josiahyan edited a comment on pull request #8214:
URL: https://github.com/apache/arrow/pull/8214#issuecomment-696452697


   @lidavidm @liyafan82 @jacques-n 
   Interpreting the results:
   This patch could be improved (performance wise) by more aggressive caching 
(option 3), at the potential expense of additional state, for an additional 35% 
speedup above this submitted patch. @liyafan82 @lidavidm I was wrong about the 
performance gain - I guess this speedup might be worth having? Please let me 
know if you prefer this variant, subject to the given tradeoff. I'll clean it 
up and resubmit if so.
   
   An append-only interface, trivially implemented, but assuming buffer 
ownership (Option 2), could get (up to) an additional 54% above the improved 
version of this patch.
   
   A type specialized variant, presumably created by templating out the various 
types (Option 4), could get (up to, these are very very rough results, it 
hasn't been verified where this speedup is coming from, it might just be the 
bit setting code) 18% above the specialized append-only interface.
   
   It might be possible to dive into the ASM to figure out what's actually 
being generated, but that might be a tad extreme for now (and unstable).


----------------------------------------------------------------
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


Reply via email to