mirror of
https://github.com/paboyle/Grid.git
synced 2025-04-03 18:55:56 +01:00
Inline asm for KNL, KNC, Skylake Xeon
This commit is contained in:
parent
ac7d1f26ad
commit
5aafdd7e1a
293
tests/Test_zmm.cc
Normal file
293
tests/Test_zmm.cc
Normal file
@ -0,0 +1,293 @@
|
||||
#include <Grid.h>
|
||||
#include <PerfCount.h>
|
||||
#include <simd/Avx512Asm.h>
|
||||
|
||||
|
||||
using namespace Grid;
|
||||
using namespace Grid::QCD;
|
||||
void WilsonDslashAvx512(void *ptr1,void *ptr2,void *ptr3);
|
||||
void WilsonDslashAvx512F(void *ptr1,void *ptr2,void *ptr3);
|
||||
void TimesIAvx512F(void *ptr1,void *ptr3);
|
||||
void TimesIAvx512(void *ptr1,void *ptr3);
|
||||
|
||||
|
||||
|
||||
int main(int argc,char **argv)
|
||||
{
|
||||
Grid_init(&argc,&argv);
|
||||
|
||||
|
||||
std::vector<int> latt4 = GridDefaultLatt();
|
||||
const int Ls=16;
|
||||
|
||||
GridCartesian * UGrid = SpaceTimeGrid::makeFourDimGrid(latt4, GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
|
||||
GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
|
||||
GridCartesian * FGrid = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
|
||||
GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
|
||||
|
||||
std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
|
||||
std::vector<int> mpi_layout = GridDefaultMpi();
|
||||
int threads = GridThread::GetThreads();
|
||||
|
||||
std::vector<int> seeds4({1,2,3,4});
|
||||
std::vector<int> seeds5({5,6,7,8});
|
||||
|
||||
GridSerialRNG sRNG; sRNG.SeedFixedIntegers(seeds4);
|
||||
|
||||
vColourMatrixD mat;
|
||||
vHalfSpinColourVectorD vec;
|
||||
vHalfSpinColourVectorD matvec;
|
||||
vHalfSpinColourVectorD ref;
|
||||
vComplexD err;
|
||||
|
||||
random(sRNG,mat);
|
||||
random(sRNG,vec);
|
||||
|
||||
ref = mat*vec;
|
||||
|
||||
WilsonDslashAvx512((void *)&vec, (void *)&mat,(void *)&matvec);
|
||||
|
||||
ref = ref - matvec;
|
||||
err = TensorRemove(innerProduct(ref,ref));
|
||||
std::cout <<"Double SU3 x 2spin diff "<< Reduce(err)<<std::endl;
|
||||
|
||||
vColourMatrixF matF;
|
||||
vHalfSpinColourVectorF vecF;
|
||||
vHalfSpinColourVectorF matvecF;
|
||||
vHalfSpinColourVectorF refF;
|
||||
vComplexF errF;
|
||||
|
||||
random(sRNG,matF);
|
||||
random(sRNG,vecF);
|
||||
|
||||
refF = matF*vecF;
|
||||
|
||||
WilsonDslashAvx512F((void *)&vecF, (void *)&matF,(void *)&matvecF);
|
||||
|
||||
refF = refF-matvecF;
|
||||
errF = TensorRemove(innerProduct(refF,refF));
|
||||
std::cout <<"Single SU3 x 2spin diff "<< Reduce(errF)<<std::endl;
|
||||
|
||||
TimesIAvx512F((void *)&vecF,(void *)&matvecF);
|
||||
refF = timesI(vecF)-matvecF;
|
||||
errF = TensorRemove(innerProduct(refF,refF));
|
||||
std::cout <<" timesI single diff "<< Reduce(errF)<<std::endl;
|
||||
|
||||
TimesIAvx512((void *)&vec,(void *)&matvec);
|
||||
|
||||
ref = timesI(vec)-matvec;
|
||||
err = TensorRemove(innerProduct(ref,ref));
|
||||
std::cout <<" timesI double diff "<< Reduce(err)<<std::endl;
|
||||
|
||||
LatticeFermion src (FGrid);
|
||||
LatticeFermion srce(FrbGrid);
|
||||
|
||||
LatticeFermion resulto(FrbGrid); resulto=zero;
|
||||
LatticeFermion resulta(FrbGrid); resulta=zero;
|
||||
LatticeFermion diff(FrbGrid);
|
||||
LatticeGaugeField Umu(UGrid);
|
||||
|
||||
#if 1
|
||||
GridParallelRNG RNG4(UGrid); RNG4.SeedFixedIntegers(seeds4);
|
||||
GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
|
||||
random(RNG5,src);
|
||||
random(RNG4,Umu);
|
||||
#else
|
||||
int mmu=3;
|
||||
std::vector<LatticeColourMatrix> U(4,UGrid);
|
||||
for(int mu=0;mu<Nd;mu++){
|
||||
U[mu] = PeekIndex<LorentzIndex>(Umu,mu);
|
||||
if ( mu!=mmu ) U[mu] = zero;
|
||||
if ( mu==mmu ) U[mu] = 1.0;
|
||||
PokeIndex<LorentzIndex>(Umu,U[mu],mu);
|
||||
}
|
||||
#endif
|
||||
pickCheckerboard(Even,srce,src);
|
||||
|
||||
RealD mass=0.1;
|
||||
RealD M5 =1.8;
|
||||
DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
|
||||
|
||||
std::cout<<GridLogMessage << "Calling Dw"<<std::endl;
|
||||
int ncall=50;
|
||||
double t0=usecond();
|
||||
for(int i=0;i<ncall;i++){
|
||||
Dw.DhopOE(srce,resulto,0);
|
||||
}
|
||||
double t1=usecond();
|
||||
|
||||
double volume=Ls; for(int mu=0;mu<Nd;mu++) volume=volume*latt4[mu];
|
||||
double flops=1344*volume/2;
|
||||
|
||||
std::cout<<GridLogMessage << "Called Dw"<<std::endl;
|
||||
std::cout<<GridLogMessage << "norm result "<< norm2(resulto)<<std::endl;
|
||||
std::cout<<GridLogMessage << "mflop/s = "<< flops*ncall/(t1-t0)<<std::endl;
|
||||
|
||||
QCD::WilsonFermion5DStatic::AsmOptDslash=1;
|
||||
t0=usecond();
|
||||
for(int i=0;i<ncall;i++){
|
||||
Dw.DhopOE(srce,resulta,0);
|
||||
}
|
||||
t1=usecond();
|
||||
|
||||
|
||||
for(int i=0;i< PerformanceCounter::NumTypes(); i++ ){
|
||||
Dw.DhopOE(srce,resulta,0);
|
||||
PerformanceCounter Counter(i);
|
||||
Counter.Start();
|
||||
Dw.DhopOE(srce,resulta,0);
|
||||
Counter.Stop();
|
||||
Counter.Report();
|
||||
}
|
||||
resulta = (-0.5) * resulta;
|
||||
|
||||
std::cout<<GridLogMessage << "Called Asm Dw"<<std::endl;
|
||||
std::cout<<GridLogMessage << "norm result "<< norm2(resulta)<<std::endl;
|
||||
std::cout<<GridLogMessage << "mflop/s = "<< flops*ncall/(t1-t0)<<std::endl;
|
||||
diff = resulto-resulta;
|
||||
std::cout<<GridLogMessage << "diff "<< norm2(diff)<<std::endl;
|
||||
|
||||
}
|
||||
|
||||
#undef VLOAD
|
||||
#undef VSTORE
|
||||
#undef VMUL
|
||||
#undef VMADD
|
||||
#undef ZEND1
|
||||
#undef ZEND2
|
||||
#undef ZLOAD
|
||||
#undef ZMUL
|
||||
#undef ZMADD
|
||||
|
||||
#define VZERO(A) VZEROd(A)
|
||||
#define VTIMESI(A,B,C) VTIMESId(A,B,C)
|
||||
#define VTIMESMINUSI(A,B,C) VTIMESMINUSId(A,B,C)
|
||||
|
||||
#define VLOAD(OFF,PTR,DEST) VLOADd(OFF,PTR,DEST)
|
||||
#define VSTORE(OFF,PTR,SRC) VSTOREd(OFF,PTR,SRC)
|
||||
#define VMUL(Uri,Uir,Chi,UChi,Z) VMULd(Uri,Uir,Chi,UChi,Z)
|
||||
#define VMADD(Uri,Uir,Chi,UChi,Z) VMADDd(Uri,Uir,Chi,UChi,Z)
|
||||
#define ZEND1(A,B,C) ZEND1d(A,B,C)
|
||||
#define ZEND2(A,B,C) ZEND2d(A,B,C)
|
||||
#define ZLOAD(A,B,C,D) ZLOADd(A,B,C,D)
|
||||
#define ZMUL(A,B,C,D,E) ZMULd(A,B,C,D,E)
|
||||
#define ZMADD(A,B,C,D,E) ZMADDd(A,B,C,D,E)
|
||||
#define ZMULMEM2SP(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr) ZMULMEM2SPd(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)
|
||||
#define ZMADDMEM2SP(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr) ZMADDMEM2SPd(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)
|
||||
|
||||
#define zz Z0
|
||||
|
||||
void TimesIAvx512(void *ptr1,void *ptr3)
|
||||
{
|
||||
__asm__ ("mov $0xAAAA, %%eax " : : :"%eax");
|
||||
__asm__ ("kmov %%eax, %%k6 " : : :);
|
||||
__asm__ ("knot %%k6, %%k7 " : : :);
|
||||
|
||||
|
||||
MASK_REGS;
|
||||
|
||||
LOAD_CHI(ptr1);
|
||||
|
||||
__asm__ (
|
||||
VZERO(zz)
|
||||
VTIMESI(Chi_00,UChi_00,zz)
|
||||
VTIMESI(Chi_01,UChi_01,zz)
|
||||
VTIMESI(Chi_02,UChi_02,zz)
|
||||
VTIMESI(Chi_10,UChi_10,zz)
|
||||
VTIMESI(Chi_11,UChi_11,zz)
|
||||
VTIMESI(Chi_12,UChi_12,zz)
|
||||
);
|
||||
|
||||
SAVE_UCHI(ptr3);
|
||||
}
|
||||
|
||||
void WilsonDslashAvx512(void *ptr1,void *ptr2,void *ptr3)
|
||||
{
|
||||
int return_address;
|
||||
// prototype computed goto to eliminate ABI save restore on call/return in
|
||||
// generated assembly.
|
||||
static void * table[] = { &&save, &&mult };
|
||||
|
||||
MASK_REGS;
|
||||
|
||||
LOAD_CHI(ptr1);
|
||||
|
||||
return_address = 0;
|
||||
goto mult;
|
||||
|
||||
save:
|
||||
SAVE_UCHI(ptr3);
|
||||
return;
|
||||
|
||||
mult:
|
||||
MULT_2SPIN(ptr2);
|
||||
goto *table[return_address];
|
||||
|
||||
}
|
||||
|
||||
#undef VLOAD
|
||||
#undef VSTORE
|
||||
#undef VMUL
|
||||
#undef VMADD
|
||||
#undef ZEND1
|
||||
#undef ZEND2
|
||||
#undef ZLOAD
|
||||
#undef ZMUL
|
||||
#undef ZMADD
|
||||
#undef VZERO
|
||||
#undef VTIMESI
|
||||
#undef VTIMESI0
|
||||
#undef VTIMESI1
|
||||
#undef VTIMESI2
|
||||
#undef VTIMESMINUSI
|
||||
#undef ZMULMEM2SP
|
||||
#undef ZMADDMEM2SP
|
||||
#define VZERO(A) VZEROf(A)
|
||||
#define VMOV(A,B) VMOVf(A,B)
|
||||
#define VADD(A,B,C) VADDf(A,B,C)
|
||||
#define VSUB(A,B,C) VSUBf(A,B,C)
|
||||
#define VTIMESI(A,B,C) VTIMESIf(A,B,C)
|
||||
#define VTIMESMINUSI(A,B,C) VTIMESMINUSIf(A,B,C)
|
||||
|
||||
#define VLOAD(OFF,PTR,DEST) VLOADf(OFF,PTR,DEST)
|
||||
#define VSTORE(OFF,PTR,SRC) VSTOREf(OFF,PTR,SRC)
|
||||
#define VMUL(Uri,Uir,Chi,UChi,Z) VMULf(Uri,Uir,Chi,UChi,Z)
|
||||
#define VMADD(Uri,Uir,Chi,UChi,Z) VMADDf(Uri,Uir,Chi,UChi,Z)
|
||||
#define ZEND1(A,B,C) ZEND1f(A,B,C)
|
||||
#define ZEND2(A,B,C) ZEND2f(A,B,C)
|
||||
#define ZLOAD(A,B,C,D) ZLOADf(A,B,C,D)
|
||||
#define ZMUL(A,B,C,D,E) ZMULf(A,B,C,D,E)
|
||||
#define ZMADD(A,B,C,D,E) ZMADDf(A,B,C,D,E)
|
||||
#define ZMULMEM2SP(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr) ZMULMEM2SPf(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)
|
||||
#define ZMADDMEM2SP(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr) ZMADDMEM2SPf(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)
|
||||
|
||||
void TimesIAvx512F(void *ptr1,void *ptr3)
|
||||
{
|
||||
MASK_REGS;
|
||||
|
||||
LOAD_CHI(ptr1);
|
||||
__asm__ (
|
||||
VZERO(zz)
|
||||
VTIMESI(Chi_00,UChi_00,zz)
|
||||
VTIMESI(Chi_01,UChi_01,zz)
|
||||
VTIMESI(Chi_02,UChi_02,zz)
|
||||
VTIMESI(Chi_10,UChi_10,zz)
|
||||
VTIMESI(Chi_11,UChi_11,zz)
|
||||
VTIMESI(Chi_12,UChi_12,zz)
|
||||
);
|
||||
SAVE_UCHI(ptr3);
|
||||
}
|
||||
|
||||
void WilsonDslashAvx512F(void *ptr1,void *ptr2,void *ptr3)
|
||||
{
|
||||
MASK_REGS;
|
||||
|
||||
LOAD_CHI(ptr1);
|
||||
|
||||
MULT_2SPIN(ptr2);
|
||||
|
||||
SAVE_UCHI(ptr3);
|
||||
|
||||
return;
|
||||
}
|
||||
|
Loading…
x
Reference in New Issue
Block a user