[Dune] Read in matrix and rhs, solve linear system

[Markus Blatt]

Hi,

On Tue, Sep 01, 2015 at 08:27:16AM +0000, Buesing, Henrik wrote:
> [Buesing, Henrik] I now get the following output (see below). The solver does not crash anymore, but gets NaNs instead...
> 
> %%MatrixMarket                                                                                              
> %%MatrixMarket                                                                                              
> Level 0 has 6.3058e+06 unknowns, 6.3058e+06 unknowns per proc (procs=1)                                     
> aggregating finished.                                                                                       
> Level 1 has 2.45706e+06 unknowns, 2.45706e+06 unknowns per proc (procs=1)                                   
> aggregating finished.                                                                                       
> Level 2 has 904947 unknowns, 904947 unknowns per proc (procs=1)                                          
> aggregating finished.                                                                                    
> Level 3 has 500740 unknowns, 500740 unknowns per proc (procs=1)                                          
> aggregating finished.                                                                                    
> Level 4 has 179584 unknowns, 179584 unknowns per proc (procs=1)                                          
> aggregating finished.                                                                                    
> Level 5 has 85507 unknowns, 85507 unknowns per proc (procs=1)                                            
> aggregating finished.                                                                                    
> Level 6 has 39132 unknowns, 39132 unknowns per proc (procs=1)                                            
> aggregating finished.                                                                                    
> Level 7 has 21563 unknowns, 21563 unknowns per proc (procs=1)                                            
> aggregating finished.                                                                                    
> Level 8 has 13195 unknowns, 13195 unknowns per proc (procs=1)                                            
> aggregating finished.                                                                                    
> Level 9 has 10225 unknowns, 10225 unknowns per proc (procs=1)                                            
> aggregating finished.                                                                                    
> Stopped coarsening because of rate breakdown 10225/9256=1.10469<1.2                                      
> operator complexity: 1.62054                                                                             
> Using a direct coarse solver (UMFPACK)                                                                   
> Building hierarchy of 10 levels (inclusive coarse solver) took 390.709 seconds.                          
> Building hierarchy took 390.709 seconds                                                                  
> === GeneralizedPCGSolver                                                                                 
>  Iter          Defect            Rate                                                                    
>     0        0.0181497                                                                                   
>     1              nan              nan                                                                  
>     2             -nan             -nan

That is weired and I have no idea why this happens. Can you try
another solver (e.g. BiCGSTABSolver instead of GeneralizedPCGSolver)?

Is it possible to publish your test matrix and right hand side for
testing?

Markus
-- 
Join us at the DUNE User Meeting Sept. 28-29, 2015, Heidelberg, Germany
http://users.dune-project.org/projects/dune-user-meeting-2015/wiki

Dr. Markus Blatt - HPC-Simulation-Software & Services http://www.dr-blatt.de
Hans-Bunte-Str. 8-10, 69123 Heidelberg, Germany
Tel.: +49 (0) 160 97590858
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL: <https://lists.dune-project.org/pipermail/dune/attachments/20150901/821be68e/attachment.sig>