Path: news.mathworks.com!newsfeed-00.mathworks.com!panix!bloom-beacon.mit.edu!newsswitch.lcs.mit.edu!newsfeed.cwix.com!dc3peer2.nntp.savvis.net!peer.nntp.savvis.net!bigfeed.bellsouth.net!news.bellsouth.net!cyclone1.gnilink.net!gnilink.net!nx02.iad.newshosting.com!newshosting.com!216.196.98.140.MISMATCH!Xl.tags.giganews.com!border1.nntp.dca.giganews.com!nntp.giganews.com!local02.nntp.dca.giganews.com!nntp.supernews.com!news.supernews.com.POSTED!not-for-mail
NNTP-Posting-Date: Mon, 22 Dec 2008 09:49:35 -0600
Newsgroups: comp.soft-sys.matlab
Subject: FMINCON implemetation speed, also any experience with NAG?
From: "Eric J. Holtman" <ejh@ericholtman.com>
Message-ID: <Xns9B7C63F60A951ejhericholtmamcom@216.168.3.30>
User-Agent: Xnews/5.04.25
Date: Mon, 22 Dec 2008 09:49:35 -0600
Lines: 22
X-Trace: sv3-uoO3BRhU64JoGzAEyUhzZ56iLmvzsx+gSPEW1jgIjLccUgEs1xuKvDxMucoqs8eUY4QNiXVZRsEXie6!4xYhgkGpib6UtK0mNZmSlBU+92QrhqihkGBRjBNG6TyaHzOwsN81tV4H7jyEFSrOQnzuowkGfq7r!HwxGZfK9RNI=
X-Complaints-To: www.supernews.com/docs/abuse.html
X-DMCA-Complaints-To: www.supernews.com/docs/dmca.html
X-Abuse-and-DMCA-Info: Please be sure to forward a copy of ALL headers
X-Abuse-and-DMCA-Info: Otherwise we will be unable to process your complaint properly
X-Postfilter: 1.3.39
Xref: news.mathworks.com comp.soft-sys.matlab:508354


I've been poking around the .m files for the Optimization Toolbox,
and I'm wondering.... is fmincon actually implemented anywhere in
.m code, or does it, after parameter checking, call into a routine
coded in C, Fortran or assembler?

Seems like (for the interior point method), it finally calls
"barrier", which I can't find.


I'm concerned that if it's in .m code, it might not be fast
enough for some large problems I want to solve.  (Yes, I'm getting
a framework and samples together to actually test this).

In that vein, does anyone here have experience with Numerical
Analysis Group's (NAG) MATLAB extensions.  I've been very happy
using NAG products in the past, maybe this would be a better
fit for me (my problem isn't general non-linear constrained
optimization, it's quadratic objective, linear constraints,
structured hessian, which quadprog won't solve efficiently 
because I can't use large-scale).  I know one the NAG functions for
quadratics does what I want, since I used it before (directly from
C++, not MATLAB), at a previous job