cyb70289 commented on PR #13661: URL: https://github.com/apache/arrow/pull/13661#issuecomment-1190968570
Looks great. I believe the big regression from some tests are not real. E.g., `arrow-bit-util-benchmark : BenchmarkBitmapVisitUInt8And/32768/0` drops from 13.983 GiB/s (O3) to 503.375 MiB/s (O2). Tested on my local host with clang-12, the result is 134MB/s, both O2 and O3. The huge gap is probably due to aggressive inline and optimization which makes the micro-benchmark far from reality. One catch is gcc -O2 disables vectorization, while clang -O2 keeps it. We may need additional -fxxxx if want to keep some useful features. -- 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]
