Path: news.mathworks.com!not-for-mail
From: "Tim Davis" <davis@cise.ufl.edu>
Newsgroups: comp.soft-sys.matlab
Subject: Re: Modifying MEX arguments in place?
Date: Fri, 15 Feb 2008 01:50:17 +0000 (UTC)
Organization: University of Florida
Lines: 45
Message-ID: <fp2r4p$s2u$1@fred.mathworks.com>
References: <fp0fqv$3gd$1@fred.mathworks.com> <fp0utr$4gr$1@fred.mathworks.com> <fp1g61$elr$1@fred.mathworks.com> <fp1kf6$pgq$1@fred.mathworks.com> <fp1lbe$8v7$1@fred.mathworks.com>
Reply-To: "Tim Davis" <davis@cise.ufl.edu>
NNTP-Posting-Host: webapp-05-blr.mathworks.com
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 8bit
X-Trace: fred.mathworks.com 1203040217 28766 172.30.248.35 (15 Feb 2008 01:50:17 GMT)
X-Complaints-To: news@mathworks.com
NNTP-Posting-Date: Fri, 15 Feb 2008 01:50:17 +0000 (UTC)
X-Newsreader: MATLAB Central Newsreader 45902
Xref: news.mathworks.com comp.soft-sys.matlab:451541



"Petr Krysl" <pkryslNOSP@Mucsd.edu> wrote in message
<fp1lbe$8v7$1@fred.mathworks.com>...
> "Titus" <titus.edelhofer@mathworks.de> wrote in message
> <fp1kf6$pgq$1@fred.mathworks.com>...
> > 
> > "Tim Davis" <davis@cise.ufl.edu> schrieb im Newsbeitrag 
> > news:fp1g61$elr$1@fred.mathworks.com...
> > > "Titus" <titus.edelhofer@mathworks.de> wrote in message
> > no doubt, MEX files are often quite handy, it's not that I
> want
> > to put this in question. The only thing I try to keep
> pointing at,
> > is, that I've seen quite a few MEX files of people who could
> > have avoided the effort when using (o.k., in newer versions)
> > the capabilities e.g. JIT has added to the MATLAB
programming
> > language.
> > 
> > Titus 
> 
> Titus,
> 
> Actually, my solver already modifies the matrix in place
> when I use pure Matlab code.  The factorization is about an
> order of magnitude slower than the Matlab Choleski sparse
> solver (which is a compiled mex code). Therefore, I see a
> potential to speed up while at the same time preserving the
> in-place aspect...
> 
> Petr
> 

<Advance notice of particular bias>
I'm the author of sparse Cholesky in MATLAB
</end notice>

I wouldn't recommend using a skyline code.  Even written in
C, theres no way any skyline code can beat a true sparse
Cholesky code.   For more details, see:

http://doi.acm.org/10.1145/1236463.1236465

The question about how to modify a mxArray in place is still
a good one, but you shouldn't try to do it for this case.