philnovi wrote: > > This method for square roots is EXACT for the number of digits that you have > the patience to use it to. [snip a nice description of the method]
Yep, that's the method I learned. And I believe I saw an analysis once that shows how it is equivalent to Newton-Raphson, and how an analogous method can be developed for cube roots. But on second thought, I'm not sure I believe that connection, since Newton provably doubles the number of bits of accuracy at each iteration, and this method converges more slowly.