A comparison of methods for terminating polynomial iterations (Q1113616)

The author compares three different methods of terminating an iterative method for the approximate computation of a zero of a polynomial, one based on the explicit calculation of rounding errors, the second on differences of the iterates, and the third on different methods of calculating values of the polynomial. By means of extensive experiments with randomly generated polynomials it turns out that the simplest method - that based on differences of the iterates - usually gives the lowest actual error of the zero.