portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-10 07:55:35 +00:00
Code Releases Activity

Aligned allocate on CPU put through this interface

Browse Source
This commit is contained in:
Peter Boyle 2020-06-20 14:34:29 -04:00
parent 0d2f913a1a
commit 6c5fa8dcd8
1 changed files with 9 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
Grid/threads/Accelerator.h
View File

@ -336,12 +336,11 @@ inline void acceleratorCopyFromDevice(void *from,void *to,size_t bytes){ hipMemc
////////////////////////////////////////////// //////////////////////////////////////////////
// CPU Target - No accelerator just thread instead // CPU Target - No accelerator just thread instead
////////////////////////////////////////////// //////////////////////////////////////////////
#define GRID_ALLOC_ALIGN (2*1024*1024) // 2MB aligned
#if ( (!defined(GRID_SYCL)) && (!defined(GRID_CUDA)) && (!defined(GRID_HIP)) ) #if ( (!defined(GRID_SYCL)) && (!defined(GRID_CUDA)) && (!defined(GRID_HIP)) )
#undef GRID_SIMT #undef GRID_SIMT
#define GRID_ALLOC_ALIGN (2*1024*1024) // 2MB aligned
#define accelerator #define accelerator
#define accelerator_inline strong_inline #define accelerator_inline strong_inline
#define accelerator_for(iterator,num,nsimd, ... ) thread_for(iterator, num, { __VA_ARGS__ }); #define accelerator_for(iterator,num,nsimd, ... ) thread_for(iterator, num, { __VA_ARGS__ });
@ -367,6 +366,14 @@ inline void acceleratorFreeDevice(void *ptr){free(ptr);};
#endif // CPU target #endif // CPU target
#ifdef HAVE_MM_MALLOC_H
inline void *acceleratorAllocCpu(size_t bytes){return _mm_malloc(bytes,GRID_ALLOC_ALIGN);};
inline void acceleratorFreeCpu (void *ptr){_mm_free(ptr);};
#else
inline void *acceleratorAllocCpu(size_t bytes){return memalign(GRID_ALLOC_ALIGN,bytes);};
inline void acceleratorFreeCpu (void *ptr){free(ptr);};
#endif
/////////////////////////////////////////////////// ///////////////////////////////////////////////////
// Synchronise across local threads for divergence resynch // Synchronise across local threads for divergence resynch
/////////////////////////////////////////////////// ///////////////////////////////////////////////////