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

Merge remote-tracking branch 'nccl/master' into no-target-id

Sfoglia il codice sorgente
This commit is contained in:
Wenkai Du
2020-12-01 11:33:47 -05:00
parent 2e8b3a0857 920dbe5b35
commit d469947641
106 ha cambiato i file con 11943 aggiunte e 4104 eliminazioni
Scarica il file Patch Scarica il file Diff Expand all files Collapse all files
tools/topo_expl/model.cpp
+10 -10
Vedi File
@@ -83,7 +83,7 @@ ncclResult_t p2pCanConnect(int* ret, struct ncclTopoSystem* topo, struct ncclTop
}
/* Send: Create and return connect structures for this peer to connect to me */
ncclResult_t p2pSendSetup(struct ncclTopoSystem* topo, struct ncclTopoGraph* graph, struct ncclPeerInfo* myInfo, struct ncclPeerInfo* peerInfo,
ncclResult_t p2pSendSetup(struct ncclComm* comm, struct ncclTopoGraph* graph, struct ncclPeerInfo* myInfo, struct ncclPeerInfo* peerInfo,
struct ncclConnect* connectInfo, struct ncclConnector* send, int channelId) {
if (myInfo->pidHash == peerInfo->pidHash) {
if (myInfo->cudaDev == peerInfo->cudaDev) {
@@ -102,7 +102,7 @@ ncclResult_t p2pSendSetup(struct ncclTopoSystem* topo, struct ncclTopoGraph* gra
}
/* Create and return connect structures for this peer to connect to me */
ncclResult_t p2pRecvSetup(struct ncclTopoSystem* topo, struct ncclTopoGraph* graph, struct ncclPeerInfo* myInfo, struct ncclPeerInfo* peerInfo,
ncclResult_t p2pRecvSetup(struct ncclComm* comm, struct ncclTopoGraph* graph, struct ncclPeerInfo* myInfo, struct ncclPeerInfo* peerInfo,
struct ncclConnect* connectInfo, struct ncclConnector * recv, int channelId) {
return ncclSuccess;
}
@@ -126,12 +126,12 @@ ncclResult_t shmCanConnect(int* ret, struct ncclTopoSystem* topo, struct ncclTop
}
/* Create and return connect structures for this peer to connect to me */
ncclResult_t shmSendSetup(struct ncclTopoSystem* topo, struct ncclTopoGraph* graph, struct ncclPeerInfo* myInfo, struct ncclPeerInfo* peerInfo, struct ncclConnect* connectInfo, struct ncclConnector* send, int channelId) {
ncclResult_t shmSendSetup(struct ncclComm* comm, struct ncclTopoGraph* graph, struct ncclPeerInfo* myInfo, struct ncclPeerInfo* peerInfo, struct ncclConnect* connectInfo, struct ncclConnector* send, int channelId) {
INFO(NCCL_INIT|NCCL_SHM,"Ring %02d : %d[%lx] -> %d[%lx] via direct shared memory", channelId, myInfo->rank, myInfo->busId, peerInfo->rank, peerInfo->busId);
return ncclSuccess;
}
ncclResult_t shmRecvSetup(struct ncclTopoSystem* topo, struct ncclTopoGraph* graph, struct ncclPeerInfo* myInfo, struct ncclPeerInfo* peerInfo, struct ncclConnect* connectInfo, struct ncclConnector* recv, int channelId) {
ncclResult_t shmRecvSetup(struct ncclComm* comm, struct ncclTopoGraph* graph, struct ncclPeerInfo* myInfo, struct ncclPeerInfo* peerInfo, struct ncclConnect* connectInfo, struct ncclConnector* recv, int channelId) {
return ncclSuccess;
}
@@ -148,11 +148,11 @@ ncclResult_t netCanConnect(int* ret, struct ncclTopoSystem* topo, struct ncclTop
return ncclSuccess;
}
ncclResult_t netSendSetup(struct ncclTopoSystem* topo, struct ncclTopoGraph* graph, struct ncclPeerInfo* myInfo, struct ncclPeerInfo* peerInfo, struct ncclConnect* connectInfo, struct ncclConnector* send, int channelId) {
ncclResult_t netSendSetup(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(topo, myInfo->rank, graph, channelId, &netDev));
NCCLCHECK(ncclTopoCheckGdr(topo, myInfo->busId, netDev, 1, &useGdr));
NCCLCHECK(ncclTopoGetNetDev(comm->topo, myInfo->rank, graph, channelId, &netDev));
NCCLCHECK(ncclTopoCheckGdr(comm->topo, myInfo->busId, netDev, 1, &useGdr));
INFO(NCCL_INIT|NCCL_NET,"Ring %02d : %d[%lx] -> %d[%lx] [send] via NET/%s/%d%s", channelId, myInfo->rank, myInfo->busId, peerInfo->rank, peerInfo->busId, ncclNetName(), netDev,
useGdr ? "/GDRDMA" : "");
@@ -161,11 +161,11 @@ ncclResult_t netSendSetup(struct ncclTopoSystem* topo, struct ncclTopoGraph* gra
NCCL_PARAM(NetGdrLevel, "NET_GDR_LEVEL", PATH_PHB);
ncclResult_t netRecvSetup(struct ncclTopoSystem* topo, struct ncclTopoGraph* graph, struct ncclPeerInfo* myInfo, struct ncclPeerInfo* peerInfo, struct ncclConnect* connectInfo, struct ncclConnector* recv, int channelId) {
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;
NCCLCHECK(ncclTopoGetNetDev(topo, myInfo->rank, graph, channelId, &netDev));
NCCLCHECK(ncclTopoCheckGdr(topo, myInfo->busId, netDev, 0, &useGdr));
NCCLCHECK(ncclTopoGetNetDev(comm->topo, myInfo->rank, graph, channelId, &netDev));
NCCLCHECK(ncclTopoCheckGdr(comm->topo, myInfo->busId, netDev, 0, &useGdr));
INFO(NCCL_INIT|NCCL_NET,"Ring %02d : %d[%lx] -> %d[%lx] [receive] via NET/%s/%d%s", channelId, peerInfo->rank, peerInfo->busId, myInfo->rank, myInfo->busId, ncclNetName(), netDev,
useGdr ? "/GDRDMA" : "");