Extending the mpn interface

[Bill Allombert]

On Mon, Feb 18, 2013 at 11:06:01AM +0100, Niels Möller wrote:
> For powm, I think it uses redc (in the flavor redc_n) also for very large
> (odd) modulo, even if the difference to plain euclidean division is
> going to be pretty small.

In my mind redc is a quadratic algorithm. you probably use another variation.

What would be very useful would be to provide lower and upper half product.
This way I can implement faster Barrett reduction.

Cheers,
Bill.