fast inverse and division

[Paul Zimmermann]

Here is a followup to the "optimizing 1/x" thread from last month.
I have written a document describing a first algorithm that computes
(an approximation of) 1/x in O(M(n)), more precisely in 3M(n) in the FFT range.
See <http://www.loria.fr/~zimmerma/papers/index.html#invrem>.
(This is work in progress. More algorithms will be added later.)

Paul Zimmermann