mikemccand commented on issue #12358: URL: https://github.com/apache/lucene/issues/12358#issuecomment-1585597128
> The default G1GC works better when you hammer a server multithreaded, but if there's only one thread doing queries, ParallelGC is better. Ahh great catch @uschindler! I just [made that change](https://github.com/Tony-X/search-benchmark-game/commit/4402d42c906830e85d8d79a30ae776f204ade770) in the benchy. > Hi. It looks like the DoQuery.java code does not do a parallel throughput measurement, but instead it runs all queries in a single thread one after each other with Nanotime before and after (thanks for the fix, Mike) Right -- we are only measuring single threaded query latency (just like luceneutil). I agree a red-line test (saturate CPU) is also really important -- we have [an issue](https://github.com/Tony-X/search-benchmark-game/issues/19) open for this. It's a bit trickier to do... Note that neither tantivy nor Lucene is using more than one thread for each query, nor any OS level concurrency like async IO (I think?), so the "lightly loaded single query latency" metric should be fair to compare across both engines. -- 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
