portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-09-20 01:05:38 +01:00
Code Releases Activity

Hadrons: meson field threaded cache copy

Browse Source
This commit is contained in:
Antonin Portelli 2018-08-14 14:02:37 +01:00
parent 89d2fac92e
commit f4878d3a13
5 changed files with 21 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
extras/Hadrons/Modules/MContraction/A2AMesonField.hpp
View File

@ -296,7 +296,8 @@ void TA2AMesonField<FImpl>::execute(void)
+ vol * ( 2.0 * sizeof(Complex) *nmom ) * N_iii*N_jjj* ngamma;
startTimer("cache copy");
for(int iii=0;iii< N_iii;iii++)
parallel_for_nest(5)(int iii=0;iii< N_iii;iii++)
for(int jjj=0;jjj< N_jjj;jjj++)
for(int m =0;m< nmom;m++)
for(int g =0;g< ngamma;g++)
@ -310,30 +311,30 @@ void TA2AMesonField<FImpl>::execute(void)
// IO
if (!par().output.empty())
{
double blockSize, ioTime;
double blockSize, ioTime;
LOG(Message) << "Writing block to disk" << std::endl;
ioTime = -getDTimer("IO: write block");
startTimer("IO: total");
for(int m = 0; m < nmom; m++)
for(int g = 0; g < ngamma; g++)
{
if ((i == 0) and (j == 0))
for(int m = 0; m < nmom; m++)
for(int g = 0; g < ngamma; g++)
{
if ((i == 0) and (j == 0))
{
startTimer("IO: file creation");
initFile(m, g);
initFile(m, g);
stopTimer("IO: file creation");
}
}
startTimer("IO: write block");
saveBlock(mfBlock, m, g, i, j);
saveBlock(mfBlock, m, g, i, j);
stopTimer("IO: write block");
}
}
stopTimer("IO: total");
blockSize = static_cast<double>(nmom*ngamma*nt*N_ii*N_jj*sizeof(Complex));
ioTime += getDTimer("IO: write block");
LOG(Message) << "HDF5 IO done " << blockSize/ioTime*1.0e6/1024/1024
<< " MB/s" << std::endl;
}
<< " MB/s" << std::endl;
}
}
double nodes = env().getGrid()->NodeCount();

2
extras/Hadrons/Modules/MContraction/A2AMesonFieldKernels.hpp
View File

@ -174,7 +174,7 @@ void makeMesonFieldBlock(MesonField &mat,
if (caller) caller->startTimer("contraction: spin trace");
int pd = grid->_processors[orthogdim];
int pc = grid->_processor_coor[orthogdim];
parallel_for_nest2(int lt=0;lt<ld;lt++)
parallel_for_nest(2)(int lt=0;lt<ld;lt++)
{
for(int pt=0;pt<pd;pt++)
{

6
lib/cshift/Cshift_common.h
View File

@ -94,7 +94,7 @@ Gather_plane_extract(const Lattice<vobj> &rhs,std::vector<typename vobj::scalar_
int n1=rhs._grid->_slice_stride[dimension];
if ( cbmask ==0x3){
parallel_for_nest2(int n=0;n<e1;n++){
parallel_for_nest(2)(int n=0;n<e1;n++){
for(int b=0;b<e2;b++){
int o = n*n1;
@ -110,7 +110,7 @@ Gather_plane_extract(const Lattice<vobj> &rhs,std::vector<typename vobj::scalar_
// Case of SIMD split AND checker dim cannot currently be hit, except in
// Test_cshift_red_black code.
std::cout << " Dense packed buffer WARNING " <<std::endl;
parallel_for_nest2(int n=0;n<e1;n++){
parallel_for_nest(2)(int n=0;n<e1;n++){
for(int b=0;b<e2;b++){
int o=n*n1;
@ -191,7 +191,7 @@ template<class vobj> void Scatter_plane_merge(Lattice<vobj> &rhs,std::vector<typ
int e2=rhs._grid->_slice_block[dimension];
if(cbmask ==0x3 ) {
parallel_for_nest2(int n=0;n<e1;n++){
parallel_for_nest(2)(int n=0;n<e1;n++){
for(int b=0;b<e2;b++){
int o = n*rhs._grid->_slice_stride[dimension];
int offset = b+n*rhs._grid->_slice_block[dimension];

2
lib/lattice/Lattice_reduction.h
View File

@ -522,7 +522,7 @@ static void sliceMaddVector(Lattice<vobj> &R,std::vector<RealD> &a,const Lattice
tensor_reduced at; at=av;
parallel_for_nest2(int n=0;n<e1;n++){
parallel_for_nest(2)(int n=0;n<e1;n++){
for(int b=0;b<e2;b++){
int ss= so+n*stride+b;
R._odata[ss] = at*X._odata[ss]+Y._odata[ss];

4
lib/threads/Threads.h
View File

@ -41,12 +41,14 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
#define PARALLEL_FOR_LOOP _Pragma("omp parallel for schedule(static)")
#define PARALLEL_FOR_LOOP_INTERN _Pragma("omp for schedule(static)")
#define PARALLEL_NESTED_LOOP2 _Pragma("omp parallel for collapse(2)")
#define PARALLEL_NESTED_LOOP(n) _Pragma("omp parallel for collapse(n)")
#define PARALLEL_REGION _Pragma("omp parallel")
#define PARALLEL_CRITICAL _Pragma("omp critical")
#else
#define PARALLEL_FOR_LOOP
#define PARALLEL_FOR_LOOP_INTERN
#define PARALLEL_NESTED_LOOP2
#define PARALLEL_NESTED_LOOP(n)
#define PARALLEL_REGION
#define PARALLEL_CRITICAL
#endif
@ -54,7 +56,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
#define parallel_region PARALLEL_REGION
#define parallel_for PARALLEL_FOR_LOOP for
#define parallel_for_internal PARALLEL_FOR_LOOP_INTERN for
#define parallel_for_nest2 PARALLEL_NESTED_LOOP2 for
#define parallel_for_nest(n) PARALLEL_NESTED_LOOP(n) for
namespace Grid {