mapleFU commented on PR #55323: URL: https://github.com/apache/arrow/pull/55323#issuecomment-6631829556
On my Mac M2 Air, this code is faster. However, on my EPYC Zen4, this code might slower in some case: The bottleneck is xsimd's x86 batch_bool<int16_t>::mask(), which uses a lookup table; replacing it with packs_epi16 + movemask_epi8 makes it 2.5-3x faster than the existing AVX2 code with both compilers. See: 1. https://github.com/xtensor-stack/xsimd/blob/master/include/xsimd/arch/xsimd_sse2.hpp#L1380 2. https://github.com/xtensor-stack/xsimd/blob/master/include/xsimd/arch/xsimd_avx2.hpp#L746 -- 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]
