[aspect-devel] Aspect stops running after several time steps

Timo Heister heister at clemson.edu
Tue Jun 16 00:14:53 PDT 2015

Hey Lev,

I assume other setups/examples are running fine for you?

I don't know where the code is hanging and there is no easy way to
find out. A couple of ideas:
./ Try to disable the free surface and see if that fixes it. This
feature is tested a lot less than other parts of the code.
./ Try to make a smaller testcase: for example, do you need 32 MPI
jobs to see the problem?
./ Is it possible to run your setup (maybe on a coarser mesh) in debug
mode to see if you trigger any exceptions?
./ Does the same problem happen at the same timestep with a smaller CFL number?
./ Can you try the latest development version of ASPECT?

If everything else fails we need to run in a debugger to see where it
is hanging. For that it would be great to have a somewhat smaller
testcase that fails (I know that you tried already). If that doesn't
work you can always do "printf"-debugging (inserting print statements
to see which part of the code is the problem).

On Tue, Jun 16, 2015 at 1:22 AM, Lev Karatun <lev.karatun at gmail.com> wrote:
> Hi,
> one of the models that I'm running gets stuck after 2 time steps. The last
> message I see in the log file is: "*** Timestep 3:  t=236212 years".
> Aspect's processes are still running and taking up resources, but nothing
> happens. Similar model, but with 1.5 lower resolution and 1.5 smaller domain
> size in x direction runs just fine. I checked the memory usage - only 1/3 is
> being used, so this doesn't seem to be a problem. There is enough free space
> on hard drive, too.
> Could anyone please give me an advice on how I can find the cause of such
> stalling?
> Thanks in advance. (I attached the log file just in case)
> Best regards,
> Lev Karatun.
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel

Timo Heister

More information about the Aspect-devel mailing list