Extending pt-active signal creation to multi-aiu/table traces - #96
Merged
Merged
Conversation
…graph-breaks * observed time accumulation per rank and per job/token * event/counter overlap treatment for multi-aiu Signed-off-by: Lars Schneidenbach <schneidenbach@us.ibm.com>
Signed-off-by: Lars Schneidenbach <schneidenbach@us.ibm.com>
lasch
requested review from
WarningRan,
eklee15,
mcalman and
yuhaohaoyu
as code owners
April 14, 2026 19:56
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
#93 introduced computation of average pt-utilization based on total ideal cycles. This PR extends this approach to track the necessary status across multi-device runs.
The state tracking was extracted into separate classes and the pt-counter event creation and sanity checking refactored into separate functions. Overlap detection functionality is preserved.
self.accumulate_event_time = Truehard-coded but easy to change if necessary)101%to indicate problems if computation comes back with>100%(warnings issued as well, pt-statistics would be wrong either way)