Fast, accurate approximation algorithms for computing floating point square root