portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-13 12:47:05 +01:00
Code Releases Activity

Hand unrolled version of dslash in a separate class.

Browse Source 
Useful to compare; raises Intel compiler from 9GFlop/s to 17.5 Gflops.
                   on ivybridge core. Raises Clang form 14.5 to 17.5
This commit is contained in:
Peter Boyle
2015-05-26 19:54:03 +01:00
parent c2ffb1a098
commit 20100d0a40
9 changed files with 157 additions and 370 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
benchmarks/Grid_wilson.cc
View File

@ -31,11 +31,9 @@ int main (int argc, char ** argv)
std::cout << "Grid is setup to use "<<threads<<" threads"<<std::endl;
std::vector<int> seeds({1,2,3,4});
GridParallelRNG pRNG(&Grid);
// std::vector<int> seeds({1,2,3,4});
// pRNG.SeedFixedIntegers(seeds);
pRNG.SeedRandomDevice();
pRNG.SeedFixedIntegers(seeds);
// pRNG.SeedRandomDevice();
LatticeFermion src (&Grid); random(pRNG,src);
LatticeFermion result(&Grid); result=zero;
@ -55,8 +53,10 @@ int main (int argc, char ** argv)
Complex cone(1.0,0.0);
for(int nn=0;nn<Nd;nn++){
random(pRNG,U[nn]);
if (nn!=0) U[nn]=zero;
else U[nn] = cone;
if(0) {
if (nn==-1) { U[nn]=zero; std::cout << "zeroing gauge field in dir "<<nn<<std::endl; }
else { U[nn] = cone;std::cout << "unit gauge field in dir "<<nn<<std::endl; }
}
pokeIndex<LorentzIndex>(Umu,U[nn],nn);
}
@ -85,7 +85,7 @@ int main (int argc, char ** argv)
WilsonMatrix Dw(Umu,Grid,RBGrid,mass);
std::cout << "Calling Dw"<<std::endl;
int ncall=1000;
int ncall=10000;
double t0=usecond();
for(int i=0;i<ncall;i++){
Dw.Dhop(src,result,0);