frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 12:54:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

Deprecate overleap functions in CUDAAllocatorConfig, use AcceleratorAllocatorConfig instead (#156165)

Browse Source 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/156165
Approved by: https://github.com/albanD
ghstack dependencies: #149601, #157908, #150312
This commit is contained in:
Yu, Guangye
2025-07-30 09:14:36 +00:00
committed by PyTorch MergeBot
parent dfacf11f66
commit 1fc010a9d8
5 changed files with 44 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
c10/xpu/XPUCachingAllocator.cpp
View File

@ -1,3 +1,4 @@
#include <c10/core/AllocatorConfig.h>
#include <c10/util/flat_hash_map.h>
#include <c10/util/irange.h>
#include <c10/xpu/XPUCachingAllocator.h>
@ -20,8 +21,6 @@ constexpr size_t kMinBlockSize = 512;
constexpr size_t kSmallSize = 1048576;
// "small" allocations are packed in 2 MiB blocks
constexpr size_t kSmallBuffer = 2097152;
// "large" allocations may be packed in 20 MiB blocks
constexpr size_t kLargeBuffer = 20971520;
// allocations between 1 and 10 MiB may use kLargeBuffer
constexpr size_t kMinLargeAlloc = 10485760;
// round up large allocations to 2 MiB