We are developing numerical compuation algorithms based on empirical error estimation method in order to obtain user-required precision approximation by using existing old-style numerical algorithms or libraries.