Compiler Optimizations Are Hard Because They Forget - Faultlore

Let’s try it out on our definitely real compiler: I’m done optimizing, the final program is (x << 1) << 1 HUH? Hold on let me look at the code I didn’t write: Ughhh our peephole passes run in the wrong order for this to work. Ok well we can just swap them around. But wait oh god if we add enough passes there might be no optimal ordering and we’ll be constantly missing “obvious” optimizations! Do w