SPLITCOMM design fix in src/misc/msccl (#1715)

* Fix TOC-TOU in mcclInit

* Improving vector resize thread safety

* Initial commit rank to comm change

* Removing unwanted include header changes

* Updated CHANGELOG.md

* Update CHANGELOG.md

Co-authored-by: Jeffrey Novotny <jnovotny@amd.com>

---------

Co-authored-by: Jeffrey Novotny <jnovotny@amd.com>
Этот коммит содержится в:
Avinash
2025-06-01 21:00:38 -05:00
коммит произвёл GitHub
родитель 2f6b20c00a
Коммит e94b360246
13 изменённых файлов: 134 добавлений и 111 удалений
+4 -4
Просмотреть файл
@@ -135,7 +135,7 @@ ncclMemFree_impl(void* ptr);
ncclResult_t
mscclLoadAlgo_impl(const char* mscclAlgoFilePath, mscclAlgoHandle_t* mscclAlgoHandle,
int rank);
const ncclComm_t comm);
ncclResult_t
mscclRunAlgo_impl(const void* sendBuff, const size_t sendCounts[], const size_t sDisPls[],
@@ -380,7 +380,7 @@ NCCL_API(ncclResult_t, ncclMemAlloc, void** ptr, size_t size);
NCCL_API(ncclResult_t, ncclMemFree, void* ptr);
NCCL_API(ncclResult_t, mscclLoadAlgo, const char* mscclAlgoFilePath,
mscclAlgoHandle_t* mscclAlgoHandle, int rank);
mscclAlgoHandle_t* mscclAlgoHandle, const ncclComm_t comm);
NCCL_API(ncclResult_t, mscclRunAlgo, const void* sendBuff, const size_t sendCounts[],
const size_t sDisPls[], void* recvBuff, const size_t recvCounts[],
@@ -620,10 +620,10 @@ ncclMemFree(void* ptr)
}
ncclResult_t
mscclLoadAlgo(const char* mscclAlgoFilePath, mscclAlgoHandle_t* mscclAlgoHandle, int rank)
mscclLoadAlgo(const char* mscclAlgoFilePath, mscclAlgoHandle_t* mscclAlgoHandle, const ncclComm_t comm)
{
return ::rccl::RcclGetFunctionTable()->mscclLoadAlgo_fn(mscclAlgoFilePath,
mscclAlgoHandle, rank);
mscclAlgoHandle, comm);
}
ncclResult_t