Re:Decimating Array.Sort with AVX2
by vkaku on Sunday May 24, 2020 @03:50PM
(#23294165)
True story, I found AVX code to run faster than AVX2 despite half the overall vectorized inputs, turns out the chips clock down when running AVX2 instructions. Ditto with AVX-512 as well.
Unless they improve chip processes, reduce power consumption, reduce nm and increase density, AVX2 will not be a 100% effective, for a WHILE. This is my empirical opinion.
Unless they improve chip processes, reduce power consumption, reduce nm and increase density, AVX2 will not be a 100% effective, for a WHILE. This is my empirical opinion.