emkornfield commented on PR #39397: URL: https://github.com/apache/arrow/pull/39397#issuecomment-1875772255
> My benchmark is just a reader based on low-level parquet api reading a single dataset prefix. As it's written using static reflection - it does not use much time. And then i look on perf report and watch asm code of long functions I think turning it into a real benchmark that we can monitor for regressions, without that there is a higher risk the change gets reverted. > Maybe we can draft a read-level benchmark in parquet-column-reader-benchmark. in current parquet-column-reader-benchmark we can find some optimization, but it's a bit indirect This sounds like a reasonable approach. Having benchmarks at different levels, for different data types makes a lot of sense to me. -- 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]
