costin commented on PR #16691: URL: https://github.com/apache/lucene/pull/16691#issuecomment-6919486529
Thanks for looking into this. In #16129 PR, dedicated SIMD code for narrower widths was within noise to the C2 generated code which already vectorized the scalar loops well, hence why it was left out. The numbers in the description are hard to interpret since only 8-bit is covered and the hardware isn't stated yet the 65536 case is within noise. The PR also pins the output to 256-bit and there are no numbers for AVX-512. SIMD paths do pay off in some cases, but here it looks like diminishing returns and I'm inclined towards keeping the scalar code. -- 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]
