Torbjorn Granlund tg at
Wed Mar 16 12:15:38 CET 2011

James Wanless <bearnol at> writes:

  There is a method for finding PerfectRoots, which I discovered independently
  a while ago, but recently noticed is also in Cohen ACCANT Algorithm 1.7.5.
  When I last looked at the GMP source, this hadn't yet been implemented, but
  it's quite possible ( :) ) you're all already aware of it, and it has been
  discounted for some other reason...

Is this an algorithm for computing a^{1/b} which works just when a is an
integer (and b is an integer too)?

We don't have, but that should be very simple to do using 2^t-adic
arithmetic and lifting (increasing t).  Is that what ACCANT does?

I guess a good implementation it would be less than a page of mpn code.


More information about the gmp-discuss mailing list