Merge branch 'develop' into feature/feynman-rules

# Conflicts:
#	lib/Threads.h
#	lib/qcd/action/fermion/WilsonFermion.cc
#	lib/qcd/action/fermion/WilsonFermion.h
#	lib/qcd/utils/SUn.h
#	lib/simd/Grid_avx.h
#	lib/simd/Intel512common.h

.gitignore

@@ -94,6 +94,10 @@ build.sh
 ################
 lib/Eigen/*
 
+# FFTW source #
+################
+lib/fftw/*
+
 # libtool macros #
 ##################
 m4/lt*

.travis.yml

@@ -9,10 +9,6 @@ matrix:
     - os:        osx
       osx_image: xcode7.2
       compiler: clang
-    - os:        osx
-      osx_image: xcode7.2
-      compiler: gcc
-      env: VERSION=-5
     - compiler: gcc
       addons:
         apt:
@@ -107,3 +103,4 @@ script:
     - make -j4
     - if [[ "$TRAVIS_OS_NAME" == "linux" ]]; then mpirun.openmpi -n 2 ./benchmarks/Benchmark_dwf --threads 1 --mpi 2.1.1.1; fi
     - if [[ "$TRAVIS_OS_NAME" == "osx" ]]; then mpirun -n 2 ./benchmarks/Benchmark_dwf --threads 1 --mpi 2.1.1.1; fi
+

benchmarks/Benchmark_dwf.cc

@@ -86,18 +86,6 @@ int main (int argc, char ** argv)
   LatticeFermion    tmp(FGrid);
   LatticeFermion    err(FGrid);
 
-  /*  src=zero;
-  std::vector<int> origin(5,0);
-  SpinColourVector f=zero;
-  for(int sp=0;sp<4;sp++){
-  for(int co=0;co<3;co++){
-    f()(sp)(co)=Complex(1.0,0.0); 
-  }}
-  pokeSite(f,src,origin);
-  */
-
-  ColourMatrix cm = Complex(1.0,0.0);
-
   LatticeGaugeField Umu(UGrid); 
   random(RNG4,Umu);
 
@@ -144,10 +132,12 @@ int main (int argc, char ** argv)
 
   DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
   
-  std::cout<<GridLogMessage << "Calling Dw"<<std::endl;
+  std::cout<<GridLogMessage << "Naive wilson implementation "<<std::endl;
+  std::cout << GridLogMessage<< "Calling Dw"<<std::endl;
   int ncall =100;
   if (1) {
 
+    Dw.ZeroCounters();
     double t0=usecond();
     for(int i=0;i<ncall;i++){
       __SSC_START;
@@ -166,7 +156,7 @@ int main (int argc, char ** argv)
     std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NP<<std::endl;
     err = ref-result; 
     std::cout<<GridLogMessage << "norm diff   "<< norm2(err)<<std::endl;
-    //    Dw.Report();
+    Dw.Report();
   }
 
   if (1)
@@ -188,8 +178,9 @@ int main (int argc, char ** argv)
       peekSite(tmp,src,site);
       pokeSite(tmp,ssrc,site);
     }}}}}
-    std::cout<<"src norms "<< norm2(src)<<" " <<norm2(ssrc)<<std::endl;
+    std::cout<<GridLogMessage<< "src norms "<< norm2(src)<<" " <<norm2(ssrc)<<std::endl;
     double t0=usecond();
+    sDw.ZeroCounters();
     for(int i=0;i<ncall;i++){
       __SSC_START;
       sDw.Dhop(ssrc,sresult,0);
@@ -199,23 +190,23 @@ int main (int argc, char ** argv)
     double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
     double flops=1344*volume*ncall;
 
-    std::cout<<GridLogMessage << "Called Dw sinner "<<ncall<<" times in "<<t1-t0<<" us"<<std::endl;
+    std::cout<<GridLogMessage << "Called Dw s_inner "<<ncall<<" times in "<<t1-t0<<" us"<<std::endl;
     std::cout<<GridLogMessage << "mflop/s =   "<< flops/(t1-t0)<<std::endl;
     std::cout<<GridLogMessage << "mflop/s per node =  "<< flops/(t1-t0)/NP<<std::endl;
-    //  sDw.Report();
+    sDw.Report();
   
     if(0){
       for(int i=0;i< PerformanceCounter::NumTypes(); i++ ){
-	sDw.Dhop(ssrc,sresult,0);
+  sDw.Dhop(ssrc,sresult,0);
-	PerformanceCounter Counter(i);
+  PerformanceCounter Counter(i);
-	Counter.Start();
+  Counter.Start();
-	sDw.Dhop(ssrc,sresult,0);
+  sDw.Dhop(ssrc,sresult,0);
-	Counter.Stop();
+  Counter.Stop();
-	Counter.Report();
+  Counter.Report();
       }
     }
 
-    std::cout<<"res norms "<< norm2(result)<<" " <<norm2(sresult)<<std::endl;
+    std::cout<<GridLogMessage<< "res norms "<< norm2(result)<<" " <<norm2(sresult)<<std::endl;
 
 
     RealF sum=0;
@@ -230,12 +221,12 @@ int main (int argc, char ** argv)
       peekSite(simd,sresult,site);
       sum=sum+norm2(normal-simd);
       if (norm2(normal-simd) > 1.0e-6 ) {
-	std::cout << "site "<<x<<","<<y<<","<<z<<","<<t<<","<<s<<" "<<norm2(normal-simd)<<std::endl;
+  std::cout << "site "<<x<<","<<y<<","<<z<<","<<t<<","<<s<<" "<<norm2(normal-simd)<<std::endl;
-	std::cout << "site "<<x<<","<<y<<","<<z<<","<<t<<","<<s<<" normal "<<normal<<std::endl;
+  std::cout << "site "<<x<<","<<y<<","<<z<<","<<t<<","<<s<<" normal "<<normal<<std::endl;
-	std::cout << "site "<<x<<","<<y<<","<<z<<","<<t<<","<<s<<" simd   "<<simd<<std::endl;
+  std::cout << "site "<<x<<","<<y<<","<<z<<","<<t<<","<<s<<" simd   "<<simd<<std::endl;
       }
     }}}}}
-    std::cout<<" difference between normal and simd is "<<sum<<std::endl;
+    std::cout<<GridLogMessage<<" difference between normal and simd is "<<sum<<std::endl;
 
 
     if (1) {
@@ -259,17 +250,21 @@ int main (int argc, char ** argv)
       sr_e = zero;
       sr_o = zero;
 
+      sDw.ZeroCounters();
+      sDw.stat.init("DhopEO");
       double t0=usecond();
-      for(int i=0;i<ncall;i++){
+      for (int i = 0; i < ncall; i++) {
-	sDw.DhopEO(ssrc_o,sr_e,DaggerNo);
+        sDw.DhopEO(ssrc_o, sr_e, DaggerNo);
       }
       double t1=usecond();
+      sDw.stat.print();
 
       double volume=Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
       double flops=(1344.0*volume*ncall)/2;
 
       std::cout<<GridLogMessage << "sDeo mflop/s =   "<< flops/(t1-t0)<<std::endl;
       std::cout<<GridLogMessage << "sDeo mflop/s per node   "<< flops/(t1-t0)/NP<<std::endl;
+      sDw.Report();
 
       sDw.DhopEO(ssrc_o,sr_e,DaggerNo);
       sDw.DhopOE(ssrc_e,sr_o,DaggerNo);
@@ -294,18 +289,19 @@ int main (int argc, char ** argv)
       //    ref =  src - Gamma(Gamma::GammaX)* src ; // 1+gamma_x
       tmp = U[mu]*Cshift(src,mu+1,1);
       for(int i=0;i<ref._odata.size();i++){
-	ref._odata[i]+= tmp._odata[i] + Gamma(Gmu[mu])*tmp._odata[i]; ;
+  ref._odata[i]+= tmp._odata[i] + Gamma(Gmu[mu])*tmp._odata[i]; ;
       }
 
       tmp =adj(U[mu])*src;
       tmp =Cshift(tmp,mu+1,-1);
       for(int i=0;i<ref._odata.size();i++){
-	ref._odata[i]+= tmp._odata[i] - Gamma(Gmu[mu])*tmp._odata[i]; ;
+  ref._odata[i]+= tmp._odata[i] - Gamma(Gmu[mu])*tmp._odata[i]; ;
       }
     }
     ref = -0.5*ref;
   }
   Dw.Dhop(src,result,1);
+  std::cout << GridLogMessage << "Naive wilson implementation Dag" << std::endl;
   std::cout<<GridLogMessage << "Called DwDag"<<std::endl;
   std::cout<<GridLogMessage << "norm result "<< norm2(result)<<std::endl;
   std::cout<<GridLogMessage << "norm ref    "<< norm2(ref)<<std::endl;
@@ -327,6 +323,7 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "src_o"<<norm2(src_o)<<std::endl;
 
   {
+    Dw.ZeroCounters();
     double t0=usecond();
     for(int i=0;i<ncall;i++){
       Dw.DhopEO(src_o,r_e,DaggerNo);
@@ -338,6 +335,7 @@ int main (int argc, char ** argv)
 
     std::cout<<GridLogMessage << "Deo mflop/s =   "<< flops/(t1-t0)<<std::endl;
     std::cout<<GridLogMessage << "Deo mflop/s per node   "<< flops/(t1-t0)/NP<<std::endl;
+    Dw.Report();
   }
   Dw.DhopEO(src_o,r_e,DaggerNo);
   Dw.DhopOE(src_e,r_o,DaggerNo);

benchmarks/Benchmark_dwf_sweep.cc

@@ -51,7 +51,7 @@ int main (int argc, char ** argv)
 {
   Grid_init(&argc,&argv);
 
-  const int Ls=16;
+  const int Ls=8;
   int threads = GridThread::GetThreads();
   std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
 

bootstrap.sh

@@ -4,7 +4,7 @@ EIGEN_URL='http://bitbucket.org/eigen/eigen/get/3.2.9.tar.bz2'
 FFTW_URL=http://www.fftw.org/fftw-3.3.4.tar.gz
 
 echo "-- deploying Eigen source..."
-wget ${EIGEN_URL}
+wget ${EIGEN_URL} --no-check-certificate
 ./scripts/update_eigen.sh `basename ${EIGEN_URL}`
 rm `basename ${EIGEN_URL}`
 

configure.ac

@@ -1,15 +1,20 @@
 AC_PREREQ([2.63])
 AC_INIT([Grid], [0.5.1-dev], [https://github.com/paboyle/Grid], [Grid])
+AC_CANONICAL_BUILD
+AC_CANONICAL_HOST
+AC_CANONICAL_TARGET
 AM_INIT_AUTOMAKE(subdir-objects)
 AC_CONFIG_MACRO_DIR([m4])
 AC_CONFIG_SRCDIR([lib/Grid.h])
 AC_CONFIG_HEADERS([lib/Config.h])
 m4_ifdef([AM_SILENT_RULES], [AM_SILENT_RULES([yes])])
 
+
 ############### Checks for programs
 AC_LANG(C++)
 CXXFLAGS="-O3 $CXXFLAGS"
 AC_PROG_CXX
+AC_PROG_RANLIB
 
 ############ openmp  ###############
 AC_OPENMP
@@ -22,9 +27,6 @@ AM_CXXFLAGS="$OPENMP_CXXFLAGS $AM_CXXFLAGS"
 AM_LDFLAGS="$OPENMP_CXXFLAGS $AM_LDFLAGS"
 fi
 
-############ libtool ###############
-LT_INIT
-
 ############### Checks for header files
 AC_CHECK_HEADERS(stdint.h)
 AC_CHECK_HEADERS(mm_malloc.h)
@@ -45,7 +47,7 @@ AC_ARG_WITH([gmp],
     [AS_HELP_STRING([--with-gmp=prefix],
     [try this for a non-standard install prefix of the GMP library])],
     [AM_CXXFLAGS="-I$with_gmp/include $AM_CXXFLAGS"]
-    [AM_LDFLAGS="-L$with_gmp/lib" $AM_LDFLAGS])
+    [AM_LDFLAGS="-L$with_gmp/lib $AM_LDFLAGS"])
 AC_ARG_WITH([mpfr],
     [AS_HELP_STRING([--with-mpfr=prefix],
     [try this for a non-standard install prefix of the MPFR library])],
@@ -68,6 +70,20 @@ case ${ac_LAPACK} in
         AC_DEFINE([USE_LAPACK],[1],[use LAPACK])
 esac
 
+################## first-touch ####################
+AC_ARG_ENABLE([numa],
+    [AC_HELP_STRING([--enable-numa=yes|no|prefix], [enable first touch numa opt])], 
+    [ac_NUMA=${enable_NUMA}],[ac_NUMA=no])
+
+case ${ac_NUMA} in
+    no)
+        ;;
+    yes)
+        AC_DEFINE([GRID_NUMA],[1],[First touch numa locality]);;
+    *)
+        AC_DEFINE([GRID_NUMA],[1],[First touch numa locality]);;
+esac
+
 ################## FFTW3 ####################
 AC_ARG_WITH([fftw],    
             [AS_HELP_STRING([--with-fftw=prefix],
@@ -75,23 +91,6 @@ AC_ARG_WITH([fftw],
             [AM_CXXFLAGS="-I$with_fftw/include $AM_CXXFLAGS"]
             [AM_LDFLAGS="-L$with_fftw/lib $AM_LDFLAGS"])
 
-#
-# What about the MKL library replacement for fftw3 ?  How do we know if fftw_execute
-# can be found in MKL? 
-#
-AC_CHECK_LIB([fftw3],[fftw_execute],
-	[AC_DEFINE([HAVE_FFTW],[1],[Define to 1 if you have the `FFTW' library (-lfftw3).])] [ac_fftw=yes],
-        [ac_fftw=no])
-
-case ${ac_fftw} in
-    no)
-        echo WARNING libfftw3 not found FFT routines will not work
-        ;;
-    yes)
-        AM_LDFLAGS="$AM_LDFLAGS -lfftw3 -lfftw3f"
-esac
-
-
 ################ Get compiler informations
 AC_LANG([C++])
 AX_CXX_COMPILE_STDCXX_11([noext],[mandatory])
@@ -105,7 +104,6 @@ AC_DEFINE_UNQUOTED([GXX_VERSION],["$GXX_VERSION"],
 ############### Checks for library functions
 CXXFLAGS_CPY=$CXXFLAGS
 LDFLAGS_CPY=$LDFLAGS
-LIBS_CPY=$LIBS
 CXXFLAGS="$AM_CXXFLAGS $CXXFLAGS"
 LDFLAGS="$AM_LDFLAGS $LDFLAGS"
 AC_CHECK_FUNCS([gettimeofday])
@@ -124,11 +122,16 @@ if test "${ac_LAPACK}x" != "nox"; then
     AC_CHECK_LIB([lapack],[LAPACKE_sbdsdc],[],
                  [AC_MSG_ERROR("LAPACK enabled but library not found")])
 fi
+AC_CHECK_LIB([fftw3],[fftw_execute],
+  [AC_DEFINE([HAVE_FFTW],[1],[Define to 1 if you have the `FFTW' library (-lfftw3).])]
+  [have_fftw=true]
+  [LIBS="$LIBS -lfftw3 -lfftw3f"],
+  [AC_MSG_WARN([**** FFTW library not found, Grid can still compile but FFT-based routines will not work ****])])
 CXXFLAGS=$CXXFLAGS_CPY
 LDFLAGS=$LDFLAGS_CPY
 
 ############### SIMD instruction selection
-AC_ARG_ENABLE([simd],[AC_HELP_STRING([--enable-simd=SSE4|AVX|AVXFMA4|AVX2|AVX512|AVX512MIC|IMCI|KNL|KNC],\
+AC_ARG_ENABLE([simd],[AC_HELP_STRING([--enable-simd=SSE4|AVX|AVXFMA4|AVXFMA|AVX2|AVX512|AVX512MIC|IMCI|KNL|KNC],\
 	[Select instructions to be SSE4.0, AVX 1.0, AVX 2.0+FMA, AVX 512, IMCI])],\
 	[ac_SIMD=${enable_simd}],[ac_SIMD=GEN])
 
@@ -144,6 +147,9 @@ case ${ax_cv_cxx_compiler_vendor} in
       AVXFMA4)
         AC_DEFINE([AVXFMA4],[1],[AVX intrinsics with FMA4])
         SIMD_FLAGS='-mavx -mfma4';;
+      AVXFMA)
+        AC_DEFINE([AVXFMA],[1],[AVX intrinsics with FMA3])
+        SIMD_FLAGS='-mavx -mfma';;
       AVX2)
         AC_DEFINE([AVX2],[1],[AVX2 intrinsics])
         SIMD_FLAGS='-mavx2 -mfma';;
@@ -151,11 +157,14 @@ case ${ax_cv_cxx_compiler_vendor} in
         AC_DEFINE([AVX512],[1],[AVX512 intrinsics])
         SIMD_FLAGS='-mavx512f -mavx512pf -mavx512er -mavx512cd';;
       IMCI|KNC)
-        AC_DEFINE([IMCI],[1],[IMCI Intrinsics for Knights Corner])
+        AC_DEFINE([IMCI],[1],[IMCI intrinsics for Knights Corner])
         SIMD_FLAGS='';;
       GEN)
         AC_DEFINE([GENERIC_VEC],[1],[generic vector code])
         SIMD_FLAGS='';;
+      QPX|BGQ)
+        AC_DEFINE([QPX],[1],[QPX intrinsics for BG/Q])
+        SIMD_FLAGS='';;
       *)
         AC_MSG_ERROR(["SIMD option ${ac_SIMD} not supported by the GCC/Clang compiler"]);;
     esac;;
@@ -169,7 +178,10 @@ case ${ax_cv_cxx_compiler_vendor} in
         SIMD_FLAGS='-mavx -xavx';;
       AVXFMA4)
         AC_DEFINE([AVXFMA4],[1],[AVX intrinsics with FMA4])
-        SIMD_FLAGS='-mavx -xavx -mfma';;
+        SIMD_FLAGS='-mavx -mfma';;
+      AVXFMA)
+        AC_DEFINE([AVXFMA],[1],[AVX intrinsics with FMA4])
+        SIMD_FLAGS='-mavx -mfma';;
       AVX2)
         AC_DEFINE([AVX2],[1],[AVX2 intrinsics])
         SIMD_FLAGS='-march=core-avx2 -xcore-avx2';;
@@ -336,12 +348,15 @@ Summary of configuration for $PACKAGE v$VERSION
 - RNG choice                    : ${ac_RNG} 
 - GMP                           : `if test "x$have_gmp" = xtrue; then echo yes; else echo no; fi`
 - LAPACK                        : ${ac_LAPACK}
-- FFTW                          : ${ac_fftw}
+- FFTW                          : `if test "x$have_fftw" = xtrue; then echo yes; else echo no; fi`
 - build DOXYGEN documentation   : `if test "x$enable_doc" = xyes; then echo yes; else echo no; fi`
 - graphs and diagrams           : `if test "x$enable_dot" = xyes; then echo yes; else echo no; fi`
 ----- BUILD FLAGS -------------------------------------
-- CXXFLAGS:  "${AM_CXXFLAGS} ${CXXFLAGS}"
+- CXXFLAGS:
-- LDFLAGS:   "${AM_LDFLAGS} ${LDFLAGS}"
+`echo ${AM_CXXFLAGS} ${CXXFLAGS} | tr ' ' '\n' | sed 's/^-/    -/g'`
-- LIBS:      "${LIBS} "
+- LDFLAGS:
+`echo ${AM_LDFLAGS} ${LDFLAGS} | tr ' ' '\n' | sed 's/^-/    -/g'`
+- LIBS:
+`echo ${LIBS} | tr ' ' '\n' | sed 's/^-/    -/g'`
 -------------------------------------------------------
 "

lib/AlignedAllocator.h

@@ -113,9 +113,8 @@ public:
 
 #endif
     _Tp tmp;
-#undef FIRST_TOUCH_OPTIMISE
+#ifdef GRID_NUMA
-#ifdef FIRST_TOUCH_OPTIMISE
+#pragma omp parallel for schedule(static)
-#pragma omp parallel for 
   for(int i=0;i<__n;i++){
     ptr[i]=tmp;
   }

lib/Init.cc

@@ -270,15 +270,15 @@ void Grid_init(int *argc,char ***argv)
   std::cout <<std::endl;
   std::cout <<COL_RED  << "__|__|__|__|__"<<             "|__|__|_"<<COL_PURPLE<<"_|__|__|"<<                "__|__|__|__|__"<<std::endl; 
   std::cout <<COL_RED  << "__|__|__|__|__"<<             "|__|__|_"<<COL_PURPLE<<"_|__|__|"<<                "__|__|__|__|__"<<std::endl; 
-  std::cout <<COL_RED  << "__|__|  |  |  "<<             "|  |  | "<<COL_PURPLE<<" |  |  |"<<                "  |  |  | _|__"<<std::endl; 
+  std::cout <<COL_RED  << "__|_ |  |  |  "<<             "|  |  | "<<COL_PURPLE<<" |  |  |"<<                "  |  |  | _|__"<<std::endl; 
-  std::cout <<COL_RED  << "__|__         "<<             "        "<<COL_PURPLE<<"        "<<                "          _|__"<<std::endl; 
+  std::cout <<COL_RED  << "__|_          "<<             "        "<<COL_PURPLE<<"        "<<                "          _|__"<<std::endl; 
   std::cout <<COL_RED  << "__|_  "<<COL_GREEN<<" GGGG   "<<COL_RED<<" RRRR   "<<COL_BLUE  <<" III    "<<COL_PURPLE<<"DDDD  "<<COL_PURPLE<<"    _|__"<<std::endl;
   std::cout <<COL_RED  << "__|_  "<<COL_GREEN<<"G       "<<COL_RED<<" R   R  "<<COL_BLUE  <<"  I     "<<COL_PURPLE<<"D   D "<<COL_PURPLE<<"    _|__"<<std::endl;
   std::cout <<COL_RED  << "__|_  "<<COL_GREEN<<"G       "<<COL_RED<<" R   R  "<<COL_BLUE  <<"  I     "<<COL_PURPLE<<"D    D"<<COL_PURPLE<<"    _|__"<<std::endl;
   std::cout <<COL_BLUE << "__|_  "<<COL_GREEN<<"G  GG   "<<COL_RED<<" RRRR   "<<COL_BLUE  <<"  I     "<<COL_PURPLE<<"D    D"<<COL_GREEN <<"    _|__"<<std::endl;
   std::cout <<COL_BLUE << "__|_  "<<COL_GREEN<<"G   G   "<<COL_RED<<" R  R   "<<COL_BLUE  <<"  I     "<<COL_PURPLE<<"D   D "<<COL_GREEN <<"    _|__"<<std::endl;
   std::cout <<COL_BLUE << "__|_  "<<COL_GREEN<<" GGGG   "<<COL_RED<<" R   R  "<<COL_BLUE  <<" III    "<<COL_PURPLE<<"DDDD  "<<COL_GREEN <<"    _|__"<<std::endl;
-  std::cout <<COL_BLUE << "__|__         "<<             "        "<<COL_GREEN <<"        "<<                "          _|__"<<std::endl; 
+  std::cout <<COL_BLUE << "__|_          "<<             "        "<<COL_GREEN <<"        "<<                "          _|__"<<std::endl; 
   std::cout <<COL_BLUE << "__|__|__|__|__"<<             "|__|__|_"<<COL_GREEN <<"_|__|__|"<<                "__|__|__|__|__"<<std::endl; 
   std::cout <<COL_BLUE << "__|__|__|__|__"<<             "|__|__|_"<<COL_GREEN <<"_|__|__|"<<                "__|__|__|__|__"<<std::endl; 
   std::cout <<COL_BLUE << "  |  |  |  |  "<<             "|  |  | "<<COL_GREEN <<" |  |  |"<<                "  |  |  |  |  "<<std::endl; 

lib/Makefile.am

@@ -17,8 +17,8 @@ endif
 include Make.inc
 include Eigen.inc
 
-lib_LTLIBRARIES = libGrid.la
+lib_LIBRARIES = libGrid.a
 
-libGrid_la_SOURCES             = $(CCFILES) $(extra_sources)
+libGrid_a_SOURCES              = $(CCFILES) $(extra_sources)
-libGrid_ladir                  = $(pkgincludedir)
+libGrid_adir                   = $(pkgincludedir)
 nobase_dist_pkginclude_HEADERS = $(HFILES) $(eigen_files) Config.h

lib/Stat.cc

@@ -0,0 +1,247 @@
+#include <Grid.h>
+#include <PerfCount.h>
+#include <Stat.h>
+
+
+namespace Grid { 
+
+
+bool PmuStat::pmu_initialized=false;
+
+
+void PmuStat::init(const char *regname)
+{
+#ifdef __x86_64__
+  name = regname;
+  if (!pmu_initialized)
+    {
+      std::cout<<"initialising pmu"<<std::endl;
+      pmu_initialized = true;
+      pmu_init();
+    }
+  clear();
+#endif
+}
+void PmuStat::clear(void)
+{
+#ifdef __x86_64__
+  count = 0;
+  tregion = 0;
+  pmc0 = 0;
+  pmc1 = 0;
+  inst = 0;
+  cyc = 0;
+  ref = 0;
+  tcycles = 0;
+  reads = 0;
+  writes = 0;
+#endif
+}
+void PmuStat::print(void)
+{
+#ifdef __x86_64__
+  std::cout <<"Reg "<<std::string(name)<<":\n";
+  std::cout <<"  region "<<tregion<<std::endl;
+  std::cout <<"  cycles "<<tcycles<<std::endl;
+  std::cout <<"  inst   "<<inst   <<std::endl;
+  std::cout <<"  cyc    "<<cyc    <<std::endl;
+  std::cout <<"  ref    "<<ref    <<std::endl;
+  std::cout <<"  pmc0   "<<pmc0   <<std::endl;
+  std::cout <<"  pmc1   "<<pmc1   <<std::endl;
+  std::cout <<"  count  "<<count  <<std::endl;
+  std::cout <<"  reads  "<<reads  <<std::endl;
+  std::cout <<"  writes "<<writes <<std::endl;
+#endif
+}
+void PmuStat::start(void)
+{
+#ifdef __x86_64__
+  pmu_start();
+  ++count;
+  xmemctrs(&mrstart, &mwstart);
+  tstart = __rdtsc();
+#endif
+}
+void PmuStat::enter(int t)
+{
+#ifdef __x86_64__
+  counters[0][t] = __rdpmc(0);
+  counters[1][t] = __rdpmc(1);
+  counters[2][t] = __rdpmc((1<<30)|0);
+  counters[3][t] = __rdpmc((1<<30)|1);
+  counters[4][t] = __rdpmc((1<<30)|2);
+  counters[5][t] = __rdtsc();
+#endif
+}
+void PmuStat::exit(int t)
+{
+#ifdef __x86_64__
+  counters[0][t] = __rdpmc(0) - counters[0][t];
+  counters[1][t] = __rdpmc(1) - counters[1][t];
+  counters[2][t] = __rdpmc((1<<30)|0) - counters[2][t];
+  counters[3][t] = __rdpmc((1<<30)|1) - counters[3][t];
+  counters[4][t] = __rdpmc((1<<30)|2) - counters[4][t];
+  counters[5][t] = __rdtsc() - counters[5][t];
+#endif
+}
+void PmuStat::accum(int nthreads)
+{
+#ifdef __x86_64__
+  tend = __rdtsc();
+  xmemctrs(&mrend, &mwend);
+  pmu_stop();
+  for (int t = 0; t < nthreads; ++t) {
+    pmc0 += counters[0][t];
+    pmc1 += counters[1][t];
+    inst += counters[2][t];
+    cyc += counters[3][t];
+    ref += counters[4][t];
+    tcycles += counters[5][t];
+  }
+  uint64_t region = tend - tstart;
+  tregion += region;
+  uint64_t mreads = mrend - mrstart;
+  reads += mreads;
+  uint64_t mwrites = mwend - mwstart;
+  writes += mwrites;
+#endif
+}
+
+
+void PmuStat::pmu_fini(void) {}
+void PmuStat::pmu_start(void) {};
+void PmuStat::pmu_stop(void) {};
+void PmuStat::pmu_init(void)
+{
+#ifdef _KNIGHTS_LANDING_
+  KNLsetup();
+#endif
+}
+void PmuStat::xmemctrs(uint64_t *mr, uint64_t *mw)
+{
+#ifdef _KNIGHTS_LANDING_
+  ctrs c;
+  KNLreadctrs(c);
+  uint64_t emr = 0, emw = 0;
+  for (int i = 0; i < NEDC; ++i)
+    {
+      emr += c.edcrd[i];
+      emw += c.edcwr[i];
+    }
+  *mr = emr;
+  *mw = emw;
+#else
+  *mr = *mw = 0;
+#endif
+}
+
+#ifdef _KNIGHTS_LANDING_
+
+struct knl_gbl_ PmuStat::gbl;
+
+#define PMU_MEM
+
+void PmuStat::KNLevsetup(const char *ename, int &fd, int event, int umask)
+{
+  char fname[1024];
+  snprintf(fname, sizeof(fname), "%s/type", ename);
+  FILE *fp = fopen(fname, "r");
+  if (fp == 0) {
+    ::printf("open %s", fname);
+    ::exit(0);
+  }
+  int type;
+  int ret = fscanf(fp, "%d", &type);
+  assert(ret == 1);
+  fclose(fp);
+  //  std::cout << "Using PMU type "<<type<<" from " << std::string(ename) <<std::endl;
+
+  struct perf_event_attr hw = {};
+  hw.size = sizeof(hw);
+  hw.type = type;
+  // see /sys/devices/uncore_*/format/*
+  // All of the events we are interested in are configured the same way, but
+  // that isn't always true. Proper code would parse the format files
+  hw.config = event | (umask << 8);
+  //hw.read_format = PERF_FORMAT_GROUP;
+  // unfortunately the above only works within a single PMU; might
+  // as well just read them one at a time
+  int cpu = 0;
+  fd = perf_event_open(&hw, -1, cpu, -1, 0);
+  if (fd == -1) {
+    ::printf("CPU %d, box %s, event 0x%lx", cpu, ename, hw.config);
+    ::exit(0);
+  } else { 
+    //    std::cout << "event "<<std::string(ename)<<" set up for fd "<<fd<<" hw.config "<<hw.config <<std::endl;
+  }
+}
+
+
+ void PmuStat::KNLsetup(void){
+
+   int ret;
+   char fname[1024];
+
+   // MC RPQ inserts and WPQ inserts (reads & writes)
+   for (int mc = 0; mc < NMC; ++mc)
+     {
+       ::snprintf(fname, sizeof(fname), "/sys/devices/uncore_imc_%d",mc);
+       // RPQ Inserts
+       KNLevsetup(fname, gbl.mc_rd[mc], 0x1, 0x1);
+       // WPQ Inserts
+       KNLevsetup(fname, gbl.mc_wr[mc], 0x2, 0x1);
+     }
+   // EDC RPQ inserts and WPQ inserts
+   for (int edc=0; edc < NEDC; ++edc)
+     {
+       ::snprintf(fname, sizeof(fname), "/sys/devices/uncore_edc_eclk_%d",edc);
+       // RPQ inserts
+       KNLevsetup(fname, gbl.edc_rd[edc], 0x1, 0x1);
+       // WPQ inserts
+       KNLevsetup(fname, gbl.edc_wr[edc], 0x2, 0x1);
+     }
+   // EDC HitE, HitM, MissE, MissM
+   for (int edc=0; edc < NEDC; ++edc)
+     {
+       ::snprintf(fname, sizeof(fname), "/sys/devices/uncore_edc_uclk_%d", edc);
+       KNLevsetup(fname, gbl.edc_hite[edc], 0x2, 0x1);
+       KNLevsetup(fname, gbl.edc_hitm[edc], 0x2, 0x2);
+       KNLevsetup(fname, gbl.edc_misse[edc], 0x2, 0x4);
+       KNLevsetup(fname, gbl.edc_missm[edc], 0x2, 0x8);
+     }
+ }
+
+uint64_t PmuStat::KNLreadctr(int fd)
+{
+  uint64_t data;
+  size_t s = ::read(fd, &data, sizeof(data));
+  if (s != sizeof(uint64_t)){
+    ::printf("read counter %lu", s);
+    ::exit(0);
+  }
+  return data;
+}
+
+void PmuStat::KNLreadctrs(ctrs &c)
+{
+  for (int i = 0; i < NMC; ++i)
+    {
+      c.mcrd[i] = KNLreadctr(gbl.mc_rd[i]);
+      c.mcwr[i] = KNLreadctr(gbl.mc_wr[i]);
+    }
+  for (int i = 0; i < NEDC; ++i)
+    {
+      c.edcrd[i] = KNLreadctr(gbl.edc_rd[i]);
+      c.edcwr[i] = KNLreadctr(gbl.edc_wr[i]);
+    }
+  for (int i = 0; i < NEDC; ++i)
+    {
+      c.edchite[i] = KNLreadctr(gbl.edc_hite[i]);
+      c.edchitm[i] = KNLreadctr(gbl.edc_hitm[i]);
+      c.edcmisse[i] = KNLreadctr(gbl.edc_misse[i]);
+      c.edcmissm[i] = KNLreadctr(gbl.edc_missm[i]);
+    }
+}
+
+#endif
+}

lib/Stat.h

@@ -0,0 +1,104 @@
+#ifndef _GRID_STAT_H
+#define _GRID_STAT_H
+
+#ifdef AVX512
+#define _KNIGHTS_LANDING_ROOTONLY
+#endif
+
+namespace Grid { 
+
+///////////////////////////////////////////////////////////////////////////////
+// Extra KNL counters from MCDRAM
+///////////////////////////////////////////////////////////////////////////////
+#ifdef _KNIGHTS_LANDING_
+#define NMC 6
+#define NEDC 8
+struct ctrs
+{
+    uint64_t mcrd[NMC];
+    uint64_t mcwr[NMC];
+    uint64_t edcrd[NEDC]; 
+    uint64_t edcwr[NEDC];
+    uint64_t edchite[NEDC];
+    uint64_t edchitm[NEDC];
+    uint64_t edcmisse[NEDC];
+    uint64_t edcmissm[NEDC];
+};
+// Peter/Azusa:
+// Our modification of a code provided by Larry Meadows from Intel
+// Verified by email exchange non-NDA, ok for github. Should be as uses /sys/devices/ FS
+// so is already public and in the linux kernel for KNL.
+struct knl_gbl_
+{
+  int mc_rd[NMC];
+  int mc_wr[NMC];
+  int edc_rd[NEDC];
+  int edc_wr[NEDC];
+  int edc_hite[NEDC];
+  int edc_hitm[NEDC];
+  int edc_misse[NEDC];
+  int edc_missm[NEDC];
+};
+#endif
+///////////////////////////////////////////////////////////////////////////////
+
+class PmuStat
+{
+    uint64_t counters[8][256];
+#ifdef _KNIGHTS_LANDING_
+    static struct knl_gbl_ gbl;
+#endif
+    const char *name;
+
+    uint64_t reads;     // memory reads
+    uint64_t writes;    // memory writes
+    uint64_t mrstart;   // memory read counter at start of parallel region
+    uint64_t mrend;     // memory read counter at end of parallel region
+    uint64_t mwstart;   // memory write counter at start of parallel region
+    uint64_t mwend;     // memory write counter at end of parallel region
+
+    // cumulative counters
+    uint64_t count;     // number of invocations
+    uint64_t tregion;   // total time in parallel region (from thread 0)
+    uint64_t tcycles;   // total cycles inside parallel region
+    uint64_t inst, ref, cyc;   // fixed counters
+    uint64_t pmc0, pmc1;// pmu
+    // add memory counters here
+    // temp variables
+    uint64_t tstart;    // tsc at start of parallel region
+    uint64_t tend;      // tsc at end of parallel region
+    // map for ctrs values
+    // 0 pmc0 start
+    // 1 pmc0 end
+    // 2 pmc1 start
+    // 3 pmc1 end
+    // 4 tsc start
+    // 5 tsc end
+    static bool pmu_initialized;
+public:
+    static bool is_init(void){ return pmu_initialized;}
+    static void pmu_init(void);
+    static void pmu_fini(void);
+    static void pmu_start(void);
+    static void pmu_stop(void);
+    void accum(int nthreads);
+    static void xmemctrs(uint64_t *mr, uint64_t *mw);
+    void start(void);
+    void enter(int t);
+    void exit(int t);
+    void print(void);
+    void init(const char *regname);
+    void clear(void);
+#ifdef _KNIGHTS_LANDING_
+    static void     KNLsetup(void);
+    static uint64_t KNLreadctr(int fd);
+    static void     KNLreadctrs(ctrs &c);
+    static void     KNLevsetup(const char *ename, int &fd, int event, int umask);
+#endif
+    
+  };
+
+}
+#endif
+
+

lib/Stencil.h

@@ -70,9 +70,70 @@
 
  namespace Grid {
 
+template<class vobj,class cobj,class compressor> void 
+Gather_plane_simple_table_compute (const Lattice<vobj> &rhs,std::vector<cobj,alignedAllocator<cobj> > &buffer,int dimension,int plane,int cbmask,compressor &compress, int off,std::vector<std::pair<int,int> >& table)
+{
+  table.resize(0);
+  int rd = rhs._grid->_rdimensions[dimension];
+
+  if ( !rhs._grid->CheckerBoarded(dimension) ) {
+    cbmask = 0x3;
+  }
+  int so= plane*rhs._grid->_ostride[dimension]; // base offset for start of plane 
+  int e1=rhs._grid->_slice_nblock[dimension];
+  int e2=rhs._grid->_slice_block[dimension];
+
+  int stride=rhs._grid->_slice_stride[dimension];
+  if ( cbmask == 0x3 ) { 
+    table.resize(e1*e2);
+    for(int n=0;n<e1;n++){
+      for(int b=0;b<e2;b++){
+	int o  = n*stride;
+	int bo = n*e2;
+	table[bo+b]=std::pair<int,int>(bo+b,o+b);
+      }
+    }
+  } else { 
+     int bo=0;
+     table.resize(e1*e2/2);
+     for(int n=0;n<e1;n++){
+       for(int b=0;b<e2;b++){
+	 int o  = n*stride;
+	 int ocb=1<<rhs._grid->CheckerBoardFromOindexTable(o+b);
+	 if ( ocb &cbmask ) {
+	   table[bo]=std::pair<int,int>(bo,o+b); bo++;
+	 }
+       }
+     }
+  }
+}
+
+template<class vobj,class cobj,class compressor> void 
+Gather_plane_simple_table (std::vector<std::pair<int,int> >& table,const Lattice<vobj> &rhs,std::vector<cobj,alignedAllocator<cobj> > &buffer,
+			   compressor &compress, int off,int so)
+{
+PARALLEL_FOR_LOOP     
+     for(int i=0;i<table.size();i++){
+       buffer[off+table[i].first]=compress(rhs._odata[so+table[i].second]);
+     }
+}
+
+template<class vobj,class cobj,class compressor> void 
+Gather_plane_simple_stencil (const Lattice<vobj> &rhs,std::vector<cobj,alignedAllocator<cobj> > &buffer,int dimension,int plane,int cbmask,compressor &compress, int off,
+			     double &t_table ,double & t_data )
+{
+  std::vector<std::pair<int,int> > table;
+  Gather_plane_simple_table_compute (rhs, buffer,dimension,plane,cbmask,compress,off,table);
+  int so  = plane*rhs._grid->_ostride[dimension]; // base offset for start of plane 
+  Gather_plane_simple_table         (table,rhs,buffer,compress,off,so);
+}
+
+
+
+
    struct StencilEntry { 
-     uint32_t _offset;
+     uint64_t _offset;
-     uint32_t _byte_offset;
+     uint64_t _byte_offset;
      uint16_t _is_local;
      uint16_t _permute;
      uint32_t _around_the_world; //256 bits, 32 bytes, 1/2 cacheline
@@ -101,12 +162,14 @@
        };
 
        std::vector<Packet> Packets;
+
+       int face_table_computed;
+       std::vector<std::vector<std::pair<int,int> > > face_table ;
        
- #define SEND_IMMEDIATE
+#define SEND_IMMEDIATE
- #define SERIAL_SENDS
+#define SERIAL_SENDS
 
        void AddPacket(void *xmit,void * rcv, Integer to,Integer from,Integer bytes){
-	 comms_bytes+=2.0*bytes;
  #ifdef SEND_IMMEDIATE
 	 commtime-=usecond();
 	 _grid->SendToRecvFrom(xmit,to,rcv,from,bytes);
@@ -256,7 +319,8 @@
 	   if( _entries[i]._is_local ) {
 	     _entries[i]._byte_offset = _entries[i]._offset*sizeof(vobj);
 	   } else { 
-	     _entries[i]._byte_offset =(uint64_t)&comm_buf[0]+ _entries[i]._offset*sizeof(cobj);
+	     // PrecomputeByteOffsets [5] 16384/32768 140735768678528 140735781261056 2581581952
+	     _entries[i]._byte_offset = _entries[i]._offset*sizeof(cobj);
 	   }
 	 }
        };
@@ -265,17 +329,21 @@
 	 //	 _mm_prefetch((char *)&_entries[ent],_MM_HINT_T0);
        }
        inline uint64_t GetInfo(int &ptype,int &local,int &perm,int point,int ent,uint64_t base) {
-	 _mm_prefetch((char *)&_entries[ent+1],_MM_HINT_T0);
+	 uint64_t cbase = (uint64_t)&comm_buf[0];
 	 local = _entries[ent]._is_local;
 	 perm  = _entries[ent]._permute;
 	 if (perm)  ptype = _permute_type[point]; 
-	 if (local) return base + _entries[ent]._byte_offset;
+	 if (local) {
-	 else       return _entries[ent]._byte_offset;
+	   return  base + _entries[ent]._byte_offset;
+	 } else {
+	   return cbase + _entries[ent]._byte_offset;
+	 }
        }
        inline uint64_t GetPFInfo(int ent,uint64_t base) {
+	 uint64_t cbase = (uint64_t)&comm_buf[0];
 	 int local = _entries[ent]._is_local;
-	 if (local) return base + _entries[ent]._byte_offset;
+	 if (local) return  base + _entries[ent]._byte_offset;
-	 else       return        _entries[ent]._byte_offset;
+	 else       return cbase + _entries[ent]._byte_offset;
        }
 
        // Comms buffers
@@ -301,6 +369,48 @@
        double gathermtime;
        double splicetime;
        double nosplicetime;
+       double t_data;
+       double t_table;
+       double calls;
+
+       void ZeroCounters(void) {
+         gathertime = 0.;
+         jointime = 0.;
+         commtime = 0.;
+         halogtime = 0.;
+         mergetime = 0.;
+         spintime = 0.;
+         gathermtime = 0.;
+         splicetime = 0.;
+         nosplicetime = 0.;
+	 t_data = 0.0;
+         t_table= 0.0;
+         comms_bytes = 0.;
+         calls = 0.;
+       };
+
+       void Report(void) {
+#define PRINTIT(A)	\
+ std::cout << GridLogMessage << " Stencil " << #A << " "<< A/calls<<std::endl;
+	 if ( calls > 0. ) {
+	   std::cout << GridLogMessage << " Stencil calls "<<calls<<std::endl;
+	   PRINTIT(halogtime);
+	   PRINTIT(gathertime);
+	   PRINTIT(gathermtime);
+	   PRINTIT(mergetime);
+	   if(comms_bytes>1.0){
+	     PRINTIT(comms_bytes);
+	     PRINTIT(commtime);
+	     std::cout << GridLogMessage << " Stencil " << comms_bytes/commtime/1000. << " GB/s "<<std::endl;
+	   }
+	   PRINTIT(jointime);
+	   PRINTIT(spintime);
+	   PRINTIT(splicetime);
+	   PRINTIT(nosplicetime);
+	   PRINTIT(t_table);
+	   PRINTIT(t_data);
+	 }
+       };
  #endif
 
    CartesianStencil(GridBase *grid,
@@ -310,18 +420,7 @@
 				      const std::vector<int> &distances) 
      :   _permute_type(npoints), _comm_buf_size(npoints)
      {
- #ifdef TIMING_HACK
+       face_table_computed=0;
-       gathertime=0;
-       jointime=0;
-       commtime=0;
-       halogtime=0;
-       mergetime=0;
-       spintime=0;
-       gathermtime=0;
-       splicetime=0;
-       nosplicetime=0;
-       comms_bytes=0;
- #endif
        _npoints = npoints;
        _grid    = grid;
        _directions = directions;
@@ -623,6 +722,7 @@
        template<class compressor>
        void HaloExchange(const Lattice<vobj> &source,compressor &compress) 
        {
+	 calls++;
 	 Mergers.resize(0);
          Packets.resize(0);
          HaloGather(source,compress);
@@ -648,7 +748,7 @@
        }
 #endif
        template<class compressor>
-       void HaloGatherDir(const Lattice<vobj> &source,compressor &compress,int point)
+       void HaloGatherDir(const Lattice<vobj> &source,compressor &compress,int point,int & face_idx)
        {
 	   int dimension    = _directions[point];
 	   int displacement = _distances[point];
@@ -676,23 +776,23 @@
 	     if ( sshift[0] == sshift[1] ) {
 	       if (splice_dim) {
 		 splicetime-=usecond();
-		 GatherSimd(source,dimension,shift,0x3,compress);
+		 GatherSimd(source,dimension,shift,0x3,compress,face_idx);
 		 splicetime+=usecond();
 	       } else { 
 		 nosplicetime-=usecond();
-		 Gather(source,dimension,shift,0x3,compress);
+		 Gather(source,dimension,shift,0x3,compress,face_idx);
 		 nosplicetime+=usecond();
 	       }
 	     } else {
 	       if(splice_dim){
 		 splicetime-=usecond();
-		 GatherSimd(source,dimension,shift,0x1,compress);// if checkerboard is unfavourable take two passes
+		 GatherSimd(source,dimension,shift,0x1,compress,face_idx);// if checkerboard is unfavourable take two passes
-		 GatherSimd(source,dimension,shift,0x2,compress);// both with block stride loop iteration
+		 GatherSimd(source,dimension,shift,0x2,compress,face_idx);// both with block stride loop iteration
 		 splicetime+=usecond();
 	       } else {
 		 nosplicetime-=usecond();
-		 Gather(source,dimension,shift,0x1,compress);
+		 Gather(source,dimension,shift,0x1,compress,face_idx);
-		 Gather(source,dimension,shift,0x2,compress);
+		 Gather(source,dimension,shift,0x2,compress,face_idx);
 		 nosplicetime+=usecond();
 	       }
 	     }
@@ -710,17 +810,19 @@
 	 u_comm_offset=0;
 
 	 // Gather all comms buffers
+	 int face_idx=0;
 	 for(int point = 0 ; point < _npoints; point++) {
 	   compress.Point(point);
-	   HaloGatherDir(source,compress,point);
+	   HaloGatherDir(source,compress,point,face_idx);
 	 }
+	 face_table_computed=1;
 
 	 assert(u_comm_offset==_unified_buffer_size);
 	 halogtime+=usecond();
        }
 
        template<class compressor>
-	 void Gather(const Lattice<vobj> &rhs,int dimension,int shift,int cbmask,compressor & compress)
+	 void Gather(const Lattice<vobj> &rhs,int dimension,int shift,int cbmask,compressor & compress,int &face_idx)
 	 {
 	   typedef typename cobj::vector_type vector_type;
 	   typedef typename cobj::scalar_type scalar_type;
@@ -757,8 +859,20 @@
 	       int bytes = words * sizeof(cobj);
 
 	       gathertime-=usecond();
-	       Gather_plane_simple (rhs,u_send_buf,dimension,sx,cbmask,compress,u_comm_offset);
+	       int so  = sx*rhs._grid->_ostride[dimension]; // base offset for start of plane 
+	       if ( !face_table_computed ) {
+		 t_table-=usecond();
+		 face_table.resize(face_idx+1);
+		 Gather_plane_simple_table_compute (rhs,u_send_buf,dimension,sx,cbmask,compress,u_comm_offset,face_table[face_idx]);
+		 t_table+=usecond();
+	       }
+	       t_data-=usecond();
+	       Gather_plane_simple_table         (face_table[face_idx],rhs,u_send_buf,compress,u_comm_offset,so);
+	       face_idx++;
+	       t_data+=usecond();
 	       gathertime+=usecond();
+	       
+	       //	       Gather_plane_simple_stencil (rhs,u_send_buf,dimension,sx,cbmask,compress,u_comm_offset,t_table,t_data);
 
 	       int rank           = _grid->_processor;
 	       int recv_from_rank;
@@ -781,7 +895,7 @@
 
 
        template<class compressor>
-	 void  GatherSimd(const Lattice<vobj> &rhs,int dimension,int shift,int cbmask,compressor &compress)
+	 void  GatherSimd(const Lattice<vobj> &rhs,int dimension,int shift,int cbmask,compressor &compress,int & face_idx)
 	 {
 	   const int Nsimd = _grid->Nsimd();
 

lib/Threads.h

@@ -37,7 +37,11 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 
 #ifdef GRID_OMP
 #include <omp.h>
+#ifdef GRID_NUMA
+#define PARALLEL_FOR_LOOP _Pragma("omp parallel for schedule(static)")
+#else
 #define PARALLEL_FOR_LOOP _Pragma("omp parallel for schedule(runtime)")
+#endif
 #define PARALLEL_NESTED_LOOP2 _Pragma("omp parallel for collapse(2)")
 #else
 #define PARALLEL_FOR_LOOP 

lib/algorithms/iterative/ConjugateGradient.h

@@ -1,153 +1,168 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./lib/algorithms/iterative/ConjugateGradient.h
+Source file: ./lib/algorithms/iterative/ConjugateGradient.h
 
-    Copyright (C) 2015
+Copyright (C) 2015
 
 Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: paboyle <paboyle@ph.ed.ac.uk>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution
-    *************************************************************************************/
+directory
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef GRID_CONJUGATE_GRADIENT_H
 #define GRID_CONJUGATE_GRADIENT_H
 
 namespace Grid {
 
-    /////////////////////////////////////////////////////////////
+/////////////////////////////////////////////////////////////
-    // Base classes for iterative processes based on operators
+// Base classes for iterative processes based on operators
-    // single input vec, single output vec.
+// single input vec, single output vec.
-    /////////////////////////////////////////////////////////////
+/////////////////////////////////////////////////////////////
 
-  template<class Field> 
+template <class Field>
-    class ConjugateGradient : public OperatorFunction<Field> {
+class ConjugateGradient : public OperatorFunction<Field> {
-public:                                                
+ public:
-    bool ErrorOnNoConverge; //throw an assert when the CG fails to converge. Defaults true.
+  bool ErrorOnNoConverge;  // throw an assert when the CG fails to converge.
-    RealD   Tolerance;
+                           // Defaults true.
-    Integer MaxIterations;
+  RealD Tolerance;
-  ConjugateGradient(RealD tol,Integer maxit, bool err_on_no_conv = true) : Tolerance(tol), MaxIterations(maxit), ErrorOnNoConverge(err_on_no_conv){ 
+  Integer MaxIterations;
-    };
+  ConjugateGradient(RealD tol, Integer maxit, bool err_on_no_conv = true)
+      : Tolerance(tol),
+        MaxIterations(maxit),
+        ErrorOnNoConverge(err_on_no_conv){};
 
+  void operator()(LinearOperatorBase<Field> &Linop, const Field &src,
+                  Field &psi) {
+    psi.checkerboard = src.checkerboard;
+    conformable(psi, src);
 
-    void operator() (LinearOperatorBase<Field> &Linop,const Field &src, Field &psi){
+    RealD cp, c, a, d, b, ssq, qq, b_pred;
 
-      psi.checkerboard = src.checkerboard;
+    Field p(src);
-      conformable(psi,src);
+    Field mmp(src);
+    Field r(src);
 
-      RealD cp,c,a,d,b,ssq,qq,b_pred;
+    // Initial residual computation & set up
-      
+    RealD guess = norm2(psi);
-      Field   p(src);
+    assert(std::isnan(guess) == 0);
-      Field mmp(src);
-      Field   r(src);
-      
-      //Initial residual computation & set up
-      RealD guess = norm2(psi);
-      assert(std::isnan(guess)==0);
 
-      Linop.HermOpAndNorm(psi,mmp,d,b);
+    
-      
+    Linop.HermOpAndNorm(psi, mmp, d, b);
-      r= src-mmp;
+    
-      p= r;
-      
-      a  =norm2(p);
-      cp =a;
-      ssq=norm2(src);
 
-      std::cout<<GridLogIterative <<std::setprecision(4)<< "ConjugateGradient: guess "<<guess<<std::endl;
+    r = src - mmp;
-      std::cout<<GridLogIterative <<std::setprecision(4)<< "ConjugateGradient:   src "<<ssq  <<std::endl;
+    p = r;
-      std::cout<<GridLogIterative <<std::setprecision(4)<< "ConjugateGradient:    mp "<<d    <<std::endl;
-      std::cout<<GridLogIterative <<std::setprecision(4)<< "ConjugateGradient:   mmp "<<b    <<std::endl;
-      std::cout<<GridLogIterative <<std::setprecision(4)<< "ConjugateGradient:  cp,r "<<cp   <<std::endl;
-      std::cout<<GridLogIterative <<std::setprecision(4)<< "ConjugateGradient:     p "<<a    <<std::endl;
 
-      RealD rsq =  Tolerance* Tolerance*ssq;
+    a = norm2(p);
-      
+    cp = a;
-      //Check if guess is really REALLY good :)
+    ssq = norm2(src);
-      if ( cp <= rsq ) {
-	return;
-      }
-      
-      std::cout<<GridLogIterative << std::setprecision(4)<< "ConjugateGradient: k=0 residual "<<cp<<" target "<<rsq<<std::endl;
 
-      GridStopWatch LinalgTimer;
+    std::cout << GridLogIterative << std::setprecision(4)
-      GridStopWatch MatrixTimer;
+              << "ConjugateGradient: guess " << guess << std::endl;
-      GridStopWatch SolverTimer;
+    std::cout << GridLogIterative << std::setprecision(4)
+              << "ConjugateGradient:   src " << ssq << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4)
+              << "ConjugateGradient:    mp " << d << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4)
+              << "ConjugateGradient:   mmp " << b << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4)
+              << "ConjugateGradient:  cp,r " << cp << std::endl;
+    std::cout << GridLogIterative << std::setprecision(4)
+              << "ConjugateGradient:     p " << a << std::endl;
 
-      SolverTimer.Start();
+    RealD rsq = Tolerance * Tolerance * ssq;
-      int k;
-      for (k=1;k<=MaxIterations;k++){
-	
-	c=cp;
 
-	MatrixTimer.Start();
+    // Check if guess is really REALLY good :)
-	Linop.HermOpAndNorm(p,mmp,d,qq);
+    if (cp <= rsq) {
-	MatrixTimer.Stop();
+      return;
-
-	LinalgTimer.Start();
-	//	RealD    qqck = norm2(mmp);
-	//	ComplexD dck  = innerProduct(p,mmp);
-      
-	a      = c/d;
-	b_pred = a*(a*qq-d)/c;
-
-	cp = axpy_norm(r,-a,mmp,r);
-	b = cp/c;
-	
-	// Fuse these loops ; should be really easy
-	psi= a*p+psi;
-	p  = p*b+r;
-	  
-	LinalgTimer.Stop();
-	std::cout<<GridLogIterative<<"ConjugateGradient: Iteration " <<k<<" residual "<<cp<< " target "<< rsq<<std::endl;
-	
-	// Stopping condition
-	if ( cp <= rsq ) { 
-	  
-	  SolverTimer.Stop();
-	  Linop.HermOpAndNorm(psi,mmp,d,qq);
-	  p=mmp-src;
-	  
-	  RealD mmpnorm = sqrt(norm2(mmp));
-	  RealD psinorm = sqrt(norm2(psi));
-	  RealD srcnorm = sqrt(norm2(src));
-	  RealD resnorm = sqrt(norm2(p));
-	  RealD true_residual = resnorm/srcnorm;
-
-	  std::cout<<GridLogMessage<<"ConjugateGradient: Converged on iteration " <<k
-		   <<" computed residual "<<sqrt(cp/ssq)
-		   <<" true residual "    <<true_residual
-		   <<" target "<<Tolerance<<std::endl;
-	  std::cout<<GridLogMessage<<"Time elapsed: Total "<< SolverTimer.Elapsed() << " Matrix  "<<MatrixTimer.Elapsed() << " Linalg "<<LinalgTimer.Elapsed();
-	  std::cout<<std::endl;
-	  
-	  if(ErrorOnNoConverge)
-	    assert(true_residual/Tolerance < 1000.0);
-
-	  return;
-	}
-      }
-      std::cout<<GridLogMessage<<"ConjugateGradient did NOT converge"<<std::endl;
-      if(ErrorOnNoConverge)	
-	assert(0);
     }
-  };
+
+    std::cout << GridLogIterative << std::setprecision(4)
+              << "ConjugateGradient: k=0 residual " << cp << " target " << rsq
+              << std::endl;
+
+    GridStopWatch LinalgTimer;
+    GridStopWatch MatrixTimer;
+    GridStopWatch SolverTimer;
+
+    SolverTimer.Start();
+    int k;
+    for (k = 1; k <= MaxIterations; k++) {
+      c = cp;
+
+      MatrixTimer.Start();
+      Linop.HermOpAndNorm(p, mmp, d, qq);
+      MatrixTimer.Stop();
+
+      LinalgTimer.Start();
+      //  RealD    qqck = norm2(mmp);
+      //  ComplexD dck  = innerProduct(p,mmp);
+
+      a = c / d;
+      b_pred = a * (a * qq - d) / c;
+
+      cp = axpy_norm(r, -a, mmp, r);
+      b = cp / c;
+
+      // Fuse these loops ; should be really easy
+      psi = a * p + psi;
+      p = p * b + r;
+
+      LinalgTimer.Stop();
+      std::cout << GridLogIterative << "ConjugateGradient: Iteration " << k
+                << " residual " << cp << " target " << rsq << std::endl;
+
+      // Stopping condition
+      if (cp <= rsq) {
+        SolverTimer.Stop();
+        Linop.HermOpAndNorm(psi, mmp, d, qq);
+        p = mmp - src;
+
+        RealD mmpnorm = sqrt(norm2(mmp));
+        RealD psinorm = sqrt(norm2(psi));
+        RealD srcnorm = sqrt(norm2(src));
+        RealD resnorm = sqrt(norm2(p));
+        RealD true_residual = resnorm / srcnorm;
+
+        std::cout << GridLogMessage
+                  << "ConjugateGradient: Converged on iteration " << k << std::endl;
+        std::cout << GridLogMessage << "Computed residual " << sqrt(cp / ssq)
+                  << " true residual " << true_residual << " target "
+                  << Tolerance << std::endl;
+        std::cout << GridLogMessage << "Time elapsed: Iterations "
+                  << SolverTimer.Elapsed() << " Matrix  "
+                  << MatrixTimer.Elapsed() << " Linalg "
+                  << LinalgTimer.Elapsed();
+        std::cout << std::endl;
+
+        if (ErrorOnNoConverge) assert(true_residual / Tolerance < 1000.0);
+
+        return;
+      }
+    }
+    std::cout << GridLogMessage << "ConjugateGradient did NOT converge"
+              << std::endl;
+    if (ErrorOnNoConverge) assert(0);
+  }
+};
 }
 #endif

lib/cartesian/Cartesian_base.h

@@ -81,11 +81,8 @@ public:
     virtual int CheckerBoardDestination(int source_cb,int shift,int dim)=0;
     virtual int CheckerBoardShift(int source_cb,int dim,int shift,int osite)=0;
     virtual int CheckerBoardShiftForCB(int source_cb,int dim,int shift,int cb)=0;
-    int  CheckerBoardFromOindex (int Oindex){
+    virtual int CheckerBoardFromOindex (int Oindex)=0;
-      std::vector<int> ocoor;
+    virtual int CheckerBoardFromOindexTable (int Oindex)=0;
-      oCoorFromOindex(ocoor,Oindex); 
-      return CheckerBoard(ocoor);
-    }
 
     //////////////////////////////////////////////////////////////////////////////////////////////
     // Local layout calculations

lib/cartesian/Cartesian_full.h

@@ -39,6 +39,13 @@ class GridCartesian: public GridBase {
 
 public:
 
+    virtual int  CheckerBoardFromOindexTable (int Oindex) {
+      return 0;
+    }
+    virtual int  CheckerBoardFromOindex (int Oindex)
+    {
+      return 0;
+    }
     virtual int CheckerBoarded(int dim){
       return 0;
     }

lib/cartesian/Cartesian_red_black.h

@@ -43,6 +43,7 @@ class GridRedBlackCartesian : public GridBase
 public:
     std::vector<int> _checker_dim_mask;
     int              _checker_dim;
+    std::vector<int> _checker_board;
 
     virtual int CheckerBoarded(int dim){
       if( dim==_checker_dim) return 1;
@@ -72,12 +73,20 @@ public:
       // or by looping over x,y,z and multiply rather than computing checkerboard.
 	  
       if ( (source_cb+ocb)&1 ) {
-
 	return (shift)/2;
       } else {
 	return (shift+1)/2;
       }
     }
+    virtual int  CheckerBoardFromOindexTable (int Oindex) {
+      return _checker_board[Oindex];
+    }
+    virtual int  CheckerBoardFromOindex (int Oindex)
+    {
+      std::vector<int> ocoor;
+      oCoorFromOindex(ocoor,Oindex);
+      return CheckerBoard(ocoor);
+    }
     virtual int CheckerBoardShift(int source_cb,int dim,int shift,int osite){
 
       if(dim != _checker_dim) return shift;
@@ -169,7 +178,7 @@ public:
 	// all elements of a simd vector must have same checkerboard.
 	// If Ls vectorised, this must still be the case; e.g. dwf rb5d
 	if ( _simd_layout[d]>1 ) {
-	  if ( d != _checker_dim ) { 
+	  if ( checker_dim_mask[d] ) { 
 	    assert( (_rdimensions[d]&0x1) == 0 );
 	  }
 	}
@@ -185,6 +194,8 @@ public:
 	  _ostride[d] = _ostride[d-1]*_rdimensions[d-1];
 	  _istride[d] = _istride[d-1]*_simd_layout[d-1];
 	}
+
+
       }
             
       ////////////////////////////////////////////////////////////////////////////////////////////
@@ -205,6 +216,18 @@ public:
 	_slice_nblock[d]=nblock;
 	block = block*_rdimensions[d];
       }
+
+      ////////////////////////////////////////////////
+      // Create a checkerboard lookup table
+      ////////////////////////////////////////////////
+      int rvol = 1;
+      for(int d=0;d<_ndimension;d++){
+	rvol=rvol * _rdimensions[d];
+      }
+      _checker_board.resize(rvol);
+      for(int osite=0;osite<_osites;osite++){
+	_checker_board[osite] = CheckerBoardFromOindex (osite);
+      }
       
     };
 protected:

lib/cshift/Cshift_common.h

@@ -1,3 +1,4 @@
+
     /*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
@@ -56,6 +57,7 @@ Gather_plane_simple (const Lattice<vobj> &rhs,std::vector<cobj,alignedAllocator<
   
   int e1=rhs._grid->_slice_nblock[dimension];
   int e2=rhs._grid->_slice_block[dimension];
+
   int stride=rhs._grid->_slice_stride[dimension];
   if ( cbmask == 0x3 ) { 
 PARALLEL_NESTED_LOOP2
@@ -68,15 +70,20 @@ PARALLEL_NESTED_LOOP2
     }
   } else { 
      int bo=0;
+     std::vector<std::pair<int,int> > table;
      for(int n=0;n<e1;n++){
        for(int b=0;b<e2;b++){
 	 int o  = n*stride;
-	 int ocb=1<<rhs._grid->CheckerBoardFromOindex(o+b);// Could easily be a table lookup
+	 int ocb=1<<rhs._grid->CheckerBoardFromOindexTable(o+b);
 	 if ( ocb &cbmask ) {
-	   buffer[off+bo++]=compress(rhs._odata[so+o+b]);
+	   table.push_back(std::pair<int,int> (bo++,o+b));
 	 }
        }
      }
+PARALLEL_FOR_LOOP     
+     for(int i=0;i<table.size();i++){
+       buffer[off+table[i].first]=compress(rhs._odata[so+table[i].second]);
+     }
   }
 }
 

lib/fftw/fftw3.h

@@ -0,0 +1,412 @@
+/*
+ * Copyright (c) 2003, 2007-14 Matteo Frigo
+ * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
+ *
+ * The following statement of license applies *only* to this header file,
+ * and *not* to the other files distributed with FFTW or derived therefrom:
+ * 
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS
+ * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+ * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
+ * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE
+ * GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
+ * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+ * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+ * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/***************************** NOTE TO USERS *********************************
+ *
+ *                 THIS IS A HEADER FILE, NOT A MANUAL
+ *
+ *    If you want to know how to use FFTW, please read the manual,
+ *    online at http://www.fftw.org/doc/ and also included with FFTW.
+ *    For a quick start, see the manual's tutorial section.
+ *
+ *   (Reading header files to learn how to use a library is a habit
+ *    stemming from code lacking a proper manual.  Arguably, it's a
+ *    *bad* habit in most cases, because header files can contain
+ *    interfaces that are not part of the public, stable API.)
+ *
+ ****************************************************************************/
+
+#ifndef FFTW3_H
+#define FFTW3_H
+
+#include <stdio.h>
+
+#ifdef __cplusplus
+extern "C"
+{
+#endif /* __cplusplus */
+
+/* If <complex.h> is included, use the C99 complex type.  Otherwise
+   define a type bit-compatible with C99 complex */
+#if !defined(FFTW_NO_Complex) && defined(_Complex_I) && defined(complex) && defined(I)
+#  define FFTW_DEFINE_COMPLEX(R, C) typedef R _Complex C
+#else
+#  define FFTW_DEFINE_COMPLEX(R, C) typedef R C[2]
+#endif
+
+#define FFTW_CONCAT(prefix, name) prefix ## name
+#define FFTW_MANGLE_DOUBLE(name) FFTW_CONCAT(fftw_, name)
+#define FFTW_MANGLE_FLOAT(name) FFTW_CONCAT(fftwf_, name)
+#define FFTW_MANGLE_LONG_DOUBLE(name) FFTW_CONCAT(fftwl_, name)
+#define FFTW_MANGLE_QUAD(name) FFTW_CONCAT(fftwq_, name)
+
+/* IMPORTANT: for Windows compilers, you should add a line
+        #define FFTW_DLL
+   here and in kernel/ifftw.h if you are compiling/using FFTW as a
+   DLL, in order to do the proper importing/exporting, or
+   alternatively compile with -DFFTW_DLL or the equivalent
+   command-line flag.  This is not necessary under MinGW/Cygwin, where
+   libtool does the imports/exports automatically. */
+#if defined(FFTW_DLL) && (defined(_WIN32) || defined(__WIN32__))
+   /* annoying Windows syntax for shared-library declarations */
+#  if defined(COMPILING_FFTW) /* defined in api.h when compiling FFTW */
+#    define FFTW_EXTERN extern __declspec(dllexport) 
+#  else /* user is calling FFTW; import symbol */
+#    define FFTW_EXTERN extern __declspec(dllimport) 
+#  endif
+#else
+#  define FFTW_EXTERN extern
+#endif
+
+enum fftw_r2r_kind_do_not_use_me {
+     FFTW_R2HC=0, FFTW_HC2R=1, FFTW_DHT=2,
+     FFTW_REDFT00=3, FFTW_REDFT01=4, FFTW_REDFT10=5, FFTW_REDFT11=6,
+     FFTW_RODFT00=7, FFTW_RODFT01=8, FFTW_RODFT10=9, FFTW_RODFT11=10
+};
+
+struct fftw_iodim_do_not_use_me {
+     int n;                     /* dimension size */
+     int is;			/* input stride */
+     int os;			/* output stride */
+};
+
+#include <stddef.h> /* for ptrdiff_t */
+struct fftw_iodim64_do_not_use_me {
+     ptrdiff_t n;                     /* dimension size */
+     ptrdiff_t is;			/* input stride */
+     ptrdiff_t os;			/* output stride */
+};
+
+typedef void (*fftw_write_char_func_do_not_use_me)(char c, void *);
+typedef int (*fftw_read_char_func_do_not_use_me)(void *);
+
+/*
+  huge second-order macro that defines prototypes for all API
+  functions.  We expand this macro for each supported precision
+ 
+  X: name-mangling macro
+  R: real data type
+  C: complex data type
+*/
+
+#define FFTW_DEFINE_API(X, R, C)					   \
+									   \
+FFTW_DEFINE_COMPLEX(R, C);						   \
+									   \
+typedef struct X(plan_s) *X(plan);					   \
+									   \
+typedef struct fftw_iodim_do_not_use_me X(iodim);			   \
+typedef struct fftw_iodim64_do_not_use_me X(iodim64);			   \
+									   \
+typedef enum fftw_r2r_kind_do_not_use_me X(r2r_kind);			   \
+									   \
+typedef fftw_write_char_func_do_not_use_me X(write_char_func);		   \
+typedef fftw_read_char_func_do_not_use_me X(read_char_func);		   \
+									   \
+FFTW_EXTERN void X(execute)(const X(plan) p);				   \
+									   \
+FFTW_EXTERN X(plan) X(plan_dft)(int rank, const int *n,			   \
+		    C *in, C *out, int sign, unsigned flags);		   \
+									   \
+FFTW_EXTERN X(plan) X(plan_dft_1d)(int n, C *in, C *out, int sign,	   \
+		       unsigned flags);					   \
+FFTW_EXTERN X(plan) X(plan_dft_2d)(int n0, int n1,			   \
+		       C *in, C *out, int sign, unsigned flags);	   \
+FFTW_EXTERN X(plan) X(plan_dft_3d)(int n0, int n1, int n2,		   \
+		       C *in, C *out, int sign, unsigned flags);	   \
+									   \
+FFTW_EXTERN X(plan) X(plan_many_dft)(int rank, const int *n,		   \
+                         int howmany,					   \
+                         C *in, const int *inembed,			   \
+                         int istride, int idist,			   \
+                         C *out, const int *onembed,			   \
+                         int ostride, int odist,			   \
+                         int sign, unsigned flags);			   \
+									   \
+FFTW_EXTERN X(plan) X(plan_guru_dft)(int rank, const X(iodim) *dims,	   \
+			 int howmany_rank,				   \
+			 const X(iodim) *howmany_dims,			   \
+			 C *in, C *out,					   \
+			 int sign, unsigned flags);			   \
+FFTW_EXTERN X(plan) X(plan_guru_split_dft)(int rank, const X(iodim) *dims, \
+			 int howmany_rank,				   \
+			 const X(iodim) *howmany_dims,			   \
+			 R *ri, R *ii, R *ro, R *io,			   \
+			 unsigned flags);				   \
+									   \
+FFTW_EXTERN X(plan) X(plan_guru64_dft)(int rank,			   \
+                         const X(iodim64) *dims,			   \
+			 int howmany_rank,				   \
+			 const X(iodim64) *howmany_dims,		   \
+			 C *in, C *out,					   \
+			 int sign, unsigned flags);			   \
+FFTW_EXTERN X(plan) X(plan_guru64_split_dft)(int rank,			   \
+                         const X(iodim64) *dims,			   \
+			 int howmany_rank,				   \
+			 const X(iodim64) *howmany_dims,		   \
+			 R *ri, R *ii, R *ro, R *io,			   \
+			 unsigned flags);				   \
+									   \
+FFTW_EXTERN void X(execute_dft)(const X(plan) p, C *in, C *out);	   \
+FFTW_EXTERN void X(execute_split_dft)(const X(plan) p, R *ri, R *ii,	   \
+                                      R *ro, R *io);			   \
+									   \
+FFTW_EXTERN X(plan) X(plan_many_dft_r2c)(int rank, const int *n,	   \
+                             int howmany,				   \
+                             R *in, const int *inembed,			   \
+                             int istride, int idist,			   \
+                             C *out, const int *onembed,		   \
+                             int ostride, int odist,			   \
+                             unsigned flags);				   \
+									   \
+FFTW_EXTERN X(plan) X(plan_dft_r2c)(int rank, const int *n,		   \
+                        R *in, C *out, unsigned flags);			   \
+									   \
+FFTW_EXTERN X(plan) X(plan_dft_r2c_1d)(int n,R *in,C *out,unsigned flags); \
+FFTW_EXTERN X(plan) X(plan_dft_r2c_2d)(int n0, int n1,			   \
+			   R *in, C *out, unsigned flags);		   \
+FFTW_EXTERN X(plan) X(plan_dft_r2c_3d)(int n0, int n1,			   \
+			   int n2,					   \
+			   R *in, C *out, unsigned flags);		   \
+									   \
+									   \
+FFTW_EXTERN X(plan) X(plan_many_dft_c2r)(int rank, const int *n,	   \
+			     int howmany,				   \
+			     C *in, const int *inembed,			   \
+			     int istride, int idist,			   \
+			     R *out, const int *onembed,		   \
+			     int ostride, int odist,			   \
+			     unsigned flags);				   \
+									   \
+FFTW_EXTERN X(plan) X(plan_dft_c2r)(int rank, const int *n,		   \
+                        C *in, R *out, unsigned flags);			   \
+									   \
+FFTW_EXTERN X(plan) X(plan_dft_c2r_1d)(int n,C *in,R *out,unsigned flags); \
+FFTW_EXTERN X(plan) X(plan_dft_c2r_2d)(int n0, int n1,			   \
+			   C *in, R *out, unsigned flags);		   \
+FFTW_EXTERN X(plan) X(plan_dft_c2r_3d)(int n0, int n1,			   \
+			   int n2,					   \
+			   C *in, R *out, unsigned flags);		   \
+									   \
+FFTW_EXTERN X(plan) X(plan_guru_dft_r2c)(int rank, const X(iodim) *dims,   \
+			     int howmany_rank,				   \
+			     const X(iodim) *howmany_dims,		   \
+			     R *in, C *out,				   \
+			     unsigned flags);				   \
+FFTW_EXTERN X(plan) X(plan_guru_dft_c2r)(int rank, const X(iodim) *dims,   \
+			     int howmany_rank,				   \
+			     const X(iodim) *howmany_dims,		   \
+			     C *in, R *out,				   \
+			     unsigned flags);				   \
+									   \
+FFTW_EXTERN X(plan) X(plan_guru_split_dft_r2c)(				   \
+                             int rank, const X(iodim) *dims,		   \
+			     int howmany_rank,				   \
+			     const X(iodim) *howmany_dims,		   \
+			     R *in, R *ro, R *io,			   \
+			     unsigned flags);				   \
+FFTW_EXTERN X(plan) X(plan_guru_split_dft_c2r)(				   \
+                             int rank, const X(iodim) *dims,		   \
+			     int howmany_rank,				   \
+			     const X(iodim) *howmany_dims,		   \
+			     R *ri, R *ii, R *out,			   \
+			     unsigned flags);				   \
+									   \
+FFTW_EXTERN X(plan) X(plan_guru64_dft_r2c)(int rank,			   \
+                             const X(iodim64) *dims,			   \
+			     int howmany_rank,				   \
+			     const X(iodim64) *howmany_dims,		   \
+			     R *in, C *out,				   \
+			     unsigned flags);				   \
+FFTW_EXTERN X(plan) X(plan_guru64_dft_c2r)(int rank,			   \
+                             const X(iodim64) *dims,			   \
+			     int howmany_rank,				   \
+			     const X(iodim64) *howmany_dims,		   \
+			     C *in, R *out,				   \
+			     unsigned flags);				   \
+									   \
+FFTW_EXTERN X(plan) X(plan_guru64_split_dft_r2c)(			   \
+                             int rank, const X(iodim64) *dims,		   \
+			     int howmany_rank,				   \
+			     const X(iodim64) *howmany_dims,		   \
+			     R *in, R *ro, R *io,			   \
+			     unsigned flags);				   \
+FFTW_EXTERN X(plan) X(plan_guru64_split_dft_c2r)(			   \
+                             int rank, const X(iodim64) *dims,		   \
+			     int howmany_rank,				   \
+			     const X(iodim64) *howmany_dims,		   \
+			     R *ri, R *ii, R *out,			   \
+			     unsigned flags);				   \
+									   \
+FFTW_EXTERN void X(execute_dft_r2c)(const X(plan) p, R *in, C *out);	   \
+FFTW_EXTERN void X(execute_dft_c2r)(const X(plan) p, C *in, R *out);	   \
+									   \
+FFTW_EXTERN void X(execute_split_dft_r2c)(const X(plan) p,		   \
+                                          R *in, R *ro, R *io);		   \
+FFTW_EXTERN void X(execute_split_dft_c2r)(const X(plan) p,		   \
+                                          R *ri, R *ii, R *out);	   \
+									   \
+FFTW_EXTERN X(plan) X(plan_many_r2r)(int rank, const int *n,		   \
+                         int howmany,					   \
+                         R *in, const int *inembed,			   \
+                         int istride, int idist,			   \
+                         R *out, const int *onembed,			   \
+                         int ostride, int odist,			   \
+                         const X(r2r_kind) *kind, unsigned flags);	   \
+									   \
+FFTW_EXTERN X(plan) X(plan_r2r)(int rank, const int *n, R *in, R *out,	   \
+                    const X(r2r_kind) *kind, unsigned flags);		   \
+									   \
+FFTW_EXTERN X(plan) X(plan_r2r_1d)(int n, R *in, R *out,		   \
+                       X(r2r_kind) kind, unsigned flags);		   \
+FFTW_EXTERN X(plan) X(plan_r2r_2d)(int n0, int n1, R *in, R *out,	   \
+                       X(r2r_kind) kind0, X(r2r_kind) kind1,		   \
+                       unsigned flags);					   \
+FFTW_EXTERN X(plan) X(plan_r2r_3d)(int n0, int n1, int n2,		   \
+                       R *in, R *out, X(r2r_kind) kind0,		   \
+                       X(r2r_kind) kind1, X(r2r_kind) kind2,		   \
+                       unsigned flags);					   \
+									   \
+FFTW_EXTERN X(plan) X(plan_guru_r2r)(int rank, const X(iodim) *dims,	   \
+                         int howmany_rank,				   \
+                         const X(iodim) *howmany_dims,			   \
+                         R *in, R *out,					   \
+                         const X(r2r_kind) *kind, unsigned flags);	   \
+									   \
+FFTW_EXTERN X(plan) X(plan_guru64_r2r)(int rank, const X(iodim64) *dims,   \
+                         int howmany_rank,				   \
+                         const X(iodim64) *howmany_dims,		   \
+                         R *in, R *out,					   \
+                         const X(r2r_kind) *kind, unsigned flags);	   \
+									   \
+FFTW_EXTERN void X(execute_r2r)(const X(plan) p, R *in, R *out);	   \
+									   \
+FFTW_EXTERN void X(destroy_plan)(X(plan) p);				   \
+FFTW_EXTERN void X(forget_wisdom)(void);				   \
+FFTW_EXTERN void X(cleanup)(void);					   \
+									   \
+FFTW_EXTERN void X(set_timelimit)(double t);				   \
+									   \
+FFTW_EXTERN void X(plan_with_nthreads)(int nthreads);			   \
+FFTW_EXTERN int X(init_threads)(void);					   \
+FFTW_EXTERN void X(cleanup_threads)(void);				   \
+									   \
+FFTW_EXTERN int X(export_wisdom_to_filename)(const char *filename);	   \
+FFTW_EXTERN void X(export_wisdom_to_file)(FILE *output_file);		   \
+FFTW_EXTERN char *X(export_wisdom_to_string)(void);			   \
+FFTW_EXTERN void X(export_wisdom)(X(write_char_func) write_char,   	   \
+                                  void *data);				   \
+FFTW_EXTERN int X(import_system_wisdom)(void);				   \
+FFTW_EXTERN int X(import_wisdom_from_filename)(const char *filename);	   \
+FFTW_EXTERN int X(import_wisdom_from_file)(FILE *input_file);		   \
+FFTW_EXTERN int X(import_wisdom_from_string)(const char *input_string);	   \
+FFTW_EXTERN int X(import_wisdom)(X(read_char_func) read_char, void *data); \
+									   \
+FFTW_EXTERN void X(fprint_plan)(const X(plan) p, FILE *output_file);	   \
+FFTW_EXTERN void X(print_plan)(const X(plan) p);			   \
+FFTW_EXTERN char *X(sprint_plan)(const X(plan) p);			   \
+									   \
+FFTW_EXTERN void *X(malloc)(size_t n);					   \
+FFTW_EXTERN R *X(alloc_real)(size_t n);					   \
+FFTW_EXTERN C *X(alloc_complex)(size_t n);				   \
+FFTW_EXTERN void X(free)(void *p);					   \
+									   \
+FFTW_EXTERN void X(flops)(const X(plan) p,				   \
+                          double *add, double *mul, double *fmas);	   \
+FFTW_EXTERN double X(estimate_cost)(const X(plan) p);			   \
+FFTW_EXTERN double X(cost)(const X(plan) p);				   \
+									   \
+FFTW_EXTERN int X(alignment_of)(R *p);                                     \
+FFTW_EXTERN const char X(version)[];                                       \
+FFTW_EXTERN const char X(cc)[];						   \
+FFTW_EXTERN const char X(codelet_optim)[];
+
+
+/* end of FFTW_DEFINE_API macro */
+
+FFTW_DEFINE_API(FFTW_MANGLE_DOUBLE, double, fftw_complex)
+FFTW_DEFINE_API(FFTW_MANGLE_FLOAT, float, fftwf_complex)
+FFTW_DEFINE_API(FFTW_MANGLE_LONG_DOUBLE, long double, fftwl_complex)
+
+/* __float128 (quad precision) is a gcc extension on i386, x86_64, and ia64
+   for gcc >= 4.6 (compiled in FFTW with --enable-quad-precision) */
+#if (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 6)) \
+ && !(defined(__ICC) || defined(__INTEL_COMPILER)) \
+ && (defined(__i386__) || defined(__x86_64__) || defined(__ia64__))
+#  if !defined(FFTW_NO_Complex) && defined(_Complex_I) && defined(complex) && defined(I)
+/* note: __float128 is a typedef, which is not supported with the _Complex
+         keyword in gcc, so instead we use this ugly __attribute__ version.
+         However, we can't simply pass the __attribute__ version to
+         FFTW_DEFINE_API because the __attribute__ confuses gcc in pointer
+         types.  Hence redefining FFTW_DEFINE_COMPLEX.  Ugh. */
+#    undef FFTW_DEFINE_COMPLEX
+#    define FFTW_DEFINE_COMPLEX(R, C) typedef _Complex float __attribute__((mode(TC))) C
+#  endif
+FFTW_DEFINE_API(FFTW_MANGLE_QUAD, __float128, fftwq_complex)
+#endif
+
+#define FFTW_FORWARD (-1)
+#define FFTW_BACKWARD (+1)
+
+#define FFTW_NO_TIMELIMIT (-1.0)
+
+/* documented flags */
+#define FFTW_MEASURE (0U)
+#define FFTW_DESTROY_INPUT (1U << 0)
+#define FFTW_UNALIGNED (1U << 1)
+#define FFTW_CONSERVE_MEMORY (1U << 2)
+#define FFTW_EXHAUSTIVE (1U << 3) /* NO_EXHAUSTIVE is default */
+#define FFTW_PRESERVE_INPUT (1U << 4) /* cancels FFTW_DESTROY_INPUT */
+#define FFTW_PATIENT (1U << 5) /* IMPATIENT is default */
+#define FFTW_ESTIMATE (1U << 6)
+#define FFTW_WISDOM_ONLY (1U << 21)
+
+/* undocumented beyond-guru flags */
+#define FFTW_ESTIMATE_PATIENT (1U << 7)
+#define FFTW_BELIEVE_PCOST (1U << 8)
+#define FFTW_NO_DFT_R2HC (1U << 9)
+#define FFTW_NO_NONTHREADED (1U << 10)
+#define FFTW_NO_BUFFERING (1U << 11)
+#define FFTW_NO_INDIRECT_OP (1U << 12)
+#define FFTW_ALLOW_LARGE_GENERIC (1U << 13) /* NO_LARGE_GENERIC is default */
+#define FFTW_NO_RANK_SPLITS (1U << 14)
+#define FFTW_NO_VRANK_SPLITS (1U << 15)
+#define FFTW_NO_VRECURSE (1U << 16)
+#define FFTW_NO_SIMD (1U << 17)
+#define FFTW_NO_SLOW (1U << 18)
+#define FFTW_NO_FIXED_RADIX_LARGE_N (1U << 19)
+#define FFTW_ALLOW_PRUNING (1U << 20)
+
+#ifdef __cplusplus
+}  /* extern "C" */
+#endif /* __cplusplus */
+
+#endif /* FFTW3_H */

lib/parallelIO/BinaryIO.h

@@ -194,22 +194,22 @@ class BinaryIO {
 
       std::vector<int> site({x,y,z,t});
 
-      if ( grid->IsBoss() ) {
+      if (grid->IsBoss()) {
-	fin.read((char *)&file_object,sizeof(file_object));
+        fin.read((char *)&file_object, sizeof(file_object));
-	bytes += sizeof(file_object);
+        bytes += sizeof(file_object);
-	if(ieee32big) be32toh_v((void *)&file_object,sizeof(file_object));
+        if (ieee32big) be32toh_v((void *)&file_object, sizeof(file_object));
-	if(ieee32)    le32toh_v((void *)&file_object,sizeof(file_object));
+        if (ieee32) le32toh_v((void *)&file_object, sizeof(file_object));
-	if(ieee64big) be64toh_v((void *)&file_object,sizeof(file_object));
+        if (ieee64big) be64toh_v((void *)&file_object, sizeof(file_object));
-	if(ieee64)    le64toh_v((void *)&file_object,sizeof(file_object));
+        if (ieee64) le64toh_v((void *)&file_object, sizeof(file_object));
 
-	munge(file_object,munged,csum);
+        munge(file_object, munged, csum);
       }
       // The boss who read the file has their value poked
       pokeSite(munged,Umu,site);
     }}}}
     timer.Stop();
     std::cout<<GridLogPerformance<<"readObjectSerial: read "<< bytes <<" bytes in "<<timer.Elapsed() <<" "
-	     << (double)bytes/ (double)timer.useconds() <<" MB/s "  <<std::endl;
+       << (double)bytes/ (double)timer.useconds() <<" MB/s "  <<std::endl;
 
     return csum;
   }
@@ -254,20 +254,20 @@ class BinaryIO {
 
       
       if ( grid->IsBoss() ) {
-	
+  
-	if(ieee32big) htobe32_v((void *)&file_object,sizeof(file_object));
+  if(ieee32big) htobe32_v((void *)&file_object,sizeof(file_object));
-	if(ieee32)    htole32_v((void *)&file_object,sizeof(file_object));
+  if(ieee32)    htole32_v((void *)&file_object,sizeof(file_object));
-	if(ieee64big) htobe64_v((void *)&file_object,sizeof(file_object));
+  if(ieee64big) htobe64_v((void *)&file_object,sizeof(file_object));
-	if(ieee64)    htole64_v((void *)&file_object,sizeof(file_object));
+  if(ieee64)    htole64_v((void *)&file_object,sizeof(file_object));
 
-	// NB could gather an xstrip as an optimisation.
+  // NB could gather an xstrip as an optimisation.
-	fout.write((char *)&file_object,sizeof(file_object));
+  fout.write((char *)&file_object,sizeof(file_object));
-	bytes+=sizeof(file_object);
+  bytes+=sizeof(file_object);
       }
     }}}}
     timer.Stop();
     std::cout<<GridLogPerformance<<"writeObjectSerial: wrote "<< bytes <<" bytes in "<<timer.Elapsed() <<" "
-	     << (double)bytes/timer.useconds() <<" MB/s "  <<std::endl;
+       << (double)bytes/timer.useconds() <<" MB/s "  <<std::endl;
 
     return csum;
   }
@@ -305,15 +305,15 @@ class BinaryIO {
       int l_idx=parallel.generator_idx(o_idx,i_idx);
 
       if( rank == grid->ThisRank() ){
-	//	std::cout << "rank" << rank<<" Getting state for index "<<l_idx<<std::endl;
+  //  std::cout << "rank" << rank<<" Getting state for index "<<l_idx<<std::endl;
-	parallel.GetState(saved,l_idx);
+  parallel.GetState(saved,l_idx);
       }
 
       grid->Broadcast(rank,(void *)&saved[0],bytes);
 
       if ( grid->IsBoss() ) {
-	Uint32Checksum((uint32_t *)&saved[0],bytes,csum);
+  Uint32Checksum((uint32_t *)&saved[0],bytes,csum);
-	fout.write((char *)&saved[0],bytes);
+  fout.write((char *)&saved[0],bytes);
       }
 
     }
@@ -355,14 +355,14 @@ class BinaryIO {
       int l_idx=parallel.generator_idx(o_idx,i_idx);
 
       if ( grid->IsBoss() ) {
-	fin.read((char *)&saved[0],bytes);
+  fin.read((char *)&saved[0],bytes);
-	Uint32Checksum((uint32_t *)&saved[0],bytes,csum);
+  Uint32Checksum((uint32_t *)&saved[0],bytes,csum);
       }
 
       grid->Broadcast(0,(void *)&saved[0],bytes);
 
       if( rank == grid->ThisRank() ){
-	parallel.SetState(saved,l_idx);
+  parallel.SetState(saved,l_idx);
       }
 
     }
@@ -415,15 +415,15 @@ class BinaryIO {
 
       if ( d == 0 ) parallel[d] = 0;
       if (parallel[d]) {
-	range[d] = grid->_ldimensions[d];
+  range[d] = grid->_ldimensions[d];
-	start[d] = grid->_processor_coor[d]*range[d];
+  start[d] = grid->_processor_coor[d]*range[d];
-	ioproc[d]= grid->_processor_coor[d];
+  ioproc[d]= grid->_processor_coor[d];
       } else {
-	range[d] = grid->_gdimensions[d];
+  range[d] = grid->_gdimensions[d];
-	start[d] = 0;
+  start[d] = 0;
-	ioproc[d]= 0;
+  ioproc[d]= 0;
 
-	if ( grid->_processor_coor[d] != 0 ) IOnode = 0;
+  if ( grid->_processor_coor[d] != 0 ) IOnode = 0;
       }
       slice_vol = slice_vol * range[d];
     }
@@ -434,9 +434,9 @@ class BinaryIO {
       std::cout<< std::dec ;
       std::cout<< GridLogMessage<< "Parallel read I/O to "<< file << " with " <<tmp<< " IOnodes for subslice ";
       for(int d=0;d<grid->_ndimension;d++){
-	std::cout<< range[d];
+  std::cout<< range[d];
-	if( d< grid->_ndimension-1 ) 
+  if( d< grid->_ndimension-1 ) 
-	  std::cout<< " x ";
+    std::cout<< " x ";
       }
       std::cout << std::endl;
     }
@@ -463,7 +463,7 @@ class BinaryIO {
 
       // need to implement these loops in Nd independent way with a lexico conversion
     for(int tlex=0;tlex<slice_vol;tlex++){
-	
+  
       std::vector<int> tsite(nd); // temporary mixed up site
       std::vector<int> gsite(nd);
       std::vector<int> lsite(nd);
@@ -472,8 +472,8 @@ class BinaryIO {
       Lexicographic::CoorFromIndex(tsite,tlex,range);
 
       for(int d=0;d<nd;d++){
-	lsite[d] = tsite[d]%grid->_ldimensions[d];  // local site
+  lsite[d] = tsite[d]%grid->_ldimensions[d];  // local site
-	gsite[d] = tsite[d]+start[d];               // global site
+  gsite[d] = tsite[d]+start[d];               // global site
       }
 
       /////////////////////////
@@ -487,29 +487,29 @@ class BinaryIO {
       // iorank reads from the seek
       ////////////////////////////////
       if (myrank == iorank) {
-	
+  
-	fin.seekg(offset+g_idx*sizeof(fileObj));
+  fin.seekg(offset+g_idx*sizeof(fileObj));
-	fin.read((char *)&fileObj,sizeof(fileObj));
+  fin.read((char *)&fileObj,sizeof(fileObj));
-	bytes+=sizeof(fileObj);
+  bytes+=sizeof(fileObj);
-	
+  
-	if(ieee32big) be32toh_v((void *)&fileObj,sizeof(fileObj));
+  if(ieee32big) be32toh_v((void *)&fileObj,sizeof(fileObj));
-	if(ieee32)    le32toh_v((void *)&fileObj,sizeof(fileObj));
+  if(ieee32)    le32toh_v((void *)&fileObj,sizeof(fileObj));
-	if(ieee64big) be64toh_v((void *)&fileObj,sizeof(fileObj));
+  if(ieee64big) be64toh_v((void *)&fileObj,sizeof(fileObj));
-	if(ieee64)    le64toh_v((void *)&fileObj,sizeof(fileObj));
+  if(ieee64)    le64toh_v((void *)&fileObj,sizeof(fileObj));
-	
+  
-	munge(fileObj,siteObj,csum);
+  munge(fileObj,siteObj,csum);
 
-      }	
+      } 
 
       // Possibly do transport through pt2pt 
       if ( rank != iorank ) { 
-	if ( (myrank == rank) || (myrank==iorank) ) {
+  if ( (myrank == rank) || (myrank==iorank) ) {
-	  grid->SendRecvPacket((void *)&siteObj,(void *)&siteObj,iorank,rank,sizeof(siteObj));
+    grid->SendRecvPacket((void *)&siteObj,(void *)&siteObj,iorank,rank,sizeof(siteObj));
-	}
+  }
       }
       // Poke at destination
       if ( myrank == rank ) {
-	  pokeLocalSite(siteObj,Umu,lsite);
+    pokeLocalSite(siteObj,Umu,lsite);
       }
       grid->Barrier(); // necessary?
     }
@@ -520,7 +520,7 @@ class BinaryIO {
 
     timer.Stop();
     std::cout<<GridLogPerformance<<"readObjectParallel: read "<< bytes <<" bytes in "<<timer.Elapsed() <<" "
-	     << (double)bytes/timer.useconds() <<" MB/s "  <<std::endl;
+       << (double)bytes/timer.useconds() <<" MB/s "  <<std::endl;
     
     return csum;
   }
@@ -558,15 +558,15 @@ class BinaryIO {
       if ( d!= grid->_ndimension-1 ) parallel[d] = 0;
 
       if (parallel[d]) {
-	range[d] = grid->_ldimensions[d];
+  range[d] = grid->_ldimensions[d];
-	start[d] = grid->_processor_coor[d]*range[d];
+  start[d] = grid->_processor_coor[d]*range[d];
-	ioproc[d]= grid->_processor_coor[d];
+  ioproc[d]= grid->_processor_coor[d];
       } else {
-	range[d] = grid->_gdimensions[d];
+  range[d] = grid->_gdimensions[d];
-	start[d] = 0;
+  start[d] = 0;
-	ioproc[d]= 0;
+  ioproc[d]= 0;
 
-	if ( grid->_processor_coor[d] != 0 ) IOnode = 0;
+  if ( grid->_processor_coor[d] != 0 ) IOnode = 0;
       }
 
       slice_vol = slice_vol * range[d];
@@ -577,9 +577,9 @@ class BinaryIO {
       grid->GlobalSum(tmp);
       std::cout<< GridLogMessage<< "Parallel write I/O from "<< file << " with " <<tmp<< " IOnodes for subslice ";
       for(int d=0;d<grid->_ndimension;d++){
-	std::cout<< range[d];
+  std::cout<< range[d];
-	if( d< grid->_ndimension-1 ) 
+  if( d< grid->_ndimension-1 ) 
-	  std::cout<< " x ";
+    std::cout<< " x ";
       }
       std::cout << std::endl;
     }
@@ -610,7 +610,7 @@ class BinaryIO {
     // should aggregate a whole chunk and then write.
     // need to implement these loops in Nd independent way with a lexico conversion
     for(int tlex=0;tlex<slice_vol;tlex++){
-	
+  
       std::vector<int> tsite(nd); // temporary mixed up site
       std::vector<int> gsite(nd);
       std::vector<int> lsite(nd);
@@ -619,8 +619,8 @@ class BinaryIO {
       Lexicographic::CoorFromIndex(tsite,tlex,range);
 
       for(int d=0;d<nd;d++){
-	lsite[d] = tsite[d]%grid->_ldimensions[d];  // local site
+  lsite[d] = tsite[d]%grid->_ldimensions[d];  // local site
-	gsite[d] = tsite[d]+start[d];               // global site
+  gsite[d] = tsite[d]+start[d];               // global site
       }
 
 
@@ -640,26 +640,26 @@ class BinaryIO {
 
       // Pair of nodes may need to do pt2pt send
       if ( rank != iorank ) { // comms is necessary
-	if ( (myrank == rank) || (myrank==iorank) ) { // and we have to do it
+  if ( (myrank == rank) || (myrank==iorank) ) { // and we have to do it
-	  // Send to IOrank 
+    // Send to IOrank 
-	  grid->SendRecvPacket((void *)&siteObj,(void *)&siteObj,rank,iorank,sizeof(siteObj));
+    grid->SendRecvPacket((void *)&siteObj,(void *)&siteObj,rank,iorank,sizeof(siteObj));
-	}
+  }
       }
 
       grid->Barrier(); // necessary?
 
       if (myrank == iorank) {
-	
+  
-	munge(siteObj,fileObj,csum);
+  munge(siteObj,fileObj,csum);
 
-	if(ieee32big) htobe32_v((void *)&fileObj,sizeof(fileObj));
+  if(ieee32big) htobe32_v((void *)&fileObj,sizeof(fileObj));
-	if(ieee32)    htole32_v((void *)&fileObj,sizeof(fileObj));
+  if(ieee32)    htole32_v((void *)&fileObj,sizeof(fileObj));
-	if(ieee64big) htobe64_v((void *)&fileObj,sizeof(fileObj));
+  if(ieee64big) htobe64_v((void *)&fileObj,sizeof(fileObj));
-	if(ieee64)    htole64_v((void *)&fileObj,sizeof(fileObj));
+  if(ieee64)    htole64_v((void *)&fileObj,sizeof(fileObj));
-	
+  
-	fout.seekp(offset+g_idx*sizeof(fileObj));
+  fout.seekp(offset+g_idx*sizeof(fileObj));
-	fout.write((char *)&fileObj,sizeof(fileObj));
+  fout.write((char *)&fileObj,sizeof(fileObj));
-	bytes+=sizeof(fileObj);
+  bytes+=sizeof(fileObj);
       }
     }
 
@@ -668,7 +668,7 @@ class BinaryIO {
 
     timer.Stop();
     std::cout<<GridLogPerformance<<"writeObjectParallel: wrote "<< bytes <<" bytes in "<<timer.Elapsed() <<" "
-	     << (double)bytes/timer.useconds() <<" MB/s "  <<std::endl;
+       << (double)bytes/timer.useconds() <<" MB/s "  <<std::endl;
 
     return csum;
   }

lib/qcd/QCD.h

@@ -55,11 +55,14 @@ namespace QCD {
     //////////////////////////////////////////////////////////////////////////////
     // QCD iMatrix types
     // Index conventions:                            Lorentz x Spin x Colour
+    // note: static const int or constexpr will work for type deductions
+    //       with the intel compiler (up to version 17)
     //////////////////////////////////////////////////////////////////////////////
-    static const int ColourIndex = 2;
+    #define ColourIndex  2
-    static const int SpinIndex   = 1;
+    #define SpinIndex    1
-    static const int LorentzIndex= 0;
+    #define LorentzIndex 0
 
+  
     // Also should make these a named enum type
     static const int DaggerNo=0;
     static const int DaggerYes=1;
@@ -490,16 +493,27 @@ namespace QCD {
 }   //namespace QCD
 } // Grid
 
+
 #include <Grid/qcd/utils/SpaceTimeGrid.h>
 #include <Grid/qcd/spin/Dirac.h>
 #include <Grid/qcd/spin/TwoSpinor.h>
 #include <Grid/qcd/utils/LinalgUtils.h>
 #include <Grid/qcd/utils/CovariantCshift.h>
+
+// Include representations 	
 #include <Grid/qcd/utils/SUn.h>
+#include <Grid/qcd/utils/SUnAdjoint.h>
+#include <Grid/qcd/utils/SUnTwoIndex.h>
+#include <Grid/qcd/representations/hmc_types.h>
+
 #include <Grid/qcd/action/Actions.h>
+
+#include <Grid/qcd/smearing/Smearing.h>
+
 #include <Grid/qcd/hmc/integrators/Integrator.h>
 #include <Grid/qcd/hmc/integrators/Integrator_algorithm.h>
 #include <Grid/qcd/hmc/HMC.h>
-#include <Grid/qcd/smearing/Smearing.h>
+
+
 
 #endif

lib/qcd/action/ActionBase.h

@@ -1,87 +1,153 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./lib/qcd/action/ActionBase.h
+Source file: ./lib/qcd/action/ActionBase.h
 
-    Copyright (C) 2015
+Copyright (C) 2015
 
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: neo <cossu@post.kek.jp>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution
-    *************************************************************************************/
+directory
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef QCD_ACTION_BASE
 #define QCD_ACTION_BASE
 namespace Grid {
-namespace QCD{
+namespace QCD {
-
-template<class GaugeField>
-class Action { 
 
+template <class GaugeField>
+class Action {
  public:
   bool is_smeared = false;
   // Boundary conditions? // Heatbath?
-  virtual void  refresh(const GaugeField &U, GridParallelRNG& pRNG) = 0;// refresh pseudofermions
+  virtual void refresh(const GaugeField& U,
-  virtual RealD S    (const GaugeField &U)                        = 0;  // evaluate the action
+                       GridParallelRNG& pRNG) = 0;  // refresh pseudofermions
-  virtual void  deriv(const GaugeField &U,GaugeField & dSdU )     = 0;  // evaluate the action derivative
+  virtual RealD S(const GaugeField& U) = 0;         // evaluate the action
-  virtual ~Action() {};
+  virtual void deriv(const GaugeField& U,
+                     GaugeField& dSdU) = 0;  // evaluate the action derivative
+  virtual ~Action(){};
+};
+
+// Indexing of tuple types
+template <class T, class Tuple>
+struct Index;
+
+template <class T, class... Types>
+struct Index<T, std::tuple<T, Types...>> {
+  static const std::size_t value = 0;
+};
+
+template <class T, class U, class... Types>
+struct Index<T, std::tuple<U, Types...>> {
+  static const std::size_t value = 1 + Index<T, std::tuple<Types...>>::value;
 };
 
-// Could derive PseudoFermion action with a PF field, FermionField, and a Grid; implement refresh
 /*
-template<class GaugeField, class FermionField>
+template <class GaugeField>
-class PseudoFermionAction : public Action<GaugeField> {
+struct ActionLevel {
  public:
-  FermionField Phi;
+  typedef Action<GaugeField>*
-  GridParallelRNG &pRNG;
+      ActPtr;  // now force the same colours as the rest of the code
-  GridBase &Grid;
 
-  PseudoFermionAction(GridBase &_Grid,GridParallelRNG &_pRNG) : Grid(_Grid), Phi(&_Grid), pRNG(_pRNG) {
+  //Add supported representations here
-  };
 
-  virtual void refresh(const GaugeField &gauge) {
-    gaussian(Phi,pRNG);
-  };
 
-};
+  unsigned int multiplier;
-*/
-
-template<class GaugeField> struct ActionLevel{
-public:
-   
-  typedef Action<GaugeField>*  ActPtr; // now force the same colours as the rest of the code
-
-  int multiplier;
 
   std::vector<ActPtr> actions;
 
-  ActionLevel(int mul = 1) : multiplier(mul) {
+  ActionLevel(unsigned int mul = 1) : actions(0), multiplier(mul) {
-    assert (mul > 0);
+    assert(mul >= 1);
   };
-   
+
-  void push_back(ActPtr ptr){
+  void push_back(ActPtr ptr) { actions.push_back(ptr); }
-    actions.push_back(ptr);
+};
+*/
+
+template <class GaugeField, class Repr = NoHirep >
+struct ActionLevel {
+ public:
+  unsigned int multiplier; 
+
+  // Fundamental repr actions separated because of the smearing
+  typedef Action<GaugeField>* ActPtr;
+
+  // construct a tuple of vectors of the actions for the corresponding higher
+  // representation fields
+  typedef typename AccessTypes<Action, Repr>::VectorCollection action_collection;
+  action_collection actions_hirep;
+  typedef typename  AccessTypes<Action, Repr>::FieldTypeCollection action_hirep_types;
+
+  std::vector<ActPtr>& actions;
+
+  // Temporary conversion between ActionLevel and ActionLevelHirep
+  //ActionLevelHirep(ActionLevel<GaugeField>& AL ):actions(AL.actions), multiplier(AL.multiplier){}
+
+  ActionLevel(unsigned int mul = 1) : actions(std::get<0>(actions_hirep)), multiplier(mul) {
+    // initialize the hirep vectors to zero.
+    //apply(this->resize, actions_hirep, 0); //need a working resize
+    assert(mul >= 1);
+  };
+
+  //void push_back(ActPtr ptr) { actions.push_back(ptr); }
+
+
+
+  template < class Field >
+  void push_back(Action<Field>* ptr) {
+    // insert only in the correct vector
+    std::get< Index < Field, action_hirep_types>::value >(actions_hirep).push_back(ptr);
+  };
+
+ 
+
+  template < class ActPtr>
+  static void resize(ActPtr ap, unsigned int n){
+    ap->resize(n);
+
   }
+
+  //template <std::size_t I>
+  //auto getRepresentation(Repr& R)->decltype(std::get<I>(R).U)  {return std::get<I>(R).U;}
+
+  // Loop on tuple for a callable function
+  template <std::size_t I = 1, typename Callable, typename ...Args>
+  inline typename std::enable_if<I == std::tuple_size<action_collection>::value, void>::type apply(
+      Callable, Repr& R,Args&...) const {}
+
+  template <std::size_t I = 1, typename Callable, typename ...Args>
+  inline typename std::enable_if<I < std::tuple_size<action_collection>::value, void>::type apply(
+      Callable fn, Repr& R, Args&... arguments) const {
+    fn(std::get<I>(actions_hirep), std::get<I>(R.rep), arguments...);
+    apply<I + 1>(fn, R, arguments...);
+  }  
+
 };
 
-template<class GaugeField> using ActionSet = std::vector<ActionLevel< GaugeField > >;
 
+//template <class GaugeField>
+//using ActionSet = std::vector<ActionLevel<GaugeField> >;
 
-}}
+template <class GaugeField, class R>
+using ActionSet = std::vector<ActionLevel<GaugeField, R> >;
+
+}
+}
 #endif

lib/qcd/action/Actions.h

@@ -116,6 +116,14 @@ typedef SymanzikGaugeAction<ConjugateGimplD>        ConjugateSymanzikGaugeAction
   template class A<GparityWilsonImplF>;		\
   template class A<GparityWilsonImplD>;		
 
+#define AdjointFermOpTemplateInstantiate(A) \
+  template class A<WilsonAdjImplF>; \
+  template class A<WilsonAdjImplD>; 
+
+#define TwoIndexFermOpTemplateInstantiate(A) \
+  template class A<WilsonTwoIndexSymmetricImplF>; \
+  template class A<WilsonTwoIndexSymmetricImplD>; 
+
 #define FermOp5dVecTemplateInstantiate(A) \
   template class A<DomainWallVec5dImplF>;	\
   template class A<DomainWallVec5dImplD>;	\
@@ -126,6 +134,7 @@ typedef SymanzikGaugeAction<ConjugateGimplD>        ConjugateSymanzikGaugeAction
  FermOp4dVecTemplateInstantiate(A) \
  FermOp5dVecTemplateInstantiate(A) 
 
+
 #define GparityFermOpTemplateInstantiate(A) 
 
 ////////////////////////////////////////////
@@ -171,6 +180,14 @@ typedef WilsonFermion<WilsonImplR> WilsonFermionR;
 typedef WilsonFermion<WilsonImplF> WilsonFermionF;
 typedef WilsonFermion<WilsonImplD> WilsonFermionD;
 
+typedef WilsonFermion<WilsonAdjImplR> WilsonAdjFermionR;
+typedef WilsonFermion<WilsonAdjImplF> WilsonAdjFermionF;
+typedef WilsonFermion<WilsonAdjImplD> WilsonAdjFermionD;
+
+typedef WilsonFermion<WilsonTwoIndexSymmetricImplR> WilsonTwoIndexSymmetricFermionR;
+typedef WilsonFermion<WilsonTwoIndexSymmetricImplF> WilsonTwoIndexSymmetricFermionF;
+typedef WilsonFermion<WilsonTwoIndexSymmetricImplD> WilsonTwoIndexSymmetricFermionD;
+
 typedef WilsonTMFermion<WilsonImplR> WilsonTMFermionR;
 typedef WilsonTMFermion<WilsonImplF> WilsonTMFermionF;
 typedef WilsonTMFermion<WilsonImplD> WilsonTMFermionD;

lib/qcd/action/fermion/FermionOperatorImpl.h

@@ -1,35 +1,36 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./lib/qcd/action/fermion/FermionOperatorImpl.h
+Source file: ./lib/qcd/action/fermion/FermionOperatorImpl.h
 
-    Copyright (C) 2015
+Copyright (C) 2015
 
 Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
 Author: paboyle <paboyle@ph.ed.ac.uk>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution
-    *************************************************************************************/
+directory
-    /*  END LEGAL */
+*************************************************************************************/
-#ifndef  GRID_QCD_FERMION_OPERATOR_IMPL_H
+/*  END LEGAL */
-#define  GRID_QCD_FERMION_OPERATOR_IMPL_H
+#ifndef GRID_QCD_FERMION_OPERATOR_IMPL_H
+#define GRID_QCD_FERMION_OPERATOR_IMPL_H
 
 namespace Grid {
 
@@ -99,258 +100,281 @@ namespace Grid {
     typedef typename Impl::SiteSpinor               SiteSpinor;		\
     typedef typename Impl::SiteHalfSpinor       SiteHalfSpinor;		\
     typedef typename Impl::Compressor               Compressor;		\
-    typedef typename Impl::StencilImpl             StencilImpl;	  \
+    typedef typename Impl::StencilImpl             StencilImpl;		\
-    typedef typename Impl::ImplParams ImplParams; \
+    typedef typename Impl::ImplParams ImplParams;			\
     typedef typename Impl::Coeff_t       Coeff_t;
 
 #define INHERIT_IMPL_TYPES(Base) \
-    INHERIT_GIMPL_TYPES(Base)\
+    INHERIT_GIMPL_TYPES(Base)	 \
     INHERIT_FIMPL_TYPES(Base)
-
+    
     ///////
     // Single flavour four spinors with colour index
     ///////
-    template<class S,int Nrepresentation=Nc,class _Coeff_t = RealD>
+    template <class S, class Representation = FundamentalRepresentation,class _Coeff_t = RealD >
-    class WilsonImpl :  public PeriodicGaugeImpl< GaugeImplTypes< S, Nrepresentation> > { 
+    class WilsonImpl
+      : public PeriodicGaugeImpl<GaugeImplTypes<S, Representation::Dimension > > {
     public:
-
+      static const int Dimension = Representation::Dimension;
+      typedef PeriodicGaugeImpl<GaugeImplTypes<S, Dimension > > Gimpl;
+      
+      //Necessary?
+      constexpr bool is_fundamental() const{return Dimension == Nc ? 1 : 0;}
 
       const bool LsVectorised=false;
-
       typedef _Coeff_t Coeff_t;
-      typedef PeriodicGaugeImpl< GaugeImplTypes< S,Nrepresentation> > Gimpl;
+
 
       INHERIT_GIMPL_TYPES(Gimpl);
-
+      
-      template<typename vtype> using iImplSpinor             = iScalar<iVector<iVector<vtype, Nrepresentation>, Ns> >;
+      template <typename vtype> using iImplSpinor            = iScalar<iVector<iVector<vtype, Dimension>, Ns> >;
-      template<typename vtype> using iImplHalfSpinor         = iScalar<iVector<iVector<vtype, Nrepresentation>, Nhs> >;
+      template <typename vtype> using iImplHalfSpinor        = iScalar<iVector<iVector<vtype, Dimension>, Nhs> >;
-      template<typename vtype> using iImplDoubledGaugeField  = iVector<iScalar<iMatrix<vtype, Nrepresentation> >, Nds >;
+      template <typename vtype> using iImplDoubledGaugeField = iVector<iScalar<iMatrix<vtype, Dimension> >, Nds>;
-    
+      
-      typedef iImplSpinor    <Simd>           SiteSpinor;
+      typedef iImplSpinor<Simd>            SiteSpinor;
-      typedef iImplHalfSpinor<Simd>           SiteHalfSpinor;
+      typedef iImplHalfSpinor<Simd>        SiteHalfSpinor;
-      typedef iImplDoubledGaugeField<Simd>    SiteDoubledGaugeField;
+      typedef iImplDoubledGaugeField<Simd> SiteDoubledGaugeField;
-
+      
-      typedef Lattice<SiteSpinor>                 FermionField;
+      typedef Lattice<SiteSpinor>            FermionField;
       typedef Lattice<SiteDoubledGaugeField> DoubledGaugeField;
-
+      
-      typedef WilsonCompressor<SiteHalfSpinor,SiteSpinor> Compressor;
+      typedef WilsonCompressor<SiteHalfSpinor, SiteSpinor> Compressor;
       typedef WilsonImplParams ImplParams;
-      typedef WilsonStencil<SiteSpinor,SiteHalfSpinor> StencilImpl;
+      typedef WilsonStencil<SiteSpinor, SiteHalfSpinor> StencilImpl;
-
+      
       ImplParams Params;
-
+      
-      WilsonImpl(const ImplParams &p= ImplParams()) : Params(p) {}; 
+      WilsonImpl(const ImplParams &p = ImplParams()) : Params(p){};
-
+      
       bool overlapCommsCompute(void) { return Params.overlapCommsCompute; };
-    
+      
-      inline void multLink(SiteHalfSpinor &phi,const SiteDoubledGaugeField &U,const SiteHalfSpinor &chi,int mu,StencilEntry *SE,StencilImpl &St){
+      inline void multLink(SiteHalfSpinor &phi,
-        mult(&phi(),&U(mu),&chi());
+			   const SiteDoubledGaugeField &U,
-      }
+			   const SiteHalfSpinor &chi,
-
+			   int mu,
-      template<class ref>
+			   StencilEntry *SE,
-      inline void loadLinkElement(Simd & reg,ref &memory){
+			   StencilImpl &St) {
-	reg = memory;
+	mult(&phi(), &U(mu), &chi());
-      }
-      inline void DoubleStore(GridBase *GaugeGrid,DoubledGaugeField &Uds,const GaugeField &Umu)
-      {
-        conformable(Uds._grid,GaugeGrid);
-        conformable(Umu._grid,GaugeGrid);
-        GaugeLinkField U(GaugeGrid);
-        for(int mu=0;mu<Nd;mu++){
-  	  U = PeekIndex<LorentzIndex>(Umu,mu);
-	  PokeIndex<LorentzIndex>(Uds,U,mu);
-	  U = adj(Cshift(U,mu,-1));
-	  PokeIndex<LorentzIndex>(Uds,U,mu+4);
-	}
       }
       
+      template <class ref>
+      inline void loadLinkElement(Simd &reg,
+				  ref &memory) {
+	reg = memory;
+      }
+      
+      inline void DoubleStore(GridBase *GaugeGrid,
+			      DoubledGaugeField &Uds,
+			      const GaugeField &Umu) {
+	conformable(Uds._grid, GaugeGrid);
+	conformable(Umu._grid, GaugeGrid);
+	GaugeLinkField U(GaugeGrid);
+	for (int mu = 0; mu < Nd; mu++) {
+	  U = PeekIndex<LorentzIndex>(Umu, mu);
+	  PokeIndex<LorentzIndex>(Uds, U, mu);
+	  U = adj(Cshift(U, mu, -1));
+	  PokeIndex<LorentzIndex>(Uds, U, mu + 4);
+	}
+      }
+
       inline void InsertForce4D(GaugeField &mat, FermionField &Btilde, FermionField &A,int mu){
 	GaugeLinkField link(mat._grid);
 	link = TraceIndex<SpinIndex>(outerProduct(Btilde,A)); 
 	PokeIndex<LorentzIndex>(mat,link,mu);
       }   
-
+      
       inline void InsertForce5D(GaugeField &mat, FermionField &Btilde, FermionField &Atilde,int mu){
-
+	
 	int Ls=Btilde._grid->_fdimensions[0];
-
 	GaugeLinkField tmp(mat._grid);
 	tmp = zero;
-PARALLEL_FOR_LOOP
+
-	for(int sss=0;sss<tmp._grid->oSites();sss++){
+        PARALLEL_FOR_LOOP
-	  int sU=sss;
+	  for(int sss=0;sss<tmp._grid->oSites();sss++){
-	  for(int s=0;s<Ls;s++){
+	    int sU=sss;
-	    int sF = s+Ls*sU;
+	    for(int s=0;s<Ls;s++){
-	    tmp[sU] = tmp[sU]+ traceIndex<SpinIndex>(outerProduct(Btilde[sF],Atilde[sF])); // ordering here
+	      int sF = s+Ls*sU;
+	      tmp[sU] = tmp[sU]+ traceIndex<SpinIndex>(outerProduct(Btilde[sF],Atilde[sF])); // ordering here
+	    }
 	  }
-	}
 	PokeIndex<LorentzIndex>(mat,tmp,mu);
-	
+
       }
-
     };
 
-
-
     ///////
     // Single flavour four spinors with colour index, 5d redblack
     ///////
     template<class S,int Nrepresentation=Nc,class _Coeff_t = RealD>
     class DomainWallVec5dImpl :  public PeriodicGaugeImpl< GaugeImplTypes< S,Nrepresentation> > { 
     public:
-    
+      
+      static const int Dimension = Nrepresentation;
       const bool LsVectorised=true;
-
+      typedef _Coeff_t Coeff_t;      
-      typedef _Coeff_t Coeff_t;
+      typedef PeriodicGaugeImpl<GaugeImplTypes<S, Nrepresentation> > Gimpl;
-      typedef PeriodicGaugeImpl< GaugeImplTypes< S,Nrepresentation> > Gimpl;
 
       INHERIT_GIMPL_TYPES(Gimpl);
       
-      template<typename vtype> using iImplSpinor             = iScalar<iVector<iVector<vtype, Nrepresentation>, Ns> >;
+      template <typename vtype> using iImplSpinor            = iScalar<iVector<iVector<vtype, Nrepresentation>, Ns> >;
-      template<typename vtype> using iImplHalfSpinor         = iScalar<iVector<iVector<vtype, Nrepresentation>, Nhs> >;
+      template <typename vtype> using iImplHalfSpinor        = iScalar<iVector<iVector<vtype, Nrepresentation>, Nhs> >;
-      template<typename vtype> using iImplDoubledGaugeField  = iVector<iScalar<iMatrix<vtype, Nrepresentation> >, Nds >;
+      template <typename vtype> using iImplDoubledGaugeField = iVector<iScalar<iMatrix<vtype, Nrepresentation> >, Nds>;
-      template<typename vtype> using iImplGaugeField         = iVector<iScalar<iMatrix<vtype, Nrepresentation> >, Nd >;
+      template <typename vtype> using iImplGaugeField        = iVector<iScalar<iMatrix<vtype, Nrepresentation> >, Nd>;
-      template<typename vtype> using iImplGaugeLink          = iScalar<iScalar<iMatrix<vtype, Nrepresentation> > >;
+      template <typename vtype> using iImplGaugeLink         = iScalar<iScalar<iMatrix<vtype, Nrepresentation> > >;
-    
+      
-      typedef iImplSpinor    <Simd>           SiteSpinor;
+      typedef iImplSpinor<Simd> SiteSpinor;
-      typedef iImplHalfSpinor<Simd>           SiteHalfSpinor;
+      typedef iImplHalfSpinor<Simd> SiteHalfSpinor;
-      typedef Lattice<SiteSpinor>             FermionField;
+      typedef Lattice<SiteSpinor> FermionField;
-
+      
       // Make the doubled gauge field a *scalar*
-      typedef iImplDoubledGaugeField<typename Simd::scalar_type>    SiteDoubledGaugeField; // This is a scalar
+      typedef iImplDoubledGaugeField<typename Simd::scalar_type>
-      typedef iImplGaugeField<typename Simd::scalar_type>           SiteScalarGaugeField;  // scalar
+      SiteDoubledGaugeField;  // This is a scalar
-      typedef iImplGaugeLink <typename Simd::scalar_type>           SiteScalarGaugeLink;   // scalar
+      typedef iImplGaugeField<typename Simd::scalar_type>
-
+      SiteScalarGaugeField;  // scalar
-      typedef Lattice<SiteDoubledGaugeField>                  DoubledGaugeField;
+      typedef iImplGaugeLink<typename Simd::scalar_type>
-
+      SiteScalarGaugeLink;  // scalar
-      typedef WilsonCompressor<SiteHalfSpinor,SiteSpinor> Compressor;
+      
+      typedef Lattice<SiteDoubledGaugeField> DoubledGaugeField;
+      
+      typedef WilsonCompressor<SiteHalfSpinor, SiteSpinor> Compressor;
       typedef WilsonImplParams ImplParams;
-      typedef WilsonStencil<SiteSpinor,SiteHalfSpinor> StencilImpl;
+      typedef WilsonStencil<SiteSpinor, SiteHalfSpinor> StencilImpl;
-
+      
       ImplParams Params;
-
+      
-      DomainWallVec5dImpl(const ImplParams &p= ImplParams()) : Params(p) {}; 
+      DomainWallVec5dImpl(const ImplParams &p = ImplParams()) : Params(p){};
-
+      
       bool overlapCommsCompute(void) { return false; };
-    
+      
-      template<class ref>
+      template <class ref>
-      inline void loadLinkElement(Simd & reg,ref &memory){
+      inline void loadLinkElement(Simd &reg, ref &memory) {
-	vsplat(reg,memory);
+	vsplat(reg, memory);
       }
-      inline void multLink(SiteHalfSpinor &phi,const SiteDoubledGaugeField &U,const SiteHalfSpinor &chi,int mu,StencilEntry *SE,StencilImpl &St)
+      inline void multLink(SiteHalfSpinor &phi, const SiteDoubledGaugeField &U,
-      {
+			   const SiteHalfSpinor &chi, int mu, StencilEntry *SE,
+			   StencilImpl &St) {
 	SiteGaugeLink UU;
-	for(int i=0;i<Nrepresentation;i++){
+	for (int i = 0; i < Nrepresentation; i++) {
-	  for(int j=0;j<Nrepresentation;j++){
+	  for (int j = 0; j < Nrepresentation; j++) {
-	    vsplat(UU()()(i,j),U(mu)()(i,j));
+	    vsplat(UU()()(i, j), U(mu)()(i, j));
 	  }
 	}
-        mult(&phi(),&UU(),&chi());
+	mult(&phi(), &UU(), &chi());
       }
-
+      
-      inline void DoubleStore(GridBase *GaugeGrid,DoubledGaugeField &Uds,const GaugeField &Umu)
+      inline void DoubleStore(GridBase *GaugeGrid, DoubledGaugeField &Uds,
-      {
+			      const GaugeField &Umu) {
-	SiteScalarGaugeField  ScalarUmu;
+	SiteScalarGaugeField ScalarUmu;
 	SiteDoubledGaugeField ScalarUds;
-
-        GaugeLinkField U   (Umu._grid);
-	GaugeField     Uadj(Umu._grid);
-        for(int mu=0;mu<Nd;mu++){
-  	  U = PeekIndex<LorentzIndex>(Umu,mu);
-	  U = adj(Cshift(U,mu,-1));
-	  PokeIndex<LorentzIndex>(Uadj,U,mu);
-	}
-
-	for(int lidx=0;lidx<GaugeGrid->lSites();lidx++){
-	  std::vector<int> lcoor;
-	  GaugeGrid->LocalIndexToLocalCoor(lidx,lcoor);
-
-	  peekLocalSite(ScalarUmu,Umu,lcoor);
-	  for(int mu=0;mu<4;mu++) ScalarUds(mu) = ScalarUmu(mu);
-
-	  peekLocalSite(ScalarUmu,Uadj,lcoor);
-	  for(int mu=0;mu<4;mu++) ScalarUds(mu+4) = ScalarUmu(mu);
-
-	  pokeLocalSite(ScalarUds,Uds,lcoor);
-	}
-
-      }
 	
-      inline void InsertForce4D(GaugeField &mat, FermionField &Btilde, FermionField &A,int mu){
+	GaugeLinkField U(Umu._grid);
-	assert(0);
+	GaugeField Uadj(Umu._grid);
-      }   
+	for (int mu = 0; mu < Nd; mu++) {
-
+	  U = PeekIndex<LorentzIndex>(Umu, mu);
-      inline void InsertForce5D(GaugeField &mat, FermionField &Btilde, FermionField &Atilde,int mu){
+	  U = adj(Cshift(U, mu, -1));
+	  PokeIndex<LorentzIndex>(Uadj, U, mu);
+	}
+	
+	for (int lidx = 0; lidx < GaugeGrid->lSites(); lidx++) {
+	  std::vector<int> lcoor;
+	  GaugeGrid->LocalIndexToLocalCoor(lidx, lcoor);
+	  
+	  peekLocalSite(ScalarUmu, Umu, lcoor);
+	  for (int mu = 0; mu < 4; mu++) ScalarUds(mu) = ScalarUmu(mu);
+	  
+	  peekLocalSite(ScalarUmu, Uadj, lcoor);
+	  for (int mu = 0; mu < 4; mu++) ScalarUds(mu + 4) = ScalarUmu(mu);
+	  
+	  pokeLocalSite(ScalarUds, Uds, lcoor);
+	}
+      }
+      
+      inline void InsertForce4D(GaugeField &mat, FermionField &Btilde,
+				FermionField &A, int mu) {
+	assert(0);
+      }
+      
+      inline void InsertForce5D(GaugeField &mat, FermionField &Btilde,
+				FermionField &Atilde, int mu) {
 	assert(0);
       }
-
     };
-
+    
-
     ////////////////////////////////////////////////////////////////////////////////////////
     // Flavour doubled spinors; is Gparity the only? what about C*?
     ////////////////////////////////////////////////////////////////////////////////////////
-
+    
-    template<class S,int Nrepresentation,class _Coeff_t = RealD>
+    template <class S, int Nrepresentation,class _Coeff_t = RealD>
-    class GparityWilsonImpl : public ConjugateGaugeImpl< GaugeImplTypes<S,Nrepresentation> >{ 
+    class GparityWilsonImpl
+      : public ConjugateGaugeImpl<GaugeImplTypes<S, Nrepresentation> > {
     public:
+      static const int Dimension = Nrepresentation;
 
       const bool LsVectorised=false;
 
       typedef _Coeff_t Coeff_t;
       typedef ConjugateGaugeImpl< GaugeImplTypes<S,Nrepresentation> > Gimpl;
-
+      
       INHERIT_GIMPL_TYPES(Gimpl);
-
+      
-      template<typename vtype> using iImplSpinor             = iVector<iVector<iVector<vtype, Nrepresentation>, Ns>, Ngp >;
+      template <typename vtype>
-      template<typename vtype> using iImplHalfSpinor         = iVector<iVector<iVector<vtype, Nrepresentation>, Nhs>, Ngp >;
+      using iImplSpinor =
-      template<typename vtype> using iImplDoubledGaugeField  = iVector<iVector<iScalar<iMatrix<vtype, Nrepresentation> >, Nds >, Ngp >;
+      iVector<iVector<iVector<vtype, Nrepresentation>, Ns>, Ngp>;
-    
+      template <typename vtype>
-      typedef iImplSpinor    <Simd>           SiteSpinor;
+      using iImplHalfSpinor =
-      typedef iImplHalfSpinor<Simd>           SiteHalfSpinor;
+	iVector<iVector<iVector<vtype, Nrepresentation>, Nhs>, Ngp>;
-      typedef iImplDoubledGaugeField<Simd>    SiteDoubledGaugeField;
+      template <typename vtype>
-
+      using iImplDoubledGaugeField =
-      typedef Lattice<SiteSpinor>                 FermionField;
+	iVector<iVector<iScalar<iMatrix<vtype, Nrepresentation> >, Nds>, Ngp>;
+      
+      typedef iImplSpinor<Simd> SiteSpinor;
+      typedef iImplHalfSpinor<Simd> SiteHalfSpinor;
+      typedef iImplDoubledGaugeField<Simd> SiteDoubledGaugeField;
+      
+      typedef Lattice<SiteSpinor> FermionField;
       typedef Lattice<SiteDoubledGaugeField> DoubledGaugeField;
-
+      
-      typedef WilsonCompressor<SiteHalfSpinor,SiteSpinor> Compressor;
+      typedef WilsonCompressor<SiteHalfSpinor, SiteSpinor> Compressor;
-      typedef WilsonStencil<SiteSpinor,SiteHalfSpinor> StencilImpl;
+      typedef WilsonStencil<SiteSpinor, SiteHalfSpinor> StencilImpl;
 
       typedef GparityWilsonImplParams ImplParams;
-
+      
       ImplParams Params;
 
-      GparityWilsonImpl(const ImplParams &p= ImplParams()) : Params(p) {}; 
+
-      
+      GparityWilsonImpl(const ImplParams &p = ImplParams()) : Params(p){};
+
       bool overlapCommsCompute(void) { return Params.overlapCommsCompute; };
 
-      // provide the multiply by link that is differentiated between Gparity (with flavour index) and non-Gparity
+      // provide the multiply by link that is differentiated between Gparity (with
-      inline void multLink(SiteHalfSpinor &phi,const SiteDoubledGaugeField &U,const SiteHalfSpinor &chi,int mu,StencilEntry *SE,StencilImpl &St){
+      // flavour index) and non-Gparity
-
+      inline void multLink(SiteHalfSpinor &phi, const SiteDoubledGaugeField &U,
+			   const SiteHalfSpinor &chi, int mu, StencilEntry *SE,
+			   StencilImpl &St) {
 	typedef SiteHalfSpinor vobj;
 	typedef typename SiteHalfSpinor::scalar_object sobj;
-
+	
 	vobj vtmp;
 	sobj stmp;
 	
 	GridBase *grid = St._grid;
-      
+	
 	const int Nsimd = grid->Nsimd();
 	
-	int direction    = St._directions[mu];
+	int direction = St._directions[mu];
-	int distance     = St._distances[mu];
+	int distance = St._distances[mu];
-	int ptype        = St._permute_type[mu]; 
+	int ptype = St._permute_type[mu];
-	int sl           = St._grid->_simd_layout[direction];
+	int sl = St._grid->_simd_layout[direction];
-
+	
 	// Fixme X.Y.Z.T hardcode in stencil
-	int mmu          = mu % Nd;
+	int mmu = mu % Nd;
-
+	
 	// assert our assumptions
-	assert((distance==1)||(distance==-1)); // nearest neighbour stencil hard code
+	assert((distance == 1) || (distance == -1));  // nearest neighbour stencil hard code
-	assert((sl==1)||(sl==2));
+	assert((sl == 1) || (sl == 2));
 	
 	std::vector<int> icoor;
-      
+	
 	if ( SE->_around_the_world && Params.twists[mmu] ) {
 
 	  if ( sl == 2 ) {
@@ -391,7 +415,7 @@ PARALLEL_FOR_LOOP
 	  mult(&phi(1),&U(1)(mu),&chi(1));
 	}
 	
-      }
+  }
 
       inline void DoubleStore(GridBase *GaugeGrid,DoubledGaugeField &Uds,const GaugeField &Umu)
       {
@@ -404,7 +428,7 @@ PARALLEL_FOR_LOOP
 	GaugeLinkField Uconj(GaugeGrid);
 	
 	Lattice<iScalar<vInteger> > coor(GaugeGrid);
-
+	
 	
 	for(int mu=0;mu<Nd;mu++){
 	  
@@ -412,19 +436,19 @@ PARALLEL_FOR_LOOP
 	  
 	  U     = PeekIndex<LorentzIndex>(Umu,mu);
 	  Uconj = conjugate(U);
-
+	  
 	  // This phase could come from a simple bc 1,1,-1,1 ..
 	  int neglink = GaugeGrid->GlobalDimensions()[mu]-1;
 	  if ( Params.twists[mu] ) { 
 	    Uconj = where(coor==neglink,-Uconj,Uconj);
 	  }
-
 	  
-PARALLEL_FOR_LOOP
+	  
-	  for(auto ss=U.begin();ss<U.end();ss++){
+	  PARALLEL_FOR_LOOP
-	    Uds[ss](0)(mu) = U[ss]();
+	    for(auto ss=U.begin();ss<U.end();ss++){
-	    Uds[ss](1)(mu) = Uconj[ss]();
+	      Uds[ss](0)(mu) = U[ss]();
-	  }
+	      Uds[ss](1)(mu) = Uconj[ss]();
+	    }
 	  
 	  U     = adj(Cshift(U    ,mu,-1));      // correct except for spanning the boundary
 	  Uconj = adj(Cshift(Uconj,mu,-1));
@@ -434,80 +458,86 @@ PARALLEL_FOR_LOOP
 	    Utmp = where(coor==0,Uconj,Utmp);
 	  }
 	  
-PARALLEL_FOR_LOOP
+	  PARALLEL_FOR_LOOP
-	  for(auto ss=U.begin();ss<U.end();ss++){
+	    for(auto ss=U.begin();ss<U.end();ss++){
-	    Uds[ss](0)(mu+4) = Utmp[ss]();
+	      Uds[ss](0)(mu+4) = Utmp[ss]();
-	  }
+	    }
 	  
 	  Utmp = Uconj;
 	  if ( Params.twists[mu] ) { 
 	    Utmp = where(coor==0,U,Utmp);
 	  }
 	  
-PARALLEL_FOR_LOOP
+	  PARALLEL_FOR_LOOP
-	  for(auto ss=U.begin();ss<U.end();ss++){
+	    for(auto ss=U.begin();ss<U.end();ss++){
-	    Uds[ss](1)(mu+4) = Utmp[ss]();
+	      Uds[ss](1)(mu+4) = Utmp[ss]();
-	  }
+	    }
 	  
 	}
       }
-
+      
-      inline void InsertForce4D(GaugeField &mat, FermionField &Btilde, FermionField &A,int mu){
+      
-	
+      inline void InsertForce4D(GaugeField &mat, FermionField &Btilde,
+				FermionField &A, int mu) {
 	// DhopDir provides U or Uconj depending on coor/flavour.
 	GaugeLinkField link(mat._grid);
 	// use lorentz for flavour as hack.
-PARALLEL_FOR_LOOP
+	auto tmp = TraceIndex<SpinIndex>(outerProduct(Btilde, A));
-        for(auto ss=link.begin();ss<link.end();ss++){
+	PARALLEL_FOR_LOOP
-	  auto ttmp = traceIndex<SpinIndex>(outerProduct(Btilde[ss],A[ss]));  
+	  for (auto ss = tmp.begin(); ss < tmp.end(); ss++) {
-	  link[ss]() = ttmp(0,0) + conjugate(ttmp(1,1)) ; 
+	    link[ss]() = tmp[ss](0, 0) - conjugate(tmp[ss](1, 1));
-	}
+	  }
-	PokeIndex<LorentzIndex>(mat,link,mu);
+	PokeIndex<LorentzIndex>(mat, link, mu);
 	return;
       }
-      inline void InsertForce5D(GaugeField &mat, FermionField &Btilde, FermionField &Atilde,int mu){
+      
-
+      inline void InsertForce5D(GaugeField &mat, FermionField &Btilde,
-	int Ls=Btilde._grid->_fdimensions[0];
+				FermionField &Atilde, int mu) {
-
+	int Ls = Btilde._grid->_fdimensions[0];
+	
 	GaugeLinkField tmp(mat._grid);
 	tmp = zero;
-PARALLEL_FOR_LOOP
+	PARALLEL_FOR_LOOP
-	for(int ss=0;ss<tmp._grid->oSites();ss++){
+	  for (int ss = 0; ss < tmp._grid->oSites(); ss++) {
-	  for(int s=0;s<Ls;s++){
+	    for (int s = 0; s < Ls; s++) {
-	    int sF = s+Ls*ss;
+	      int sF = s + Ls * ss;
-	    auto ttmp = traceIndex<SpinIndex>(outerProduct(Btilde[sF],Atilde[sF]));
+	      auto ttmp = traceIndex<SpinIndex>(outerProduct(Btilde[sF], Atilde[sF]));
-	    tmp[ss]() = tmp[ss]()+ ttmp(0,0) + conjugate(ttmp(1,1));
+	      tmp[ss]() = tmp[ss]() + ttmp(0, 0) + conjugate(ttmp(1, 1));
+	    }
 	  }
-	}
+	PokeIndex<LorentzIndex>(mat, tmp, mu);
-	PokeIndex<LorentzIndex>(mat,tmp,mu);
 	return;
       }
     };
 
-    typedef WilsonImpl<vComplex ,Nc> WilsonImplR; // Real.. whichever prec
+    typedef WilsonImpl<vComplex,  FundamentalRepresentation > WilsonImplR;   // Real.. whichever prec
-    typedef WilsonImpl<vComplexF,Nc> WilsonImplF; // Float
+    typedef WilsonImpl<vComplexF, FundamentalRepresentation > WilsonImplF;  // Float
-    typedef WilsonImpl<vComplexD,Nc> WilsonImplD; // Double
+    typedef WilsonImpl<vComplexD, FundamentalRepresentation > WilsonImplD;  // Double
 
-    typedef WilsonImpl<vComplex ,Nc,ComplexD> ZWilsonImplR; // Real.. whichever prec
+
-    typedef WilsonImpl<vComplexF,Nc,ComplexD> ZWilsonImplF; // Float
+    typedef WilsonImpl<vComplex,  FundamentalRepresentation, ComplexD > ZWilsonImplR; // Real.. whichever prec
-    typedef WilsonImpl<vComplexD,Nc,ComplexD> ZWilsonImplD; // Double
+    typedef WilsonImpl<vComplexF, FundamentalRepresentation, ComplexD > ZWilsonImplF; // Float
+    typedef WilsonImpl<vComplexD, FundamentalRepresentation, ComplexD > ZWilsonImplD; // Double
+
+    typedef WilsonImpl<vComplex,  AdjointRepresentation > WilsonAdjImplR;   // Real.. whichever prec
+    typedef WilsonImpl<vComplexF, AdjointRepresentation > WilsonAdjImplF;  // Float
+    typedef WilsonImpl<vComplexD, AdjointRepresentation > WilsonAdjImplD;  // Double
+
+    typedef WilsonImpl<vComplex,  TwoIndexSymmetricRepresentation > WilsonTwoIndexSymmetricImplR;   // Real.. whichever prec
+    typedef WilsonImpl<vComplexF, TwoIndexSymmetricRepresentation > WilsonTwoIndexSymmetricImplF;  // Float
+    typedef WilsonImpl<vComplexD, TwoIndexSymmetricRepresentation > WilsonTwoIndexSymmetricImplD;  // Double
 
     typedef DomainWallVec5dImpl<vComplex ,Nc> DomainWallVec5dImplR; // Real.. whichever prec
     typedef DomainWallVec5dImpl<vComplexF,Nc> DomainWallVec5dImplF; // Float
     typedef DomainWallVec5dImpl<vComplexD,Nc> DomainWallVec5dImplD; // Double
-
+    
     typedef DomainWallVec5dImpl<vComplex ,Nc,ComplexD> ZDomainWallVec5dImplR; // Real.. whichever prec
     typedef DomainWallVec5dImpl<vComplexF,Nc,ComplexD> ZDomainWallVec5dImplF; // Float
     typedef DomainWallVec5dImpl<vComplexD,Nc,ComplexD> ZDomainWallVec5dImplD; // Double
 
-    typedef DomainWallVec5dImpl<vComplex ,Nc> DomainWallVec5dImplR; // Real.. whichever prec
+    typedef GparityWilsonImpl<vComplex, Nc>  GparityWilsonImplR;  // Real.. whichever prec
-    typedef DomainWallVec5dImpl<vComplexF,Nc> DomainWallVec5dImplF; // Float
+    typedef GparityWilsonImpl<vComplexF, Nc> GparityWilsonImplF;  // Float
-    typedef DomainWallVec5dImpl<vComplexD,Nc> DomainWallVec5dImplD; // Double
+    typedef GparityWilsonImpl<vComplexD, Nc> GparityWilsonImplD;  // Double
-
+}
-    typedef GparityWilsonImpl<vComplex ,Nc> GparityWilsonImplR; // Real.. whichever prec
-    typedef GparityWilsonImpl<vComplexF,Nc> GparityWilsonImplF; // Float
-    typedef GparityWilsonImpl<vComplexD,Nc> GparityWilsonImplD; // Double
-
-  }
 }
 #endif

lib/qcd/action/fermion/WilsonFermion.cc

@@ -1,130 +1,129 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./lib/qcd/action/fermion/WilsonFermion.cc
+Source file: ./lib/qcd/action/fermion/WilsonFermion.cc
 
-    Copyright (C) 2015
+Copyright (C) 2015
 
 Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
 Author: paboyle <paboyle@ph.ed.ac.uk>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution
-    *************************************************************************************/
+directory
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #include <Grid.h>
 
 namespace Grid {
 namespace QCD {
 
-  const std::vector<int> WilsonFermionStatic::directions   ({0,1,2,3, 0, 1, 2, 3});
+const std::vector<int> WilsonFermionStatic::directions({0, 1, 2, 3, 0, 1, 2,
-  const std::vector<int> WilsonFermionStatic::displacements({1,1,1,1,-1,-1,-1,-1});
+                                                        3});
-  int WilsonFermionStatic::HandOptDslash;
+const std::vector<int> WilsonFermionStatic::displacements({1, 1, 1, 1, -1, -1,
+                                                           -1, -1});
+int WilsonFermionStatic::HandOptDslash;
 
-  /////////////////////////////////
+/////////////////////////////////
-  // Constructor and gauge import
+// Constructor and gauge import
-  /////////////////////////////////
+/////////////////////////////////
 
-  template<class Impl>
+template <class Impl>
-  WilsonFermion<Impl>::WilsonFermion(GaugeField &_Umu,
+WilsonFermion<Impl>::WilsonFermion(GaugeField &_Umu, GridCartesian &Fgrid,
-				     GridCartesian         &Fgrid,
+                                   GridRedBlackCartesian &Hgrid, RealD _mass,
-				     GridRedBlackCartesian &Hgrid, 
+                                   const ImplParams &p)
-				     RealD _mass,const ImplParams &p) :
+    : Kernels(p),
-        Kernels(p),
+      _grid(&Fgrid),
-        _grid(&Fgrid),
+      _cbgrid(&Hgrid),
-	_cbgrid(&Hgrid),
+      Stencil(&Fgrid, npoint, Even, directions, displacements),
-	Stencil    (&Fgrid,npoint,Even,directions,displacements),
+      StencilEven(&Hgrid, npoint, Even, directions,
-	StencilEven(&Hgrid,npoint,Even,directions,displacements), // source is Even
+                  displacements),  // source is Even
-	StencilOdd (&Hgrid,npoint,Odd ,directions,displacements), // source is Odd
+      StencilOdd(&Hgrid, npoint, Odd, directions,
-	mass(_mass),
+                 displacements),  // source is Odd
-	Lebesgue(_grid),
+      mass(_mass),
-	LebesgueEvenOdd(_cbgrid),
+      Lebesgue(_grid),
-	Umu(&Fgrid),
+      LebesgueEvenOdd(_cbgrid),
-	UmuEven(&Hgrid),
+      Umu(&Fgrid),
-	UmuOdd (&Hgrid) 
+      UmuEven(&Hgrid),
-  {
+      UmuOdd(&Hgrid) {
-    // Allocate the required comms buffer
+  // Allocate the required comms buffer
-    ImportGauge(_Umu);
+  ImportGauge(_Umu);
+}
+
+template <class Impl>
+void WilsonFermion<Impl>::ImportGauge(const GaugeField &_Umu) {
+  GaugeField HUmu(_Umu._grid);
+  HUmu = _Umu * (-0.5);
+  Impl::DoubleStore(GaugeGrid(), Umu, HUmu);
+  pickCheckerboard(Even, UmuEven, Umu);
+  pickCheckerboard(Odd, UmuOdd, Umu);
+}
+
+/////////////////////////////
+// Implement the interface
+/////////////////////////////
+
+template <class Impl>
+RealD WilsonFermion<Impl>::M(const FermionField &in, FermionField &out) {
+  out.checkerboard = in.checkerboard;
+  Dhop(in, out, DaggerNo);
+  return axpy_norm(out, 4 + mass, in, out);
+}
+
+template <class Impl>
+RealD WilsonFermion<Impl>::Mdag(const FermionField &in, FermionField &out) {
+  out.checkerboard = in.checkerboard;
+  Dhop(in, out, DaggerYes);
+  return axpy_norm(out, 4 + mass, in, out);
+}
+
+template <class Impl>
+void WilsonFermion<Impl>::Meooe(const FermionField &in, FermionField &out) {
+  if (in.checkerboard == Odd) {
+    DhopEO(in, out, DaggerNo);
+  } else {
+    DhopOE(in, out, DaggerNo);
   }
+}
 
-  template<class Impl>
+template <class Impl>
-  void WilsonFermion<Impl>::ImportGauge(const GaugeField &_Umu)
+void WilsonFermion<Impl>::MeooeDag(const FermionField &in, FermionField &out) {
-  {
+  if (in.checkerboard == Odd) {
-    GaugeField HUmu(_Umu._grid);
+    DhopEO(in, out, DaggerYes);
-    HUmu = _Umu*(-0.5);
+  } else {
-    Impl::DoubleStore(GaugeGrid(),Umu,HUmu);
+    DhopOE(in, out, DaggerYes);
-    pickCheckerboard(Even,UmuEven,Umu);
-    pickCheckerboard(Odd ,UmuOdd,Umu);
   }
+}
   
-  /////////////////////////////
+  template <class Impl>
-  // Implement the interface
-  /////////////////////////////
-      
-  template<class Impl>
-  RealD WilsonFermion<Impl>::M(const FermionField &in, FermionField &out) 
-  {
-    out.checkerboard=in.checkerboard;
-    Dhop(in,out,DaggerNo);
-    return axpy_norm(out,4+mass,in,out);
-  }
-
-  template<class Impl>
-  RealD WilsonFermion<Impl>::Mdag(const FermionField &in, FermionField &out) 
-  {
-    out.checkerboard=in.checkerboard;
-    Dhop(in,out,DaggerYes);
-    return axpy_norm(out,4+mass,in,out);
-  }
-
-  template<class Impl>
-  void WilsonFermion<Impl>::Meooe(const FermionField &in, FermionField &out) 
-  {
-    if ( in.checkerboard == Odd ) {
-      DhopEO(in,out,DaggerNo);
-    } else {
-      DhopOE(in,out,DaggerNo);
-    }
-  }
-  template<class Impl>
-  void WilsonFermion<Impl>::MeooeDag(const FermionField &in, FermionField &out) 
-  {
-    if ( in.checkerboard == Odd ) {
-      DhopEO(in,out,DaggerYes);
-    } else {
-      DhopOE(in,out,DaggerYes);
-    }
-  }
-
-  template<class Impl>
   void WilsonFermion<Impl>::Mooee(const FermionField &in, FermionField &out) {
     out.checkerboard = in.checkerboard;
-    typename FermionField::scalar_type scal(4.0+mass);
+    typename FermionField::scalar_type scal(4.0 + mass);
-    out = scal*in;
+    out = scal * in;
   }
-  
+
-  template<class Impl>
+  template <class Impl>
   void WilsonFermion<Impl>::MooeeDag(const FermionField &in, FermionField &out) {
     out.checkerboard = in.checkerboard;
-    Mooee(in,out);
+    Mooee(in, out);
   }
-  
+
   template<class Impl>
   void WilsonFermion<Impl>::MooeeInv(const FermionField &in, FermionField &out) {
     out.checkerboard = in.checkerboard;
@@ -136,8 +135,9 @@ namespace QCD {
     out.checkerboard = in.checkerboard;
     MooeeInv(in,out);
   }
+
   template<class Impl>
-  void WilsonFermion<Impl>:: MomentumSpacePropagator(FermionField &out, const FermionField &in,RealD _m) {
+  void WilsonFermion<Impl>::MomentumSpacePropagator(FermionField &out, const FermionField &in,RealD _m) {
 
     // what type LatticeComplex 
     conformable(_grid,out._grid);
@@ -190,184 +190,182 @@ namespace QCD {
     out = num*denom; // [ -i gmu sin k + 2 sin^2 k/2 + m] / [ sin^2 k + (2 sin^2 k/2 + m)^2 ]
 
   }
-  
-  ///////////////////////////////////
-  // Internal
-  ///////////////////////////////////
-
-  template<class Impl>
-  void WilsonFermion<Impl>::DerivInternal(StencilImpl & st,
-					  DoubledGaugeField & U,
-					  GaugeField &mat,
-					  const FermionField &A,
-					  const FermionField &B,int dag) {
-	
-    assert((dag==DaggerNo) ||(dag==DaggerYes));
-    
-    Compressor compressor(dag);
-    
-    FermionField Btilde(B._grid);
-    FermionField Atilde(B._grid);
-    Atilde = A;
-
-    st.HaloExchange(B,compressor);
-    
-    for(int mu=0;mu<Nd;mu++){
-      
-      ////////////////////////////////////////////////////////////////////////
-      // Flip gamma (1+g)<->(1-g) if dag
-      ////////////////////////////////////////////////////////////////////////
-      int gamma = mu;
-      if ( !dag ) gamma+= Nd;
-      
-      ////////////////////////
-      // Call the single hop
-      ////////////////////////
-PARALLEL_FOR_LOOP
-	for(int sss=0;sss<B._grid->oSites();sss++){
-	  Kernels::DiracOptDhopDir(st,U,st.comm_buf,sss,sss,B,Btilde,mu,gamma);
-	}
-      
-      //////////////////////////////////////////////////
-      // spin trace outer product
-      //////////////////////////////////////////////////
-      Impl::InsertForce4D(mat,Btilde,Atilde,mu);
-
-    }
-  }
-  
-  template<class Impl>
-  void WilsonFermion<Impl>::DhopDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
-  {
-    conformable(U._grid,_grid);  
-    conformable(U._grid,V._grid);
-    conformable(U._grid,mat._grid);
-    
-    mat.checkerboard = U.checkerboard;
-    
-    DerivInternal(Stencil,Umu,mat,U,V,dag);
-  }
-  
-  template<class Impl>
-  void WilsonFermion<Impl>::DhopDerivOE(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
-  {
-    conformable(U._grid,_cbgrid);  
-    conformable(U._grid,V._grid);
-    conformable(U._grid,mat._grid);
-    
-    assert(V.checkerboard==Even);
-    assert(U.checkerboard==Odd);
-    mat.checkerboard = Odd;
-    
-    DerivInternal(StencilEven,UmuOdd,mat,U,V,dag);
-  }
-  
-  template<class Impl>
-  void WilsonFermion<Impl>::DhopDerivEO(GaugeField &mat,const FermionField &U,const FermionField &V,int dag)
-  {
-    conformable(U._grid,_cbgrid);  
-    conformable(U._grid,V._grid);
-    conformable(U._grid,mat._grid);
-	
-    assert(V.checkerboard==Odd);
-    assert(U.checkerboard==Even);
-    mat.checkerboard = Even;
-	
-    DerivInternal(StencilOdd,UmuEven,mat,U,V,dag);
-  }
-  
-
-  template<class Impl>
-  void WilsonFermion<Impl>::Dhop(const FermionField &in, FermionField &out,int dag) {
-    conformable(in._grid,_grid); // verifies full grid
-    conformable(in._grid,out._grid);
-    
-    out.checkerboard = in.checkerboard;
-    
-    DhopInternal(Stencil,Lebesgue,Umu,in,out,dag);
-  }
-  
-  template<class Impl>
-  void WilsonFermion<Impl>::DhopOE(const FermionField &in, FermionField &out,int dag) {
-    conformable(in._grid,_cbgrid);    // verifies half grid
-    conformable(in._grid,out._grid); // drops the cb check
-    
-    assert(in.checkerboard==Even);
-    out.checkerboard = Odd;
-    
-    DhopInternal(StencilEven,LebesgueEvenOdd,UmuOdd,in,out,dag);
-  }
-  
-  template<class Impl>
-  void WilsonFermion<Impl>::DhopEO(const FermionField &in, FermionField &out,int dag) {
-    conformable(in._grid,_cbgrid);    // verifies half grid
-    conformable(in._grid,out._grid); // drops the cb check
-    
-    assert(in.checkerboard==Odd);
-    out.checkerboard = Even;
-    
-    DhopInternal(StencilOdd,LebesgueEvenOdd,UmuEven,in,out,dag);
-  }
-  
-  template<class Impl>
-  void WilsonFermion<Impl>::Mdir (const FermionField &in, FermionField &out,int dir,int disp) {
-    DhopDir(in,out,dir,disp);
-  }
-  
-
-  template<class Impl>
-  void WilsonFermion<Impl>::DhopDir(const FermionField &in, FermionField &out,int dir,int disp){
-    
-    int skip = (disp==1) ? 0 : 1;
-    int dirdisp  = dir+skip*4;
-    int gamma    = dir+(1-skip)*4;
-    
-    DhopDirDisp(in,out,dirdisp,gamma,DaggerNo);
-    
-  };
-  
-  template<class Impl>
-  void WilsonFermion<Impl>::DhopDirDisp(const FermionField &in, FermionField &out,int dirdisp,int gamma,int dag) {
-    
-    Compressor compressor(dag);
-    
-    Stencil.HaloExchange(in,compressor);
-    
-PARALLEL_FOR_LOOP
-      for(int sss=0;sss<in._grid->oSites();sss++){
-	Kernels::DiracOptDhopDir(Stencil,Umu,Stencil.comm_buf,sss,sss,in,out,dirdisp,gamma);
-      }
-    
-  };
-
-  template<class Impl>
-  void WilsonFermion<Impl>::DhopInternal(StencilImpl & st,LebesgueOrder& lo,DoubledGaugeField & U,
-					 const FermionField &in, FermionField &out,int dag) 
-  {
-    assert((dag==DaggerNo) ||(dag==DaggerYes));
-
-    Compressor compressor(dag);
-    st.HaloExchange(in,compressor);
-    
-    if ( dag == DaggerYes ) {
-PARALLEL_FOR_LOOP
-      for(int sss=0;sss<in._grid->oSites();sss++){
-	Kernels::DiracOptDhopSiteDag(st,lo,U,st.comm_buf,sss,sss,1,1,in,out);
-      }
-    } else {
-PARALLEL_FOR_LOOP
-      for(int sss=0;sss<in._grid->oSites();sss++){
-	Kernels::DiracOptDhopSite(st,lo,U,st.comm_buf,sss,sss,1,1,in,out);
-      }
-    }
-  };
-
  
-  FermOpTemplateInstantiate(WilsonFermion);
-  GparityFermOpTemplateInstantiate(WilsonFermion);
 
+///////////////////////////////////
+// Internal
+///////////////////////////////////
 
-}}
+template <class Impl>
+void WilsonFermion<Impl>::DerivInternal(StencilImpl &st, DoubledGaugeField &U,
+                                        GaugeField &mat, const FermionField &A,
+                                        const FermionField &B, int dag) {
+  assert((dag == DaggerNo) || (dag == DaggerYes));
 
+  Compressor compressor(dag);
 
+  FermionField Btilde(B._grid);
+  FermionField Atilde(B._grid);
+  Atilde = A;
 
+  st.HaloExchange(B, compressor);
+
+  for (int mu = 0; mu < Nd; mu++) {
+    ////////////////////////////////////////////////////////////////////////
+    // Flip gamma (1+g)<->(1-g) if dag
+    ////////////////////////////////////////////////////////////////////////
+    int gamma = mu;
+    if (!dag) gamma += Nd;
+
+    ////////////////////////
+    // Call the single hop
+    ////////////////////////
+    PARALLEL_FOR_LOOP
+    for (int sss = 0; sss < B._grid->oSites(); sss++) {
+      Kernels::DiracOptDhopDir(st, U, st.comm_buf, sss, sss, B, Btilde, mu,
+                               gamma);
+    }
+
+    //////////////////////////////////////////////////
+    // spin trace outer product
+    //////////////////////////////////////////////////
+    Impl::InsertForce4D(mat, Btilde, Atilde, mu);
+  }
+}
+
+template <class Impl>
+void WilsonFermion<Impl>::DhopDeriv(GaugeField &mat, const FermionField &U,
+                                    const FermionField &V, int dag) {
+  conformable(U._grid, _grid);
+  conformable(U._grid, V._grid);
+  conformable(U._grid, mat._grid);
+
+  mat.checkerboard = U.checkerboard;
+
+  DerivInternal(Stencil, Umu, mat, U, V, dag);
+}
+
+template <class Impl>
+void WilsonFermion<Impl>::DhopDerivOE(GaugeField &mat, const FermionField &U,
+                                      const FermionField &V, int dag) {
+  conformable(U._grid, _cbgrid);
+  conformable(U._grid, V._grid);
+  conformable(U._grid, mat._grid);
+
+  assert(V.checkerboard == Even);
+  assert(U.checkerboard == Odd);
+  mat.checkerboard = Odd;
+
+  DerivInternal(StencilEven, UmuOdd, mat, U, V, dag);
+}
+
+template <class Impl>
+void WilsonFermion<Impl>::DhopDerivEO(GaugeField &mat, const FermionField &U,
+                                      const FermionField &V, int dag) {
+  conformable(U._grid, _cbgrid);
+  conformable(U._grid, V._grid);
+  conformable(U._grid, mat._grid);
+
+  assert(V.checkerboard == Odd);
+  assert(U.checkerboard == Even);
+  mat.checkerboard = Even;
+
+  DerivInternal(StencilOdd, UmuEven, mat, U, V, dag);
+}
+
+template <class Impl>
+void WilsonFermion<Impl>::Dhop(const FermionField &in, FermionField &out,
+                               int dag) {
+  conformable(in._grid, _grid);  // verifies full grid
+  conformable(in._grid, out._grid);
+
+  out.checkerboard = in.checkerboard;
+
+  DhopInternal(Stencil, Lebesgue, Umu, in, out, dag);
+}
+
+template <class Impl>
+void WilsonFermion<Impl>::DhopOE(const FermionField &in, FermionField &out,
+                                 int dag) {
+  conformable(in._grid, _cbgrid);    // verifies half grid
+  conformable(in._grid, out._grid);  // drops the cb check
+
+  assert(in.checkerboard == Even);
+  out.checkerboard = Odd;
+
+  DhopInternal(StencilEven, LebesgueEvenOdd, UmuOdd, in, out, dag);
+}
+
+template <class Impl>
+void WilsonFermion<Impl>::DhopEO(const FermionField &in, FermionField &out,
+                                 int dag) {
+  conformable(in._grid, _cbgrid);    // verifies half grid
+  conformable(in._grid, out._grid);  // drops the cb check
+
+  assert(in.checkerboard == Odd);
+  out.checkerboard = Even;
+
+  DhopInternal(StencilOdd, LebesgueEvenOdd, UmuEven, in, out, dag);
+}
+
+template <class Impl>
+void WilsonFermion<Impl>::Mdir(const FermionField &in, FermionField &out,
+                               int dir, int disp) {
+  DhopDir(in, out, dir, disp);
+}
+
+template <class Impl>
+void WilsonFermion<Impl>::DhopDir(const FermionField &in, FermionField &out,
+                                  int dir, int disp) {
+  int skip = (disp == 1) ? 0 : 1;
+  int dirdisp = dir + skip * 4;
+  int gamma = dir + (1 - skip) * 4;
+
+  DhopDirDisp(in, out, dirdisp, gamma, DaggerNo);
+};
+
+template <class Impl>
+void WilsonFermion<Impl>::DhopDirDisp(const FermionField &in, FermionField &out,
+                                      int dirdisp, int gamma, int dag) {
+  Compressor compressor(dag);
+
+  Stencil.HaloExchange(in, compressor);
+
+  PARALLEL_FOR_LOOP
+  for (int sss = 0; sss < in._grid->oSites(); sss++) {
+    Kernels::DiracOptDhopDir(Stencil, Umu, Stencil.comm_buf, sss, sss, in, out,
+                             dirdisp, gamma);
+  }
+};
+
+template <class Impl>
+void WilsonFermion<Impl>::DhopInternal(StencilImpl &st, LebesgueOrder &lo,
+                                       DoubledGaugeField &U,
+                                       const FermionField &in,
+                                       FermionField &out, int dag) {
+  assert((dag == DaggerNo) || (dag == DaggerYes));
+
+  Compressor compressor(dag);
+  st.HaloExchange(in, compressor);
+
+  if (dag == DaggerYes) {
+    PARALLEL_FOR_LOOP
+    for (int sss = 0; sss < in._grid->oSites(); sss++) {
+      Kernels::DiracOptDhopSiteDag(st, lo, U, st.comm_buf, sss, sss, 1, 1, in,
+                                   out);
+    }
+  } else {
+    PARALLEL_FOR_LOOP
+    for (int sss = 0; sss < in._grid->oSites(); sss++) {
+      Kernels::DiracOptDhopSite(st, lo, U, st.comm_buf, sss, sss, 1, 1, in,
+                                out);
+    }
+  }
+};
+
+FermOpTemplateInstantiate(WilsonFermion);
+AdjointFermOpTemplateInstantiate(WilsonFermion);
+TwoIndexFermOpTemplateInstantiate(WilsonFermion);
+GparityFermOpTemplateInstantiate(WilsonFermion);
+}
+}

lib/qcd/action/fermion/WilsonFermion.h

@@ -1,163 +1,154 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./lib/qcd/action/fermion/WilsonFermion.h
+Source file: ./lib/qcd/action/fermion/WilsonFermion.h
 
-    Copyright (C) 2015
+Copyright (C) 2015
 
 Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: paboyle <paboyle@ph.ed.ac.uk>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution
-    *************************************************************************************/
+directory
-    /*  END LEGAL */
+*************************************************************************************/
-#ifndef  GRID_QCD_WILSON_FERMION_H
+/*  END LEGAL */
-#define  GRID_QCD_WILSON_FERMION_H
+#ifndef GRID_QCD_WILSON_FERMION_H
+#define GRID_QCD_WILSON_FERMION_H
 
 namespace Grid {
 
-  namespace QCD {
+namespace QCD {
 
-    class WilsonFermionStatic {
+class WilsonFermionStatic {
-    public:
+ public:
-      static int HandOptDslash; // these are a temporary hack
+  static int HandOptDslash;  // these are a temporary hack
-      static int MortonOrder;
+  static int MortonOrder;
-      static const std::vector<int> directions   ;
+  static const std::vector<int> directions;
-      static const std::vector<int> displacements;
+  static const std::vector<int> displacements;
-      static const int npoint=8;
+  static const int npoint = 8;
-    };
+};
 
-    template<class Impl>
+template <class Impl>
-    class WilsonFermion : public WilsonKernels<Impl>, public WilsonFermionStatic
+class WilsonFermion : public WilsonKernels<Impl>, public WilsonFermionStatic {
-    {
+ public:
-    public:
+  INHERIT_IMPL_TYPES(Impl);
-    INHERIT_IMPL_TYPES(Impl);
+  typedef WilsonKernels<Impl> Kernels;
-    typedef WilsonKernels<Impl> Kernels;
 
-      ///////////////////////////////////////////////////////////////
+  ///////////////////////////////////////////////////////////////
-      // Implement the abstract base
+  // Implement the abstract base
-      ///////////////////////////////////////////////////////////////
+  ///////////////////////////////////////////////////////////////
-      GridBase *GaugeGrid(void)              { return _grid ;}
+  GridBase *GaugeGrid(void) { return _grid; }
-      GridBase *GaugeRedBlackGrid(void)      { return _cbgrid ;}
+  GridBase *GaugeRedBlackGrid(void) { return _cbgrid; }
-      GridBase *FermionGrid(void)            { return _grid;}
+  GridBase *FermionGrid(void) { return _grid; }
-      GridBase *FermionRedBlackGrid(void)    { return _cbgrid;}
+  GridBase *FermionRedBlackGrid(void) { return _cbgrid; }
 
-      //////////////////////////////////////////////////////////////////
+  //////////////////////////////////////////////////////////////////
-      // override multiply; cut number routines if pass dagger argument
+  // override multiply; cut number routines if pass dagger argument
-      // and also make interface more uniformly consistent
+  // and also make interface more uniformly consistent
-      //////////////////////////////////////////////////////////////////
+  //////////////////////////////////////////////////////////////////
-      RealD M(const FermionField &in, FermionField &out);
+  RealD M(const FermionField &in, FermionField &out);
-      RealD Mdag(const FermionField &in, FermionField &out);
+  RealD Mdag(const FermionField &in, FermionField &out);
 
-      /////////////////////////////////////////////////////////
+  /////////////////////////////////////////////////////////
-      // half checkerboard operations
+  // half checkerboard operations
-      // could remain virtual so we  can derive Clover from Wilson base
+  // could remain virtual so we  can derive Clover from Wilson base
-      /////////////////////////////////////////////////////////
+  /////////////////////////////////////////////////////////
-      void Meooe(const FermionField &in, FermionField &out) ;
+  void Meooe(const FermionField &in, FermionField &out);
-      void MeooeDag(const FermionField &in, FermionField &out) ;
+  void MeooeDag(const FermionField &in, FermionField &out);
 
-      // allow override for twisted mass and clover
+  // allow override for twisted mass and clover
-      virtual void Mooee(const FermionField &in, FermionField &out) ;
+  virtual void Mooee(const FermionField &in, FermionField &out);
-      virtual void MooeeDag(const FermionField &in, FermionField &out) ;
+  virtual void MooeeDag(const FermionField &in, FermionField &out);
-      virtual void MooeeInv(const FermionField &in, FermionField &out) ;
+  virtual void MooeeInv(const FermionField &in, FermionField &out);
-      virtual void MooeeInvDag(const FermionField &in, FermionField &out) ;
+  virtual void MooeeInvDag(const FermionField &in, FermionField &out);
 
-      virtual void  MomentumSpacePropagator(FermionField &out,const FermionField &in,RealD _mass) ;
+  virtual void  MomentumSpacePropagator(FermionField &out,const FermionField &in,RealD _mass) ;
 
-      ////////////////////////
+  ////////////////////////
-      // Derivative interface
+  // Derivative interface
-      ////////////////////////
+  ////////////////////////
-      // Interface calls an internal routine
+  // Interface calls an internal routine
-      void DhopDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  void DhopDeriv(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
-      void DhopDerivOE(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  void DhopDerivOE(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
-      void DhopDerivEO(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+  void DhopDerivEO(GaugeField &mat,const FermionField &U,const FermionField &V,int dag);
+
+  ///////////////////////////////////////////////////////////////
+  // non-hermitian hopping term; half cb or both
+  ///////////////////////////////////////////////////////////////
+  void Dhop(const FermionField &in, FermionField &out, int dag);
+  void DhopOE(const FermionField &in, FermionField &out, int dag);
+  void DhopEO(const FermionField &in, FermionField &out, int dag);
+
+  ///////////////////////////////////////////////////////////////
+  // Multigrid assistance; force term uses too
+  ///////////////////////////////////////////////////////////////
+  void Mdir(const FermionField &in, FermionField &out, int dir, int disp);
+  void DhopDir(const FermionField &in, FermionField &out, int dir, int disp);
+  void DhopDirDisp(const FermionField &in, FermionField &out, int dirdisp,
+                   int gamma, int dag);
+
+  ///////////////////////////////////////////////////////////////
+  // Extra methods added by derived
+  ///////////////////////////////////////////////////////////////
+  void DerivInternal(StencilImpl &st, DoubledGaugeField &U, GaugeField &mat,
+                     const FermionField &A, const FermionField &B, int dag);
+
+  void DhopInternal(StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
+                    const FermionField &in, FermionField &out, int dag);
+
+  // Constructor
+  WilsonFermion(GaugeField &_Umu, GridCartesian &Fgrid,
+                GridRedBlackCartesian &Hgrid, RealD _mass,
+                const ImplParams &p = ImplParams());
+
+  // DoubleStore impl dependent
+  void ImportGauge(const GaugeField &_Umu);
+
+  ///////////////////////////////////////////////////////////////
+  // Data members require to support the functionality
+  ///////////////////////////////////////////////////////////////
+
+  //    protected:
+ public:
+  RealD mass;
+
+  GridBase *_grid;
+  GridBase *_cbgrid;
+
+  // Defines the stencils for even and odd
+  StencilImpl Stencil;
+  StencilImpl StencilEven;
+  StencilImpl StencilOdd;
+
+  // Copy of the gauge field , with even and odd subsets
+  DoubledGaugeField Umu;
+  DoubledGaugeField UmuEven;
+  DoubledGaugeField UmuOdd;
+
+  LebesgueOrder Lebesgue;
+  LebesgueOrder LebesgueEvenOdd;
+};
+
+typedef WilsonFermion<WilsonImplF> WilsonFermionF;
+typedef WilsonFermion<WilsonImplD> WilsonFermionD;
 
 
-      ///////////////////////////////////////////////////////////////
+}
-      // non-hermitian hopping term; half cb or both
-      ///////////////////////////////////////////////////////////////
-      void Dhop(const FermionField &in, FermionField &out,int dag) ;
-      void DhopOE(const FermionField &in, FermionField &out,int dag) ;
-      void DhopEO(const FermionField &in, FermionField &out,int dag) ;
-
-      ///////////////////////////////////////////////////////////////
-      // Multigrid assistance; force term uses too
-      ///////////////////////////////////////////////////////////////
-      void Mdir (const FermionField &in, FermionField &out,int dir,int disp) ;
-      void DhopDir(const FermionField &in, FermionField &out,int dir,int disp);
-      void DhopDirDisp(const FermionField &in, FermionField &out,int dirdisp,int gamma,int dag) ;
-
-      ///////////////////////////////////////////////////////////////
-      // Extra methods added by derived
-      ///////////////////////////////////////////////////////////////
-      void DerivInternal(StencilImpl & st,
-			 DoubledGaugeField & U,
-			 GaugeField &mat,
-			 const FermionField &A,
-			 const FermionField &B,
-			 int dag);
-
-      void DhopInternal(StencilImpl & st,LebesgueOrder & lo,DoubledGaugeField & U,
-			const FermionField &in, FermionField &out,int dag) ;
-
-      // Constructor
-      WilsonFermion(GaugeField &_Umu,
-		    GridCartesian         &Fgrid,
-		    GridRedBlackCartesian &Hgrid, 
-		    RealD _mass,
-		    const ImplParams &p= ImplParams()
-		    ) ;
-
-      // DoubleStore impl dependent
-      void ImportGauge(const GaugeField &_Umu);
-
-      ///////////////////////////////////////////////////////////////
-      // Data members require to support the functionality
-      ///////////////////////////////////////////////////////////////
-
-      //    protected:
-    public:
-
-      RealD                        mass;
-
-      GridBase                     *    _grid; 
-      GridBase                     *  _cbgrid;
-
-      //Defines the stencils for even and odd
-      StencilImpl Stencil; 
-      StencilImpl StencilEven; 
-      StencilImpl StencilOdd; 
-
-      // Copy of the gauge field , with even and odd subsets
-      DoubledGaugeField Umu;
-      DoubledGaugeField UmuEven;
-      DoubledGaugeField UmuOdd;
-
-      LebesgueOrder Lebesgue;
-      LebesgueOrder LebesgueEvenOdd;
-
-      
-    };
-
-    typedef WilsonFermion<WilsonImplF> WilsonFermionF;
-    typedef WilsonFermion<WilsonImplD> WilsonFermionD;
-
-  }
 }
 #endif

lib/qcd/action/fermion/WilsonFermion5D.cc

@@ -42,11 +42,11 @@ const std::vector<int> WilsonFermion5DStatic::displacements({1,1,1,1,-1,-1,-1,-1
   // 5d lattice for DWF.
 template<class Impl>
 WilsonFermion5D<Impl>::WilsonFermion5D(GaugeField &_Umu,
-				       GridCartesian         &FiveDimGrid,
+               GridCartesian         &FiveDimGrid,
-				       GridRedBlackCartesian &FiveDimRedBlackGrid,
+               GridRedBlackCartesian &FiveDimRedBlackGrid,
-				       GridCartesian         &FourDimGrid,
+               GridCartesian         &FourDimGrid,
-				       GridRedBlackCartesian &FourDimRedBlackGrid,
+               GridRedBlackCartesian &FourDimRedBlackGrid,
-				       RealD _M5,const ImplParams &p) :
+               RealD _M5,const ImplParams &p) :
   Kernels(p),
   _FiveDimGrid        (&FiveDimGrid),
   _FiveDimRedBlackGrid(&FiveDimRedBlackGrid),
@@ -135,10 +135,10 @@ WilsonFermion5D<Impl>::WilsonFermion5D(GaugeField &_Umu,
   /*
 template<class Impl>
 WilsonFermion5D<Impl>::WilsonFermion5D(int simd,GaugeField &_Umu,
-				       GridCartesian         &FiveDimGrid,
+               GridCartesian         &FiveDimGrid,
-				       GridRedBlackCartesian &FiveDimRedBlackGrid,
+               GridRedBlackCartesian &FiveDimRedBlackGrid,
-				       GridCartesian         &FourDimGrid,
+               GridCartesian         &FourDimGrid,
-				       RealD _M5,const ImplParams &p) :
+               RealD _M5,const ImplParams &p) :
 {
   int nsimd = Simd::Nsimd();
 
@@ -175,6 +175,73 @@ WilsonFermion5D<Impl>::WilsonFermion5D(int simd,GaugeField &_Umu,
 }  
   */
      
+template<class Impl>
+void WilsonFermion5D<Impl>::Report(void)
+{
+    std::vector<int> latt = GridDefaultLatt();          
+    RealD volume = Ls;  for(int mu=0;mu<Nd;mu++) volume=volume*latt[mu];
+    RealD NP = _FourDimGrid->_Nprocessors;
+
+  if ( DhopCalls > 0 ) {
+    std::cout << GridLogMessage << "#### Dhop calls report " << std::endl;
+    std::cout << GridLogMessage << "WilsonFermion5D Number of Dhop Calls     : " << DhopCalls  << std::endl;
+    std::cout << GridLogMessage << "WilsonFermion5D Total Communication time : " << DhopCommTime
+              << " us" << std::endl;
+    std::cout << GridLogMessage << "WilsonFermion5D CommTime/Calls           : "
+              << DhopCommTime / DhopCalls << " us" << std::endl;
+    std::cout << GridLogMessage << "WilsonFermion5D Total Compute time       : "
+              << DhopComputeTime << " us" << std::endl;
+    std::cout << GridLogMessage << "WilsonFermion5D ComputeTime/Calls        : "
+              << DhopComputeTime / DhopCalls << " us" << std::endl;
+
+    RealD mflops = 1344*volume*DhopCalls/DhopComputeTime;
+    std::cout << GridLogMessage << "Average mflops/s per call                : " << mflops << std::endl;
+    std::cout << GridLogMessage << "Average mflops/s per call per node       : " << mflops/NP << std::endl;
+
+   }
+
+  if ( DerivCalls > 0 ) {
+  std::cout << GridLogMessage << "#### Deriv calls report "<< std::endl;
+  std::cout << GridLogMessage << "WilsonFermion5D Number of Deriv Calls    : " <<DerivCalls <<std::endl;
+  std::cout << GridLogMessage << "WilsonFermion5D Total Communication time : " <<DerivCommTime <<" us"<<std::endl;
+  std::cout << GridLogMessage << "WilsonFermion5D CommTime/Calls           : " <<DerivCommTime/DerivCalls<<" us" <<std::endl;
+  std::cout << GridLogMessage << "WilsonFermion5D Total Compute time       : " <<DerivComputeTime <<" us"<<std::endl;
+  std::cout << GridLogMessage << "WilsonFermion5D ComputeTime/Calls        : " <<DerivComputeTime/DerivCalls<<" us" <<std::endl;
+  std::cout << GridLogMessage << "WilsonFermion5D Total Dhop Compute time  : " <<DerivDhopComputeTime <<" us"<<std::endl;
+  std::cout << GridLogMessage << "WilsonFermion5D Dhop ComputeTime/Calls   : " <<DerivDhopComputeTime/DerivCalls<<" us" <<std::endl;
+
+
+
+  RealD mflops = 144*volume*DerivCalls/DerivDhopComputeTime;
+  std::cout << GridLogMessage << "Average mflops/s per call                : " << mflops << std::endl;
+  std::cout << GridLogMessage << "Average mflops/s per call per node       : " << mflops/NP << std::endl;
+
+  }
+
+  if (DerivCalls > 0 || DhopCalls > 0){
+  std::cout << GridLogMessage << "WilsonFermion5D Stencil"<<std::endl;  Stencil.Report();
+  std::cout << GridLogMessage << "WilsonFermion5D StencilEven"<<std::endl;  StencilEven.Report();
+  std::cout << GridLogMessage << "WilsonFermion5D StencilOdd"<<std::endl;  StencilOdd.Report();
+  }
+}
+
+template<class Impl>
+void WilsonFermion5D<Impl>::ZeroCounters(void) {
+  DhopCalls       = 0;
+  DhopCommTime    = 0;
+  DhopComputeTime = 0;
+
+  DerivCalls       = 0;
+  DerivCommTime    = 0;
+  DerivComputeTime = 0;
+  DerivDhopComputeTime = 0;
+
+  Stencil.ZeroCounters();
+  StencilEven.ZeroCounters();
+  StencilOdd.ZeroCounters();
+}
+
+
 template<class Impl>
 void WilsonFermion5D<Impl>::ImportGauge(const GaugeField &_Umu)
 {
@@ -215,12 +282,13 @@ PARALLEL_FOR_LOOP
 
 template<class Impl>
 void WilsonFermion5D<Impl>::DerivInternal(StencilImpl & st,
-					  DoubledGaugeField & U,
+            DoubledGaugeField & U,
-					  GaugeField &mat,
+            GaugeField &mat,
-					  const FermionField &A,
+            const FermionField &A,
-					  const FermionField &B,
+            const FermionField &B,
-					  int dag)
+            int dag)
 {
+  DerivCalls++;
   assert((dag==DaggerNo) ||(dag==DaggerYes));
 
   conformable(st._grid,A._grid);
@@ -231,51 +299,53 @@ void WilsonFermion5D<Impl>::DerivInternal(StencilImpl & st,
   FermionField Btilde(B._grid);
   FermionField Atilde(B._grid);
 
+  DerivCommTime-=usecond();
   st.HaloExchange(B,compressor);
+  DerivCommTime+=usecond();
 
   Atilde=A;
 
-  for(int mu=0;mu<Nd;mu++){
+  DerivComputeTime-=usecond();
-      
+  for (int mu = 0; mu < Nd; mu++) {
     ////////////////////////////////////////////////////////////////////////
     // Flip gamma if dag
     ////////////////////////////////////////////////////////////////////////
     int gamma = mu;
-    if ( !dag ) gamma+= Nd;
+    if (!dag) gamma += Nd;
 
     ////////////////////////
     // Call the single hop
     ////////////////////////
 
-PARALLEL_FOR_LOOP
+    DerivDhopComputeTime -= usecond();
-    for(int sss=0;sss<U._grid->oSites();sss++){
+    PARALLEL_FOR_LOOP
-      for(int s=0;s<Ls;s++){
+    for (int sss = 0; sss < U._grid->oSites(); sss++) {
-	int sU=sss;
+      for (int s = 0; s < Ls; s++) {
-	int sF = s+Ls*sU;
+        int sU = sss;
+        int sF = s + Ls * sU;
 
-	assert ( sF< B._grid->oSites());
+        assert(sF < B._grid->oSites());
-	assert ( sU< U._grid->oSites());
+        assert(sU < U._grid->oSites());
 
-	Kernels::DiracOptDhopDir(st,U,st.comm_buf,sF,sU,B,Btilde,mu,gamma);
+        Kernels::DiracOptDhopDir(st, U, st.comm_buf, sF, sU, B, Btilde, mu,
-
+                                 gamma);
-    ////////////////////////////
-    // spin trace outer product
-    ////////////////////////////
 
+        ////////////////////////////
+        // spin trace outer product
+        ////////////////////////////
       }
-
     }
-
+    DerivDhopComputeTime += usecond();
-    Impl::InsertForce5D(mat,Btilde,Atilde,mu);
+    Impl::InsertForce5D(mat, Btilde, Atilde, mu);
-
   }
+  DerivComputeTime += usecond();
 }
 
 template<class Impl>
 void WilsonFermion5D<Impl>::DhopDeriv(      GaugeField &mat,
-					    const FermionField &A,
+              const FermionField &A,
-					    const FermionField &B,
+              const FermionField &B,
-					    int dag)
+              int dag)
 {
   conformable(A._grid,FermionGrid());  
   conformable(A._grid,B._grid);
@@ -288,9 +358,9 @@ void WilsonFermion5D<Impl>::DhopDeriv(      GaugeField &mat,
 
 template<class Impl>
 void WilsonFermion5D<Impl>::DhopDerivEO(GaugeField &mat,
-					const FermionField &A,
+          const FermionField &A,
-					const FermionField &B,
+          const FermionField &B,
-					int dag)
+          int dag)
 {
   conformable(A._grid,FermionRedBlackGrid());
   conformable(GaugeRedBlackGrid(),mat._grid);
@@ -306,9 +376,9 @@ void WilsonFermion5D<Impl>::DhopDerivEO(GaugeField &mat,
 
 template<class Impl>
 void WilsonFermion5D<Impl>::DhopDerivOE(GaugeField &mat,
-				  const FermionField &A,
+          const FermionField &A,
-				  const FermionField &B,
+          const FermionField &B,
-				  int dag)
+          int dag)
 {
   conformable(A._grid,FermionRedBlackGrid());
   conformable(GaugeRedBlackGrid(),mat._grid);
@@ -323,32 +393,61 @@ void WilsonFermion5D<Impl>::DhopDerivOE(GaugeField &mat,
 
 template<class Impl>
 void WilsonFermion5D<Impl>::DhopInternal(StencilImpl & st, LebesgueOrder &lo,
-					 DoubledGaugeField & U,
+           DoubledGaugeField & U,
-					 const FermionField &in, FermionField &out,int dag)
+           const FermionField &in, FermionField &out,int dag)
 {
+  DhopCalls++;
   //  assert((dag==DaggerNo) ||(dag==DaggerYes));
   Compressor compressor(dag);
 
   int LLs = in._grid->_rdimensions[0];
   
+  DhopCommTime-=usecond();
   st.HaloExchange(in,compressor);
+  DhopCommTime+=usecond();
   
+  DhopComputeTime-=usecond();
   // Dhop takes the 4d grid from U, and makes a 5d index for fermion
-  if ( dag == DaggerYes ) {
+  if (dag == DaggerYes) {
-PARALLEL_FOR_LOOP
+    PARALLEL_FOR_LOOP
-    for(int ss=0;ss<U._grid->oSites();ss++){
+    for (int ss = 0; ss < U._grid->oSites(); ss++) {
-	int sU=ss;
+      int sU = ss;
-	int sF=LLs*sU;
+      int sF = LLs * sU;
-	Kernels::DiracOptDhopSiteDag(st,lo,U,st.comm_buf,sF,sU,LLs,1,in,out);
+      Kernels::DiracOptDhopSiteDag(st, lo, U, st.comm_buf, sF, sU, LLs, 1, in,
+                                   out);
     }
+#ifdef AVX512
+  } else if (stat.is_init() ) {
+
+    int nthreads;
+    stat.start();
+    #pragma omp parallel
+    {
+    #pragma omp master
+    nthreads = omp_get_num_threads();
+    int mythread = omp_get_thread_num();
+    stat.enter(mythread);
+    #pragma omp for nowait
+   for(int ss=0;ss<U._grid->oSites();ss++)
+    {
+       int sU=ss;
+       int sF=LLs*sU;
+       Kernels::DiracOptDhopSite(st,lo,U,st.comm_buf,sF,sU,LLs,1,in,out);
+     }
+    stat.exit(mythread);
+    }
+    stat.accum(nthreads);
+#endif
   } else {
-PARALLEL_FOR_LOOP
+    PARALLEL_FOR_LOOP
-    for(int ss=0;ss<U._grid->oSites();ss++){
+    for (int ss = 0; ss < U._grid->oSites(); ss++) {
-      int sU=ss;
+      int sU = ss;
-      int sF=LLs*sU;
+      int sF = LLs * sU;
-      Kernels::DiracOptDhopSite(st,lo,U,st.comm_buf,sF,sU,LLs,1,in,out);
+      Kernels::DiracOptDhopSite(st, lo, U, st.comm_buf, sF, sU, LLs, 1, in,
+                                out);
     }
   }
+  DhopComputeTime+=usecond();
 }
 
 

lib/qcd/action/fermion/WilsonFermion5D.h

@@ -31,6 +31,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #ifndef  GRID_QCD_WILSON_FERMION_5D_H
 #define  GRID_QCD_WILSON_FERMION_5D_H
 
+#include <Grid/Stat.h>
+
 namespace Grid {
 
   namespace QCD {
@@ -60,6 +62,18 @@ namespace Grid {
     public:
      INHERIT_IMPL_TYPES(Impl);
      typedef WilsonKernels<Impl> Kernels;
+     PmuStat stat;
+
+     void Report(void);
+     void ZeroCounters(void);
+     double DhopCalls;
+     double DhopCommTime;
+     double DhopComputeTime;
+
+     double DerivCalls;
+     double DerivCommTime;
+     double DerivComputeTime;
+     double DerivDhopComputeTime;
 
       ///////////////////////////////////////////////////////////////
       // Implement the abstract base

lib/qcd/action/fermion/WilsonKernels.cc

@@ -1,103 +1,54 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
+Source file: ./lib/qcd/action/fermion/WilsonKernels.cc
 
-    Copyright (C) 2015
+Copyright (C) 2015
 
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
 Author: paboyle <paboyle@ph.ed.ac.uk>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution
-    *************************************************************************************/
+directory
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #include <Grid.h>
 namespace Grid {
 namespace QCD {
 
-  int WilsonKernelsStatic::HandOpt;
+int WilsonKernelsStatic::HandOpt;
-  int WilsonKernelsStatic::AsmOpt;
+int WilsonKernelsStatic::AsmOpt;
 
-template<class Impl> 
+template <class Impl>
-WilsonKernels<Impl>::WilsonKernels(const ImplParams &p): Base(p) {};
+WilsonKernels<Impl>::WilsonKernels(const ImplParams &p) : Base(p){};
 
-template<class Impl> 
+////////////////////////////////////////////
-void WilsonKernels<Impl>::DiracOptDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
+// Generic implementation; move to different file?
-						  std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+////////////////////////////////////////////
-						  int sF,int sU,int Ls, int Ns, const FermionField &in, FermionField &out)
-{
-#ifdef AVX512
-  if ( AsmOpt ) {
 
-    WilsonKernels<Impl>::DiracOptAsmDhopSite(st,lo,U,buf,sF,sU,Ls,Ns,in,out);
+template <class Impl>
-
+void WilsonKernels<Impl>::DiracOptGenericDhopSiteDag(
-  } else {
+    StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
-#else
+    std::vector<SiteHalfSpinor, alignedAllocator<SiteHalfSpinor> > &buf, int sF,
-  {  
+    int sU, const FermionField &in, FermionField &out) {
-#endif
+  SiteHalfSpinor tmp;
-    for(int site=0;site<Ns;site++) {
+  SiteHalfSpinor chi;
-      for(int s=0;s<Ls;s++) {
-	if (HandOpt) WilsonKernels<Impl>::DiracOptHandDhopSite(st,lo,U,buf,sF,sU,in,out);
-	else         WilsonKernels<Impl>::DiracOptGenericDhopSite(st,lo,U,buf,sF,sU,in,out);
-	sF++;
-      }
-      sU++;
-    }
-
-  }
-}
-
-template<class Impl> 
-void WilsonKernels<Impl>::DiracOptDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
-					   std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-					   int sF,int sU,int Ls, int Ns, const FermionField &in, FermionField &out)
-{
-#ifdef AVX512
-  if ( AsmOpt ) {
-    WilsonKernels<Impl>::DiracOptAsmDhopSiteDag(st,lo,U,buf,sF,sU,Ls,Ns,in,out);
-  } else {
-#else
-  {  
-#endif
-    for(int site=0;site<Ns;site++) {
-      for(int s=0;s<Ls;s++) {
-	if (HandOpt) WilsonKernels<Impl>::DiracOptHandDhopSiteDag(st,lo,U,buf,sF,sU,in,out);
-	else         WilsonKernels<Impl>::DiracOptGenericDhopSiteDag(st,lo,U,buf,sF,sU,in,out);
-	sF++;
-      }
-      sU++;
-    }
-  }
-}
-
-
-  ////////////////////////////////////////////
-  // Generic implementation; move to different file?
-  ////////////////////////////////////////////
-
-template<class Impl> 
-void WilsonKernels<Impl>::DiracOptGenericDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
-					   std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-					   int sF,int sU,const FermionField &in, FermionField &out)
-{
-  SiteHalfSpinor  tmp;    
-  SiteHalfSpinor  chi;    
   SiteHalfSpinor *chi_p;
   SiteHalfSpinor Uchi;
   SiteSpinor result;
@@ -107,176 +58,175 @@ void WilsonKernels<Impl>::DiracOptGenericDhopSiteDag(StencilImpl &st,LebesgueOrd
   ///////////////////////////
   // Xp
   ///////////////////////////
-  SE=st.GetEntry(ptype,Xp,sF);
+  SE = st.GetEntry(ptype, Xp, sF);
 
-  if (SE->_is_local ) { 
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjXp(tmp,in._odata[SE->_offset]);
+      spProjXp(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
     } else {
-      spProjXp(chi,in._odata[SE->_offset]);
+      spProjXp(chi, in._odata[SE->_offset]);
     }
-  } else { 
+  } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
-  
+
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Xp,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Xp, SE, st);
-  spReconXp(result,Uchi);
+  spReconXp(result, Uchi);
-    
+
   ///////////////////////////
   // Yp
   ///////////////////////////
-  SE=st.GetEntry(ptype,Yp,sF);
+  SE = st.GetEntry(ptype, Yp, sF);
 
-  if ( SE->_is_local ) { 
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjYp(tmp,in._odata[SE->_offset]);
+      spProjYp(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
     } else {
-      spProjYp(chi,in._odata[SE->_offset]);
+      spProjYp(chi, in._odata[SE->_offset]);
     }
-  } else { 
+  } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
 
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Yp,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Yp, SE, st);
-  accumReconYp(result,Uchi);
+  accumReconYp(result, Uchi);
 
   ///////////////////////////
   // Zp
   ///////////////////////////
-  SE=st.GetEntry(ptype,Zp,sF);
+  SE = st.GetEntry(ptype, Zp, sF);
 
-  if ( SE->_is_local ) { 
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjZp(tmp,in._odata[SE->_offset]);
+      spProjZp(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
     } else {
-      spProjZp(chi,in._odata[SE->_offset]);
+      spProjZp(chi, in._odata[SE->_offset]);
     }
-  } else { 
+  } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
 
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Zp,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Zp, SE, st);
-  accumReconZp(result,Uchi);
+  accumReconZp(result, Uchi);
 
   ///////////////////////////
   // Tp
   ///////////////////////////
-  SE=st.GetEntry(ptype,Tp,sF);
+  SE = st.GetEntry(ptype, Tp, sF);
 
-  if ( SE->_is_local ) {
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjTp(tmp,in._odata[SE->_offset]);
+      spProjTp(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
     } else {
-      spProjTp(chi,in._odata[SE->_offset]);
+      spProjTp(chi, in._odata[SE->_offset]);
     }
   } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
 
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Tp,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Tp, SE, st);
-  accumReconTp(result,Uchi);
+  accumReconTp(result, Uchi);
 
   ///////////////////////////
   // Xm
   ///////////////////////////
-  SE=st.GetEntry(ptype,Xm,sF);
+  SE = st.GetEntry(ptype, Xm, sF);
 
-  if ( SE->_is_local ) {
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjXm(tmp,in._odata[SE->_offset]);
+      spProjXm(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
     } else {
-      spProjXm(chi,in._odata[SE->_offset]);
+      spProjXm(chi, in._odata[SE->_offset]);
     }
   } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
 
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Xm,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Xm, SE, st);
-  accumReconXm(result,Uchi);
+  accumReconXm(result, Uchi);
-  
+
   ///////////////////////////
   // Ym
   ///////////////////////////
-  SE=st.GetEntry(ptype,Ym,sF);
+  SE = st.GetEntry(ptype, Ym, sF);
 
-  if ( SE->_is_local ) {
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjYm(tmp,in._odata[SE->_offset]);
+      spProjYm(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
     } else {
-      spProjYm(chi,in._odata[SE->_offset]);
+      spProjYm(chi, in._odata[SE->_offset]);
     }
   } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
 
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Ym,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Ym, SE, st);
-  accumReconYm(result,Uchi);
+  accumReconYm(result, Uchi);
-  
+
   ///////////////////////////
   // Zm
   ///////////////////////////
-  SE=st.GetEntry(ptype,Zm,sF);
+  SE = st.GetEntry(ptype, Zm, sF);
 
-  if ( SE->_is_local ) {
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjZm(tmp,in._odata[SE->_offset]);
+      spProjZm(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
     } else {
-      spProjZm(chi,in._odata[SE->_offset]);
+      spProjZm(chi, in._odata[SE->_offset]);
     }
   } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
 
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Zm,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Zm, SE, st);
-  accumReconZm(result,Uchi);
+  accumReconZm(result, Uchi);
 
   ///////////////////////////
   // Tm
   ///////////////////////////
-  SE=st.GetEntry(ptype,Tm,sF);
+  SE = st.GetEntry(ptype, Tm, sF);
 
-  if ( SE->_is_local ) {
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjTm(tmp,in._odata[SE->_offset]);
+      spProjTm(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
-    } else { 
+    } else {
-      spProjTm(chi,in._odata[SE->_offset]);
+      spProjTm(chi, in._odata[SE->_offset]);
     }
   } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
 
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Tm,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Tm, SE, st);
-  accumReconTm(result,Uchi);
+  accumReconTm(result, Uchi);
 
-  vstream(out._odata[sF],result);
+  vstream(out._odata[sF], result);
 };
 
-
+// Need controls to do interior, exterior, or both
-  // Need controls to do interior, exterior, or both
+template <class Impl>
-template<class Impl> 
+void WilsonKernels<Impl>::DiracOptGenericDhopSite(
-void WilsonKernels<Impl>::DiracOptGenericDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
+    StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
-						  std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+    std::vector<SiteHalfSpinor, alignedAllocator<SiteHalfSpinor> > &buf, int sF,
-						  int sF,int sU,const FermionField &in, FermionField &out)
+    int sU, const FermionField &in, FermionField &out) {
-{
+  SiteHalfSpinor tmp;
-  SiteHalfSpinor  tmp;    
+  SiteHalfSpinor chi;
-  SiteHalfSpinor  chi;    
+  SiteHalfSpinor *chi_p;
-  SiteHalfSpinor *chi_p;    
   SiteHalfSpinor Uchi;
   SiteSpinor result;
   StencilEntry *SE;
@@ -285,296 +235,298 @@ void WilsonKernels<Impl>::DiracOptGenericDhopSite(StencilImpl &st,LebesgueOrder
   ///////////////////////////
   // Xp
   ///////////////////////////
-  SE=st.GetEntry(ptype,Xm,sF);
+  SE = st.GetEntry(ptype, Xm, sF);
 
-  if ( SE->_is_local ) { 
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjXp(tmp,in._odata[SE->_offset]);
+      spProjXp(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
     } else {
-      spProjXp(chi,in._odata[SE->_offset]);
+      spProjXp(chi, in._odata[SE->_offset]);
     }
-  } else { 
+  } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
-  
+
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Xm,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Xm, SE, st);
-  spReconXp(result,Uchi);
+  spReconXp(result, Uchi);
-    
+
   ///////////////////////////
   // Yp
   ///////////////////////////
-  SE=st.GetEntry(ptype,Ym,sF);
+  SE = st.GetEntry(ptype, Ym, sF);
 
-  if ( SE->_is_local ) { 
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjYp(tmp,in._odata[SE->_offset]);
+      spProjYp(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
     } else {
-      spProjYp(chi,in._odata[SE->_offset]);
+      spProjYp(chi, in._odata[SE->_offset]);
     }
-  } else { 
+  } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
 
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Ym,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Ym, SE, st);
-  accumReconYp(result,Uchi);
+  accumReconYp(result, Uchi);
 
   ///////////////////////////
   // Zp
   ///////////////////////////
-  SE=st.GetEntry(ptype,Zm,sF);
+  SE = st.GetEntry(ptype, Zm, sF);
 
-  if ( SE->_is_local ) { 
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjZp(tmp,in._odata[SE->_offset]);
+      spProjZp(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
     } else {
-      spProjZp(chi,in._odata[SE->_offset]);
+      spProjZp(chi, in._odata[SE->_offset]);
     }
-  } else { 
+  } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
 
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Zm,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Zm, SE, st);
-  accumReconZp(result,Uchi);
+  accumReconZp(result, Uchi);
 
   ///////////////////////////
   // Tp
   ///////////////////////////
-  SE=st.GetEntry(ptype,Tm,sF);
+  SE = st.GetEntry(ptype, Tm, sF);
 
-  if ( SE->_is_local ) {
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjTp(tmp,in._odata[SE->_offset]);
+      spProjTp(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
     } else {
-      spProjTp(chi,in._odata[SE->_offset]);
+      spProjTp(chi, in._odata[SE->_offset]);
     }
   } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
 
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Tm,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Tm, SE, st);
-  accumReconTp(result,Uchi);
+  accumReconTp(result, Uchi);
 
   ///////////////////////////
   // Xm
   ///////////////////////////
-  SE=st.GetEntry(ptype,Xp,sF);
+  SE = st.GetEntry(ptype, Xp, sF);
 
-  if ( SE->_is_local ) {
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjXm(tmp,in._odata[SE->_offset]);
+      spProjXm(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
     } else {
-      spProjXm(chi,in._odata[SE->_offset]);
+      spProjXm(chi, in._odata[SE->_offset]);
     }
   } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
 
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Xp,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Xp, SE, st);
-  accumReconXm(result,Uchi);
+  accumReconXm(result, Uchi);
 
   ///////////////////////////
   // Ym
   ///////////////////////////
-  SE=st.GetEntry(ptype,Yp,sF);
+  SE = st.GetEntry(ptype, Yp, sF);
 
-  if ( SE->_is_local ) {
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjYm(tmp,in._odata[SE->_offset]);
+      spProjYm(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
     } else {
-      spProjYm(chi,in._odata[SE->_offset]);
+      spProjYm(chi, in._odata[SE->_offset]);
     }
   } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
 
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Yp,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Yp, SE, st);
-  accumReconYm(result,Uchi);
+  accumReconYm(result, Uchi);
-  
+
   ///////////////////////////
   // Zm
   ///////////////////////////
-  SE=st.GetEntry(ptype,Zp,sF);
+  SE = st.GetEntry(ptype, Zp, sF);
 
-  if ( SE->_is_local ) {
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjZm(tmp,in._odata[SE->_offset]);
+      spProjZm(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
     } else {
-      spProjZm(chi,in._odata[SE->_offset]);
+      spProjZm(chi, in._odata[SE->_offset]);
     }
   } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
 
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Zp,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Zp, SE, st);
-  accumReconZm(result,Uchi);
+  accumReconZm(result, Uchi);
 
   ///////////////////////////
   // Tm
   ///////////////////////////
-  SE=st.GetEntry(ptype,Tp,sF);
+  SE = st.GetEntry(ptype, Tp, sF);
 
-  if ( SE->_is_local ) {
+  if (SE->_is_local) {
     chi_p = &chi;
-    if ( SE->_permute ) {
+    if (SE->_permute) {
-      spProjTm(tmp,in._odata[SE->_offset]);
+      spProjTm(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
-    } else { 
+    } else {
-      spProjTm(chi,in._odata[SE->_offset]);
+      spProjTm(chi, in._odata[SE->_offset]);
     }
   } else {
-    chi_p=&buf[SE->_offset];
+    chi_p = &buf[SE->_offset];
   }
 
-  Impl::multLink(Uchi,U._odata[sU],*chi_p,Tp,SE,st);
+  Impl::multLink(Uchi, U._odata[sU], *chi_p, Tp, SE, st);
-  accumReconTm(result,Uchi);
+  accumReconTm(result, Uchi);
 
-  vstream(out._odata[sF],result);
+  vstream(out._odata[sF], result);
 };
 
-template<class Impl> 
+template <class Impl>
-void WilsonKernels<Impl>::DiracOptDhopDir(StencilImpl &st,DoubledGaugeField &U,
+void WilsonKernels<Impl>::DiracOptDhopDir(
-					  std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+    StencilImpl &st, DoubledGaugeField &U,
-					  int sF,int sU,const FermionField &in, FermionField &out,int dir,int gamma)
+    std::vector<SiteHalfSpinor, alignedAllocator<SiteHalfSpinor> > &buf, int sF,
-{
+    int sU, const FermionField &in, FermionField &out, int dir, int gamma) {
-  SiteHalfSpinor  tmp;    
+  SiteHalfSpinor tmp;
-  SiteHalfSpinor  chi;    
+  SiteHalfSpinor chi;
-  SiteSpinor   result;
+  SiteSpinor result;
   SiteHalfSpinor Uchi;
   StencilEntry *SE;
   int ptype;
 
-  SE=st.GetEntry(ptype,dir,sF);
+  SE = st.GetEntry(ptype, dir, sF);
 
   // Xp
-  if(gamma==Xp){
+  if (gamma == Xp) {
-    if (  SE->_is_local && SE->_permute ) {
+    if (SE->_is_local && SE->_permute) {
-      spProjXp(tmp,in._odata[SE->_offset]);
+      spProjXp(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
-    } else if ( SE->_is_local ) {
+    } else if (SE->_is_local) {
-      spProjXp(chi,in._odata[SE->_offset]);
+      spProjXp(chi, in._odata[SE->_offset]);
-    } else { 
+    } else {
-      chi=buf[SE->_offset];
+      chi = buf[SE->_offset];
     }
-    Impl::multLink(Uchi,U._odata[sU],chi,dir,SE,st);
+    Impl::multLink(Uchi, U._odata[sU], chi, dir, SE, st);
-    spReconXp(result,Uchi);
+    spReconXp(result, Uchi);
   }
 
   // Yp
-  if ( gamma==Yp ){
+  if (gamma == Yp) {
-    if (  SE->_is_local && SE->_permute ) {
+    if (SE->_is_local && SE->_permute) {
-      spProjYp(tmp,in._odata[SE->_offset]);
+      spProjYp(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
-    } else if ( SE->_is_local ) {
+    } else if (SE->_is_local) {
-      spProjYp(chi,in._odata[SE->_offset]);
+      spProjYp(chi, in._odata[SE->_offset]);
-    } else { 
+    } else {
-      chi=buf[SE->_offset];
+      chi = buf[SE->_offset];
     }
-    Impl::multLink(Uchi,U._odata[sU],chi,dir,SE,st);
+    Impl::multLink(Uchi, U._odata[sU], chi, dir, SE, st);
-    spReconYp(result,Uchi);
+    spReconYp(result, Uchi);
   }
-  
+
   // Zp
-  if ( gamma ==Zp ){
+  if (gamma == Zp) {
-    if (  SE->_is_local && SE->_permute ) {
+    if (SE->_is_local && SE->_permute) {
-      spProjZp(tmp,in._odata[SE->_offset]);
+      spProjZp(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
-    } else if ( SE->_is_local ) {
+    } else if (SE->_is_local) {
-      spProjZp(chi,in._odata[SE->_offset]);
+      spProjZp(chi, in._odata[SE->_offset]);
-    } else { 
+    } else {
-      chi=buf[SE->_offset];
+      chi = buf[SE->_offset];
     }
-    Impl::multLink(Uchi,U._odata[sU],chi,dir,SE,st);
+    Impl::multLink(Uchi, U._odata[sU], chi, dir, SE, st);
-    spReconZp(result,Uchi);
+    spReconZp(result, Uchi);
   }
-  
+
   // Tp
-  if ( gamma ==Tp ){
+  if (gamma == Tp) {
-    if (  SE->_is_local && SE->_permute ) {
+    if (SE->_is_local && SE->_permute) {
-      spProjTp(tmp,in._odata[SE->_offset]);
+      spProjTp(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
-    } else if ( SE->_is_local ) {
+    } else if (SE->_is_local) {
-      spProjTp(chi,in._odata[SE->_offset]);
+      spProjTp(chi, in._odata[SE->_offset]);
-    } else { 
+    } else {
-      chi=buf[SE->_offset];
+      chi = buf[SE->_offset];
     }
-    Impl::multLink(Uchi,U._odata[sU],chi,dir,SE,st);
+    Impl::multLink(Uchi, U._odata[sU], chi, dir, SE, st);
-    spReconTp(result,Uchi);
+    spReconTp(result, Uchi);
   }
 
   // Xm
-  if ( gamma==Xm ){
+  if (gamma == Xm) {
-    if (  SE->_is_local && SE->_permute ) {
+    if (SE->_is_local && SE->_permute) {
-      spProjXm(tmp,in._odata[SE->_offset]);
+      spProjXm(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
-    } else if ( SE->_is_local ) {
+    } else if (SE->_is_local) {
-      spProjXm(chi,in._odata[SE->_offset]);
+      spProjXm(chi, in._odata[SE->_offset]);
-    } else { 
+    } else {
-      chi=buf[SE->_offset];
+      chi = buf[SE->_offset];
     }
-    Impl::multLink(Uchi,U._odata[sU],chi,dir,SE,st);
+    Impl::multLink(Uchi, U._odata[sU], chi, dir, SE, st);
-    spReconXm(result,Uchi);
+    spReconXm(result, Uchi);
   }
 
   // Ym
-  if ( gamma == Ym ){
+  if (gamma == Ym) {
-    if (  SE->_is_local && SE->_permute ) {
+    if (SE->_is_local && SE->_permute) {
-      spProjYm(tmp,in._odata[SE->_offset]);
+      spProjYm(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
-    } else if ( SE->_is_local ) {
+    } else if (SE->_is_local) {
-      spProjYm(chi,in._odata[SE->_offset]);
+      spProjYm(chi, in._odata[SE->_offset]);
-    } else { 
+    } else {
-      chi=buf[SE->_offset];
+      chi = buf[SE->_offset];
     }
-    Impl::multLink(Uchi,U._odata[sU],chi,dir,SE,st);
+    Impl::multLink(Uchi, U._odata[sU], chi, dir, SE, st);
-    spReconYm(result,Uchi);
+    spReconYm(result, Uchi);
   }
 
   // Zm
-  if ( gamma == Zm ){
+  if (gamma == Zm) {
-    if (  SE->_is_local && SE->_permute ) {
+    if (SE->_is_local && SE->_permute) {
-      spProjZm(tmp,in._odata[SE->_offset]);
+      spProjZm(tmp, in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
+      permute(chi, tmp, ptype);
-    } else if ( SE->_is_local ) {
+    } else if (SE->_is_local) {
-      spProjZm(chi,in._odata[SE->_offset]);
+      spProjZm(chi, in._odata[SE->_offset]);
-    } else { 
+    } else {
-      chi=buf[SE->_offset];
+      chi = buf[SE->_offset];
     }
-    Impl::multLink(Uchi,U._odata[sU],chi,dir,SE,st);
+    Impl::multLink(Uchi, U._odata[sU], chi, dir, SE, st);
-    spReconZm(result,Uchi);
+    spReconZm(result, Uchi);
-  }
-  
-  // Tm
-  if ( gamma==Tm ) {
-    if (  SE->_is_local && SE->_permute ) {
-      spProjTm(tmp,in._odata[SE->_offset]);
-      permute(chi,tmp,ptype);
-    } else if ( SE->_is_local ) {
-      spProjTm(chi,in._odata[SE->_offset]);
-    } else { 
-      chi=buf[SE->_offset];
-    }
-    Impl::multLink(Uchi,U._odata[sU],chi,dir,SE,st);
-    spReconTm(result,Uchi);
   }
 
-  vstream(out._odata[sF],result);
+  // Tm
+  if (gamma == Tm) {
+    if (SE->_is_local && SE->_permute) {
+      spProjTm(tmp, in._odata[SE->_offset]);
+      permute(chi, tmp, ptype);
+    } else if (SE->_is_local) {
+      spProjTm(chi, in._odata[SE->_offset]);
+    } else {
+      chi = buf[SE->_offset];
+    }
+    Impl::multLink(Uchi, U._odata[sU], chi, dir, SE, st);
+    spReconTm(result, Uchi);
+  }
+
+  vstream(out._odata[sF], result);
 }
 
-
+FermOpTemplateInstantiate(WilsonKernels);
-  FermOpTemplateInstantiate(WilsonKernels);
+AdjointFermOpTemplateInstantiate(WilsonKernels);
+TwoIndexFermOpTemplateInstantiate(WilsonKernels);
 
 }}
+

lib/qcd/action/fermion/WilsonKernels.h

@@ -1,34 +1,35 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./lib/qcd/action/fermion/WilsonKernels.h
+Source file: ./lib/qcd/action/fermion/WilsonKernels.h
 
-    Copyright (C) 2015
+Copyright (C) 2015
 
 Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: paboyle <paboyle@ph.ed.ac.uk>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution
-    *************************************************************************************/
+directory
-    /*  END LEGAL */
+*************************************************************************************/
-#ifndef  GRID_QCD_DHOP_H
+/*  END LEGAL */
-#define  GRID_QCD_DHOP_H
+#ifndef GRID_QCD_DHOP_H
+#define GRID_QCD_DHOP_H
 
 namespace Grid {
 
@@ -48,55 +49,158 @@ namespace Grid {
     template<class Impl> class WilsonKernels : public FermionOperator<Impl> , public WilsonKernelsStatic { 
     public:
 
-     INHERIT_IMPL_TYPES(Impl);
+      INHERIT_IMPL_TYPES(Impl);
-     typedef FermionOperator<Impl> Base;
+      typedef FermionOperator<Impl> Base;
      
     public:
 
-     void DiracOptDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
+      template <bool EnableBool = true>
-			   std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+      typename std::enable_if<Impl::Dimension == 3 && Nc == 3 &&EnableBool, void>::type
-			   int sF, int sU,int Ls, int Ns, const FermionField &in, FermionField &out);
+	DiracOptDhopSite(
-      
+			 StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
-     void DiracOptDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
+			 std::vector<SiteHalfSpinor, alignedAllocator<SiteHalfSpinor> > &buf,
-			      std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+			 int sF, int sU, int Ls, int Ns, const FermionField &in,
-			      int sF,int sU,int Ls, int Ns, const FermionField &in,FermionField &out);
+			 FermionField &out) {
+#ifdef AVX512
+	if (AsmOpt) {
+	  WilsonKernels<Impl>::DiracOptAsmDhopSite(st, lo, U, buf, sF, sU, Ls, Ns,
+						   in, out);
+
+	} else {
+#else
+	  {
+#endif
+	    for (int site = 0; site < Ns; site++) {
+	      for (int s = 0; s < Ls; s++) {
+		if (HandOpt)
+		  WilsonKernels<Impl>::DiracOptHandDhopSite(st, lo, U, buf, sF, sU,
+							    in, out);
+		else
+		  WilsonKernels<Impl>::DiracOptGenericDhopSite(st, lo, U, buf, sF, sU,
+							       in, out);
+		sF++;
+	      }
+	      sU++;
+	    }
+	  }
+	}
+
+	template <bool EnableBool = true>
+	  typename std::enable_if<(Impl::Dimension != 3 || (Impl::Dimension == 3 && Nc != 3)) && EnableBool, void>::type
+	  DiracOptDhopSite(
+			   StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
+			   std::vector<SiteHalfSpinor, alignedAllocator<SiteHalfSpinor> > &buf,
+			   int sF, int sU, int Ls, int Ns, const FermionField &in,
+			   FermionField &out) {
+	  for (int site = 0; site < Ns; site++) {
+	    for (int s = 0; s < Ls; s++) {
+	      WilsonKernels<Impl>::DiracOptGenericDhopSite(st, lo, U, buf, sF, sU, in,
+							   out);
+	      sF++;
+	    }
+	    sU++;
+	  }
+	}
+
+	template <bool EnableBool = true>
+	  typename std::enable_if<Impl::Dimension == 3 && Nc == 3 && EnableBool,
+				  void>::type
+	  DiracOptDhopSiteDag(
+			      StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
+			      std::vector<SiteHalfSpinor, alignedAllocator<SiteHalfSpinor> > &buf,
+			      int sF, int sU, int Ls, int Ns, const FermionField &in,
+			      FermionField &out) {
+#ifdef AVX512
+				    if (AsmOpt) {
+				      WilsonKernels<Impl>::DiracOptAsmDhopSiteDag(st, lo, U, buf, sF, sU, Ls,
+										  Ns, in, out);
+				    } else {
+#else
+				      {
+#endif
+					for (int site = 0; site < Ns; site++) {
+					  for (int s = 0; s < Ls; s++) {
+					    if (HandOpt)
+					      WilsonKernels<Impl>::DiracOptHandDhopSiteDag(st, lo, U, buf, sF, sU,
+											   in, out);
+					    else
+					      WilsonKernels<Impl>::DiracOptGenericDhopSiteDag(st, lo, U, buf, sF,
+											      sU, in, out);
+					    sF++;
+					  }
+					  sU++;
+					}
+				      }
+				    }
+
+				    template <bool EnableBool = true>
+				      typename std::enable_if<
+				      (Impl::Dimension != 3 || (Impl::Dimension == 3 && Nc != 3)) && EnableBool,
+				      void>::type
+				      DiracOptDhopSiteDag(
+							  StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
+							  std::vector<SiteHalfSpinor, alignedAllocator<SiteHalfSpinor> > &buf,
+							  int sF, int sU, int Ls, int Ns, const FermionField &in,
+							  FermionField &out) {
+					for (int site = 0; site < Ns; site++) {
+					  for (int s = 0; s < Ls; s++) {
+					    WilsonKernels<Impl>::DiracOptGenericDhopSiteDag(st, lo, U, buf, sF, sU,
+											    in, out);
+					    sF++;
+					  }
+					  sU++;
+					}
+				      }
+
+				    void DiracOptDhopDir(
+							 StencilImpl &st, DoubledGaugeField &U,
+							 std::vector<SiteHalfSpinor, alignedAllocator<SiteHalfSpinor> > &buf,
+							 int sF, int sU, const FermionField &in, FermionField &out, int dirdisp,
+							 int gamma);
+
+	private:
+				    // Specialised variants
+				    void DiracOptGenericDhopSite(
+								 StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
+								 std::vector<SiteHalfSpinor, alignedAllocator<SiteHalfSpinor> > &buf,
+								 int sF, int sU, const FermionField &in, FermionField &out);
+
+				    void DiracOptGenericDhopSiteDag(
+								    StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
+								    std::vector<SiteHalfSpinor, alignedAllocator<SiteHalfSpinor> > &buf,
+								    int sF, int sU, const FermionField &in, FermionField &out);
+
+				    void DiracOptAsmDhopSite(
+							     StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
+							     std::vector<SiteHalfSpinor, alignedAllocator<SiteHalfSpinor> > &buf,
+							     int sF, int sU, int Ls, int Ns, const FermionField &in,
+							     FermionField &out);
+
+				    void DiracOptAsmDhopSiteDag(
+								StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
+								std::vector<SiteHalfSpinor, alignedAllocator<SiteHalfSpinor> > &buf,
+								int sF, int sU, int Ls, int Ns, const FermionField &in,
+								FermionField &out);
+
+				    void DiracOptHandDhopSite(
+							      StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
+							      std::vector<SiteHalfSpinor, alignedAllocator<SiteHalfSpinor> > &buf,
+							      int sF, int sU, const FermionField &in, FermionField &out);
+
+				    void DiracOptHandDhopSiteDag(
+								 StencilImpl &st, LebesgueOrder &lo, DoubledGaugeField &U,
+								 std::vector<SiteHalfSpinor, alignedAllocator<SiteHalfSpinor> > &buf,
+								 int sF, int sU, const FermionField &in, FermionField &out);
+
+	public:
+				    WilsonKernels(const ImplParams &p = ImplParams());
+				  };
+    
+      }
+    }
+
+
+
+
 
-     void DiracOptDhopDir(StencilImpl &st,DoubledGaugeField &U,
-			  std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-			  int sF,int sU,const FermionField &in, FermionField &out,int dirdisp,int gamma);
-
-    private:
-     // Specialised variants
-     void DiracOptGenericDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
-			   std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-			   int sF,int sU, const FermionField &in, FermionField &out);
-      
-     void DiracOptGenericDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
-			      std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-			      int sF,int sU,const FermionField &in,FermionField &out);
-
-     void DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
-			      std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-			      int sF,int sU,int Ls, int Ns, const FermionField &in, FermionField &out);
-
-     void DiracOptAsmDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
-			      std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-			      int sF,int sU,int Ls, int Ns, const FermionField &in, FermionField &out);
-
-
-     void DiracOptHandDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
-			      std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-			      int sF,int sU,const FermionField &in, FermionField &out);
-     
-     void DiracOptHandDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
-				 std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-				 int sF,int sU,const FermionField &in, FermionField &out);
-    public:
-
-     WilsonKernels(const ImplParams &p= ImplParams());
-     
-    };
-
-  }
-}
 #endif

lib/qcd/action/fermion/WilsonKernelsAsm.cc

@@ -1,4 +1,4 @@
-    /*************************************************************************************
+/*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -26,68 +26,77 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
     51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
     See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
+*************************************************************************************/
-    /*  END LEGAL */
+/*  END LEGAL */
 
 #include <Grid.h>
 
+
 namespace Grid {
-namespace QCD {
+  namespace QCD {
+    
+    ///////////////////////////////////////////////////////////
+    // Default to no assembler implementation
+    ///////////////////////////////////////////////////////////
+    template<class Impl>
+      void WilsonKernels<Impl >::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,
+                             std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+                             int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
+    {
+      assert(0);
+    }
+    template<class Impl>
+      void WilsonKernels<Impl >::DiracOptAsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,
+                                std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+                                int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
+    {
+      assert(0);
+    }
 
 
-  ///////////////////////////////////////////////////////////
-  // Default to no assembler implementation
-  ///////////////////////////////////////////////////////////
-template<class Impl>
-void WilsonKernels<Impl >::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,
-					       std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-					       int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
-{
-  assert(0);
-}
 
 #if defined(AVX512) 
-
+    
-
+    
-  ///////////////////////////////////////////////////////////
+    ///////////////////////////////////////////////////////////
-  // If we are AVX512 specialise the single precision routine
+    // If we are AVX512 specialise the single precision routine
-  ///////////////////////////////////////////////////////////
+    ///////////////////////////////////////////////////////////
-
+    
 #include <simd/Intel512wilson.h>
 #include <simd/Intel512single.h>
-
+    
-static Vector<vComplexF> signs;
+    static Vector<vComplexF> signs;
-
+    
-int setupSigns(void ){
+    int setupSigns(void ){
-  Vector<vComplexF> bother(2);
+      Vector<vComplexF> bother(2);
-  signs = bother;
+      signs = bother;
-  vrsign(signs[0]);
+      vrsign(signs[0]);
-  visign(signs[1]);
+      visign(signs[1]);
-  return 1;
+      return 1;
-}
+    }
-static int signInit = setupSigns();
+    static int signInit = setupSigns();
-
+  
 #define label(A)  ilabel(A)
 #define ilabel(A) ".globl\n"  #A ":\n" 
-
+  
 #define MAYBEPERM(A,perm) if (perm) { A ; }
 #define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN(ptr,pf)
 #define FX(A) WILSONASM_ ##A
-
+  
 #undef KERNEL_DAG
-template<>
+    template<>
-void WilsonKernels<WilsonImplF>::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,
+    void WilsonKernels<WilsonImplF>::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,
-						     std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+							 std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-						     int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
+							 int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
 #include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
+      
 #define KERNEL_DAG
-template<>
+    template<>
-void WilsonKernels<WilsonImplF>::DiracOptAsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,
+    void WilsonKernels<WilsonImplF>::DiracOptAsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,
-						     std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+							    std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-						     int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
+							    int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
 #include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
+				    
 #undef VMOVIDUP
 #undef VMOVRDUP
 #undef MAYBEPERM
@@ -98,43 +107,43 @@ void WilsonKernels<WilsonImplF>::DiracOptAsmDhopSiteDag(StencilImpl &st,Lebesgue
 #define VMOVIDUP(A,B,C)                                  VBCASTIDUPf(A,B,C)
 #define VMOVRDUP(A,B,C)                                  VBCASTRDUPf(A,B,C)
 #define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN_LS(ptr,pf)
-
+				    
 #undef KERNEL_DAG
-template<>
+    template<>
-void WilsonKernels<DomainWallVec5dImplF>::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,
+    void WilsonKernels<DomainWallVec5dImplF>::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,
-								   std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+								  std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-								   int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
+								  int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
 #include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
+				    
 #define KERNEL_DAG
-template<>
+    template<>
-void WilsonKernels<DomainWallVec5dImplF>::DiracOptAsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,
+    void WilsonKernels<DomainWallVec5dImplF>::DiracOptAsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,
-								   std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+								     std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-								   int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
+								     int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out)
 #include <qcd/action/fermion/WilsonKernelsAsmBody.h>
-
+				    
 #endif
 
 
+#define INSTANTIATE_ASM(A)\
+template void WilsonKernels<A>::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,\
+                                   std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,\
+                                  int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out);\
+template void WilsonKernels<A>::DiracOptAsmDhopSiteDag(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,\
+                                   std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,\
+                                  int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out);\
 
-template void WilsonKernels<WilsonImplF>::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,
-							       std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-							      int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out);		
 
-template void WilsonKernels<WilsonImplD>::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, 
+INSTANTIATE_ASM(WilsonImplF);
-							       std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+INSTANTIATE_ASM(WilsonImplD);
-							       int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out);		
+INSTANTIATE_ASM(ZWilsonImplF);
-template void WilsonKernels<GparityWilsonImplF>::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, 
+INSTANTIATE_ASM(ZWilsonImplD);
-							       std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+INSTANTIATE_ASM(GparityWilsonImplF);
-							       int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out);		
+INSTANTIATE_ASM(GparityWilsonImplD);
-template void WilsonKernels<GparityWilsonImplD>::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, 
+INSTANTIATE_ASM(DomainWallVec5dImplF);
-							       std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+INSTANTIATE_ASM(DomainWallVec5dImplD);
-							       int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out);		
+INSTANTIATE_ASM(ZDomainWallVec5dImplF);
-template void WilsonKernels<DomainWallVec5dImplF>::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, 
+INSTANTIATE_ASM(ZDomainWallVec5dImplD);
-							       std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
+  }
-							       int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out);		
+}
-template void WilsonKernels<DomainWallVec5dImplD>::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U, 
-							       std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
-							       int ss,int ssU,int Ls,int Ns,const FermionField &in, FermionField &out);		
-}}
 

lib/qcd/action/fermion/WilsonKernelsAsmBody.h

@@ -134,7 +134,9 @@
   ////////////////////////////////
   // Xm
   ////////////////////////////////
+#ifndef STREAM_STORE
   basep= (uint64_t) &out._odata[ss];
+#endif
   //  basep= st.GetPFInfo(nent,plocal); nent++;
   if ( local ) {
     LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
@@ -229,7 +231,9 @@
     LOAD_CHI(base);
   }
   base= (uint64_t) &out._odata[ss];
+#ifndef STREAM_STORE
   PREFETCH_CHIMU(base);
+#endif
   {
     MULT_2SPIN_DIR_PFTM(Tm,basep);
   }

lib/qcd/action/fermion/WilsonKernelsHand.cc

@@ -311,8 +311,8 @@ namespace Grid {
 namespace QCD {
 
 
-template<class Impl>
+  template<class Impl>
-void WilsonKernels<Impl >::DiracOptHandDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
+  void WilsonKernels<Impl>::DiracOptHandDhopSite(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
 					       std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
 					       int ss,int sU,const FermionField &in, FermionField &out)
 {
@@ -554,8 +554,8 @@ void WilsonKernels<Impl >::DiracOptHandDhopSite(StencilImpl &st,LebesgueOrder &l
   }
 }
 
-template<class Impl>
+  template<class Impl>
-void WilsonKernels<Impl >::DiracOptHandDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
+  void WilsonKernels<Impl>::DiracOptHandDhopSiteDag(StencilImpl &st,LebesgueOrder &lo,DoubledGaugeField &U,
 					       std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
 					       int ss,int sU,const FermionField &in, FermionField &out)
 {

lib/qcd/action/gauge/GaugeImpl.h

@@ -30,7 +30,6 @@ directory
 #define GRID_QCD_GAUGE_IMPL_H
 
 namespace Grid {
-
 namespace QCD {
 
 ////////////////////////////////////////////////////////////////////////
@@ -52,7 +51,7 @@ public:
   typedef S Simd;
 
   template <typename vtype>
-  using iImplGaugeLink = iScalar<iScalar<iMatrix<vtype, Nrepresentation>>>;
+  using iImplGaugeLink  = iScalar<iScalar<iMatrix<vtype, Nrepresentation>>>;
   template <typename vtype>
   using iImplGaugeField = iVector<iScalar<iMatrix<vtype, Nrepresentation>>, Nd>;
 
@@ -64,7 +63,7 @@ public:
                                                  // ugly
   typedef Lattice<SiteGaugeField> GaugeField;
 
-  // Move this elsewhere?
+  // Move this elsewhere? FIXME
   static inline void AddGaugeLink(GaugeField &U, GaugeLinkField &W,
                                   int mu) { // U[mu] += W
     PARALLEL_FOR_LOOP
@@ -174,12 +173,19 @@ typedef GaugeImplTypes<vComplex, Nc> GimplTypesR;
 typedef GaugeImplTypes<vComplexF, Nc> GimplTypesF;
 typedef GaugeImplTypes<vComplexD, Nc> GimplTypesD;
 
+typedef GaugeImplTypes<vComplex, SU<Nc>::AdjointDimension> GimplAdjointTypesR;
+typedef GaugeImplTypes<vComplexF, SU<Nc>::AdjointDimension> GimplAdjointTypesF;
+typedef GaugeImplTypes<vComplexD, SU<Nc>::AdjointDimension> GimplAdjointTypesD;
+
 typedef PeriodicGaugeImpl<GimplTypesR> PeriodicGimplR; // Real.. whichever prec
 typedef PeriodicGaugeImpl<GimplTypesF> PeriodicGimplF; // Float
 typedef PeriodicGaugeImpl<GimplTypesD> PeriodicGimplD; // Double
 
-typedef ConjugateGaugeImpl<GimplTypesR>
+typedef PeriodicGaugeImpl<GimplAdjointTypesR> PeriodicGimplAdjR; // Real.. whichever prec
-    ConjugateGimplR; // Real.. whichever prec
+typedef PeriodicGaugeImpl<GimplAdjointTypesF> PeriodicGimplAdjF; // Float
+typedef PeriodicGaugeImpl<GimplAdjointTypesD> PeriodicGimplAdjD; // Double
+
+typedef ConjugateGaugeImpl<GimplTypesR> ConjugateGimplR; // Real.. whichever prec
 typedef ConjugateGaugeImpl<GimplTypesF> ConjugateGimplF; // Float
 typedef ConjugateGaugeImpl<GimplTypesD> ConjugateGimplD; // Double
 }

lib/qcd/action/pseudofermion/TwoFlavour.h

@@ -1,149 +1,151 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./lib/qcd/action/pseudofermion/TwoFlavour.h
+Source file: ./lib/qcd/action/pseudofermion/TwoFlavour.h
 
-    Copyright (C) 2015
+Copyright (C) 2015
 
 Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution
-    *************************************************************************************/
+directory
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef QCD_PSEUDOFERMION_TWO_FLAVOUR_H
 #define QCD_PSEUDOFERMION_TWO_FLAVOUR_H
 
-namespace Grid{
+namespace Grid {
-  namespace QCD{
+namespace QCD {
 
-    ////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////
-    // Two flavour pseudofermion action for any dop
+// Two flavour pseudofermion action for any dop
-    ////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////
-    template<class Impl>
+template <class Impl>
-    class TwoFlavourPseudoFermionAction : public Action<typename Impl::GaugeField> {
+class TwoFlavourPseudoFermionAction : public Action<typename Impl::GaugeField> {
-    public:
+ public:
-      INHERIT_IMPL_TYPES(Impl);
+  INHERIT_IMPL_TYPES(Impl);
 
-    private:
+ private:
-      
+  FermionOperator<Impl> &FermOp;  // the basic operator
-      FermionOperator<Impl> & FermOp;// the basic operator
 
-      OperatorFunction<FermionField> &DerivativeSolver;
+  OperatorFunction<FermionField> &DerivativeSolver;
 
-      OperatorFunction<FermionField> &ActionSolver;
+  OperatorFunction<FermionField> &ActionSolver;
 
-      FermionField Phi; // the pseudo fermion field for this trajectory
+  FermionField Phi;  // the pseudo fermion field for this trajectory
 
-    public:
+ public:
-      /////////////////////////////////////////////////
+  /////////////////////////////////////////////////
-      // Pass in required objects.
+  // Pass in required objects.
-      /////////////////////////////////////////////////
+  /////////////////////////////////////////////////
-    TwoFlavourPseudoFermionAction(FermionOperator<Impl>  &Op, 
+  TwoFlavourPseudoFermionAction(FermionOperator<Impl> &Op,
-				  OperatorFunction<FermionField> & DS,
+                                OperatorFunction<FermionField> &DS,
-				  OperatorFunction<FermionField> & AS
+                                OperatorFunction<FermionField> &AS)
-				  ) : FermOp(Op), DerivativeSolver(DS), ActionSolver(AS), Phi(Op.FermionGrid()) {
+      : FermOp(Op),
-      };
+        DerivativeSolver(DS),
-      
+        ActionSolver(AS),
-      //////////////////////////////////////////////////////////////////////////////////////
+        Phi(Op.FermionGrid()){};
-      // Push the gauge field in to the dops. Assume any BC's and smearing already applied
-      //////////////////////////////////////////////////////////////////////////////////////
-      virtual void refresh(const GaugeField &U, GridParallelRNG& pRNG) {
 
-	// P(phi) = e^{- phi^dag (MdagM)^-1 phi}
+  //////////////////////////////////////////////////////////////////////////////////////
-	// Phi = Mdag eta 
+  // Push the gauge field in to the dops. Assume any BC's and smearing already
-	// P(eta) = e^{- eta^dag eta}
+  // applied
-	//
+  //////////////////////////////////////////////////////////////////////////////////////
-	// e^{x^2/2 sig^2} => sig^2 = 0.5.
+  virtual void refresh(const GaugeField &U, GridParallelRNG &pRNG) {
-	// 
+    // P(phi) = e^{- phi^dag (MdagM)^-1 phi}
-	// So eta should be of width sig = 1/sqrt(2).
+    // Phi = Mdag eta
-	// and must multiply by 0.707....
+    // P(eta) = e^{- eta^dag eta}
-	//
+    //
-	// Chroma has this scale factor: two_flavor_monomial_w.h
+    // e^{x^2/2 sig^2} => sig^2 = 0.5.
-	// IroIro: does not use this scale. It is absorbed by a change of vars
+    //
-	//         in the Phi integral, and thus is only an irrelevant prefactor for the partition function.
+    // So eta should be of width sig = 1/sqrt(2).
-	//
+    // and must multiply by 0.707....
-	RealD scale = std::sqrt(0.5);
+    //
-	FermionField eta(FermOp.FermionGrid());
+    // Chroma has this scale factor: two_flavor_monomial_w.h
+    // IroIro: does not use this scale. It is absorbed by a change of vars
+    //         in the Phi integral, and thus is only an irrelevant prefactor for
+    //         the partition function.
+    //
+    RealD scale = std::sqrt(0.5);
+    FermionField eta(FermOp.FermionGrid());
 
-	gaussian(pRNG,eta);
+    gaussian(pRNG, eta);
 
-	FermOp.ImportGauge(U);
+    FermOp.ImportGauge(U);
-	FermOp.Mdag(eta,Phi);
+    FermOp.Mdag(eta, Phi);
 
-	Phi=Phi*scale;
+    Phi = Phi * scale;
-	
+  };
-      };
 
-      //////////////////////////////////////////////////////
+  //////////////////////////////////////////////////////
-      // S = phi^dag (Mdag M)^-1 phi
+  // S = phi^dag (Mdag M)^-1 phi
-      //////////////////////////////////////////////////////
+  //////////////////////////////////////////////////////
-      virtual RealD S(const GaugeField &U) {
+  virtual RealD S(const GaugeField &U) {
+    FermOp.ImportGauge(U);
 
-	FermOp.ImportGauge(U);
+    FermionField X(FermOp.FermionGrid());
+    FermionField Y(FermOp.FermionGrid());
 
-	FermionField X(FermOp.FermionGrid());
+    MdagMLinearOperator<FermionOperator<Impl>, FermionField> MdagMOp(FermOp);
-	FermionField Y(FermOp.FermionGrid());
+    X = zero;
-	
+    ActionSolver(MdagMOp, Phi, X);
-	MdagMLinearOperator<FermionOperator<Impl> ,FermionField> MdagMOp(FermOp);
+    MdagMOp.Op(X, Y);
-	X=zero;
-	ActionSolver(MdagMOp,Phi,X);
-	MdagMOp.Op(X,Y);
 
-	RealD action = norm2(Y);
+    RealD action = norm2(Y);
-	std::cout << GridLogMessage << "Pseudofermion action "<<action<<std::endl;
+    std::cout << GridLogMessage << "Pseudofermion action " << action
-	return action;
+              << std::endl;
-      };
+    return action;
+  };
 
-      //////////////////////////////////////////////////////
+  //////////////////////////////////////////////////////
-      // dS/du = - phi^dag  (Mdag M)^-1 [ Mdag dM + dMdag M ]  (Mdag M)^-1 phi
+  // dS/du = - phi^dag  (Mdag M)^-1 [ Mdag dM + dMdag M ]  (Mdag M)^-1 phi
-      //       = - phi^dag M^-1 dM (MdagM)^-1 phi -  phi^dag (MdagM)^-1 dMdag dM (Mdag)^-1 phi 
+  //       = - phi^dag M^-1 dM (MdagM)^-1 phi -  phi^dag (MdagM)^-1 dMdag dM
-      //
+  //       (Mdag)^-1 phi
-      //       = - Ydag dM X  - Xdag dMdag Y
+  //
-      //
+  //       = - Ydag dM X  - Xdag dMdag Y
-      //////////////////////////////////////////////////////
+  //
-      virtual void deriv(const GaugeField &U,GaugeField & dSdU) {
+  //////////////////////////////////////////////////////
+  virtual void deriv(const GaugeField &U, GaugeField &dSdU) {
+    FermOp.ImportGauge(U);
 
-	FermOp.ImportGauge(U);
+    FermionField X(FermOp.FermionGrid());
+    FermionField Y(FermOp.FermionGrid());
+    GaugeField tmp(FermOp.GaugeGrid());
 
-	FermionField X(FermOp.FermionGrid());
+    MdagMLinearOperator<FermionOperator<Impl>, FermionField> MdagMOp(FermOp);
-	FermionField Y(FermOp.FermionGrid());
-	GaugeField   tmp(FermOp.GaugeGrid());
 
-	MdagMLinearOperator<FermionOperator<Impl> ,FermionField> MdagMOp(FermOp);
+    X = zero;
+    DerivativeSolver(MdagMOp, Phi, X); // X = (MdagM)^-1 phi    
+    MdagMOp.Op(X, Y);                  // Y = M X = (Mdag)^-1 phi
 
-	X=zero;
+    // Our conventions really make this UdSdU; We do not differentiate wrt Udag
-	DerivativeSolver(MdagMOp,Phi,X);
+    // here.
-	MdagMOp.Op(X,Y);
+    // So must take dSdU - adj(dSdU) and left multiply by mom to get dS/dt.
 
-	// Our conventions really make this UdSdU; We do not differentiate wrt Udag here.
+    FermOp.MDeriv(tmp, Y, X, DaggerNo);
-	// So must take dSdU - adj(dSdU) and left multiply by mom to get dS/dt.
+    dSdU = tmp;
+    FermOp.MDeriv(tmp, X, Y, DaggerYes);
+    dSdU = dSdU + tmp;
 
-	FermOp.MDeriv(tmp , Y, X,DaggerNo );  dSdU=tmp;
+    // not taking here the traceless antihermitian component
-	FermOp.MDeriv(tmp , X, Y,DaggerYes);  dSdU=dSdU+tmp;
+  };
-	
+};
-	//dSdU = Ta(dSdU);
+}
-
-      };
-
-    };
-    
-  }
 }
 
 #endif

lib/qcd/action/pseudofermion/TwoFlavourEvenOdd.h

@@ -1,70 +1,66 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./lib/qcd/action/pseudofermion/TwoFlavourEvenOdd.h
+Source file: ./lib/qcd/action/pseudofermion/TwoFlavourEvenOdd.h
 
-    Copyright (C) 2015
+Copyright (C) 2015
 
 Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution
-    *************************************************************************************/
+directory
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef QCD_PSEUDOFERMION_TWO_FLAVOUR_EVEN_ODD_H
 #define QCD_PSEUDOFERMION_TWO_FLAVOUR_EVEN_ODD_H
 
-namespace Grid{
+namespace Grid {
-  namespace QCD{
+namespace QCD {
 
+////////////////////////////////////////////////////////////////////////
+// Two flavour pseudofermion action for any EO prec dop
+////////////////////////////////////////////////////////////////////////
+template <class Impl>
+class TwoFlavourEvenOddPseudoFermionAction
+    : public Action<typename Impl::GaugeField> {
+ public:
+  INHERIT_IMPL_TYPES(Impl);
 
+ private:
+  FermionOperator<Impl> &FermOp;  // the basic operator
 
-    ////////////////////////////////////////////////////////////////////////
+  OperatorFunction<FermionField> &DerivativeSolver;
-    // Two flavour pseudofermion action for any EO prec dop
+  OperatorFunction<FermionField> &ActionSolver;
-    ////////////////////////////////////////////////////////////////////////
-    template<class Impl>
-    class TwoFlavourEvenOddPseudoFermionAction : public Action<typename Impl::GaugeField> {
 
-    public:
+  FermionField PhiOdd;   // the pseudo fermion field for this trajectory
+  FermionField PhiEven;  // the pseudo fermion field for this trajectory
 
-      INHERIT_IMPL_TYPES(Impl);
+ public:
-
+  /////////////////////////////////////////////////
-    private:
+  // Pass in required objects.
-      
+  /////////////////////////////////////////////////
-      FermionOperator<Impl> & FermOp;// the basic operator
+  TwoFlavourEvenOddPseudoFermionAction(FermionOperator<Impl> &Op,
-
+                                       OperatorFunction<FermionField> &DS,
-      OperatorFunction<FermionField> &DerivativeSolver;
+                                       OperatorFunction<FermionField> &AS)
-      OperatorFunction<FermionField> &ActionSolver;
+      : FermOp(Op),
-
+        DerivativeSolver(DS),
-      FermionField PhiOdd;   // the pseudo fermion field for this trajectory
+        ActionSolver(AS),
-      FermionField PhiEven;  // the pseudo fermion field for this trajectory
-
-    public:
-      /////////////////////////////////////////////////
-      // Pass in required objects.
-      /////////////////////////////////////////////////
-      TwoFlavourEvenOddPseudoFermionAction(FermionOperator<Impl>  &Op, 
-					 OperatorFunction<FermionField> & DS,
-					 OperatorFunction<FermionField> & AS
-					   ) : 
-        FermOp(Op), 
-	DerivativeSolver(DS), 
-	ActionSolver(AS), 
         PhiEven(Op.FermionRedBlackGrid()),
 	PhiOdd(Op.FermionRedBlackGrid())
 		  {};

lib/qcd/action/pseudofermion/TwoFlavourEvenOddRatio.h

@@ -131,9 +131,11 @@ namespace Grid{
 	Vpc.MpcDag(PhiOdd,Y);           // Y= Vdag phi
 	X=zero;
 	ActionSolver(Mpc,Y,X);          // X= (MdagM)^-1 Vdag phi
-	Mpc.Mpc(X,Y);                   // Y=  Mdag^-1 Vdag phi
+	//Mpc.Mpc(X,Y);                   // Y=  Mdag^-1 Vdag phi
+	// Multiply by Ydag
+	RealD action = real(innerProduct(Y,X));
 
-	RealD action = norm2(Y);
+	//RealD action = norm2(Y);
 
 	// The EE factorised block; normally can replace with zero if det is constant (gauge field indept)
 	// Only really clover term that creates this. Leave the EE portion as a future to do to make most

lib/qcd/hmc/HmcRunner.h

@@ -1,179 +1,191 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./lib/qcd/hmc/HmcRunner.h
+Source file: ./lib/qcd/hmc/HmcRunner.h
 
-    Copyright (C) 2015
+Copyright (C) 2015
 
 Author: paboyle <paboyle@ph.ed.ac.uk>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution
-    *************************************************************************************/
+directory
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #ifndef HMC_RUNNER
 #define HMC_RUNNER
 
-namespace Grid{
+namespace Grid {
-  namespace QCD{
+namespace QCD {
 
-
+template <class Gimpl, class RepresentationsPolicy = NoHirep >
-template<class Gimpl>
 class NerscHmcRunnerTemplate {
-public:
+ public:
-
   INHERIT_GIMPL_TYPES(Gimpl);
 
   enum StartType_t { ColdStart, HotStart, TepidStart, CheckpointStart };
 
-  ActionSet<GaugeField> TheAction;
+  ActionSet<GaugeField, RepresentationsPolicy> TheAction;
 
-  GridCartesian         * UGrid   ;
+  GridCartesian *UGrid;
-  GridCartesian         * FGrid   ;
+  GridCartesian *FGrid;
-  GridRedBlackCartesian * UrbGrid ;
+  GridRedBlackCartesian *UrbGrid;
-  GridRedBlackCartesian * FrbGrid ;
+  GridRedBlackCartesian *FrbGrid;
 
-  virtual void BuildTheAction (int argc, char **argv) = 0; // necessary?
+  virtual void BuildTheAction(int argc, char **argv) = 0;  // necessary?
-
-  
-  void Run (int argc, char  **argv){
 
+  void Run(int argc, char **argv) {
     StartType_t StartType = HotStart;
 
     std::string arg;
 
-    if( GridCmdOptionExists(argv,argv+argc,"--StartType") ){
+    if (GridCmdOptionExists(argv, argv + argc, "--StartType")) {
-      arg = GridCmdOptionPayload(argv,argv+argc,"--StartType");
+      arg = GridCmdOptionPayload(argv, argv + argc, "--StartType");
-      if ( arg == "HotStart" ) { StartType = HotStart; }
+      if (arg == "HotStart") {
-      else if ( arg == "ColdStart" ) { StartType = ColdStart; }
+        StartType = HotStart;
-      else if ( arg == "TepidStart" ) { StartType = TepidStart; }
+      } else if (arg == "ColdStart") {
-      else if ( arg == "CheckpointStart" ) { StartType = CheckpointStart; }
+        StartType = ColdStart;
-      else assert(0);
+      } else if (arg == "TepidStart") {
+        StartType = TepidStart;
+      } else if (arg == "CheckpointStart") {
+        StartType = CheckpointStart;
+      } else {
+        std::cout << GridLogError << "Unrecognized option in --StartType\n";
+        std::cout << GridLogError << "Valid [HotStart, ColdStart, TepidStart, CheckpointStart]\n";
+        assert(0);
+      }
     }
 
     int StartTraj = 0;
-    if( GridCmdOptionExists(argv,argv+argc,"--StartTrajectory") ){
+    if (GridCmdOptionExists(argv, argv + argc, "--StartTrajectory")) {
-      arg= GridCmdOptionPayload(argv,argv+argc,"--StartTrajectory");
+      arg = GridCmdOptionPayload(argv, argv + argc, "--StartTrajectory");
       std::vector<int> ivec(0);
-      GridCmdOptionIntVector(arg,ivec);
+      GridCmdOptionIntVector(arg, ivec);
       StartTraj = ivec[0];
-    }    
+    }
 
     int NumTraj = 1;
-    if( GridCmdOptionExists(argv,argv+argc,"--Trajectories") ){
+    if (GridCmdOptionExists(argv, argv + argc, "--Trajectories")) {
-      arg= GridCmdOptionPayload(argv,argv+argc,"--Trajectories");
+      arg = GridCmdOptionPayload(argv, argv + argc, "--Trajectories");
       std::vector<int> ivec(0);
-      GridCmdOptionIntVector(arg,ivec);
+      GridCmdOptionIntVector(arg, ivec);
       NumTraj = ivec[0];
     }
 
     int NumThermalizations = 10;
-    if( GridCmdOptionExists(argv,argv+argc,"--Thermalizations") ){
+    if (GridCmdOptionExists(argv, argv + argc, "--Thermalizations")) {
-      arg= GridCmdOptionPayload(argv,argv+argc,"--Thermalizations");
+      arg = GridCmdOptionPayload(argv, argv + argc, "--Thermalizations");
       std::vector<int> ivec(0);
-      GridCmdOptionIntVector(arg,ivec);
+      GridCmdOptionIntVector(arg, ivec);
       NumThermalizations = ivec[0];
     }
 
+    GridSerialRNG sRNG;
+    GridParallelRNG pRNG(UGrid);
+    LatticeGaugeField U(UGrid);  // change this to an extended field (smearing class)?
 
-    GridSerialRNG    sRNG;
+    std::vector<int> SerSeed({1, 2, 3, 4, 5});
-    GridParallelRNG  pRNG(UGrid);
+    std::vector<int> ParSeed({6, 7, 8, 9, 10});
-    LatticeGaugeField  U(UGrid); // change this to an extended field (smearing class)
 
-    std::vector<int> SerSeed({1,2,3,4,5});
-    std::vector<int> ParSeed({6,7,8,9,10});
-
-    
     // Create integrator, including the smearing policy
-    // Smearing policy
+    // Smearing policy, only defined for Nc=3
+    /*
     std::cout << GridLogDebug << " Creating the Stout class\n";
-    double rho = 0.1; // smearing parameter, now hardcoded
+    double rho = 0.1;  // smearing parameter, now hardcoded
-    int Nsmear = 1;   // number of smearing levels
+    int Nsmear = 1;    // number of smearing levels
     Smear_Stout<Gimpl> Stout(rho);
     std::cout << GridLogDebug << " Creating the SmearedConfiguration class\n";
-    SmearedConfiguration<Gimpl> SmearingPolicy(UGrid, Nsmear, Stout);
+    //SmearedConfiguration<Gimpl> SmearingPolicy(UGrid, Nsmear, Stout);
     std::cout << GridLogDebug << " done\n";
+    */
     //////////////
-    typedef MinimumNorm2<GaugeField, SmearedConfiguration<Gimpl> >  IntegratorType;// change here to change the algorithm
+    NoSmearing<Gimpl> SmearingPolicy;
-    IntegratorParameters MDpar(20);
+    typedef MinimumNorm2<GaugeField, NoSmearing<Gimpl>, RepresentationsPolicy >
+        IntegratorType;  // change here to change the algorithm
+    IntegratorParameters MDpar(20, 1.0);
     IntegratorType MDynamics(UGrid, MDpar, TheAction, SmearingPolicy);
 
-    
     // Checkpoint strategy
-    NerscHmcCheckpointer<Gimpl> Checkpoint(std::string("ckpoint_lat"),std::string("ckpoint_rng"),1);
+    NerscHmcCheckpointer<Gimpl> Checkpoint(std::string("ckpoint_lat"),
-    PlaquetteLogger<Gimpl>      PlaqLog(std::string("plaq"));
+                                           std::string("ckpoint_rng"), 1);
+    PlaquetteLogger<Gimpl> PlaqLog(std::string("plaq"));
 
     HMCparameters HMCpar;
-    HMCpar.StartTrajectory   = StartTraj;
+    HMCpar.StartTrajectory = StartTraj;
-    HMCpar.Trajectories      = NumTraj;
+    HMCpar.Trajectories = NumTraj;
     HMCpar.NoMetropolisUntil = NumThermalizations;
-    
 
-    if ( StartType == HotStart ) {
+    if (StartType == HotStart) {
       // Hot start
       HMCpar.MetropolisTest = true;
       sRNG.SeedFixedIntegers(SerSeed);
       pRNG.SeedFixedIntegers(ParSeed);
-      SU3::HotConfiguration(pRNG, U);
+      SU<Nc>::HotConfiguration(pRNG, U);
-    } else if ( StartType == ColdStart ) { 
+    } else if (StartType == ColdStart) {
       // Cold start
       HMCpar.MetropolisTest = true;
       sRNG.SeedFixedIntegers(SerSeed);
       pRNG.SeedFixedIntegers(ParSeed);
-      SU3::ColdConfiguration(pRNG, U);
+      SU<Nc>::ColdConfiguration(pRNG, U);
-    } else if ( StartType == TepidStart ) {       
+    } else if (StartType == TepidStart) {
       // Tepid start
       HMCpar.MetropolisTest = true;
       sRNG.SeedFixedIntegers(SerSeed);
       pRNG.SeedFixedIntegers(ParSeed);
-      SU3::TepidConfiguration(pRNG, U);
+      SU<Nc>::TepidConfiguration(pRNG, U);
-    } else if ( StartType == CheckpointStart ) { 
+    } else if (StartType == CheckpointStart) {
       HMCpar.MetropolisTest = true;
       // CheckpointRestart
       Checkpoint.CheckpointRestore(StartTraj, U, sRNG, pRNG);
     }
 
-    // Attach the gauge field to the smearing Policy and create the fill the smeared set
+    // Attach the gauge field to the smearing Policy and create the fill the
+    // smeared set
     // notice that the unit configuration is singular in this procedure
-    std::cout << GridLogMessage << "Filling the smeared set\n"; 
+    std::cout << GridLogMessage << "Filling the smeared set\n";
     SmearingPolicy.set_GaugeField(U);
-    
+
-    HybridMonteCarlo<GaugeField,IntegratorType>  HMC(HMCpar, MDynamics,sRNG,pRNG,U); 
+    HybridMonteCarlo<GaugeField, IntegratorType> HMC(HMCpar, MDynamics, sRNG,
+                                                     pRNG, U);
     HMC.AddObservable(&Checkpoint);
     HMC.AddObservable(&PlaqLog);
-    
+
     // Run it
     HMC.evolve();
-    
   }
-  
 };
 
- typedef NerscHmcRunnerTemplate<PeriodicGimplR> NerscHmcRunner;
+typedef NerscHmcRunnerTemplate<PeriodicGimplR> NerscHmcRunner;
- typedef NerscHmcRunnerTemplate<PeriodicGimplF> NerscHmcRunnerF;
+typedef NerscHmcRunnerTemplate<PeriodicGimplF> NerscHmcRunnerF;
- typedef NerscHmcRunnerTemplate<PeriodicGimplD> NerscHmcRunnerD;
+typedef NerscHmcRunnerTemplate<PeriodicGimplD> NerscHmcRunnerD;
 
- typedef NerscHmcRunnerTemplate<PeriodicGimplR> PeriodicNerscHmcRunner;
+typedef NerscHmcRunnerTemplate<PeriodicGimplR> PeriodicNerscHmcRunner;
- typedef NerscHmcRunnerTemplate<PeriodicGimplF> PeriodicNerscHmcRunnerF;
+typedef NerscHmcRunnerTemplate<PeriodicGimplF> PeriodicNerscHmcRunnerF;
- typedef NerscHmcRunnerTemplate<PeriodicGimplD> PeriodicNerscHmcRunnerD;
+typedef NerscHmcRunnerTemplate<PeriodicGimplD> PeriodicNerscHmcRunnerD;
 
- typedef NerscHmcRunnerTemplate<ConjugateGimplR> ConjugateNerscHmcRunner;
+typedef NerscHmcRunnerTemplate<ConjugateGimplR> ConjugateNerscHmcRunner;
- typedef NerscHmcRunnerTemplate<ConjugateGimplF> ConjugateNerscHmcRunnerF;
+typedef NerscHmcRunnerTemplate<ConjugateGimplF> ConjugateNerscHmcRunnerF;
- typedef NerscHmcRunnerTemplate<ConjugateGimplD> ConjugateNerscHmcRunnerD;
+typedef NerscHmcRunnerTemplate<ConjugateGimplD> ConjugateNerscHmcRunnerD;
 
-}}
+template <class RepresentationsPolicy>
+using NerscHmcRunnerHirep = NerscHmcRunnerTemplate<PeriodicGimplR, RepresentationsPolicy>;
+
+
+
+}
+}
 #endif

lib/qcd/hmc/integrators/Integrator.h

@@ -1,33 +1,34 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./lib/qcd/hmc/integrators/Integrator.h
+Source file: ./lib/qcd/hmc/integrators/Integrator.h
 
-    Copyright (C) 2015
+Copyright (C) 2015
 
 Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: neo <cossu@post.kek.jp>
 Author: paboyle <paboyle@ph.ed.ac.uk>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution
-    *************************************************************************************/
+directory
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 //--------------------------------------------------------------------
 /*! @file Integrator.h
  * @brief Classes for the Molecular Dynamics integrator
@@ -40,208 +41,278 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #ifndef INTEGRATOR_INCLUDED
 #define INTEGRATOR_INCLUDED
 
-//class Observer;
+// class Observer;
 
 #include <memory>
 
- namespace Grid{
+namespace Grid {
- 	namespace QCD{
+namespace QCD {
 
- 		struct IntegratorParameters{
+struct IntegratorParameters {
+  int Nexp;
+  int MDsteps;  // number of outer steps
+  RealD trajL;  // trajectory length
+  RealD stepsize;
 
- 			int Nexp;
+  IntegratorParameters(int MDsteps_, RealD trajL_ = 1.0, int Nexp_ = 12)
-      int MDsteps;  // number of outer steps
+      : Nexp(Nexp_),
-      RealD trajL;  // trajectory length 
+        MDsteps(MDsteps_),
-      RealD stepsize;
+        trajL(trajL_),
-
+        stepsize(trajL / MDsteps){
-      IntegratorParameters(int MDsteps_, 
+            // empty body constructor
-      	RealD trajL_=1.0,
+        };
-      	int Nexp_=12):
-      Nexp(Nexp_),
-      MDsteps(MDsteps_),
-      trajL(trajL_),
-      stepsize(trajL/MDsteps)
-      {
-	  // empty body constructor
-      };
-
-  };
-
-    /*! @brief Class for Molecular Dynamics management */   
-    template<class GaugeField, class SmearingPolicy>
-  class Integrator {
-
-  protected:
-
-  	typedef IntegratorParameters ParameterType;
-
-  	IntegratorParameters Params;
-
-  	const ActionSet<GaugeField> as;
-
-      int levels;              //
-      double t_U;              // Track time passing on each level and for U and for P
-      std::vector<double> t_P; //
-
-      GaugeField P;
-
-      SmearingPolicy &Smearer;
-      
-      // Should match any legal (SU(n)) gauge field
-      // Need to use this template to match Ncol to pass to SU<N> class
-      template<int Ncol,class vec> void generate_momenta(Lattice< iVector< iScalar< iMatrix<vec,Ncol> >, Nd> > & P,GridParallelRNG& pRNG){
-      typedef Lattice< iScalar< iScalar< iMatrix<vec,Ncol> > > > GaugeLinkField;
-      GaugeLinkField Pmu(P._grid);
-      Pmu = zero;
-      for(int mu=0;mu<Nd;mu++){
-      	SU<Ncol>::GaussianLieAlgebraMatrix(pRNG, Pmu);
-      	PokeIndex<LorentzIndex>(P, Pmu, mu);
-      }
-  }
-
-
-      //ObserverList observers; // not yet
-      //      typedef std::vector<Observer*> ObserverList;
-      //      void register_observers();
-      //      void notify_observers();
-
-  void update_P(GaugeField&U, int level, double ep){
-  	t_P[level]+=ep;
-  	update_P(P,U,level,ep);
-
-  	std::cout<<GridLogIntegrator<<"["<<level<<"] P " << " dt "<< ep <<" : t_P "<< t_P[level] <<std::endl;
-  }
-
-  void update_P(GaugeField &Mom,GaugeField&U, int level,double ep){
-  	// input U actually not used... 
-  	for(int a=0; a<as[level].actions.size(); ++a){
-  		GaugeField force(U._grid);
-  		GaugeField& Us = Smearer.get_U(as[level].actions.at(a)->is_smeared);
-  		as[level].actions.at(a)->deriv(Us,force); // deriv should NOT include Ta
-
-	  	std::cout<< GridLogIntegrator << "Smearing (on/off): "<<as[level].actions.at(a)->is_smeared <<std::endl;
-	  	if (as[level].actions.at(a)->is_smeared) Smearer.smeared_force(force);
-	  	force = Ta(force);
-	  	std::cout<< GridLogIntegrator << "Force average: "<< norm2(force)/(U._grid->gSites()) <<std::endl;
-	  	Mom -= force*ep;
-	  }
-	}
-
-	void update_U(GaugeField&U, double ep){
-		update_U(P,U,ep);
-
-		t_U+=ep;
-		int fl = levels-1;
-		std::cout<< GridLogIntegrator <<"   "<<"["<<fl<<"] U " << " dt "<< ep <<" : t_U "<< t_U <<std::endl;
-
-	}
-	void update_U(GaugeField &Mom, GaugeField&U, double ep){
-	//rewrite exponential to deal automatically  with the lorentz index?
-	//	GaugeLinkField Umu(U._grid);
-	//	GaugeLinkField Pmu(U._grid);
-		for (int mu = 0; mu < Nd; mu++){
-			auto Umu=PeekIndex<LorentzIndex>(U, mu);
-			auto Pmu=PeekIndex<LorentzIndex>(Mom, mu);
-			Umu = expMat(Pmu, ep, Params.Nexp)*Umu;
-			ProjectOnGroup(Umu);
-			PokeIndex<LorentzIndex>(U, Umu, mu);
-		}
-	// Update the smeared fields, can be implemented as observer
-		Smearer.set_GaugeField(U);
-	}
-
-	virtual void step (GaugeField& U,int level, int first,int last)=0;
-
-public:
-
-	Integrator(GridBase* grid, 
-		IntegratorParameters Par,
-		ActionSet<GaugeField> & Aset,
-		SmearingPolicy &Sm):
-	Params(Par),
-	as(Aset),
-	P(grid),
-	levels(Aset.size()),
-	Smearer(Sm)
-	{
-		t_P.resize(levels,0.0);
-		t_U=0.0;
-	// initialization of smearer delegated outside of Integrator
-	};
-
-	virtual ~Integrator(){}
-
-      //Initialization of momenta and actions
-	void refresh(GaugeField& U,GridParallelRNG &pRNG){
-		std::cout<<GridLogIntegrator<< "Integrator refresh\n";
-		generate_momenta(P,pRNG);
-		for(int level=0; level< as.size(); ++level){
-			for(int actionID=0; actionID<as[level].actions.size(); ++actionID){
-	    // get gauge field from the SmearingPolicy and
-	    // based on the boolean is_smeared in actionID
-				GaugeField& Us = Smearer.get_U(as[level].actions.at(actionID)->is_smeared);
-				as[level].actions.at(actionID)->refresh(Us, pRNG);
-			}
-		}
-	}
-
-      // Calculate action
-	RealD S(GaugeField& U){// here also U not used
-
-		LatticeComplex Hloc(U._grid);	Hloc = zero;
-	// Momenta
-		for (int mu=0; mu <Nd; mu++){
-			auto Pmu = PeekIndex<LorentzIndex>(P, mu);
-			Hloc -= trace(Pmu*Pmu);
-		}
-		Complex Hsum = sum(Hloc);
-
-		RealD H = Hsum.real();
-		RealD Hterm;
-		std::cout<<GridLogMessage << "Momentum action H_p = "<< H << "\n";
-
-	// Actions
-		for(int level=0; level<as.size(); ++level){
-			for(int actionID=0; actionID<as[level].actions.size(); ++actionID){
-	    // get gauge field from the SmearingPolicy and
-	    // based on the boolean is_smeared in actionID
-				GaugeField& Us = Smearer.get_U(as[level].actions.at(actionID)->is_smeared);
-				Hterm = as[level].actions.at(actionID)->S(Us);
-				std::cout<<GridLogMessage << "S Level "<<level<<" term "<<actionID<<" H = "<<Hterm<<std::endl;
-				H += Hterm;
-			}
-		}
-
-		return H;
-	}
-
-	void integrate(GaugeField& U){
-
-	// reset the clocks
-		t_U=0;
-		for(int level=0; level<as.size(); ++level){
-			t_P[level]=0;
-		}	
-
-	for(int step=0; step< Params.MDsteps; ++step){   // MD step
-		int first_step = (step==0);
-		int  last_step = (step==Params.MDsteps-1);
-		this->step(U,0,first_step,last_step);
-	}
-
-	// Check the clocks all match on all levels
-	for(int level=0; level<as.size(); ++level){
-	  assert(fabs(t_U - t_P[level])<1.0e-6); // must be the same
-	  std::cout<<GridLogIntegrator<<" times["<<level<<"]= "<<t_P[level]<< " " << t_U <<std::endl;
-	}	
-
-	// and that we indeed got to the end of the trajectory
-	assert(fabs(t_U-Params.trajL) < 1.0e-6);
-
-
-}
 };
 
+/*! @brief Class for Molecular Dynamics management */
+template <class GaugeField, class SmearingPolicy, class RepresentationPolicy>
+class Integrator {
+ protected:
+  typedef IntegratorParameters ParameterType;
+
+  IntegratorParameters Params;
+
+  const ActionSet<GaugeField, RepresentationPolicy> as;
+
+  int levels;  //
+  double t_U;  // Track time passing on each level and for U and for P
+  std::vector<double> t_P;  //
+
+  GaugeField P;
+
+  SmearingPolicy& Smearer;
+
+  RepresentationPolicy Representations;
+
+  // Should match any legal (SU(n)) gauge field
+  // Need to use this template to match Ncol to pass to SU<N> class
+  template <int Ncol, class vec>
+  void generate_momenta(Lattice<iVector<iScalar<iMatrix<vec, Ncol> >, Nd> >& P,
+                        GridParallelRNG& pRNG) {
+    typedef Lattice<iScalar<iScalar<iMatrix<vec, Ncol> > > > GaugeLinkField;
+    GaugeLinkField Pmu(P._grid);
+    Pmu = zero;
+    for (int mu = 0; mu < Nd; mu++) {
+      SU<Ncol>::GaussianFundamentalLieAlgebraMatrix(pRNG, Pmu);
+      PokeIndex<LorentzIndex>(P, Pmu, mu);
+    }
+  }
+
+  // ObserverList observers; // not yet
+  //      typedef std::vector<Observer*> ObserverList;
+  //      void register_observers();
+  //      void notify_observers();
+
+  void update_P(GaugeField& U, int level, double ep) {
+    t_P[level] += ep;
+    update_P(P, U, level, ep);
+
+    std::cout << GridLogIntegrator << "[" << level << "] P "
+              << " dt " << ep << " : t_P " << t_P[level] << std::endl;
+  }
+
+  // to be used by the actionlevel class to iterate
+  // over the representations
+  struct _updateP {
+    template <class FieldType, class GF, class Repr>
+    void operator()(std::vector<Action<FieldType>*> repr_set, Repr& Rep,
+                    GF& Mom, GF& U, double ep) {
+      for (int a = 0; a < repr_set.size(); ++a) {
+        FieldType forceR(U._grid);
+        // Implement smearing only for the fundamental representation now
+        repr_set.at(a)->deriv(Rep.U, forceR);
+        GF force =
+            Rep.RtoFundamentalProject(forceR);  // Ta for the fundamental rep
+        std::cout << GridLogIntegrator << "Hirep Force average: "
+                  << norm2(force) / (U._grid->gSites()) << std::endl;
+        Mom -= force * ep ;
+      }
+    }
+  } update_P_hireps{};
+
+  void update_P(GaugeField& Mom, GaugeField& U, int level, double ep) {
+    // input U actually not used in the fundamental case
+    // Fundamental updates, include smearing
+    for (int a = 0; a < as[level].actions.size(); ++a) {
+      GaugeField force(U._grid);
+      GaugeField& Us = Smearer.get_U(as[level].actions.at(a)->is_smeared);
+      as[level].actions.at(a)->deriv(Us, force);  // deriv should NOT include Ta
+
+      std::cout << GridLogIntegrator
+                << "Smearing (on/off): " << as[level].actions.at(a)->is_smeared
+                << std::endl;
+      if (as[level].actions.at(a)->is_smeared) Smearer.smeared_force(force);
+      force = Ta(force);
+      std::cout << GridLogIntegrator
+                << "Force average: " << norm2(force) / (U._grid->gSites())
+                << std::endl;
+      Mom -= force * ep;
+    }
+
+    // Force from the other representations
+    as[level].apply(update_P_hireps, Representations, Mom, U, ep);
+  }
+
+  void update_U(GaugeField& U, double ep) {
+    update_U(P, U, ep);
+
+    t_U += ep;
+    int fl = levels - 1;
+    std::cout << GridLogIntegrator << "   "
+              << "[" << fl << "] U "
+              << " dt " << ep << " : t_U " << t_U << std::endl;
+  }
+  void update_U(GaugeField& Mom, GaugeField& U, double ep) {
+    // rewrite exponential to deal internally with the lorentz index?
+    for (int mu = 0; mu < Nd; mu++) {
+      auto Umu = PeekIndex<LorentzIndex>(U, mu);
+      auto Pmu = PeekIndex<LorentzIndex>(Mom, mu);
+      Umu = expMat(Pmu, ep, Params.Nexp) * Umu;
+      PokeIndex<LorentzIndex>(U, ProjectOnGroup(Umu), mu);
+    }
+    
+    // Update the smeared fields, can be implemented as observer
+    Smearer.set_GaugeField(U);
+    // Update the higher representations fields
+    Representations.update(U);  // void functions if fundamental representation
+  }
+
+  virtual void step(GaugeField& U, int level, int first, int last) = 0;
+
+ public:
+  Integrator(GridBase* grid, IntegratorParameters Par,
+             ActionSet<GaugeField, RepresentationPolicy>& Aset,
+             SmearingPolicy& Sm)
+      : Params(Par),
+        as(Aset),
+        P(grid),
+        levels(Aset.size()),
+        Smearer(Sm),
+        Representations(grid) {
+    t_P.resize(levels, 0.0);
+    t_U = 0.0;
+    // initialization of smearer delegated outside of Integrator
+  };
+
+  virtual ~Integrator() {}
+
+  // to be used by the actionlevel class to iterate
+  // over the representations
+  struct _refresh {
+    template <class FieldType, class Repr>
+    void operator()(std::vector<Action<FieldType>*> repr_set, Repr& Rep,
+                    GridParallelRNG& pRNG) {
+      for (int a = 0; a < repr_set.size(); ++a){
+        repr_set.at(a)->refresh(Rep.U, pRNG);
+      
+      std::cout << GridLogDebug << "Hirep refreshing pseudofermions" << std::endl;
+    }
+    }
+  } refresh_hireps{};
+
+  // Initialization of momenta and actions
+  void refresh(GaugeField& U, GridParallelRNG& pRNG) {
+    std::cout << GridLogIntegrator << "Integrator refresh\n";
+    generate_momenta(P, pRNG);
+
+    // Update the smeared fields, can be implemented as observer
+    // necessary to keep the fields updated even after a reject
+    // of the Metropolis
+    Smearer.set_GaugeField(U);
+    // Set the (eventual) representations gauge fields
+    Representations.update(U);
+
+    // The Smearer is attached to a pointer of the gauge field
+    // automatically gets the correct field
+    // whether or not has been accepted in the previous sweep
+    for (int level = 0; level < as.size(); ++level) {
+      for (int actionID = 0; actionID < as[level].actions.size(); ++actionID) {
+        // get gauge field from the SmearingPolicy and
+        // based on the boolean is_smeared in actionID
+        GaugeField& Us =
+            Smearer.get_U(as[level].actions.at(actionID)->is_smeared);
+        as[level].actions.at(actionID)->refresh(Us, pRNG);
+      }
+
+      // Refresh the higher representation actions
+      as[level].apply(refresh_hireps, Representations, pRNG);
+    }
+  }
+
+  // to be used by the actionlevel class to iterate
+  // over the representations
+  struct _S {
+    template <class FieldType, class Repr>
+    void operator()(std::vector<Action<FieldType>*> repr_set, Repr& Rep,
+                    int level, RealD& H) {
+      
+      for (int a = 0; a < repr_set.size(); ++a) {
+        RealD Hterm = repr_set.at(a)->S(Rep.U);
+        std::cout << GridLogMessage << "S Level " << level << " term " << a
+                  << " H Hirep = " << Hterm << std::endl;
+        H += Hterm;
+
+      }
+    }
+  } S_hireps{};
+
+  // Calculate action
+  RealD S(GaugeField& U) {  // here also U not used
+
+    LatticeComplex Hloc(U._grid);
+    Hloc = zero;
+    // Momenta
+    for (int mu = 0; mu < Nd; mu++) {
+      auto Pmu = PeekIndex<LorentzIndex>(P, mu);
+      Hloc -= trace(Pmu * Pmu);
+    }
+    Complex Hsum = sum(Hloc);
+
+    RealD H = Hsum.real();
+    RealD Hterm;
+    std::cout << GridLogMessage << "Momentum action H_p = " << H << "\n";
+
+    // Actions
+    for (int level = 0; level < as.size(); ++level) {
+      for (int actionID = 0; actionID < as[level].actions.size(); ++actionID) {
+        // get gauge field from the SmearingPolicy and
+        // based on the boolean is_smeared in actionID
+        GaugeField& Us =
+            Smearer.get_U(as[level].actions.at(actionID)->is_smeared);
+        Hterm = as[level].actions.at(actionID)->S(Us);
+        std::cout << GridLogMessage << "S Level " << level << " term "
+                  << actionID << " H = " << Hterm << std::endl;
+        H += Hterm;
+      }
+      as[level].apply(S_hireps, Representations, level, H);
+    }
+
+    return H;
+  }
+
+  void integrate(GaugeField& U) {
+    // reset the clocks
+    t_U = 0;
+    for (int level = 0; level < as.size(); ++level) {
+      t_P[level] = 0;
+    }
+
+    for (int step = 0; step < Params.MDsteps; ++step) {  // MD step
+      int first_step = (step == 0);
+      int last_step = (step == Params.MDsteps - 1);
+      this->step(U, 0, first_step, last_step);
+    }
+
+    // Check the clocks all match on all levels
+    for (int level = 0; level < as.size(); ++level) {
+      assert(fabs(t_U - t_P[level]) < 1.0e-6);  // must be the same
+      std::cout << GridLogIntegrator << " times[" << level
+                << "]= " << t_P[level] << " " << t_U << std::endl;
+    }
+
+    // and that we indeed got to the end of the trajectory
+    assert(fabs(t_U - Params.trajL) < 1.0e-6);
+  }
+};
 }
 }
-#endif//INTEGRATOR_INCLUDED
+#endif  // INTEGRATOR_INCLUDED

lib/qcd/hmc/integrators/Integrator_algorithm.h

@@ -91,17 +91,19 @@ namespace Grid{
     *  P 1/2                            P 1/2
     */    
 
-    template<class GaugeField, class SmearingPolicy> class LeapFrog :
+    template<class GaugeField,
-      public Integrator<GaugeField, SmearingPolicy> {
+	     class SmearingPolicy,
+	     class RepresentationPolicy = Representations< FundamentalRepresentation > > class LeapFrog :
+      public Integrator<GaugeField, SmearingPolicy, RepresentationPolicy> {
     public:
 
-      typedef LeapFrog<GaugeField, SmearingPolicy> Algorithm;
+      typedef LeapFrog<GaugeField, SmearingPolicy, RepresentationPolicy> Algorithm;
 
       LeapFrog(GridBase* grid, 
 	       IntegratorParameters Par,
-	       ActionSet<GaugeField> & Aset,
+	       ActionSet<GaugeField, RepresentationPolicy> & Aset,
 	       SmearingPolicy & Sm):
-	Integrator<GaugeField, SmearingPolicy>(grid,Par,Aset,Sm) {};
+	Integrator<GaugeField, SmearingPolicy, RepresentationPolicy>(grid,Par,Aset,Sm) {};
 
 
       void step (GaugeField& U, int level,int _first, int _last){
@@ -138,8 +140,10 @@ namespace Grid{
       }
     };
 
-    template<class GaugeField, class SmearingPolicy> class MinimumNorm2 :
+    template<class GaugeField,
-      public Integrator<GaugeField, SmearingPolicy> {
+	     class SmearingPolicy,
+	     class RepresentationPolicy = Representations < FundamentalRepresentation > > class MinimumNorm2 :
+      public Integrator<GaugeField, SmearingPolicy, RepresentationPolicy> {
     private:
       const RealD lambda = 0.1931833275037836;
 
@@ -147,9 +151,9 @@ namespace Grid{
 
       MinimumNorm2(GridBase* grid, 
 		   IntegratorParameters Par,
-		   ActionSet<GaugeField> & Aset,
+		   ActionSet<GaugeField, RepresentationPolicy> & Aset,
 		   SmearingPolicy& Sm):
-	Integrator<GaugeField, SmearingPolicy>(grid,Par,Aset,Sm) {};
+	Integrator<GaugeField, SmearingPolicy, RepresentationPolicy>(grid,Par,Aset,Sm) {};
 
       void step (GaugeField& U, int level, int _first,int _last){
 
@@ -197,8 +201,10 @@ namespace Grid{
     };
 
 
-    template<class GaugeField, class SmearingPolicy> class ForceGradient :
+    template<class GaugeField,
-      public Integrator<GaugeField, SmearingPolicy> {
+	     class SmearingPolicy,
+	     class RepresentationPolicy = Representations< FundamentalRepresentation > > class ForceGradient :
+      public Integrator<GaugeField, SmearingPolicy, RepresentationPolicy> {
     private:
       const RealD lambda = 1.0/6.0;;
       const RealD chi    = 1.0/72.0;
@@ -209,9 +215,9 @@ namespace Grid{
       // Looks like dH scales as dt^4. tested wilson/wilson 2 level.
     ForceGradient(GridBase* grid, 
 		  IntegratorParameters Par,
-		  ActionSet<GaugeField> & Aset,
+		  ActionSet<GaugeField, RepresentationPolicy> & Aset,
 		  SmearingPolicy &Sm):
-      Integrator<GaugeField, SmearingPolicy>(grid,Par,Aset, Sm) {};
+      Integrator<GaugeField, SmearingPolicy, RepresentationPolicy>(grid,Par,Aset, Sm) {};
 
 
       void FG_update_P(GaugeField&U, int level,double fg_dt,double ep){

lib/qcd/representations/adjoint.h

@@ -0,0 +1,115 @@
+/*
+ *  Policy classes for the HMC
+ *  Author: Guido Cossu
+*/
+
+#ifndef ADJOINT_H
+#define ADJOINT_H
+
+namespace Grid {
+namespace QCD {
+
+/*
+* This is an helper class for the HMC
+* Should contain only the data for the adjoint representation
+* and the facility to convert from the fundamental -> adjoint
+*/
+
+template <int ncolour>
+class AdjointRep {
+ public:
+  // typdef to be used by the Representations class in HMC to get the
+  // types for the higher representation fields
+  typedef typename SU_Adjoint<ncolour>::LatticeAdjMatrix LatticeMatrix;
+  typedef typename SU_Adjoint<ncolour>::LatticeAdjField LatticeField;
+  static const int Dimension = ncolour * ncolour - 1;
+
+  LatticeField U;
+
+  explicit AdjointRep(GridBase *grid) : U(grid) {}
+
+  void update_representation(const LatticeGaugeField &Uin) {
+    std::cout << GridLogDebug << "Updating adjoint representation\n";
+    // Uin is in the fundamental representation
+    // get the U in AdjointRep
+    // (U_adj)_B = tr[e^a U e^b U^dag]
+    // e^a = t^a/sqrt(T_F)
+    // where t^a is the generator in the fundamental
+    // T_F is 1/2 for the fundamental representation
+    conformable(U, Uin);
+    U = zero;
+    LatticeColourMatrix tmp(Uin._grid);
+
+    Vector<typename SU<ncolour>::Matrix> ta(Dimension);
+
+    // Debug lines
+    // LatticeMatrix uno(Uin._grid);
+    // uno = 1.0;
+    ////////////////
+
+    // FIXME probably not very efficient to get all the generators
+    // everytime
+    for (int a = 0; a < Dimension; a++) SU<ncolour>::generator(a, ta[a]);
+
+    for (int mu = 0; mu < Nd; mu++) {
+      auto Uin_mu = peekLorentz(Uin, mu);
+      auto U_mu = peekLorentz(U, mu);
+      for (int a = 0; a < Dimension; a++) {
+        tmp = 2.0 * adj(Uin_mu) * ta[a] * Uin_mu;
+        for (int b = 0; b < Dimension; b++)
+          pokeColour(U_mu, trace(tmp * ta[b]), a, b);
+      }
+      pokeLorentz(U, U_mu, mu);
+      // Check matrix U_mu, must be real orthogonal
+      // reality
+      /*
+      LatticeMatrix Ucheck = U_mu - conjugate(U_mu);
+      std::cout << GridLogMessage << "Reality check: " << norm2(Ucheck) <<
+      std::endl;
+
+      Ucheck = U_mu * adj(U_mu) - uno;
+      std::cout << GridLogMessage << "orthogonality check: " << norm2(Ucheck) <<
+      std::endl;
+      */
+    }
+  }
+
+  LatticeGaugeField RtoFundamentalProject(const LatticeField &in,
+                                          Real scale = 1.0) const {
+    LatticeGaugeField out(in._grid);
+    out = zero;
+
+    for (int mu = 0; mu < Nd; mu++) {
+      LatticeColourMatrix out_mu(in._grid);  // fundamental representation
+      LatticeMatrix in_mu = peekLorentz(in, mu);
+
+      out_mu = zero;
+
+      typename SU<ncolour>::LatticeAlgebraVector h(in._grid);
+      projectOnAlgebra(h, in_mu, double(Nc) * 2.0);  // factor C(r)/C(fund)
+      FundamentalLieAlgebraMatrix(h, out_mu);   // apply scale only once
+      pokeLorentz(out, out_mu, mu);
+      // Returns traceless antihermitian matrix Nc * Nc.
+      // Confirmed
+    }
+    return out;
+  }
+
+ private:
+  void projectOnAlgebra(typename SU<ncolour>::LatticeAlgebraVector &h_out,
+                        const LatticeMatrix &in, Real scale = 1.0) const {
+    SU_Adjoint<ncolour>::projectOnAlgebra(h_out, in, scale);
+  }
+
+  void FundamentalLieAlgebraMatrix(
+      typename SU<ncolour>::LatticeAlgebraVector &h,
+      typename SU<ncolour>::LatticeMatrix &out, Real scale = 1.0) const {
+    SU<ncolour>::FundamentalLieAlgebraMatrix(h, out, scale);
+  }
+};
+
+typedef AdjointRep<Nc> AdjointRepresentation;
+}
+}
+
+#endif

lib/qcd/representations/fundamental.h

@@ -0,0 +1,45 @@
+/*
+ *	Policy classes for the HMC
+ *	Author: Guido Cossu
+*/	
+
+#ifndef FUNDAMENTAL_H
+#define FUNDAMENTAL_H
+
+
+namespace Grid {
+namespace QCD {
+
+/*
+* This is an helper class for the HMC
+* Empty since HMC updates already the fundamental representation 
+*/
+
+template <int ncolour>
+class FundamentalRep {
+ public:
+  static const int Dimension = ncolour;
+
+  // typdef to be used by the Representations class in HMC to get the
+  // types for the higher representation fields
+  typedef typename SU<ncolour>::LatticeMatrix LatticeMatrix;
+  typedef LatticeGaugeField LatticeField;
+  
+  explicit FundamentalRep(GridBase* grid) {} //do nothing
+  void update_representation(const LatticeGaugeField& Uin) {} // do nothing
+
+  LatticeField RtoFundamentalProject(const LatticeField& in, Real scale = 1.0) const{
+    return (scale * in);
+  }
+
+};
+
+typedef	 FundamentalRep<Nc> FundamentalRepresentation;
+
+}
+}
+
+
+
+
+#endif

lib/qcd/representations/hmc_types.h

@@ -0,0 +1,91 @@
+#ifndef HMC_TYPES_H
+#define HMC_TYPES_H
+
+#include <Grid/qcd/representations/adjoint.h>
+#include <Grid/qcd/representations/two_index.h>
+#include <Grid/qcd/representations/fundamental.h>
+#include <tuple>
+#include <utility>
+
+namespace Grid {
+namespace QCD {
+
+// Supported types
+// enum {Fundamental, Adjoint} repr_type;
+
+// Utility to add support to the HMC for representations other than the
+// fundamental
+template <class... Reptypes>
+class Representations {
+ public:
+  typedef std::tuple<Reptypes...> Representation_type;
+
+  // Size of the tuple, known at compile time
+  static const int tuple_size = sizeof...(Reptypes);
+  // The collection of types for the gauge fields
+  typedef std::tuple<typename Reptypes::LatticeField...> Representation_Fields;
+
+  // To access the Reptypes (FundamentalRepresentation, AdjointRepresentation)
+  template <std::size_t N>
+  using repr_type = typename std::tuple_element<N, Representation_type>::type;
+  // in order to get the typename of the field use
+  // type repr_type<I>::LatticeField
+
+  Representation_type rep;
+
+  // Multiple types constructor
+  explicit Representations(GridBase* grid) : rep(Reptypes(grid)...){};
+
+  int size() { return tuple_size; }
+
+  // update the fields
+  template <std::size_t I = 0>
+  inline typename std::enable_if<(I == tuple_size), void>::type update(
+      LatticeGaugeField& U) {}
+
+  template <std::size_t I = 0>
+  inline typename std::enable_if<(I < tuple_size), void>::type update(
+      LatticeGaugeField& U) {
+    std::get<I>(rep).update_representation(U);
+    update<I + 1>(U);
+  }
+
+
+
+};
+
+typedef Representations<FundamentalRepresentation> NoHirep;
+
+// Helper classes to access the elements
+// Strips the first N parameters from the tuple
+// sequence of classes to obtain the S sequence
+// Creates a type that is a tuple of vectors of the template type A
+template <template <typename> class A, class TupleClass,
+          size_t N = TupleClass::tuple_size, size_t... S>
+struct AccessTypes : AccessTypes<A, TupleClass, N - 1, N - 1, S...> {};
+
+template <template <typename> class A, class TupleClass, size_t... S>
+struct AccessTypes<A, TupleClass, 0, S...> {
+ public:
+  typedef typename TupleClass::Representation_Fields Rfields;
+
+  template <std::size_t N>
+  using elem = typename std::tuple_element<N, Rfields>::type;  // fields types
+
+  typedef std::tuple<std::vector< A< elem<S> >* > ... > VectorCollection;
+  typedef std::tuple< elem<S> ... > FieldTypeCollection;
+
+  // Debug
+  void return_size() {
+    std::cout << GridLogMessage
+              << "Access:" << std::tuple_size<std::tuple<elem<S>...> >::value
+              << "\n";
+    std::cout << GridLogMessage
+              << "Access vectors:" << std::tuple_size<VectorCollection>::value
+              << "\n";
+  }
+};
+}
+}
+
+#endif

lib/qcd/representations/two_index.h

@@ -0,0 +1,99 @@
+/*
+ *  Policy classes for the HMC
+ *  Authors: Guido Cossu, David Preti
+*/
+
+#ifndef SUN2INDEX_H_H
+#define SUN2INDEX_H_H
+
+namespace Grid {
+namespace QCD {
+
+/*
+ * This is an helper class for the HMC
+ * Should contain only the data for the two index representations
+ * and the facility to convert from the fundamental -> two index
+ * The templated parameter TwoIndexSymmetry choses between the 
+ * symmetric and antisymmetric representations
+ * 
+ * There is an 
+ * enum TwoIndexSymmetry { Symmetric = 1, AntiSymmetric = -1 };
+ * in the SUnTwoIndex.h file
+ */
+
+template <int ncolour, TwoIndexSymmetry S>
+class TwoIndexRep {
+ public:
+  // typdef to be used by the Representations class in HMC to get the
+  // types for the higher representation fields
+  typedef typename SU_TwoIndex<ncolour, S>::LatticeTwoIndexMatrix LatticeMatrix;
+  typedef typename SU_TwoIndex<ncolour, S>::LatticeTwoIndexField LatticeField;
+  static const int Dimension = ncolour * (ncolour + S) / 2;
+
+  LatticeField U;
+
+  explicit TwoIndexRep(GridBase *grid) : U(grid) {}
+
+  void update_representation(const LatticeGaugeField &Uin) {
+    std::cout << GridLogDebug << "Updating TwoIndex representation\n";
+    // Uin is in the fundamental representation
+    // get the U in TwoIndexRep
+    // (U)_{(ij)(lk)} = tr [ adj(e^(ij)) U e^(lk) transpose(U) ]
+    conformable(U, Uin);
+    U = zero;
+    LatticeColourMatrix tmp(Uin._grid);
+
+    Vector<typename SU<ncolour>::Matrix> eij(Dimension);
+
+    for (int a = 0; a < Dimension; a++)
+      SU_TwoIndex<ncolour, S>::base(a, eij[a]);
+
+    for (int mu = 0; mu < Nd; mu++) {
+      auto Uin_mu = peekLorentz(Uin, mu);
+      auto U_mu = peekLorentz(U, mu);
+      for (int a = 0; a < Dimension; a++) {
+        tmp = transpose(Uin_mu) * adj(eij[a]) * Uin_mu;
+        for (int b = 0; b < Dimension; b++)
+          pokeColour(U_mu, trace(tmp * eij[b]), a, b);
+      }
+      pokeLorentz(U, U_mu, mu);
+    }
+  }
+
+  LatticeGaugeField RtoFundamentalProject(const LatticeField &in,
+                                          Real scale = 1.0) const {
+    LatticeGaugeField out(in._grid);
+    out = zero;
+
+    for (int mu = 0; mu < Nd; mu++) {
+      LatticeColourMatrix out_mu(in._grid);  // fundamental representation
+      LatticeMatrix in_mu = peekLorentz(in, mu);
+
+      out_mu = zero;
+
+      typename SU<ncolour>::LatticeAlgebraVector h(in._grid);
+      projectOnAlgebra(h, in_mu, double(Nc + 2 * S));  // factor T(r)/T(fund)
+      FundamentalLieAlgebraMatrix(h, out_mu);          // apply scale only once
+      pokeLorentz(out, out_mu, mu);
+    }
+    return out;
+  }
+
+ private:
+  void projectOnAlgebra(typename SU<ncolour>::LatticeAlgebraVector &h_out,
+                        const LatticeMatrix &in, Real scale = 1.0) const {
+    SU_TwoIndex<ncolour, S>::projectOnAlgebra(h_out, in, scale);
+  }
+
+  void FundamentalLieAlgebraMatrix(
+      typename SU<ncolour>::LatticeAlgebraVector &h,
+      typename SU<ncolour>::LatticeMatrix &out, Real scale = 1.0) const {
+    SU<ncolour>::FundamentalLieAlgebraMatrix(h, out, scale);
+  }
+};
+
+typedef TwoIndexRep<Nc, Symmetric> TwoIndexSymmetricRepresentation;
+typedef TwoIndexRep<Nc, AntiSymmetric> TwoIndexAntiSymmetricRepresentation;
+}
+}
+#endif

lib/qcd/smearing/GaugeConfiguration.h

@@ -10,6 +10,29 @@ namespace Grid {
 
 namespace QCD {
 
+  //trivial class for no smearing
+  template< class Gimpl >
+class NoSmearing {
+public:
+  INHERIT_GIMPL_TYPES(Gimpl);
+
+  GaugeField*
+      ThinLinks;
+
+  NoSmearing(): ThinLinks(NULL) {}
+
+  void set_GaugeField(GaugeField& U) { ThinLinks = &U; }
+
+  void smeared_force(GaugeField& SigmaTilde) const {}
+
+  GaugeField& get_SmearedU() { return *ThinLinks; }
+
+  GaugeField& get_U(bool smeared = false) {
+    return *ThinLinks;
+  }
+
+};
+
 /*!
   @brief Smeared configuration container
 
@@ -201,6 +224,8 @@ class SmearedConfiguration {
   SmearedConfiguration()
       : smearingLevels(0), StoutSmearing(), SmearedSet(), ThinLinks(NULL) {}
 
+
+  
   // attach the smeared routines to the thin links U and fill the smeared set
   void set_GaugeField(GaugeField& U) { fill_smearedSet(U); }
 

lib/qcd/smearing/StoutSmearing.h

@@ -18,14 +18,12 @@ class Smear_Stout : public Smear<Gimpl> {
   INHERIT_GIMPL_TYPES(Gimpl)
 
   Smear_Stout(Smear<Gimpl>* base) : SmearBase(base) {
-    static_assert(Nc == 3,
+    assert(Nc == 3);//                  "Stout smearing currently implemented only for Nc==3");
-                  "Stout smearing currently implemented only for Nc==3");
   }
 
   /*! Default constructor */
   Smear_Stout(double rho = 1.0) : SmearBase(new Smear_APE<Gimpl>(rho)) {
-    static_assert(Nc == 3,
+    assert(Nc == 3);//                  "Stout smearing currently implemented only for Nc==3");
-                  "Stout smearing currently implemented only for Nc==3");
   }
 
   ~Smear_Stout() {}  // delete SmearBase...

lib/qcd/utils/SUnAdjoint.h

@@ -0,0 +1,182 @@
+#ifndef QCD_UTIL_SUNADJOINT_H
+#define QCD_UTIL_SUNADJOINT_H
+
+////////////////////////////////////////////////////////////////////////
+//
+// * Adjoint representation generators
+//
+// * Normalisation for the fundamental generators: 
+//   trace ta tb = 1/2 delta_ab = T_F delta_ab
+//   T_F = 1/2  for SU(N) groups
+//
+//
+//   base for NxN hermitian traceless matrices
+//   normalized to 1:
+//
+//   (e_Adj)^a = t^a / sqrt(T_F)
+//
+//   then the real, antisymmetric generators for the adjoint representations
+//   are computed ( shortcut: e^a == (e_Adj)^a )
+//
+//   (iT_adj)^d_ba = i tr[e^a t^d e^b - t^d e^a e^b]
+//
+////////////////////////////////////////////////////////////////////////
+
+namespace Grid {
+namespace QCD {
+
+template <int ncolour>
+class SU_Adjoint : public SU<ncolour> {
+ public:
+  static const int Dimension = ncolour * ncolour - 1;
+
+  template <typename vtype>
+  using iSUnAdjointMatrix =
+      iScalar<iScalar<iMatrix<vtype, Dimension > > >;
+
+  // Actually the adjoint matrices are real...
+  // Consider this overhead... FIXME
+  typedef iSUnAdjointMatrix<Complex> AMatrix;
+  typedef iSUnAdjointMatrix<ComplexF> AMatrixF;
+  typedef iSUnAdjointMatrix<ComplexD> AMatrixD;
+
+  typedef iSUnAdjointMatrix<vComplex> vAMatrix;
+  typedef iSUnAdjointMatrix<vComplexF> vAMatrixF;
+  typedef iSUnAdjointMatrix<vComplexD> vAMatrixD;
+
+  typedef Lattice<vAMatrix>  LatticeAdjMatrix;
+  typedef Lattice<vAMatrixF> LatticeAdjMatrixF;
+  typedef Lattice<vAMatrixD> LatticeAdjMatrixD;
+
+  typedef Lattice<iVector<iScalar<iMatrix<vComplex, Dimension> >, Nd> >
+      LatticeAdjField;
+  typedef Lattice<iVector<iScalar<iMatrix<vComplexF, Dimension> >, Nd> >
+      LatticeAdjFieldF;
+  typedef Lattice<iVector<iScalar<iMatrix<vComplexD, Dimension> >, Nd> >
+      LatticeAdjFieldD;
+
+
+
+
+  template <class cplx>
+  static void generator(int Index, iSUnAdjointMatrix<cplx> &iAdjTa) {
+    // returns i(T_Adj)^index necessary for the projectors
+    // see definitions above
+    iAdjTa = zero;
+    Vector<typename SU<ncolour>::template iSUnMatrix<cplx> > ta(ncolour * ncolour - 1);
+    typename SU<ncolour>::template iSUnMatrix<cplx> tmp;
+
+    // FIXME not very efficient to get all the generators everytime
+    for (int a = 0; a < Dimension; a++) SU<ncolour>::generator(a, ta[a]);
+
+    for (int a = 0; a < Dimension; a++) {
+      tmp = ta[a] * ta[Index] - ta[Index] * ta[a];
+      for (int b = 0; b < (ncolour * ncolour - 1); b++) {
+        typename SU<ncolour>::template iSUnMatrix<cplx> tmp1 =
+            2.0 * tmp * ta[b];  // 2.0 from the normalization
+        Complex iTr = TensorRemove(timesI(trace(tmp1)));
+        //iAdjTa()()(b, a) = iTr;
+        iAdjTa()()(a, b) = iTr;
+      }
+    }
+  }
+
+  static void printGenerators(void) {
+    for (int gen = 0; gen < Dimension; gen++) {
+      AMatrix ta;
+      generator(gen, ta);
+      std::cout << GridLogMessage << "Nc = " << ncolour << " t_" << gen
+                << std::endl;
+      std::cout << GridLogMessage << ta << std::endl;
+    }
+  }
+
+  static void testGenerators(void) {
+    AMatrix adjTa;
+    std::cout << GridLogMessage << "Adjoint - Checking if real" << std::endl;
+    for (int a = 0; a < Dimension; a++) {
+      generator(a, adjTa);
+      std::cout << GridLogMessage << a << std::endl;
+      assert(norm2(adjTa - conjugate(adjTa)) < 1.0e-6);
+    }
+    std::cout << GridLogMessage << std::endl;
+
+    std::cout << GridLogMessage << "Adjoint - Checking if antisymmetric"
+              << std::endl;
+    for (int a = 0; a < Dimension; a++) {
+      generator(a, adjTa);
+      std::cout << GridLogMessage << a << std::endl;
+      assert(norm2(adjTa + transpose(adjTa)) < 1.0e-6);
+    }
+    std::cout << GridLogMessage << std::endl;
+  }
+
+  static void AdjointLieAlgebraMatrix(
+      const typename SU<ncolour>::LatticeAlgebraVector &h,
+      LatticeAdjMatrix &out, Real scale = 1.0) {
+    conformable(h, out);
+    GridBase *grid = out._grid;
+    LatticeAdjMatrix la(grid);
+    AMatrix iTa;
+
+    out = zero;
+    for (int a = 0; a < Dimension; a++) {
+      generator(a, iTa);
+      la = peekColour(h, a) * iTa;
+      out += la;
+    }
+    out *= scale;
+  }
+
+  // Projects the algebra components a lattice matrix (of dimension ncol*ncol -1 )
+  static void projectOnAlgebra(typename SU<ncolour>::LatticeAlgebraVector &h_out, const LatticeAdjMatrix &in, Real scale = 1.0) {
+    conformable(h_out, in);
+    h_out = zero;
+    AMatrix iTa;
+    Real coefficient = - 1.0/(ncolour) * scale;// 1/Nc for the normalization of the trace in the adj rep
+
+    for (int a = 0; a < Dimension; a++) {
+      generator(a, iTa);
+      auto tmp = real(trace(iTa * in)) * coefficient;
+      pokeColour(h_out, tmp, a);
+    }
+  }
+
+  // a projector that keeps the generators stored to avoid the overhead of recomputing them 
+  static void projector(typename SU<ncolour>::LatticeAlgebraVector &h_out, const LatticeAdjMatrix &in, Real scale = 1.0) {
+    conformable(h_out, in);
+    static std::vector<AMatrix> iTa(Dimension);  // to store the generators
+    h_out = zero;
+    static bool precalculated = false; 
+    if (!precalculated){
+      precalculated = true;
+        for (int a = 0; a < Dimension; a++) generator(a, iTa[a]);
+    }
+
+    Real coefficient = -1.0 / (ncolour) * scale;  // 1/Nc for the normalization of
+                                                // the trace in the adj rep
+
+    for (int a = 0; a < Dimension; a++) {
+      auto tmp = real(trace(iTa[a] * in)) * coefficient; 
+      pokeColour(h_out, tmp, a);
+    }
+  }
+
+
+};
+
+
+
+
+// Some useful type names
+
+typedef SU_Adjoint<2> SU2Adjoint;
+typedef SU_Adjoint<3> SU3Adjoint;
+typedef SU_Adjoint<4> SU4Adjoint;
+typedef SU_Adjoint<5> SU5Adjoint;
+
+typedef SU_Adjoint<Nc> AdjointMatrices;
+}
+}
+
+#endif

lib/qcd/utils/SUnTwoIndex.h

@@ -0,0 +1,276 @@
+////////////////////////////////////////////////////////////////////////
+//
+// * Two index representation generators
+//
+// * Normalisation for the fundamental generators:
+//   trace ta tb = 1/2 delta_ab = T_F delta_ab
+//   T_F = 1/2  for SU(N) groups
+//
+//
+//   base for NxN two index (anti-symmetric) matrices
+//   normalized to 1 (d_ij is the kroenecker delta)
+//
+//   (e^(ij)_{kl} = 1 / sqrt(2) (d_ik d_jl +/- d_jk d_il)
+//
+//   Then the generators are written as
+//
+//   (iT_a)^(ij)(lk) = i * ( tr[e^(ij)^dag e^(lk) T^trasp_a] +
+//   tr[e^(lk)e^(ij)^dag T_a] )  //
+//   
+//
+////////////////////////////////////////////////////////////////////////
+
+// Authors: David Preti, Guido Cossu
+
+#ifndef QCD_UTIL_SUN2INDEX_H
+#define QCD_UTIL_SUN2INDEX_H
+
+
+namespace Grid {
+namespace QCD {
+
+enum TwoIndexSymmetry { Symmetric = 1, AntiSymmetric = -1 };
+
+inline Real delta(int a, int b) { return (a == b) ? 1.0 : 0.0; }
+
+template <int ncolour, TwoIndexSymmetry S>
+class SU_TwoIndex : public SU<ncolour> {
+ public:
+  static const int Dimension = ncolour * (ncolour + S) / 2;
+  static const int NumGenerators = SU<ncolour>::AdjointDimension;
+
+  template <typename vtype>
+  using iSUnTwoIndexMatrix = iScalar<iScalar<iMatrix<vtype, Dimension> > >;
+
+  typedef iSUnTwoIndexMatrix<Complex> TIMatrix;
+  typedef iSUnTwoIndexMatrix<ComplexF> TIMatrixF;
+  typedef iSUnTwoIndexMatrix<ComplexD> TIMatrixD;
+
+  typedef iSUnTwoIndexMatrix<vComplex> vTIMatrix;
+  typedef iSUnTwoIndexMatrix<vComplexF> vTIMatrixF;
+  typedef iSUnTwoIndexMatrix<vComplexD> vTIMatrixD;
+
+  typedef Lattice<vTIMatrix> LatticeTwoIndexMatrix;
+  typedef Lattice<vTIMatrixF> LatticeTwoIndexMatrixF;
+  typedef Lattice<vTIMatrixD> LatticeTwoIndexMatrixD;
+
+  typedef Lattice<iVector<iScalar<iMatrix<vComplex, Dimension> >, Nd> >
+      LatticeTwoIndexField;
+  typedef Lattice<iVector<iScalar<iMatrix<vComplexF, Dimension> >, Nd> >
+      LatticeTwoIndexFieldF;
+  typedef Lattice<iVector<iScalar<iMatrix<vComplexD, Dimension> >, Nd> >
+      LatticeTwoIndexFieldD;
+
+  template <typename vtype>
+  using iSUnMatrix = iScalar<iScalar<iMatrix<vtype, ncolour> > >;
+
+  typedef iSUnMatrix<Complex> Matrix;
+  typedef iSUnMatrix<ComplexF> MatrixF;
+  typedef iSUnMatrix<ComplexD> MatrixD;
+
+  template <class cplx>
+  static void base(int Index, iSUnMatrix<cplx> &eij) {
+    // returns (e)^(ij)_{kl} necessary for change of base U_F -> U_R
+    assert(Index < NumGenerators);
+    eij = zero;
+
+    // for the linearisation of the 2 indexes 
+    static int a[ncolour * (ncolour - 1) / 2][2]; // store the a <-> i,j
+    static bool filled = false;
+    if (!filled) {
+      int counter = 0;
+      for (int i = 1; i < ncolour; i++) {
+        for (int j = 0; j < i; j++) {
+          a[counter][0] = i;
+          a[counter][1] = j;
+          counter++;
+        }
+      }
+      filled = true;
+    }
+
+    if (Index < ncolour * (ncolour - 1) / 2) {
+      baseOffDiagonal(a[Index][0], a[Index][1], eij);
+    } else {
+      baseDiagonal(Index, eij);
+    }
+  }
+
+  template <class cplx>
+  static void baseDiagonal(int Index, iSUnMatrix<cplx> &eij) {
+    eij = zero;
+    eij()()(Index - ncolour * (ncolour - 1) / 2,
+            Index - ncolour * (ncolour - 1) / 2) = 1.0;
+  }
+
+  template <class cplx>
+  static void baseOffDiagonal(int i, int j, iSUnMatrix<cplx> &eij) {
+    eij = zero;
+    for (int k = 0; k < ncolour; k++)
+      for (int l = 0; l < ncolour; l++)
+        eij()()(l, k) = delta(i, k) * delta(j, l) +
+                        S * delta(j, k) * delta(i, l);
+
+    RealD nrm = 1. / std::sqrt(2.0);
+    eij = eij * nrm;
+  }
+
+  static void printBase(void) {
+    for (int gen = 0; gen < Dimension; gen++) {
+      Matrix tmp;
+      base(gen, tmp);
+      std::cout << GridLogMessage << "Nc = " << ncolour << " t_" << gen
+                << std::endl;
+      std::cout << GridLogMessage << tmp << std::endl;
+    }
+  }
+
+  template <class cplx>
+  static void generator(int Index, iSUnTwoIndexMatrix<cplx> &i2indTa) {
+    Vector<typename SU<ncolour>::template iSUnMatrix<cplx> > ta(
+        ncolour * ncolour - 1);
+    Vector<typename SU<ncolour>::template iSUnMatrix<cplx> > eij(Dimension);
+    typename SU<ncolour>::template iSUnMatrix<cplx> tmp;
+    i2indTa = zero;
+    
+    for (int a = 0; a < ncolour * ncolour - 1; a++)
+      SU<ncolour>::generator(a, ta[a]);
+    
+    for (int a = 0; a < Dimension; a++) base(a, eij[a]);
+
+    for (int a = 0; a < Dimension; a++) {
+      tmp = transpose(ta[Index]) * adj(eij[a]) + adj(eij[a]) * ta[Index];
+      for (int b = 0; b < Dimension; b++) {
+        typename SU<ncolour>::template iSUnMatrix<cplx> tmp1 =
+            tmp * eij[b]; 
+        Complex iTr = TensorRemove(timesI(trace(tmp1)));
+        i2indTa()()(a, b) = iTr;
+      }
+    }
+  }
+
+  static void printGenerators(void) {
+    for (int gen = 0; gen < ncolour * ncolour - 1; gen++) {
+      TIMatrix i2indTa;
+      generator(gen, i2indTa);
+      std::cout << GridLogMessage << "Nc = " << ncolour << " t_" << gen
+                << std::endl;
+      std::cout << GridLogMessage << i2indTa << std::endl;
+    }
+  }
+
+  static void testGenerators(void) {
+    TIMatrix i2indTa, i2indTb;
+    std::cout << GridLogMessage << "2IndexRep - Checking if traceless"
+              << std::endl;
+    for (int a = 0; a < ncolour * ncolour - 1; a++) {
+      generator(a, i2indTa);
+      std::cout << GridLogMessage << a << std::endl;
+      assert(norm2(trace(i2indTa)) < 1.0e-6);
+    }
+    std::cout << GridLogMessage << std::endl;
+
+    std::cout << GridLogMessage << "2IndexRep - Checking if antihermitean"
+              << std::endl;
+    for (int a = 0; a < ncolour * ncolour - 1; a++) {
+      generator(a, i2indTa);
+      std::cout << GridLogMessage << a << std::endl;
+      assert(norm2(adj(i2indTa) + i2indTa) < 1.0e-6);
+    }
+
+    std::cout << GridLogMessage << std::endl;
+    std::cout << GridLogMessage
+              << "2IndexRep - Checking Tr[Ta*Tb]=delta(a,b)*(N +- 2)/2"
+              << std::endl;
+    for (int a = 0; a < ncolour * ncolour - 1; a++) {
+      for (int b = 0; b < ncolour * ncolour - 1; b++) {
+        generator(a, i2indTa);
+        generator(b, i2indTb);
+
+        // generator returns iTa, so we need a minus sign here
+        Complex Tr = -TensorRemove(trace(i2indTa * i2indTb));
+        std::cout << GridLogMessage << "a=" << a << "b=" << b << "Tr=" << Tr
+                  << std::endl;
+      }
+    }
+    std::cout << GridLogMessage << std::endl;
+  }
+
+  static void TwoIndexLieAlgebraMatrix(
+      const typename SU<ncolour>::LatticeAlgebraVector &h,
+      LatticeTwoIndexMatrix &out, Real scale = 1.0) {
+    conformable(h, out);
+    GridBase *grid = out._grid;
+    LatticeTwoIndexMatrix la(grid);
+    TIMatrix i2indTa;
+
+    out = zero;
+    for (int a = 0; a < ncolour * ncolour - 1; a++) {
+      generator(a, i2indTa);
+      la = peekColour(h, a) * i2indTa;
+      out += la;
+    }
+    out *= scale;
+  }
+
+  // Projects the algebra components 
+  // of a lattice matrix ( of dimension ncol*ncol -1 )
+  static void projectOnAlgebra(
+      typename SU<ncolour>::LatticeAlgebraVector &h_out,
+      const LatticeTwoIndexMatrix &in, Real scale = 1.0) {
+    conformable(h_out, in);
+    h_out = zero;
+    TIMatrix i2indTa;
+    Real coefficient = -2.0 / (ncolour + 2 * S) * scale;
+    // 2/(Nc +/- 2) for the normalization of the trace in the two index rep
+    for (int a = 0; a < ncolour * ncolour - 1; a++) {
+      generator(a, i2indTa);
+      auto tmp = real(trace(i2indTa * in)) * coefficient;
+      pokeColour(h_out, tmp, a);
+    }
+  }
+
+  // a projector that keeps the generators stored to avoid the overhead of
+  // recomputing them
+  static void projector(typename SU<ncolour>::LatticeAlgebraVector &h_out,
+                        const LatticeTwoIndexMatrix &in, Real scale = 1.0) {
+    conformable(h_out, in);
+    // to store the generators
+    static std::vector<TIMatrix> i2indTa(ncolour * ncolour -1); 
+    h_out = zero;
+    static bool precalculated = false;
+    if (!precalculated) {
+      precalculated = true;
+      for (int a = 0; a < ncolour * ncolour - 1; a++) generator(a, i2indTa[a]);
+    }
+
+    Real coefficient =
+        -2.0 / (ncolour + 2 * S) * scale;  // 2/(Nc +/- 2) for the normalization
+                                           // of the trace in the two index rep
+
+    for (int a = 0; a < ncolour * ncolour - 1; a++) {
+      auto tmp = real(trace(i2indTa[a] * in)) * coefficient;
+      pokeColour(h_out, tmp, a);
+    }
+  }
+};
+
+// Some useful type names
+typedef SU_TwoIndex<Nc, Symmetric> TwoIndexSymmMatrices;
+typedef SU_TwoIndex<Nc, AntiSymmetric> TwoIndexAntiSymmMatrices;
+
+typedef SU_TwoIndex<2, Symmetric> SU2TwoIndexSymm;
+typedef SU_TwoIndex<3, Symmetric> SU3TwoIndexSymm;
+typedef SU_TwoIndex<4, Symmetric> SU4TwoIndexSymm;
+typedef SU_TwoIndex<5, Symmetric> SU5TwoIndexSymm;
+
+typedef SU_TwoIndex<2, AntiSymmetric> SU2TwoIndexAntiSymm;
+typedef SU_TwoIndex<3, AntiSymmetric> SU3TwoIndexAntiSymm;
+typedef SU_TwoIndex<4, AntiSymmetric> SU4TwoIndexAntiSymm;
+typedef SU_TwoIndex<5, AntiSymmetric> SU5TwoIndexAntiSymm;
+
+
+}
+}
+
+#endif

lib/simd/Grid_avx.h

@@ -454,7 +454,7 @@ namespace Optimization {
 #define _mm256_alignr_epi64_grid(ret,a,b,n) ret=(__m256d) _mm256_alignr_epi8((__m256i)a,(__m256i)b,(n*8)%16)
 #endif
 
-#if defined (AVX1) || defined (AVXFMA4)  
+#if defined (AVX1) || defined (AVXFMA)  
 #define _mm256_alignr_epi32_grid(ret,a,b,n) {	\
     __m128 aa, bb;				\
 						\

lib/simd/Grid_avx512.h

@@ -370,7 +370,67 @@ namespace Optimization {
 
   //////////////////////////////////////////////
   // Some Template specialization
+
+  // Hack for CLANG until mm512_reduce_add_ps etc... are implemented in GCC and Clang releases
+#undef GNU_CLANG_COMPILER 
+#ifdef GNU_CLANG_COMPILER
+  //Complex float Reduce
+  template<>
+    inline Grid::ComplexF Reduce<Grid::ComplexF, __m512>::operator()(__m512 in){
+    __m512 v1,v2;
+    v1=Optimization::Permute::Permute0(in); // avx 512; quad complex single
+    v1= _mm512_add_ps(v1,in);
+    v2=Optimization::Permute::Permute1(v1); 
+    v1 = _mm512_add_ps(v1,v2);
+    v2=Optimization::Permute::Permute2(v1); 
+    v1 = _mm512_add_ps(v1,v2);
+    u512f conv; conv.v = v1;
+    return Grid::ComplexF(conv.f[0],conv.f[1]);
+  }
   
+  //Real float Reduce
+  template<>
+    inline Grid::RealF Reduce<Grid::RealF, __m512>::operator()(__m512 in){
+    __m512 v1,v2;
+    v1 = Optimization::Permute::Permute0(in); // avx 512; octo-double
+    v1 = _mm512_add_ps(v1,in);
+    v2 = Optimization::Permute::Permute1(v1); 
+    v1 = _mm512_add_ps(v1,v2);
+    v2 = Optimization::Permute::Permute2(v1); 
+    v1 = _mm512_add_ps(v1,v2);
+    v2 = Optimization::Permute::Permute3(v1); 
+    v1 = _mm512_add_ps(v1,v2);
+    u512f conv; conv.v=v1;
+    return conv.f[0];
+  }
+  
+  
+  //Complex double Reduce
+  template<>
+    inline Grid::ComplexD Reduce<Grid::ComplexD, __m512d>::operator()(__m512d in){
+    __m512d v1;
+    v1 = Optimization::Permute::Permute0(in); // sse 128; paired complex single
+    v1 = _mm512_add_pd(v1,in);
+    v1 = Optimization::Permute::Permute1(in); // sse 128; paired complex single
+    v1 = _mm512_add_pd(v1,in);
+    u512d conv; conv.v = v1;
+    return Grid::ComplexD(conv.f[0],conv.f[1]);
+  }
+  
+  //Real double Reduce
+  template<>
+    inline Grid::RealD Reduce<Grid::RealD, __m512d>::operator()(__m512d in){
+    __m512d v1,v2;
+    v1 = Optimization::Permute::Permute0(in); // avx 512; quad double
+    v1 = _mm512_add_pd(v1,in);
+      v2 = Optimization::Permute::Permute1(v1); 
+      v1 = _mm512_add_pd(v1,v2);
+      v2 = Optimization::Permute::Permute2(v1); 
+      v1 = _mm512_add_pd(v1,v2);
+     u512d conv; conv.v = v1;
+     return conv.f[0];
+  }
+#else
   //Complex float Reduce
   template<>
   inline Grid::ComplexF Reduce<Grid::ComplexF, __m512>::operator()(__m512 in){
@@ -382,7 +442,6 @@ namespace Optimization {
     return _mm512_reduce_add_ps(in);
   }
   
-  
   //Complex double Reduce
   template<>
   inline Grid::ComplexD Reduce<Grid::ComplexD, __m512d>::operator()(__m512d in){
@@ -402,6 +461,7 @@ namespace Optimization {
     printf("Reduce : Missing integer implementation -> FIX\n");
     assert(0);
   }
+#endif
   
   
 }

lib/simd/Grid_qpx.h

@@ -1,300 +1,421 @@
-    /*************************************************************************************
+/*******************************************************************************
-
+ 
-    Grid physics library, www.github.com/paboyle/Grid 
+ Grid physics library, www.github.com/paboyle/Grid
-
+ 
-    Source file: ./lib/simd/Grid_qpx.h
+ Source file: ./lib/simd/Grid_qpx.h
-
+ 
-    Copyright (C) 2015
+ Copyright (C) 2016
-
+ 
-Author: neo <cossu@post.kek.jp>
+ Author: Antonin Portelli <antonin.portelli@me.com>
-
+ 
-    This program is free software; you can redistribute it and/or modify
+ 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
+ it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+ the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+ (at your option) any later version.
-
+ 
-    This program is distributed in the hope that it will be useful,
+ This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+ GNU General Public License for more details.
-
+ 
-    You should have received a copy of the GNU General Public License along
+ 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.,
+ with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+ 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
-
+ 
-    See the full license in the file "LICENSE" in the top level distribution directory
+ See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
+ ******************************************************************************/
-    /*  END LEGAL */
-//----------------------------------------------------------------------
-/*! @file Grid_qpx.h
-  @brief Optimization libraries for QPX instructions set for BG/Q
-
-  Using intrinsics
-*/
-// Time-stamp: <2015-05-27 11:30:21 neo>
-//----------------------------------------------------------------------
-
-// lot of undefined functions
 
+namespace Grid {
 namespace Optimization {
+  typedef struct 
+  {
+    float v0,v1,v2,v3;
+  } vector4float;
+
+  inline std::ostream & operator<<(std::ostream& stream, const vector4double a)
+  {
+    stream << "{"<<vec_extract(a,0)<<","<<vec_extract(a,1)<<","<<vec_extract(a,2)<<","<<vec_extract(a,3)<<"}";
+    return stream;
+  };
+
+  inline std::ostream & operator<<(std::ostream& stream, const vector4float a)
+  {
+    stream << "{"<< a.v0 <<","<< a.v1 <<","<< a.v2 <<","<< a.v3 <<"}";
+    return stream;
+  };
   
   struct Vsplat{
     //Complex float
-    inline float operator()(float a, float b){
+    inline vector4float operator()(float a, float b){
-      return {a,b,a,b};
+      return (vector4float){a, b, a, b};
     }
     // Real float
-    inline float operator()(float a){
+    inline vector4float operator()(float a){
-      return {a,a,a,a};
+      return (vector4float){a, a, a, a};
     }
     //Complex double
     inline vector4double operator()(double a, double b){
-      return {a,b,a,b};
+      return (vector4double){a, b, a, b};
     }
     //Real double
     inline vector4double operator()(double a){
-      return {a,a,a,a};
+      return (vector4double){a, a, a, a};
     }
     //Integer
     inline int operator()(Integer a){
-#error
+      return a;
     }
   };
-
+  
   struct Vstore{
-    //Float 
+    //Float
-    inline void operator()(float a, float* F){
+    inline void operator()(vector4double a, float *f){
-      assert(0);
+      vec_st(a, 0, f);
     }
+
+    inline void operator()(vector4double a, vector4float &f){
+      vec_st(a, 0, (float *)(&f));
+    }
+
+    inline void operator()(vector4float a, float *f){
+      f[0] = a.v0;
+      f[1] = a.v1;
+      f[2] = a.v2;
+      f[3] = a.v3;
+    }
+
     //Double
-    inline void operator()(vector4double a, double* D){
+    inline void operator()(vector4double a, double *d){
-      assert(0);
+      vec_st(a, 0, d);
     }
     //Integer
-    inline void operator()(int a, Integer* I){
+    inline void operator()(int a, Integer *i){
-      assert(0);
+      i[0] = a;
     }
-
   };
-
+  
-
   struct Vstream{
     //Float
-    inline void operator()(float * a, float b){
+    inline void operator()(float *f, vector4double a){
-      assert(0);
+      vec_st(a, 0, f);
-    }
-    //Double
-    inline void operator()(double * a, vector4double b){
-      assert(0);
     }
 
+    inline void operator()(vector4float f, vector4double a){
+      vec_st(a, 0, (float *)(&f));
+    }
+
+    inline void operator()(float *f, vector4float a){
+      f[0] = a.v0;
+      f[1] = a.v1;
+      f[2] = a.v2;
+      f[3] = a.v3;
+    }
+
+    //Double
+    inline void operator()(double *d, vector4double a){
+      vec_st(a, 0, d);
+    }
 
   };
-
+  
-
-
   struct Vset{
-    // Complex float 
+    // Complex float
-    inline float operator()(Grid::ComplexF *a){
+    inline vector4float operator()(Grid::ComplexF *a){
-      return {a[0].real(),a[0].imag(),a[1].real(),a[1].imag(),a[2].real(),a[2].imag(),a[3].real(),a[3].imag()};
+      return (vector4float){a[0].real(), a[0].imag(), a[1].real(), a[1].imag()};
     }
-    // Complex double 
+    // Complex double
     inline vector4double operator()(Grid::ComplexD *a){
-      return {a[0].real(),a[0].imag(),a[1].real(),a[1].imag(),a[2].real(),a[2].imag(),a[3].real(),a[3].imag()};
+      return vec_ld(0, (double *)a);
     }
-    // Real float 
+
-    inline float operator()(float *a){
+    // Real float
-      return {a[0],a[1],a[2],a[3],a[4],a[5],a[6],a[7]};
+    inline vector4float operator()(float *a){
+      return (vector4float){a[0], a[1], a[2], a[3]};
     }
+
+    inline vector4double operator()(vector4float a){
+      return vec_ld(0, (float *)(&a));
+    }
+
     // Real double
     inline vector4double operator()(double *a){
-      return {a[0],a[1],a[2],a[3],a[4],a[5],a[6],a[7]};
+      return vec_ld(0, a);
     }
     // Integer
     inline int operator()(Integer *a){
-#error
+      return a[0];
     }
-
+    
-
+    
   };
-
+  
   template <typename Out_type, typename In_type>
-    struct Reduce{
+  struct Reduce{
-      //Need templated class to overload output type
+    //Need templated class to overload output type
-      //General form must generate error if compiled
+    //General form must generate error if compiled
-      inline Out_type operator()(In_type in){
+    inline Out_type operator()(In_type in){
-	printf("Error, using wrong Reduce function\n");
+      printf("Error, using wrong Reduce function\n");
-	exit(1);
+      exit(1);
-	return 0;
+      return 0;
-      }
+    }
-    };
+  };
-
+  
-
- 
-
   /////////////////////////////////////////////////////
   // Arithmetic operations
   /////////////////////////////////////////////////////
+  #define FLOAT_WRAP_2(fn, pref)\
+  pref vector4float fn(vector4float a, vector4float b)\
+  {\
+    vector4double ad, bd, rd;\
+    vector4float  r;\
+    \
+    ad = Vset()(a);\
+    bd = Vset()(b);\
+    rd = fn(ad, bd);\
+    Vstore()(rd, r);\
+    \
+    return r;\
+  }
+
+  #define FLOAT_WRAP_1(fn, pref)\
+  pref vector4float fn(vector4float a)\
+  {\
+    vector4double ad, rd;\
+    vector4float  r;\
+    \
+    ad = Vset()(a);\
+    rd = fn(ad);\
+    Vstore()(rd, r);\
+    \
+    return r;\
+  }
+
   struct Sum{
-    //Complex/Real float
-    inline float operator()(float a, float b){
-#error
-    }
     //Complex/Real double
     inline vector4double operator()(vector4double a, vector4double b){
-      return vec_add(a,b);
+      return vec_add(a, b);
     }
+
+    //Complex/Real float
+    FLOAT_WRAP_2(operator(), inline)
+
     //Integer
     inline int operator()(int a, int b){
-#error
+      return a + b;
     }
   };
-
+  
   struct Sub{
-    //Complex/Real float
-    inline float operator()(float a, float b){
-#error
-    }
     //Complex/Real double
     inline vector4double operator()(vector4double a, vector4double b){
-#error
+      return vec_sub(a, b);
     }
+
+    //Complex/Real float
+    FLOAT_WRAP_2(operator(), inline)
+
     //Integer
-    inline floati operator()(int a, int b){
+    inline int operator()(int a, int b){
-#error
+      return a - b;
     }
   };
-
+  
-
   struct MultComplex{
-    // Complex float
-    inline float operator()(float a, float b){
-#error
-    }
     // Complex double
     inline vector4double operator()(vector4double a, vector4double b){
-#error
+      return vec_xxnpmadd(a, b, vec_xmul(b, a));
     }
-  };
 
+    // Complex float
+    FLOAT_WRAP_2(operator(), inline)
+  };
+  
   struct Mult{
-    // Real float
-    inline float operator()(float a, float b){
-#error
-    }
     // Real double
     inline vector4double operator()(vector4double a, vector4double b){
-#error
+      return vec_mul(a, b);
     }
+
+    // Real float
+    FLOAT_WRAP_2(operator(), inline)
+
     // Integer
     inline int operator()(int a, int b){
-#error
+      return a*b;
     }
   };
-
+  
-
   struct Conj{
-    // Complex single
-    inline float operator()(float in){
-      assert(0);
-    }
     // Complex double
-    inline vector4double operator()(vector4double in){
+    inline vector4double operator()(vector4double v){
-      assert(0);
+      return vec_mul(v, (vector4double){1., -1., 1., -1.});
     }
-    // do not define for integer input
-  };
 
+    // Complex float
+    FLOAT_WRAP_1(operator(), inline)
+  };
+  
   struct TimesMinusI{
-    //Complex single
-    inline float operator()(float in, float ret){
-      assert(0);
-    }
     //Complex double
-    inline vector4double operator()(vector4double in, vector4double ret){
+    inline vector4double operator()(vector4double v, vector4double ret){
-      assert(0);
+      return vec_xxcpnmadd(v, (vector4double){1., 1., 1., 1.},
+                               (vector4double){0., 0., 0., 0.});
     }
 
-
+    // Complex float
+    FLOAT_WRAP_2(operator(), inline)
   };
-
+  
   struct TimesI{
-    //Complex single
-    inline float operator()(float in, float ret){
-  
-    }
     //Complex double
-    inline vector4double operator()(vector4double in, vector4double ret){
+    inline vector4double operator()(vector4double v, vector4double ret){
-  
+      return vec_xxcpnmadd(v, (vector4double){-1., -1., -1., -1.},
+                              (vector4double){0., 0., 0., 0.});
     }
 
-
+    // Complex float
+    FLOAT_WRAP_2(operator(), inline)
   };
 
+  struct Permute{
+    //Complex double
+    static inline vector4double Permute0(vector4double v){ //0123 -> 2301
+      return vec_perm(v, v, vec_gpci(02301));
+    };
+    static inline vector4double Permute1(vector4double v){ //0123 -> 1032
+      return vec_perm(v, v, vec_gpci(01032));
+    };
+    static inline vector4double Permute2(vector4double v){
+      return v;
+    };
+    static inline vector4double Permute3(vector4double v){
+      return v;
+    };
 
+    // Complex float
+    FLOAT_WRAP_1(Permute0, static inline)
+    FLOAT_WRAP_1(Permute1, static inline)
+    FLOAT_WRAP_1(Permute2, static inline)
+    FLOAT_WRAP_1(Permute3, static inline)
+  };
   
+  struct Rotate{
+    static inline vector4double rotate(vector4double v, int n){
+      switch(n){
+        case 0:
+          return v;
+          break;
+        case 1:
+          return vec_perm(v, v, vec_gpci(01230));
+          break;
+        case 2:
+          return vec_perm(v, v, vec_gpci(02301));
+          break;
+        case 3:
+          return vec_perm(v, v, vec_gpci(03012));
+          break;
+        default: assert(0);
+      }
+    }
 
+    static inline vector4float rotate(vector4float v, int n){
+      vector4double vd, rd;
+      vector4float  r;
 
-  //////////////////////////////////////////////
+      vd = Vset()(v);
-  // Some Template specialization
+      rd = rotate(vd, n);
+      Vstore()(rd, r);
+
+      return r;
+    }
+  };
   
   //Complex float Reduce
   template<>
-    inline Grid::ComplexF Reduce<Grid::ComplexF, float>::operator()(float in){
+  inline Grid::ComplexF
-    assert(0);
+  Reduce<Grid::ComplexF, vector4float>::operator()(vector4float v) { //2 complex
+    vector4float v1,v2;
+    
+    v1 = Optimization::Permute::Permute0(v);
+    v1 = Optimization::Sum()(v1, v);
+    
+    return Grid::ComplexF(v1.v0, v1.v1);
   }
   //Real float Reduce
   template<>
-    inline Grid::RealF Reduce<Grid::RealF, float>::operator()(float in){
+  inline Grid::RealF
-    assert(0);
+  Reduce<Grid::RealF, vector4float>::operator()(vector4float v){ //4 floats
+    vector4float v1,v2;
+    
+    v1 = Optimization::Permute::Permute0(v);
+    v1 = Optimization::Sum()(v1, v);
+    v2 = Optimization::Permute::Permute1(v1);
+    v1 = Optimization::Sum()(v1, v2);
+    
+    return v1.v0;
   }
   
   
   //Complex double Reduce
   template<>
-    inline Grid::ComplexD Reduce<Grid::ComplexD, vector4double>::operator()(vector4double in){
+  inline Grid::ComplexD
-    assert(0);
+  Reduce<Grid::ComplexD, vector4double>::operator()(vector4double v){ //2 complex
+    vector4double v1;
+    
+    v1 = Optimization::Permute::Permute0(v);
+    v1 = vec_add(v1, v);
+    
+    return Grid::ComplexD(vec_extract(v1, 0), vec_extract(v1, 1));
   }
   
   //Real double Reduce
   template<>
-    inline Grid::RealD Reduce<Grid::RealD, vector4double>::operator()(vector4double in){
+  inline Grid::RealD
-    assert(0);
+  Reduce<Grid::RealD, vector4double>::operator()(vector4double v){ //4 doubles
-  }
+    vector4double v1,v2;
+    
+    v1 = Optimization::Permute::Permute0(v);
+    v1 = vec_add(v1, v);
+    v2 = Optimization::Permute::Permute1(v1);
+    v1 = vec_add(v1, v2);
 
+    return vec_extract(v1, 0);
+  }
+  
   //Integer Reduce
   template<>
-    inline Integer Reduce<Integer, floati>::operator()(float in){
+  inline Integer Reduce<Integer, int>::operator()(int in){
+    // FIXME unimplemented
+    printf("Reduce : Missing integer implementation -> FIX\n");
     assert(0);
   }
-  
-  
 }
 
-//////////////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////////
-// Here assign types 
+// Here assign types
-namespace Grid {
+typedef Optimization::vector4float SIMD_Ftype;  // Single precision type
-  typedef float SIMD_Ftype  __attribute__ ((vector_size (16)));         // Single precision type
+typedef vector4double              SIMD_Dtype; // Double precision type
-  typedef vector4double SIMD_Dtype; // Double precision type
+typedef int                        SIMD_Itype; // Integer type
-  typedef int SIMD_Itype;           // Integer type
 
-  inline void v_prefetch0(int size, const char *ptr){};
+// prefetch utilities
-
+inline void v_prefetch0(int size, const char *ptr){};
-  // Function name aliases
+inline void prefetch_HINT_T0(const char *ptr){};
-  typedef Optimization::Vsplat   VsplatSIMD;
-  typedef Optimization::Vstore   VstoreSIMD;
-  typedef Optimization::Vset     VsetSIMD;
-  typedef Optimization::Vstream  VstreamSIMD;
-  template <typename S, typename T> using ReduceSIMD = Optimization::Reduce<S,T>;
 
 
-  // Arithmetic operations
+// Function name aliases
-  typedef Optimization::Sum         SumSIMD;
+typedef Optimization::Vsplat   VsplatSIMD;
-  typedef Optimization::Sub         SubSIMD;
+typedef Optimization::Vstore   VstoreSIMD;
-  typedef Optimization::Mult        MultSIMD;
+typedef Optimization::Vset     VsetSIMD;
-  typedef Optimization::MultComplex MultComplexSIMD;
+typedef Optimization::Vstream  VstreamSIMD;
-  typedef Optimization::Conj        ConjSIMD;
+template <typename S, typename T> using ReduceSIMD = Optimization::Reduce<S,T>;
-  typedef Optimization::TimesMinusI TimesMinusISIMD;
-  typedef Optimization::TimesI      TimesISIMD;
 
+// Arithmetic operations
+typedef Optimization::Sum         SumSIMD;
+typedef Optimization::Sub         SubSIMD;
+typedef Optimization::Mult        MultSIMD;
+typedef Optimization::MultComplex MultComplexSIMD;
+typedef Optimization::Conj        ConjSIMD;
+typedef Optimization::TimesMinusI TimesMinusISIMD;
+typedef Optimization::TimesI      TimesISIMD;
+  
 }

lib/simd/Intel512common.h

@@ -138,9 +138,14 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
 #define ZLOADf(OFF,PTR,ri,ir)  VLOADf(OFF,PTR,ir)  VSHUFf(ir,ri)
 #define ZLOADd(OFF,PTR,ri,ir)  VLOADd(OFF,PTR,ir)  VSHUFd(ir,ri)
 
-
+#define STREAM_STORE
+#ifdef STREAM_STORE
 #define VSTOREf(OFF,PTR,SRC)   "vmovntps " #SRC "," #OFF "*64(" #PTR ")"  ";\n"
 #define VSTOREd(OFF,PTR,SRC)   "vmovntpd " #SRC "," #OFF "*64(" #PTR ")"  ";\n"
+#else
+#define VSTOREf(OFF,PTR,SRC)   "vmovaps " #SRC "," #OFF "*64(" #PTR ")"  ";\n"
+#define VSTOREd(OFF,PTR,SRC)   "vmovapd " #SRC "," #OFF "*64(" #PTR ")"  ";\n"
+#endif
 
 // Swaps Re/Im ; could unify this with IMCI
 #define VSHUFd(A,DEST)         "vpshufd  $0x4e," #A "," #DEST  ";\n"    

lib/tensors/Tensor_exp.h

@@ -56,7 +56,7 @@ namespace Grid {
 	temp = unit + temp*arg;
       }
       
-      return ProjectOnGroup(temp);//maybe not strictly necessary
+      return temp;
       
     }
 

scripts/filelist

@@ -13,21 +13,22 @@ echo CCFILES=$CCFILES >> Make.inc
 
 # tests Make.inc
 cd $home/tests
-dirs=`find . -type d `
+dirs=`find . -type d -not -path '*/\.*'`
 for subdir in $dirs; do
-	cd $home/tests/$subdir
+    cd $home/tests/$subdir
-	TESTS=`ls T*.cc`
+    pwd
-	TESTLIST=`echo ${TESTS} | sed s/.cc//g `
+    TESTS=`ls T*.cc`
-	PREF=`[ $subdir = '.' ] && echo noinst || echo EXTRA`
+    TESTLIST=`echo ${TESTS} | sed s/.cc//g `
-	echo "tests: ${TESTLIST}" > Make.inc
+    PREF=`[ $subdir = '.' ] && echo noinst || echo EXTRA`
-	echo ${PREF}_PROGRAMS = ${TESTLIST} >> Make.inc
+    echo "tests: ${TESTLIST}" > Make.inc
+    echo ${PREF}_PROGRAMS = ${TESTLIST} >> Make.inc
+    echo >> Make.inc
+    for f in $TESTS; do
+	BNAME=`basename $f .cc`
+	echo ${BNAME}_SOURCES=$f  >> Make.inc
+	echo ${BNAME}_LDADD=-lGrid>> Make.inc
 	echo >> Make.inc
-	for f in $TESTS; do
+    done
-		BNAME=`basename $f .cc`
-		echo ${BNAME}_SOURCES=$f  >> Make.inc
-		echo ${BNAME}_LDADD=-lGrid>> Make.inc
-		echo >> Make.inc
-	done
 done
 
 # benchmarks Make.inc
@@ -38,10 +39,10 @@ TESTLIST=`echo ${TESTS} | sed s/.cc//g `
 echo bin_PROGRAMS = ${TESTLIST} > Make.inc
 echo >> Make.inc
 for f in $TESTS; do
-	BNAME=`basename $f .cc`
+    BNAME=`basename $f .cc`
-	echo ${BNAME}_SOURCES=$f  >> Make.inc
+    echo ${BNAME}_SOURCES=$f  >> Make.inc
-	echo ${BNAME}_LDADD=-lGrid>> Make.inc
+    echo ${BNAME}_LDADD=-lGrid>> Make.inc
-	echo >> Make.inc
+    echo >> Make.inc
 done
 
 cd ..

scripts/reconfigure_script

@@ -0,0 +1,4 @@
+aclocal -I m4
+autoheader -f
+automake -f --add-missing
+autoconf -f

tests/Test_simd.cc

@@ -157,10 +157,9 @@ void Tester(const functor &func)
   std::cout << GridLogMessage << " " << func.name() << std::endl;
 
   std::cout << GridLogDebug << v_input1 << std::endl;
+  std::cout << GridLogDebug << v_input2 << std::endl;
   std::cout << GridLogDebug << v_result << std::endl;
 
-
-
   int ok=0;
   for(int i=0;i<Nsimd;i++){
     if ( abs(reference[i]-result[i])>1.0e-7){

tests/core/Test_lie_generators.cc

@@ -1,76 +1,532 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./tests/Test_lie_generators.cc
+Source file: ./tests/Test_lie_generators.cc
 
-    Copyright (C) 2015
+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>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution
-    *************************************************************************************/
+directory
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #include <Grid/Grid.h>
 
+#include <Grid/qcd/utils/CovariantCshift.h>
+
+#include <Grid/qcd/utils/SUn.h>
+#include <Grid/qcd/utils/SUnAdjoint.h>
+#include <Grid/qcd/utils/SUnTwoIndex.h>
+
+#include <Grid/qcd/representations/adjoint.h>
+#include <Grid/qcd/representations/two_index.h>
+#include <Grid/qcd/utils/WilsonLoops.h>
 
 using namespace std;
 using namespace Grid;
 using namespace Grid::QCD;
 
+int main(int argc, char** argv) {
+  Grid_init(&argc, &argv);
 
-int main (int argc, char ** argv)
+  std::vector<int> latt({4, 4, 4, 8});
-{
+  GridCartesian* grid = SpaceTimeGrid::makeFourDimGrid(
-  Grid_init(&argc,&argv);
+      latt, GridDefaultSimd(Nd, vComplex::Nsimd()), GridDefaultMpi());
 
-  std::vector<int> latt({4,4,4,8});
+  GridRedBlackCartesian* rbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(grid);
-  GridCartesian * grid = SpaceTimeGrid::makeFourDimGrid(latt, 
-							GridDefaultSimd(Nd,vComplex::Nsimd()),
-							GridDefaultMpi());
-  
-  GridRedBlackCartesian * rbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(grid);
 
-  std::cout<<GridLogMessage<<"*********************************************"<<std::endl;
+  std::cout << GridLogMessage << "*********************************************"
-  std::cout<<GridLogMessage<<"* Generators for SU(2)"<<std::endl;
+            << std::endl;
-  std::cout<<GridLogMessage<<"*********************************************"<<std::endl;
+  std::cout << GridLogMessage << "* Generators for SU(2)" << std::endl;
+  std::cout << GridLogMessage << "*********************************************"
+            << std::endl;
   SU2::printGenerators();
+  std::cout << "Dimension of adjoint representation: "<< SU2Adjoint::Dimension << std::endl;
+  SU2Adjoint::printGenerators();
   SU2::testGenerators();
+  SU2Adjoint::testGenerators();
 
-  std::cout<<GridLogMessage<<"*********************************************"<<std::endl;
+  std::cout << GridLogMessage << "*********************************************"
-  std::cout<<GridLogMessage<<"* Generators for SU(3)"<<std::endl;
+            << std::endl;
-  std::cout<<GridLogMessage<<"*********************************************"<<std::endl;
+  std::cout << GridLogMessage << "* Generators for SU(3)" << std::endl;
+  std::cout << GridLogMessage << "*********************************************"
+            << std::endl;
   SU3::printGenerators();
+  std::cout << "Dimension of adjoint representation: "<< SU3Adjoint::Dimension << std::endl;
+  SU3Adjoint::printGenerators();
   SU3::testGenerators();
+  SU3Adjoint::testGenerators();
 
-  //  std::cout<<GridLogMessage<<"*********************************************"<<std::endl;
+  std::cout<<GridLogMessage<<"*********************************************"<<std::endl;
-  //  std::cout<<GridLogMessage<<"* Generators for SU(4)"<<std::endl;
+  std::cout<<GridLogMessage<<"* Generators for SU(4)"<<std::endl;
-  //  std::cout<<GridLogMessage<<"*********************************************"<<std::endl;
+  std::cout<<GridLogMessage<<"*********************************************"<<std::endl;
-  //  SU4::printGenerators();
+  SU4::printGenerators();
-  //  SU4::testGenerators();
+  std::cout << "Dimension of adjoint representation: "<< SU4Adjoint::Dimension << std::endl;
+  SU4Adjoint::printGenerators();
+  SU4::testGenerators();
+  SU4Adjoint::testGenerators();
 
-  //  std::cout<<GridLogMessage<<"*********************************************"<<std::endl;
+  // Projectors 
-  //  std::cout<<GridLogMessage<<"* Generators for SU(5)"<<std::endl;
+  GridParallelRNG gridRNG(grid);
-  //  std::cout<<GridLogMessage<<"*********************************************"<<std::endl;
+  gridRNG.SeedRandomDevice();
-  //  SU5::printGenerators();
+  SU3Adjoint::LatticeAdjMatrix Gauss(grid);
-  //  SU5::testGenerators();
+  SU3::LatticeAlgebraVector ha(grid);
+  SU3::LatticeAlgebraVector hb(grid);
+  random(gridRNG,Gauss);
+
+  std::cout << GridLogMessage << "Start projectOnAlgebra" << std::endl;
+  SU3Adjoint::projectOnAlgebra(ha, Gauss);
+  std::cout << GridLogMessage << "end projectOnAlgebra" << std::endl;
+  std::cout << GridLogMessage << "Start projector" << std::endl;
+  SU3Adjoint::projector(hb, Gauss);
+  std::cout << GridLogMessage << "end projector" << std::endl;
+
+  std::cout << GridLogMessage << "ReStart projector" << std::endl;
+  SU3Adjoint::projector(hb, Gauss);
+  std::cout << GridLogMessage << "end projector" << std::endl;
+  SU3::LatticeAlgebraVector diff = ha -hb;
+  std::cout << GridLogMessage << "Difference: " << norm2(diff) << std::endl;
 
 
-  Grid_finalize();
+  // Testing HMC representation classes
+  AdjointRep<Nc> AdjRep(grid);
+
+  // AdjointRepresentation has the predefined number of colours Nc
+  Representations<FundamentalRepresentation, AdjointRepresentation, TwoIndexSymmetricRepresentation> RepresentationTypes(grid);  
+
+  
+  LatticeGaugeField U(grid), V(grid);
+  SU<Nc>::HotConfiguration<LatticeGaugeField>(gridRNG, U);
+  SU<Nc>::HotConfiguration<LatticeGaugeField>(gridRNG, V);
+
+  // Adjoint representation
+  // Test group structure
+  // (U_f * V_f)_r = U_r * V_r
+  LatticeGaugeField UV(grid);
+  UV = zero;
+  for (int mu = 0; mu < Nd; mu++) {
+    SU<Nc>::LatticeMatrix Umu = peekLorentz(U,mu);
+    SU<Nc>::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++) {
+    typename AdjointRep<Nc>::LatticeMatrix Urmu = peekLorentz(Ur,mu);
+    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
+  SU<Nc>::LatticeAlgebraVector h_adj(grid);
+  typename AdjointRep<Nc>::LatticeMatrix Ar(grid);
+  random(gridRNG,h_adj);
+  h_adj = real(h_adj);
+  SU_Adjoint<Nc>::AdjointLieAlgebraMatrix(h_adj,Ar);
+
+  // Re-extract h_adj
+  SU<Nc>::LatticeAlgebraVector h_adj2(grid);
+  SU_Adjoint<Nc>::projectOnAlgebra(h_adj2, Ar);
+  SU<Nc>::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;
+
+  Ucheck = Uadj * adj(Uadj) - uno;
+  std::cout << GridLogMessage << "orthogonality check 1: " << norm2(Ucheck)
+            << std::endl;
+  Ucheck = adj(Uadj) * Uadj - uno;
+  std::cout << GridLogMessage << "orthogonality check 2: " << norm2(Ucheck)
+            << std::endl;
+      
+
+  // Construct the fundamental matrix in the group
+  SU<Nc>::LatticeMatrix Af(grid);
+  SU<Nc>::FundamentalLieAlgebraMatrix(h_adj,Af);
+  SU<Nc>::LatticeMatrix Ufund(grid);
+  Ufund  = expMat(Af, 1.0, 16);
+  // Check unitarity
+  SU<Nc>::LatticeMatrix uno_f(grid);
+  uno_f = 1.0;
+  SU<Nc>::LatticeMatrix UnitCheck(grid);
+  UnitCheck = Ufund * adj(Ufund) - uno_f;
+  std::cout << GridLogMessage << "unitarity check 1: " << norm2(UnitCheck)
+            << std::endl;
+  UnitCheck = adj(Ufund) * Ufund - uno_f;
+  std::cout << GridLogMessage << "unitarity check 2: " << norm2(UnitCheck)
+            << 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::cout << GridLogMessage << "*********************************************"
+            << std::endl;
+
+
+  
+  std::cout << GridLogMessage << "* eS^{ij} base for SU(2)" << std::endl;
+  std::cout << GridLogMessage << "*********************************************"
+            << std::endl;
+  std::cout << GridLogMessage << "Dimension of Two Index Symmetric representation: "<< SU2TwoIndexSymm::Dimension << std::endl;
+  SU2TwoIndexSymm::printBase();
+      std::cout << GridLogMessage << "*********************************************"
+            << 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;
+  SU2TwoIndexSymm::testGenerators();
+  std::cout << GridLogMessage << "*********************************************"
+            << std::endl;
+
+
+  
+  std::cout << GridLogMessage << "*********************************************"
+            << std::endl;
+  std::cout << GridLogMessage << "* eAS^{ij} base for SU(2)" << std::endl;
+  
+  std::cout << GridLogMessage << "*********************************************"
+            << std::endl;
+  std::cout << GridLogMessage << "Dimension of Two Index anti-Symmetric representation: "<< SU2TwoIndexAntiSymm::Dimension << std::endl;
+  SU2TwoIndexAntiSymm::printBase();
+      std::cout << GridLogMessage << "*********************************************"
+            << 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::cout << GridLogMessage << "Test for the Two Index Symmetric projectors"
+      << std::endl;
+  // Projectors 
+  SU3TwoIndexSymm::LatticeTwoIndexMatrix Gauss2(grid);
+  random(gridRNG,Gauss2);
+  
+  std::cout << GridLogMessage << "Start projectOnAlgebra" << std::endl;
+  SU3TwoIndexSymm::projectOnAlgebra(ha, Gauss2);
+  std::cout << GridLogMessage << "end projectOnAlgebra" << std::endl;
+  std::cout << GridLogMessage << "Start projector" << std::endl;
+  SU3TwoIndexSymm::projector(hb, Gauss2);
+  std::cout << GridLogMessage << "end projector" << std::endl;
+  
+  std::cout << GridLogMessage << "ReStart projector" << std::endl;
+  SU3TwoIndexSymm::projector(hb, Gauss2);
+  std::cout << GridLogMessage << "end projector" << std::endl;
+  SU3::LatticeAlgebraVector diff2 = ha - hb;
+  std::cout << GridLogMessage << "Difference: " << norm2(diff) << std::endl;
+  std::cout << GridLogMessage << "*********************************************"
+      << std::endl;
+
+  
+    std::cout << GridLogMessage << "*********************************************"
+      << std::endl;
+  std::cout << GridLogMessage << "Test for the Two index anti-Symmetric projectors"
+      << std::endl;
+  // Projectors
+  SU3TwoIndexAntiSymm::LatticeTwoIndexMatrix Gauss2a(grid);
+  random(gridRNG,Gauss2a);
+  
+  std::cout << GridLogMessage << "Start projectOnAlgebra" << std::endl;
+  SU3TwoIndexAntiSymm::projectOnAlgebra(ha, Gauss2a);
+  std::cout << GridLogMessage << "end projectOnAlgebra" << std::endl;
+  std::cout << GridLogMessage << "Start projector" << std::endl;
+  SU3TwoIndexAntiSymm::projector(hb, Gauss2a);
+  std::cout << GridLogMessage << "end projector" << std::endl;
+  
+  std::cout << GridLogMessage << "ReStart projector" << std::endl;
+  SU3TwoIndexAntiSymm::projector(hb, Gauss2a);
+  std::cout << GridLogMessage << "end projector" << std::endl;
+  SU3::LatticeAlgebraVector diff2a = ha - hb;
+  std::cout << GridLogMessage << "Difference: " << norm2(diff2a) << std::endl;
+  std::cout << GridLogMessage << "*********************************************"
+      << std::endl;
+    
+  
+  std::cout << GridLogMessage << "Two index Symmetric: Checking Group Structure"
+      << std::endl;
+  // Testing HMC representation classes
+  TwoIndexRep< Nc, Symmetric > TIndexRep(grid);
+
+  // Test group structure
+  // (U_f * V_f)_r = U_r * V_r
+  LatticeGaugeField U2(grid), V2(grid);
+  SU<Nc>::HotConfiguration<LatticeGaugeField>(gridRNG, U2);
+  SU<Nc>::HotConfiguration<LatticeGaugeField>(gridRNG, V2);
+  
+  LatticeGaugeField UV2(grid);
+  UV2 = zero;
+  for (int mu = 0; mu < Nd; mu++) {
+    SU<Nc>::LatticeMatrix Umu2 = peekLorentz(U2,mu);
+    SU<Nc>::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++) {
+    typename TwoIndexRep< Nc, Symmetric >::LatticeMatrix Urmu2 = peekLorentz(Ur2,mu);
+    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
+  SU<Nc>::LatticeAlgebraVector h_sym(grid);
+  typename TwoIndexRep< Nc, Symmetric>::LatticeMatrix Ar_sym(grid);
+  random(gridRNG,h_sym);
+  h_sym = real(h_sym);
+  SU_TwoIndex<Nc,Symmetric>::TwoIndexLieAlgebraMatrix(h_sym,Ar_sym);
+  
+  // Re-extract h_sym
+  SU<Nc>::LatticeAlgebraVector h_sym2(grid);
+  SU_TwoIndex< Nc, Symmetric>::projectOnAlgebra(h_sym2, Ar_sym);
+  SU<Nc>::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;
+  
+  Ucheck2iS = U2iS * adj(U2iS) - uno2iS;
+  std::cout << GridLogMessage << "orthogonality check 1: " << norm2(Ucheck2iS)
+      << std::endl;
+  Ucheck2iS = adj(U2iS) * U2iS - uno2iS;
+  std::cout << GridLogMessage << "orthogonality check 2: " << norm2(Ucheck2iS)
+      << std::endl;
+  
+  
+  
+  // Construct the fundamental matrix in the group
+  SU<Nc>::LatticeMatrix Af_sym(grid);
+  SU<Nc>::FundamentalLieAlgebraMatrix(h_sym,Af_sym);
+  SU<Nc>::LatticeMatrix Ufund2(grid);
+  Ufund2  = expMat(Af_sym, 1.0, 16);
+  SU<Nc>::LatticeMatrix UnitCheck2(grid);
+  UnitCheck2 = Ufund2 * adj(Ufund2) - uno_f;
+  std::cout << GridLogMessage << "unitarity check 1: " << norm2(UnitCheck2)
+      << std::endl;
+  UnitCheck2 = adj(Ufund2) * Ufund2 - uno_f;
+  std::cout << GridLogMessage << "unitarity check 2: " << norm2(UnitCheck2)
+      << 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::endl;
+    
+  
+  std::cout << GridLogMessage << "Two Index anti-Symmetric: Check Group Structure"
+      << std::endl;
+  // Testing HMC representation classes
+  TwoIndexRep< Nc, AntiSymmetric > TIndexRepA(grid);
+
+
+  // Test group structure
+  // (U_f * V_f)_r = U_r * V_r
+  LatticeGaugeField U2A(grid), V2A(grid);
+  SU<Nc>::HotConfiguration<LatticeGaugeField>(gridRNG, U2A);
+  SU<Nc>::HotConfiguration<LatticeGaugeField>(gridRNG, V2A);
+  
+  LatticeGaugeField UV2A(grid);
+  UV2A = zero;
+  for (int mu = 0; mu < Nd; mu++) {
+    SU<Nc>::LatticeMatrix Umu2A = peekLorentz(U2,mu);
+    SU<Nc>::LatticeMatrix Vmu2A = peekLorentz(V2,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
+  SU<Nc>::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
+  SU<Nc>::LatticeAlgebraVector h_Asym2(grid);
+  SU_TwoIndex< Nc, AntiSymmetric>::projectOnAlgebra(h_Asym2, Ar_Asym);
+  SU<Nc>::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
+  SU<Nc>::LatticeMatrix Af_Asym(grid);
+  SU<Nc>::FundamentalLieAlgebraMatrix(h_Asym,Af_Asym);
+  SU<Nc>::LatticeMatrix Ufund2A(grid);
+  Ufund2A  = expMat(Af_Asym, 1.0, 16);
+  SU<Nc>::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;
 }
 
 
+
+
+
+
+
+
+
+  
+  Grid_finalize();
+}

tests/debug/Test_zmm.cc

@@ -1,30 +1,30 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./tests/Test_zmm.cc
+Source file: ./tests/Test_zmm.cc
 
-    Copyright (C) 2015
+Copyright (C) 2015
 
 Author: paboyle <paboyle@ph.ed.ac.uk>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
+*************************************************************************************/
-    /*  END LEGAL */
+/*  END LEGAL */
 #include <Grid/Grid.h>
 #include <Grid/PerfCount.h>
 

tests/forces/Test_contfrac_force.cc

@@ -96,7 +96,7 @@ int main (int argc, char ** argv)
 
   for(int mu=0;mu<Nd;mu++){
 
-    SU3::GaussianLieAlgebraMatrix(RNG4, mommu); // Traceless antihermitian momentum; gaussian in lie alg
+    SU3::GaussianFundamentalLieAlgebraMatrix(RNG4, mommu); // Traceless antihermitian momentum; gaussian in lie alg
 
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 

tests/forces/Test_dwf_force.cc

@@ -96,7 +96,7 @@ int main (int argc, char ** argv)
 
   for(int mu=0;mu<Nd;mu++){
 
-    SU3::GaussianLieAlgebraMatrix(RNG4, mommu); // Traceless antihermitian momentum; gaussian in lie alg
+    SU3::GaussianFundamentalLieAlgebraMatrix(RNG4, mommu); // Traceless antihermitian momentum; gaussian in lie alg
 
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 

tests/forces/Test_dwf_gpforce.cc

@@ -113,7 +113,7 @@ int main (int argc, char ** argv)
 
   for(int mu=0;mu<Nd;mu++){
 
-    SU3::GaussianLieAlgebraMatrix(RNG4, mommu); // Traceless antihermitian momentum; gaussian in lie alg
+    SU3::GaussianFundamentalLieAlgebraMatrix(RNG4, mommu); // Traceless antihermitian momentum; gaussian in lie alg
 
     Hmom -= real(sum(trace(mommu*mommu)));
 

tests/forces/Test_gp_rect_force.cc

@@ -82,7 +82,7 @@ int main (int argc, char ** argv)
 
   for(int mu=0;mu<Nd;mu++){
 
-    SU3::GaussianLieAlgebraMatrix(pRNG, mommu); // Traceless antihermitian momentum; gaussian in lie alg
+    SU3::GaussianFundamentalLieAlgebraMatrix(pRNG, mommu); // Traceless antihermitian momentum; gaussian in lie alg
 
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 

tests/forces/Test_gpdwf_force.cc

@@ -100,7 +100,7 @@ int main (int argc, char ** argv)
 
   for(int mu=0;mu<Nd;mu++){
 
-    SU3::GaussianLieAlgebraMatrix(RNG4, mommu); // Traceless antihermitian momentum; gaussian in lie alg
+    SU3::GaussianFundamentalLieAlgebraMatrix(RNG4, mommu); // Traceless antihermitian momentum; gaussian in lie alg
 
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 
@@ -169,7 +169,7 @@ int main (int argc, char ** argv)
   //
   //	Pmu = zero;
   //	for(int mu=0;mu<Nd;mu++){
-  //	  SU<Ncol>::GaussianLieAlgebraMatrix(pRNG, Pmu);
+  //	  SU<Ncol>::GaussianFundamentalLieAlgebraMatrix(pRNG, Pmu);
   //	  PokeIndex<LorentzIndex>(P, Pmu, mu);
   //	}
   //

tests/forces/Test_gpwilson_force.cc

@@ -100,7 +100,7 @@ int main (int argc, char ** argv)
 
   for(int mu=0;mu<Nd;mu++){
 
-    SU3::GaussianLieAlgebraMatrix(RNG4, mommu); // Traceless antihermitian momentum; gaussian in lie alg
+    SU3::GaussianFundamentalLieAlgebraMatrix(RNG4, mommu); // Traceless antihermitian momentum; gaussian in lie alg
 
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 

tests/forces/Test_partfrac_force.cc

@@ -96,7 +96,7 @@ int main (int argc, char ** argv)
 
   for(int mu=0;mu<Nd;mu++){
 
-    SU3::GaussianLieAlgebraMatrix(RNG4, mommu); // Traceless antihermitian momentum; gaussian in lie alg
+    SU3::GaussianFundamentalLieAlgebraMatrix(RNG4, mommu); // Traceless antihermitian momentum; gaussian in lie alg
 
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 

tests/forces/Test_rect_force.cc

@@ -81,7 +81,7 @@ int main (int argc, char ** argv)
 
   for(int mu=0;mu<Nd;mu++){
 
-    SU3::GaussianLieAlgebraMatrix(pRNG, mommu); // Traceless antihermitian momentum; gaussian in lie alg
+    SU3::GaussianFundamentalLieAlgebraMatrix(pRNG, mommu); // Traceless antihermitian momentum; gaussian in lie alg
 
     PokeIndex<LorentzIndex>(mom,mommu,mu);
 

tests/forces/Test_wilson_force.cc

@@ -58,8 +58,8 @@ int main (int argc, char ** argv)
 
   LatticeGaugeField U(&Grid);
 
-  SU3::HotConfiguration(pRNG,U);
+  //SU2::HotConfiguration(pRNG,U);
-  //  SU3::ColdConfiguration(pRNG,U);
+  SU3::ColdConfiguration(pRNG,U);
   
   ////////////////////////////////////
   // Unmodified matrix element
@@ -76,6 +76,8 @@ int main (int argc, char ** argv)
 
   Dw.MDeriv(tmp , Mphi,  phi,DaggerNo );  UdSdU=tmp;
   Dw.MDeriv(tmp , phi,  Mphi,DaggerYes ); UdSdU=(UdSdU+tmp);
+  // Take the trace
+  UdSdU = Ta(UdSdU);
   
   LatticeFermion Ftmp      (&Grid);
 
@@ -93,7 +95,7 @@ int main (int argc, char ** argv)
 
   for(int mu=0;mu<Nd;mu++){
 
-    SU3::GaussianLieAlgebraMatrix(pRNG, mommu); // Traceless antihermitian momentum; gaussian in lie alg
+    SU3::GaussianFundamentalLieAlgebraMatrix(pRNG, mommu); // Traceless antihermitian momentum; gaussian in lie alg
 
     Hmom -= real(sum(trace(mommu*mommu)));
 
@@ -140,11 +142,13 @@ int main (int argc, char ** argv)
   LatticeComplex dS(&Grid); dS = zero;
   LatticeComplex dSmom(&Grid); dSmom = zero;
   LatticeComplex dSmom2(&Grid); dSmom2 = zero;
+  
   for(int mu=0;mu<Nd;mu++){
     mommu   = PeekIndex<LorentzIndex>(UdSdU,mu);
     mommu=Ta(mommu)*2.0;
     PokeIndex<LorentzIndex>(UdSdU,mommu,mu);
   }
+  
 
   for(int mu=0;mu<Nd;mu++){
     mommu   = PeekIndex<LorentzIndex>(mom,mu);
@@ -168,7 +172,7 @@ int main (int argc, char ** argv)
     dSmom2 = dSmom2 - trace(forcemu*forcemu) *(0.25* dt*dt);
 
     // Update mom action density
-    mommu = mommu + forcemu*(dt*0.5);
+    mommu = mommu + forcemu*(dt * 0.5);
 
     Hmomprime -= real(sum(trace(mommu*mommu)));
 

tests/forces/Test_wilson_force_phiMdagMphi.cc

@@ -103,7 +103,7 @@ int main (int argc, char ** argv)
 
   for(int mu=0;mu<Nd;mu++){
 
-    SU3::GaussianLieAlgebraMatrix(pRNG, mommu); // Traceless antihermitian momentum; gaussian in lie alg
+    SU3::GaussianFundamentalLieAlgebraMatrix(pRNG, mommu); // Traceless antihermitian momentum; gaussian in lie alg
 
     //    Dw.DoubleStore(Dw.Umu,Uprime); // update U _and_ Udag
     Dw.DhopDirDisp(phi,Ftmp,mu,mu+4,DaggerYes); 

tests/forces/Test_wilson_force_phiMphi.cc

@@ -86,7 +86,7 @@ int main (int argc, char ** argv)
   LatticeColourMatrix Umu_save(&Grid);
   LatticeColourMatrix dU (&Grid);
   LatticeColourMatrix mom(&Grid); 
-  SU3::GaussianLieAlgebraMatrix(pRNG, mom); // Traceless antihermitian momentum; gaussian in lie alg
+  SU3::GaussianFundamentalLieAlgebraMatrix(pRNG, mom); // Traceless antihermitian momentum; gaussian in lie alg
 
 
   // check mom is as i expect

tests/hmc/Make.inc

@@ -0,0 +1,63 @@
+tests: Test_hmc_EODWFRatio Test_hmc_EODWFRatio_Gparity Test_hmc_EOWilsonFermionGauge Test_hmc_EOWilsonRatio Test_hmc_GparityIwasakiGauge Test_hmc_GparityWilsonGauge Test_hmc_IwasakiGauge Test_hmc_RectGauge Test_hmc_WilsonAdjointFermionGauge Test_hmc_WilsonFermionGauge Test_hmc_WilsonGauge Test_hmc_WilsonMixedRepresentationsFermionGauge Test_hmc_WilsonRatio Test_hmc_WilsonTwoIndexSymmetricFermionGauge Test_multishift_sqrt Test_remez Test_rhmc_EOWilson1p1 Test_rhmc_EOWilsonRatio Test_rhmc_Wilson1p1 Test_rhmc_WilsonRatio
+EXTRA_PROGRAMS = Test_hmc_EODWFRatio Test_hmc_EODWFRatio_Gparity Test_hmc_EOWilsonFermionGauge Test_hmc_EOWilsonRatio Test_hmc_GparityIwasakiGauge Test_hmc_GparityWilsonGauge Test_hmc_IwasakiGauge Test_hmc_RectGauge Test_hmc_WilsonAdjointFermionGauge Test_hmc_WilsonFermionGauge Test_hmc_WilsonGauge Test_hmc_WilsonMixedRepresentationsFermionGauge Test_hmc_WilsonRatio Test_hmc_WilsonTwoIndexSymmetricFermionGauge Test_multishift_sqrt Test_remez Test_rhmc_EOWilson1p1 Test_rhmc_EOWilsonRatio Test_rhmc_Wilson1p1 Test_rhmc_WilsonRatio
+
+Test_hmc_EODWFRatio_SOURCES=Test_hmc_EODWFRatio.cc
+Test_hmc_EODWFRatio_LDADD=-lGrid
+
+Test_hmc_EODWFRatio_Gparity_SOURCES=Test_hmc_EODWFRatio_Gparity.cc
+Test_hmc_EODWFRatio_Gparity_LDADD=-lGrid
+
+Test_hmc_EOWilsonFermionGauge_SOURCES=Test_hmc_EOWilsonFermionGauge.cc
+Test_hmc_EOWilsonFermionGauge_LDADD=-lGrid
+
+Test_hmc_EOWilsonRatio_SOURCES=Test_hmc_EOWilsonRatio.cc
+Test_hmc_EOWilsonRatio_LDADD=-lGrid
+
+Test_hmc_GparityIwasakiGauge_SOURCES=Test_hmc_GparityIwasakiGauge.cc
+Test_hmc_GparityIwasakiGauge_LDADD=-lGrid
+
+Test_hmc_GparityWilsonGauge_SOURCES=Test_hmc_GparityWilsonGauge.cc
+Test_hmc_GparityWilsonGauge_LDADD=-lGrid
+
+Test_hmc_IwasakiGauge_SOURCES=Test_hmc_IwasakiGauge.cc
+Test_hmc_IwasakiGauge_LDADD=-lGrid
+
+Test_hmc_RectGauge_SOURCES=Test_hmc_RectGauge.cc
+Test_hmc_RectGauge_LDADD=-lGrid
+
+Test_hmc_WilsonAdjointFermionGauge_SOURCES=Test_hmc_WilsonAdjointFermionGauge.cc
+Test_hmc_WilsonAdjointFermionGauge_LDADD=-lGrid
+
+Test_hmc_WilsonFermionGauge_SOURCES=Test_hmc_WilsonFermionGauge.cc
+Test_hmc_WilsonFermionGauge_LDADD=-lGrid
+
+Test_hmc_WilsonGauge_SOURCES=Test_hmc_WilsonGauge.cc
+Test_hmc_WilsonGauge_LDADD=-lGrid
+
+Test_hmc_WilsonMixedRepresentationsFermionGauge_SOURCES=Test_hmc_WilsonMixedRepresentationsFermionGauge.cc
+Test_hmc_WilsonMixedRepresentationsFermionGauge_LDADD=-lGrid
+
+Test_hmc_WilsonRatio_SOURCES=Test_hmc_WilsonRatio.cc
+Test_hmc_WilsonRatio_LDADD=-lGrid
+
+Test_hmc_WilsonTwoIndexSymmetricFermionGauge_SOURCES=Test_hmc_WilsonTwoIndexSymmetricFermionGauge.cc
+Test_hmc_WilsonTwoIndexSymmetricFermionGauge_LDADD=-lGrid
+
+Test_multishift_sqrt_SOURCES=Test_multishift_sqrt.cc
+Test_multishift_sqrt_LDADD=-lGrid
+
+Test_remez_SOURCES=Test_remez.cc
+Test_remez_LDADD=-lGrid
+
+Test_rhmc_EOWilson1p1_SOURCES=Test_rhmc_EOWilson1p1.cc
+Test_rhmc_EOWilson1p1_LDADD=-lGrid
+
+Test_rhmc_EOWilsonRatio_SOURCES=Test_rhmc_EOWilsonRatio.cc
+Test_rhmc_EOWilsonRatio_LDADD=-lGrid
+
+Test_rhmc_Wilson1p1_SOURCES=Test_rhmc_Wilson1p1.cc
+Test_rhmc_Wilson1p1_LDADD=-lGrid
+
+Test_rhmc_WilsonRatio_SOURCES=Test_rhmc_WilsonRatio.cc
+Test_rhmc_WilsonRatio_LDADD=-lGrid
+

tests/hmc/Test_hmc_EODWFRatio.cc

@@ -75,7 +75,15 @@ public:
     Level1.push_back(&Waction);
     TheAction.push_back(Level1);
 
+    NumOp.ZeroCounters();
+    DenOp.ZeroCounters();
     Run(argc,argv);
+    
+    std::cout << GridLogMessage << "Numerator report, Pauli-Villars term         : " << std::endl;
+    NumOp.Report();
+    std::cout << GridLogMessage << "Denominator report, Dw(m) term (includes CG) : " << std::endl;
+    DenOp.Report();
+    
   };
 
 };

tests/hmc/Test_hmc_EOWilsonFermionGauge.cc

@@ -67,7 +67,7 @@ public:
     TwoFlavourEvenOddPseudoFermionAction<ImplPolicy> Nf2(FermOp,CG,CG);
 
     //Set smearing (true/false), default: false
-    Nf2.is_smeared=false;
+    Nf2.is_smeared=true;
     
     //Collect actions
     ActionLevel<LatticeGaugeField> Level1(1);

tests/hmc/Test_hmc_WilsonAdjointFermionGauge.cc

@@ -0,0 +1,107 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./tests/Test_hmc_WilsonAdjointFermionGauge.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: neo <cossu@post.kek.jp>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include "Grid/Grid.h"
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+namespace Grid {
+namespace QCD {
+
+// Here change the allowed (higher) representations
+typedef Representations< FundamentalRepresentation, AdjointRepresentation > TheRepresentations;
+
+
+class HmcRunner : public NerscHmcRunnerHirep< TheRepresentations > {
+ public:
+  void BuildTheAction(int argc, char **argv)
+
+  {
+    typedef WilsonAdjImplR ImplPolicy; // gauge field implemetation for the pseudofermions
+    typedef WilsonAdjFermionR FermionAction; // type of lattice fermions (Wilson, DW, ...)
+    typedef typename FermionAction::FermionField FermionField;
+
+    UGrid = SpaceTimeGrid::makeFourDimGrid(
+        GridDefaultLatt(), GridDefaultSimd(Nd, vComplex::Nsimd()),
+        GridDefaultMpi());
+    UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+
+    FGrid = UGrid;
+    FrbGrid = UrbGrid;
+
+    // temporarily need a gauge field
+    //LatticeGaugeField U(UGrid);
+    AdjointRepresentation::LatticeField U(UGrid);
+
+    // Gauge action
+    WilsonGaugeActionR Waction(2.25);
+
+    Real mass = -0.95;
+    FermionAction FermOp(U, *FGrid, *FrbGrid, mass);
+
+    ConjugateGradient<FermionField> CG(1.0e-8, 10000, false);
+    ConjugateResidual<FermionField> CR(1.0e-8, 10000);
+
+    // Pass two solvers: one for the force computation and one for the action
+    TwoFlavourPseudoFermionAction<ImplPolicy> Nf2(FermOp, CG, CG);
+
+    // Set smearing (true/false), default: false
+    Nf2.is_smeared = false;
+
+    // Collect actions
+    ActionLevel<LatticeGaugeField, TheRepresentations > Level1(1);
+    Level1.push_back(&Nf2);
+
+    ActionLevel<LatticeGaugeField, TheRepresentations > Level2(4);
+    Level2.push_back(&Waction);
+
+    TheAction.push_back(Level1);
+    TheAction.push_back(Level2);
+
+    Run(argc, argv);
+  };
+};
+}
+}
+
+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;
+
+  HmcRunner TheHMC;
+
+  TheHMC.BuildTheAction(argc, argv);
+}

tests/hmc/Test_hmc_WilsonFermionGauge.cc

@@ -1,77 +1,76 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./tests/Test_hmc_WilsonFermionGauge.cc
+Source file: ./tests/Test_hmc_WilsonFermionGauge.cc
 
-    Copyright (C) 2015
+Copyright (C) 2015
 
 Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 Author: neo <cossu@post.kek.jp>
 Author: paboyle <paboyle@ph.ed.ac.uk>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution directory
-    *************************************************************************************/
+*************************************************************************************/
-    /*  END LEGAL */
+/*  END LEGAL */
 #include <Grid/Grid.h>
 
 using namespace std;
 using namespace Grid;
 using namespace Grid::QCD;
 
-namespace Grid { 
+namespace Grid {
-  namespace QCD { 
+namespace QCD {
-
 
 class HmcRunner : public NerscHmcRunner {
-public:
+ public:
-
+  void BuildTheAction(int argc, char **argv)
-  void BuildTheAction (int argc, char **argv)
 
   {
     typedef WilsonImplR ImplPolicy;
     typedef WilsonFermionR FermionAction;
     typedef typename FermionAction::FermionField FermionField;
 
-    UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
+    UGrid = SpaceTimeGrid::makeFourDimGrid(
+        GridDefaultLatt(), GridDefaultSimd(Nd, vComplex::Nsimd()),
+        GridDefaultMpi());
     UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
-  
+
-    FGrid   = UGrid;
+    FGrid = UGrid;
     FrbGrid = UrbGrid;
 
     // temporarily need a gauge field
-    LatticeGaugeField  U(UGrid);
+    LatticeGaugeField U(UGrid);
 
     // Gauge action
     WilsonGaugeActionR Waction(5.6);
 
-    Real mass=-0.77;
+    Real mass = -0.77;
-    FermionAction FermOp(U,*FGrid,*FrbGrid,mass);
+    FermionAction FermOp(U, *FGrid, *FrbGrid, mass);
-  
-    ConjugateGradient<FermionField>  CG(1.0e-8,10000);
 
-    TwoFlavourPseudoFermionAction<ImplPolicy> Nf2(FermOp,CG,CG);
+    ConjugateGradient<FermionField> CG(1.0e-8, 10000);
-  
+
-    //Set smearing (true/false), default: false
+    TwoFlavourPseudoFermionAction<ImplPolicy> Nf2(FermOp, CG, CG);
+
+    // Set smearing (true/false), default: false
     Nf2.is_smeared = true;
 
-
+    // Collect actions
-    //Collect actions
     ActionLevel<LatticeGaugeField> Level1(1);
     Level1.push_back(&Nf2);
 
@@ -81,24 +80,20 @@ public:
     TheAction.push_back(Level1);
     TheAction.push_back(Level2);
 
-    Run(argc,argv);
+    Run(argc, argv);
   };
-
 };
-
+}
-}}
-
-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;
-
-  HmcRunner TheHMC;
-  
-  TheHMC.BuildTheAction(argc,argv);
-
 }
 
+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;
+
+  HmcRunner TheHMC;
+
+  TheHMC.BuildTheAction(argc, argv);
+}

tests/hmc/Test_hmc_WilsonMixedRepresentationsFermionGauge.cc

@@ -0,0 +1,113 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./tests/Test_hmc_WilsonAdjointFermionGauge.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: neo <cossu@post.kek.jp>
+Author: paboyle <paboyle@ph.ed.ac.uk>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include "Grid/Grid.h"
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+namespace Grid {
+namespace QCD {
+
+// Here change the allowed (higher) representations
+typedef Representations< FundamentalRepresentation, AdjointRepresentation , TwoIndexSymmetricRepresentation> TheRepresentations;
+
+class HmcRunner : public NerscHmcRunnerHirep< TheRepresentations > {
+ public:
+  void BuildTheAction(int argc, char **argv)
+
+  {
+    typedef WilsonAdjImplR AdjImplPolicy; // gauge field implemetation for the pseudofermions
+    typedef WilsonAdjFermionR AdjFermionAction; // type of lattice fermions (Wilson, DW, ...)
+    typedef WilsonTwoIndexSymmetricImplR SymmImplPolicy; 
+    typedef WilsonTwoIndexSymmetricFermionR SymmFermionAction; 
+
+ 
+    typedef typename AdjFermionAction::FermionField AdjFermionField;
+    typedef typename SymmFermionAction::FermionField SymmFermionField;
+
+    UGrid = SpaceTimeGrid::makeFourDimGrid(
+        GridDefaultLatt(), GridDefaultSimd(Nd, vComplex::Nsimd()),
+        GridDefaultMpi());
+    UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+
+    FGrid = UGrid;
+    FrbGrid = UrbGrid;
+
+    // temporarily need a gauge field
+    //LatticeGaugeField U(UGrid);
+    AdjointRepresentation::LatticeField UA(UGrid);
+    TwoIndexSymmetricRepresentation::LatticeField US(UGrid);
+
+    // Gauge action
+    WilsonGaugeActionR Waction(5.6);
+
+    Real adjoint_mass = -0.1;
+    Real symm_mass = -0.5;
+    AdjFermionAction AdjFermOp(UA, *FGrid, *FrbGrid, adjoint_mass);
+    SymmFermionAction SymmFermOp(US, *FGrid, *FrbGrid, symm_mass);
+
+    ConjugateGradient<AdjFermionField> CG_adj(1.0e-8, 10000, false);
+    ConjugateGradient<SymmFermionField> CG_symm(1.0e-8, 10000, false);
+
+    // Pass two solvers: one for the force computation and one for the action
+    TwoFlavourPseudoFermionAction<AdjImplPolicy> Nf2_Adj(AdjFermOp, CG_adj, CG_adj);
+    TwoFlavourPseudoFermionAction<SymmImplPolicy> Nf2_Symm(SymmFermOp, CG_symm, CG_symm);
+
+    // Collect actions
+    ActionLevel<LatticeGaugeField, TheRepresentations > Level1(1);
+    Level1.push_back(&Nf2_Adj);
+    Level1.push_back(&Nf2_Symm);
+
+    ActionLevel<LatticeGaugeField, TheRepresentations > Level2(4);
+    Level2.push_back(&Waction);
+
+    TheAction.push_back(Level1);
+    TheAction.push_back(Level2);
+
+    Run(argc, argv);
+  };
+};
+}
+}
+
+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;
+
+  HmcRunner TheHMC;
+
+  TheHMC.BuildTheAction(argc, argv);
+}

tests/hmc/Test_hmc_WilsonTwoIndexSymmetricFermionGauge.cc

@@ -0,0 +1,103 @@
+/*************************************************************************************
+
+Grid physics library, www.github.com/paboyle/Grid
+
+Source file: ./tests/Test_hmc_WilsonAdjointFermionGauge.cc
+
+Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+Author: neo <cossu@post.kek.jp>
+
+This program is free software; you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation; either version 2 of the License, or
+(at your option) any later version.
+
+This program is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License along
+with this program; if not, write to the Free Software Foundation, Inc.,
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+See the full license in the file "LICENSE" in the top level distribution
+directory
+*************************************************************************************/
+/*  END LEGAL */
+#include "Grid/Grid.h"
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+namespace Grid {
+namespace QCD {
+
+// Here change the allowed (higher) representations
+typedef Representations< FundamentalRepresentation, TwoIndexSymmetricRepresentation > TheRepresentations;
+
+
+class HmcRunner : public NerscHmcRunnerHirep< TheRepresentations > {
+ public:
+  void BuildTheAction(int argc, char **argv)
+
+  {
+    typedef WilsonTwoIndexSymmetricImplR ImplPolicy; // gauge field implemetation for the pseudofermions
+    typedef WilsonTwoIndexSymmetricFermionR FermionAction; // type of lattice fermions (Wilson, DW, ...)
+    typedef typename FermionAction::FermionField FermionField;
+
+    UGrid = SpaceTimeGrid::makeFourDimGrid(
+        GridDefaultLatt(), GridDefaultSimd(Nd, vComplex::Nsimd()),
+        GridDefaultMpi());
+    UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+
+    FGrid = UGrid;
+    FrbGrid = UrbGrid;
+
+    // temporarily need a gauge field
+    TwoIndexSymmetricRepresentation::LatticeField U(UGrid);
+
+    // Gauge action
+    WilsonGaugeActionR Waction(2.0);
+
+    Real mass = -0.0;
+    FermionAction FermOp(U, *FGrid, *FrbGrid, mass);
+
+    ConjugateGradient<FermionField> CG(1.0e-8, 10000, false);
+
+    // Pass two solvers: one for the force computation and one for the action
+    TwoFlavourPseudoFermionAction<ImplPolicy> Nf2(FermOp, CG, CG);
+
+    // Set smearing (true/false), default: false
+    Nf2.is_smeared = false;
+
+    // Collect actions
+    ActionLevel<LatticeGaugeField, TheRepresentations > Level1(1);
+    Level1.push_back(&Nf2);
+
+    ActionLevel<LatticeGaugeField, TheRepresentations > Level2(4);
+    Level2.push_back(&Waction);
+
+    TheAction.push_back(Level1);
+    TheAction.push_back(Level2);
+
+    Run(argc, argv);
+  };
+};
+}
+}
+
+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;
+
+  HmcRunner TheHMC;
+
+  TheHMC.BuildTheAction(argc, argv);
+}

tests/solver/Test_dwf_cg_prec.cc

@@ -1,87 +1,105 @@
-    /*************************************************************************************
+/*************************************************************************************
 
-    Grid physics library, www.github.com/paboyle/Grid 
+Grid physics library, www.github.com/paboyle/Grid
 
-    Source file: ./tests/Test_dwf_cg_prec.cc
+Source file: ./tests/Test_dwf_cg_prec.cc
 
-    Copyright (C) 2015
+Copyright (C) 2015
 
 Author: Peter Boyle <paboyle@ph.ed.ac.uk>
 
-    This program is free software; you can redistribute it and/or modify
+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
+it under the terms of the GNU General Public License as published by
-    the Free Software Foundation; either version 2 of the License, or
+the Free Software Foundation; either version 2 of the License, or
-    (at your option) any later version.
+(at your option) any later version.
 
-    This program is distributed in the hope that it will be useful,
+This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
+but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
+GNU General Public License for more details.
 
-    You should have received a copy of the GNU General Public License along
+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.,
+with this program; if not, write to the Free Software Foundation, Inc.,
-    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 
-    See the full license in the file "LICENSE" in the top level distribution directory
+See the full license in the file "LICENSE" in the top level distribution
-    *************************************************************************************/
+directory
-    /*  END LEGAL */
+*************************************************************************************/
+/*  END LEGAL */
 #include <Grid/Grid.h>
 
 using namespace std;
 using namespace Grid;
 using namespace Grid::QCD;
 
-template<class d>
+template <class d>
 struct scal {
   d internal;
 };
 
-  Gamma::GammaMatrix Gmu [] = {
+Gamma::GammaMatrix Gmu[] = {Gamma::GammaX, Gamma::GammaY, Gamma::GammaZ,
-    Gamma::GammaX,
+                            Gamma::GammaT};
-    Gamma::GammaY,
-    Gamma::GammaZ,
-    Gamma::GammaT
-  };
 
-int main (int argc, char ** argv)
+int main(int argc, char** argv) {
-{
+  Grid_init(&argc, &argv);
-  Grid_init(&argc,&argv);
 
-  const int Ls=8;
+  const int Ls = 16;
 
-  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
+  GridCartesian* UGrid = SpaceTimeGrid::makeFourDimGrid(
-  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+      GridDefaultLatt(), GridDefaultSimd(Nd, vComplex::Nsimd()),
-  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
+      GridDefaultMpi());
-  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
+  GridRedBlackCartesian* UrbGrid =
+      SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+  GridCartesian* FGrid = SpaceTimeGrid::makeFiveDimGrid(Ls, UGrid);
+  GridRedBlackCartesian* FrbGrid =
+      SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls, UGrid);
 
-  std::vector<int> seeds4({1,2,3,4});
+  std::vector<int> seeds4({1, 2, 3, 4});
-  std::vector<int> seeds5({5,6,7,8});
+  std::vector<int> seeds5({5, 6, 7, 8});
-  GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
+  GridParallelRNG RNG5(FGrid);
-  GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
+  RNG5.SeedFixedIntegers(seeds5);
+  GridParallelRNG RNG4(UGrid);
+  RNG4.SeedFixedIntegers(seeds4);
 
-  LatticeFermion    src(FGrid); random(RNG5,src);
+  LatticeFermion src(FGrid);
-  LatticeFermion result(FGrid); result=zero;
+  random(RNG5, src);
-  LatticeGaugeField Umu(UGrid); 
+  LatticeFermion result(FGrid);
+  result = zero;
+  LatticeGaugeField Umu(UGrid);
 
-  SU3::HotConfiguration(RNG4,Umu);
+  SU3::HotConfiguration(RNG4, Umu);
 
-  std::vector<LatticeColourMatrix> U(4,UGrid);
+  std::cout << GridLogMessage << "Lattice dimensions: " << GridDefaultLatt()
-  for(int mu=0;mu<Nd;mu++){
+            << "   Ls: " << Ls << std::endl;
-    U[mu] = PeekIndex<LorentzIndex>(Umu,mu);
+
+  std::vector<LatticeColourMatrix> U(4, UGrid);
+  for (int mu = 0; mu < Nd; mu++) {
+    U[mu] = PeekIndex<LorentzIndex>(Umu, mu);
   }
-  
-  RealD mass=0.1;
-  RealD M5=1.8;
-  DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
 
-  LatticeFermion    src_o(FrbGrid);
+  RealD mass = 0.01;
+  RealD M5 = 1.8;
+  DomainWallFermionR Ddwf(Umu, *FGrid, *FrbGrid, *UGrid, *UrbGrid, mass, M5);
+
+  LatticeFermion src_o(FrbGrid);
   LatticeFermion result_o(FrbGrid);
-  pickCheckerboard(Odd,src_o,src);
+  pickCheckerboard(Odd, src_o, src);
-  result_o=zero;
+  result_o = zero;
 
-  SchurDiagMooeeOperator<DomainWallFermionR,LatticeFermion> HermOpEO(Ddwf);
+  GridStopWatch CGTimer;
-  ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
+
-  CG(HermOpEO,src_o,result_o);
+  SchurDiagMooeeOperator<DomainWallFermionR, LatticeFermion> HermOpEO(Ddwf);
+  ConjugateGradient<LatticeFermion> CG(1.0e-8, 10000, 0);// switch off the assert
+
+  CGTimer.Start();
+  CG(HermOpEO, src_o, result_o);
+  CGTimer.Stop();
+
+  std::cout << GridLogMessage << "Total CG time : " << CGTimer.Elapsed()
+            << std::endl;
+
+  std::cout << GridLogMessage << "######## Dhop calls summary" << std::endl;
+  Ddwf.Report();
 
   Grid_finalize();
 }

tests/solver/Test_wilson_cg_prec.cc

@@ -83,6 +83,7 @@ int main (int argc, char ** argv)
   SchurDiagMooeeOperator<WilsonFermionR,LatticeFermion> HermOpEO(Dw);
   ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
   CG(HermOpEO,src_o,result_o);
+  
 
   Grid_finalize();
 }