help-octave
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: using gmres.m from Matlab


From: David Bateman
Subject: Re: using gmres.m from Matlab
Date: Tue, 08 May 2007 10:12:42 +0200
User-agent: Thunderbird 1.5.0.7 (X11/20060921)

baptiste AuguiƩ wrote:
> Hi,
> 
> 
> I'm using a Matlab code which deals with huge matrices. There is some
> inversion required, for which I have several options:
> 
> direct inversion
> gmres
> minres
> qmr

The iterative solvers I don't need and so never bothered writing. There
is bicg in octave-forge. The article

http://arxiv.org/pdf/hep-lat/0610054

describes QCDLAB which has all of these in very simple forms (ie missing
all the error correction, and bells and whistles). You could use though,
or even better use them as the basis to contribute compatible functions
to Octave as its is GPLed...

> 
> The last three don't seem to be available in Octave. Since I own a
> licence of Matlab but can only run Octave on a remote supercomputer, is
> it OK for me to copy these Matlab files and use them with Octave? Is
> that a stupid thing to do?

Don't..

> 
> Another option I'll look into is using FFT techniques since my matrix
> has a particular Block Toeplitz structure, but that will probably
> require quite a bit of work.

There was some code for Toeplitz matrices running around on the lists a
few months ago. The conclusion was that its not a good idea to include
it in the generic matrix solvers and you loose a significant portion of
the effiicency. Its better to have the solver work directly on a vector
representation of the toeplitz matrix.. See the thread for details

http://www.cae.wisc.edu/pipermail/octave-maintainers/2006-November/001339.html

D.



reply via email to

[Prev in Thread] Current Thread [Next in Thread]