portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-12 20:27:06 +01:00
Code Releases Activity

Merge pull request #210 from grid-test-organisation/feature/gpu-port-develop

Browse Source 
Cayley fermion functions for GPUs
This commit is contained in:
Peter Boyle
2019-05-18 19:06:20 +01:00
committed by GitHub
parent 4e9df9e93c 1a82533d22
commit ee6f96d85c
25 changed files with 516 additions and 242 deletions
Download Patch File Download Diff File Expand all files Collapse all files

50
benchmarks/Benchmark_mooee.cc
View File

@ -76,13 +76,20 @@ int main (int argc, char ** argv)
std::cout << GridLogMessage<< "* Benchmarking DomainWallFermionR::Dhop "<<std::endl;
std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
GridParallelRNG RNG5(FGrid);
GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
LatticeFermion src(FGrid); random(RNG5,src);
LatticeFermion result(FGrid);
DomainWallFermionR Dw(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
double t0,t1;
typedef typename DomainWallFermionR::Coeff_t Coeff_t;
Vector<Coeff_t> diag = Dw.bs;
Vector<Coeff_t> upper= Dw.cs;
Vector<Coeff_t> lower= Dw.cs;
upper[Ls-1]=-Dw.mass*upper[Ls-1];
lower[0] =-Dw.mass*lower[0];
LatticeFermion r_eo(FGrid);
LatticeFermion src_e (FrbGrid);
LatticeFermion src_o (FrbGrid);
@ -99,13 +106,13 @@ int main (int argc, char ** argv)
r_o = Zero();
#define BENCH_DW(A,in,out) \
Dw.CayleyZeroCounters(); \
Dw. A (in,out); \
#define BENCH_DW(A,...) \
Dw. A (__VA_ARGS__); \
FGrid->Barrier(); \
Dw.CayleyZeroCounters(); \
t0=usecond(); \
for(int i=0;i<ncall;i++){ \
Dw. A (in,out); \
Dw. A (__VA_ARGS__); \
} \
t1=usecond(); \
FGrid->Barrier(); \
@ -114,9 +121,9 @@ int main (int argc, char ** argv)
std::cout<<GridLogMessage << "******************"<<std::endl;
#define BENCH_ZDW(A,in,out) \
zDw.CayleyZeroCounters(); \
zDw. A (in,out); \
FGrid->Barrier(); \
zDw.CayleyZeroCounters(); \
t0=usecond(); \
for(int i=0;i<ncall;i++){ \
zDw. A (in,out); \
@ -128,9 +135,9 @@ int main (int argc, char ** argv)
std::cout<<GridLogMessage << "******************"<<std::endl;
#define BENCH_DW_SSC(A,in,out) \
Dw.CayleyZeroCounters(); \
Dw. A (in,out); \
FGrid->Barrier(); \
Dw.CayleyZeroCounters(); \
t0=usecond(); \
for(int i=0;i<ncall;i++){ \
__SSC_START ; \
@ -143,23 +150,10 @@ int main (int argc, char ** argv)
std::cout<<GridLogMessage << "Called " #A " "<< (t1-t0)/ncall<<" us"<<std::endl;\
std::cout<<GridLogMessage << "******************"<<std::endl;
#define BENCH_DW_MEO(A,in,out) \
Dw.CayleyZeroCounters(); \
Dw. A (in,out,0); \
FGrid->Barrier(); \
t0=usecond(); \
for(int i=0;i<ncall;i++){ \
Dw. A (in,out,0); \
} \
t1=usecond(); \
FGrid->Barrier(); \
Dw.CayleyReport(); \
std::cout<<GridLogMessage << "Called " #A " "<< (t1-t0)/ncall<<" us"<<std::endl;\
std::cout<<GridLogMessage << "******************"<<std::endl;
BENCH_DW_MEO(Dhop ,src,result);
BENCH_DW_MEO(DhopEO ,src_o,r_e);
BENCH_DW(Dhop ,src,result,0);
BENCH_DW(DhopEO ,src_o,r_e,0);
BENCH_DW(Meooe ,src_o,r_e);
BENCH_DW(M5D ,src_o,src_o,r_e,lower,diag,upper);
BENCH_DW(Mooee ,src_o,r_o);
BENCH_DW(MooeeInv,src_o,r_o);
@ -173,7 +167,7 @@ int main (int argc, char ** argv)
std::cout << GridLogMessage<< "* Benchmarking DomainWallFermionVec5dR::Dhop "<<std::endl;
std::cout << GridLogMessage<< "*********************************************************" <<std::endl;
GridParallelRNG RNG5(sFGrid);
GridParallelRNG RNG5(sFGrid); RNG5.SeedFixedIntegers(seeds5);
LatticeFermion src(sFGrid); random(RNG5,src);
LatticeFermion sref(sFGrid);
LatticeFermion result(sFGrid);
@ -184,7 +178,7 @@ int main (int argc, char ** argv)
RealD b=1.5;// Scale factor b+c=2, b-c=1
RealD c=0.5;
std::vector<ComplexD> gamma(Ls,std::complex<double>(1.0,0.0));
Vector<ComplexD> gamma(Ls,std::complex<double>(1.0,0.0));
ZMobiusFermionVec5dR zDw(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,gamma,b,c);
std::cout<<GridLogMessage << "Calling Dhop "<<std::endl;
@ -207,8 +201,8 @@ int main (int argc, char ** argv)
r_e = Zero();
r_o = Zero();
BENCH_DW_MEO(Dhop ,src,result);
BENCH_DW_MEO(DhopEO ,src_o,r_e);
BENCH_DW(Dhop ,src,result,0);
BENCH_DW(DhopEO ,src_o,r_e,0);
BENCH_DW_SSC(Meooe ,src_o,r_e);
BENCH_DW(Mooee ,src_o,r_o);
BENCH_DW(MooeeInv,src_o,r_o);