Update to static data

Grid/DisableWarnings.h

@@ -44,22 +44,14 @@ directory
 #ifdef __NVCC__
  //disables nvcc specific warning in json.hpp
 #pragma clang diagnostic ignored "-Wdeprecated-register"
-
-#if (__CUDACC_VER_MAJOR__ >= 11) && (__CUDACC_VER_MINOR__ >= 5)
- //disables nvcc specific warning in json.hpp
-#pragma nv_diag_suppress unsigned_compare_with_zero
-#pragma nv_diag_suppress cast_to_qualified_type
- //disables nvcc specific warning in many files
-#pragma nv_diag_suppress esa_on_defaulted_function_ignored
-#pragma nv_diag_suppress extra_semicolon
-#else
- //disables nvcc specific warning in json.hpp
 #pragma diag_suppress unsigned_compare_with_zero
 #pragma diag_suppress cast_to_qualified_type
+
  //disables nvcc specific warning in many files
 #pragma diag_suppress esa_on_defaulted_function_ignored
 #pragma diag_suppress extra_semicolon
-#endif
+
+//Eigen only
 #endif
 
 // Disable vectorisation in Eigen on the Power8/9 and PowerPC

Grid/GridStd.h

@@ -16,7 +16,6 @@
 #include <functional>
 #include <stdio.h>
 #include <stdlib.h>
-#include <strings.h>
 #include <stdio.h>
 #include <signal.h>
 #include <ctime>

Grid/Grid_Eigen_Dense.h

@@ -14,11 +14,7 @@
 /* NVCC save and restore compile environment*/
 #ifdef __NVCC__
 #pragma push
-#if (__CUDACC_VER_MAJOR__ >= 11) && (__CUDACC_VER_MINOR__ >= 5)
-#pragma nv_diag_suppress code_is_unreachable
-#else
 #pragma diag_suppress code_is_unreachable
-#endif
 #pragma push_macro("__CUDA_ARCH__")
 #pragma push_macro("__NVCC__")
 #pragma push_macro("__CUDACC__")

Grid/algorithms/CoarsenedMatrix.h

@@ -262,7 +262,7 @@ public:
 	autoView( Tnp_v , (*Tnp), AcceleratorWrite);
 	autoView( Tnm_v , (*Tnm), AcceleratorWrite);
 	const int Nsimd = CComplex::Nsimd();
-	accelerator_for(ss, FineGrid->oSites(), Nsimd, {
+	accelerator_forNB(ss, FineGrid->oSites(), Nsimd, {
 	  coalescedWrite(y_v[ss],xscale*y_v(ss)+mscale*Tn_v(ss));
 	  coalescedWrite(Tnp_v[ss],2.0*y_v(ss)-Tnm_v(ss));
         });

Grid/algorithms/approx/Chebyshev.h

@@ -264,7 +264,7 @@ public:
       auto Tnp_v = Tnp->View();
       auto Tnm_v = Tnm->View();
       constexpr int Nsimd = vector_type::Nsimd();
-      accelerator_for(ss, in.Grid()->oSites(), Nsimd, {
+      accelerator_forNB(ss, in.Grid()->oSites(), Nsimd, {
 	  coalescedWrite(y_v[ss],xscale*y_v(ss)+mscale*Tn_v(ss));
 	  coalescedWrite(Tnp_v[ss],2.0*y_v(ss)-Tnm_v(ss));
       });

Grid/algorithms/iterative/Deflation.h

@@ -113,43 +113,7 @@ public:
     blockPromote(guess_coarse,guess,subspace);
     guess.Checkerboard() = src.Checkerboard();
   };
-
+};
-  void operator()(const std::vector<FineField> &src,std::vector<FineField> &guess) {
-    int Nevec = (int)evec_coarse.size();
-    int Nsrc = (int)src.size();
-    // make temp variables
-    std::vector<CoarseField> src_coarse(Nsrc,evec_coarse[0].Grid());
-    std::vector<CoarseField> guess_coarse(Nsrc,evec_coarse[0].Grid());    
-    //Preporcessing
-    std::cout << GridLogMessage << "Start BlockProject for loop" << std::endl;
-    for (int j=0;j<Nsrc;j++)
-    {
-    guess_coarse[j] = Zero();
-    std::cout << GridLogMessage << "BlockProject iter: " << j << std::endl;
-    blockProject(src_coarse[j],src[j],subspace);
-    }
-    //deflation set up for eigen vector batchsize 1 and source batch size equal number of sources
-    std::cout << GridLogMessage << "Start ProjectAccum for loop" << std::endl;
-    for (int i=0;i<Nevec;i++)
-    {
-      std::cout << GridLogMessage << "ProjectAccum Nvec: " << i << std::endl;
-      const CoarseField & tmp = evec_coarse[i];
-      for (int j=0;j<Nsrc;j++)
-      {
-        axpy(guess_coarse[j],TensorRemove(innerProduct(tmp,src_coarse[j])) / eval_coarse[i],tmp,guess_coarse[j]);
-      }
-    }
-    //postprocessing
-    std::cout << GridLogMessage << "Start BlockPromote for loop" << std::endl;
-    for (int j=0;j<Nsrc;j++)
-    {
-    std::cout << GridLogMessage << "BlockProject iter: " << j << std::endl;
-    blockPromote(guess_coarse[j],guess[j],subspace);
-    guess[j].Checkerboard() = src[j].Checkerboard();
-    }
-  };
-
-  };
 
 
 

Grid/allocator/MemoryManager.h

@@ -113,6 +113,11 @@ 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
@@ -152,6 +157,7 @@ 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

Grid/allocator/MemoryManagerCache.cc

@@ -23,6 +23,11 @@ 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
@@ -86,6 +91,14 @@ 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);
@@ -129,6 +142,7 @@ 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);
   }
@@ -231,6 +245,9 @@ 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) {
@@ -349,6 +366,10 @@ void MemoryManager::CpuViewClose(uint64_t CpuPtr)
   assert(AccCache.accLock==0);
 
   AccCache.cpuLock--;
+
+  if(AccCache.cpuLock==0) {
+    LRUinsertback(AccCache);
+  }
 }
 /*
  *  Action  State   StateNext         Flush    Clone
@@ -371,6 +392,9 @@ 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;
 
@@ -416,6 +440,12 @@ 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)

Grid/allocator/MemoryManagerShared.cc

@@ -12,6 +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; };

Grid/communicator/Communicator_mpi3.cc

@@ -392,9 +392,9 @@ double CartesianCommunicator::StencilSendToRecvFromBegin(std::vector<CommsReques
     acceleratorCopyDeviceToDeviceAsynch(xmit,shm,bytes);
   }
 
-  //  if ( CommunicatorPolicy == CommunicatorPolicySequential ) {
+  if ( CommunicatorPolicy == CommunicatorPolicySequential ) {
-  //    this->StencilSendToRecvFromComplete(list,dir);
+    this->StencilSendToRecvFromComplete(list,dir);
-  //  }
+  }
 
   return off_node_bytes;
 }

Grid/lattice/Lattice_reduction.h

@@ -142,15 +142,6 @@ inline typename vobj::scalar_objectD sumD(const vobj *arg, Integer osites)
   return sumD_cpu(arg,osites);
 #endif  
 }
-template<class vobj>
-inline typename vobj::scalar_objectD sumD_large(const vobj *arg, Integer osites)
-{
-#if defined(GRID_CUDA)||defined(GRID_HIP)
-  return sumD_gpu_large(arg,osites);
-#else
-  return sumD_cpu(arg,osites);
-#endif  
-}
 
 template<class vobj>
 inline typename vobj::scalar_object sum(const Lattice<vobj> &arg)
@@ -168,22 +159,6 @@ inline typename vobj::scalar_object sum(const Lattice<vobj> &arg)
   return ssum;
 }
 
-template<class vobj>
-inline typename vobj::scalar_object sum_large(const Lattice<vobj> &arg)
-{
-#if defined(GRID_CUDA)||defined(GRID_HIP)
-  autoView( arg_v, arg, AcceleratorRead);
-  Integer osites = arg.Grid()->oSites();
-  auto ssum= sum_gpu_large(&arg_v[0],osites);
-#else
-  autoView(arg_v, arg, CpuRead);
-  Integer osites = arg.Grid()->oSites();
-  auto ssum= sum_cpu(&arg_v[0],osites);
-#endif
-  arg.Grid()->GlobalSum(ssum);
-  return ssum;
-}
-
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 // Deterministic Reduction operations
 ////////////////////////////////////////////////////////////////////////////////////////////////////

Grid/lattice/Lattice_reduction_gpu.h

@@ -23,7 +23,7 @@ unsigned int nextPow2(Iterator x) {
 }
 
 template <class Iterator>
-int getNumBlocksAndThreads(const Iterator n, const size_t sizeofsobj, Iterator &threads, Iterator &blocks) {
+void getNumBlocksAndThreads(const Iterator n, const size_t sizeofsobj, Iterator &threads, Iterator &blocks) {
   
   int device;
 #ifdef GRID_CUDA
@@ -37,13 +37,13 @@ int getNumBlocksAndThreads(const Iterator n, const size_t sizeofsobj, Iterator &
   Iterator sharedMemPerBlock   = gpu_props[device].sharedMemPerBlock;
   Iterator maxThreadsPerBlock  = gpu_props[device].maxThreadsPerBlock;
   Iterator multiProcessorCount = gpu_props[device].multiProcessorCount;
-  /*  
+  
   std::cout << GridLogDebug << "GPU has:" << std::endl;
   std::cout << GridLogDebug << "\twarpSize            = " << warpSize << std::endl;
   std::cout << GridLogDebug << "\tsharedMemPerBlock   = " << sharedMemPerBlock << std::endl;
   std::cout << GridLogDebug << "\tmaxThreadsPerBlock  = " << maxThreadsPerBlock << std::endl;
   std::cout << GridLogDebug << "\tmultiProcessorCount = " << multiProcessorCount << std::endl;
-  */  
+  
   if (warpSize != WARP_SIZE) {
     std::cout << GridLogError << "The warp size of the GPU in use does not match the warp size set when compiling Grid." << std::endl;
     exit(EXIT_FAILURE);
@@ -53,12 +53,12 @@ int getNumBlocksAndThreads(const Iterator n, const size_t sizeofsobj, Iterator &
   threads = warpSize;
   if ( threads*sizeofsobj > sharedMemPerBlock ) {
     std::cout << GridLogError << "The object is too large for the shared memory." << std::endl;
-    return 0;
+    exit(EXIT_FAILURE);
   }
   while( 2*threads*sizeofsobj < sharedMemPerBlock && 2*threads <= maxThreadsPerBlock ) threads *= 2;
   // keep all the streaming multiprocessors busy
   blocks = nextPow2(multiProcessorCount);
-  return 1;
+  
 }
 
 template <class sobj, class Iterator>
@@ -198,7 +198,7 @@ __global__ void reduceKernel(const vobj *lat, sobj *buffer, Iterator n) {
 // Possibly promote to double and sum
 /////////////////////////////////////////////////////////////////////////////////////////////////////////
 template <class vobj>
-inline typename vobj::scalar_objectD sumD_gpu_small(const vobj *lat, Integer osites) 
+inline typename vobj::scalar_objectD sumD_gpu(const vobj *lat, Integer osites) 
 {
   typedef typename vobj::scalar_objectD sobj;
   typedef decltype(lat) Iterator;
@@ -207,9 +207,7 @@ inline typename vobj::scalar_objectD sumD_gpu_small(const vobj *lat, Integer osi
   Integer size = osites*nsimd;
 
   Integer numThreads, numBlocks;
-  int ok = getNumBlocksAndThreads(size, sizeof(sobj), numThreads, numBlocks);
+  getNumBlocksAndThreads(size, sizeof(sobj), numThreads, numBlocks);
-  assert(ok);
-
   Integer smemSize = numThreads * sizeof(sobj);
 
   Vector<sobj> buffer(numBlocks);
@@ -220,54 +218,6 @@ inline typename vobj::scalar_objectD sumD_gpu_small(const vobj *lat, Integer osi
   auto result = buffer_v[0];
   return result;
 }
-
-template <class vobj>
-inline typename vobj::scalar_objectD sumD_gpu_large(const vobj *lat, Integer osites)
-{
-  typedef typename vobj::vector_type  vector;
-  typedef typename vobj::scalar_typeD scalarD;
-  typedef typename vobj::scalar_objectD sobj;
-  sobj ret;
-  scalarD *ret_p = (scalarD *)&ret;
-  
-  const int words = sizeof(vobj)/sizeof(vector);
-
-  Vector<vector> buffer(osites);
-  vector *dat = (vector *)lat;
-  vector *buf = &buffer[0];
-  iScalar<vector> *tbuf =(iScalar<vector> *)  &buffer[0];
-  for(int w=0;w<words;w++) {
-
-    accelerator_for(ss,osites,1,{
-	buf[ss] = dat[ss*words+w];
-      });
-      
-    ret_p[w] = sumD_gpu_small(tbuf,osites);
-  }
-  return ret;
-}
-
-template <class vobj>
-inline typename vobj::scalar_objectD sumD_gpu(const vobj *lat, Integer osites)
-{
-  typedef typename vobj::vector_type  vector;
-  typedef typename vobj::scalar_typeD scalarD;
-  typedef typename vobj::scalar_objectD sobj;
-  sobj ret;
-  
-  Integer nsimd= vobj::Nsimd();
-  Integer size = osites*nsimd;
-  Integer numThreads, numBlocks;
-  int ok = getNumBlocksAndThreads(size, sizeof(sobj), numThreads, numBlocks);
-  
-  if ( ok ) {
-    ret = sumD_gpu_small(lat,osites);
-  } else {
-    ret = sumD_gpu_large(lat,osites);
-  }
-  return ret;
-}
-
 /////////////////////////////////////////////////////////////////////////////////////////////////////////
 // Return as same precision as input performing reduction in double precision though
 /////////////////////////////////////////////////////////////////////////////////////////////////////////
@@ -280,13 +230,6 @@ inline typename vobj::scalar_object sum_gpu(const vobj *lat, Integer osites)
   return result;
 }
 
-template <class vobj>
+
-inline typename vobj::scalar_object sum_gpu_large(const vobj *lat, Integer osites)
-{
-  typedef typename vobj::scalar_object sobj;
-  sobj result;
-  result = sumD_gpu_large(lat,osites);
-  return result;
-}
 
 NAMESPACE_END(Grid);

Grid/parallelIO/IldgIO.h

@@ -31,7 +31,6 @@ directory
 #include <fstream>
 #include <iomanip>
 #include <iostream>
-#include <string>
 #include <map>
 
 #include <pwd.h>
@@ -655,8 +654,7 @@ class IldgWriter : public ScidacWriter {
     // Fill ILDG header data struct
     //////////////////////////////////////////////////////
     ildgFormat ildgfmt ;
-    const std::string stNC = std::to_string( Nc ) ;
+    ildgfmt.field     = std::string("su3gauge");
-    ildgfmt.field          = std::string("su"+stNC+"gauge");
 
     if ( format == std::string("IEEE32BIG") ) { 
       ildgfmt.precision = 32;
@@ -873,8 +871,7 @@ class IldgReader : public GridLimeReader {
     } else { 
 
       assert(found_ildgFormat);
-      const std::string stNC = std::to_string( Nc ) ;
+      assert ( ildgFormat_.field == std::string("su3gauge") );
-      assert ( ildgFormat_.field == std::string("su"+stNC+"gauge") );
 
       ///////////////////////////////////////////////////////////////////////////////////////
       // Populate our Grid metadata as best we can
@@ -882,7 +879,7 @@ class IldgReader : public GridLimeReader {
 
       std::ostringstream vers; vers << ildgFormat_.version;
       FieldMetaData_.hdr_version = vers.str();
-      FieldMetaData_.data_type = std::string("4D_SU"+stNC+"_GAUGE_"+stNC+"x"+stNC);
+      FieldMetaData_.data_type = std::string("4D_SU3_GAUGE_3X3");
 
       FieldMetaData_.nd=4;
       FieldMetaData_.dimension.resize(4);

Grid/parallelIO/MetaData.h

@@ -6,8 +6,8 @@
 
     Copyright (C) 2015
 
+
     Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-    Author: Jamie Hudspith <renwick.james.hudspth@gmail.com>
 
     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
@@ -182,8 +182,8 @@ class GaugeStatistics
 public:
   void operator()(Lattice<vLorentzColourMatrixD> & data,FieldMetaData &header)
   {
-    header.link_trace = WilsonLoops<Impl>::linkTrace(data);
+    header.link_trace=WilsonLoops<Impl>::linkTrace(data);
-    header.plaquette  = WilsonLoops<Impl>::avgPlaquette(data);
+    header.plaquette =WilsonLoops<Impl>::avgPlaquette(data);
   }
 };
 typedef GaugeStatistics<PeriodicGimplD> PeriodicGaugeStatistics;
@@ -203,24 +203,20 @@ template<> inline void PrepareMetaData<vLorentzColourMatrixD>(Lattice<vLorentzCo
 //////////////////////////////////////////////////////////////////////
 inline void reconstruct3(LorentzColourMatrix & cm)
 {
-  assert( Nc < 4 && Nc > 1 ) ;
+  const int x=0;
+  const int y=1;
+  const int z=2;
   for(int mu=0;mu<Nd;mu++){
-    #if Nc == 2
+    cm(mu)()(2,x) = adj(cm(mu)()(0,y)*cm(mu)()(1,z)-cm(mu)()(0,z)*cm(mu)()(1,y)); //x= yz-zy
-      cm(mu)()(1,0) = -adj(cm(mu)()(0,y)) ;
+    cm(mu)()(2,y) = adj(cm(mu)()(0,z)*cm(mu)()(1,x)-cm(mu)()(0,x)*cm(mu)()(1,z)); //y= zx-xz
-      cm(mu)()(1,1) =  adj(cm(mu)()(0,x)) ;
+    cm(mu)()(2,z) = adj(cm(mu)()(0,x)*cm(mu)()(1,y)-cm(mu)()(0,y)*cm(mu)()(1,x)); //z= xy-yx
-    #else
-      const int x=0 , y=1 , z=2 ; // a little disinenuous labelling
-      cm(mu)()(2,x) = adj(cm(mu)()(0,y)*cm(mu)()(1,z)-cm(mu)()(0,z)*cm(mu)()(1,y)); //x= yz-zy
-      cm(mu)()(2,y) = adj(cm(mu)()(0,z)*cm(mu)()(1,x)-cm(mu)()(0,x)*cm(mu)()(1,z)); //y= zx-xz
-      cm(mu)()(2,z) = adj(cm(mu)()(0,x)*cm(mu)()(1,y)-cm(mu)()(0,y)*cm(mu)()(1,x)); //z= xy-yx
-    #endif
   }
 }
 
 ////////////////////////////////////////////////////////////////////////////////
 // Some data types for intermediate storage
 ////////////////////////////////////////////////////////////////////////////////
-template<typename vtype> using iLorentzColour2x3 = iVector<iVector<iVector<vtype, Nc>, Nc-1>, Nd >;
+template<typename vtype> using iLorentzColour2x3 = iVector<iVector<iVector<vtype, Nc>, 2>, Nd >;
 
 typedef iLorentzColour2x3<Complex>  LorentzColour2x3;
 typedef iLorentzColour2x3<ComplexF> LorentzColour2x3F;
@@ -282,6 +278,7 @@ struct GaugeSimpleMunger{
 
 template <class fobj, class sobj>
 struct GaugeSimpleUnmunger {
+
   void operator()(sobj &in, fobj &out) {
     for (int mu = 0; mu < Nd; mu++) {
       for (int i = 0; i < Nc; i++) {
@@ -320,8 +317,8 @@ template<class fobj,class sobj>
 struct Gauge3x2munger{
   void operator() (fobj &in,sobj &out){
     for(int mu=0;mu<Nd;mu++){
-      for(int i=0;i<Nc-1;i++){
+      for(int i=0;i<2;i++){
-	for(int j=0;j<Nc;j++){
+	for(int j=0;j<3;j++){
 	  out(mu)()(i,j) = in(mu)(i)(j);
 	}}
     }
@@ -333,8 +330,8 @@ template<class fobj,class sobj>
 struct Gauge3x2unmunger{
   void operator() (sobj &in,fobj &out){
     for(int mu=0;mu<Nd;mu++){
-      for(int i=0;i<Nc-1;i++){
+      for(int i=0;i<2;i++){
-	for(int j=0;j<Nc;j++){
+	for(int j=0;j<3;j++){
 	  out(mu)(i)(j) = in(mu)()(i,j);
 	}}
     }

Grid/parallelIO/NerscIO.h

@@ -9,7 +9,6 @@
     Author: Matt Spraggs <matthew.spraggs@gmail.com>
     Author: Peter Boyle <paboyle@ph.ed.ac.uk>
     Author: paboyle <paboyle@ph.ed.ac.uk>
-    Author: Jamie Hudspith <renwick.james.hudspth@gmail.com>
 
     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
@@ -31,8 +30,6 @@
 #ifndef GRID_NERSC_IO_H
 #define GRID_NERSC_IO_H
 
-#include <string>
-
 NAMESPACE_BEGIN(Grid);
 
 using namespace Grid;
@@ -148,17 +145,15 @@ public:
 
     std::string format(header.floating_point);
 
-    const int ieee32big = (format == std::string("IEEE32BIG"));
+    int ieee32big = (format == std::string("IEEE32BIG"));
-    const int ieee32    = (format == std::string("IEEE32"));
+    int ieee32    = (format == std::string("IEEE32"));
-    const int ieee64big = (format == std::string("IEEE64BIG"));
+    int ieee64big = (format == std::string("IEEE64BIG"));
-    const int ieee64    = (format == std::string("IEEE64") || \
+    int ieee64    = (format == std::string("IEEE64") || format == std::string("IEEE64LITTLE"));
-			   format == std::string("IEEE64LITTLE"));
 
     uint32_t nersc_csum,scidac_csuma,scidac_csumb;
     // depending on datatype, set up munger;
     // munger is a function of <floating point, Real, data_type>
-    const std::string stNC = std::to_string( Nc ) ;
+    if ( header.data_type == std::string("4D_SU3_GAUGE") ) {
-    if ( header.data_type == std::string("4D_SU"+stNC+"_GAUGE") ) {
       if ( ieee32 || ieee32big ) {
 	BinaryIO::readLatticeObject<vLorentzColourMatrixD, LorentzColour2x3F> 
 	  (Umu,file,Gauge3x2munger<LorentzColour2x3F,LorentzColourMatrix>(), offset,format,
@@ -169,7 +164,7 @@ public:
 	  (Umu,file,Gauge3x2munger<LorentzColour2x3D,LorentzColourMatrix>(),offset,format,
 	   nersc_csum,scidac_csuma,scidac_csumb);
       }
-    } else if ( header.data_type == std::string("4D_SU"+stNC+"_GAUGE_"+stNC+"x"+stNC) ) {
+    } else if ( header.data_type == std::string("4D_SU3_GAUGE_3x3") ) {
       if ( ieee32 || ieee32big ) {
 	BinaryIO::readLatticeObject<vLorentzColourMatrixD,LorentzColourMatrixF>
 	  (Umu,file,GaugeSimpleMunger<LorentzColourMatrixF,LorentzColourMatrix>(),offset,format,
@@ -214,29 +209,27 @@ public:
   template<class GaugeStats=PeriodicGaugeStatistics>
   static inline void writeConfiguration(Lattice<vLorentzColourMatrixD > &Umu,
 					std::string file, 
-					std::string ens_label = std::string("DWF"),
+					std::string ens_label = std::string("DWF"))
-					std::string ens_id = std::string("UKQCD"),
-					unsigned int sequence_number = 1)
   {
-    writeConfiguration(Umu,file,0,1,ens_label,ens_id,sequence_number);
+    writeConfiguration(Umu,file,0,1,ens_label);
   }
   template<class GaugeStats=PeriodicGaugeStatistics>
   static inline void writeConfiguration(Lattice<vLorentzColourMatrixD > &Umu,
 					std::string file, 
 					int two_row,
 					int bits32,
-					std::string ens_label = std::string("DWF"),
+					std::string ens_label = std::string("DWF"))
-					std::string ens_id = std::string("UKQCD"),
-					unsigned int sequence_number = 1)
   {
     typedef vLorentzColourMatrixD vobj;
     typedef typename vobj::scalar_object sobj;
 
     FieldMetaData header;
-    header.sequence_number = sequence_number;
+    ///////////////////////////////////////////
-    header.ensemble_id     = ens_id;
+    // Following should become arguments
+    ///////////////////////////////////////////
+    header.sequence_number = 1;
+    header.ensemble_id     = std::string("UKQCD");
     header.ensemble_label  = ens_label;
-    header.hdr_version     = "1.0" ;
 
     typedef LorentzColourMatrixD fobj3D;
     typedef LorentzColour2x3D    fobj2D;
@@ -250,14 +243,10 @@ public:
 
     uint64_t offset;
 
-    // Sod it -- always write NcxNc double
+    // Sod it -- always write 3x3 double
-    header.floating_point  = std::string("IEEE64BIG");
+    header.floating_point = std::string("IEEE64BIG");
-    const std::string stNC = std::to_string( Nc ) ;
+    header.data_type      = std::string("4D_SU3_GAUGE_3x3");
-    if( two_row ) {
+    GaugeSimpleUnmunger<fobj3D,sobj> munge;
-      header.data_type = std::string("4D_SU" + stNC + "_GAUGE" );
-    } else {
-      header.data_type = std::string("4D_SU" + stNC + "_GAUGE_" + stNC + "x" + stNC );
-    }
     if ( grid->IsBoss() ) { 
       truncate(file);
       offset = writeHeader(header,file);
@@ -265,15 +254,8 @@ public:
     grid->Broadcast(0,(void *)&offset,sizeof(offset));
 
     uint32_t nersc_csum,scidac_csuma,scidac_csumb;
-    if( two_row ) {
+    BinaryIO::writeLatticeObject<vobj,fobj3D>(Umu,file,munge,offset,header.floating_point,
-      Gauge3x2unmunger<fobj2D,sobj> munge;
+					      nersc_csum,scidac_csuma,scidac_csumb);
-      BinaryIO::writeLatticeObject<vobj,fobj2D>(Umu,file,munge,offset,header.floating_point,
-						nersc_csum,scidac_csuma,scidac_csumb);
-    } else {
-      GaugeSimpleUnmunger<fobj3D,sobj> munge;
-      BinaryIO::writeLatticeObject<vobj,fobj3D>(Umu,file,munge,offset,header.floating_point,
-						nersc_csum,scidac_csuma,scidac_csumb);
-    }
     header.checksum = nersc_csum;
     if ( grid->IsBoss() ) { 
       writeHeader(header,file);
@@ -305,7 +287,8 @@ public:
     header.plaquette=0.0;
     MachineCharacteristics(header);
 
-    uint64_t offset;
+	uint64_t offset;
+  
 #ifdef RNG_RANLUX
     header.floating_point = std::string("UINT64");
     header.data_type      = std::string("RANLUX48");
@@ -345,7 +328,7 @@ public:
 
     GridBase *grid = parallel.Grid();
 
-    uint64_t offset = readHeader(file,grid,header);
+	uint64_t offset = readHeader(file,grid,header);
 
     FieldMetaData clone(header);
 

Grid/pugixml/pugixml.cc

@@ -16,12 +16,8 @@
 
 #ifdef __NVCC__
 #pragma push
-#if (__CUDACC_VER_MAJOR__ >= 11) && (__CUDACC_VER_MINOR__ >= 5)
-#pragma nv_diag_suppress declared_but_not_referenced // suppress "function was declared but never referenced warning"
-#else
 #pragma diag_suppress declared_but_not_referenced // suppress "function was declared but never referenced warning"
 #endif
-#endif
 
 #include "pugixml.h"
 

Grid/qcd/action/fermion/CayleyFermion5D.h

@@ -68,16 +68,9 @@ public:
   ///////////////////////////////////////////////////////////////
   // Support for MADWF tricks
   ///////////////////////////////////////////////////////////////
-  RealD Mass(void) { return (mass_plus + mass_minus) / 2.0; };
+  RealD Mass(void) { return mass; };
-  RealD MassPlus(void) { return mass_plus; };
-  RealD MassMinus(void) { return mass_minus; };
   void  SetMass(RealD _mass) { 
-    mass_plus=mass_minus=_mass; 
+    mass=_mass; 
-    SetCoefficientsInternal(_zolo_hi,_gamma,_b,_c);  // Reset coeffs
-  } ;
-  void  SetMass(RealD _mass_plus, RealD _mass_minus) { 
-    mass_plus=_mass_plus;
-    mass_minus=_mass_minus;
     SetCoefficientsInternal(_zolo_hi,_gamma,_b,_c);  // Reset coeffs
   } ;
   void  P(const FermionField &psi, FermionField &chi);
@@ -115,7 +108,7 @@ public:
   void   MeooeDag5D    (const FermionField &in, FermionField &out);
 
   //    protected:
-  RealD mass_plus, mass_minus;
+  RealD mass;
 
   // Save arguments to SetCoefficientsInternal
   Vector<Coeff_t> _gamma;

Grid/qcd/action/fermion/CloverHelpers.h

@@ -1,435 +0,0 @@
-/*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid
-
-    Source file: ./lib/qcd/action/fermion/WilsonCloverFermionImplementation.h
-
-    Copyright (C) 2017 - 2022
-
-    Author: paboyle <paboyle@ph.ed.ac.uk>
-    Author: Daniel Richtmann <daniel.richtmann@gmail.com>
-    Author: Mattia Bruno <mattia.bruno@cern.ch>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-/*  END LEGAL */
-
-#pragma once
-
-#include <Grid/Grid.h>
-#include <Grid/qcd/spin/Dirac.h>
-#include <Grid/qcd/action/fermion/WilsonCloverHelpers.h>
-
-////////////////////////////////////////////
-// Standard Clover
-//   (4+m0) + csw * clover_term
-// Exp Clover
-//   (4+m0) * exp(csw/(4+m0) clover_term)
-//   = (4+m0) + csw * clover_term + ...
-////////////////////////////////////////////
-
-NAMESPACE_BEGIN(Grid);
-
-
-//////////////////////////////////
-// Generic Standard Clover
-//////////////////////////////////
-
-template<class Impl>
-class CloverHelpers: public WilsonCloverHelpers<Impl> {
-public:
-
-  INHERIT_IMPL_TYPES(Impl);
-  INHERIT_CLOVER_TYPES(Impl);
-
-  typedef WilsonCloverHelpers<Impl> Helpers;
-
-  static void Instantiate(CloverField& CloverTerm, CloverField& CloverTermInv, RealD csw_t, RealD diag_mass) {
-    GridBase *grid = CloverTerm.Grid();
-    CloverTerm += diag_mass;
-
-    int lvol = grid->lSites();
-    int DimRep = Impl::Dimension;
-    {
-      autoView(CTv,CloverTerm,CpuRead);
-      autoView(CTIv,CloverTermInv,CpuWrite);
-      thread_for(site, lvol, {
-        Coordinate lcoor;
-        grid->LocalIndexToLocalCoor(site, lcoor);
-        Eigen::MatrixXcd EigenCloverOp = Eigen::MatrixXcd::Zero(Ns * DimRep, Ns * DimRep);
-        Eigen::MatrixXcd EigenInvCloverOp = Eigen::MatrixXcd::Zero(Ns * DimRep, Ns * DimRep);
-        typename SiteClover::scalar_object Qx = Zero(), Qxinv = Zero();
-        peekLocalSite(Qx, CTv, lcoor);
-
-        for (int j = 0; j < Ns; j++)
-          for (int k = 0; k < Ns; k++)
-            for (int a = 0; a < DimRep; a++)
-              for (int b = 0; b < DimRep; b++){
-                auto zz =  Qx()(j, k)(a, b);
-                EigenCloverOp(a + j * DimRep, b + k * DimRep) = std::complex<double>(zz);
-              }
-
-        EigenInvCloverOp = EigenCloverOp.inverse();
-        for (int j = 0; j < Ns; j++)
-          for (int k = 0; k < Ns; k++)
-            for (int a = 0; a < DimRep; a++)
-              for (int b = 0; b < DimRep; b++)
-                Qxinv()(j, k)(a, b) = EigenInvCloverOp(a + j * DimRep, b + k * DimRep);
-               pokeLocalSite(Qxinv, CTIv, lcoor);
-      });
-    }
-  }
-
-  static GaugeLinkField Cmunu(std::vector<GaugeLinkField> &U, GaugeLinkField &lambda, int mu, int nu) {
-    return Helpers::Cmunu(U, lambda, mu, nu);
-  }
-
-};
-
-
-//////////////////////////////////
-// Generic Exp Clover
-//////////////////////////////////
-
-template<class Impl>
-class ExpCloverHelpers: public WilsonCloverHelpers<Impl> {
-public:
-
-  INHERIT_IMPL_TYPES(Impl);
-  INHERIT_CLOVER_TYPES(Impl);
-
-  template <typename vtype> using iImplClover = iScalar<iMatrix<iMatrix<vtype, Impl::Dimension>, Ns>>;
-  typedef WilsonCloverHelpers<Impl> Helpers;
-
-  // Can this be avoided?
-  static void IdentityTimesC(const CloverField& in, RealD c) {
-    int DimRep = Impl::Dimension;
-
-    autoView(in_v, in, AcceleratorWrite);
-
-    accelerator_for(ss, in.Grid()->oSites(), 1, {
-      for (int sa=0; sa<Ns; sa++)
-        for (int ca=0; ca<DimRep; ca++)
-          in_v[ss]()(sa,sa)(ca,ca) = c;
-    });
-  }
-
-  static int getNMAX(RealD prec, RealD R) {
-    /* compute stop condition for exponential */
-    int NMAX=1;
-    RealD cond=R*R/2.;
-
-    while (cond*std::exp(R)>prec) {
-      NMAX++;
-      cond*=R/(double)(NMAX+1);
-    }
-    return NMAX;
-  }
-
-  static int getNMAX(Lattice<iImplClover<vComplexD>> &t, RealD R) {return getNMAX(1e-12,R);}
-  static int getNMAX(Lattice<iImplClover<vComplexF>> &t, RealD R) {return getNMAX(1e-6,R);}
-
-  static void Instantiate(CloverField& Clover, CloverField& CloverInv, RealD csw_t, RealD diag_mass) {
-    GridBase* grid = Clover.Grid();
-    CloverField ExpClover(grid);
-
-    int NMAX = getNMAX(Clover, 3.*csw_t/diag_mass);
-
-    Clover *= (1.0/diag_mass);
-
-    // Taylor expansion, slow but generic
-    // Horner scheme: a0 + a1 x + a2 x^2 + .. = a0 + x (a1 + x(...))
-    // qN = cN
-    // qn = cn + qn+1 X
-    std::vector<RealD> cn(NMAX+1);
-    cn[0] = 1.0;
-    for (int i=1; i<=NMAX; i++)
-      cn[i] = cn[i-1] / RealD(i);
-
-    ExpClover = Zero();
-    IdentityTimesC(ExpClover, cn[NMAX]);
-    for (int i=NMAX-1; i>=0; i--)
-      ExpClover = ExpClover * Clover + cn[i];
-
-    // prepare inverse
-    CloverInv = (-1.0)*Clover;
-
-    Clover = ExpClover * diag_mass;
-
-    ExpClover = Zero();
-    IdentityTimesC(ExpClover, cn[NMAX]);
-    for (int i=NMAX-1; i>=0; i--)
-      ExpClover = ExpClover * CloverInv + cn[i];
-
-    CloverInv = ExpClover * (1.0/diag_mass);
-
-  }
-
-  static GaugeLinkField Cmunu(std::vector<GaugeLinkField> &U, GaugeLinkField &lambda, int mu, int nu) {
-    assert(0);
-    return lambda;
-  }
-
-};
-
-
-//////////////////////////////////
-// Compact Standard Clover
-//////////////////////////////////
-
-
-template<class Impl>
-class CompactCloverHelpers: public CompactWilsonCloverHelpers<Impl>,
-                            public WilsonCloverHelpers<Impl> {
-public:
-
-  INHERIT_IMPL_TYPES(Impl);
-  INHERIT_CLOVER_TYPES(Impl);
-  INHERIT_COMPACT_CLOVER_TYPES(Impl);
-
-  typedef WilsonCloverHelpers<Impl> Helpers;
-  typedef CompactWilsonCloverHelpers<Impl> CompactHelpers;
-
-  static void MassTerm(CloverField& Clover, RealD diag_mass) {
-    Clover += diag_mass;
-  }
-
-  static void Exponentiate_Clover(CloverDiagonalField& Diagonal,
-                          CloverTriangleField& Triangle,
-                          RealD csw_t, RealD diag_mass) {
-
-    // Do nothing
-  }
-
-  // TODO: implement Cmunu for better performances with compact layout, but don't do it
-  // here, but rather in WilsonCloverHelpers.h -> CompactWilsonCloverHelpers
-  static GaugeLinkField Cmunu(std::vector<GaugeLinkField> &U, GaugeLinkField &lambda, int mu, int nu) {
-    return Helpers::Cmunu(U, lambda, mu, nu);
-  }
-};
-
-//////////////////////////////////
-// Compact Exp Clover
-//////////////////////////////////
-
-template<class Impl>
-class CompactExpCloverHelpers: public CompactWilsonCloverHelpers<Impl> {
-public:
-
-  INHERIT_IMPL_TYPES(Impl);
-  INHERIT_CLOVER_TYPES(Impl);
-  INHERIT_COMPACT_CLOVER_TYPES(Impl);
-
-  template <typename vtype> using iImplClover = iScalar<iMatrix<iMatrix<vtype, Impl::Dimension>, Ns>>;
-  typedef CompactWilsonCloverHelpers<Impl> CompactHelpers;
-
-  static void MassTerm(CloverField& Clover, RealD diag_mass) {
-    // do nothing!
-    // mass term is multiplied to exp(Clover) below
-  }
-
-  static int getNMAX(RealD prec, RealD R) {
-    /* compute stop condition for exponential */
-    int NMAX=1;
-    RealD cond=R*R/2.;
-
-    while (cond*std::exp(R)>prec) {
-      NMAX++;
-      cond*=R/(double)(NMAX+1);
-    }
-    return NMAX;
-  }
-
-  static int getNMAX(Lattice<iImplCloverDiagonal<vComplexD>> &t, RealD R) {return getNMAX(1e-12,R);}
-  static int getNMAX(Lattice<iImplCloverDiagonal<vComplexF>> &t, RealD R) {return getNMAX(1e-6,R);}
-
-  static void ExponentiateHermitean6by6(const iMatrix<ComplexD,6> &arg, const RealD& alpha, const std::vector<RealD>& cN, const int Niter, iMatrix<ComplexD,6>& dest){
-
-  	  typedef iMatrix<ComplexD,6> mat;
-
-  	  RealD qn[6];
-  	  RealD qnold[6];
-  	  RealD p[5];
-  	  RealD trA2, trA3, trA4;
-
-  	  mat A2, A3, A4, A5;
-  	  A2 = alpha * alpha * arg * arg;
-  	  A3 = alpha * arg * A2;
-  	  A4 = A2 * A2;
-  	  A5 = A2 * A3;
-
-  	  trA2 = toReal( trace(A2) );
-  	  trA3 = toReal( trace(A3) );
-  	  trA4 = toReal( trace(A4));
-
-  	  p[0] = toReal( trace(A3 * A3)) / 6.0 - 0.125 * trA4 * trA2 - trA3 * trA3 / 18.0 + trA2 * trA2 * trA2/ 48.0;
-  	  p[1] = toReal( trace(A5)) / 5.0 - trA3 * trA2 / 6.0;
-  	  p[2] = toReal( trace(A4)) / 4.0 - 0.125 * trA2 * trA2;
-  	  p[3] = trA3 / 3.0;
-  	  p[4] = 0.5 * trA2;
-
-  	  qnold[0] = cN[Niter];
-  	  qnold[1] = 0.0;
-  	  qnold[2] = 0.0;
-  	  qnold[3] = 0.0;
-  	  qnold[4] = 0.0;
-  	  qnold[5] = 0.0;
-
-  	  for(int i = Niter-1; i >= 0; i--)
-  	  {
-  	   qn[0] = p[0] * qnold[5] + cN[i];
-  	   qn[1] = p[1] * qnold[5] + qnold[0];
-  	   qn[2] = p[2] * qnold[5] + qnold[1];
-  	   qn[3] = p[3] * qnold[5] + qnold[2];
-  	   qn[4] = p[4] * qnold[5] + qnold[3];
-  	   qn[5] = qnold[4];
-
-  	   qnold[0] = qn[0];
-  	   qnold[1] = qn[1];
-  	   qnold[2] = qn[2];
-  	   qnold[3] = qn[3];
-  	   qnold[4] = qn[4];
-  	   qnold[5] = qn[5];
-  	  }
-
-  	  mat unit(1.0);
-
-  	  dest = (qn[0] * unit + qn[1] * alpha * arg + qn[2] * A2 + qn[3] * A3 + qn[4] * A4 + qn[5] * A5);
-
-    }
-
-  static void Exponentiate_Clover(CloverDiagonalField& Diagonal, CloverTriangleField& Triangle, RealD csw_t, RealD diag_mass) {
-
-    GridBase* grid = Diagonal.Grid();
-    int NMAX = getNMAX(Diagonal, 3.*csw_t/diag_mass);
-
-    //
-    // Implementation completely in Daniel's layout
-    //
-
-    // Taylor expansion with Cayley-Hamilton recursion
-    // underlying Horner scheme as above
-    std::vector<RealD> cn(NMAX+1);
-    cn[0] = 1.0;
-    for (int i=1; i<=NMAX; i++){
-      cn[i] = cn[i-1] / RealD(i);
-    }
-
-      // Taken over from Daniel's implementation
-      conformable(Diagonal, Triangle);
-
-      long lsites = grid->lSites();
-    {
-      typedef typename SiteCloverDiagonal::scalar_object scalar_object_diagonal;
-      typedef typename SiteCloverTriangle::scalar_object scalar_object_triangle;
-      typedef iMatrix<ComplexD,6> mat;
-
-      autoView(diagonal_v,  Diagonal,  CpuRead);
-      autoView(triangle_v,  Triangle,  CpuRead);
-      autoView(diagonalExp_v, Diagonal, CpuWrite);
-      autoView(triangleExp_v, Triangle, CpuWrite);
-
-      thread_for(site, lsites, { // NOTE: Not on GPU because of (peek/poke)LocalSite
-
-    	  mat srcCloverOpUL(0.0); // upper left block
-    	  mat srcCloverOpLR(0.0); // lower right block
-    	  mat ExpCloverOp;
-
-        scalar_object_diagonal diagonal_tmp     = Zero();
-        scalar_object_diagonal diagonal_exp_tmp = Zero();
-        scalar_object_triangle triangle_tmp     = Zero();
-        scalar_object_triangle triangle_exp_tmp = Zero();
-
-        Coordinate lcoor;
-        grid->LocalIndexToLocalCoor(site, lcoor);
-
-        peekLocalSite(diagonal_tmp, diagonal_v, lcoor);
-        peekLocalSite(triangle_tmp, triangle_v, lcoor);
-
-        int block;
-        block = 0;
-        for(int i = 0; i < 6; i++){
-        	for(int j = 0; j < 6; j++){
-        		if (i == j){
-        			srcCloverOpUL(i,j) = static_cast<ComplexD>(TensorRemove(diagonal_tmp()(block)(i)));
-        		}
-        		else{
-        			srcCloverOpUL(i,j) = static_cast<ComplexD>(TensorRemove(CompactHelpers::triangle_elem(triangle_tmp, block, i, j)));
-        		}
-        	}
-        }
-        block = 1;
-        for(int i = 0; i < 6; i++){
-          	for(int j = 0; j < 6; j++){
-           		if (i == j){
-           			srcCloverOpLR(i,j) = static_cast<ComplexD>(TensorRemove(diagonal_tmp()(block)(i)));
-           		}
-           		else{
-           			srcCloverOpLR(i,j) = static_cast<ComplexD>(TensorRemove(CompactHelpers::triangle_elem(triangle_tmp, block, i, j)));
-           		}
-            }
-        }
-
-        // exp(Clover)
-
-        ExponentiateHermitean6by6(srcCloverOpUL,1.0/diag_mass,cn,NMAX,ExpCloverOp);
-
-        block = 0;
-        for(int i = 0; i < 6; i++){
-        	for(int j = 0; j < 6; j++){
-            	if (i == j){
-            		diagonal_exp_tmp()(block)(i) = ExpCloverOp(i,j);
-            	}
-            	else if(i < j){
-            		triangle_exp_tmp()(block)(CompactHelpers::triangle_index(i, j)) = ExpCloverOp(i,j);
-            	}
-           	}
-        }
-
-        ExponentiateHermitean6by6(srcCloverOpLR,1.0/diag_mass,cn,NMAX,ExpCloverOp);
-
-        block = 1;
-        for(int i = 0; i < 6; i++){
-        	for(int j = 0; j < 6; j++){
-              	if (i == j){
-              		diagonal_exp_tmp()(block)(i) = ExpCloverOp(i,j);
-               	}
-               	else if(i < j){
-               		triangle_exp_tmp()(block)(CompactHelpers::triangle_index(i, j)) = ExpCloverOp(i,j);
-               	}
-            }
-        }
-
-        pokeLocalSite(diagonal_exp_tmp, diagonalExp_v, lcoor);
-        pokeLocalSite(triangle_exp_tmp, triangleExp_v, lcoor);
-      });
-    }
-
-    Diagonal *= diag_mass;
-    Triangle *= diag_mass;
-  }
-
-
-  static GaugeLinkField Cmunu(std::vector<GaugeLinkField> &U, GaugeLinkField &lambda, int mu, int nu) {
-    assert(0);
-    return lambda;
-  }
-
-};
-
-
-NAMESPACE_END(Grid);

Grid/qcd/action/fermion/CompactWilsonCloverFermion.h

@@ -1,241 +0,0 @@
-/*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid
-
-    Source file: ./lib/qcd/action/fermion/CompactWilsonCloverFermion.h
-
-    Copyright (C) 2020 - 2022
-
-    Author: Daniel Richtmann <daniel.richtmann@gmail.com>
-    Author: Nils Meyer <nils.meyer@ur.de>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-/*  END LEGAL */
-
-#pragma once
-
-#include <Grid/qcd/action/fermion/WilsonCloverTypes.h>
-#include <Grid/qcd/action/fermion/WilsonCloverHelpers.h>
-#include <Grid/qcd/action/fermion/CloverHelpers.h>
-
-NAMESPACE_BEGIN(Grid);
-
-// see Grid/qcd/action/fermion/WilsonCloverFermion.h for description
-//
-// Modifications done here:
-//
-// Original: clover term = 12x12 matrix per site
-//
-// But: Only two diagonal 6x6 hermitian blocks are non-zero (also true for original, verified by running)
-// Sufficient to store/transfer only the real parts of the diagonal and one triangular part
-// 2 * (6 + 15 * 2) = 72 real or 36 complex words to be stored/transfered
-//
-// Here: Above but diagonal as complex numbers, i.e., need to store/transfer
-// 2 * (6 * 2 + 15 * 2) = 84 real or 42 complex words
-//
-// Words per site and improvement compared to original (combined with the input and output spinors):
-//
-// - Original: 2*12 + 12*12 = 168 words -> 1.00 x less
-// - Minimal:  2*12 + 36    =  60 words -> 2.80 x less
-// - Here:     2*12 + 42    =  66 words -> 2.55 x less
-//
-// These improvements directly translate to wall-clock time
-//
-// Data layout:
-//
-// - diagonal and triangle part as separate lattice fields,
-//   this was faster than as 1 combined field on all tested machines
-// - diagonal: as expected
-// - triangle: store upper right triangle in row major order
-// - graphical:
-//        0  1  2  3  4
-//           5  6  7  8
-//              9 10 11 = upper right triangle indices
-//                12 13
-//                   14
-//     0
-//        1
-//           2
-//              3       = diagonal indices
-//                 4
-//                    5
-//     0
-//     1  5
-//     2  6  9          = lower left triangle indices
-//     3  7 10 12
-//     4  8 11 13 14
-//
-// Impact on total memory consumption:
-// - Original: (2 * 1 + 8 * 1/2) 12x12 matrices = 6 12x12 matrices = 864 complex words per site
-// - Here:     (2 * 1 + 4 * 1/2) diagonal parts = 4 diagonal parts =  24 complex words per site
-//           + (2 * 1 + 4 * 1/2) triangle parts = 4 triangle parts =  60 complex words per site
-//                                                                 =  84 complex words per site
-
-template<class Impl, class CloverHelpers>
-class CompactWilsonCloverFermion : public WilsonFermion<Impl>,
-                                   public WilsonCloverHelpers<Impl>,
-                                   public CompactWilsonCloverHelpers<Impl> {
-  /////////////////////////////////////////////
-  // Sizes
-  /////////////////////////////////////////////
-
-public:
-
-  INHERIT_COMPACT_CLOVER_SIZES(Impl);
-
-  /////////////////////////////////////////////
-  // Type definitions
-  /////////////////////////////////////////////
-
-public:
-
-  INHERIT_IMPL_TYPES(Impl);
-  INHERIT_CLOVER_TYPES(Impl);
-  INHERIT_COMPACT_CLOVER_TYPES(Impl);
-
-  typedef WilsonFermion<Impl>              WilsonBase;
-  typedef WilsonCloverHelpers<Impl>        Helpers;
-  typedef CompactWilsonCloverHelpers<Impl> CompactHelpers;
-
-  /////////////////////////////////////////////
-  // Constructors
-  /////////////////////////////////////////////
-
-public:
-
-  CompactWilsonCloverFermion(GaugeField& _Umu,
-			    GridCartesian& Fgrid,
-			    GridRedBlackCartesian& Hgrid,
-			    const RealD _mass,
-			    const RealD _csw_r = 0.0,
-			    const RealD _csw_t = 0.0,
-			    const RealD _cF = 1.0,
-			    const WilsonAnisotropyCoefficients& clover_anisotropy = WilsonAnisotropyCoefficients(),
-			    const ImplParams& impl_p = ImplParams());
-
-  /////////////////////////////////////////////
-  // Member functions (implementing interface)
-  /////////////////////////////////////////////
-
-public:
-
-  virtual void Instantiatable() {};
-  int          ConstEE()     override { return 0; };
-  int          isTrivialEE() override { return 0; };
-
-  void Dhop(const FermionField& in, FermionField& out, int dag) override;
-
-  void DhopOE(const FermionField& in, FermionField& out, int dag) override;
-
-  void DhopEO(const FermionField& in, FermionField& out, int dag) override;
-
-  void DhopDir(const FermionField& in, FermionField& out, int dir, int disp) override;
-
-  void DhopDirAll(const FermionField& in, std::vector<FermionField>& out) /* override */;
-
-  void M(const FermionField& in, FermionField& out) override;
-
-  void Mdag(const FermionField& in, FermionField& out) override;
-
-  void Meooe(const FermionField& in, FermionField& out) override;
-
-  void MeooeDag(const FermionField& in, FermionField& out) override;
-
-  void Mooee(const FermionField& in, FermionField& out) override;
-
-  void MooeeDag(const FermionField& in, FermionField& out) override;
-
-  void MooeeInv(const FermionField& in, FermionField& out) override;
-
-  void MooeeInvDag(const FermionField& in, FermionField& out) override;
-
-  void Mdir(const FermionField& in, FermionField& out, int dir, int disp) override;
-
-  void MdirAll(const FermionField& in, std::vector<FermionField>& out) override;
-
-  void MDeriv(GaugeField& force, const FermionField& X, const FermionField& Y, int dag) override;
-
-  void MooDeriv(GaugeField& mat, const FermionField& U, const FermionField& V, int dag) override;
-
-  void MeeDeriv(GaugeField& mat, const FermionField& U, const FermionField& V, int dag) override;
-
-  /////////////////////////////////////////////
-  // Member functions (internals)
-  /////////////////////////////////////////////
-
-  void MooeeInternal(const FermionField&        in,
-                     FermionField&              out,
-                     const CloverDiagonalField& diagonal,
-                     const CloverTriangleField& triangle);
-
-  /////////////////////////////////////////////
-  // Helpers
-  /////////////////////////////////////////////
-
-  void ImportGauge(const GaugeField& _Umu) override;
-
-  /////////////////////////////////////////////
-  // Helpers
-  /////////////////////////////////////////////
-
-private:
-
-  template<class Field>
-  const MaskField* getCorrectMaskField(const Field &in) const {
-    if(in.Grid()->_isCheckerBoarded) {
-      if(in.Checkerboard() == Odd) {
-        return &this->BoundaryMaskOdd;
-      } else {
-        return &this->BoundaryMaskEven;
-      }
-    } else {
-      return &this->BoundaryMask;
-    }
-  }
-
-  template<class Field>
-  void ApplyBoundaryMask(Field& f) {
-    const MaskField* m = getCorrectMaskField(f); assert(m != nullptr);
-    assert(m != nullptr);
-    CompactHelpers::ApplyBoundaryMask(f, *m);
-  }
-
-  /////////////////////////////////////////////
-  // Member Data
-  /////////////////////////////////////////////
-
-public:
-
-  RealD csw_r;
-  RealD csw_t;
-  RealD cF;
-
-  bool open_boundaries;
-
-  CloverDiagonalField Diagonal,    DiagonalEven,    DiagonalOdd;
-  CloverDiagonalField DiagonalInv, DiagonalInvEven, DiagonalInvOdd;
-
-  CloverTriangleField Triangle,    TriangleEven,    TriangleOdd;
-  CloverTriangleField TriangleInv, TriangleInvEven, TriangleInvOdd;
-
-  FermionField Tmp;
-
-  MaskField BoundaryMask, BoundaryMaskEven, BoundaryMaskOdd;
-};
-
-NAMESPACE_END(Grid);

Grid/qcd/action/fermion/Fermion.h

@@ -53,7 +53,6 @@ NAMESPACE_CHECK(Wilson);
 #include <Grid/qcd/action/fermion/WilsonTMFermion.h>       // 4d wilson like
 NAMESPACE_CHECK(WilsonTM);
 #include <Grid/qcd/action/fermion/WilsonCloverFermion.h> // 4d wilson clover fermions
-#include <Grid/qcd/action/fermion/CompactWilsonCloverFermion.h> // 4d compact wilson clover fermions
 NAMESPACE_CHECK(WilsonClover);
 #include <Grid/qcd/action/fermion/WilsonFermion5D.h>     // 5d base used by all 5d overlap types
 NAMESPACE_CHECK(Wilson5D);
@@ -138,52 +137,21 @@ typedef WilsonTMFermion<WilsonImplF> WilsonTMFermionF;
 typedef WilsonTMFermion<WilsonImplD> WilsonTMFermionD;
 
 // Clover fermions
-template <typename WImpl> using WilsonClover = WilsonCloverFermion<WImpl, CloverHelpers<WImpl>>;
+typedef WilsonCloverFermion<WilsonImplR> WilsonCloverFermionR;
-template <typename WImpl> using WilsonExpClover = WilsonCloverFermion<WImpl, ExpCloverHelpers<WImpl>>;
+typedef WilsonCloverFermion<WilsonImplF> WilsonCloverFermionF;
+typedef WilsonCloverFermion<WilsonImplD> WilsonCloverFermionD;
 
-typedef WilsonClover<WilsonImplR> WilsonCloverFermionR;
+typedef WilsonCloverFermion<WilsonAdjImplR> WilsonCloverAdjFermionR;
-typedef WilsonClover<WilsonImplF> WilsonCloverFermionF;
+typedef WilsonCloverFermion<WilsonAdjImplF> WilsonCloverAdjFermionF;
-typedef WilsonClover<WilsonImplD> WilsonCloverFermionD;
+typedef WilsonCloverFermion<WilsonAdjImplD> WilsonCloverAdjFermionD;
 
-typedef WilsonExpClover<WilsonImplR> WilsonExpCloverFermionR;
+typedef WilsonCloverFermion<WilsonTwoIndexSymmetricImplR> WilsonCloverTwoIndexSymmetricFermionR;
-typedef WilsonExpClover<WilsonImplF> WilsonExpCloverFermionF;
+typedef WilsonCloverFermion<WilsonTwoIndexSymmetricImplF> WilsonCloverTwoIndexSymmetricFermionF;
-typedef WilsonExpClover<WilsonImplD> WilsonExpCloverFermionD;
+typedef WilsonCloverFermion<WilsonTwoIndexSymmetricImplD> WilsonCloverTwoIndexSymmetricFermionD;
 
-typedef WilsonClover<WilsonAdjImplR> WilsonCloverAdjFermionR;
+typedef WilsonCloverFermion<WilsonTwoIndexAntiSymmetricImplR> WilsonCloverTwoIndexAntiSymmetricFermionR;
-typedef WilsonClover<WilsonAdjImplF> WilsonCloverAdjFermionF;
+typedef WilsonCloverFermion<WilsonTwoIndexAntiSymmetricImplF> WilsonCloverTwoIndexAntiSymmetricFermionF;
-typedef WilsonClover<WilsonAdjImplD> WilsonCloverAdjFermionD;
+typedef WilsonCloverFermion<WilsonTwoIndexAntiSymmetricImplD> WilsonCloverTwoIndexAntiSymmetricFermionD;
-
-typedef WilsonClover<WilsonTwoIndexSymmetricImplR> WilsonCloverTwoIndexSymmetricFermionR;
-typedef WilsonClover<WilsonTwoIndexSymmetricImplF> WilsonCloverTwoIndexSymmetricFermionF;
-typedef WilsonClover<WilsonTwoIndexSymmetricImplD> WilsonCloverTwoIndexSymmetricFermionD;
-
-typedef WilsonClover<WilsonTwoIndexAntiSymmetricImplR> WilsonCloverTwoIndexAntiSymmetricFermionR;
-typedef WilsonClover<WilsonTwoIndexAntiSymmetricImplF> WilsonCloverTwoIndexAntiSymmetricFermionF;
-typedef WilsonClover<WilsonTwoIndexAntiSymmetricImplD> WilsonCloverTwoIndexAntiSymmetricFermionD;
-
-// Compact Clover fermions
-template <typename WImpl> using CompactWilsonClover = CompactWilsonCloverFermion<WImpl, CompactCloverHelpers<WImpl>>;
-template <typename WImpl> using CompactWilsonExpClover = CompactWilsonCloverFermion<WImpl, CompactExpCloverHelpers<WImpl>>;
-
-typedef CompactWilsonClover<WilsonImplR> CompactWilsonCloverFermionR;
-typedef CompactWilsonClover<WilsonImplF> CompactWilsonCloverFermionF;
-typedef CompactWilsonClover<WilsonImplD> CompactWilsonCloverFermionD;
-
-typedef CompactWilsonExpClover<WilsonImplR> CompactWilsonExpCloverFermionR;
-typedef CompactWilsonExpClover<WilsonImplF> CompactWilsonExpCloverFermionF;
-typedef CompactWilsonExpClover<WilsonImplD> CompactWilsonExpCloverFermionD;
-
-typedef CompactWilsonClover<WilsonAdjImplR> CompactWilsonCloverAdjFermionR;
-typedef CompactWilsonClover<WilsonAdjImplF> CompactWilsonCloverAdjFermionF;
-typedef CompactWilsonClover<WilsonAdjImplD> CompactWilsonCloverAdjFermionD;
-
-typedef CompactWilsonClover<WilsonTwoIndexSymmetricImplR> CompactWilsonCloverTwoIndexSymmetricFermionR;
-typedef CompactWilsonClover<WilsonTwoIndexSymmetricImplF> CompactWilsonCloverTwoIndexSymmetricFermionF;
-typedef CompactWilsonClover<WilsonTwoIndexSymmetricImplD> CompactWilsonCloverTwoIndexSymmetricFermionD;
-
-typedef CompactWilsonClover<WilsonTwoIndexAntiSymmetricImplR> CompactWilsonCloverTwoIndexAntiSymmetricFermionR;
-typedef CompactWilsonClover<WilsonTwoIndexAntiSymmetricImplF> CompactWilsonCloverTwoIndexAntiSymmetricFermionF;
-typedef CompactWilsonClover<WilsonTwoIndexAntiSymmetricImplD> CompactWilsonCloverTwoIndexAntiSymmetricFermionD;
 
 // Domain Wall fermions
 typedef DomainWallFermion<WilsonImplR> DomainWallFermionR;

Grid/qcd/action/fermion/WilsonCloverFermion.h

@@ -4,11 +4,10 @@
 
     Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.h
 
-    Copyright (C) 2017 - 2022
+    Copyright (C) 2017
 
     Author: Guido Cossu <guido.cossu@ed.ac.uk>
     Author: David Preti <>
-    Author: Daniel Richtmann <daniel.richtmann@gmail.com>
 
     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
@@ -30,9 +29,7 @@
 
 #pragma once
 
-#include <Grid/qcd/action/fermion/WilsonCloverTypes.h>
+#include <Grid/Grid.h>
-#include <Grid/qcd/action/fermion/WilsonCloverHelpers.h>
-#include <Grid/qcd/action/fermion/CloverHelpers.h>
 
 NAMESPACE_BEGIN(Grid);
 
@@ -52,16 +49,19 @@ NAMESPACE_BEGIN(Grid);
 // csw_r = csw_t to recover the isotropic version
 //////////////////////////////////////////////////////////////////
 
-template<class Impl, class CloverHelpers>
+template <class Impl>
-class WilsonCloverFermion : public WilsonFermion<Impl>,
+class WilsonCloverFermion : public WilsonFermion<Impl>
-                            public WilsonCloverHelpers<Impl>
 {
 public:
+  // Types definitions
   INHERIT_IMPL_TYPES(Impl);
-  INHERIT_CLOVER_TYPES(Impl);
+  template <typename vtype>
+  using iImplClover = iScalar<iMatrix<iMatrix<vtype, Impl::Dimension>, Ns>>;
+  typedef iImplClover<Simd> SiteCloverType;
+  typedef Lattice<SiteCloverType> CloverFieldType;
 
-  typedef WilsonFermion<Impl>       WilsonBase;
+public:
-  typedef WilsonCloverHelpers<Impl> Helpers;
+  typedef WilsonFermion<Impl> WilsonBase;
 
   virtual int    ConstEE(void)     { return 0; };
   virtual void Instantiatable(void){};
@@ -72,7 +72,42 @@ public:
                       const RealD _csw_r = 0.0,
                       const RealD _csw_t = 0.0,
                       const WilsonAnisotropyCoefficients &clover_anisotropy = WilsonAnisotropyCoefficients(),
-                      const ImplParams &impl_p = ImplParams());
+                      const ImplParams &impl_p = ImplParams()) : WilsonFermion<Impl>(_Umu,
+                                                                                     Fgrid,
+                                                                                     Hgrid,
+                                                                                     _mass, impl_p, clover_anisotropy),
+                                                                 CloverTerm(&Fgrid),
+                                                                 CloverTermInv(&Fgrid),
+                                                                 CloverTermEven(&Hgrid),
+                                                                 CloverTermOdd(&Hgrid),
+                                                                 CloverTermInvEven(&Hgrid),
+                                                                 CloverTermInvOdd(&Hgrid),
+                                                                 CloverTermDagEven(&Hgrid),
+                                                                 CloverTermDagOdd(&Hgrid),
+                                                                 CloverTermInvDagEven(&Hgrid),
+                                                                 CloverTermInvDagOdd(&Hgrid)
+  {
+    assert(Nd == 4); // require 4 dimensions
+
+    if (clover_anisotropy.isAnisotropic)
+    {
+      csw_r = _csw_r * 0.5 / clover_anisotropy.xi_0;
+      diag_mass = _mass + 1.0 + (Nd - 1) * (clover_anisotropy.nu / clover_anisotropy.xi_0);
+    }
+    else
+    {
+      csw_r = _csw_r * 0.5;
+      diag_mass = 4.0 + _mass;
+    }
+    csw_t = _csw_t * 0.5;
+
+    if (csw_r == 0)
+      std::cout << GridLogWarning << "Initializing WilsonCloverFermion with csw_r = 0" << std::endl;
+    if (csw_t == 0)
+      std::cout << GridLogWarning << "Initializing WilsonCloverFermion with csw_t = 0" << std::endl;
+
+    ImportGauge(_Umu);
+  }
 
   virtual void M(const FermionField &in, FermionField &out);
   virtual void Mdag(const FermionField &in, FermionField &out);
@@ -89,21 +124,250 @@ public:
   void ImportGauge(const GaugeField &_Umu);
 
   // Derivative parts unpreconditioned pseudofermions
-  void MDeriv(GaugeField &force, const FermionField &X, const FermionField &Y, int dag);
+  void MDeriv(GaugeField &force, const FermionField &X, const FermionField &Y, int dag)
+  {
+    conformable(X.Grid(), Y.Grid());
+    conformable(X.Grid(), force.Grid());
+    GaugeLinkField force_mu(force.Grid()), lambda(force.Grid());
+    GaugeField clover_force(force.Grid());
+    PropagatorField Lambda(force.Grid());
 
-public:
+    // Guido: Here we are hitting some performance issues:
+    // need to extract the components of the DoubledGaugeField
+    // for each call
+    // Possible solution
+    // Create a vector object to store them? (cons: wasting space)
+    std::vector<GaugeLinkField> U(Nd, this->Umu.Grid());
+
+    Impl::extractLinkField(U, this->Umu);
+
+    force = Zero();
+    // Derivative of the Wilson hopping term
+    this->DhopDeriv(force, X, Y, dag);
+
+    ///////////////////////////////////////////////////////////
+    // Clover term derivative
+    ///////////////////////////////////////////////////////////
+    Impl::outerProductImpl(Lambda, X, Y);
+    //std::cout << "Lambda:" << Lambda << std::endl;
+
+    Gamma::Algebra sigma[] = {
+        Gamma::Algebra::SigmaXY,
+        Gamma::Algebra::SigmaXZ,
+        Gamma::Algebra::SigmaXT,
+        Gamma::Algebra::MinusSigmaXY,
+        Gamma::Algebra::SigmaYZ,
+        Gamma::Algebra::SigmaYT,
+        Gamma::Algebra::MinusSigmaXZ,
+        Gamma::Algebra::MinusSigmaYZ,
+        Gamma::Algebra::SigmaZT,
+        Gamma::Algebra::MinusSigmaXT,
+        Gamma::Algebra::MinusSigmaYT,
+        Gamma::Algebra::MinusSigmaZT};
+
+    /*
+      sigma_{\mu \nu}=
+      | 0         sigma[0]  sigma[1]  sigma[2] |
+      | sigma[3]    0       sigma[4]  sigma[5] |
+      | sigma[6]  sigma[7]     0      sigma[8] |
+      | sigma[9]  sigma[10] sigma[11]   0      |
+    */
+
+    int count = 0;
+    clover_force = Zero();
+    for (int mu = 0; mu < 4; mu++)
+    {
+      force_mu = Zero();
+      for (int nu = 0; nu < 4; nu++)
+      {
+        if (mu == nu)
+        continue;
+        
+        RealD factor;
+        if (nu == 4 || mu == 4)
+        {
+          factor = 2.0 * csw_t;
+        }
+        else
+        {
+          factor = 2.0 * csw_r;
+        }
+        PropagatorField Slambda = Gamma(sigma[count]) * Lambda; // sigma checked
+        Impl::TraceSpinImpl(lambda, Slambda);                   // traceSpin ok
+        force_mu -= factor*Cmunu(U, lambda, mu, nu);                   // checked
+        count++;
+      }
+
+      pokeLorentz(clover_force, U[mu] * force_mu, mu);
+    }
+    //clover_force *= csw;
+    force += clover_force;
+  }
+
+  // Computing C_{\mu \nu}(x) as in Eq.(B.39) in Zbigniew Sroczynski's PhD thesis
+  GaugeLinkField Cmunu(std::vector<GaugeLinkField> &U, GaugeLinkField &lambda, int mu, int nu)
+  {
+    conformable(lambda.Grid(), U[0].Grid());
+    GaugeLinkField out(lambda.Grid()), tmp(lambda.Grid());
+    // insertion in upper staple
+    // please check redundancy of shift operations
+
+    // C1+
+    tmp = lambda * U[nu];
+    out = Impl::ShiftStaple(Impl::CovShiftForward(tmp, nu, Impl::CovShiftBackward(U[mu], mu, Impl::CovShiftIdentityBackward(U[nu], nu))), mu);
+
+    // C2+
+    tmp = U[mu] * Impl::ShiftStaple(adj(lambda), mu);
+    out += Impl::ShiftStaple(Impl::CovShiftForward(U[nu], nu, Impl::CovShiftBackward(tmp, mu, Impl::CovShiftIdentityBackward(U[nu], nu))), mu);
+
+    // C3+
+    tmp = U[nu] * Impl::ShiftStaple(adj(lambda), nu);
+    out += Impl::ShiftStaple(Impl::CovShiftForward(U[nu], nu, Impl::CovShiftBackward(U[mu], mu, Impl::CovShiftIdentityBackward(tmp, nu))), mu);
+
+    // C4+
+    out += Impl::ShiftStaple(Impl::CovShiftForward(U[nu], nu, Impl::CovShiftBackward(U[mu], mu, Impl::CovShiftIdentityBackward(U[nu], nu))), mu) * lambda;
+
+    // insertion in lower staple
+    // C1-
+    out -= Impl::ShiftStaple(lambda, mu) * Impl::ShiftStaple(Impl::CovShiftBackward(U[nu], nu, Impl::CovShiftBackward(U[mu], mu, U[nu])), mu);
+
+    // C2-
+    tmp = adj(lambda) * U[nu];
+    out -= Impl::ShiftStaple(Impl::CovShiftBackward(tmp, nu, Impl::CovShiftBackward(U[mu], mu, U[nu])), mu);
+
+    // C3-
+    tmp = lambda * U[nu];
+    out -= Impl::ShiftStaple(Impl::CovShiftBackward(U[nu], nu, Impl::CovShiftBackward(U[mu], mu, tmp)), mu);
+
+    // C4-
+    out -= Impl::ShiftStaple(Impl::CovShiftBackward(U[nu], nu, Impl::CovShiftBackward(U[mu], mu, U[nu])), mu) * lambda;
+
+    return out;
+  }
+
+protected:
   // here fixing the 4 dimensions, make it more general?
 
   RealD csw_r;                                               // Clover coefficient - spatial
   RealD csw_t;                                               // Clover coefficient - temporal
   RealD diag_mass;                                           // Mass term
-  CloverField CloverTerm, CloverTermInv;                     // Clover term
+  CloverFieldType CloverTerm, CloverTermInv;                 // Clover term
-  CloverField CloverTermEven, CloverTermOdd;                 // Clover term EO
+  CloverFieldType CloverTermEven, CloverTermOdd;             // Clover term EO
-  CloverField CloverTermInvEven, CloverTermInvOdd;           // Clover term Inv EO
+  CloverFieldType CloverTermInvEven, CloverTermInvOdd;       // Clover term Inv EO
-  CloverField CloverTermDagEven, CloverTermDagOdd;           // Clover term Dag EO
+  CloverFieldType CloverTermDagEven, CloverTermDagOdd;       // Clover term Dag EO
-  CloverField CloverTermInvDagEven, CloverTermInvDagOdd;     // Clover term Inv Dag EO
+  CloverFieldType CloverTermInvDagEven, CloverTermInvDagOdd; // Clover term Inv Dag EO
-};
 
+ public:
+  // eventually these can be compressed into 6x6 blocks instead of the 12x12
+  // using the DeGrand-Rossi basis for the gamma matrices
+  CloverFieldType fillCloverYZ(const GaugeLinkField &F)
+  {
+    CloverFieldType T(F.Grid());
+    T = Zero();
+    autoView(T_v,T,AcceleratorWrite);
+    autoView(F_v,F,AcceleratorRead);
+    accelerator_for(i, CloverTerm.Grid()->oSites(),1,
+    {
+      T_v[i]()(0, 1) = timesMinusI(F_v[i]()());
+      T_v[i]()(1, 0) = timesMinusI(F_v[i]()());
+      T_v[i]()(2, 3) = timesMinusI(F_v[i]()());
+      T_v[i]()(3, 2) = timesMinusI(F_v[i]()());
+    });
+
+    return T;
+  }
+
+  CloverFieldType fillCloverXZ(const GaugeLinkField &F)
+  {
+    CloverFieldType T(F.Grid());
+    T = Zero();
+    
+    autoView(T_v, T,AcceleratorWrite);
+    autoView(F_v, F,AcceleratorRead);
+    accelerator_for(i, CloverTerm.Grid()->oSites(),1,
+    {
+      T_v[i]()(0, 1) = -F_v[i]()();
+      T_v[i]()(1, 0) = F_v[i]()();
+      T_v[i]()(2, 3) = -F_v[i]()();
+      T_v[i]()(3, 2) = F_v[i]()();
+    });
+
+    return T;
+  }
+
+  CloverFieldType fillCloverXY(const GaugeLinkField &F)
+  {
+    CloverFieldType T(F.Grid());
+    T = Zero();
+
+    autoView(T_v,T,AcceleratorWrite);
+    autoView(F_v,F,AcceleratorRead);
+    accelerator_for(i, CloverTerm.Grid()->oSites(),1,
+    {
+      T_v[i]()(0, 0) = timesMinusI(F_v[i]()());
+      T_v[i]()(1, 1) = timesI(F_v[i]()());
+      T_v[i]()(2, 2) = timesMinusI(F_v[i]()());
+      T_v[i]()(3, 3) = timesI(F_v[i]()());
+    });
+
+    return T;
+  }
+
+  CloverFieldType fillCloverXT(const GaugeLinkField &F)
+  {
+    CloverFieldType T(F.Grid());
+    T = Zero();
+
+    autoView( T_v , T, AcceleratorWrite);
+    autoView( F_v , F, AcceleratorRead);
+    accelerator_for(i, CloverTerm.Grid()->oSites(),1,
+    {
+      T_v[i]()(0, 1) = timesI(F_v[i]()());
+      T_v[i]()(1, 0) = timesI(F_v[i]()());
+      T_v[i]()(2, 3) = timesMinusI(F_v[i]()());
+      T_v[i]()(3, 2) = timesMinusI(F_v[i]()());
+    });
+
+    return T;
+  }
+
+  CloverFieldType fillCloverYT(const GaugeLinkField &F)
+  {
+    CloverFieldType T(F.Grid());
+    T = Zero();
+    
+    autoView( T_v ,T,AcceleratorWrite);
+    autoView( F_v ,F,AcceleratorRead);
+    accelerator_for(i, CloverTerm.Grid()->oSites(),1,
+    {
+      T_v[i]()(0, 1) = -(F_v[i]()());
+      T_v[i]()(1, 0) = (F_v[i]()());
+      T_v[i]()(2, 3) = (F_v[i]()());
+      T_v[i]()(3, 2) = -(F_v[i]()());
+    });
+
+    return T;
+  }
+
+  CloverFieldType fillCloverZT(const GaugeLinkField &F)
+  {
+    CloverFieldType T(F.Grid());
+
+    T = Zero();
+
+    autoView( T_v , T,AcceleratorWrite);
+    autoView( F_v , F,AcceleratorRead);
+    accelerator_for(i, CloverTerm.Grid()->oSites(),1,
+    {
+      T_v[i]()(0, 0) = timesI(F_v[i]()());
+      T_v[i]()(1, 1) = timesMinusI(F_v[i]()());
+      T_v[i]()(2, 2) = timesMinusI(F_v[i]()());
+      T_v[i]()(3, 3) = timesI(F_v[i]()());
+    });
+
+    return T;
+  }
+};
 NAMESPACE_END(Grid);
 
 

Grid/qcd/action/fermion/WilsonCloverHelpers.h

@@ -1,763 +0,0 @@
-/*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid
-
-    Source file: ./lib/qcd/action/fermion/WilsonCloverHelpers.h
-
-    Copyright (C) 2021 - 2022
-
-    Author: Daniel Richtmann <daniel.richtmann@gmail.com>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-#pragma once
-
-// Helper routines that implement common clover functionality
-
-NAMESPACE_BEGIN(Grid);
-
-template<class Impl> class WilsonCloverHelpers {
-public:
-
-  INHERIT_IMPL_TYPES(Impl);
-  INHERIT_CLOVER_TYPES(Impl);
-
-  // Computing C_{\mu \nu}(x) as in Eq.(B.39) in Zbigniew Sroczynski's PhD thesis
-  static GaugeLinkField Cmunu(std::vector<GaugeLinkField> &U, GaugeLinkField &lambda, int mu, int nu)
-  {
-    conformable(lambda.Grid(), U[0].Grid());
-    GaugeLinkField out(lambda.Grid()), tmp(lambda.Grid());
-    // insertion in upper staple
-    // please check redundancy of shift operations
-
-    // C1+
-    tmp = lambda * U[nu];
-    out = Impl::ShiftStaple(Impl::CovShiftForward(tmp, nu, Impl::CovShiftBackward(U[mu], mu, Impl::CovShiftIdentityBackward(U[nu], nu))), mu);
-
-    // C2+
-    tmp = U[mu] * Impl::ShiftStaple(adj(lambda), mu);
-    out += Impl::ShiftStaple(Impl::CovShiftForward(U[nu], nu, Impl::CovShiftBackward(tmp, mu, Impl::CovShiftIdentityBackward(U[nu], nu))), mu);
-
-    // C3+
-    tmp = U[nu] * Impl::ShiftStaple(adj(lambda), nu);
-    out += Impl::ShiftStaple(Impl::CovShiftForward(U[nu], nu, Impl::CovShiftBackward(U[mu], mu, Impl::CovShiftIdentityBackward(tmp, nu))), mu);
-
-    // C4+
-    out += Impl::ShiftStaple(Impl::CovShiftForward(U[nu], nu, Impl::CovShiftBackward(U[mu], mu, Impl::CovShiftIdentityBackward(U[nu], nu))), mu) * lambda;
-
-    // insertion in lower staple
-    // C1-
-    out -= Impl::ShiftStaple(lambda, mu) * Impl::ShiftStaple(Impl::CovShiftBackward(U[nu], nu, Impl::CovShiftBackward(U[mu], mu, U[nu])), mu);
-
-    // C2-
-    tmp = adj(lambda) * U[nu];
-    out -= Impl::ShiftStaple(Impl::CovShiftBackward(tmp, nu, Impl::CovShiftBackward(U[mu], mu, U[nu])), mu);
-
-    // C3-
-    tmp = lambda * U[nu];
-    out -= Impl::ShiftStaple(Impl::CovShiftBackward(U[nu], nu, Impl::CovShiftBackward(U[mu], mu, tmp)), mu);
-
-    // C4-
-    out -= Impl::ShiftStaple(Impl::CovShiftBackward(U[nu], nu, Impl::CovShiftBackward(U[mu], mu, U[nu])), mu) * lambda;
-
-    return out;
-  }
-
-  static CloverField fillCloverYZ(const GaugeLinkField &F)
-  {
-    CloverField T(F.Grid());
-    T = Zero();
-    autoView(T_v,T,AcceleratorWrite);
-    autoView(F_v,F,AcceleratorRead);
-    accelerator_for(i, T.Grid()->oSites(),CloverField::vector_type::Nsimd(),
-    {
-      coalescedWrite(T_v[i]()(0, 1), coalescedRead(timesMinusI(F_v[i]()())));
-      coalescedWrite(T_v[i]()(1, 0), coalescedRead(timesMinusI(F_v[i]()())));
-      coalescedWrite(T_v[i]()(2, 3), coalescedRead(timesMinusI(F_v[i]()())));
-      coalescedWrite(T_v[i]()(3, 2), coalescedRead(timesMinusI(F_v[i]()())));
-    });
-
-    return T;
-  }
-
-  static CloverField fillCloverXZ(const GaugeLinkField &F)
-  {
-    CloverField T(F.Grid());
-    T = Zero();
-    
-    autoView(T_v, T,AcceleratorWrite);
-    autoView(F_v, F,AcceleratorRead);
-    accelerator_for(i, T.Grid()->oSites(),CloverField::vector_type::Nsimd(),
-    {
-      coalescedWrite(T_v[i]()(0, 1), coalescedRead(-F_v[i]()()));
-      coalescedWrite(T_v[i]()(1, 0), coalescedRead(F_v[i]()()));
-      coalescedWrite(T_v[i]()(2, 3), coalescedRead(-F_v[i]()()));
-      coalescedWrite(T_v[i]()(3, 2), coalescedRead(F_v[i]()()));
-    });
-
-    return T;
-  }
-
-  static CloverField fillCloverXY(const GaugeLinkField &F)
-  {
-    CloverField T(F.Grid());
-    T = Zero();
-
-    autoView(T_v,T,AcceleratorWrite);
-    autoView(F_v,F,AcceleratorRead);
-    accelerator_for(i, T.Grid()->oSites(),CloverField::vector_type::Nsimd(),
-    {
-      coalescedWrite(T_v[i]()(0, 0), coalescedRead(timesMinusI(F_v[i]()())));
-      coalescedWrite(T_v[i]()(1, 1), coalescedRead(timesI(F_v[i]()())));
-      coalescedWrite(T_v[i]()(2, 2), coalescedRead(timesMinusI(F_v[i]()())));
-      coalescedWrite(T_v[i]()(3, 3), coalescedRead(timesI(F_v[i]()())));
-    });
-
-    return T;
-  }
-
-  static CloverField fillCloverXT(const GaugeLinkField &F)
-  {
-    CloverField T(F.Grid());
-    T = Zero();
-
-    autoView( T_v , T, AcceleratorWrite);
-    autoView( F_v , F, AcceleratorRead);
-    accelerator_for(i, T.Grid()->oSites(),CloverField::vector_type::Nsimd(),
-    {
-      coalescedWrite(T_v[i]()(0, 1), coalescedRead(timesI(F_v[i]()())));
-      coalescedWrite(T_v[i]()(1, 0), coalescedRead(timesI(F_v[i]()())));
-      coalescedWrite(T_v[i]()(2, 3), coalescedRead(timesMinusI(F_v[i]()())));
-      coalescedWrite(T_v[i]()(3, 2), coalescedRead(timesMinusI(F_v[i]()())));
-    });
-
-    return T;
-  }
-
-  static CloverField fillCloverYT(const GaugeLinkField &F)
-  {
-    CloverField T(F.Grid());
-    T = Zero();
-    
-    autoView( T_v ,T,AcceleratorWrite);
-    autoView( F_v ,F,AcceleratorRead);
-    accelerator_for(i, T.Grid()->oSites(),CloverField::vector_type::Nsimd(),
-    {
-      coalescedWrite(T_v[i]()(0, 1), coalescedRead(-(F_v[i]()())));
-      coalescedWrite(T_v[i]()(1, 0), coalescedRead((F_v[i]()())));
-      coalescedWrite(T_v[i]()(2, 3), coalescedRead((F_v[i]()())));
-      coalescedWrite(T_v[i]()(3, 2), coalescedRead(-(F_v[i]()())));
-    });
-
-    return T;
-  }
-
-  static CloverField fillCloverZT(const GaugeLinkField &F)
-  {
-    CloverField T(F.Grid());
-
-    T = Zero();
-
-    autoView( T_v , T,AcceleratorWrite);
-    autoView( F_v , F,AcceleratorRead);
-    accelerator_for(i, T.Grid()->oSites(),CloverField::vector_type::Nsimd(),
-    {
-      coalescedWrite(T_v[i]()(0, 0), coalescedRead(timesI(F_v[i]()())));
-      coalescedWrite(T_v[i]()(1, 1), coalescedRead(timesMinusI(F_v[i]()())));
-      coalescedWrite(T_v[i]()(2, 2), coalescedRead(timesMinusI(F_v[i]()())));
-      coalescedWrite(T_v[i]()(3, 3), coalescedRead(timesI(F_v[i]()())));
-    });
-
-    return T;
-  }
-
-  template<class _Spinor>
-  static accelerator_inline void multClover(_Spinor& phi, const SiteClover& C, const _Spinor& chi) {
-    auto CC = coalescedRead(C);
-    mult(&phi, &CC, &chi);
-  }
-
-  template<class _SpinorField>
-  inline void multCloverField(_SpinorField& out, const CloverField& C, const _SpinorField& phi) {
-    const int Nsimd = SiteSpinor::Nsimd();
-    autoView(out_v, out, AcceleratorWrite);
-    autoView(phi_v, phi, AcceleratorRead);
-    autoView(C_v,   C,   AcceleratorRead);
-    typedef decltype(coalescedRead(out_v[0])) calcSpinor;
-    accelerator_for(sss,out.Grid()->oSites(),Nsimd,{
-      calcSpinor tmp;
-      multClover(tmp,C_v[sss],phi_v(sss));
-      coalescedWrite(out_v[sss],tmp);
-    });
-  }
-};
-
-
-////////////////////////////////////////////////////////
-
-template<class Impl> class CompactWilsonCloverHelpers {
-public:
-
-  INHERIT_COMPACT_CLOVER_SIZES(Impl);
-
-  INHERIT_IMPL_TYPES(Impl);
-  INHERIT_CLOVER_TYPES(Impl);
-  INHERIT_COMPACT_CLOVER_TYPES(Impl);
-
-  #if 0
-  static accelerator_inline typename SiteCloverTriangle::vector_type triangle_elem(const SiteCloverTriangle& triangle, int block, int i, int j) {
-    assert(i != j);
-    if(i < j) {
-      return triangle()(block)(triangle_index(i, j));
-    } else { // i > j
-      return conjugate(triangle()(block)(triangle_index(i, j)));
-    }
-  }
-  #else
-  template<typename vobj>
-  static accelerator_inline vobj triangle_elem(const iImplCloverTriangle<vobj>& triangle, int block, int i, int j) {
-    assert(i != j);
-    if(i < j) {
-      return triangle()(block)(triangle_index(i, j));
-    } else { // i > j
-      return conjugate(triangle()(block)(triangle_index(i, j)));
-    }
-  }
-  #endif
-
-  static accelerator_inline int triangle_index(int i, int j) {
-    if(i == j)
-      return 0;
-    else if(i < j)
-      return Nred * (Nred - 1) / 2 - (Nred - i) * (Nred - i - 1) / 2 + j - i - 1;
-    else // i > j
-      return Nred * (Nred - 1) / 2 - (Nred - j) * (Nred - j - 1) / 2 + i - j - 1;
-  }
-
-  static void MooeeKernel_gpu(int                        Nsite,
-                              int                        Ls,
-                              const FermionField&        in,
-                              FermionField&              out,
-                              const CloverDiagonalField& diagonal,
-                              const CloverTriangleField& triangle) {
-    autoView(diagonal_v, diagonal, AcceleratorRead);
-    autoView(triangle_v, triangle, AcceleratorRead);
-    autoView(in_v,       in,       AcceleratorRead);
-    autoView(out_v,      out,      AcceleratorWrite);
-
-    typedef decltype(coalescedRead(out_v[0])) CalcSpinor;
-
-    const uint64_t NN = Nsite * Ls;
-
-    accelerator_for(ss, NN, Simd::Nsimd(), {
-      int sF = ss;
-      int sU = ss/Ls;
-      CalcSpinor res;
-      CalcSpinor in_t = in_v(sF);
-      auto diagonal_t = diagonal_v(sU);
-      auto triangle_t = triangle_v(sU);
-      for(int block=0; block<Nhs; block++) {
-        int s_start = block*Nhs;
-        for(int i=0; i<Nred; i++) {
-          int si = s_start + i/Nc, ci = i%Nc;
-          res()(si)(ci) = diagonal_t()(block)(i) * in_t()(si)(ci);
-          for(int j=0; j<Nred; j++) {
-            if (j == i) continue;
-            int sj = s_start + j/Nc, cj = j%Nc;
-            res()(si)(ci) = res()(si)(ci) + triangle_elem(triangle_t, block, i, j) * in_t()(sj)(cj);
-          };
-        };
-      };
-      coalescedWrite(out_v[sF], res);
-    });
-  }
-
-  static void MooeeKernel_cpu(int                        Nsite,
-                              int                        Ls,
-                              const FermionField&        in,
-                              FermionField&              out,
-                              const CloverDiagonalField& diagonal,
-                              const CloverTriangleField& triangle) {
-    autoView(diagonal_v, diagonal, CpuRead);
-    autoView(triangle_v, triangle, CpuRead);
-    autoView(in_v,       in,       CpuRead);
-    autoView(out_v,      out,      CpuWrite);
-
-    typedef SiteSpinor CalcSpinor;
-
-#if defined(A64FX) || defined(A64FXFIXEDSIZE)
-#define PREFETCH_CLOVER(BASE) {                                     \
-    uint64_t base;                                                  \
-    int pf_dist_L1 = 1;                                             \
-    int pf_dist_L2 = -5; /* -> penalty -> disable */                \
-                                                                    \
-    if ((pf_dist_L1 >= 0) && (sU + pf_dist_L1 < Nsite)) {           \
-      base = (uint64_t)&diag_t()(pf_dist_L1+BASE)(0);               \
-      svprfd(svptrue_b64(), (int64_t*)(base +    0), SV_PLDL1STRM); \
-      svprfd(svptrue_b64(), (int64_t*)(base +  256), SV_PLDL1STRM); \
-      svprfd(svptrue_b64(), (int64_t*)(base +  512), SV_PLDL1STRM); \
-      svprfd(svptrue_b64(), (int64_t*)(base +  768), SV_PLDL1STRM); \
-      svprfd(svptrue_b64(), (int64_t*)(base + 1024), SV_PLDL1STRM); \
-      svprfd(svptrue_b64(), (int64_t*)(base + 1280), SV_PLDL1STRM); \
-    }                                                               \
-                                                                    \
-    if ((pf_dist_L2 >= 0) && (sU + pf_dist_L2 < Nsite)) {           \
-      base = (uint64_t)&diag_t()(pf_dist_L2+BASE)(0);               \
-      svprfd(svptrue_b64(), (int64_t*)(base +    0), SV_PLDL2STRM); \
-      svprfd(svptrue_b64(), (int64_t*)(base +  256), SV_PLDL2STRM); \
-      svprfd(svptrue_b64(), (int64_t*)(base +  512), SV_PLDL2STRM); \
-      svprfd(svptrue_b64(), (int64_t*)(base +  768), SV_PLDL2STRM); \
-      svprfd(svptrue_b64(), (int64_t*)(base + 1024), SV_PLDL2STRM); \
-      svprfd(svptrue_b64(), (int64_t*)(base + 1280), SV_PLDL2STRM); \
-    }                                                               \
-  }
-// TODO: Implement/generalize this for other architectures
-// I played around a bit on KNL (see below) but didn't bring anything
-// #elif defined(AVX512)
-// #define PREFETCH_CLOVER(BASE) {                              \
-//     uint64_t base;                                           \
-//     int pf_dist_L1 = 1;                                      \
-//     int pf_dist_L2 = +4;                                     \
-//                                                              \
-//     if ((pf_dist_L1 >= 0) && (sU + pf_dist_L1 < Nsite)) {    \
-//       base = (uint64_t)&diag_t()(pf_dist_L1+BASE)(0);        \
-//       _mm_prefetch((const char*)(base +    0), _MM_HINT_T0); \
-//       _mm_prefetch((const char*)(base +   64), _MM_HINT_T0); \
-//       _mm_prefetch((const char*)(base +  128), _MM_HINT_T0); \
-//       _mm_prefetch((const char*)(base +  192), _MM_HINT_T0); \
-//       _mm_prefetch((const char*)(base +  256), _MM_HINT_T0); \
-//       _mm_prefetch((const char*)(base +  320), _MM_HINT_T0); \
-//     }                                                        \
-//                                                              \
-//     if ((pf_dist_L2 >= 0) && (sU + pf_dist_L2 < Nsite)) {    \
-//       base = (uint64_t)&diag_t()(pf_dist_L2+BASE)(0);        \
-//       _mm_prefetch((const char*)(base +    0), _MM_HINT_T1); \
-//       _mm_prefetch((const char*)(base +   64), _MM_HINT_T1); \
-//       _mm_prefetch((const char*)(base +  128), _MM_HINT_T1); \
-//       _mm_prefetch((const char*)(base +  192), _MM_HINT_T1); \
-//       _mm_prefetch((const char*)(base +  256), _MM_HINT_T1); \
-//       _mm_prefetch((const char*)(base +  320), _MM_HINT_T1); \
-//     }                                                        \
-//   }
-#else
-#define PREFETCH_CLOVER(BASE)
-#endif
-
-    const uint64_t NN = Nsite * Ls;
-
-    thread_for(ss, NN, {
-      int sF = ss;
-      int sU = ss/Ls;
-      CalcSpinor res;
-      CalcSpinor in_t = in_v[sF];
-      auto diag_t     = diagonal_v[sU]; // "diag" instead of "diagonal" here to make code below easier to read
-      auto triangle_t = triangle_v[sU];
-
-      // upper half
-      PREFETCH_CLOVER(0);
-
-      auto in_cc_0_0 = conjugate(in_t()(0)(0)); // Nils: reduces number
-      auto in_cc_0_1 = conjugate(in_t()(0)(1)); // of conjugates from
-      auto in_cc_0_2 = conjugate(in_t()(0)(2)); // 30 to 20
-      auto in_cc_1_0 = conjugate(in_t()(1)(0));
-      auto in_cc_1_1 = conjugate(in_t()(1)(1));
-
-      res()(0)(0) =               diag_t()(0)( 0) * in_t()(0)(0)
-                  +           triangle_t()(0)( 0) * in_t()(0)(1)
-                  +           triangle_t()(0)( 1) * in_t()(0)(2)
-                  +           triangle_t()(0)( 2) * in_t()(1)(0)
-                  +           triangle_t()(0)( 3) * in_t()(1)(1)
-                  +           triangle_t()(0)( 4) * in_t()(1)(2);
-
-      res()(0)(1) =           triangle_t()(0)( 0) * in_cc_0_0;
-      res()(0)(1) =               diag_t()(0)( 1) * in_t()(0)(1)
-                  +           triangle_t()(0)( 5) * in_t()(0)(2)
-                  +           triangle_t()(0)( 6) * in_t()(1)(0)
-                  +           triangle_t()(0)( 7) * in_t()(1)(1)
-                  +           triangle_t()(0)( 8) * in_t()(1)(2)
-                  + conjugate(       res()(0)( 1));
-
-      res()(0)(2) =           triangle_t()(0)( 1) * in_cc_0_0
-                  +           triangle_t()(0)( 5) * in_cc_0_1;
-      res()(0)(2) =               diag_t()(0)( 2) * in_t()(0)(2)
-                  +           triangle_t()(0)( 9) * in_t()(1)(0)
-                  +           triangle_t()(0)(10) * in_t()(1)(1)
-                  +           triangle_t()(0)(11) * in_t()(1)(2)
-                  + conjugate(       res()(0)( 2));
-
-      res()(1)(0) =           triangle_t()(0)( 2) * in_cc_0_0
-                  +           triangle_t()(0)( 6) * in_cc_0_1
-                  +           triangle_t()(0)( 9) * in_cc_0_2;
-      res()(1)(0) =               diag_t()(0)( 3) * in_t()(1)(0)
-                  +           triangle_t()(0)(12) * in_t()(1)(1)
-                  +           triangle_t()(0)(13) * in_t()(1)(2)
-                  + conjugate(       res()(1)( 0));
-
-      res()(1)(1) =           triangle_t()(0)( 3) * in_cc_0_0
-                  +           triangle_t()(0)( 7) * in_cc_0_1
-                  +           triangle_t()(0)(10) * in_cc_0_2
-                  +           triangle_t()(0)(12) * in_cc_1_0;
-      res()(1)(1) =               diag_t()(0)( 4) * in_t()(1)(1)
-                  +           triangle_t()(0)(14) * in_t()(1)(2)
-                  + conjugate(       res()(1)( 1));
-
-      res()(1)(2) =           triangle_t()(0)( 4) * in_cc_0_0
-                  +           triangle_t()(0)( 8) * in_cc_0_1
-                  +           triangle_t()(0)(11) * in_cc_0_2
-                  +           triangle_t()(0)(13) * in_cc_1_0
-                  +           triangle_t()(0)(14) * in_cc_1_1;
-      res()(1)(2) =               diag_t()(0)( 5) * in_t()(1)(2)
-                  + conjugate(       res()(1)( 2));
-
-      vstream(out_v[sF]()(0)(0), res()(0)(0));
-      vstream(out_v[sF]()(0)(1), res()(0)(1));
-      vstream(out_v[sF]()(0)(2), res()(0)(2));
-      vstream(out_v[sF]()(1)(0), res()(1)(0));
-      vstream(out_v[sF]()(1)(1), res()(1)(1));
-      vstream(out_v[sF]()(1)(2), res()(1)(2));
-
-      // lower half
-      PREFETCH_CLOVER(1);
-
-      auto in_cc_2_0 = conjugate(in_t()(2)(0));
-      auto in_cc_2_1 = conjugate(in_t()(2)(1));
-      auto in_cc_2_2 = conjugate(in_t()(2)(2));
-      auto in_cc_3_0 = conjugate(in_t()(3)(0));
-      auto in_cc_3_1 = conjugate(in_t()(3)(1));
-
-      res()(2)(0) =               diag_t()(1)( 0) * in_t()(2)(0)
-                  +           triangle_t()(1)( 0) * in_t()(2)(1)
-                  +           triangle_t()(1)( 1) * in_t()(2)(2)
-                  +           triangle_t()(1)( 2) * in_t()(3)(0)
-                  +           triangle_t()(1)( 3) * in_t()(3)(1)
-                  +           triangle_t()(1)( 4) * in_t()(3)(2);
-
-      res()(2)(1) =           triangle_t()(1)( 0) * in_cc_2_0;
-      res()(2)(1) =               diag_t()(1)( 1) * in_t()(2)(1)
-                  +           triangle_t()(1)( 5) * in_t()(2)(2)
-                  +           triangle_t()(1)( 6) * in_t()(3)(0)
-                  +           triangle_t()(1)( 7) * in_t()(3)(1)
-                  +           triangle_t()(1)( 8) * in_t()(3)(2)
-                  + conjugate(       res()(2)( 1));
-
-      res()(2)(2) =           triangle_t()(1)( 1) * in_cc_2_0
-                  +           triangle_t()(1)( 5) * in_cc_2_1;
-      res()(2)(2) =               diag_t()(1)( 2) * in_t()(2)(2)
-                  +           triangle_t()(1)( 9) * in_t()(3)(0)
-                  +           triangle_t()(1)(10) * in_t()(3)(1)
-                  +           triangle_t()(1)(11) * in_t()(3)(2)
-                  + conjugate(       res()(2)( 2));
-
-      res()(3)(0) =           triangle_t()(1)( 2) * in_cc_2_0
-                  +           triangle_t()(1)( 6) * in_cc_2_1
-                  +           triangle_t()(1)( 9) * in_cc_2_2;
-      res()(3)(0) =               diag_t()(1)( 3) * in_t()(3)(0)
-                  +           triangle_t()(1)(12) * in_t()(3)(1)
-                  +           triangle_t()(1)(13) * in_t()(3)(2)
-                  + conjugate(       res()(3)( 0));
-
-      res()(3)(1) =           triangle_t()(1)( 3) * in_cc_2_0
-                  +           triangle_t()(1)( 7) * in_cc_2_1
-                  +           triangle_t()(1)(10) * in_cc_2_2
-                  +           triangle_t()(1)(12) * in_cc_3_0;
-      res()(3)(1) =               diag_t()(1)( 4) * in_t()(3)(1)
-                  +           triangle_t()(1)(14) * in_t()(3)(2)
-                  + conjugate(       res()(3)( 1));
-
-      res()(3)(2) =           triangle_t()(1)( 4) * in_cc_2_0
-                  +           triangle_t()(1)( 8) * in_cc_2_1
-                  +           triangle_t()(1)(11) * in_cc_2_2
-                  +           triangle_t()(1)(13) * in_cc_3_0
-                  +           triangle_t()(1)(14) * in_cc_3_1;
-      res()(3)(2) =               diag_t()(1)( 5) * in_t()(3)(2)
-                  + conjugate(       res()(3)( 2));
-
-      vstream(out_v[sF]()(2)(0), res()(2)(0));
-      vstream(out_v[sF]()(2)(1), res()(2)(1));
-      vstream(out_v[sF]()(2)(2), res()(2)(2));
-      vstream(out_v[sF]()(3)(0), res()(3)(0));
-      vstream(out_v[sF]()(3)(1), res()(3)(1));
-      vstream(out_v[sF]()(3)(2), res()(3)(2));
-    });
-  }
-
-  static void MooeeKernel(int                        Nsite,
-                          int                        Ls,
-                          const FermionField&        in,
-                          FermionField&              out,
-                          const CloverDiagonalField& diagonal,
-                          const CloverTriangleField& triangle) {
-#if defined(GRID_CUDA) || defined(GRID_HIP)
-    MooeeKernel_gpu(Nsite, Ls, in, out, diagonal, triangle);
-#else
-    MooeeKernel_cpu(Nsite, Ls, in, out, diagonal, triangle);
-#endif
-  }
-
-  static void Invert(const CloverDiagonalField& diagonal,
-                     const CloverTriangleField& triangle,
-                     CloverDiagonalField&       diagonalInv,
-                     CloverTriangleField&       triangleInv) {
-    conformable(diagonal, diagonalInv);
-    conformable(triangle, triangleInv);
-    conformable(diagonal, triangle);
-
-    diagonalInv.Checkerboard() = diagonal.Checkerboard();
-    triangleInv.Checkerboard() = triangle.Checkerboard();
-
-    GridBase* grid = diagonal.Grid();
-
-    long lsites = grid->lSites();
-
-    typedef typename SiteCloverDiagonal::scalar_object scalar_object_diagonal;
-    typedef typename SiteCloverTriangle::scalar_object scalar_object_triangle;
-
-    autoView(diagonal_v,  diagonal,  CpuRead);
-    autoView(triangle_v,  triangle,  CpuRead);
-    autoView(diagonalInv_v, diagonalInv, CpuWrite);
-    autoView(triangleInv_v, triangleInv, CpuWrite);
-
-    thread_for(site, lsites, { // NOTE: Not on GPU because of Eigen & (peek/poke)LocalSite
-      Eigen::MatrixXcd clover_inv_eigen = Eigen::MatrixXcd::Zero(Ns*Nc, Ns*Nc);
-      Eigen::MatrixXcd clover_eigen = Eigen::MatrixXcd::Zero(Ns*Nc, Ns*Nc);
-
-      scalar_object_diagonal diagonal_tmp     = Zero();
-      scalar_object_diagonal diagonal_inv_tmp = Zero();
-      scalar_object_triangle triangle_tmp     = Zero();
-      scalar_object_triangle triangle_inv_tmp = Zero();
-
-      Coordinate lcoor;
-      grid->LocalIndexToLocalCoor(site, lcoor);
-
-      peekLocalSite(diagonal_tmp, diagonal_v, lcoor);
-      peekLocalSite(triangle_tmp, triangle_v, lcoor);
-
-      // TODO: can we save time here by inverting the two 6x6 hermitian matrices separately?
-      for (long s_row=0;s_row<Ns;s_row++) {
-        for (long s_col=0;s_col<Ns;s_col++) {
-          if(abs(s_row - s_col) > 1 || s_row + s_col == 3) continue;
-          int block       = s_row / Nhs;
-          int s_row_block = s_row % Nhs;
-          int s_col_block = s_col % Nhs;
-          for (long c_row=0;c_row<Nc;c_row++) {
-            for (long c_col=0;c_col<Nc;c_col++) {
-              int i = s_row_block * Nc + c_row;
-              int j = s_col_block * Nc + c_col;
-              if(i == j)
-                clover_eigen(s_row*Nc+c_row, s_col*Nc+c_col) = static_cast<ComplexD>(TensorRemove(diagonal_tmp()(block)(i)));
-              else
-                clover_eigen(s_row*Nc+c_row, s_col*Nc+c_col) = static_cast<ComplexD>(TensorRemove(triangle_elem(triangle_tmp, block, i, j)));
-            }
-          }
-        }
-      }
-
-      clover_inv_eigen = clover_eigen.inverse();
-
-      for (long s_row=0;s_row<Ns;s_row++) {
-        for (long s_col=0;s_col<Ns;s_col++) {
-          if(abs(s_row - s_col) > 1 || s_row + s_col == 3) continue;
-          int block       = s_row / Nhs;
-          int s_row_block = s_row % Nhs;
-          int s_col_block = s_col % Nhs;
-          for (long c_row=0;c_row<Nc;c_row++) {
-            for (long c_col=0;c_col<Nc;c_col++) {
-              int i = s_row_block * Nc + c_row;
-              int j = s_col_block * Nc + c_col;
-              if(i == j)
-                diagonal_inv_tmp()(block)(i) = clover_inv_eigen(s_row*Nc+c_row, s_col*Nc+c_col);
-              else if(i < j)
-                triangle_inv_tmp()(block)(triangle_index(i, j)) = clover_inv_eigen(s_row*Nc+c_row, s_col*Nc+c_col);
-              else
-                continue;
-            }
-          }
-        }
-      }
-
-      pokeLocalSite(diagonal_inv_tmp, diagonalInv_v, lcoor);
-      pokeLocalSite(triangle_inv_tmp, triangleInv_v, lcoor);
-    });
-  }
-
-  static void ConvertLayout(const CloverField&   full,
-                            CloverDiagonalField& diagonal,
-                            CloverTriangleField& triangle) {
-    conformable(full, diagonal);
-    conformable(full, triangle);
-
-    diagonal.Checkerboard() = full.Checkerboard();
-    triangle.Checkerboard() = full.Checkerboard();
-
-    autoView(full_v,     full,     AcceleratorRead);
-    autoView(diagonal_v, diagonal, AcceleratorWrite);
-    autoView(triangle_v, triangle, AcceleratorWrite);
-
-    // NOTE: this function cannot be 'private' since nvcc forbids this for kernels
-    accelerator_for(ss, full.Grid()->oSites(), 1, {
-      for(int s_row = 0; s_row < Ns; s_row++) {
-        for(int s_col = 0; s_col < Ns; s_col++) {
-          if(abs(s_row - s_col) > 1 || s_row + s_col == 3) continue;
-          int block       = s_row / Nhs;
-          int s_row_block = s_row % Nhs;
-          int s_col_block = s_col % Nhs;
-          for(int c_row = 0; c_row < Nc; c_row++) {
-            for(int c_col = 0; c_col < Nc; c_col++) {
-              int i = s_row_block * Nc + c_row;
-              int j = s_col_block * Nc + c_col;
-              if(i == j)
-                diagonal_v[ss]()(block)(i) = full_v[ss]()(s_row, s_col)(c_row, c_col);
-              else if(i < j)
-                triangle_v[ss]()(block)(triangle_index(i, j)) = full_v[ss]()(s_row, s_col)(c_row, c_col);
-              else
-                continue;
-            }
-          }
-        }
-      }
-    });
-  }
-
-
-  static void ConvertLayout(const CloverDiagonalField& diagonal,
-                            const CloverTriangleField& triangle,
-                            CloverField&               full) {
-    conformable(full, diagonal);
-    conformable(full, triangle);
-
-    full.Checkerboard() = diagonal.Checkerboard();
-
-    full = Zero();
-
-    autoView(diagonal_v, diagonal, AcceleratorRead);
-    autoView(triangle_v, triangle, AcceleratorRead);
-    autoView(full_v,     full,     AcceleratorWrite);
-
-    // NOTE: this function cannot be 'private' since nvcc forbids this for kernels
-    accelerator_for(ss, full.Grid()->oSites(), 1, {
-      for(int s_row = 0; s_row < Ns; s_row++) {
-        for(int s_col = 0; s_col < Ns; s_col++) {
-          if(abs(s_row - s_col) > 1 || s_row + s_col == 3) continue;
-          int block       = s_row / Nhs;
-          int s_row_block = s_row % Nhs;
-          int s_col_block = s_col % Nhs;
-          for(int c_row = 0; c_row < Nc; c_row++) {
-            for(int c_col = 0; c_col < Nc; c_col++) {
-              int i = s_row_block * Nc + c_row;
-              int j = s_col_block * Nc + c_col;
-              if(i == j)
-                full_v[ss]()(s_row, s_col)(c_row, c_col) = diagonal_v[ss]()(block)(i);
-              else
-                full_v[ss]()(s_row, s_col)(c_row, c_col) = triangle_elem(triangle_v[ss], block, i, j);
-            }
-          }
-        }
-      }
-    });
-  }
-
-  static void ModifyBoundaries(CloverDiagonalField& diagonal, CloverTriangleField& triangle, RealD csw_t, RealD cF, RealD diag_mass) {
-    // Checks/grid
-    double t0 = usecond();
-    conformable(diagonal, triangle);
-    GridBase* grid = diagonal.Grid();
-
-    // Determine the boundary coordinates/sites
-    double t1 = usecond();
-    int t_dir = Nd - 1;
-    Lattice<iScalar<vInteger>> t_coor(grid);
-    LatticeCoordinate(t_coor, t_dir);
-    int T = grid->GlobalDimensions()[t_dir];
-
-    // Set off-diagonal parts at boundary to zero -- OK
-    double t2 = usecond();
-    CloverTriangleField zeroTriangle(grid);
-    zeroTriangle.Checkerboard() = triangle.Checkerboard();
-    zeroTriangle = Zero();
-    triangle = where(t_coor == 0,   zeroTriangle, triangle);
-    triangle = where(t_coor == T-1, zeroTriangle, triangle);
-
-    // Set diagonal to unity (scaled correctly) -- OK
-    double t3 = usecond();
-    CloverDiagonalField tmp(grid);
-    tmp.Checkerboard() = diagonal.Checkerboard();
-    tmp                = -1.0 * csw_t + diag_mass;
-    diagonal           = where(t_coor == 0,   tmp, diagonal);
-    diagonal           = where(t_coor == T-1, tmp, diagonal);
-
-    // Correct values next to boundary
-    double t4 = usecond();
-    if(cF != 1.0) {
-      tmp = cF - 1.0;
-      tmp += diagonal;
-      diagonal = where(t_coor == 1,   tmp, diagonal);
-      diagonal = where(t_coor == T-2, tmp, diagonal);
-    }
-
-    // Report timings
-    double t5 = usecond();
-#if 0
-    std::cout << GridLogMessage << "CompactWilsonCloverHelpers::ModifyBoundaries timings:"
-              << " checks = "          << (t1 - t0) / 1e6
-              << ", coordinate = "     << (t2 - t1) / 1e6
-              << ", off-diag zero = "  << (t3 - t2) / 1e6
-              << ", diagonal unity = " << (t4 - t3) / 1e6
-              << ", near-boundary = "  << (t5 - t4) / 1e6
-              << ", total = "          << (t5 - t0) / 1e6
-              << std::endl;
-#endif
-  }
-
-  template<class Field, class Mask>
-  static strong_inline void ApplyBoundaryMask(Field& f, const Mask& m) {
-    conformable(f, m);
-    auto grid  = f.Grid();
-    const uint32_t Nsite = grid->oSites();
-    const uint32_t Nsimd = grid->Nsimd();
-    autoView(f_v, f, AcceleratorWrite);
-    autoView(m_v, m, AcceleratorRead);
-    // NOTE: this function cannot be 'private' since nvcc forbids this for kernels
-    accelerator_for(ss, Nsite, Nsimd, {
-      coalescedWrite(f_v[ss], m_v(ss) * f_v(ss));
-    });
-  }
-
-  template<class MaskField>
-  static void SetupMasks(MaskField& full, MaskField& even, MaskField& odd) {
-    assert(even.Grid()->_isCheckerBoarded && even.Checkerboard() == Even);
-    assert(odd.Grid()->_isCheckerBoarded  && odd.Checkerboard()  == Odd);
-    assert(!full.Grid()->_isCheckerBoarded);
-
-    GridBase* grid = full.Grid();
-    int t_dir = Nd-1;
-    Lattice<iScalar<vInteger>> t_coor(grid);
-    LatticeCoordinate(t_coor, t_dir);
-    int T = grid->GlobalDimensions()[t_dir];
-
-    MaskField zeroMask(grid); zeroMask = Zero();
-    full = 1.0;
-    full = where(t_coor == 0,   zeroMask, full);
-    full = where(t_coor == T-1, zeroMask, full);
-
-    pickCheckerboard(Even, even, full);
-    pickCheckerboard(Odd,  odd,  full);
-  }
-};
-
-NAMESPACE_END(Grid);

Grid/qcd/action/fermion/WilsonCloverTypes.h

@@ -1,90 +0,0 @@
-/*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid
-
-    Source file: ./lib/qcd/action/fermion/WilsonCloverTypes.h
-
-    Copyright (C) 2021 - 2022
-
-    Author: Daniel Richtmann <daniel.richtmann@gmail.com>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-#pragma once
-
-NAMESPACE_BEGIN(Grid);
-
-template<class Impl>
-class WilsonCloverTypes {
-public:
-  INHERIT_IMPL_TYPES(Impl);
-
-  template <typename vtype> using iImplClover = iScalar<iMatrix<iMatrix<vtype, Impl::Dimension>, Ns>>;
-
-  typedef iImplClover<Simd> SiteClover;
-
-  typedef Lattice<SiteClover> CloverField;
-};
-
-template<class Impl>
-class CompactWilsonCloverTypes {
-public:
-  INHERIT_IMPL_TYPES(Impl);
-
-  static constexpr int Nred      = Nc * Nhs;        // 6
-  static constexpr int Nblock    = Nhs;             // 2
-  static constexpr int Ndiagonal = Nred;            // 6
-  static constexpr int Ntriangle = (Nred - 1) * Nc; // 15
-
-  template<typename vtype> using iImplCloverDiagonal = iScalar<iVector<iVector<vtype, Ndiagonal>, Nblock>>;
-  template<typename vtype> using iImplCloverTriangle = iScalar<iVector<iVector<vtype, Ntriangle>, Nblock>>;
-
-  typedef iImplCloverDiagonal<Simd> SiteCloverDiagonal;
-  typedef iImplCloverTriangle<Simd> SiteCloverTriangle;
-  typedef iSinglet<Simd>            SiteMask;
-
-  typedef Lattice<SiteCloverDiagonal> CloverDiagonalField;
-  typedef Lattice<SiteCloverTriangle> CloverTriangleField;
-  typedef Lattice<SiteMask>           MaskField;
-};
-
-#define INHERIT_CLOVER_TYPES(Impl)                                 \
-  typedef typename WilsonCloverTypes<Impl>::SiteClover SiteClover; \
-  typedef typename WilsonCloverTypes<Impl>::CloverField CloverField;
-
-#define INHERIT_COMPACT_CLOVER_TYPES(Impl) \
-  typedef typename CompactWilsonCloverTypes<Impl>::SiteCloverDiagonal  SiteCloverDiagonal; \
-  typedef typename CompactWilsonCloverTypes<Impl>::SiteCloverTriangle  SiteCloverTriangle; \
-  typedef typename CompactWilsonCloverTypes<Impl>::SiteMask            SiteMask; \
-  typedef typename CompactWilsonCloverTypes<Impl>::CloverDiagonalField CloverDiagonalField; \
-  typedef typename CompactWilsonCloverTypes<Impl>::CloverTriangleField CloverTriangleField; \
-  typedef typename CompactWilsonCloverTypes<Impl>::MaskField           MaskField; \
-  /* ugly duplication but needed inside functionality classes */ \
-  template<typename vtype> using iImplCloverDiagonal = \
-    iScalar<iVector<iVector<vtype, CompactWilsonCloverTypes<Impl>::Ndiagonal>, CompactWilsonCloverTypes<Impl>::Nblock>>; \
-  template<typename vtype> using iImplCloverTriangle = \
-    iScalar<iVector<iVector<vtype, CompactWilsonCloverTypes<Impl>::Ntriangle>, CompactWilsonCloverTypes<Impl>::Nblock>>;
-
-#define INHERIT_COMPACT_CLOVER_SIZES(Impl)                                    \
-  static constexpr int Nred      = CompactWilsonCloverTypes<Impl>::Nred;      \
-  static constexpr int Nblock    = CompactWilsonCloverTypes<Impl>::Nblock;    \
-  static constexpr int Ndiagonal = CompactWilsonCloverTypes<Impl>::Ndiagonal; \
-  static constexpr int Ntriangle = CompactWilsonCloverTypes<Impl>::Ntriangle;
-
-NAMESPACE_END(Grid);

Grid/qcd/action/fermion/implementation/CayleyFermion5DImplementation.h

@@ -47,7 +47,7 @@ CayleyFermion5D<Impl>::CayleyFermion5D(GaugeField &_Umu,
 			FiveDimRedBlackGrid,
 			FourDimGrid,
 			FourDimRedBlackGrid,_M5,p),
-  mass_plus(_mass), mass_minus(_mass)
+  mass(_mass)
 { 
 }
 
@@ -209,8 +209,8 @@ void CayleyFermion5D<Impl>::M5D   (const FermionField &psi, FermionField &chi)
 {
   int Ls=this->Ls;
   Vector<Coeff_t> diag (Ls,1.0);
-  Vector<Coeff_t> upper(Ls,-1.0); upper[Ls-1]=mass_minus;
+  Vector<Coeff_t> upper(Ls,-1.0); upper[Ls-1]=mass;
-  Vector<Coeff_t> lower(Ls,-1.0); lower[0]   =mass_plus;
+  Vector<Coeff_t> lower(Ls,-1.0); lower[0]   =mass;
   M5D(psi,chi,chi,lower,diag,upper);
 }
 template<class Impl>
@@ -220,8 +220,8 @@ void CayleyFermion5D<Impl>::Meooe5D    (const FermionField &psi, FermionField &D
   Vector<Coeff_t> diag = bs;
   Vector<Coeff_t> upper= cs;
   Vector<Coeff_t> lower= cs; 
-  upper[Ls-1]=-mass_minus*upper[Ls-1];
+  upper[Ls-1]=-mass*upper[Ls-1];
-  lower[0]   =-mass_plus*lower[0];
+  lower[0]   =-mass*lower[0];
   M5D(psi,psi,Din,lower,diag,upper);
 }
 // FIXME Redunant with the above routine; check this and eliminate
@@ -235,8 +235,8 @@ template<class Impl> void CayleyFermion5D<Impl>::Meo5D     (const FermionField &
     upper[i]=-ceo[i];
     lower[i]=-ceo[i];
   }
-  upper[Ls-1]=-mass_minus*upper[Ls-1];
+  upper[Ls-1]=-mass*upper[Ls-1];
-  lower[0]   =-mass_plus*lower[0];
+  lower[0]   =-mass*lower[0];
   M5D(psi,psi,chi,lower,diag,upper);
 }
 template<class Impl>
@@ -250,8 +250,8 @@ void CayleyFermion5D<Impl>::Mooee       (const FermionField &psi, FermionField &
     upper[i]=-cee[i];
     lower[i]=-cee[i];
   }
-  upper[Ls-1]=-mass_minus*upper[Ls-1];
+  upper[Ls-1]=-mass*upper[Ls-1];
-  lower[0]   =-mass_plus*lower[0];
+  lower[0]   =-mass*lower[0];
   M5D(psi,psi,chi,lower,diag,upper);
 }
 template<class Impl>
@@ -266,9 +266,9 @@ void CayleyFermion5D<Impl>::MooeeDag    (const FermionField &psi, FermionField &
     // Assemble the 5d matrix
     if ( s==0 ) {
       upper[s] = -cee[s+1] ;
-      lower[s] = mass_minus*cee[Ls-1];
+      lower[s] = mass*cee[Ls-1];
     } else if ( s==(Ls-1)) { 
-      upper[s] = mass_plus*cee[0];
+      upper[s] = mass*cee[0];
       lower[s] = -cee[s-1];
     } else {
       upper[s]=-cee[s+1];
@@ -291,8 +291,8 @@ void CayleyFermion5D<Impl>::M5Ddag (const FermionField &psi, FermionField &chi)
   Vector<Coeff_t> diag(Ls,1.0);
   Vector<Coeff_t> upper(Ls,-1.0);
   Vector<Coeff_t> lower(Ls,-1.0);
-  upper[Ls-1]=-mass_plus*upper[Ls-1];
+  upper[Ls-1]=-mass*upper[Ls-1];
-  lower[0]   =-mass_minus*lower[0];
+  lower[0]   =-mass*lower[0];
   M5Ddag(psi,chi,chi,lower,diag,upper);
 }
 
@@ -307,9 +307,9 @@ void CayleyFermion5D<Impl>::MeooeDag5D    (const FermionField &psi, FermionField
   for (int s=0;s<Ls;s++){
     if ( s== 0 ) {
       upper[s] = cs[s+1];
-      lower[s] =-mass_minus*cs[Ls-1];
+      lower[s] =-mass*cs[Ls-1];
     } else if ( s==(Ls-1) ) { 
-      upper[s] =-mass_plus*cs[0];
+      upper[s] =-mass*cs[0];
       lower[s] = cs[s-1];
     } else { 
       upper[s] = cs[s+1];
@@ -552,7 +552,7 @@ void CayleyFermion5D<Impl>::SetCoefficientsInternal(RealD zolo_hi,Vector<Coeff_t
       
       lee[i] =-cee[i+1]/bee[i]; // sub-diag entry on the ith column
       
-      leem[i]=mass_minus*cee[Ls-1]/bee[0];
+      leem[i]=mass*cee[Ls-1]/bee[0];
       for(int j=0;j<i;j++) {
 	assert(bee[j+1]!=Coeff_t(0.0));
 	leem[i]*= aee[j]/bee[j+1];
@@ -560,7 +560,7 @@ void CayleyFermion5D<Impl>::SetCoefficientsInternal(RealD zolo_hi,Vector<Coeff_t
       
       uee[i] =-aee[i]/bee[i];   // up-diag entry on the ith row
       
-      ueem[i]=mass_plus;
+      ueem[i]=mass;
       for(int j=1;j<=i;j++) ueem[i]*= cee[j]/bee[j];
       ueem[i]*= aee[0]/bee[0];
       
@@ -573,7 +573,7 @@ void CayleyFermion5D<Impl>::SetCoefficientsInternal(RealD zolo_hi,Vector<Coeff_t
   }
 	
   { 
-    Coeff_t delta_d=mass_minus*cee[Ls-1];
+    Coeff_t delta_d=mass*cee[Ls-1];
     for(int j=0;j<Ls-1;j++) {
       assert(bee[j] != Coeff_t(0.0));
       delta_d *= cee[j]/bee[j];
@@ -642,10 +642,6 @@ void CayleyFermion5D<Impl>::ContractConservedCurrent( PropagatorField &q_in_1,
 						      Current curr_type,
 						      unsigned int mu)
 {
-
-  assert(mass_plus == mass_minus);
-  RealD mass = mass_plus;
-  
 #if (!defined(GRID_HIP))
   Gamma::Algebra Gmu [] = {
     Gamma::Algebra::GammaX,
@@ -781,8 +777,6 @@ void CayleyFermion5D<Impl>::SeqConservedCurrent(PropagatorField &q_in,
   assert(mu>=0);
   assert(mu<Nd);
 
-  assert(mass_plus == mass_minus);
-  RealD mass = mass_plus;
 
 #if 0
   int tshift = (mu == Nd-1) ? 1 : 0;

Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h

@@ -1,373 +0,0 @@
-/*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid
-
-    Source file: ./lib/qcd/action/fermion/CompactWilsonCloverFermionImplementation.h
-
-    Copyright (C) 2017 - 2022
-
-    Author: paboyle <paboyle@ph.ed.ac.uk>
-    Author: Guido Cossu <guido.cossu@ed.ac.uk>
-    Author: Daniel Richtmann <daniel.richtmann@gmail.com>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-/*  END LEGAL */
-
-#include <Grid/Grid.h>
-#include <Grid/qcd/spin/Dirac.h>
-#include <Grid/qcd/action/fermion/CompactWilsonCloverFermion.h>
-
-
-NAMESPACE_BEGIN(Grid);
-template<class Impl, class CloverHelpers>
-CompactWilsonCloverFermion<Impl, CloverHelpers>::CompactWilsonCloverFermion(GaugeField& _Umu,
-                                                                            GridCartesian& Fgrid,
-                                                                            GridRedBlackCartesian& Hgrid,
-                                                                            const RealD _mass,
-                                                                            const RealD _csw_r,
-                                                                            const RealD _csw_t,
-                                                                            const RealD _cF,
-                                                                            const WilsonAnisotropyCoefficients& clover_anisotropy,
-                                                                            const ImplParams& impl_p)
-  : WilsonBase(_Umu, Fgrid, Hgrid, _mass, impl_p, clover_anisotropy)
-  , csw_r(_csw_r)
-  , csw_t(_csw_t)
-  , cF(_cF)
-  , open_boundaries(impl_p.boundary_phases[Nd-1] == 0.0)
-  , Diagonal(&Fgrid),        Triangle(&Fgrid)
-  , DiagonalEven(&Hgrid),    TriangleEven(&Hgrid)
-  , DiagonalOdd(&Hgrid),     TriangleOdd(&Hgrid)
-  , DiagonalInv(&Fgrid),     TriangleInv(&Fgrid)
-  , DiagonalInvEven(&Hgrid), TriangleInvEven(&Hgrid)
-  , DiagonalInvOdd(&Hgrid),  TriangleInvOdd(&Hgrid)
-  , Tmp(&Fgrid)
-  , BoundaryMask(&Fgrid)
-  , BoundaryMaskEven(&Hgrid), BoundaryMaskOdd(&Hgrid)
-{
-  assert(Nd == 4 && Nc == 3 && Ns == 4 && Impl::Dimension == 3);
-
-  csw_r *= 0.5;
-  csw_t *= 0.5;
-  if (clover_anisotropy.isAnisotropic)
-    csw_r /= clover_anisotropy.xi_0;
-
-  ImportGauge(_Umu);
-  if (open_boundaries) {
-    this->BoundaryMaskEven.Checkerboard() = Even;
-    this->BoundaryMaskOdd.Checkerboard() = Odd;
-    CompactHelpers::SetupMasks(this->BoundaryMask, this->BoundaryMaskEven, this->BoundaryMaskOdd);
-  }
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::Dhop(const FermionField& in, FermionField& out, int dag) {
-  WilsonBase::Dhop(in, out, dag);
-  if(open_boundaries) ApplyBoundaryMask(out);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::DhopOE(const FermionField& in, FermionField& out, int dag) {
-  WilsonBase::DhopOE(in, out, dag);
-  if(open_boundaries) ApplyBoundaryMask(out);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::DhopEO(const FermionField& in, FermionField& out, int dag) {
-  WilsonBase::DhopEO(in, out, dag);
-  if(open_boundaries) ApplyBoundaryMask(out);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::DhopDir(const FermionField& in, FermionField& out, int dir, int disp) {
-  WilsonBase::DhopDir(in, out, dir, disp);
-  if(this->open_boundaries) ApplyBoundaryMask(out);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::DhopDirAll(const FermionField& in, std::vector<FermionField>& out) {
-  WilsonBase::DhopDirAll(in, out);
-  if(this->open_boundaries) {
-    for(auto& o : out) ApplyBoundaryMask(o);
-  }
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::M(const FermionField& in, FermionField& out) {
-  out.Checkerboard() = in.Checkerboard();
-  WilsonBase::Dhop(in, out, DaggerNo); // call base to save applying bc
-  Mooee(in, Tmp);
-  axpy(out, 1.0, out, Tmp);
-  if(open_boundaries) ApplyBoundaryMask(out);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::Mdag(const FermionField& in, FermionField& out) {
-  out.Checkerboard() = in.Checkerboard();
-  WilsonBase::Dhop(in, out, DaggerYes);  // call base to save applying bc
-  MooeeDag(in, Tmp);
-  axpy(out, 1.0, out, Tmp);
-  if(open_boundaries) ApplyBoundaryMask(out);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::Meooe(const FermionField& in, FermionField& out) {
-  WilsonBase::Meooe(in, out);
-  if(open_boundaries) ApplyBoundaryMask(out);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::MeooeDag(const FermionField& in, FermionField& out) {
-  WilsonBase::MeooeDag(in, out);
-  if(open_boundaries) ApplyBoundaryMask(out);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::Mooee(const FermionField& in, FermionField& out) {
-  if(in.Grid()->_isCheckerBoarded) {
-    if(in.Checkerboard() == Odd) {
-      MooeeInternal(in, out, DiagonalOdd, TriangleOdd);
-    } else {
-      MooeeInternal(in, out, DiagonalEven, TriangleEven);
-    }
-  } else {
-    MooeeInternal(in, out, Diagonal, Triangle);
-  }
-  if(open_boundaries) ApplyBoundaryMask(out);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::MooeeDag(const FermionField& in, FermionField& out) {
-  Mooee(in, out); // blocks are hermitian
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::MooeeInv(const FermionField& in, FermionField& out) {
-  if(in.Grid()->_isCheckerBoarded) {
-    if(in.Checkerboard() == Odd) {
-      MooeeInternal(in, out, DiagonalInvOdd, TriangleInvOdd);
-    } else {
-      MooeeInternal(in, out, DiagonalInvEven, TriangleInvEven);
-    }
-  } else {
-    MooeeInternal(in, out, DiagonalInv, TriangleInv);
-  }
-  if(open_boundaries) ApplyBoundaryMask(out);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::MooeeInvDag(const FermionField& in, FermionField& out) {
-  MooeeInv(in, out); // blocks are hermitian
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::Mdir(const FermionField& in, FermionField& out, int dir, int disp) {
-  DhopDir(in, out, dir, disp);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::MdirAll(const FermionField& in, std::vector<FermionField>& out) {
-  DhopDirAll(in, out);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::MDeriv(GaugeField& force, const FermionField& X, const FermionField& Y, int dag) {
-  assert(!open_boundaries); // TODO check for changes required for open bc
-
-  // NOTE: code copied from original clover term
-  conformable(X.Grid(), Y.Grid());
-  conformable(X.Grid(), force.Grid());
-  GaugeLinkField force_mu(force.Grid()), lambda(force.Grid());
-  GaugeField clover_force(force.Grid());
-  PropagatorField Lambda(force.Grid());
-
-  // Guido: Here we are hitting some performance issues:
-  // need to extract the components of the DoubledGaugeField
-  // for each call
-  // Possible solution
-  // Create a vector object to store them? (cons: wasting space)
-  std::vector<GaugeLinkField> U(Nd, this->Umu.Grid());
-
-  Impl::extractLinkField(U, this->Umu);
-
-  force = Zero();
-  // Derivative of the Wilson hopping term
-  this->DhopDeriv(force, X, Y, dag);
-
-  ///////////////////////////////////////////////////////////
-  // Clover term derivative
-  ///////////////////////////////////////////////////////////
-  Impl::outerProductImpl(Lambda, X, Y);
-  //std::cout << "Lambda:" << Lambda << std::endl;
-
-  Gamma::Algebra sigma[] = {
-      Gamma::Algebra::SigmaXY,
-      Gamma::Algebra::SigmaXZ,
-      Gamma::Algebra::SigmaXT,
-      Gamma::Algebra::MinusSigmaXY,
-      Gamma::Algebra::SigmaYZ,
-      Gamma::Algebra::SigmaYT,
-      Gamma::Algebra::MinusSigmaXZ,
-      Gamma::Algebra::MinusSigmaYZ,
-      Gamma::Algebra::SigmaZT,
-      Gamma::Algebra::MinusSigmaXT,
-      Gamma::Algebra::MinusSigmaYT,
-      Gamma::Algebra::MinusSigmaZT};
-
-  /*
-    sigma_{\mu \nu}=
-    | 0         sigma[0]  sigma[1]  sigma[2] |
-    | sigma[3]    0       sigma[4]  sigma[5] |
-    | sigma[6]  sigma[7]     0      sigma[8] |
-    | sigma[9]  sigma[10] sigma[11]   0      |
-  */
-
-  int count = 0;
-  clover_force = Zero();
-  for (int mu = 0; mu < 4; mu++)
-  {
-    force_mu = Zero();
-    for (int nu = 0; nu < 4; nu++)
-    {
-      if (mu == nu)
-        continue;
-
-      RealD factor;
-      if (nu == 4 || mu == 4)
-      {
-        factor = 2.0 * csw_t;
-      }
-      else
-      {
-        factor = 2.0 * csw_r;
-      }
-      PropagatorField Slambda = Gamma(sigma[count]) * Lambda; // sigma checked
-      Impl::TraceSpinImpl(lambda, Slambda);                   // traceSpin ok
-      force_mu -= factor*CloverHelpers::Cmunu(U, lambda, mu, nu);   // checked
-      count++;
-    }
-
-    pokeLorentz(clover_force, U[mu] * force_mu, mu);
-  }
-  //clover_force *= csw;
-  force += clover_force;
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::MooDeriv(GaugeField& mat, const FermionField& U, const FermionField& V, int dag) {
-  assert(0);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::MeeDeriv(GaugeField& mat, const FermionField& U, const FermionField& V, int dag) {
-  assert(0);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::MooeeInternal(const FermionField&        in,
-                    FermionField&              out,
-                    const CloverDiagonalField& diagonal,
-                    const CloverTriangleField& triangle) {
-  assert(in.Checkerboard() == Odd || in.Checkerboard() == Even);
-  out.Checkerboard() = in.Checkerboard();
-  conformable(in, out);
-  conformable(in, diagonal);
-  conformable(in, triangle);
-
-  CompactHelpers::MooeeKernel(diagonal.oSites(), 1, in, out, diagonal, triangle);
-}
-
-template<class Impl, class CloverHelpers>
-void CompactWilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeField& _Umu) {
-  // NOTE: parts copied from original implementation
-
-  // Import gauge into base class
-  double t0 = usecond();
-  WilsonBase::ImportGauge(_Umu); // NOTE: called here and in wilson constructor -> performed twice, but can't avoid that
-
-  // Initialize temporary variables
-  double t1 = usecond();
-  conformable(_Umu.Grid(), this->GaugeGrid());
-  GridBase* grid = _Umu.Grid();
-  typename Impl::GaugeLinkField Bx(grid), By(grid), Bz(grid), Ex(grid), Ey(grid), Ez(grid);
-  CloverField TmpOriginal(grid);
-
-  // Compute the field strength terms mu>nu
-  double t2 = usecond();
-  WilsonLoops<Impl>::FieldStrength(Bx, _Umu, Zdir, Ydir);
-  WilsonLoops<Impl>::FieldStrength(By, _Umu, Zdir, Xdir);
-  WilsonLoops<Impl>::FieldStrength(Bz, _Umu, Ydir, Xdir);
-  WilsonLoops<Impl>::FieldStrength(Ex, _Umu, Tdir, Xdir);
-  WilsonLoops<Impl>::FieldStrength(Ey, _Umu, Tdir, Ydir);
-  WilsonLoops<Impl>::FieldStrength(Ez, _Umu, Tdir, Zdir);
-
-  // Compute the Clover Operator acting on Colour and Spin
-  // multiply here by the clover coefficients for the anisotropy
-  double t3 = usecond();
-  TmpOriginal  = Helpers::fillCloverYZ(Bx) * csw_r;
-  TmpOriginal += Helpers::fillCloverXZ(By) * csw_r;
-  TmpOriginal += Helpers::fillCloverXY(Bz) * csw_r;
-  TmpOriginal += Helpers::fillCloverXT(Ex) * csw_t;
-  TmpOriginal += Helpers::fillCloverYT(Ey) * csw_t;
-  TmpOriginal += Helpers::fillCloverZT(Ez) * csw_t;
-  // Handle mass term based on clover policy
-  CloverHelpers::MassTerm(TmpOriginal, this->diag_mass);
-  
-  // Convert the data layout of the clover term
-  double t4 = usecond();
-  CompactHelpers::ConvertLayout(TmpOriginal, Diagonal, Triangle);
-
-  // Exponentiate the clover (nothing happens in case of the standard clover)
-  double t5 = usecond();
-  CloverHelpers::Exponentiate_Clover(Diagonal, Triangle, csw_t, this->diag_mass);
-
-  // Possible modify the boundary values
-  double t6 = usecond();
-  if(open_boundaries) CompactHelpers::ModifyBoundaries(Diagonal, Triangle, csw_t, cF, this->diag_mass);
-
-  // Invert the Clover term (explicit inversion needed for the improvement in case of open boundary conditions)
-  double t7 = usecond();
-  CompactHelpers::Invert(Diagonal, Triangle, DiagonalInv, TriangleInv);
-
-  // Fill the remaining clover fields
-  double t8 = usecond();
-  pickCheckerboard(Even, DiagonalEven,    Diagonal);
-  pickCheckerboard(Even, TriangleEven,    Triangle);
-  pickCheckerboard(Odd,  DiagonalOdd,     Diagonal);
-  pickCheckerboard(Odd,  TriangleOdd,     Triangle);
-  pickCheckerboard(Even, DiagonalInvEven, DiagonalInv);
-  pickCheckerboard(Even, TriangleInvEven, TriangleInv);
-  pickCheckerboard(Odd,  DiagonalInvOdd,  DiagonalInv);
-  pickCheckerboard(Odd,  TriangleInvOdd,  TriangleInv);
-
-  // Report timings
-  double t9 = usecond();
-
-  std::cout << GridLogDebug << "CompactWilsonCloverFermion::ImportGauge timings:" << std::endl;
-  std::cout << GridLogDebug << "WilsonFermion::Importgauge = " << (t1 - t0) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "allocations =                " << (t2 - t1) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "field strength =             " << (t3 - t2) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "fill clover =                " << (t4 - t3) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "convert =                    " << (t5 - t4) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "exponentiation =             " << (t6 - t5) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "boundaries =                 " << (t7 - t6) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "inversions =                 " << (t8 - t7) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "pick cbs =                   " << (t9 - t8) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "total =                      " << (t9 - t0) / 1e6 << std::endl;
-}
-
-NAMESPACE_END(Grid);

Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h

@@ -2,13 +2,12 @@
 
     Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./lib/qcd/action/fermion/WilsonCloverFermionImplementation.h
+    Source file: ./lib/qcd/action/fermion/WilsonCloverFermion.cc
 
-    Copyright (C) 2017 - 2022
+    Copyright (C) 2017
 
     Author: paboyle <paboyle@ph.ed.ac.uk>
     Author: Guido Cossu <guido.cossu@ed.ac.uk>
-    Author: Daniel Richtmann <daniel.richtmann@gmail.com>
 
     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
@@ -34,48 +33,9 @@
 
 NAMESPACE_BEGIN(Grid);
 
-template<class Impl, class CloverHelpers>
-WilsonCloverFermion<Impl, CloverHelpers>::WilsonCloverFermion(GaugeField&                         _Umu,
-                                               GridCartesian&                      Fgrid,
-                                               GridRedBlackCartesian&              Hgrid,
-                                               const RealD                         _mass,
-                                               const RealD                         _csw_r,
-                                               const RealD                         _csw_t,
-                                               const WilsonAnisotropyCoefficients& clover_anisotropy,
-                                               const ImplParams&                   impl_p)
-  : WilsonFermion<Impl>(_Umu, Fgrid, Hgrid, _mass, impl_p, clover_anisotropy)
-  , CloverTerm(&Fgrid)
-  , CloverTermInv(&Fgrid)
-  , CloverTermEven(&Hgrid)
-  , CloverTermOdd(&Hgrid)
-  , CloverTermInvEven(&Hgrid)
-  , CloverTermInvOdd(&Hgrid)
-  , CloverTermDagEven(&Hgrid)
-  , CloverTermDagOdd(&Hgrid)
-  , CloverTermInvDagEven(&Hgrid)
-  , CloverTermInvDagOdd(&Hgrid) {
-  assert(Nd == 4); // require 4 dimensions
-
-  if(clover_anisotropy.isAnisotropic) {
-    csw_r     = _csw_r * 0.5 / clover_anisotropy.xi_0;
-    diag_mass = _mass + 1.0 + (Nd - 1) * (clover_anisotropy.nu / clover_anisotropy.xi_0);
-  } else {
-    csw_r     = _csw_r * 0.5;
-    diag_mass = 4.0 + _mass;
-  }
-  csw_t = _csw_t * 0.5;
-
-  if(csw_r == 0)
-    std::cout << GridLogWarning << "Initializing WilsonCloverFermion with csw_r = 0" << std::endl;
-  if(csw_t == 0)
-    std::cout << GridLogWarning << "Initializing WilsonCloverFermion with csw_t = 0" << std::endl;
-
-  ImportGauge(_Umu);
-}
-
 // *NOT* EO
-template<class Impl, class CloverHelpers>
+template <class Impl>
-void WilsonCloverFermion<Impl, CloverHelpers>::M(const FermionField &in, FermionField &out)
+void WilsonCloverFermion<Impl>::M(const FermionField &in, FermionField &out)
 {
   FermionField temp(out.Grid());
 
@@ -89,8 +49,8 @@ void WilsonCloverFermion<Impl, CloverHelpers>::M(const FermionField &in, Fermion
   out += temp;
 }
 
-template<class Impl, class CloverHelpers>
+template <class Impl>
-void WilsonCloverFermion<Impl, CloverHelpers>::Mdag(const FermionField &in, FermionField &out)
+void WilsonCloverFermion<Impl>::Mdag(const FermionField &in, FermionField &out)
 {
   FermionField temp(out.Grid());
 
@@ -104,16 +64,13 @@ void WilsonCloverFermion<Impl, CloverHelpers>::Mdag(const FermionField &in, Ferm
   out += temp;
 }
 
-template<class Impl, class CloverHelpers>
+template <class Impl>
-void WilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeField &_Umu)
+void WilsonCloverFermion<Impl>::ImportGauge(const GaugeField &_Umu)
 {
-  double t0 = usecond();
   WilsonFermion<Impl>::ImportGauge(_Umu);
-  double t1 = usecond();
   GridBase *grid = _Umu.Grid();
   typename Impl::GaugeLinkField Bx(grid), By(grid), Bz(grid), Ex(grid), Ey(grid), Ez(grid);
 
-  double t2 = usecond();
   // Compute the field strength terms mu>nu
   WilsonLoops<Impl>::FieldStrength(Bx, _Umu, Zdir, Ydir);
   WilsonLoops<Impl>::FieldStrength(By, _Umu, Zdir, Xdir);
@@ -122,20 +79,52 @@ void WilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeField &_Um
   WilsonLoops<Impl>::FieldStrength(Ey, _Umu, Tdir, Ydir);
   WilsonLoops<Impl>::FieldStrength(Ez, _Umu, Tdir, Zdir);
 
-  double t3 = usecond();
   // Compute the Clover Operator acting on Colour and Spin
   // multiply here by the clover coefficients for the anisotropy
-  CloverTerm  = Helpers::fillCloverYZ(Bx) * csw_r;
+  CloverTerm  = fillCloverYZ(Bx) * csw_r;
-  CloverTerm += Helpers::fillCloverXZ(By) * csw_r;
+  CloverTerm += fillCloverXZ(By) * csw_r;
-  CloverTerm += Helpers::fillCloverXY(Bz) * csw_r;
+  CloverTerm += fillCloverXY(Bz) * csw_r;
-  CloverTerm += Helpers::fillCloverXT(Ex) * csw_t;
+  CloverTerm += fillCloverXT(Ex) * csw_t;
-  CloverTerm += Helpers::fillCloverYT(Ey) * csw_t;
+  CloverTerm += fillCloverYT(Ey) * csw_t;
-  CloverTerm += Helpers::fillCloverZT(Ez) * csw_t;
+  CloverTerm += fillCloverZT(Ez) * csw_t;
-   
+  CloverTerm += diag_mass;
-  double t4 = usecond();
+
-  CloverHelpers::Instantiate(CloverTerm, CloverTermInv, csw_t, this->diag_mass);
+  int lvol = _Umu.Grid()->lSites();
+  int DimRep = Impl::Dimension;
+
+  {
+    autoView(CTv,CloverTerm,CpuRead);
+    autoView(CTIv,CloverTermInv,CpuWrite);
+    thread_for(site, lvol, {
+      Coordinate lcoor;
+      grid->LocalIndexToLocalCoor(site, lcoor);
+      Eigen::MatrixXcd EigenCloverOp = Eigen::MatrixXcd::Zero(Ns * DimRep, Ns * DimRep);
+      Eigen::MatrixXcd EigenInvCloverOp = Eigen::MatrixXcd::Zero(Ns * DimRep, Ns * DimRep);
+      typename SiteCloverType::scalar_object Qx = Zero(), Qxinv = Zero();
+      peekLocalSite(Qx, CTv, lcoor);
+      //if (csw!=0){
+      for (int j = 0; j < Ns; j++)
+	for (int k = 0; k < Ns; k++)
+	  for (int a = 0; a < DimRep; a++)
+	    for (int b = 0; b < DimRep; b++){
+	      auto zz =  Qx()(j, k)(a, b);
+	      EigenCloverOp(a + j * DimRep, b + k * DimRep) = std::complex<double>(zz);
+	    }
+      //   if (site==0) std::cout << "site =" << site << "\n" << EigenCloverOp << std::endl;
+      
+      EigenInvCloverOp = EigenCloverOp.inverse();
+      //std::cout << EigenInvCloverOp << std::endl;
+      for (int j = 0; j < Ns; j++)
+	for (int k = 0; k < Ns; k++)
+	  for (int a = 0; a < DimRep; a++)
+	    for (int b = 0; b < DimRep; b++)
+	      Qxinv()(j, k)(a, b) = EigenInvCloverOp(a + j * DimRep, b + k * DimRep);
+      //    if (site==0) std::cout << "site =" << site << "\n" << EigenInvCloverOp << std::endl;
+      //  }
+      pokeLocalSite(Qxinv, CTIv, lcoor);
+    });
+  }
 
-  double t5 = usecond();
   // Separate the even and odd parts
   pickCheckerboard(Even, CloverTermEven, CloverTerm);
   pickCheckerboard(Odd, CloverTermOdd, CloverTerm);
@@ -148,47 +137,37 @@ void WilsonCloverFermion<Impl, CloverHelpers>::ImportGauge(const GaugeField &_Um
 
   pickCheckerboard(Even, CloverTermInvDagEven, adj(CloverTermInv));
   pickCheckerboard(Odd, CloverTermInvDagOdd, adj(CloverTermInv));
-  double t6 = usecond();
-
-  std::cout << GridLogDebug << "WilsonCloverFermion::ImportGauge timings:" << std::endl;
-  std::cout << GridLogDebug << "WilsonFermion::Importgauge = " << (t1 - t0) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "allocations =                " << (t2 - t1) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "field strength =             " << (t3 - t2) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "fill clover =                " << (t4 - t3) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "instantiation =              " << (t5 - t4) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "pick cbs =                   " << (t6 - t5) / 1e6 << std::endl;
-  std::cout << GridLogDebug << "total =                      " << (t6 - t0) / 1e6 << std::endl;
 }
 
-template<class Impl, class CloverHelpers>
+template <class Impl>
-void WilsonCloverFermion<Impl, CloverHelpers>::Mooee(const FermionField &in, FermionField &out)
+void WilsonCloverFermion<Impl>::Mooee(const FermionField &in, FermionField &out)
 {
   this->MooeeInternal(in, out, DaggerNo, InverseNo);
 }
 
-template<class Impl, class CloverHelpers>
+template <class Impl>
-void WilsonCloverFermion<Impl, CloverHelpers>::MooeeDag(const FermionField &in, FermionField &out)
+void WilsonCloverFermion<Impl>::MooeeDag(const FermionField &in, FermionField &out)
 {
   this->MooeeInternal(in, out, DaggerYes, InverseNo);
 }
 
-template<class Impl, class CloverHelpers>
+template <class Impl>
-void WilsonCloverFermion<Impl, CloverHelpers>::MooeeInv(const FermionField &in, FermionField &out)
+void WilsonCloverFermion<Impl>::MooeeInv(const FermionField &in, FermionField &out)
 {
   this->MooeeInternal(in, out, DaggerNo, InverseYes);
 }
 
-template<class Impl, class CloverHelpers>
+template <class Impl>
-void WilsonCloverFermion<Impl, CloverHelpers>::MooeeInvDag(const FermionField &in, FermionField &out)
+void WilsonCloverFermion<Impl>::MooeeInvDag(const FermionField &in, FermionField &out)
 {
   this->MooeeInternal(in, out, DaggerYes, InverseYes);
 }
 
-template<class Impl, class CloverHelpers>
+template <class Impl>
-void WilsonCloverFermion<Impl, CloverHelpers>::MooeeInternal(const FermionField &in, FermionField &out, int dag, int inv)
+void WilsonCloverFermion<Impl>::MooeeInternal(const FermionField &in, FermionField &out, int dag, int inv)
 {
   out.Checkerboard() = in.Checkerboard();
-  CloverField *Clover;
+  CloverFieldType *Clover;
   assert(in.Checkerboard() == Odd || in.Checkerboard() == Even);
 
   if (dag)
@@ -203,12 +182,12 @@ void WilsonCloverFermion<Impl, CloverHelpers>::MooeeInternal(const FermionField
       {
         Clover = (inv) ? &CloverTermInvDagEven : &CloverTermDagEven;
       }
-      Helpers::multCloverField(out, *Clover, in);
+      out = *Clover * in;
     }
     else
     {
       Clover = (inv) ? &CloverTermInv : &CloverTerm;
-      Helpers::multCloverField(out, *Clover, in); // don't bother with adj, hermitian anyway
+      out = adj(*Clover) * in;
     }
   }
   else
@@ -226,109 +205,29 @@ void WilsonCloverFermion<Impl, CloverHelpers>::MooeeInternal(const FermionField
         //  std::cout << "Calling clover term Even" << std::endl;
         Clover = (inv) ? &CloverTermInvEven : &CloverTermEven;
       }
-      Helpers::multCloverField(out, *Clover, in);
+      out = *Clover * in;
       //  std::cout << GridLogMessage << "*Clover.Checkerboard() "  << (*Clover).Checkerboard() << std::endl;
     }
     else
     {
       Clover = (inv) ? &CloverTermInv : &CloverTerm;
-      Helpers::multCloverField(out, *Clover, in);
+      out = *Clover * in;
     }
   }
+
 } // MooeeInternal
 
-// Derivative parts unpreconditioned pseudofermions
-template<class Impl, class CloverHelpers>
-void WilsonCloverFermion<Impl, CloverHelpers>::MDeriv(GaugeField &force, const FermionField &X, const FermionField &Y, int dag)
-{
-  conformable(X.Grid(), Y.Grid());
-  conformable(X.Grid(), force.Grid());
-  GaugeLinkField force_mu(force.Grid()), lambda(force.Grid());
-  GaugeField clover_force(force.Grid());
-  PropagatorField Lambda(force.Grid());
-
-  // Guido: Here we are hitting some performance issues:
-  // need to extract the components of the DoubledGaugeField
-  // for each call
-  // Possible solution
-  // Create a vector object to store them? (cons: wasting space)
-  std::vector<GaugeLinkField> U(Nd, this->Umu.Grid());
-
-  Impl::extractLinkField(U, this->Umu);
-
-  force = Zero();
-  // Derivative of the Wilson hopping term
-  this->DhopDeriv(force, X, Y, dag);
-
-  ///////////////////////////////////////////////////////////
-  // Clover term derivative
-  ///////////////////////////////////////////////////////////
-  Impl::outerProductImpl(Lambda, X, Y);
-  //std::cout << "Lambda:" << Lambda << std::endl;
-
-  Gamma::Algebra sigma[] = {
-      Gamma::Algebra::SigmaXY,
-      Gamma::Algebra::SigmaXZ,
-      Gamma::Algebra::SigmaXT,
-      Gamma::Algebra::MinusSigmaXY,
-      Gamma::Algebra::SigmaYZ,
-      Gamma::Algebra::SigmaYT,
-      Gamma::Algebra::MinusSigmaXZ,
-      Gamma::Algebra::MinusSigmaYZ,
-      Gamma::Algebra::SigmaZT,
-      Gamma::Algebra::MinusSigmaXT,
-      Gamma::Algebra::MinusSigmaYT,
-      Gamma::Algebra::MinusSigmaZT};
-
-  /*
-    sigma_{\mu \nu}=
-    | 0         sigma[0]  sigma[1]  sigma[2] |
-    | sigma[3]    0       sigma[4]  sigma[5] |
-    | sigma[6]  sigma[7]     0      sigma[8] |
-    | sigma[9]  sigma[10] sigma[11]   0      |
-  */
-
-  int count = 0;
-  clover_force = Zero();
-  for (int mu = 0; mu < 4; mu++)
-  {
-    force_mu = Zero();
-    for (int nu = 0; nu < 4; nu++)
-    {
-      if (mu == nu)
-      continue;
-
-      RealD factor;
-      if (nu == 4 || mu == 4)
-      {
-        factor = 2.0 * csw_t;
-      }
-      else
-      {
-        factor = 2.0 * csw_r;
-      }
-      PropagatorField Slambda = Gamma(sigma[count]) * Lambda; // sigma checked
-      Impl::TraceSpinImpl(lambda, Slambda);                   // traceSpin ok
-      force_mu -= factor*CloverHelpers::Cmunu(U, lambda, mu, nu);                   // checked
-      count++;
-    }
-
-    pokeLorentz(clover_force, U[mu] * force_mu, mu);
-  }
-  //clover_force *= csw;
-  force += clover_force;
-}
 
 // Derivative parts
-template<class Impl, class CloverHelpers>
+template <class Impl>
-void WilsonCloverFermion<Impl, CloverHelpers>::MooDeriv(GaugeField &mat, const FermionField &X, const FermionField &Y, int dag)
+void WilsonCloverFermion<Impl>::MooDeriv(GaugeField &mat, const FermionField &X, const FermionField &Y, int dag)
 {
   assert(0);
 }
 
 // Derivative parts
-template<class Impl, class CloverHelpers>
+template <class Impl>
-void WilsonCloverFermion<Impl, CloverHelpers>::MeeDeriv(GaugeField &mat, const FermionField &U, const FermionField &V, int dag)
+void WilsonCloverFermion<Impl>::MeeDeriv(GaugeField &mat, const FermionField &U, const FermionField &V, int dag)
 {
   assert(0); // not implemented yet
 }

Grid/qcd/action/fermion/implementation/WilsonFermionImplementation.h

@@ -4,13 +4,12 @@ Grid physics library, www.github.com/paboyle/Grid
 
 Source file: ./lib/qcd/action/fermion/WilsonFermion.cc
 
-Copyright (C) 2022
+Copyright (C) 2015
 
 Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
 Author: paboyle <paboyle@ph.ed.ac.uk>
-Author: Fabian Joswig <fabian.joswig@ed.ac.uk>
 
 This program is free software; you can redistribute it and/or modify
 it under the terms of the GNU General Public License as published by
@@ -600,47 +599,11 @@ void WilsonFermion<Impl>::ContractConservedCurrent(PropagatorField &q_in_1,
                                                    Current curr_type,
                                                    unsigned int mu)
 {
-  if(curr_type != Current::Vector)
-  {
-    std::cout << GridLogError << "Only the conserved vector current is implemented so far." << std::endl;
-    exit(1);
-  }
-
   Gamma g5(Gamma::Algebra::Gamma5);
   conformable(_grid, q_in_1.Grid());
   conformable(_grid, q_in_2.Grid());
   conformable(_grid, q_out.Grid());
-  auto UGrid= this->GaugeGrid();
+  assert(0);
-
-  PropagatorField tmp_shifted(UGrid);
-  PropagatorField g5Lg5(UGrid);
-  PropagatorField R(UGrid);
-  PropagatorField gmuR(UGrid);
-
-    Gamma::Algebra Gmu [] = {
-    Gamma::Algebra::GammaX,
-    Gamma::Algebra::GammaY,
-    Gamma::Algebra::GammaZ,
-    Gamma::Algebra::GammaT,
-  };
-  Gamma gmu=Gamma(Gmu[mu]);
-
-  g5Lg5=g5*q_in_1*g5;
-  tmp_shifted=Cshift(q_in_2,mu,1);
-  Impl::multLinkField(R,this->Umu,tmp_shifted,mu);
-  gmuR=gmu*R;
-
-  q_out=adj(g5Lg5)*R;
-  q_out-=adj(g5Lg5)*gmuR;
-
-  tmp_shifted=Cshift(q_in_1,mu,1);
-  Impl::multLinkField(g5Lg5,this->Umu,tmp_shifted,mu);
-  g5Lg5=g5*g5Lg5*g5;
-  R=q_in_2;
-  gmuR=gmu*R;
-
-  q_out-=adj(g5Lg5)*R;
-  q_out-=adj(g5Lg5)*gmuR;
 }
 
 
@@ -654,51 +617,9 @@ void WilsonFermion<Impl>::SeqConservedCurrent(PropagatorField &q_in,
                                               unsigned int tmax,
 					      ComplexField &lattice_cmplx)
 {
-  if(curr_type != Current::Vector)
-  {
-    std::cout << GridLogError << "Only the conserved vector current is implemented so far." << std::endl;
-    exit(1);
-  }
-
-  int tshift = (mu == Nd-1) ? 1 : 0;
-  unsigned int LLt    = GridDefaultLatt()[Tp];
   conformable(_grid, q_in.Grid());
   conformable(_grid, q_out.Grid());
-  auto UGrid= this->GaugeGrid();
+  assert(0);
-
-  PropagatorField tmp(UGrid);
-  PropagatorField Utmp(UGrid);
-  PropagatorField L(UGrid);
-  PropagatorField zz (UGrid);
-  zz=Zero();
-  LatticeInteger lcoor(UGrid); LatticeCoordinate(lcoor,Nd-1);
-
-    Gamma::Algebra Gmu [] = {
-    Gamma::Algebra::GammaX,
-    Gamma::Algebra::GammaY,
-    Gamma::Algebra::GammaZ,
-    Gamma::Algebra::GammaT,
-  };
-  Gamma gmu=Gamma(Gmu[mu]);
-
-  tmp = Cshift(q_in,mu,1);
-  Impl::multLinkField(Utmp,this->Umu,tmp,mu);
-  tmp = ( Utmp*lattice_cmplx - gmu*Utmp*lattice_cmplx ); // Forward hop
-  tmp = where((lcoor>=tmin),tmp,zz); // Mask the time
-  q_out = where((lcoor<=tmax),tmp,zz); // Position of current complicated
-
-  tmp = q_in *lattice_cmplx;
-  tmp = Cshift(tmp,mu,-1);
-  Impl::multLinkField(Utmp,this->Umu,tmp,mu+Nd); // Adjoint link
-  tmp = -( 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);
-  }
-  q_out+= where((lcoor<=tmax+tshift),tmp,zz); // Position of current complicated
 }
 
 NAMESPACE_END(Grid);

Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h

@@ -498,7 +498,6 @@ void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField
 #ifndef GRID_CUDA
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDag);     return;}
 #endif
-     acceleratorFenceComputeStream();
    } else if( interior ) {
      if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteDagInt); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteDagInt);    return;}
@@ -506,13 +505,11 @@ void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDagInt);     return;}
 #endif
    } else if( exterior ) {
-     acceleratorFenceComputeStream();
      if (Opt == WilsonKernelsStatic::OptGeneric    ) { KERNEL_CALL(GenericDhopSiteDagExt); return;}
      if (Opt == WilsonKernelsStatic::OptHandUnroll ) { KERNEL_CALL(HandDhopSiteDagExt);    return;}
 #ifndef GRID_CUDA
      if (Opt == WilsonKernelsStatic::OptInlineAsm  ) {  ASM_CALL(AsmDhopSiteDagExt);     return;}
 #endif
-     acceleratorFenceComputeStream();
    }
    assert(0 && " Kernel optimisation case not covered ");
   }

Grid/qcd/action/fermion/instantiation/CompactWilsonCloverFermionInstantiation.cc.master

@@ -1,44 +0,0 @@
-/*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid
-
-    Source file: ./lib/ qcd/action/fermion/instantiation/CompactWilsonCloverFermionInstantiation.cc.master
-
-    Copyright (C) 2017 - 2022
-
-    Author: paboyle <paboyle@ph.ed.ac.uk>
-    Author: Guido Cossu <guido.cossu@ed.ac.uk>
-    Author: Daniel Richtmann <daniel.richtmann@gmail.com>
-    Author: Mattia Bruno <mattia.bruno@cern.ch>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-/*  END LEGAL */
-
-#include <Grid/Grid.h>
-#include <Grid/qcd/spin/Dirac.h>
-#include <Grid/qcd/action/fermion/CompactWilsonCloverFermion.h>
-#include <Grid/qcd/action/fermion/implementation/CompactWilsonCloverFermionImplementation.h>
-#include <Grid/qcd/action/fermion/CloverHelpers.h>
-
-NAMESPACE_BEGIN(Grid);
-
-#include "impl.h"
-template class CompactWilsonCloverFermion<IMPLEMENTATION, CompactCloverHelpers<IMPLEMENTATION>>; 
-template class CompactWilsonCloverFermion<IMPLEMENTATION, CompactExpCloverHelpers<IMPLEMENTATION>>; 
-
-NAMESPACE_END(Grid);

Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/WilsonKernelsInstantiationWilsonAdjImplD.cc

@@ -1 +0,0 @@
-../WilsonKernelsInstantiation.cc.master

Grid/qcd/action/fermion/instantiation/WilsonAdjImplD/WilsonKernelsInstantiationWilsonAdjImplD.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);

Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/WilsonKernelsInstantiationWilsonAdjImplF.cc

@@ -1 +0,0 @@
-../WilsonKernelsInstantiation.cc.master

Grid/qcd/action/fermion/instantiation/WilsonAdjImplF/WilsonKernelsInstantiationWilsonAdjImplF.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);

Grid/qcd/action/fermion/instantiation/WilsonCloverFermionInstantiation.cc.master

@@ -8,8 +8,7 @@
 
     Author: paboyle <paboyle@ph.ed.ac.uk>
     Author: Guido Cossu <guido.cossu@ed.ac.uk>
-    Author: Mattia Bruno <mattia.bruno@cern.ch>
+
-    
     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
@@ -32,12 +31,10 @@
 #include <Grid/qcd/spin/Dirac.h>
 #include <Grid/qcd/action/fermion/WilsonCloverFermion.h>
 #include <Grid/qcd/action/fermion/implementation/WilsonCloverFermionImplementation.h>
-#include <Grid/qcd/action/fermion/CloverHelpers.h>
 
 NAMESPACE_BEGIN(Grid);
 
 #include "impl.h"
-template class WilsonCloverFermion<IMPLEMENTATION, CloverHelpers<IMPLEMENTATION>>; 
+template class WilsonCloverFermion<IMPLEMENTATION>; 
-template class WilsonCloverFermion<IMPLEMENTATION, ExpCloverHelpers<IMPLEMENTATION>>; 
 
 NAMESPACE_END(Grid);

Grid/qcd/action/fermion/instantiation/WilsonImplD/CompactWilsonCloverFermionInstantiationWilsonImplD.cc

@@ -1 +0,0 @@
-../CompactWilsonCloverFermionInstantiation.cc.master

Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonKernelsInstantiationWilsonImplD.cc

@@ -1 +0,0 @@
-../WilsonKernelsInstantiation.cc.master

Grid/qcd/action/fermion/instantiation/WilsonImplD/WilsonKernelsInstantiationWilsonImplD.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);

Grid/qcd/action/fermion/instantiation/WilsonImplF/CompactWilsonCloverFermionInstantiationWilsonImplF.cc

@@ -1 +0,0 @@
-../CompactWilsonCloverFermionInstantiation.cc.master

Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonKernelsInstantiationWilsonImplF.cc

@@ -1 +0,0 @@
-../WilsonKernelsInstantiation.cc.master

Grid/qcd/action/fermion/instantiation/WilsonImplF/WilsonKernelsInstantiationWilsonImplF.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);

Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/WilsonKernelsInstantiationWilsonTwoIndexAntiSymmetricImplD.cc

@@ -1 +0,0 @@
-../WilsonKernelsInstantiation.cc.master

Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplD/WilsonKernelsInstantiationWilsonTwoIndexAntiSymmetricImplD.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);

Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/WilsonKernelsInstantiationWilsonTwoIndexAntiSymmetricImplF.cc

@@ -1 +0,0 @@
-../WilsonKernelsInstantiation.cc.master

Grid/qcd/action/fermion/instantiation/WilsonTwoIndexAntiSymmetricImplF/WilsonKernelsInstantiationWilsonTwoIndexAntiSymmetricImplF.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);

Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/WilsonKernelsInstantiationWilsonTwoIndexSymmetricImplD.cc

@@ -1 +0,0 @@
-../WilsonKernelsInstantiation.cc.master

Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplD/WilsonKernelsInstantiationWilsonTwoIndexSymmetricImplD.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);

Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/WilsonKernelsInstantiationWilsonTwoIndexSymmetricImplF.cc

@@ -1 +0,0 @@
-../WilsonKernelsInstantiation.cc.master

Grid/qcd/action/fermion/instantiation/WilsonTwoIndexSymmetricImplF/WilsonKernelsInstantiationWilsonTwoIndexSymmetricImplF.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);

Grid/qcd/action/fermion/instantiation/ZWilsonImplD/WilsonKernelsInstantiationZWilsonImplD.cc

@@ -1 +0,0 @@
-../WilsonKernelsInstantiation.cc.master

Grid/qcd/action/fermion/instantiation/ZWilsonImplD/WilsonKernelsInstantiationZWilsonImplD.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);

Grid/qcd/action/fermion/instantiation/ZWilsonImplF/WilsonKernelsInstantiationZWilsonImplF.cc

@@ -1 +0,0 @@
-../WilsonKernelsInstantiation.cc.master

Grid/qcd/action/fermion/instantiation/ZWilsonImplF/WilsonKernelsInstantiationZWilsonImplF.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);

Grid/qcd/action/fermion/instantiation/generate_instantiations.sh

@@ -18,10 +18,6 @@ WILSON_IMPL_LIST=" \
 	   GparityWilsonImplF \
 	   GparityWilsonImplD "
 
-COMPACT_WILSON_IMPL_LIST=" \
-	   WilsonImplF \
-	   WilsonImplD "
-
 DWF_IMPL_LIST=" \
 	   WilsonImplF \
 	   WilsonImplD \
@@ -50,17 +46,7 @@ for impl in $WILSON_IMPL_LIST
 do
 for f in $CC_LIST
 do
-  ln -f -s ../$f.cc.master $impl/$f$impl.cc
+  ln -f -s ../$f.cc.master $impl/$f$impl.cc 
-done
-done
-
-CC_LIST="CompactWilsonCloverFermionInstantiation"
-
-for impl in $COMPACT_WILSON_IMPL_LIST
-do
-for f in $CC_LIST
-do
-  ln -f -s ../$f.cc.master $impl/$f$impl.cc
 done
 done
 
@@ -77,14 +63,14 @@ for impl in $DWF_IMPL_LIST $GDWF_IMPL_LIST
 do
 for f in $CC_LIST
 do
-  ln -f -s ../$f.cc.master $impl/$f$impl.cc
+  ln -f -s ../$f.cc.master $impl/$f$impl.cc 
 done
 done
 
 # overwrite the .cc file in Gparity directories
 for impl in $GDWF_IMPL_LIST
 do
-  ln -f -s ../WilsonKernelsInstantiationGparity.cc.master $impl/WilsonKernelsInstantiation$impl.cc
+  ln -f -s ../WilsonKernelsInstantiationGparity.cc.master $impl/WilsonKernelsInstantiation$impl.cc 
 done
 
 
@@ -98,7 +84,7 @@ for impl in $STAG_IMPL_LIST
 do
 for f in $CC_LIST
 do
-  ln -f -s ../$f.cc.master $impl/$f$impl.cc
+  ln -f -s ../$f.cc.master $impl/$f$impl.cc 
 done
 done
 

Grid/qcd/utils/GaugeFix.h

@@ -55,12 +55,12 @@ public:
     }
   }  
 
-  static void SteepestDescentGaugeFix(GaugeLorentz &Umu,Real & alpha,int maxiter,Real Omega_tol, Real Phi_tol,bool Fourier=false,int orthog=-1,bool err_on_no_converge=true) {
+  static void SteepestDescentGaugeFix(GaugeLorentz &Umu,Real & alpha,int maxiter,Real Omega_tol, Real Phi_tol,bool Fourier=false,int orthog=-1) {
     GridBase *grid = Umu.Grid();
     GaugeMat xform(grid);
-    SteepestDescentGaugeFix(Umu,xform,alpha,maxiter,Omega_tol,Phi_tol,Fourier,orthog,err_on_no_converge);
+    SteepestDescentGaugeFix(Umu,xform,alpha,maxiter,Omega_tol,Phi_tol,Fourier,orthog);
   }
-  static void SteepestDescentGaugeFix(GaugeLorentz &Umu,GaugeMat &xform,Real & alpha,int maxiter,Real Omega_tol, Real Phi_tol,bool Fourier=false,int orthog=-1,bool err_on_no_converge=true) {
+  static void SteepestDescentGaugeFix(GaugeLorentz &Umu,GaugeMat &xform,Real & alpha,int maxiter,Real Omega_tol, Real Phi_tol,bool Fourier=false,int orthog=-1) {
 
     GridBase *grid = Umu.Grid();
 
@@ -122,8 +122,6 @@ public:
 
       }
     }
-    std::cout << GridLogError << "Gauge fixing did not converge in " << maxiter << " iterations." << std::endl;
-    if (err_on_no_converge) assert(0);
   };
   static Real SteepestDescentStep(std::vector<GaugeMat> &U,GaugeMat &xform,Real & alpha, GaugeMat & dmuAmu,int orthog) {
     GridBase *grid = U[0].Grid();

Grid/qcd/utils/WilsonLoops.h

@@ -125,6 +125,7 @@ public:
     return sumplaq / vol / faces / Nc; // Nd , Nc dependent... FIXME
   }
 
+
   //////////////////////////////////////////////////
   // average over all x,y,z the temporal loop
   //////////////////////////////////////////////////
@@ -164,7 +165,7 @@ public:
 
     double vol = Umu.Grid()->gSites();
 
-    return p.real() / vol / (4.0 * Nc ) ;
+    return p.real() / vol / 4.0 / 3.0;
   };
 
   //////////////////////////////////////////////////

Grid/serialisation/JSON_IO.cc

@@ -26,7 +26,7 @@
     *************************************************************************************/
     /*  END LEGAL */
 #include <Grid/Grid.h>
-#ifndef GRID_HIP
+#if (!defined(GRID_CUDA)) && (!defined(GRID_HIP))
 
 NAMESPACE_BEGIN(Grid);
 
@@ -82,7 +82,7 @@ void JSONWriter::writeDefault(const std::string &s,	const std::string &x)
   if (s.size())
     ss_ << "\""<< s << "\" : \"" << os.str() << "\" ," ;
   else
-    ss_ << "\""<< os.str() << "\" ," ;
+    ss_ << os.str() << " ," ;
 }
 
 // Reader implementation ///////////////////////////////////////////////////////

Grid/serialisation/JSON_IO.h

@@ -54,7 +54,7 @@ namespace Grid
     void pop(void);
     template <typename U>
     void writeDefault(const std::string &s, const U &x);
-#if defined(GRID_CUDA) || defined(GRID_HIP)
+#ifdef __NVCC__
     void writeDefault(const std::string &s, const Grid::ComplexD &x) 
     { 
       std::complex<double> z(real(x),imag(x));
@@ -101,7 +101,7 @@ namespace Grid
     void readDefault(const std::string &s, std::vector<U> &output);
     template <typename U, typename P>
     void readDefault(const std::string &s, std::pair<U,P> &output);
-#if defined(GRID_CUDA) || defined(GRID_HIP)
+#ifdef __NVCC__
     void readDefault(const std::string &s, ComplexD &output)
     { 
       std::complex<double> z;

Grid/serialisation/Serialisation.h

@@ -36,7 +36,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 #include "BinaryIO.h"
 #include "TextIO.h"
 #include "XmlIO.h"
-#ifndef GRID_HIP
+#if (!defined(GRID_CUDA)) && (!defined(GRID_HIP))
 #include "JSON_IO.h"
 #endif
 

Grid/threads/Accelerator.cc

@@ -195,15 +195,12 @@ void acceleratorInit(void)
 #ifdef GRID_SYCL
 
 cl::sycl::queue *theGridAccelerator;
-cl::sycl::queue *theCopyAccelerator;
 void acceleratorInit(void)
 {
   int nDevices = 1;
   cl::sycl::gpu_selector selector;
   cl::sycl::device selectedDevice { selector };
   theGridAccelerator = new sycl::queue (selectedDevice);
-  //  theCopyAccelerator = new sycl::queue (selectedDevice);
-  theCopyAccelerator = theGridAccelerator; // Should proceed concurrenlty anyway.
 
 #ifdef GRID_SYCL_LEVEL_ZERO_IPC
   zeInit(0);

Grid/threads/Accelerator.h

@@ -247,6 +247,7 @@ inline int  acceleratorIsCommunicable(void *ptr)
 //////////////////////////////////////////////
 // SyCL acceleration
 //////////////////////////////////////////////
+
 #ifdef GRID_SYCL
 NAMESPACE_END(Grid);
 #include <CL/sycl.hpp>
@@ -261,7 +262,6 @@ NAMESPACE_END(Grid);
 NAMESPACE_BEGIN(Grid);
 
 extern cl::sycl::queue *theGridAccelerator;
-extern cl::sycl::queue *theCopyAccelerator;
 
 #ifdef __SYCL_DEVICE_ONLY__
 #define GRID_SIMT
@@ -289,7 +289,7 @@ accelerator_inline int acceleratorSIMTlane(int Nsimd) {
       cgh.parallel_for(					\
       cl::sycl::nd_range<3>(global,local), \
       [=] (cl::sycl::nd_item<3> item) /*mutable*/     \
-      [[intel::reqd_sub_group_size(16)]]	      \
+      [[intel::reqd_sub_group_size(8)]]	      \
       {						      \
       auto iter1    = item.get_global_id(0);	      \
       auto iter2    = item.get_global_id(1);	      \
@@ -298,19 +298,19 @@ accelerator_inline int acceleratorSIMTlane(int Nsimd) {
      });	   			              \
     });
 
-#define accelerator_barrier(dummy) { printf(" theGridAccelerator::wait()\n");  theGridAccelerator->wait(); }
+#define accelerator_barrier(dummy) theGridAccelerator->wait();
 
 inline void *acceleratorAllocShared(size_t bytes){ return malloc_shared(bytes,*theGridAccelerator);};
 inline void *acceleratorAllocDevice(size_t bytes){ return malloc_device(bytes,*theGridAccelerator);};
 inline void acceleratorFreeShared(void *ptr){free(ptr,*theGridAccelerator);};
 inline void acceleratorFreeDevice(void *ptr){free(ptr,*theGridAccelerator);};
-
+inline void acceleratorCopyDeviceToDeviceAsynch(void *from,void *to,size_t bytes)  {
-inline void acceleratorCopySynchronise(void) {  printf(" theCopyAccelerator::wait()\n"); theCopyAccelerator->wait(); }
+  theGridAccelerator->memcpy(to,from,bytes);
-inline void acceleratorCopyDeviceToDeviceAsynch(void *from,void *to,size_t bytes)  {  theCopyAccelerator->memcpy(to,from,bytes);}
+}
-inline void acceleratorCopyToDevice(void *from,void *to,size_t bytes)  { theCopyAccelerator->memcpy(to,from,bytes); theCopyAccelerator->wait();}
+inline void acceleratorCopySynchronise(void) {  theGridAccelerator->wait(); std::cout<<"acceleratorCopySynchronise() wait "<<std::endl; }
-inline void acceleratorCopyFromDevice(void *from,void *to,size_t bytes){ theCopyAccelerator->memcpy(to,from,bytes); theCopyAccelerator->wait();}
+inline void acceleratorCopyToDevice(void *from,void *to,size_t bytes)  { theGridAccelerator->memcpy(to,from,bytes); theGridAccelerator->wait();}
-inline void acceleratorMemSet(void *base,int value,size_t bytes) { theCopyAccelerator->memset(base,value,bytes); theCopyAccelerator->wait();}
+inline void acceleratorCopyFromDevice(void *from,void *to,size_t bytes){ theGridAccelerator->memcpy(to,from,bytes); theGridAccelerator->wait();}
-
+inline void acceleratorMemSet(void *base,int value,size_t bytes) { theGridAccelerator->memset(base,value,bytes); theGridAccelerator->wait();}
 inline int  acceleratorIsCommunicable(void *ptr)
 {
 #if 0
@@ -342,7 +342,7 @@ extern hipStream_t copyStream;
 /*These routines define mapping from thread grid to loop & vector lane indexing */
 accelerator_inline int acceleratorSIMTlane(int Nsimd) {
 #ifdef GRID_SIMT
-  return hipThreadIdx_x; 
+  return hipThreadIdx_z; 
 #else
   return 0;
 #endif
@@ -356,41 +356,19 @@ accelerator_inline int acceleratorSIMTlane(int Nsimd) {
       { __VA_ARGS__;}							\
     };									\
     int nt=acceleratorThreads();					\
-    dim3 hip_threads(nsimd, nt, 1);					 \
+    dim3 hip_threads(nt,1,nsimd);					\
-    dim3 hip_blocks ((num1+nt-1)/nt,num2,1); \
+    dim3 hip_blocks ((num1+nt-1)/nt,num2,1);				\
-    if(hip_threads.x * hip_threads.y * hip_threads.z <= 64){ \
+    hipLaunchKernelGGL(LambdaApply,hip_blocks,hip_threads,		\
-      hipLaunchKernelGGL(LambdaApply64,hip_blocks,hip_threads,		\
+		       0,0,						\
-            0,0,						\
+		       num1,num2,nsimd,lambda);				\
-            num1,num2,nsimd, lambda);				\
-    } else { \
-      hipLaunchKernelGGL(LambdaApply,hip_blocks,hip_threads,		\
-            0,0,						\
-            num1,num2,nsimd, lambda);				\
-    } \
   }
 
-
 template<typename lambda>  __global__
-__launch_bounds__(64,1)
-void LambdaApply64(uint64_t numx, uint64_t numy, uint64_t numz, lambda Lambda)
-{
-  // Following the same scheme as CUDA for now
-  uint64_t x = threadIdx.y + blockDim.y*blockIdx.x;
-  uint64_t y = threadIdx.z + blockDim.z*blockIdx.y;
-  uint64_t z = threadIdx.x;
-  if ( (x < numx) && (y<numy) && (z<numz) ) {
-    Lambda(x,y,z);
-  }
-}
-
-template<typename lambda>  __global__
-__launch_bounds__(1024,1)
 void LambdaApply(uint64_t numx, uint64_t numy, uint64_t numz, lambda Lambda)
 {
-  // Following the same scheme as CUDA for now
+  uint64_t x = hipThreadIdx_x + hipBlockDim_x*hipBlockIdx_x;
-  uint64_t x = threadIdx.y + blockDim.y*blockIdx.x;
+  uint64_t y = hipThreadIdx_y + hipBlockDim_y*hipBlockIdx_y;
-  uint64_t y = threadIdx.z + blockDim.z*blockIdx.y;
+  uint64_t z = hipThreadIdx_z ;//+ hipBlockDim_z*hipBlockIdx_z;
-  uint64_t z = threadIdx.x;
   if ( (x < numx) && (y<numy) && (z<numz) ) {
     Lambda(x,y,z);
   }
@@ -481,10 +459,9 @@ inline void acceleratorCopySynchronise(void) { hipStreamSynchronize(copyStream);
 #define accelerator_for2d(iter1, num1, iter2, num2, nsimd, ... ) thread_for2d(iter1,num1,iter2,num2,{ __VA_ARGS__ });
 
 accelerator_inline int acceleratorSIMTlane(int Nsimd) { return 0; } // CUDA specific
-
+inline void acceleratorCopyToDevice(void *from,void *to,size_t bytes)  { memcpy(to,from,bytes);}
-inline void acceleratorCopyToDevice(void *from,void *to,size_t bytes)  { thread_bcopy(from,to,bytes); }
+inline void acceleratorCopyFromDevice(void *from,void *to,size_t bytes){ memcpy(to,from,bytes);}
-inline void acceleratorCopyFromDevice(void *from,void *to,size_t bytes){ thread_bcopy(from,to,bytes);}
+inline void acceleratorCopyDeviceToDeviceAsynch(void *from,void *to,size_t bytes)  { memcpy(to,from,bytes);}
-inline void acceleratorCopyDeviceToDeviceAsynch(void *from,void *to,size_t bytes)  { thread_bcopy(from,to,bytes);}
 inline void acceleratorCopySynchronise(void) {};
 
 inline int  acceleratorIsCommunicable(void *ptr){ return 1; }
@@ -511,16 +488,7 @@ inline void *acceleratorAllocCpu(size_t bytes){return memalign(GRID_ALLOC_ALIGN,
 inline void acceleratorFreeCpu  (void *ptr){free(ptr);};
 #endif
 
-//////////////////////////////////////////////
-// Fencing needed ONLY for SYCL
-//////////////////////////////////////////////
 
-#ifdef GRID_SYCL
-inline void acceleratorFenceComputeStream(void){ accelerator_barrier();};
-#else
-// Ordering within a stream guaranteed on Nvidia & AMD
-inline void acceleratorFenceComputeStream(void){ };
-#endif
 
 ///////////////////////////////////////////////////
 // Synchronise across local threads for divergence resynch

Grid/threads/Threads.h

@@ -72,20 +72,3 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define thread_region                                       DO_PRAGMA(omp parallel)
 #define thread_critical                                     DO_PRAGMA(omp critical)
 
-#ifdef GRID_OMP
-inline void thread_bcopy(void *from, void *to,size_t bytes)
-{
-  uint64_t *ufrom = (uint64_t *)from;
-  uint64_t *uto   = (uint64_t *)to;
-  assert(bytes%8==0);
-  uint64_t words=bytes/8;
-  thread_for(w,words,{
-      uto[w] = ufrom[w];
-  });
-}
-#else
-inline void thread_bcopy(void *from, void *to,size_t bytes)
-{
-  bcopy(from,to,bytes);
-}
-#endif

Grid/util/Init.cc

@@ -167,13 +167,6 @@ void GridCmdOptionInt(std::string &str,int & val)
   return;
 }
 
-void GridCmdOptionFloat(std::string &str,float & val)
-{
-  std::stringstream ss(str);
-  ss>>val;
-  return;
-}
-
 
 void GridParseLayout(char **argv,int argc,
 		     Coordinate &latt_c,
@@ -534,7 +527,6 @@ void Grid_init(int *argc,char ***argv)
 void Grid_finalize(void)
 {
 #if defined (GRID_COMMS_MPI) || defined (GRID_COMMS_MPI3) || defined (GRID_COMMS_MPIT)
-  MPI_Barrier(MPI_COMM_WORLD);
   MPI_Finalize();
   Grid_unquiesce_nodes();
 #endif

Grid/util/Init.h

@@ -57,7 +57,6 @@ void GridCmdOptionCSL(std::string str,std::vector<std::string> & vec);
 template<class VectorInt>
 void GridCmdOptionIntVector(const std::string &str,VectorInt & vec);
 void GridCmdOptionInt(std::string &str,int & val);
-void GridCmdOptionFloat(std::string &str,float & val);
 
 
 void GridParseLayout(char **argv,int argc,

README.md

@@ -148,7 +148,7 @@ If you want to build all the tests at once just use `make tests`.
 - `--enable-mkl[=<path>]`: use Intel MKL for FFT (and LAPACK if enabled) routines. A UNIX prefix containing the library can be specified (optional).
 - `--enable-numa`: enable NUMA first touch optimisation
 - `--enable-simd=<code>`: setup Grid for the SIMD target `<code>` (default: `GEN`). A list of possible SIMD targets is detailed in a section below.
-- `--enable-gen-simd-width=<size>`: select the size (in bytes) of the generic SIMD vector type (default: 64 bytes).
+- `--enable-gen-simd-width=<size>`: select the size (in bytes) of the generic SIMD vector type (default: 32 bytes).
 - `--enable-comms=<comm>`: Use `<comm>` for message passing (default: `none`). A list of possible SIMD targets is detailed in a section below.
 - `--enable-rng={sitmo|ranlux48|mt19937}`: choose the RNG (default: `sitmo `).
 - `--disable-timers`: disable system dependent high-resolution timers.

benchmarks/Benchmark_mooee.cc

@@ -81,8 +81,8 @@ int main (int argc, char ** argv)
     Vector<Coeff_t> diag = Dw.bs;
     Vector<Coeff_t> upper= Dw.cs;
     Vector<Coeff_t> lower= Dw.cs;
-    upper[Ls-1]=-Dw.mass_minus*upper[Ls-1];
+    upper[Ls-1]=-Dw.mass*upper[Ls-1];
-    lower[0]   =-Dw.mass_plus*lower[0];
+    lower[0]   =-Dw.mass*lower[0];
     
     LatticeFermion r_eo(FGrid);
     LatticeFermion src_e (FrbGrid);

benchmarks/Benchmark_wilson_sweep.cc

@@ -44,13 +44,6 @@ void bench_wilson (
 		   double const     volume,
 		   int const           dag );
 
-void bench_wilson_eo (
-       LatticeFermion &    src,
-       LatticeFermion & result,
-       WilsonFermionR &     Dw,
-       double const     volume,
-       int const           dag );
-
 int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
@@ -117,8 +110,8 @@ int main (int argc, char ** argv)
 	  bench_wilson(src,result,Dw,volume,DaggerYes);
 	  std::cout << "\t";
     // EO
-    bench_wilson_eo(src_o,result_e,Dw,volume,DaggerNo);
+	  bench_wilson(src,result,Dw,volume,DaggerNo);
-    bench_wilson_eo(src_o,result_e,Dw,volume,DaggerYes);
+	  bench_wilson(src,result,Dw,volume,DaggerYes);
 	  std::cout << std::endl;
 	}
     }

configure.ac

@@ -159,7 +159,7 @@ case ${ac_ZMOBIUS} in
 esac
 ############### Nc
 AC_ARG_ENABLE([Nc],
-    [AC_HELP_STRING([--enable-Nc=2|3|4|5], [enable number of colours])],
+    [AC_HELP_STRING([--enable-Nc=2|3|4], [enable number of colours])],
     [ac_Nc=${enable_Nc}], [ac_Nc=3])
 
 case ${ac_Nc} in

systems/Crusher/config-command

@@ -1,12 +0,0 @@
-../../configure --enable-comms=mpi-auto \
---enable-unified=no \
---enable-shm=nvlink \
---enable-accelerator=hip \
---enable-gen-simd-width=64 \
---enable-simd=GPU \
---disable-fermion-reps \
---disable-gparity \
-CXX=hipcc MPICXX=mpicxx \
-CXXFLAGS="-fPIC -I/opt/rocm-4.5.0/include/ -std=c++14 -I${MPICH_DIR}/include " \
- LDFLAGS=" -L${MPICH_DIR}/lib -lmpi -L${CRAY_MPICH_ROOTDIR}/gtl/lib -lmpi_gtl_hsa "
-HIPFLAGS = --amdgpu-target=gfx90a

systems/Crusher/dwf.slurm

@@ -1,30 +0,0 @@
-#!/bin/bash
-# Begin LSF Directives
-#SBATCH -A LGT104
-#SBATCH -t 01:00:00
-##SBATCH -U openmpThu
-##SBATCH -p ecp
-#SBATCH -J DWF
-#SBATCH -o DWF.%J
-#SBATCH -e DWF.%J
-#SBATCH -N 1
-#SBATCH -n 1
-#SBATCH --exclusive  
-
-DIR=.
-module list
-#export MPIR_CVAR_GPU_EAGER_DEVICE_MEM=0
-export MPICH_GPU_SUPPORT_ENABLED=1
-export MPICH_SMP_SINGLE_COPY_MODE=XPMEM
-#export MPICH_SMP_SINGLE_COPY_MODE=NONE
-#export MPICH_SMP_SINGLE_COPY_MODE=CMA
-export OMP_NUM_THREADS=1
-
-AT=8
-echo MPICH_SMP_SINGLE_COPY_MODE $MPICH_SMP_SINGLE_COPY_MODE
-
-PARAMS=" --accelerator-threads ${AT} --grid 24.24.24.24 --shm-mpi 0 --mpi 1.1.1.1"
-
-srun --gpus-per-task 1 -n1 ./benchmarks/Benchmark_dwf_fp32 $PARAMS
-
-

systems/Crusher/dwf4.slurm

@@ -1,27 +0,0 @@
-#!/bin/bash
-# Begin LSF Directives
-#SBATCH -A LGT104
-#SBATCH -t 01:00:00
-##SBATCH -U openmpThu
-#SBATCH -J DWF
-#SBATCH -o DWF.%J
-#SBATCH -e DWF.%J
-#SBATCH -N 1
-#SBATCH -n 4
-#SBATCH --exclusive
-
-DIR=.
-module list
-export MPIR_CVAR_GPU_EAGER_DEVICE_MEM=0
-export MPICH_GPU_SUPPORT_ENABLED=1
-#export MPICH_SMP_SINGLE_COPY_MODE=XPMEM
-export MPICH_SMP_SINGLE_COPY_MODE=NONE
-#export MPICH_SMP_SINGLE_COPY_MODE=CMA
-export OMP_NUM_THREADS=4
-
-echo MPICH_SMP_SINGLE_COPY_MODE $MPICH_SMP_SINGLE_COPY_MODE
-PARAMS=" --accelerator-threads 8 --grid 32.32.64.64 --mpi 1.1.2.2 --comms-overlap --shm 2048 --shm-mpi 0"
-
-srun --gpus-per-task 1 -n4 ./mpiwrapper.sh ./benchmarks/Benchmark_dwf_fp32 $PARAMS
-
-

systems/Crusher/dwf8.slurm

@@ -1,27 +0,0 @@
-#!/bin/bash
-# Begin LSF Directives
-#SBATCH -A LGT104
-#SBATCH -t 01:00:00
-##SBATCH -U openmpThu
-#SBATCH -J DWF
-#SBATCH -o DWF.%J
-#SBATCH -e DWF.%J
-#SBATCH -N 1
-#SBATCH -n 8
-#SBATCH --exclusive
-
-DIR=.
-module list
-export MPIR_CVAR_GPU_EAGER_DEVICE_MEM=0
-export MPICH_GPU_SUPPORT_ENABLED=1
-export MPICH_SMP_SINGLE_COPY_MODE=XPMEM
-#export MPICH_SMP_SINGLE_COPY_MODE=NONE
-#export MPICH_SMP_SINGLE_COPY_MODE=CMA
-export OMP_NUM_THREADS=1
-
-echo MPICH_SMP_SINGLE_COPY_MODE $MPICH_SMP_SINGLE_COPY_MODE
-PARAMS=" --accelerator-threads 8 --grid 32.64.64.64 --mpi 1.2.2.2 --comms-overlap --shm 2048 --shm-mpi 0"
-
-srun --gpus-per-task 1 -n8 ./mpiwrapper.sh ./benchmarks/Benchmark_dwf_fp32 $PARAMS
-
-

systems/Crusher/mpiwrapper.sh

@@ -1,12 +0,0 @@
-#!/bin/bash
-
-lrank=$SLURM_LOCALID
-
-export ROCR_VISIBLE_DEVICES=$SLURM_LOCALID
-
-echo "`hostname` - $lrank device=$ROCR_VISIBLE_DEVICES binding=$BINDING"
-
-$*
-
-
-

systems/Crusher/sourceme.sh

@@ -1,5 +0,0 @@
-module load PrgEnv-gnu
-module load rocm/4.5.0
-module load gmp
-module load cray-fftw
-module load craype-accel-amd-gfx90a

systems/PVC/benchmarks/run-1tile.sh

@@ -1,62 +0,0 @@
-#!/bin/sh
-##SBATCH -p PVC-SPR-QZEH 
-##SBATCH -p PVC-ICX-QZNW
-#SBATCH -p QZ1J-ICX-PVC
-##SBATCH -p QZ1J-SPR-PVC-2C
-
-source /nfs/site/home/paboylex/ATS/GridNew/Grid/systems/PVC-nightly/setup.sh
-
-export NT=16
-
-export I_MPI_OFFLOAD=1
-export I_MPI_OFFLOAD_TOPOLIB=level_zero
-export I_MPI_OFFLOAD_DOMAIN_SIZE=-1
-
-# export IGC_EnableLSCFenceUGMBeforeEOT=0
-# export SYCL_PROGRAM_COMPILE_OPTIONS="-ze-opt-large-register-file=False"
-export SYCL_DEVICE_FILTER=gpu,level_zero
-#export IGC_ShaderDumpEnable=1 
-#export IGC_DumpToCurrentDir=1
-export I_MPI_OFFLOAD_CELL=tile
-export EnableImplicitScaling=0
-export EnableWalkerPartition=0
-export ZE_AFFINITY_MASK=0.0
-mpiexec -launcher ssh -n 1 -host localhost  ./Benchmark_dwf_fp32 --mpi 1.1.1.1 --grid 32.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1 --cacheblocking 8.8.8.8
-
-export ZE_AFFINITY_MASK=0
-export I_MPI_OFFLOAD_CELL=device
-export EnableImplicitScaling=1
-export EnableWalkerPartition=1
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-#mpiexec -launcher ssh -n 2 -host localhost  vtune -collect gpu-hotspots -knob gpu-sampling-interval=1 -data-limit=0 -r ./vtune_run4 -- ./wrap.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-overlap --shm-mpi 1
-
-#mpiexec  -launcher ssh -n 1 -host localhost ./wrap.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-overlap --shm-mpi 1
-
-#mpiexec  -launcher ssh -n 2 -host localhost ./wrap.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1
-
-#mpiexec  -launcher ssh -n 2 -host localhost ./wrap.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-overlap --shm-mpi 1
-
-#mpiexec  -launcher ssh -n 2 -host localhost ./wrap.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 0
-
-#mpirun -np 2 ./wrap.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.2 --grid 16.32.32.64 --accelerator-threads $NT --comms-sequential --shm-mpi 0
-#mpirun -np 2 ./wrap.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.2 --grid 32.32.32.64 --accelerator-threads $NT --comms-sequential --shm-mpi 1
-

systems/PVC/benchmarks/run-2tile-mpi.sh

@@ -1,26 +0,0 @@
-#!/bin/bash
-##SBATCH -p PVC-SPR-QZEH 
-##SBATCH -p PVC-ICX-QZNW
-#SBATCH -p QZ1J-ICX-PVC
-
-source /nfs/site/home/paboylex/ATS/GridNew/Grid/systems/PVC-nightly/setup.sh
-
-export NT=16
-
-
-# export IGC_EnableLSCFenceUGMBeforeEOT=0
-# export SYCL_PROGRAM_COMPILE_OPTIONS="-ze-opt-large-register-file=False"
-#export IGC_ShaderDumpEnable=1 
-#export IGC_DumpToCurrentDir=1
-export I_MPI_OFFLOAD=1
-export I_MPI_OFFLOAD_TOPOLIB=level_zero
-export I_MPI_OFFLOAD_DOMAIN_SIZE=-1
-export SYCL_DEVICE_FILTER=gpu,level_zero
-export I_MPI_OFFLOAD_CELL=tile
-export EnableImplicitScaling=0
-export EnableWalkerPartition=0
-
-mpiexec -launcher ssh -n 1 -host localhost  ./wrap.sh ./Benchmark_dwf_fp32 --mpi 1.1.1.1 --grid 32.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1 > 1tile.log
-
-mpiexec -launcher ssh -n 2 -host localhost  ./wrap.sh ./Benchmark_dwf_fp32 --mpi 2.1.1.1 --grid 64.32.32.32 --accelerator-threads $NT --comms-sequential --shm-mpi 1 > 2tile.log
-

systems/PVC/benchmarks/wrap.sh

@@ -1,14 +0,0 @@
-#!/bin/sh
-
-export ZE_AFFINITY_MASK=0.$MPI_LOCALRANKID
-
-echo Ranke $MPI_LOCALRANKID ZE_AFFINITY_MASK is $ZE_AFFINITY_MASK
-
-
-if [ $MPI_LOCALRANKID = "0" ] 
-then
-#  ~psteinbr/build_pti/ze_tracer -c $@
-  onetrace --chrome-kernel-timeline $@
-else
-  $@
-fi

systems/PVC/config-command

@@ -1,15 +0,0 @@
-INSTALL=/nfs/site/home/azusayax/install
-../../configure \
-	--enable-simd=GPU \
-	--enable-gen-simd-width=64 \
-	--enable-comms=mpi \
-	--disable-accelerator-cshift \
-	--disable-gparity \
-	--disable-fermion-reps \
-	--enable-shm=nvlink \
-	--enable-accelerator=sycl \
-	--enable-unified=yes \
-	CXX=mpicxx \
-	LDFLAGS="-fsycl-device-code-split=per_kernel -fsycl-device-lib=all -lze_loader -L$INSTALL/lib" \
-	CXXFLAGS="-cxx=dpcpp -fsycl-unnamed-lambda -fsycl -no-fma -I$INSTALL/include -Wtautological-constant-compare"
-

systems/PVC/setup.sh

@@ -1,11 +0,0 @@
-export https_proxy=http://proxy-chain.intel.com:911
-export LD_LIBRARY_PATH=/nfs/site/home/azusayax/install/lib:$LD_LIBRARY_PATH
-
-module load intel-release
-source /opt/intel/oneapi/PVC_setup.sh
-#source /opt/intel/oneapi/ATS_setup.sh
-module load intel/mpich/pvc45.3
-export PATH=~/ATS/pti-gpu/tools/onetrace/:$PATH
-
-#clsh embargo-ci-neo-022845
-#source /opt/intel/vtune_amplifier/amplxe-vars.sh

systems/mac-arm/config-command-mpi

@@ -1 +0,0 @@
-CXX=mpicxx-openmpi-mp CXXFLAGS=-I/opt/local/include/ LDFLAGS=-L/opt/local/lib/ ../../configure --enable-simd=GEN --enable-debug --enable-comms=mpi

tests/IO/Test_nersc_io.cc

@@ -147,7 +147,7 @@ int main (int argc, char ** argv)
   Complex p  = TensorRemove(Tp);
   std::cout<<GridLogMessage << "calculated plaquettes " <<p*PlaqScale<<std::endl;
 
-  Complex LinkTraceScale(1.0/vol/4.0/(Real)Nc);
+  Complex LinkTraceScale(1.0/vol/4.0/3.0);
   TComplex Tl = sum(LinkTrace);
   Complex l  = TensorRemove(Tl);
   std::cout<<GridLogMessage << "calculated link trace " <<l*LinkTraceScale<<std::endl;
@@ -157,10 +157,8 @@ int main (int argc, char ** argv)
   Complex ll= TensorRemove(TcP);
   std::cout<<GridLogMessage << "coarsened plaquettes sum to " <<ll*PlaqScale<<std::endl;
 
-  const string stNc   = to_string( Nc   ) ;
+  std::string clone2x3("./ckpoint_clone2x3.4000");
-  const string stNcM1 = to_string( Nc-1 ) ;
+  std::string clone3x3("./ckpoint_clone3x3.4000");
-  std::string clone2x3("./ckpoint_clone"+stNcM1+"x"+stNc+".4000");
-  std::string clone3x3("./ckpoint_clone"+stNc+"x"+stNc+".4000");
 
   NerscIO::writeConfiguration(Umu,clone3x3,0,precision32);
   NerscIO::writeConfiguration(Umu,clone2x3,1,precision32);

tests/core/Test_compact_wilson_clover_speedup.cc

@@ -1,226 +0,0 @@
-/*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid
-
-    Source file: ./tests/core/Test_compact_wilson_clover_speedup.cc
-
-    Copyright (C) 2020 - 2022
-
-    Author: Daniel Richtmann <daniel.richtmann@gmail.com>
-    Author: Nils Meyer       <nils.meyer@ur.de>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-
-#include <Grid/Grid.h>
-
-using namespace Grid;
-
-NAMESPACE_BEGIN(CommandlineHelpers);
-
-static bool checkPresent(int* argc, char*** argv, const std::string& option) {
-  return GridCmdOptionExists(*argv, *argv + *argc, option);
-}
-
-static std::string getContent(int* argc, char*** argv, const std::string& option) {
-  return GridCmdOptionPayload(*argv, *argv + *argc, option);
-}
-
-static int readInt(int* argc, char*** argv, std::string&& option, int defaultValue) {
-  std::string arg;
-  int         ret = defaultValue;
-  if(checkPresent(argc, argv, option)) {
-    arg = getContent(argc, argv, option);
-    GridCmdOptionInt(arg, ret);
-  }
-  return ret;
-}
-
-static float readFloat(int* argc, char*** argv, std::string&& option, float defaultValue) {
-  std::string arg;
-  float       ret = defaultValue;
-  if(checkPresent(argc, argv, option)) {
-    arg = getContent(argc, argv, option);
-    GridCmdOptionFloat(arg, ret);
-  }
-  return ret;
-}
-
-NAMESPACE_END(CommandlineHelpers);
-
-
-#define _grid_printf(LOGGER, ...) \
-  { \
-    if((LOGGER).isActive()) { /* this makes it safe to put, e.g., norm2 in the calling code w.r.t. performance */ \
-      char _printf_buf[1024]; \
-      std::sprintf(_printf_buf, __VA_ARGS__); \
-      std::cout << (LOGGER) << _printf_buf; \
-      fflush(stdout); \
-    } \
-  }
-#define grid_printf_msg(...) _grid_printf(GridLogMessage, __VA_ARGS__)
-
-
-template<typename Field>
-bool resultsAgree(const Field& ref, const Field& res, const std::string& name) {
-  RealD checkTolerance = (getPrecision<Field>::value == 2) ? 1e-15 : 1e-7;
-  Field diff(ref.Grid());
-  diff = ref - res;
-  auto absDev = norm2(diff);
-  auto relDev = absDev / norm2(ref);
-  std::cout << GridLogMessage
-            << "norm2(reference), norm2(" << name << "), abs. deviation, rel. deviation: " << norm2(ref) << " "
-            << norm2(res) << " " << absDev << " " << relDev << " -> check "
-            << ((relDev < checkTolerance) ? "passed" : "failed") << std::endl;
-
-  return relDev <= checkTolerance;
-}
-
-
-template<typename vCoeff_t>
-void runBenchmark(int* argc, char*** argv) {
-  // read from command line
-  const int   nIter        = CommandlineHelpers::readInt(     argc, argv, "--niter", 1000);
-  const RealD mass         = CommandlineHelpers::readFloat(   argc, argv, "--mass",  0.5);
-  const RealD csw          = CommandlineHelpers::readFloat(   argc, argv, "--csw",   1.0);
-  const RealD cF           = CommandlineHelpers::readFloat(   argc, argv, "--cF",    1.0);
-  const bool  antiPeriodic = CommandlineHelpers::checkPresent(argc, argv, "--antiperiodic");
-
-  // precision
-  static_assert(getPrecision<vCoeff_t>::value == 2 || getPrecision<vCoeff_t>::value == 1, "Incorrect precision"); // double or single
-  std::string precision = (getPrecision<vCoeff_t>::value == 2 ? "double" : "single");
-
-  // setup grids
-  GridCartesian*         UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd, vCoeff_t::Nsimd()), GridDefaultMpi());
-  GridRedBlackCartesian* UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
-  // clang-format on
-
-  // setup rng
-  std::vector<int> seeds({1, 2, 3, 4});
-  GridParallelRNG  pRNG(UGrid);
-  pRNG.SeedFixedIntegers(seeds);
-
-  // type definitions
-  typedef WilsonImpl<vCoeff_t, FundamentalRepresentation, CoeffReal> WImpl;
-  typedef WilsonCloverFermion<WImpl, CloverHelpers<WImpl>> WilsonCloverOperator;
-  typedef CompactWilsonCloverFermion<WImpl, CompactCloverHelpers<WImpl>> CompactWilsonCloverOperator;
-  typedef typename WilsonCloverOperator::FermionField Fermion;
-  typedef typename WilsonCloverOperator::GaugeField Gauge;
-
-  // setup fields
-  Fermion src(UGrid); random(pRNG, src);
-  Fermion ref(UGrid); ref = Zero();
-  Fermion res(UGrid); res = Zero();
-  Fermion hop(UGrid); hop = Zero();
-  Fermion diff(UGrid); diff = Zero();
-  Gauge   Umu(UGrid); SU3::HotConfiguration(pRNG, Umu);
-
-  // setup boundary phases
-  typename WilsonCloverOperator::ImplParams implParams;
-  std::vector<Complex> boundary_phases(Nd, 1.);
-  if(antiPeriodic) boundary_phases[Nd-1] = -1.;
-  implParams.boundary_phases = boundary_phases;
-  WilsonAnisotropyCoefficients anisParams;
-
-  // misc stuff needed for benchmarks
-  double volume=1.0; for(int mu=0; mu<Nd; mu++) volume*=UGrid->_fdimensions[mu];
-
-  // setup fermion operators
-  WilsonCloverOperator        Dwc(        Umu, *UGrid, *UrbGrid, mass, csw, csw,     anisParams, implParams);
-  CompactWilsonCloverOperator Dwc_compact(Umu, *UGrid, *UrbGrid, mass, csw, csw, cF, anisParams, implParams);
-
-  // now test the conversions
-  typename CompactWilsonCloverOperator::CloverField         tmp_ref(UGrid);  tmp_ref  = Dwc.CloverTerm;
-  typename CompactWilsonCloverOperator::CloverField         tmp_res(UGrid);  tmp_res  = Zero();
-  typename CompactWilsonCloverOperator::CloverField         tmp_diff(UGrid); tmp_diff = Zero();
-  typename CompactWilsonCloverOperator::CloverDiagonalField diagonal(UGrid); diagonal = Zero();
-  typename CompactWilsonCloverOperator::CloverTriangleField triangle(UGrid); diagonal = Zero();
-  CompactWilsonCloverOperator::CompactHelpers::ConvertLayout(tmp_ref, diagonal, triangle);
-  CompactWilsonCloverOperator::CompactHelpers::ConvertLayout(diagonal, triangle, tmp_res);
-  tmp_diff = tmp_ref - tmp_res;
-  std::cout << GridLogMessage << "conversion: ref, res, diff, eps"
-            << " " << norm2(tmp_ref)
-            << " " << norm2(tmp_res)
-            << " " << norm2(tmp_diff)
-            << " " << norm2(tmp_diff) / norm2(tmp_ref)
-            << std::endl;
-
-  // performance per site (use minimal values necessary)
-  double hop_flop_per_site            = 1320; // Rich's Talk + what Peter uses
-  double hop_byte_per_site            = (8 * 9 + 9 * 12) * 2 * getPrecision<vCoeff_t>::value * 4;
-  double clov_flop_per_site           = 504; // Rich's Talk and 1412.2629
-  double clov_byte_per_site           = (2 * 18 + 12 + 12) * 2 * getPrecision<vCoeff_t>::value * 4;
-  double clov_flop_per_site_performed = 1128;
-  double clov_byte_per_site_performed = (12 * 12 + 12 + 12) * 2 * getPrecision<vCoeff_t>::value * 4;
-
-  // total performance numbers
-  double hop_gflop_total            = volume * nIter * hop_flop_per_site / 1e9;
-  double hop_gbyte_total            = volume * nIter * hop_byte_per_site / 1e9;
-  double clov_gflop_total           = volume * nIter * clov_flop_per_site / 1e9;
-  double clov_gbyte_total           = volume * nIter * clov_byte_per_site / 1e9;
-  double clov_gflop_performed_total = volume * nIter * clov_flop_per_site_performed / 1e9;
-  double clov_gbyte_performed_total = volume * nIter * clov_byte_per_site_performed / 1e9;
-
-  // warmup + measure dhop
-  for(auto n : {1, 2, 3, 4, 5}) Dwc.Dhop(src, hop, 0);
-  double t0 = usecond();
-  for(int n = 0; n < nIter; n++) Dwc.Dhop(src, hop, 0);
-  double t1 = usecond();
-  double secs_hop = (t1-t0)/1e6;
-  grid_printf_msg("Performance(%35s, %s): %2.4f s, %6.0f GFlop/s, %6.0f GByte/s, speedup vs ref = %.2f, fraction of hop = %.2f\n",
-              "hop", precision.c_str(), secs_hop, hop_gflop_total/secs_hop, hop_gbyte_total/secs_hop, 0.0, secs_hop/secs_hop);
-
-#define BENCH_CLOVER_KERNEL(KERNEL) { \
-  /* warmup + measure reference clover */ \
-  for(auto n : {1, 2, 3, 4, 5}) Dwc.KERNEL(src, ref); \
-  double t2 = usecond(); \
-  for(int n = 0; n < nIter; n++) Dwc.KERNEL(src, ref); \
-  double t3 = usecond(); \
-  double secs_ref = (t3-t2)/1e6; \
-  grid_printf_msg("Performance(%35s, %s): %2.4f s, %6.0f GFlop/s, %6.0f GByte/s, speedup vs ref = %.2f, fraction of hop = %.2f\n", \
-                  "reference_"#KERNEL, precision.c_str(), secs_ref, clov_gflop_total/secs_ref, clov_gbyte_total/secs_ref, secs_ref/secs_ref, secs_ref/secs_hop); \
-  grid_printf_msg("Performance(%35s, %s): %2.4f s, %6.0f GFlop/s, %6.0f GByte/s, speedup vs ref = %.2f, fraction of hop = %.2f\n", /* to see how well the ET performs */  \
-                  "reference_"#KERNEL"_performed", precision.c_str(), secs_ref, clov_gflop_performed_total/secs_ref, clov_gbyte_performed_total/secs_ref, secs_ref/secs_ref, secs_ref/secs_hop); \
-\
-  /* warmup + measure compact clover */ \
-  for(auto n : {1, 2, 3, 4, 5}) Dwc_compact.KERNEL(src, res); \
-  double t4 = usecond(); \
-  for(int n = 0; n < nIter; n++) Dwc_compact.KERNEL(src, res); \
-  double t5 = usecond(); \
-  double secs_res = (t5-t4)/1e6; \
-  grid_printf_msg("Performance(%35s, %s): %2.4f s, %6.0f GFlop/s, %6.0f GByte/s, speedup vs ref = %.2f, fraction of hop = %.2f\n", \
-                  "compact_"#KERNEL, precision.c_str(), secs_res, clov_gflop_total/secs_res, clov_gbyte_total/secs_res, secs_ref/secs_res, secs_res/secs_hop); \
-  assert(resultsAgree(ref, res, #KERNEL)); \
-}
-
-  BENCH_CLOVER_KERNEL(Mooee);
-  BENCH_CLOVER_KERNEL(MooeeDag);
-  BENCH_CLOVER_KERNEL(MooeeInv);
-  BENCH_CLOVER_KERNEL(MooeeInvDag);
-
-  grid_printf_msg("finalize %s\n", precision.c_str());
-}
-
-int main(int argc, char** argv) {
-  Grid_init(&argc, &argv);
-
-  runBenchmark<vComplexD>(&argc, &argv);
-  runBenchmark<vComplexF>(&argc, &argv);
-
-  Grid_finalize();
-}

tests/core/Test_lie_generators.cc

@@ -9,7 +9,6 @@ Copyright (C) 2015
 Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: Guido Cossu <guido.cossu@ed.ac.uk>
-Author: Jamie Hudspith <renwick.james.hudspth@gmail.com>
 
 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
@@ -43,14 +42,14 @@ directory
 
 using namespace std;
 using namespace Grid;
-;
+ ;
 
 int main(int argc, char** argv) {
   Grid_init(&argc, &argv);
 
   std::vector<int> latt({4, 4, 4, 8});
   GridCartesian* grid = SpaceTimeGrid::makeFourDimGrid(
-						       latt, GridDefaultSimd(Nd, vComplex::Nsimd()), GridDefaultMpi());
+      latt, GridDefaultSimd(Nd, vComplex::Nsimd()), GridDefaultMpi());
 
   GridRedBlackCartesian* rbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(grid);
 
@@ -61,19 +60,15 @@ int main(int argc, char** argv) {
             << std::endl;
   SU2::printGenerators();
   std::cout << "Dimension of adjoint representation: "<< SU2Adjoint::Dimension << std::endl;
-
-  // guard as this code fails to compile for Nc != 3
-#if (Nc == 3)
-    
   SU2Adjoint::printGenerators();
   SU2::testGenerators();
   SU2Adjoint::testGenerators();
-    
+
   std::cout << GridLogMessage << "*********************************************"
-	    << std::endl;
+            << std::endl;
   std::cout << GridLogMessage << "* Generators for SU(Nc" << std::endl;
   std::cout << GridLogMessage << "*********************************************"
-	    << std::endl;
+            << std::endl;
   SU3::printGenerators();
   std::cout << "Dimension of adjoint representation: "<< SU3Adjoint::Dimension << std::endl;
   SU3Adjoint::printGenerators();
@@ -116,10 +111,12 @@ int main(int argc, char** argv) {
 
   // AdjointRepresentation has the predefined number of colours Nc
   //  Representations<FundamentalRepresentation, AdjointRepresentation, TwoIndexSymmetricRepresentation> RepresentationTypes(grid);  
+
+  
   LatticeGaugeField U(grid), V(grid);
   SU3::HotConfiguration<LatticeGaugeField>(gridRNG, U);
   SU3::HotConfiguration<LatticeGaugeField>(gridRNG, V);
-    
+
   // Adjoint representation
   // Test group structure
   // (U_f * V_f)_r = U_r * V_r
@@ -130,17 +127,17 @@ int main(int argc, char** argv) {
     SU3::LatticeMatrix Vmu = peekLorentz(V,mu);
     pokeLorentz(UV,Umu*Vmu, mu);
   }
-    
+
   AdjRep.update_representation(UV);
   typename AdjointRep<Nc>::LatticeField UVr = AdjRep.U;  // (U_f * V_f)_r 
-    
+
-    
+
   AdjRep.update_representation(U);
   typename AdjointRep<Nc>::LatticeField Ur = AdjRep.U;  // U_r
-    
+
   AdjRep.update_representation(V);
   typename AdjointRep<Nc>::LatticeField Vr = AdjRep.U;  // V_r
-    
+
   typename AdjointRep<Nc>::LatticeField UrVr(grid);
   UrVr = Zero();
   for (int mu = 0; mu < Nd; mu++) {
@@ -148,10 +145,10 @@ int main(int argc, char** argv) {
     typename AdjointRep<Nc>::LatticeMatrix Vrmu = peekLorentz(Vr,mu);
     pokeLorentz(UrVr,Urmu*Vrmu, mu);
   }
-    
+
   typename AdjointRep<Nc>::LatticeField Diff_check = UVr - UrVr;
   std::cout << GridLogMessage << "Group structure SU("<<Nc<<") check difference (Adjoint representation) : " << norm2(Diff_check) << std::endl;
-    
+
   // Check correspondence of algebra and group transformations
   // Create a random vector
   SU3::LatticeAlgebraVector h_adj(grid);
@@ -159,31 +156,32 @@ int main(int argc, char** argv) {
   random(gridRNG,h_adj);
   h_adj = real(h_adj);
   SU_Adjoint<Nc>::AdjointLieAlgebraMatrix(h_adj,Ar);
-    
+
   // Re-extract h_adj
   SU3::LatticeAlgebraVector h_adj2(grid);
   SU_Adjoint<Nc>::projectOnAlgebra(h_adj2, Ar);
   SU3::LatticeAlgebraVector h_diff = h_adj - h_adj2;
   std::cout << GridLogMessage << "Projections structure check vector difference (Adjoint representation) : " << norm2(h_diff) << std::endl;
-    
+
   // Exponentiate
   typename AdjointRep<Nc>::LatticeMatrix Uadj(grid);
   Uadj  = expMat(Ar, 1.0, 16);
-    
+
   typename AdjointRep<Nc>::LatticeMatrix uno(grid);
   uno = 1.0;
   // Check matrix Uadj, must be real orthogonal
   typename AdjointRep<Nc>::LatticeMatrix Ucheck = Uadj - conjugate(Uadj);
   std::cout << GridLogMessage << "Reality check: " << norm2(Ucheck)
-	    << std::endl;
+            << std::endl;
-    
+
   Ucheck = Uadj * adj(Uadj) - uno;
   std::cout << GridLogMessage << "orthogonality check 1: " << norm2(Ucheck)
-	    << std::endl;
+            << std::endl;
   Ucheck = adj(Uadj) * Uadj - uno;
   std::cout << GridLogMessage << "orthogonality check 2: " << norm2(Ucheck)
-	    << std::endl;
+            << std::endl;
-    
+      
+
   // Construct the fundamental matrix in the group
   SU3::LatticeMatrix Af(grid);
   SU3::FundamentalLieAlgebraMatrix(h_adj,Af);
@@ -195,65 +193,72 @@ int main(int argc, char** argv) {
   SU3::LatticeMatrix UnitCheck(grid);
   UnitCheck = Ufund * adj(Ufund) - uno_f;
   std::cout << GridLogMessage << "unitarity check 1: " << norm2(UnitCheck)
-	    << std::endl;
+            << std::endl;
   UnitCheck = adj(Ufund) * Ufund - uno_f;
   std::cout << GridLogMessage << "unitarity check 2: " << norm2(UnitCheck)
-	    << std::endl;
+            << std::endl;
+
 
   // Tranform to the adjoint representation
   U = Zero(); // fill this with only one direction
   pokeLorentz(U,Ufund,0); // the representation transf acts on full gauge fields
-    
+
   AdjRep.update_representation(U);
   Ur = AdjRep.U;  // U_r  
   typename AdjointRep<Nc>::LatticeMatrix Ur0 = peekLorentz(Ur,0); // this should be the same as Uadj
-    
+
   typename AdjointRep<Nc>::LatticeMatrix Diff_check_mat = Ur0 - Uadj;
   std::cout << GridLogMessage << "Projections structure check group difference : " << norm2(Diff_check_mat) << std::endl;
 
+
+
+
   // TwoIndexRep tests
+
   std::cout << GridLogMessage << "*********************************************"
-	    << std::endl;
+            << std::endl;
   std::cout << GridLogMessage << "*********************************************"
-	    << std::endl;
+            << std::endl;
 
 
   
   std::cout << GridLogMessage << "* eS^{ij} base for SU(2)" << std::endl;
   std::cout << GridLogMessage << "*********************************************"
-	    << std::endl;
+            << std::endl;
   std::cout << GridLogMessage << "Dimension of Two Index Symmetric representation: "<< SU2TwoIndexSymm::Dimension << std::endl;
   SU2TwoIndexSymm::printBase();
-  std::cout << GridLogMessage << "*********************************************"
+      std::cout << GridLogMessage << "*********************************************"
-	    << std::endl;
+            << std::endl;
-  std::cout << GridLogMessage << "Generators of Two Index Symmetric representation: "<< SU2TwoIndexSymm::Dimension << std::endl;
+        std::cout << GridLogMessage << "Generators of Two Index Symmetric representation: "<< SU2TwoIndexSymm::Dimension << std::endl;
   SU2TwoIndexSymm::printGenerators();
-  std::cout << GridLogMessage << "Test of Two Index Symmetric Generators: "<< SU2TwoIndexSymm::Dimension << std::endl;
+        std::cout << GridLogMessage << "Test of Two Index Symmetric Generators: "<< SU2TwoIndexSymm::Dimension << std::endl;
   SU2TwoIndexSymm::testGenerators();
   std::cout << GridLogMessage << "*********************************************"
-	    << std::endl;
+            << std::endl;
 
 
   
   std::cout << GridLogMessage << "*********************************************"
-	    << std::endl;
+            << std::endl;
   std::cout << GridLogMessage << "* eAS^{ij} base for SU(2)" << std::endl;
   
   std::cout << GridLogMessage << "*********************************************"
-	    << std::endl;
+            << std::endl;
   std::cout << GridLogMessage << "Dimension of Two Index anti-Symmetric representation: "<< SU2TwoIndexAntiSymm::Dimension << std::endl;
   SU2TwoIndexAntiSymm::printBase();
-  std::cout << GridLogMessage << "*********************************************"
+      std::cout << GridLogMessage << "*********************************************"
-	    << std::endl;
+            << std::endl;
-  std::cout << GridLogMessage << "Dimension of Two Index anti-Symmetric representation: "<< SU2TwoIndexAntiSymm::Dimension << std::endl;
+        std::cout << GridLogMessage << "Dimension of Two Index anti-Symmetric representation: "<< SU2TwoIndexAntiSymm::Dimension << std::endl;
   SU2TwoIndexAntiSymm::printGenerators();
   std::cout << GridLogMessage << "Test of Two Index anti-Symmetric Generators: "<< SU2TwoIndexAntiSymm::Dimension << std::endl;
   SU2TwoIndexAntiSymm::testGenerators();
   
+  
+  
   std::cout << GridLogMessage << "*********************************************"
-	    << std::endl;
+      << std::endl;
   std::cout << GridLogMessage << "Test for the Two Index Symmetric projectors"
-	    << std::endl;
+      << std::endl;
   // Projectors 
   SU3TwoIndexSymm::LatticeTwoIndexMatrix Gauss2(grid);
   random(gridRNG,Gauss2);
@@ -271,13 +276,13 @@ int main(int argc, char** argv) {
   SU3::LatticeAlgebraVector diff2 = ha - hb;
   std::cout << GridLogMessage << "Difference: " << norm2(diff) << std::endl;
   std::cout << GridLogMessage << "*********************************************"
-	    << std::endl;
+      << std::endl;
 
   
-  std::cout << GridLogMessage << "*********************************************"
+    std::cout << GridLogMessage << "*********************************************"
-	    << std::endl;
+      << std::endl;
   std::cout << GridLogMessage << "Test for the Two index anti-Symmetric projectors"
-	    << std::endl;
+      << std::endl;
   // Projectors
   SU3TwoIndexAntiSymm::LatticeTwoIndexMatrix Gauss2a(grid);
   random(gridRNG,Gauss2a);
@@ -295,11 +300,11 @@ int main(int argc, char** argv) {
   SU3::LatticeAlgebraVector diff2a = ha - hb;
   std::cout << GridLogMessage << "Difference: " << norm2(diff2a) << std::endl;
   std::cout << GridLogMessage << "*********************************************"
-	    << std::endl;
+      << std::endl;
     
   
   std::cout << GridLogMessage << "Two index Symmetric: Checking Group Structure"
-	    << std::endl;
+      << std::endl;
   // Testing HMC representation classes
   TwoIndexRep< Nc, Symmetric > TIndexRep(grid);
 
@@ -308,7 +313,7 @@ int main(int argc, char** argv) {
   LatticeGaugeField U2(grid), V2(grid);
   SU3::HotConfiguration<LatticeGaugeField>(gridRNG, U2);
   SU3::HotConfiguration<LatticeGaugeField>(gridRNG, V2);
-    
+  
   LatticeGaugeField UV2(grid);
   UV2 = Zero();
   for (int mu = 0; mu < Nd; mu++) {
@@ -316,16 +321,16 @@ int main(int argc, char** argv) {
     SU3::LatticeMatrix Vmu2 = peekLorentz(V2,mu);
     pokeLorentz(UV2,Umu2*Vmu2, mu);
   }
-    
+  
   TIndexRep.update_representation(UV2);
   typename TwoIndexRep< Nc, Symmetric >::LatticeField UVr2 = TIndexRep.U;  // (U_f * V_f)_r
-    
+  
   TIndexRep.update_representation(U2);
   typename TwoIndexRep< Nc, Symmetric >::LatticeField Ur2 = TIndexRep.U;  // U_r
-    
+  
   TIndexRep.update_representation(V2);
   typename TwoIndexRep< Nc, Symmetric >::LatticeField Vr2 = TIndexRep.U;  // V_r
-    
+  
   typename TwoIndexRep< Nc, Symmetric >::LatticeField Ur2Vr2(grid);
   Ur2Vr2 = Zero();
   for (int mu = 0; mu < Nd; mu++) {
@@ -333,11 +338,11 @@ int main(int argc, char** argv) {
     typename TwoIndexRep< Nc, Symmetric >::LatticeMatrix Vrmu2 = peekLorentz(Vr2,mu);
     pokeLorentz(Ur2Vr2,Urmu2*Vrmu2, mu);
   }
-    
+  
   typename TwoIndexRep< Nc, Symmetric >::LatticeField Diff_check2 = UVr2 - Ur2Vr2;
   std::cout << GridLogMessage << "Group structure SU("<<Nc<<") check difference (Two Index Symmetric): " << norm2(Diff_check2) << std::endl;
-    
+
-    
+  
   // Check correspondence of algebra and group transformations
   // Create a random vector
   SU3::LatticeAlgebraVector h_sym(grid);
@@ -345,31 +350,34 @@ int main(int argc, char** argv) {
   random(gridRNG,h_sym);
   h_sym = real(h_sym);
   SU_TwoIndex<Nc,Symmetric>::TwoIndexLieAlgebraMatrix(h_sym,Ar_sym);
-    
+  
   // Re-extract h_sym
   SU3::LatticeAlgebraVector h_sym2(grid);
   SU_TwoIndex< Nc, Symmetric>::projectOnAlgebra(h_sym2, Ar_sym);
   SU3::LatticeAlgebraVector h_diff_sym = h_sym - h_sym2;
   std::cout << GridLogMessage << "Projections structure check vector difference (Two Index Symmetric): " << norm2(h_diff_sym) << std::endl;
-    
+
+  
   // Exponentiate
   typename TwoIndexRep< Nc, Symmetric>::LatticeMatrix U2iS(grid);
   U2iS  = expMat(Ar_sym, 1.0, 16);
-    
+  
   typename TwoIndexRep< Nc, Symmetric>::LatticeMatrix uno2iS(grid);
   uno2iS = 1.0;
   // Check matrix U2iS, must be real orthogonal
   typename TwoIndexRep< Nc, Symmetric>::LatticeMatrix Ucheck2iS = U2iS - conjugate(U2iS);
   std::cout << GridLogMessage << "Reality check: " << norm2(Ucheck2iS)
-	    << std::endl;
+      << std::endl;
-    
+  
   Ucheck2iS = U2iS * adj(U2iS) - uno2iS;
   std::cout << GridLogMessage << "orthogonality check 1: " << norm2(Ucheck2iS)
-	    << std::endl;
+      << std::endl;
   Ucheck2iS = adj(U2iS) * U2iS - uno2iS;
   std::cout << GridLogMessage << "orthogonality check 2: " << norm2(Ucheck2iS)
-	    << std::endl;
+      << std::endl;
-    
+  
+  
+  
   // Construct the fundamental matrix in the group
   SU3::LatticeMatrix Af_sym(grid);
   SU3::FundamentalLieAlgebraMatrix(h_sym,Af_sym);
@@ -378,137 +386,147 @@ int main(int argc, char** argv) {
   SU3::LatticeMatrix UnitCheck2(grid);
   UnitCheck2 = Ufund2 * adj(Ufund2) - uno_f;
   std::cout << GridLogMessage << "unitarity check 1: " << norm2(UnitCheck2)
-	    << std::endl;
+      << std::endl;
   UnitCheck2 = adj(Ufund2) * Ufund2 - uno_f;
   std::cout << GridLogMessage << "unitarity check 2: " << norm2(UnitCheck2)
-	    << std::endl;
+      << std::endl;
-    
+  
-    
+
   // Tranform to the 2Index Sym representation
   U = Zero(); // fill this with only one direction
   pokeLorentz(U,Ufund2,0); // the representation transf acts on full gauge fields
-    
+  
   TIndexRep.update_representation(U);
   Ur2 = TIndexRep.U;  // U_r  
   typename TwoIndexRep< Nc, Symmetric>::LatticeMatrix Ur02 = peekLorentz(Ur2,0); // this should be the same as U2iS
-    
+  
   typename TwoIndexRep< Nc, Symmetric>::LatticeMatrix Diff_check_mat2 = Ur02 - U2iS;
   std::cout << GridLogMessage << "Projections structure check group difference (Two Index Symmetric): " << norm2(Diff_check_mat2) << std::endl;
-    
+  
+
+
+
   if (TwoIndexRep<Nc, AntiSymmetric >::Dimension != 1){
 
-    std::cout << GridLogMessage << "*********************************************"
+  std::cout << GridLogMessage << "*********************************************"
-	      << std::endl;
+      << std::endl;
     
   
-    std::cout << GridLogMessage << "Two Index anti-Symmetric: Check Group Structure"
+  std::cout << GridLogMessage << "Two Index anti-Symmetric: Check Group Structure"
-	      << std::endl;
+      << std::endl;
-    // Testing HMC representation classes
+  // Testing HMC representation classes
-    TwoIndexRep< Nc, AntiSymmetric > TIndexRepA(grid);
+  TwoIndexRep< Nc, AntiSymmetric > TIndexRepA(grid);
 
 
-    // Test group structure
+  // Test group structure
-    // (U_f * V_f)_r = U_r * V_r
+  // (U_f * V_f)_r = U_r * V_r
-    LatticeGaugeField U2A(grid), V2A(grid);
+  LatticeGaugeField U2A(grid), V2A(grid);
-    SU3::HotConfiguration<LatticeGaugeField>(gridRNG, U2A);
+  SU3::HotConfiguration<LatticeGaugeField>(gridRNG, U2A);
-    SU3::HotConfiguration<LatticeGaugeField>(gridRNG, V2A);
+  SU3::HotConfiguration<LatticeGaugeField>(gridRNG, V2A);
   
-    LatticeGaugeField UV2A(grid);
+  LatticeGaugeField UV2A(grid);
-    UV2A = Zero();
+  UV2A = Zero();
-    for (int mu = 0; mu < Nd; mu++) {
+  for (int mu = 0; mu < Nd; mu++) {
-      SU3::LatticeMatrix Umu2A = peekLorentz(U2,mu);
+    SU3::LatticeMatrix Umu2A = peekLorentz(U2,mu);
-      SU3::LatticeMatrix Vmu2A = peekLorentz(V2,mu);
+    SU3::LatticeMatrix Vmu2A = peekLorentz(V2,mu);
-      pokeLorentz(UV2A,Umu2A*Vmu2A, mu);
+    pokeLorentz(UV2A,Umu2A*Vmu2A, mu);
-    }
-  
-    TIndexRep.update_representation(UV2A);
-    typename TwoIndexRep< Nc, AntiSymmetric >::LatticeField UVr2A = TIndexRepA.U;  // (U_f * V_f)_r
-  
-    TIndexRep.update_representation(U2A);
-    typename TwoIndexRep< Nc, AntiSymmetric >::LatticeField Ur2A = TIndexRepA.U;  // U_r
-  
-    TIndexRep.update_representation(V2A);
-    typename TwoIndexRep< Nc, AntiSymmetric >::LatticeField Vr2A = TIndexRepA.U;  // V_r
-  
-    typename TwoIndexRep< Nc, AntiSymmetric >::LatticeField Ur2Vr2A(grid);
-    Ur2Vr2A = Zero();
-    for (int mu = 0; mu < Nd; mu++) {
-      typename TwoIndexRep< Nc, AntiSymmetric >::LatticeMatrix Urmu2A = peekLorentz(Ur2A,mu);
-      typename TwoIndexRep< Nc, AntiSymmetric >::LatticeMatrix Vrmu2A = peekLorentz(Vr2A,mu);
-      pokeLorentz(Ur2Vr2A,Urmu2A*Vrmu2A, mu);
-    }
-  
-    typename TwoIndexRep< Nc, AntiSymmetric >::LatticeField Diff_check2A = UVr2A - Ur2Vr2A;
-    std::cout << GridLogMessage << "Group structure SU("<<Nc<<") check difference (Two Index anti-Symmetric): " << norm2(Diff_check2A) << std::endl;
-
-  
-    // Check correspondence of algebra and group transformations
-    // Create a random vector
-    SU3::LatticeAlgebraVector h_Asym(grid);
-    typename TwoIndexRep< Nc, AntiSymmetric>::LatticeMatrix Ar_Asym(grid);
-    random(gridRNG,h_Asym);
-    h_Asym = real(h_Asym);
-    SU_TwoIndex< Nc, AntiSymmetric>::TwoIndexLieAlgebraMatrix(h_Asym,Ar_Asym);
-  
-    // Re-extract h_sym
-    SU3::LatticeAlgebraVector h_Asym2(grid);
-    SU_TwoIndex< Nc, AntiSymmetric>::projectOnAlgebra(h_Asym2, Ar_Asym);
-    SU3::LatticeAlgebraVector h_diff_Asym = h_Asym - h_Asym2;
-    std::cout << GridLogMessage << "Projections structure check vector difference (Two Index anti-Symmetric): " << norm2(h_diff_Asym) << std::endl;
-
-  
-    // Exponentiate
-    typename TwoIndexRep< Nc, AntiSymmetric>::LatticeMatrix U2iAS(grid);
-    U2iAS  = expMat(Ar_Asym, 1.0, 16);
-  
-    typename TwoIndexRep< Nc, AntiSymmetric>::LatticeMatrix uno2iAS(grid);
-    uno2iAS = 1.0;
-    // Check matrix U2iS, must be real orthogonal
-    typename TwoIndexRep< Nc, AntiSymmetric>::LatticeMatrix Ucheck2iAS = U2iAS - conjugate(U2iAS);
-    std::cout << GridLogMessage << "Reality check: " << norm2(Ucheck2iAS)
-	      << std::endl;
-  
-    Ucheck2iAS = U2iAS * adj(U2iAS) - uno2iAS;
-    std::cout << GridLogMessage << "orthogonality check 1: " << norm2(Ucheck2iAS)
-	      << std::endl;
-    Ucheck2iAS = adj(U2iAS) * U2iAS - uno2iAS;
-    std::cout << GridLogMessage << "orthogonality check 2: " << norm2(Ucheck2iAS)
-	      << std::endl;
-  
-  
-  
-    // Construct the fundamental matrix in the group
-    SU3::LatticeMatrix Af_Asym(grid);
-    SU3::FundamentalLieAlgebraMatrix(h_Asym,Af_Asym);
-    SU3::LatticeMatrix Ufund2A(grid);
-    Ufund2A  = expMat(Af_Asym, 1.0, 16);
-    SU3::LatticeMatrix UnitCheck2A(grid);
-    UnitCheck2A = Ufund2A * adj(Ufund2A) - uno_f;
-    std::cout << GridLogMessage << "unitarity check 1: " << norm2(UnitCheck2A)
-	      << std::endl;
-    UnitCheck2A = adj(Ufund2A) * Ufund2A - uno_f;
-    std::cout << GridLogMessage << "unitarity check 2: " << norm2(UnitCheck2A)
-	      << std::endl;
-  
-
-    // Tranform to the 2Index Sym representation
-    U = Zero(); // fill this with only one direction
-    pokeLorentz(U,Ufund2A,0); // the representation transf acts on full gauge fields
-  
-    TIndexRepA.update_representation(U);
-    Ur2A = TIndexRepA.U;  // U_r  
-    typename TwoIndexRep< Nc, AntiSymmetric>::LatticeMatrix Ur02A = peekLorentz(Ur2A,0); // this should be the same as U2iS
-  
-    typename TwoIndexRep< Nc, AntiSymmetric>::LatticeMatrix Diff_check_mat2A = Ur02A - U2iAS;
-    std::cout << GridLogMessage << "Projections structure check group difference (Two Index anti-Symmetric): " << norm2(Diff_check_mat2A) << std::endl;
-  
-  } else  {
-    std::cout << GridLogMessage << "Skipping Two Index anti-Symmetric tests "
-      "because representation is trivial (dim = 1)"
-	      << std::endl;
   }
+  
+  TIndexRep.update_representation(UV2A);
+  typename TwoIndexRep< Nc, AntiSymmetric >::LatticeField UVr2A = TIndexRepA.U;  // (U_f * V_f)_r
+  
+  TIndexRep.update_representation(U2A);
+  typename TwoIndexRep< Nc, AntiSymmetric >::LatticeField Ur2A = TIndexRepA.U;  // U_r
+  
+  TIndexRep.update_representation(V2A);
+  typename TwoIndexRep< Nc, AntiSymmetric >::LatticeField Vr2A = TIndexRepA.U;  // V_r
+  
+  typename TwoIndexRep< Nc, AntiSymmetric >::LatticeField Ur2Vr2A(grid);
+  Ur2Vr2A = Zero();
+  for (int mu = 0; mu < Nd; mu++) {
+    typename TwoIndexRep< Nc, AntiSymmetric >::LatticeMatrix Urmu2A = peekLorentz(Ur2A,mu);
+    typename TwoIndexRep< Nc, AntiSymmetric >::LatticeMatrix Vrmu2A = peekLorentz(Vr2A,mu);
+    pokeLorentz(Ur2Vr2A,Urmu2A*Vrmu2A, mu);
+  }
+  
+  typename TwoIndexRep< Nc, AntiSymmetric >::LatticeField Diff_check2A = UVr2A - Ur2Vr2A;
+  std::cout << GridLogMessage << "Group structure SU("<<Nc<<") check difference (Two Index anti-Symmetric): " << norm2(Diff_check2A) << std::endl;
+
+  
+  // Check correspondence of algebra and group transformations
+  // Create a random vector
+  SU3::LatticeAlgebraVector h_Asym(grid);
+  typename TwoIndexRep< Nc, AntiSymmetric>::LatticeMatrix Ar_Asym(grid);
+  random(gridRNG,h_Asym);
+  h_Asym = real(h_Asym);
+  SU_TwoIndex< Nc, AntiSymmetric>::TwoIndexLieAlgebraMatrix(h_Asym,Ar_Asym);
+  
+  // Re-extract h_sym
+  SU3::LatticeAlgebraVector h_Asym2(grid);
+  SU_TwoIndex< Nc, AntiSymmetric>::projectOnAlgebra(h_Asym2, Ar_Asym);
+  SU3::LatticeAlgebraVector h_diff_Asym = h_Asym - h_Asym2;
+  std::cout << GridLogMessage << "Projections structure check vector difference (Two Index anti-Symmetric): " << norm2(h_diff_Asym) << std::endl;
+
+  
+  // Exponentiate
+  typename TwoIndexRep< Nc, AntiSymmetric>::LatticeMatrix U2iAS(grid);
+  U2iAS  = expMat(Ar_Asym, 1.0, 16);
+  
+  typename TwoIndexRep< Nc, AntiSymmetric>::LatticeMatrix uno2iAS(grid);
+  uno2iAS = 1.0;
+  // Check matrix U2iS, must be real orthogonal
+  typename TwoIndexRep< Nc, AntiSymmetric>::LatticeMatrix Ucheck2iAS = U2iAS - conjugate(U2iAS);
+  std::cout << GridLogMessage << "Reality check: " << norm2(Ucheck2iAS)
+      << std::endl;
+  
+  Ucheck2iAS = U2iAS * adj(U2iAS) - uno2iAS;
+  std::cout << GridLogMessage << "orthogonality check 1: " << norm2(Ucheck2iAS)
+      << std::endl;
+  Ucheck2iAS = adj(U2iAS) * U2iAS - uno2iAS;
+  std::cout << GridLogMessage << "orthogonality check 2: " << norm2(Ucheck2iAS)
+      << std::endl;
+  
+  
+  
+  // Construct the fundamental matrix in the group
+  SU3::LatticeMatrix Af_Asym(grid);
+  SU3::FundamentalLieAlgebraMatrix(h_Asym,Af_Asym);
+  SU3::LatticeMatrix Ufund2A(grid);
+  Ufund2A  = expMat(Af_Asym, 1.0, 16);
+  SU3::LatticeMatrix UnitCheck2A(grid);
+  UnitCheck2A = Ufund2A * adj(Ufund2A) - uno_f;
+  std::cout << GridLogMessage << "unitarity check 1: " << norm2(UnitCheck2A)
+      << std::endl;
+  UnitCheck2A = adj(Ufund2A) * Ufund2A - uno_f;
+  std::cout << GridLogMessage << "unitarity check 2: " << norm2(UnitCheck2A)
+      << std::endl;
+  
+
+  // Tranform to the 2Index Sym representation
+  U = Zero(); // fill this with only one direction
+  pokeLorentz(U,Ufund2A,0); // the representation transf acts on full gauge fields
+  
+  TIndexRepA.update_representation(U);
+  Ur2A = TIndexRepA.U;  // U_r  
+  typename TwoIndexRep< Nc, AntiSymmetric>::LatticeMatrix Ur02A = peekLorentz(Ur2A,0); // this should be the same as U2iS
+  
+  typename TwoIndexRep< Nc, AntiSymmetric>::LatticeMatrix Diff_check_mat2A = Ur02A - U2iAS;
+  std::cout << GridLogMessage << "Projections structure check group difference (Two Index anti-Symmetric): " << norm2(Diff_check_mat2A) << std::endl;
+  
+} else  {
+  std::cout << GridLogMessage << "Skipping Two Index anti-Symmetric tests "
+                                 "because representation is trivial (dim = 1)"
+            << std::endl;
+}
+
+
+
+
+
+
+
+
 
-#endif
   
   Grid_finalize();
 }

tests/core/Test_reunitarise.cc

@@ -122,15 +122,14 @@ int main (int argc, char ** argv)
   std::cout << "Determinant defect before projection " <<norm2(detU)<<std::endl;
   tmp = U*adj(U) - ident;
   std::cout << "Unitarity check before projection    " << norm2(tmp)<<std::endl; 
-#if (Nc == 3)
+  
   ProjectSU3(U);
   detU= Determinant(U) ;
   detU= detU -1.0;
   std::cout << "Determinant ProjectSU3 defect " <<norm2(detU)<<std::endl;
   tmp = U*adj(U) - ident;
   std::cout << "Unitarity check after projection    " << norm2(tmp)<<std::endl; 
-#endif
+
-  
   ProjectSUn(UU);
   detUU= Determinant(UU);
   detUU= detUU -1.0;

tests/core/Test_wilson_clover.cc

@@ -2,12 +2,11 @@
 
     Grid physics library, www.github.com/paboyle/Grid 
 
-    Source file: ./tests/core/Test_wilson_clover.cc
+    Source file: ./benchmarks/Benchmark_wilson.cc
 
     Copyright (C) 2015
 
     Author: Guido Cossu <guido.cossu@ed.ac.uk>
-            Fabian Joswig <fabian.joswig@ed.ac.uk>
 
     This program is free software; you can redistribute it and/or modify
     it under the terms of the GNU General Public License as published by
@@ -68,6 +67,8 @@ int main(int argc, char **argv)
   tmp = Zero();
   FermionField err(&Grid);
   err = Zero();
+  FermionField err2(&Grid);
+  err2 = Zero();
   FermionField phi(&Grid);
   random(pRNG, phi);
   FermionField chi(&Grid);
@@ -76,8 +77,6 @@ int main(int argc, char **argv)
   SU<Nc>::HotConfiguration(pRNG, Umu);
   std::vector<LatticeColourMatrix> U(4, &Grid);
 
-  double tolerance = 1e-4;
-
   double volume = 1;
   for (int mu = 0; mu < Nd; mu++)
   {
@@ -89,7 +88,7 @@ int main(int argc, char **argv)
   RealD csw_t = 1.0;
 
   WilsonCloverFermionR Dwc(Umu, Grid, RBGrid, mass, csw_r, csw_t, anis, params);
-  CompactWilsonCloverFermionR Dwc_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 1.0, anis, params);
+  //Dwc.ImportGauge(Umu); // not necessary, included in the constructor
 
   std::cout << GridLogMessage << "==========================================================" << std::endl;
   std::cout << GridLogMessage << "= Testing that Deo + Doe = Dunprec " << std::endl;
@@ -113,24 +112,7 @@ int main(int argc, char **argv)
   setCheckerboard(r_eo, r_e);
 
   err = ref - r_eo;
-  std::cout << GridLogMessage << "EO norm diff\t" << norm2(err) << " (" << norm2(ref) << " - " << norm2(r_eo) << ")" << std::endl;
+  std::cout << GridLogMessage << "EO norm diff   " << norm2(err) << " " << norm2(ref) << " " << norm2(r_eo) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-
-
-  Dwc_compact.Meooe(src_e, r_o);
-  std::cout << GridLogMessage << "Applied Meo" << std::endl;
-  Dwc_compact.Meooe(src_o, r_e);
-  std::cout << GridLogMessage << "Applied Moe" << std::endl;
-  Dwc_compact.Dhop(src, ref, DaggerNo);
-
-  setCheckerboard(r_eo, r_o);
-  setCheckerboard(r_eo, r_e);
-
-  err = ref - r_eo;
-  std::cout << GridLogMessage << "EO norm diff compact\t" << norm2(err) << " (" << norm2(ref) << " - " << norm2(r_eo) << ")" << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
 
   std::cout << GridLogMessage << "==============================================================" << std::endl;
   std::cout << GridLogMessage << "= Test Ddagger is the dagger of D by requiring                " << std::endl;
@@ -170,22 +152,6 @@ int main(int argc, char **argv)
   std::cout << GridLogMessage << "pDce - conj(cDpo) " << pDce - conj(cDpo) << std::endl;
   std::cout << GridLogMessage << "pDco - conj(cDpe) " << pDco - conj(cDpe) << std::endl;
 
-  Dwc_compact.Meooe(chi_e, dchi_o);
-  Dwc_compact.Meooe(chi_o, dchi_e);
-  Dwc_compact.MeooeDag(phi_e, dphi_o);
-  Dwc_compact.MeooeDag(phi_o, dphi_e);
-
-  pDce = innerProduct(phi_e, dchi_e);
-  pDco = innerProduct(phi_o, dchi_o);
-  cDpe = innerProduct(chi_e, dphi_e);
-  cDpo = innerProduct(chi_o, dphi_o);
-
-  std::cout << GridLogMessage << "e compact " << pDce << " " << cDpe << std::endl;
-  std::cout << GridLogMessage << "o compact " << pDco << " " << cDpo << std::endl;
-
-  std::cout << GridLogMessage << "pDce - conj(cDpo) compact " << pDce - conj(cDpo) << std::endl;
-  std::cout << GridLogMessage << "pDco - conj(cDpe) compact " << pDco - conj(cDpe) << std::endl;
-
   std::cout << GridLogMessage << "==============================================================" << std::endl;
   std::cout << GridLogMessage << "= Test MeeInv Mee = 1   (if csw!=0)                           " << std::endl;
   std::cout << GridLogMessage << "==============================================================" << std::endl;
@@ -203,21 +169,7 @@ int main(int argc, char **argv)
   setCheckerboard(phi, phi_o);
 
   err = phi - chi;
-  std::cout << GridLogMessage << "norm diff " << norm2(err) << std::endl;
+  std::cout << GridLogMessage << "norm diff   " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  Dwc_compact.Mooee(chi_e, src_e);
-  Dwc_compact.MooeeInv(src_e, phi_e);
-
-  Dwc_compact.Mooee(chi_o, src_o);
-  Dwc_compact.MooeeInv(src_o, phi_o);
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = phi - chi;
-  std::cout << GridLogMessage << "norm diff compact " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
 
   std::cout << GridLogMessage << "==============================================================" << std::endl;
   std::cout << GridLogMessage << "= Test MeeDag MeeInvDag = 1    (if csw!=0)                    " << std::endl;
@@ -236,21 +188,7 @@ int main(int argc, char **argv)
   setCheckerboard(phi, phi_o);
 
   err = phi - chi;
-  std::cout << GridLogMessage << "norm diff " << norm2(err) << std::endl;
+  std::cout << GridLogMessage << "norm diff   " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  Dwc_compact.MooeeDag(chi_e, src_e);
-  Dwc_compact.MooeeInvDag(src_e, phi_e);
-
-  Dwc_compact.MooeeDag(chi_o, src_o);
-  Dwc_compact.MooeeInvDag(src_o, phi_o);
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = phi - chi;
-  std::cout << GridLogMessage << "norm diff compact " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
 
   std::cout << GridLogMessage << "==============================================================" << std::endl;
   std::cout << GridLogMessage << "= Test MeeInv MeeDag = 1      (if csw!=0)                     " << std::endl;
@@ -269,21 +207,7 @@ int main(int argc, char **argv)
   setCheckerboard(phi, phi_o);
 
   err = phi - chi;
-  std::cout << GridLogMessage << "norm diff " << norm2(err) << std::endl;
+  std::cout << GridLogMessage << "norm diff   " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  Dwc_compact.MooeeDag(chi_e, src_e);
-  Dwc_compact.MooeeInv(src_e, phi_e);
-
-  Dwc_compact.MooeeDag(chi_o, src_o);
-  Dwc_compact.MooeeInv(src_o, phi_o);
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = phi - chi;
-  std::cout << GridLogMessage << "norm diff compact " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
 
   std::cout << GridLogMessage << "================================================================" << std::endl;
   std::cout << GridLogMessage << "= Testing gauge covariance Clover term with EO preconditioning  " << std::endl;
@@ -325,7 +249,7 @@ int main(int argc, char **argv)
   /////////////////
 
   WilsonCloverFermionR Dwc_prime(U_prime, Grid, RBGrid, mass, csw_r, csw_t, anis, params);
-  CompactWilsonCloverFermionR Dwc_compact_prime(U_prime, Grid, RBGrid, mass, csw_r, csw_t, 1.0, anis, params);
+  Dwc_prime.ImportGauge(U_prime);
 
   tmp = Omega * src;
   pickCheckerboard(Even, src_e, tmp);
@@ -338,37 +262,7 @@ int main(int argc, char **argv)
   setCheckerboard(phi, phi_o);
 
   err = chi - adj(Omega) * phi;
-  std::cout << GridLogMessage << "norm diff " << norm2(err) << std::endl;
+  std::cout << GridLogMessage << "norm diff   " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  chi = Zero();
-  phi = Zero();
-  tmp = Zero();
-  pickCheckerboard(Even, chi_e, chi);
-  pickCheckerboard(Odd, chi_o, chi);
-  pickCheckerboard(Even, phi_e, phi);
-  pickCheckerboard(Odd, phi_o, phi);
-
-  Dwc_compact.Mooee(src_e, chi_e);
-  Dwc_compact.Mooee(src_o, chi_o);
-  setCheckerboard(chi, chi_e);
-  setCheckerboard(chi, chi_o);
-  setCheckerboard(src, src_e);
-  setCheckerboard(src, src_o);
-
-  tmp = Omega * src;
-  pickCheckerboard(Even, src_e, tmp);
-  pickCheckerboard(Odd, src_o, tmp);
-
-  Dwc_compact_prime.Mooee(src_e, phi_e);
-  Dwc_compact_prime.Mooee(src_o, phi_o);
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = chi - adj(Omega) * phi;
-  std::cout << GridLogMessage << "norm diff compact " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
 
   std::cout << GridLogMessage << "=================================================================" << std::endl;
   std::cout << GridLogMessage << "= Testing gauge covariance Clover term w/o EO preconditioning  " << std::endl;
@@ -378,6 +272,7 @@ int main(int argc, char **argv)
   phi = Zero();
 
   WilsonFermionR Dw(Umu, Grid, RBGrid, mass, params);
+  Dw.ImportGauge(Umu);
 
   Dw.M(src, result);
   Dwc.M(src, chi);
@@ -385,24 +280,13 @@ int main(int argc, char **argv)
   Dwc_prime.M(Omega * src, phi);
 
   WilsonFermionR Dw_prime(U_prime, Grid, RBGrid, mass, params);
+  Dw_prime.ImportGauge(U_prime);
   Dw_prime.M(Omega * src, result2);
 
-  err = result - adj(Omega) * result2;
-  std::cout << GridLogMessage << "norm diff Wilson              " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
   err = chi - adj(Omega) * phi;
-  std::cout << GridLogMessage << "norm diff WilsonClover        " << norm2(err) << std::endl;
+  err2 = result - adj(Omega) * result2;
-  assert(fabs(norm2(err)) < tolerance);
+  std::cout << GridLogMessage << "norm diff Wilson   " << norm2(err) << std::endl;
-
+  std::cout << GridLogMessage << "norm diff WilsonClover  " << norm2(err2) << std::endl;
-  chi = Zero();
-  phi = Zero();
-
-  Dwc_compact.M(src, chi);
-  Dwc_compact_prime.M(Omega * src, phi);
-
-  err = chi - adj(Omega) * phi;
-  std::cout << GridLogMessage << "norm diff CompactWilsonClover " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
 
   std::cout << GridLogMessage << "==========================================================" << std::endl;
   std::cout << GridLogMessage << "= Testing Mooee(csw=0) Clover to reproduce Mooee Wilson   " << std::endl;
@@ -412,6 +296,7 @@ int main(int argc, char **argv)
   phi = Zero();
   err = Zero();
   WilsonCloverFermionR Dwc_csw0(Umu, Grid, RBGrid, mass, 0.0, 0.0, anis, params); //  <-- Notice: csw=0
+  Dwc_csw0.ImportGauge(Umu);
 
   pickCheckerboard(Even, phi_e, phi);
   pickCheckerboard(Odd, phi_o, phi);
@@ -431,34 +316,7 @@ int main(int argc, char **argv)
   setCheckerboard(src, src_o);
 
   err = chi - phi;
-  std::cout << GridLogMessage << "norm diff " << norm2(err) << std::endl;
+  std::cout << GridLogMessage << "norm diff  " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  chi = Zero();
-  phi = Zero();
-  err = Zero();
-  CompactWilsonCloverFermionR Dwc_compact_csw0(Umu, Grid, RBGrid, mass, 0.0, 0.0, 1.0, anis, params); //  <-- Notice: csw=0
-
-  pickCheckerboard(Even, phi_e, phi);
-  pickCheckerboard(Odd, phi_o, phi);
-  pickCheckerboard(Even, chi_e, chi);
-  pickCheckerboard(Odd, chi_o, chi);
-
-  Dw.Mooee(src_e, chi_e);
-  Dw.Mooee(src_o, chi_o);
-  Dwc_compact_csw0.Mooee(src_e, phi_e);
-  Dwc_compact_csw0.Mooee(src_o, phi_o);
-
-  setCheckerboard(chi, chi_e);
-  setCheckerboard(chi, chi_o);
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-  setCheckerboard(src, src_e);
-  setCheckerboard(src, src_o);
-
-  err = chi - phi;
-  std::cout << GridLogMessage << "norm diff compact " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
 
   std::cout << GridLogMessage << "==========================================================" << std::endl;
   std::cout << GridLogMessage << "= Testing EO operator is equal to the unprec              " << std::endl;
@@ -490,41 +348,9 @@ int main(int argc, char **argv)
   setCheckerboard(phi, phi_o);
 
   err = ref - phi;
-  std::cout << GridLogMessage << "ref (unpreconditioned operator) diff         : " << norm2(ref) << std::endl;
+  std::cout << GridLogMessage << "ref (unpreconditioned operator) diff  :" << norm2(ref) << std::endl;
-  std::cout << GridLogMessage << "phi (EO decomposition)          diff         : " << norm2(phi) << std::endl;
+  std::cout << GridLogMessage << "phi (EO decomposition)          diff  :" << norm2(phi) << std::endl;
-  std::cout << GridLogMessage << "norm diff                                    : " << norm2(err) << std::endl;
+  std::cout << GridLogMessage << "norm diff                             :" << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  chi = Zero();
-  phi = Zero();
-  err = Zero();
-
-  pickCheckerboard(Even, phi_e, phi);
-  pickCheckerboard(Odd, phi_o, phi);
-  pickCheckerboard(Even, chi_e, chi);
-  pickCheckerboard(Odd, chi_o, chi);
-
-  // M phi = (Mooee src_e + Meooe src_o , Meooe src_e + Mooee src_o)
-
-  Dwc_compact.M(src, ref); // Reference result from the unpreconditioned operator
-
-  // EO matrix
-  Dwc_compact.Mooee(src_e, chi_e); 
-  Dwc_compact.Mooee(src_o, chi_o);
-  Dwc_compact.Meooe(src_o, phi_e);
-  Dwc_compact.Meooe(src_e, phi_o);
-
-  phi_o += chi_o;
-  phi_e += chi_e;
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = ref - phi;
-  std::cout << GridLogMessage << "ref (unpreconditioned operator) diff compact : " << norm2(ref) << std::endl;
-  std::cout << GridLogMessage << "phi (EO decomposition)          diff compact : " << norm2(phi) << std::endl;
-  std::cout << GridLogMessage << "norm diff compact                            : " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
 
   Grid_finalize();
 }

tests/core/Test_wilson_conserved_current.cc

@@ -1,253 +0,0 @@
-/*************************************************************************************
-Grid physics library, www.github.com/paboyle/Grid
-
-Source file: ./tests/Test_cayley_cg.cc
-
-Copyright (C) 2022
-
-Author: Peter Boyle <paboyle@ph.ed.ac.uk>
-Author: Fabian Joswig <fabian.joswig@ed.ac.uk>
-
-This program is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2 of the License, or
-(at your option) any later version.
-
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
-
-You should have received a copy of the GNU General Public License along
-with this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-See the full license in the file "LICENSE" in the top level distribution directory
-*************************************************************************************/
-/*  END LEGAL */
-#include <Grid/Grid.h>
-
-using namespace std;
-using namespace Grid;
-
-
-template<class What>
-void  TestConserved(What & Dw,
-		    LatticeGaugeField &Umu,
-		    GridCartesian         * UGrid,	       GridRedBlackCartesian * UrbGrid,
-		    GridParallelRNG *RNG4);
-
-  Gamma::Algebra Gmu [] = {
-    Gamma::Algebra::GammaX,
-    Gamma::Algebra::GammaY,
-    Gamma::Algebra::GammaZ,
-    Gamma::Algebra::GammaT,
-    Gamma::Algebra::Gamma5
-  };
-
-int main (int argc, char ** argv)
-{
-  Grid_init(&argc,&argv);
-
-  int threads = GridThread::GetThreads();
-  std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
-
-  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(),
-								   GridDefaultSimd(Nd,vComplex::Nsimd()),
-								   GridDefaultMpi());
-  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
-
-  std::vector<int> seeds5({5,6,7,8});
-  GridParallelRNG          RNG4(UGrid);
-  std::vector<int> seeds4({1,2,3,4}); RNG4.SeedFixedIntegers(seeds4);
-
-  LatticeGaugeField Umu(UGrid);
-  if( argc > 1 && argv[1][0] != '-' )
-  {
-    std::cout<<GridLogMessage <<"Loading configuration from "<<argv[1]<<std::endl;
-    FieldMetaData header;
-    NerscIO::readConfiguration(Umu, header, argv[1]);
-  }
-  else
-  {
-    std::cout<<GridLogMessage <<"Using hot configuration"<<std::endl;
-    SU<Nc>::HotConfiguration(RNG4,Umu);
-  }
-
-  typename WilsonCloverFermionR::ImplParams params;
-  WilsonAnisotropyCoefficients anis;
-  RealD mass = 0.1;
-  RealD csw_r = 1.0;
-  RealD csw_t = 1.0;
-
-  std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  std::cout<<GridLogMessage <<"WilsonFermion test"<<std::endl;
-  std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  WilsonFermionR Dw(Umu,*UGrid,*UrbGrid,mass,params);
-  TestConserved<WilsonFermionR>(Dw,Umu,UGrid,UrbGrid,&RNG4);
-
-  std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  std::cout<<GridLogMessage <<"WilsonCloverFermion test"<<std::endl;
-  std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  WilsonCloverFermionR Dwc(Umu, *UGrid, *UrbGrid, mass, csw_r, csw_t, anis, params);
-  TestConserved<WilsonCloverFermionR>(Dwc,Umu,UGrid,UrbGrid,&RNG4);
-
-  std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  std::cout<<GridLogMessage <<"CompactWilsonCloverFermion test"<<std::endl;
-  std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  CompactWilsonCloverFermionR Dwcc(Umu, *UGrid, *UrbGrid, mass, csw_r, csw_t, 1.0, anis, params);
-  TestConserved<CompactWilsonCloverFermionR>(Dwcc,Umu,UGrid,UrbGrid,&RNG4);
-
-  std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  std::cout<<GridLogMessage <<"WilsonExpCloverFermion test"<<std::endl;
-  std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  WilsonExpCloverFermionR Dewc(Umu, *UGrid, *UrbGrid, mass, csw_r, csw_t, anis, params);
-  TestConserved<WilsonExpCloverFermionR>(Dewc,Umu,UGrid,UrbGrid,&RNG4);
-
-  std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  std::cout<<GridLogMessage <<"CompactWilsonExpCloverFermion test"<<std::endl;
-  std::cout<<GridLogMessage <<"=================================="<<std::endl;
-  CompactWilsonExpCloverFermionR Dewcc(Umu, *UGrid, *UrbGrid, mass, csw_r, csw_t, 1.0, anis, params);
-  TestConserved<CompactWilsonExpCloverFermionR>(Dewcc,Umu,UGrid,UrbGrid,&RNG4);
-
-  Grid_finalize();
-}
-
-
-
-template<class Action>
-void  TestConserved(Action & Dw,
-		    LatticeGaugeField &Umu,
-		    GridCartesian         * UGrid,	       GridRedBlackCartesian * UrbGrid,
-		    GridParallelRNG *RNG4)
-{
-  LatticePropagator phys_src(UGrid);
-  LatticePropagator seqsrc(UGrid);
-  LatticePropagator prop4(UGrid);
-  LatticePropagator Vector_mu(UGrid);
-  LatticeComplex    SV (UGrid);
-  LatticeComplex    VV (UGrid);
-  LatticePropagator seqprop(UGrid);
-
-  SpinColourMatrix kronecker; kronecker=1.0;
-  Coordinate coor({0,0,0,0});
-  phys_src=Zero();
-  pokeSite(kronecker,phys_src,coor);
-
-  ConjugateGradient<LatticeFermion> CG(1.0e-16,100000);
-  SchurRedBlackDiagTwoSolve<LatticeFermion> schur(CG);
-  ZeroGuesser<LatticeFermion> zpg;
-  for(int s=0;s<Nd;s++){
-    for(int c=0;c<Nc;c++){
-      LatticeFermion src4  (UGrid);
-      PropToFerm<Action>(src4,phys_src,s,c);
-
-      LatticeFermion result4(UGrid); result4=Zero();
-      schur(Dw,src4,result4,zpg);
-      std::cout<<GridLogMessage<<"spin "<<s<<" color "<<c<<" norm2(sourc4d) "<<norm2(src4)
-               <<" norm2(result4d) "<<norm2(result4)<<std::endl;
-      FermToProp<Action>(prop4,result4,s,c);
-    }
-  }
-
-  auto curr = Current::Vector;
-  const int mu_J=0;
-  const int t_J=0;
-
-  LatticeComplex    ph (UGrid); ph=1.0;
-
-  Dw.SeqConservedCurrent(prop4,
-			   seqsrc,
-			   phys_src,
-			   curr,
-			   mu_J,
-			   t_J,
-			   t_J,// whole lattice
-			   ph);
-
-  for(int s=0;s<Nd;s++){
-    for(int c=0;c<Nc;c++){
-
-      LatticeFermion src4  (UGrid);
-      PropToFerm<Action>(src4,seqsrc,s,c);
-
-      LatticeFermion result4(UGrid); result4=Zero();
-      schur(Dw,src4,result4,zpg);
-
-      FermToProp<Action>(seqprop,result4,s,c);
-    }
-  }
-
-  Gamma g5(Gamma::Algebra::Gamma5);
-  Gamma gT(Gamma::Algebra::GammaT);
-
-  std::vector<TComplex> sumSV;
-  std::vector<TComplex> sumVV;
-
-  Dw.ContractConservedCurrent(prop4,prop4,Vector_mu,phys_src,Current::Vector,Tdir);
-
-  SV       = trace(Vector_mu);        // Scalar-Vector conserved current
-  VV       = trace(gT*Vector_mu);     // (local) Vector-Vector conserved current
-
-  // Spatial sum
-  sliceSum(SV,sumSV,Tdir);
-  sliceSum(VV,sumVV,Tdir);
-
-  const int Nt{static_cast<int>(sumSV.size())};
-
-  std::cout<<GridLogMessage<<"Vector Ward identity by timeslice (~ 0)"<<std::endl;
-  for(int t=0;t<Nt;t++){
-    std::cout<<GridLogMessage <<" t "<<t<<" SV "<<real(TensorRemove(sumSV[t]))<<" VV "<<real(TensorRemove(sumVV[t]))<<std::endl;
-    assert(abs(real(TensorRemove(sumSV[t]))) < 1e-10);
-    assert(abs(real(TensorRemove(sumVV[t]))) < 1e-2);
-  }
-
-  ///////////////////////////////
-  // 3pt vs 2pt check
-  ///////////////////////////////
-  {
-    Gamma::Algebra        gA = Gamma::Algebra::Identity;
-    Gamma                 g(gA);
-
-    LatticePropagator cur(UGrid);
-    LatticePropagator tmp(UGrid);
-    LatticeComplex c(UGrid);
-    SpinColourMatrix qSite;
-    peekSite(qSite, seqprop, coor);
-
-    Complex               test_S, test_V, check_S, check_V;
-
-    std::vector<TComplex> check_buf;
-
-    test_S = trace(qSite*g);
-    test_V = trace(qSite*g*Gamma::gmu[mu_J]);
-
-    Dw.ContractConservedCurrent(prop4,prop4,cur,phys_src,curr,mu_J);
-
-    c = trace(cur*g);
-    sliceSum(c, check_buf, Tp);
-    check_S = TensorRemove(check_buf[t_J]);
-
-    auto gmu=Gamma::gmu[mu_J];
-    c = trace(cur*g*gmu);
-    sliceSum(c, check_buf, Tp);
-    check_V = TensorRemove(check_buf[t_J]);
-
-
-    std::cout<<GridLogMessage << std::setprecision(14)<<"Test S  = " << abs(test_S)   << std::endl;
-    std::cout<<GridLogMessage << "Test V  = " << abs(test_V) << std::endl;
-    std::cout<<GridLogMessage << "Check S = " << abs(check_S) << std::endl;
-    std::cout<<GridLogMessage << "Check V = " << abs(check_V) << std::endl;
-
-    // Check difference = 0
-    check_S = check_S - test_S;
-    check_V = check_V - test_V;
-
-    std::cout<<GridLogMessage << "Consistency check for sequential conserved " <<std::endl;
-    std::cout<<GridLogMessage << "Diff S  = " << abs(check_S) << std::endl;
-    assert(abs(check_S) < 1e-8);
-    std::cout<<GridLogMessage << "Diff V  = " << abs(check_V) << std::endl;
-    assert(abs(check_V) < 1e-8);
-  }
-
-}

tests/core/Test_wilson_exp_clover.cc

@@ -1,530 +0,0 @@
-/*************************************************************************************
-
-    Grid physics library, www.github.com/paboyle/Grid 
-
-    Source file: ./tests/core/Test_wilson_exp_clover.cc
-
-    Copyright (C) 2022
-
-    Author: Guido Cossu <guido.cossu@ed.ac.uk>
-            Fabian Joswig <fabian.joswig@ed.ac.uk>
-
-    This program is free software; you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License along
-    with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
-    See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
-/*  END LEGAL */
-#include <Grid/Grid.h>
-
-using namespace std;
-using namespace Grid;
-
-int main(int argc, char **argv)
-{
-  Grid_init(&argc, &argv);
-
-  auto latt_size = GridDefaultLatt();
-  auto simd_layout = GridDefaultSimd(Nd, vComplex::Nsimd());
-  auto mpi_layout = GridDefaultMpi();
-  GridCartesian Grid(latt_size, simd_layout, mpi_layout);
-  GridRedBlackCartesian RBGrid(&Grid);
-
-  int threads = GridThread::GetThreads();
-  std::cout << GridLogMessage << "Grid is setup to use " << threads << " threads" << std::endl;
-  std::cout << GridLogMessage << "Grid floating point word size is REALF" << sizeof(RealF) << std::endl;
-  std::cout << GridLogMessage << "Grid floating point word size is REALD" << sizeof(RealD) << std::endl;
-  std::cout << GridLogMessage << "Grid floating point word size is REAL" << sizeof(Real) << std::endl;
-
-  std::vector<int> seeds({1, 2, 3, 4});
-  GridParallelRNG pRNG(&Grid);
-  pRNG.SeedFixedIntegers(seeds);
-  //  pRNG.SeedFixedIntegers(std::vector<int>({45,12,81,9});
-
-  typedef typename WilsonExpCloverFermionR::FermionField FermionField;
-  typename WilsonExpCloverFermionR::ImplParams params;
-  WilsonAnisotropyCoefficients anis;
-
-  FermionField src(&Grid);
-  random(pRNG, src);
-  FermionField result(&Grid);
-  result = Zero();
-  FermionField result2(&Grid);
-  result2 = Zero();
-  FermionField ref(&Grid);
-  ref = Zero();
-  FermionField tmp(&Grid);
-  tmp = Zero();
-  FermionField err(&Grid);
-  err = Zero();
-  FermionField phi(&Grid);
-  random(pRNG, phi);
-  FermionField chi(&Grid);
-  random(pRNG, chi);
-  LatticeGaugeField Umu(&Grid);
-  SU<Nc>::HotConfiguration(pRNG, Umu);
-  std::vector<LatticeColourMatrix> U(4, &Grid);
-
-  double tolerance = 1e-4;
-
-  double volume = 1;
-  for (int mu = 0; mu < Nd; mu++)
-  {
-    volume = volume * latt_size[mu];
-  }
-
-  RealD mass = 0.1;
-  RealD csw_r = 1.0;
-  RealD csw_t = 1.0;
-
-  WilsonExpCloverFermionR Dwc(Umu, Grid, RBGrid, mass, csw_r, csw_t, anis, params);
-  CompactWilsonExpCloverFermionR Dwc_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 1.0, anis, params);
-
-  std::cout << GridLogMessage << "==========================================================" << std::endl;
-  std::cout << GridLogMessage << "= Testing that Deo + Doe = Dunprec " << std::endl;
-  std::cout << GridLogMessage << "==========================================================" << std::endl;
-
-  FermionField src_e(&RBGrid);
-  FermionField src_o(&RBGrid);
-  FermionField r_e(&RBGrid);
-  FermionField r_o(&RBGrid);
-  FermionField r_eo(&Grid);
-  pickCheckerboard(Even, src_e, src);
-  pickCheckerboard(Odd, src_o, src);
-
-  Dwc.Meooe(src_e, r_o);
-  std::cout << GridLogMessage << "Applied Meo" << std::endl;
-  Dwc.Meooe(src_o, r_e);
-  std::cout << GridLogMessage << "Applied Moe" << std::endl;
-  Dwc.Dhop(src, ref, DaggerNo);
-
-  setCheckerboard(r_eo, r_o);
-  setCheckerboard(r_eo, r_e);
-
-  err = ref - r_eo;
-  std::cout << GridLogMessage << "EO norm diff\t" << norm2(err) << " (" << norm2(ref) << " - " << norm2(r_eo) << ")" << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-
-
-  Dwc_compact.Meooe(src_e, r_o);
-  std::cout << GridLogMessage << "Applied Meo" << std::endl;
-  Dwc_compact.Meooe(src_o, r_e);
-  std::cout << GridLogMessage << "Applied Moe" << std::endl;
-  Dwc_compact.Dhop(src, ref, DaggerNo);
-
-  setCheckerboard(r_eo, r_o);
-  setCheckerboard(r_eo, r_e);
-
-  err = ref - r_eo;
-  std::cout << GridLogMessage << "EO norm diff compact\t" << norm2(err) << " (" << norm2(ref) << " - " << norm2(r_eo) << ")" << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-
-  std::cout << GridLogMessage << "==============================================================" << std::endl;
-  std::cout << GridLogMessage << "= Test Ddagger is the dagger of D by requiring                " << std::endl;
-  std::cout << GridLogMessage << "=  < phi | Deo | chi > * = < chi | Deo^dag| phi>  " << std::endl;
-  std::cout << GridLogMessage << "==============================================================" << std::endl;
-
-  FermionField chi_e(&RBGrid);
-  FermionField chi_o(&RBGrid);
-
-  FermionField dchi_e(&RBGrid);
-  FermionField dchi_o(&RBGrid);
-
-  FermionField phi_e(&RBGrid);
-  FermionField phi_o(&RBGrid);
-
-  FermionField dphi_e(&RBGrid);
-  FermionField dphi_o(&RBGrid);
-
-  pickCheckerboard(Even, chi_e, chi);
-  pickCheckerboard(Odd, chi_o, chi);
-  pickCheckerboard(Even, phi_e, phi);
-  pickCheckerboard(Odd, phi_o, phi);
-
-  Dwc.Meooe(chi_e, dchi_o);
-  Dwc.Meooe(chi_o, dchi_e);
-  Dwc.MeooeDag(phi_e, dphi_o);
-  Dwc.MeooeDag(phi_o, dphi_e);
-
-  ComplexD pDce = innerProduct(phi_e, dchi_e);
-  ComplexD pDco = innerProduct(phi_o, dchi_o);
-  ComplexD cDpe = innerProduct(chi_e, dphi_e);
-  ComplexD cDpo = innerProduct(chi_o, dphi_o);
-
-  std::cout << GridLogMessage << "e " << pDce << " " << cDpe << std::endl;
-  std::cout << GridLogMessage << "o " << pDco << " " << cDpo << std::endl;
-
-  std::cout << GridLogMessage << "pDce - conj(cDpo) " << pDce - conj(cDpo) << std::endl;
-  std::cout << GridLogMessage << "pDco - conj(cDpe) " << pDco - conj(cDpe) << std::endl;
-
-  Dwc_compact.Meooe(chi_e, dchi_o);
-  Dwc_compact.Meooe(chi_o, dchi_e);
-  Dwc_compact.MeooeDag(phi_e, dphi_o);
-  Dwc_compact.MeooeDag(phi_o, dphi_e);
-
-  pDce = innerProduct(phi_e, dchi_e);
-  pDco = innerProduct(phi_o, dchi_o);
-  cDpe = innerProduct(chi_e, dphi_e);
-  cDpo = innerProduct(chi_o, dphi_o);
-
-  std::cout << GridLogMessage << "e compact " << pDce << " " << cDpe << std::endl;
-  std::cout << GridLogMessage << "o compact " << pDco << " " << cDpo << std::endl;
-
-  std::cout << GridLogMessage << "pDce - conj(cDpo) compact " << pDce - conj(cDpo) << std::endl;
-  std::cout << GridLogMessage << "pDco - conj(cDpe) compact " << pDco - conj(cDpe) << std::endl;
-
-  std::cout << GridLogMessage << "==============================================================" << std::endl;
-  std::cout << GridLogMessage << "= Test MeeInv Mee = 1   (if csw!=0)                           " << std::endl;
-  std::cout << GridLogMessage << "==============================================================" << std::endl;
-
-  pickCheckerboard(Even, chi_e, chi);
-  pickCheckerboard(Odd, chi_o, chi);
-
-  Dwc.Mooee(chi_e, src_e);
-  Dwc.MooeeInv(src_e, phi_e);
-
-  Dwc.Mooee(chi_o, src_o);
-  Dwc.MooeeInv(src_o, phi_o);
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = phi - chi;
-  std::cout << GridLogMessage << "norm diff " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  Dwc_compact.Mooee(chi_e, src_e);
-  Dwc_compact.MooeeInv(src_e, phi_e);
-
-  Dwc_compact.Mooee(chi_o, src_o);
-  Dwc_compact.MooeeInv(src_o, phi_o);
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = phi - chi;
-  std::cout << GridLogMessage << "norm diff compact " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  std::cout << GridLogMessage << "==============================================================" << std::endl;
-  std::cout << GridLogMessage << "= Test MeeDag MeeInvDag = 1    (if csw!=0)                    " << std::endl;
-  std::cout << GridLogMessage << "==============================================================" << std::endl;
-
-  pickCheckerboard(Even, chi_e, chi);
-  pickCheckerboard(Odd, chi_o, chi);
-
-  Dwc.MooeeDag(chi_e, src_e);
-  Dwc.MooeeInvDag(src_e, phi_e);
-
-  Dwc.MooeeDag(chi_o, src_o);
-  Dwc.MooeeInvDag(src_o, phi_o);
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = phi - chi;
-  std::cout << GridLogMessage << "norm diff " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  Dwc_compact.MooeeDag(chi_e, src_e);
-  Dwc_compact.MooeeInvDag(src_e, phi_e);
-
-  Dwc_compact.MooeeDag(chi_o, src_o);
-  Dwc_compact.MooeeInvDag(src_o, phi_o);
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = phi - chi;
-  std::cout << GridLogMessage << "norm diff compact " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  std::cout << GridLogMessage << "==============================================================" << std::endl;
-  std::cout << GridLogMessage << "= Test MeeInv MeeDag = 1      (if csw!=0)                     " << std::endl;
-  std::cout << GridLogMessage << "==============================================================" << std::endl;
-
-  pickCheckerboard(Even, chi_e, chi);
-  pickCheckerboard(Odd, chi_o, chi);
-
-  Dwc.MooeeDag(chi_e, src_e);
-  Dwc.MooeeInv(src_e, phi_e);
-
-  Dwc.MooeeDag(chi_o, src_o);
-  Dwc.MooeeInv(src_o, phi_o);
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = phi - chi;
-  std::cout << GridLogMessage << "norm diff " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  Dwc_compact.MooeeDag(chi_e, src_e);
-  Dwc_compact.MooeeInv(src_e, phi_e);
-
-  Dwc_compact.MooeeDag(chi_o, src_o);
-  Dwc_compact.MooeeInv(src_o, phi_o);
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = phi - chi;
-  std::cout << GridLogMessage << "norm diff compact " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  std::cout << GridLogMessage << "================================================================" << std::endl;
-  std::cout << GridLogMessage << "= Testing gauge covariance Clover term with EO preconditioning  " << std::endl;
-  std::cout << GridLogMessage << "================================================================" << std::endl;
-
-  chi = Zero();
-  phi = Zero();
-  tmp = Zero();
-  pickCheckerboard(Even, chi_e, chi);
-  pickCheckerboard(Odd, chi_o, chi);
-  pickCheckerboard(Even, phi_e, phi);
-  pickCheckerboard(Odd, phi_o, phi);
-
-  Dwc.Mooee(src_e, chi_e);
-  Dwc.Mooee(src_o, chi_o);
-  setCheckerboard(chi, chi_e);
-  setCheckerboard(chi, chi_o);
-  setCheckerboard(src, src_e);
-  setCheckerboard(src, src_o);
-
-  ////////////////////// Gauge Transformation
-  std::vector<int> seeds2({5, 6, 7, 8});
-  GridParallelRNG pRNG2(&Grid);
-  pRNG2.SeedFixedIntegers(seeds2);
-  LatticeColourMatrix Omega(&Grid);
-  LatticeColourMatrix ShiftedOmega(&Grid);
-  LatticeGaugeField U_prime(&Grid);
-  U_prime = Zero();
-  LatticeColourMatrix U_prime_mu(&Grid);
-  U_prime_mu = Zero();
-  SU<Nc>::LieRandomize(pRNG2, Omega, 1.0);
-  for (int mu = 0; mu < Nd; mu++)
-  {
-    U[mu] = peekLorentz(Umu, mu);
-    ShiftedOmega = Cshift(Omega, mu, 1);
-    U_prime_mu = Omega * U[mu] * adj(ShiftedOmega);
-    pokeLorentz(U_prime, U_prime_mu, mu);
-  }
-  /////////////////
-
-  WilsonExpCloverFermionR Dwc_prime(U_prime, Grid, RBGrid, mass, csw_r, csw_t, anis, params);
-  CompactWilsonExpCloverFermionR Dwc_compact_prime(U_prime, Grid, RBGrid, mass, csw_r, csw_t, 1.0, anis, params);
-
-  tmp = Omega * src;
-  pickCheckerboard(Even, src_e, tmp);
-  pickCheckerboard(Odd, src_o, tmp);
-
-  Dwc_prime.Mooee(src_e, phi_e);
-  Dwc_prime.Mooee(src_o, phi_o);
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = chi - adj(Omega) * phi;
-  std::cout << GridLogMessage << "norm diff " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  chi = Zero();
-  phi = Zero();
-  tmp = Zero();
-  pickCheckerboard(Even, chi_e, chi);
-  pickCheckerboard(Odd, chi_o, chi);
-  pickCheckerboard(Even, phi_e, phi);
-  pickCheckerboard(Odd, phi_o, phi);
-
-  Dwc_compact.Mooee(src_e, chi_e);
-  Dwc_compact.Mooee(src_o, chi_o);
-  setCheckerboard(chi, chi_e);
-  setCheckerboard(chi, chi_o);
-  setCheckerboard(src, src_e);
-  setCheckerboard(src, src_o);
-
-  tmp = Omega * src;
-  pickCheckerboard(Even, src_e, tmp);
-  pickCheckerboard(Odd, src_o, tmp);
-
-  Dwc_compact_prime.Mooee(src_e, phi_e);
-  Dwc_compact_prime.Mooee(src_o, phi_o);
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = chi - adj(Omega) * phi;
-  std::cout << GridLogMessage << "norm diff compact " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  std::cout << GridLogMessage << "=================================================================" << std::endl;
-  std::cout << GridLogMessage << "= Testing gauge covariance Clover term w/o EO preconditioning  " << std::endl;
-  std::cout << GridLogMessage << "================================================================" << std::endl;
-
-  chi = Zero();
-  phi = Zero();
-
-  WilsonFermionR Dw(Umu, Grid, RBGrid, mass, params);
-
-  Dw.M(src, result);
-  Dwc.M(src, chi);
-
-  Dwc_prime.M(Omega * src, phi);
-
-  WilsonFermionR Dw_prime(U_prime, Grid, RBGrid, mass, params);
-  Dw_prime.M(Omega * src, result2);
-
-  err = result - adj(Omega) * result2;
-  std::cout << GridLogMessage << "norm diff Wilson                 " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-  err = chi - adj(Omega) * phi;
-  std::cout << GridLogMessage << "norm diff WilsonExpClover        " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  chi = Zero();
-  phi = Zero();
-
-  Dwc_compact.M(src, chi);
-  Dwc_compact_prime.M(Omega * src, phi);
-
-  err = chi - adj(Omega) * phi;
-  std::cout << GridLogMessage << "norm diff CompactWilsonExpClover " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  std::cout << GridLogMessage << "==========================================================" << std::endl;
-  std::cout << GridLogMessage << "= Testing Mooee(csw=0) Clover to reproduce Mooee Wilson   " << std::endl;
-  std::cout << GridLogMessage << "==========================================================" << std::endl;
-
-  chi = Zero();
-  phi = Zero();
-  err = Zero();
-  WilsonExpCloverFermionR Dwc_csw0(Umu, Grid, RBGrid, mass, 0.0, 0.0, anis, params); //  <-- Notice: csw=0
-
-  pickCheckerboard(Even, phi_e, phi);
-  pickCheckerboard(Odd, phi_o, phi);
-  pickCheckerboard(Even, chi_e, chi);
-  pickCheckerboard(Odd, chi_o, chi);
-
-  Dw.Mooee(src_e, chi_e);
-  Dw.Mooee(src_o, chi_o);
-  Dwc_csw0.Mooee(src_e, phi_e);
-  Dwc_csw0.Mooee(src_o, phi_o);
-
-  setCheckerboard(chi, chi_e);
-  setCheckerboard(chi, chi_o);
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-  setCheckerboard(src, src_e);
-  setCheckerboard(src, src_o);
-
-  err = chi - phi;
-  std::cout << GridLogMessage << "norm diff " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  chi = Zero();
-  phi = Zero();
-  err = Zero();
-  CompactWilsonExpCloverFermionR Dwc_compact_csw0(Umu, Grid, RBGrid, mass, 0.0, 0.0, 1.0, anis, params); //  <-- Notice: csw=0
-
-  pickCheckerboard(Even, phi_e, phi);
-  pickCheckerboard(Odd, phi_o, phi);
-  pickCheckerboard(Even, chi_e, chi);
-  pickCheckerboard(Odd, chi_o, chi);
-
-  Dw.Mooee(src_e, chi_e);
-  Dw.Mooee(src_o, chi_o);
-  Dwc_compact_csw0.Mooee(src_e, phi_e);
-  Dwc_compact_csw0.Mooee(src_o, phi_o);
-
-  setCheckerboard(chi, chi_e);
-  setCheckerboard(chi, chi_o);
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-  setCheckerboard(src, src_e);
-  setCheckerboard(src, src_o);
-
-  err = chi - phi;
-  std::cout << GridLogMessage << "norm diff compact " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  std::cout << GridLogMessage << "==========================================================" << std::endl;
-  std::cout << GridLogMessage << "= Testing EO operator is equal to the unprec              " << std::endl;
-  std::cout << GridLogMessage << "==========================================================" << std::endl;
-
-  chi = Zero();
-  phi = Zero();
-  err = Zero();
-
-  pickCheckerboard(Even, phi_e, phi);
-  pickCheckerboard(Odd, phi_o, phi);
-  pickCheckerboard(Even, chi_e, chi);
-  pickCheckerboard(Odd, chi_o, chi);
-
-  // M phi = (Mooee src_e + Meooe src_o , Meooe src_e + Mooee src_o)
-
-  Dwc.M(src, ref); // Reference result from the unpreconditioned operator
-
-  // EO matrix
-  Dwc.Mooee(src_e, chi_e); 
-  Dwc.Mooee(src_o, chi_o);
-  Dwc.Meooe(src_o, phi_e);
-  Dwc.Meooe(src_e, phi_o);
-
-  phi_o += chi_o;
-  phi_e += chi_e;
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = ref - phi;
-  std::cout << GridLogMessage << "ref (unpreconditioned operator) diff         : " << norm2(ref) << std::endl;
-  std::cout << GridLogMessage << "phi (EO decomposition)          diff         : " << norm2(phi) << std::endl;
-  std::cout << GridLogMessage << "norm diff                                    : " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  chi = Zero();
-  phi = Zero();
-  err = Zero();
-
-  pickCheckerboard(Even, phi_e, phi);
-  pickCheckerboard(Odd, phi_o, phi);
-  pickCheckerboard(Even, chi_e, chi);
-  pickCheckerboard(Odd, chi_o, chi);
-
-  // M phi = (Mooee src_e + Meooe src_o , Meooe src_e + Mooee src_o)
-
-  Dwc_compact.M(src, ref); // Reference result from the unpreconditioned operator
-
-  // EO matrix
-  Dwc_compact.Mooee(src_e, chi_e); 
-  Dwc_compact.Mooee(src_o, chi_o);
-  Dwc_compact.Meooe(src_o, phi_e);
-  Dwc_compact.Meooe(src_e, phi_o);
-
-  phi_o += chi_o;
-  phi_e += chi_e;
-
-  setCheckerboard(phi, phi_e);
-  setCheckerboard(phi, phi_o);
-
-  err = ref - phi;
-  std::cout << GridLogMessage << "ref (unpreconditioned operator) diff compact : " << norm2(ref) << std::endl;
-  std::cout << GridLogMessage << "phi (EO decomposition)          diff compact : " << norm2(phi) << std::endl;
-  std::cout << GridLogMessage << "norm diff compact                            : " << norm2(err) << std::endl;
-  assert(fabs(norm2(err)) < tolerance);
-
-  Grid_finalize();
-}

tests/hmc/Test_hmc_ScalarActionNxN.cc

@@ -132,8 +132,8 @@ int main(int argc, char **argv) {
 
   // Checkpointer definition
   CheckpointerParameters CPparams(Reader);
-  TheHMC.Resources.LoadBinaryCheckpointer(CPparams);
+  //TheHMC.Resources.LoadBinaryCheckpointer(CPparams);
-  //TheHMC.Resources.LoadScidacCheckpointer(CPparams, SPar); this breaks for compilation without lime
+  TheHMC.Resources.LoadScidacCheckpointer(CPparams, SPar);
 
   RNGModuleParameters RNGpar(Reader);
   TheHMC.Resources.SetRNGSeeds(RNGpar);

tests/hmc/Test_hmc_WG_Production.cc

@@ -74,10 +74,10 @@ int main(int argc, char **argv) {
 
   // Checkpointer definition
   CheckpointerParameters CPparams(Reader);
-  TheHMC.Resources.LoadNerscCheckpointer(CPparams);
+  //TheHMC.Resources.LoadNerscCheckpointer(CPparams);
 
-  // Store metadata in the Scidac checkpointer - obviously breaks without LIME
+  // Store metadata in the Scidac checkpointer
-  //TheHMC.Resources.LoadScidacCheckpointer(CPparams, WilsonPar);
+  TheHMC.Resources.LoadScidacCheckpointer(CPparams, WilsonPar);
 
   RNGModuleParameters RNGpar(Reader);
   TheHMC.Resources.SetRNGSeeds(RNGpar);

tests/lanczos/Test_compressed_lanczos.cc

@@ -37,8 +37,6 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 using namespace std;
 using namespace Grid;
 
-#ifdef HAVE_LIME
-
 template<class Fobj,class CComplex,int nbasis>
 class LocalCoherenceLanczosScidac : public LocalCoherenceLanczos<Fobj,CComplex,nbasis>
 { 
@@ -251,11 +249,3 @@ int main (int argc, char ** argv) {
   Grid_finalize();
 }
 
-#else
-
-int main( void )
-{
-  return 0 ;
-}
-
-#endif // HAVE_LIME_H

tests/lanczos/Test_dwf_compressed_lanczos_reorg.cc

@@ -36,8 +36,7 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
 using namespace std;
 using namespace Grid;
-
+ ;
-#ifdef HAVE_LIME
 
 template<class Fobj,class CComplex,int nbasis>
 class LocalCoherenceLanczosScidac : public LocalCoherenceLanczos<Fobj,CComplex,nbasis>
@@ -251,11 +250,3 @@ int main (int argc, char ** argv) {
   Grid_finalize();
 }
 
-#else
-
-int main( void )
-{
-  return 0 ;
-}
-
-#endif

tests/solver/Test_coarse_even_odd.cc

@@ -93,16 +93,8 @@ int main(int argc, char** argv) {
   //                    Setup of Dirac Matrix and Operator                   //
   /////////////////////////////////////////////////////////////////////////////
 
-  LatticeGaugeField Umu(Grid_f);
+  LatticeGaugeField Umu(Grid_f); SU3::HotConfiguration(pRNG_f, Umu);
-#if (Nc==2)
+
-  SU2::HotConfiguration(pRNG_f, Umu);
-#elif (defined Nc==3)
-  SU3::HotConfiguration(pRNG_f, Umu);
-#elif (defined Nc==4)
-  SU4::HotConfiguration(pRNG_f, Umu);
-#elif (defined Nc==5)
-  SU5::HotConfiguration(pRNG_f, Umu);
-#endif
   RealD checkTolerance = (getPrecision<LatticeFermion>::value == 1) ? 1e-7 : 1e-15;
 
   RealD mass = -0.30;

tests/solver/Test_dwf_mrhs_cg.cc

@@ -34,7 +34,6 @@ using namespace Grid;
 
 int main (int argc, char ** argv)
 {
-#ifdef HAVE_LIME 
   typedef typename DomainWallFermionR::FermionField FermionField; 
   typedef typename DomainWallFermionR::ComplexField ComplexField; 
   typename DomainWallFermionR::ImplParams params; 
@@ -238,5 +237,4 @@ int main (int argc, char ** argv)
   }
 
   Grid_finalize();
-#endif // HAVE_LIME
 }

tests/solver/Test_wilsonclover_cg_prec.cc

@@ -71,12 +71,7 @@ int main (int argc, char ** argv)
   RealD mass = -0.1;
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
-  RealD cF = 1.0;
   WilsonCloverFermionR Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
-  CompactWilsonCloverFermionR Dw_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
-  WilsonExpCloverFermionR Dwe(Umu, Grid, RBGrid, mass, csw_r, csw_t);
-  CompactWilsonExpCloverFermionR Dwe_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
-
 
   //  HermitianOperator<WilsonFermion,LatticeFermion> HermOp(Dw);
   //  ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
@@ -85,28 +80,12 @@ int main (int argc, char ** argv)
   LatticeFermion    src_o(&RBGrid);
   LatticeFermion result_o(&RBGrid);
   pickCheckerboard(Odd,src_o,src);
+  result_o=Zero();
 
-  ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
-
-  std::cout << GridLogMessage << "Testing Wilson Clover" << std::endl;
   SchurDiagMooeeOperator<WilsonCloverFermionR,LatticeFermion> HermOpEO(Dw);
-  result_o=Zero();
+  ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
   CG(HermOpEO,src_o,result_o);
-
+  
-  std::cout << GridLogMessage << "Testing Compact Wilson Clover" << std::endl;
-  SchurDiagMooeeOperator<CompactWilsonCloverFermionR,LatticeFermion> HermOpEO_compact(Dw_compact);
-  result_o=Zero();
-  CG(HermOpEO_compact,src_o,result_o);
-
-  std::cout << GridLogMessage << "Testing Wilson Exp Clover" << std::endl;
-  SchurDiagMooeeOperator<WilsonExpCloverFermionR,LatticeFermion> HermOpEO_exp(Dwe);
-  result_o=Zero();
-  CG(HermOpEO_exp,src_o,result_o);
-
-  std::cout << GridLogMessage << "Testing Compact Wilson Exp Clover" << std::endl;
-  SchurDiagMooeeOperator<CompactWilsonExpCloverFermionR,LatticeFermion> HermOpEO_exp_compact(Dwe_compact);
-  result_o=Zero();
-  CG(HermOpEO_exp_compact,src_o,result_o);
 
   Grid_finalize();
 }

tests/solver/Test_wilsonclover_cg_schur.cc

@@ -60,36 +60,18 @@ int main (int argc, char ** argv)
   LatticeGaugeField Umu(&Grid); SU<Nc>::HotConfiguration(pRNG,Umu);
 
   LatticeFermion    src(&Grid); random(pRNG,src);
-  LatticeFermion result(&Grid);
+  LatticeFermion result(&Grid); result=Zero();
-  LatticeFermion resid(&Grid);
+  LatticeFermion resid(&Grid); 
-
-  ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
-  SchurRedBlackDiagMooeeSolve<LatticeFermion> SchurSolver(CG);
 
   RealD mass = -0.1;
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
-  RealD cF = 1.0;
-
-  std::cout << GridLogMessage << "Testing Wilson Clover" << std::endl;
   WilsonCloverFermionR Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
-  result=Zero();
+
+  ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
+  SchurRedBlackDiagMooeeSolve<LatticeFermion> SchurSolver(CG);
+
   SchurSolver(Dw,src,result);
-
-  std::cout << GridLogMessage << "Testing Compact Wilson Clover" << std::endl;
-  CompactWilsonCloverFermionR Dw_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
-  result=Zero();
-  SchurSolver(Dw_compact,src,result);
-
-  std::cout << GridLogMessage << "Testing Wilson Exp Clover" << std::endl;
-  WilsonExpCloverFermionR Dwe(Umu, Grid, RBGrid, mass, csw_r, csw_t);
-  result=Zero();
-  SchurSolver(Dwe,src,result);
-
-  std::cout << GridLogMessage << "Testing Compact Wilson Exp Clover" << std::endl;
-  CompactWilsonExpCloverFermionR Dwe_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
-  result=Zero();
-  SchurSolver(Dwe_compact,src,result);
   
   Grid_finalize();
 }

tests/solver/Test_wilsonclover_cg_unprec.cc

@@ -59,7 +59,7 @@ int main (int argc, char ** argv)
 
   LatticeFermion src(&Grid); random(pRNG,src);
   RealD nrm = norm2(src);
-  LatticeFermion result(&Grid);
+  LatticeFermion result(&Grid); result=Zero();
   LatticeGaugeField Umu(&Grid); SU<Nc>::HotConfiguration(pRNG,Umu);
 
   double volume=1;
@@ -70,34 +70,11 @@ int main (int argc, char ** argv)
   RealD mass = -0.1;
   RealD csw_r = 1.0;
   RealD csw_t = 1.0;
-  RealD cF = 1.0;
   WilsonCloverFermionR Dw(Umu, Grid, RBGrid, mass, csw_r, csw_t);
-  CompactWilsonCloverFermionR Dw_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
-  WilsonExpCloverFermionR Dwe(Umu, Grid, RBGrid, mass, csw_r, csw_t);
-  CompactWilsonExpCloverFermionR Dwe_compact(Umu, Grid, RBGrid, mass, csw_r, csw_t, 0.0);
 
+  MdagMLinearOperator<WilsonFermionR,LatticeFermion> HermOp(Dw);
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
-
-  std::cout << GridLogMessage << "Testing Wilson Clover" << std::endl;
-  MdagMLinearOperator<WilsonCloverFermionR,LatticeFermion> HermOp(Dw);
-  result=Zero();
   CG(HermOp,src,result);
 
-  std::cout << GridLogMessage << "Testing Compact Wilson Clover" << std::endl;
-  MdagMLinearOperator<CompactWilsonCloverFermionR,LatticeFermion> HermOp_compact(Dw_compact);
-  result=Zero();
-  CG(HermOp_compact,src,result);
-
-
-  std::cout << GridLogMessage << "Testing Wilson Exp Clover" << std::endl;
-  MdagMLinearOperator<WilsonExpCloverFermionR,LatticeFermion> HermOp_exp(Dwe);
-  result=Zero();
-  CG(HermOp_exp,src,result);
-
-  std::cout << GridLogMessage << "Testing Compact Wilson Exp Clover" << std::endl;
-  MdagMLinearOperator<CompactWilsonExpCloverFermionR,LatticeFermion> HermOp_exp_compact(Dwe_compact);
-  result=Zero();
-  CG(HermOp_exp_compact,src,result);
-
   Grid_finalize();
 }