[GitHub] [arrow] alamb commented on pull request #8664: ARROW-10588: [Rust] Safe bit operations for Arrow

[GitBox <gi...@apache.org>]

alamb commented on pull request #8664:
URL: https://github.com/apache/arrow/pull/8664#issuecomment-729965696


   I ran the filtering selectivity benchmark on my laptop again and this time it shows a ~400% slowdown... So I wonder if that unsafe code is actually giving us a performance gain?
   
   results on master @ 0e2681fe52ea48cade10171a018554da05f95285
   ```
   cargo bench -- "filter u8 high selectivity"
   
   filter u8 high selectivity
                           time:   [3.5905 us 3.6379 us 3.6942 us]
                           time:   [3.3929 us 3.4179 us 3.4457 us]
                           time:   [3.3876 us 3.4167 us 3.4470 us]
                           time:   [3.4520 us 3.4747 us 3.4981 us]
                           time:   [3.4288 us 3.4538 us 3.4820 us]
   
   ```
   
   Then on this PR at @457d8c7303864673c25bcb4e95e002b9585e9c39
   
   ```
   filter u8 high selectivity
                           time:   [16.951 us 17.164 us 17.395 us]
                           time:   [16.649 us 16.886 us 17.136 us]
                           time:   [17.175 us 17.751 us 18.520 us]
                           time:   [16.028 us 16.109 us 16.196 us]
                           time:   [16.224 us 16.341 us 16.456 us]
   ```
   
   Which does seem like a substantial and significant performance regression.
   
   I double checked that I could reproduce the faster master numbers and I can
   


----------------------------------------------------------------
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:
users@infra.apache.org