[PGNCCL] Use long/short wait for different non-blocking calls #142291
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/142291
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ✅ No FailuresAs of commit 0265eac with merge base 0610b97 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
pytorch-bot
bot
added
oncall: distributed
eqy
approved these changes
Dec 7, 2024
…alls" In nonblocking mode, we always check if the NCCL communicator is ready between issuing commands to it. Today this is done by the `waitReady()` function. Unfortunately, the `waitReady()` function is burned with `C10D_NCCL_CHECK_TIMEOUT_SLEEP` which would sleep for an interval between two consecutive checks. While this is nice when waiting for comm init or finalize to complete, it degrades performance of collective calls (which would almost certainly return success immediately.) This PR adds a `bool longInterval` argument to `waitReady` and let call site determine whether long wait is likely; if not, `waitReady` would use `sched_yield()` to more eagerly check for readiness. Thanks eqy for reporting an issue that small collectives has perf impact in nonblocking mode. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o [ghstack-poisoned]
|
Thanks @eqy , yeah, just fixed them |
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Merge failedReason: 1 mandatory check(s) failed. The first few are: Dig deeper by viewing the failures on hud |
fduwjj
approved these changes
Dec 9, 2024
| waitReady(); | ||
| // Wait with long interval if communicator is being initialized. | ||
| bool longInterval = !initialized_; | ||
| waitReady(longInterval); |
There was a problem hiding this comment.
can you maybe directly pass in (!initialized_)?
…alls" In nonblocking mode, we always check if the NCCL communicator is ready between issuing commands to it. Today this is done by the `waitReady()` function. Unfortunately, the `waitReady()` function is burned with `C10D_NCCL_CHECK_TIMEOUT_SLEEP` which would sleep for an interval between two consecutive checks. While this is nice when waiting for comm init or finalize, it degrades performance of collective calls (which would almost certainly return success immediately.) This PR adds a `bool longInterval` argument to `waitReady` and let call site determine whether long wait is likely; if not, `waitReady` would use `sched_yield()` to more eagerly check for readiness. Thanks eqy for reporting an issue that small collectives has perf impact in nonblocking mode. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o [ghstack-poisoned]
kwen2501
added
the
keep-going
kwen2501
changed the title
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
bluenote10
pushed a commit
to bluenote10/pytorch
that referenced
this pull request
Dec 14, 2024
…h#142291) In nonblocking mode, we always check if the NCCL communicator is ready between issuing commands to it. Today this is done by the `waitReady()` function. Unfortunately, the `waitReady()` function is burned with `C10D_NCCL_CHECK_TIMEOUT_SLEEP` which would sleep for an interval between two consecutive checks. While this is nice when waiting for comm init or finalize, it degrades performance of collective calls (which would almost certainly return success immediately.) This PR adds a `bool longInterval` argument to `waitReady` and let call site determine whether long wait is likely; if not, `waitReady` would use `sched_yield()` to more eagerly check for readiness. Thanks @eqy for reporting an issue that small collectives has perf impact in nonblocking mode. Pull Request resolved: pytorch#142291 Approved by: https://github.com/eqy, https://github.com/fduwjj
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Stack from ghstack (oldest at bottom):
In nonblocking mode, we always check if the NCCL communicator is ready between issuing commands to it.
Today this is done by the
waitReady()function.Unfortunately, the
waitReady()function is burned withC10D_NCCL_CHECK_TIMEOUT_SLEEPwhich would sleep for an interval between two consecutive checks.While this is nice when waiting for comm init or finalize, it degrades performance of collective calls (which would almost certainly return success immediately.)
This PR adds a
bool longIntervalargument towaitReadyand let call site determine whether long wait is likely; if not,waitReadywould usesched_yield()to more eagerly check for readiness.Thanks @eqy for reporting an issue that small collectives has perf impact in nonblocking mode.
cc @H-Huang @awgu @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k @c-p-i-o