portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-15 14:27:06 +01:00
Code Releases Activity

Improve the matching of stencil coarsening

Browse Source
This commit is contained in:
Peter Boyle
2020-09-08 15:36:58 -04:00
parent 729882827c
commit 3064c9a6e2
12 changed files with 62 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
benchmarks/Benchmark_su3.cc
View File

@ -191,7 +191,7 @@ int main (int argc, char ** argv)
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
std::cout<<GridLogMessage << "= Benchmarking SU3xSU3 CovShiftForward(z,x,y)"<<std::endl;
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
std::cout<<GridLogMessage << " L "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
std::cout<<GridLogMessage << " L "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GB/s (incl Cshift)\t\t GFlop/s"<<std::endl;
std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
for(int lat=LMIN;lat<=LMAX;lat+=LADD){
@ -216,15 +216,16 @@ int main (int argc, char ** argv)
double bytes=3*vol*Nc*Nc*sizeof(Complex);
double ncbytes=5*vol*Nc*Nc*sizeof(Complex);
double flops=Nc*Nc*(6+8+8)*vol;
std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<" \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<" \t\t"<<bytes/time<<"\t\t"<<ncbytes/time<<"\t\t" << flops/time<<std::endl;
}
}
#if 1
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
std::cout<<GridLogMessage << "= Benchmarking SU3xSU3 z= x * Cshift(y)"<<std::endl;
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
std::cout<<GridLogMessage << " L "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
std::cout<<GridLogMessage << " L "<<"\t\t"<<"bytes"<<"\t\t\t"<<"GB/s\t\t GB/s (incl Cshift)\t\t GFlop/s"<<std::endl;
std::cout<<GridLogMessage << "----------------------------------------------------------"<<std::endl;
for(int lat=LMIN;lat<=LMAX;lat+=LADD){
@ -258,10 +259,11 @@ int main (int argc, char ** argv)
tmult = tmult /Nloop;
double bytes=3*vol*Nc*Nc*sizeof(Complex);
double ncbytes=5*vol*Nc*Nc*sizeof(Complex);
double flops=Nc*Nc*(6+8+8)*vol;
std::cout<<GridLogMessage<<std::setprecision(3) << "total us "<<time<<" shift "<<tshift <<" mult "<<tmult<<std::endl;
time = time * 1000; // convert to NS for GB/s
std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<" \t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
std::cout<<GridLogMessage<<std::setprecision(3) << lat<<"\t\t"<<bytes<<" \t\t"<<bytes/time<<"\t\t" <<ncbytes/time<<"\t\t" << flops/time<<std::endl;
}
}
#endif