AI/rocm-systems
2
0
Forka 0
Codice Problemi Pull Request Actions Pacchetti Progetti Rilasci Wiki Attività

Enable collnet in RCCL (#333)

Sfoglia il codice sorgente 
* Enable CollNet and use different number of channels

* topo_expl: enable collnet
This commit is contained in:
Wenkai Du
2021-03-19 12:58:13 -07:00
committato da GitHub
parent 244e25d980
commit 1d6244b18d
14 ha cambiato i file con 393 aggiunte e 24 eliminazioni
Scarica il file Patch Scarica il file Diff Expand all files Collapse all files
tools/topo_expl/model.cpp
+33 -2
Vedi File
@@ -159,8 +159,6 @@ ncclResult_t netSendSetup(struct ncclComm* comm, struct ncclTopoGraph* graph, st
return ncclSuccess;
}
NCCL_PARAM(NetGdrLevel, "NET_GDR_LEVEL", PATH_PHB);
ncclResult_t netRecvSetup(struct ncclComm* comm, struct ncclTopoGraph* graph, struct ncclPeerInfo* myInfo, struct ncclPeerInfo* peerInfo, struct ncclConnect* connectInfo, struct ncclConnector* recv, int channelId) {
int netDev, useGdr = 0;
@@ -179,6 +177,39 @@ struct ncclTransport netTransport = {
{ netRecvSetup, NULL, NULL, NULL }
};
/* Determine if two peers can communicate with NET */
ncclResult_t collNetCanConnect(int* ret, struct ncclTopoSystem* topo, struct ncclTopoGraph* graph, struct ncclPeerInfo* info1, struct ncclPeerInfo* info2) {
*ret = 1;
return ncclSuccess;
}
ncclResult_t collNetSendSetup(struct ncclComm* comm, struct ncclTopoGraph* graph, struct ncclPeerInfo* myInfo, struct ncclPeerInfo* peerInfo, struct ncclConnect* connectInfo, struct ncclConnector* send, int channelId) {
int netDev, useGdr = 0;
NCCLCHECK(ncclTopoGetNetDev(comm->topo, myInfo->rank, graph, channelId, &netDev));
NCCLCHECK(ncclTopoCheckGdr(comm->topo, myInfo->busId, netDev, 1, &useGdr));
INFO(NCCL_INIT|NCCL_NET,"Coll %02d : %d [send] via COLLNET/%s/%d%s", channelId, myInfo->rank, "SHARP", netDev, useGdr ? "/GDRDMA" : "");
return ncclSuccess;
}
ncclResult_t collNetRecvSetup(struct ncclComm* comm, struct ncclTopoGraph* graph, struct ncclPeerInfo* myInfo, struct ncclPeerInfo* peerInfo, struct ncclConnect* connectInfo, struct ncclConnector* recv, int channelId) {
int netDev, useGdr = 0;
NCCLCHECK(ncclTopoGetNetDev(comm->topo, myInfo->rank, graph, channelId, &netDev));
NCCLCHECK(ncclTopoCheckGdr(comm->topo, myInfo->busId, netDev, 0, &useGdr));
INFO(NCCL_INIT|NCCL_NET,"Coll %02d : %d [receive] via COLLNET/%s/%d%s", channelId, myInfo->rank, "SHARP", netDev, useGdr ? "/GDRDMA" : "");
return ncclSuccess;
}
struct ncclTransport collNetTransport = {
"COL",
collNetCanConnect,
{ collNetSendSetup, NULL, NULL, NULL },
{ collNetRecvSetup, NULL, NULL, NULL }
};
struct ncclTransport ncclTransports[NTRANSPORTS] = {
p2pTransport,
shmTransport,