bharadwaj-pendyala commented on PR #11257: URL: https://github.com/apache/arrow-rs/pull/11257#issuecomment-5945225607
Both reruns agree on the only-valid rows: 2.6x at kept 1/4, 3.1-3.2x at 1023/2048, 1.8x at 1/2048. The nulls-at-end dense row came out worse than my PR description says, though. It went from 5.5-5.6 µs to 6.4 µs (14% and 18%), against the ~7% I measured on my M1. That row is the worst case on purpose, with the one selected null in the last word, so the scan reads both bitmaps in full before `filter_bits` reads them again. I've fixed the number in the description. -- 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]
