Suppose I have a char buf[12]; which I know will always have a right justified unsigned number padded on the left with spaces. So, for example: _________329 (where the _ stands for space). The fastest way I can think of parsing it is with something like this:
while (*buf == ' ') buf++;
atoi(buf);
but I was wondering if there was a faster way, particularly with atoi given that we know it is unsigned which atoi does not assume..
I assume the first char is reserved for a "potential sign" and is always "space"? Because otherwise, you would only need a char[11] instead of a char[12]. Anyway, fixed size allows manual loop unrolling:
unsigned parse(const char(&b)[12])
{
return ((((((((((b[1] & 15))
* 10 + (b[2] & 15))
* 10 + (b[3] & 15))
* 10 + (b[4] & 15))
* 10 + (b[5] & 15))
* 10 + (b[6] & 15))
* 10 + (b[7] & 15))
* 10 + (b[8] & 15))
* 10 + (b[9] & 15))
* 10 + (b[10]& 15);
}
Note that the & 15 trick treats spaces and zeros the same and will work with both ASCII (space = 32, zero = 48) and EBCDIC (space = 48, zero = 240). I haven't checked out other character encodings yet :)
Will this actually be faster or slower than atoi? The only way to find out is to measure. But I would probably stay with atoi in any case, because using a standard function always improves readability.