Skip to content

Conversation

kyle-yh-kim
Copy link

Issue #, if available:
N/A

Description of changes:
Following the previous thread, this PR aims to reduce unnecessary re-compilation by updating NVTX trace dimensions into a run-time config.

Like any other run-time config, it is activated via environment variable with OFI_NCCL prefix.

export OFI_NCCL_NVTX_TRACE_DIMENSION="PER_COMM" or "PER_DEV" or "NONE"

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

unncessary re-compilation.

Signed-off-by: Kyle Kim <[email protected]>
@kyle-yh-kim kyle-yh-kim requested review from a team and bwbarrett as code owners October 3, 2025 04:53
/* Create nvtx domain */
char buf[64];
snprintf(buf, 64, "aws-ofi-nccl dev %d_%d", dev_id, i);
this->nvtx_domain[i] = nvtxDomainCreateA(buf);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Are we missing the destruction of this nvtx_domain currently ?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants