miniblog.

Rust is doing constant propagation on its internal IR before monomorphisation. This can give faster compiles because LLVM is given less code!

I used to think DCE and constant propagation were pointless: eg clean code shouldn't have dead sections. But inlining creates opportunities!

What does the 80/20 rule look like in compiler optimisation? I suspect basic constant propagation and DCE gets you a respectable speedup.