It's not so much branch prediction as the increase of CPU speed relative to bus speed. Loop unrolling increases code size. Unless you're running a lot of iter

It's not so much branch prediction as the increase of CPU speed relative to bus speed. Loop unrolling increases code size. Unless you're running a lot of iterations of a very tight loop, fetching the extra code from RAM will take longer than unrolling will save you. If you are running a lot of iterations of a very tight loop, it might still be worth trying. Just remember to profile.