Greetings!

   Well, when I say up to 5 digits - I mean that while looking on the
   printouts of both solutions (where the double are by default printed
   to std::out with 5-6 digits ) I see that they coincide for the
   few first time steps.

   Of course in the solution control I've specified tolerances from
   1.e-12 to 1.e-30. Did not work, still "diverge".

Regards,
Slawa

Friday, June 26, 2009, 2:23, you wrote:


WB> Slawa,

>>    Well, that was my guess also that minor changes in the solution of
>>    the linear system might lead to these instabilities with the
>>    explicit solution. In fact solution "diverge" even if I run program
>>    it on single process.

WB> OK, so that at least gives you a better handle on debugging the problem.
WB> Debugging things that don't work in parallel is a pain.


>>    Probably these differences in the solution I 
>>    get out of PETSC CG solver are quite small, at least vector norm of
>>    the solution (up to 5 digits) coincide for the first few steps with
>>    the solution I get using deal's CG.

WB> So what happens if you tighten the convergence tolerance in your 
WB> SolverControl object? 5 digits appears not enough to me.


>>    How do you think which settings I should set for PETSC CG solver so
>>    that it will run with the same parameters as deal.II build-in CG
>>    solver?

WB> I don't exactly know -- all of these iterative solvers have slightly 
WB> different ways to compute residual norms (some use the preconditioned 
WB> residual, some the unpreconditioned) and the only way I see in your case
WB> is to start to work with very small tolerances. You can then start to 
WB> relax the convergence tolerance until you see that the time stepping 
WB> scheme starts to diverge.

WB> Best
WB>  W.

_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii

Reply via email to