jpountz commented on PR #12664: URL: https://github.com/apache/lucene/pull/12664#issuecomment-1759583212
I tried to re-inline advance() but it still suggested a small slowdown for `HighTermDayOfYearSort` and `CountAndHighMed` too so I suspect that there is something else. There is indeed one more check when one needs to move to a different block, but I would expect the cost to be dominated by the decoding of a full block in that case so that it shouldn't make a big difference. The slowdown on `CountOrHighHigh` and `CountOrHighMed` reported above should be either noise or due to the JVM compiling things differently because of other changes, given that `PostingsEnum#nextDoc()` wasn't changed. I'm tempted to not think too hard about `HighTermDayOfYearSort` and `CountAndHighMed` since there are more queries that get faster than slower with this change. While it's not big, I would expect the speedup on `CountAndHighHigh` to be real since it generally needs to advance postings by small increments, which is what the change to `PostingsEnum#advance` is supposed to help with. -- 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
