tyrelr commented on pull request #9215: URL: https://github.com/apache/arrow/pull/9215#issuecomment-766153344
The 10% performance impact looks like mostly improvements: ``` critcmp master-499b6d0c array-slice-78f360b7c -t 10 group array-slice-78f360b7c master-499b6d0c ----- --------------------- --------------- add 512 1.00 282.9±0.41ns ? B/sec 1.37 387.7±0.71ns ? B/sec add_nulls_512 1.00 344.7±0.54ns ? B/sec 1.30 449.3±1.08ns ? B/sec bench_primitive 1.42 3.1±0.00ms 1279.3 MB/sec 1.00 2.2±0.01ms 1815.1 MB/sec buffer_bit_ops or 1.00 590.9±0.95ns ? B/sec 1.14 673.4±0.93ns ? B/sec cast int64 to int32 512 1.15 2.5±0.02µs ? B/sec 1.00 2.2±0.01µs ? B/sec equal_bool_512 1.11 43.1±0.08ns ? B/sec 1.00 38.8±0.06ns ? B/sec like_utf8 scalar equals 1.00 72.5±0.88µs ? B/sec 1.22 88.4±0.18µs ? B/sec max nulls 512 1.00 1758.9±6.23ns ? B/sec 1.31 2.3±0.01µs ? B/sec min string 512 1.00 3.4±0.00µs ? B/sec 1.32 4.5±0.07µs ? B/sec multiply 512 1.00 335.6±0.67ns ? B/sec 1.19 398.6±5.74ns ? B/sec mutable str 1024 1.00 2.7±0.00ms ? B/sec 1.14 3.0±0.01ms ? B/sec subtract 512 1.00 345.3±1.26ns ? B/sec 1.12 385.0±0.77ns ? B/sec take bool nulls 1024 1.15 4.2±0.03µs ? B/sec 1.00 3.7±0.02µs ? B/sec ``` I don't believe that the list/binary/string array would be involved in most of the benchmarks listed here? (multiply, subtract, take bool, add, add_nulls, bench_primitive should be operations on boolean/primitive arrays, which have no functional changes)... ---------------------------------------------------------------- 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]
