I think a more realistic thing would be adding a few custom OPs to make binary rewriting more performant. Another interesting thing would be implementing a sub

I think a more realistic thing would be adding a few custom OPs to make binary rewriting more performant. Another interesting thing would be implementing a subset of RVV with AVX512 hardware primitives, although idk if/how they are exposed in micro code, amd you likely would need to use a different instruction encoding.