[Coin-ipopt] Bug report on GAMS-IPOPT

[Stefan Vigerske]

Hi,

Andreas wrote:
> From the output it seems that you are using Bonmin with the outer 
> approximation option for a nonconvex problem?  As the message says: This
> is very experimental and has a good chance of not working.
> 
> If you have questions to Bonmin, please send a message to the Bonmin
> mailing list.
> 
> As for the error message that you see, it means that the linear solver
> in Ipopt ran out of memory.

If you continue to have problems with MUMPS when using different options
in Bonmin (e.g., set bonmin.algorithm to B-BB), you might consider to
compile a Gams/Bonmin with a different linear solver (e.g., HSL,
Paradiso...).
They have limitations in their licence, but since you are at a
university it should be ok for you to use them (but please read the
licence information before download).
Unfortunately, GAMS ships Bonmin and Ipopt with MUMPS only, which is not
always the best choice, but it is the only one which licence is generous
enough to allow shipping inside a commercial product.
For compiling Gams/Bonmin by yourself see the GAMSlinks webpage at
https://projects.coin-or.org/GAMSlinks. You will need to get the stable
0.3 version via subversion.

Best,
Stefan