Re:Zen4's AVX512 Teardown
by ComputerGuru on Monday October 03, 2022 @08:34PM
(#33075419)
Because it's hard to say "conditionally enable AVX-512 only on platforms supporting it BUT not on platforms where it actually brings performance penalties."
Can you give a real example of AVX-512 actually causing a net performance penalty on any CPU?
The only way I can see that happening is using AVX-512 in small, infrequently called functions such as strcmp, and the solution is: don't do that.
If proper SIMD code runs for say 1ms at a time, it's pretty much guaranteed to benefit from any implementation of AVX-512.
Why is that sentence present tense instead of past tense? I suppose it continues to be a problem for Intel, but your comment appears to be presenting Intel downsides as if they were universal. Zen 4 apparently implements AVX-512 efficiently, without the problems Intel implementations experienced. That's what this whole discussion is about, and that's what Phoronix found as well.[0]
Hopefully Intel will catch up to AMD on AVX-512, but in the mean time, people optimizing software should be aware that AVX-512 has few (if any) downsides on certain platforms. Phoronix found zero performance penalty, but perhaps more testing is required.
[0]: https://www.phoronix.com/review/amd-zen4-avx512/6