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]
This commit is contained in:
@@ -126,8 +126,10 @@ ncclResult_t ncclTopoConnectNodes(struct ncclTopoNode* node, struct ncclTopoNode
|
||||
ncclResult_t ncclTopoPrintPaths(struct ncclTopoSystem* system);
|
||||
ncclResult_t ncclTopoLoadSystem(const char* xmlTopoFile, struct ncclTopoSystem* system);
|
||||
|
||||
ncclResult_t ncclTopoGetLocalNet(struct ncclTopoSystem* system, int rank, int64_t* id, int rr);
|
||||
|
||||
ncclResult_t ncclTopoGetSystemFromXml(struct ncclXml* xml, struct ncclTopoSystem** topoSystem);
|
||||
ncclResult_t ncclTopoGetGraphFromXml(struct ncclXmlNode *xmlGraphs, struct ncclTopoSystem* system, struct ncclTopoGraph* graph);
|
||||
ncclResult_t ncclTopoGetGraphFromXml(struct ncclXmlNode *xmlGraphs, struct ncclTopoSystem* system, struct ncclTopoGraph* graph, int* nChannels);
|
||||
ncclResult_t ncclTopoGetXmlFromGraphs(int ngraphs, struct ncclTopoGraph** graphs, struct ncclTopoSystem* system, struct ncclXml *xml);
|
||||
|
||||
static ncclResult_t ncclTopoIdToIndex(struct ncclTopoSystem* system, int type, int64_t id, int* index) {
|
||||
@@ -141,4 +143,15 @@ static ncclResult_t ncclTopoIdToIndex(struct ncclTopoSystem* system, int type, i
|
||||
return ncclInternalError;
|
||||
}
|
||||
|
||||
static ncclResult_t ncclTopoRankToIndex(struct ncclTopoSystem* system, int rank, int* index) {
|
||||
*index = -1;
|
||||
for (int i=0; i<system->nodes[GPU].count; i++) {
|
||||
if (system->nodes[GPU].nodes[i].gpu.rank == rank) {
|
||||
*index = i;
|
||||
return ncclSuccess;
|
||||
}
|
||||
}
|
||||
return ncclInternalError;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
Reference in New Issue
Block a user