kosiew commented on PR #25358:
URL: https://github.com/apache/datafusion/pull/25358#issuecomment-6677221434
@alamb,
do these results justify continuing staged sort as a default-off
experimental feature?
- [Strong targeted
gains](https://github.com/apache/datafusion/pull/27128#issuecomment-6658082877):
14.1× for unique prefixes, 4.0× for partial
ties, 6.4× across multiple stages. All-tied case improved 1.16×.
- [Limited multi-run benefit:
1.08×](https://github.com/apache/datafusion/pull/27128#issuecomment-6658082877)
versus 4.0× for the same query
with a single-run-oriented configuration; merge-path work may limit
gains.
- [cheap-key and TopK controls slowed
~2%](https://github.com/apache/datafusion/pull/27128#issuecomment-6675126085).
TPC-H TopK aggregate timings stayed within ~1%, but
noisy per-query means flagged 8–17% slowdowns.
--
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]
