| Re: [eigen] Opinions on some minor features for 3.2 |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/eigen Archives
]
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] Opinions on some minor features for 3.2
- From: Billy Araujo <billyaraujo@xxxxxxxxx>
- Date: Wed, 20 Mar 2013 15:56:22 +0000
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=mime-version:x-received:in-reply-to:references:date:message-id :subject:from:to:content-type; bh=NOrCFwQdtFY1J+ctvIabywNDMozscHx0nIKj1YGg6sg=; b=fU6FYcR1RCDK0yq4cDM0wVCz4e2yH4E2z7hfC3eCtUXyLQKEjm0uF5DLI1D4eOljzq qnlRD6fnH7mVmsISHRPNak+UK62OYNbSKF7MBFDclAiVjcOpdyLtEfjUIXUa172f544n ohhX3eZXrDphvHG6WpKN4f5uDHlHellJInU9ArPAIcl49pO/2D31WNNQA5Offk/Zf0+n g4O1Dymttu59/eQhOxen1Cdvt4mxkqBm7ICeBN0XjXSVXDovq7SXvYIlrXkyWllJEjjz W829Oi5BoGtQhJIgOvHU2iYu+oDbSk0qzXiFnziHQlBJigaLC/5OHbahyMWWmqudQ+BV 2sAQ==
It would be nice if Eigen had a feature in Sparse and Dense Matrices
to eliminate a row/column thus reducing size of matrix and obtain
permutation vector and a right hand vector with the know values.
similiar suggested by:
http://eigen.tuxfamily.org/bz/show_bug.cgi?id=273
this would be useful for finite element analysis. I have done this on
top of eigen and can contribute my code for further clean up and
testing as I am not too familiar with eigen's inner architecture.
Billy.
On Wed, Mar 20, 2013 at 3:26 PM, Gael Guennebaud
<gael.guennebaud@xxxxxxxxx> wrote:
> Hi,
>
> Since a beta for 3.2 has already been released, we should limit as
> much as possible the new features in 3.2 to make sure we can release
> it asap. However there are few minor ones that I'd like to suggest.
> Please give your opinion.
>
>
> 1 - Add support for the following products: SelfadjointView * Sparse
> and Sparse * SelfadjointView
> This is straightforward to add because there is no obvious solution
> than evaluating the SelfadjointView into a full SparseMatrix. This
> copy will be documented. This copy is not a big deal from my point of
> view because SparseMatrix * SparseMatrix already require many explicit
> transpositions in most cases.
> (see http://eigen.tuxfamily.org/bz/show_bug.cgi?id=553)
>
>
> 2 - Make the sparse inner iterator copyable -> trivial and harmless
> change; This is useful to store a list of iterators.
> (see http://eigen.tuxfamily.org/bz/show_bug.cgi?id=557)
>
>
> 3 - Regarding this second change and related bug report, I also
> proposed to add new members to sparse InnerIterators:
> a) add a pos() member returning the number of non-zeros that have been
> iterated on
> b) add a += operator to skip many non-zeros at once.
>
> That's easy to add, but I'm not really sure about it. Clearly the name
> "pos" is not good, but perhaps we could wait for a clear need for
> that.
>
>
> Feel free to extend this discussion with other suggestions too.
>
> cheers,
> Gael.
>
>