dungba88 commented on PR #14226:
URL: https://github.com/apache/lucene/pull/14226#issuecomment-2710421110

   > this is where we first search pro-rated across segments, and then if any 
segment has its output queue full with competitive hits, we revisit it using 
the prior hits as entry points and a queue size of K (the full original K).
   
   Yes that's correct
   
   > My only question is whether lambda=2 is correct - the graphs show lambda 
6+ looking more reasonable, and indeed they don't seem to show a data point for 
2 at all?
   
   I've benchmarked with the minimum of 3 so there is no data points for 2. For 
lambda=3, it's interesting that the simplified version is slower than the 
optimistic version. This is (somewhat) expected as with the simplified version 
on the second pass we immediately use the queue size of k (k=100 here) while 
the optimistic version will gradually increase the queue size, so that it may 
end up with a size much less than k. When we increase lambda, the difference 
becomes diminished as there is less re-entry needed and the queue size 
difference would also be smaller.


-- 
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