portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-07-26 09:17:08 +01:00
Code Releases Activity

Merge branch 'develop' into feature/hirep

Browse Source
This commit is contained in:
Guido Cossu
2016-09-01 12:59:53 +01:00
parent f45ef8d114 8c89391c02
commit 0fd179fb33
75 changed files with 16078 additions and 2795 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
benchmarks/Benchmark_comms.cc
View File

@@ -194,7 +194,7 @@ int main (int argc, char ** argv)
}
}
#if 0
std::cout<<GridLogMessage << "===================================================================================================="<<std::endl;
std::cout<<GridLogMessage << "= Benchmarking sequential persistent halo exchange in "<<nmu<<" dimensions"<<std::endl;
@@ -315,7 +315,7 @@ int main (int argc, char ** argv)
}
}
#endif
Grid_finalize();
}

2
benchmarks/Benchmark_dwf_sweep.cc
View File

@@ -61,6 +61,8 @@ int main (int argc, char ** argv)
QCD::WilsonKernelsStatic::AsmOpt=0;
}
std::cout<<GridLogMessage << "=========================================================================="<<std::endl;
std::cout<<GridLogMessage << "= Benchmarking DWF"<<std::endl;
std::cout<<GridLogMessage << "=========================================================================="<<std::endl;
std::cout<<GridLogMessage << "Volume \t\t\tProcs \t Dw \t eoDw \t sDw \t eosDw (Mflop/s) "<<std::endl;
std::cout<<GridLogMessage << "=========================================================================="<<std::endl;

117
benchmarks/Benchmark_wilson_sweep.cc Normal file
View File

@@ -0,0 +1,117 @@
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
Source file: ./benchmarks/Benchmark_wilson.cc
Copyright (C) 2015
Author: Peter Boyle <paboyle@ph.ed.ac.uk>
Author: paboyle <paboyle@ph.ed.ac.uk>
Author: Richard Rollins <rprollins@users.noreply.github.com>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
#include <Grid/Grid.h>
using namespace std;
using namespace Grid;
using namespace Grid::QCD;
template<class d>
struct scal {
d internal;
};
Gamma::GammaMatrix Gmu [] = {
Gamma::GammaX,
Gamma::GammaY,
Gamma::GammaZ,
Gamma::GammaT
};
bool overlapComms = false;
void bench_wilson (
LatticeFermion & src,
LatticeFermion & result,
WilsonFermionR & Dw,
double const volume,
int const dag );
int main (int argc, char ** argv)
{
Grid_init(&argc,&argv);
if( GridCmdOptionExists(argv,argv+argc,"--asynch") ){ overlapComms = true; }
typename WilsonFermionR::ImplParams params;
params.overlapCommsCompute = overlapComms;
std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplex::Nsimd());
std::vector<int> mpi_layout = GridDefaultMpi();
std::vector<int> seeds({1,2,3,4});
RealD mass = 0.1;
std::cout<<GridLogMessage << "============================================================================="<< std::endl;
std::cout<<GridLogMessage << "= Benchmarking Wilson" << std::endl;
std::cout<<GridLogMessage << "============================================================================="<< std::endl;
std::cout<<GridLogMessage << "Volume\t\t\tWilson/MFLOPs\tWilsonDag/MFLOPs" << std::endl;
std::cout<<GridLogMessage << "============================================================================="<< std::endl;
int Lmax = 32;
int dmin = 0;
if ( getenv("LMAX") ) Lmax=atoi(getenv("LMAX"));
if ( getenv("DMIN") ) dmin=atoi(getenv("DMIN"));
for (int L=8; L<=Lmax; L*=2)
{
std::vector<int> latt_size = std::vector<int>(4,L);
for(int d=4; d>dmin; d--)
{
if ( d<=3 ) { latt_size[d] *= 2; }
std::cout << GridLogMessage;
std::copy( latt_size.begin(), --latt_size.end(), std::ostream_iterator<int>( std::cout, std::string("x").c_str() ) );
std::cout << latt_size.back() << "\t\t";
GridCartesian Grid(latt_size,simd_layout,mpi_layout);
GridRedBlackCartesian RBGrid(latt_size,simd_layout,mpi_layout);
GridParallelRNG pRNG(&Grid); pRNG.SeedFixedIntegers(seeds);
LatticeGaugeField Umu(&Grid); random(pRNG,Umu);
LatticeFermion src(&Grid); random(pRNG,src);
LatticeFermion result(&Grid); result=zero;
double volume = std::accumulate(latt_size.begin(),latt_size.end(),1,std::multiplies<int>());
WilsonFermionR Dw(Umu,Grid,RBGrid,mass,params);
bench_wilson(src,result,Dw,volume,DaggerNo);
bench_wilson(src,result,Dw,volume,DaggerYes);
std::cout << std::endl;
}
}
std::cout<<GridLogMessage << "============================================================================="<< std::endl;
Grid_finalize();
}
void bench_wilson (
LatticeFermion & src,
LatticeFermion & result,
WilsonFermionR & Dw,
double const volume,
int const dag )
{
int ncall = 1000;
double t0 = usecond();
for(int i=0; i<ncall; i++) { Dw.Dhop(src,result,dag); }
double t1 = usecond();
double flops = 1344 * volume * ncall;
std::cout << flops/(t1-t0) << "\t\t";
}

16
benchmarks/Benchmark_zmm.cc
View File

@@ -40,14 +40,20 @@ int main(int argc,char **argv)
std::ofstream os("zmm.dat");
os << "#V Ls Lxy Lzt C++ Asm OMP L1 " <<std::endl;
std::cout<<GridLogMessage << "====================================================================="<<std::endl;
std::cout<<GridLogMessage << "= Benchmarking ZMM"<<std::endl;
std::cout<<GridLogMessage << "====================================================================="<<std::endl;
std::cout<<GridLogMessage << "Volume \t\t\t\tC++DW/MFLOPs\tASM-DW/MFLOPs\tdiff"<<std::endl;
std::cout<<GridLogMessage << "====================================================================="<<std::endl;
for(int L=4;L<=32;L+=4){
for(int m=1;m<=2;m++){
for(int Ls=8;Ls<=16;Ls+=8){
std::vector<int> grid({L,L,m*L,m*L});
std::cout << GridLogMessage <<"\t";
for(int i=0;i<4;i++) {
std::cout << grid[i]<<"x";
}
std::cout << Ls<<std::endl;
std::cout << Ls<<"\t\t";
bench(os,grid,Ls);
}
}
@@ -104,7 +110,6 @@ int bench(std::ofstream &os, std::vector<int> &latt4,int Ls)
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++){
@@ -116,7 +121,7 @@ int bench(std::ofstream &os, std::vector<int> &latt4,int Ls)
double flops=1344*volume/2;
mfc = flops*ncall/(t1-t0);
std::cout<<GridLogMessage << "Called C++ Dw"<< " mflop/s = "<< mfc<<std::endl;
std::cout<<mfc<<"\t\t";
QCD::WilsonKernelsStatic::AsmOpt=1;
t0=usecond();
@@ -125,7 +130,7 @@ int bench(std::ofstream &os, std::vector<int> &latt4,int Ls)
}
t1=usecond();
mfa = flops*ncall/(t1-t0);
std::cout<<GridLogMessage << "Called ASM Dw"<< " mflop/s = "<< mfa<<std::endl;
std::cout<<mfa<<"\t\t";
/*
int dag=DaggerNo;
t0=usecond();
@@ -163,8 +168,7 @@ int bench(std::ofstream &os, std::vector<int> &latt4,int Ls)
//resulta = (-0.5) * resulta;
diff = resulto-resulta;
std::cout<<GridLogMessage << "diff "<< norm2(diff)<<std::endl;
std::cout<<std::endl;
std::cout<<norm2(diff)<<std::endl;
return 0;
}

39
benchmarks/Make.inc
View File

@@ -1,39 +0,0 @@
bin_PROGRAMS = Benchmark_comms Benchmark_dwf Benchmark_dwf_ntpf Benchmark_dwf_sweep Benchmark_memory_asynch Benchmark_memory_bandwidth Benchmark_su3 Benchmark_wilson Benchmark_zmm
Benchmark_comms_SOURCES=Benchmark_comms.cc
Benchmark_comms_LDADD=-lGrid
Benchmark_dwf_SOURCES=Benchmark_dwf.cc
Benchmark_dwf_LDADD=-lGrid
Benchmark_dwf_ntpf_SOURCES=Benchmark_dwf_ntpf.cc
Benchmark_dwf_ntpf_LDADD=-lGrid
Benchmark_dwf_sweep_SOURCES=Benchmark_dwf_sweep.cc
Benchmark_dwf_sweep_LDADD=-lGrid
Benchmark_memory_asynch_SOURCES=Benchmark_memory_asynch.cc
Benchmark_memory_asynch_LDADD=-lGrid
Benchmark_memory_bandwidth_SOURCES=Benchmark_memory_bandwidth.cc
Benchmark_memory_bandwidth_LDADD=-lGrid
Benchmark_su3_SOURCES=Benchmark_su3.cc
Benchmark_su3_LDADD=-lGrid
Benchmark_wilson_SOURCES=Benchmark_wilson.cc
Benchmark_wilson_LDADD=-lGrid
Benchmark_zmm_SOURCES=Benchmark_zmm.cc
Benchmark_zmm_LDADD=-lGrid

7
benchmarks/Makefile.am
View File

@@ -1,8 +1 @@
# additional include paths necessary to compile the C++ library
AM_CXXFLAGS = -I$(top_srcdir)/include
AM_LDFLAGS = -L$(top_builddir)/lib
#
# Test code
#
include Make.inc