MMX these days costs extra instructions on both AMD and Intel CPUs. This has been the case since SSE and SSE2 were introduced and later AVX extensions. There q

MMX these days costs extra instructions on both AMD and Intel CPUs. This has been the case since SSE and SSE2 were introduced and later AVX extensions. There quite possibly been a point in time where certain MMX instructions run faster on some AMD CPUs and required less uOps to achieve. https://agner.org/optimize/ https://agner.org/optimize/ This is a pretty good source for uOps and instruction cycles.