Accelerate conditional inner joins with larger right tables #9523
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
vyasr
added
3 - Ready for Review
Benchmarks
Before: After: |
Codecov Report
@@ Coverage Diff @@
## branch-21.12 #9523 +/- ##
================================================
- Coverage 10.79% 10.67% -0.12%
================================================
Files 116 117 +1
Lines 18869 19716 +847
================================================
+ Hits 2036 2104 +68
- Misses 16833 17612 +779
Continue to review full report at Codecov.
|
|
@vyasr I ran with this patch locally and it does what we need. Before the patch: After the patch: I can reproduce similar fast times than with our spark-side prototype I had discussed here #9461. Thanks! |
abellina
approved these changes
Oct 28, 2021
Thank you for finding all my find-and-replace and copy-paste errors! |
|
@gpucibot merge |
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Outdated suggestions cannot be applied.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.
This PR introduces reconfigures the conditional join kernels to launch one thread for each row in the larger of the two tables rather than always launching one thread for each row in the left table. Swapping the table on which the kernel is ordered helps improve performance in cases where the right table is significantly larger than the left table.
This PR is related to #9461 but it does not completely address its since this change only works for inner joins. Other join kinds are less straightforward since they require keeping track of rows for which no matches were found, which is easy to do if all the work for a given row is performed on the same thread (the current approach) but will require a significantly modified or entirely new kernel to also incorporate the inter-thread communication needed if the table ordering is swapped.