Re: [Qucs-devel] Offer to help concerning the numerical difficulties

[Frans S. <fra...@gm...>]

On 22-06-13 10:01, Soeren D. Schulze wrote:
> Am 20.06.2013 10:18, schrieb Richard Crozier:
>> Hi All,
>>
>> Bear in mind however, that there is in progress work to add an option to
>> use the gnucap solver as the Qucs simulation backend. The gnucap solver
>> is much faster, and also more memory efficient, being based on a sparse
>> matrix representation. Working to help add this option might be a faster
>> route to a 'better' solver algorithm than modifying the existing solver.
> Ah, that's interesting.  Does the gnucap solver support everything that
> we need, so it can replace the Qucs solver, or is it just supposed to be
> an option?
qucs-core has some features that can not be found in gnucap, like 
s-parameter simulations. I don't think it will replace qucs-core any 
time soon. We could start by supporting gnucap through a 
transient_gnucap simulation for example.

> Right now, I'm only interestested in convergence, not so much in speed.
>   But do you know if convergence is also better in gnucap?
>
>> Also I would also really be wanting to see how these changes would
>> impact my integration of the Qucs solvers with Octave, so if we can keep
>> anything related to this quite firmly on a feature branch or fork until
>> everyone agrees to merge anything I would really appreciate it.
> Yes, we should really do that.
>
>
> Sören
>
> ------------------------------------------------------------------------------
> This SF.net email is sponsored by Windows:
>
> Build for Windows Store.
>
> http://p.sf.net/sfu/windows-dev2dev
> _______________________________________________
> Qucs-devel mailing list
> Quc...@li...
> https://lists.sourceforge.net/lists/listinfo/qucs-devel