shatejas commented on PR #13985: URL: https://github.com/apache/lucene/pull/13985#issuecomment-2474147590
> I'm curious how much this actually helps, and I know that you said that benchmark results would be posted. @ChrisHegarty Preliminary results showed approximately 40mins (~13%) reduction in total force merge time for 10m dataset. You should be able to find details here https://github.com/opensearch-project/k-NN/issues/2134#issuecomment-2420793541. Though the setup does not use LuceneHNSW it does use Lucene99FlatVectorsReader. > What is unclear is how much the sequential advise here helps over something like a load (similar to MemorySegment::load), that touches every page. I did try preload which does `MemorySegment::load` and there was an improvement in force merge time. The advantage I see with updateAdvice is being able to delay the loading of vectors till needed. -- 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
