zhuqi-lucas commented on PR #7537:
URL: https://github.com/apache/arrow-rs/pull/7537#issuecomment-2906792622

   > > > Ok, the latest benchmark result I think are now better and show no 
regression thanks to @zhuqi-lucas 's suggestion. I will try @Dandandan 's idea 
to use a Vec and see if that helps
   > > 
   > > 
   > > Thank you @alamb, it's a good news!
   > 
   > This PR has shown me that for some queries the dispatch logic for 
RowSelection is quite high (as in just doing an extra compare in that loop made 
a measurable difference).
   > 
   > @zhuqi-lucas in your testing, did you measure where the cutoff for using a 
bitmap vs a RowSelector was? I think I remember seeing a value of `10` somewhere
   
   I agree @alamb , i was testing with 10 for cutoff for using a bitmap vs a 
RowSelector, it's a very basic cutoff:
   
   avg_size_of_selector = total row /  selectors
   
   if avg_size_of_selector > 10  using selector
   
   if avg_size_of_selector  <= 10 using bitmap
   
   
   And the default is selector because i use it to compute avg_size_of_selector.
   


-- 
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-unsubscr...@arrow.apache.org

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org

Reply via email to