[Coin-ipopt] Re: Questions about IPOPT in a commercial application

[damien at khubla.com]

All,

having been through commercial delivery of IPOPT with MUMPS once already,
and soon to be going through it again in a different area, I'll throw this
out there.  MUMPS has a generous & liberal license that lets you use it
commercially for free, although the MUMPS team would like an
acknowledgement somewhere in the application's documentation or Help
screen.  The MUMPS team also requires that you tell them that you're going
to use it in something (commercially or otherwise).  Informing them that
you're using it within an application of IPOPT should be enough (note that
I'm not a lawyer either and your organisation might have specific policies
on this).  The official MUMPS availability spiel is here:

http://graal.ens-lyon.fr/MUMPS/avail.html

Damien

> Dear Jean-Philippe,
>
> I'm copying my reply to the Ipopt mailing list.
>
> Of course, I'm very happy to hear that you find Ipopt useful and consider
> to use it in a commercial application.
>
> Currently, the only linear solver that works with Ipopt and has a "free"
> license is MUMPS, which is public domain, see
> http://graal.ens-lyon.fr/MUMPS/avail.html
>
> In my own limited experience, it seems that MUMPS does not perform as fast
> in Ipopt as MA27 on small to medium-sized problems, but has roughly the
> same robustness.
>
> As for Pardiso, yes, the version of Pardiso currently in the MKL doesn't
> have some features that are required for the use in Ipopt.  You might be
> able to link and run the code, but if you are solving non-convex problems
> with Ipopt, the non-convexity might not be handled well.  As far as I
> know, future versions of the MKL will include a version of Pardiso that
> works with Ipopt.
>
> Having written all this, please note that I'm not a legal expert, so
> please check the MUMPS license carefully yourself.  Also, if you want to
> know more about Pardiso and which version will be in the MKL, please
> contact the Pardiso developer Olaf Schenk.
>
> I hope this helps,
>
> Andreas
>
>> Dear Dr. Wächter,
>>
>> First of all I would like to congratulate you for your work on IPOPT,
>> which I think is a great piece of software what will grow in popularity.
>>
>> We are considering using the IPOPT solver in a commercial application.
>> We need however to procure a licence for a linear solver. IPOPT supports
>> several such solvers, but I would like to ask you whether you would
>> recommend a particular one for use in a commercial product.
>>
>> Also, IPOPT documentations mentions that the Pardiso solver sold by
>> Intel in the MKL package is currently incompatible with IPOPT. I was
>> interested in acquiring MKL because of Intel's support and flexible
>> licensing terms. Do you know whether (or when) future versions of this
>> library will be compatible with IPOPT?
>>
>> Thank you very much,
>>
>> Jean-Philippe
>>
>> _____________________________
>>
>> Jean-Philippe Côté, Ph.D.
>> CTO, ExPretio
>>
>> Tel/Fax: +1 514-392-1119
>> www.expretio.com
>>
>>
>>
>>
>>_______________________________________________
> Coin-ipopt mailing list
> Coin-ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/coin-ipopt
>