[Trilinos-Users] Aztec00 Question

Heroux, Michael A maherou at sandia.gov
Wed Jun 30 21:57:08 MDT 2010


John,

It is normal for the local representation of data to swap the block columns.
We do this so that local rows and corresponding columns contain the diagonal
values.

Are your matrix and vector maps compatible?  They should both be BlockMaps,
since you are using a block matrix.  Alternatively, you can wrap your
VbrMatrix as an Epetra_VbrRowMatrix and stick with standard maps.

If none of these ideas work, can you send me a small code to test?

Mike

On 6/30/10 12:57 PM, "John Mitchell" <jamitch at sandia.gov> wrote:

> Hello,
> 
> I am having trouble understanding why the following "diagonal" linear
> system will not solve?
> 
> The entire system is a 6x6 matrix and is run on two processors.  Each
> processor owns a row of  2 3x3 matrices.  The matrix and RHS vector are
> given below and the Aztec "warning" message is also printed.
> By inspection, the solution for GID=3 should be .001 but that is not
> produced?
> 
> In this case I have dirichlet boundary conditions on all points and so
> there are 1's on the diagonal and the rest of the entries are zero.
> Note that the 2nd row has the columns swapped?  I'm hoping thats not an
> issue with the solver or graph?  Of course when I fill these entries
> this is specified and I get no return errors.
> 
> Any thoughts/suggestions?
> John
> 
> local row Id = 0; global rowId = 0; col gIds = 0, 1
>     1.000000e+00 0.000000e+00 0.000000e+00     0.000000e+00 0.000000e+00
> 0.000000e+00
>     0.000000e+00 1.000000e+00 0.000000e+00     0.000000e+00 0.000000e+00
> 0.000000e+00
>     0.000000e+00 0.000000e+00 1.000000e+00     0.000000e+00 0.000000e+00
> 0.000000e+00
> 
> 
> 
> 
> local row Id = 0; global rowId = 1; col gIds = 1, 0
>     1.000000e+00 0.000000e+00 0.000000e+00     0.000000e+00 0.000000e+00
> 0.000000e+00
>     0.000000e+00 1.000000e+00 0.000000e+00     0.000000e+00 0.000000e+00
> 0.000000e+00
>     0.000000e+00 0.000000e+00 1.000000e+00     0.000000e+00 0.000000e+00
> 0.000000e+00
> 
> 
> Rhs Vector
> ----------
> Epetra::Vector     MyPID           GID               Value
>          0             0                       0
>          0             1                       0
>          0             2                       0
> 
> Epetra::Vector   
>          1             3                   0.001
>          1             4                       0
>          1             5                       0
> 
> 
==============================================================================>
=
> AZ_check_input: Setup information on processor 0
> 
>     solver:                    0
>     convergence flag:            0
>     maximum iterations:            500
>     reordering:                    1
>     preconditioner:                1
>     polynomial order:            1
>     tolerance:                1.0e-06
>     drop:                    0.0e+00
> 
>     Number of internal unknowns:        0
>     Number of border  unknowns:        3
>     Total number of unknowns:        3
>     Number of external unknowns:        3
>     Number of internal blocks:        0
>     Number of border  blocks:        3
>     Total number of blocks:            3
>     Number of external blocks:        3
>     Number of processors:            2
>     Node number:                0
>     Number of neighbors:            1
>     Number of unknowns sent to neighbors:    3
> 
==============================================================================>
=
> AZ_check_input: Setup information on processor 0
> 
>     solver:                    0
>     convergence flag:            0
>     maximum iterations:            500
>     reordering:                    1
>     preconditioner:                1
>     polynomial order:            1
>     tolerance:                1.0e-06
>     drop:                    0.0e+00
> 
>     Number of internal unknowns:        0
>     Number of border  unknowns:        3
>     Total number of unknowns:        3
>     Number of external unknowns:        3
>     Number of internal blocks:        0
>     Number of border  blocks:        3
>     Total number of blocks:            3
>     Number of external blocks:        3
>     Number of processors:            2
>     Node number:                0
>     Number of neighbors:            1
>     Number of unknowns sent to neighbors:    3
> 
==============================================================================>
=
> 
>         *******************************************************
>         ***** Problem: Epetra::VbrMatrix
>         ***** Preconditioned CG solution
>         ***** 1 step block Jacobi
>         ***** No scaling
>         *******************************************************
> 
>                 iter:    0           residual = 1.000000e+00
> 
>     Solver:            cg
>     number of iterations:    1
> 
>     Actual residual =  1.0000e-03    Recursive residual =  1.0000e-03
> 
>     Calculated Norms                Requested Norm
>     --------------------------------------------    --------------
> 
>     ||r||_2 / ||r0||_2:        1.000000e+00    1.000000e-06
> 
> 
>         Solution time: 0.000106 (sec.)
>         total iterations: 1
> 
> 
>     ***************************************************************
> 
>     Warning: direction vector is no longer A-conjugate
>     to previous vector but solution has not converged.
> 
>     ***************************************************************
> 
> 
> _______________________________________________
> Trilinos-Users mailing list
> Trilinos-Users at software.sandia.gov
> http://software.sandia.gov/mailman/listinfo/trilinos-users
> 




More information about the Trilinos-Users mailing list