Yes, if you have two AVX512 FMA units, you use them for GEMM; with a current microarchitecture implementation, if you only have one, you don't. https://github.c

Yes, if you have two AVX512 FMA units, you use them for GEMM; with a current microarchitecture implementation, if you only have one, you don't. https://github.com/flame/blis/blob/2d8ec164e7ae4f0c461c27309dc1f5d1966eb003/frame/base/bli_cpuid.c#L173 https://github.com/flame/blis/blob/2d8ec164e7ae4f0c461c27309... GEMM is one of relatively few computations with sufficient computational intensity.