2.7.3-1
Add support for A100 GPU and related platforms.
Add support for CUDA 11.
Add support for send/receive operations (beta).
[ROCm/rccl commit: 5949d96f36]
이 커밋은 다음에 포함됨:
@@ -25,10 +25,11 @@ ncclResult_t ncclTopoPrint(struct ncclTopoSystem* system);
|
||||
ncclResult_t ncclTopoComputePaths(struct ncclTopoSystem* system, struct ncclPeerInfo* info);
|
||||
void ncclTopoFree(struct ncclTopoSystem* system);
|
||||
ncclResult_t ncclTopoTrimSystem(struct ncclTopoSystem* system, struct ncclComm* comm);
|
||||
ncclResult_t ncclTopoComputeP2pChannels(struct ncclComm* comm);
|
||||
|
||||
// Query topology
|
||||
ncclResult_t ncclTopoGetNetDev(struct ncclTopoSystem* system, struct ncclTopoGraph* graph, int rank, int channelId, int* net);
|
||||
ncclResult_t ncclTopoCheckP2p(struct ncclTopoSystem* system, int64_t id1, int64_t id2, int* p2p);
|
||||
ncclResult_t ncclTopoGetNetDev(struct ncclTopoSystem* system, int rank, struct ncclTopoGraph* graph, int channelId, int* net);
|
||||
ncclResult_t ncclTopoCheckP2p(struct ncclTopoSystem* system, int64_t id1, int64_t id2, int* p2p, int *read);
|
||||
ncclResult_t ncclTopoCheckGdr(struct ncclTopoSystem* topo, int64_t busId, int netDev, int read, int* useGdr);
|
||||
|
||||
// Set CPU affinity
|
||||
@@ -96,7 +97,7 @@ ncclResult_t ncclTopoPostset(struct ncclComm* comm, int* firstRanks,
|
||||
|
||||
ncclResult_t ncclTopoConnectCollNet(struct ncclComm* comm, struct ncclTopoGraph* collNetGraph, int rank);
|
||||
|
||||
ncclResult_t ncclTopoSetThresholds(struct ncclComm* comm, int minCompCap, int maxCompCap, struct ncclTopoGraph* treeGraph, struct ncclTopoGraph* ringGraph, struct ncclTopoGraph* collNetGraph);
|
||||
ncclResult_t ncclTopoTuneModel(struct ncclComm* comm, int minCompCap, int maxCompCap, struct ncclTopoGraph* treeGraph, struct ncclTopoGraph* ringGraph, struct ncclTopoGraph* collNetGraph);
|
||||
#include "info.h"
|
||||
ncclResult_t ncclTopoGetAlgoTime(struct ncclInfo* info, int algorithm, int protocol, float* time);
|
||||
|
||||
|
||||
새 이슈에서 참조
사용자 차단