portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-04-04 19:25:56 +01:00
Code Releases Activity

Ugly hack to speed up compile on GPU; we don't use the hand kernels on GPU anyway so why compile

Browse Source
This commit is contained in:
paboyle 2018-06-13 20:35:28 +01:00
parent b710fec6ea
commit 73bb2d5128
1 changed files with 16 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
lib/qcd/action/fermion/WilsonKernelsHand.cc
View File

@ -576,6 +576,7 @@ template<class Impl> void accelerator
WilsonKernels<Impl>::HandDhopSite(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf, WilsonKernels<Impl>::HandDhopSite(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
int ss,int sU,const FermionFieldView &in, FermionFieldView &out) int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
{ {
#ifndef GRID_NVCC
// T==0, Z==1, Y==2, Z==3 expect 1,2,2,2 simd layout etc... // T==0, Z==1, Y==2, Z==3 expect 1,2,2,2 simd layout etc...
typedef typename Simd::scalar_type S; typedef typename Simd::scalar_type S;
typedef typename Simd::vector_type V; typedef typename Simd::vector_type V;
@ -597,12 +598,14 @@ WilsonKernels<Impl>::HandDhopSite(StencilView &st, DoubledGaugeFieldView &U,Site
HAND_RESULT(ss,F) HAND_RESULT(ss,F)
HAND_DOP_SITE(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN); HAND_DOP_SITE(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN);
#endif
} }
template<class Impl> accelerator template<class Impl> accelerator
void WilsonKernels<Impl>::HandDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf, void WilsonKernels<Impl>::HandDhopSiteDag(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
int ss,int sU,const FermionFieldView &in, FermionFieldView &out) int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
{ {
#ifndef GRID_NVCC
typedef typename Simd::scalar_type S; typedef typename Simd::scalar_type S;
typedef typename Simd::vector_type V; typedef typename Simd::vector_type V;
@ -623,12 +626,14 @@ void WilsonKernels<Impl>::HandDhopSiteDag(StencilView &st, DoubledGaugeFieldView
HAND_RESULT(ss,F) HAND_RESULT(ss,F)
HAND_DOP_SITE_DAG(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN); HAND_DOP_SITE_DAG(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN);
#endif
} }
template<class Impl> void accelerator template<class Impl> void accelerator
WilsonKernels<Impl>::HandDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf, WilsonKernels<Impl>::HandDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
int ss,int sU,const FermionFieldView &in, FermionFieldView &out) int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
{ {
#ifndef GRID_NVCC
// T==0, Z==1, Y==2, Z==3 expect 1,2,2,2 simd layout etc... // T==0, Z==1, Y==2, Z==3 expect 1,2,2,2 simd layout etc...
typedef typename Simd::scalar_type S; typedef typename Simd::scalar_type S;
typedef typename Simd::vector_type V; typedef typename Simd::vector_type V;
@ -651,12 +656,14 @@ WilsonKernels<Impl>::HandDhopSiteInt(StencilView &st, DoubledGaugeFieldView &U,S
HAND_RESULT(ss,F) HAND_RESULT(ss,F)
HAND_DOP_SITE_INT(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN); HAND_DOP_SITE_INT(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN);
#endif
} }
template<class Impl> accelerator template<class Impl> accelerator
void WilsonKernels<Impl>::HandDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf, void WilsonKernels<Impl>::HandDhopSiteDagInt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
int ss,int sU,const FermionFieldView &in, FermionFieldView &out) int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
{ {
#ifndef GRID_NVCC
typedef typename Simd::scalar_type S; typedef typename Simd::scalar_type S;
typedef typename Simd::vector_type V; typedef typename Simd::vector_type V;
@ -678,12 +685,14 @@ void WilsonKernels<Impl>::HandDhopSiteDagInt(StencilView &st, DoubledGaugeFieldV
HAND_RESULT(ss,F) HAND_RESULT(ss,F)
HAND_DOP_SITE_DAG_INT(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN); HAND_DOP_SITE_DAG_INT(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN);
#endif
} }
template<class Impl> void accelerator template<class Impl> void accelerator
WilsonKernels<Impl>::HandDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf, WilsonKernels<Impl>::HandDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
int ss,int sU,const FermionFieldView &in, FermionFieldView &out) int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
{ {
#ifndef GRID_NVCC
// T==0, Z==1, Y==2, Z==3 expect 1,2,2,2 simd layout etc... // T==0, Z==1, Y==2, Z==3 expect 1,2,2,2 simd layout etc...
typedef typename Simd::scalar_type S; typedef typename Simd::scalar_type S;
typedef typename Simd::vector_type V; typedef typename Simd::vector_type V;
@ -708,12 +717,14 @@ WilsonKernels<Impl>::HandDhopSiteExt(StencilView &st, DoubledGaugeFieldView &U,S
HAND_DOP_SITE_EXT(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN); HAND_DOP_SITE_EXT(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN);
perm++; perm++;
#endif
} }
template<class Impl> template<class Impl>
accelerator void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf, accelerator void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf,
int ss,int sU,const FermionFieldView &in, FermionFieldView &out) int ss,int sU,const FermionFieldView &in, FermionFieldView &out)
{ {
#ifndef GRID_NVCC
typedef typename Simd::scalar_type S; typedef typename Simd::scalar_type S;
typedef typename Simd::vector_type V; typedef typename Simd::vector_type V;
@ -737,6 +748,7 @@ accelerator void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilView &st, Double
HAND_DOP_SITE_DAG_EXT(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN); HAND_DOP_SITE_DAG_EXT(, LOAD_CHI,LOAD_CHIMU,MULT_2SPIN);
perm++; perm++;
#endif
} }
//////////////////////////////////////////////// ////////////////////////////////////////////////
@ -793,7 +805,9 @@ accelerator void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilView &st, Double
FermionFieldView &out){ assert(0); } \ FermionFieldView &out){ assert(0); } \
#ifdef GRID_NVCC
#define HAND_SPECIALISE_GPARITY(IMPL) HAND_SPECIALISE_EMPTY(IMPL)
#else
#define HAND_SPECIALISE_GPARITY(IMPL) \ #define HAND_SPECIALISE_GPARITY(IMPL) \
template<> void \ template<> void \
WilsonKernels<IMPL>::HandDhopSite(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf, \ WilsonKernels<IMPL>::HandDhopSite(StencilView &st, DoubledGaugeFieldView &U,SiteHalfSpinor *buf, \
@ -893,7 +907,7 @@ accelerator void WilsonKernels<Impl>::HandDhopSiteDagExt(StencilView &st, Double
nmu = 0; \ nmu = 0; \
HAND_DOP_SITE_DAG_EXT(1, LOAD_CHI_GPARITY,LOAD_CHIMU_GPARITY,MULT_2SPIN_GPARITY); \ HAND_DOP_SITE_DAG_EXT(1, LOAD_CHI_GPARITY,LOAD_CHIMU_GPARITY,MULT_2SPIN_GPARITY); \
} }
#endif
HAND_SPECIALISE_GPARITY(GparityWilsonImplF); HAND_SPECIALISE_GPARITY(GparityWilsonImplF);
HAND_SPECIALISE_GPARITY(GparityWilsonImplD); HAND_SPECIALISE_GPARITY(GparityWilsonImplD);