[Ipopt] Ipopt with Intel MKL Pardiso

[Olaf Schenk]

Hi Neil,

IPOPT needs an accurate pivoting strategy that is described in

O. Schenk, A. Wächter, and M. Hagemann, Matching-based Preprocessing 
Algorithms to the Solution of Saddle-Point Problems in Large-Scale 
Nonconvex Interior-Point Optimization. Journal of Computational 
Optimization and Applications, pp. 321-341, Volume 36, Numbers 2-3 / 
April, 2007

This pivoting option is not available in MKL and therefore the inertia 
computation in MKL is not reliable.

Regards,
Olaf


On 09/16/2011 12:22 PM, Neil Windelinckx wrote:
>
> For a number of years it has been said that the version of Pardiso 
> provided with Intel MKL is not compatible with IPOpt. (eg 
> http://list.coin-or.org/pipermail/ipopt/2009-June/001583.html )
>
> However, looking back through these discussions they say that it is 
> not compatible because the MKL version it does not compute the inertia 
> of the linear system (or from the IPOpt pre-requisites page "Note that 
> IPOPT requires that the linear solver is able to provide the inertia 
> (number of positive and negative eigenvalues) of the symmetric matrix 
> that is factorized.").
>
> Having looked at the Intel MKL documentation, the inertia is provided 
> in MKL version 10.3 (it was not provided with version 10.0).
>
> Does this now mean that the Intel MKL version of pardiso is now 
> compatible with IPOpt? Or is there some other reason that it will not 
> work correctly?
>
> Thanks
>
> Neil.
>
>
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt


-- 

---
URL: http://fgb.informatik.unibas.ch/people/oschenk/
---

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/ipopt/attachments/20110916/94a94132/attachment.html>