In certain cases, this level of optimization makes perfect sense. In other cases, arguably more common when writing C++, it makes more sense to invest that time

In certain cases, this level of optimization makes perfect sense. In other cases, arguably more common when writing C++, it makes more sense to invest that time in understanding the object model, i.e. what actually happens when you pass a struct by value, return an object (NRV), where invisible object copying occurs (STL, I'm looking at you) or similar issues. Lippman's "C++ Object Model" is a good reference for that.