Allow intranode use of network connection (#383)

* Allow intranode use of network connection

* Checking for graph for null pointer

[ROCm/rccl commit: a3a8c2d56b]
このコミットが含まれているのは:
Wenkai Du
2021-06-08 07:37:59 -07:00
committed by GitHub
コミット c8a432dc25
7個のファイルの変更120行の追加49行の削除
+3
ファイルの表示
@@ -33,6 +33,7 @@ ncclResult_t ncclTopoGetNvbGpus(struct ncclTopoSystem* system, int rank, int* nr
ncclResult_t ncclTopoGetNetDev(struct ncclTopoSystem* system, int rank, struct ncclTopoGraph* graph, int channelId, int rr, int* net);
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);
ncclResult_t ncclTopoGetIntraNetDev(struct ncclTopoSystem* system, int rank, struct ncclTopoGraph* graph, int channelId, int type, int* dev);
// Set CPU affinity
ncclResult_t ncclTopoSetAffinity(struct ncclTopoSystem* system, int rank);
@@ -78,6 +79,8 @@ struct ncclTopoGraph {
int nHops;
int intra[MAXCHANNELS*NCCL_TOPO_MAX_NODES];
int inter[MAXCHANNELS*2];
int nIntraChannels;
int intraNets[MAXCHANNELS*NCCL_TOPO_MAX_NODES*2];
};
ncclResult_t ncclTopoCompute(struct ncclTopoSystem* system, struct ncclTopoGraph* graph);