You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by GitBox <gi...@apache.org> on 2022/02/13 15:11:17 UTC
[GitHub] [arrow-rs] alamb commented on pull request #1284: Vectorized DeltaBitPackDecoder (#1281)
alamb commented on pull request #1284:
URL: https://github.com/apache/arrow-rs/pull/1284#issuecomment-1038184104
🤔 I also ran the performance tests
On a google cloud machine (c2-standard-16, CPU Intel Cascade Lake)
`before`: https://github.com/alamb/arrow-rs/commit/3125344abbce5d7f7ca83cacc44b7eacf68caddb / https://github.com/alamb/arrow-rs/tree/alamb/delta-packed-reader-benches branch, based off master 936ed5e4762871e5fa889e76c95acc07b955f0c2 with the new benches
`after`: on this branch https://github.com/apache/arrow-rs/pull/1284/commits/f987da919314dbc75d5d6843f44bc89cad1f4c20 / https://github.com/tustvold/arrow-rs/tree/delta-packed-reader
I ran the benchmarks like this
```shell
cargo bench -p parquet --bench arrow_reader --features=test_common,experimental -- --save-baseline after
```
And then compared:
```shell
alamb@instance-1:/data/arrow-rs$ critcmp before after
group after before
----- ----- ------
arrow_array_reader/Int32Array/binary packed, mandatory, no NULLs 1.00 31.5±0.04µs ? ?/sec 3.15 99.1±0.22µs ? ?/sec
arrow_array_reader/Int32Array/binary packed, optional, half NULLs 1.00 45.8±0.18µs ? ?/sec 1.73 79.5±0.32µs ? ?/sec
arrow_array_reader/Int32Array/binary packed, optional, no NULLs 1.00 49.1±0.11µs ? ?/sec 2.38 116.8±0.28µs ? ?/sec
arrow_array_reader/Int32Array/dictionary encoded, mandatory, no NULLs 1.00 36.1±0.07µs ? ?/sec 1.02 36.8±0.08µs ? ?/sec
arrow_array_reader/Int32Array/dictionary encoded, optional, half NULLs 1.00 48.6±0.11µs ? ?/sec 1.01 49.0±0.16µs ? ?/sec
arrow_array_reader/Int32Array/dictionary encoded, optional, no NULLs 1.00 53.6±0.13µs ? ?/sec 1.02 54.7±0.10µs ? ?/sec
arrow_array_reader/Int32Array/plain encoded, mandatory, no NULLs 1.01 4.7±0.24µs ? ?/sec 1.00 4.7±0.11µs ? ?/sec
arrow_array_reader/Int32Array/plain encoded, optional, half NULLs 1.00 32.2±0.09µs ? ?/sec 1.01 32.5±0.19µs ? ?/sec
arrow_array_reader/Int32Array/plain encoded, optional, no NULLs 1.00 22.4±0.64µs ? ?/sec 1.03 23.0±0.35µs ? ?/sec
arrow_array_reader/Int64Array/binary packed, mandatory, no NULLs 1.00 40.5±0.05µs ? ?/sec 5.77 233.9±0.68µs ? ?/sec
arrow_array_reader/Int64Array/binary packed, optional, half NULLs 1.00 51.3±0.08µs ? ?/sec 2.87 147.0±0.45µs ? ?/sec
arrow_array_reader/Int64Array/binary packed, optional, no NULLs 1.00 58.3±0.06µs ? ?/sec 4.31 251.3±0.24µs ? ?/sec
arrow_array_reader/Int64Array/dictionary encoded, mandatory, no NULLs 1.00 37.5±0.07µs ? ?/sec 1.02 38.3±0.15µs ? ?/sec
arrow_array_reader/Int64Array/dictionary encoded, optional, half NULLs 1.00 49.5±0.16µs ? ?/sec 1.02 50.4±0.21µs ? ?/sec
arrow_array_reader/Int64Array/dictionary encoded, optional, no NULLs 1.00 54.7±0.11µs ? ?/sec 1.02 56.0±0.11µs ? ?/sec
arrow_array_reader/Int64Array/plain encoded, mandatory, no NULLs 1.00 7.9±0.44µs ? ?/sec 1.02 8.0±0.26µs ? ?/sec
arrow_array_reader/Int64Array/plain encoded, optional, half NULLs 1.00 34.3±0.12µs ? ?/sec 1.00 34.3±0.06µs ? ?/sec
arrow_array_reader/Int64Array/plain encoded, optional, no NULLs 1.01 26.3±1.57µs ? ?/sec 1.00 26.1±0.44µs ? ?/sec
arrow_array_reader/StringArray/dictionary encoded, mandatory, no NULLs 1.00 204.7±1.03µs ? ?/sec 1.01 205.8±0.93µs ? ?/sec
arrow_array_reader/StringArray/dictionary encoded, optional, half NULLs 1.00 226.6±2.75µs ? ?/sec 1.00 226.9±0.76µs ? ?/sec
arrow_array_reader/StringArray/dictionary encoded, optional, no NULLs 1.01 223.2±1.35µs ? ?/sec 1.00 221.9±1.55µs ? ?/sec
arrow_array_reader/StringArray/plain encoded, mandatory, no NULLs 1.00 213.5±0.77µs ? ?/sec 1.00 212.9±0.61µs ? ?/sec
arrow_array_reader/StringArray/plain encoded, optional, half NULLs 1.01 233.5±0.63µs ? ?/sec 1.00 230.7±0.96µs ? ?/sec
arrow_array_reader/StringArray/plain encoded, optional, no NULLs 1.00 235.8±1.39µs ? ?/sec 1.00 236.1±0.86µs ? ?/sec
arrow_array_reader/StringDictionary/dictionary encoded, mandatory, no NULLs - new 1.00 28.4±0.07µs ? ?/sec 1.01 28.8±0.08µs ? ?/sec
arrow_array_reader/StringDictionary/dictionary encoded, mandatory, no NULLs - old 1.00 1910.0±3.25µs ? ?/sec 1.00 1913.3±8.46µs ? ?/sec
arrow_array_reader/StringDictionary/dictionary encoded, optional, half NULLs - new 1.00 45.8±0.39µs ? ?/sec 1.02 46.7±0.10µs ? ?/sec
arrow_array_reader/StringDictionary/dictionary encoded, optional, half NULLs - old 1.00 1726.2±3.24µs ? ?/sec 1.00 1731.8±6.23µs ? ?/sec
arrow_array_reader/StringDictionary/dictionary encoded, optional, no NULLs - new 1.00 46.0±0.11µs ? ?/sec 1.01 46.6±0.20µs ? ?/sec
arrow_array_reader/StringDictionary/dictionary encoded, optional, no NULLs - old 1.00 1963.8±3.34µs ? ?/sec 1.00 1968.7±4.28µs ? ?/sec
```
There are some very nice improvements (3x - 5x) in binary packed decoding 👨🍳 👌
![Screen Shot 2022-02-13 at 10 10 15 AM](https://user-images.githubusercontent.com/490673/153759627-f2c698c9-5f88-494f-a9dc-70f4d03b2901.png)
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org