tang-hi commented on PR #12417: URL: https://github.com/apache/lucene/pull/12417#issuecomment-1629033196
> It's silly to represent the decoded values as int[], and then coerce then into long[] for prefix sum - we should just vectorise prefix sum to work on int[] (in a reasonable way) Yes, I realized this issue and then tried to directly sum the int[] using a simple for loop. Comparing it with the baseline using jmh, the effect doesn't seem to be much better. Of course, I haven't tested how much improvement there will be in luceneutil's bench after making this change. Currently, I have only found two places that need to be modified, namely the two occurrences of decodeTo32 in PForUtil. Maybe later we can directly sum and then see what the benchmark looks like. -- 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
