[Csdp] Failure: return code is 8

[Brian Borchers]

On Thu, Apr 11, 2013 at 12:25 PM, Michael Poss <michael.poss at hds.utc.fr>wrote:

> Dear all,
>
> I'm working on a complex network design problem involving robust
> constraints, reformulated as linear matrix inequalities.
> I'm trying to solve the problem using csdp. For small instances,
> everything works fine, but for larger ones I get very quickly an error,
> given below. Most of my constraints are linear constraints, more precisely,
> flow conservation constraints.
> Any clue on why this happens ?
>
> Best regards,
>
> Michael.
>
>
> --------------------------------------------------------------------------------------------------------------------------------------
>
> Iter:  0 Ap: 0.00e+000 Pobj:  2.2969834e+008 Ad: 0.00e+000 Dobj:
>  0.0000000e+000
> Iter:  1 Ap: 8.79e-001 Pobj:  2.5279101e+008 Ad: 8.91e-001 Dobj:
>  1.2389219e+011
> Iter:  2 Ap: 8.09e-001 Pobj:  2.8075298e+008 Ad: 8.85e-001 Dobj:
>  1.6902705e+011
> Iter:  3 Ap: 5.08e-001 Pobj:  2.7716885e+008 Ad: 6.89e-001 Dobj:
>  3.4354074e+011
> Iter:  4 Ap: 3.80e-001 Pobj:  2.3436358e+008 Ad: 2.19e-001 Dobj:
>  3.4354074e+011
> Factorization of the system matrix failed, giving up.
> Failure: return code is 8
> Primal objective value: 2.3436358e+008
> Dual objective value: 3.4354074e+011
> Relative primal infeasibility: 1.28e+007
> Relative dual infeasibility: 1.29e+004
> Real Relative Gap: 9.99e-001
> XZ Relative Gap: 4.26e+002
> DIMACS error measures: 7.08e+007 0.00e+000 7.25e+005 0.00e+000 9.99e-001
> 4.26e+002
>

Did you attempt to solve this problem using the "csdp" stand alone solver,
or did you use the library interface to pass the problem in to the code?

It's basically impossible to tell what could have gone wrong here without
looking at the problem in more detail and without running the solver with
some debugging options turned on.  If you'd be willing to share your .dat-s
file with me, I'd be happy to take a look at it.  Did you attempt to solve
this problem using the "csdp" stand alone solver, or did you use the
library interface to pass the problem in to the code?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/csdp/attachments/20130416/606eafaf/attachment.html>