portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-19 08:17:05 +01:00
Code Releases Activity

Systematise the accelerator primitives and locate to Grid/threads/Accelerator.h / Accelerator.cc

Browse Source 
Aim to reduce the amount of cuda and other code variations floating around all over the place.

Will move GpuInit iinto Accelerator.cc from Init.cc
Need to worry about SharedMemoryMPI.cc and the Peer2Peer windows
This commit is contained in:
Peter Boyle
2020-05-08 06:23:55 -07:00
parent 28a1fcaaff
commit f8b8e00090
13 changed files with 557 additions and 718 deletions
Download Patch File Download Diff File Expand all files Collapse all files

67
Grid/allocator/AlignedAllocator.h
View File

@ -29,28 +29,16 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
#ifndef GRID_ALIGNED_ALLOCATOR_H
#define GRID_ALIGNED_ALLOCATOR_H
#ifdef HAVE_MALLOC_MALLOC_H
#include <malloc/malloc.h>
#endif
#ifdef HAVE_MALLOC_H
#include <malloc.h>
#endif
#ifdef HAVE_MM_MALLOC_H
#include <mm_malloc.h>
#endif
#define POINTER_CACHE
#define GRID_ALLOC_ALIGN (2*1024*1024)
NAMESPACE_BEGIN(Grid);
// Move control to configure.ac and Config.h?
/*Move control to configure.ac and Config.h*/
#define POINTER_CACHE
/*Pinning pages is costly*/
/*Could maintain separate large and small allocation caches*/
#ifdef POINTER_CACHE
class PointerCache {
private:
/*Pinning pages is costly*/
/*Could maintain separate large and small allocation caches*/
static const int Ncache=128;
static int victim;
@ -159,44 +147,16 @@ public:
size_type bytes = __n*sizeof(_Tp);
profilerAllocate(bytes);
#ifdef POINTER_CACHE
_Tp *ptr = (_Tp *) PointerCache::Lookup(bytes);
#else
pointer ptr = nullptr;
#endif
#ifdef GRID_CUDA
////////////////////////////////////
// Unified (managed) memory
////////////////////////////////////
if ( ptr == (_Tp *) NULL ) {
// printf(" alignedAllocater cache miss %ld bytes ",bytes); BACKTRACEFP(stdout);
auto err = cudaMallocManaged((void **)&ptr,bytes);
if( err != cudaSuccess ) {
ptr = (_Tp *) NULL;
std::cerr << " cudaMallocManaged failed for " << bytes<<" bytes " <<cudaGetErrorString(err)<< std::endl;
assert(0);
}
}
assert( ptr != (_Tp *)NULL);
#endif
if ( ptr == (_Tp *) NULL ) ptr = (_Tp *) acceleratorAllocShared(bytes);
#ifdef GRID_SYCL
if ( ptr == (_Tp *) NULL ) ptr = (_Tp *) malloc_shared(bytes,*theGridAccelerator);
#endif
#if ( !defined(GRID_CUDA)) && (!defined(GRID_SYCL))
//////////////////////////////////////////////////////////////////////////////////////////
// 2MB align; could make option probably doesn't need configurability
//////////////////////////////////////////////////////////////////////////////////////////
#ifdef HAVE_MM_MALLOC_H
if ( ptr == (_Tp *) NULL ) ptr = (_Tp *) _mm_malloc(bytes,GRID_ALLOC_ALIGN);
#else
if ( ptr == (_Tp *) NULL ) ptr = (_Tp *) memalign(GRID_ALLOC_ALIGN,bytes);
#endif
assert( ptr != (_Tp *)NULL);
#endif
return ptr;
}
@ -211,20 +171,7 @@ public:
pointer __freeme = __p;
#endif
#ifdef GRID_CUDA
if ( __freeme ) cudaFree((void *)__freeme);
#endif
#ifdef GRID_SYCL
if ( __freeme ) free((void *)__freeme,*theGridAccelerator);
#endif
#if ( !defined(GRID_CUDA)) && (!defined(GRID_SYCL))
#ifdef HAVE_MM_MALLOC_H
if ( __freeme ) _mm_free((void *)__freeme);
#else
if ( __freeme ) free((void *)__freeme);
#endif
#endif
if ( __freeme ) acceleratorFreeShared((void *)__freeme);
}
// FIXME: hack for the copy constructor, eventually it must be avoided