It's very interesting to see how those hash function finalizers are constructed/discovered behind the scene. It seems to me that instruction count is not the be

It's very interesting to see how those hash function finalizers are constructed/discovered behind the scene. It seems to me that instruction count is not the best metric (since slow instructions and fast instructions have drastically different speeds, and CPU internally can pipeline instructions), probably sorting the candidates by real clock time performance is more reasonable?