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

Reorganise of file naming

Browse Source
This commit is contained in:
Peter Boyle
2015-06-03 12:47:05 +01:00
parent a3b599ae30
commit 1d0df449e8
75 changed files with 85 additions and 231 deletions
Download Patch File Download Diff File Expand all files Collapse all files

86
lib/Threads.h Normal file
View File

@ -0,0 +1,86 @@
#ifndef GRID_THREADS_H
#define GRID_THREADS_H
#ifdef _OPENMP
#define GRID_OMP
#endif
#define UNROLL _Pragma("unroll")
#ifdef GRID_OMP
#include <omp.h>
#define PARALLEL_FOR_LOOP _Pragma("omp parallel for ")
#define PARALLEL_NESTED_LOOP2 _Pragma("omp parallel for collapse(2)")
#else
#define PARALLEL_FOR_LOOP
#define PARALLEL_NESTED_LOOP2
#endif
namespace Grid {
// Introduce a class to gain deterministic bit reproducible reduction.
// make static; perhaps just a namespace is required.
class GridThread {
public:
static int _threads;
static void SetThreads(int thr) {
#ifdef GRID_OMP
_threads = MIN(thr,omp_get_max_threads()) ;
omp_set_num_threads(_threads);
#else
_threads = 1;
#endif
};
static void SetMaxThreads(void) {
#ifdef GRID_OMP
_threads = omp_get_max_threads();
omp_set_num_threads(_threads);
#else
_threads = 1;
#endif
};
static int GetThreads(void) { return _threads; };
static int SumArraySize(void) {return _threads;};
static void GetWork(int nwork, int me, int & mywork, int & myoff){
int basework = nwork/_threads;
int backfill = _threads-(nwork%_threads);
if ( me >= _threads ) {
mywork = myoff = 0;
} else {
mywork = (nwork+me)/_threads;
myoff = basework * me;
if ( me > backfill )
myoff+= (me-backfill);
}
return;
};
static void GetWorkBarrier(int nwork, int &me, int & mywork, int & myoff){
me = ThreadBarrier();
GetWork(nwork,me,mywork,myoff);
};
static int ThreadBarrier(void) {
#ifdef GRID_OMP
#pragma omp barrier
return omp_get_thread_num();
#else
return 0;
#endif
};
template<class obj> static void ThreadSum( std::vector<obj> &sum_array,obj &val,int me){
sum_array[me] = val;
val=zero;
ThreadBarrier();
for(int i=0;i<_threads;i++) val+= sum_array[i];
ThreadBarrier();
};
};
}
#endif