[Ipopt] Matlab Interface

tony at kelman.net tony at kelman.net
Sun Nov 8 17:36:12 EST 2015

Matlab has always removed zeros from sparse matrices, and the old matlab interface did quite a bit of copying work to handle this.

Can you please post the code somewhere like github where it is easier to look at than the file exchange? It would be best if you had a clear history of incremental changes relative to the current matlab interface.

On Sun, Nov 8, 2015 at 12:30 PM -0800, "Enrico Bertolazzi" <enrico.bertolazzi at unitn.it> wrote:

I rewrite the Peter Carbonetto MATLAB interface for IPOPT.
The new interface is available at:


I rewrite the interface cause I have some problem in compiling it
on OSX and find that interface fails on some problems.
In the rewriting I tried to use STL for dynamic allocation and
added some extra checks to improve robustness.
The old interface assume that sparse pattern of sparse matrices
is fixed while recent version of MATLAB seem to delete
zero element on the pattern when passed to the interface.
This explain why old interface sometimes fails.
The new interface tries to manage this behaviour.

I compiled successfully the interface in OSX and LINUX using
MUMPS and MA57 (available in MATLAB).

Instructions on how to compile the interface are contained in some
readme files.

The interface compile also in WINDOWS with Visual Studio (2013)
but I am not able to link successfully with precompiled DLL
available at http://www.coin-or.org.
I successfully compiled and linked the interface using cygwin
but the resulting mex-interface crashes, I am working to undestand why.
I am not a user of windows OS, hence, the help of a WINDOWS expert
would be appreciated.

Ipopt mailing list
Ipopt at list.coin-or.org
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/ipopt/attachments/20151108/eafc71ad/attachment-0001.html>

More information about the Ipopt mailing list