cyb70289 commented on issue #6986: URL: https://github.com/apache/arrow/pull/6986#issuecomment-616915079
@pitrou @wesm Oops, I only checked case "BitmapReader" from benchmark [arrow-bit-util-benchmark](https://github.com/apache/arrow/blob/master/cpp/src/arrow/util/bit_util_benchmark.cc). Obviously it's not enough. I compared all cases just now and see huge performance drop from below 4 tests: Before this patch: ```bash BenchmarkBitmapAnd/32768/1 563496 ns 563260 ns 1243 bytes_per_second=55.4806M/s BenchmarkBitmapAnd/131072/1 2219810 ns 2218984 ns 318 bytes_per_second=56.3321M/s BenchmarkBitmapAnd/32768/2 561738 ns 561467 ns 1265 bytes_per_second=55.6577M/s BenchmarkBitmapAnd/131072/2 2246229 ns 2245119 ns 305 bytes_per_second=55.6763M/s ``` After this patch: ```bash BenchmarkBitmapAnd/32768/1 1653467 ns 1652680 ns 422 bytes_per_second=18.9087M/s BenchmarkBitmapAnd/131072/1 6665501 ns 6661561 ns 105 bytes_per_second=18.7644M/s BenchmarkBitmapAnd/32768/2 1670793 ns 1670246 ns 423 bytes_per_second=18.7098M/s BenchmarkBitmapAnd/131072/2 6702369 ns 6698957 ns 103 bytes_per_second=18.6596M/s ``` Before reverting this patch, I would like to understand why it happens. BTW: we definitely need continuous benchmark tools to detect these things early. ---------------------------------------------------------------- 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]
