Need help determining output of a short C program containing ASM [duplicate]

#include <stdio.h>int main(void){    int var1 = 12;    int var2 = 5;    asm volatile("mov %3, %0 \n""add %2, %1" : "=r"(var1), "=r"(var2) : "r"(var1), "r"(var2));    /*    asm volatile("mov %1, %0 \n""add %0, %1" : "+r"(var1), "+r"(var2));*/    /*    asm volatile("mov %1, %0" : "=r"(var1) : "r"(var2));    asm volatile("add %1, %0" : "=r"(var2) : "r"(var1));*/    printf("var1 = %d, var2 = %d\n", var1, var2);    return 0;}

The first and second pieces of commented asm both result in

var1 = 5, var2 = 10

which is as I expect -- var1 assigned using var2, then both added together to obtain new var2.

I use

gcc test.c

within centos7 on x86_64 machine to compile, so I think I'm using GAS (which means both add and mov have their src/dst reg as 1st/2nd operand).

So, My problem here is, why would the first asm generate the following result:

var1 = 5, var2 = 24

I must admit the first asm looks quite nasty (at least to me, which is why I tried the other two versions to verify my thought), but is it "wrong" ? Or, does it totally make no sense to write so?

If, for example, it's actually acceptable to write such asm code, then I could not figure out why its output looks like that. Is it machine/abi dependent or what? Why doesn't it also give var1 = 5, var2 = 10 ?

I'm really a newbie for x86 assembly, so plz forgive me for this question if it looks dumb, and THANK YOU in advance for taking time to read my question.