[Ipopt] Restoration Failed

Andreas Waechter andreasw at watson.ibm.com
Wed Sep 9 14:05:38 EDT 2009

Hi Bruno,

>From the output, there seems to be something wrong with your function 
evalution code, as Stefan has suggested.  Furthermore, you are starting 
with an almost optimal point it seems (primal and dual infeasibilities are 
very small), and yet Ipopt cannot compute a search direction in the first 
iteration - MA27 keeps saying that the KKT matrix is rank-deficient, even 
though it does perturbations of the KKT matrix that usually make that go 

But the most telling thing is that your error is not deterministic.  This 
probably means that there is a memory error (such as accessing 
uninitialized memory).  I suggest you run your program with a memory 
checker, such as valgrind (free memory checker on Linux).



On Wed, 9 Sep 2009, Bruno Guerreiro (ISR) wrote:

> Hi Ali,
> I have both 3.6.1 and 3.7.0 installed.
> However, I get exactly the same behavior with either one.
> My greatest problem with this bug, is that it seams to be non
> deterministic, in the sense that, when I get a "Restoration failed" on
> one run, if I run the solver again with exactly the same conditions, it
> is very probable that I'll get an "Optimal solution found".
> Best,
> Bruno Guerreiro
> Ali Baharev wrote, On 09/08/2009 06:28 PM:
>> Hi,
>> Which version of IPOPT are you using?
>> I had problems with the restoration phase with 3.7.0, see:
>> http://list.coin-or.org/pipermail/ipopt/2009-August/001659.html
>> Ali
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt

More information about the Ipopt mailing list