[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Getfem-users] Gmm++ problem with superlu interface.

From: Dawid Midura
Subject: Re: [Getfem-users] Gmm++ problem with superlu interface.
Date: Thu, 3 Nov 2011 18:38:05 -0400

Hi Yves,

Thanks for your reply. I do not see SuperLU_solve being used in test/ I do see it used in test/, but there we have the case of real matrices. When I run a test code with a real matrix I do not get the same error when using superlu. I did check it again, and the matrix K my code is computing is not singular. Any other thoughts/suggestions? By the way, in getfem++ scalar_type is a typedef of a double, is that correct?

Dawid M.

On Thu, Nov 3, 2011 at 1:01 PM, Yves Renard <address@hidden> wrote:

Dear Dawid,

The test program in the directory 'tests' use complex sparse
matrices and superlu and it works.

You can also export your matrix and compute the smallest eigen values in
Matlab to see if it is singular.


On jeudi 3 novembre 2011, Dawid Midura wrote:
> Hi all,
> I have the following code that gives me trouble:
>   typedef gmm::wsvector<std::complex<double> > complex_sparse_wvector;
>   typedef gmm::col_matrix<complex_sparse_wvector> complex_sparse_wmatrix;
>   complex_sparse_wvector F(N);
>   complex_sparse_wmatrix K(N,N);
>   ...
>   // Assemble stiffness matrix K, and load vector F.
>   ...
>   double condition;
>   vector<complex<double> > U(N);
>   gmm::SuperLU_solve(K, U, F, condition);
> I compile with appropriate flags for superlu, when I run the code I get the
> following:
> ============================================
> |      An error has been detected !!!      |
> ============================================
> Error in /usr/include/gmm/gmm_superlu_interface.h, line 244 :
> SuperLU solve failed: info=-2
> From what I have read about SuperLU, info=-2 means that there is a problem
> with the second argument the the SuperLU_gssvx which is matrix K in the
> SuperLU format SuperMatrix, I am not using getfem to assemble K or F, I
> have written my own procedures to do that. I know K is not singular, when I
> make my assembly procedures work with dense matrices and vectors I don't
> have any problems. Any help would be greatly appreciated.
> Thanks,
> Dawid M.


 Yves Renard (address@hidden)       tel : (33)
 Pole de Mathematiques, INSA-Lyon             fax : (33)
 20, rue Albert Einstein
 69621 Villeurbanne Cedex, FRANCE


Getfem-users mailing list

reply via email to

[Prev in Thread] Current Thread [Next in Thread]