portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-16 23:07:05 +01:00
Code Releases Activity

Merge branch 'master' of https://github.com/paboyle/Grid

Browse Source
This commit is contained in:
Azusa Yamaguchi
2015-12-09 12:48:44 +00:00
parent 967be91692 26161addd0
commit a32a59fc43
24 changed files with 406 additions and 221 deletions
Download Patch File Download Diff File Expand all files Collapse all files

51
lib/qcd/action/fermion/WilsonFermion5D.cc
View File

@ -266,11 +266,8 @@ void WilsonFermion5D<Impl>::DhopInternal(StencilImpl & st, LebesgueOrder &lo,
if( this->HandOptDslash ) {
#pragma omp parallel for schedule(static)
for(int ss=0;ss<U._grid->oSites();ss++){
int sU=ss;
for(int s=0;s<Ls;s++){
int sU=ss;
if ( LebesgueOrder::UseLebesgueOrder ) {
sU=lo.Reorder(ss);
}
int sF = s+Ls*sU;
Kernels::DiracOptHandDhopSiteDag(st,U,comm_buf,sF,sU,in,out);
}
@ -323,52 +320,42 @@ PARALLEL_FOR_LOOP
// Counter.Report();
// }
} else if( this->HandOptDslash ) {
/*
#pragma omp parallel for
#pragma omp parallel for schedule(static)
for(int t=0;t<threads;t++){
int hyperthread = t%HT;
int core = t/HT;
int sswork, swork,soff, sU,sF;
sswork = (nwork + cores-1)/cores;
int sswork, swork,soff,ssoff, sU,sF;
GridThread::GetWork(nwork,core,sswork,ssoff,cores);
GridThread::GetWork(Ls , hyperthread, swork, soff,HT);
for(int ss=0;ss<sswork;ss++){
sU=ss+core*sswork; // max locality within an L2 slice
if ( LebesgueOrder::UseLebesgueOrder ) {
sU = lo.Reorder(sU);
sU=ss+ ssoff;
for(int s=soff;s<soff+swork;s++){
sF = s+Ls*sU;
Kernels::DiracOptHandDhopSite(st,U,comm_buf,sF,sU,in,out);
}
if ( sU < nwork ) {
for(int s=soff;s<soff+swork;s++){
sF = s+Ls*sU;
Kernels::DiracOptHandDhopSite(st,U,comm_buf,sF,sU,in,out);
}
}
}
}
/*
#pragma omp parallel for schedule(static)
for(int ss=0;ss<U._grid->oSites();ss++){
for(int s=0;s<Ls;s++){
int sU=ss;
if ( LebesgueOrder::UseLebesgueOrder ) {
sU=lo.Reorder(ss);
}
int sF = s+Ls*sU;
Kernels::DiracOptHandDhopSite(st,U,comm_buf,sF,sU,in,out);
}
}
*/
#pragma omp parallel for schedule(static)
for(int ss=0;ss<U._grid->oSites();ss++){
int sU=ss;
for(int s=0;s<Ls;s++){
int sF = s+Ls*sU;
Kernels::DiracOptHandDhopSite(st,U,comm_buf,sF,sU,in,out);
}
}
} else {
PARALLEL_FOR_LOOP
for(int ss=0;ss<U._grid->oSites();ss++){
int sU=ss;
for(int s=0;s<Ls;s++){
// int sU=lo.Reorder(ss);
int sU=ss;
int sF = s+Ls*sU;
Kernels::DiracOptDhopSite(st,U,comm_buf,sF,sU,in,out);
}

2
lib/qcd/action/fermion/WilsonKernels.h
View File

@ -29,7 +29,7 @@ namespace Grid {
std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> > &buf,
int sF,int sU,const FermionField &in, FermionField &out,int dirdisp,int gamma);
#if defined(AVX512) || defined(IMCI)
void DiracOptAsmDhopSite(CartesianStencil &st,DoubledGaugeField &U,
void DiracOptAsmDhopSite(StencilImpl &st,DoubledGaugeField &U,
std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> > &buf,
int sF,int sU,const FermionField &in, FermionField &out,uint64_t *);
#else