Comment by dang

Can anyone give an example of one of the optimizations that make this compiler so fast, and how it's connected to the whole-program approach—i.e., what invariants it's able to exploit that aren't available to more traditional compilers?