Wraparound multiplicaton vs mullo

[Niels Möller]

Torbjorn Granlund <tg at gmplib.org> writes:

> Not just that, I think.  One would get a carry-out from these addmul
> loops to handle somehow. A negacyclic basecase would need the addmul to
> become a submul halfway through.

For B^n - 1, I think it should be all additions. It should be the same
operations as in mul_basecase, but with mod n arithmetic for all
indexing of rp.

Regards,
/Niels