AMD has begun their open-source compiler enablement upstreaming effort for Zen 6 processors! The first “Znver6” patch was sent out on Friday in preparing for new instructions to be found with these next-generation AMD Ryzen and EPYC processors.
The GNU Binutils patch plumbing the new “Znver6” target also confirms the new CPU ISA additions over current-generation Zen 5 processors. On top of Znver5, the new Znver6 target adds: AVX512_BMM, AVX_NE_CONVERT, AVX_IFMA, AVX_VNNI_INT8, AVX512_FP16. Namely more Advanced Vector Extensions (AVX) additions and continuing to build out their AVX-512 offerings inline with Intel.
Most interesting is AMD’s new AVX-512 BMM. The new AVX-512 BMM instructions are new bit manipulation instructions for matrix multiply and bit reversal. Besides AVX-512 BMM (AVX512_BMM), the other ISA features have been found with other Intel (Xeon) processors.
The GNU Binutils patch can be found on the mailing list for AMD Zen 6. Hopefully the GCC compiler and LLVM/Clang patches for Zen 6 aren’t too far behind now. Great seeing them get out this support early so will hopefully be part of GCC 16 and LLVM/Clang 22 ahead of AMD Zen 6 processors shipping later in 2026.
