ggml : support AVX512VNNI (#6280)

Commit

1 year ago

ggml : support AVX512VNNI (#6280) This change causes some quants (e.g. Q4_0, Q8_0) to go faster on some architectures (e.g. AMD Zen 4).