rmuir commented on issue #12621: URL: https://github.com/apache/lucene/issues/12621#issuecomment-1747002969
the type conversions are what makes it slow. for float case it is the equiv of: ``` float x = something; float y = something; float z = something; // no conversions float result = x + y * z; ``` for the binary case it is the equivalent of: ``` byte a = something; byte b = something; byte c = something; // multiply b + c avoiding overflow short z1 = (short)b * (short)c; // add a + z1 avoiding overflow int z = (int)a + (int)z1; ``` You can understand the limitations at the hardware level better by reading https://en.wikichip.org/wiki/x86/avx512_vnni `VPDPBUSD` is currently not supported by openjdk either: https://bugs.openjdk.org/browse/JDK-8215891 -- 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
