Question

I studied 32 bit machine, and I know sub $12, %esp as same as three time on pop %eax and mov $0, %eax.

But I don’t know why we use add $18, %rsp with start procedure call or sub $18, %rsp start with procedure call at x86-64 machine.

Why does compiler use this rule for procedure call? Argument rule?

0x0000000000400df3 <main+191>:  mov    $0x402227,%edi
0x0000000000400df8 <main+196>:  callq  0x400ab8 <puts@plt>
0x0000000000400dfd <main+201>:  callq  0x4014c0 <read_line>
0x0000000000400e02 <main+206>:  mov    %rax,%rdi
0x0000000000400e05 <main+209>:  callq  0x401139 <phase_3>

-----------------------------------------------------------------------
example:

0x0000000000401139 <example+0>: sub    $0x18,%rsp
0x000000000040113d <example+4>: lea    0x8(%rsp),%rcx
0x0000000000401142 <example+9>: lea    0xc(%rsp),%rdx
0x0000000000401147 <example+14>:        mov    $0x40248a,%esi
0x000000000040114c <example+19>:        mov    $0x0,%eax
0x0000000000401151 <example+24>:        callq  0x400ac8 <__isoc99_sscanf@plt>
0x0000000000401156 <example+29>:        cmp    $0x1,%eax

Answer 1

“I know sub $12, %esp as same as three time on pop %eax and mov $0, %eax.” You know wrong. sub $12, %esp does not change eax, neither by popping into it nor by moving a 0 into it.

The primary/direct reason for using something like sub $12, %esp is to reserve some space on the stack. That can be used to give some space to local variables of a subroutine. That can also be used to align the stack pointer if it’s desirable for performance or other reasons.

add $18, %rsp does the reverse, it releases the space previously allocated on the stack. This can be used to free the space occupied by local variables or by the parameters that have been previously passed to another subroutine and aren’t needed on the stack anymore.