uschindler commented on issue #15079: URL: https://github.com/apache/lucene/issues/15079#issuecomment-3213229338
Hi, I got the new results, from what I see at first: - many boolean queries got an improvement and restored to the "pre-pollution" state. So It looks like the GroupVInt Varhandle Change and removal of the Lambda helped a lot (see also the problem with the benchmark that @jpountz discovered). Actually in a micro-.benchmark the VarHandle change has a bit of negative impact (but still much faster than the plain group vint!). See #15104 for details. IMHO, the change is good, although Adrien sees a slowdown (which may be caused by too-short warmup time). <img width="1335" height="614" alt="Image" src="https://github.com/user-attachments/assets/a2f06764-8dae-4ff6-b614-a84c69f87679"; /> - I don't see to much change in the indexing time with KNN. I hjave not yet looked at allocation profile. But to me it is a bit slower than before the pollution added by @jpountz. <img width="3055" height="1282" alt="Image" src="https://github.com/user-attachments/assets/16635732-2f44-4a2f-a5a6-21d55910ef60"; /> To me the change looks fine, or am I misinterpreting something? -- 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
