Return the smallest positive floating point number representable with the specified precision on this computer. Double precision is assumed if no floating point type is specified.