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

2.3.5-5

Browse Source 
Add support for inter-node communication using sockets and InfiniBand/RoCE.
Improve latency.
Add support for aggregation.
Improve LL/regular tuning.
Remove tests as those are now at github.com/nvidia/nccl-tests .


[ROCm/rccl commit: f93fe9bfd9]
This commit is contained in:
Sylvain Jeaugey
2018-09-24 16:06:59 -07:00
parent 63ab6df5b3
commit 8ffcfac437
132 changed files with 12424 additions and 9415 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rccl/src/include/enqueue.h
+26
View File
@@ -0,0 +1,26 @@
/*************************************************************************
* Copyright (c) 2015-2018, NVIDIA CORPORATION. All rights reserved.
*
* See LICENSE.txt for license information
************************************************************************/
#ifndef NCCL_ENQUEUE_H_
#define NCCL_ENQUEUE_H_
#include "core.h"
#include "group.h"
typedef ncclResult_t(*ncclFunc_t)(const void* sendbuff, void* recvbuff, size_t count,
ncclDataType_t type, ncclRedOp_t op, int root, ncclComm_t comm, cudaStream_t stream);
ncclResult_t ncclEnqueueCheck(ncclFunc_t func, const char* primName, const void* sendbuff,
void* recvbuff, size_t count, ncclDataType_t type, ncclRedOp_t op, int root,
ncclComm_t comm, cudaStream_t stream);
ncclResult_t ncclCpuBarrierIn(ncclComm_t comm, int* isLast);
ncclResult_t ncclCpuBarrierLast(ncclComm_t comm);
ncclResult_t ncclCpuBarrierOut(ncclComm_t comm);
ncclResult_t ncclBarrierEnqueue(ncclComm_t comm);
ncclResult_t ncclBarrierEnqueueWait(ncclComm_t comm);
ncclResult_t ncclEnqueueEvents(ncclComm_t comm);
#endif // End include guard