rmuir commented on issue #14042: URL: https://github.com/apache/lucene/issues/14042#issuecomment-2548442349
I think we are having a communication issue over terminology. I don't care about unrolling, i care about superscalar execution. JVM doesn't allow it, which means the hardware sits there idle and wasted. so we unroll the code to make this parallelism possible. Does it make sense? The "unrolling" the JVM does where it uses same registers over and over and forces serial execution is something irrelevant for this purpose. I totally see why they don't do this, besides the fact there is no "fast math" type flag I am aware of, it would be insanely confusing in general to start getting different answers after JIT compilation "changed". -- 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
