Dear Wolfgang, 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. 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.
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? Regards, Slawa Friday, June 19, 2009, 22:55, you wrote: >> Then I modified this code (following tutorial 17) to run it in >> parallel using PETSC (just as a sand-box be sure that I understood >> how the parallel tools work). Principally nothing have changed >> except CG solver (for "inversion" of the mass matrix) now is used >> from PETSC and it uses another preconditioner. WB> So does the solver work if you run the program on a single machine, or does WB> it work there as well? WB> PETSc's CG solver by default uses a BlockJacobi preconditioner which WB> computes an ILU of the locally stored block on each processor. In other WB> words, on a single machine, the BlockJacobi preconditioner is equivalent WB> to an ILU of the entire matrix, but if you have more than one machine WB> communicating via MPI, the BlockJacobi preconditioner moves away from ILU WB> and towards a simple Jacobi preconditioner, i.e. the preconditioner WB> becomes worse and worse as there are more processors. As a consequence, WB> you have to expect more linear iterations, and you have to make sure you WB> allow more iterations to achieve the same accuracy. WB> I would not be surprised if a lack of accuracy in the CG solver can lead to WB> instabilities even if the time step was otherwise small enough. WB> Best WB> W. -- Best regards, 21.06.2009 Svyatoslav Gladkov, Kharkov, UKRAINE 19:23 _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
