It really depends, to be honest. Donald Knuth uses a fictional assembly language in his Art of Computer Programming because he believes that any abstraction provides a place for bad analysis to hide under the covers. For instance, in your example, you use an addition operator whose implementation is undefined. It might be implemented as a string of increments and run in linear time. It might use a scheme of bit shifts and run in some logarithmic time. It might be implemented to run as a look-up table in constant time. The point is that you don't know what the processor's actually going to do, and this is what Knuth was getting at.

However, in nearly every analysis I've done, mathematical operations are assumed to run in constant time because it can cloud the overall picture of an algorithm's complexity with an implementation detail. Unless you're specifically working on the implementation of those operations, you're probably safe to assume that as well.