OSSForks/pytorch
1
0
Fork 0
mirror of https://github.com/zebrajr/pytorch.git synced 2025-12-06 12:20:52 +01:00
Code Issues Actions 38 Packages Projects Releases Wiki Activity

Replace insert with std::rotate_copy for RingBuffer (#165348)

Browse Source 
Pull Request resolved: https://github.com/pytorch/pytorch/pull/165348
Approved by: https://github.com/eqy, https://github.com/Skylion007
This commit is contained in:
Lakshay Garg 2025-10-14 05:11:25 +00:00 committed by PyTorch MergeBot
parent 33bfec27ff
commit 496adf9f9c
1 changed files with 7 additions and 19 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
c10/cuda/CUDACachingAllocator.cpp
View File

@ -1080,19 +1080,12 @@ class RingBuffer {
void getEntries(std::vector<T>& result) const { void getEntries(std::vector<T>& result) const {
std::lock_guard<std::mutex> lk(alloc_trace_lock); std::lock_guard<std::mutex> lk(alloc_trace_lock);
result.reserve(alloc_trace->size()); result.reserve(result.size() + alloc_trace->size());
result.insert( std::rotate_copy(
result.end(),
alloc_trace->begin() +
static_cast<typename std::vector<T>::difference_type>(
alloc_trace_next),
alloc_trace->end());
result.insert(
result.end(),
alloc_trace->begin(), alloc_trace->begin(),
alloc_trace->begin() + std::next(alloc_trace->begin(), alloc_trace_next),
static_cast<typename std::vector<T>::difference_type>( alloc_trace->end(),
alloc_trace_next)); std::back_inserter(result));
} }
void clear() { void clear() {
@ -4466,10 +4459,7 @@ struct BackendStaticInitializer {
if (kv[0] == "backend") { if (kv[0] == "backend") {
#ifdef USE_ROCM #ifdef USE_ROCM
// convenience for ROCm users to allow either CUDA or HIP env var // convenience for ROCm users to allow either CUDA or HIP env var
if (kv[1] == if (kv[1] == "cudaMallocAsync" || kv[1] == "hipMallocAsync")
"cud"
"aMallocAsync" ||
kv[1] == "hipMallocAsync")
#else #else
if (kv[1] == "cudaMallocAsync") if (kv[1] == "cudaMallocAsync")
#endif #endif
@ -4491,9 +4481,7 @@ struct BackendStaticInitializer {
// HIPAllocatorMasqueradingAsCUDA because it needs to happen during static // HIPAllocatorMasqueradingAsCUDA because it needs to happen during static
// initialization, and doing so there may introduce static initialization // initialization, and doing so there may introduce static initialization
// order (SIOF) issues. // order (SIOF) issues.
#define HIP_MASQUERADING_AS_CUDA \ #define HIP_MASQUERADING_AS_CUDA "cuda"
"cud" \
"a"
at::SetAllocator(c10::Device(HIP_MASQUERADING_AS_CUDA).type(), r, 0); at::SetAllocator(c10::Device(HIP_MASQUERADING_AS_CUDA).type(), r, 0);
allocator.store(r); allocator.store(r);
#undef HIP_MASQUERADING_AS_CUDA #undef HIP_MASQUERADING_AS_CUDA