bit_trick.c

static char*
str_chr(char * str, int chr)
{
    char *s = str;
    int c = chr & 0xFF;
    unsigned m = (c << 24)|(c << 16)|(c << 8)|c;
    for (;;) {
        while (((uintptr_t)s) & 3) {
        chk1: if (s[0] == c) return s;
        chk2: if (s[0] == 0) return s;
            ++s;
        }
        for (;;) {
            unsigned v = *(unsigned*)s;
            unsigned c = (~v) & 0x80808080;
            if (((v ^ m) - 0x01010101) & c) goto chk1;
            if ((v - 0x01010101) & c) goto chk2;
            s += 4;
         }
    }
}
static int
str_len(const char *str)
{
    const char *s = str;
    while (((uintptr_t)s & 3)) {
        chk: if (*s == 0) return s-str;
        s++;
    }
    for (;;) {
        unsigned int v = *(unsigned int*)s;
        unsigned int c = ~v & 0x10101010;
        if ((v - 0x01010101) & c) goto chk;
        s += 4;
    } return s-str;
}

Yeah this whole thing is undefined behavior. In theory this line:

 unsigned v = *(unsigned*)s;

Is already undefined. As for unsigned m = (c << 24)|(c << 16)|(c << 8)|c; at this point all
architectures outside microcontrollers have int/unsigned 4 byte so I don't care anymore. Heck if you
look at all commonly supported hardware for linux you will not find a single instance with int/unsigned being smaller
than 4 byte.

...is already undefined.

Didn't read that far ;)

Btw where do you (want to) use this code? I'm certain, that the library functions are faster on most/all CPUs.

Btw where do you (want to) use this code? I'm certain, that the library functions are faster on most/all CPUs.

The trick is from here: https://github.com/nothings/stb/blob/master/stb_sprintf.h#L306.
Based on: https://graphics.stanford.edu/~seander/bithacks.html#HasLessInWord.

I spend some time trying to understanding how it works. So I wrote these two to:
1.) Confirm I understood how it works
2.) Create a reference if I ever need it