dholland@ wrote: > ok, so naively I would think that *reordering* blocks wouldn't change > their size, or at least not much... does anyone know where this 10% > overhead comes from? Is it all jump instructions?
Probably it moves and aligns reordered "branch taken" blocks at cacheline and there might be something wrong around MD parameters (but I have not checked at all). --- Izumi Tsutsui
