neilconway opened a new pull request, #23984: URL: https://github.com/apache/datafusion/pull/23984
## Which issue does this PR close? - Closes: #23983 ## Rationale for this change `PartitionWindowAggStates` and `PartitionBatches` were using IndexMap's default SipHash hasher. Switching to foldhash yields a small but measurable performance improvement. Benchmark below; there is a small but measurable performance win. Note that the improvement is modest in part because linear mode is very inefficient when there are many partitions (#23982); I will send a PR for that separately. Benchmarks: - linear, 10,000 partitions: 216.7 ms -> 204.1 ms (~6%) - linear, 100 partitions: 45.0 ms -> 45.4 ms (within drift) - sorted, 10,000 partitions: 35.0 ms -> 34.8 ms (control) ## What changes are included in this PR? * Add new benchmark * Switch from using SipHash to foldhash ## Are these changes tested? Yes, covered by existing tests. ## Are there any user-facing changes? No functional change at all. -- 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]
