The bitcount() routine can be optimised for speed and code size to exploit an algorithm for which the operation n & n-1 clears the least significant bit set.