On 3 Jun 2016, at 16:21, Anna Kalogirou <A.Kalogirou@leeds.ac.uk> wrote:

Hi Lawrence,

Thanks for the suggestions. Indeed, it doesn’t give any segmentation faults when running it from my mac laptop. On my linux machine, I ran firedrake-update, firedrake-clean and make alltest without any problems, but it still complains about those segmentation faults. Do you have any idea what the problem on linux might be?

Not sure.  I would run under gdb and see where the segfault occurs:

gdb --args python buoy-swe.py
(gdb) run
... wait for segfault
(gdb) bt

Now regarding the code itself, I fixed a few mistakes and updated it on bitbucket. It is now running but it’s blowing up. I suspect is due to bad preconditioning (or lack of any). How can I set up the preconditioner for the mixed system via setting the correct PETSc options?

If you're blowing up, but the solves are converging, then preconditioning is not likely to be the problem.  Instead the numerics might be screwy.  Or, by blowing up, do you mean that the solvers don't converge?

Lawrence
_______________________________________________
firedrake mailing list
firedrake@imperial.ac.uk
https://mailman.ic.ac.uk/mailman/listinfo/firedrake