Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

update trace replay mode for post analysis #20

Closed
wants to merge 1 commit into from

Conversation

kingchc
Copy link
Contributor

@kingchc kingchc commented Mar 11, 2021

Summary:

  • Output new traces with sequence number and latency for post analysis (one file per rank since there could be skew due to imbalanced workloads)
  • Remove extra barrier in non-blocking mode
  • add option --allow-list to specified desired collectives to be replayed and ignore the rest in the trace
  • cleaner summary output in the end (move to logging.info)
  • Add workaround to correctly complete UCC collectives in non-blocking mode

Differential Revision: D26730965

@facebook-github-bot facebook-github-bot added CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. fb-exported labels Mar 11, 2021
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D26730965

kingchc added a commit to kingchc/param-1 that referenced this pull request Mar 11, 2021
Summary:
Pull Request resolved: facebookresearch#20

- Output new traces with sequence number and latency for post analysis (one file per rank since there could be skew due to imbalanced workloads)
- Remove extra barrier in non-blocking mode
- add option `--allow-list` to specified desired collectives to be replayed and ignore the rest in the trace
- cleaner summary output in the end (move to logging.info)
- Add workaround to correctly complete UCC collectives in non-blocking mode

Differential Revision: D26730965

fbshipit-source-id: efc0102bc4b46bb50afcf42e2ceead13a38c9dea
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D26730965

kingchc added a commit to kingchc/param-1 that referenced this pull request Mar 12, 2021
Summary:
Pull Request resolved: facebookresearch#20

- Output new traces with sequence number and latency for post analysis (one file per rank since there could be skew due to imbalanced workloads)
- Remove extra barrier in non-blocking mode
- add option `--allow-list` to specified desired collectives to be replayed and ignore the rest in the trace
- cleaner summary output in the end (move to logging.info)
- Add workaround to correctly complete UCC collectives in non-blocking mode

Differential Revision: D26730965

fbshipit-source-id: 951e295a695310f2a3883572f4eabe0a8e1d604b
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D26730965

Summary:
Pull Request resolved: facebookresearch#20

- Output new traces with sequence number and latency for post analysis (one file per rank since there could be skew due to imbalanced workloads)
- Remove extra barrier in non-blocking mode
- add option `--allow-list`/`--allow-ops` to specified desired collectives to be replayed and ignore the rest in the trace
- cleaner summary output in the end (move to logging.info)
- Add workaround to correctly complete UCC collectives in non-blocking mode

Differential Revision: D26730965

fbshipit-source-id: 7d0ee80903245bed3390b42e3f8072e27bafb7e2
@facebook-github-bot
Copy link
Contributor

This pull request was exported from Phabricator. Differential Revision: D26730965

@facebook-github-bot
Copy link
Contributor

This pull request has been merged in 4a9e7ee.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. fb-exported Merged
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

2 participants