内联函数中的分支

Question

I think I have a serious compiler mistrust. Do branches inside inline functions get optimized out if they have constant results?

For the example function:

#define MODE_FROM_X_TO_Y 0
#define MODE_FROM_Y_TO_X 1

inline void swapValues(int & x, int & y, int mode) {
    switch(mode) {
        case MODE_FROM_X_TO_Y:
            y = x;
            break;
        case MODE_FROM_Y_TO_X:
            x = y;
            break;
    }
}

Would:

swapValues(n,m,MODE_FROM_X_TO_Y);

be optimized as:

n = m;

Answer 1

First off, it won't even compile (until you add a return type).

Secondly, swap is an awesomely badly chosen name (since it doesn't do a swap , and conflicts with the std::swap name).

Thirdly, head over to http://gcc.godbolt.org/ :

Live On Godbolt

Answer 2

Generally speaking, the answers to these questions are compiler dependent.

To get an answer to your question with your code, with your compiler (and compiler version), and compiler settings (eg optimisation flags) you will need to examine the code output by the compiler.

Branches in any code - not just within inlined functions - can potentially be "optimized out" if the compiler can detect that the same branch is always followed.

Some modern compilers are also smart enough to not inline a function declared inline if it evaluates another function as a better candidate for inlining. A number of modern compilers can do a better job making such decisions than a typical C/C++ programmer.