0lai0 commented on issue #5091:
URL:
https://github.com/apache/datafusion-comet/issues/5091#issuecomment-5303040304
Tried and reverted (bench regressed):
- temporal.rs::days_to_date → Date32Type::to_naive_date_opt. date_trunc
regressed +15% to +29% on three of
four shapes because chrono::NaiveDate::from_ymd_opt(1970, 1, 1).unwrap()
is not const, so the epoch is
reconstructed per row. The old const i32 offset let LLVM fold it. Worth
re-trying once the epoch becomes const
upstream.
- covariance.rs::{update_batch,retract_batch} → and(is_not_null,
is_not_null) + filter. Regressed sparse-null
shapes by +18% to +33% because filter allocates two new Float64Arrays per
batch. Only +12% win on the no-null
path, not enough to justify a per-batch heuristic.
--
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]
