How can I substract two numbers (char to be exact) in assembly?

Asked Dec 08 '20 at 18:32

Active Dec 08 '20 at 20:08

Viewed 169 times

I'm trying to re-write the function strcmp in assembly.

I wrote this code:

global _ft_strcmp

section .text

_ft_strcmp:
            mov rax, 0
            mov rdx, -1
_loop:
    inc rdx
    mov dh, BYTE[rsi + rdx]
    mov al, BYTE[rdi + rdx]
    cmp dh, al
    je  _loop
    jmp _exit

_exit:
    sub dh, al
    mov rax, dh
    ret

main.c:

#include <stdio.h>
#include <string.h>

int ft_strcmp(char *s1, char *s2);

int     main()
{
    printf("|-%d-||-%d-|\n",ft_strcmp("mehdi", "Mehdi"), strcmp("mehdi", "Mehdi"));
    return (0);
}

when I compile the file ft_strcmp.s I get this error:

test nasm -f macho64 ft_strcmp.s
ft_strcmp.s:18: error: invalid combination of opcode and operands

EDIT: I'm using Intel x86_64 syntax :)

edited Dec 08 '20 at 18:42

Peter Cordes

328,167
45
605
847

asked Dec 08 '20 at 18:32

DarkSide77

3

`mov rax, dh` should be `movzx eax, dh` – Aplet123 Dec 08 '20 at 18:37
The error is fixed, but the return value is wrong – DarkSide77 Dec 08 '20 at 18:42
2

DH is part of RDX, so that would crash if it did assemble. Also, normally you'd use `.asm` for NASM source files, `.S` for GAS source files that gcc/clang can handle directly. – Peter Cordes Dec 08 '20 at 18:42
@PeterCordes I have to use `.s` file – DarkSide77 Dec 08 '20 at 18:50
If you want the sign of the result to correctly reflect the compare result, you need to zero-extend the *inputs* before subtract. signed-overflow is possible for `sub` of arbitrary 8-bit inputs. – Peter Cordes Dec 08 '20 at 20:07

How can I substract two numbers (char to be exact) in assembly?

0 Answers0