bfloat roots of polynomials



Richard Fateman wrote:
> I think the smart money on this computation is heading towards finding the
> eigenvalues of the companion matrix.

Is it?  I thought Jenkins-Traub was still pretty much state of the art.

Ray