pitrou commented on issue #43687: URL: https://github.com/apache/arrow/issues/43687#issuecomment-2290854565
So, first a general reminder: these SIMD-specialized variants rely on autovectorization. I don't think we have any explicit SIMD code (either by way of intrinsics or using xsimd) in our compute kernels, though I may misremember. Now to the question that is asked here, I think the answer is: we don't know. Perhaps the person who contributed this had access to a AVX512 machine and decided that AVX512 was neutral or even detrimental for those data types? It could be worth redoing such measurements, as CPUs have changed quite a bit since then (especially, the CPU frequency penalty for running AVX512 code may be significantly smaller nowadays). -- 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]
