cyb70289 commented on pull request #9635: URL: https://github.com/apache/arrow/pull/9635#issuecomment-791325415
List benchmark comparison of non-null case, which drops most. See PR comment for performance analysis. gcc-7.5 ``` Before ------ SumKernelFloat/1048576/0 29.6 us 29.6 us 23768 bytes_per_second=33.0098G/s null_percent=0 size=1048.58k SumKernelDouble/1048576/0 19.8 us 19.8 us 35733 bytes_per_second=49.3665G/s null_percent=0 size=1048.58k SumKernelInt8/1048576/0 242 us 242 us 2895 bytes_per_second=4.03683G/s null_percent=0 size=1048.58k SumKernelInt16/1048576/0 55.9 us 55.9 us 12446 bytes_per_second=17.4673G/s null_percent=0 size=1048.58k SumKernelInt32/1048576/0 28.8 us 28.8 us 24615 bytes_per_second=33.9168G/s null_percent=0 size=1048.58k SumKernelInt64/1048576/0 22.9 us 22.9 us 30336 bytes_per_second=42.6622G/s null_percent=0 size=1048.58k After ----- SumKernelFloat/1048576/0 115 us 115 us 6113 bytes_per_second=8.49307G/s null_percent=0 size=1048.58k SumKernelDouble/1048576/0 47.5 us 47.5 us 14814 bytes_per_second=20.5647G/s null_percent=0 size=1048.58k SumKernelInt8/1048576/0 358 us 358 us 1957 bytes_per_second=2.72924G/s null_percent=0 size=1048.58k SumKernelInt16/1048576/0 155 us 155 us 4531 bytes_per_second=6.31986G/s null_percent=0 size=1048.58k SumKernelInt32/1048576/0 53.9 us 53.9 us 12996 bytes_per_second=18.1266G/s null_percent=0 size=1048.58k SumKernelInt64/1048576/0 36.2 us 36.2 us 19274 bytes_per_second=26.9443G/s null_percent=0 size=1048.58k ``` clang-9 ``` Before ------ SumKernelFloat/1048576/0 28.9 us 28.9 us 24207 bytes_per_second=33.7537G/s null_percent=0 size=1048.58k SumKernelDouble/1048576/0 24.1 us 24.1 us 29240 bytes_per_second=40.5844G/s null_percent=0 size=1048.58k SumKernelInt8/1048576/0 60.9 us 60.9 us 11492 bytes_per_second=16.0346G/s null_percent=0 size=1048.58k SumKernelInt16/1048576/0 35.8 us 35.8 us 19570 bytes_per_second=27.284G/s null_percent=0 size=1048.58k SumKernelInt32/1048576/0 23.9 us 23.9 us 29181 bytes_per_second=40.8518G/s null_percent=0 size=1048.58k SumKernelInt64/1048576/0 55.9 us 55.9 us 12516 bytes_per_second=17.4578G/s null_percent=0 size=1048.58k After ----- SumKernelFloat/1048576/0 139 us 139 us 5053 bytes_per_second=7.04828G/s null_percent=0 size=1048.58k SumKernelDouble/1048576/0 46.5 us 46.5 us 14836 bytes_per_second=20.9803G/s null_percent=0 size=1048.58k SumKernelInt8/1048576/0 322 us 322 us 2173 bytes_per_second=3.03406G/s null_percent=0 size=1048.58k SumKernelInt16/1048576/0 71.0 us 71.0 us 9839 bytes_per_second=13.7602G/s null_percent=0 size=1048.58k SumKernelInt32/1048576/0 40.9 us 40.9 us 17058 bytes_per_second=23.8892G/s null_percent=0 size=1048.58k SumKernelInt64/1048576/0 45.1 us 45.1 us 15500 bytes_per_second=21.6414G/s null_percent=0 size=1048.58k ``` ---------------------------------------------------------------- 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: us...@infra.apache.org
