zhaih commented on PR #12660: URL: https://github.com/apache/lucene/pull/12660#issuecomment-1774469691
> I would be curious to see the contention times and also understand how this changes CPU usage vs. single-threaded. @msokolov as for CPU usage, I just tested with 1M docs, and on my laptop I can see when it's doing forceMerge it constantly ranges from 790~810%, while if a merge triggers when adding the document, the CPU can goes to 900%. I would like to see how it scales to, say 20 or even 40 cores, but unfortunately this laptop with 10 cores (including 2 efficient core) is the only one I have for now... BTW the forceMerge of the 1M index took `362125 ms`, while the contention time for that merge was `81847 ms` in CPU time, so average `10230 ms` wall time and thus `~2.8%` of total time. And array copy time was `12151 ms` in CPU time, and average `1519 ms` wall time and thus `~0.4%` of total time. So you're right, comparing with result of 100k documents, the contention overhead goes down when the graph is getting bigger. -- 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
