,Chi_01);\
+      permute##dir(Chi_02,Chi_02);\
+      permute##dir(Chi_10,Chi_10);	\
+      permute##dir(Chi_11,Chi_11);\
+      permute##dir(Chi_12,Chi_12);

 #endif

@@ -371,91 +371,88 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
  result_32-= UChi_12;

 #define HAND_STENCIL_LEGB(PROJ,PERM,DIR,RECON)	\
-  {int ptype;					\
-   SE=st.GetEntry(ptype,DIR,ss);		\
-   auto offset = SE->_offset;			\
-   auto local  = SE->_is_local;			\
-   auto perm   = SE->_permute;			\
-   if ( local ) {				\
-     LOAD_CHIMU(PERM);				\
-     PROJ;					\
-     if ( perm) {				\
-       PERMUTE_DIR(PERM);			\
-     }						\
-   } else {					\
-     LOAD_CHI;					\
-   }						\
-   acceleratorSynchronise();			\
-   MULT_2SPIN(DIR);				\
-   RECON;					}
+  SE=st.GetEntry(ptype,DIR,ss);			\
+  offset = SE->_offset;				\
+  local  = SE->_is_local;			\
+  perm   = SE->_permute;			\
+  if ( local ) {				\
+    LOAD_CHIMU(PERM);				\
+    PROJ;					\
+    if ( perm) {				\
+      PERMUTE_DIR(PERM);			\
+    }						\
+  } else {					\
+    LOAD_CHI;					\
+  }						\
+  acceleratorSynchronise();			\
+  MULT_2SPIN(DIR);				\
+  RECON;					

-#define HAND_STENCIL_LEG(PROJ,PERM,DIR,RECON)		\
-  { SE=&st_p[DIR+8*ss];						\
-  auto ptype=st_perm[DIR];					\
-  auto offset = SE->_offset;					\
-  auto local  = SE->_is_local;					\
-  auto perm   = SE->_permute;					\
-  if ( local ) {						\
-    LOAD_CHIMU(PERM);						\
-    PROJ;							\
-    if ( perm) {						\
-      PERMUTE_DIR(PERM);					\
-    }								\
-  } else {							\
-    LOAD_CHI;							\
-  }								\
-  acceleratorSynchronise();					\
-  MULT_2SPIN(DIR);						\
-  RECON;					}
+#define HAND_STENCIL_LEG(PROJ,PERM,DIR,RECON)	\
+  SE=&st_p[DIR+8*ss];				\
+  ptype=st_perm[DIR];				\
+  offset = SE->_offset;				\
+  local  = SE->_is_local;			\
+  perm   = SE->_permute;			\
+  if ( local ) {				\
+    LOAD_CHIMU(PERM);				\
+    PROJ;					\
+    if ( perm) {				\
+      PERMUTE_DIR(PERM);			\
+    }						\
+  } else {					\
+    LOAD_CHI;					\
+  }						\
+  acceleratorSynchronise();			\
+  MULT_2SPIN(DIR);				\
+  RECON;					

 #define HAND_STENCIL_LEGA(PROJ,PERM,DIR,RECON)				\
-  { SE=&st_p[DIR+8*ss];							\
-    auto ptype=st_perm[DIR];						\
-    /*SE=st.GetEntry(ptype,DIR,ss);*/					\
-    auto offset = SE->_offset;						\
-    auto perm   = SE->_permute;						\
-    LOAD_CHIMU(PERM);							\
-    PROJ;								\
-    MULT_2SPIN(DIR);							\
-    RECON;					}
+  SE=&st_p[DIR+8*ss];							\
+  ptype=st_perm[DIR];							\
+ /*SE=st.GetEntry(ptype,DIR,ss);*/					\
+  offset = SE->_offset;				\
+  perm   = SE->_permute;			\
+  LOAD_CHIMU(PERM);				\
+  PROJ;						\
+  MULT_2SPIN(DIR);				\
+  RECON;					

 #define HAND_STENCIL_LEG_INT(PROJ,PERM,DIR,RECON)	\
-  { int ptype;						\
-  SE=st.GetEntry(ptype,DIR,ss);				\
-  auto offset = SE->_offset;					\
-  auto local  = SE->_is_local;					\
-  auto perm   = SE->_permute;					\
-  if ( local ) {						\
-    LOAD_CHIMU(PERM);						\
-    PROJ;							\
-    if ( perm) {						\
-      PERMUTE_DIR(PERM);					\
-    }								\
-  } else if ( st.same_node[DIR] ) {				\
-    LOAD_CHI;							\
-  }								\
-  acceleratorSynchronise();					\
-  if (local || st.same_node[DIR] ) {				\
-    MULT_2SPIN(DIR);						\
-    RECON;							\
-  }								\
-  acceleratorSynchronise();			}
+  SE=st.GetEntry(ptype,DIR,ss);			\
+  offset = SE->_offset;				\
+  local  = SE->_is_local;			\
+  perm   = SE->_permute;			\
+  if ( local ) {				\
+    LOAD_CHIMU(PERM);				\
+    PROJ;					\
+    if ( perm) {				\
+      PERMUTE_DIR(PERM);			\
+    }						\
+  } else if ( st.same_node[DIR] ) {		\
+    LOAD_CHI;					\
+  }						\
+  acceleratorSynchronise();			\
+  if (local || st.same_node[DIR] ) {		\
+    MULT_2SPIN(DIR);				\
+    RECON;					\
+  }						\
+  acceleratorSynchronise();			

 #define HAND_STENCIL_LEG_EXT(PROJ,PERM,DIR,RECON)	\
-  { int ptype;						\
-  SE=st.GetEntry(ptype,DIR,ss);				\
-  auto offset = SE->_offset;				\
-  if((!SE->_is_local)&&(!st.same_node[DIR]) ) {		\
-    LOAD_CHI;						\
-    MULT_2SPIN(DIR);					\
-    RECON;						\
-    nmu++;						\
-  }							\
-  acceleratorSynchronise();			}
+  SE=st.GetEntry(ptype,DIR,ss);			\
+  offset = SE->_offset;				\
+  if((!SE->_is_local)&&(!st.same_node[DIR]) ) {	\
+    LOAD_CHI;					\
+    MULT_2SPIN(DIR);				\
+    RECON;					\
+    nmu++;					\
+  }						\
+  acceleratorSynchronise();			

-#define HAND_RESULT(ss)					\
-  {							\
-    SiteSpinor & ref (out[ss]);				\
+#define HAND_RESULT(ss)				\
+  {						\
+    SiteSpinor & ref (out[ss]);			\
    coalescedWrite(ref()(0)(0),result_00,lane);		\
    coalescedWrite(ref()(0)(1),result_01,lane);		\
    coalescedWrite(ref()(0)(2),result_02,lane);		\
@@ -566,6 +563,7 @@ WilsonKernels<Impl>::HandDhopSiteSycl(StencilVector st_perm,StencilEntry *st_p,

  HAND_DECLARATIONS(Simt);

+  int offset,local,perm, ptype;
  StencilEntry *SE;
  HAND_STENCIL_LEG(XM_PROJ,3,Xp,XM_RECON);
  HAND_STENCIL_LEG(YM_PROJ,2,Yp,YM_RECON_ACCUM);
@@ -595,7 +593,9 @@ WilsonKernels<Impl>::HandDhopSite(StencilView &st, DoubledGaugeFieldView &U,Site

  HAND_DECLARATIONS(Simt);

+  int offset,local,perm, ptype;
  StencilEntry *SE;
+
  HAND_STENCIL_LEG(XM_PROJ,3,Xp,XM_RECON);
  HAND_STENCIL_LEG(YM_PROJ,2,Yp,YM_RECON_ACCUM);
  HAND_STENCIL_LEG(ZM_PROJ,1,Zp,ZM_RECON_ACCUM);
@@ -623,6 +623,8 @@ void WilsonKernels<Impl>::HandDhopSiteDag(StencilView &st,DoubledGaugeFieldView
  HAND_DECLARATIONS(Simt);

  StencilEntry *SE;
+  int offset,local,perm, ptype;
+  
  HAND_STENCIL_LEG(XP_PROJ,3,Xp,XP_RECON);
  HAND_STENCIL_LEG(YP_PROJ,2,Yp,YP_RECON_ACCUM);
  HAND_STENCIL_LEG(ZP_PROJ,1,Zp,ZP_RECON_ACCUM);
@@ -638,8 +640,8 @@ template<class Impl>  accelerator_inline void
 WilsonKernels<Impl>::HandDhopSiteInt(StencilView &st,DoubledGaugeFieldView &U,SiteHalfSpinor  *buf,
 					  int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
 {
-  //  auto st_p = st._entries_p;						
-  //  auto st_perm = st._permute_type;					
+  auto st_p = st._entries_p;						
+  auto st_perm = st._permute_type;					
 // 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;
@@ -650,6 +652,7 @@ WilsonKernels<Impl>::HandDhopSiteInt(StencilView &st,DoubledGaugeFieldView &U,Si

  HAND_DECLARATIONS(Simt);

+  int offset,local,perm, ptype;
  StencilEntry *SE;
  ZERO_RESULT;
  HAND_STENCIL_LEG_INT(XM_PROJ,3,Xp,XM_RECON_ACCUM);
@@ -667,8 +670,8 @@ template<class Impl> accelerator_inline
 void WilsonKernels<Impl>::HandDhopSiteDagInt(StencilView &st,DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
 						  int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
 {
-  //  auto st_p = st._entries_p;						
-  //  auto st_perm = st._permute_type;					
+  auto st_p = st._entries_p;						
+  auto st_perm = st._permute_type;					
  typedef typename Simd::scalar_type S;
  typedef typename Simd::vector_type V;
  typedef decltype( coalescedRead( in[0]()(0)(0) )) Simt;
@@ -679,6 +682,7 @@ void WilsonKernels<Impl>::HandDhopSiteDagInt(StencilView &st,DoubledGaugeFieldVi
  HAND_DECLARATIONS(Simt);

  StencilEntry *SE;
+  int offset,local,perm, ptype;
  ZERO_RESULT;
  HAND_STENCIL_LEG_INT(XP_PROJ,3,Xp,XP_RECON_ACCUM);
  HAND_STENCIL_LEG_INT(YP_PROJ,2,Yp,YP_RECON_ACCUM);
@@ -695,8 +699,8 @@ template<class Impl>  accelerator_inline void
 WilsonKernels<Impl>::HandDhopSiteExt(StencilView &st,DoubledGaugeFieldView &U,SiteHalfSpinor  *buf,
 					  int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
 {
-  //  auto st_p = st._entries_p;						
-  //  auto st_perm = st._permute_type;					
+  auto st_p = st._entries_p;						
+  auto st_perm = st._permute_type;					
 // 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;
@@ -707,7 +711,7 @@ WilsonKernels<Impl>::HandDhopSiteExt(StencilView &st,DoubledGaugeFieldView &U,Si

  HAND_DECLARATIONS(Simt);

-  //  int offset, ptype;
+  int offset, ptype;
  StencilEntry *SE;
  int nmu=0;
  ZERO_RESULT;
@@ -726,8 +730,8 @@ template<class Impl>  accelerator_inline
 void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilView &st,DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
 						  int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
 {
-  //  auto st_p = st._entries_p;						
-  //  auto st_perm = st._permute_type;					
+  auto st_p = st._entries_p;						
+  auto st_perm = st._permute_type;					
  typedef typename Simd::scalar_type S;
  typedef typename Simd::vector_type V;
  typedef decltype( coalescedRead( in[0]()(0)(0) )) Simt;
@@ -738,7 +742,7 @@ void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilView &st,DoubledGaugeFieldVi
  HAND_DECLARATIONS(Simt);

  StencilEntry *SE;
-  //  int offset, ptype;
+  int offset, ptype;
  int nmu=0;
  ZERO_RESULT;
  HAND_STENCIL_LEG_EXT(XP_PROJ,3,Xp,XP_RECON_ACCUM);
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/CayleyFermion5DInstantiationGparityWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/CayleyFermion5DInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1 @@
+../CayleyFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/ContinuedFractionFermion5DInstantiationGparityWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/ContinuedFractionFermion5DInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1 @@
+../ContinuedFractionFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/DomainWallEOFAFermionInstantiationGparityWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/DomainWallEOFAFermionInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1 @@
+../DomainWallEOFAFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/MobiusEOFAFermionInstantiationGparityWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/MobiusEOFAFermionInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1 @@
+../MobiusEOFAFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/PartialFractionFermion5DInstantiationGparityWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/PartialFractionFermion5DInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1 @@
+../PartialFractionFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonCloverFermionInstantiationGparityWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonCloverFermionInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1 @@
+../WilsonCloverFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonFermion5DInstantiationGparityWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonFermion5DInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1 @@
+../WilsonFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonFermionInstantiationGparityWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonFermionInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1 @@
+../WilsonFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonKernelsInstantiationGparityWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonKernelsInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1 @@
+../WilsonKernelsInstantiationGparity.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonTMFermionInstantiationGparityWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/WilsonTMFermionInstantiationGparityWilsonImplDF.cc
@@ -0,0 +1 @@
+../WilsonTMFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/impl.h
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplDF/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION GparityWilsonImplDF
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/CayleyFermion5DInstantiationGparityWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/CayleyFermion5DInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1 @@
+../CayleyFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/ContinuedFractionFermion5DInstantiationGparityWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/ContinuedFractionFermion5DInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1 @@
+../ContinuedFractionFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/DomainWallEOFAFermionInstantiationGparityWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/DomainWallEOFAFermionInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1 @@
+../DomainWallEOFAFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/MobiusEOFAFermionInstantiationGparityWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/MobiusEOFAFermionInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1 @@
+../MobiusEOFAFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/PartialFractionFermion5DInstantiationGparityWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/PartialFractionFermion5DInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1 @@
+../PartialFractionFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonCloverFermionInstantiationGparityWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonCloverFermionInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1 @@
+../WilsonCloverFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonFermion5DInstantiationGparityWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonFermion5DInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1 @@
+../WilsonFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonFermionInstantiationGparityWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonFermionInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1 @@
+../WilsonFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonKernelsInstantiationGparityWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonKernelsInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1 @@
+../WilsonKernelsInstantiationGparity.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonTMFermionInstantiationGparityWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/WilsonTMFermionInstantiationGparityWilsonImplFH.cc
@@ -0,0 +1 @@
+../WilsonTMFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/impl.h
+++ b/Grid/qcd/action/fermion/instantiation/GparityWilsonImplFH/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION GparityWilsonImplFH
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/CayleyFermion5DInstantiationWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/CayleyFermion5DInstantiationWilsonImplDF.cc
@@ -0,0 +1 @@
+../CayleyFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/ContinuedFractionFermion5DInstantiationWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/ContinuedFractionFermion5DInstantiationWilsonImplDF.cc
@@ -0,0 +1 @@
+../ContinuedFractionFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/DomainWallEOFAFermionInstantiationWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/DomainWallEOFAFermionInstantiationWilsonImplDF.cc
@@ -0,0 +1 @@
+../DomainWallEOFAFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/MobiusEOFAFermionInstantiationWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/MobiusEOFAFermionInstantiationWilsonImplDF.cc
@@ -0,0 +1 @@
+../MobiusEOFAFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/PartialFractionFermion5DInstantiationWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/PartialFractionFermion5DInstantiationWilsonImplDF.cc
@@ -0,0 +1 @@
+../PartialFractionFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonCloverFermionInstantiationWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonCloverFermionInstantiationWilsonImplDF.cc
@@ -0,0 +1 @@
+../WilsonCloverFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonFermion5DInstantiationWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonFermion5DInstantiationWilsonImplDF.cc
@@ -0,0 +1 @@
+../WilsonFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonFermionInstantiationWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonFermionInstantiationWilsonImplDF.cc
@@ -0,0 +1 @@
+../WilsonFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonKernelsInstantiationWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonKernelsInstantiationWilsonImplDF.cc
@@ -2,11 +2,14 @@

 Grid physics library, www.github.com/paboyle/Grid

-Source file: ./lib/serialisation/BaseIO.h
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc

-Copyright (C) 2015
+Copyright (C) 2015, 2020

-Author: Michael Marshall <michael.marshall@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: Nils Meyer <nils.meyer@ur.de> Regensburg University

 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
@@ -22,14 +25,27 @@ 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 */
+#include <Grid/qcd/action/fermion/FermionCore.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h>
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsHandImplementation.h>

-#include <Grid/GridCore.h>
+#ifndef AVX512
+#ifndef QPX
+#ifndef A64FX
+#ifndef A64FXFIXEDSIZE
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+#endif
+#endif

-NAMESPACE_BEGIN(Grid)
+NAMESPACE_BEGIN(Grid);

-std::uint64_t EigenIO::EigenResizeCounter(0);
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>;

-NAMESPACE_END(Grid)
+NAMESPACE_END(Grid);
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonTMFermionInstantiationWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/WilsonTMFermionInstantiationWilsonImplDF.cc
@@ -0,0 +1 @@
+../WilsonTMFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplDF/impl.h
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplDF/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION WilsonImplDF
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/CayleyFermion5DInstantiationWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/CayleyFermion5DInstantiationWilsonImplFH.cc
@@ -0,0 +1 @@
+../CayleyFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/ContinuedFractionFermion5DInstantiationWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/ContinuedFractionFermion5DInstantiationWilsonImplFH.cc
@@ -0,0 +1 @@
+../ContinuedFractionFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/DomainWallEOFAFermionInstantiationWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/DomainWallEOFAFermionInstantiationWilsonImplFH.cc
@@ -0,0 +1 @@
+../DomainWallEOFAFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/MobiusEOFAFermionInstantiationWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/MobiusEOFAFermionInstantiationWilsonImplFH.cc
@@ -0,0 +1 @@
+../MobiusEOFAFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/PartialFractionFermion5DInstantiationWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/PartialFractionFermion5DInstantiationWilsonImplFH.cc
@@ -0,0 +1 @@
+../PartialFractionFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonCloverFermionInstantiationWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonCloverFermionInstantiationWilsonImplFH.cc
@@ -0,0 +1 @@
+../WilsonCloverFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonFermion5DInstantiationWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonFermion5DInstantiationWilsonImplFH.cc
@@ -0,0 +1 @@
+../WilsonFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonFermionInstantiationWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonFermionInstantiationWilsonImplFH.cc
@@ -0,0 +1 @@
+../WilsonFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonKernelsInstantiationWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonKernelsInstantiationWilsonImplFH.cc
@@ -0,0 +1,51 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015, 2020
+
+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: Nils Meyer <nils.meyer@ur.de> Regensburg University
+
+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
+#ifndef A64FX
+#ifndef A64FXFIXEDSIZE
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>;
+
+NAMESPACE_END(Grid);
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonTMFermionInstantiationWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/WilsonTMFermionInstantiationWilsonImplFH.cc
@@ -0,0 +1 @@
+../WilsonTMFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/WilsonImplFH/impl.h
+++ b/Grid/qcd/action/fermion/instantiation/WilsonImplFH/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION WilsonImplFH
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/CayleyFermion5DInstantiationZWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/CayleyFermion5DInstantiationZWilsonImplDF.cc
@@ -0,0 +1 @@
+../CayleyFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/ContinuedFractionFermion5DInstantiationZWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/ContinuedFractionFermion5DInstantiationZWilsonImplDF.cc
@@ -0,0 +1 @@
+../ContinuedFractionFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/DomainWallEOFAFermionInstantiationZWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/DomainWallEOFAFermionInstantiationZWilsonImplDF.cc
@@ -0,0 +1 @@
+../DomainWallEOFAFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/MobiusEOFAFermionInstantiationZWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/MobiusEOFAFermionInstantiationZWilsonImplDF.cc
@@ -0,0 +1 @@
+../MobiusEOFAFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/PartialFractionFermion5DInstantiationZWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/PartialFractionFermion5DInstantiationZWilsonImplDF.cc
@@ -0,0 +1 @@
+../PartialFractionFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/WilsonFermion5DInstantiationZWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/WilsonFermion5DInstantiationZWilsonImplDF.cc
@@ -0,0 +1 @@
+../WilsonFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/WilsonKernelsInstantiationZWilsonImplDF.cc
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/WilsonKernelsInstantiationZWilsonImplDF.cc
@@ -0,0 +1,51 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015, 2020
+
+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: Nils Meyer <nils.meyer@ur.de> Regensburg University
+
+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
+#ifndef A64FX
+#ifndef A64FXFIXEDSIZE
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>;
+
+NAMESPACE_END(Grid);
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/impl.h
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplDF/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION ZWilsonImplDF
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/CayleyFermion5DInstantiationZWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/CayleyFermion5DInstantiationZWilsonImplFH.cc
@@ -0,0 +1 @@
+../CayleyFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/ContinuedFractionFermion5DInstantiationZWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/ContinuedFractionFermion5DInstantiationZWilsonImplFH.cc
@@ -0,0 +1 @@
+../ContinuedFractionFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/DomainWallEOFAFermionInstantiationZWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/DomainWallEOFAFermionInstantiationZWilsonImplFH.cc
@@ -0,0 +1 @@
+../DomainWallEOFAFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/MobiusEOFAFermionInstantiationZWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/MobiusEOFAFermionInstantiationZWilsonImplFH.cc
@@ -0,0 +1 @@
+../MobiusEOFAFermionInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/PartialFractionFermion5DInstantiationZWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/PartialFractionFermion5DInstantiationZWilsonImplFH.cc
@@ -0,0 +1 @@
+../PartialFractionFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/WilsonFermion5DInstantiationZWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/WilsonFermion5DInstantiationZWilsonImplFH.cc
@@ -0,0 +1 @@
+../WilsonFermion5DInstantiation.cc.master
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/WilsonKernelsInstantiationZWilsonImplFH.cc
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/WilsonKernelsInstantiationZWilsonImplFH.cc
@@ -0,0 +1,51 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+
+Copyright (C) 2015, 2020
+
+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: Nils Meyer <nils.meyer@ur.de> Regensburg University
+
+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
+#ifndef A64FX
+#ifndef A64FXFIXEDSIZE
+#include <Grid/qcd/action/fermion/implementation/WilsonKernelsAsmImplementation.h>
+#endif
+#endif
+#endif
+#endif
+
+NAMESPACE_BEGIN(Grid);
+
+#include "impl.h"
+template class WilsonKernels<IMPLEMENTATION>;
+
+NAMESPACE_END(Grid);
--- a/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/impl.h
+++ b/Grid/qcd/action/fermion/instantiation/ZWilsonImplFH/impl.h
@@ -0,0 +1 @@
+#define IMPLEMENTATION ZWilsonImplFH
--- a/Grid/qcd/action/fermion/instantiation/generate_instantiations.sh
+++ b/Grid/qcd/action/fermion/instantiation/generate_instantiations.sh
@@ -9,6 +9,8 @@ STAG5_IMPL_LIST=""
 WILSON_IMPL_LIST=" \
 	   WilsonImplF \
 	   WilsonImplD \
+	   WilsonImplFH \
+	   WilsonImplDF \
 	   WilsonAdjImplF \
 	   WilsonAdjImplD \
 	   WilsonTwoIndexSymmetricImplF \
@@ -16,17 +18,26 @@ WILSON_IMPL_LIST=" \
 	   WilsonTwoIndexAntiSymmetricImplF \
 	   WilsonTwoIndexAntiSymmetricImplD \
 	   GparityWilsonImplF \
-	   GparityWilsonImplD "
+	   GparityWilsonImplD \
+	   GparityWilsonImplFH \
+	   GparityWilsonImplDF"

 DWF_IMPL_LIST=" \
 	   WilsonImplF \
 	   WilsonImplD \
+	   WilsonImplFH \
+	   WilsonImplDF \
 	   ZWilsonImplF \
-	   ZWilsonImplD "
+	   ZWilsonImplD \
+	   ZWilsonImplFH \
+	   ZWilsonImplDF "

 GDWF_IMPL_LIST=" \
 	   GparityWilsonImplF \
-	   GparityWilsonImplD "
+	   GparityWilsonImplD \
+	   GparityWilsonImplFH \
+	   GparityWilsonImplDF"
+

 IMPL_LIST="$STAG_IMPL_LIST  $WILSON_IMPL_LIST $DWF_IMPL_LIST $GDWF_IMPL_LIST"

--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Chulwoo Jung	b2493d6d25	Switching Block lanczos precision to explicitly single Adding sample run script and input file	2022-03-08 10:18:36 -08:00
Chulwoo Jung	8fd16686dc	Checking block lanczos deleted some diag outputs	2021-09-05 23:04:41 -04:00
Chulwoo Jung	23b9c6b5f5	Merge branch 'develop' of https://github.com/paboyle/Grid into feature/block_lanczos	2021-09-03 17:38:10 -04:00
Chulwoo Jung	43943008bf	Merge branch 'develop' of https://github.com/paboyle/Grid into feature/block_lanczos	2020-10-01 15:35:29 -04:00
Chulwoo Jung	8ee26f9112	Checking in before pulling develop	2020-10-01 15:35:03 -04:00
Chulwoo Jung	f91e3af97f	Checking in before trying to reduce memory footprint	2020-08-08 22:11:14 -04:00
Chulwoo Jung	43298ef681	Make precision switchable for cublas routines	2020-07-22 14:49:32 -04:00
Chulwoo Jung	7e70df27e4	Confirmed double precision working	2020-07-21 00:48:46 -04:00
Chulwoo Jung	c55d657736	Merge branch 'dev-BlockLanczosOpt' of https://github.com/yongchull/Grid into feature/block_lanczos	2020-07-20 16:36:46 -04:00
Yong-Chull Jang	b89b1280d5	use gemm twice to complete the Gram Schmidt	2020-03-31 05:39:31 -04:00
Yong-Chull Jang	ac7090e6d3	block Lanczos cublas buffer is set at the inital step; buffer width is fixed to the block size then cublas Zgemm is called multiple times	2020-03-30 22:25:50 -04:00
Yong-Chull Jang	02edbe624f	first working version of Gram Schmidt using cublas gemm; explicit data type and site vector size has to be removed	2020-03-30 18:36:21 -04:00
Yong-Chull Jang	9266b89ad8	fix rngs issue; block Lanczos is working	2020-03-25 15:45:50 -04:00
Yong-Chull Jang	2db7e6f8ab	merge manually Block Lanczos files from Chulwoo's update (last state = commit 731a05 + untracked files) to develop branch; namespace QCD is removed; FIXME: multiple starting vectors result in nan after initial orthogonalization	2020-03-24 01:03:24 -04:00
				`@@ -0,0 +1 @@`
				`../ContinuedFractionFermion5DInstantiation.cc.master`
				`@@ -0,0 +1 @@`
				`../DomainWallEOFAFermionInstantiation.cc.master`
				`@@ -0,0 +1 @@`
				`../MobiusEOFAFermionInstantiation.cc.master`
				`@@ -0,0 +1 @@`
				`../PartialFractionFermion5DInstantiation.cc.master`
				`@@ -0,0 +1 @@`
				`../WilsonCloverFermionInstantiation.cc.master`
				`@@ -0,0 +1 @@`
				`../WilsonKernelsInstantiationGparity.cc.master`