tang-hi commented on PR #12417: URL: https://github.com/apache/lucene/pull/12417#issuecomment-1628889057
I tested the code for vectors and scalars on wikimediumall. Comparing them to the baseline, it shows that neither of them performs better than the baseline. I found that the bottleneck in performance is not in the encoding and decoding, but rather in the prefix sum calculation. The baseline code uses some tricks, such as using a long to store two ints, which allows it to calculate the prefix sum faster. I will post the performance test results below. I have to say that the implementation of PforUtil in the baseline is indeed very clever. My question now is whether we should keep the original compression format or continue seeking a faster prefix sum. Because currently, I don't have any optimization ideas for the current PR. If you have any good ideas, please feel free to suggest. -- 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
