From: fateman@cs.berkeley.edu   
      
   On 4/10/2016 10:06 AM, clicliclic@freenet.de wrote:   
   >   
      
   >>   
   >   
   > Polynomial may be ill-conditioned such that numerical root finders have   
   > severe difficulties locating some of the roots and placing error bounds   
   > on them.   
      
   I think that increasing the precision of the floating-point arithmetic   
   sufficiently solves all the difficulties (but not necessarily fast.)   
      
   It has been my experience that using a symbolic method when a numerical   
   method can do the job, has low payoff.  Also, (I don't know about this   
   particular algorithm) it is sometimes the case that a "fast" algorithm   
   is not really fast, but only "asymptotically fast" for problems whose   
   size grows far beyond tasks of interest.   Implementation and timing   
   are needed..   
      
   RJF   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)