From 9cfc180334ab7aa16ce90405d7d11f51d164a94a Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Fri, 29 May 2015 14:11:34 +0100
Subject: [PATCH 01/20] Integer wrap problem fixed.
---
benchmarks/Grid_comms.cc | 8 +++++---
1 file changed, 5 insertions(+), 3 deletions(-)
diff --git a/benchmarks/Grid_comms.cc b/benchmarks/Grid_comms.cc
index 59d709a0..c218070b 100644
--- a/benchmarks/Grid_comms.cc
+++ b/benchmarks/Grid_comms.cc
@@ -77,7 +77,8 @@ int main (int argc, char ** argv)
}
double stop=usecond();
- double xbytes = Nloop*bytes*2*ncomm;
+ double dbytes = bytes;
+ double xbytes = Nloop*dbytes*2.0*ncomm;
double rbytes = xbytes;
double bidibytes = xbytes+rbytes;
@@ -151,8 +152,9 @@ int main (int argc, char ** argv)
}
double stop=usecond();
-
- double xbytes = Nloop*bytes*2*ncomm;
+
+ double dbytes = bytes;
+ double xbytes = Nloop*dbytes*2.0*ncomm;
double rbytes = xbytes;
double bidibytes = xbytes+rbytes;
From a75b6f6e78fab299b6c08f983ca92b4c4ebfcaf8 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Sun, 31 May 2015 15:09:02 +0100
Subject: [PATCH 02/20] Large scale change to support 5d fermion formulations.
Have 5d replicated wilson with 4d gauge working and matrix regressing to Ls
copies of wilson.
---
benchmarks/Grid_comms.cc | 2 +-
benchmarks/Grid_dwf.cc | 145 +++++++++++
benchmarks/Grid_wilson.cc | 9 +-
benchmarks/Grid_wilson_cg_prec.cc | 6 +-
benchmarks/Grid_wilson_cg_schur.cc | 2 +-
benchmarks/Grid_wilson_cg_unprec.cc | 4 +-
benchmarks/Grid_wilson_evenodd.cc | 2 +-
benchmarks/Makefile.am | 5 +-
lib/Grid.h | 2 +-
lib/Grid_init.cc | 6 +-
lib/Grid_stencil.h | 33 +--
lib/Grid_threads.h | 2 +-
lib/Makefile.am | 166 +++++++------
lib/algorithms/SparseMatrix.h | 7 +-
lib/algorithms/iterative/SchurRedBlack.h | 4 +-
lib/cartesian/Grid_cartesian_base.h | 11 +-
lib/cartesian/Grid_cartesian_full.h | 7 +-
lib/cartesian/Grid_cartesian_red_black.h | 196 ++++++++++-----
lib/cshift/Grid_cshift_common.h | 14 +-
lib/cshift/Grid_cshift_mpi.h | 16 +-
lib/cshift/Grid_cshift_none.h | 2 +-
lib/lattice/Grid_lattice_peekpoke.h | 2 +-
lib/lattice/Grid_lattice_transfer.h | 3 +-
lib/qcd/{Grid_qcd_dirac.cc => Dirac.cc} | 0
lib/qcd/{Grid_qcd_dirac.h => Dirac.h} | 0
lib/qcd/Grid_qcd_wilson_dop.cc | 217 -----------------
lib/qcd/Grid_qcd_wilson_dop.h | 105 --------
lib/qcd/{Grid_qcd.h => QCD.h} | 17 +-
lib/qcd/{Grid_qcd_2spinor.h => TwoSpinor.h} | 0
lib/qcd/action/Actions.h | 10 +
lib/qcd/action/fermion/FermionAction.h | 47 ++++
.../action/fermion/FiveDimWilsonFermion.cc | 228 ++++++++++++++++++
lib/qcd/action/fermion/FiveDimWilsonFermion.h | 108 +++++++++
lib/qcd/action/fermion/WilsonCompressor.h | 61 +++++
lib/qcd/action/fermion/WilsonFermion.cc | 163 +++++++++++++
lib/qcd/action/fermion/WilsonFermion.h | 87 +++++++
.../fermion/WilsonKernels.cc} | 46 ++--
lib/qcd/action/fermion/WilsonKernels.h | 42 ++++
.../fermion/WilsonKernelsHand.cc} | 57 ++---
lib/stencil/Grid_lebesgue.cc | 103 ++++++++
lib/stencil/Grid_lebesgue.h | 29 +++
lib/stencil/Grid_stencil_common.cc | 103 +-------
scripts/linecount | 2 +-
tests/Grid_cshift_red_black.cc | 165 +++++++++++++
tests/Makefile.am | 5 +-
45 files changed, 1549 insertions(+), 692 deletions(-)
create mode 100644 benchmarks/Grid_dwf.cc
rename lib/qcd/{Grid_qcd_dirac.cc => Dirac.cc} (100%)
rename lib/qcd/{Grid_qcd_dirac.h => Dirac.h} (100%)
delete mode 100644 lib/qcd/Grid_qcd_wilson_dop.cc
delete mode 100644 lib/qcd/Grid_qcd_wilson_dop.h
rename lib/qcd/{Grid_qcd.h => QCD.h} (97%)
rename lib/qcd/{Grid_qcd_2spinor.h => TwoSpinor.h} (100%)
create mode 100644 lib/qcd/action/Actions.h
create mode 100644 lib/qcd/action/fermion/FermionAction.h
create mode 100644 lib/qcd/action/fermion/FiveDimWilsonFermion.cc
create mode 100644 lib/qcd/action/fermion/FiveDimWilsonFermion.h
create mode 100644 lib/qcd/action/fermion/WilsonCompressor.h
create mode 100644 lib/qcd/action/fermion/WilsonFermion.cc
create mode 100644 lib/qcd/action/fermion/WilsonFermion.h
rename lib/qcd/{Grid_qcd_dhop.cc => action/fermion/WilsonKernels.cc} (87%)
create mode 100644 lib/qcd/action/fermion/WilsonKernels.h
rename lib/qcd/{Grid_qcd_dhop_hand.cc => action/fermion/WilsonKernelsHand.cc} (92%)
create mode 100644 lib/stencil/Grid_lebesgue.cc
create mode 100644 lib/stencil/Grid_lebesgue.h
create mode 100644 tests/Grid_cshift_red_black.cc
diff --git a/benchmarks/Grid_comms.cc b/benchmarks/Grid_comms.cc
index c218070b..a24e7ed8 100644
--- a/benchmarks/Grid_comms.cc
+++ b/benchmarks/Grid_comms.cc
@@ -82,7 +82,7 @@ int main (int argc, char ** argv)
double rbytes = xbytes;
double bidibytes = xbytes+rbytes;
- double time = stop-start;
+ double time = stop-start; // microseconds
std::cout << lat<<"\t\t"<<Ls<<"\t\t"<<bytes<<"\t\t"<<xbytes/time<<"\t\t"<<bidibytes/time<<std::endl;
}
diff --git a/benchmarks/Grid_dwf.cc b/benchmarks/Grid_dwf.cc
new file mode 100644
index 00000000..eb1d9299
--- /dev/null
+++ b/benchmarks/Grid_dwf.cc
@@ -0,0 +1,145 @@
+#include <Grid.h>
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+template<class d>
+struct scal {
+ d internal;
+};
+
+ Gamma::GammaMatrix Gmu [] = {
+ Gamma::GammaX,
+ Gamma::GammaY,
+ Gamma::GammaZ,
+ Gamma::GammaT
+ };
+
+int main (int argc, char ** argv)
+{
+ Grid_init(&argc,&argv);
+
+ int threads = GridThread::GetThreads();
+ std::cout << "Grid is setup to use "<<threads<<" threads"<<std::endl;
+
+ std::vector<int> latt4 = GridDefaultLatt();
+ std::vector<int> simd4 = GridDefaultSimd(Nd,vComplexF::Nsimd());
+ std::vector<int> mpi4 = GridDefaultMpi();
+
+ assert(latt4.size()==4 );
+ assert(simd4.size()==4 );
+ assert(mpi4.size() ==4 );
+
+ const int Ls=1;
+ std::vector<int> latt5({Ls,latt4[0],latt4[1],latt4[2],latt4[3]});
+ std::vector<int> simd5({1 ,simd4[0],simd4[1],simd4[2],simd4[3]});
+ std::vector<int> mpi5({1 , mpi4[0], mpi4[1], mpi4[2], mpi4[3]});
+ std::vector<int> cb5({0,1,1,1,1}); // Checkerboard 4d only
+ int cbd=1; // use dim-1 to reduce
+
+ // Four dim grid for gauge field U
+ GridCartesian UGrid(latt4,simd4,mpi4);
+ GridRedBlackCartesian UrbGrid(&UGrid);
+
+ // Five dim grid for fermions F
+ GridCartesian FGrid(latt5,simd5,mpi5);
+ GridRedBlackCartesian FrbGrid(latt5,simd5,mpi5,cb5,cbd);
+
+ std::vector<int> seeds4({1,2,3,4});
+ std::vector<int> seeds5({5,6,7,8});
+
+ GridParallelRNG RNG5(&FGrid); RNG5.SeedFixedIntegers(seeds5);
+ LatticeFermion src (&FGrid); random(RNG5,src);
+ LatticeFermion result(&FGrid); result=zero;
+ LatticeFermion ref(&FGrid); ref=zero;
+ LatticeFermion tmp(&FGrid);
+ LatticeFermion err(&FGrid);
+
+ ColourMatrix cm = Complex(1.0,0.0);
+
+ GridParallelRNG RNG4(&UGrid); RNG4.SeedFixedIntegers(seeds4);
+ LatticeGaugeField Umu(&UGrid); random(RNG4,Umu);
+ LatticeGaugeField Umu5d(&FGrid);
+
+ // replicate across fifth dimension
+ for(int ss=0;ss<Umu._grid->oSites();ss++){
+ for(int s=0;s<Ls;s++){
+ Umu5d._odata[Ls*ss+s] = Umu._odata[ss];
+ }
+ }
+
+ ////////////////////////////////////
+ // Naive wilson implementation
+ ////////////////////////////////////
+ std::vector<LatticeColourMatrix> U(4,&FGrid);
+ for(int mu=0;mu<Nd;mu++){
+ U[mu] = peekIndex<LorentzIndex>(Umu5d,mu);
+ }
+
+ if (1)
+ {
+ ref = zero;
+ for(int mu=0;mu<Nd;mu++){
+
+ tmp = U[mu]*Cshift(src,mu+1,1);
+ ref=ref + tmp + Gamma(Gmu[mu])*tmp;
+
+ tmp =adj(U[mu])*src;
+ tmp =Cshift(tmp,mu+1,-1);
+ ref=ref + tmp - Gamma(Gmu[mu])*tmp;
+ }
+ ref = -0.5*ref;
+ }
+
+ RealD mass=0.1;
+ FiveDimWilsonFermion Dw(Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass);
+
+ std::cout << "Calling Dw"<<std::endl;
+ int ncall=1000;
+ double t0=usecond();
+ for(int i=0;i<ncall;i++){
+ Dw.Dhop(src,result,0);
+ }
+ double t1=usecond();
+
+
+ double volume=Ls; for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
+ double flops=1344*volume*ncall;
+
+ std::cout << "Called Dw"<<std::endl;
+ std::cout << "norm result "<< norm2(result)<<std::endl;
+ std::cout << "norm ref "<< norm2(ref)<<std::endl;
+ std::cout << "mflop/s = "<< flops/(t1-t0)<<std::endl;
+ err = ref-result;
+ std::cout << "norm diff "<< norm2(err)<<std::endl;
+
+
+ if (1)
+ { // Naive wilson dag implementation
+ ref = zero;
+ for(int mu=0;mu<Nd;mu++){
+
+ // ref = src - Gamma(Gamma::GammaX)* src ; // 1+gamma_x
+ tmp = U[mu]*Cshift(src,mu+1,1);
+ for(int i=0;i<ref._odata.size();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 = -0.5*ref;
+ }
+ Dw.Dhop(src,result,1);
+ std::cout << "Called DwDag"<<std::endl;
+ std::cout << "norm result "<< norm2(result)<<std::endl;
+ std::cout << "norm ref "<< norm2(ref)<<std::endl;
+ err = ref-result;
+ std::cout << "norm diff "<< norm2(err)<<std::endl;
+
+ Grid_finalize();
+}
diff --git a/benchmarks/Grid_wilson.cc b/benchmarks/Grid_wilson.cc
index 3b0d04bc..c1050098 100644
--- a/benchmarks/Grid_wilson.cc
+++ b/benchmarks/Grid_wilson.cc
@@ -49,6 +49,7 @@ int main (int argc, char ** argv)
}
// Only one non-zero (y)
+#if 0
Umu=zero;
Complex cone(1.0,0.0);
for(int nn=0;nn<Nd;nn++){
@@ -59,6 +60,7 @@ int main (int argc, char ** argv)
}
pokeIndex<LorentzIndex>(Umu,U[nn],nn);
}
+#endif
for(int mu=0;mu<Nd;mu++){
U[mu] = peekIndex<LorentzIndex>(Umu,mu);
@@ -80,9 +82,9 @@ int main (int argc, char ** argv)
}
}
}
-
+ ref = -0.5*ref;
RealD mass=0.1;
- WilsonMatrix Dw(Umu,Grid,RBGrid,mass);
+ WilsonFermion Dw(Umu,Grid,RBGrid,mass);
std::cout << "Calling Dw"<<std::endl;
int ncall=10000;
@@ -91,7 +93,7 @@ int main (int argc, char ** argv)
Dw.Dhop(src,result,0);
}
double t1=usecond();
- double flops=1320*volume*ncall;
+ double flops=1344*volume*ncall;
std::cout << "Called Dw"<<std::endl;
std::cout << "norm result "<< norm2(result)<<std::endl;
@@ -129,6 +131,7 @@ int main (int argc, char ** argv)
}
}
}
+ ref = -0.5*ref;
Dw.Dhop(src,result,1);
std::cout << "Called DwDag"<<std::endl;
std::cout << "norm result "<< norm2(result)<<std::endl;
diff --git a/benchmarks/Grid_wilson_cg_prec.cc b/benchmarks/Grid_wilson_cg_prec.cc
index e86ec820..e376349c 100644
--- a/benchmarks/Grid_wilson_cg_prec.cc
+++ b/benchmarks/Grid_wilson_cg_prec.cc
@@ -42,9 +42,9 @@ int main (int argc, char ** argv)
}
RealD mass=0.5;
- WilsonMatrix Dw(Umu,Grid,RBGrid,mass);
+ WilsonFermion Dw(Umu,Grid,RBGrid,mass);
- // HermitianOperator<WilsonMatrix,LatticeFermion> HermOp(Dw);
+ // HermitianOperator<WilsonFermion,LatticeFermion> HermOp(Dw);
// ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
// CG(HermOp,src,result);
@@ -53,7 +53,7 @@ int main (int argc, char ** argv)
pickCheckerboard(Odd,src_o,src);
result_o=zero;
- HermitianCheckerBoardedOperator<WilsonMatrix,LatticeFermion> HermOpEO(Dw);
+ HermitianCheckerBoardedOperator<WilsonFermion,LatticeFermion> HermOpEO(Dw);
ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
CG(HermOpEO,src_o,result_o);
diff --git a/benchmarks/Grid_wilson_cg_schur.cc b/benchmarks/Grid_wilson_cg_schur.cc
index af630ae1..28db1d4b 100644
--- a/benchmarks/Grid_wilson_cg_schur.cc
+++ b/benchmarks/Grid_wilson_cg_schur.cc
@@ -37,7 +37,7 @@ int main (int argc, char ** argv)
LatticeFermion resid(&Grid);
RealD mass=0.5;
- WilsonMatrix Dw(Umu,Grid,RBGrid,mass);
+ WilsonFermion Dw(Umu,Grid,RBGrid,mass);
ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
SchurRedBlackSolve<LatticeFermion> SchurSolver(CG);
diff --git a/benchmarks/Grid_wilson_cg_unprec.cc b/benchmarks/Grid_wilson_cg_unprec.cc
index 15302aab..905dfde5 100644
--- a/benchmarks/Grid_wilson_cg_unprec.cc
+++ b/benchmarks/Grid_wilson_cg_unprec.cc
@@ -47,9 +47,9 @@ int main (int argc, char ** argv)
}
RealD mass=0.5;
- WilsonMatrix Dw(Umu,Grid,RBGrid,mass);
+ WilsonFermion Dw(Umu,Grid,RBGrid,mass);
- HermitianOperator<WilsonMatrix,LatticeFermion> HermOp(Dw);
+ HermitianOperator<WilsonFermion,LatticeFermion> HermOp(Dw);
ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
CG(HermOp,src,result);
diff --git a/benchmarks/Grid_wilson_evenodd.cc b/benchmarks/Grid_wilson_evenodd.cc
index 4bc8c357..a073139d 100644
--- a/benchmarks/Grid_wilson_evenodd.cc
+++ b/benchmarks/Grid_wilson_evenodd.cc
@@ -60,7 +60,7 @@ int main (int argc, char ** argv)
RealD mass=0.1;
- WilsonMatrix Dw(Umu,Grid,RBGrid,mass);
+ WilsonFermion Dw(Umu,Grid,RBGrid,mass);
LatticeFermion src_e (&RBGrid);
LatticeFermion src_o (&RBGrid);
diff --git a/benchmarks/Makefile.am b/benchmarks/Makefile.am
index 95ae5eca..b2649669 100644
--- a/benchmarks/Makefile.am
+++ b/benchmarks/Makefile.am
@@ -5,11 +5,14 @@ AM_LDFLAGS = -L$(top_builddir)/lib
#
# Test code
#
-bin_PROGRAMS = Grid_wilson Grid_comms Grid_memory_bandwidth Grid_su3 Grid_wilson_cg_unprec Grid_wilson_evenodd Grid_wilson_cg_prec Grid_wilson_cg_schur
+bin_PROGRAMS = Grid_wilson Grid_comms Grid_memory_bandwidth Grid_su3 Grid_wilson_cg_unprec Grid_wilson_evenodd Grid_wilson_cg_prec Grid_wilson_cg_schur Grid_dwf
Grid_wilson_SOURCES = Grid_wilson.cc
Grid_wilson_LDADD = -lGrid
+Grid_dwf_SOURCES = Grid_dwf.cc
+Grid_dwf_LDADD = -lGrid
+
Grid_wilson_evenodd_SOURCES = Grid_wilson_evenodd.cc
Grid_wilson_evenodd_LDADD = -lGrid
diff --git a/lib/Grid.h b/lib/Grid.h
index 1e513a46..7fa56892 100644
--- a/lib/Grid.h
+++ b/lib/Grid.h
@@ -60,7 +60,7 @@
#include <Grid_algorithms.h>// subdir aggregate
-#include <qcd/Grid_qcd.h>
+#include <qcd/QCD.h>
#include <parallelIO/GridNerscIO.h>
namespace Grid {
diff --git a/lib/Grid_init.cc b/lib/Grid_init.cc
index f72393cb..580e602a 100644
--- a/lib/Grid_init.cc
+++ b/lib/Grid_init.cc
@@ -142,7 +142,11 @@ void Grid_init(int *argc,char ***argv)
Grid_quiesce_nodes();
}
if( GridCmdOptionExists(*argv,*argv+*argc,"--dslash-opt") ){
- WilsonMatrix::HandOptDslash=1;
+ WilsonFermion::HandOptDslash=1;
+ FiveDimWilsonFermion::HandOptDslash=1;
+ }
+ if( GridCmdOptionExists(*argv,*argv+*argc,"--lebesgue") ){
+ LebesgueOrder::UseLebesgueOrder=1;
}
GridParseLayout(*argv,*argc,
Grid_default_latt,
diff --git a/lib/Grid_stencil.h b/lib/Grid_stencil.h
index fa22361b..50d22453 100644
--- a/lib/Grid_stencil.h
+++ b/lib/Grid_stencil.h
@@ -1,6 +1,8 @@
#ifndef GRID_STENCIL_H
#define GRID_STENCIL_H
+#include <stencil/Grid_lebesgue.h> // subdir aggregate
+
//////////////////////////////////////////////////////////////////////////////////////////
// Must not lose sight that goal is to be able to construct really efficient
// gather to a point stencil code. CSHIFT is not the best way, so need
@@ -38,29 +40,12 @@
//////////////////////////////////////////////////////////////////////////////////////////
namespace Grid {
-
-
+
class CartesianStencil { // Stencil runs along coordinate axes only; NO diagonal fill in.
public:
typedef uint32_t StencilInteger;
-
-
-
- StencilInteger alignup(StencilInteger n){
- n--; // 1000 0011 --> 1000 0010
- n |= n >> 1; // 1000 0010 | 0100 0001 = 1100 0011
- n |= n >> 2; // 1100 0011 | 0011 0000 = 1111 0011
- n |= n >> 4; // 1111 0011 | 0000 1111 = 1111 1111
- n |= n >> 8; // ... (At this point all bits are 1, so further bitwise-or
- n |= n >> 16; // operations produce no effect.)
- n++; // 1111 1111 --> 1 0000 0000
- return n;
- };
- void LebesgueOrder(void);
-
- std::vector<StencilInteger> _LebesgueReorder;
int _checkerboard;
int _npoints; // Move to template param?
@@ -131,8 +116,8 @@ namespace Grid {
// Gather phase
int sshift [2];
if ( comm_dim ) {
- sshift[0] = _grid->CheckerBoardShift(_checkerboard,dimension,shift,0);
- sshift[1] = _grid->CheckerBoardShift(_checkerboard,dimension,shift,1);
+ sshift[0] = _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,Even);
+ sshift[1] = _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,Odd);
if ( sshift[0] == sshift[1] ) {
if (splice_dim) {
GatherStartCommsSimd(source,dimension,shift,0x3,u_comm_buf,u_comm_offset,compress);
@@ -179,8 +164,8 @@ namespace Grid {
std::vector<cobj,alignedAllocator<cobj> > send_buf(buffer_size); // hmm...
std::vector<cobj,alignedAllocator<cobj> > recv_buf(buffer_size);
- int cb= (cbmask==0x2)? 1 : 0;
- int sshift= _grid->CheckerBoardShift(rhs.checkerboard,dimension,shift,cb);
+ int cb= (cbmask==0x2)? Odd : Even;
+ int sshift= _grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,cb);
for(int x=0;x<rd;x++){
@@ -266,8 +251,8 @@ namespace Grid {
// Work out what to send where
///////////////////////////////////////////
- int cb = (cbmask==0x2)? 1 : 0;
- int sshift= _grid->CheckerBoardShift(rhs.checkerboard,dimension,shift,cb);
+ int cb = (cbmask==0x2)? Odd : Even;
+ int sshift= _grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,cb);
// loop over outer coord planes orthog to dim
for(int x=0;x<rd;x++){
diff --git a/lib/Grid_threads.h b/lib/Grid_threads.h
index 24581855..f453b860 100644
--- a/lib/Grid_threads.h
+++ b/lib/Grid_threads.h
@@ -9,7 +9,7 @@
#ifdef GRID_OMP
#include <omp.h>
-#define PARALLEL_FOR_LOOP _Pragma("omp parallel for")
+#define PARALLEL_FOR_LOOP _Pragma("omp parallel for ")
#define PARALLEL_NESTED_LOOP2 _Pragma("omp parallel for collapse(2)")
#else
#define PARALLEL_FOR_LOOP
diff --git a/lib/Makefile.am b/lib/Makefile.am
index a5b89c0a..aa531df7 100644
--- a/lib/Makefile.am
+++ b/lib/Makefile.am
@@ -16,87 +16,103 @@ endif
lib_LIBRARIES = libGrid.a
libGrid_a_SOURCES = \
Grid_init.cc \
+ stencil/Grid_lebesgue.cc \
stencil/Grid_stencil_common.cc \
- qcd/Grid_qcd_dirac.cc \
- qcd/Grid_qcd_dhop.cc \
- qcd/Grid_qcd_dhop_hand.cc \
- qcd/Grid_qcd_wilson_dop.cc \
algorithms/approx/Zolotarev.cc \
algorithms/approx/Remez.cc \
+ qcd/action/fermion/FiveDimWilsonFermion.cc\
+ qcd/action/fermion/WilsonFermion.cc\
+ qcd/action/fermion/WilsonKernels.cc\
+ qcd/action/fermion/WilsonKernelsHand.cc\
+ qcd/Dirac.cc\
$(extra_sources)
#
# Include files
#
-nobase_include_HEADERS = algorithms/approx/bigfloat.h \
- algorithms/approx/Chebyshev.h \
- algorithms/approx/Remez.h \
- algorithms/approx/Zolotarev.h \
- algorithms/iterative/ConjugateGradient.h \
- algorithms/iterative/NormalEquations.h \
- algorithms/iterative/SchurRedBlack.h \
- algorithms/LinearOperator.h \
- algorithms/SparseMatrix.h \
- cartesian/Grid_cartesian_base.h \
- cartesian/Grid_cartesian_full.h \
- cartesian/Grid_cartesian_red_black.h \
- communicator/Grid_communicator_base.h \
- cshift/Grid_cshift_common.h \
- cshift/Grid_cshift_mpi.h \
- cshift/Grid_cshift_none.h \
- Grid.h \
- Grid_algorithms.h \
- Grid_aligned_allocator.h \
- Grid_cartesian.h \
- Grid_communicator.h \
- Grid_comparison.h \
- Grid_cshift.h \
- Grid_extract.h \
- Grid_lattice.h \
- Grid_math.h \
- Grid_simd.h \
- Grid_stencil.h \
- Grid_threads.h \
- lattice/Grid_lattice_arith.h \
- lattice/Grid_lattice_base.h \
- lattice/Grid_lattice_comparison.h \
- lattice/Grid_lattice_conformable.h \
- lattice/Grid_lattice_coordinate.h \
- lattice/Grid_lattice_ET.h \
- lattice/Grid_lattice_local.h \
- lattice/Grid_lattice_overload.h \
- lattice/Grid_lattice_peekpoke.h \
- lattice/Grid_lattice_reality.h \
- lattice/Grid_lattice_reduction.h \
- lattice/Grid_lattice_rng.h \
- lattice/Grid_lattice_trace.h \
- lattice/Grid_lattice_transfer.h \
- lattice/Grid_lattice_transpose.h \
- lattice/Grid_lattice_where.h \
- math/Grid_math_arith.h \
- math/Grid_math_arith_add.h \
- math/Grid_math_arith_mac.h \
- math/Grid_math_arith_mul.h \
- math/Grid_math_arith_scalar.h \
- math/Grid_math_arith_sub.h \
- math/Grid_math_inner.h \
- math/Grid_math_outer.h \
- math/Grid_math_peek.h \
- math/Grid_math_poke.h \
- math/Grid_math_reality.h \
- math/Grid_math_tensors.h \
- math/Grid_math_trace.h \
- math/Grid_math_traits.h \
- math/Grid_math_transpose.h \
- parallelIO/GridNerscIO.h \
- qcd/Grid_qcd.h \
- qcd/Grid_qcd_2spinor.h \
- qcd/Grid_qcd_dirac.h \
- qcd/Grid_qcd_wilson_dop.h \
- simd/Grid_vector_types.h \
- simd/Grid_sse4.h \
- simd/Grid_avx.h \
- simd/Grid_avx512.h
-
-
+nobase_include_HEADERS=\
+ ./algorithms/approx/bigfloat.h\
+ ./algorithms/approx/bigfloat_double.h\
+ ./algorithms/approx/Chebyshev.h\
+ ./algorithms/approx/Remez.h\
+ ./algorithms/approx/Zolotarev.h\
+ ./algorithms/iterative/ConjugateGradient.h\
+ ./algorithms/iterative/NormalEquations.h\
+ ./algorithms/iterative/SchurRedBlack.h\
+ ./algorithms/LinearOperator.h\
+ ./algorithms/SparseMatrix.h\
+ ./cartesian/Grid_cartesian_base.h\
+ ./cartesian/Grid_cartesian_full.h\
+ ./cartesian/Grid_cartesian_red_black.h\
+ ./communicator/Grid_communicator_base.h\
+ ./cshift/Grid_cshift_common.h\
+ ./cshift/Grid_cshift_mpi.h\
+ ./cshift/Grid_cshift_none.h\
+ ./Grid.h\
+ ./Grid_algorithms.h\
+ ./Grid_aligned_allocator.h\
+ ./Grid_cartesian.h\
+ ./Grid_communicator.h\
+ ./Grid_comparison.h\
+ ./Grid_config.h\
+ ./Grid_cshift.h\
+ ./Grid_extract.h\
+ ./Grid_lattice.h\
+ ./Grid_math.h\
+ ./Grid_simd.h\
+ ./Grid_stencil.h\
+ ./Grid_threads.h\
+ ./lattice/Grid_lattice_arith.h\
+ ./lattice/Grid_lattice_base.h\
+ ./lattice/Grid_lattice_comparison.h\
+ ./lattice/Grid_lattice_conformable.h\
+ ./lattice/Grid_lattice_coordinate.h\
+ ./lattice/Grid_lattice_ET.h\
+ ./lattice/Grid_lattice_local.h\
+ ./lattice/Grid_lattice_overload.h\
+ ./lattice/Grid_lattice_peekpoke.h\
+ ./lattice/Grid_lattice_reality.h\
+ ./lattice/Grid_lattice_reduction.h\
+ ./lattice/Grid_lattice_rng.h\
+ ./lattice/Grid_lattice_trace.h\
+ ./lattice/Grid_lattice_transfer.h\
+ ./lattice/Grid_lattice_transpose.h\
+ ./lattice/Grid_lattice_where.h\
+ ./math/Grid_math_arith.h\
+ ./math/Grid_math_arith_add.h\
+ ./math/Grid_math_arith_mac.h\
+ ./math/Grid_math_arith_mul.h\
+ ./math/Grid_math_arith_scalar.h\
+ ./math/Grid_math_arith_sub.h\
+ ./math/Grid_math_inner.h\
+ ./math/Grid_math_outer.h\
+ ./math/Grid_math_peek.h\
+ ./math/Grid_math_poke.h\
+ ./math/Grid_math_reality.h\
+ ./math/Grid_math_tensors.h\
+ ./math/Grid_math_trace.h\
+ ./math/Grid_math_traits.h\
+ ./math/Grid_math_transpose.h\
+ ./parallelIO/GridNerscIO.h\
+ ./qcd/action/Actions.h\
+ ./qcd/action/fermion/FermionAction.h\
+ ./qcd/action/fermion/FiveDimWilsonFermion.h\
+ ./qcd/action/fermion/WilsonCompressor.h\
+ ./qcd/action/fermion/WilsonFermion.h\
+ ./qcd/action/fermion/WilsonKernels.h\
+ ./qcd/Dirac.h\
+ ./qcd/QCD.h\
+ ./qcd/TwoSpinor.h\
+ ./qcd/FermionAction.h\
+ ./simd/Grid_avx.h\
+ ./simd/Grid_avx512.h\
+ ./simd/Grid_qpx.h\
+ ./simd/Grid_sse4.h\
+ ./simd/Grid_vector_types.h\
+ ./simd/Old/Grid_vComplexD.h\
+ ./simd/Old/Grid_vComplexF.h\
+ ./simd/Old/Grid_vInteger.h\
+ ./simd/Old/Grid_vRealD.h\
+ ./simd/Old/Grid_vRealF.h\
+ ./stencil/Grid_lebesgue.h
diff --git a/lib/algorithms/SparseMatrix.h b/lib/algorithms/SparseMatrix.h
index 7bfe959b..9c955e9a 100644
--- a/lib/algorithms/SparseMatrix.h
+++ b/lib/algorithms/SparseMatrix.h
@@ -10,7 +10,7 @@ namespace Grid {
/////////////////////////////////////////////////////////////////////////////////////////////
template<class Field> class SparseMatrixBase {
public:
- GridBase *_grid;
+ virtual GridBase *Grid(void) =0;
// Full checkerboar operations
virtual RealD M (const Field &in, Field &out)=0;
virtual RealD Mdag (const Field &in, Field &out)=0;
@@ -19,7 +19,6 @@ namespace Grid {
ni=M(in,tmp);
no=Mdag(tmp,out);
}
- SparseMatrixBase(GridBase *grid) : _grid(grid) {};
};
/////////////////////////////////////////////////////////////////////////////////////////////
@@ -27,7 +26,7 @@ namespace Grid {
/////////////////////////////////////////////////////////////////////////////////////////////
template<class Field> class CheckerBoardedSparseMatrixBase : public SparseMatrixBase<Field> {
public:
- GridBase *_cbgrid;
+ virtual GridBase *RedBlackGrid(void)=0;
// half checkerboard operaions
virtual void Meooe (const Field &in, Field &out)=0;
virtual void Mooee (const Field &in, Field &out)=0;
@@ -62,9 +61,7 @@ namespace Grid {
Field tmp(in._grid);
ni=Mpc(in,tmp);
no=MpcDag(tmp,out);
- // std::cout<<"MpcDagMpc "<<ni<<" "<<no<<std::endl;
}
- CheckerBoardedSparseMatrixBase(GridBase *grid,GridBase *cbgrid) : SparseMatrixBase<Field>(grid), _cbgrid(cbgrid) {};
};
}
diff --git a/lib/algorithms/iterative/SchurRedBlack.h b/lib/algorithms/iterative/SchurRedBlack.h
index d3a76d7f..109f554b 100644
--- a/lib/algorithms/iterative/SchurRedBlack.h
+++ b/lib/algorithms/iterative/SchurRedBlack.h
@@ -60,8 +60,8 @@ namespace Grid {
// FIXME CGdiagonalMee not implemented virtual function
// FIXME use CBfactorise to control schur decomp
- GridBase *grid = _Matrix._cbgrid;
- GridBase *fgrid= _Matrix._grid;
+ GridBase *grid = _Matrix.RedBlackGrid();
+ GridBase *fgrid= _Matrix.Grid();
Field src_e(grid);
Field src_o(grid);
diff --git a/lib/cartesian/Grid_cartesian_base.h b/lib/cartesian/Grid_cartesian_base.h
index a74773f8..e93125c1 100644
--- a/lib/cartesian/Grid_cartesian_base.h
+++ b/lib/cartesian/Grid_cartesian_base.h
@@ -52,16 +52,13 @@ public:
////////////////////////////////////////////////////////////////
virtual int CheckerBoarded(int dim)=0;
virtual int CheckerBoard(std::vector<int> site)=0;
- virtual int CheckerBoardDestination(int source_cb,int shift)=0;
+ virtual int CheckerBoardDestination(int source_cb,int shift,int dim)=0;
virtual int CheckerBoardShift(int source_cb,int dim,int shift,int osite)=0;
- inline int CheckerBoardFromOindex (int Oindex){
+ virtual int CheckerBoardShiftForCB(int source_cb,int dim,int shift,int cb)=0;
+ int CheckerBoardFromOindex (int Oindex){
std::vector<int> ocoor;
oCoorFromOindex(ocoor,Oindex);
- int ss=0;
- for(int d=0;d<_ndimension;d++){
- ss=ss+ocoor[d];
- }
- return ss&0x1;
+ return CheckerBoard(ocoor);
}
//////////////////////////////////////////////////////////////////////////////////////////////
diff --git a/lib/cartesian/Grid_cartesian_full.h b/lib/cartesian/Grid_cartesian_full.h
index 73bd08b3..330bbfaf 100644
--- a/lib/cartesian/Grid_cartesian_full.h
+++ b/lib/cartesian/Grid_cartesian_full.h
@@ -18,11 +18,14 @@ public:
virtual int CheckerBoard(std::vector<int> site){
return 0;
}
- virtual int CheckerBoardDestination(int cb,int shift){
+ virtual int CheckerBoardDestination(int cb,int shift,int dim){
return 0;
}
+ virtual int CheckerBoardShiftForCB(int source_cb,int dim,int shift, int ocb){
+ return shift;
+ }
virtual int CheckerBoardShift(int source_cb,int dim,int shift, int osite){
- return shift;
+ return shift;
}
GridCartesian(std::vector<int> &dimensions,
std::vector<int> &simd_layout,
diff --git a/lib/cartesian/Grid_cartesian_red_black.h b/lib/cartesian/Grid_cartesian_red_black.h
index ff2d3ba8..ace36edb 100644
--- a/lib/cartesian/Grid_cartesian_red_black.h
+++ b/lib/cartesian/Grid_cartesian_red_black.h
@@ -8,6 +8,10 @@ namespace Grid {
static const int CbBlack=1;
static const int Even =CbRed;
static const int Odd =CbBlack;
+
+ // Perhaps these are misplaced and
+ // should be in sparse matrix.
+ // Also should make these a named enum type
static const int DaggerNo=0;
static const int DaggerYes=1;
@@ -15,116 +19,174 @@ namespace Grid {
class GridRedBlackCartesian : public GridBase
{
public:
+ std::vector<int> _checker_dim_mask;
+ int _checker_dim;
+
virtual int CheckerBoarded(int dim){
- if( dim==0) return 1;
+ if( dim==_checker_dim) return 1;
else return 0;
}
virtual int CheckerBoard(std::vector<int> site){
- return (site[0]+site[1]+site[2]+site[3])&0x1;
+ int linear=0;
+ assert(site.size()==_ndimension);
+ for(int d=0;d<_ndimension;d++){
+ if(_checker_dim_mask[d])
+ linear=linear+site[d];
+ }
+ return (linear&0x1);
}
+
// Depending on the cb of site, we toggle source cb.
// for block #b, element #e = (b, e)
// we need
- virtual int CheckerBoardShift(int source_cb,int dim,int shift,int osite){
+ virtual int CheckerBoardShiftForCB(int source_cb,int dim,int shift,int ocb){
+ if(dim != _checker_dim) return shift;
- if(dim != 0) return shift;
-
- int fulldim =_fdimensions[0];
+ int fulldim =_fdimensions[dim];
shift = (shift+fulldim)%fulldim;
// Probably faster with table lookup;
// or by looping over x,y,z and multiply rather than computing checkerboard.
- int ocb=CheckerBoardFromOindex(osite);
if ( (source_cb+ocb)&1 ) {
+
return (shift)/2;
} else {
return (shift+1)/2;
}
}
+ virtual int CheckerBoardShift(int source_cb,int dim,int shift,int osite){
- virtual int CheckerBoardDestination(int source_cb,int shift){
- if ((shift+_fdimensions[0])&0x1) {
+ if(dim != _checker_dim) return shift;
+
+ int ocb=CheckerBoardFromOindex(osite);
+
+ return CheckerBoardShiftForCB(source_cb,dim,shift,ocb);
+ }
+
+ virtual int CheckerBoardDestination(int source_cb,int shift,int dim){
+ if ( _checker_dim_mask[dim] ) {
+ // If _fdimensions[checker_dim] is odd, then shifting by 1 in other dims
+ // does NOT cause a parity hop.
+ int add=(dim==_checker_dim) ? 0 : _fdimensions[_checker_dim];
+ if ( (shift+add) &0x1) {
return 1-source_cb;
} else {
return source_cb;
}
+ } else {
+ return source_cb;
+
+ }
};
GridRedBlackCartesian(GridBase *base) : GridRedBlackCartesian(base->_fdimensions,base->_simd_layout,base->_processors) {};
GridRedBlackCartesian(std::vector<int> &dimensions,
std::vector<int> &simd_layout,
- std::vector<int> &processor_grid ) : GridBase(processor_grid)
+ std::vector<int> &processor_grid,
+ std::vector<int> &checker_dim_mask,
+ int checker_dim
+ ) : GridBase(processor_grid)
+ {
+ Init(dimensions,simd_layout,processor_grid,checker_dim_mask,checker_dim);
+ }
+ GridRedBlackCartesian(std::vector<int> &dimensions,
+ std::vector<int> &simd_layout,
+ std::vector<int> &processor_grid) : GridBase(processor_grid)
+ {
+ std::vector<int> checker_dim_mask(dimensions.size(),1);
+ Init(dimensions,simd_layout,processor_grid,checker_dim_mask,0);
+ }
+ void Init(std::vector<int> &dimensions,
+ std::vector<int> &simd_layout,
+ std::vector<int> &processor_grid,
+ std::vector<int> &checker_dim_mask,
+ int checker_dim)
{
///////////////////////
// Grid information
///////////////////////
- _ndimension = dimensions.size();
-
- _fdimensions.resize(_ndimension);
- _gdimensions.resize(_ndimension);
- _ldimensions.resize(_ndimension);
- _rdimensions.resize(_ndimension);
- _simd_layout.resize(_ndimension);
-
- _ostride.resize(_ndimension);
- _istride.resize(_ndimension);
-
- _fsites = _gsites = _osites = _isites = 1;
+ _checker_dim = checker_dim;
+ assert(checker_dim_mask[checker_dim]==1);
+ _ndimension = dimensions.size();
+ assert(checker_dim_mask.size()==_ndimension);
+ assert(processor_grid.size()==_ndimension);
+ assert(simd_layout.size()==_ndimension);
+
+ _fdimensions.resize(_ndimension);
+ _gdimensions.resize(_ndimension);
+ _ldimensions.resize(_ndimension);
+ _rdimensions.resize(_ndimension);
+ _simd_layout.resize(_ndimension);
+
+ _ostride.resize(_ndimension);
+ _istride.resize(_ndimension);
+
+ _fsites = _gsites = _osites = _isites = 1;
+
+ _checker_dim_mask=checker_dim_mask;
- for(int d=0;d<_ndimension;d++){
- _fdimensions[d] = dimensions[d];
- _gdimensions[d] = _fdimensions[d];
- _fsites = _fsites * _fdimensions[d];
- _gsites = _gsites * _gdimensions[d];
-
- if (d==0) _gdimensions[0] = _gdimensions[0]/2; // Remove a checkerboard
- _ldimensions[d] = _gdimensions[d]/_processors[d];
-
- // Use a reduced simd grid
- _simd_layout[d] = simd_layout[d];
- _rdimensions[d]= _ldimensions[d]/_simd_layout[d];
-
- _osites *= _rdimensions[d];
- _isites *= _simd_layout[d];
-
- // Addressing support
- if ( d==0 ) {
- _ostride[d] = 1;
- _istride[d] = 1;
- } else {
- _ostride[d] = _ostride[d-1]*_rdimensions[d-1];
- _istride[d] = _istride[d-1]*_simd_layout[d-1];
- }
- }
-
- ////////////////////////////////////////////////////////////////////////////////////////////
- // subplane information
- ////////////////////////////////////////////////////////////////////////////////////////////
- _slice_block.resize(_ndimension);
- _slice_stride.resize(_ndimension);
- _slice_nblock.resize(_ndimension);
+ for(int d=0;d<_ndimension;d++){
+ _fdimensions[d] = dimensions[d];
+ _gdimensions[d] = _fdimensions[d];
+ _fsites = _fsites * _fdimensions[d];
+ _gsites = _gsites * _gdimensions[d];
- int block =1;
- int nblock=1;
- for(int d=0;d<_ndimension;d++) nblock*=_rdimensions[d];
-
- for(int d=0;d<_ndimension;d++){
- nblock/=_rdimensions[d];
- _slice_block[d] =block;
- _slice_stride[d]=_ostride[d]*_rdimensions[d];
- _slice_nblock[d]=nblock;
- block = block*_rdimensions[d];
- }
+ if (d==_checker_dim) {
+ _gdimensions[d] = _gdimensions[d]/2; // Remove a checkerboard
+ }
+ _ldimensions[d] = _gdimensions[d]/_processors[d];
+
+ // Use a reduced simd grid
+ _simd_layout[d] = simd_layout[d];
+ _rdimensions[d]= _ldimensions[d]/_simd_layout[d];
+
+ _osites *= _rdimensions[d];
+ _isites *= _simd_layout[d];
+
+ // Addressing support
+ if ( d==0 ) {
+ _ostride[d] = 1;
+ _istride[d] = 1;
+ } else {
+ _ostride[d] = _ostride[d-1]*_rdimensions[d-1];
+ _istride[d] = _istride[d-1]*_simd_layout[d-1];
+ }
+ }
+ ////////////////////////////////////////////////////////////////////////////////////////////
+ // subplane information
+ ////////////////////////////////////////////////////////////////////////////////////////////
+ _slice_block.resize(_ndimension);
+ _slice_stride.resize(_ndimension);
+ _slice_nblock.resize(_ndimension);
+
+ int block =1;
+ int nblock=1;
+ for(int d=0;d<_ndimension;d++) nblock*=_rdimensions[d];
+
+ for(int d=0;d<_ndimension;d++){
+ nblock/=_rdimensions[d];
+ _slice_block[d] =block;
+ _slice_stride[d]=_ostride[d]*_rdimensions[d];
+ _slice_nblock[d]=nblock;
+ block = block*_rdimensions[d];
+ }
+
};
protected:
virtual int oIndex(std::vector<int> &coor)
{
- int idx=_ostride[0]*((coor[0]/2)%_rdimensions[0]);
- for(int d=1;d<_ndimension;d++) idx+=_ostride[d]*(coor[d]%_rdimensions[d]);
+ int idx=0;
+ for(int d=0;d<_ndimension;d++) {
+ if( d==_checker_dim ) {
+ idx+=_ostride[d]*((coor[d]/2)%_rdimensions[d]);
+ } else {
+ idx+=_ostride[d]*(coor[d]%_rdimensions[d]);
+ }
+ }
return idx;
};
diff --git a/lib/cshift/Grid_cshift_common.h b/lib/cshift/Grid_cshift_common.h
index c369fe1c..06e812d9 100644
--- a/lib/cshift/Grid_cshift_common.h
+++ b/lib/cshift/Grid_cshift_common.h
@@ -175,7 +175,6 @@ PARALLEL_NESTED_LOOP2
if ( ocb&cbmask ) {
//lhs._odata[lo+o]=rhs._odata[ro+o];
vstream(lhs._odata[lo+o],rhs._odata[ro+o]);
-
}
}
@@ -217,8 +216,8 @@ template<class vobj> void Cshift_local(Lattice<vobj>& ret,Lattice<vobj> &rhs,int
{
int sshift[2];
- sshift[0] = rhs._grid->CheckerBoardShift(rhs.checkerboard,dimension,shift,0);
- sshift[1] = rhs._grid->CheckerBoardShift(rhs.checkerboard,dimension,shift,1);
+ sshift[0] = rhs._grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,Even);
+ sshift[1] = rhs._grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,Odd);
if ( sshift[0] == sshift[1] ) {
Cshift_local(ret,rhs,dimension,shift,0x3);
@@ -239,8 +238,7 @@ template<class vobj> Lattice<vobj> Cshift_local(Lattice<vobj> &ret,Lattice<vobj>
// Map to always positive shift modulo global full dimension.
shift = (shift+fd)%fd;
- ret.checkerboard = grid->CheckerBoardDestination(rhs.checkerboard,shift);
-
+ ret.checkerboard = grid->CheckerBoardDestination(rhs.checkerboard,shift,dimension);
// the permute type
int permute_dim =grid->PermuteDim(dimension);
int permute_type=grid->PermuteType(dimension);
@@ -250,11 +248,11 @@ template<class vobj> Lattice<vobj> Cshift_local(Lattice<vobj> &ret,Lattice<vobj>
int o = 0;
int bo = x * grid->_ostride[dimension];
- int cb= (cbmask==0x2)? 1 : 0;
+ int cb= (cbmask==0x2)? Odd : Even;
- int sshift = grid->CheckerBoardShift(rhs.checkerboard,dimension,shift,cb);
+ int sshift = grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,cb);
int sx = (x+sshift)%rd;
-
+
int permute_slice=0;
if(permute_dim){
int wrap = sshift/rd;
diff --git a/lib/cshift/Grid_cshift_mpi.h b/lib/cshift/Grid_cshift_mpi.h
index 569017d8..8c0badcd 100644
--- a/lib/cshift/Grid_cshift_mpi.h
+++ b/lib/cshift/Grid_cshift_mpi.h
@@ -39,8 +39,8 @@ template<class vobj> void Cshift_comms(Lattice<vobj>& ret,Lattice<vobj> &rhs,int
{
int sshift[2];
- sshift[0] = rhs._grid->CheckerBoardShift(rhs.checkerboard,dimension,shift,0);
- sshift[1] = rhs._grid->CheckerBoardShift(rhs.checkerboard,dimension,shift,1);
+ sshift[0] = rhs._grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,Even);
+ sshift[1] = rhs._grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,Odd);
if ( sshift[0] == sshift[1] ) {
Cshift_comms(ret,rhs,dimension,shift,0x3);
@@ -54,8 +54,8 @@ template<class vobj> void Cshift_comms_simd(Lattice<vobj>& ret,Lattice<vobj> &rh
{
int sshift[2];
- sshift[0] = rhs._grid->CheckerBoardShift(rhs.checkerboard,dimension,shift,0);
- sshift[1] = rhs._grid->CheckerBoardShift(rhs.checkerboard,dimension,shift,1);
+ sshift[0] = rhs._grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,Even);
+ sshift[1] = rhs._grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,Odd);
if ( sshift[0] == sshift[1] ) {
Cshift_comms_simd(ret,rhs,dimension,shift,0x3);
@@ -87,8 +87,8 @@ template<class vobj> void Cshift_comms(Lattice<vobj> &ret,Lattice<vobj> &rhs,int
std::vector<vobj,alignedAllocator<vobj> > send_buf(buffer_size);
std::vector<vobj,alignedAllocator<vobj> > recv_buf(buffer_size);
- int cb= (cbmask==0x2)? 1 : 0;
- int sshift= rhs._grid->CheckerBoardShift(rhs.checkerboard,dimension,shift,cb);
+ int cb= (cbmask==0x2)? Odd : Even;
+ int sshift= rhs._grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,cb);
for(int x=0;x<rd;x++){
@@ -162,8 +162,8 @@ template<class vobj> void Cshift_comms_simd(Lattice<vobj> &ret,Lattice<vobj> &r
///////////////////////////////////////////
// Work out what to send where
///////////////////////////////////////////
- int cb = (cbmask==0x2)? 1 : 0;
- int sshift= grid->CheckerBoardShift(rhs.checkerboard,dimension,shift,cb);
+ int cb = (cbmask==0x2)? Odd : Even;
+ int sshift= grid->CheckerBoardShiftForCB(rhs.checkerboard,dimension,shift,cb);
// loop over outer coord planes orthog to dim
for(int x=0;x<rd;x++){
diff --git a/lib/cshift/Grid_cshift_none.h b/lib/cshift/Grid_cshift_none.h
index 3485a63e..be9467b1 100644
--- a/lib/cshift/Grid_cshift_none.h
+++ b/lib/cshift/Grid_cshift_none.h
@@ -4,7 +4,7 @@ namespace Grid {
template<class vobj> Lattice<vobj> Cshift(Lattice<vobj> &rhs,int dimension,int shift)
{
Lattice<vobj> ret(rhs._grid);
- ret.checkerboard = rhs._grid->CheckerBoardDestination(rhs.checkerboard,shift);
+ ret.checkerboard = rhs._grid->CheckerBoardDestination(rhs.checkerboard,shift,dimension);
Cshift_local(ret,rhs,dimension,shift);
return ret;
}
diff --git a/lib/lattice/Grid_lattice_peekpoke.h b/lib/lattice/Grid_lattice_peekpoke.h
index 7bffdbb8..9416226a 100644
--- a/lib/lattice/Grid_lattice_peekpoke.h
+++ b/lib/lattice/Grid_lattice_peekpoke.h
@@ -116,7 +116,7 @@ PARALLEL_FOR_LOOP
int Nsimd = grid->Nsimd();
- assert( l.checkerboard== l._grid->CheckerBoard(site));
+ assert( l.checkerboard == l._grid->CheckerBoard(site));
assert( sizeof(sobj)*Nsimd == sizeof(vobj));
int rank,odx,idx;
diff --git a/lib/lattice/Grid_lattice_transfer.h b/lib/lattice/Grid_lattice_transfer.h
index 27ae27e1..f0046b0e 100644
--- a/lib/lattice/Grid_lattice_transfer.h
+++ b/lib/lattice/Grid_lattice_transfer.h
@@ -32,7 +32,6 @@ PARALLEL_FOR_LOOP
cbos=half._grid->CheckerBoard(coor);
if (cbos==cb) {
-
half._odata[ssh] = full._odata[ss];
ssh++;
}
@@ -45,7 +44,7 @@ PARALLEL_FOR_LOOP
for(int ss=0;ss<full._grid->oSites();ss++){
std::vector<int> coor;
int cbos;
-
+
full._grid->oCoorFromOindex(coor,ss);
cbos=half._grid->CheckerBoard(coor);
diff --git a/lib/qcd/Grid_qcd_dirac.cc b/lib/qcd/Dirac.cc
similarity index 100%
rename from lib/qcd/Grid_qcd_dirac.cc
rename to lib/qcd/Dirac.cc
diff --git a/lib/qcd/Grid_qcd_dirac.h b/lib/qcd/Dirac.h
similarity index 100%
rename from lib/qcd/Grid_qcd_dirac.h
rename to lib/qcd/Dirac.h
diff --git a/lib/qcd/Grid_qcd_wilson_dop.cc b/lib/qcd/Grid_qcd_wilson_dop.cc
deleted file mode 100644
index 9a3f5f6a..00000000
--- a/lib/qcd/Grid_qcd_wilson_dop.cc
+++ /dev/null
@@ -1,217 +0,0 @@
-#include <Grid.h>
-
-namespace Grid {
-namespace QCD {
-
-const std::vector<int> WilsonMatrix::directions ({0,1,2,3, 0, 1, 2, 3});
-const std::vector<int> WilsonMatrix::displacements({1,1,1,1,-1,-1,-1,-1});
-
- int WilsonMatrix::HandOptDslash;
-
- class WilsonCompressor {
- public:
- int mu;
- int dag;
-
- WilsonCompressor(int _dag){
- mu=0;
- dag=_dag;
- assert((dag==0)||(dag==1));
- }
- void Point(int p) {
- mu=p;
- };
-
- vHalfSpinColourVector operator () (const vSpinColourVector &in)
- {
- vHalfSpinColourVector ret;
- int mudag=mu;
- if (dag) {
- mudag=(mu+Nd)%(2*Nd);
- }
- switch(mudag) {
- case Xp:
- spProjXp(ret,in);
- break;
- case Yp:
- spProjYp(ret,in);
- break;
- case Zp:
- spProjZp(ret,in);
- break;
- case Tp:
- spProjTp(ret,in);
- break;
- case Xm:
- spProjXm(ret,in);
- break;
- case Ym:
- spProjYm(ret,in);
- break;
- case Zm:
- spProjZm(ret,in);
- break;
- case Tm:
- spProjTm(ret,in);
- break;
- default:
- assert(0);
- break;
- }
- return ret;
- }
- };
-
- WilsonMatrix::WilsonMatrix(LatticeGaugeField &_Umu,GridCartesian &Fgrid,GridRedBlackCartesian &Hgrid, double _mass) :
- CheckerBoardedSparseMatrixBase<LatticeFermion>(&Fgrid,&Hgrid),
- Stencil ( _grid,npoint,Even,directions,displacements),
- StencilEven(_cbgrid,npoint,Even,directions,displacements), // source is Even
- StencilOdd (_cbgrid,npoint,Odd ,directions,displacements), // source is Odd
- mass(_mass),
- Umu(_grid),
- UmuEven(_cbgrid),
- UmuOdd (_cbgrid)
- {
- // Allocate the required comms buffer
- comm_buf.resize(Stencil._unified_buffer_size); // this is always big enough to contain EO
-
- DoubleStore(Umu,_Umu);
- pickCheckerboard(Even,UmuEven,Umu);
- pickCheckerboard(Odd ,UmuOdd,Umu);
- }
-
-void WilsonMatrix::DoubleStore(LatticeDoubledGaugeField &Uds,const LatticeGaugeField &Umu)
-{
- LatticeColourMatrix U(_grid);
-
- 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);
- }
-}
-
-RealD WilsonMatrix::M(const LatticeFermion &in, LatticeFermion &out)
-{
- out.checkerboard=in.checkerboard;
- Dhop(in,out,DaggerNo);
- out = (4+mass)*in - 0.5*out ; // FIXME : axpby_norm! fusion fun
- return norm2(out);
-}
-RealD WilsonMatrix::Mdag(const LatticeFermion &in, LatticeFermion &out)
-{
- out.checkerboard=in.checkerboard;
- Dhop(in,out,DaggerYes);
- out = (4+mass)*in - 0.5*out ; // FIXME : axpby_norm! fusion fun
- return norm2(out);
-}
-
-void WilsonMatrix::Meooe(const LatticeFermion &in, LatticeFermion &out)
-{
- if ( in.checkerboard == Odd ) {
- DhopEO(in,out,DaggerNo);
- } else {
- DhopOE(in,out,DaggerNo);
- }
- out = (-0.5)*out; // FIXME : scale factor in Dhop
-}
-void WilsonMatrix::MeooeDag(const LatticeFermion &in, LatticeFermion &out)
-{
- if ( in.checkerboard == Odd ) {
- DhopEO(in,out,DaggerYes);
- } else {
- DhopOE(in,out,DaggerYes);
- }
- out = (-0.5)*out; // FIXME : scale factor in Dhop
-}
-void WilsonMatrix::Mooee(const LatticeFermion &in, LatticeFermion &out)
-{
- out.checkerboard = in.checkerboard;
- out = (4.0+mass)*in;
- return ;
-}
-void WilsonMatrix::MooeeDag(const LatticeFermion &in, LatticeFermion &out)
-{
- out.checkerboard = in.checkerboard;
- Mooee(in,out);
-}
-void WilsonMatrix::MooeeInv(const LatticeFermion &in, LatticeFermion &out)
-{
- out.checkerboard = in.checkerboard;
- out = (1.0/(4.0+mass))*in;
- return ;
-}
-void WilsonMatrix::MooeeInvDag(const LatticeFermion &in, LatticeFermion &out)
-{
- out.checkerboard = in.checkerboard;
- MooeeInv(in,out);
-}
-
-void WilsonMatrix::DhopInternal(CartesianStencil & st,LatticeDoubledGaugeField & U,
- const LatticeFermion &in, LatticeFermion &out,int dag)
-{
- assert((dag==DaggerNo) ||(dag==DaggerYes));
- WilsonCompressor compressor(dag);
- st.HaloExchange<vSpinColourVector,vHalfSpinColourVector,WilsonCompressor>(in,comm_buf,compressor);
-
- if ( dag == DaggerYes ) {
- if( HandOptDslash ) {
-PARALLEL_FOR_LOOP
- for(int sss=0;sss<in._grid->oSites();sss++){
- DiracOptHand::DhopSiteDag(st,U,comm_buf,sss,in,out);
- }
- } else {
-PARALLEL_FOR_LOOP
- for(int sss=0;sss<in._grid->oSites();sss++){
- DiracOpt::DhopSiteDag(st,U,comm_buf,sss,in,out);
- }
- }
- } else {
- if( HandOptDslash ) {
-PARALLEL_FOR_LOOP
- for(int sss=0;sss<in._grid->oSites();sss++){
- DiracOptHand::DhopSite(st,U,comm_buf,sss,in,out);
- }
- } else {
-PARALLEL_FOR_LOOP
- for(int sss=0;sss<in._grid->oSites();sss++){
- DiracOpt::DhopSite(st,U,comm_buf,sss,in,out);
- }
- }
- }
-}
-void WilsonMatrix::DhopOE(const LatticeFermion &in, LatticeFermion &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,UmuOdd,in,out,dag);
-}
-void WilsonMatrix::DhopEO(const LatticeFermion &in, LatticeFermion &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,UmuEven,in,out,dag);
-}
-void WilsonMatrix::Dhop(const LatticeFermion &in, LatticeFermion &out,int dag)
-{
- conformable(in._grid,_grid); // verifies full grid
- conformable(in._grid,out._grid);
-
- out.checkerboard = in.checkerboard;
-
- DhopInternal(Stencil,Umu,in,out,dag);
-}
-
-}}
-
-
-
diff --git a/lib/qcd/Grid_qcd_wilson_dop.h b/lib/qcd/Grid_qcd_wilson_dop.h
deleted file mode 100644
index 87418603..00000000
--- a/lib/qcd/Grid_qcd_wilson_dop.h
+++ /dev/null
@@ -1,105 +0,0 @@
-#ifndef GRID_QCD_WILSON_DOP_H
-#define GRID_QCD_WILSON_DOP_H
-
-
-namespace Grid {
-
- namespace QCD {
-
- // Should be in header?
- const int Xp = 0;
- const int Yp = 1;
- const int Zp = 2;
- const int Tp = 3;
- const int Xm = 4;
- const int Ym = 5;
- const int Zm = 6;
- const int Tm = 7;
-
- class WilsonMatrix : public CheckerBoardedSparseMatrixBase<LatticeFermion>
- {
- //NB r=1;
- public:
- static int HandOptDslash;
-
- double mass;
- // GridBase * grid; // Inherited
- // GridBase * cbgrid;
-
- //Defines the stencils for even and odd
- CartesianStencil Stencil;
- CartesianStencil StencilEven;
- CartesianStencil StencilOdd;
-
- // Copy of the gauge field , with even and odd subsets
- LatticeDoubledGaugeField Umu;
- LatticeDoubledGaugeField UmuEven;
- LatticeDoubledGaugeField UmuOdd;
-
- static const int npoint=8;
- static const std::vector<int> directions ;
- static const std::vector<int> displacements;
- static const int Xp,Xm,Yp,Ym,Zp,Zm,Tp,Tm;
-
- // Comms buffer
- std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > comm_buf;
-
- // Constructor
- WilsonMatrix(LatticeGaugeField &_Umu,GridCartesian &Fgrid,GridRedBlackCartesian &Hgrid,double _mass);
-
- // DoubleStore
- void DoubleStore(LatticeDoubledGaugeField &Uds,const LatticeGaugeField &Umu);
-
- // override multiply
- virtual RealD M (const LatticeFermion &in, LatticeFermion &out);
- virtual RealD Mdag (const LatticeFermion &in, LatticeFermion &out);
-
- // half checkerboard operaions
- virtual void Meooe (const LatticeFermion &in, LatticeFermion &out);
- virtual void MeooeDag (const LatticeFermion &in, LatticeFermion &out);
- virtual void Mooee (const LatticeFermion &in, LatticeFermion &out);
- virtual void MooeeDag (const LatticeFermion &in, LatticeFermion &out);
- virtual void MooeeInv (const LatticeFermion &in, LatticeFermion &out);
- virtual void MooeeInvDag (const LatticeFermion &in, LatticeFermion &out);
-
- // non-hermitian hopping term; half cb or both
- void Dhop (const LatticeFermion &in, LatticeFermion &out,int dag);
- void DhopOE(const LatticeFermion &in, LatticeFermion &out,int dag);
- void DhopEO(const LatticeFermion &in, LatticeFermion &out,int dag);
- void DhopInternal(CartesianStencil & st,LatticeDoubledGaugeField &U,
- const LatticeFermion &in, LatticeFermion &out,int dag);
-
- typedef iScalar<iMatrix<vComplex, Nc> > matrix;
-
-
- };
-
-
- class DiracOpt {
- public:
- // These ones will need to be package intelligently. WilsonType base class
- // for use by DWF etc..
- static void DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
- std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > &buf,
- int ss,const LatticeFermion &in, LatticeFermion &out);
- static void DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
- std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > &buf,
- int ss,const LatticeFermion &in, LatticeFermion &out);
-
- };
- class DiracOptHand {
- public:
- // These ones will need to be package intelligently. WilsonType base class
- // for use by DWF etc..
- static void DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
- std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > &buf,
- int ss,const LatticeFermion &in, LatticeFermion &out);
- static void DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
- std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > &buf,
- int ss,const LatticeFermion &in, LatticeFermion &out);
-
- };
-
- }
-}
-#endif
diff --git a/lib/qcd/Grid_qcd.h b/lib/qcd/QCD.h
similarity index 97%
rename from lib/qcd/Grid_qcd.h
rename to lib/qcd/QCD.h
index 959f3529..7c45eb23 100644
--- a/lib/qcd/Grid_qcd.h
+++ b/lib/qcd/QCD.h
@@ -4,6 +4,16 @@ namespace Grid{
namespace QCD {
+
+ static const int Xp = 0;
+ static const int Yp = 1;
+ static const int Zp = 2;
+ static const int Tp = 3;
+ static const int Xm = 4;
+ static const int Ym = 5;
+ static const int Zm = 6;
+ static const int Tm = 7;
+
static const int Nc=3;
static const int Ns=4;
static const int Nd=4;
@@ -297,9 +307,8 @@ namespace QCD {
} //namespace QCD
} // Grid
-#include <qcd/Grid_qcd_dirac.h>
-#include <qcd/Grid_qcd_2spinor.h>
-//#include <qcd/Grid_qcd_pauli.h>
-#include <qcd/Grid_qcd_wilson_dop.h>
+#include <qcd/Dirac.h>
+#include <qcd/TwoSpinor.h>
+#include <qcd/action/Actions.h>
#endif
diff --git a/lib/qcd/Grid_qcd_2spinor.h b/lib/qcd/TwoSpinor.h
similarity index 100%
rename from lib/qcd/Grid_qcd_2spinor.h
rename to lib/qcd/TwoSpinor.h
diff --git a/lib/qcd/action/Actions.h b/lib/qcd/action/Actions.h
new file mode 100644
index 00000000..c4e8a2f0
--- /dev/null
+++ b/lib/qcd/action/Actions.h
@@ -0,0 +1,10 @@
+#ifndef GRID_QCD_ACTIONS_H
+#define GRID_QCD_ACTIONS_H
+
+#include <qcd/action/fermion/FermionAction.h>
+#include <qcd/action/fermion/WilsonCompressor.h>
+#include <qcd/action/fermion/WilsonKernels.h>
+#include <qcd/action/fermion/WilsonFermion.h>
+#include <qcd/action/fermion/FiveDimWilsonFermion.h>
+
+#endif
diff --git a/lib/qcd/action/fermion/FermionAction.h b/lib/qcd/action/fermion/FermionAction.h
new file mode 100644
index 00000000..1b05174b
--- /dev/null
+++ b/lib/qcd/action/fermion/FermionAction.h
@@ -0,0 +1,47 @@
+#ifndef GRID_QCD_WILSON_DOP_H
+#define GRID_QCD_WILSON_DOP_H
+
+namespace Grid {
+
+ namespace QCD {
+
+ //////////////////////////////////////////////////////////////////////////////
+ // Four component fermions
+ // Should type template the vector and gauge types
+ // Think about multiple representations
+ //////////////////////////////////////////////////////////////////////////////
+ template<class FermionField,class GaugeField>
+ class FermionAction : public CheckerBoardedSparseMatrixBase<FermionField>
+ {
+ public:
+
+ GridBase * Grid(void) { return FermionGrid(); }; // this is all the linalg routines need to know
+ GridBase * RedBlackGrid(void) { return FermionRedBlackGrid(); };
+
+ virtual GridBase *FermionGrid(void) =0;
+ virtual GridBase *FermionRedBlackGrid(void) =0;
+ virtual GridBase *GaugeGrid(void) =0;
+ virtual GridBase *GaugeRedBlackGrid(void) =0;
+
+ // override multiply
+ virtual RealD M (const FermionField &in, FermionField &out)=0;
+ virtual RealD Mdag (const FermionField &in, FermionField &out)=0;
+
+ // half checkerboard operaions
+ virtual void Meooe (const FermionField &in, FermionField &out)=0;
+ virtual void MeooeDag (const FermionField &in, FermionField &out)=0;
+ virtual void Mooee (const FermionField &in, FermionField &out)=0;
+ virtual void MooeeDag (const FermionField &in, FermionField &out)=0;
+ virtual void MooeeInv (const FermionField &in, FermionField &out)=0;
+ virtual void MooeeInvDag (const FermionField &in, FermionField &out)=0;
+
+ // non-hermitian hopping term; half cb or both
+ virtual void Dhop (const FermionField &in, FermionField &out,int dag)=0;
+ virtual void DhopOE(const FermionField &in, FermionField &out,int dag)=0;
+ virtual void DhopEO(const FermionField &in, FermionField &out,int dag)=0;
+
+ };
+
+ }
+}
+#endif
diff --git a/lib/qcd/action/fermion/FiveDimWilsonFermion.cc b/lib/qcd/action/fermion/FiveDimWilsonFermion.cc
new file mode 100644
index 00000000..43645899
--- /dev/null
+++ b/lib/qcd/action/fermion/FiveDimWilsonFermion.cc
@@ -0,0 +1,228 @@
+#include <Grid.h>
+
+namespace Grid {
+namespace QCD {
+
+ // S-direction is INNERMOST and takes no part in the parity.
+ const std::vector<int> FiveDimWilsonFermion::directions ({1,2,3,4, 1, 2, 3, 4});
+ const std::vector<int> FiveDimWilsonFermion::displacements({1,1,1,1,-1,-1,-1,-1});
+
+ int FiveDimWilsonFermion::HandOptDslash;
+
+ // 5d lattice for DWF.
+ FiveDimWilsonFermion::FiveDimWilsonFermion(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ double _mass) :
+ _FiveDimGrid(&FiveDimGrid),
+ _FiveDimRedBlackGrid(&FiveDimRedBlackGrid),
+ _FourDimGrid(&FourDimGrid),
+ _FourDimRedBlackGrid(&FourDimRedBlackGrid),
+ Stencil (_FiveDimGrid,npoint,Even,directions,displacements),
+ StencilEven(_FiveDimRedBlackGrid,npoint,Even,directions,displacements), // source is Even
+ StencilOdd (_FiveDimRedBlackGrid,npoint,Odd ,directions,displacements), // source is Odd
+ mass(_mass),
+ Umu(_FourDimGrid),
+ UmuEven(_FourDimRedBlackGrid),
+ UmuOdd (_FourDimRedBlackGrid),
+ Lebesgue(_FourDimGrid),
+ LebesgueEvenOdd(_FourDimRedBlackGrid)
+{
+ // some assertions
+ assert(FiveDimGrid._ndimension==5);
+ assert(FourDimGrid._ndimension==4);
+
+ assert(FiveDimRedBlackGrid._ndimension==5);
+ assert(FourDimRedBlackGrid._ndimension==4);
+
+ assert(FiveDimRedBlackGrid._checker_dim==1);
+
+ // Dimension zero of the five-d is the Ls direction
+ Ls=FiveDimGrid._fdimensions[0];
+ assert(FiveDimRedBlackGrid._fdimensions[0]==Ls);
+ assert(FiveDimRedBlackGrid._processors[0] ==1);
+ assert(FiveDimRedBlackGrid._simd_layout[0]==1);
+ assert(FiveDimGrid._processors[0] ==1);
+ assert(FiveDimGrid._simd_layout[0] ==1);
+
+ // Other dimensions must match the decomposition of the four-D fields
+ for(int d=0;d<4;d++){
+ assert(FourDimRedBlackGrid._fdimensions[d] ==FourDimGrid._fdimensions[d]);
+ assert(FiveDimRedBlackGrid._fdimensions[d+1]==FourDimGrid._fdimensions[d]);
+
+ assert(FourDimRedBlackGrid._processors[d] ==FourDimGrid._processors[d]);
+ assert(FiveDimRedBlackGrid._processors[d+1] ==FourDimGrid._processors[d]);
+
+ assert(FourDimRedBlackGrid._simd_layout[d] ==FourDimGrid._simd_layout[d]);
+ assert(FiveDimRedBlackGrid._simd_layout[d+1]==FourDimGrid._simd_layout[d]);
+
+ assert(FiveDimGrid._fdimensions[d+1] ==FourDimGrid._fdimensions[d]);
+ assert(FiveDimGrid._processors[d+1] ==FourDimGrid._processors[d]);
+ assert(FiveDimGrid._simd_layout[d+1] ==FourDimGrid._simd_layout[d]);
+ }
+
+ // Allocate the required comms buffer
+ comm_buf.resize(Stencil._unified_buffer_size); // this is always big enough to contain EO
+
+ DoubleStore(Umu,_Umu);
+ pickCheckerboard(Even,UmuEven,Umu);
+ pickCheckerboard(Odd ,UmuOdd,Umu);
+}
+void FiveDimWilsonFermion::DoubleStore(LatticeDoubledGaugeField &Uds,const LatticeGaugeField &Umu)
+{
+ conformable(Uds._grid,GaugeGrid());
+ conformable(Umu._grid,GaugeGrid());
+ LatticeColourMatrix 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);
+ }
+}
+
+RealD FiveDimWilsonFermion::M(const LatticeFermion &in, LatticeFermion &out)
+{
+ out.checkerboard=in.checkerboard;
+ Dhop(in,out,DaggerNo);
+ return axpy_norm(out,5.0-M5,in,out);
+}
+RealD FiveDimWilsonFermion::Mdag(const LatticeFermion &in, LatticeFermion &out)
+{
+ out.checkerboard=in.checkerboard;
+ Dhop(in,out,DaggerYes);
+ return axpy_norm(out,5.0-M5,in,out);
+}
+void FiveDimWilsonFermion::Meooe(const LatticeFermion &in, LatticeFermion &out)
+{
+ if ( in.checkerboard == Odd ) {
+ DhopEO(in,out,DaggerNo);
+ } else {
+ DhopOE(in,out,DaggerNo);
+ }
+}
+void FiveDimWilsonFermion::MeooeDag(const LatticeFermion &in, LatticeFermion &out)
+{
+ if ( in.checkerboard == Odd ) {
+ DhopEO(in,out,DaggerYes);
+ } else {
+ DhopOE(in,out,DaggerYes);
+ }
+}
+void FiveDimWilsonFermion::Mooee(const LatticeFermion &in, LatticeFermion &out)
+{
+ out.checkerboard = in.checkerboard;
+ out = (5.0-M5)*in;
+ return ;
+}
+void FiveDimWilsonFermion::MooeeDag(const LatticeFermion &in, LatticeFermion &out)
+{
+ out.checkerboard = in.checkerboard;
+ Mooee(in,out);
+}
+void FiveDimWilsonFermion::MooeeInv(const LatticeFermion &in, LatticeFermion &out)
+{
+ out.checkerboard = in.checkerboard;
+ out = (1.0/(5.0-M5))*in;
+ return ;
+}
+void FiveDimWilsonFermion::MooeeInvDag(const LatticeFermion &in, LatticeFermion &out)
+{
+ out.checkerboard = in.checkerboard;
+ MooeeInv(in,out);
+}
+void FiveDimWilsonFermion::DhopInternal(CartesianStencil & st, LebesgueOrder &lo,
+ LatticeDoubledGaugeField & U,
+ const LatticeFermion &in, LatticeFermion &out,int dag)
+{
+ assert((dag==DaggerNo) ||(dag==DaggerYes));
+
+ WilsonCompressor compressor(dag);
+
+ st.HaloExchange<vSpinColourVector,vHalfSpinColourVector,WilsonCompressor>(in,comm_buf,compressor);
+
+ // Dhop takes the 4d grid from U, and makes a 5d index for fermion
+ // Not loop ordering and data layout.
+ // Designed to create
+ // - per thread reuse in L1 cache for U
+ // - 8 linear access unit stride streams per thread for Fermion for hw prefetchable.
+ if ( dag == DaggerYes ) {
+ if( HandOptDslash ) {
+ for(int ss=0;ss<U._grid->oSites();ss++){
+ int sU=lo.Reorder(ss);
+PARALLEL_FOR_LOOP
+ for(int s=0;s<Ls;s++){
+ int sF = s+Ls*sU;
+ DiracOptHand::DhopSiteDag(st,U,comm_buf,sF,sU,in,out);
+ }
+ }
+ } else {
+ for(int ss=0;ss<U._grid->oSites();ss++){
+ int sU=lo.Reorder(ss);
+PARALLEL_FOR_LOOP
+ for(int s=0;s<Ls;s++){
+ int sF = s+Ls*sU;
+ DiracOpt::DhopSiteDag(st,U,comm_buf,sF,sU,in,out);
+ }
+ }
+ }
+ } else {
+ if( HandOptDslash ) {
+
+PARALLEL_FOR_LOOP
+ for(int ss=0;ss<U._grid->oSites();ss++){
+ int sU=lo.Reorder(ss);
+ for(int s=0;s<Ls;s++){
+ int sF = s+Ls*sU;
+ DiracOptHand::DhopSite(st,U,comm_buf,sF,sU,in,out);
+ }
+ }
+
+ } else {
+ for(int ss=0;ss<U._grid->oSites();ss++){
+ int sU=lo.Reorder(ss);
+PARALLEL_FOR_LOOP
+ for(int s=0;s<Ls;s++){
+ int sF = s+Ls*sU;
+ DiracOpt::DhopSite(st,U,comm_buf,sF,sU,in,out);
+ }
+ }
+ }
+ }
+}
+void FiveDimWilsonFermion::DhopOE(const LatticeFermion &in, LatticeFermion &out,int dag)
+{
+ conformable(in._grid,FermionRedBlackGrid()); // 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);
+}
+void FiveDimWilsonFermion::DhopEO(const LatticeFermion &in, LatticeFermion &out,int dag)
+{
+ conformable(in._grid,FermionRedBlackGrid()); // 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);
+}
+void FiveDimWilsonFermion::Dhop(const LatticeFermion &in, LatticeFermion &out,int dag)
+{
+ conformable(in._grid,FermionGrid()); // verifies full grid
+ conformable(in._grid,out._grid);
+
+ out.checkerboard = in.checkerboard;
+
+ DhopInternal(Stencil,Lebesgue,Umu,in,out,dag);
+}
+
+}}
+
+
+
diff --git a/lib/qcd/action/fermion/FiveDimWilsonFermion.h b/lib/qcd/action/fermion/FiveDimWilsonFermion.h
new file mode 100644
index 00000000..3aa85023
--- /dev/null
+++ b/lib/qcd/action/fermion/FiveDimWilsonFermion.h
@@ -0,0 +1,108 @@
+#ifndef GRID_QCD_DWF_H
+#define GRID_QCD_DWF_H
+
+namespace Grid {
+
+ namespace QCD {
+
+ ////////////////////////////////////////////////////////////////////////////////
+ // This is the 4d red black case appropriate to support
+ //
+ // parity = (x+y+z+t)|2;
+ // generalised five dim fermions like mobius, zolotarev etc..
+ //
+ // i.e. even even contains fifth dim hopping term.
+ //
+ // [DIFFERS from original CPS red black implementation parity = (x+y+z+t+s)|2 ]
+ ////////////////////////////////////////////////////////////////////////////////
+ class FiveDimWilsonFermion : public FermionAction<LatticeFermion,LatticeGaugeField>
+ {
+ public:
+ ///////////////////////////////////////////////////////////////
+ // Implement the abstract base
+ ///////////////////////////////////////////////////////////////
+ GridBase *GaugeGrid(void) { return _FourDimGrid ;}
+ GridBase *GaugeRedBlackGrid(void) { return _FourDimRedBlackGrid ;}
+ GridBase *FermionGrid(void) { return _FiveDimGrid;}
+ GridBase *FermionRedBlackGrid(void) { return _FiveDimRedBlackGrid;}
+
+ // override multiply
+ virtual RealD M (const LatticeFermion &in, LatticeFermion &out);
+ virtual RealD Mdag (const LatticeFermion &in, LatticeFermion &out);
+
+ // half checkerboard operaions
+ virtual void Meooe (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MeooeDag (const LatticeFermion &in, LatticeFermion &out);
+ virtual void Mooee (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeDag (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeInv (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeInvDag (const LatticeFermion &in, LatticeFermion &out);
+
+ // non-hermitian hopping term; half cb or both
+ void Dhop (const LatticeFermion &in, LatticeFermion &out,int dag);
+ void DhopOE(const LatticeFermion &in, LatticeFermion &out,int dag);
+ void DhopEO(const LatticeFermion &in, LatticeFermion &out,int dag);
+
+ ///////////////////////////////////////////////////////////////
+ // New methods added
+ ///////////////////////////////////////////////////////////////
+ void DhopInternal(CartesianStencil & st,
+ LebesgueOrder &lo,
+ LatticeDoubledGaugeField &U,
+ const LatticeFermion &in,
+ LatticeFermion &out,
+ int dag);
+
+ // Constructors
+ FiveDimWilsonFermion(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ double _mass);
+
+ // DoubleStore
+ void DoubleStore(LatticeDoubledGaugeField &Uds,const LatticeGaugeField &Umu);
+
+ ///////////////////////////////////////////////////////////////
+ // Data members require to support the functionality
+ ///////////////////////////////////////////////////////////////
+ static int HandOptDslash; // these are a temporary hack
+
+ protected:
+
+ // Add these to the support from Wilson
+ GridBase *_FourDimGrid;
+ GridBase *_FourDimRedBlackGrid;
+ GridBase *_FiveDimGrid;
+ GridBase *_FiveDimRedBlackGrid;
+
+ static const int npoint=8;
+ static const std::vector<int> directions ;
+ static const std::vector<int> displacements;
+
+ double M5;
+ double mass;
+ int Ls;
+
+ //Defines the stencils for even and odd
+ CartesianStencil Stencil;
+ CartesianStencil StencilEven;
+ CartesianStencil StencilOdd;
+
+ // Copy of the gauge field , with even and odd subsets
+ LatticeDoubledGaugeField Umu;
+ LatticeDoubledGaugeField UmuEven;
+ LatticeDoubledGaugeField UmuOdd;
+
+ LebesgueOrder Lebesgue;
+ LebesgueOrder LebesgueEvenOdd;
+
+ // Comms buffer
+ std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > comm_buf;
+
+ };
+ }
+}
+
+#endif
diff --git a/lib/qcd/action/fermion/WilsonCompressor.h b/lib/qcd/action/fermion/WilsonCompressor.h
new file mode 100644
index 00000000..d9fe977a
--- /dev/null
+++ b/lib/qcd/action/fermion/WilsonCompressor.h
@@ -0,0 +1,61 @@
+#ifndef GRID_QCD_WILSON_COMPRESSOR_H
+#define GRID_QCD_WILSON_COMPRESSOR_H
+
+namespace Grid {
+namespace QCD {
+
+ class WilsonCompressor {
+ public:
+ int mu;
+ int dag;
+
+ WilsonCompressor(int _dag){
+ mu=0;
+ dag=_dag;
+ assert((dag==0)||(dag==1));
+ }
+ void Point(int p) {
+ mu=p;
+ };
+
+ vHalfSpinColourVector operator () (const vSpinColourVector &in)
+ {
+ vHalfSpinColourVector ret;
+ int mudag=mu;
+ if (dag) {
+ mudag=(mu+Nd)%(2*Nd);
+ }
+ switch(mudag) {
+ case Xp:
+ spProjXp(ret,in);
+ break;
+ case Yp:
+ spProjYp(ret,in);
+ break;
+ case Zp:
+ spProjZp(ret,in);
+ break;
+ case Tp:
+ spProjTp(ret,in);
+ break;
+ case Xm:
+ spProjXm(ret,in);
+ break;
+ case Ym:
+ spProjYm(ret,in);
+ break;
+ case Zm:
+ spProjZm(ret,in);
+ break;
+ case Tm:
+ spProjTm(ret,in);
+ break;
+ default:
+ assert(0);
+ break;
+ }
+ return ret;
+ }
+ };
+}} // namespace close
+#endif
diff --git a/lib/qcd/action/fermion/WilsonFermion.cc b/lib/qcd/action/fermion/WilsonFermion.cc
new file mode 100644
index 00000000..aa30a7fa
--- /dev/null
+++ b/lib/qcd/action/fermion/WilsonFermion.cc
@@ -0,0 +1,163 @@
+#include <Grid.h>
+
+namespace Grid {
+namespace QCD {
+
+const std::vector<int> WilsonFermion::directions ({0,1,2,3, 0, 1, 2, 3});
+const std::vector<int> WilsonFermion::displacements({1,1,1,1,-1,-1,-1,-1});
+
+int WilsonFermion::HandOptDslash;
+
+WilsonFermion::WilsonFermion(LatticeGaugeField &_Umu,
+ GridCartesian &Fgrid,
+ GridRedBlackCartesian &Hgrid,
+ double _mass) :
+ _grid(&Fgrid),
+ _cbgrid(&Hgrid),
+ Stencil (&Fgrid,npoint,Even,directions,displacements),
+ StencilEven(&Hgrid,npoint,Even,directions,displacements), // source is Even
+ StencilOdd (&Hgrid,npoint,Odd ,directions,displacements), // source is Odd
+ mass(_mass),
+ Umu(&Fgrid),
+ UmuEven(&Hgrid),
+ UmuOdd (&Hgrid)
+{
+ // Allocate the required comms buffer
+ comm_buf.resize(Stencil._unified_buffer_size); // this is always big enough to contain EO
+ DoubleStore(Umu,_Umu);
+ pickCheckerboard(Even,UmuEven,Umu);
+ pickCheckerboard(Odd ,UmuOdd,Umu);
+}
+
+void WilsonFermion::DoubleStore(LatticeDoubledGaugeField &Uds,const LatticeGaugeField &Umu)
+{
+ conformable(Uds._grid,GaugeGrid());
+ conformable(Umu._grid,GaugeGrid());
+ LatticeColourMatrix 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);
+ }
+}
+
+RealD WilsonFermion::M(const LatticeFermion &in, LatticeFermion &out)
+{
+ out.checkerboard=in.checkerboard;
+ Dhop(in,out,DaggerNo);
+ return axpy_norm(out,4+mass,in,out);
+}
+RealD WilsonFermion::Mdag(const LatticeFermion &in, LatticeFermion &out)
+{
+ out.checkerboard=in.checkerboard;
+ Dhop(in,out,DaggerYes);
+ return axpy_norm(out,4+mass,in,out);
+}
+
+void WilsonFermion::Meooe(const LatticeFermion &in, LatticeFermion &out)
+{
+ if ( in.checkerboard == Odd ) {
+ DhopEO(in,out,DaggerNo);
+ } else {
+ DhopOE(in,out,DaggerNo);
+ }
+}
+void WilsonFermion::MeooeDag(const LatticeFermion &in, LatticeFermion &out)
+{
+ if ( in.checkerboard == Odd ) {
+ DhopEO(in,out,DaggerYes);
+ } else {
+ DhopOE(in,out,DaggerYes);
+ }
+}
+void WilsonFermion::Mooee(const LatticeFermion &in, LatticeFermion &out)
+{
+ out.checkerboard = in.checkerboard;
+ out = (4.0+mass)*in;
+ return ;
+}
+void WilsonFermion::MooeeDag(const LatticeFermion &in, LatticeFermion &out)
+{
+ out.checkerboard = in.checkerboard;
+ Mooee(in,out);
+}
+void WilsonFermion::MooeeInv(const LatticeFermion &in, LatticeFermion &out)
+{
+ out.checkerboard = in.checkerboard;
+ out = (1.0/(4.0+mass))*in;
+ return ;
+}
+void WilsonFermion::MooeeInvDag(const LatticeFermion &in, LatticeFermion &out)
+{
+ out.checkerboard = in.checkerboard;
+ MooeeInv(in,out);
+}
+
+void WilsonFermion::DhopInternal(CartesianStencil & st,LatticeDoubledGaugeField & U,
+ const LatticeFermion &in, LatticeFermion &out,int dag)
+{
+ assert((dag==DaggerNo) ||(dag==DaggerYes));
+ WilsonCompressor compressor(dag);
+ st.HaloExchange<vSpinColourVector,vHalfSpinColourVector,WilsonCompressor>(in,comm_buf,compressor);
+
+ if ( dag == DaggerYes ) {
+ if( HandOptDslash ) {
+PARALLEL_FOR_LOOP
+ for(int sss=0;sss<in._grid->oSites();sss++){
+ DiracOptHand::DhopSiteDag(st,U,comm_buf,sss,sss,in,out);
+ }
+ } else {
+PARALLEL_FOR_LOOP
+ for(int sss=0;sss<in._grid->oSites();sss++){
+ DiracOpt::DhopSiteDag(st,U,comm_buf,sss,sss,in,out);
+ }
+ }
+ } else {
+ if( HandOptDslash ) {
+PARALLEL_FOR_LOOP
+ for(int sss=0;sss<in._grid->oSites();sss++){
+ DiracOptHand::DhopSite(st,U,comm_buf,sss,sss,in,out);
+ }
+ } else {
+PARALLEL_FOR_LOOP
+ for(int sss=0;sss<in._grid->oSites();sss++){
+ DiracOpt::DhopSite(st,U,comm_buf,sss,sss,in,out);
+ }
+ }
+ }
+}
+void WilsonFermion::DhopOE(const LatticeFermion &in, LatticeFermion &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,UmuOdd,in,out,dag);
+}
+void WilsonFermion::DhopEO(const LatticeFermion &in, LatticeFermion &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,UmuEven,in,out,dag);
+}
+void WilsonFermion::Dhop(const LatticeFermion &in, LatticeFermion &out,int dag)
+{
+ conformable(in._grid,_grid); // verifies full grid
+ conformable(in._grid,out._grid);
+
+ out.checkerboard = in.checkerboard;
+
+ DhopInternal(Stencil,Umu,in,out,dag);
+}
+
+}}
+
+
+
diff --git a/lib/qcd/action/fermion/WilsonFermion.h b/lib/qcd/action/fermion/WilsonFermion.h
new file mode 100644
index 00000000..5c208131
--- /dev/null
+++ b/lib/qcd/action/fermion/WilsonFermion.h
@@ -0,0 +1,87 @@
+#ifndef GRID_QCD_WILSON_FERMION_H
+#define GRID_QCD_WILSON_FERMION_H
+
+namespace Grid {
+
+ namespace QCD {
+
+ class WilsonFermion : public FermionAction<LatticeFermion,LatticeGaugeField>
+ {
+ public:
+
+ ///////////////////////////////////////////////////////////////
+ // Implement the abstract base
+ ///////////////////////////////////////////////////////////////
+ GridBase *GaugeGrid(void) { return _grid ;}
+ GridBase *GaugeRedBlackGrid(void) { return _cbgrid ;}
+ GridBase *FermionGrid(void) { return _grid;}
+ GridBase *FermionRedBlackGrid(void) { return _cbgrid;}
+
+ // override multiply
+ virtual RealD M (const LatticeFermion &in, LatticeFermion &out);
+ virtual RealD Mdag (const LatticeFermion &in, LatticeFermion &out);
+
+ // half checkerboard operaions
+ void Meooe (const LatticeFermion &in, LatticeFermion &out);
+ void MeooeDag (const LatticeFermion &in, LatticeFermion &out);
+ virtual void Mooee (const LatticeFermion &in, LatticeFermion &out); // remain virtual so we
+ virtual void MooeeDag (const LatticeFermion &in, LatticeFermion &out); // can derive Clover
+ virtual void MooeeInv (const LatticeFermion &in, LatticeFermion &out); // from Wilson bas
+ virtual void MooeeInvDag (const LatticeFermion &in, LatticeFermion &out);
+
+ // non-hermitian hopping term; half cb or both
+ void Dhop (const LatticeFermion &in, LatticeFermion &out,int dag);
+ void DhopOE(const LatticeFermion &in, LatticeFermion &out,int dag);
+ void DhopEO(const LatticeFermion &in, LatticeFermion &out,int dag);
+
+ ///////////////////////////////////////////////////////////////
+ // Extra methods added by derived
+ ///////////////////////////////////////////////////////////////
+ void DhopInternal(CartesianStencil & st,
+ LatticeDoubledGaugeField &U,
+ const LatticeFermion &in,
+ LatticeFermion &out,
+ int dag);
+
+ // Constructor
+ WilsonFermion(LatticeGaugeField &_Umu,GridCartesian &Fgrid,GridRedBlackCartesian &Hgrid,double _mass);
+
+ // DoubleStore
+ void DoubleStore(LatticeDoubledGaugeField &Uds,const LatticeGaugeField &Umu);
+
+ ///////////////////////////////////////////////////////////////
+ // Data members require to support the functionality
+ ///////////////////////////////////////////////////////////////
+ static int HandOptDslash; // these are a temporary hack
+ static int MortonOrder;
+
+ protected:
+
+ double mass;
+
+ GridBase * _grid;
+ GridBase * _cbgrid;
+
+ static const int npoint=8;
+ static const std::vector<int> directions ;
+ static const std::vector<int> displacements;
+
+ //Defines the stencils for even and odd
+ CartesianStencil Stencil;
+ CartesianStencil StencilEven;
+ CartesianStencil StencilOdd;
+
+ // Copy of the gauge field , with even and odd subsets
+ LatticeDoubledGaugeField Umu;
+ LatticeDoubledGaugeField UmuEven;
+ LatticeDoubledGaugeField UmuOdd;
+
+ // Comms buffer
+ std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > comm_buf;
+
+
+ };
+
+ }
+}
+#endif
diff --git a/lib/qcd/Grid_qcd_dhop.cc b/lib/qcd/action/fermion/WilsonKernels.cc
similarity index 87%
rename from lib/qcd/Grid_qcd_dhop.cc
rename to lib/qcd/action/fermion/WilsonKernels.cc
index 1e5dcd16..879bbb7a 100644
--- a/lib/qcd/Grid_qcd_dhop.cc
+++ b/lib/qcd/action/fermion/WilsonKernels.cc
@@ -4,8 +4,8 @@ namespace Grid {
namespace QCD {
void DiracOpt::DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
- std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > &buf,
- int ss,const LatticeFermion &in, LatticeFermion &out)
+ std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > &buf,
+ int sF,int sU,const LatticeFermion &in, LatticeFermion &out)
{
vHalfSpinColourVector tmp;
vHalfSpinColourVector chi;
@@ -16,6 +16,7 @@ void DiracOpt::DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
//#define VERBOSE( A) if ( ss<10 ) { std::cout << "site " <<ss << " " #A " neigh " << offset << " perm "<< perm <<std::endl;}
// Xp
+ int ss = sF;
offset = st._offsets [Xp][ss];
local = st._is_local[Xp][ss];
perm = st._permute[Xp][ss];
@@ -29,7 +30,7 @@ void DiracOpt::DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Xp),&chi());
+ mult(&Uchi(),&U._odata[sU](Xp),&chi());
spReconXp(result,Uchi);
// std::cout << "XP_RECON"<<std::endl;
@@ -51,7 +52,7 @@ void DiracOpt::DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Yp),&chi());
+ mult(&Uchi(),&U._odata[sU](Yp),&chi());
accumReconYp(result,Uchi);
// Zp
@@ -67,7 +68,7 @@ void DiracOpt::DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Zp),&chi());
+ mult(&Uchi(),&U._odata[sU](Zp),&chi());
accumReconZp(result,Uchi);
// Tp
@@ -83,7 +84,7 @@ void DiracOpt::DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Tp),&chi());
+ mult(&Uchi(),&U._odata[sU](Tp),&chi());
accumReconTp(result,Uchi);
// Xm
@@ -101,7 +102,7 @@ void DiracOpt::DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Xm),&chi());
+ mult(&Uchi(),&U._odata[sU](Xm),&chi());
accumReconXm(result,Uchi);
// std::cout << "XM_RECON_ACCUM"<<std::endl;
// std::cout << result()(0)(0) <<" "<<result()(0)(1) <<" "<<result()(0)(2) <<std::endl;
@@ -124,7 +125,7 @@ void DiracOpt::DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Ym),&chi());
+ mult(&Uchi(),&U._odata[sU](Ym),&chi());
accumReconYm(result,Uchi);
// Zm
@@ -140,7 +141,7 @@ void DiracOpt::DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Zm),&chi());
+ mult(&Uchi(),&U._odata[sU](Zm),&chi());
accumReconZm(result,Uchi);
// Tm
@@ -156,15 +157,15 @@ void DiracOpt::DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Tm),&chi());
+ mult(&Uchi(),&U._odata[sU](Tm),&chi());
accumReconTm(result,Uchi);
- vstream(out._odata[ss],result);
+ vstream(out._odata[ss],result*(-0.5));
}
void DiracOpt::DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
- std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > &buf,
- int ss,const LatticeFermion &in, LatticeFermion &out)
+ std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > &buf,
+ int sF,int sU,const LatticeFermion &in, LatticeFermion &out)
{
vHalfSpinColourVector tmp;
vHalfSpinColourVector chi;
@@ -173,6 +174,7 @@ void DiracOpt::DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
int offset,local,perm, ptype;
// Xp
+ int ss=sF;
offset = st._offsets [Xm][ss];
local = st._is_local[Xm][ss];
perm = st._permute[Xm][ss];
@@ -186,7 +188,7 @@ void DiracOpt::DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Xm),&chi());
+ mult(&Uchi(),&U._odata[sU](Xm),&chi());
spReconXp(result,Uchi);
// Yp
@@ -202,7 +204,7 @@ void DiracOpt::DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Ym),&chi());
+ mult(&Uchi(),&U._odata[sU](Ym),&chi());
accumReconYp(result,Uchi);
// Zp
@@ -218,7 +220,7 @@ void DiracOpt::DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Zm),&chi());
+ mult(&Uchi(),&U._odata[sU](Zm),&chi());
accumReconZp(result,Uchi);
// Tp
@@ -234,7 +236,7 @@ void DiracOpt::DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Tm),&chi());
+ mult(&Uchi(),&U._odata[sU](Tm),&chi());
accumReconTp(result,Uchi);
// Xm
@@ -252,7 +254,7 @@ void DiracOpt::DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Xp),&chi());
+ mult(&Uchi(),&U._odata[sU](Xp),&chi());
accumReconXm(result,Uchi);
// Ym
@@ -269,7 +271,7 @@ void DiracOpt::DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Yp),&chi());
+ mult(&Uchi(),&U._odata[sU](Yp),&chi());
accumReconYm(result,Uchi);
// Zm
@@ -285,7 +287,7 @@ void DiracOpt::DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Zp),&chi());
+ mult(&Uchi(),&U._odata[sU](Zp),&chi());
accumReconZm(result,Uchi);
// Tm
@@ -301,9 +303,9 @@ void DiracOpt::DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
} else {
chi=buf[offset];
}
- mult(&Uchi(),&U._odata[ss](Tp),&chi());
+ mult(&Uchi(),&U._odata[sU](Tp),&chi());
accumReconTm(result,Uchi);
- vstream(out._odata[ss],result);
+ vstream(out._odata[ss],result*(-0.5));
}
}}
diff --git a/lib/qcd/action/fermion/WilsonKernels.h b/lib/qcd/action/fermion/WilsonKernels.h
new file mode 100644
index 00000000..cc535022
--- /dev/null
+++ b/lib/qcd/action/fermion/WilsonKernels.h
@@ -0,0 +1,42 @@
+#ifndef GRID_QCD_DHOP_H
+#define GRID_QCD_DHOP_H
+
+namespace Grid {
+
+ namespace QCD {
+
+ ////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+ // Helper classes that implement Wilson stencil for a single site.
+ ////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+ // Generic version works for any Nc and with extra flavour indices
+ class DiracOpt {
+ public:
+ // These ones will need to be package intelligently. WilsonType base class
+ // for use by DWF etc..
+ static void DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
+ std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > &buf,
+ int sF,int sU,const LatticeFermion &in, LatticeFermion &out);
+ static void DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
+ std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > &buf,
+ int sF,int sU,const LatticeFermion &in, LatticeFermion &out);
+
+ };
+
+ // Hand unrolled for Nc=3, one flavour
+ class DiracOptHand {
+ public:
+ // These ones will need to be package intelligently. WilsonType base class
+ // for use by DWF etc..
+ static void DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
+ std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > &buf,
+ int sF,int sU,const LatticeFermion &in, LatticeFermion &out);
+ static void DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
+ std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > &buf,
+ int sF,int sU,const LatticeFermion &in, LatticeFermion &out);
+
+ };
+
+ }
+}
+#endif
diff --git a/lib/qcd/Grid_qcd_dhop_hand.cc b/lib/qcd/action/fermion/WilsonKernelsHand.cc
similarity index 92%
rename from lib/qcd/Grid_qcd_dhop_hand.cc
rename to lib/qcd/action/fermion/WilsonKernelsHand.cc
index f8d464fb..019b668b 100644
--- a/lib/qcd/Grid_qcd_dhop_hand.cc
+++ b/lib/qcd/action/fermion/WilsonKernelsHand.cc
@@ -27,7 +27,7 @@
Chi_12 = ref()(1)(2);
#define MULT_2SPIN(A)\
- auto & ref(U._odata[ss](A)); \
+ auto & ref(U._odata[sU](A)); \
U_00 = ref()(0,0);\
U_10 = ref()(1,0);\
U_20 = ref()(2,0);\
@@ -282,7 +282,7 @@ namespace QCD {
void DiracOptHand::DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > &buf,
- int ss,const LatticeFermion &in, LatticeFermion &out)
+ int sF,int sU,const LatticeFermion &in, LatticeFermion &out)
{
REGISTER vComplex result_00; // 12 regs on knc
REGISTER vComplex result_01;
@@ -338,7 +338,8 @@ void DiracOptHand::DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
int offset,local,perm, ptype;
-
+ int ss=sF;
+
// Xp
offset = st._offsets [Xp][ss];
local = st._is_local[Xp][ss];
@@ -514,24 +515,24 @@ void DiracOptHand::DhopSite(CartesianStencil &st,LatticeDoubledGaugeField &U,
{
vSpinColourVector & ref (out._odata[ss]);
- vstream(ref()(0)(0),result_00);
- vstream(ref()(0)(1),result_01);
- vstream(ref()(0)(2),result_02);
- vstream(ref()(1)(0),result_10);
- vstream(ref()(1)(1),result_11);
- vstream(ref()(1)(2),result_12);
- vstream(ref()(2)(0),result_20);
- vstream(ref()(2)(1),result_21);
- vstream(ref()(2)(2),result_22);
- vstream(ref()(3)(0),result_30);
- vstream(ref()(3)(1),result_31);
- vstream(ref()(3)(2),result_32);
+ vstream(ref()(0)(0),result_00*(-0.5));
+ vstream(ref()(0)(1),result_01*(-0.5));
+ vstream(ref()(0)(2),result_02*(-0.5));
+ vstream(ref()(1)(0),result_10*(-0.5));
+ vstream(ref()(1)(1),result_11*(-0.5));
+ vstream(ref()(1)(2),result_12*(-0.5));
+ vstream(ref()(2)(0),result_20*(-0.5));
+ vstream(ref()(2)(1),result_21*(-0.5));
+ vstream(ref()(2)(2),result_22*(-0.5));
+ vstream(ref()(3)(0),result_30*(-0.5));
+ vstream(ref()(3)(1),result_31*(-0.5));
+ vstream(ref()(3)(2),result_32*(-0.5));
}
}
void DiracOptHand::DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
std::vector<vHalfSpinColourVector,alignedAllocator<vHalfSpinColourVector> > &buf,
- int ss,const LatticeFermion &in, LatticeFermion &out)
+ int ss,int sU,const LatticeFermion &in, LatticeFermion &out)
{
REGISTER vComplex result_00; // 12 regs on knc
REGISTER vComplex result_01;
@@ -752,18 +753,18 @@ void DiracOptHand::DhopSiteDag(CartesianStencil &st,LatticeDoubledGaugeField &U,
{
vSpinColourVector & ref (out._odata[ss]);
- vstream(ref()(0)(0),result_00);
- vstream(ref()(0)(1),result_01);
- vstream(ref()(0)(2),result_02);
- vstream(ref()(1)(0),result_10);
- vstream(ref()(1)(1),result_11);
- vstream(ref()(1)(2),result_12);
- vstream(ref()(2)(0),result_20);
- vstream(ref()(2)(1),result_21);
- vstream(ref()(2)(2),result_22);
- vstream(ref()(3)(0),result_30);
- vstream(ref()(3)(1),result_31);
- vstream(ref()(3)(2),result_32);
+ vstream(ref()(0)(0),result_00*(-0.5));
+ vstream(ref()(0)(1),result_01*(-0.5));
+ vstream(ref()(0)(2),result_02*(-0.5));
+ vstream(ref()(1)(0),result_10*(-0.5));
+ vstream(ref()(1)(1),result_11*(-0.5));
+ vstream(ref()(1)(2),result_12*(-0.5));
+ vstream(ref()(2)(0),result_20*(-0.5));
+ vstream(ref()(2)(1),result_21*(-0.5));
+ vstream(ref()(2)(2),result_22*(-0.5));
+ vstream(ref()(3)(0),result_30*(-0.5));
+ vstream(ref()(3)(1),result_31*(-0.5));
+ vstream(ref()(3)(2),result_32*(-0.5));
}
}
}}
diff --git a/lib/stencil/Grid_lebesgue.cc b/lib/stencil/Grid_lebesgue.cc
new file mode 100644
index 00000000..977e3562
--- /dev/null
+++ b/lib/stencil/Grid_lebesgue.cc
@@ -0,0 +1,103 @@
+#include <Grid.h>
+
+namespace Grid {
+
+int LebesgueOrder::UseLebesgueOrder;
+
+LebesgueOrder::IndexInteger LebesgueOrder::alignup(IndexInteger n){
+ n--; // 1000 0011 --> 1000 0010
+ n |= n >> 1; // 1000 0010 | 0100 0001 = 1100 0011
+ n |= n >> 2; // 1100 0011 | 0011 0000 = 1111 0011
+ n |= n >> 4; // 1111 0011 | 0000 1111 = 1111 1111
+ n |= n >> 8; // ... (At this point all bits are 1, so further bitwise-or
+ n |= n >> 16; // operations produce no effect.)
+ n++; // 1111 1111 --> 1 0000 0000
+ return n;
+};
+
+LebesgueOrder::LebesgueOrder(GridBase *grid)
+{
+ _LebesgueReorder.resize(0);
+
+ // Align up dimensions to power of two.
+ const IndexInteger one=1;
+ IndexInteger ND = grid->_ndimension;
+ std::vector<IndexInteger> dims(ND);
+ std::vector<IndexInteger> adims(ND);
+ std::vector<std::vector<IndexInteger> > bitlist(ND);
+
+ for(IndexInteger mu=0;mu<ND;mu++){
+ dims[mu] = grid->_rdimensions[mu];
+ assert ( dims[mu] != 0 );
+ adims[mu] = alignup(dims[mu]);
+ }
+
+ // List which bits of padded volume coordinate contribute; this strategy
+ // i) avoids recursion
+ // ii) has loop lengths at most the width of a 32 bit word.
+ int sitebit=0;
+ int split=2;
+ for(int mu=0;mu<ND;mu++){ // mu 0 takes bit 0; mu 1 takes bit 1 etc...
+ for(int bit=0;bit<split;bit++){
+ IndexInteger mask = one<<bit;
+ if ( mask&(adims[mu]-1) ){
+ bitlist[mu].push_back(sitebit);
+ sitebit++;
+ }
+ }
+ }
+ for(int bit=split;bit<32;bit++){
+ IndexInteger mask = one<<bit;
+ for(int mu=0;mu<ND;mu++){ // mu 0 takes bit 0; mu 1 takes bit 1 etc...
+ if ( mask&(adims[mu]-1) ){
+ bitlist[mu].push_back(sitebit);
+ sitebit++;
+ }
+ }
+ }
+
+ // Work out padded and unpadded volumes
+ IndexInteger avol = 1;
+ for(int mu=0;mu<ND;mu++) avol = avol * adims[mu];
+
+ IndexInteger vol = 1;
+ for(int mu=0;mu<ND;mu++) vol = vol * dims[mu];
+
+ // Loop over padded volume, following Lebesgue curve
+ // We interleave the bits from sequential "mu".
+ std::vector<IndexInteger> ax(ND);
+
+ for(IndexInteger asite=0;asite<avol;asite++){
+
+ // Start with zero and collect bits
+ for(int mu=0;mu<ND;mu++) ax[mu] = 0;
+
+ int contained = 1;
+ for(int mu=0;mu<ND;mu++){
+
+ // Build the coordinate on the aligned volume
+ for(int bit=0;bit<bitlist[mu].size();bit++){
+ int sbit=bitlist[mu][bit];
+
+ if(asite&(one<<sbit)){
+ ax[mu]|=one<<bit;
+ }
+ }
+
+ // Is it contained in original box
+ if ( ax[mu]>dims[mu]-1 ) contained = 0;
+
+ }
+
+ if ( contained ) {
+ int site = ax[0]
+ + dims[0]*ax[1]
+ +dims[0]*dims[1]*ax[2]
+ +dims[0]*dims[1]*dims[2]*ax[3];
+
+ _LebesgueReorder.push_back(site);
+ }
+ }
+ assert( _LebesgueReorder.size() == vol );
+}
+}
diff --git a/lib/stencil/Grid_lebesgue.h b/lib/stencil/Grid_lebesgue.h
new file mode 100644
index 00000000..1d59b127
--- /dev/null
+++ b/lib/stencil/Grid_lebesgue.h
@@ -0,0 +1,29 @@
+#ifndef GRID_LEBESGUE_H
+#define GRID_LEBESGUE_H
+
+#include<vector>
+
+// Lebesgue, Morton, Z-graph ordering assistance
+namespace Grid {
+
+ class LebesgueOrder {
+ public:
+
+ static int UseLebesgueOrder;
+
+ typedef uint32_t IndexInteger;
+
+ inline IndexInteger Reorder(IndexInteger ss) {
+ return UseLebesgueOrder ? _LebesgueReorder[ss] : ss;
+ };
+
+ IndexInteger alignup(IndexInteger n);
+
+ LebesgueOrder(GridBase *grid);
+
+ private:
+ std::vector<IndexInteger> _LebesgueReorder;
+
+ };
+}
+#endif
diff --git a/lib/stencil/Grid_stencil_common.cc b/lib/stencil/Grid_stencil_common.cc
index 6cbcc890..f0f8c581 100644
--- a/lib/stencil/Grid_stencil_common.cc
+++ b/lib/stencil/Grid_stencil_common.cc
@@ -3,95 +3,6 @@
namespace Grid {
-
-void CartesianStencil::LebesgueOrder(void)
-{
- _LebesgueReorder.resize(0);
-
- // Align up dimensions to power of two.
- const StencilInteger one=1;
- StencilInteger ND = _grid->_ndimension;
- std::vector<StencilInteger> dims(ND);
- std::vector<StencilInteger> adims(ND);
- std::vector<std::vector<StencilInteger> > bitlist(ND);
-
-
- for(StencilInteger mu=0;mu<ND;mu++){
- dims[mu] = _grid->_rdimensions[mu];
- assert ( dims[mu] != 0 );
- adims[mu] = alignup(dims[mu]);
- }
-
- // List which bits of padded volume coordinate contribute; this strategy
- // i) avoids recursion
- // ii) has loop lengths at most the width of a 32 bit word.
- int sitebit=0;
- int split=24;
- for(int mu=0;mu<ND;mu++){ // mu 0 takes bit 0; mu 1 takes bit 1 etc...
- for(int bit=0;bit<split;bit++){
- StencilInteger mask = one<<bit;
- if ( mask&(adims[mu]-1) ){
- bitlist[mu].push_back(sitebit);
- sitebit++;
- }
- }
- }
- for(int bit=split;bit<32;bit++){
- StencilInteger mask = one<<bit;
- for(int mu=0;mu<ND;mu++){ // mu 0 takes bit 0; mu 1 takes bit 1 etc...
- if ( mask&(adims[mu]-1) ){
- bitlist[mu].push_back(sitebit);
- sitebit++;
- }
- }
- }
-
- // Work out padded and unpadded volumes
- StencilInteger avol = 1;
- for(int mu=0;mu<ND;mu++) avol = avol * adims[mu];
-
- StencilInteger vol = 1;
- for(int mu=0;mu<ND;mu++) vol = vol * dims[mu];
-
- // Loop over padded volume, following Lebesgue curve
- // We interleave the bits from sequential "mu".
- std::vector<StencilInteger> ax(ND);
-
- for(StencilInteger asite=0;asite<avol;asite++){
-
- // Start with zero and collect bits
- for(int mu=0;mu<ND;mu++) ax[mu] = 0;
-
- int contained = 1;
- for(int mu=0;mu<ND;mu++){
-
- // Build the coordinate on the aligned volume
- for(int bit=0;bit<bitlist[mu].size();bit++){
- int sbit=bitlist[mu][bit];
-
- if(asite&(one<<sbit)){
- ax[mu]|=one<<bit;
- }
- }
-
- // Is it contained in original box
- if ( ax[mu]>dims[mu]-1 ) contained = 0;
-
- }
-
- if ( contained ) {
- int site = ax[0]
- + dims[0]*ax[1]
- +dims[0]*dims[1]*ax[2]
- +dims[0]*dims[1]*dims[2]*ax[3];
-
- _LebesgueReorder.push_back(site);
- }
- }
-
- assert( _LebesgueReorder.size() == vol );
-}
-
CartesianStencil::CartesianStencil(GridBase *grid,
int npoints,
int checkerboard,
@@ -110,8 +21,6 @@ void CartesianStencil::LebesgueOrder(void)
_unified_buffer_size=0;
_request_count =0;
- LebesgueOrder();
-
int osites = _grid->oSites();
for(int i=0;i<npoints;i++){
@@ -143,8 +52,8 @@ void CartesianStencil::LebesgueOrder(void)
// up a table containing the npoint "neighbours" and whether they
// live in lattice or a comms buffer.
if ( !comm_dim ) {
- sshift[0] = _grid->CheckerBoardShift(_checkerboard,dimension,shift,0);
- sshift[1] = _grid->CheckerBoardShift(_checkerboard,dimension,shift,1);
+ sshift[0] = _grid->CheckerBoardShift(_checkerboard,dimension,shift,Even);
+ sshift[1] = _grid->CheckerBoardShift(_checkerboard,dimension,shift,Odd);
if ( sshift[0] == sshift[1] ) {
Local(point,dimension,shift,0x3);
@@ -154,8 +63,8 @@ void CartesianStencil::LebesgueOrder(void)
}
} else { // All permute extract done in comms phase prior to Stencil application
// So tables are the same whether comm_dim or splice_dim
- sshift[0] = _grid->CheckerBoardShift(_checkerboard,dimension,shift,0);
- sshift[1] = _grid->CheckerBoardShift(_checkerboard,dimension,shift,1);
+ sshift[0] = _grid->CheckerBoardShift(_checkerboard,dimension,shift,Even);
+ sshift[1] = _grid->CheckerBoardShift(_checkerboard,dimension,shift,Odd);
if ( sshift[0] == sshift[1] ) {
Comms(point,dimension,shift,0x3);
} else {
@@ -185,7 +94,7 @@ void CartesianStencil::LebesgueOrder(void)
int o = 0;
int bo = x * _grid->_ostride[dimension];
- int cb= (cbmask==0x2)? 1 : 0;
+ int cb= (cbmask==0x2)? Odd : Even;
int sshift = _grid->CheckerBoardShift(_checkerboard,dimension,shift,cb);
int sx = (x+sshift)%rd;
@@ -224,7 +133,7 @@ void CartesianStencil::LebesgueOrder(void)
_comm_buf_size[point] = buffer_size; // Size of _one_ plane. Multiple planes may be gathered and
// send to one or more remote nodes.
- int cb= (cbmask==0x2)? 1 : 0;
+ int cb= (cbmask==0x2)? Odd : Even;
int sshift= _grid->CheckerBoardShift(_checkerboard,dimension,shift,cb);
for(int x=0;x<rd;x++){
diff --git a/scripts/linecount b/scripts/linecount
index c0731b23..350abf1c 100755
--- a/scripts/linecount
+++ b/scripts/linecount
@@ -1,3 +1,3 @@
#!/bin/sh
-wc -l */*.h */*/*.h */*/*/*.h */*.cc */*/*.cc */*/*/*.cc
\ No newline at end of file
+wc -l lib/*.h lib/*/*.h lib/*/*/*.h lib/*.cc lib/*/*.cc lib/*/*/*.cc tests/*.cc benchmarks/*.cc
\ No newline at end of file
diff --git a/tests/Grid_cshift_red_black.cc b/tests/Grid_cshift_red_black.cc
new file mode 100644
index 00000000..212e9b6e
--- /dev/null
+++ b/tests/Grid_cshift_red_black.cc
@@ -0,0 +1,165 @@
+#include <Grid.h>
+#include <parallelIO/GridNerscIO.h>
+
+using namespace Grid;
+using namespace Grid::QCD;
+
+int main (int argc, char ** argv)
+{
+ Grid_init(&argc,&argv);
+
+ std::vector<int> latt_size = GridDefaultLatt();
+ int Nd = latt_size.size();
+ std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplexF::Nsimd());
+ std::vector<int> mpi_layout = GridDefaultMpi();
+
+ std::vector<int> mask(Nd,1);
+ mask[0]=0;
+
+ GridCartesian Fine (latt_size,simd_layout,mpi_layout);
+ GridRedBlackCartesian RBFine(latt_size,simd_layout,mpi_layout,mask,1);
+
+ GridParallelRNG FineRNG(&Fine); FineRNG.SeedRandomDevice();
+
+ LatticeComplex U(&Fine);
+ LatticeComplex ShiftU(&Fine);
+ LatticeComplex rbShiftU(&Fine);
+ LatticeComplex Ue(&RBFine);
+ LatticeComplex Uo(&RBFine);
+ LatticeComplex ShiftUe(&RBFine);
+ LatticeComplex ShiftUo(&RBFine);
+ LatticeComplex lex(&Fine);
+ lex=zero;
+ Integer stride =1;
+ {
+ double nrm;
+ LatticeComplex coor(&Fine);
+
+ for(int d=0;d<Nd;d++){
+ // Integer i=10000;
+ Integer i=0;
+ LatticeCoordinate(coor,d);
+ lex = lex + coor*stride+i;
+ stride=stride*latt_size[d];
+ }
+ U=lex;
+ }
+
+ pickCheckerboard(Even,Ue,U);
+ pickCheckerboard(Odd,Uo,U);
+
+ // std::cout << U<<std::endl;
+ std::cout << "Ue " <<norm2(Ue)<<std::endl;
+ std::cout << "Uo " <<norm2(Uo)<<std::endl;
+
+
+ TComplex cm;
+ for(int dir=0;dir<Nd;dir++){
+ if ( dir!=1 ) continue;
+ for(int shift=0;shift<latt_size[dir];shift++){
+
+ std::cout<<"Shifting by "<<shift<<" in direction"<<dir<<std::endl;
+
+ // std::cout<<"Even grid"<<std::endl;
+ ShiftUe = Cshift(Ue,dir,shift); // Shift everything cb by cb
+ // std::cout << "\tShiftUe " <<norm2(ShiftUe)<<std::endl;
+
+ // std::cout<<"Odd grid"<<std::endl;
+ ShiftUo = Cshift(Uo,dir,shift);
+ // std::cout << "\tShiftUo " <<norm2(ShiftUo)<<std::endl;
+
+ // std::cout<<"Recombined Even/Odd grids"<<std::endl;
+ setCheckerboard(rbShiftU,ShiftUe);
+ setCheckerboard(rbShiftU,ShiftUo);
+ // std::cout << "\trbShiftU " <<norm2(rbShiftU)<<std::endl;
+
+ // std::cout<<"Full grid shift"<<std::endl;
+ ShiftU = Cshift(U,dir,shift); // Shift everything
+ // std::cout << "\tShiftU " <<norm2(rbShiftU)<<std::endl;
+
+ std::vector<int> coor(4);
+
+ std::cout << "Checking the non-checkerboard shift"<<std::endl;
+ for(coor[3]=0;coor[3]<latt_size[3];coor[3]++){
+ for(coor[2]=0;coor[2]<latt_size[2];coor[2]++){
+ for(coor[1]=0;coor[1]<latt_size[1];coor[1]++){
+ for(coor[0]=0;coor[0]<latt_size[0];coor[0]++){
+
+ peekSite(cm,ShiftU,coor);
+
+ ///////// double nrm=norm2(U);
+
+ std::vector<int> scoor(coor);
+ scoor[dir] = (scoor[dir]+shift)%latt_size[dir];
+
+ Integer slex = scoor[0]
+ + latt_size[0]*scoor[1]
+ + latt_size[0]*latt_size[1]*scoor[2]
+ + latt_size[0]*latt_size[1]*latt_size[2]*scoor[3];
+
+ Complex scm(slex);
+
+ double nrm = abs(scm-cm()()());
+ std::vector<int> peer(4);
+ int index=real(cm);
+ Fine.CoorFromIndex(peer,index,latt_size);
+
+ if (nrm > 0){
+ std::cerr<<"FAIL shift "<< shift<<" in dir "<< dir
+ <<" ["<<coor[0]<<","<<coor[1]<<","<<coor[2]<<","<<coor[3]<<"] = "
+ << cm()()()<<" expect "<<scm<<" "<<nrm<<std::endl;
+ std::cerr<<"Got "<<index<<" " << peer[0]<<","<<peer[1]<<","<<peer[2]<<","<<peer[3]<<std::endl;
+ index=real(scm);
+ Fine.CoorFromIndex(peer,index,latt_size);
+ std::cerr<<"Expect "<<index<<" " << peer[0]<<","<<peer[1]<<","<<peer[2]<<","<<peer[3]<<std::endl;
+ exit(-1);
+ }
+ }}}}
+
+
+ std::cout << "Checking the checkerboard shift"<<std::endl;
+ for(coor[3]=0;coor[3]<latt_size[3];coor[3]++){
+ for(coor[2]=0;coor[2]<latt_size[2];coor[2]++){
+ for(coor[1]=0;coor[1]<latt_size[1];coor[1]++){
+ for(coor[0]=0;coor[0]<latt_size[0];coor[0]++){
+
+ peekSite(cm,rbShiftU,coor);
+
+ double nrm=norm2(U);
+
+ std::vector<int> scoor(coor);
+ scoor[dir] = (scoor[dir]+shift)%latt_size[dir];
+
+ Integer slex = scoor[0]
+ + latt_size[0]*scoor[1]
+ + latt_size[0]*latt_size[1]*scoor[2]
+ + latt_size[0]*latt_size[1]*latt_size[2]*scoor[3];
+
+ Complex scm(slex);
+
+ nrm = abs(scm-cm()()());
+ std::vector<int> peer(4);
+ int index=real(cm);
+ Fine.CoorFromIndex(peer,index,latt_size);
+
+ if (nrm > 0){
+ std::cerr<<"FAIL shift "<< shift<<" in dir "<< dir
+ <<" ["<<coor[0]<<","<<coor[1]<<","<<coor[2]<<","<<coor[3]<<"] = "
+ << cm()()()<<" expect "<<scm<<" "<<nrm<<std::endl;
+ std::cerr<<"Got "<<index<<" " << peer[0]<<","<<peer[1]<<","<<peer[2]<<","<<peer[3]<<std::endl;
+ index=real(scm);
+ Fine.CoorFromIndex(peer,index,latt_size);
+ std::cerr<<"Expect "<<index<<" " << peer[0]<<","<<peer[1]<<","<<peer[2]<<","<<peer[3]<<std::endl;
+ exit(-1);
+ } else if (0) {
+ std::cout<<"PASS shift "<< shift<<" in dir "<< dir
+ <<" ["<<coor[0]<<","<<coor[1]<<","<<coor[2]<<","<<coor[3]<<"] = "
+ << cm()()()<<" expect "<<scm<<" "<<nrm<<std::endl;
+ }
+ }}}}
+
+ }
+ }
+
+ Grid_finalize();
+}
diff --git a/tests/Makefile.am b/tests/Makefile.am
index 3c79191c..7ef4eb1f 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -5,7 +5,7 @@ AM_LDFLAGS = -L$(top_builddir)/lib
#
# Test code
#
-bin_PROGRAMS = Grid_main Grid_stencil Grid_nersc_io Grid_cshift Grid_gamma Grid_simd Grid_rng Grid_remez Grid_rng_fixed
+bin_PROGRAMS = Grid_main Grid_stencil Grid_nersc_io Grid_cshift Grid_gamma Grid_simd Grid_rng Grid_remez Grid_rng_fixed Grid_cshift_red_black
Grid_main_SOURCES = Grid_main.cc
Grid_main_LDADD = -lGrid
@@ -25,6 +25,9 @@ Grid_nersc_io_LDADD = -lGrid
Grid_cshift_SOURCES = Grid_cshift.cc
Grid_cshift_LDADD = -lGrid
+Grid_cshift_red_black_SOURCES = Grid_cshift_red_black.cc
+Grid_cshift_red_black_LDADD = -lGrid
+
Grid_gamma_SOURCES = Grid_gamma.cc
Grid_gamma_LDADD = -lGrid
From 8272e15bd65e439864b91877d69641c1fb5a1062 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Sun, 31 May 2015 15:11:09 +0100
Subject: [PATCH 03/20] Updated line counter
---
scripts/linecount | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/scripts/linecount b/scripts/linecount
index 350abf1c..7624c15f 100755
--- a/scripts/linecount
+++ b/scripts/linecount
@@ -1,3 +1,3 @@
#!/bin/sh
-wc -l lib/*.h lib/*/*.h lib/*/*/*.h lib/*.cc lib/*/*.cc lib/*/*/*.cc tests/*.cc benchmarks/*.cc
\ No newline at end of file
+wc -l lib/*.h lib/*/*.h lib/*/*/*.h lib/*.cc lib/*/*.cc lib/*/*/*.cc tests/*.cc benchmarks/*.cc lib/*/*/*/*.cc lib/*/*/*/*.h
\ No newline at end of file
From ce8c7a77b60ffa2d418097663125fc61163bf018 Mon Sep 17 00:00:00 2001
From: azusayamaguchi <ayamaguc@YAMAKAZE.local>
Date: Sun, 31 May 2015 18:50:08 +0100
Subject: [PATCH 04/20] Bug in Makefile.am fixed
---
lib/Makefile.am | 1 -
1 file changed, 1 deletion(-)
diff --git a/lib/Makefile.am b/lib/Makefile.am
index aa531df7..e983b432 100644
--- a/lib/Makefile.am
+++ b/lib/Makefile.am
@@ -103,7 +103,6 @@ nobase_include_HEADERS=\
./qcd/Dirac.h\
./qcd/QCD.h\
./qcd/TwoSpinor.h\
- ./qcd/FermionAction.h\
./simd/Grid_avx.h\
./simd/Grid_avx512.h\
./simd/Grid_qpx.h\
From 9ea64767b05bc4c468da089e2c49bbc6813c2fa5 Mon Sep 17 00:00:00 2001
From: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
Date: Sun, 31 May 2015 22:50:03 +0100
Subject: [PATCH 05/20] No compile fix on mpi target
---
lib/cshift/Grid_cshift_mpi.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/cshift/Grid_cshift_mpi.h b/lib/cshift/Grid_cshift_mpi.h
index 8c0badcd..3b3fad29 100644
--- a/lib/cshift/Grid_cshift_mpi.h
+++ b/lib/cshift/Grid_cshift_mpi.h
@@ -17,7 +17,7 @@ template<class vobj> Lattice<vobj> Cshift(Lattice<vobj> &rhs,int dimension,int s
// Map to always positive shift modulo global full dimension.
shift = (shift+fd)%fd;
- ret.checkerboard = rhs._grid->CheckerBoardDestination(rhs.checkerboard,shift);
+ ret.checkerboard = rhs._grid->CheckerBoardDestination(rhs.checkerboard,shift,dimension);
// the permute type
int simd_layout = rhs._grid->_simd_layout[dimension];
From 4c617c36430977fecec6366104e3d3b27d69caf0 Mon Sep 17 00:00:00 2001
From: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
Date: Mon, 1 Jun 2015 12:25:59 +0100
Subject: [PATCH 06/20] Const safety
---
lib/cshift/Grid_cshift_common.h | 8 ++++----
lib/cshift/Grid_cshift_mpi.h | 10 +++++-----
lib/cshift/Grid_cshift_none.h | 2 +-
3 files changed, 10 insertions(+), 10 deletions(-)
diff --git a/lib/cshift/Grid_cshift_common.h b/lib/cshift/Grid_cshift_common.h
index 06e812d9..ffa3f771 100644
--- a/lib/cshift/Grid_cshift_common.h
+++ b/lib/cshift/Grid_cshift_common.h
@@ -153,7 +153,7 @@ PARALLEL_NESTED_LOOP2
//////////////////////////////////////////////////////
// local to node block strided copies
//////////////////////////////////////////////////////
-template<class vobj> void Copy_plane(Lattice<vobj>& lhs,Lattice<vobj> &rhs, int dimension,int lplane,int rplane,int cbmask)
+template<class vobj> void Copy_plane(Lattice<vobj>& lhs,const Lattice<vobj> &rhs, int dimension,int lplane,int rplane,int cbmask)
{
int rd = rhs._grid->_rdimensions[dimension];
@@ -182,7 +182,7 @@ PARALLEL_NESTED_LOOP2
}
-template<class vobj> void Copy_plane_permute(Lattice<vobj>& lhs,Lattice<vobj> &rhs, int dimension,int lplane,int rplane,int cbmask,int permute_type)
+template<class vobj> void Copy_plane_permute(Lattice<vobj>& lhs,const Lattice<vobj> &rhs, int dimension,int lplane,int rplane,int cbmask,int permute_type)
{
int rd = rhs._grid->_rdimensions[dimension];
@@ -212,7 +212,7 @@ PARALLEL_NESTED_LOOP2
//////////////////////////////////////////////////////
// Local to node Cshift
//////////////////////////////////////////////////////
-template<class vobj> void Cshift_local(Lattice<vobj>& ret,Lattice<vobj> &rhs,int dimension,int shift)
+template<class vobj> void Cshift_local(Lattice<vobj>& ret,const Lattice<vobj> &rhs,int dimension,int shift)
{
int sshift[2];
@@ -227,7 +227,7 @@ template<class vobj> void Cshift_local(Lattice<vobj>& ret,Lattice<vobj> &rhs,int
}
}
-template<class vobj> Lattice<vobj> Cshift_local(Lattice<vobj> &ret,Lattice<vobj> &rhs,int dimension,int shift,int cbmask)
+template<class vobj> Lattice<vobj> Cshift_local(Lattice<vobj> &ret,const Lattice<vobj> &rhs,int dimension,int shift,int cbmask)
{
GridBase *grid = rhs._grid;
int fd = grid->_fdimensions[dimension];
diff --git a/lib/cshift/Grid_cshift_mpi.h b/lib/cshift/Grid_cshift_mpi.h
index 3b3fad29..9cdfd316 100644
--- a/lib/cshift/Grid_cshift_mpi.h
+++ b/lib/cshift/Grid_cshift_mpi.h
@@ -4,7 +4,7 @@
namespace Grid {
-template<class vobj> Lattice<vobj> Cshift(Lattice<vobj> &rhs,int dimension,int shift)
+template<class vobj> Lattice<vobj> Cshift(const Lattice<vobj> &rhs,int dimension,int shift)
{
typedef typename vobj::vector_type vector_type;
typedef typename vobj::scalar_type scalar_type;
@@ -35,7 +35,7 @@ template<class vobj> Lattice<vobj> Cshift(Lattice<vobj> &rhs,int dimension,int s
return ret;
}
-template<class vobj> void Cshift_comms(Lattice<vobj>& ret,Lattice<vobj> &rhs,int dimension,int shift)
+template<class vobj> void Cshift_comms(Lattice<vobj>& ret,const Lattice<vobj> &rhs,int dimension,int shift)
{
int sshift[2];
@@ -50,7 +50,7 @@ template<class vobj> void Cshift_comms(Lattice<vobj>& ret,Lattice<vobj> &rhs,int
}
}
-template<class vobj> void Cshift_comms_simd(Lattice<vobj>& ret,Lattice<vobj> &rhs,int dimension,int shift)
+template<class vobj> void Cshift_comms_simd(Lattice<vobj>& ret,const Lattice<vobj> &rhs,int dimension,int shift)
{
int sshift[2];
@@ -65,7 +65,7 @@ template<class vobj> void Cshift_comms_simd(Lattice<vobj>& ret,Lattice<vobj> &rh
}
}
-template<class vobj> void Cshift_comms(Lattice<vobj> &ret,Lattice<vobj> &rhs,int dimension,int shift,int cbmask)
+template<class vobj> void Cshift_comms(Lattice<vobj> &ret,const Lattice<vobj> &rhs,int dimension,int shift,int cbmask)
{
typedef typename vobj::vector_type vector_type;
typedef typename vobj::scalar_type scalar_type;
@@ -124,7 +124,7 @@ template<class vobj> void Cshift_comms(Lattice<vobj> &ret,Lattice<vobj> &rhs,int
}
}
-template<class vobj> void Cshift_comms_simd(Lattice<vobj> &ret,Lattice<vobj> &rhs,int dimension,int shift,int cbmask)
+template<class vobj> void Cshift_comms_simd(Lattice<vobj> &ret,const Lattice<vobj> &rhs,int dimension,int shift,int cbmask)
{
GridBase *grid=rhs._grid;
const int Nsimd = grid->Nsimd();
diff --git a/lib/cshift/Grid_cshift_none.h b/lib/cshift/Grid_cshift_none.h
index be9467b1..a8c6f6c0 100644
--- a/lib/cshift/Grid_cshift_none.h
+++ b/lib/cshift/Grid_cshift_none.h
@@ -1,7 +1,7 @@
#ifndef _GRID_CSHIFT_NONE_H_
#define _GRID_CSHIFT_NONE_H_
namespace Grid {
-template<class vobj> Lattice<vobj> Cshift(Lattice<vobj> &rhs,int dimension,int shift)
+template<class vobj> Lattice<vobj> Cshift(const Lattice<vobj> &rhs,int dimension,int shift)
{
Lattice<vobj> ret(rhs._grid);
ret.checkerboard = rhs._grid->CheckerBoardDestination(rhs.checkerboard,shift,dimension);
From 8bd9fb4427350dcf900825da19a630ba6e89ad46 Mon Sep 17 00:00:00 2001
From: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
Date: Mon, 1 Jun 2015 12:26:20 +0100
Subject: [PATCH 07/20] FIx miistake
---
lib/lattice/Grid_lattice_base.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/lattice/Grid_lattice_base.h b/lib/lattice/Grid_lattice_base.h
index 4a6d3180..15ef0944 100644
--- a/lib/lattice/Grid_lattice_base.h
+++ b/lib/lattice/Grid_lattice_base.h
@@ -176,7 +176,7 @@ PARALLEL_FOR_LOOP
PARALLEL_FOR_LOOP
for(int ss=0;ss<_grid->oSites();ss++){
#ifdef STREAMING_STORES
- vobj tmp = eval(tmp,ss,expr);
+ vobj tmp = eval(ss,expr);
vstream(_odata[ss] ,tmp);
#else
_odata[ss]=eval(ss,expr);
From 2583570e17e95591d0f9dac4bddf18d3c1f6260b Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 2 Jun 2015 16:57:12 +0100
Subject: [PATCH 08/20] Domain wall fermions now invert ; have the basis set up
for Tanh/Zolo * (Cayley/PartFrac/ContFrac) * (Mobius/Shamir/Wilson) Approx
Representation Kernel.
All are done with space-time taking part in checkerboarding, Ls uncheckerboarded
Have only so far tested the Domain Wall limit of mobius, and at that only checked
that it
i) Inverts
ii) 5dim DW == Ls copies of 4dim D2
iii) MeeInv Mee == 1
iv) Meo+Mee+Moe+Moo == M unprec.
v) MpcDagMpc is hermitan
vi) Mdag is the adjoint of M between stochastic vectors.
That said, the RB schur solve, RB MpcDagMpc solve, Unprec solve
all converge and the true residual becomes small; so pretty good tests.
---
benchmarks/Grid_dwf.cc | 79 +++---
benchmarks/Grid_dwf_cg_prec.cc | 58 +++++
benchmarks/Grid_dwf_cg_schur.cc | 53 ++++
benchmarks/Grid_dwf_cg_unprec.cc | 53 ++++
benchmarks/Grid_dwf_even_odd.cc | 207 +++++++++++++++
benchmarks/Grid_wilson_evenodd.cc | 5 +-
benchmarks/Makefile.am | 46 +++-
lib/Grid_init.cc | 2 +-
lib/Makefile.am | 16 +-
lib/algorithms/LinearOperator.h | 32 ---
lib/algorithms/approx/Zolotarev.cc | 9 +-
lib/algorithms/approx/Zolotarev.h | 9 +-
lib/cartesian/Grid_cartesian_base.h | 2 +-
lib/cartesian/Grid_cartesian_full.h | 6 +-
lib/cartesian/Grid_cartesian_red_black.h | 24 +-
lib/communicator/Grid_communicator_base.h | 2 +-
lib/communicator/Grid_communicator_mpi.cc | 2 +-
lib/communicator/Grid_communicator_none.cc | 2 +-
lib/qcd/LinalgUtils.h | 113 +++++++++
lib/qcd/QCD.h | 2 +
lib/qcd/SpaceTimeGrid.cc | 52 ++++
lib/qcd/SpaceTimeGrid.h | 18 ++
lib/qcd/action/Actions.h | 78 +++++-
lib/qcd/action/fermion/CayleyFermion5D.cc | 235 ++++++++++++++++++
lib/qcd/action/fermion/CayleyFermion5D.h | 61 +++++
.../fermion/ContinuedFractionFermion5D.cc | 119 +++++++++
.../fermion/ContinuedFractionFermion5D.h | 53 ++++
lib/qcd/action/fermion/DomainWallFermion.h | 118 +++++++++
.../{FermionAction.h => FermionOperator.h} | 7 +-
.../fermion/PartialFractionFermion5D.cc | 47 ++++
.../action/fermion/PartialFractionFermion5D.h | 49 ++++
lib/qcd/action/fermion/WilsonFermion.cc | 6 +-
lib/qcd/action/fermion/WilsonFermion.h | 6 +-
...DimWilsonFermion.cc => WilsonFermion5D.cc} | 106 +++-----
...veDimWilsonFermion.h => WilsonFermion5D.h} | 31 +--
lib/stencil/Grid_stencil_common.cc | 12 +-
36 files changed, 1500 insertions(+), 220 deletions(-)
create mode 100644 benchmarks/Grid_dwf_cg_prec.cc
create mode 100644 benchmarks/Grid_dwf_cg_schur.cc
create mode 100644 benchmarks/Grid_dwf_cg_unprec.cc
create mode 100644 benchmarks/Grid_dwf_even_odd.cc
create mode 100644 lib/qcd/LinalgUtils.h
create mode 100644 lib/qcd/SpaceTimeGrid.cc
create mode 100644 lib/qcd/SpaceTimeGrid.h
create mode 100644 lib/qcd/action/fermion/CayleyFermion5D.cc
create mode 100644 lib/qcd/action/fermion/CayleyFermion5D.h
create mode 100644 lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
create mode 100644 lib/qcd/action/fermion/ContinuedFractionFermion5D.h
create mode 100644 lib/qcd/action/fermion/DomainWallFermion.h
rename lib/qcd/action/fermion/{FermionAction.h => FermionOperator.h} (92%)
create mode 100644 lib/qcd/action/fermion/PartialFractionFermion5D.cc
create mode 100644 lib/qcd/action/fermion/PartialFractionFermion5D.h
rename lib/qcd/action/fermion/{FiveDimWilsonFermion.cc => WilsonFermion5D.cc} (68%)
rename lib/qcd/action/fermion/{FiveDimWilsonFermion.h => WilsonFermion5D.h} (72%)
diff --git a/benchmarks/Grid_dwf.cc b/benchmarks/Grid_dwf.cc
index eb1d9299..62869b17 100644
--- a/benchmarks/Grid_dwf.cc
+++ b/benchmarks/Grid_dwf.cc
@@ -24,43 +24,28 @@ int main (int argc, char ** argv)
std::cout << "Grid is setup to use "<<threads<<" threads"<<std::endl;
std::vector<int> latt4 = GridDefaultLatt();
- std::vector<int> simd4 = GridDefaultSimd(Nd,vComplexF::Nsimd());
- std::vector<int> mpi4 = GridDefaultMpi();
-
- assert(latt4.size()==4 );
- assert(simd4.size()==4 );
- assert(mpi4.size() ==4 );
-
- const int Ls=1;
- std::vector<int> latt5({Ls,latt4[0],latt4[1],latt4[2],latt4[3]});
- std::vector<int> simd5({1 ,simd4[0],simd4[1],simd4[2],simd4[3]});
- std::vector<int> mpi5({1 , mpi4[0], mpi4[1], mpi4[2], mpi4[3]});
- std::vector<int> cb5({0,1,1,1,1}); // Checkerboard 4d only
- int cbd=1; // use dim-1 to reduce
-
- // Four dim grid for gauge field U
- GridCartesian UGrid(latt4,simd4,mpi4);
- GridRedBlackCartesian UrbGrid(&UGrid);
-
- // Five dim grid for fermions F
- GridCartesian FGrid(latt5,simd5,mpi5);
- GridRedBlackCartesian FrbGrid(latt5,simd5,mpi5,cb5,cbd);
+ const int Ls=8;
+ GridCartesian * UGrid = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+ 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> seeds5({5,6,7,8});
- GridParallelRNG RNG5(&FGrid); RNG5.SeedFixedIntegers(seeds5);
- LatticeFermion src (&FGrid); random(RNG5,src);
- LatticeFermion result(&FGrid); result=zero;
- LatticeFermion ref(&FGrid); ref=zero;
- LatticeFermion tmp(&FGrid);
- LatticeFermion err(&FGrid);
+ GridParallelRNG RNG4(UGrid); RNG4.SeedFixedIntegers(seeds4);
+ GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
+
+ LatticeFermion src (FGrid); random(RNG5,src);
+ LatticeFermion result(FGrid); result=zero;
+ LatticeFermion ref(FGrid); ref=zero;
+ LatticeFermion tmp(FGrid);
+ LatticeFermion err(FGrid);
ColourMatrix cm = Complex(1.0,0.0);
- GridParallelRNG RNG4(&UGrid); RNG4.SeedFixedIntegers(seeds4);
- LatticeGaugeField Umu(&UGrid); random(RNG4,Umu);
- LatticeGaugeField Umu5d(&FGrid);
+ LatticeGaugeField Umu(UGrid); random(RNG4,Umu);
+ LatticeGaugeField Umu5d(FGrid);
// replicate across fifth dimension
for(int ss=0;ss<Umu._grid->oSites();ss++){
@@ -72,7 +57,7 @@ int main (int argc, char ** argv)
////////////////////////////////////
// Naive wilson implementation
////////////////////////////////////
- std::vector<LatticeColourMatrix> U(4,&FGrid);
+ std::vector<LatticeColourMatrix> U(4,FGrid);
for(int mu=0;mu<Nd;mu++){
U[mu] = peekIndex<LorentzIndex>(Umu5d,mu);
}
@@ -93,17 +78,17 @@ int main (int argc, char ** argv)
}
RealD mass=0.1;
- FiveDimWilsonFermion Dw(Umu,FGrid,FrbGrid,UGrid,UrbGrid,mass);
+ RealD M5 =1.8;
+ DomainWallFermion Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
std::cout << "Calling Dw"<<std::endl;
- int ncall=1000;
+ int ncall=10;
double t0=usecond();
for(int i=0;i<ncall;i++){
Dw.Dhop(src,result,0);
}
double t1=usecond();
-
double volume=Ls; for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
double flops=1344*volume*ncall;
@@ -141,5 +126,31 @@ int main (int argc, char ** argv)
err = ref-result;
std::cout << "norm diff "<< norm2(err)<<std::endl;
+ LatticeFermion src_e (FrbGrid);
+ LatticeFermion src_o (FrbGrid);
+ LatticeFermion r_e (FrbGrid);
+ LatticeFermion r_o (FrbGrid);
+ LatticeFermion r_eo (FGrid);
+
+
+ std::cout << "Calling Deo and Doe"<<std::endl;
+ pickCheckerboard(Even,src_e,src);
+ pickCheckerboard(Odd,src_o,src);
+
+ Dw.DhopEO(src_o,r_e,DaggerNo);
+ Dw.DhopOE(src_e,r_o,DaggerNo);
+ Dw.Dhop(src,result,DaggerNo);
+
+ setCheckerboard(r_eo,r_o);
+ setCheckerboard(r_eo,r_e);
+
+ err = r_eo-result;
+ std::cout << "norm diff "<< norm2(err)<<std::endl;
+
+ pickCheckerboard(Even,src_e,err);
+ pickCheckerboard(Odd,src_o,err);
+ std::cout << "norm diff even "<< norm2(src_e)<<std::endl;
+ std::cout << "norm diff odd "<< norm2(src_o)<<std::endl;
+
Grid_finalize();
}
diff --git a/benchmarks/Grid_dwf_cg_prec.cc b/benchmarks/Grid_dwf_cg_prec.cc
new file mode 100644
index 00000000..0cf86a19
--- /dev/null
+++ b/benchmarks/Grid_dwf_cg_prec.cc
@@ -0,0 +1,58 @@
+#include <Grid.h>
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+template<class d>
+struct scal {
+ d internal;
+};
+
+ Gamma::GammaMatrix Gmu [] = {
+ Gamma::GammaX,
+ Gamma::GammaY,
+ Gamma::GammaZ,
+ Gamma::GammaT
+ };
+
+int main (int argc, char ** argv)
+{
+ Grid_init(&argc,&argv);
+
+ const int Ls=8;
+
+ GridCartesian * UGrid = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+ 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> seeds5({5,6,7,8});
+ GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
+ GridParallelRNG RNG4(UGrid); RNG4.SeedFixedIntegers(seeds4);
+
+ LatticeFermion src(FGrid); random(RNG5,src);
+ LatticeFermion result(FGrid); result=zero;
+ LatticeGaugeField Umu(UGrid); random(RNG4,Umu);
+
+ 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;
+ DomainWallFermion Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+
+ LatticeFermion src_o(FrbGrid);
+ LatticeFermion result_o(FrbGrid);
+ pickCheckerboard(Odd,src_o,src);
+ result_o=zero;
+
+ HermitianCheckerBoardedOperator<DomainWallFermion,LatticeFermion> HermOpEO(Ddwf);
+ ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
+ CG(HermOpEO,src_o,result_o);
+
+ Grid_finalize();
+}
diff --git a/benchmarks/Grid_dwf_cg_schur.cc b/benchmarks/Grid_dwf_cg_schur.cc
new file mode 100644
index 00000000..aac4d3fd
--- /dev/null
+++ b/benchmarks/Grid_dwf_cg_schur.cc
@@ -0,0 +1,53 @@
+#include <Grid.h>
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+template<class d>
+struct scal {
+ d internal;
+};
+
+ Gamma::GammaMatrix Gmu [] = {
+ Gamma::GammaX,
+ Gamma::GammaY,
+ Gamma::GammaZ,
+ Gamma::GammaT
+ };
+
+int main (int argc, char ** argv)
+{
+ Grid_init(&argc,&argv);
+
+ const int Ls=8;
+
+ GridCartesian * UGrid = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+ 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> seeds5({5,6,7,8});
+ GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
+ GridParallelRNG RNG4(UGrid); RNG4.SeedFixedIntegers(seeds4);
+
+ LatticeFermion src(FGrid); random(RNG5,src);
+ LatticeFermion result(FGrid); result=zero;
+ LatticeGaugeField Umu(UGrid); random(RNG4,Umu);
+
+ 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;
+ DomainWallFermion Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+
+ ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
+ SchurRedBlackSolve<LatticeFermion> SchurSolver(CG);
+ SchurSolver(Ddwf,src,result);
+
+ Grid_finalize();
+}
diff --git a/benchmarks/Grid_dwf_cg_unprec.cc b/benchmarks/Grid_dwf_cg_unprec.cc
new file mode 100644
index 00000000..5c9e7ad3
--- /dev/null
+++ b/benchmarks/Grid_dwf_cg_unprec.cc
@@ -0,0 +1,53 @@
+#include <Grid.h>
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+template<class d>
+struct scal {
+ d internal;
+};
+
+ Gamma::GammaMatrix Gmu [] = {
+ Gamma::GammaX,
+ Gamma::GammaY,
+ Gamma::GammaZ,
+ Gamma::GammaT
+ };
+
+int main (int argc, char ** argv)
+{
+ Grid_init(&argc,&argv);
+
+ const int Ls=8;
+
+ GridCartesian * UGrid = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+ 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> seeds5({5,6,7,8});
+ GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
+ GridParallelRNG RNG4(UGrid); RNG4.SeedFixedIntegers(seeds4);
+
+ LatticeFermion src(FGrid); random(RNG5,src);
+ LatticeFermion result(FGrid); result=zero;
+ LatticeGaugeField Umu(UGrid); random(RNG4,Umu);
+
+ 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;
+ DomainWallFermion Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+
+ HermitianOperator<DomainWallFermion,LatticeFermion> HermOp(Ddwf);
+ ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
+ CG(HermOp,src,result);
+
+ Grid_finalize();
+}
diff --git a/benchmarks/Grid_dwf_even_odd.cc b/benchmarks/Grid_dwf_even_odd.cc
new file mode 100644
index 00000000..ac47bbf9
--- /dev/null
+++ b/benchmarks/Grid_dwf_even_odd.cc
@@ -0,0 +1,207 @@
+#include <Grid.h>
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+template<class d>
+struct scal {
+ d internal;
+};
+
+ Gamma::GammaMatrix Gmu [] = {
+ Gamma::GammaX,
+ Gamma::GammaY,
+ Gamma::GammaZ,
+ Gamma::GammaT
+ };
+
+
+int main (int argc, char ** argv)
+{
+ Grid_init(&argc,&argv);
+
+ int threads = GridThread::GetThreads();
+ std::cout << "Grid is setup to use "<<threads<<" threads"<<std::endl;
+
+
+ const int Ls=8;
+ GridCartesian * UGrid = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+ 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> seeds5({5,6,7,8});
+
+ GridParallelRNG RNG4(UGrid); RNG4.SeedFixedIntegers(seeds4);
+ GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
+
+ LatticeFermion src (FGrid); random(RNG5,src);
+ LatticeFermion phi (FGrid); random(RNG5,phi);
+ LatticeFermion chi (FGrid); random(RNG5,chi);
+ LatticeFermion result(FGrid); result=zero;
+ LatticeFermion ref(FGrid); ref=zero;
+ LatticeFermion tmp(FGrid); tmp=zero;
+ LatticeFermion err(FGrid); tmp=zero;
+ LatticeGaugeField Umu(UGrid); random(RNG4,Umu);
+ std::vector<LatticeColourMatrix> U(4,UGrid);
+
+ // Only one non-zero (y)
+ Umu=zero;
+ for(int nn=0;nn<Nd;nn++){
+ random(RNG4,U[nn]);
+ if ( nn>0 )
+ U[nn]=zero;
+ pokeIndex<LorentzIndex>(Umu,U[nn],nn);
+ }
+
+ RealD mass=0.1;
+ RealD M5 =1.8;
+ DomainWallFermion Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+
+ LatticeFermion src_e (FrbGrid);
+ LatticeFermion src_o (FrbGrid);
+ LatticeFermion r_e (FrbGrid);
+ LatticeFermion r_o (FrbGrid);
+ LatticeFermion r_eo (FGrid);
+ LatticeFermion r_eeoo(FGrid);
+
+ std::cout<<"=========================================================="<<std::endl;
+ std::cout<<"= Testing that Meo + Moe + Moo + Mee = Munprec "<<std::endl;
+ std::cout<<"=========================================================="<<std::endl;
+
+ pickCheckerboard(Even,src_e,src);
+ pickCheckerboard(Odd,src_o,src);
+
+ Ddwf.Meooe(src_e,r_o); std::cout<<"Applied Meo"<<std::endl;
+ Ddwf.Meooe(src_o,r_e); std::cout<<"Applied Moe"<<std::endl;
+ setCheckerboard(r_eo,r_o);
+ setCheckerboard(r_eo,r_e);
+
+ Ddwf.Mooee(src_e,r_e); std::cout<<"Applied Mee"<<std::endl;
+ Ddwf.Mooee(src_o,r_o); std::cout<<"Applied Moo"<<std::endl;
+ setCheckerboard(r_eeoo,r_e);
+ setCheckerboard(r_eeoo,r_o);
+
+ r_eo=r_eo+r_eeoo;
+ Ddwf.M(src,ref);
+
+ // std::cout << r_eo<<std::endl;
+ // std::cout << ref <<std::endl;
+
+ err= ref - r_eo;
+ std::cout << "EO norm diff "<< norm2(err)<< " "<<norm2(ref)<< " " << norm2(r_eo) <<std::endl;
+
+ LatticeComplex cerr(FGrid);
+ cerr = localInnerProduct(err,err);
+ // std::cout << cerr<<std::endl;
+
+ std::cout<<"=============================================================="<<std::endl;
+ std::cout<<"= Test Ddagger is the dagger of D by requiring "<<std::endl;
+ std::cout<<"= < phi | Deo | chi > * = < chi | Deo^dag| phi> "<<std::endl;
+ std::cout<<"=============================================================="<<std::endl;
+
+ LatticeFermion chi_e (FrbGrid);
+ LatticeFermion chi_o (FrbGrid);
+
+ LatticeFermion dchi_e (FrbGrid);
+ LatticeFermion dchi_o (FrbGrid);
+
+ LatticeFermion phi_e (FrbGrid);
+ LatticeFermion phi_o (FrbGrid);
+
+ LatticeFermion dphi_e (FrbGrid);
+ LatticeFermion dphi_o (FrbGrid);
+
+
+ pickCheckerboard(Even,chi_e,chi);
+ pickCheckerboard(Odd ,chi_o,chi);
+ pickCheckerboard(Even,phi_e,phi);
+ pickCheckerboard(Odd ,phi_o,phi);
+
+ Ddwf.Meooe(chi_e,dchi_o);
+ Ddwf.Meooe(chi_o,dchi_e);
+ Ddwf.MeooeDag(phi_e,dphi_o);
+ Ddwf.MeooeDag(phi_o,dphi_e);
+
+ ComplexD pDce = innerProduct(phi_e,dchi_e);
+ ComplexD pDco = innerProduct(phi_o,dchi_o);
+ ComplexD cDpe = innerProduct(chi_e,dphi_e);
+ ComplexD cDpo = innerProduct(chi_o,dphi_o);
+
+ std::cout <<"e "<<pDce<<" "<<cDpe <<std::endl;
+ std::cout <<"o "<<pDco<<" "<<cDpo <<std::endl;
+
+ std::cout <<"pDce - conj(cDpo) "<< pDce-conj(cDpo) <<std::endl;
+ std::cout <<"pDco - conj(cDpe) "<< pDco-conj(cDpe) <<std::endl;
+
+ std::cout<<"=============================================================="<<std::endl;
+ std::cout<<"= Test MeeInv Mee = 1 "<<std::endl;
+ std::cout<<"=============================================================="<<std::endl;
+
+ pickCheckerboard(Even,chi_e,chi);
+ pickCheckerboard(Odd ,chi_o,chi);
+
+ Ddwf.Mooee(chi_e,src_e);
+ Ddwf.MooeeInv(src_e,phi_e);
+
+ Ddwf.Mooee(chi_o,src_o);
+ Ddwf.MooeeInv(src_o,phi_o);
+
+ setCheckerboard(phi,phi_e);
+ setCheckerboard(phi,phi_o);
+
+ err = phi-chi;
+ std::cout << "norm diff "<< norm2(err)<< std::endl;
+
+ std::cout<<"=============================================================="<<std::endl;
+ std::cout<<"= Test MeeInvDag MeeDag = 1 "<<std::endl;
+ std::cout<<"=============================================================="<<std::endl;
+
+ pickCheckerboard(Even,chi_e,chi);
+ pickCheckerboard(Odd ,chi_o,chi);
+
+ Ddwf.MooeeDag(chi_e,src_e);
+ Ddwf.MooeeInvDag(src_e,phi_e);
+
+ Ddwf.MooeeDag(chi_o,src_o);
+ Ddwf.MooeeInvDag(src_o,phi_o);
+
+ setCheckerboard(phi,phi_e);
+ setCheckerboard(phi,phi_o);
+
+ err = phi-chi;
+ std::cout << "norm diff "<< norm2(err)<< std::endl;
+
+ std::cout<<"=============================================================="<<std::endl;
+ std::cout<<"= Test MpcDagMpc is Hermitian "<<std::endl;
+ std::cout<<"=============================================================="<<std::endl;
+
+ random(RNG5,phi);
+ random(RNG5,chi);
+ pickCheckerboard(Even,chi_e,chi);
+ pickCheckerboard(Odd ,chi_o,chi);
+ pickCheckerboard(Even,phi_e,phi);
+ pickCheckerboard(Odd ,phi_o,phi);
+ RealD t1,t2;
+
+ Ddwf.MpcDagMpc(chi_e,dchi_e,t1,t2);
+ Ddwf.MpcDagMpc(chi_o,dchi_o,t1,t2);
+
+ Ddwf.MpcDagMpc(phi_e,dphi_e,t1,t2);
+ Ddwf.MpcDagMpc(phi_o,dphi_o,t1,t2);
+
+ pDce = innerProduct(phi_e,dchi_e);
+ pDco = innerProduct(phi_o,dchi_o);
+ cDpe = innerProduct(chi_e,dphi_e);
+ cDpo = innerProduct(chi_o,dphi_o);
+
+ std::cout <<"e "<<pDce<<" "<<cDpe <<std::endl;
+ std::cout <<"o "<<pDco<<" "<<cDpo <<std::endl;
+
+ std::cout <<"pDce - conj(cDpo) "<< pDco-conj(cDpo) <<std::endl;
+ std::cout <<"pDco - conj(cDpe) "<< pDce-conj(cDpe) <<std::endl;
+
+ Grid_finalize();
+}
diff --git a/benchmarks/Grid_wilson_evenodd.cc b/benchmarks/Grid_wilson_evenodd.cc
index a073139d..3ebc4709 100644
--- a/benchmarks/Grid_wilson_evenodd.cc
+++ b/benchmarks/Grid_wilson_evenodd.cc
@@ -68,8 +68,6 @@ int main (int argc, char ** argv)
LatticeFermion r_o (&RBGrid);
LatticeFermion r_eo (&Grid);
- const int Even=0;
- const int Odd=1;
std::cout<<"=========================================================="<<std::endl;
std::cout<<"= Testing that Deo + Doe = Dunprec "<<std::endl;
std::cout<<"=========================================================="<<std::endl;
@@ -79,12 +77,11 @@ int main (int argc, char ** argv)
Dw.Meooe(src_e,r_o); std::cout<<"Applied Meo"<<std::endl;
Dw.Meooe(src_o,r_e); std::cout<<"Applied Moe"<<std::endl;
- Dw.Dhop (src,ref,0);
+ Dw.Dhop (src,ref,DaggerNo);
setCheckerboard(r_eo,r_o);
setCheckerboard(r_eo,r_e);
- ref = (-0.5)*ref;
err= ref - r_eo;
std::cout << "EO norm diff "<< norm2(err)<< " "<<norm2(ref)<< " " << norm2(r_eo) <<std::endl;
diff --git a/benchmarks/Makefile.am b/benchmarks/Makefile.am
index b2649669..e92f2e84 100644
--- a/benchmarks/Makefile.am
+++ b/benchmarks/Makefile.am
@@ -5,14 +5,34 @@ AM_LDFLAGS = -L$(top_builddir)/lib
#
# Test code
#
-bin_PROGRAMS = Grid_wilson Grid_comms Grid_memory_bandwidth Grid_su3 Grid_wilson_cg_unprec Grid_wilson_evenodd Grid_wilson_cg_prec Grid_wilson_cg_schur Grid_dwf
+bin_PROGRAMS = \
+ Grid_comms \
+ Grid_memory_bandwidth \
+ Grid_su3 \
+ Grid_wilson \
+ Grid_wilson_evenodd \
+ Grid_wilson_cg_unprec \
+ Grid_wilson_cg_prec \
+ Grid_wilson_cg_schur \
+ Grid_dwf\
+ Grid_dwf_even_odd\
+ Grid_dwf_cg_unprec\
+ Grid_dwf_cg_prec\
+ Grid_dwf_cg_schur
+
+Grid_comms_SOURCES = Grid_comms.cc
+Grid_comms_LDADD = -lGrid
+
+Grid_su3_SOURCES = Grid_su3.cc Grid_su3_test.cc Grid_su3_expr.cc
+Grid_su3_LDADD = -lGrid
+
+Grid_memory_bandwidth_SOURCES = Grid_memory_bandwidth.cc
+Grid_memory_bandwidth_LDADD = -lGrid
+
Grid_wilson_SOURCES = Grid_wilson.cc
Grid_wilson_LDADD = -lGrid
-Grid_dwf_SOURCES = Grid_dwf.cc
-Grid_dwf_LDADD = -lGrid
-
Grid_wilson_evenodd_SOURCES = Grid_wilson_evenodd.cc
Grid_wilson_evenodd_LDADD = -lGrid
@@ -25,12 +45,18 @@ Grid_wilson_cg_prec_LDADD = -lGrid
Grid_wilson_cg_schur_SOURCES = Grid_wilson_cg_schur.cc
Grid_wilson_cg_schur_LDADD = -lGrid
-Grid_comms_SOURCES = Grid_comms.cc
-Grid_comms_LDADD = -lGrid
+Grid_dwf_SOURCES = Grid_dwf.cc
+Grid_dwf_LDADD = -lGrid
-Grid_su3_SOURCES = Grid_su3.cc Grid_su3_test.cc Grid_su3_expr.cc
-Grid_su3_LDADD = -lGrid
+Grid_dwf_even_odd_SOURCES = Grid_dwf_even_odd.cc
+Grid_dwf_even_odd_LDADD = -lGrid
-Grid_memory_bandwidth_SOURCES = Grid_memory_bandwidth.cc
-Grid_memory_bandwidth_LDADD = -lGrid
+Grid_dwf_cg_unprec_SOURCES = Grid_dwf_cg_unprec.cc
+Grid_dwf_cg_unprec_LDADD = -lGrid
+
+Grid_dwf_cg_prec_SOURCES = Grid_dwf_cg_prec.cc
+Grid_dwf_cg_prec_LDADD = -lGrid
+
+Grid_dwf_cg_schur_SOURCES = Grid_dwf_cg_schur.cc
+Grid_dwf_cg_schur_LDADD = -lGrid
diff --git a/lib/Grid_init.cc b/lib/Grid_init.cc
index 580e602a..20371a4b 100644
--- a/lib/Grid_init.cc
+++ b/lib/Grid_init.cc
@@ -143,7 +143,7 @@ void Grid_init(int *argc,char ***argv)
}
if( GridCmdOptionExists(*argv,*argv+*argc,"--dslash-opt") ){
WilsonFermion::HandOptDslash=1;
- FiveDimWilsonFermion::HandOptDslash=1;
+ WilsonFermion5D::HandOptDslash=1;
}
if( GridCmdOptionExists(*argv,*argv+*argc,"--lebesgue") ){
LebesgueOrder::UseLebesgueOrder=1;
diff --git a/lib/Makefile.am b/lib/Makefile.am
index e983b432..b688d8b2 100644
--- a/lib/Makefile.am
+++ b/lib/Makefile.am
@@ -20,13 +20,17 @@ libGrid_a_SOURCES = \
stencil/Grid_stencil_common.cc \
algorithms/approx/Zolotarev.cc \
algorithms/approx/Remez.cc \
- qcd/action/fermion/FiveDimWilsonFermion.cc\
- qcd/action/fermion/WilsonFermion.cc\
+ qcd/SpaceTimeGrid.cc\
+ qcd/Dirac.cc\
qcd/action/fermion/WilsonKernels.cc\
qcd/action/fermion/WilsonKernelsHand.cc\
- qcd/Dirac.cc\
+ qcd/action/fermion/WilsonFermion.cc\
+ qcd/action/fermion/WilsonFermion5D.cc\
+ qcd/action/fermion/CayleyFermion5D.cc \
+ qcd/action/fermion/ContinuedFractionFermion5D.cc \
$(extra_sources)
+# qcd/action/fermion/PartialFractionFermion5D.cc \
#
# Include files
#
@@ -95,11 +99,11 @@ nobase_include_HEADERS=\
./math/Grid_math_transpose.h\
./parallelIO/GridNerscIO.h\
./qcd/action/Actions.h\
- ./qcd/action/fermion/FermionAction.h\
- ./qcd/action/fermion/FiveDimWilsonFermion.h\
+ ./qcd/action/fermion/FermionOperator.h\
./qcd/action/fermion/WilsonCompressor.h\
- ./qcd/action/fermion/WilsonFermion.h\
./qcd/action/fermion/WilsonKernels.h\
+ ./qcd/action/fermion/WilsonFermion.h\
+ ./qcd/action/fermion/WilsonFermion5D.h\
./qcd/Dirac.h\
./qcd/QCD.h\
./qcd/TwoSpinor.h\
diff --git a/lib/algorithms/LinearOperator.h b/lib/algorithms/LinearOperator.h
index cb6240ba..5839f2c1 100644
--- a/lib/algorithms/LinearOperator.h
+++ b/lib/algorithms/LinearOperator.h
@@ -125,39 +125,7 @@ namespace Grid {
};
*/
- // Chroma interface defining GaugeAction
- /*
- template<typename P, typename Q> class GaugeAction
- virtual const CreateGaugeState<P,Q>& getCreateState() const = 0;
- virtual GaugeState<P,Q>* createState(const Q& q) const
- virtual const GaugeBC<P,Q>& getGaugeBC() const
- virtual const Set& getSet(void) const = 0;
- virtual void deriv(P& result, const Handle< GaugeState<P,Q> >& state) const
- virtual Double S(const Handle< GaugeState<P,Q> >& state) const = 0;
- class LinearGaugeAction : public GaugeAction< multi1d<LatticeColorMatrix>, multi1d<LatticeColorMatrix> >
- typedef multi1d<LatticeColorMatrix> P;
- typedef multi1d<LatticeColorMatrix> Q;
- virtual void staple(LatticeColorMatrix& result,
- const Handle< GaugeState<P,Q> >& state,
- int mu, int cb) const = 0;
- */
-
- // Chroma interface defining FermionAction
- /*
- template<typename T, typename P, typename Q> class FermAct4D : public FermionAction<T,P,Q>
- virtual LinearOperator<T>* linOp(Handle< FermState<T,P,Q> > state) const = 0;
- virtual LinearOperator<T>* lMdagM(Handle< FermState<T,P,Q> > state) const = 0;
- virtual LinOpSystemSolver<T>* invLinOp(Handle< FermState<T,P,Q> > state,
- virtual MdagMSystemSolver<T>* invMdagM(Handle< FermState<T,P,Q> > state,
- virtual LinOpMultiSystemSolver<T>* mInvLinOp(Handle< FermState<T,P,Q> > state,
- virtual MdagMMultiSystemSolver<T>* mInvMdagM(Handle< FermState<T,P,Q> > state,
- virtual MdagMMultiSystemSolverAccumulate<T>* mInvMdagMAcc(Handle< FermState<T,P,Q> > state,
- virtual SystemSolver<T>* qprop(Handle< FermState<T,P,Q> > state,
- class DiffFermAct4D : public FermAct4D<T,P,Q>
- virtual DiffLinearOperator<T,Q,P>* linOp(Handle< FermState<T,P,Q> > state) const = 0;
- virtual DiffLinearOperator<T,Q,P>* lMdagM(Handle< FermState<T,P,Q> > state) const = 0;
- */
}
#endif
diff --git a/lib/algorithms/approx/Zolotarev.cc b/lib/algorithms/approx/Zolotarev.cc
index 7629bbde..c73a3436 100644
--- a/lib/algorithms/approx/Zolotarev.cc
+++ b/lib/algorithms/approx/Zolotarev.cc
@@ -58,6 +58,8 @@
/* Compute the partial fraction expansion coefficients (alpha) from the
* factored form */
+namespace Grid {
+namespace Approx {
static void construct_partfrac(izd *z) {
int dn = z -> dn, dd = z -> dd, type = z -> type;
@@ -291,7 +293,7 @@ static void sncndnFK(INTERNAL_PRECISION u, INTERNAL_PRECISION k,
* Set type = 0 for the Zolotarev approximation, which is zero at x = 0, and
* type = 1 for the approximation which is infinite at x = 0. */
-zolotarev_data* bfm_zolotarev(PRECISION epsilon, int n, int type) {
+zolotarev_data* grid_zolotarev(PRECISION epsilon, int n, int type) {
INTERNAL_PRECISION A, c, cp, kp, ksq, sn, cn, dn, Kp, Kj, z, z0, t, M, F,
l, invlambda, xi, xisq, *tv, s, opl;
int m, czero, ts;
@@ -412,7 +414,7 @@ zolotarev_data* bfm_zolotarev(PRECISION epsilon, int n, int type) {
return zd;
}
-zolotarev_data* bfm_higham(PRECISION epsilon, int n) {
+zolotarev_data* grid_higham(PRECISION epsilon, int n) {
INTERNAL_PRECISION A, M, c, cp, z, z0, t, epssq;
int m, czero;
zolotarev_data *zd;
@@ -502,6 +504,7 @@ zolotarev_data* bfm_higham(PRECISION epsilon, int n) {
free(d);
return zd;
}
+}}
#ifdef TEST
@@ -707,4 +710,6 @@ int main(int argc, char** argv) {
return EXIT_SUCCESS;
}
+
+
#endif /* TEST */
diff --git a/lib/algorithms/approx/Zolotarev.h b/lib/algorithms/approx/Zolotarev.h
index 3f0dc58e..869e5a89 100644
--- a/lib/algorithms/approx/Zolotarev.h
+++ b/lib/algorithms/approx/Zolotarev.h
@@ -1,7 +1,8 @@
/* -*- Mode: C; comment-column: 22; fill-column: 79; -*- */
#ifdef __cplusplus
-extern "C" {
+namespace Grid {
+namespace Approx {
#endif
#define HVERSION Header Time-stamp: <14-OCT-2004 09:26:51.00 adk@MISSCONTRARY>
@@ -76,10 +77,10 @@ typedef struct {
* zolotarev_data structure. The arguments must satisfy the constraints that
* epsilon > 0, n > 0, and type = 0 or 1. */
-ZOLOTAREV_DATA* bfm_higham(PRECISION epsilon, int n) ;
-ZOLOTAREV_DATA* bfm_zolotarev(PRECISION epsilon, int n, int type);
+ZOLOTAREV_DATA* grid_higham(PRECISION epsilon, int n) ;
+ZOLOTAREV_DATA* grid_zolotarev(PRECISION epsilon, int n, int type);
#endif
#ifdef __cplusplus
-}
+}}
#endif
diff --git a/lib/cartesian/Grid_cartesian_base.h b/lib/cartesian/Grid_cartesian_base.h
index e93125c1..66339648 100644
--- a/lib/cartesian/Grid_cartesian_base.h
+++ b/lib/cartesian/Grid_cartesian_base.h
@@ -21,7 +21,7 @@ public:
// Give Lattice access
template<class object> friend class Lattice;
- GridBase(std::vector<int> & processor_grid) : CartesianCommunicator(processor_grid) {};
+ GridBase(const std::vector<int> & processor_grid) : CartesianCommunicator(processor_grid) {};
// Physics Grid information.
diff --git a/lib/cartesian/Grid_cartesian_full.h b/lib/cartesian/Grid_cartesian_full.h
index 330bbfaf..2a9e0be8 100644
--- a/lib/cartesian/Grid_cartesian_full.h
+++ b/lib/cartesian/Grid_cartesian_full.h
@@ -27,9 +27,9 @@ public:
virtual int CheckerBoardShift(int source_cb,int dim,int shift, int osite){
return shift;
}
- GridCartesian(std::vector<int> &dimensions,
- std::vector<int> &simd_layout,
- std::vector<int> &processor_grid
+ GridCartesian(const std::vector<int> &dimensions,
+ const std::vector<int> &simd_layout,
+ const std::vector<int> &processor_grid
) : GridBase(processor_grid)
{
///////////////////////
diff --git a/lib/cartesian/Grid_cartesian_red_black.h b/lib/cartesian/Grid_cartesian_red_black.h
index ace36edb..3a84ed49 100644
--- a/lib/cartesian/Grid_cartesian_red_black.h
+++ b/lib/cartesian/Grid_cartesian_red_black.h
@@ -81,28 +81,28 @@ public:
}
};
- GridRedBlackCartesian(GridBase *base) : GridRedBlackCartesian(base->_fdimensions,base->_simd_layout,base->_processors) {};
+ GridRedBlackCartesian(const GridBase *base) : GridRedBlackCartesian(base->_fdimensions,base->_simd_layout,base->_processors) {};
- GridRedBlackCartesian(std::vector<int> &dimensions,
- std::vector<int> &simd_layout,
- std::vector<int> &processor_grid,
- std::vector<int> &checker_dim_mask,
+ GridRedBlackCartesian(const std::vector<int> &dimensions,
+ const std::vector<int> &simd_layout,
+ const std::vector<int> &processor_grid,
+ const std::vector<int> &checker_dim_mask,
int checker_dim
) : GridBase(processor_grid)
{
Init(dimensions,simd_layout,processor_grid,checker_dim_mask,checker_dim);
}
- GridRedBlackCartesian(std::vector<int> &dimensions,
- std::vector<int> &simd_layout,
- std::vector<int> &processor_grid) : GridBase(processor_grid)
+ GridRedBlackCartesian(const std::vector<int> &dimensions,
+ const std::vector<int> &simd_layout,
+ const std::vector<int> &processor_grid) : GridBase(processor_grid)
{
std::vector<int> checker_dim_mask(dimensions.size(),1);
Init(dimensions,simd_layout,processor_grid,checker_dim_mask,0);
}
- void Init(std::vector<int> &dimensions,
- std::vector<int> &simd_layout,
- std::vector<int> &processor_grid,
- std::vector<int> &checker_dim_mask,
+ void Init(const std::vector<int> &dimensions,
+ const std::vector<int> &simd_layout,
+ const std::vector<int> &processor_grid,
+ const std::vector<int> &checker_dim_mask,
int checker_dim)
{
///////////////////////
diff --git a/lib/communicator/Grid_communicator_base.h b/lib/communicator/Grid_communicator_base.h
index 47c1f525..61e19993 100644
--- a/lib/communicator/Grid_communicator_base.h
+++ b/lib/communicator/Grid_communicator_base.h
@@ -27,7 +27,7 @@ class CartesianCommunicator {
#endif
// Constructor
- CartesianCommunicator(std::vector<int> &pdimensions_in);
+ CartesianCommunicator(const std::vector<int> &pdimensions_in);
// Wraps MPI_Cart routines
void ShiftedRanks(int dim,int shift,int & source, int & dest);
diff --git a/lib/communicator/Grid_communicator_mpi.cc b/lib/communicator/Grid_communicator_mpi.cc
index 6ef05c3d..5dd34705 100644
--- a/lib/communicator/Grid_communicator_mpi.cc
+++ b/lib/communicator/Grid_communicator_mpi.cc
@@ -5,7 +5,7 @@ namespace Grid {
// Should error check all MPI calls.
-CartesianCommunicator::CartesianCommunicator(std::vector<int> &processors)
+CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors)
{
_ndimension = processors.size();
std::vector<int> periodic(_ndimension,1);
diff --git a/lib/communicator/Grid_communicator_none.cc b/lib/communicator/Grid_communicator_none.cc
index 90eb26cc..d7eb9453 100644
--- a/lib/communicator/Grid_communicator_none.cc
+++ b/lib/communicator/Grid_communicator_none.cc
@@ -1,7 +1,7 @@
#include "Grid.h"
namespace Grid {
-CartesianCommunicator::CartesianCommunicator(std::vector<int> &processors)
+CartesianCommunicator::CartesianCommunicator(const std::vector<int> &processors)
{
_processors = processors;
_ndimension = processors.size();
diff --git a/lib/qcd/LinalgUtils.h b/lib/qcd/LinalgUtils.h
new file mode 100644
index 00000000..2b83d115
--- /dev/null
+++ b/lib/qcd/LinalgUtils.h
@@ -0,0 +1,113 @@
+#ifndef GRID_QCD_LINALG_UTILS_H
+#define GRID_QCD_LINALG_UTILS_H
+
+namespace Grid{
+namespace QCD{
+////////////////////////////////////////////////////////////////////////
+//This file brings additional linear combination assist that is helpful
+//to QCD such as chiral projectors and spin matrices applied to one of the inputs.
+//These routines support five-D chiral fermions and contain s-subslice indexing
+//on the 5d (rb4d) checkerboarded lattices
+////////////////////////////////////////////////////////////////////////
+template<class vobj>
+void axpby_ssp(Lattice<vobj> &z, RealD a,const Lattice<vobj> &x,RealD b,const Lattice<vobj> &y,int s,int sp)
+{
+ z.checkerboard = x.checkerboard;
+ conformable(x,y);
+ conformable(x,z);
+ GridBase *grid=x._grid;
+ int Ls = grid->_rdimensions[0];
+PARALLEL_FOR_LOOP
+ for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
+ vobj tmp = a*x._odata[ss+s]+b*y._odata[ss+sp];
+ vstream(z._odata[ss+s],tmp);
+ }
+}
+
+template<class vobj>
+void ag5xpby_ssp(Lattice<vobj> &z,RealD a,const Lattice<vobj> &x,RealD b,const Lattice<vobj> &y,int s,int sp)
+{
+ z.checkerboard = x.checkerboard;
+ conformable(x,y);
+ conformable(x,z);
+ GridBase *grid=x._grid;
+ int Ls = grid->_rdimensions[0];
+PARALLEL_FOR_LOOP
+ for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
+ vobj tmp;
+ multGamma5(tmp(),a*x._odata[ss+s]());
+ tmp = tmp + b*y._odata[ss+sp];
+ vstream(z._odata[ss+s],tmp);
+ }
+}
+
+template<class vobj>
+void axpbg5y_ssp(Lattice<vobj> &z,RealD a,const Lattice<vobj> &x,RealD b,const Lattice<vobj> &y,int s,int sp)
+{
+ z.checkerboard = x.checkerboard;
+ conformable(x,y);
+ conformable(x,z);
+ GridBase *grid=x._grid;
+ int Ls = grid->_rdimensions[0];
+PARALLEL_FOR_LOOP
+ for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
+ vobj tmp;
+ multGamma5(tmp(),b*y._odata[ss+sp]());
+ tmp = tmp + a*x._odata[ss+s];
+ vstream(z._odata[ss+s],tmp);
+ }
+}
+
+template<class vobj>
+void ag5xpbg5y_ssp(Lattice<vobj> &z,RealD a,const Lattice<vobj> &x,RealD b,const Lattice<vobj> &y,int s,int sp)
+{
+ z.checkerboard = x.checkerboard;
+ conformable(x,y);
+ conformable(x,z);
+ GridBase *grid=x._grid;
+ int Ls = grid->_rdimensions[0];
+PARALLEL_FOR_LOOP
+ for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
+ vobj tmp1;
+ vobj tmp2;
+ tmp1 = a*x._odata[ss+s]+b*y._odata[ss+sp];
+ multGamma5(tmp2(),tmp1());
+ vstream(z._odata[ss+s],tmp2);
+ }
+}
+
+template<class vobj>
+void axpby_ssp_pminus(Lattice<vobj> &z,RealD a,const Lattice<vobj> &x,RealD b,const Lattice<vobj> &y,int s,int sp)
+{
+ z.checkerboard = x.checkerboard;
+ conformable(x,y);
+ conformable(x,z);
+ GridBase *grid=x._grid;
+ int Ls = grid->_rdimensions[0];
+PARALLEL_FOR_LOOP
+ for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
+ vobj tmp;
+ spProj5m(tmp,y._odata[ss+sp]);
+ tmp = a*x._odata[ss+s]+b*tmp;
+ vstream(z._odata[ss+s],tmp);
+ }
+}
+
+template<class vobj>
+void axpby_ssp_pplus(Lattice<vobj> &z,RealD a,const Lattice<vobj> &x,RealD b,const Lattice<vobj> &y,int s,int sp)
+{
+ z.checkerboard = x.checkerboard;
+ conformable(x,y);
+ conformable(x,z);
+ GridBase *grid=x._grid;
+ int Ls = grid->_rdimensions[0];
+PARALLEL_FOR_LOOP
+ for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
+ vobj tmp;
+ spProj5p(tmp,y._odata[ss+sp]);
+ tmp = a*x._odata[ss+s]+b*tmp;
+ vstream(z._odata[ss+s],tmp);
+ }
+}
+}}
+#endif
diff --git a/lib/qcd/QCD.h b/lib/qcd/QCD.h
index 7c45eb23..4b5edc5c 100644
--- a/lib/qcd/QCD.h
+++ b/lib/qcd/QCD.h
@@ -307,8 +307,10 @@ namespace QCD {
} //namespace QCD
} // Grid
+#include <qcd/SpaceTimeGrid.h>
#include <qcd/Dirac.h>
#include <qcd/TwoSpinor.h>
+#include <qcd/LinalgUtils.h>
#include <qcd/action/Actions.h>
#endif
diff --git a/lib/qcd/SpaceTimeGrid.cc b/lib/qcd/SpaceTimeGrid.cc
new file mode 100644
index 00000000..284c5771
--- /dev/null
+++ b/lib/qcd/SpaceTimeGrid.cc
@@ -0,0 +1,52 @@
+#include <Grid.h>
+
+namespace Grid {
+ namespace QCD {
+
+/////////////////////////////////////////////////////////////////
+// Public interface
+/////////////////////////////////////////////////////////////////
+GridCartesian *SpaceTimeGrid::makeFourDimGrid(const std::vector<int> & latt,const std::vector<int> &simd,const std::vector<int> &mpi)
+{
+ return new GridCartesian(latt,simd,mpi);
+}
+GridRedBlackCartesian *SpaceTimeGrid::makeFourDimRedBlackGrid(const GridCartesian *FourDimGrid)
+{
+ return new GridRedBlackCartesian(FourDimGrid);
+}
+
+GridCartesian *SpaceTimeGrid::makeFiveDimGrid(int Ls,const GridCartesian *FourDimGrid)
+{
+ int N4=FourDimGrid->_ndimension;
+
+ std::vector<int> latt5(1,Ls);
+ std::vector<int> simd5(1,1);
+ std::vector<int> mpi5(1,1);
+
+ for(int d=0;d<N4;d++){
+ latt5.push_back(FourDimGrid->_fdimensions[d]);
+ simd5.push_back(FourDimGrid->_simd_layout[d]);
+ mpi5.push_back(FourDimGrid->_processors[d]);
+ }
+ return new GridCartesian(latt5,simd5,mpi5);
+}
+
+GridRedBlackCartesian *SpaceTimeGrid::makeFiveDimRedBlackGrid(int Ls,const GridCartesian *FourDimGrid)
+{
+ int N4=FourDimGrid->_ndimension;
+ int cbd=1;
+ std::vector<int> latt5(1,Ls);
+ std::vector<int> simd5(1,1);
+ std::vector<int> mpi5(1,1);
+ std::vector<int> cb5(1,0);
+
+ for(int d=0;d<N4;d++){
+ latt5.push_back(FourDimGrid->_fdimensions[d]);
+ simd5.push_back(FourDimGrid->_simd_layout[d]);
+ mpi5.push_back(FourDimGrid->_processors[d]);
+ cb5.push_back( 1);
+ }
+ return new GridRedBlackCartesian(latt5,simd5,mpi5,cb5,cbd);
+}
+
+}}
diff --git a/lib/qcd/SpaceTimeGrid.h b/lib/qcd/SpaceTimeGrid.h
new file mode 100644
index 00000000..0b386a0e
--- /dev/null
+++ b/lib/qcd/SpaceTimeGrid.h
@@ -0,0 +1,18 @@
+#ifndef GRID_QCD_SPACE_TIME_GRID_H
+#define GRID_QCD_SPACE_TIME_GRID_H
+namespace Grid {
+namespace QCD {
+
+class SpaceTimeGrid {
+ public:
+
+ static GridCartesian *makeFourDimGrid(const std::vector<int> & latt,const std::vector<int> &simd,const std::vector<int> &mpi);
+ static GridRedBlackCartesian *makeFourDimRedBlackGrid (const GridCartesian *FourDimGrid);
+ static GridCartesian *makeFiveDimGrid (int Ls,const GridCartesian *FourDimGrid);
+ static GridRedBlackCartesian *makeFiveDimRedBlackGrid(int Ls,const GridCartesian *FourDimGrid);
+
+};
+
+}}
+
+#endif
diff --git a/lib/qcd/action/Actions.h b/lib/qcd/action/Actions.h
index c4e8a2f0..acbf027c 100644
--- a/lib/qcd/action/Actions.h
+++ b/lib/qcd/action/Actions.h
@@ -1,10 +1,80 @@
#ifndef GRID_QCD_ACTIONS_H
#define GRID_QCD_ACTIONS_H
-#include <qcd/action/fermion/FermionAction.h>
-#include <qcd/action/fermion/WilsonCompressor.h>
-#include <qcd/action/fermion/WilsonKernels.h>
+
+// Some reorganisation likely required as both Chroma and IroIro
+// are separating the concept of the operator from that of action.
+//
+// The FermAction contains methods to create
+//
+// * Linear operators (Hermitian and non-hermitian) .. my LinearOperator
+// * System solvers (Hermitian and non-hermitian) .. my OperatorFunction
+// * MultiShift System solvers (Hermitian and non-hermitian) .. my OperatorFunction
+
+
+////////////////////////////////////////////
+// Abstract base interface
+////////////////////////////////////////////
+#include <qcd/action/fermion/FermionOperator.h>
+
+////////////////////////////////////////////
+// Utility functions
+////////////////////////////////////////////
+#include <qcd/action/fermion/WilsonCompressor.h> //used by all wilson type fermions
+#include <qcd/action/fermion/WilsonKernels.h> //used by all wilson type fermions
+
+////////////////////////////////////////////
+// 4D formulations
+////////////////////////////////////////////
#include <qcd/action/fermion/WilsonFermion.h>
-#include <qcd/action/fermion/FiveDimWilsonFermion.h>
+//#include <qcd/action/fermion/CloverFermion.h>
+
+////////////////////////////////////////////
+// 5D formulations
+////////////////////////////////////////////
+#include <qcd/action/fermion/WilsonFermion5D.h> // used by all 5d overlap types
+#include <qcd/action/fermion/CayleyFermion5D.h>
+#include <qcd/action/fermion/ContinuedFractionFermion5D.h>
+//#include <qcd/action/fermion/PartialFraction.h>
+
+#include <qcd/action/fermion/DomainWallFermion.h>
+//#include <qcd/action/fermion/ScaledShamirCayleyTanh.h>
+
+
+ // Chroma interface defining FermionAction
+ /*
+ template<typename T, typename P, typename Q> class FermAct4D : public FermionAction<T,P,Q>
+ virtual LinearOperator<T>* linOp(Handle< FermState<T,P,Q> > state) const = 0;
+ virtual LinearOperator<T>* lMdagM(Handle< FermState<T,P,Q> > state) const = 0;
+ virtual LinOpSystemSolver<T>* invLinOp(Handle< FermState<T,P,Q> > state,
+ virtual MdagMSystemSolver<T>* invMdagM(Handle< FermState<T,P,Q> > state,
+ virtual LinOpMultiSystemSolver<T>* mInvLinOp(Handle< FermState<T,P,Q> > state,
+ virtual MdagMMultiSystemSolver<T>* mInvMdagM(Handle< FermState<T,P,Q> > state,
+ virtual MdagMMultiSystemSolverAccumulate<T>* mInvMdagMAcc(Handle< FermState<T,P,Q> > state,
+ virtual SystemSolver<T>* qprop(Handle< FermState<T,P,Q> > state,
+ class DiffFermAct4D : public FermAct4D<T,P,Q>
+ virtual DiffLinearOperator<T,Q,P>* linOp(Handle< FermState<T,P,Q> > state) const = 0;
+ virtual DiffLinearOperator<T,Q,P>* lMdagM(Handle< FermState<T,P,Q> > state) const = 0;
+ */
+
+
+ // Chroma interface defining GaugeAction
+ /*
+ template<typename P, typename Q> class GaugeAction
+ virtual const CreateGaugeState<P,Q>& getCreateState() const = 0;
+ virtual GaugeState<P,Q>* createState(const Q& q) const
+ virtual const GaugeBC<P,Q>& getGaugeBC() const
+ virtual const Set& getSet(void) const = 0;
+ virtual void deriv(P& result, const Handle< GaugeState<P,Q> >& state) const
+ virtual Double S(const Handle< GaugeState<P,Q> >& state) const = 0;
+
+ class LinearGaugeAction : public GaugeAction< multi1d<LatticeColorMatrix>, multi1d<LatticeColorMatrix> >
+ typedef multi1d<LatticeColorMatrix> P;
+ typedef multi1d<LatticeColorMatrix> Q;
+ virtual void staple(LatticeColorMatrix& result,
+ const Handle< GaugeState<P,Q> >& state,
+ int mu, int cb) const = 0;
+ */
+
#endif
diff --git a/lib/qcd/action/fermion/CayleyFermion5D.cc b/lib/qcd/action/fermion/CayleyFermion5D.cc
new file mode 100644
index 00000000..263cc28b
--- /dev/null
+++ b/lib/qcd/action/fermion/CayleyFermion5D.cc
@@ -0,0 +1,235 @@
+#include <Grid.h>
+namespace Grid {
+namespace QCD {
+
+ CayleyFermion5D::CayleyFermion5D(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD _M5) :
+ WilsonFermion5D(_Umu,
+ FiveDimGrid,
+ FiveDimRedBlackGrid,
+ FourDimGrid,
+ FourDimRedBlackGrid,_M5),
+ mass(_mass)
+ {
+ std::cout << "Constructing a CayleyFermion5D"<<std::endl;
+ }
+
+ // override multiply
+ RealD CayleyFermion5D::M (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ LatticeFermion Din(psi._grid);
+
+ // Assemble Din
+ for(int s=0;s<Ls;s++){
+ if ( s==0 ) {
+ // Din = bs psi[s] + cs[s] psi[s+1}
+ axpby_ssp_pminus(Din,bs[s],psi,cs[s],psi,s,s+1);
+ // Din+= -mass*cs[s] psi[s+1}
+ axpby_ssp_pplus (Din,1.0,Din,-mass*cs[s],psi,s,Ls-1);
+ } else if ( s==(Ls-1)) {
+ axpby_ssp_pminus(Din,bs[s],psi,-mass*cs[s],psi,s,0);
+ axpby_ssp_pplus (Din,1.0,Din,cs[s],psi,s,s-1);
+ } else {
+ axpby_ssp_pminus(Din,bs[s],psi,cs[s],psi,s,s+1);
+ axpby_ssp_pplus(Din,1.0,Din,cs[s],psi,s,s-1);
+ }
+ }
+
+ DW(Din,chi,DaggerNo);
+ // ((b D_W + D_w hop terms +1) on s-diag
+ axpby(chi,1.0,1.0,chi,psi);
+
+ for(int s=0;s<Ls;s++){
+ if ( s==0 ){
+ axpby_ssp_pminus(chi,1.0,chi,-1.0,psi,s,s+1);
+ axpby_ssp_pplus (chi,1.0,chi,mass,psi,s,Ls-1);
+ } else if ( s==(Ls-1)) {
+ axpby_ssp_pminus(chi,1.0,chi,mass,psi,s,0);
+ axpby_ssp_pplus (chi,1.0,chi,-1.0,psi,s,s-1);
+ } else {
+ axpby_ssp_pminus(chi,1.0,chi,-1.0,psi,s,s+1);
+ axpby_ssp_pplus (chi,1.0,chi,-1.0,psi,s,s-1);
+ }
+ }
+ return norm2(chi);
+ }
+
+ RealD CayleyFermion5D::Mdag (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ // Under adjoint
+ //D1+ D1- P- -> D1+^dag P+ D2-^dag
+ //D2- P+ D2+ P-D1-^dag D2+dag
+
+ LatticeFermion Din(psi._grid);
+ // Apply Dw
+ DW(psi,Din,DaggerYes);
+
+ for(int s=0;s<Ls;s++){
+ // Collect the terms in DW
+ // Chi = bs Din[s] + cs[s] Din[s+1}
+ // Chi+= -mass*cs[s] psi[s+1}
+ if ( s==0 ) {
+ axpby_ssp_pplus (chi,bs[s],Din,cs[s+1],Din,s,s+1);
+ axpby_ssp_pminus(chi,1.0,chi,-mass*cs[Ls-1],Din,s,Ls-1);
+ } else if ( s==(Ls-1)) {
+ axpby_ssp_pplus (chi,bs[s],Din,-mass*cs[0],Din,s,0);
+ axpby_ssp_pminus(chi,1.0,chi,cs[s-1],Din,s,s-1);
+ } else {
+ axpby_ssp_pplus (chi,bs[s],Din,cs[s+1],Din,s,s+1);
+ axpby_ssp_pminus(chi,1.0,chi,cs[s-1],Din,s,s-1);
+ }
+ // Collect the terms indept of DW
+ if ( s==0 ){
+ axpby_ssp_pplus (chi,1.0,chi,-1.0,psi,s,s+1);
+ axpby_ssp_pminus(chi,1.0,chi,mass,psi,s,Ls-1);
+ } else if ( s==(Ls-1)) {
+ axpby_ssp_pplus (chi,1.0,chi,mass,psi,s,0);
+ axpby_ssp_pminus(chi,1.0,chi,-1.0,psi,s,s-1);
+ } else {
+ axpby_ssp_pplus(chi,1.0,chi,-1.0,psi,s,s+1);
+ axpby_ssp_pminus(chi,1.0,chi,-1.0,psi,s,s-1);
+ }
+ }
+ // ((b D_W + D_w hop terms +1) on s-diag
+ axpby (chi,1.0,1.0,chi,psi);
+ return norm2(chi);
+ }
+
+ // half checkerboard operations
+ void CayleyFermion5D::Meooe (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ LatticeFermion tmp(psi._grid);
+ // Assemble the 5d matrix
+ for(int s=0;s<Ls;s++){
+ if ( s==0 ) {
+ // tmp = bs psi[s] + cs[s] psi[s+1}
+ // tmp+= -mass*cs[s] psi[s+1}
+ axpby_ssp_pminus(tmp,beo[s],psi,-ceo[s],psi ,s, s+1);
+ axpby_ssp_pplus(tmp,1.0,tmp,mass*ceo[s],psi,s,Ls-1);
+ } else if ( s==(Ls-1)) {
+ axpby_ssp_pminus(tmp,beo[s],psi,mass*ceo[s],psi,s,0);
+ axpby_ssp_pplus(tmp,1.0,tmp,-ceo[s],psi,s,s-1);
+ } else {
+ axpby_ssp_pminus(tmp,beo[s],psi,-ceo[s],psi,s,s+1);
+ axpby_ssp_pplus (tmp,1.0,tmp,-ceo[s],psi,s,s-1);
+ }
+ }
+ // Apply 4d dslash
+ if ( psi.checkerboard == Odd ) {
+ DhopEO(tmp,chi,DaggerNo);
+ } else {
+ DhopOE(tmp,chi,DaggerNo);
+ }
+ }
+
+ void CayleyFermion5D::MeooeDag (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ LatticeFermion tmp(psi._grid);
+ // Apply 4d dslash
+ if ( psi.checkerboard == Odd ) {
+ DhopEO(psi,tmp,DaggerYes);
+ } else {
+ DhopOE(psi,tmp,DaggerYes);
+ }
+ // Assemble the 5d matrix
+ for(int s=0;s<Ls;s++){
+ if ( s==0 ) {
+ axpby_ssp_pplus(chi,beo[s],tmp, -ceo[s+1] ,tmp,s,s+1);
+ axpby_ssp_pminus(chi, 1.0,chi,mass*ceo[Ls-1],tmp,s,Ls-1);
+ } else if ( s==(Ls-1)) {
+ axpby_ssp_pplus(chi,beo[s],tmp,mass*ceo[0],tmp,s,0);
+ axpby_ssp_pminus(chi,1.0,chi,-ceo[s-1],tmp,s,s-1);
+ } else {
+ axpby_ssp_pplus(chi,beo[s],tmp,-ceo[s+1],tmp,s,s+1);
+ axpby_ssp_pminus(chi,1.0 ,chi,-ceo[s-1],tmp,s,s-1);
+ }
+ }
+ }
+
+ void CayleyFermion5D::Mooee (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ for (int s=0;s<Ls;s++){
+ if ( s==0 ) {
+ axpby_ssp_pminus(chi,bee[s],psi ,-cee[s],psi,s,s+1);
+ axpby_ssp_pplus (chi,1.0,chi,mass*cee[s],psi,s,Ls-1);
+ } else if ( s==(Ls-1)) {
+ axpby_ssp_pminus(chi,bee[s],psi,mass*cee[s],psi,s,0);
+ axpby_ssp_pplus (chi,1.0,chi,-cee[s],psi,s,s-1);
+ } else {
+ axpby_ssp_pminus(chi,bee[s],psi,-cee[s],psi,s,s+1);
+ axpby_ssp_pplus (chi,1.0,chi,-cee[s],psi,s,s-1);
+ }
+ }
+ }
+
+ void CayleyFermion5D::MooeeDag (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ for (int s=0;s<Ls;s++){
+ // Assemble the 5d matrix
+ if ( s==0 ) {
+ axpby_ssp_pplus(chi,bee[s],psi,-cee[s+1] ,psi,s,s+1);
+ axpby_ssp_pminus(chi,1.0,chi,mass*cee[Ls-1],psi,s,Ls-1);
+ } else if ( s==(Ls-1)) {
+ axpby_ssp_pplus(chi,bee[s],psi,mass*cee[0],psi,s,0);
+ axpby_ssp_pminus(chi,1.0,chi,-cee[s-1],psi,s,s-1);
+ } else {
+ axpby_ssp_pplus(chi,bee[s],psi,-cee[s+1],psi,s,s+1);
+ axpby_ssp_pminus(chi,1.0 ,chi,-cee[s-1],psi,s,s-1);
+ }
+ }
+ }
+
+ void CayleyFermion5D::MooeeInv (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ // Apply (L^{\prime})^{-1}
+ axpby_ssp (chi,1.0,psi, 0.0,psi,0,0); // chi[0]=psi[0]
+ for (int s=1;s<Ls;s++){
+ axpby_ssp_pplus(chi,1.0,psi,-lee[s-1],chi,s,s-1);// recursion Psi[s] -lee P_+ chi[s-1]
+ }
+ // L_m^{-1}
+ for (int s=0;s<Ls-1;s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
+ axpby_ssp_pminus(chi,1.0,chi,-leem[s],chi,Ls-1,s);
+ }
+ // U_m^{-1} D^{-1}
+ for (int s=0;s<Ls-1;s++){
+ // Chi[s] + 1/d chi[s]
+ axpby_ssp_pplus(chi,1.0/dee[s],chi,-ueem[s]/dee[Ls-1],chi,s,Ls-1);
+ }
+ axpby_ssp(chi,1.0/dee[Ls-1],chi,0.0,chi,Ls-1,Ls-1); // Modest avoidable
+
+ // Apply U^{-1}
+ for (int s=Ls-2;s>=0;s--){
+ axpby_ssp_pminus (chi,1.0,chi,-uee[s],chi,s,s+1); // chi[Ls]
+ }
+ }
+
+ void CayleyFermion5D::MooeeInvDag (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ // Apply (U^{\prime})^{-dagger}
+ axpby_ssp (chi,1.0,psi, 0.0,psi,0,0); // chi[0]=psi[0]
+ for (int s=1;s<Ls;s++){
+ axpby_ssp_pminus(chi,1.0,psi,-uee[s-1],chi,s,s-1);
+ }
+ // U_m^{-\dagger}
+ for (int s=0;s<Ls-1;s++){
+ axpby_ssp_pplus(chi,1.0,chi,-ueem[s],chi,Ls-1,s);
+ }
+ // L_m^{-\dagger} D^{-dagger}
+ for (int s=0;s<Ls-1;s++){
+ axpby_ssp_pminus(chi,1.0/dee[s],chi,-leem[s]/dee[Ls-1],chi,s,Ls-1);
+ }
+ axpby_ssp(chi,1.0/dee[Ls-1],chi,0.0,chi,Ls-1,Ls-1); // Modest avoidable
+
+ // Apply L^{-dagger}
+ for (int s=Ls-2;s>=0;s--){
+ axpby_ssp_pplus (chi,1.0,chi,-lee[s],chi,s,s+1); // chi[Ls]
+ }
+ }
+
+}
+}
+
diff --git a/lib/qcd/action/fermion/CayleyFermion5D.h b/lib/qcd/action/fermion/CayleyFermion5D.h
new file mode 100644
index 00000000..857ac124
--- /dev/null
+++ b/lib/qcd/action/fermion/CayleyFermion5D.h
@@ -0,0 +1,61 @@
+#ifndef GRID_QCD_CAYLEY_FERMION_H
+#define GRID_QCD_CAYLEY_FERMION_H
+
+namespace Grid {
+
+ namespace QCD {
+
+ class CayleyFermion5D : public WilsonFermion5D
+ {
+ public:
+
+ // override multiply
+ virtual RealD M (const LatticeFermion &in, LatticeFermion &out);
+ virtual RealD Mdag (const LatticeFermion &in, LatticeFermion &out);
+
+ // half checkerboard operations
+ virtual void Meooe (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MeooeDag (const LatticeFermion &in, LatticeFermion &out);
+ virtual void Mooee (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeDag (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeInv (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeInvDag (const LatticeFermion &in, LatticeFermion &out);
+
+ // protected:
+
+ Approx::zolotarev_data *zdata;
+
+ RealD mass;
+ // Cayley form Moebius (tanh and zolotarev)
+ std::vector<RealD> omega;
+ std::vector<RealD> bs; // S dependent coeffs
+ std::vector<RealD> cs;
+ std::vector<RealD> as;
+ // For preconditioning Cayley form
+ std::vector<RealD> bee;
+ std::vector<RealD> cee;
+ std::vector<RealD> aee;
+ std::vector<RealD> beo;
+ std::vector<RealD> ceo;
+ std::vector<RealD> aeo;
+ // LDU factorisation of the eeoo matrix
+ std::vector<RealD> lee;
+ std::vector<RealD> leem;
+ std::vector<RealD> uee;
+ std::vector<RealD> ueem;
+ std::vector<RealD> dee;
+
+ // Constructors
+ CayleyFermion5D(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD _M5);
+
+ };
+
+ }
+}
+
+#endif
diff --git a/lib/qcd/action/fermion/ContinuedFractionFermion5D.cc b/lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
new file mode 100644
index 00000000..c281b486
--- /dev/null
+++ b/lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
@@ -0,0 +1,119 @@
+#include <Grid.h>
+
+namespace Grid {
+
+ namespace QCD {
+
+ RealD ContinuedFractionFermion5D::M (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ LatticeFermion D(psi._grid);
+
+ DW(psi,D,DaggerNo);
+
+ int sign=1;
+ for(int s=0;s<Ls;s++){
+ if ( s==0 ) {
+ ag5xpby_ssp(chi,cc[0]*Beta[0]*sign*scale,D,sqrt_cc[0],psi,s,s+1); // Multiplies Dw by G5 so Hw
+ } else if ( s==(Ls-1) ){
+ RealD R=(1.0+mass)/(1.0-mass);
+ ag5xpby_ssp(chi,Beta[s]*scale,D,sqrt_cc[s-1],psi,s,s-1);
+ ag5xpby_ssp(chi,R,psi,1.0,chi,s,s);
+ } else {
+ ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*scale,D,sqrt_cc[s],psi,s,s+1);
+ axpby_ssp(chi,1.0,chi,sqrt_cc[s-1],psi,s,s-1);
+ }
+ sign=-sign;
+ }
+ return norm2(chi);
+ }
+ RealD ContinuedFractionFermion5D::Mdag (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ // This matrix is already hermitian. (g5 Dw) = Dw dag g5 = (g5 Dw)dag
+ // The rest of matrix is symmetric.
+ // Can ignore "dag"
+ return M(psi,chi);
+ }
+ void ContinuedFractionFermion5D::Meooe (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ Dhop(psi,chi,DaggerNo); // Dslash on diagonal. g5 Dslash is hermitian
+
+ int sign=1;
+ for(int s=0;s<Ls;s++){
+ if ( s==(Ls-1) ){
+ ag5xpby_ssp(chi,Beta[s]*scale,chi,0.0,chi,s,s);
+ } else {
+ ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*scale,chi,0.0,chi,s,s);
+ }
+ sign=-sign;
+ }
+
+ }
+ void ContinuedFractionFermion5D::MeooeDag (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ Meooe(psi,chi);
+ }
+ void ContinuedFractionFermion5D::Mooee (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ double dw_diag = (4.0-this->M5)*scale;
+
+ int sign=1;
+ for(int s=0;s<Ls;s++){
+ if ( s==0 ) {
+ ag5xpby_ssp(chi,cc[0]*Beta[0]*sign*dw_diag,psi,sqrt_cc[0],psi,s,s+1); // Multiplies Dw by G5 so Hw
+ } else if ( s==(Ls-1) ){
+ // Drop the CC here.
+ double R=(1+this->mass)/(1-this->mass);
+ ag5xpby_ssp(chi,Beta[s]*dw_diag,psi,sqrt_cc[s-1],psi,s,s-1);
+ ag5xpby_ssp(chi,R,psi,1.0,chi,s,s);
+ } else {
+ ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*dw_diag,psi,sqrt_cc[s],psi,s,s+1);
+ axpby_ssp(chi,1.0,chi,sqrt_cc[s-1],psi,s,s-1);
+ }
+ sign=-sign;
+ }
+ }
+
+ void ContinuedFractionFermion5D::MooeeDag (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ Mooee(psi,chi);
+ }
+ void ContinuedFractionFermion5D::MooeeInv (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ // Apply Linv
+ axpby_ssp(chi,1.0/cc_d[0],psi,0.0,psi,0,0);
+ for(int s=1;s<Ls;s++){
+ axpbg5y_ssp(chi,1.0/cc_d[s],psi,-1.0/See[s-1],chi,s,s-1);
+ }
+ // Apply Dinv
+ for(int s=0;s<Ls;s++){
+ ag5xpby_ssp(chi,1.0/See[s],chi,0.0,chi,s,s); //only appearance of See[0]
+ }
+ // Apply Uinv = (Linv)^T
+ axpby_ssp(chi,1.0/cc_d[Ls-1],chi,0.0,chi,this->Ls-1,this->Ls-1);
+ for(int s=Ls-2;s>=0;s--){
+ axpbg5y_ssp(chi,1.0/cc_d[s],chi,-1.0*cc_d[s+1]/See[s]/cc_d[s],chi,s,s+1);
+ }
+ }
+ void ContinuedFractionFermion5D::MooeeInvDag (const LatticeFermion &psi, LatticeFermion &chi)
+ {
+ MooeeInv(psi,chi);
+ }
+
+ // Constructors
+ ContinuedFractionFermion5D::ContinuedFractionFermion5D(
+ LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD M5) :
+ WilsonFermion5D(_Umu,
+ FiveDimGrid, FiveDimRedBlackGrid,
+ FourDimGrid, FourDimRedBlackGrid,M5),
+ mass(_mass)
+ {
+ }
+
+ }
+}
+
diff --git a/lib/qcd/action/fermion/ContinuedFractionFermion5D.h b/lib/qcd/action/fermion/ContinuedFractionFermion5D.h
new file mode 100644
index 00000000..7f5c022a
--- /dev/null
+++ b/lib/qcd/action/fermion/ContinuedFractionFermion5D.h
@@ -0,0 +1,53 @@
+#ifndef GRID_QCD_CONTINUED_FRACTION_H
+#define GRID_QCD_CONTINUED_FRACTION_H
+
+namespace Grid {
+
+ namespace QCD {
+
+ class ContinuedFractionFermion5D : public WilsonFermion5D
+ {
+ public:
+
+ // override multiply
+ virtual RealD M (const LatticeFermion &in, LatticeFermion &out);
+ virtual RealD Mdag (const LatticeFermion &in, LatticeFermion &out);
+
+ // half checkerboard operaions
+ virtual void Meooe (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MeooeDag (const LatticeFermion &in, LatticeFermion &out);
+ virtual void Mooee (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeDag (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeInv (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeInvDag (const LatticeFermion &in, LatticeFermion &out);
+
+ private:
+
+ Approx::zolotarev_data *zdata;
+
+ // Cont frac
+ RealD mass;
+ RealD R;
+ RealD scale;
+ std::vector<double> Beta;
+ std::vector<double> cc;;
+ std::vector<double> cc_d;;
+ std::vector<double> sqrt_cc;
+ std::vector<double> See;
+ std::vector<double> Aee;
+
+ // Constructors
+ ContinuedFractionFermion5D(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD M5);
+
+ };
+
+
+ }
+}
+
+#endif
diff --git a/lib/qcd/action/fermion/DomainWallFermion.h b/lib/qcd/action/fermion/DomainWallFermion.h
new file mode 100644
index 00000000..2abb6eb2
--- /dev/null
+++ b/lib/qcd/action/fermion/DomainWallFermion.h
@@ -0,0 +1,118 @@
+#ifndef GRID_QCD_DOMAIN_WALL_FERMION_H
+#define GRID_QCD_DOMAIN_WALL_FERMION_H
+
+#include <Grid.h>
+
+namespace Grid {
+
+ namespace QCD {
+
+ class DomainWallFermion : public CayleyFermion5D
+ {
+ public:
+
+ // Constructors
+ DomainWallFermion(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD _M5) :
+
+ CayleyFermion5D(_Umu,
+ FiveDimGrid,
+ FiveDimRedBlackGrid,
+ FourDimGrid,
+ FourDimRedBlackGrid,_mass,_M5)
+
+ {
+ RealD eps = 1.0;
+
+ zdata = Approx::grid_higham(eps,this->Ls);// eps is ignored for higham
+ assert(zdata->n==this->Ls);
+
+ ///////////////////////////////////////////////////////////
+ // The Cayley coeffs (unprec)
+ ///////////////////////////////////////////////////////////
+ this->omega.resize(this->Ls);
+ this->bs.resize(this->Ls);
+ this->cs.resize(this->Ls);
+ this->as.resize(this->Ls);
+
+ for(int i=0; i < this->Ls; i++){
+ this->as[i] = 1.0;
+ this->omega[i] = ((double)zdata -> gamma[i]);
+ double bb=1.0;
+ this->bs[i] = 0.5*(bb/(this->omega[i]) + 1.0);
+ this->cs[i] = 0.5*(bb/(this->omega[i]) - 1.0);
+ }
+
+ ////////////////////////////////////////////////////////
+ // Constants for the preconditioned matrix Cayley form
+ ////////////////////////////////////////////////////////
+ this->bee.resize(this->Ls);
+ this->cee.resize(this->Ls);
+ this->beo.resize(this->Ls);
+ this->ceo.resize(this->Ls);
+
+ for(int i=0;i<this->Ls;i++){
+ this->bee[i]=as[i]*(bs[i]*(4.0-M5) +1.0);
+ this->cee[i]=as[i]*(1.0-cs[i]*(4.0-M5));
+ this->beo[i]=as[i]*bs[i];
+ this->ceo[i]=-as[i]*cs[i];
+ }
+
+ aee.resize(this->Ls);
+ aeo.resize(this->Ls);
+ for(int i=0;i<this->Ls;i++){
+ aee[i]=cee[i];
+ aeo[i]=ceo[i];
+ }
+
+ //////////////////////////////////////////
+ // LDU decomposition of eeoo
+ //////////////////////////////////////////
+ dee.resize(this->Ls);
+ lee.resize(this->Ls);
+ leem.resize(this->Ls);
+ uee.resize(this->Ls);
+ ueem.resize(this->Ls);
+
+ for(int i=0;i<this->Ls;i++){
+
+ dee[i] = bee[i];
+
+ if ( i < this->Ls-1 ) {
+
+ lee[i] =-cee[i+1]/bee[i]; // sub-diag entry on the ith column
+
+ leem[i]=this->mass*cee[this->Ls-1]/bee[0];
+ for(int j=0;j<i;j++) leem[i]*= aee[j]/bee[j+1];
+
+ uee[i] =-aee[i]/bee[i]; // up-diag entry on the ith row
+
+ ueem[i]=this->mass;
+ for(int j=1;j<=i;j++) ueem[i]*= cee[j]/bee[j];
+ ueem[i]*= aee[0]/bee[0];
+
+ } else {
+ lee[i] =0.0;
+ leem[i]=0.0;
+ uee[i] =0.0;
+ ueem[i]=0.0;
+ }
+ }
+
+ {
+ double delta_d=mass*cee[this->Ls-1];
+ for(int j=0;j<this->Ls-1;j++) delta_d *= cee[j]/bee[j];
+ dee[this->Ls-1] += delta_d;
+ }
+ }
+
+ };
+
+ }
+}
+
+#endif
diff --git a/lib/qcd/action/fermion/FermionAction.h b/lib/qcd/action/fermion/FermionOperator.h
similarity index 92%
rename from lib/qcd/action/fermion/FermionAction.h
rename to lib/qcd/action/fermion/FermionOperator.h
index 1b05174b..47c47478 100644
--- a/lib/qcd/action/fermion/FermionAction.h
+++ b/lib/qcd/action/fermion/FermionOperator.h
@@ -1,5 +1,5 @@
-#ifndef GRID_QCD_WILSON_DOP_H
-#define GRID_QCD_WILSON_DOP_H
+#ifndef GRID_QCD_FERMION_OPERATOR_H
+#define GRID_QCD_FERMION_OPERATOR_H
namespace Grid {
@@ -11,7 +11,7 @@ namespace Grid {
// Think about multiple representations
//////////////////////////////////////////////////////////////////////////////
template<class FermionField,class GaugeField>
- class FermionAction : public CheckerBoardedSparseMatrixBase<FermionField>
+ class FermionOperator : public CheckerBoardedSparseMatrixBase<FermionField>
{
public:
@@ -40,6 +40,7 @@ namespace Grid {
virtual void DhopOE(const FermionField &in, FermionField &out,int dag)=0;
virtual void DhopEO(const FermionField &in, FermionField &out,int dag)=0;
+
};
}
diff --git a/lib/qcd/action/fermion/PartialFractionFermion5D.cc b/lib/qcd/action/fermion/PartialFractionFermion5D.cc
new file mode 100644
index 00000000..21f62a75
--- /dev/null
+++ b/lib/qcd/action/fermion/PartialFractionFermion5D.cc
@@ -0,0 +1,47 @@
+#ifndef GRID_QCD_PARTIAL_FRACTION_H
+#define GRID_QCD_PARTIAL_FRACTION_H
+
+namespace Grid {
+
+ namespace QCD {
+
+ class PartialFractionFermion5D : public WilsonFermion5D
+ {
+ public:
+
+ // override multiply
+ virtual RealD M (const LatticeFermion &in, LatticeFermion &out);
+ virtual RealD Mdag (const LatticeFermion &in, LatticeFermion &out);
+
+ // half checkerboard operaions
+ virtual void Meooe (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MeooeDag (const LatticeFermion &in, LatticeFermion &out);
+ virtual void Mooee (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeDag (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeInv (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeInvDag (const LatticeFermion &in, LatticeFermion &out);
+
+ private:
+
+ zolotarev_data *zdata;
+
+ // Part frac
+ double R=(1+this->mass)/(1-this->mass);
+ std::vector<double> p;
+ std::vector<double> q;
+
+ // Constructors
+ PartialFractionFermion5D(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD M5);
+
+ };
+
+
+ }
+}
+
+#endif
diff --git a/lib/qcd/action/fermion/PartialFractionFermion5D.h b/lib/qcd/action/fermion/PartialFractionFermion5D.h
new file mode 100644
index 00000000..c51aa563
--- /dev/null
+++ b/lib/qcd/action/fermion/PartialFractionFermion5D.h
@@ -0,0 +1,49 @@
+#ifndef GRID_QCD_PARTIAL_FRACTION_H
+#define GRID_QCD_PARTIAL_FRACTION_H
+
+namespace Grid {
+
+ namespace QCD {
+
+ class PartialFractionFermion5D : public WilsonFermion5D
+ {
+ public:
+
+ // override multiply
+ virtual RealD M (const LatticeFermion &in, LatticeFermion &out);
+ virtual RealD Mdag (const LatticeFermion &in, LatticeFermion &out);
+
+ // half checkerboard operaions
+ virtual void Meooe (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MeooeDag (const LatticeFermion &in, LatticeFermion &out);
+ virtual void Mooee (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeDag (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeInv (const LatticeFermion &in, LatticeFermion &out);
+ virtual void MooeeInvDag (const LatticeFermion &in, LatticeFermion &out);
+
+ private:
+
+ virtual void PartialFractionCoefficients(void);
+
+ zolotarev_data *zdata;
+
+ // Part frac
+ double R=(1+this->mass)/(1-this->mass);
+ std::vector<double> p;
+ std::vector<double> q;
+
+ // Constructors
+ PartialFractionFermion5D(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD M5);
+
+ };
+
+
+ }
+}
+
+#endif
diff --git a/lib/qcd/action/fermion/WilsonFermion.cc b/lib/qcd/action/fermion/WilsonFermion.cc
index aa30a7fa..9f2da251 100644
--- a/lib/qcd/action/fermion/WilsonFermion.cc
+++ b/lib/qcd/action/fermion/WilsonFermion.cc
@@ -9,9 +9,9 @@ const std::vector<int> WilsonFermion::displacements({1,1,1,1,-1,-1,-1,-1});
int WilsonFermion::HandOptDslash;
WilsonFermion::WilsonFermion(LatticeGaugeField &_Umu,
- GridCartesian &Fgrid,
- GridRedBlackCartesian &Hgrid,
- double _mass) :
+ GridCartesian &Fgrid,
+ GridRedBlackCartesian &Hgrid,
+ RealD _mass) :
_grid(&Fgrid),
_cbgrid(&Hgrid),
Stencil (&Fgrid,npoint,Even,directions,displacements),
diff --git a/lib/qcd/action/fermion/WilsonFermion.h b/lib/qcd/action/fermion/WilsonFermion.h
index 5c208131..6040d328 100644
--- a/lib/qcd/action/fermion/WilsonFermion.h
+++ b/lib/qcd/action/fermion/WilsonFermion.h
@@ -5,7 +5,7 @@ namespace Grid {
namespace QCD {
- class WilsonFermion : public FermionAction<LatticeFermion,LatticeGaugeField>
+ class WilsonFermion : public FermionOperator<LatticeFermion,LatticeGaugeField>
{
public:
@@ -44,7 +44,7 @@ namespace Grid {
int dag);
// Constructor
- WilsonFermion(LatticeGaugeField &_Umu,GridCartesian &Fgrid,GridRedBlackCartesian &Hgrid,double _mass);
+ WilsonFermion(LatticeGaugeField &_Umu,GridCartesian &Fgrid,GridRedBlackCartesian &Hgrid,RealD _mass);
// DoubleStore
void DoubleStore(LatticeDoubledGaugeField &Uds,const LatticeGaugeField &Umu);
@@ -57,7 +57,7 @@ namespace Grid {
protected:
- double mass;
+ RealD mass;
GridBase * _grid;
GridBase * _cbgrid;
diff --git a/lib/qcd/action/fermion/FiveDimWilsonFermion.cc b/lib/qcd/action/fermion/WilsonFermion5D.cc
similarity index 68%
rename from lib/qcd/action/fermion/FiveDimWilsonFermion.cc
rename to lib/qcd/action/fermion/WilsonFermion5D.cc
index 43645899..d22701b0 100644
--- a/lib/qcd/action/fermion/FiveDimWilsonFermion.cc
+++ b/lib/qcd/action/fermion/WilsonFermion5D.cc
@@ -4,18 +4,18 @@ namespace Grid {
namespace QCD {
// S-direction is INNERMOST and takes no part in the parity.
- const std::vector<int> FiveDimWilsonFermion::directions ({1,2,3,4, 1, 2, 3, 4});
- const std::vector<int> FiveDimWilsonFermion::displacements({1,1,1,1,-1,-1,-1,-1});
+ const std::vector<int> WilsonFermion5D::directions ({1,2,3,4, 1, 2, 3, 4});
+ const std::vector<int> WilsonFermion5D::displacements({1,1,1,1,-1,-1,-1,-1});
- int FiveDimWilsonFermion::HandOptDslash;
+ int WilsonFermion5D::HandOptDslash;
// 5d lattice for DWF.
- FiveDimWilsonFermion::FiveDimWilsonFermion(LatticeGaugeField &_Umu,
+ WilsonFermion5D::WilsonFermion5D(LatticeGaugeField &_Umu,
GridCartesian &FiveDimGrid,
GridRedBlackCartesian &FiveDimRedBlackGrid,
GridCartesian &FourDimGrid,
GridRedBlackCartesian &FourDimRedBlackGrid,
- double _mass) :
+ RealD _M5) :
_FiveDimGrid(&FiveDimGrid),
_FiveDimRedBlackGrid(&FiveDimRedBlackGrid),
_FourDimGrid(&FourDimGrid),
@@ -23,7 +23,7 @@ namespace QCD {
Stencil (_FiveDimGrid,npoint,Even,directions,displacements),
StencilEven(_FiveDimRedBlackGrid,npoint,Even,directions,displacements), // source is Even
StencilOdd (_FiveDimRedBlackGrid,npoint,Odd ,directions,displacements), // source is Odd
- mass(_mass),
+ M5(_M5),
Umu(_FourDimGrid),
UmuEven(_FourDimRedBlackGrid),
UmuOdd (_FourDimRedBlackGrid),
@@ -70,7 +70,7 @@ namespace QCD {
pickCheckerboard(Even,UmuEven,Umu);
pickCheckerboard(Odd ,UmuOdd,Umu);
}
-void FiveDimWilsonFermion::DoubleStore(LatticeDoubledGaugeField &Uds,const LatticeGaugeField &Umu)
+void WilsonFermion5D::DoubleStore(LatticeDoubledGaugeField &Uds,const LatticeGaugeField &Umu)
{
conformable(Uds._grid,GaugeGrid());
conformable(Umu._grid,GaugeGrid());
@@ -82,60 +82,9 @@ void FiveDimWilsonFermion::DoubleStore(LatticeDoubledGaugeField &Uds,const Latti
pokeIndex<LorentzIndex>(Uds,U,mu+4);
}
}
-
-RealD FiveDimWilsonFermion::M(const LatticeFermion &in, LatticeFermion &out)
-{
- out.checkerboard=in.checkerboard;
- Dhop(in,out,DaggerNo);
- return axpy_norm(out,5.0-M5,in,out);
-}
-RealD FiveDimWilsonFermion::Mdag(const LatticeFermion &in, LatticeFermion &out)
-{
- out.checkerboard=in.checkerboard;
- Dhop(in,out,DaggerYes);
- return axpy_norm(out,5.0-M5,in,out);
-}
-void FiveDimWilsonFermion::Meooe(const LatticeFermion &in, LatticeFermion &out)
-{
- if ( in.checkerboard == Odd ) {
- DhopEO(in,out,DaggerNo);
- } else {
- DhopOE(in,out,DaggerNo);
- }
-}
-void FiveDimWilsonFermion::MeooeDag(const LatticeFermion &in, LatticeFermion &out)
-{
- if ( in.checkerboard == Odd ) {
- DhopEO(in,out,DaggerYes);
- } else {
- DhopOE(in,out,DaggerYes);
- }
-}
-void FiveDimWilsonFermion::Mooee(const LatticeFermion &in, LatticeFermion &out)
-{
- out.checkerboard = in.checkerboard;
- out = (5.0-M5)*in;
- return ;
-}
-void FiveDimWilsonFermion::MooeeDag(const LatticeFermion &in, LatticeFermion &out)
-{
- out.checkerboard = in.checkerboard;
- Mooee(in,out);
-}
-void FiveDimWilsonFermion::MooeeInv(const LatticeFermion &in, LatticeFermion &out)
-{
- out.checkerboard = in.checkerboard;
- out = (1.0/(5.0-M5))*in;
- return ;
-}
-void FiveDimWilsonFermion::MooeeInvDag(const LatticeFermion &in, LatticeFermion &out)
-{
- out.checkerboard = in.checkerboard;
- MooeeInv(in,out);
-}
-void FiveDimWilsonFermion::DhopInternal(CartesianStencil & st, LebesgueOrder &lo,
- LatticeDoubledGaugeField & U,
- const LatticeFermion &in, LatticeFermion &out,int dag)
+void WilsonFermion5D::DhopInternal(CartesianStencil & st, LebesgueOrder &lo,
+ LatticeDoubledGaugeField & U,
+ const LatticeFermion &in, LatticeFermion &out,int dag)
{
assert((dag==DaggerNo) ||(dag==DaggerYes));
@@ -150,19 +99,21 @@ void FiveDimWilsonFermion::DhopInternal(CartesianStencil & st, LebesgueOrder &lo
// - 8 linear access unit stride streams per thread for Fermion for hw prefetchable.
if ( dag == DaggerYes ) {
if( HandOptDslash ) {
- for(int ss=0;ss<U._grid->oSites();ss++){
- int sU=lo.Reorder(ss);
PARALLEL_FOR_LOOP
+ for(int ss=0;ss<U._grid->oSites();ss++){
for(int s=0;s<Ls;s++){
+ //int sU=lo.Reorder(ss);
+ int sU=ss;
int sF = s+Ls*sU;
DiracOptHand::DhopSiteDag(st,U,comm_buf,sF,sU,in,out);
}
}
} else {
- for(int ss=0;ss<U._grid->oSites();ss++){
- int sU=lo.Reorder(ss);
PARALLEL_FOR_LOOP
+ for(int ss=0;ss<U._grid->oSites();ss++){
for(int s=0;s<Ls;s++){
+ // int sU=lo.Reorder(ss);
+ int sU=ss;
int sF = s+Ls*sU;
DiracOpt::DhopSiteDag(st,U,comm_buf,sF,sU,in,out);
}
@@ -170,21 +121,22 @@ PARALLEL_FOR_LOOP
}
} else {
if( HandOptDslash ) {
-
PARALLEL_FOR_LOOP
for(int ss=0;ss<U._grid->oSites();ss++){
- int sU=lo.Reorder(ss);
for(int s=0;s<Ls;s++){
+ // int sU=lo.Reorder(ss);
+ int sU=ss;
int sF = s+Ls*sU;
DiracOptHand::DhopSite(st,U,comm_buf,sF,sU,in,out);
}
}
} else {
- for(int ss=0;ss<U._grid->oSites();ss++){
- int sU=lo.Reorder(ss);
PARALLEL_FOR_LOOP
+ for(int ss=0;ss<U._grid->oSites();ss++){
for(int s=0;s<Ls;s++){
+ // int sU=lo.Reorder(ss);
+ int sU=ss;
int sF = s+Ls*sU;
DiracOpt::DhopSite(st,U,comm_buf,sF,sU,in,out);
}
@@ -192,7 +144,7 @@ PARALLEL_FOR_LOOP
}
}
}
-void FiveDimWilsonFermion::DhopOE(const LatticeFermion &in, LatticeFermion &out,int dag)
+void WilsonFermion5D::DhopOE(const LatticeFermion &in, LatticeFermion &out,int dag)
{
conformable(in._grid,FermionRedBlackGrid()); // verifies half grid
conformable(in._grid,out._grid); // drops the cb check
@@ -202,7 +154,7 @@ void FiveDimWilsonFermion::DhopOE(const LatticeFermion &in, LatticeFermion &out,
DhopInternal(StencilEven,LebesgueEvenOdd,UmuOdd,in,out,dag);
}
-void FiveDimWilsonFermion::DhopEO(const LatticeFermion &in, LatticeFermion &out,int dag)
+void WilsonFermion5D::DhopEO(const LatticeFermion &in, LatticeFermion &out,int dag)
{
conformable(in._grid,FermionRedBlackGrid()); // verifies half grid
conformable(in._grid,out._grid); // drops the cb check
@@ -212,7 +164,7 @@ void FiveDimWilsonFermion::DhopEO(const LatticeFermion &in, LatticeFermion &out,
DhopInternal(StencilOdd,LebesgueEvenOdd,UmuEven,in,out,dag);
}
-void FiveDimWilsonFermion::Dhop(const LatticeFermion &in, LatticeFermion &out,int dag)
+void WilsonFermion5D::Dhop(const LatticeFermion &in, LatticeFermion &out,int dag)
{
conformable(in._grid,FermionGrid()); // verifies full grid
conformable(in._grid,out._grid);
@@ -221,8 +173,14 @@ void FiveDimWilsonFermion::Dhop(const LatticeFermion &in, LatticeFermion &out,in
DhopInternal(Stencil,Lebesgue,Umu,in,out,dag);
}
-
-}}
+void WilsonFermion5D::DW(const LatticeFermion &in, LatticeFermion &out,int dag)
+{
+ out.checkerboard=in.checkerboard;
+ Dhop(in,out,dag); // -0.5 is included
+ axpy(out,4.0-M5,in,out);
+}
+}
+}
diff --git a/lib/qcd/action/fermion/FiveDimWilsonFermion.h b/lib/qcd/action/fermion/WilsonFermion5D.h
similarity index 72%
rename from lib/qcd/action/fermion/FiveDimWilsonFermion.h
rename to lib/qcd/action/fermion/WilsonFermion5D.h
index 3aa85023..d4777d01 100644
--- a/lib/qcd/action/fermion/FiveDimWilsonFermion.h
+++ b/lib/qcd/action/fermion/WilsonFermion5D.h
@@ -15,7 +15,7 @@ namespace Grid {
//
// [DIFFERS from original CPS red black implementation parity = (x+y+z+t+s)|2 ]
////////////////////////////////////////////////////////////////////////////////
- class FiveDimWilsonFermion : public FermionAction<LatticeFermion,LatticeGaugeField>
+ class WilsonFermion5D : public FermionOperator<LatticeFermion,LatticeGaugeField>
{
public:
///////////////////////////////////////////////////////////////
@@ -26,19 +26,21 @@ namespace Grid {
GridBase *FermionGrid(void) { return _FiveDimGrid;}
GridBase *FermionRedBlackGrid(void) { return _FiveDimRedBlackGrid;}
- // override multiply
- virtual RealD M (const LatticeFermion &in, LatticeFermion &out);
- virtual RealD Mdag (const LatticeFermion &in, LatticeFermion &out);
+ // full checkerboard operations; leave unimplemented as abstract for now
+ //virtual RealD M (const LatticeFermion &in, LatticeFermion &out)=0;
+ //virtual RealD Mdag (const LatticeFermion &in, LatticeFermion &out)=0;
- // half checkerboard operaions
- virtual void Meooe (const LatticeFermion &in, LatticeFermion &out);
- virtual void MeooeDag (const LatticeFermion &in, LatticeFermion &out);
- virtual void Mooee (const LatticeFermion &in, LatticeFermion &out);
- virtual void MooeeDag (const LatticeFermion &in, LatticeFermion &out);
- virtual void MooeeInv (const LatticeFermion &in, LatticeFermion &out);
- virtual void MooeeInvDag (const LatticeFermion &in, LatticeFermion &out);
+ // half checkerboard operations; leave unimplemented as abstract for now
+ // virtual void Meooe (const LatticeFermion &in, LatticeFermion &out)=0;
+ // virtual void MeooeDag (const LatticeFermion &in, LatticeFermion &out)=0;
+ // virtual void Mooee (const LatticeFermion &in, LatticeFermion &out)=0;
+ // virtual void MooeeDag (const LatticeFermion &in, LatticeFermion &out)=0;
+ // virtual void MooeeInv (const LatticeFermion &in, LatticeFermion &out)=0;
+ // virtual void MooeeInvDag (const LatticeFermion &in, LatticeFermion &out)=0;
- // non-hermitian hopping term; half cb or both
+ // Implement hopping term non-hermitian hopping term; half cb or both
+ // Implement s-diagonal DW
+ void DW (const LatticeFermion &in, LatticeFermion &out,int dag);
void Dhop (const LatticeFermion &in, LatticeFermion &out,int dag);
void DhopOE(const LatticeFermion &in, LatticeFermion &out,int dag);
void DhopEO(const LatticeFermion &in, LatticeFermion &out,int dag);
@@ -54,12 +56,12 @@ namespace Grid {
int dag);
// Constructors
- FiveDimWilsonFermion(LatticeGaugeField &_Umu,
+ WilsonFermion5D(LatticeGaugeField &_Umu,
GridCartesian &FiveDimGrid,
GridRedBlackCartesian &FiveDimRedBlackGrid,
GridCartesian &FourDimGrid,
GridRedBlackCartesian &FourDimRedBlackGrid,
- double _mass);
+ double _M5);
// DoubleStore
void DoubleStore(LatticeDoubledGaugeField &Uds,const LatticeGaugeField &Umu);
@@ -82,7 +84,6 @@ namespace Grid {
static const std::vector<int> displacements;
double M5;
- double mass;
int Ls;
//Defines the stencils for even and odd
diff --git a/lib/stencil/Grid_stencil_common.cc b/lib/stencil/Grid_stencil_common.cc
index f0f8c581..7f894faf 100644
--- a/lib/stencil/Grid_stencil_common.cc
+++ b/lib/stencil/Grid_stencil_common.cc
@@ -52,8 +52,8 @@ namespace Grid {
// up a table containing the npoint "neighbours" and whether they
// live in lattice or a comms buffer.
if ( !comm_dim ) {
- sshift[0] = _grid->CheckerBoardShift(_checkerboard,dimension,shift,Even);
- sshift[1] = _grid->CheckerBoardShift(_checkerboard,dimension,shift,Odd);
+ sshift[0] = _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,Even);
+ sshift[1] = _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,Odd);
if ( sshift[0] == sshift[1] ) {
Local(point,dimension,shift,0x3);
@@ -63,8 +63,8 @@ namespace Grid {
}
} else { // All permute extract done in comms phase prior to Stencil application
// So tables are the same whether comm_dim or splice_dim
- sshift[0] = _grid->CheckerBoardShift(_checkerboard,dimension,shift,Even);
- sshift[1] = _grid->CheckerBoardShift(_checkerboard,dimension,shift,Odd);
+ sshift[0] = _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,Even);
+ sshift[1] = _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,Odd);
if ( sshift[0] == sshift[1] ) {
Comms(point,dimension,shift,0x3);
} else {
@@ -96,7 +96,7 @@ namespace Grid {
int cb= (cbmask==0x2)? Odd : Even;
- int sshift = _grid->CheckerBoardShift(_checkerboard,dimension,shift,cb);
+ int sshift = _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,cb);
int sx = (x+sshift)%rd;
int permute_slice=0;
@@ -134,7 +134,7 @@ namespace Grid {
// send to one or more remote nodes.
int cb= (cbmask==0x2)? Odd : Even;
- int sshift= _grid->CheckerBoardShift(_checkerboard,dimension,shift,cb);
+ int sshift= _grid->CheckerBoardShiftForCB(_checkerboard,dimension,shift,cb);
for(int x=0;x<rd;x++){
From 35fdba81dd5bf8546b9d93dcba869a7ba1ed915a Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Tue, 2 Jun 2015 17:25:26 +0100
Subject: [PATCH 09/20] Reorg; moving prec/unprec/schur CG for Wilson and DWF
into tests as these are really tests and not benchmarks (no performance
reports, only convergence test).
---
benchmarks/Makefile.am | 39 ++-------------
{benchmarks => tests}/Grid_dwf_cg_prec.cc | 0
{benchmarks => tests}/Grid_dwf_cg_schur.cc | 0
{benchmarks => tests}/Grid_dwf_cg_unprec.cc | 0
{benchmarks => tests}/Grid_dwf_even_odd.cc | 0
{benchmarks => tests}/Grid_wilson_cg_prec.cc | 0
{benchmarks => tests}/Grid_wilson_cg_schur.cc | 0
.../Grid_wilson_cg_unprec.cc | 0
{benchmarks => tests}/Grid_wilson_evenodd.cc | 0
tests/Makefile.am | 50 ++++++++++++++++++-
10 files changed, 52 insertions(+), 37 deletions(-)
rename {benchmarks => tests}/Grid_dwf_cg_prec.cc (100%)
rename {benchmarks => tests}/Grid_dwf_cg_schur.cc (100%)
rename {benchmarks => tests}/Grid_dwf_cg_unprec.cc (100%)
rename {benchmarks => tests}/Grid_dwf_even_odd.cc (100%)
rename {benchmarks => tests}/Grid_wilson_cg_prec.cc (100%)
rename {benchmarks => tests}/Grid_wilson_cg_schur.cc (100%)
rename {benchmarks => tests}/Grid_wilson_cg_unprec.cc (100%)
rename {benchmarks => tests}/Grid_wilson_evenodd.cc (100%)
diff --git a/benchmarks/Makefile.am b/benchmarks/Makefile.am
index e92f2e84..e1e092f4 100644
--- a/benchmarks/Makefile.am
+++ b/benchmarks/Makefile.am
@@ -10,53 +10,20 @@ bin_PROGRAMS = \
Grid_memory_bandwidth \
Grid_su3 \
Grid_wilson \
- Grid_wilson_evenodd \
- Grid_wilson_cg_unprec \
- Grid_wilson_cg_prec \
- Grid_wilson_cg_schur \
- Grid_dwf\
- Grid_dwf_even_odd\
- Grid_dwf_cg_unprec\
- Grid_dwf_cg_prec\
- Grid_dwf_cg_schur
+ Grid_dwf
Grid_comms_SOURCES = Grid_comms.cc
Grid_comms_LDADD = -lGrid
-Grid_su3_SOURCES = Grid_su3.cc Grid_su3_test.cc Grid_su3_expr.cc
-Grid_su3_LDADD = -lGrid
-
Grid_memory_bandwidth_SOURCES = Grid_memory_bandwidth.cc
Grid_memory_bandwidth_LDADD = -lGrid
+Grid_su3_SOURCES = Grid_su3.cc Grid_su3_test.cc Grid_su3_expr.cc
+Grid_su3_LDADD = -lGrid
Grid_wilson_SOURCES = Grid_wilson.cc
Grid_wilson_LDADD = -lGrid
-Grid_wilson_evenodd_SOURCES = Grid_wilson_evenodd.cc
-Grid_wilson_evenodd_LDADD = -lGrid
-
-Grid_wilson_cg_unprec_SOURCES = Grid_wilson_cg_unprec.cc
-Grid_wilson_cg_unprec_LDADD = -lGrid
-
-Grid_wilson_cg_prec_SOURCES = Grid_wilson_cg_prec.cc
-Grid_wilson_cg_prec_LDADD = -lGrid
-
-Grid_wilson_cg_schur_SOURCES = Grid_wilson_cg_schur.cc
-Grid_wilson_cg_schur_LDADD = -lGrid
-
Grid_dwf_SOURCES = Grid_dwf.cc
Grid_dwf_LDADD = -lGrid
-Grid_dwf_even_odd_SOURCES = Grid_dwf_even_odd.cc
-Grid_dwf_even_odd_LDADD = -lGrid
-
-Grid_dwf_cg_unprec_SOURCES = Grid_dwf_cg_unprec.cc
-Grid_dwf_cg_unprec_LDADD = -lGrid
-
-Grid_dwf_cg_prec_SOURCES = Grid_dwf_cg_prec.cc
-Grid_dwf_cg_prec_LDADD = -lGrid
-
-Grid_dwf_cg_schur_SOURCES = Grid_dwf_cg_schur.cc
-Grid_dwf_cg_schur_LDADD = -lGrid
-
diff --git a/benchmarks/Grid_dwf_cg_prec.cc b/tests/Grid_dwf_cg_prec.cc
similarity index 100%
rename from benchmarks/Grid_dwf_cg_prec.cc
rename to tests/Grid_dwf_cg_prec.cc
diff --git a/benchmarks/Grid_dwf_cg_schur.cc b/tests/Grid_dwf_cg_schur.cc
similarity index 100%
rename from benchmarks/Grid_dwf_cg_schur.cc
rename to tests/Grid_dwf_cg_schur.cc
diff --git a/benchmarks/Grid_dwf_cg_unprec.cc b/tests/Grid_dwf_cg_unprec.cc
similarity index 100%
rename from benchmarks/Grid_dwf_cg_unprec.cc
rename to tests/Grid_dwf_cg_unprec.cc
diff --git a/benchmarks/Grid_dwf_even_odd.cc b/tests/Grid_dwf_even_odd.cc
similarity index 100%
rename from benchmarks/Grid_dwf_even_odd.cc
rename to tests/Grid_dwf_even_odd.cc
diff --git a/benchmarks/Grid_wilson_cg_prec.cc b/tests/Grid_wilson_cg_prec.cc
similarity index 100%
rename from benchmarks/Grid_wilson_cg_prec.cc
rename to tests/Grid_wilson_cg_prec.cc
diff --git a/benchmarks/Grid_wilson_cg_schur.cc b/tests/Grid_wilson_cg_schur.cc
similarity index 100%
rename from benchmarks/Grid_wilson_cg_schur.cc
rename to tests/Grid_wilson_cg_schur.cc
diff --git a/benchmarks/Grid_wilson_cg_unprec.cc b/tests/Grid_wilson_cg_unprec.cc
similarity index 100%
rename from benchmarks/Grid_wilson_cg_unprec.cc
rename to tests/Grid_wilson_cg_unprec.cc
diff --git a/benchmarks/Grid_wilson_evenodd.cc b/tests/Grid_wilson_evenodd.cc
similarity index 100%
rename from benchmarks/Grid_wilson_evenodd.cc
rename to tests/Grid_wilson_evenodd.cc
diff --git a/tests/Makefile.am b/tests/Makefile.am
index 7ef4eb1f..6f92a94d 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -5,7 +5,30 @@ AM_LDFLAGS = -L$(top_builddir)/lib
#
# Test code
#
-bin_PROGRAMS = Grid_main Grid_stencil Grid_nersc_io Grid_cshift Grid_gamma Grid_simd Grid_rng Grid_remez Grid_rng_fixed Grid_cshift_red_black
+bin_PROGRAMS = Grid_main \
+ Grid_simd \
+ Grid_gamma \
+ Grid_cshift \
+ Grid_cshift_red_black \
+ Grid_stencil \
+ Grid_nersc_io \
+ Grid_rng \
+ Grid_remez \
+ Grid_rng_fixed \
+ Grid_wilson_evenodd \
+ Grid_wilson_cg_unprec \
+ Grid_wilson_cg_prec \
+ Grid_wilson_cg_schur \
+ Grid_dwf_even_odd\
+ Grid_dwf_cg_unprec\
+ Grid_dwf_cg_prec\
+ Grid_dwf_cg_schur
+
+test:
+ for f in $bin_PROGRAMS
+ do
+ ./$f > $f.log
+ done
Grid_main_SOURCES = Grid_main.cc
Grid_main_LDADD = -lGrid
@@ -39,3 +62,28 @@ Grid_simd_LDADD = -lGrid
#Grid_simd_new_SOURCES = Grid_simd_new.cc
#Grid_simd_new_LDADD = -lGrid
+
+Grid_wilson_evenodd_SOURCES = Grid_wilson_evenodd.cc
+Grid_wilson_evenodd_LDADD = -lGrid
+
+Grid_wilson_cg_unprec_SOURCES = Grid_wilson_cg_unprec.cc
+Grid_wilson_cg_unprec_LDADD = -lGrid
+
+Grid_wilson_cg_prec_SOURCES = Grid_wilson_cg_prec.cc
+Grid_wilson_cg_prec_LDADD = -lGrid
+
+Grid_wilson_cg_schur_SOURCES = Grid_wilson_cg_schur.cc
+Grid_wilson_cg_schur_LDADD = -lGrid
+
+Grid_dwf_even_odd_SOURCES = Grid_dwf_even_odd.cc
+Grid_dwf_even_odd_LDADD = -lGrid
+
+Grid_dwf_cg_unprec_SOURCES = Grid_dwf_cg_unprec.cc
+Grid_dwf_cg_unprec_LDADD = -lGrid
+
+Grid_dwf_cg_prec_SOURCES = Grid_dwf_cg_prec.cc
+Grid_dwf_cg_prec_LDADD = -lGrid
+
+Grid_dwf_cg_schur_SOURCES = Grid_dwf_cg_schur.cc
+Grid_dwf_cg_schur_LDADD = -lGrid
+
From 59163862420a81a5e9d41bd11eae2c5c329e96b9 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 3 Jun 2015 09:36:26 +0100
Subject: [PATCH 10/20] Mobius Caley form, Mobius Zolotarev operators. Pass
Even Odd vs unprec test and hermiticity checks in tests/Grid_any_evenodd.cc;
will work on inversion tests shortly.
---
lib/qcd/action/Actions.h | 8 +-
lib/qcd/action/fermion/CayleyFermion5D.cc | 110 ++++++++-
lib/qcd/action/fermion/CayleyFermion5D.h | 6 +-
lib/qcd/action/fermion/DomainWallFermion.h | 83 +------
lib/qcd/action/fermion/MobiusFermion.h | 46 ++++
.../action/fermion/MobiusZolotarevFermion.h | 48 ++++
lib/qcd/action/fermion/ScaledShamir.h | 51 ++++
lib/qcd/action/fermion/WilsonFermion5D.h | 12 +
tests/Grid_any_evenodd.cc | 226 ++++++++++++++++++
tests/Makefile.am | 12 +-
10 files changed, 507 insertions(+), 95 deletions(-)
create mode 100644 lib/qcd/action/fermion/MobiusFermion.h
create mode 100644 lib/qcd/action/fermion/MobiusZolotarevFermion.h
create mode 100644 lib/qcd/action/fermion/ScaledShamir.h
create mode 100644 tests/Grid_any_evenodd.cc
diff --git a/lib/qcd/action/Actions.h b/lib/qcd/action/Actions.h
index acbf027c..d37d1cd4 100644
--- a/lib/qcd/action/Actions.h
+++ b/lib/qcd/action/Actions.h
@@ -38,6 +38,9 @@
//#include <qcd/action/fermion/PartialFraction.h>
#include <qcd/action/fermion/DomainWallFermion.h>
+#include <qcd/action/fermion/DomainWallFermion.h>
+#include <qcd/action/fermion/MobiusFermion.h>
+#include <qcd/action/fermion/MobiusZolotarevFermion.h>
//#include <qcd/action/fermion/ScaledShamirCayleyTanh.h>
@@ -70,11 +73,6 @@
class LinearGaugeAction : public GaugeAction< multi1d<LatticeColorMatrix>, multi1d<LatticeColorMatrix> >
typedef multi1d<LatticeColorMatrix> P;
- typedef multi1d<LatticeColorMatrix> Q;
- virtual void staple(LatticeColorMatrix& result,
- const Handle< GaugeState<P,Q> >& state,
- int mu, int cb) const = 0;
*/
-
#endif
diff --git a/lib/qcd/action/fermion/CayleyFermion5D.cc b/lib/qcd/action/fermion/CayleyFermion5D.cc
index 263cc28b..be528e79 100644
--- a/lib/qcd/action/fermion/CayleyFermion5D.cc
+++ b/lib/qcd/action/fermion/CayleyFermion5D.cc
@@ -15,7 +15,6 @@ namespace QCD {
FourDimRedBlackGrid,_M5),
mass(_mass)
{
- std::cout << "Constructing a CayleyFermion5D"<<std::endl;
}
// override multiply
@@ -229,7 +228,112 @@ namespace QCD {
axpby_ssp_pplus (chi,1.0,chi,-lee[s],chi,s,s+1); // chi[Ls]
}
}
+
+ void CayleyFermion5D::SetCoefficients(RealD scale,Approx::zolotarev_data *zdata,RealD b,RealD c)
+ {
+
+ ///////////////////////////////////////////////////////////
+ // The Cayley coeffs (unprec)
+ ///////////////////////////////////////////////////////////
+ omega.resize(Ls);
+ bs.resize(Ls);
+ cs.resize(Ls);
+ as.resize(Ls);
+
+ //
+ // Ts = ( [bs+cs]Dw )^-1 ( (bs+cs) Dw )
+ // -(g5 ------- -1 ) ( g5 --------- + 1 )
+ // ( {2+(bs-cs)Dw} ) ( 2+(bs-cs) Dw )
+ //
+ // bs = 1/2( (1/omega_s + 1)*b + (1/omega - 1)*c ) = 1/2( 1/omega(b+c) + (b-c) )
+ // cs = 1/2( (1/omega_s - 1)*b + (1/omega + 1)*c ) = 1/2( 1/omega(b+c) - (b-c) )
+ //
+ // bs+cs = 0.5*( 1/omega(b+c) + (b-c) + 1/omega(b+c) - (b-c) ) = 1/omega(b+c)
+ // bs-cs = 0.5*( 1/omega(b+c) + (b-c) - 1/omega(b+c) + (b-c) ) = b-c
+ //
+ // So
+ //
+ // Ts = ( [b+c]Dw/omega_s )^-1 ( (b+c) Dw /omega_s )
+ // -(g5 ------- -1 ) ( g5 --------- + 1 )
+ // ( {2+(b-c)Dw} ) ( 2+(b-c) Dw )
+ //
+ // Ts = ( [b+c]Dw )^-1 ( (b+c) Dw )
+ // -(g5 ------- -omega_s) ( g5 --------- + omega_s )
+ // ( {2+(b-c)Dw} ) ( 2+(b-c) Dw )
+ //
+
+ double bpc = b+c;
+ double bmc = b-c;
+ for(int i=0; i < Ls; i++){
+ as[i] = 1.0;
+ omega[i] = ((double)zdata->gamma[i]); //NB reciprocal relative to Chroma NEF code
+ bs[i] = 0.5*(bpc/omega[i] + bmc);
+ cs[i] = 0.5*(bpc/omega[i] - bmc);
+ }
+
+ ////////////////////////////////////////////////////////
+ // Constants for the preconditioned matrix Cayley form
+ ////////////////////////////////////////////////////////
+ bee.resize(Ls);
+ cee.resize(Ls);
+ beo.resize(Ls);
+ ceo.resize(Ls);
+
+ for(int i=0;i<Ls;i++){
+ bee[i]=as[i]*(bs[i]*(4.0-M5) +1.0);
+ cee[i]=as[i]*(1.0-cs[i]*(4.0-M5));
+ beo[i]=as[i]*bs[i];
+ ceo[i]=-as[i]*cs[i];
+ }
+
+ aee.resize(Ls);
+ aeo.resize(Ls);
+ for(int i=0;i<Ls;i++){
+ aee[i]=cee[i];
+ aeo[i]=ceo[i];
+ }
+
+ //////////////////////////////////////////
+ // LDU decomposition of eeoo
+ //////////////////////////////////////////
+ dee.resize(Ls);
+ lee.resize(Ls);
+ leem.resize(Ls);
+ uee.resize(Ls);
+ ueem.resize(Ls);
+
+ for(int i=0;i<Ls;i++){
+
+ dee[i] = bee[i];
+
+ if ( i < Ls-1 ) {
+
+ lee[i] =-cee[i+1]/bee[i]; // sub-diag entry on the ith column
+
+ leem[i]=mass*cee[Ls-1]/bee[0];
+ for(int j=0;j<i;j++) leem[i]*= aee[j]/bee[j+1];
+
+ uee[i] =-aee[i]/bee[i]; // up-diag entry on the ith row
+
+ ueem[i]=mass;
+ for(int j=1;j<=i;j++) ueem[i]*= cee[j]/bee[j];
+ ueem[i]*= aee[0]/bee[0];
+
+ } else {
+ lee[i] =0.0;
+ leem[i]=0.0;
+ uee[i] =0.0;
+ ueem[i]=0.0;
+ }
+ }
+
+ {
+ double delta_d=mass*cee[Ls-1];
+ for(int j=0;j<Ls-1;j++) delta_d *= cee[j]/bee[j];
+ dee[Ls-1] += delta_d;
+ }
+ }
+
+}}
-}
-}
diff --git a/lib/qcd/action/fermion/CayleyFermion5D.h b/lib/qcd/action/fermion/CayleyFermion5D.h
index 857ac124..57c71992 100644
--- a/lib/qcd/action/fermion/CayleyFermion5D.h
+++ b/lib/qcd/action/fermion/CayleyFermion5D.h
@@ -22,10 +22,8 @@ namespace Grid {
virtual void MooeeInvDag (const LatticeFermion &in, LatticeFermion &out);
// protected:
-
- Approx::zolotarev_data *zdata;
-
RealD mass;
+
// Cayley form Moebius (tanh and zolotarev)
std::vector<RealD> omega;
std::vector<RealD> bs; // S dependent coeffs
@@ -53,6 +51,8 @@ namespace Grid {
GridRedBlackCartesian &FourDimRedBlackGrid,
RealD _mass,RealD _M5);
+ protected:
+ void SetCoefficients(RealD scale,Approx::zolotarev_data *zdata,RealD b,RealD c);
};
}
diff --git a/lib/qcd/action/fermion/DomainWallFermion.h b/lib/qcd/action/fermion/DomainWallFermion.h
index 2abb6eb2..3e6a9739 100644
--- a/lib/qcd/action/fermion/DomainWallFermion.h
+++ b/lib/qcd/action/fermion/DomainWallFermion.h
@@ -28,86 +28,13 @@ namespace Grid {
{
RealD eps = 1.0;
- zdata = Approx::grid_higham(eps,this->Ls);// eps is ignored for higham
+ Approx::zolotarev_data *zdata = Approx::grid_higham(eps,this->Ls);// eps is ignored for higham
assert(zdata->n==this->Ls);
+
+ std::cout << "DomainWallFermion with Ls="<<Ls<<std::endl;
+ // Call base setter
+ this->CayleyFermion5D::SetCoefficients(1.0,zdata,1.0,0.0);
- ///////////////////////////////////////////////////////////
- // The Cayley coeffs (unprec)
- ///////////////////////////////////////////////////////////
- this->omega.resize(this->Ls);
- this->bs.resize(this->Ls);
- this->cs.resize(this->Ls);
- this->as.resize(this->Ls);
-
- for(int i=0; i < this->Ls; i++){
- this->as[i] = 1.0;
- this->omega[i] = ((double)zdata -> gamma[i]);
- double bb=1.0;
- this->bs[i] = 0.5*(bb/(this->omega[i]) + 1.0);
- this->cs[i] = 0.5*(bb/(this->omega[i]) - 1.0);
- }
-
- ////////////////////////////////////////////////////////
- // Constants for the preconditioned matrix Cayley form
- ////////////////////////////////////////////////////////
- this->bee.resize(this->Ls);
- this->cee.resize(this->Ls);
- this->beo.resize(this->Ls);
- this->ceo.resize(this->Ls);
-
- for(int i=0;i<this->Ls;i++){
- this->bee[i]=as[i]*(bs[i]*(4.0-M5) +1.0);
- this->cee[i]=as[i]*(1.0-cs[i]*(4.0-M5));
- this->beo[i]=as[i]*bs[i];
- this->ceo[i]=-as[i]*cs[i];
- }
-
- aee.resize(this->Ls);
- aeo.resize(this->Ls);
- for(int i=0;i<this->Ls;i++){
- aee[i]=cee[i];
- aeo[i]=ceo[i];
- }
-
- //////////////////////////////////////////
- // LDU decomposition of eeoo
- //////////////////////////////////////////
- dee.resize(this->Ls);
- lee.resize(this->Ls);
- leem.resize(this->Ls);
- uee.resize(this->Ls);
- ueem.resize(this->Ls);
-
- for(int i=0;i<this->Ls;i++){
-
- dee[i] = bee[i];
-
- if ( i < this->Ls-1 ) {
-
- lee[i] =-cee[i+1]/bee[i]; // sub-diag entry on the ith column
-
- leem[i]=this->mass*cee[this->Ls-1]/bee[0];
- for(int j=0;j<i;j++) leem[i]*= aee[j]/bee[j+1];
-
- uee[i] =-aee[i]/bee[i]; // up-diag entry on the ith row
-
- ueem[i]=this->mass;
- for(int j=1;j<=i;j++) ueem[i]*= cee[j]/bee[j];
- ueem[i]*= aee[0]/bee[0];
-
- } else {
- lee[i] =0.0;
- leem[i]=0.0;
- uee[i] =0.0;
- ueem[i]=0.0;
- }
- }
-
- {
- double delta_d=mass*cee[this->Ls-1];
- for(int j=0;j<this->Ls-1;j++) delta_d *= cee[j]/bee[j];
- dee[this->Ls-1] += delta_d;
- }
}
};
diff --git a/lib/qcd/action/fermion/MobiusFermion.h b/lib/qcd/action/fermion/MobiusFermion.h
new file mode 100644
index 00000000..4c291fad
--- /dev/null
+++ b/lib/qcd/action/fermion/MobiusFermion.h
@@ -0,0 +1,46 @@
+#ifndef GRID_QCD_MOBIUS_FERMION_H
+#define GRID_QCD_MOBIUS_FERMION_H
+
+#include <Grid.h>
+
+namespace Grid {
+
+ namespace QCD {
+
+ class MobiusFermion : public CayleyFermion5D
+ {
+ public:
+
+ // Constructors
+ MobiusFermion(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD _M5,
+ RealD b, RealD c) :
+
+ CayleyFermion5D(_Umu,
+ FiveDimGrid,
+ FiveDimRedBlackGrid,
+ FourDimGrid,
+ FourDimRedBlackGrid,_mass,_M5)
+
+ {
+ RealD eps = 1.0;
+
+ std::cout << "MobiusFermion (b="<<b<<",c="<<c<<") with Ls= "<<Ls<<" Tanh approx"<<std::endl;
+ Approx::zolotarev_data *zdata = Approx::grid_higham(eps,this->Ls);// eps is ignored for higham
+ assert(zdata->n==this->Ls);
+
+ // Call base setter
+ this->CayleyFermion5D::SetCoefficients(1.0,zdata,b,c);
+
+ }
+
+ };
+
+ }
+}
+
+#endif
diff --git a/lib/qcd/action/fermion/MobiusZolotarevFermion.h b/lib/qcd/action/fermion/MobiusZolotarevFermion.h
new file mode 100644
index 00000000..866d0c39
--- /dev/null
+++ b/lib/qcd/action/fermion/MobiusZolotarevFermion.h
@@ -0,0 +1,48 @@
+#ifndef GRID_QCD_MOBIUS_ZOLOTAREV_FERMION_H
+#define GRID_QCD_MOBIUS_ZOLOTAREV_FERMION_H
+
+#include <Grid.h>
+
+namespace Grid {
+
+ namespace QCD {
+
+ class MobiusZolotarevFermion : public CayleyFermion5D
+ {
+ public:
+
+ // Constructors
+ MobiusZolotarevFermion(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD _M5,
+ RealD b, RealD c,
+ RealD lo, RealD hi) :
+
+ CayleyFermion5D(_Umu,
+ FiveDimGrid,
+ FiveDimRedBlackGrid,
+ FourDimGrid,
+ FourDimRedBlackGrid,_mass,_M5)
+
+ {
+ RealD eps = lo/hi;
+
+ Approx::zolotarev_data *zdata = Approx::grid_zolotarev(eps,this->Ls,0);// eps is ignored for higham
+ assert(zdata->n==this->Ls);
+
+ std::cout << "MobiusZolotarevFermion (b="<<b<<",c="<<c<<") with Ls= "<<Ls<<" Zolotarev range ["<<lo<<","<<hi<<"]"<<std::endl;
+
+ // Call base setter
+ this->CayleyFermion5D::SetCoefficients(1.0,zdata,b,c);
+
+ }
+
+ };
+
+ }
+}
+
+#endif
diff --git a/lib/qcd/action/fermion/ScaledShamir.h b/lib/qcd/action/fermion/ScaledShamir.h
new file mode 100644
index 00000000..a1fd33d0
--- /dev/null
+++ b/lib/qcd/action/fermion/ScaledShamir.h
@@ -0,0 +1,51 @@
+#ifndef GRID_QCD_DOMAIN_WALL_FERMION_H
+#define GRID_QCD_DOMAIN_WALL_FERMION_H
+
+#include <Grid.h>
+
+namespace Grid {
+
+ namespace QCD {
+
+ class ScaledShamirFermion : public CayleyFermion5D
+ {
+ public:
+
+ // Constructors
+ ScaledShamirFermion(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD _M5, RealD scale) :
+
+ CayleyFermion5D(_Umu,
+ FiveDimGrid,
+ FiveDimRedBlackGrid,
+ FourDimGrid,
+ FourDimRedBlackGrid,_mass,_M5,
+ RealD b,
+ RealD c)
+
+ {
+ RealD eps = 1.0;
+
+ Approx::zolotarev_data *zdata = Approx::grid_higham(eps,this->Ls);// eps is ignored for higham
+ assert(zdata->n==this->Ls);
+
+ //b+c = scale;
+ //b-c = 1
+ //b = 0.5(scale+1);
+ //c = 0.5(scale-1);
+
+ // Call base setter
+ this->CayleyFermion5D::SetCoefficients(1.0,zdata,0.5*(scale+1.0),0.5*(scale-1.0));
+
+ }
+
+ };
+
+ }
+}
+
+#endif
diff --git a/lib/qcd/action/fermion/WilsonFermion5D.h b/lib/qcd/action/fermion/WilsonFermion5D.h
index d4777d01..062c4d82 100644
--- a/lib/qcd/action/fermion/WilsonFermion5D.h
+++ b/lib/qcd/action/fermion/WilsonFermion5D.h
@@ -14,6 +14,18 @@ namespace Grid {
// i.e. even even contains fifth dim hopping term.
//
// [DIFFERS from original CPS red black implementation parity = (x+y+z+t+s)|2 ]
+ ////////////////////////////
+ //ContFrac:
+ // Ls always odd. Rational poly deg is either Ls or Ls-1
+ //PartFrac
+ // Ls always odd. Rational poly deg is either Ls or Ls-1
+ //
+ //Cayley: Ls always even, Rational poly deg is Ls
+ //
+ // Just set nrational as Ls. Forget about Ls-1 cases.
+ //
+ // Require odd Ls for cont and part frac
+ ////////////////////////////
////////////////////////////////////////////////////////////////////////////////
class WilsonFermion5D : public FermionOperator<LatticeFermion,LatticeGaugeField>
{
diff --git a/tests/Grid_any_evenodd.cc b/tests/Grid_any_evenodd.cc
new file mode 100644
index 00000000..8d8580c1
--- /dev/null
+++ b/tests/Grid_any_evenodd.cc
@@ -0,0 +1,226 @@
+#include <Grid.h>
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+template<class d>
+struct scal {
+ d internal;
+};
+
+ Gamma::GammaMatrix Gmu [] = {
+ Gamma::GammaX,
+ Gamma::GammaY,
+ Gamma::GammaZ,
+ Gamma::GammaT
+ };
+
+
+template<class What>
+void TestWhat(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4, GridParallelRNG *RNG5);
+
+int main (int argc, char ** argv)
+{
+ Grid_init(&argc,&argv);
+
+ int threads = GridThread::GetThreads();
+ std::cout << "Grid is setup to use "<<threads<<" threads"<<std::endl;
+
+ const int Ls=8;
+ GridCartesian * UGrid = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+ 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> seeds5({5,6,7,8});
+ GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
+ GridParallelRNG RNG4(UGrid); RNG4.SeedFixedIntegers(seeds4);
+
+ LatticeGaugeField Umu(UGrid); random(RNG4,Umu);
+ std::vector<LatticeColourMatrix> U(4,UGrid);
+
+ RealD mass=0.1;
+ RealD M5 =1.8;
+ DomainWallFermion Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+ TestWhat<DomainWallFermion>(Ddwf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ RealD b=1.5;// Scale factor b+c=2, b-c=1
+ RealD c=0.5;
+ MobiusFermion Dmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c);
+ TestWhat<MobiusFermion>(Dmob,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ MobiusZolotarevFermion Dzolo(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,0.1,4.0);
+ TestWhat<MobiusZolotarevFermion>(Dzolo,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+
+ Grid_finalize();
+}
+
+template<class What>
+void TestWhat(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5)
+{
+
+ LatticeFermion src (FGrid); random(*RNG5,src);
+ LatticeFermion phi (FGrid); random(*RNG5,phi);
+ LatticeFermion chi (FGrid); random(*RNG5,chi);
+ LatticeFermion result(FGrid); result=zero;
+ LatticeFermion ref(FGrid); ref=zero;
+ LatticeFermion tmp(FGrid); tmp=zero;
+ LatticeFermion err(FGrid); tmp=zero;
+
+ LatticeFermion src_e (FrbGrid);
+ LatticeFermion src_o (FrbGrid);
+ LatticeFermion r_e (FrbGrid);
+ LatticeFermion r_o (FrbGrid);
+ LatticeFermion r_eo (FGrid);
+ LatticeFermion r_eeoo(FGrid);
+
+ std::cout<<"=========================================================="<<std::endl;
+ std::cout<<"= Testing that Meo + Moe + Moo + Mee = Munprec "<<std::endl;
+ std::cout<<"=========================================================="<<std::endl;
+
+ pickCheckerboard(Even,src_e,src);
+ pickCheckerboard(Odd,src_o,src);
+
+ Ddwf.Meooe(src_e,r_o); std::cout<<"Applied Meo"<<std::endl;
+ Ddwf.Meooe(src_o,r_e); std::cout<<"Applied Moe"<<std::endl;
+ setCheckerboard(r_eo,r_o);
+ setCheckerboard(r_eo,r_e);
+
+ Ddwf.Mooee(src_e,r_e); std::cout<<"Applied Mee"<<std::endl;
+ Ddwf.Mooee(src_o,r_o); std::cout<<"Applied Moo"<<std::endl;
+ setCheckerboard(r_eeoo,r_e);
+ setCheckerboard(r_eeoo,r_o);
+
+ r_eo=r_eo+r_eeoo;
+ Ddwf.M(src,ref);
+
+ // std::cout << r_eo<<std::endl;
+ // std::cout << ref <<std::endl;
+
+ err= ref - r_eo;
+ std::cout << "EO norm diff "<< norm2(err)<< " "<<norm2(ref)<< " " << norm2(r_eo) <<std::endl;
+
+ LatticeComplex cerr(FGrid);
+ cerr = localInnerProduct(err,err);
+ // std::cout << cerr<<std::endl;
+
+ std::cout<<"=============================================================="<<std::endl;
+ std::cout<<"= Test Ddagger is the dagger of D by requiring "<<std::endl;
+ std::cout<<"= < phi | Deo | chi > * = < chi | Deo^dag| phi> "<<std::endl;
+ std::cout<<"=============================================================="<<std::endl;
+
+ LatticeFermion chi_e (FrbGrid);
+ LatticeFermion chi_o (FrbGrid);
+
+ LatticeFermion dchi_e (FrbGrid);
+ LatticeFermion dchi_o (FrbGrid);
+
+ LatticeFermion phi_e (FrbGrid);
+ LatticeFermion phi_o (FrbGrid);
+
+ LatticeFermion dphi_e (FrbGrid);
+ LatticeFermion dphi_o (FrbGrid);
+
+
+ pickCheckerboard(Even,chi_e,chi);
+ pickCheckerboard(Odd ,chi_o,chi);
+ pickCheckerboard(Even,phi_e,phi);
+ pickCheckerboard(Odd ,phi_o,phi);
+
+ Ddwf.Meooe(chi_e,dchi_o);
+ Ddwf.Meooe(chi_o,dchi_e);
+ Ddwf.MeooeDag(phi_e,dphi_o);
+ Ddwf.MeooeDag(phi_o,dphi_e);
+
+ ComplexD pDce = innerProduct(phi_e,dchi_e);
+ ComplexD pDco = innerProduct(phi_o,dchi_o);
+ ComplexD cDpe = innerProduct(chi_e,dphi_e);
+ ComplexD cDpo = innerProduct(chi_o,dphi_o);
+
+ std::cout <<"e "<<pDce<<" "<<cDpe <<std::endl;
+ std::cout <<"o "<<pDco<<" "<<cDpo <<std::endl;
+
+ std::cout <<"pDce - conj(cDpo) "<< pDce-conj(cDpo) <<std::endl;
+ std::cout <<"pDco - conj(cDpe) "<< pDco-conj(cDpe) <<std::endl;
+
+ std::cout<<"=============================================================="<<std::endl;
+ std::cout<<"= Test MeeInv Mee = 1 "<<std::endl;
+ std::cout<<"=============================================================="<<std::endl;
+
+ pickCheckerboard(Even,chi_e,chi);
+ pickCheckerboard(Odd ,chi_o,chi);
+
+ Ddwf.Mooee(chi_e,src_e);
+ Ddwf.MooeeInv(src_e,phi_e);
+
+ Ddwf.Mooee(chi_o,src_o);
+ Ddwf.MooeeInv(src_o,phi_o);
+
+ setCheckerboard(phi,phi_e);
+ setCheckerboard(phi,phi_o);
+
+ err = phi-chi;
+ std::cout << "norm diff "<< norm2(err)<< std::endl;
+
+ std::cout<<"=============================================================="<<std::endl;
+ std::cout<<"= Test MeeInvDag MeeDag = 1 "<<std::endl;
+ std::cout<<"=============================================================="<<std::endl;
+
+ pickCheckerboard(Even,chi_e,chi);
+ pickCheckerboard(Odd ,chi_o,chi);
+
+ Ddwf.MooeeDag(chi_e,src_e);
+ Ddwf.MooeeInvDag(src_e,phi_e);
+
+ Ddwf.MooeeDag(chi_o,src_o);
+ Ddwf.MooeeInvDag(src_o,phi_o);
+
+ setCheckerboard(phi,phi_e);
+ setCheckerboard(phi,phi_o);
+
+ err = phi-chi;
+ std::cout << "norm diff "<< norm2(err)<< std::endl;
+
+ std::cout<<"=============================================================="<<std::endl;
+ std::cout<<"= Test MpcDagMpc is Hermitian "<<std::endl;
+ std::cout<<"=============================================================="<<std::endl;
+
+ random(*RNG5,phi);
+ random(*RNG5,chi);
+ pickCheckerboard(Even,chi_e,chi);
+ pickCheckerboard(Odd ,chi_o,chi);
+ pickCheckerboard(Even,phi_e,phi);
+ pickCheckerboard(Odd ,phi_o,phi);
+ RealD t1,t2;
+
+ Ddwf.MpcDagMpc(chi_e,dchi_e,t1,t2);
+ Ddwf.MpcDagMpc(chi_o,dchi_o,t1,t2);
+
+ Ddwf.MpcDagMpc(phi_e,dphi_e,t1,t2);
+ Ddwf.MpcDagMpc(phi_o,dphi_o,t1,t2);
+
+ pDce = innerProduct(phi_e,dchi_e);
+ pDco = innerProduct(phi_o,dchi_o);
+ cDpe = innerProduct(chi_e,dphi_e);
+ cDpo = innerProduct(chi_o,dphi_o);
+
+ std::cout <<"e "<<pDce<<" "<<cDpe <<std::endl;
+ std::cout <<"o "<<pDco<<" "<<cDpo <<std::endl;
+
+ std::cout <<"pDce - conj(cDpo) "<< pDco-conj(cDpo) <<std::endl;
+ std::cout <<"pDco - conj(cDpe) "<< pDce-conj(cDpe) <<std::endl;
+
+}
diff --git a/tests/Makefile.am b/tests/Makefile.am
index 6f92a94d..f8f5df29 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -22,13 +22,10 @@ bin_PROGRAMS = Grid_main \
Grid_dwf_even_odd\
Grid_dwf_cg_unprec\
Grid_dwf_cg_prec\
- Grid_dwf_cg_schur
+ Grid_dwf_cg_schur\
+ Grid_any_evenodd
+
-test:
- for f in $bin_PROGRAMS
- do
- ./$f > $f.log
- done
Grid_main_SOURCES = Grid_main.cc
Grid_main_LDADD = -lGrid
@@ -66,6 +63,9 @@ Grid_simd_LDADD = -lGrid
Grid_wilson_evenodd_SOURCES = Grid_wilson_evenodd.cc
Grid_wilson_evenodd_LDADD = -lGrid
+Grid_any_evenodd_SOURCES = Grid_any_evenodd.cc
+Grid_any_evenodd_LDADD = -lGrid
+
Grid_wilson_cg_unprec_SOURCES = Grid_wilson_cg_unprec.cc
Grid_wilson_cg_unprec_LDADD = -lGrid
From 343d039b378647f494a9a15f66f14b8e0b1ab4db Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 3 Jun 2015 09:51:06 +0100
Subject: [PATCH 11/20] Scaled Shamir and Scaled Shamir Zolotarev aliases for
special cases of Mobius.
---
lib/qcd/action/Actions.h | 3 +-
lib/qcd/action/fermion/ScaledShamir.h | 51 -------------------
lib/qcd/action/fermion/ScaledShamirFermion.h | 37 ++++++++++++++
.../fermion/ScaledShamirZolotarevFermion.h | 39 ++++++++++++++
tests/Grid_any_evenodd.cc | 8 ++-
5 files changed, 85 insertions(+), 53 deletions(-)
delete mode 100644 lib/qcd/action/fermion/ScaledShamir.h
create mode 100644 lib/qcd/action/fermion/ScaledShamirFermion.h
create mode 100644 lib/qcd/action/fermion/ScaledShamirZolotarevFermion.h
diff --git a/lib/qcd/action/Actions.h b/lib/qcd/action/Actions.h
index d37d1cd4..893564e0 100644
--- a/lib/qcd/action/Actions.h
+++ b/lib/qcd/action/Actions.h
@@ -41,7 +41,8 @@
#include <qcd/action/fermion/DomainWallFermion.h>
#include <qcd/action/fermion/MobiusFermion.h>
#include <qcd/action/fermion/MobiusZolotarevFermion.h>
-//#include <qcd/action/fermion/ScaledShamirCayleyTanh.h>
+#include <qcd/action/fermion/ScaledShamirFermion.h>
+#include <qcd/action/fermion/ScaledShamirZolotarevFermion.h>
// Chroma interface defining FermionAction
diff --git a/lib/qcd/action/fermion/ScaledShamir.h b/lib/qcd/action/fermion/ScaledShamir.h
deleted file mode 100644
index a1fd33d0..00000000
--- a/lib/qcd/action/fermion/ScaledShamir.h
+++ /dev/null
@@ -1,51 +0,0 @@
-#ifndef GRID_QCD_DOMAIN_WALL_FERMION_H
-#define GRID_QCD_DOMAIN_WALL_FERMION_H
-
-#include <Grid.h>
-
-namespace Grid {
-
- namespace QCD {
-
- class ScaledShamirFermion : public CayleyFermion5D
- {
- public:
-
- // Constructors
- ScaledShamirFermion(LatticeGaugeField &_Umu,
- GridCartesian &FiveDimGrid,
- GridRedBlackCartesian &FiveDimRedBlackGrid,
- GridCartesian &FourDimGrid,
- GridRedBlackCartesian &FourDimRedBlackGrid,
- RealD _mass,RealD _M5, RealD scale) :
-
- CayleyFermion5D(_Umu,
- FiveDimGrid,
- FiveDimRedBlackGrid,
- FourDimGrid,
- FourDimRedBlackGrid,_mass,_M5,
- RealD b,
- RealD c)
-
- {
- RealD eps = 1.0;
-
- Approx::zolotarev_data *zdata = Approx::grid_higham(eps,this->Ls);// eps is ignored for higham
- assert(zdata->n==this->Ls);
-
- //b+c = scale;
- //b-c = 1
- //b = 0.5(scale+1);
- //c = 0.5(scale-1);
-
- // Call base setter
- this->CayleyFermion5D::SetCoefficients(1.0,zdata,0.5*(scale+1.0),0.5*(scale-1.0));
-
- }
-
- };
-
- }
-}
-
-#endif
diff --git a/lib/qcd/action/fermion/ScaledShamirFermion.h b/lib/qcd/action/fermion/ScaledShamirFermion.h
new file mode 100644
index 00000000..59fb16a8
--- /dev/null
+++ b/lib/qcd/action/fermion/ScaledShamirFermion.h
@@ -0,0 +1,37 @@
+#ifndef GRID_QCD_SCALED_SHAMIR_FERMION_H
+#define GRID_QCD_SCALED_SHAMIR_FERMION_H
+
+#include <Grid.h>
+
+namespace Grid {
+
+ namespace QCD {
+
+ class ScaledShamirFermion : public MobiusFermion
+ {
+ public:
+
+ // Constructors
+ ScaledShamirFermion(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD _M5,
+ RealD scale) :
+
+ // b+c=scale, b-c = 1 <=> 2b = scale+1; 2c = scale-1
+ MobiusFermion(_Umu,
+ FiveDimGrid,
+ FiveDimRedBlackGrid,
+ FourDimGrid,
+ FourDimRedBlackGrid,_mass,_M5,0.5*(scale+1.0),0.5*(scale-1.0))
+ {
+ }
+
+ };
+
+ }
+}
+
+#endif
diff --git a/lib/qcd/action/fermion/ScaledShamirZolotarevFermion.h b/lib/qcd/action/fermion/ScaledShamirZolotarevFermion.h
new file mode 100644
index 00000000..a4c88d5d
--- /dev/null
+++ b/lib/qcd/action/fermion/ScaledShamirZolotarevFermion.h
@@ -0,0 +1,39 @@
+#ifndef GRID_QCD_SCALED_SHAMIR_ZOLOTAREV_FERMION_H
+#define GRID_QCD_SCALED_SHAMIR_ZOLOTAREV_FERMION_H
+
+#include <Grid.h>
+
+namespace Grid {
+
+ namespace QCD {
+
+ class ScaledShamirZolotarevFermion : public MobiusZolotarevFermion
+ {
+ public:
+
+ // Constructors
+
+
+ ScaledShamirZolotarevFermion(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD _M5,
+ RealD scale,
+ RealD lo, RealD hi) :
+
+ MobiusZolotarevFermion(_Umu,
+ FiveDimGrid,
+ FiveDimRedBlackGrid,
+ FourDimGrid,
+ FourDimRedBlackGrid,_mass,_M5,0.5*(scale+1.0),0.5*(scale-1.0),lo,hi)
+
+ {}
+
+ };
+
+ }
+}
+
+#endif
diff --git a/tests/Grid_any_evenodd.cc b/tests/Grid_any_evenodd.cc
index 8d8580c1..de1e3c78 100644
--- a/tests/Grid_any_evenodd.cc
+++ b/tests/Grid_any_evenodd.cc
@@ -56,9 +56,15 @@ int main (int argc, char ** argv)
MobiusFermion Dmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c);
TestWhat<MobiusFermion>(Dmob,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
- MobiusZolotarevFermion Dzolo(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,0.1,4.0);
+ MobiusZolotarevFermion Dzolo(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,0.1,2.0);
TestWhat<MobiusZolotarevFermion>(Dzolo,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+ ScaledShamirFermion Dsham(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0);
+ TestWhat<ScaledShamirFermion>(Dsham,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ ScaledShamirZolotarevFermion Dshamz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0,0.1,2.0);
+ TestWhat<ScaledShamirZolotarevFermion>(Dshamz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
Grid_finalize();
}
From 26e9b04fab8c9e6db1b902ef0a9564785dea172b Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 3 Jun 2015 10:54:03 +0100
Subject: [PATCH 12/20] CG test written and passes i.e. converges with small
true residual in RedBlack MpcDagMpc, Unprec MdagM and Schur red black solver
for each of.
DomainWallFermion
MobiusFermion
MobiusZolotarevFermion
ScaledShamirFermion
ScaledShamirZolotarevFermion
---
.../{Grid_comms.cc => Benchmark_comms.cc} | 0
benchmarks/{Grid_dwf.cc => Benchmark_dwf.cc} | 0
...width.cc => Benchmark_memory_bandwidth.cc} | 0
benchmarks/{Grid_su3.cc => Benchmark_su3.cc} | 0
...Grid_su3_expr.cc => Benchmark_su3_expr.cc} | 0
...Grid_su3_test.cc => Benchmark_su3_test.cc} | 0
.../{Grid_wilson.cc => Benchmark_wilson.cc} | 0
benchmarks/Makefile.am | 30 ++--
tests/Makefile.am | 121 ++++++-------
tests/{Grid_cshift.cc => Test_cshift.cc} | 0
..._red_black.cc => Test_cshift_red_black.cc} | 0
...rid_dwf_cg_prec.cc => Test_dwf_cg_prec.cc} | 0
...d_dwf_cg_schur.cc => Test_dwf_cg_schur.cc} | 0
...dwf_cg_unprec.cc => Test_dwf_cg_unprec.cc} | 0
...d_dwf_even_odd.cc => Test_dwf_even_odd.cc} | 0
tests/{Grid_gamma.cc => Test_gamma.cc} | 0
tests/{Grid_main.cc => Test_main.cc} | 0
tests/Test_many_cg.cc | 164 ++++++++++++++++++
...id_any_evenodd.cc => Test_many_evenodd.cc} | 0
tests/{Grid_nersc_io.cc => Test_nersc_io.cc} | 0
tests/{Grid_remez.cc => Test_remez.cc} | 0
tests/{Grid_rng.cc => Test_rng.cc} | 0
.../{Grid_rng_fixed.cc => Test_rng_fixed.cc} | 0
tests/{Grid_simd.cc => Test_simd.cc} | 0
tests/{Grid_simd_new.cc => Test_simd_new.cc} | 0
tests/{Grid_stencil.cc => Test_stencil.cc} | 0
...lson_cg_prec.cc => Test_wilson_cg_prec.cc} | 0
...on_cg_schur.cc => Test_wilson_cg_schur.cc} | 0
..._cg_unprec.cc => Test_wilson_cg_unprec.cc} | 0
...lson_evenodd.cc => Test_wilson_evenodd.cc} | 0
30 files changed, 241 insertions(+), 74 deletions(-)
rename benchmarks/{Grid_comms.cc => Benchmark_comms.cc} (100%)
rename benchmarks/{Grid_dwf.cc => Benchmark_dwf.cc} (100%)
rename benchmarks/{Grid_memory_bandwidth.cc => Benchmark_memory_bandwidth.cc} (100%)
rename benchmarks/{Grid_su3.cc => Benchmark_su3.cc} (100%)
rename benchmarks/{Grid_su3_expr.cc => Benchmark_su3_expr.cc} (100%)
rename benchmarks/{Grid_su3_test.cc => Benchmark_su3_test.cc} (100%)
rename benchmarks/{Grid_wilson.cc => Benchmark_wilson.cc} (100%)
rename tests/{Grid_cshift.cc => Test_cshift.cc} (100%)
rename tests/{Grid_cshift_red_black.cc => Test_cshift_red_black.cc} (100%)
rename tests/{Grid_dwf_cg_prec.cc => Test_dwf_cg_prec.cc} (100%)
rename tests/{Grid_dwf_cg_schur.cc => Test_dwf_cg_schur.cc} (100%)
rename tests/{Grid_dwf_cg_unprec.cc => Test_dwf_cg_unprec.cc} (100%)
rename tests/{Grid_dwf_even_odd.cc => Test_dwf_even_odd.cc} (100%)
rename tests/{Grid_gamma.cc => Test_gamma.cc} (100%)
rename tests/{Grid_main.cc => Test_main.cc} (100%)
create mode 100644 tests/Test_many_cg.cc
rename tests/{Grid_any_evenodd.cc => Test_many_evenodd.cc} (100%)
rename tests/{Grid_nersc_io.cc => Test_nersc_io.cc} (100%)
rename tests/{Grid_remez.cc => Test_remez.cc} (100%)
rename tests/{Grid_rng.cc => Test_rng.cc} (100%)
rename tests/{Grid_rng_fixed.cc => Test_rng_fixed.cc} (100%)
rename tests/{Grid_simd.cc => Test_simd.cc} (100%)
rename tests/{Grid_simd_new.cc => Test_simd_new.cc} (100%)
rename tests/{Grid_stencil.cc => Test_stencil.cc} (100%)
rename tests/{Grid_wilson_cg_prec.cc => Test_wilson_cg_prec.cc} (100%)
rename tests/{Grid_wilson_cg_schur.cc => Test_wilson_cg_schur.cc} (100%)
rename tests/{Grid_wilson_cg_unprec.cc => Test_wilson_cg_unprec.cc} (100%)
rename tests/{Grid_wilson_evenodd.cc => Test_wilson_evenodd.cc} (100%)
diff --git a/benchmarks/Grid_comms.cc b/benchmarks/Benchmark_comms.cc
similarity index 100%
rename from benchmarks/Grid_comms.cc
rename to benchmarks/Benchmark_comms.cc
diff --git a/benchmarks/Grid_dwf.cc b/benchmarks/Benchmark_dwf.cc
similarity index 100%
rename from benchmarks/Grid_dwf.cc
rename to benchmarks/Benchmark_dwf.cc
diff --git a/benchmarks/Grid_memory_bandwidth.cc b/benchmarks/Benchmark_memory_bandwidth.cc
similarity index 100%
rename from benchmarks/Grid_memory_bandwidth.cc
rename to benchmarks/Benchmark_memory_bandwidth.cc
diff --git a/benchmarks/Grid_su3.cc b/benchmarks/Benchmark_su3.cc
similarity index 100%
rename from benchmarks/Grid_su3.cc
rename to benchmarks/Benchmark_su3.cc
diff --git a/benchmarks/Grid_su3_expr.cc b/benchmarks/Benchmark_su3_expr.cc
similarity index 100%
rename from benchmarks/Grid_su3_expr.cc
rename to benchmarks/Benchmark_su3_expr.cc
diff --git a/benchmarks/Grid_su3_test.cc b/benchmarks/Benchmark_su3_test.cc
similarity index 100%
rename from benchmarks/Grid_su3_test.cc
rename to benchmarks/Benchmark_su3_test.cc
diff --git a/benchmarks/Grid_wilson.cc b/benchmarks/Benchmark_wilson.cc
similarity index 100%
rename from benchmarks/Grid_wilson.cc
rename to benchmarks/Benchmark_wilson.cc
diff --git a/benchmarks/Makefile.am b/benchmarks/Makefile.am
index e1e092f4..74ff03c6 100644
--- a/benchmarks/Makefile.am
+++ b/benchmarks/Makefile.am
@@ -6,24 +6,24 @@ AM_LDFLAGS = -L$(top_builddir)/lib
# Test code
#
bin_PROGRAMS = \
- Grid_comms \
- Grid_memory_bandwidth \
- Grid_su3 \
- Grid_wilson \
- Grid_dwf
+ Benchmark_comms \
+ Benchmark_memory_bandwidth \
+ Benchmark_su3 \
+ Benchmark_wilson \
+ Benchmark_dwf
-Grid_comms_SOURCES = Grid_comms.cc
-Grid_comms_LDADD = -lGrid
+Benchmark_comms_SOURCES = Benchmark_comms.cc
+Benchmark_comms_LDADD = -lGrid
-Grid_memory_bandwidth_SOURCES = Grid_memory_bandwidth.cc
-Grid_memory_bandwidth_LDADD = -lGrid
+Benchmark_memory_bandwidth_SOURCES = Benchmark_memory_bandwidth.cc
+Benchmark_memory_bandwidth_LDADD = -lGrid
-Grid_su3_SOURCES = Grid_su3.cc Grid_su3_test.cc Grid_su3_expr.cc
-Grid_su3_LDADD = -lGrid
+Benchmark_su3_SOURCES = Benchmark_su3.cc Benchmark_su3_test.cc Benchmark_su3_expr.cc
+Benchmark_su3_LDADD = -lGrid
-Grid_wilson_SOURCES = Grid_wilson.cc
-Grid_wilson_LDADD = -lGrid
+Benchmark_wilson_SOURCES = Benchmark_wilson.cc
+Benchmark_wilson_LDADD = -lGrid
-Grid_dwf_SOURCES = Grid_dwf.cc
-Grid_dwf_LDADD = -lGrid
+Benchmark_dwf_SOURCES = Benchmark_dwf.cc
+Benchmark_dwf_LDADD = -lGrid
diff --git a/tests/Makefile.am b/tests/Makefile.am
index f8f5df29..bb3448f1 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -5,85 +5,88 @@ AM_LDFLAGS = -L$(top_builddir)/lib
#
# Test code
#
-bin_PROGRAMS = Grid_main \
- Grid_simd \
- Grid_gamma \
- Grid_cshift \
- Grid_cshift_red_black \
- Grid_stencil \
- Grid_nersc_io \
- Grid_rng \
- Grid_remez \
- Grid_rng_fixed \
- Grid_wilson_evenodd \
- Grid_wilson_cg_unprec \
- Grid_wilson_cg_prec \
- Grid_wilson_cg_schur \
- Grid_dwf_even_odd\
- Grid_dwf_cg_unprec\
- Grid_dwf_cg_prec\
- Grid_dwf_cg_schur\
- Grid_any_evenodd
+bin_PROGRAMS = Test_main \
+ Test_simd \
+ Test_gamma \
+ Test_cshift \
+ Test_cshift_red_black \
+ Test_stencil \
+ Test_nersc_io \
+ Test_rng \
+ Test_remez \
+ Test_rng_fixed \
+ Test_wilson_evenodd \
+ Test_wilson_cg_unprec \
+ Test_wilson_cg_prec \
+ Test_wilson_cg_schur \
+ Test_dwf_even_odd\
+ Test_dwf_cg_unprec\
+ Test_dwf_cg_prec\
+ Test_dwf_cg_schur\
+ Test_many_evenodd\
+ Test_many_cg
-Grid_main_SOURCES = Grid_main.cc
-Grid_main_LDADD = -lGrid
+Test_main_SOURCES = Test_main.cc
+Test_main_LDADD = -lGrid
-Grid_rng_SOURCES = Grid_rng.cc
-Grid_rng_LDADD = -lGrid
+Test_rng_SOURCES = Test_rng.cc
+Test_rng_LDADD = -lGrid
-Grid_rng_fixed_SOURCES = Grid_rng_fixed.cc
-Grid_rng_fixed_LDADD = -lGrid
+Test_rng_fixed_SOURCES = Test_rng_fixed.cc
+Test_rng_fixed_LDADD = -lGrid
-Grid_remez_SOURCES = Grid_remez.cc
-Grid_remez_LDADD = -lGrid
+Test_remez_SOURCES = Test_remez.cc
+Test_remez_LDADD = -lGrid
-Grid_nersc_io_SOURCES = Grid_nersc_io.cc
-Grid_nersc_io_LDADD = -lGrid
+Test_nersc_io_SOURCES = Test_nersc_io.cc
+Test_nersc_io_LDADD = -lGrid
-Grid_cshift_SOURCES = Grid_cshift.cc
-Grid_cshift_LDADD = -lGrid
+Test_cshift_SOURCES = Test_cshift.cc
+Test_cshift_LDADD = -lGrid
-Grid_cshift_red_black_SOURCES = Grid_cshift_red_black.cc
-Grid_cshift_red_black_LDADD = -lGrid
+Test_cshift_red_black_SOURCES = Test_cshift_red_black.cc
+Test_cshift_red_black_LDADD = -lGrid
-Grid_gamma_SOURCES = Grid_gamma.cc
-Grid_gamma_LDADD = -lGrid
+Test_gamma_SOURCES = Test_gamma.cc
+Test_gamma_LDADD = -lGrid
-Grid_stencil_SOURCES = Grid_stencil.cc
-Grid_stencil_LDADD = -lGrid
+Test_stencil_SOURCES = Test_stencil.cc
+Test_stencil_LDADD = -lGrid
-Grid_simd_SOURCES = Grid_simd.cc
-Grid_simd_LDADD = -lGrid
+Test_simd_SOURCES = Test_simd.cc
+Test_simd_LDADD = -lGrid
-#Grid_simd_new_SOURCES = Grid_simd_new.cc
-#Grid_simd_new_LDADD = -lGrid
+#Test_simd_new_SOURCES = Test_simd_new.cc
+#Test_simd_new_LDADD = -lGrid
-Grid_wilson_evenodd_SOURCES = Grid_wilson_evenodd.cc
-Grid_wilson_evenodd_LDADD = -lGrid
+Test_wilson_evenodd_SOURCES = Test_wilson_evenodd.cc
+Test_wilson_evenodd_LDADD = -lGrid
-Grid_any_evenodd_SOURCES = Grid_any_evenodd.cc
-Grid_any_evenodd_LDADD = -lGrid
+Test_wilson_cg_unprec_SOURCES = Test_wilson_cg_unprec.cc
+Test_wilson_cg_unprec_LDADD = -lGrid
-Grid_wilson_cg_unprec_SOURCES = Grid_wilson_cg_unprec.cc
-Grid_wilson_cg_unprec_LDADD = -lGrid
+Test_wilson_cg_prec_SOURCES = Test_wilson_cg_prec.cc
+Test_wilson_cg_prec_LDADD = -lGrid
-Grid_wilson_cg_prec_SOURCES = Grid_wilson_cg_prec.cc
-Grid_wilson_cg_prec_LDADD = -lGrid
+Test_wilson_cg_schur_SOURCES = Test_wilson_cg_schur.cc
+Test_wilson_cg_schur_LDADD = -lGrid
-Grid_wilson_cg_schur_SOURCES = Grid_wilson_cg_schur.cc
-Grid_wilson_cg_schur_LDADD = -lGrid
+Test_dwf_even_odd_SOURCES = Test_dwf_even_odd.cc
+Test_dwf_even_odd_LDADD = -lGrid
-Grid_dwf_even_odd_SOURCES = Grid_dwf_even_odd.cc
-Grid_dwf_even_odd_LDADD = -lGrid
+Test_dwf_cg_unprec_SOURCES = Test_dwf_cg_unprec.cc
+Test_dwf_cg_unprec_LDADD = -lGrid
-Grid_dwf_cg_unprec_SOURCES = Grid_dwf_cg_unprec.cc
-Grid_dwf_cg_unprec_LDADD = -lGrid
+Test_dwf_cg_prec_SOURCES = Test_dwf_cg_prec.cc
+Test_dwf_cg_prec_LDADD = -lGrid
-Grid_dwf_cg_prec_SOURCES = Grid_dwf_cg_prec.cc
-Grid_dwf_cg_prec_LDADD = -lGrid
+Test_dwf_cg_schur_SOURCES = Test_dwf_cg_schur.cc
+Test_dwf_cg_schur_LDADD = -lGrid
-Grid_dwf_cg_schur_SOURCES = Grid_dwf_cg_schur.cc
-Grid_dwf_cg_schur_LDADD = -lGrid
+Test_many_evenodd_SOURCES = Test_many_evenodd.cc
+Test_many_evenodd_LDADD = -lGrid
+Test_many_cg_SOURCES = Test_many_cg.cc
+Test_many_cg_LDADD = -lGrid
diff --git a/tests/Grid_cshift.cc b/tests/Test_cshift.cc
similarity index 100%
rename from tests/Grid_cshift.cc
rename to tests/Test_cshift.cc
diff --git a/tests/Grid_cshift_red_black.cc b/tests/Test_cshift_red_black.cc
similarity index 100%
rename from tests/Grid_cshift_red_black.cc
rename to tests/Test_cshift_red_black.cc
diff --git a/tests/Grid_dwf_cg_prec.cc b/tests/Test_dwf_cg_prec.cc
similarity index 100%
rename from tests/Grid_dwf_cg_prec.cc
rename to tests/Test_dwf_cg_prec.cc
diff --git a/tests/Grid_dwf_cg_schur.cc b/tests/Test_dwf_cg_schur.cc
similarity index 100%
rename from tests/Grid_dwf_cg_schur.cc
rename to tests/Test_dwf_cg_schur.cc
diff --git a/tests/Grid_dwf_cg_unprec.cc b/tests/Test_dwf_cg_unprec.cc
similarity index 100%
rename from tests/Grid_dwf_cg_unprec.cc
rename to tests/Test_dwf_cg_unprec.cc
diff --git a/tests/Grid_dwf_even_odd.cc b/tests/Test_dwf_even_odd.cc
similarity index 100%
rename from tests/Grid_dwf_even_odd.cc
rename to tests/Test_dwf_even_odd.cc
diff --git a/tests/Grid_gamma.cc b/tests/Test_gamma.cc
similarity index 100%
rename from tests/Grid_gamma.cc
rename to tests/Test_gamma.cc
diff --git a/tests/Grid_main.cc b/tests/Test_main.cc
similarity index 100%
rename from tests/Grid_main.cc
rename to tests/Test_main.cc
diff --git a/tests/Test_many_cg.cc b/tests/Test_many_cg.cc
new file mode 100644
index 00000000..3241a025
--- /dev/null
+++ b/tests/Test_many_cg.cc
@@ -0,0 +1,164 @@
+#include <Grid.h>
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+template<class d>
+struct scal {
+ d internal;
+};
+
+ Gamma::GammaMatrix Gmu [] = {
+ Gamma::GammaX,
+ Gamma::GammaY,
+ Gamma::GammaZ,
+ Gamma::GammaT
+ };
+
+template<class What>
+void TestCGinversions(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5);
+template<class What>
+void TestCGschur(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5);
+
+template<class What>
+void TestCGunprec(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5);
+
+template<class What>
+void TestCGprec(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5);
+
+int main (int argc, char ** argv)
+{
+ Grid_init(&argc,&argv);
+
+ int threads = GridThread::GetThreads();
+ std::cout << "Grid is setup to use "<<threads<<" threads"<<std::endl;
+
+ const int Ls=8;
+ GridCartesian * UGrid = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+ 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> seeds5({5,6,7,8});
+ GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
+ GridParallelRNG RNG4(UGrid); RNG4.SeedFixedIntegers(seeds4);
+
+ LatticeGaugeField Umu(UGrid); random(RNG4,Umu);
+ std::vector<LatticeColourMatrix> U(4,UGrid);
+
+ RealD mass=0.1;
+ RealD M5 =1.8;
+ std::cout <<"DomainWallFermion test"<<std::endl;
+ DomainWallFermion Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+ TestCGinversions<DomainWallFermion>(Ddwf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ RealD b=1.5;// Scale factor b+c=2, b-c=1
+ RealD c=0.5;
+ std::cout <<"MobiusFermion test"<<std::endl;
+ MobiusFermion Dmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c);
+ TestCGinversions<MobiusFermion>(Dmob,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ std::cout <<"MobiusZolotarevFermion test"<<std::endl;
+ MobiusZolotarevFermion Dzolo(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,0.1,2.0);
+ TestCGinversions<MobiusZolotarevFermion>(Dzolo,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ std::cout <<"ScaledShamirFermion test"<<std::endl;
+ ScaledShamirFermion Dsham(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0);
+ TestCGinversions<ScaledShamirFermion>(Dsham,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ std::cout <<"ScaledShamirZolotarevFermion test"<<std::endl;
+ ScaledShamirZolotarevFermion Dshamz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0,0.1,2.0);
+ TestCGinversions<ScaledShamirZolotarevFermion>(Dshamz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ Grid_finalize();
+}
+template<class What>
+void TestCGinversions(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5)
+{
+ std::cout << "Testing unpreconditioned inverter"<<std::endl;
+ TestCGunprec<What>(Ddwf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,RNG4,RNG5);
+ std::cout << "Testing red black preconditioned inverter"<<std::endl;
+ TestCGprec<What>(Ddwf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,RNG4,RNG5);
+ std::cout << "Testing red black Schur inverter"<<std::endl;
+ TestCGschur<What>(Ddwf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,RNG4,RNG5);
+}
+
+template<class What>
+void TestCGunprec(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5)
+{
+ LatticeFermion src (FGrid); random(*RNG5,src);
+ LatticeFermion result(FGrid); result=zero;
+
+ HermitianOperator<What,LatticeFermion> HermOp(Ddwf);
+ ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
+ CG(HermOp,src,result);
+
+}
+template<class What>
+void TestCGprec(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5)
+{
+ LatticeFermion src (FGrid); random(*RNG5,src);
+ LatticeFermion src_o(FrbGrid);
+ LatticeFermion result_o(FrbGrid);
+ pickCheckerboard(Odd,src_o,src);
+ result_o=zero;
+
+ HermitianCheckerBoardedOperator<What,LatticeFermion> HermOpEO(Ddwf);
+ ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
+ CG(HermOpEO,src_o,result_o);
+}
+
+
+template<class What>
+void TestCGschur(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5)
+{
+ LatticeFermion src (FGrid); random(*RNG5,src);
+ LatticeFermion result(FGrid); result=zero;
+
+ ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
+ SchurRedBlackSolve<LatticeFermion> SchurSolver(CG);
+ SchurSolver(Ddwf,src,result);
+}
diff --git a/tests/Grid_any_evenodd.cc b/tests/Test_many_evenodd.cc
similarity index 100%
rename from tests/Grid_any_evenodd.cc
rename to tests/Test_many_evenodd.cc
diff --git a/tests/Grid_nersc_io.cc b/tests/Test_nersc_io.cc
similarity index 100%
rename from tests/Grid_nersc_io.cc
rename to tests/Test_nersc_io.cc
diff --git a/tests/Grid_remez.cc b/tests/Test_remez.cc
similarity index 100%
rename from tests/Grid_remez.cc
rename to tests/Test_remez.cc
diff --git a/tests/Grid_rng.cc b/tests/Test_rng.cc
similarity index 100%
rename from tests/Grid_rng.cc
rename to tests/Test_rng.cc
diff --git a/tests/Grid_rng_fixed.cc b/tests/Test_rng_fixed.cc
similarity index 100%
rename from tests/Grid_rng_fixed.cc
rename to tests/Test_rng_fixed.cc
diff --git a/tests/Grid_simd.cc b/tests/Test_simd.cc
similarity index 100%
rename from tests/Grid_simd.cc
rename to tests/Test_simd.cc
diff --git a/tests/Grid_simd_new.cc b/tests/Test_simd_new.cc
similarity index 100%
rename from tests/Grid_simd_new.cc
rename to tests/Test_simd_new.cc
diff --git a/tests/Grid_stencil.cc b/tests/Test_stencil.cc
similarity index 100%
rename from tests/Grid_stencil.cc
rename to tests/Test_stencil.cc
diff --git a/tests/Grid_wilson_cg_prec.cc b/tests/Test_wilson_cg_prec.cc
similarity index 100%
rename from tests/Grid_wilson_cg_prec.cc
rename to tests/Test_wilson_cg_prec.cc
diff --git a/tests/Grid_wilson_cg_schur.cc b/tests/Test_wilson_cg_schur.cc
similarity index 100%
rename from tests/Grid_wilson_cg_schur.cc
rename to tests/Test_wilson_cg_schur.cc
diff --git a/tests/Grid_wilson_cg_unprec.cc b/tests/Test_wilson_cg_unprec.cc
similarity index 100%
rename from tests/Grid_wilson_cg_unprec.cc
rename to tests/Test_wilson_cg_unprec.cc
diff --git a/tests/Grid_wilson_evenodd.cc b/tests/Test_wilson_evenodd.cc
similarity index 100%
rename from tests/Grid_wilson_evenodd.cc
rename to tests/Test_wilson_evenodd.cc
From 8fe3d4f97126a300f76da35f3cbe9d943f6541c7 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 3 Jun 2015 11:26:54 +0100
Subject: [PATCH 13/20] Overlap Wilson Cayley tanh & zolo
---
lib/qcd/action/Actions.h | 27 ++++++++++---
.../action/fermion/MobiusZolotarevFermion.h | 1 +
.../fermion/OverlapWilsonCayleyTanhFermion.h | 34 ++++++++++++++++
.../OverlapWilsonCayleyZolotarevFermion.h | 37 ++++++++++++++++++
.../action/fermion/PartialFractionFermion5D.h | 4 +-
.../fermion/ScaledShamirZolotarevFermion.h | 39 -------------------
.../action/fermion/ShamirZolotarevFermion.h | 39 +++++++++++++++++++
tests/Test_many_cg.cc | 14 +++++--
tests/Test_many_evenodd.cc | 17 +++++++-
9 files changed, 161 insertions(+), 51 deletions(-)
create mode 100644 lib/qcd/action/fermion/OverlapWilsonCayleyTanhFermion.h
create mode 100644 lib/qcd/action/fermion/OverlapWilsonCayleyZolotarevFermion.h
delete mode 100644 lib/qcd/action/fermion/ScaledShamirZolotarevFermion.h
create mode 100644 lib/qcd/action/fermion/ShamirZolotarevFermion.h
diff --git a/lib/qcd/action/Actions.h b/lib/qcd/action/Actions.h
index 893564e0..8a8c4642 100644
--- a/lib/qcd/action/Actions.h
+++ b/lib/qcd/action/Actions.h
@@ -30,19 +30,36 @@
//#include <qcd/action/fermion/CloverFermion.h>
////////////////////////////////////////////
-// 5D formulations
+// 5D formulations...
////////////////////////////////////////////
+
#include <qcd/action/fermion/WilsonFermion5D.h> // used by all 5d overlap types
+
+//////////
+// Cayley
+//////////
#include <qcd/action/fermion/CayleyFermion5D.h>
-#include <qcd/action/fermion/ContinuedFractionFermion5D.h>
-//#include <qcd/action/fermion/PartialFraction.h>
#include <qcd/action/fermion/DomainWallFermion.h>
#include <qcd/action/fermion/DomainWallFermion.h>
+
#include <qcd/action/fermion/MobiusFermion.h>
-#include <qcd/action/fermion/MobiusZolotarevFermion.h>
#include <qcd/action/fermion/ScaledShamirFermion.h>
-#include <qcd/action/fermion/ScaledShamirZolotarevFermion.h>
+#include <qcd/action/fermion/OverlapWilsonCayleyTanhFermion.h>
+
+#include <qcd/action/fermion/MobiusZolotarevFermion.h>
+#include <qcd/action/fermion/ShamirZolotarevFermion.h>
+#include <qcd/action/fermion/OverlapWilsonCayleyZolotarevFermion.h>
+
+//////////////////////
+// Continued fraction
+//////////////////////
+#include <qcd/action/fermion/ContinuedFractionFermion5D.h>
+
+//////////////////////
+// Partial fraction
+//////////////////////
+#include <qcd/action/fermion/PartialFractionFermion5D.h>
// Chroma interface defining FermionAction
diff --git a/lib/qcd/action/fermion/MobiusZolotarevFermion.h b/lib/qcd/action/fermion/MobiusZolotarevFermion.h
index 866d0c39..9ac795d9 100644
--- a/lib/qcd/action/fermion/MobiusZolotarevFermion.h
+++ b/lib/qcd/action/fermion/MobiusZolotarevFermion.h
@@ -34,6 +34,7 @@ namespace Grid {
assert(zdata->n==this->Ls);
std::cout << "MobiusZolotarevFermion (b="<<b<<",c="<<c<<") with Ls= "<<Ls<<" Zolotarev range ["<<lo<<","<<hi<<"]"<<std::endl;
+ std::cout << "MobiusZolotarevFermion : note there is a degeneracy between (b+c) and Zolo param hi"<<std::endl;
// Call base setter
this->CayleyFermion5D::SetCoefficients(1.0,zdata,b,c);
diff --git a/lib/qcd/action/fermion/OverlapWilsonCayleyTanhFermion.h b/lib/qcd/action/fermion/OverlapWilsonCayleyTanhFermion.h
new file mode 100644
index 00000000..e764c8ae
--- /dev/null
+++ b/lib/qcd/action/fermion/OverlapWilsonCayleyTanhFermion.h
@@ -0,0 +1,34 @@
+#ifndef OVERLAP_WILSON_CAYLEY_TANH_FERMION_H
+#define OVERLAP_WILSON_CAYLEY_TANH_FERMION_H
+
+#include <Grid.h>
+
+namespace Grid {
+
+ namespace QCD {
+
+ class OverlapWilsonCayleyTanhFermion : public MobiusFermion
+ {
+ public:
+
+ // Constructors
+ OverlapWilsonCayleyTanhFermion(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD _M5,
+ RealD scale) :
+
+ // b+c=scale, b-c = 0 <=> b =c = scale/2
+ MobiusFermion(_Umu,
+ FiveDimGrid,
+ FiveDimRedBlackGrid,
+ FourDimGrid,
+ FourDimRedBlackGrid,_mass,_M5,0.5*scale,0.5*scale)
+ {
+ }
+ };
+ }
+}
+#endif
diff --git a/lib/qcd/action/fermion/OverlapWilsonCayleyZolotarevFermion.h b/lib/qcd/action/fermion/OverlapWilsonCayleyZolotarevFermion.h
new file mode 100644
index 00000000..82c43fb7
--- /dev/null
+++ b/lib/qcd/action/fermion/OverlapWilsonCayleyZolotarevFermion.h
@@ -0,0 +1,37 @@
+#ifndef OVERLAP_WILSON_CAYLEY_ZOLOTAREV_FERMION_H
+#define OVERLAP_WILSON_CAYLEY_ZOLOTAREV_FERMION_H
+
+#include <Grid.h>
+
+namespace Grid {
+
+ namespace QCD {
+
+ class OverlapWilsonCayleyZolotarevFermion : public MobiusZolotarevFermion
+ {
+ public:
+
+ // Constructors
+
+ OverlapWilsonCayleyZolotarevFermion(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD _M5,
+ RealD lo, RealD hi) :
+ // b+c=1.0, b-c = 0 <=> b =c = 1/2
+ MobiusZolotarevFermion(_Umu,
+ FiveDimGrid,
+ FiveDimRedBlackGrid,
+ FourDimGrid,
+ FourDimRedBlackGrid,_mass,_M5,0.5,0.5,lo,hi)
+
+ {}
+
+ };
+
+ }
+}
+
+#endif
diff --git a/lib/qcd/action/fermion/PartialFractionFermion5D.h b/lib/qcd/action/fermion/PartialFractionFermion5D.h
index c51aa563..95f8c0f9 100644
--- a/lib/qcd/action/fermion/PartialFractionFermion5D.h
+++ b/lib/qcd/action/fermion/PartialFractionFermion5D.h
@@ -25,10 +25,10 @@ namespace Grid {
virtual void PartialFractionCoefficients(void);
- zolotarev_data *zdata;
+ Approx::zolotarev_data *zdata;
// Part frac
- double R=(1+this->mass)/(1-this->mass);
+ double R;
std::vector<double> p;
std::vector<double> q;
diff --git a/lib/qcd/action/fermion/ScaledShamirZolotarevFermion.h b/lib/qcd/action/fermion/ScaledShamirZolotarevFermion.h
deleted file mode 100644
index a4c88d5d..00000000
--- a/lib/qcd/action/fermion/ScaledShamirZolotarevFermion.h
+++ /dev/null
@@ -1,39 +0,0 @@
-#ifndef GRID_QCD_SCALED_SHAMIR_ZOLOTAREV_FERMION_H
-#define GRID_QCD_SCALED_SHAMIR_ZOLOTAREV_FERMION_H
-
-#include <Grid.h>
-
-namespace Grid {
-
- namespace QCD {
-
- class ScaledShamirZolotarevFermion : public MobiusZolotarevFermion
- {
- public:
-
- // Constructors
-
-
- ScaledShamirZolotarevFermion(LatticeGaugeField &_Umu,
- GridCartesian &FiveDimGrid,
- GridRedBlackCartesian &FiveDimRedBlackGrid,
- GridCartesian &FourDimGrid,
- GridRedBlackCartesian &FourDimRedBlackGrid,
- RealD _mass,RealD _M5,
- RealD scale,
- RealD lo, RealD hi) :
-
- MobiusZolotarevFermion(_Umu,
- FiveDimGrid,
- FiveDimRedBlackGrid,
- FourDimGrid,
- FourDimRedBlackGrid,_mass,_M5,0.5*(scale+1.0),0.5*(scale-1.0),lo,hi)
-
- {}
-
- };
-
- }
-}
-
-#endif
diff --git a/lib/qcd/action/fermion/ShamirZolotarevFermion.h b/lib/qcd/action/fermion/ShamirZolotarevFermion.h
new file mode 100644
index 00000000..6a7df439
--- /dev/null
+++ b/lib/qcd/action/fermion/ShamirZolotarevFermion.h
@@ -0,0 +1,39 @@
+#ifndef GRID_QCD_SHAMIR_ZOLOTAREV_FERMION_H
+#define GRID_QCD_SHAMIR_ZOLOTAREV_FERMION_H
+
+#include <Grid.h>
+
+namespace Grid {
+
+ namespace QCD {
+
+ class ShamirZolotarevFermion : public MobiusZolotarevFermion
+ {
+ public:
+
+ // Constructors
+
+
+ ShamirZolotarevFermion(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD _M5,
+ RealD lo, RealD hi) :
+
+ // b+c = 1; b-c = 1 => b=1, c=0
+ MobiusZolotarevFermion(_Umu,
+ FiveDimGrid,
+ FiveDimRedBlackGrid,
+ FourDimGrid,
+ FourDimRedBlackGrid,_mass,_M5,1.0,0.0,lo,hi)
+
+ {}
+
+ };
+
+ }
+}
+
+#endif
diff --git a/tests/Test_many_cg.cc b/tests/Test_many_cg.cc
index 3241a025..4510d4b5 100644
--- a/tests/Test_many_cg.cc
+++ b/tests/Test_many_cg.cc
@@ -89,9 +89,17 @@ int main (int argc, char ** argv)
ScaledShamirFermion Dsham(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0);
TestCGinversions<ScaledShamirFermion>(Dsham,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
- std::cout <<"ScaledShamirZolotarevFermion test"<<std::endl;
- ScaledShamirZolotarevFermion Dshamz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0,0.1,2.0);
- TestCGinversions<ScaledShamirZolotarevFermion>(Dshamz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+ std::cout <<"ShamirZolotarevFermion test"<<std::endl;
+ ShamirZolotarevFermion Dshamz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
+ TestCGinversions<ShamirZolotarevFermion>(Dshamz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ std::cout <<"OverlapWilsonCayleyTanhFermion test"<<std::endl;
+ OverlapWilsonCayleyTanhFermion Dov(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+ TestCGinversions<OverlapWilsonCayleyTanhFermion>(Dov,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ std::cout <<"OverlapWilsonCayleyZolotarevFermion test"<<std::endl;
+ OverlapWilsonCayleyZolotarevFermion Dovz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
+ TestCGinversions<OverlapWilsonCayleyZolotarevFermion>(Dovz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
Grid_finalize();
}
diff --git a/tests/Test_many_evenodd.cc b/tests/Test_many_evenodd.cc
index de1e3c78..df28981b 100644
--- a/tests/Test_many_evenodd.cc
+++ b/tests/Test_many_evenodd.cc
@@ -48,23 +48,36 @@ int main (int argc, char ** argv)
RealD mass=0.1;
RealD M5 =1.8;
+ std::cout <<"DomainWallFermion test"<<std::endl;
DomainWallFermion Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
TestWhat<DomainWallFermion>(Ddwf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
RealD b=1.5;// Scale factor b+c=2, b-c=1
RealD c=0.5;
+ std::cout <<"MobiusFermion test"<<std::endl;
MobiusFermion Dmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c);
TestWhat<MobiusFermion>(Dmob,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+ std::cout <<"MobiusZolotarevFermion test"<<std::endl;
MobiusZolotarevFermion Dzolo(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,0.1,2.0);
TestWhat<MobiusZolotarevFermion>(Dzolo,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+ std::cout <<"ScaledShamirFermion test"<<std::endl;
ScaledShamirFermion Dsham(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0);
TestWhat<ScaledShamirFermion>(Dsham,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
- ScaledShamirZolotarevFermion Dshamz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0,0.1,2.0);
- TestWhat<ScaledShamirZolotarevFermion>(Dshamz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+ std::cout <<"ShamirZolotarevFermion test"<<std::endl;
+ ShamirZolotarevFermion Dshamz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
+ TestWhat<ShamirZolotarevFermion>(Dshamz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ std::cout <<"OverlapWilsonCayleyTanhFermion test"<<std::endl;
+ OverlapWilsonCayleyTanhFermion Dov(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+ TestWhat<OverlapWilsonCayleyTanhFermion>(Dov,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ std::cout <<"OverlapWilsonCayleyZolotarevFermion test"<<std::endl;
+ OverlapWilsonCayleyZolotarevFermion Dovz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
+ TestWhat<OverlapWilsonCayleyZolotarevFermion>(Dovz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
Grid_finalize();
}
From 4bcc319e11c2ac4babbbed22fb65a51f1c3787b1 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 3 Jun 2015 12:47:05 +0100
Subject: [PATCH 14/20] Reorganise of file naming
---
Makefile.am | 2 +
configure | 4 +-
configure.ac | 2 +-
lib/{Grid_algorithms.h => Algorithms.h} | 0
...aligned_allocator.h => AlignedAllocator.h} | 0
lib/Cartesian.h | 8 ++
lib/{Grid_communicator.h => Communicator.h} | 2 +-
lib/{Grid_comparison.h => Comparison.h} | 4 +-
lib/{Grid_cshift.h => Cshift.h} | 6 +-
lib/Grid.h | 26 ++--
lib/{Grid_config.h => GridConfig.h} | 0
lib/{Grid_config.h.in => GridConfig.h.in} | 2 +-
lib/{Grid_init.cc => GridInit.cc} | 0
lib/Grid_cartesian.h | 8 --
lib/Grid_math.h | 16 ---
lib/{Grid_lattice.h => Lattice.h} | 2 +-
lib/Makefile.am | 113 ++----------------
lib/{Grid_simd.h => Simd.h} | 0
lib/{Grid_stencil.h => Stencil.h} | 2 +-
lib/Tensors.h | 16 +++
lib/{Grid_threads.h => Threads.h} | 0
...Grid_cartesian_base.h => Cartesian_base.h} | 1 -
...Grid_cartesian_full.h => Cartesian_full.h} | 0
...sian_red_black.h => Cartesian_red_black.h} | 0
...ommunicator_base.h => Communicator_base.h} | 0
...ommunicator_mpi.cc => Communicator_mpi.cc} | 0
...municator_none.cc => Communicator_none.cc} | 0
.../{Grid_cshift_common.h => Cshift_common.h} | 0
.../{Grid_cshift_mpi.h => Cshift_mpi.h} | 0
.../{Grid_cshift_none.h => Cshift_none.h} | 0
.../{Grid_lattice_ET.h => Lattice_ET.h} | 0
.../{Grid_lattice_arith.h => Lattice_arith.h} | 0
.../{Grid_lattice_base.h => Lattice_base.h} | 27 ++---
...tice_comparison.h => Lattice_comparison.h} | 0
...ce_conformable.h => Lattice_conformable.h} | 0
...tice_coordinate.h => Lattice_coordinate.h} | 0
.../{Grid_lattice_local.h => Lattice_local.h} | 0
..._lattice_overload.h => Lattice_overload.h} | 0
..._lattice_peekpoke.h => Lattice_peekpoke.h} | 0
...id_lattice_reality.h => Lattice_reality.h} | 0
...attice_reduction.h => Lattice_reduction.h} | 0
.../{Grid_lattice_rng.h => Lattice_rng.h} | 0
.../{Grid_lattice_trace.h => Lattice_trace.h} | 0
..._lattice_transfer.h => Lattice_transfer.h} | 0
...attice_transpose.h => Lattice_transpose.h} | 0
.../{Grid_lattice_where.h => Lattice_where.h} | 0
lib/math/Grid_math_arith.h | 11 --
lib/parallelIO/{GridNerscIO.h => NerscIO.h} | 0
.../fermion/PartialFractionFermion5D.cc | 46 -------
lib/stencil/{Grid_lebesgue.cc => Lebesgue.cc} | 0
lib/stencil/{Grid_lebesgue.h => Lebesgue.h} | 0
...id_stencil_common.cc => Stencil_common.cc} | 0
lib/tensors/Tensor_arith.h | 11 ++
.../Tensor_arith_add.h} | 0
.../Tensor_arith_mac.h} | 0
.../Tensor_arith_mul.h} | 0
.../Tensor_arith_scalar.h} | 0
.../Tensor_arith_sub.h} | 0
.../Tensor_class.h} | 0
.../Tensor_extract_merge.h} | 0
.../Tensor_inner.h} | 0
.../Tensor_outer.h} | 0
.../Tensor_peek.h} | 0
.../Tensor_poke.h} | 0
.../Tensor_reality.h} | 0
.../Tensor_trace.h} | 0
.../Tensor_traits.h} | 0
.../Tensor_transpose.h} | 0
tests/Test_cshift.cc | 1 -
tests/Test_cshift_red_black.cc | 1 -
tests/Test_gamma.cc | 1 -
tests/Test_nersc_io.cc | 1 -
tests/Test_rng.cc | 1 -
tests/Test_rng_fixed.cc | 1 -
tests/Test_simd.cc | 1 -
75 files changed, 85 insertions(+), 231 deletions(-)
rename lib/{Grid_algorithms.h => Algorithms.h} (100%)
rename lib/{Grid_aligned_allocator.h => AlignedAllocator.h} (100%)
create mode 100644 lib/Cartesian.h
rename lib/{Grid_communicator.h => Communicator.h} (57%)
rename lib/{Grid_comparison.h => Comparison.h} (98%)
rename lib/{Grid_cshift.h => Cshift.h} (51%)
rename lib/{Grid_config.h => GridConfig.h} (100%)
rename lib/{Grid_config.h.in => GridConfig.h.in} (97%)
rename lib/{Grid_init.cc => GridInit.cc} (100%)
delete mode 100644 lib/Grid_cartesian.h
delete mode 100644 lib/Grid_math.h
rename lib/{Grid_lattice.h => Lattice.h} (58%)
rename lib/{Grid_simd.h => Simd.h} (100%)
rename lib/{Grid_stencil.h => Stencil.h} (99%)
create mode 100644 lib/Tensors.h
rename lib/{Grid_threads.h => Threads.h} (100%)
rename lib/cartesian/{Grid_cartesian_base.h => Cartesian_base.h} (99%)
rename lib/cartesian/{Grid_cartesian_full.h => Cartesian_full.h} (100%)
rename lib/cartesian/{Grid_cartesian_red_black.h => Cartesian_red_black.h} (100%)
rename lib/communicator/{Grid_communicator_base.h => Communicator_base.h} (100%)
rename lib/communicator/{Grid_communicator_mpi.cc => Communicator_mpi.cc} (100%)
rename lib/communicator/{Grid_communicator_none.cc => Communicator_none.cc} (100%)
rename lib/cshift/{Grid_cshift_common.h => Cshift_common.h} (100%)
rename lib/cshift/{Grid_cshift_mpi.h => Cshift_mpi.h} (100%)
rename lib/cshift/{Grid_cshift_none.h => Cshift_none.h} (100%)
rename lib/lattice/{Grid_lattice_ET.h => Lattice_ET.h} (100%)
rename lib/lattice/{Grid_lattice_arith.h => Lattice_arith.h} (100%)
rename lib/lattice/{Grid_lattice_base.h => Lattice_base.h} (93%)
rename lib/lattice/{Grid_lattice_comparison.h => Lattice_comparison.h} (100%)
rename lib/lattice/{Grid_lattice_conformable.h => Lattice_conformable.h} (100%)
rename lib/lattice/{Grid_lattice_coordinate.h => Lattice_coordinate.h} (100%)
rename lib/lattice/{Grid_lattice_local.h => Lattice_local.h} (100%)
rename lib/lattice/{Grid_lattice_overload.h => Lattice_overload.h} (100%)
rename lib/lattice/{Grid_lattice_peekpoke.h => Lattice_peekpoke.h} (100%)
rename lib/lattice/{Grid_lattice_reality.h => Lattice_reality.h} (100%)
rename lib/lattice/{Grid_lattice_reduction.h => Lattice_reduction.h} (100%)
rename lib/lattice/{Grid_lattice_rng.h => Lattice_rng.h} (100%)
rename lib/lattice/{Grid_lattice_trace.h => Lattice_trace.h} (100%)
rename lib/lattice/{Grid_lattice_transfer.h => Lattice_transfer.h} (100%)
rename lib/lattice/{Grid_lattice_transpose.h => Lattice_transpose.h} (100%)
rename lib/lattice/{Grid_lattice_where.h => Lattice_where.h} (100%)
delete mode 100644 lib/math/Grid_math_arith.h
rename lib/parallelIO/{GridNerscIO.h => NerscIO.h} (100%)
rename lib/stencil/{Grid_lebesgue.cc => Lebesgue.cc} (100%)
rename lib/stencil/{Grid_lebesgue.h => Lebesgue.h} (100%)
rename lib/stencil/{Grid_stencil_common.cc => Stencil_common.cc} (100%)
create mode 100644 lib/tensors/Tensor_arith.h
rename lib/{math/Grid_math_arith_add.h => tensors/Tensor_arith_add.h} (100%)
rename lib/{math/Grid_math_arith_mac.h => tensors/Tensor_arith_mac.h} (100%)
rename lib/{math/Grid_math_arith_mul.h => tensors/Tensor_arith_mul.h} (100%)
rename lib/{math/Grid_math_arith_scalar.h => tensors/Tensor_arith_scalar.h} (100%)
rename lib/{math/Grid_math_arith_sub.h => tensors/Tensor_arith_sub.h} (100%)
rename lib/{math/Grid_math_tensors.h => tensors/Tensor_class.h} (100%)
rename lib/{Grid_extract.h => tensors/Tensor_extract_merge.h} (100%)
rename lib/{math/Grid_math_inner.h => tensors/Tensor_inner.h} (100%)
rename lib/{math/Grid_math_outer.h => tensors/Tensor_outer.h} (100%)
rename lib/{math/Grid_math_peek.h => tensors/Tensor_peek.h} (100%)
rename lib/{math/Grid_math_poke.h => tensors/Tensor_poke.h} (100%)
rename lib/{math/Grid_math_reality.h => tensors/Tensor_reality.h} (100%)
rename lib/{math/Grid_math_trace.h => tensors/Tensor_trace.h} (100%)
rename lib/{math/Grid_math_traits.h => tensors/Tensor_traits.h} (100%)
rename lib/{math/Grid_math_transpose.h => tensors/Tensor_transpose.h} (100%)
diff --git a/Makefile.am b/Makefile.am
index fc3f6a0a..3b1d5690 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -1,3 +1,5 @@
# additional include paths necessary to compile the C++ library
AM_CXXFLAGS = -I$(top_srcdir)/
SUBDIRS = lib tests benchmarks
+
+filelist: $(SUBDIRS)
\ No newline at end of file
diff --git a/configure b/configure
index 615ba987..20328052 100755
--- a/configure
+++ b/configure
@@ -3064,7 +3064,7 @@ fi
-ac_config_headers="$ac_config_headers lib/Grid_config.h"
+ac_config_headers="$ac_config_headers lib/GridConfig.h"
# Check whether --enable-silent-rules was given.
if test "${enable_silent_rules+set}" = set; then :
@@ -5814,7 +5814,7 @@ cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
for ac_config_target in $ac_config_targets
do
case $ac_config_target in
- "lib/Grid_config.h") CONFIG_HEADERS="$CONFIG_HEADERS lib/Grid_config.h" ;;
+ "lib/GridConfig.h") CONFIG_HEADERS="$CONFIG_HEADERS lib/GridConfig.h" ;;
"depfiles") CONFIG_COMMANDS="$CONFIG_COMMANDS depfiles" ;;
"docs/doxy.cfg") CONFIG_FILES="$CONFIG_FILES docs/doxy.cfg" ;;
"Makefile") CONFIG_FILES="$CONFIG_FILES Makefile" ;;
diff --git a/configure.ac b/configure.ac
index 5dcbea36..03fd7900 100644
--- a/configure.ac
+++ b/configure.ac
@@ -11,7 +11,7 @@ AC_CANONICAL_SYSTEM
AM_INIT_AUTOMAKE(subdir-objects)
AC_CONFIG_MACRO_DIR([m4])
AC_CONFIG_SRCDIR([lib/Grid.h])
-AC_CONFIG_HEADERS([lib/Grid_config.h])
+AC_CONFIG_HEADERS([lib/GridConfig.h])
m4_ifdef([AM_SILENT_RULES], [AM_SILENT_RULES([yes])])
AC_MSG_NOTICE([
diff --git a/lib/Grid_algorithms.h b/lib/Algorithms.h
similarity index 100%
rename from lib/Grid_algorithms.h
rename to lib/Algorithms.h
diff --git a/lib/Grid_aligned_allocator.h b/lib/AlignedAllocator.h
similarity index 100%
rename from lib/Grid_aligned_allocator.h
rename to lib/AlignedAllocator.h
diff --git a/lib/Cartesian.h b/lib/Cartesian.h
new file mode 100644
index 00000000..db384b53
--- /dev/null
+++ b/lib/Cartesian.h
@@ -0,0 +1,8 @@
+#ifndef GRID_CARTESIAN_H
+#define GRID_CARTESIAN_H
+
+#include <cartesian/Cartesian_base.h>
+#include <cartesian/Cartesian_full.h>
+#include <cartesian/Cartesian_red_black.h>
+
+#endif
diff --git a/lib/Grid_communicator.h b/lib/Communicator.h
similarity index 57%
rename from lib/Grid_communicator.h
rename to lib/Communicator.h
index cfa6e0a7..6880adda 100644
--- a/lib/Grid_communicator.h
+++ b/lib/Communicator.h
@@ -1,6 +1,6 @@
#ifndef GRID_COMMUNICATOR_H
#define GRID_COMMUNICATOR_H
-#include <communicator/Grid_communicator_base.h>
+#include <communicator/Communicator_base.h>
#endif
diff --git a/lib/Grid_comparison.h b/lib/Comparison.h
similarity index 98%
rename from lib/Grid_comparison.h
rename to lib/Comparison.h
index 3f9c206d..ecd6ece0 100644
--- a/lib/Grid_comparison.h
+++ b/lib/Comparison.h
@@ -141,7 +141,7 @@ namespace Grid {
}
}
-#include <lattice/Grid_lattice_comparison.h>
-#include <lattice/Grid_lattice_where.h>
+#include <lattice/Lattice_comparison.h>
+#include <lattice/Lattice_where.h>
#endif
diff --git a/lib/Grid_cshift.h b/lib/Cshift.h
similarity index 51%
rename from lib/Grid_cshift.h
rename to lib/Cshift.h
index 10c7a3c4..3caccbf9 100644
--- a/lib/Grid_cshift.h
+++ b/lib/Cshift.h
@@ -1,13 +1,13 @@
#ifndef _GRID_CSHIFT_H_
#define _GRID_CSHIFT_H_
-#include <cshift/Grid_cshift_common.h>
+#include <cshift/Cshift_common.h>
#ifdef GRID_COMMS_NONE
-#include <cshift/Grid_cshift_none.h>
+#include <cshift/Cshift_none.h>
#endif
#ifdef GRID_COMMS_MPI
-#include <cshift/Grid_cshift_mpi.h>
+#include <cshift/Cshift_mpi.h>
#endif
#endif
diff --git a/lib/Grid.h b/lib/Grid.h
index 7fa56892..16530434 100644
--- a/lib/Grid.h
+++ b/lib/Grid.h
@@ -33,7 +33,7 @@
#define strong_inline __attribute__((always_inline)) inline
-#include <Grid_config.h>
+#include <GridConfig.h>
////////////////////////////////////////////////////////////
// Tunable header includes
@@ -46,22 +46,22 @@
#include <malloc.h>
#endif
-#include <Grid_aligned_allocator.h>
+#include <AlignedAllocator.h>
-#include <Grid_simd.h>
-#include <Grid_threads.h>
+#include <Simd.h>
+#include <Threads.h>
-#include <Grid_cartesian.h> // subdir aggregate
-#include <Grid_math.h> // subdir aggregate
-#include <Grid_lattice.h> // subdir aggregate
-#include <Grid_comparison.h>
-#include <Grid_cshift.h> // subdir aggregate
-#include <Grid_stencil.h> // subdir aggregate
-
-#include <Grid_algorithms.h>// subdir aggregate
+#include <Communicator.h> // subdir aggregate
+#include <Cartesian.h> // subdir aggregate
+#include <Tensors.h> // subdir aggregate
+#include <Lattice.h> // subdir aggregate
+#include <Comparison.h>
+#include <Cshift.h> // subdir aggregate
+#include <Stencil.h> // subdir aggregate
+#include <Algorithms.h>// subdir aggregate
#include <qcd/QCD.h>
-#include <parallelIO/GridNerscIO.h>
+#include <parallelIO/NerscIO.h>
namespace Grid {
diff --git a/lib/Grid_config.h b/lib/GridConfig.h
similarity index 100%
rename from lib/Grid_config.h
rename to lib/GridConfig.h
diff --git a/lib/Grid_config.h.in b/lib/GridConfig.h.in
similarity index 97%
rename from lib/Grid_config.h.in
rename to lib/GridConfig.h.in
index 6f05d6cb..03f91ff4 100644
--- a/lib/Grid_config.h.in
+++ b/lib/GridConfig.h.in
@@ -1,4 +1,4 @@
-/* lib/Grid_config.h.in. Generated from configure.ac by autoheader. */
+/* lib/GridConfig.h.in. Generated from configure.ac by autoheader. */
/* AVX */
#undef AVX1
diff --git a/lib/Grid_init.cc b/lib/GridInit.cc
similarity index 100%
rename from lib/Grid_init.cc
rename to lib/GridInit.cc
diff --git a/lib/Grid_cartesian.h b/lib/Grid_cartesian.h
deleted file mode 100644
index c01be20a..00000000
--- a/lib/Grid_cartesian.h
+++ /dev/null
@@ -1,8 +0,0 @@
-#ifndef GRID_CARTESIAN_H
-#define GRID_CARTESIAN_H
-
-#include <cartesian/Grid_cartesian_base.h>
-#include <cartesian/Grid_cartesian_full.h>
-#include <cartesian/Grid_cartesian_red_black.h>
-
-#endif
diff --git a/lib/Grid_math.h b/lib/Grid_math.h
deleted file mode 100644
index 17bc09a5..00000000
--- a/lib/Grid_math.h
+++ /dev/null
@@ -1,16 +0,0 @@
-#ifndef GRID_MATH_H
-#define GRID_MATH_H
-
-#include <math/Grid_math_traits.h>
-#include <math/Grid_math_tensors.h>
-#include <math/Grid_math_arith.h>
-#include <math/Grid_math_inner.h>
-#include <math/Grid_math_outer.h>
-#include <math/Grid_math_transpose.h>
-#include <math/Grid_math_trace.h>
-#include <math/Grid_math_peek.h>
-#include <math/Grid_math_poke.h>
-#include <math/Grid_math_reality.h>
-
-
-#endif
diff --git a/lib/Grid_lattice.h b/lib/Lattice.h
similarity index 58%
rename from lib/Grid_lattice.h
rename to lib/Lattice.h
index 35664aee..1f29a908 100644
--- a/lib/Grid_lattice.h
+++ b/lib/Lattice.h
@@ -1,6 +1,6 @@
#ifndef GRID_LATTICE_H
#define GRID_LATTICE_H
-#include <lattice/Grid_lattice_base.h>
+#include <lattice/Lattice_base.h>
#endif
diff --git a/lib/Makefile.am b/lib/Makefile.am
index b688d8b2..7ee52e0e 100644
--- a/lib/Makefile.am
+++ b/lib/Makefile.am
@@ -3,119 +3,26 @@ AM_CXXFLAGS = -I$(top_srcdir)/
extra_sources=
if BUILD_COMMS_MPI
- extra_sources+=communicator/Grid_communicator_mpi.cc
+ extra_sources+=communicator/Communicator_mpi.cc
endif
if BUILD_COMMS_NONE
- extra_sources+=communicator/Grid_communicator_none.cc
+ extra_sources+=communicator/Communicator_none.cc
endif
#
# Libraries
#
-lib_LIBRARIES = libGrid.a
-libGrid_a_SOURCES = \
- Grid_init.cc \
- stencil/Grid_lebesgue.cc \
- stencil/Grid_stencil_common.cc \
- algorithms/approx/Zolotarev.cc \
- algorithms/approx/Remez.cc \
- qcd/SpaceTimeGrid.cc\
- qcd/Dirac.cc\
- qcd/action/fermion/WilsonKernels.cc\
- qcd/action/fermion/WilsonKernelsHand.cc\
- qcd/action/fermion/WilsonFermion.cc\
- qcd/action/fermion/WilsonFermion5D.cc\
- qcd/action/fermion/CayleyFermion5D.cc \
- qcd/action/fermion/ContinuedFractionFermion5D.cc \
- $(extra_sources)
-# qcd/action/fermion/PartialFractionFermion5D.cc \
+include Make.inc
+
+lib_LIBRARIES = libGrid.a
+libGrid_a_SOURCES = $(CCFILES) $(extra_sources)
+
+
+# qcd/action/fermion/PartialFractionFermion5D.cc\ \
#
# Include files
#
-nobase_include_HEADERS=\
- ./algorithms/approx/bigfloat.h\
- ./algorithms/approx/bigfloat_double.h\
- ./algorithms/approx/Chebyshev.h\
- ./algorithms/approx/Remez.h\
- ./algorithms/approx/Zolotarev.h\
- ./algorithms/iterative/ConjugateGradient.h\
- ./algorithms/iterative/NormalEquations.h\
- ./algorithms/iterative/SchurRedBlack.h\
- ./algorithms/LinearOperator.h\
- ./algorithms/SparseMatrix.h\
- ./cartesian/Grid_cartesian_base.h\
- ./cartesian/Grid_cartesian_full.h\
- ./cartesian/Grid_cartesian_red_black.h\
- ./communicator/Grid_communicator_base.h\
- ./cshift/Grid_cshift_common.h\
- ./cshift/Grid_cshift_mpi.h\
- ./cshift/Grid_cshift_none.h\
- ./Grid.h\
- ./Grid_algorithms.h\
- ./Grid_aligned_allocator.h\
- ./Grid_cartesian.h\
- ./Grid_communicator.h\
- ./Grid_comparison.h\
- ./Grid_config.h\
- ./Grid_cshift.h\
- ./Grid_extract.h\
- ./Grid_lattice.h\
- ./Grid_math.h\
- ./Grid_simd.h\
- ./Grid_stencil.h\
- ./Grid_threads.h\
- ./lattice/Grid_lattice_arith.h\
- ./lattice/Grid_lattice_base.h\
- ./lattice/Grid_lattice_comparison.h\
- ./lattice/Grid_lattice_conformable.h\
- ./lattice/Grid_lattice_coordinate.h\
- ./lattice/Grid_lattice_ET.h\
- ./lattice/Grid_lattice_local.h\
- ./lattice/Grid_lattice_overload.h\
- ./lattice/Grid_lattice_peekpoke.h\
- ./lattice/Grid_lattice_reality.h\
- ./lattice/Grid_lattice_reduction.h\
- ./lattice/Grid_lattice_rng.h\
- ./lattice/Grid_lattice_trace.h\
- ./lattice/Grid_lattice_transfer.h\
- ./lattice/Grid_lattice_transpose.h\
- ./lattice/Grid_lattice_where.h\
- ./math/Grid_math_arith.h\
- ./math/Grid_math_arith_add.h\
- ./math/Grid_math_arith_mac.h\
- ./math/Grid_math_arith_mul.h\
- ./math/Grid_math_arith_scalar.h\
- ./math/Grid_math_arith_sub.h\
- ./math/Grid_math_inner.h\
- ./math/Grid_math_outer.h\
- ./math/Grid_math_peek.h\
- ./math/Grid_math_poke.h\
- ./math/Grid_math_reality.h\
- ./math/Grid_math_tensors.h\
- ./math/Grid_math_trace.h\
- ./math/Grid_math_traits.h\
- ./math/Grid_math_transpose.h\
- ./parallelIO/GridNerscIO.h\
- ./qcd/action/Actions.h\
- ./qcd/action/fermion/FermionOperator.h\
- ./qcd/action/fermion/WilsonCompressor.h\
- ./qcd/action/fermion/WilsonKernels.h\
- ./qcd/action/fermion/WilsonFermion.h\
- ./qcd/action/fermion/WilsonFermion5D.h\
- ./qcd/Dirac.h\
- ./qcd/QCD.h\
- ./qcd/TwoSpinor.h\
- ./simd/Grid_avx.h\
- ./simd/Grid_avx512.h\
- ./simd/Grid_qpx.h\
- ./simd/Grid_sse4.h\
- ./simd/Grid_vector_types.h\
- ./simd/Old/Grid_vComplexD.h\
- ./simd/Old/Grid_vComplexF.h\
- ./simd/Old/Grid_vInteger.h\
- ./simd/Old/Grid_vRealD.h\
- ./simd/Old/Grid_vRealF.h\
- ./stencil/Grid_lebesgue.h
+nobase_include_HEADERS=$(HFILES)
diff --git a/lib/Grid_simd.h b/lib/Simd.h
similarity index 100%
rename from lib/Grid_simd.h
rename to lib/Simd.h
diff --git a/lib/Grid_stencil.h b/lib/Stencil.h
similarity index 99%
rename from lib/Grid_stencil.h
rename to lib/Stencil.h
index 50d22453..8529e73a 100644
--- a/lib/Grid_stencil.h
+++ b/lib/Stencil.h
@@ -1,7 +1,7 @@
#ifndef GRID_STENCIL_H
#define GRID_STENCIL_H
-#include <stencil/Grid_lebesgue.h> // subdir aggregate
+#include <stencil/Lebesgue.h> // subdir aggregate
//////////////////////////////////////////////////////////////////////////////////////////
// Must not lose sight that goal is to be able to construct really efficient
diff --git a/lib/Tensors.h b/lib/Tensors.h
new file mode 100644
index 00000000..e812149f
--- /dev/null
+++ b/lib/Tensors.h
@@ -0,0 +1,16 @@
+#ifndef GRID_MATH_H
+#define GRID_MATH_H
+
+#include <tensors/Tensor_traits.h>
+#include <tensors/Tensor_class.h>
+#include <tensors/Tensor_arith.h>
+#include <tensors/Tensor_inner.h>
+#include <tensors/Tensor_outer.h>
+#include <tensors/Tensor_transpose.h>
+#include <tensors/Tensor_trace.h>
+#include <tensors/Tensor_peek.h>
+#include <tensors/Tensor_poke.h>
+#include <tensors/Tensor_reality.h>
+#include <tensors/Tensor_extract_merge.h>
+
+#endif
diff --git a/lib/Grid_threads.h b/lib/Threads.h
similarity index 100%
rename from lib/Grid_threads.h
rename to lib/Threads.h
diff --git a/lib/cartesian/Grid_cartesian_base.h b/lib/cartesian/Cartesian_base.h
similarity index 99%
rename from lib/cartesian/Grid_cartesian_base.h
rename to lib/cartesian/Cartesian_base.h
index 66339648..6303e38e 100644
--- a/lib/cartesian/Grid_cartesian_base.h
+++ b/lib/cartesian/Cartesian_base.h
@@ -2,7 +2,6 @@
#define GRID_CARTESIAN_BASE_H
#include <Grid.h>
-#include <Grid_communicator.h>
namespace Grid{
diff --git a/lib/cartesian/Grid_cartesian_full.h b/lib/cartesian/Cartesian_full.h
similarity index 100%
rename from lib/cartesian/Grid_cartesian_full.h
rename to lib/cartesian/Cartesian_full.h
diff --git a/lib/cartesian/Grid_cartesian_red_black.h b/lib/cartesian/Cartesian_red_black.h
similarity index 100%
rename from lib/cartesian/Grid_cartesian_red_black.h
rename to lib/cartesian/Cartesian_red_black.h
diff --git a/lib/communicator/Grid_communicator_base.h b/lib/communicator/Communicator_base.h
similarity index 100%
rename from lib/communicator/Grid_communicator_base.h
rename to lib/communicator/Communicator_base.h
diff --git a/lib/communicator/Grid_communicator_mpi.cc b/lib/communicator/Communicator_mpi.cc
similarity index 100%
rename from lib/communicator/Grid_communicator_mpi.cc
rename to lib/communicator/Communicator_mpi.cc
diff --git a/lib/communicator/Grid_communicator_none.cc b/lib/communicator/Communicator_none.cc
similarity index 100%
rename from lib/communicator/Grid_communicator_none.cc
rename to lib/communicator/Communicator_none.cc
diff --git a/lib/cshift/Grid_cshift_common.h b/lib/cshift/Cshift_common.h
similarity index 100%
rename from lib/cshift/Grid_cshift_common.h
rename to lib/cshift/Cshift_common.h
diff --git a/lib/cshift/Grid_cshift_mpi.h b/lib/cshift/Cshift_mpi.h
similarity index 100%
rename from lib/cshift/Grid_cshift_mpi.h
rename to lib/cshift/Cshift_mpi.h
diff --git a/lib/cshift/Grid_cshift_none.h b/lib/cshift/Cshift_none.h
similarity index 100%
rename from lib/cshift/Grid_cshift_none.h
rename to lib/cshift/Cshift_none.h
diff --git a/lib/lattice/Grid_lattice_ET.h b/lib/lattice/Lattice_ET.h
similarity index 100%
rename from lib/lattice/Grid_lattice_ET.h
rename to lib/lattice/Lattice_ET.h
diff --git a/lib/lattice/Grid_lattice_arith.h b/lib/lattice/Lattice_arith.h
similarity index 100%
rename from lib/lattice/Grid_lattice_arith.h
rename to lib/lattice/Lattice_arith.h
diff --git a/lib/lattice/Grid_lattice_base.h b/lib/lattice/Lattice_base.h
similarity index 93%
rename from lib/lattice/Grid_lattice_base.h
rename to lib/lattice/Lattice_base.h
index 15ef0944..6b5fe261 100644
--- a/lib/lattice/Grid_lattice_base.h
+++ b/lib/lattice/Lattice_base.h
@@ -283,24 +283,23 @@ PARALLEL_FOR_LOOP
-#include <lattice/Grid_lattice_conformable.h>
+#include <lattice/Lattice_conformable.h>
#define GRID_LATTICE_EXPRESSION_TEMPLATES
#ifdef GRID_LATTICE_EXPRESSION_TEMPLATES
-#include <lattice/Grid_lattice_ET.h>
+#include <lattice/Lattice_ET.h>
#else
-#include <lattice/Grid_lattice_overload.h>
+#include <lattice/Lattice_overload.h>
#endif
-#include <lattice/Grid_lattice_arith.h>
-#include <lattice/Grid_lattice_trace.h>
-#include <lattice/Grid_lattice_transpose.h>
-#include <lattice/Grid_lattice_local.h>
-#include <lattice/Grid_lattice_reduction.h>
-#include <lattice/Grid_lattice_peekpoke.h>
-#include <lattice/Grid_lattice_reality.h>
-#include <Grid_extract.h>
-#include <lattice/Grid_lattice_coordinate.h>
-#include <lattice/Grid_lattice_rng.h>
-#include <lattice/Grid_lattice_transfer.h>
+#include <lattice/Lattice_arith.h>
+#include <lattice/Lattice_trace.h>
+#include <lattice/Lattice_transpose.h>
+#include <lattice/Lattice_local.h>
+#include <lattice/Lattice_reduction.h>
+#include <lattice/Lattice_peekpoke.h>
+#include <lattice/Lattice_reality.h>
+#include <lattice/Lattice_coordinate.h>
+#include <lattice/Lattice_rng.h>
+#include <lattice/Lattice_transfer.h>
diff --git a/lib/lattice/Grid_lattice_comparison.h b/lib/lattice/Lattice_comparison.h
similarity index 100%
rename from lib/lattice/Grid_lattice_comparison.h
rename to lib/lattice/Lattice_comparison.h
diff --git a/lib/lattice/Grid_lattice_conformable.h b/lib/lattice/Lattice_conformable.h
similarity index 100%
rename from lib/lattice/Grid_lattice_conformable.h
rename to lib/lattice/Lattice_conformable.h
diff --git a/lib/lattice/Grid_lattice_coordinate.h b/lib/lattice/Lattice_coordinate.h
similarity index 100%
rename from lib/lattice/Grid_lattice_coordinate.h
rename to lib/lattice/Lattice_coordinate.h
diff --git a/lib/lattice/Grid_lattice_local.h b/lib/lattice/Lattice_local.h
similarity index 100%
rename from lib/lattice/Grid_lattice_local.h
rename to lib/lattice/Lattice_local.h
diff --git a/lib/lattice/Grid_lattice_overload.h b/lib/lattice/Lattice_overload.h
similarity index 100%
rename from lib/lattice/Grid_lattice_overload.h
rename to lib/lattice/Lattice_overload.h
diff --git a/lib/lattice/Grid_lattice_peekpoke.h b/lib/lattice/Lattice_peekpoke.h
similarity index 100%
rename from lib/lattice/Grid_lattice_peekpoke.h
rename to lib/lattice/Lattice_peekpoke.h
diff --git a/lib/lattice/Grid_lattice_reality.h b/lib/lattice/Lattice_reality.h
similarity index 100%
rename from lib/lattice/Grid_lattice_reality.h
rename to lib/lattice/Lattice_reality.h
diff --git a/lib/lattice/Grid_lattice_reduction.h b/lib/lattice/Lattice_reduction.h
similarity index 100%
rename from lib/lattice/Grid_lattice_reduction.h
rename to lib/lattice/Lattice_reduction.h
diff --git a/lib/lattice/Grid_lattice_rng.h b/lib/lattice/Lattice_rng.h
similarity index 100%
rename from lib/lattice/Grid_lattice_rng.h
rename to lib/lattice/Lattice_rng.h
diff --git a/lib/lattice/Grid_lattice_trace.h b/lib/lattice/Lattice_trace.h
similarity index 100%
rename from lib/lattice/Grid_lattice_trace.h
rename to lib/lattice/Lattice_trace.h
diff --git a/lib/lattice/Grid_lattice_transfer.h b/lib/lattice/Lattice_transfer.h
similarity index 100%
rename from lib/lattice/Grid_lattice_transfer.h
rename to lib/lattice/Lattice_transfer.h
diff --git a/lib/lattice/Grid_lattice_transpose.h b/lib/lattice/Lattice_transpose.h
similarity index 100%
rename from lib/lattice/Grid_lattice_transpose.h
rename to lib/lattice/Lattice_transpose.h
diff --git a/lib/lattice/Grid_lattice_where.h b/lib/lattice/Lattice_where.h
similarity index 100%
rename from lib/lattice/Grid_lattice_where.h
rename to lib/lattice/Lattice_where.h
diff --git a/lib/math/Grid_math_arith.h b/lib/math/Grid_math_arith.h
deleted file mode 100644
index ca90ba88..00000000
--- a/lib/math/Grid_math_arith.h
+++ /dev/null
@@ -1,11 +0,0 @@
-#ifndef GRID_MATH_ARITH_H
-#define GRID_MATH_ARITH_H
-
-#include <math/Grid_math_arith_add.h>
-#include <math/Grid_math_arith_sub.h>
-#include <math/Grid_math_arith_mac.h>
-#include <math/Grid_math_arith_mul.h>
-#include <math/Grid_math_arith_scalar.h>
-
-#endif
-
diff --git a/lib/parallelIO/GridNerscIO.h b/lib/parallelIO/NerscIO.h
similarity index 100%
rename from lib/parallelIO/GridNerscIO.h
rename to lib/parallelIO/NerscIO.h
diff --git a/lib/qcd/action/fermion/PartialFractionFermion5D.cc b/lib/qcd/action/fermion/PartialFractionFermion5D.cc
index 21f62a75..8b137891 100644
--- a/lib/qcd/action/fermion/PartialFractionFermion5D.cc
+++ b/lib/qcd/action/fermion/PartialFractionFermion5D.cc
@@ -1,47 +1 @@
-#ifndef GRID_QCD_PARTIAL_FRACTION_H
-#define GRID_QCD_PARTIAL_FRACTION_H
-namespace Grid {
-
- namespace QCD {
-
- class PartialFractionFermion5D : public WilsonFermion5D
- {
- public:
-
- // override multiply
- virtual RealD M (const LatticeFermion &in, LatticeFermion &out);
- virtual RealD Mdag (const LatticeFermion &in, LatticeFermion &out);
-
- // half checkerboard operaions
- virtual void Meooe (const LatticeFermion &in, LatticeFermion &out);
- virtual void MeooeDag (const LatticeFermion &in, LatticeFermion &out);
- virtual void Mooee (const LatticeFermion &in, LatticeFermion &out);
- virtual void MooeeDag (const LatticeFermion &in, LatticeFermion &out);
- virtual void MooeeInv (const LatticeFermion &in, LatticeFermion &out);
- virtual void MooeeInvDag (const LatticeFermion &in, LatticeFermion &out);
-
- private:
-
- zolotarev_data *zdata;
-
- // Part frac
- double R=(1+this->mass)/(1-this->mass);
- std::vector<double> p;
- std::vector<double> q;
-
- // Constructors
- PartialFractionFermion5D(LatticeGaugeField &_Umu,
- GridCartesian &FiveDimGrid,
- GridRedBlackCartesian &FiveDimRedBlackGrid,
- GridCartesian &FourDimGrid,
- GridRedBlackCartesian &FourDimRedBlackGrid,
- RealD _mass,RealD M5);
-
- };
-
-
- }
-}
-
-#endif
diff --git a/lib/stencil/Grid_lebesgue.cc b/lib/stencil/Lebesgue.cc
similarity index 100%
rename from lib/stencil/Grid_lebesgue.cc
rename to lib/stencil/Lebesgue.cc
diff --git a/lib/stencil/Grid_lebesgue.h b/lib/stencil/Lebesgue.h
similarity index 100%
rename from lib/stencil/Grid_lebesgue.h
rename to lib/stencil/Lebesgue.h
diff --git a/lib/stencil/Grid_stencil_common.cc b/lib/stencil/Stencil_common.cc
similarity index 100%
rename from lib/stencil/Grid_stencil_common.cc
rename to lib/stencil/Stencil_common.cc
diff --git a/lib/tensors/Tensor_arith.h b/lib/tensors/Tensor_arith.h
new file mode 100644
index 00000000..853a19a5
--- /dev/null
+++ b/lib/tensors/Tensor_arith.h
@@ -0,0 +1,11 @@
+#ifndef GRID_MATH_ARITH_H
+#define GRID_MATH_ARITH_H
+
+#include <tensors/Tensor_arith_add.h>
+#include <tensors/Tensor_arith_sub.h>
+#include <tensors/Tensor_arith_mac.h>
+#include <tensors/Tensor_arith_mul.h>
+#include <tensors/Tensor_arith_scalar.h>
+
+#endif
+
diff --git a/lib/math/Grid_math_arith_add.h b/lib/tensors/Tensor_arith_add.h
similarity index 100%
rename from lib/math/Grid_math_arith_add.h
rename to lib/tensors/Tensor_arith_add.h
diff --git a/lib/math/Grid_math_arith_mac.h b/lib/tensors/Tensor_arith_mac.h
similarity index 100%
rename from lib/math/Grid_math_arith_mac.h
rename to lib/tensors/Tensor_arith_mac.h
diff --git a/lib/math/Grid_math_arith_mul.h b/lib/tensors/Tensor_arith_mul.h
similarity index 100%
rename from lib/math/Grid_math_arith_mul.h
rename to lib/tensors/Tensor_arith_mul.h
diff --git a/lib/math/Grid_math_arith_scalar.h b/lib/tensors/Tensor_arith_scalar.h
similarity index 100%
rename from lib/math/Grid_math_arith_scalar.h
rename to lib/tensors/Tensor_arith_scalar.h
diff --git a/lib/math/Grid_math_arith_sub.h b/lib/tensors/Tensor_arith_sub.h
similarity index 100%
rename from lib/math/Grid_math_arith_sub.h
rename to lib/tensors/Tensor_arith_sub.h
diff --git a/lib/math/Grid_math_tensors.h b/lib/tensors/Tensor_class.h
similarity index 100%
rename from lib/math/Grid_math_tensors.h
rename to lib/tensors/Tensor_class.h
diff --git a/lib/Grid_extract.h b/lib/tensors/Tensor_extract_merge.h
similarity index 100%
rename from lib/Grid_extract.h
rename to lib/tensors/Tensor_extract_merge.h
diff --git a/lib/math/Grid_math_inner.h b/lib/tensors/Tensor_inner.h
similarity index 100%
rename from lib/math/Grid_math_inner.h
rename to lib/tensors/Tensor_inner.h
diff --git a/lib/math/Grid_math_outer.h b/lib/tensors/Tensor_outer.h
similarity index 100%
rename from lib/math/Grid_math_outer.h
rename to lib/tensors/Tensor_outer.h
diff --git a/lib/math/Grid_math_peek.h b/lib/tensors/Tensor_peek.h
similarity index 100%
rename from lib/math/Grid_math_peek.h
rename to lib/tensors/Tensor_peek.h
diff --git a/lib/math/Grid_math_poke.h b/lib/tensors/Tensor_poke.h
similarity index 100%
rename from lib/math/Grid_math_poke.h
rename to lib/tensors/Tensor_poke.h
diff --git a/lib/math/Grid_math_reality.h b/lib/tensors/Tensor_reality.h
similarity index 100%
rename from lib/math/Grid_math_reality.h
rename to lib/tensors/Tensor_reality.h
diff --git a/lib/math/Grid_math_trace.h b/lib/tensors/Tensor_trace.h
similarity index 100%
rename from lib/math/Grid_math_trace.h
rename to lib/tensors/Tensor_trace.h
diff --git a/lib/math/Grid_math_traits.h b/lib/tensors/Tensor_traits.h
similarity index 100%
rename from lib/math/Grid_math_traits.h
rename to lib/tensors/Tensor_traits.h
diff --git a/lib/math/Grid_math_transpose.h b/lib/tensors/Tensor_transpose.h
similarity index 100%
rename from lib/math/Grid_math_transpose.h
rename to lib/tensors/Tensor_transpose.h
diff --git a/tests/Test_cshift.cc b/tests/Test_cshift.cc
index 76125977..cc1da333 100644
--- a/tests/Test_cshift.cc
+++ b/tests/Test_cshift.cc
@@ -1,5 +1,4 @@
#include <Grid.h>
-#include <parallelIO/GridNerscIO.h>
using namespace Grid;
using namespace Grid::QCD;
diff --git a/tests/Test_cshift_red_black.cc b/tests/Test_cshift_red_black.cc
index 212e9b6e..9ffa66b1 100644
--- a/tests/Test_cshift_red_black.cc
+++ b/tests/Test_cshift_red_black.cc
@@ -1,5 +1,4 @@
#include <Grid.h>
-#include <parallelIO/GridNerscIO.h>
using namespace Grid;
using namespace Grid::QCD;
diff --git a/tests/Test_gamma.cc b/tests/Test_gamma.cc
index e803029b..bae77ecc 100644
--- a/tests/Test_gamma.cc
+++ b/tests/Test_gamma.cc
@@ -1,5 +1,4 @@
#include <Grid.h>
-#include <parallelIO/GridNerscIO.h>
using namespace std;
using namespace Grid;
diff --git a/tests/Test_nersc_io.cc b/tests/Test_nersc_io.cc
index 6fe587a6..80d78291 100644
--- a/tests/Test_nersc_io.cc
+++ b/tests/Test_nersc_io.cc
@@ -1,5 +1,4 @@
#include <Grid.h>
-#include <parallelIO/GridNerscIO.h>
using namespace std;
using namespace Grid;
diff --git a/tests/Test_rng.cc b/tests/Test_rng.cc
index 97f6c6b7..1731b740 100644
--- a/tests/Test_rng.cc
+++ b/tests/Test_rng.cc
@@ -1,5 +1,4 @@
#include <Grid.h>
-#include <parallelIO/GridNerscIO.h>
using namespace std;
using namespace Grid;
diff --git a/tests/Test_rng_fixed.cc b/tests/Test_rng_fixed.cc
index 01e3315b..c836c93f 100644
--- a/tests/Test_rng_fixed.cc
+++ b/tests/Test_rng_fixed.cc
@@ -1,5 +1,4 @@
#include <Grid.h>
-#include <parallelIO/GridNerscIO.h>
using namespace std;
using namespace Grid;
diff --git a/tests/Test_simd.cc b/tests/Test_simd.cc
index db600fe4..f4858d06 100644
--- a/tests/Test_simd.cc
+++ b/tests/Test_simd.cc
@@ -1,5 +1,4 @@
#include <Grid.h>
-#include <parallelIO/GridNerscIO.h>
using namespace std;
using namespace Grid;
From 98dcb6831bc110e85106168453d3c2400d9ff2c7 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 3 Jun 2015 12:47:46 +0100
Subject: [PATCH 15/20] Convenience script to build the list of headers and .cc
files in the library
---
scripts/filelist | 9 +++++++++
1 file changed, 9 insertions(+)
create mode 100755 scripts/filelist
diff --git a/scripts/filelist b/scripts/filelist
new file mode 100755
index 00000000..73c6db51
--- /dev/null
+++ b/scripts/filelist
@@ -0,0 +1,9 @@
+#!/bin/bash
+
+cd lib
+HFILES=`find . -type f -name '*.h'`
+CCFILES=`find . -type f -name '*.cc' -not -name '*ommunicator*.cc'`
+echo HFILES=$HFILES > Make.inc
+echo >> Make.inc
+echo CCFILES=$CCFILES >> Make.inc
+cd ..
From 4ef11d96e958627deda5c06dabce9d1a5d2b5bfc Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 3 Jun 2015 12:49:36 +0100
Subject: [PATCH 16/20] Make.inc needed in repo
---
lib/Make.inc | 3 +++
1 file changed, 3 insertions(+)
create mode 100644 lib/Make.inc
diff --git a/lib/Make.inc b/lib/Make.inc
new file mode 100644
index 00000000..31683256
--- /dev/null
+++ b/lib/Make.inc
@@ -0,0 +1,3 @@
+HFILES=./algorithms/approx/bigfloat.h ./algorithms/approx/bigfloat_double.h ./algorithms/approx/Chebyshev.h ./algorithms/approx/Remez.h ./algorithms/approx/Zolotarev.h ./algorithms/iterative/ConjugateGradient.h ./algorithms/iterative/NormalEquations.h ./algorithms/iterative/SchurRedBlack.h ./algorithms/LinearOperator.h ./algorithms/SparseMatrix.h ./Algorithms.h ./AlignedAllocator.h ./cartesian/Cartesian_base.h ./cartesian/Cartesian_full.h ./cartesian/Cartesian_red_black.h ./Cartesian.h ./communicator/Communicator_base.h ./Communicator.h ./Comparison.h ./cshift/Cshift_common.h ./cshift/Cshift_mpi.h ./cshift/Cshift_none.h ./Cshift.h ./Grid.h ./GridConfig.h ./lattice/Lattice_arith.h ./lattice/Lattice_base.h ./lattice/Lattice_comparison.h ./lattice/Lattice_conformable.h ./lattice/Lattice_coordinate.h ./lattice/Lattice_ET.h ./lattice/Lattice_local.h ./lattice/Lattice_overload.h ./lattice/Lattice_peekpoke.h ./lattice/Lattice_reality.h ./lattice/Lattice_reduction.h ./lattice/Lattice_rng.h ./lattice/Lattice_trace.h ./lattice/Lattice_transfer.h ./lattice/Lattice_transpose.h ./lattice/Lattice_where.h ./Lattice.h ./parallelIO/NerscIO.h ./qcd/action/Actions.h ./qcd/action/DiffAction.h ./qcd/action/fermion/CayleyFermion5D.h ./qcd/action/fermion/ContinuedFractionFermion5D.h ./qcd/action/fermion/DomainWallFermion.h ./qcd/action/fermion/FermionOperator.h ./qcd/action/fermion/MobiusFermion.h ./qcd/action/fermion/MobiusZolotarevFermion.h ./qcd/action/fermion/OverlapWilsonCayleyTanhFermion.h ./qcd/action/fermion/OverlapWilsonCayleyZolotarevFermion.h ./qcd/action/fermion/PartialFractionFermion5D.h ./qcd/action/fermion/ScaledShamirFermion.h ./qcd/action/fermion/ShamirZolotarevFermion.h ./qcd/action/fermion/WilsonCompressor.h ./qcd/action/fermion/WilsonFermion.h ./qcd/action/fermion/WilsonFermion5D.h ./qcd/action/fermion/WilsonKernels.h ./qcd/Dirac.h ./qcd/LinalgUtils.h ./qcd/QCD.h ./qcd/SpaceTimeGrid.h ./qcd/TwoSpinor.h ./simd/Grid_avx.h ./simd/Grid_avx512.h ./simd/Grid_qpx.h ./simd/Grid_sse4.h ./simd/Grid_vector_types.h ./simd/Old/Grid_vComplexD.h ./simd/Old/Grid_vComplexF.h ./simd/Old/Grid_vInteger.h ./simd/Old/Grid_vRealD.h ./simd/Old/Grid_vRealF.h ./Simd.h ./stencil/Lebesgue.h ./Stencil.h ./tensors/Tensor_arith.h ./tensors/Tensor_arith_add.h ./tensors/Tensor_arith_mac.h ./tensors/Tensor_arith_mul.h ./tensors/Tensor_arith_scalar.h ./tensors/Tensor_arith_sub.h ./tensors/Tensor_class.h ./tensors/Tensor_extract_merge.h ./tensors/Tensor_inner.h ./tensors/Tensor_outer.h ./tensors/Tensor_peek.h ./tensors/Tensor_poke.h ./tensors/Tensor_reality.h ./tensors/Tensor_trace.h ./tensors/Tensor_traits.h ./tensors/Tensor_transpose.h ./Tensors.h ./Threads.h
+
+CCFILES=./algorithms/approx/Remez.cc ./algorithms/approx/Zolotarev.cc ./GridInit.cc ./qcd/action/fermion/CayleyFermion5D.cc ./qcd/action/fermion/ContinuedFractionFermion5D.cc ./qcd/action/fermion/PartialFractionFermion5D.cc ./qcd/action/fermion/WilsonFermion.cc ./qcd/action/fermion/WilsonFermion5D.cc ./qcd/action/fermion/WilsonKernels.cc ./qcd/action/fermion/WilsonKernelsHand.cc ./qcd/Dirac.cc ./qcd/SpaceTimeGrid.cc ./stencil/Lebesgue.cc ./stencil/Stencil_common.cc
From eaa3e6aaf6aff695eb8bc60e384133864092a0d0 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Wed, 3 Jun 2015 13:07:00 +0100
Subject: [PATCH 17/20] Assist for generating file lists contained in Make.inc
files for convenience when things are added
---
benchmarks/Make.inc | 23 +++
benchmarks/Makefile.am | 23 +--
...nchmark_su3_expr.cc => simple_su3_expr.cc} | 0
...nchmark_su3_test.cc => simple_su3_test.cc} | 0
lib/Make.inc | 1 +
scripts/filelist | 49 +++++-
tests/Make.inc | 83 +++++++++
tests/Makefile.am | 89 +---------
tests/Test_simd_new.cc | 165 ------------------
9 files changed, 157 insertions(+), 276 deletions(-)
create mode 100644 benchmarks/Make.inc
rename benchmarks/{Benchmark_su3_expr.cc => simple_su3_expr.cc} (100%)
rename benchmarks/{Benchmark_su3_test.cc => simple_su3_test.cc} (100%)
create mode 100644 tests/Make.inc
delete mode 100644 tests/Test_simd_new.cc
diff --git a/benchmarks/Make.inc b/benchmarks/Make.inc
new file mode 100644
index 00000000..9871149c
--- /dev/null
+++ b/benchmarks/Make.inc
@@ -0,0 +1,23 @@
+
+bin_PROGRAMS = Benchmark_comms Benchmark_dwf Benchmark_memory_bandwidth Benchmark_su3 Benchmark_wilson
+
+
+Benchmark_comms_SOURCES=Benchmark_comms.cc
+Benchmark_comms_LDADD=-lGrid
+
+
+Benchmark_dwf_SOURCES=Benchmark_dwf.cc
+Benchmark_dwf_LDADD=-lGrid
+
+
+Benchmark_memory_bandwidth_SOURCES=Benchmark_memory_bandwidth.cc
+Benchmark_memory_bandwidth_LDADD=-lGrid
+
+
+Benchmark_su3_SOURCES=Benchmark_su3.cc
+Benchmark_su3_LDADD=-lGrid
+
+
+Benchmark_wilson_SOURCES=Benchmark_wilson.cc
+Benchmark_wilson_LDADD=-lGrid
+
diff --git a/benchmarks/Makefile.am b/benchmarks/Makefile.am
index 74ff03c6..8081844f 100644
--- a/benchmarks/Makefile.am
+++ b/benchmarks/Makefile.am
@@ -5,25 +5,4 @@ AM_LDFLAGS = -L$(top_builddir)/lib
#
# Test code
#
-bin_PROGRAMS = \
- Benchmark_comms \
- Benchmark_memory_bandwidth \
- Benchmark_su3 \
- Benchmark_wilson \
- Benchmark_dwf
-
-Benchmark_comms_SOURCES = Benchmark_comms.cc
-Benchmark_comms_LDADD = -lGrid
-
-Benchmark_memory_bandwidth_SOURCES = Benchmark_memory_bandwidth.cc
-Benchmark_memory_bandwidth_LDADD = -lGrid
-
-Benchmark_su3_SOURCES = Benchmark_su3.cc Benchmark_su3_test.cc Benchmark_su3_expr.cc
-Benchmark_su3_LDADD = -lGrid
-
-Benchmark_wilson_SOURCES = Benchmark_wilson.cc
-Benchmark_wilson_LDADD = -lGrid
-
-Benchmark_dwf_SOURCES = Benchmark_dwf.cc
-Benchmark_dwf_LDADD = -lGrid
-
+include Make.inc
diff --git a/benchmarks/Benchmark_su3_expr.cc b/benchmarks/simple_su3_expr.cc
similarity index 100%
rename from benchmarks/Benchmark_su3_expr.cc
rename to benchmarks/simple_su3_expr.cc
diff --git a/benchmarks/Benchmark_su3_test.cc b/benchmarks/simple_su3_test.cc
similarity index 100%
rename from benchmarks/Benchmark_su3_test.cc
rename to benchmarks/simple_su3_test.cc
diff --git a/lib/Make.inc b/lib/Make.inc
index 31683256..dd5269ac 100644
--- a/lib/Make.inc
+++ b/lib/Make.inc
@@ -1,3 +1,4 @@
+
HFILES=./algorithms/approx/bigfloat.h ./algorithms/approx/bigfloat_double.h ./algorithms/approx/Chebyshev.h ./algorithms/approx/Remez.h ./algorithms/approx/Zolotarev.h ./algorithms/iterative/ConjugateGradient.h ./algorithms/iterative/NormalEquations.h ./algorithms/iterative/SchurRedBlack.h ./algorithms/LinearOperator.h ./algorithms/SparseMatrix.h ./Algorithms.h ./AlignedAllocator.h ./cartesian/Cartesian_base.h ./cartesian/Cartesian_full.h ./cartesian/Cartesian_red_black.h ./Cartesian.h ./communicator/Communicator_base.h ./Communicator.h ./Comparison.h ./cshift/Cshift_common.h ./cshift/Cshift_mpi.h ./cshift/Cshift_none.h ./Cshift.h ./Grid.h ./GridConfig.h ./lattice/Lattice_arith.h ./lattice/Lattice_base.h ./lattice/Lattice_comparison.h ./lattice/Lattice_conformable.h ./lattice/Lattice_coordinate.h ./lattice/Lattice_ET.h ./lattice/Lattice_local.h ./lattice/Lattice_overload.h ./lattice/Lattice_peekpoke.h ./lattice/Lattice_reality.h ./lattice/Lattice_reduction.h ./lattice/Lattice_rng.h ./lattice/Lattice_trace.h ./lattice/Lattice_transfer.h ./lattice/Lattice_transpose.h ./lattice/Lattice_where.h ./Lattice.h ./parallelIO/NerscIO.h ./qcd/action/Actions.h ./qcd/action/DiffAction.h ./qcd/action/fermion/CayleyFermion5D.h ./qcd/action/fermion/ContinuedFractionFermion5D.h ./qcd/action/fermion/DomainWallFermion.h ./qcd/action/fermion/FermionOperator.h ./qcd/action/fermion/MobiusFermion.h ./qcd/action/fermion/MobiusZolotarevFermion.h ./qcd/action/fermion/OverlapWilsonCayleyTanhFermion.h ./qcd/action/fermion/OverlapWilsonCayleyZolotarevFermion.h ./qcd/action/fermion/PartialFractionFermion5D.h ./qcd/action/fermion/ScaledShamirFermion.h ./qcd/action/fermion/ShamirZolotarevFermion.h ./qcd/action/fermion/WilsonCompressor.h ./qcd/action/fermion/WilsonFermion.h ./qcd/action/fermion/WilsonFermion5D.h ./qcd/action/fermion/WilsonKernels.h ./qcd/Dirac.h ./qcd/LinalgUtils.h ./qcd/QCD.h ./qcd/SpaceTimeGrid.h ./qcd/TwoSpinor.h ./simd/Grid_avx.h ./simd/Grid_avx512.h ./simd/Grid_qpx.h ./simd/Grid_sse4.h ./simd/Grid_vector_types.h ./simd/Old/Grid_vComplexD.h ./simd/Old/Grid_vComplexF.h ./simd/Old/Grid_vInteger.h ./simd/Old/Grid_vRealD.h ./simd/Old/Grid_vRealF.h ./Simd.h ./stencil/Lebesgue.h ./Stencil.h ./tensors/Tensor_arith.h ./tensors/Tensor_arith_add.h ./tensors/Tensor_arith_mac.h ./tensors/Tensor_arith_mul.h ./tensors/Tensor_arith_scalar.h ./tensors/Tensor_arith_sub.h ./tensors/Tensor_class.h ./tensors/Tensor_extract_merge.h ./tensors/Tensor_inner.h ./tensors/Tensor_outer.h ./tensors/Tensor_peek.h ./tensors/Tensor_poke.h ./tensors/Tensor_reality.h ./tensors/Tensor_trace.h ./tensors/Tensor_traits.h ./tensors/Tensor_transpose.h ./Tensors.h ./Threads.h
CCFILES=./algorithms/approx/Remez.cc ./algorithms/approx/Zolotarev.cc ./GridInit.cc ./qcd/action/fermion/CayleyFermion5D.cc ./qcd/action/fermion/ContinuedFractionFermion5D.cc ./qcd/action/fermion/PartialFractionFermion5D.cc ./qcd/action/fermion/WilsonFermion.cc ./qcd/action/fermion/WilsonFermion5D.cc ./qcd/action/fermion/WilsonKernels.cc ./qcd/action/fermion/WilsonKernelsHand.cc ./qcd/Dirac.cc ./qcd/SpaceTimeGrid.cc ./stencil/Lebesgue.cc ./stencil/Stencil_common.cc
diff --git a/scripts/filelist b/scripts/filelist
index 73c6db51..1b575611 100755
--- a/scripts/filelist
+++ b/scripts/filelist
@@ -1,9 +1,56 @@
#!/bin/bash
cd lib
+
HFILES=`find . -type f -name '*.h'`
CCFILES=`find . -type f -name '*.cc' -not -name '*ommunicator*.cc'`
-echo HFILES=$HFILES > Make.inc
+echo> Make.inc
+echo HFILES=$HFILES >> Make.inc
echo >> Make.inc
echo CCFILES=$CCFILES >> Make.inc
+
cd ..
+
+
+
+cd tests
+
+echo> Make.inc
+TESTS=`ls T*.cc`
+TESTLIST=`echo ${TESTS} | sed s/.cc//g `
+
+echo > Make.inc
+echo bin_PROGRAMS = ${TESTLIST} >> Make.inc
+echo >> Make.inc
+
+for f in $TESTS
+do
+BNAME=`basename $f .cc`
+echo >> Make.inc
+echo ${BNAME}_SOURCES=$f >> Make.inc
+echo ${BNAME}_LDADD=-lGrid>> Make.inc
+echo >> Make.inc
+done
+
+cd ..
+
+
+cd benchmarks
+
+
+echo> Make.inc
+TESTS=`ls B*.cc`
+TESTLIST=`echo ${TESTS} | sed s/.cc//g `
+
+echo > Make.inc
+echo bin_PROGRAMS = ${TESTLIST} >> Make.inc
+echo >> Make.inc
+
+for f in $TESTS
+do
+BNAME=`basename $f .cc`
+echo >> Make.inc
+echo ${BNAME}_SOURCES=$f >> Make.inc
+echo ${BNAME}_LDADD=-lGrid>> Make.inc
+echo >> Make.inc
+done
diff --git a/tests/Make.inc b/tests/Make.inc
new file mode 100644
index 00000000..d592f218
--- /dev/null
+++ b/tests/Make.inc
@@ -0,0 +1,83 @@
+
+bin_PROGRAMS = Test_cshift Test_cshift_red_black Test_dwf_cg_prec Test_dwf_cg_schur Test_dwf_cg_unprec Test_dwf_even_odd Test_gamma Test_main Test_many_cg Test_many_evenodd Test_nersc_io Test_remez Test_rng Test_rng_fixed Test_simd Test_stencil Test_wilson_cg_prec Test_wilson_cg_schur Test_wilson_cg_unprec Test_wilson_evenodd
+
+
+Test_cshift_SOURCES=Test_cshift.cc
+Test_cshift_LDADD=-lGrid
+
+
+Test_cshift_red_black_SOURCES=Test_cshift_red_black.cc
+Test_cshift_red_black_LDADD=-lGrid
+
+
+Test_dwf_cg_prec_SOURCES=Test_dwf_cg_prec.cc
+Test_dwf_cg_prec_LDADD=-lGrid
+
+
+Test_dwf_cg_schur_SOURCES=Test_dwf_cg_schur.cc
+Test_dwf_cg_schur_LDADD=-lGrid
+
+
+Test_dwf_cg_unprec_SOURCES=Test_dwf_cg_unprec.cc
+Test_dwf_cg_unprec_LDADD=-lGrid
+
+
+Test_dwf_even_odd_SOURCES=Test_dwf_even_odd.cc
+Test_dwf_even_odd_LDADD=-lGrid
+
+
+Test_gamma_SOURCES=Test_gamma.cc
+Test_gamma_LDADD=-lGrid
+
+
+Test_main_SOURCES=Test_main.cc
+Test_main_LDADD=-lGrid
+
+
+Test_many_cg_SOURCES=Test_many_cg.cc
+Test_many_cg_LDADD=-lGrid
+
+
+Test_many_evenodd_SOURCES=Test_many_evenodd.cc
+Test_many_evenodd_LDADD=-lGrid
+
+
+Test_nersc_io_SOURCES=Test_nersc_io.cc
+Test_nersc_io_LDADD=-lGrid
+
+
+Test_remez_SOURCES=Test_remez.cc
+Test_remez_LDADD=-lGrid
+
+
+Test_rng_SOURCES=Test_rng.cc
+Test_rng_LDADD=-lGrid
+
+
+Test_rng_fixed_SOURCES=Test_rng_fixed.cc
+Test_rng_fixed_LDADD=-lGrid
+
+
+Test_simd_SOURCES=Test_simd.cc
+Test_simd_LDADD=-lGrid
+
+
+Test_stencil_SOURCES=Test_stencil.cc
+Test_stencil_LDADD=-lGrid
+
+
+Test_wilson_cg_prec_SOURCES=Test_wilson_cg_prec.cc
+Test_wilson_cg_prec_LDADD=-lGrid
+
+
+Test_wilson_cg_schur_SOURCES=Test_wilson_cg_schur.cc
+Test_wilson_cg_schur_LDADD=-lGrid
+
+
+Test_wilson_cg_unprec_SOURCES=Test_wilson_cg_unprec.cc
+Test_wilson_cg_unprec_LDADD=-lGrid
+
+
+Test_wilson_evenodd_SOURCES=Test_wilson_evenodd.cc
+Test_wilson_evenodd_LDADD=-lGrid
+
diff --git a/tests/Makefile.am b/tests/Makefile.am
index bb3448f1..83385001 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -2,91 +2,4 @@
AM_CXXFLAGS = -I$(top_srcdir)/lib
AM_LDFLAGS = -L$(top_builddir)/lib
-#
-# Test code
-#
-bin_PROGRAMS = Test_main \
- Test_simd \
- Test_gamma \
- Test_cshift \
- Test_cshift_red_black \
- Test_stencil \
- Test_nersc_io \
- Test_rng \
- Test_remez \
- Test_rng_fixed \
- Test_wilson_evenodd \
- Test_wilson_cg_unprec \
- Test_wilson_cg_prec \
- Test_wilson_cg_schur \
- Test_dwf_even_odd\
- Test_dwf_cg_unprec\
- Test_dwf_cg_prec\
- Test_dwf_cg_schur\
- Test_many_evenodd\
- Test_many_cg
-
-
-
-Test_main_SOURCES = Test_main.cc
-Test_main_LDADD = -lGrid
-
-Test_rng_SOURCES = Test_rng.cc
-Test_rng_LDADD = -lGrid
-
-Test_rng_fixed_SOURCES = Test_rng_fixed.cc
-Test_rng_fixed_LDADD = -lGrid
-
-Test_remez_SOURCES = Test_remez.cc
-Test_remez_LDADD = -lGrid
-
-Test_nersc_io_SOURCES = Test_nersc_io.cc
-Test_nersc_io_LDADD = -lGrid
-
-Test_cshift_SOURCES = Test_cshift.cc
-Test_cshift_LDADD = -lGrid
-
-Test_cshift_red_black_SOURCES = Test_cshift_red_black.cc
-Test_cshift_red_black_LDADD = -lGrid
-
-Test_gamma_SOURCES = Test_gamma.cc
-Test_gamma_LDADD = -lGrid
-
-Test_stencil_SOURCES = Test_stencil.cc
-Test_stencil_LDADD = -lGrid
-
-Test_simd_SOURCES = Test_simd.cc
-Test_simd_LDADD = -lGrid
-
-#Test_simd_new_SOURCES = Test_simd_new.cc
-#Test_simd_new_LDADD = -lGrid
-
-Test_wilson_evenodd_SOURCES = Test_wilson_evenodd.cc
-Test_wilson_evenodd_LDADD = -lGrid
-
-Test_wilson_cg_unprec_SOURCES = Test_wilson_cg_unprec.cc
-Test_wilson_cg_unprec_LDADD = -lGrid
-
-Test_wilson_cg_prec_SOURCES = Test_wilson_cg_prec.cc
-Test_wilson_cg_prec_LDADD = -lGrid
-
-Test_wilson_cg_schur_SOURCES = Test_wilson_cg_schur.cc
-Test_wilson_cg_schur_LDADD = -lGrid
-
-Test_dwf_even_odd_SOURCES = Test_dwf_even_odd.cc
-Test_dwf_even_odd_LDADD = -lGrid
-
-Test_dwf_cg_unprec_SOURCES = Test_dwf_cg_unprec.cc
-Test_dwf_cg_unprec_LDADD = -lGrid
-
-Test_dwf_cg_prec_SOURCES = Test_dwf_cg_prec.cc
-Test_dwf_cg_prec_LDADD = -lGrid
-
-Test_dwf_cg_schur_SOURCES = Test_dwf_cg_schur.cc
-Test_dwf_cg_schur_LDADD = -lGrid
-
-Test_many_evenodd_SOURCES = Test_many_evenodd.cc
-Test_many_evenodd_LDADD = -lGrid
-
-Test_many_cg_SOURCES = Test_many_cg.cc
-Test_many_cg_LDADD = -lGrid
+include Make.inc
diff --git a/tests/Test_simd_new.cc b/tests/Test_simd_new.cc
deleted file mode 100644
index 41781304..00000000
--- a/tests/Test_simd_new.cc
+++ /dev/null
@@ -1,165 +0,0 @@
-#include <Grid.h>
-#include "simd/Grid_vector_types.h"
-#include <parallelIO/GridNerscIO.h>
-
-using namespace std;
-using namespace Grid;
-using namespace Grid::QCD;
-
-class funcPlus {
-public:
- funcPlus() {};
- template<class vec> void operator()(vec &rr,vec &i1,vec &i2) const { rr = i1+i2;}
- std::string name(void) const { return std::string("Plus"); }
-};
-class funcMinus {
-public:
- funcMinus() {};
- template<class vec> void operator()(vec &rr,vec &i1,vec &i2) const { rr = i1-i2;}
- std::string name(void) const { return std::string("Minus"); }
-};
-class funcTimes {
-public:
- funcTimes() {};
- template<class vec> void operator()(vec &rr,vec &i1,vec &i2) const { rr = i1*i2;}
- std::string name(void) const { return std::string("Times"); }
-};
-class funcConj {
-public:
- funcConj() {};
- template<class vec> void operator()(vec &rr,vec &i1,vec &i2) const { rr = conjugate(i1);}
- std::string name(void) const { return std::string("Conj"); }
-};
-class funcAdj {
-public:
- funcAdj() {};
- template<class vec> void operator()(vec &rr,vec &i1,vec &i2) const { rr = adj(i1);}
- std::string name(void) const { return std::string("Adj"); }
-};
-
-class funcTimesI {
-public:
- funcTimesI() {};
- template<class vec> void operator()(vec &rr,vec &i1,vec &i2) const { rr = timesI(i1);}
- std::string name(void) const { return std::string("timesI"); }
-};
-
-class funcTimesMinusI {
-public:
- funcTimesMinusI() {};
- template<class vec> void operator()(vec &rr,vec &i1,vec &i2) const { rr = timesMinusI(i1);}
- std::string name(void) const { return std::string("timesMinusI"); }
-};
-
-template<class scal, class vec,class functor >
-void Tester(const functor &func)
-{
- GridSerialRNG sRNG;
- sRNG.SeedRandomDevice();
-
- int Nsimd = vec::Nsimd();
-
- std::vector<scal> input1(Nsimd);
- std::vector<scal> input2(Nsimd);
- std::vector<scal> result(Nsimd);
- std::vector<scal> reference(Nsimd);
-
- std::vector<vec,alignedAllocator<vec> > buf(3);
- vec & v_input1 = buf[0];
- vec & v_input2 = buf[1];
- vec & v_result = buf[2];
-
-
- for(int i=0;i<Nsimd;i++){
- random(sRNG,input1[i]);
- random(sRNG,input2[i]);
- random(sRNG,result[i]);
- }
-
- merge<vec,scal>(v_input1,input1);
- merge<vec,scal>(v_input2,input2);
- merge<vec,scal>(v_result,result);
-
- func(v_result,v_input1,v_input2);
-
- for(int i=0;i<Nsimd;i++) {
- func(reference[i],input1[i],input2[i]);
- }
-
- extract<vec,scal>(v_result,result);
- std::cout << " " << func.name()<<std::endl;
-
- int ok=0;
- for(int i=0;i<Nsimd;i++){
- if ( abs(reference[i]-result[i])>0){
- std::cout<< "*****" << std::endl;
- std::cout<< "["<<i<<"] "<< abs(reference[i]-result[i]) << " " <<reference[i]<< " " << result[i]<<std::endl;
- ok++;
- }
- }
- if ( ok==0 ) std::cout << " OK!" <<std::endl;
-
-}
-
-
-
-int main (int argc, char ** argv)
-{
- Grid_init(&argc,&argv);
-
- std::vector<int> latt_size = GridDefaultLatt();
- std::vector<int> simd_layout = GridDefaultSimd(4,MyComplexF::Nsimd());
- std::vector<int> mpi_layout = GridDefaultMpi();
-
- GridCartesian Grid(latt_size,simd_layout,mpi_layout);
- std::vector<int> seeds({1,2,3,4});
-
- // Insist that operations on random scalars gives
- // identical results to on vectors.
-
- std::cout << "==================================="<< std::endl;
- std::cout << "Testing MyComplexF "<<std::endl;
- std::cout << "==================================="<< std::endl;
-
- Tester<ComplexF,MyComplexF>(funcTimesI());
- Tester<ComplexF,MyComplexF>(funcTimesMinusI());
- Tester<ComplexF,MyComplexF>(funcPlus());
- Tester<ComplexF,MyComplexF>(funcMinus());
- Tester<ComplexF,MyComplexF>(funcTimes());
- Tester<ComplexF,MyComplexF>(funcConj());
- Tester<ComplexF,MyComplexF>(funcAdj());
-
- std::cout << "==================================="<< std::endl;
- std::cout << "Testing MyComplexD "<<std::endl;
- std::cout << "==================================="<< std::endl;
-
-
- Tester<ComplexD,MyComplexD>(funcTimesI());
- Tester<ComplexD,MyComplexD>(funcTimesMinusI());
- Tester<ComplexD,MyComplexD>(funcPlus());
- Tester<ComplexD,MyComplexD>(funcMinus());
- Tester<ComplexD,MyComplexD>(funcTimes());
- Tester<ComplexD,MyComplexD>(funcConj());
- Tester<ComplexD,MyComplexD>(funcAdj());
-
- std::cout << "==================================="<< std::endl;
- std::cout << "Testing MyRealF "<<std::endl;
- std::cout << "==================================="<< std::endl;
-
-
- Tester<RealF,MyRealF>(funcPlus());
- Tester<RealF,MyRealF>(funcMinus());
- Tester<RealF,MyRealF>(funcTimes());
- Tester<RealF,MyRealF>(funcAdj());
-
- std::cout << "==================================="<< std::endl;
- std::cout << "Testing MyRealD "<<std::endl;
- std::cout << "==================================="<< std::endl;
-
- Tester<RealD,MyRealD>(funcPlus());
- Tester<RealD,MyRealD>(funcMinus());
- Tester<RealD,MyRealD>(funcTimes());
- Tester<RealD,MyRealD>(funcAdj());
-
- Grid_finalize();
-}
From 50bd2935278d283b53fe24091aa7bb0abb05ef97 Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 4 Jun 2015 00:00:45 +0100
Subject: [PATCH 18/20] First pass at continued fraction; solver and even odd
decomposition tests pass. Have to make ContFrac class virtual and derive end
non-abstract actions for the particular cases.
---
lib/qcd/action/fermion/CayleyFermion5D.cc | 11 +-
lib/qcd/action/fermion/CayleyFermion5D.h | 5 +-
.../fermion/ContinuedFractionFermion5D.cc | 81 +++++--
.../fermion/ContinuedFractionFermion5D.h | 34 +--
lib/qcd/action/fermion/DomainWallFermion.h | 3 +-
lib/qcd/action/fermion/MobiusFermion.h | 3 +-
.../action/fermion/MobiusZolotarevFermion.h | 4 +-
tests/InvSqrt.gnu | 0
tests/Make.inc | 30 ++-
tests/Sqrt.gnu | 2 -
tests/{Test_many_cg.cc => Test_cayley_cg.cc} | 0
...any_evenodd.cc => Test_cayley_even_odd.cc} | 0
tests/Test_contfrac_cg.cc | 147 ++++++++++++
tests/Test_contfrac_even_odd.cc | 218 ++++++++++++++++++
...son_evenodd.cc => Test_wilson_even_odd.cc} | 0
15 files changed, 490 insertions(+), 48 deletions(-)
delete mode 100644 tests/InvSqrt.gnu
delete mode 100644 tests/Sqrt.gnu
rename tests/{Test_many_cg.cc => Test_cayley_cg.cc} (100%)
rename tests/{Test_many_evenodd.cc => Test_cayley_even_odd.cc} (100%)
create mode 100644 tests/Test_contfrac_cg.cc
create mode 100644 tests/Test_contfrac_even_odd.cc
rename tests/{Test_wilson_evenodd.cc => Test_wilson_even_odd.cc} (100%)
diff --git a/lib/qcd/action/fermion/CayleyFermion5D.cc b/lib/qcd/action/fermion/CayleyFermion5D.cc
index be528e79..e47ff331 100644
--- a/lib/qcd/action/fermion/CayleyFermion5D.cc
+++ b/lib/qcd/action/fermion/CayleyFermion5D.cc
@@ -229,7 +229,14 @@ namespace QCD {
}
}
- void CayleyFermion5D::SetCoefficients(RealD scale,Approx::zolotarev_data *zdata,RealD b,RealD c)
+ // Tanh
+ void CayleyFermion5D::SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD b,RealD c)
+ {
+ SetCoefficientsZolotarev(1.0,zdata,b,c);
+
+ }
+ //Zolo
+ void CayleyFermion5D::SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata,RealD b,RealD c)
{
///////////////////////////////////////////////////////////
@@ -266,7 +273,7 @@ namespace QCD {
double bmc = b-c;
for(int i=0; i < Ls; i++){
as[i] = 1.0;
- omega[i] = ((double)zdata->gamma[i]); //NB reciprocal relative to Chroma NEF code
+ omega[i] = ((double)zdata->gamma[i])*zolo_hi; //NB reciprocal relative to Chroma NEF code
bs[i] = 0.5*(bpc/omega[i] + bmc);
cs[i] = 0.5*(bpc/omega[i] - bmc);
}
diff --git a/lib/qcd/action/fermion/CayleyFermion5D.h b/lib/qcd/action/fermion/CayleyFermion5D.h
index 57c71992..e2175d77 100644
--- a/lib/qcd/action/fermion/CayleyFermion5D.h
+++ b/lib/qcd/action/fermion/CayleyFermion5D.h
@@ -20,7 +20,7 @@ namespace Grid {
virtual void MooeeDag (const LatticeFermion &in, LatticeFermion &out);
virtual void MooeeInv (const LatticeFermion &in, LatticeFermion &out);
virtual void MooeeInvDag (const LatticeFermion &in, LatticeFermion &out);
-
+ virtual void Instantiatable(void)=0;
// protected:
RealD mass;
@@ -52,7 +52,8 @@ namespace Grid {
RealD _mass,RealD _M5);
protected:
- void SetCoefficients(RealD scale,Approx::zolotarev_data *zdata,RealD b,RealD c);
+ void SetCoefficientsZolotarev(RealD zolohi,Approx::zolotarev_data *zdata,RealD b,RealD c);
+ void SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD b,RealD c);
};
}
diff --git a/lib/qcd/action/fermion/ContinuedFractionFermion5D.cc b/lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
index c281b486..250e365f 100644
--- a/lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+++ b/lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
@@ -1,9 +1,56 @@
#include <Grid.h>
namespace Grid {
-
namespace QCD {
+ void ContinuedFractionFermion5D::SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD b,RealD c)
+ {
+ SetCoefficientsZolotarev(1.0,zdata,b,c);
+ }
+ void ContinuedFractionFermion5D::SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata,RealD b,RealD c)
+ {
+ R=(1+this->mass)/(1-this->mass);
+
+ Beta.resize(Ls);
+ cc.resize(Ls);
+ cc_d.resize(Ls);
+ sqrt_cc.resize(Ls);
+ for(int i=0; i < Ls ; i++){
+ Beta[i] = zdata -> beta[i];
+ cc[i] = 1.0/Beta[i];
+ cc_d[i]=sqrt(cc[i]);
+ }
+
+ cc_d[Ls-1]=1.0;
+ for(int i=0; i < Ls-1 ; i++){
+ sqrt_cc[i]= sqrt(cc[i]*cc[i+1]);
+ }
+ sqrt_cc[Ls-2]=sqrt(cc[Ls-2]);
+
+
+ ZoloHiInv =1.0/zolo_hi;
+ double dw_diag = (4.0-M5)*ZoloHiInv;
+
+ See.resize(Ls);
+ Aee.resize(Ls);
+ int sign=1;
+ for(int s=0;s<Ls;s++){
+ Aee[s] = sign * Beta[s] * dw_diag;
+ sign = - sign;
+ }
+ Aee[Ls-1] += R;
+
+ See[0] = Aee[0];
+ for(int s=1;s<Ls;s++){
+ See[s] = Aee[s] - 1.0/See[s-1];
+ }
+ for(int s=0;s<Ls;s++){
+ std::cout <<"s = "<<s<<" Beta "<<Beta[s]<<" Aee "<<Aee[s] <<" See "<<See[s] <<std::endl;
+ }
+ }
+
+
+
RealD ContinuedFractionFermion5D::M (const LatticeFermion &psi, LatticeFermion &chi)
{
LatticeFermion D(psi._grid);
@@ -13,13 +60,13 @@ namespace Grid {
int sign=1;
for(int s=0;s<Ls;s++){
if ( s==0 ) {
- ag5xpby_ssp(chi,cc[0]*Beta[0]*sign*scale,D,sqrt_cc[0],psi,s,s+1); // Multiplies Dw by G5 so Hw
+ ag5xpby_ssp(chi,cc[0]*Beta[0]*sign*ZoloHiInv,D,sqrt_cc[0],psi,s,s+1); // Multiplies Dw by G5 so Hw
} else if ( s==(Ls-1) ){
RealD R=(1.0+mass)/(1.0-mass);
- ag5xpby_ssp(chi,Beta[s]*scale,D,sqrt_cc[s-1],psi,s,s-1);
+ ag5xpby_ssp(chi,Beta[s]*ZoloHiInv,D,sqrt_cc[s-1],psi,s,s-1);
ag5xpby_ssp(chi,R,psi,1.0,chi,s,s);
} else {
- ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*scale,D,sqrt_cc[s],psi,s,s+1);
+ ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*ZoloHiInv,D,sqrt_cc[s],psi,s,s+1);
axpby_ssp(chi,1.0,chi,sqrt_cc[s-1],psi,s,s-1);
}
sign=-sign;
@@ -35,18 +82,22 @@ namespace Grid {
}
void ContinuedFractionFermion5D::Meooe (const LatticeFermion &psi, LatticeFermion &chi)
{
- Dhop(psi,chi,DaggerNo); // Dslash on diagonal. g5 Dslash is hermitian
+ // Apply 4d dslash
+ if ( psi.checkerboard == Odd ) {
+ DhopEO(psi,chi,DaggerNo); // Dslash on diagonal. g5 Dslash is hermitian
+ } else {
+ DhopOE(psi,chi,DaggerNo); // Dslash on diagonal. g5 Dslash is hermitian
+ }
int sign=1;
for(int s=0;s<Ls;s++){
if ( s==(Ls-1) ){
- ag5xpby_ssp(chi,Beta[s]*scale,chi,0.0,chi,s,s);
+ ag5xpby_ssp(chi,Beta[s]*ZoloHiInv,chi,0.0,chi,s,s);
} else {
- ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*scale,chi,0.0,chi,s,s);
+ ag5xpby_ssp(chi,cc[s]*Beta[s]*sign*ZoloHiInv,chi,0.0,chi,s,s);
}
sign=-sign;
- }
-
+ }
}
void ContinuedFractionFermion5D::MeooeDag (const LatticeFermion &psi, LatticeFermion &chi)
{
@@ -54,7 +105,7 @@ namespace Grid {
}
void ContinuedFractionFermion5D::Mooee (const LatticeFermion &psi, LatticeFermion &chi)
{
- double dw_diag = (4.0-this->M5)*scale;
+ double dw_diag = (4.0-M5)*ZoloHiInv;
int sign=1;
for(int s=0;s<Ls;s++){
@@ -62,7 +113,7 @@ namespace Grid {
ag5xpby_ssp(chi,cc[0]*Beta[0]*sign*dw_diag,psi,sqrt_cc[0],psi,s,s+1); // Multiplies Dw by G5 so Hw
} else if ( s==(Ls-1) ){
// Drop the CC here.
- double R=(1+this->mass)/(1-this->mass);
+ double R=(1+mass)/(1-mass);
ag5xpby_ssp(chi,Beta[s]*dw_diag,psi,sqrt_cc[s-1],psi,s,s-1);
ag5xpby_ssp(chi,R,psi,1.0,chi,s,s);
} else {
@@ -80,7 +131,7 @@ namespace Grid {
void ContinuedFractionFermion5D::MooeeInv (const LatticeFermion &psi, LatticeFermion &chi)
{
// Apply Linv
- axpby_ssp(chi,1.0/cc_d[0],psi,0.0,psi,0,0);
+ axpby_ssp(chi,1.0/cc_d[0],psi,0.0,psi,0,0);
for(int s=1;s<Ls;s++){
axpbg5y_ssp(chi,1.0/cc_d[s],psi,-1.0/See[s-1],chi,s,s-1);
}
@@ -89,7 +140,7 @@ namespace Grid {
ag5xpby_ssp(chi,1.0/See[s],chi,0.0,chi,s,s); //only appearance of See[0]
}
// Apply Uinv = (Linv)^T
- axpby_ssp(chi,1.0/cc_d[Ls-1],chi,0.0,chi,this->Ls-1,this->Ls-1);
+ axpby_ssp(chi,1.0/cc_d[Ls-1],chi,0.0,chi,Ls-1,Ls-1);
for(int s=Ls-2;s>=0;s--){
axpbg5y_ssp(chi,1.0/cc_d[s],chi,-1.0*cc_d[s+1]/See[s]/cc_d[s],chi,s,s+1);
}
@@ -112,6 +163,10 @@ namespace Grid {
FourDimGrid, FourDimRedBlackGrid,M5),
mass(_mass)
{
+ assert((Ls&0x1)==1); // Odd Ls required
+ int nrational=Ls-1;// Even rational order
+ zdata = Approx::grid_higham(1.0,nrational);// eps is ignored for higham
+ SetCoefficientsTanh(zdata,1.0,0.0);
}
}
diff --git a/lib/qcd/action/fermion/ContinuedFractionFermion5D.h b/lib/qcd/action/fermion/ContinuedFractionFermion5D.h
index 7f5c022a..99365009 100644
--- a/lib/qcd/action/fermion/ContinuedFractionFermion5D.h
+++ b/lib/qcd/action/fermion/ContinuedFractionFermion5D.h
@@ -21,20 +21,8 @@ namespace Grid {
virtual void MooeeInv (const LatticeFermion &in, LatticeFermion &out);
virtual void MooeeInvDag (const LatticeFermion &in, LatticeFermion &out);
- private:
-
- Approx::zolotarev_data *zdata;
-
- // Cont frac
- RealD mass;
- RealD R;
- RealD scale;
- std::vector<double> Beta;
- std::vector<double> cc;;
- std::vector<double> cc_d;;
- std::vector<double> sqrt_cc;
- std::vector<double> See;
- std::vector<double> Aee;
+ // virtual void Instantiatable(void)=0;
+ virtual void Instantiatable(void) {};
// Constructors
ContinuedFractionFermion5D(LatticeGaugeField &_Umu,
@@ -44,6 +32,24 @@ namespace Grid {
GridRedBlackCartesian &FourDimRedBlackGrid,
RealD _mass,RealD M5);
+ protected:
+
+ void SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD b,RealD c);
+ void SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata,RealD b,RealD c);
+
+ Approx::zolotarev_data *zdata;
+
+ // Cont frac
+ RealD mass;
+ RealD R;
+ RealD ZoloHiInv;
+ std::vector<double> Beta;
+ std::vector<double> cc;;
+ std::vector<double> cc_d;;
+ std::vector<double> sqrt_cc;
+ std::vector<double> See;
+ std::vector<double> Aee;
+
};
diff --git a/lib/qcd/action/fermion/DomainWallFermion.h b/lib/qcd/action/fermion/DomainWallFermion.h
index 3e6a9739..a25c0c3c 100644
--- a/lib/qcd/action/fermion/DomainWallFermion.h
+++ b/lib/qcd/action/fermion/DomainWallFermion.h
@@ -11,6 +11,7 @@ namespace Grid {
{
public:
+ virtual void Instantiatable(void) {};
// Constructors
DomainWallFermion(LatticeGaugeField &_Umu,
GridCartesian &FiveDimGrid,
@@ -33,7 +34,7 @@ namespace Grid {
std::cout << "DomainWallFermion with Ls="<<Ls<<std::endl;
// Call base setter
- this->CayleyFermion5D::SetCoefficients(1.0,zdata,1.0,0.0);
+ this->CayleyFermion5D::SetCoefficientsTanh(zdata,1.0,0.0);
}
diff --git a/lib/qcd/action/fermion/MobiusFermion.h b/lib/qcd/action/fermion/MobiusFermion.h
index 4c291fad..33f94089 100644
--- a/lib/qcd/action/fermion/MobiusFermion.h
+++ b/lib/qcd/action/fermion/MobiusFermion.h
@@ -11,6 +11,7 @@ namespace Grid {
{
public:
+ virtual void Instantiatable(void) {};
// Constructors
MobiusFermion(LatticeGaugeField &_Umu,
GridCartesian &FiveDimGrid,
@@ -34,7 +35,7 @@ namespace Grid {
assert(zdata->n==this->Ls);
// Call base setter
- this->CayleyFermion5D::SetCoefficients(1.0,zdata,b,c);
+ this->CayleyFermion5D::SetCoefficientsTanh(zdata,b,c);
}
diff --git a/lib/qcd/action/fermion/MobiusZolotarevFermion.h b/lib/qcd/action/fermion/MobiusZolotarevFermion.h
index 9ac795d9..1be61601 100644
--- a/lib/qcd/action/fermion/MobiusZolotarevFermion.h
+++ b/lib/qcd/action/fermion/MobiusZolotarevFermion.h
@@ -11,6 +11,7 @@ namespace Grid {
{
public:
+ virtual void Instantiatable(void) {};
// Constructors
MobiusZolotarevFermion(LatticeGaugeField &_Umu,
GridCartesian &FiveDimGrid,
@@ -34,10 +35,9 @@ namespace Grid {
assert(zdata->n==this->Ls);
std::cout << "MobiusZolotarevFermion (b="<<b<<",c="<<c<<") with Ls= "<<Ls<<" Zolotarev range ["<<lo<<","<<hi<<"]"<<std::endl;
- std::cout << "MobiusZolotarevFermion : note there is a degeneracy between (b+c) and Zolo param hi"<<std::endl;
// Call base setter
- this->CayleyFermion5D::SetCoefficients(1.0,zdata,b,c);
+ this->CayleyFermion5D::SetCoefficientsZolotarev(hi,zdata,b,c);
}
diff --git a/tests/InvSqrt.gnu b/tests/InvSqrt.gnu
deleted file mode 100644
index e69de29b..00000000
diff --git a/tests/Make.inc b/tests/Make.inc
index d592f218..b525874d 100644
--- a/tests/Make.inc
+++ b/tests/Make.inc
@@ -1,5 +1,21 @@
-bin_PROGRAMS = Test_cshift Test_cshift_red_black Test_dwf_cg_prec Test_dwf_cg_schur Test_dwf_cg_unprec Test_dwf_even_odd Test_gamma Test_main Test_many_cg Test_many_evenodd Test_nersc_io Test_remez Test_rng Test_rng_fixed Test_simd Test_stencil Test_wilson_cg_prec Test_wilson_cg_schur Test_wilson_cg_unprec Test_wilson_evenodd
+bin_PROGRAMS = Test_cayley_cg Test_cayley_even_odd Test_contfrac_cg Test_contfrac_even_odd Test_cshift Test_cshift_red_black Test_dwf_cg_prec Test_dwf_cg_schur Test_dwf_cg_unprec Test_dwf_even_odd Test_gamma Test_main Test_nersc_io Test_remez Test_rng Test_rng_fixed Test_simd Test_stencil Test_wilson_cg_prec Test_wilson_cg_schur Test_wilson_cg_unprec Test_wilson_even_odd
+
+
+Test_cayley_cg_SOURCES=Test_cayley_cg.cc
+Test_cayley_cg_LDADD=-lGrid
+
+
+Test_cayley_even_odd_SOURCES=Test_cayley_even_odd.cc
+Test_cayley_even_odd_LDADD=-lGrid
+
+
+Test_contfrac_cg_SOURCES=Test_contfrac_cg.cc
+Test_contfrac_cg_LDADD=-lGrid
+
+
+Test_contfrac_even_odd_SOURCES=Test_contfrac_even_odd.cc
+Test_contfrac_even_odd_LDADD=-lGrid
Test_cshift_SOURCES=Test_cshift.cc
@@ -34,14 +50,6 @@ Test_main_SOURCES=Test_main.cc
Test_main_LDADD=-lGrid
-Test_many_cg_SOURCES=Test_many_cg.cc
-Test_many_cg_LDADD=-lGrid
-
-
-Test_many_evenodd_SOURCES=Test_many_evenodd.cc
-Test_many_evenodd_LDADD=-lGrid
-
-
Test_nersc_io_SOURCES=Test_nersc_io.cc
Test_nersc_io_LDADD=-lGrid
@@ -78,6 +86,6 @@ Test_wilson_cg_unprec_SOURCES=Test_wilson_cg_unprec.cc
Test_wilson_cg_unprec_LDADD=-lGrid
-Test_wilson_evenodd_SOURCES=Test_wilson_evenodd.cc
-Test_wilson_evenodd_LDADD=-lGrid
+Test_wilson_even_odd_SOURCES=Test_wilson_even_odd.cc
+Test_wilson_even_odd_LDADD=-lGrid
diff --git a/tests/Sqrt.gnu b/tests/Sqrt.gnu
deleted file mode 100644
index ae56ab97..00000000
--- a/tests/Sqrt.gnu
+++ /dev/null
@@ -1,2 +0,0 @@
-f(x) = 6.81384+(-2.34645e-06/(x+0.000228091))+(-1.51593e-05/(x+0.00112084))+(-6.89254e-05/(x+0.003496))+(-0.000288983/(x+0.00954309))+(-0.00119277/(x+0.024928))+(-0.0050183/(x+0.0646627))+(-0.0226449/(x+0.171576))+(-0.123767/(x+0.491792))+(-1.1705/(x+1.78667))+(-102.992/(x+18.4866));
-f(x) = 0.14676+(0.00952992/(x+5.40933e-05))+(0.0115952/(x+0.000559699))+(0.0161824/(x+0.00203338))+(0.0243252/(x+0.00582831))+(0.0379533/(x+0.0154649))+(0.060699/(x+0.0401156))+(0.100345/(x+0.104788))+(0.178335/(x+0.286042))+(0.381586/(x+0.892189))+(1.42625/(x+4.38422));
diff --git a/tests/Test_many_cg.cc b/tests/Test_cayley_cg.cc
similarity index 100%
rename from tests/Test_many_cg.cc
rename to tests/Test_cayley_cg.cc
diff --git a/tests/Test_many_evenodd.cc b/tests/Test_cayley_even_odd.cc
similarity index 100%
rename from tests/Test_many_evenodd.cc
rename to tests/Test_cayley_even_odd.cc
diff --git a/tests/Test_contfrac_cg.cc b/tests/Test_contfrac_cg.cc
new file mode 100644
index 00000000..7fa0d6fc
--- /dev/null
+++ b/tests/Test_contfrac_cg.cc
@@ -0,0 +1,147 @@
+#include <Grid.h>
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+template<class d>
+struct scal {
+ d internal;
+};
+
+ Gamma::GammaMatrix Gmu [] = {
+ Gamma::GammaX,
+ Gamma::GammaY,
+ Gamma::GammaZ,
+ Gamma::GammaT
+ };
+
+
+template<class What>
+void TestCGinversions(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5);
+template<class What>
+void TestCGschur(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5);
+
+template<class What>
+void TestCGunprec(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5);
+
+template<class What>
+void TestCGprec(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5);
+
+int main (int argc, char ** argv)
+{
+ Grid_init(&argc,&argv);
+
+ int threads = GridThread::GetThreads();
+ std::cout << "Grid is setup to use "<<threads<<" threads"<<std::endl;
+
+ const int Ls=9;
+ GridCartesian * UGrid = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+ 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> seeds5({5,6,7,8});
+ GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
+ GridParallelRNG RNG4(UGrid); RNG4.SeedFixedIntegers(seeds4);
+
+ LatticeGaugeField Umu(UGrid); random(RNG4,Umu);
+ std::vector<LatticeColourMatrix> U(4,UGrid);
+
+ RealD mass=0.1;
+ RealD M5 =1.8;
+ std::cout <<"ContinuedFractionFermion test"<<std::endl;
+ ContinuedFractionFermion5D Dcf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+ TestCGinversions<ContinuedFractionFermion5D>(Dcf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ Grid_finalize();
+}
+template<class What>
+void TestCGinversions(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5)
+{
+ std::cout << "Testing unpreconditioned inverter"<<std::endl;
+ TestCGunprec<What>(Ddwf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,RNG4,RNG5);
+ std::cout << "Testing red black preconditioned inverter"<<std::endl;
+ TestCGprec<What>(Ddwf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,RNG4,RNG5);
+ std::cout << "Testing red black Schur inverter"<<std::endl;
+ TestCGschur<What>(Ddwf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,RNG4,RNG5);
+}
+
+template<class What>
+void TestCGunprec(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5)
+{
+ LatticeFermion src (FGrid); random(*RNG5,src);
+ LatticeFermion result(FGrid); result=zero;
+
+ HermitianOperator<What,LatticeFermion> HermOp(Ddwf);
+ ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
+ CG(HermOp,src,result);
+
+}
+template<class What>
+void TestCGprec(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5)
+{
+ LatticeFermion src (FGrid); random(*RNG5,src);
+ LatticeFermion src_o(FrbGrid);
+ LatticeFermion result_o(FrbGrid);
+ pickCheckerboard(Odd,src_o,src);
+ result_o=zero;
+
+ HermitianCheckerBoardedOperator<What,LatticeFermion> HermOpEO(Ddwf);
+ ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
+ CG(HermOpEO,src_o,result_o);
+}
+
+
+template<class What>
+void TestCGschur(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5)
+{
+ LatticeFermion src (FGrid); random(*RNG5,src);
+ LatticeFermion result(FGrid); result=zero;
+
+ ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
+ SchurRedBlackSolve<LatticeFermion> SchurSolver(CG);
+ SchurSolver(Ddwf,src,result);
+}
diff --git a/tests/Test_contfrac_even_odd.cc b/tests/Test_contfrac_even_odd.cc
new file mode 100644
index 00000000..801bd955
--- /dev/null
+++ b/tests/Test_contfrac_even_odd.cc
@@ -0,0 +1,218 @@
+#include <Grid.h>
+
+using namespace std;
+using namespace Grid;
+using namespace Grid::QCD;
+
+template<class d>
+struct scal {
+ d internal;
+};
+
+ Gamma::GammaMatrix Gmu [] = {
+ Gamma::GammaX,
+ Gamma::GammaY,
+ Gamma::GammaZ,
+ Gamma::GammaT
+ };
+
+
+template<class What>
+void TestWhat(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4, GridParallelRNG *RNG5);
+
+int main (int argc, char ** argv)
+{
+ Grid_init(&argc,&argv);
+
+ int threads = GridThread::GetThreads();
+ std::cout << "Grid is setup to use "<<threads<<" threads"<<std::endl;
+
+ const int Ls=9;
+ GridCartesian * UGrid = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexF::Nsimd()),GridDefaultMpi());
+ 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> seeds5({5,6,7,8});
+ GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
+ GridParallelRNG RNG4(UGrid); RNG4.SeedFixedIntegers(seeds4);
+
+ LatticeGaugeField Umu(UGrid); random(RNG4,Umu);
+ std::vector<LatticeColourMatrix> U(4,UGrid);
+
+ RealD mass=0.1;
+ RealD M5 =1.8;
+ std::cout <<"ContinuedFractionFermion test"<<std::endl;
+ ContinuedFractionFermion5D Dcf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+ TestWhat<ContinuedFractionFermion5D>(Dcf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ Grid_finalize();
+}
+
+template<class What>
+void TestWhat(What & Ddwf,
+ GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
+ GridCartesian * UGrid, GridRedBlackCartesian * UrbGrid,
+ RealD mass, RealD M5,
+ GridParallelRNG *RNG4,
+ GridParallelRNG *RNG5)
+{
+
+ LatticeFermion src (FGrid); random(*RNG5,src);
+ LatticeFermion phi (FGrid); random(*RNG5,phi);
+ LatticeFermion chi (FGrid); random(*RNG5,chi);
+ LatticeFermion result(FGrid); result=zero;
+ LatticeFermion ref(FGrid); ref=zero;
+ LatticeFermion tmp(FGrid); tmp=zero;
+ LatticeFermion err(FGrid); tmp=zero;
+
+ LatticeFermion src_e (FrbGrid);
+ LatticeFermion src_o (FrbGrid);
+ LatticeFermion r_e (FrbGrid);
+ LatticeFermion r_o (FrbGrid);
+ LatticeFermion r_eo (FGrid);
+ LatticeFermion r_eeoo(FGrid);
+
+ std::cout<<"=========================================================="<<std::endl;
+ std::cout<<"= Testing that Meo + Moe + Moo + Mee = Munprec "<<std::endl;
+ std::cout<<"=========================================================="<<std::endl;
+
+ pickCheckerboard(Even,src_e,src);
+ pickCheckerboard( Odd,src_o,src);
+
+ Ddwf.Meooe(src_e,r_o); std::cout<<"Applied Meo "<<norm2(r_o)<<std::endl;
+ Ddwf.Meooe(src_o,r_e); std::cout<<"Applied Moe "<<norm2(r_e)<<std::endl;
+ setCheckerboard(r_eo,r_o);
+ setCheckerboard(r_eo,r_e);
+
+ Ddwf.Mooee(src_e,r_e); std::cout<<"Applied Mee"<<norm2(r_e)<<std::endl;
+ Ddwf.Mooee(src_o,r_o); std::cout<<"Applied Moo"<<norm2(r_o)<<std::endl;
+ setCheckerboard(r_eeoo,r_e);
+ setCheckerboard(r_eeoo,r_o);
+
+ r_eo=r_eo+r_eeoo;
+ Ddwf.M(src,ref);
+
+ // std::cout << r_eo<<std::endl;
+ // std::cout << ref <<std::endl;
+
+ err= ref - r_eo;
+ std::cout << "EO norm diff "<< norm2(err)<< " "<<norm2(ref)<< " " << norm2(r_eo) <<std::endl;
+
+ LatticeComplex cerr(FGrid);
+ cerr = localInnerProduct(err,err);
+ // std::cout << cerr<<std::endl;
+
+ std::cout<<"=============================================================="<<std::endl;
+ std::cout<<"= Test Ddagger is the dagger of D by requiring "<<std::endl;
+ std::cout<<"= < phi | Deo | chi > * = < chi | Deo^dag| phi> "<<std::endl;
+ std::cout<<"=============================================================="<<std::endl;
+
+ LatticeFermion chi_e (FrbGrid);
+ LatticeFermion chi_o (FrbGrid);
+
+ LatticeFermion dchi_e (FrbGrid);
+ LatticeFermion dchi_o (FrbGrid);
+
+ LatticeFermion phi_e (FrbGrid);
+ LatticeFermion phi_o (FrbGrid);
+
+ LatticeFermion dphi_e (FrbGrid);
+ LatticeFermion dphi_o (FrbGrid);
+
+
+ pickCheckerboard(Even,chi_e,chi);
+ pickCheckerboard(Odd ,chi_o,chi);
+ pickCheckerboard(Even,phi_e,phi);
+ pickCheckerboard(Odd ,phi_o,phi);
+
+ Ddwf.Meooe(chi_e,dchi_o);
+ Ddwf.Meooe(chi_o,dchi_e);
+ Ddwf.MeooeDag(phi_e,dphi_o);
+ Ddwf.MeooeDag(phi_o,dphi_e);
+
+ ComplexD pDce = innerProduct(phi_e,dchi_e);
+ ComplexD pDco = innerProduct(phi_o,dchi_o);
+ ComplexD cDpe = innerProduct(chi_e,dphi_e);
+ ComplexD cDpo = innerProduct(chi_o,dphi_o);
+
+ std::cout <<"e "<<pDce<<" "<<cDpe <<std::endl;
+ std::cout <<"o "<<pDco<<" "<<cDpo <<std::endl;
+
+ std::cout <<"pDce - conj(cDpo) "<< pDce-conj(cDpo) <<std::endl;
+ std::cout <<"pDco - conj(cDpe) "<< pDco-conj(cDpe) <<std::endl;
+
+ std::cout<<"=============================================================="<<std::endl;
+ std::cout<<"= Test MeeInv Mee = 1 "<<std::endl;
+ std::cout<<"=============================================================="<<std::endl;
+
+ pickCheckerboard(Even,chi_e,chi);
+ pickCheckerboard(Odd ,chi_o,chi);
+
+ Ddwf.Mooee(chi_e,src_e);
+ Ddwf.MooeeInv(src_e,phi_e);
+
+ Ddwf.Mooee(chi_o,src_o);
+ Ddwf.MooeeInv(src_o,phi_o);
+
+ setCheckerboard(phi,phi_e);
+ setCheckerboard(phi,phi_o);
+
+ err = phi-chi;
+ std::cout << "norm diff "<< norm2(err)<< std::endl;
+
+ std::cout<<"=============================================================="<<std::endl;
+ std::cout<<"= Test MeeInvDag MeeDag = 1 "<<std::endl;
+ std::cout<<"=============================================================="<<std::endl;
+
+ pickCheckerboard(Even,chi_e,chi);
+ pickCheckerboard(Odd ,chi_o,chi);
+
+ Ddwf.MooeeDag(chi_e,src_e);
+ Ddwf.MooeeInvDag(src_e,phi_e);
+
+ Ddwf.MooeeDag(chi_o,src_o);
+ Ddwf.MooeeInvDag(src_o,phi_o);
+
+ setCheckerboard(phi,phi_e);
+ setCheckerboard(phi,phi_o);
+
+ err = phi-chi;
+ std::cout << "norm diff "<< norm2(err)<< std::endl;
+
+ std::cout<<"=============================================================="<<std::endl;
+ std::cout<<"= Test MpcDagMpc is Hermitian "<<std::endl;
+ std::cout<<"=============================================================="<<std::endl;
+
+ random(*RNG5,phi);
+ random(*RNG5,chi);
+ pickCheckerboard(Even,chi_e,chi);
+ pickCheckerboard(Odd ,chi_o,chi);
+ pickCheckerboard(Even,phi_e,phi);
+ pickCheckerboard(Odd ,phi_o,phi);
+ RealD t1,t2;
+
+ Ddwf.MpcDagMpc(chi_e,dchi_e,t1,t2);
+ Ddwf.MpcDagMpc(chi_o,dchi_o,t1,t2);
+
+ Ddwf.MpcDagMpc(phi_e,dphi_e,t1,t2);
+ Ddwf.MpcDagMpc(phi_o,dphi_o,t1,t2);
+
+ pDce = innerProduct(phi_e,dchi_e);
+ pDco = innerProduct(phi_o,dchi_o);
+ cDpe = innerProduct(chi_e,dphi_e);
+ cDpo = innerProduct(chi_o,dphi_o);
+
+ std::cout <<"e "<<pDce<<" "<<cDpe <<std::endl;
+ std::cout <<"o "<<pDco<<" "<<cDpo <<std::endl;
+
+ std::cout <<"pDce - conj(cDpo) "<< pDco-conj(cDpo) <<std::endl;
+ std::cout <<"pDco - conj(cDpe) "<< pDce-conj(cDpe) <<std::endl;
+
+}
diff --git a/tests/Test_wilson_evenodd.cc b/tests/Test_wilson_even_odd.cc
similarity index 100%
rename from tests/Test_wilson_evenodd.cc
rename to tests/Test_wilson_even_odd.cc
From c327019574dfa1e5086594a3660ab235016109cc Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 4 Jun 2015 00:23:16 +0100
Subject: [PATCH 19/20] Implementing the Hw kernel continued fraction 5d
overlap cases
---
lib/qcd/action/Actions.h | 2 +
.../fermion/ContinuedFractionFermion5D.cc | 9 ++--
.../fermion/ContinuedFractionFermion5D.h | 6 +--
.../OverlapWilsonContfracTanhFermion.h | 39 ++++++++++++++++
.../OverlapWilsonContfracZolotarevFermion.h | 44 +++++++++++++++++++
5 files changed, 91 insertions(+), 9 deletions(-)
create mode 100644 lib/qcd/action/fermion/OverlapWilsonContfracTanhFermion.h
create mode 100644 lib/qcd/action/fermion/OverlapWilsonContfracZolotarevFermion.h
diff --git a/lib/qcd/action/Actions.h b/lib/qcd/action/Actions.h
index 8a8c4642..b31e9136 100644
--- a/lib/qcd/action/Actions.h
+++ b/lib/qcd/action/Actions.h
@@ -55,6 +55,8 @@
// Continued fraction
//////////////////////
#include <qcd/action/fermion/ContinuedFractionFermion5D.h>
+#include <qcd/action/fermion/OverlapWilsonContfracTanhFermion.h>
+#include <qcd/action/fermion/OverlapWilsonContfracZolotarevFermion.h>
//////////////////////
// Partial fraction
diff --git a/lib/qcd/action/fermion/ContinuedFractionFermion5D.cc b/lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
index 250e365f..92f6473e 100644
--- a/lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
+++ b/lib/qcd/action/fermion/ContinuedFractionFermion5D.cc
@@ -3,11 +3,11 @@
namespace Grid {
namespace QCD {
- void ContinuedFractionFermion5D::SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD b,RealD c)
+ void ContinuedFractionFermion5D::SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD scale)
{
- SetCoefficientsZolotarev(1.0,zdata,b,c);
+ SetCoefficientsZolotarev(1.0/scale,zdata);
}
- void ContinuedFractionFermion5D::SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata,RealD b,RealD c)
+ void ContinuedFractionFermion5D::SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata)
{
R=(1+this->mass)/(1-this->mass);
@@ -164,9 +164,6 @@ namespace Grid {
mass(_mass)
{
assert((Ls&0x1)==1); // Odd Ls required
- int nrational=Ls-1;// Even rational order
- zdata = Approx::grid_higham(1.0,nrational);// eps is ignored for higham
- SetCoefficientsTanh(zdata,1.0,0.0);
}
}
diff --git a/lib/qcd/action/fermion/ContinuedFractionFermion5D.h b/lib/qcd/action/fermion/ContinuedFractionFermion5D.h
index 99365009..f363878f 100644
--- a/lib/qcd/action/fermion/ContinuedFractionFermion5D.h
+++ b/lib/qcd/action/fermion/ContinuedFractionFermion5D.h
@@ -22,7 +22,7 @@ namespace Grid {
virtual void MooeeInvDag (const LatticeFermion &in, LatticeFermion &out);
// virtual void Instantiatable(void)=0;
- virtual void Instantiatable(void) {};
+ virtual void Instantiatable(void) =0;
// Constructors
ContinuedFractionFermion5D(LatticeGaugeField &_Umu,
@@ -34,8 +34,8 @@ namespace Grid {
protected:
- void SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD b,RealD c);
- void SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata,RealD b,RealD c);
+ void SetCoefficientsTanh(Approx::zolotarev_data *zdata,RealD scale);
+ void SetCoefficientsZolotarev(RealD zolo_hi,Approx::zolotarev_data *zdata);;
Approx::zolotarev_data *zdata;
diff --git a/lib/qcd/action/fermion/OverlapWilsonContfracTanhFermion.h b/lib/qcd/action/fermion/OverlapWilsonContfracTanhFermion.h
new file mode 100644
index 00000000..4865f169
--- /dev/null
+++ b/lib/qcd/action/fermion/OverlapWilsonContfracTanhFermion.h
@@ -0,0 +1,39 @@
+#ifndef OVERLAP_WILSON_CAYLEY_TANH_FERMION_H
+#define OVERLAP_WILSON_CAYLEY_TANH_FERMION_H
+
+#include <Grid.h>
+
+namespace Grid {
+
+ namespace QCD {
+
+ class OverlapWilsonContFracTanhFermion : public ContinuedFractionFermion5D
+ {
+ public:
+
+ virtual void Instantiatable(void){};
+ // Constructors
+ OverlapWilsonContFracTanhFermion(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD _M5,
+ RealD scale) :
+
+ // b+c=scale, b-c = 0 <=> b =c = scale/2
+ ContinuedFractionFermion5D(_Umu,
+ FiveDimGrid,
+ FiveDimRedBlackGrid,
+ FourDimGrid,
+ FourDimRedBlackGrid,_mass)
+ {
+ assert((Ls&0x1)==1); // Odd Ls required
+ int nrational=Ls-1;// Even rational order
+ zdata = Approx::grid_higham(1.0,nrational);// eps is ignored for higham
+ SetCoefficientsTanh(zdata,scale);
+ }
+ };
+ }
+}
+#endif
diff --git a/lib/qcd/action/fermion/OverlapWilsonContfracZolotarevFermion.h b/lib/qcd/action/fermion/OverlapWilsonContfracZolotarevFermion.h
new file mode 100644
index 00000000..7478c062
--- /dev/null
+++ b/lib/qcd/action/fermion/OverlapWilsonContfracZolotarevFermion.h
@@ -0,0 +1,44 @@
+#ifndef OVERLAP_WILSON_CAYLEY_TANH_FERMION_H
+#define OVERLAP_WILSON_CAYLEY_TANH_FERMION_H
+
+#include <Grid.h>
+
+namespace Grid {
+
+ namespace QCD {
+
+ class OverlapWilsonContFracZolotarevFermion : public ContinuedFractionFermion5D
+ {
+ public:
+
+ virtual void Instantiatable(void){};
+ // Constructors
+ OverlapWilsonContFracZolotarevFermion(LatticeGaugeField &_Umu,
+ GridCartesian &FiveDimGrid,
+ GridRedBlackCartesian &FiveDimRedBlackGrid,
+ GridCartesian &FourDimGrid,
+ GridRedBlackCartesian &FourDimRedBlackGrid,
+ RealD _mass,RealD _M5,
+ RealD lo,RealD hi):
+
+ // b+c=scale, b-c = 0 <=> b =c = scale/2
+ ContinuedFractionFermion5D(_Umu,
+ FiveDimGrid,
+ FiveDimRedBlackGrid,
+ FourDimGrid,
+ FourDimRedBlackGrid,_mass)
+ {
+ assert((Ls&0x1)==1); // Odd Ls required
+
+ int nrational=Ls-1;// Even rational order
+ RealD eps = lo/hi;
+
+ Approx::zolotarev_data *zdata = Approx::grid_zolotarev(eps,nrational,0);
+
+ SetCoefficientsZolotarev(hi,zdata);
+
+ }
+ };
+ }
+}
+#endif
From 37aa74dfd2c6634a017925e5dbccfd3ecfbefc6f Mon Sep 17 00:00:00 2001
From: Peter Boyle <paboyle@ph.ed.ac.uk>
Date: Thu, 4 Jun 2015 06:02:00 +0100
Subject: [PATCH 20/20] CG Tests work for wilson kernel cont frac zolo and tanh
---
.../fermion/OverlapWilsonContfracTanhFermion.h | 12 ++++++------
.../fermion/OverlapWilsonContfracZolotarevFermion.h | 6 +++---
tests/Test_contfrac_cg.cc | 12 +++++++++---
tests/Test_contfrac_even_odd.cc | 11 ++++++++---
4 files changed, 26 insertions(+), 15 deletions(-)
diff --git a/lib/qcd/action/fermion/OverlapWilsonContfracTanhFermion.h b/lib/qcd/action/fermion/OverlapWilsonContfracTanhFermion.h
index 4865f169..ed0c24dc 100644
--- a/lib/qcd/action/fermion/OverlapWilsonContfracTanhFermion.h
+++ b/lib/qcd/action/fermion/OverlapWilsonContfracTanhFermion.h
@@ -1,5 +1,5 @@
-#ifndef OVERLAP_WILSON_CAYLEY_TANH_FERMION_H
-#define OVERLAP_WILSON_CAYLEY_TANH_FERMION_H
+#ifndef OVERLAP_WILSON_CONTFRAC_TANH_FERMION_H
+#define OVERLAP_WILSON_CONTFRAC_TANH_FERMION_H
#include <Grid.h>
@@ -23,10 +23,10 @@ namespace Grid {
// b+c=scale, b-c = 0 <=> b =c = scale/2
ContinuedFractionFermion5D(_Umu,
- FiveDimGrid,
- FiveDimRedBlackGrid,
- FourDimGrid,
- FourDimRedBlackGrid,_mass)
+ FiveDimGrid,
+ FiveDimRedBlackGrid,
+ FourDimGrid,
+ FourDimRedBlackGrid,_mass,_M5)
{
assert((Ls&0x1)==1); // Odd Ls required
int nrational=Ls-1;// Even rational order
diff --git a/lib/qcd/action/fermion/OverlapWilsonContfracZolotarevFermion.h b/lib/qcd/action/fermion/OverlapWilsonContfracZolotarevFermion.h
index 7478c062..caf01133 100644
--- a/lib/qcd/action/fermion/OverlapWilsonContfracZolotarevFermion.h
+++ b/lib/qcd/action/fermion/OverlapWilsonContfracZolotarevFermion.h
@@ -1,5 +1,5 @@
-#ifndef OVERLAP_WILSON_CAYLEY_TANH_FERMION_H
-#define OVERLAP_WILSON_CAYLEY_TANH_FERMION_H
+#ifndef OVERLAP_WILSON_CONTFRAC_ZOLOTAREV_FERMION_H
+#define OVERLAP_WILSON_CONTFRAC_ZOLOTAREV_FERMION_H
#include <Grid.h>
@@ -26,7 +26,7 @@ namespace Grid {
FiveDimGrid,
FiveDimRedBlackGrid,
FourDimGrid,
- FourDimRedBlackGrid,_mass)
+ FourDimRedBlackGrid,_mass,_M5)
{
assert((Ls&0x1)==1); // Odd Ls required
diff --git a/tests/Test_contfrac_cg.cc b/tests/Test_contfrac_cg.cc
index 7fa0d6fc..83475254 100644
--- a/tests/Test_contfrac_cg.cc
+++ b/tests/Test_contfrac_cg.cc
@@ -72,9 +72,15 @@ int main (int argc, char ** argv)
RealD mass=0.1;
RealD M5 =1.8;
- std::cout <<"ContinuedFractionFermion test"<<std::endl;
- ContinuedFractionFermion5D Dcf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
- TestCGinversions<ContinuedFractionFermion5D>(Dcf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+
+ std::cout <<"OverlapWilsonContFracTanhFermion test"<<std::endl;
+ OverlapWilsonContFracTanhFermion Dcf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+ TestCGinversions<OverlapWilsonContFracTanhFermion>(Dcf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ std::cout <<"OverlapWilsonContFracZolotarevFermion test"<<std::endl;
+ OverlapWilsonContFracZolotarevFermion Dcfz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,6.0);
+ TestCGinversions<OverlapWilsonContFracZolotarevFermion>(Dcfz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
Grid_finalize();
}
diff --git a/tests/Test_contfrac_even_odd.cc b/tests/Test_contfrac_even_odd.cc
index 801bd955..e13c1189 100644
--- a/tests/Test_contfrac_even_odd.cc
+++ b/tests/Test_contfrac_even_odd.cc
@@ -48,9 +48,14 @@ int main (int argc, char ** argv)
RealD mass=0.1;
RealD M5 =1.8;
- std::cout <<"ContinuedFractionFermion test"<<std::endl;
- ContinuedFractionFermion5D Dcf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
- TestWhat<ContinuedFractionFermion5D>(Dcf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ std::cout <<"OverlapWilsonContFracTanhFermion test"<<std::endl;
+ OverlapWilsonContFracTanhFermion Dcf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
+ TestWhat<OverlapWilsonContFracTanhFermion>(Dcf,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
+
+ std::cout <<"OverlapWilsonContFracZolotarevFermion test"<<std::endl;
+ OverlapWilsonContFracZolotarevFermion Dcfz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,6.0);
+ TestWhat<OverlapWilsonContFracZolotarevFermion>(Dcfz,FGrid,FrbGrid,UGrid,UrbGrid,mass,M5,&RNG4,&RNG5);
Grid_finalize();
}