AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

RCCL 2.4 update

Browse Source
This commit is contained in:
Wenkai Du
2019-07-05 15:43:00 -07:00
parent 4d579e51cc
commit f11c8f60cd
95 changed files with 7829 additions and 614 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/collectives/reduce.cc
+3 -2
View File
@@ -1,5 +1,6 @@
/*************************************************************************
* Copyright (c) 2015-2019, NVIDIA CORPORATION. All rights reserved.
* Modifications Copyright (c) 2019 Advanced Micro Devices, Inc. All rights reserved.
*
* See LICENSE.txt for license information
************************************************************************/
@@ -8,9 +9,9 @@
#include "collectives.h"
NCCL_API(ncclResult_t, ncclReduce, const void* sendbuff, void* recvbuff, size_t count,
ncclDataType_t datatype, ncclRedOp_t op, int root, ncclComm_t comm, cudaStream_t stream);
ncclDataType_t datatype, ncclRedOp_t op, int root, ncclComm_t comm, hipStream_t stream);
ncclResult_t ncclReduce(const void* sendbuff, void* recvbuff, size_t count,
ncclDataType_t datatype, ncclRedOp_t op, int root, ncclComm_t comm, cudaStream_t stream) {
ncclDataType_t datatype, ncclRedOp_t op, int root, ncclComm_t comm, hipStream_t stream) {
struct ncclInfo info = { ncclCollReduce, "Reduce",
sendbuff, recvbuff, count, datatype, op, root, comm, stream, /* Args */
REDUCE_CHUNKSTEPS, REDUCE_SLICESTEPS };