Dear dealers, I've faced the following problem.
As a test example I have a program which solves nondimensional heat equation using forward Euler method in time. The time step is chosen appropriately that solution does not blow up (CFL condition it is called?) dt = 0.4e-4. There are no hanging nodes, mesh is constant. 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. Problem - in this case after around 10 time steps solution blows up with the same time step used before (dt = 0.4e-4)! During these first successful steps norm of the solution for parallel version and for serial coincide up to 5 digits, but after some time in parallel version it blows up. This happens both running the code with 2 or 1 processors. Furthermore if I will set time step 1 order of magnitude smaller (dt = 0.4e-5) then everything will work fine. If take dt = 0.3,0.2,0.1 e-4 solution will fail. Could anybody provide explanation for that? Of course if needed I can provide both codes, they are simple. Regards, Slawa _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
