Why is Modulo 2 optimization different from x&1?

Asked Mar 23 '21 at 07:19

Active Mar 23 '21 at 13:50

Viewed 175 times

Why is the assembly output of this code:

code

int 
f(int n) {
        return n % 2;
}

asm

f:
        mov     edx, edi
        shr     edx, 31
        lea     eax, [rdi+rdx]
        and     eax, 1
        sub     eax, edx
        ret

different from this one:

code

int 
g(int n) {
        return n & 1;
}

asm

g:
        mov     eax, edi
        and     eax, 1
        ret

I couldn't find anything online (maybe I didn't use the right keywords).

NOTE:

edited Mar 23 '21 at 13:50

Peter Cordes

asked Mar 23 '21 at 07:19

5

Compare the output of f and g for argument `-1` – Mat Mar 23 '21 at 07:27
@Mat right, it makes sense. Thank you. – Mar 23 '21 at 07:30
2

To complete what@Mat said, to be pedantic, in first case you get the remainder, in second case the modulo – Damien Mar 23 '21 at 08:29
3

If you use `unsigned`, the compiler can optimize to just an AND. – Peter Cordes Mar 23 '21 at 13:49
@PeterCordes Correct. That idea came into my mind after testing with Mats value. – Mar 23 '21 at 15:27

0 Answers0