numeric_limits::epsilon

static T epsilon() throw();

The function returns the difference between 1 and the smallest value greater than 1 that is representable for the type (which is the value FLT_EPSILON for type float).

See the related numeric_limits sample program.