portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-10 07:55:35 +00:00
Code Releases Activity

FFT optimisation

Browse Source
This commit is contained in:
Antonin Portelli 2016-10-24 19:25:40 +01:00
parent a795b5705e
commit 13bf0482e3
1 changed files with 13 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
lib/FFT.h
View File

@ -200,18 +200,14 @@ namespace Grid {
sign,FFTW_ESTIMATE);
}
double add,mul,fma;
FFTW<scalar>::fftw_flops(p,&add,&mul,&fma);
flops_call = add+mul+2.0*fma;
GridStopWatch timer;
std::vector<int> lcoor(Nd), gcoor(Nd);
// Barrel shift and collect global pencil
for(int p=0;p<processors[dim];p++) {
for(int idx=0;idx<sgrid->lSites();idx++) {
std::vector<int> lcoor(Nd);
sgrid->LocalIndexToLocalCoor(idx,lcoor);
sobj s;
@ -228,14 +224,11 @@ namespace Grid {
// Loop over orthog coords
int NN=pencil_g.lSites();
GridStopWatch Timer;
Timer.Start();
GridStopWatch timer;
timer.Start();
PARALLEL_FOR_LOOP
for(int idx=0;idx<NN;idx++) {
std::vector<int> lcoor(Nd);
for(int idx=0;idx<NN;idx++) {
pencil_g.LocalIndexToLocalCoor(idx,lcoor);
if ( lcoor[dim] == 0 ) { // restricts loop to plane at lcoor[dim]==0
@ -245,15 +238,17 @@ PARALLEL_FOR_LOOP
}
}
Timer.Stop();
usec += Timer.useconds();
flops+= flops_call*NN;
timer.Stop();
double add,mul,fma;
FFTW<scalar>::fftw_flops(p,&add,&mul,&fma);
flops_call = add+mul+2.0*fma;
usec += timer.useconds();
flops+= flops_call*NN;
int pc = processor_coor[dim];
for(int idx=0;idx<sgrid->lSites();idx++) {
std::vector<int> lcoor(Nd);
for(int idx=0;idx<sgrid->lSites();idx++) {
sgrid->LocalIndexToLocalCoor(idx,lcoor);
std::vector<int> gcoor = lcoor;
gcoor = lcoor;
// extract the result
sobj s;
gcoor[dim] = lcoor[dim]+L*pc;