Add MSCCL Support (#658)

* Add MSCCL support

* Add alignment and message size checking

* Fix nRanks checking, in-place and out-of-place tests and group call handling

* Fix hipGraph unit test

* Change MSCCL init warning to INFO

* Revise license info

[ROCm/rccl commit: adafc0f759]
This commit is contained in:
Ziyue Yang
2022-12-13 07:51:04 +08:00
committed by GitHub
parent bb27f5c8ba
commit 431c7d2cb9
38 changed files with 40923 additions and 11 deletions
+1 -5
View File
@@ -1,5 +1,6 @@
/*************************************************************************
* Copyright (c) 2015-2017, NVIDIA CORPORATION. All rights reserved.
* Modifications Copyright (c) Microsoft Corporation. Licensed under the MIT License.
*
* See LICENSE.txt for license information
************************************************************************/
@@ -67,11 +68,6 @@ extern __thread struct ncclComm* ncclGroupCommHead;
extern __thread struct ncclComm* ncclGroupCommPreconnectHead;
extern __thread int ncclGroupBlocking;
inline ncclResult_t ncclGroupStartInternal() {
ncclGroupDepth++;
return ncclSuccess;
}
inline ncclResult_t ncclGroupErrCheck(ncclResult_t ret) {
if (ncclGroupDepth > 0) {
if (ret != ncclSuccess && ret != ncclInProgress) ncclGroupError = ret;