A perfect power, and then?

[Torbjorn Granlund]

nisse at lysator.liu.se (Niels Möller) writes:

  > Perhaps you could add some of your remarks as a comment to the file?
  
  Makes sense. Not tonight, though.
  
I saw your edits.  And I edited them sligtly, adding some more FIXMEs
and clarifying the operation of binv_root.  Please read and tell me if
you agree.

I realise that "Use a small table to get starting value" might not be
easy to implement for root since one might need k tables.  Other
possibilities would be:

(1) have a k-indexed table of (say) 32-it words, from which one extracts
a few bits of starting value from using the lowest y bits as shift
count.

(2) iterate single limb code before entering the mpn loop.

-- 
Torbjörn