is it certain in which register arguments and variables are stored?

Question

I'm still uncertain how registers are being used by the assembler

say I have a program:

int main(int rdi, int rsi, int rdx) {

    rdx = rdi;

    return 0;
}

Would this in assembly be translated into:

movq %rdx, %rdi 

ret rax;

I'm new to AT&T and have hard time predicting when a certain register will be used. Looking at this chart from Computer Systems - A programmer's perspective, third edition, R.E. Bryant and D. R. O'Hallaron:

charter

Answer 1

Is it certain in which register arguments and variables are stored?

Only at entry and exit of a function.

There is no guarantee as to what registers will be used within a function, even for variables which are parameters to the function. Compilers can (and often will) move variables around between registers to optimize register/stack usage, especially on register-starved architectures like x86.

In this case, a simple assignment operation like rdx = rdi may not compile to any assembly code at all, because the compiler will simply recognize that both values can now be found in the register %rdi. Even for a more complex operation like rdx = rdi + 1, the compiler has the freedom to store the value in any register, not specifically in %rdx. (It may even store the value back to %rdi, e.g. inc %rdi, if it recognizes that the original value is never used afterwards.)

Answer 2

No, it would be translated into:

mov %rdi, %rdx # move %rdi into %rdx
xor %eax, %eax # zero return value
ret # return

Of course, it's more than likely that rdx = rdi (and therefore mov %rdi, %rdx) will be removed by the compiler, because rdx is not used again.

Credit to @Jester for finding this out before me.