AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

2.20.3-1

Browse Source 
Add support for alternating rings, allow for cross-nic rings without
cross-rail communication.
Add support for user buffer registration for network send/recv.
Optimize aggregated operations to better utilize all channels.
Add flattening for BCM PCI gen5 switches.
Add support for inter-node NVLink communication
Add support for port fusion in NET/IB.
Add support for ReduceScatter and AllGather using Collnet.
Update net API to v8.
Fix hang during A2A connection.
This commit is contained in:
Sylvain Jeaugey
2024-02-05 05:06:02 -08:00
parent b6d7438d31
commit b6475625fb
74 changed files with 4632 additions and 2165 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/graph/xml.cc
+2 -2
View File
@@ -592,8 +592,8 @@ ncclResult_t ncclTopoGetXmlFromSys(struct ncclXmlNode* pciNode, struct ncclXml*
NCCLCHECK(xmlGetAttrStr(pciNode, "busid", &newBusId));
for (int s=0; s<parent->nSubs; s++) {
const char* busId;
NCCLCHECK(xmlGetAttrStr(parent->subs[s], "busid", &busId));
if (strcmp(newBusId, busId) < 0) { subIndex = s; break; }
NCCLCHECK(xmlGetAttr(parent->subs[s], "busid", &busId));
if (busId != NULL && strcmp(newBusId, busId) < 0) { subIndex = s; break; }
}
for (int s = parent->nSubs; s > subIndex; s--) parent->subs[s] = parent->subs[s-1];
parent->subs[subIndex] = pciNode;