[clang] [X86][AMX-AVX512] Add required/min vector width 512 constraint (PR #173532)

Phoebe Wang via cfe-commits cfe-commits at lists.llvm.org
Thu Dec 25 03:00:01 PST 2025


phoebewang wrote:

> Why diamondrapids still prefer 256-bit?

There are some discussions about preferrering 512-bit on diamondrapids. We'll check on HW and change if it behaves better.

https://github.com/llvm/llvm-project/pull/173532


More information about the cfe-commits mailing list