jpountz commented on PR #13186: URL: https://github.com/apache/lucene/pull/13186#issuecomment-2009068628
You are anticipating my next question, which was going to be about whether it helps speed up doc ID reordering. Your reasoning about why the swap logic is inefficient seems to assume that there would be lots of equal elements, but actually since we tie break on docID, there would never be any equal elements and phase 2 is a no-op in practice? > and also keep track of original positions I have a bias against this because it would further increase RAM requirements by 4 bytes (one integer) per doc ID? If we see evidence that swapping is a bottleneck, I'd be more keen to looking into using something like `RadixSelector`, which should perform less useless swapping. -- 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
