1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
|
/*************************************************************************
* Copyright (c) 2015-2018, NVIDIA CORPORATION. All rights reserved.
*
* See LICENSE.txt for license information
************************************************************************/
#include "core.h"
#include "common_coll.h"
#include "enqueue.h"
#include "collectives.h"
ncclResult_t ncclReduceFunc(const void* sendbuff, void* recvbuff, const size_t count,
ncclDataType_t datatype, ncclRedOp_t op, int root, ncclComm_t comm, cudaStream_t stream) {
size_t nbytes = count*ncclTypeSize(datatype);
INFO(COLL,"opCount %lx sendbuff %p recvbuff %p count %zi size %zi datatype %d op %d root %d comm %p [nranks=%d] stream %p", comm->opCount, sendbuff, recvbuff, count, nbytes, datatype, op, root, comm, comm->nRanks, stream);
if (comm->nRanks == 1) {
if (sendbuff != recvbuff)
CUDACHECK(cudaMemcpyAsync(recvbuff, sendbuff, nbytes, cudaMemcpyDeviceToDevice, stream));
} else {
NCCLCHECK(transportSaveProxies(REDUCE_SUBSTEPS, REDUCE_BUFCHUNKS, 1, 1, nbytes, proxyPatternTo(root), comm));
NCCLCHECK(saveKernel(ncclCollReduce, sendbuff, recvbuff, count, datatype, op, root, comm, stream, nbytes, 1));
}
return ncclSuccess;
}
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);
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) {
return ncclEnqueueCheck(ncclReduceFunc, "Reduce", sendbuff, recvbuff, count, datatype,
op, root, comm, stream);
}
|
