Allow shuffle extraction before protocol drain#1019
Open
pentschev wants to merge 2 commits into
Open
Conversation
rapidsai#927 made `Shuffler::wait()` wait until all internal protocol/send cleanup had drained before returning, which introduced a 10-20% regression in `bench_shuffle` on an 8-GPU node by delaying output extraction and unpack/concat work. For the reported 16 input partitions / 4 output partitions case, mean per-rank elapsed time improved from about 207 ms back down to about 172 ms after restoring overlap. This change makes `wait()` return when local shuffle output is ready to extract, while adding `wait_reusable()` for callers that need the stronger full-drain guarantee before reusing an op_id.
pentschev
added
bug
…before-protocol-drain
madsbk
reviewed
May 8, 2026
Member
madsbk
left a comment
madsbk
left a comment
There was a problem hiding this comment.
Do we need to update the Python bindings as well?
What about renaming to wait_drained()? Or maybe wait_for_drain()?
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
#927 made
Shuffler::wait()wait until all internal protocol/send cleanup had drained before returning, which introduced a 10-20% regression inbench_shuffleon an 8-GPU node by delaying output extraction and unpack/concat work. For the reported 16 input partitions / 4 output partitions case, mean per-rank elapsed time improved from about 207 ms back down to about 172 ms after restoring overlap. This change makeswait()return when local shuffle output is ready to extract, while addingwait_reusable()for callers that need the stronger full-drain guarantee before reusing an op_id.