portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-09-20 01:05:38 +01:00
Code Releases Activity

Added su3 matrix benchmark.

Browse Source
This commit is contained in:
Peter Boyle 2015-05-15 14:41:19 +01:00
parent 8e1b5dda4b
commit b4b70702fd
2 changed files with 143 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

139
benchmarks/Grid_su3.cc Normal file
View File

@ -0,0 +1,139 @@
#include <Grid.h>
using namespace std;
using namespace Grid;
using namespace Grid::QCD;
int main (int argc, char ** argv)
{
Grid_init(&argc,&argv);
int Nloop=1000;
std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
std::vector<int> mpi_layout = GridDefaultMpi();
std::cout << "===================================================================================================="<<std::endl;
std::cout << "= Benchmarking SU3xSU3 x= x*y"<<std::endl;
std::cout << "===================================================================================================="<<std::endl;
std::cout << " L "<<"\t\t"<<"bytes"<<"\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
std::cout << "----------------------------------------------------------"<<std::endl;
for(int lat=2;lat<=24;lat+=2){
std::vector<int> latt_size ({lat,lat,lat,lat});
GridCartesian Grid(latt_size,simd_layout,mpi_layout);
// GridParallelRNG pRNG(&Grid); pRNG.SeedRandomDevice();
LatticeColourMatrix z(&Grid);// random(pRNG,z);
LatticeColourMatrix x(&Grid);// random(pRNG,x);
LatticeColourMatrix y(&Grid);// random(pRNG,y);
double start=usecond();
for(int i=0;i<Nloop;i++){
x=x*y;
}
double stop=usecond();
double time = (stop-start)/Nloop*1000.0;
double bytes=3.0*lat*lat*lat*lat*Nc*Nc*sizeof(Complex);
double footprint=2.0*lat*lat*lat*lat*Nc*Nc*sizeof(Complex);
double flops=Nc*Nc*(6.0+8.0+8.0)*lat*lat*lat*lat;
std::cout<<std::setprecision(2) << lat<<"\t\t"<<footprint<<"\t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
}
std::cout << "===================================================================================================="<<std::endl;
std::cout << "= Benchmarking SU3xSU3 z= x*y"<<std::endl;
std::cout << "===================================================================================================="<<std::endl;
std::cout << " L "<<"\t\t"<<"bytes"<<"\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
std::cout << "----------------------------------------------------------"<<std::endl;
for(int lat=2;lat<=24;lat+=2){
std::vector<int> latt_size ({lat,lat,lat,lat});
GridCartesian Grid(latt_size,simd_layout,mpi_layout);
// GridParallelRNG pRNG(&Grid); pRNG.SeedRandomDevice();
LatticeColourMatrix z(&Grid); //random(pRNG,z);
LatticeColourMatrix x(&Grid); //random(pRNG,x);
LatticeColourMatrix y(&Grid); //random(pRNG,y);
double start=usecond();
for(int i=0;i<Nloop;i++){
z=x*y;
}
double stop=usecond();
double time = (stop-start)/Nloop*1000.0;
double bytes=3*lat*lat*lat*lat*Nc*Nc*sizeof(Complex);
double flops=Nc*Nc*(6+8+8)*lat*lat*lat*lat;
std::cout<<std::setprecision(2) << lat<<"\t\t"<<bytes<<"\t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
}
std::cout << "===================================================================================================="<<std::endl;
std::cout << "= Benchmarking SU3xSU3 mult(z,x,y)"<<std::endl;
std::cout << "===================================================================================================="<<std::endl;
std::cout << " L "<<"\t\t"<<"bytes"<<"\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
std::cout << "----------------------------------------------------------"<<std::endl;
for(int lat=2;lat<=24;lat+=2){
std::vector<int> latt_size ({lat,lat,lat,lat});
GridCartesian Grid(latt_size,simd_layout,mpi_layout);
// GridParallelRNG pRNG(&Grid); pRNG.SeedRandomDevice();
LatticeColourMatrix z(&Grid); //random(pRNG,z);
LatticeColourMatrix x(&Grid); //random(pRNG,x);
LatticeColourMatrix y(&Grid); //random(pRNG,y);
double start=usecond();
for(int i=0;i<Nloop;i++){
mult(z,x,y);
}
double stop=usecond();
double time = (stop-start)/Nloop*1000.0;
double bytes=3*lat*lat*lat*lat*Nc*Nc*sizeof(Complex);
double flops=Nc*Nc*(6+8+8)*lat*lat*lat*lat;
std::cout<<std::setprecision(2) << lat<<"\t\t"<<bytes<<"\t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
}
std::cout << "===================================================================================================="<<std::endl;
std::cout << "= Benchmarking SU3xSU3 mac(z,x,y)"<<std::endl;
std::cout << "===================================================================================================="<<std::endl;
std::cout << " L "<<"\t\t"<<"bytes"<<"\t\t"<<"GB/s\t\t GFlop/s"<<std::endl;
std::cout << "----------------------------------------------------------"<<std::endl;
for(int lat=2;lat<=24;lat+=2){
std::vector<int> latt_size ({lat,lat,lat,lat});
GridCartesian Grid(latt_size,simd_layout,mpi_layout);
// GridParallelRNG pRNG(&Grid); pRNG.SeedRandomDevice();
LatticeColourMatrix z(&Grid); //random(pRNG,z);
LatticeColourMatrix x(&Grid); //random(pRNG,x);
LatticeColourMatrix y(&Grid); //random(pRNG,y);
double start=usecond();
for(int i=0;i<Nloop;i++){
mac(z,x,y);
}
double stop=usecond();
double time = (stop-start)/Nloop*1000.0;
double bytes=3*lat*lat*lat*lat*Nc*Nc*sizeof(Complex);
double flops=Nc*Nc*(8+8+8)*lat*lat*lat*lat;
std::cout<<std::setprecision(2) << lat<<"\t\t"<<bytes<<"\t\t"<<bytes/time<<"\t\t" << flops/time<<std::endl;
}
Grid_finalize();
}

5
benchmarks/Makefile.am
View File

@ -5,7 +5,7 @@ AM_LDFLAGS = -L$(top_builddir)/lib
#
# Test code
#
bin_PROGRAMS = Grid_wilson Grid_comms Grid_memory_bandwidth
bin_PROGRAMS = Grid_wilson Grid_comms Grid_memory_bandwidth Grid_su3
Grid_wilson_SOURCES = Grid_wilson.cc
@ -14,6 +14,9 @@ Grid_wilson_LDADD = -lGrid
Grid_comms_SOURCES = Grid_comms.cc
Grid_comms_LDADD = -lGrid
Grid_su3_SOURCES = Grid_su3.cc
Grid_su3_LDADD = -lGrid
Grid_memory_bandwidth_SOURCES = Grid_memory_bandwidth.cc
Grid_memory_bandwidth_LDADD = -lGrid