roots

[Jason Moxham]

On Wednesday 23 Oct 2002 6:24 pm, Torbjorn Granlund wrote:
> [Follow-ups to gmp-devel, since this is about GMP development.]
>
> These are nice speedups for large operands, but also large
> slowdowns for the very important range of small numbers.
> That needs to be addressed.
>

I think I can get it better for small n

> The code is almost uncommented.  There aren't even any comments
> with an overview of the algorithms, or any comments before each
> of the very many functions.
>

As I may of mentioned , I have just started work on these fns , so it's a=
ll=20
still very much in development

> Is an entirely new type, float_t, and 20 functions manipulating
> that type, really necessary?  And surely, they should not be
> exported to the user?
>

I plan to dump all the float_t types , just mp_limb_t* will be left , its=
 just=20
a lot easier to work at the higher level while coming to grips with how t=
he=20
algorithm works in practice

> GMP is almost entirely volunteer work, which makes it important
> to contribute mature code.  You need to work much more on your
> code before it is ready for us GMP maintainers.
>
> This might be a discouraging reply, and I understand that it is
> much more fun hacking things that run than to mature them for GMP
> inclusion.  Unfortunately, sompebody needs to do the latter, or
> the code will remain a local hack on your computer.

I plan to mature the code for GMP inclusion , but that takes time . I thi=
nk=20
its a good idea to send preliminary code to the gmp lists , other people =
may=20
spot good ideas and save me time , or they may say that they already have=
=20
better code.