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

replayer update v0 (#1733)

Browse Source 
* First version of new replayer, with comments on future TODOs

* plus minor fixes for UT

* Updated format of recorder, especially in binary department, according to replayer's need
This commit is contained in:
Tim
2025-06-13 15:05:34 -04:00
committed by GitHub
parent 4486d091b8
commit ba97c9c18b
8 changed files with 768 additions and 1129 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/group.cc
+12 -3
View File
@@ -98,7 +98,10 @@ ncclResult_t ncclAsyncJobComplete(struct ncclAsyncJob* job) {
NCCL_API(ncclResult_t, ncclGroupStart);
ncclResult_t ncclGroupStart_impl() {
NCCLCHECK(Recorder::instance().record(rrGroupStart, ncclGroupDepth));
if (!mscclIsCaller())
{
NCCLCHECK(Recorder::instance().record(rrGroupStart, ncclGroupDepth));
}
ncclResult_t ret = ncclSuccess;
NVTX3_FUNC_RANGE_IN(nccl_domain);
@@ -117,7 +120,10 @@ ncclResult_t ncclGroupStartInternal() {
NCCL_API(ncclResult_t, ncclGroupEnd);
ncclResult_t ncclGroupEnd_impl() {
NCCLCHECK(Recorder::instance().record(rrGroupEnd, ncclGroupDepth));
if (!mscclIsCaller())
{
NCCLCHECK(Recorder::instance().record(rrGroupEnd, ncclGroupDepth));
}
ncclResult_t ret = ncclSuccess;
NVTX3_FUNC_RANGE_IN(nccl_domain);
NCCLCHECKGOTO(ncclGroupEndInternal(), ret, exit);
@@ -128,7 +134,10 @@ exit:
NCCL_API(ncclResult_t, ncclGroupSimulateEnd, ncclSimInfo_t* simInfo);
ncclResult_t ncclGroupSimulateEnd(ncclSimInfo_t* simInfo) {
Recorder::instance().record(ncclGroupDepth, simInfo);
if (!mscclIsCaller())
{
Recorder::instance().record(ncclGroupDepth, simInfo);
}
ncclResult_t ret = ncclSuccess;
NVTX3_FUNC_RANGE_IN(nccl_domain);
NCCLCHECKGOTO(ncclGroupEndInternal(simInfo), ret, exit);