Distinct left join #10520
Merged
Distinct left join #10520
Conversation
This file contains 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
Signed-off-by: Jason Lowe <[email protected]>
Signed-off-by: Jason Lowe <[email protected]>
jlowe
added
performance
revans2
previously approved these changes
Mar 4, 2024
revans2
approved these changes
Mar 5, 2024
|
build |
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Outdated suggestions cannot be applied.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.
Depends on #10503 and rapidsai/cudf#15154.
Updates the left hash join code to leverage the cudf distinct left join API where the join can be performed much more quickly when it is known there are no duplicate join keys in the hash table. Unlike distinct inner join, distinct left join does not require gathering the left table.
Running this on NDS at scale factor 3K resulted in no regressions but no statistically significant improvements either, probably because there aren't enough distinct left outer joins in practice on that benchmark. Microbenchmarks show this can be up to 35% faster than a non-distinct left outer hash join.