Yes, I've seen it before. Don't use the FFT multiplication routines in
it for integer computations that require proven results. He uses a
real FFT subject to roundoff error, with a sum of digits test, which
is not proven. There is an option to switch to NTT's which are 4 times
slower, but these are
William Stein wrote:
> Hi Sage-Devel,
>
> Have you ever wondered about the mathematics behind how MPFR, GMP,
> MPIR, etc., work under the hood? Fortunately, Paul Zimmerman and
> Richard Brent just published a new very-accessible book about exactly
> this, and has the foresight to release their bo
I refer to this book regularly, as it is a good introductory reference
and contains plenty of literature citations.
Some of the more recent algorithms are yet to be described in it, but
having it available is invaluable for the field.
Sadly, I cannot use this as an excuse to bring out the old jok
