whats happening ?

[Jason Moxham]

On Wednesday 23 Oct 2002 9:04 am, Torbjorn Granlund wrote:
> Jason Moxham <J.L.Moxham@maths.soton.ac.uk> writes:
>
>   On Tuesday 22 Oct 2002 8:25 pm, Torbjorn Granlund wrote:
>   > Jason Moxham <J.L.Moxham@maths.soton.ac.uk> writes:
>   >
>   >   I done some preliminary work on a few algorithms ,and even in a
>   > completely un-optimized form have got some good results for some of=
 the
>   > following functions , other results , not so good :(
>   >
>   >   mpz_root using varible precision , division free newton method
>   >
>   > What is needed is mpn_rootrem with these properties.  mpz_root
>   > should not change.
>
>   The varible precision root method I implemented has proven bounds , s=
o
> cannot ever be in error , I assume this is your objection , if not , wh=
at
> is it ?
>
> Well, we're clearly in misunderstanding.
>
> No, I don't object to code that is correct.  As a matter of fact,
> I much prefer correct code to incorrect code.  It is a very
> reasonable misunderstanding that you thought I liked incorrect
> code.  ;-)
>
> My point was that replacing mpz_root isn't the right change since
> the real root extraction is performed at the mpn level.  In the
> development sources, there are several call sites for
> mpn_rootrem.

No problem , in the basic method all the numbers are positive and we know=
 how=20
many bits each one is , so conversion to mpn_root should be easier than m=
ost=20
, however this just returns the root not the rem , clearly the rem part c=
an=20
got from this but it will increase the run time (perhaps as much as doubl=
e)