[Trilinos-Users] Problem with ML - TestSmoothers() method

Raymond Tuminaro rstumin at sandia.gov
Thu Jun 23 17:39:43 MDT 2011


Bram,
   Sorry for the problems. It looks like we have something that has changed ... making
TestSmoothers() a bit out of date. I've attached two files which hopefully fix the
problem. They should replace the ones in ml/src/Utils. I'm heading out on vacation ...
so I really hope I got this right. I'll be back on July 5th.
-Ray


Reps Bram wrote:
> Hello,
> 
>  
> 
> I am using the ML package of Trilinos and would like to test my 
> preconditioner with the method TestSmoothers(). For a given 
> preconditioning object it should return a nice overview of the 
> performance of different smoothers (Jacobi, Gauss-Seidel, Aztec 
> preconditioner…). And each smoother is tested with different damping 
> factors. This is described in Section 6.6.4 of the ML-guide.
> 
>  
> 
> However, I get a runtime error. The method TestSmoothers() seems to 
> start, but it returns an error when it wants to set the damping factor 
> of the first smoother, apparently because the parameter name is wrong. 
> (I’ve pasted the ouput below)
> 
> Has anyone encountered this problem yet?
> 
>  
> 
> Kind Regards,
> 
> Bram Reps
> 
>  
> 
> Research group TWNA <http://twna.ua.ac.be/>
> 
> University of Antwerp
> 
> e: bram.reps at ua.ac.be <mailto:bram.reps at ua.ac.be>
> 
> t: +32 3 265 38 95
> 
> f: +32 3 265 37 77
> 
>  
> 
>  
> 
>  
> 
> Here’s the output:
> 
>  
> 
> ------------------------------------------------------------------------------
> 
> *** Analysis of ML parameters (smoothers)
> 
>  
> 
> maximum iterations = 500
> 
> tolerance          = 1.000000000000000e-05
> 
>  
> 
> All options as in the input parameter list, except that
> 
> all levels have the same smoother
> 
>  
> 
> M: maximum iterations exceeded without convergence
> 
> N: normal exit status (convergence achieved)
> 
> B: breakdown occurred
> 
> I: matrix is ill-conditioned
> 
> L: numerical loss of precision occurred
> 
>  
> 
> count  smoother type                 its       ||r||/||r_0||  time (s)
> 
>  
> 
> - Jacobi
> 
> Error, the parameter {name="smoother: damping",type="double",value="0.25"}
> 
> in the parameter (sub)list "NSSA default values->smoother: list (level 0)"
> 
> was not found in the list of valid parameters!
> 
>  
> 
> The valid parameters and types are:
> 
>   {
> 
>     "coarse: MLS polynomial order" : int = 2
> 
>     "coarse: add to diag" : double = 0
> 
>     "coarse: polynomial order" : int = 2
> 
>     "smoother: Aztec as solver" : bool = 0
> 
>     "smoother: Aztec options" : Teuchos::RCP<std::vector<int, 
> std::allocator<int> > > = Teuchos::RCP<std::vector<int, 
> std::allocator<int> > >{ptr=0x2aefda0,node=0x2aefe90,count=1}
> 
>     "smoother: Aztec params" : Teuchos::RCP<std::vector<double, 
> std::allocator<double> > > = Teuchos::RCP<std::vector<double, 
> std::allocator<double> > >{ptr=0x2aefed0,node=0x2aefff0,count=1}
> 
>     "smoother: Block Chebyshev block list" : int* = 0
> 
>     "smoother: Block Chebyshev block starts" : int* = 0
> 
>     "smooth
> 

-- 
Ray Tuminaro                      phone: (925) 294-2564
MS 9159                           fax:   (925) 294-2234
Sandia National Laboratories      email: rstumin at sandia.gov
PO Box 969                        http://www.cs.sandia.gov/~rstumin
Livermore, CA 94551


-------------- next part --------------
A non-text attachment was scrubbed...
Name: ml_MultiLevelPreconditioner_Smoothers.cpp
Type: text/x-c++src
Size: 58825 bytes
Desc: not available
Url : https://software.sandia.gov/pipermail/trilinos-users/attachments/20110623/cae4d434/attachment.bin 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ml_MultiLevelPreconditioner_Test.cpp
Type: text/x-c++src
Size: 21175 bytes
Desc: not available
Url : https://software.sandia.gov/pipermail/trilinos-users/attachments/20110623/cae4d434/attachment-0001.bin 


More information about the Trilinos-Users mailing list