Refactor hash join with multiset#18021
Merged
rapids-bot[bot] merged 26 commits intorapidsai:branch-25.08from Jul 9, 2025
Merged
Refactor hash join with multiset#18021rapids-bot[bot] merged 26 commits intorapidsai:branch-25.08from
rapids-bot[bot] merged 26 commits intorapidsai:branch-25.08from
Conversation
|
Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually. Contributors can view more details about this message here. |
PointKernel
added a commit
to NVIDIA/cuCollections
that referenced
this pull request
Feb 27, 2025
…ance (#681) This PR enhances the CG-based device insertion by introducing an additional `SupportsErase` build-time check. When erasure is not required, the new implementation leverages this flag to select a more efficient code path, ensuring comparisons are made against an empty sentinel without loading the target slot's content into the CAS operation. This optimization gets rid of excessive local memory transactions, resulting in a 10~30% improvement in multimap performance. This PR also updates the multimap insert and count benchmarks to run with the new implementations. Unblocking rapidsai/cudf#18021
1 task
Member
Author
|
With NVIDIA/cuCollections#694, we can get about the same performance with the new cuco data structure. |
Member
Author
|
/ok to test 3d6b0a9 |
Member
Author
|
/ok to test cd19e35 |
Member
Author
|
/ok to test fcfee9f |
Member
Author
|
/ok to test 2d30df7 |
Member
Author
Performance on RTX8000 for reference. |
6 tasks
davidwendt
approved these changes
Jul 8, 2025
shrshi
approved these changes
Jul 9, 2025
Member
Author
|
/merge |
7c58187
into
rapidsai:branch-25.08
163 of 174 checks passed
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
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.
Description
Part of #12261
This PR refactors hash join to use
cuco::static_multisetin place ofcuco::static_multimap, eliminating the awkward use of cucopair_*APIs and improving overall readability.In terms of performance, it introduces up to a 20% slowdown for small datasets (e.g., ~1000 elements), but the impact diminishes as data size increases. While we should be mindful of this regression, there’s no need for alarm, the performance difference becomes negligible for larger workloads.
The slowdown has been traced to an increase in branching instructions (~15%), and addressing these extra branches on the cuco side should resolve the issue.
Checklist