portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-16 23:07:05 +01:00
Code Releases Activity

Global changes to parallel_for structure.

Browse Source 
Move the comms flags to more sensible names
This commit is contained in:
paboyle
2017-02-21 05:24:27 -05:00
parent 3906cd2149
commit 3ae92fa2e6
43 changed files with 271 additions and 513 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
lib/lattice/Lattice_reduction.h
View File

@ -57,8 +57,7 @@ namespace Grid {
sumarray[i]=zero;
}
PARALLEL_FOR_LOOP
for(int thr=0;thr<grid->SumArraySize();thr++){
parallel_for(int thr=0;thr<grid->SumArraySize();thr++){
int nwork, mywork, myoff;
GridThread::GetWork(left._grid->oSites(),thr,mywork,myoff);
@ -68,7 +67,7 @@ PARALLEL_FOR_LOOP
}
sumarray[thr]=TensorRemove(vnrm) ;
}
vector_type vvnrm; vvnrm=zero; // sum across threads
for(int i=0;i<grid->SumArraySize();i++){
vvnrm = vvnrm+sumarray[i];
@ -114,18 +113,17 @@ PARALLEL_FOR_LOOP
sumarray[i]=zero;
}
PARALLEL_FOR_LOOP
for(int thr=0;thr<grid->SumArraySize();thr++){
parallel_for(int thr=0;thr<grid->SumArraySize();thr++){
int nwork, mywork, myoff;
GridThread::GetWork(grid->oSites(),thr,mywork,myoff);
vobj vvsum=zero;
for(int ss=myoff;ss<mywork+myoff; ss++){
vvsum = vvsum + arg._odata[ss];
}
sumarray[thr]=vvsum;
}
vobj vsum=zero; // sum across threads
for(int i=0;i<grid->SumArraySize();i++){
vsum = vsum+sumarray[i];