dongjoon-hyun opened a new pull request, #58969:
URL: https://github.com/apache/spark/pull/58969

   ### What changes were proposed in this pull request?
   
   This PR adds a third shape to `AsOfJoinBenchmark`: 10K x 10K rows over 
**10** equi-key groups,
   between the existing `groups=100` and no-equi-key cases. `asOfJoinBenchmark` 
already takes
   `numGroups`, so this is purely a new invocation.
   
   ### Why are the changes needed?
   
   `SortMergeAsOfJoinScanner` buffers the right-side rows of an equi-key group 
and rescans that
   buffer once per left row, so the scan cost per left row grows with the group 
size. Neither
   existing case covers that: `groups=100` buffers only ~100 right rows per 
group, where the scan is
   swamped by the shuffle and the sort, and the no-equi-key case takes a 
different planning path
   (`AllTuples` distribution, `bufferAllRight` instead of `bufferRightGroup`).
   
   `groups=10` buffers ~1000 right rows per group on the ordinary equi-key 
path, making changes to
   the scan measurable.
   
   ### Does this PR introduce _any_ user-facing change?
   
   No. This is a test-only change to a benchmark class.
   
   ### How was this patch tested?
   
   Benchmark-only change. 
   
   ### Was this patch authored or co-authored using generative AI tooling?
   
   Generated-by: Claude Opus 5


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to