Hey, I was just profiling a Rust routine that’s pulling in about a 1.5 ms penalty per loop, and I’m staring at cache lines and branch prediction. Ever dug into the micro‑timing of your loops to squeeze out those extra milliseconds?