[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Sparse-merge
From: |
David Bateman |
Subject: |
Re: Sparse-merge |
Date: |
Thu, 10 Mar 2005 09:37:18 +0100 |
User-agent: |
Mozilla Thunderbird 0.8 (X11/20040923) |
Richard Hindmarsh wrote:
David Bateman wrote:
John W. Eaton wrote:
| Thanks. Will 2.9.x (or does the main trunk) have an interface to
ARPACK | (e.g. through "eigs"), or is anyone working on this?
Not yet. SEe the list of projects for sparse matrices in the file
http://www.octave.org/cgi-bin/viewcvs.cgi/octave/PROJECTS?rev=1.60&view=auto
ARPACK is definitely on the todo list. It is a complete lack of time
at the moment that will prevent me from doing something here.
If you want to do the ARPACK interfacing, I'd welcome the help and
would be willing to vet the code and give pointers.
I could possibly do it - I used ARPACK before "eigs" appeared in
Matlab and have written .mex files though not .oct. Am motivated but
short of time.
A quick look at "eigs" in Matlab shows that "eigs.m" calls an
interface routine called "arpackc.dll" (or whatever suffix). It also
looks as though "eigs" would be largely useable in Octave. I guess
that an Octave "arpackc" would not be too difficult to write but what
is then one expected to do - rewrite "eigs" so that it doesn't look
like the Matlab one?
Richard,
I have to agree with John on this, you can't use any matlab code
whatsoever, at the risk of being accused of copyright violation.... I
consider the help available on the mathworks website as open game for
developing compatiability code, at least in terms of a specification of
the interface, but again don't copy it when you write the help for the
functions.
I also consider that cross-comparison of the outputs for octave and
matlab implementations is fair game. However, I never look at the matlab
dot-m files at the risk of finding myself writing the same code even
accidentally. This makes life slightly harder of course...
ARPACK is needed for the eigs function and the svds function, and would
then be interfaced to the rank, etc functions to allow them to work.
This shouldn't be a major development for a person familar with ARPACK
and oct-files. The main issue would be one of packaging, the way John is
leaning I think he would expect ARPACK to be an external library with an
autoconf function to detect its presence. Then the eigs.cc and svds.cc
code would link to this library if present or build a stub version that
throws an error.
In any case, we'd welcome any inputs, and I would give pointers once you
start..
Regards
David
--
David Bateman address@hidden
Motorola Labs - Paris +33 1 69 35 48 04 (Ph)
Parc Les Algorithmes, Commune de St Aubin +33 1 69 35 77 01 (Fax)
91193 Gif-Sur-Yvette FRANCE
The information contained in this communication has been classified as:
[x] General Business Information
[ ] Motorola Internal Use Only
[ ] Motorola Confidential Proprietary
-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.
Octave's home on the web: http://www.octave.org
How to fund new projects: http://www.octave.org/funding.html
Subscription information: http://www.octave.org/archive.html
-------------------------------------------------------------
- Sparse-merge, Richard Hindmarsh, 2005/03/07
- Sparse-merge, John W. Eaton, 2005/03/08
- Re: Sparse-merge, Richard Hindmarsh, 2005/03/08
- Re: Sparse-merge, John W. Eaton, 2005/03/08
- Re: Sparse-merge, David Bateman, 2005/03/09
- Re: Sparse-merge, Richard Hindmarsh, 2005/03/09
- Re: Sparse-merge, John W. Eaton, 2005/03/09
- Re: Sparse-merge,
David Bateman <=
- Re: Sparse-merge, Richard Hindmarsh, 2005/03/10
- Re: Sparse-merge, David Bateman, 2005/03/10