rygorous

Why do compilers even bother with exploiting undefinedness signed overflow? And what are those
mysterious cases where it helps?

A lot of people (myself included) are against transforms that aggressively exploit undefined behavior, but
I think it's useful to know what compiler writers are accomplishing by this.

TL;DR: C doesn't work very well if int!=register width, but (for backwards compat) int is 32-bit on all
major 64-bit targets, and this causes quite hairy problems for code generation and optimization in some
fairly common cases. The signed overflow UB exploitation is an attempt to work around this.