mccullocht commented on issue #15697: URL: https://github.com/apache/lucene/issues/15697#issuecomment-3929444243
The "unpacked" vector is the query vector and is likely to be in the L1 cache for the duration of the query. A load from the L1 cache should be a speed bump -- a cycle or two at most. The packed vector is where you are most likely to miss on caches and we read that one linearly, although TBH issuing reads to multiple locations might be helpful for hiding memory latency. It is troubling that vector performance is not substantially better than scalar performance in this case. -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
