uschindler commented on PR #912: URL: https://github.com/apache/lucene/pull/912#issuecomment-1317618039
Could be. It may be caused by too much coping of very small arrays (especially float and long) between offheap and heap. For explanation see above. I don't think this is a huge problem because it mainly comes from missing optimizations in the JVM and not our code. Could look different in Java 20. My only guess where it might come from: The overhead of copyMemory is huge for small arrays. ByteBuffer internally loops for very short arrays. We could try to call super.readLongs() if array is short (same for floats). But before doing this I would like to figure out what exactly slows. If I could get a profile only from this test or somebody could explain what mmap dir actions are used mainly in those facets. -- 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
