portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-18 07:47:06 +01:00
Code Releases Activity

Threading support rework.

Browse Source 
Placed parallel pragmas as macros; implemented deterministic thread reduction in style of
BFM.
This commit is contained in:
Peter Boyle
2015-05-12 07:51:41 +01:00
parent 8b765be2b1
commit 65c91eae64
26 changed files with 276 additions and 184 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
lib/lattice/Grid_lattice_overload.h
View File

@ -10,7 +10,7 @@ namespace Grid {
inline Lattice<vobj> operator -(const Lattice<vobj> &r)
{
Lattice<vobj> ret(r._grid);
#pragma omp parallel for
PARALLEL_FOR_LOOP
for(int ss=0;ss<r._grid->oSites();ss++){
vstream(ret._odata[ss], -r._odata[ss]);
}
@ -47,7 +47,7 @@ namespace Grid {
inline auto operator * (const left &lhs,const Lattice<right> &rhs) -> Lattice<decltype(lhs*rhs._odata[0])>
{
Lattice<decltype(lhs*rhs._odata[0])> ret(rhs._grid);
#pragma omp parallel for
PARALLEL_FOR_LOOP
for(int ss=0;ss<rhs._grid->oSites(); ss++){
decltype(lhs*rhs._odata[0]) tmp=lhs*rhs._odata[ss];
vstream(ret._odata[ss],tmp);
@ -59,7 +59,7 @@ namespace Grid {
inline auto operator + (const left &lhs,const Lattice<right> &rhs) -> Lattice<decltype(lhs+rhs._odata[0])>
{
Lattice<decltype(lhs+rhs._odata[0])> ret(rhs._grid);
#pragma omp parallel for
PARALLEL_FOR_LOOP
for(int ss=0;ss<rhs._grid->oSites(); ss++){
decltype(lhs+rhs._odata[0]) tmp =lhs-rhs._odata[ss];
vstream(ret._odata[ss],tmp);
@ -71,7 +71,7 @@ namespace Grid {
inline auto operator - (const left &lhs,const Lattice<right> &rhs) -> Lattice<decltype(lhs-rhs._odata[0])>
{
Lattice<decltype(lhs-rhs._odata[0])> ret(rhs._grid);
#pragma omp parallel for
PARALLEL_FOR_LOOP
for(int ss=0;ss<rhs._grid->oSites(); ss++){
decltype(lhs-rhs._odata[0]) tmp=lhs-rhs._odata[ss];
vstream(ret._odata[ss],tmp);
@ -83,7 +83,7 @@ namespace Grid {
inline auto operator * (const Lattice<left> &lhs,const right &rhs) -> Lattice<decltype(lhs._odata[0]*rhs)>
{
Lattice<decltype(lhs._odata[0]*rhs)> ret(lhs._grid);
#pragma omp parallel for
PARALLEL_FOR_LOOP
for(int ss=0;ss<lhs._grid->oSites(); ss++){
decltype(lhs._odata[0]*rhs) tmp =lhs._odata[ss]*rhs;
vstream(ret._odata[ss],tmp);
@ -95,7 +95,7 @@ namespace Grid {
inline auto operator + (const Lattice<left> &lhs,const right &rhs) -> Lattice<decltype(lhs._odata[0]+rhs)>
{
Lattice<decltype(lhs._odata[0]+rhs)> ret(lhs._grid);
#pragma omp parallel for
PARALLEL_FOR_LOOP
for(int ss=0;ss<rhs._grid->oSites(); ss++){
decltype(lhs._odata[0]+rhs) tmp=lhs._odata[ss]+rhs;
vstream(ret._odata[ss],tmp);
@ -107,7 +107,7 @@ namespace Grid {
inline auto operator - (const Lattice<left> &lhs,const right &rhs) -> Lattice<decltype(lhs._odata[0]-rhs)>
{
Lattice<decltype(lhs._odata[0]-rhs)> ret(lhs._grid);
#pragma omp parallel for
PARALLEL_FOR_LOOP
for(int ss=0;ss<rhs._grid->oSites(); ss++){
decltype(lhs._odata[0]-rhs) tmp=lhs._odata[ss]-rhs;
vstream(ret._odata[ss],tmp);