I pondered this issue recently. Whilst it is very straightforward in assembler as mentioned, it doesn't make for very portable code! I trolled the web for an elegant solution but there was nothing that grabbed me. Simply, parity is the addition of all the bits - the result is a bit. Depending on if you invert this bit, you get even or odd parity. So in general 'c', one would get:

char do_parity(char val)
{
char a,parity,tmp;

tmp = val; //get a copy to play with
parity = 0; //start with parity = 1 for odd parity result
for(a=0; a<7;a++)
{
if (tmp & 0x01)
{
parity ^= 0x01; //flip the parity bit for each '1' bit
}
tmp>>=1; //shift next bit
}
return parity;
}

Not super efficient is it? Hopefully the compiler is smart enough to figure that you are calculating parity and substitutes the more optimum code!!