I'll have a look for the discussion. Any idea if the branches are more expensive than some multiplies? step = (a<=b)?1:-1; for (i = a; step*i < step*b;

I'll have a look for the discussion. Any idea if the branches are more expensive than some multiplies? step = (a<=b)?1:-1; for (i = a; step*i < step*b; i += step) { doSomething(); } I guess if the branch predictor is any good it could be slower with multiplies, but I have no idea what to expect in a dynamic language.