Re: seg fault in mpz_root

[Paul Zimmermann]

   > That might be better for the future.  Unless we change to Paul Z.'s
   > discrete variant, which if I understand it keeps intermediates below
   > the true root.

   Not too certain if Paul Z method will be very fast for large k (say >10) as

I guess that my method will only be faster for k=2 and k=3.
For k=4, it's better to call twice mpn_sqrtrem. Then for k=5,
recomputing s^5 should be faster.

Paul