[Ipopt] Linking IPOPT with .deb-packaged MUMPS on Ubuntu 9.10 Karmic?

[Stefan Vigerske]

Hi,

Damien Hocking wrote:
> MUMPS is now at 4.9.2, and the 4.9 series is supposed to be better on 
> symmetric optimisation matrices.  I've been using it in IPOPT 3.7.x for 
> a couple of months now on kinetic parameter regressions.

When I try to solve very small LPs or QCPs with Ipopt using Mumps 4.9.2,
then I get a segfault right in the beginning.
Since 4.8.3 is working fine for me, I did not find it worth to waste
time in debugging so far.

Stefan

> 
> Damien
> 
> John Pye wrote:
>> Stefan Vigerske wrote:
>>   
>>> Hi,
>>>
>>>   
>>>     
>>>> On Wednesday 16 December 2009 14:04:23 Stefan Vigerske wrote:
>>>>     
>>>>       
>>>>>> I'm currently working on the packaging of ASCEND for use on Ubuntu, and
>>>>>> as part of that, I've created a Debian package for IPOPT. Recently,
>>>>>> MUMPS has been added to the Ubuntu package repositories (version 4.8.4)
>>>>>> so I'd like to make use of that system package, rather than having a
>>>>>> separate copy compiled especially for Ipopt.
>>>>>>
>>>>>> However, when I use
>>>>>>  ./configure --with-mumps-dir=/usr
>>>>>>         
>>>>>>           
>>>>> It should be --with-mumps-lib and --with-mumps-incdir.
>>>>> Stefan
>>>>>       
>>>>>         
>> Correction: it should be (AFAICT):
>>
>> ./configure --with-mumps-lib=/usr/lib/libdmumps.a 
>> --with-mumps-incdir=/usr/include
>>
>>
>>   
>>>> Just a word of caution: My current Ubuntu (karmic) has MUMPS version 4.8.4,
>>>> However, IPOPT pulls in version 4.8.3. I believe there have been issues in the 
>>>> past with the use of MUMPS newer than expected by IPOPT.
>>>>     
>>>>       
>>> Yes, newer versions than 4.8.3 did not seem to work stable with Ipopt.
>>>   
>>>     
>> Any ideas on what the problem is here? Has there been any discussion 
>> between IPOPT and MUMPS developers on the issue?
>>
>> Is there an unstable version of IPOPT that works with new MUMPS?
>>
>> Cheers
>> JP
>>
>> _______________________________________________
>> Ipopt mailing list
>> Ipopt at list.coin-or.org
>> http://list.coin-or.org/mailman/listinfo/ipopt
>>   
> 
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt
> 


-- 
Stefan Vigerske
Humboldt University Berlin, Numerical Mathematics
http://www.math.hu-berlin.de/~stefan