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

Add ncclCommDump API (#2068)

Browse Source 
* Add ncclCommDump API

* remove trailing whitespace changes

* Add more proxy trace timestamps

* Add facebook_rccl namespace before proxyTrace timestamp call

* Clean up ProxyTrae construction

* Move updateProxyOpCounter to member function

* Move setProxyOpTimestamp to member function

* Move addNewProxyOp to member function

* Make internal methods private

* Make ProxyTrace thread safe

* Fix unit tests

* Fix overwritten ProxyTrace DONE setting in net.cc
This commit is contained in:
Ahmed Khan
2025-12-11 17:02:35 -05:00
committed by GitHub
parent 1f2f9f33ba
commit 08dd75712f
9 changed files with 230 additions and 138 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/proxy.cc
+2 -1
View File
@@ -1873,7 +1873,8 @@ ncclResult_t ncclProxyInit(struct ncclComm* comm, struct ncclSocket* sock, union
comm->proxyState->peerAddresses = peerAddresses;
comm->proxyState->peerAddressesUDS = peerAddressesUDS;
if (rcclParamEnableProxyTrace()) {
facebook_rccl::proxyTraceInit(comm->proxyState->proxyTrace, comm->rank, comm->commHash);
INFO(NCCL_PROXY, "Initializing ProxyTrace, rank: %d, commHash: %lu", comm->rank, comm->commHash);
comm->proxyState->proxyTrace = std::make_unique<facebook_rccl::ProxyTrace>(comm->rank);
}
// UDS support