I am inclined to believe that addition, subtraction, shifts, comparisons, and logical operations only take one clock cycle to execute on modern x86 processors. (Well, strictly speaking, it takes time to access the RAM, too, but the compiler will try to store variables in registers.) I see no reason why any of these instructions would benefit more from pipelining than any other. So I expect them to be equally fast and then it would be the number of these, not the kind, to matter. However, it is unlikely that trimming off a few of these would significantly improve your runtime. Better to worry about the cache; it can occasionally make the difference between AC and TLE.
Multiplication, division, and mod are certainly slower, but I would be very surprised if the compiler did not attempt to optimize integer multiplication by a constant by sometimes converting it into addition, subtraction, and shifts. For example, the compiler will probably replace "15*x" by "(x<
