If the branch is predictable, and the cpu guesses correctly, then it is faster than integer arithmetic. A branch miss on the other hand is much more expensive,

If the branch is predictable, and the cpu guesses correctly, then it is faster than integer arithmetic. A branch miss on the other hand is much more expensive, about 14 cycles last time I was testing (on a core i5 desktop cpu from a few years ago).