This is exactly how the hardware works and what micro-ops are, on any system with a u-op cache or trace cache those decoded instructions are cached and used instead of decoding again. Unfortunately you still have to decode the instructions at least once first and that bottleneck is the one being discussed here. This is all transparent to the OS and not visible outside a low level instruction cache though, which means you don’t need a major OS change, but arguably if you were willing to take that hit you could go further here.