jonkeane commented on pull request #9615: URL: https://github.com/apache/arrow/pull/9615#issuecomment-840105430
Yes! I reran the benchmarks again (comparing the last commit here with the base commit and `apache/arrow@HEAD` for today and I see some drastic improvements for floats and ints (and the improvements we saw before with dict are still there of course). The naturalistic datasets aren't seeing much (if any) speed up — they are all within the noise range for variability that we see here. I'm going to dig into those separately and see if I see any funny patterns there that might explain it. [parallel-data-conversion.html.zip](https://github.com/apache/arrow/files/6469138/parallel-data-conversion.html.zip) -- 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. For queries about this service, please contact Infrastructure at: [email protected]
