benwtrent commented on PR #14226:
URL: https://github.com/apache/lucene/pull/14226#issuecomment-2660166393

   > The latency improvement from re-using scores is surprisingly large. I 
would have expected costs to be dominated by newly-explored nodes, but this is 
cool.
   
   @msokolov this was only over 10 different query iterations on my laptop 😅 
so, I wouldn't necessarily trust my latency numbers :D.
   
   But, the visited count is attractive. The magic numbers are just weird to 
me. Why is `16` so special?
   
   FWIW, I haven't tested this on anyother datasets yet. 
   
   If we want to provide entry-point scores, I can do that refactor in a 
separate PR. It isn't so bad, and is generally useful (we score entry point 
twice on every search, which is a minimal cost, but we could prevent doing 
that).


-- 
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: issues-unsubscr...@lucene.apache.org

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


---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org

Reply via email to