admin管理员组

文章数量:1287774

I read a few years ago that gather and scatter in AVX 512 weren't very efficient. Has this changed in AMD's implementation with the Zen5 (Granite Ridge/Turin) architecture? ie Ryzen 9000 desktop or Epyc server processors.

I was looking into performance improvements for histogramming a few years ago and am wondering if it is time to revisit the benefit of the scatter intrinsics.

Thanks!!

本文标签: histogramDoes AVX512 scatter work efficiently on Zen 5Stack Overflow