Merge remote-tracking branch 'nccl/master' into develop
This commit is contained in:
+8
-3
@@ -1,6 +1,6 @@
|
||||
/*************************************************************************
|
||||
* Copyright (c) 2016-2021, NVIDIA CORPORATION. All rights reserved.
|
||||
* Modifications Copyright (c) 2019-2021 Advanced Micro Devices, Inc. All rights reserved.
|
||||
* Copyright (c) 2016-2022, NVIDIA CORPORATION. All rights reserved.
|
||||
* Modifications Copyright (c) 2019-2022 Advanced Micro Devices, Inc. All rights reserved.
|
||||
*
|
||||
* See LICENSE.txt for license information
|
||||
************************************************************************/
|
||||
@@ -31,12 +31,15 @@ ncclResult_t ncclTopoComputeP2pChannels(struct ncclComm* comm);
|
||||
ncclResult_t ncclTopoGetNvbGpus(struct ncclTopoSystem* system, int rank, int* nranks, int** ranks);
|
||||
|
||||
// Query topology
|
||||
ncclResult_t ncclTopoGetNetDev(struct ncclTopoSystem* system, int rank, struct ncclTopoGraph* graph, int channelId, int rr, int* net);
|
||||
ncclResult_t ncclTopoGetNetDev(struct ncclComm* comm, int rank, struct ncclTopoGraph* graph, int channelId, int peerRank, int* net, int* proxyRank);
|
||||
ncclResult_t ncclTopoCheckP2p(struct ncclTopoSystem* system, int64_t id1, int64_t id2, int* p2p, int *read, int* intermediateRank);
|
||||
ncclResult_t ncclTopoCheckGdr(struct ncclTopoSystem* topo, int64_t busId, int netDev, int read, int* useGdr);
|
||||
#define MAX_XGMI_INTER_GPUS 4
|
||||
ncclResult_t ncclTopoGetIntraNetDev(struct ncclTopoSystem* system, int rank, struct ncclTopoGraph* graph, int channelId, int type, int* dev);
|
||||
ncclResult_t ncclTopoGetLinkType(struct ncclTopoSystem* system, int cudaDev1, int cudaDev2, bool* isXGMI, int maxInter=MAX_XGMI_INTER_GPUS, int nInter=0, int *inter=nullptr);
|
||||
int ncclPxnDisable();
|
||||
ncclResult_t ncclTopoGetPxnRanks(struct ncclComm* comm, int** intermediateRanks, int* nranks);
|
||||
ncclResult_t ncclTopoGetLocalRank(struct ncclTopoSystem* system, int rank, int* localRank);
|
||||
|
||||
// Find CPU affinity
|
||||
ncclResult_t ncclTopoGetCpuAffinity(struct ncclTopoSystem* system, int rank, cpu_set_t* affinity);
|
||||
@@ -54,6 +57,7 @@ ncclResult_t ncclTopoGetCpuAffinity(struct ncclTopoSystem* system, int rank, cpu
|
||||
#define NCCL_TOPO_CPU_TYPE_YONGFENG 1
|
||||
ncclResult_t ncclTopoCpuType(struct ncclTopoSystem* system, int* arch, int* vendor, int* model);
|
||||
ncclResult_t ncclTopoGetNetCount(struct ncclTopoSystem* system, int* count);
|
||||
ncclResult_t ncclTopoGetLocalNet(struct ncclTopoSystem* system, int rank, int* id);
|
||||
|
||||
#define NCCL_TOPO_MAX_NODES 256
|
||||
|
||||
@@ -76,6 +80,7 @@ struct ncclTopoGraph {
|
||||
int nChannels;
|
||||
float speedIntra;
|
||||
float speedInter;
|
||||
float latencyInter;
|
||||
int typeIntra;
|
||||
int typeInter;
|
||||
int sameChannels;
|
||||
|
||||
Reference in New Issue
Block a user