costin commented on pull request #453: URL: https://github.com/apache/lucene/pull/453#issuecomment-973193877
Thanks to @jpountz I've added two more tests beside JMH. `PackedRandomTest` which does a brute force for testing performance of random locations instead of consecutive ones. Surprisingly the VHLongByte implementation comes ahead, followed by Packed64 and finally Packed64VHLongLong - the results on my machine are roughly: ``` 1. VHLongAndByte 939100 2. Packed64 1358500 3. VHLongLong 1510400 ``` Adrien suggested another workload, more realistic `PackedMacroTest` which does `SortedDocValues` merging, which creates an `OrdinalMap` backed by `PackedInts` and then recorded the merge times. In all cases I manually updated the PackedInt class to returned the desired implementation. Based on these set of results, Packed64 is on the first place, followed by VHLongAndByte with VHLongLong coming in third. I'm attaching the merging times found for reference: [p64vhll.log](https://github.com/apache/lucene/files/7565523/p64vhll.log) [p64.log](https://github.com/apache/lucene/files/7565524/p64.log) [p64vhlb.log](https://github.com/apache/lucene/files/7565525/p64vhlb.log) . -- 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
