[petsc-users] Advice Being Sought

[Barry Smith]

On Apr 22, 2013, at 11:26 AM, David Scott <d.scott at ed.ac.uk> wrote:

> Thanks for the suggestion.
> 
> I had tried '-pc_type gamg -pc_gamg_agg_nsmooths 1' with an earlier version of the code without success. I have tried it again but I get NaN's after only 90 time steps whereas with block Jacobi it runs quite happily for 36,000 time steps and produces physically sensible results.

  David,

    We would be very interested in determining what is "going wrong" with the solver here since we hope to make it robust.  Would it be possible for you to use a MatView() and VecView() on the matrix and the right hand side with a binary viewer when it "goes bad" and send us the resulting file? 

   Barry

We'd run the gamg solver on your matrix and track down what is happening.


> 
> David
> 
> On 22/04/2013 13:12, Matthew Knepley wrote:
>> On Mon, Apr 22, 2013 at 7:56 AM, David Scott <d.scott at ed.ac.uk
>> <mailto:d.scott at ed.ac.uk>> wrote:
>> 
>>    Hello,
>> 
>>    I am working on a fluid-mechanical code to solve the two-phase
>>    Navier–Stokes equations with levelset interface capturing. I have
>>    been asked to replace the pressure calculation which uses the SOR
>>    and Jacobi iterative schemes with a Krylov subspace method. I have
>>    done this and the code is working but as I have never used PETSc
>>    before I would like to know if improvements to my code, or the run
>>    time parameters that I am using, could be made.
>> 
>>    I am using GMRES with a Block Jacobi pre-conditioner. I have tried
>>    Conjugate Gradient with a Block Jacobi pre-conditioner but it
>>    diverges. If I use GMRES for the first few thousand time steps and
>>    then swap to CG it does converge but the speed of execution is
>>    somewhat reduced.
>> 
>> 
>> Krylov methods do not work with preconditioners. You have a Poisson
>> problem, so as abundantly documented in the literature, you should use
>> multigrid. The easiest thing to try is
>> 
>>   -pc_type gamg -pc_gamg_agg_nsmooths 1
>> 
>>    Thanks,
>> 
>>      Matt
>> 
>>    I have attached relevant excerpts from the code.
>> 
>>    Yours sincerely,
>> 
>>    David Scott
>>    --
>>    Dr. D. M. Scott
>>    Applications Consultant
>>    Edinburgh Parallel Computing Centre
>>    Tel. 0131 650 5921
>> 
>>    The University of Edinburgh is a charitable body, registered in
>>    Scotland, with registration number SC005336.
>> 
>> 
>> 
>> 
>> --
>> What most experimenters take for granted before they begin their
>> experiments is infinitely more interesting than any results to which
>> their experiments lead.
>> -- Norbert Wiener
> 
> 
> -- 
> Dr. D. M. Scott
> Applications Consultant
> Edinburgh Parallel Computing Centre
> Tel. 0131 650 5921
> 
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.