Skip to content

Test/pt overhead draft#1816

Draft
fengjica wants to merge 4 commits into
ai-dynamo:mainfrom
fengjica:test/pt_overhead_draft
Draft

Test/pt overhead draft#1816
fengjica wants to merge 4 commits into
ai-dynamo:mainfrom
fengjica:test/pt_overhead_draft

Conversation

@fengjica

Copy link
Copy Markdown
Collaborator

What?

Describe what this PR is doing.

Why?

Justification for the PR. If there is an existing issue/bug, please reference it. For
bug fixes, the 'Why?' and 'What?' can be merged into a single item.

How?

It is optional, but for complex PRs, please provide information about the design,
architecture, approach, etc.

fengjica added 4 commits June 19, 2026 11:43
- progressCompletionQueue: add lock_acquired and use_try_lock params
- PT uses try_lock (less likely to blocks data thread)
- Data thread uses blocking lock in postXfer/checkXfer/getNotifs
- EAGAIN retry loops drain CQ under same lock (lock_acquired=true)
- Remove if (!progress_thread_enabled_) guards so data thread always
  progresses completions
Use an RAII guard to set data_thread_active_ flag during postXfer.
The progress thread skips polling when this flag is set, since the
data thread handles its own CQ progress. Eliminates ep_mutex_
contention during active submission bursts.
@copy-pr-bot

copy-pr-bot Bot commented Jun 23, 2026

Copy link
Copy Markdown

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@github-actions

Copy link
Copy Markdown

👋 Hi fengjica! Thank you for contributing to ai-dynamo/nixl.

Your PR reviewers will review your contribution then trigger the CI to test your changes.

🚀

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

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant