etseidl commented on PR #9831: URL: https://github.com/apache/arrow-rs/pull/9831#issuecomment-4383977021
So the regression is def there. Around 5% on average for the list_primitive's. I ran the list_primitive/default through samply, and it looks like a good bit of the slowdown is in `LevelInfoBuilder::write_leaf`. I'm seeing `extend` in main at 6100 samples, being replaced by `extend_from_iter` 3480, extend 3070, append_rep_level_run 1890 (total ~8400). Looking at the code, the only real difference is passing through all the `match`s to pick the right enum variant. Would it be possible to have different versions of `write_leaf` for each variant? Honestly I'm far more tolerant of write regressions than read. And 5% just for lists seems a fair trade. -- 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]
