portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-11-02 21:14:32 +00:00
Code Releases Activity

Faster code for split unsplit

Browse Source
This commit is contained in:
paboyle
2017-12-05 11:39:26 +00:00
parent ae3b7713a9
commit d93c6760ec
1 changed files with 6 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
lib/lattice/Lattice_transfer.h
View File

@@ -822,6 +822,7 @@ void Grid_split(std::vector<Lattice<Vobj> > & full,Lattice<Vobj> & split)
// Loop over reordered data post A2A
parallel_for(int c=0;c<chunk;c++){
std::vector<int> coor(ndim);
for(int m=0;m<M;m++){
for(int s=0;s<sP;s++){
@@ -833,7 +834,6 @@ void Grid_split(std::vector<Lattice<Vobj> > & full,Lattice<Vobj> & split)
uint64_t lex_vec = lex_fvol_vec/fvol;
// which node sets an adder to the coordinate
std::vector<int> coor(ndim);
Lexicographic::CoorFromIndex(coor, lex_fvol, ldims);
coor[d] += m*ldims[d];
Lexicographic::IndexFromCoor(coor, lex_r, rdims);
@@ -940,10 +940,11 @@ void Grid_unsplit(std::vector<Lattice<Vobj> > & full,Lattice<Vobj> & split)
{
// Loop over reordered data post A2A
for(int c=0;c<chunk;c++){
parallel_for(int c=0;c<chunk;c++){
std::vector<int> coor(ndim);
for(int m=0;m<M;m++){
for(int s=0;s<sP;s++){
// addressing; use lexico
int lex_r;
uint64_t lex_c = c+chunk*m+chunk*M*s;
@@ -952,7 +953,6 @@ void Grid_unsplit(std::vector<Lattice<Vobj> > & full,Lattice<Vobj> & split)
uint64_t lex_vec = lex_fvol_vec/fvol;
// which node sets an adder to the coordinate
std::vector<int> coor(ndim);
Lexicographic::CoorFromIndex(coor, lex_fvol, ldims);
coor[d] += m*ldims[d];
Lexicographic::IndexFromCoor(coor, lex_r, rdims);
@@ -978,9 +978,9 @@ void Grid_unsplit(std::vector<Lattice<Vobj> > & full,Lattice<Vobj> & split)
lsites = full_grid->lSites();
for(int v=0;v<nvector;v++){
assert(v<full.size());
// assert(v<full.size());
parallel_for(int site=0;site<lsites;site++){
assert(v*lsites+site < alldata.size());
// assert(v*lsites+site < alldata.size());
scalardata[site] = alldata[v*lsites+site];
}
vectorizeFromLexOrdArray(scalardata,full[v]);