At Sat, 5 Jan 2008 11:31:40 -0500, Andrew W. Steiner wrote: > > I briefly took a look at this. The GSL cubic routine is failing > because 'R' and 'Q' > are not precisely zero. If the CR2==CQ3 test succeeded, the routine might > also give the correct answer, except that Q is actually negative in this > context, so that sqrt(Q) is not finite.
Some of this comes from the fact that the decimal coefficients and roots of (x-1.1)^3 do not have exact representations in binary arithmetic. I imagine it would work for polynomials based on exactly representable numbers, e.g (x-1.125)^3 _______________________________________________ Help-gsl mailing list Help-gsl@gnu.org http://lists.gnu.org/mailman/listinfo/help-gsl