This sounds nice and clean and theoretic, but in practice you get better results by matching the design to the workload. Computers aren't new, we know what the workload might be. Javascript is a couple of decades old. The challenge is getting compilers to use it - which is much easier with non-AOT languages, where you can update the compiler for existing code without involving the developer.

Otherwise you end up using too many instructions to achieve simple things. Sometimes different semantics can be much slower; you can see this when emulating architecture A on architecture B, and why Apple's switchable memory-ordering semantics are absolute genius.

Also, this is why GPUs and DSPs exist.