sqrt algorithm

[Niels Möller]

tg at gmplib.org (Torbjörn Granlund) writes:

> I haven't thought carefully about this, but we should probably provide
> plain fllor(n/q) functions as the main entry points.  We could then add
> fraction limb development functions as a parallel set.

I agree that approach makes the interface nicer. We'd want functions
that can reuse the needed inverses between the two calls, though.

In my current, case, I haven't fully sorted out the sizes I really need
(and the code I have posted is not yet correct). But I think the main
need is U/D, where both inputs are n limbs (so the quotient is a single
limb, and usually small), followed by about n fraction limbs.

Regards,
/Niels

-- 
Niels Möller. PGP-encrypted email is preferred. Keyid C0B98E26.
Internet email is subject to wholesale government surveillance.