msokolov commented on PR #13153: URL: https://github.com/apache/lucene/pull/13153#issuecomment-1997659513
right that last change seemed promising! But on luceneutil tasks it didn't show much impact, basically a regression to teh mean compared to the first revision I tested; possibly a slight negative trend overall. I'll paste the results here. I was considering creating a more targeted benchmark to exaggerate the impact using a field and query type where we would expect to see a benefit, if only as a debugging tool, but I haven't had a chance to do that yet. Another idea I had was to try to help branch prediction by using a specialized BlockDecoder or so rather than embedding if statements in every advance/nextDoc call. I'm also not sure whether writing full longs versus truncating the final long in the bitset and writing only as many bytes as needed (added in the recent revision) might be making some difference? Seems unlikely. -- 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
