Call stack

When you call a function, the compiler will create a stack frame for the new function call and it will push items on to the stack. The data put on the stack depends on your compiler and whether the code is compiled for the debug or release build; however, in general there will be information about the parameters passed to the function, the return address (the address after the function call), and the automatic variables allocated in the function.

This means that, when you make a function call at runtime, there will be a memory overhead and performance overhead from creating the stack frame before the function runs, and a performance overhead in cleaning up, after the function completes. If a function is inlined, this overhead does not occur because the function call will use the current stack frame rather than a new one. Clearly, inlined functions should be small, both in terms of code and the memory used on the stack. The compiler can ignore the inline specifier and call the function with a separate stack frame.