[aspect-devel] ASPECT on Mac

[Ian Rose]

Hi Sylvia,

​I haven't followed this thread in detail, but I notice that you are trying
to run the free surface cookbook.​  I can't really help with the
compile/linking problems, but the solver convergence issues look like they
may be related to the free surface.  Are you trying to run the cookbook
without modification?  The output that you have for the first step is
identical to what I get (on any of several systems), but the output for the
second step looks like it is quite different.  Notice the ~3 order of
magnitude jump in characteristic velocities, with a corresponding drop in
timesteps.  Looks to me like there are some serious stability issues
arising.

I know you have been having some problems with graphical output, but if you
do get that working, perhaps a look at the solution will give a clue as to
what is going on.  Do other cookbooks run without difficulty?

Best,
Ian


On Wed, Oct 1, 2014 at 1:39 PM, Sylvia Rockel <sylvia.rockel at fu-berlin.de>
wrote:

> Hi Timo,
>
> > Do you get something else when running in debug mode?
> I don't think so. Looks very similar, although the exception is again
> displayed very cryptically.
>
>
> > Also, it looks like you are running ASPECT 1.1. Can you install the
> > newest development version from git?
> I tried, but it gave an compile error, since the link.h couldn't be found.
> File is attached.
>
> Best,
> Sylvia
>
> > On Wed, Oct 1, 2014 at 3:06 PM, Sylvia Rockel
> > <sylvia.rockel at fu-berlin.de> wrote:
> >> Dear Timo,
> >>
> >> thanks a lot for the hint.
> >>
> >>> Do you have the rights to create files in the directory you are running
> >>> programs at?
> >>
> >> Perfect, deal.ii problem seems to be solved now. You were right, as I
> >> was
> >> logged in as admin, I thought I've got all rights, but I was wrong.
> >> Changed
> >> permissions and now it's running smoothly.
> >>
> >> But I still have this ASPECT problem while running mpi. In direct run,
> >> everything works just fine. Could you please have a look on the file
> >> attached?
> >>
> >> Best,
> >> Sylvia
> >>
> >>
> >>> On Wed, Oct 1, 2014 at 7:57 AM, Sylvia Rockel
> >>> <sylvia.rockel at fu-berlin.de> wrote:
> >>>> Dear Timo,
> >>>>
> >>>> I've re-installed my os x now and installed aspect again, but it's
> >>>> still
> >>>> the same.
> >>>>
> >>>> Concerning deal.ii step-32:
> >>>> It still doesn't work with mpi. I also tested to run it directly, but
> >>>> it's
> >>>> the same error, see attached file. I really have to set the graphical
> >>>> output to false and then it works fine.
> >>>>
> >>>> Concerning aspect:
> >>>> I'm currently testing all cookbooks for whether they work with mpi or
> >>>> not.
> >>>> As far as I've come right now, only stokes works with mpi. The others,
> >>>> i. e.:
> >>>>
> >>>> - composition-active,
> >>>> - convection-box
> >>>> - convection-box-3d
> >>>> - free-surface
> >>>>
> >>>> all get sooner or later the same exception, but I like to run them
> >>>> directly later, to determine, whether this is also the case without
> >>>> mpi,
> >>>> since it's something concerning the residual.
> >>>>
> >>>> So, I'll give you an update as soon as I tested everything.
> >>>>
> >>>> Best,
> >>>> Sylvia
> >>>>
> >>>>> Dear Timo,
> >>>>>
> >>>>> that's what I thought. You are perfectly right, if I run aspect
> >>>>> directly
> >>>>> everything works just fine. Honestly I haven't tried it with deal,
> >>>>> 'cause
> >>>>> it
> >>>>> run through if I canceled the graphical output. You may find the
> >>>>> output
> >>>>> of
> >>>>> "which mpirun" etc. attached. As you can see, it's the one from deal.
> >>>>> Same
> >>>>> is true for mpicc, mpic++, and mpif77/mpif90.
> >>>>>
> >>>>> The only additional mpiexe are in the petsc bin, but this is turned
> >>>>> of
> >>>>> for
> >>>>> aspect anyways. I'm slightly out of ideas right now.
> >>>>>
> >>>>> Thanks for your quick reply.
> >>>>>
> >>>>> Best,
> >>>>> Sylvia
> >>>>>
> >>>>>
> >>>>>
> >>>>>> Dear Sylvia,
> >>>>>>
> >>>>>> all these errors (and weird output) is related to your mpirun
> >>>>>> problem.
> >>>>>> It looks like the mpirun isn't from the same MPI library that is
> >>>>>> used
> >>>>>> to compile the code. I assume things work correctly if you run
> >>>>>> aspect
> >>>>>> or step-32 directly (so "./aspect cookbooks/convection.box.prm")?
> >>>>>>
> >>>>>> Can you check if there is an mpirun (or mpiexec) in one of the
> >>>>>> directories installed by the deal.II bundle?
> >>>>>>
> >>>>>> Can you please post the output of "which mpirun" and "ldd
> >>>>>> libdeal_II.so" (the last one needs to be run in the lib/ directory
> >>>>>> of
> >>>>>> the installation folder of deal.II)?
> >>>>>>
> >>>>>>
> >>>>>> --
> >>>>>> Timo Heister
> >>>>>> http://www.math.clemson.edu/~heister/
> >>>>>> _______________________________________________
> >>>>>> Aspect-devel mailing list
> >>>>>> Aspect-devel at geodynamics.org
> >>>>>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
> >>>>>>_______________________________________________
> >>>>> Aspect-devel mailing list
> >>>>> Aspect-devel at geodynamics.org
> >>>>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
> >>>>
> >>>> _______________________________________________
> >>>> Aspect-devel mailing list
> >>>> Aspect-devel at geodynamics.org
> >>>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
> >>>
> >>>
> >>>
> >>> --
> >>> Timo Heister
> >>> http://www.math.clemson.edu/~heister/
> >>> _______________________________________________
> >>> Aspect-devel mailing list
> >>> Aspect-devel at geodynamics.org
> >>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
> >>>
> >>
> >> _______________________________________________
> >> Aspect-devel mailing list
> >> Aspect-devel at geodynamics.org
> >> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
> >
> >
> > --
> > Timo Heister
> > http://www.math.clemson.edu/~heister/
> > _______________________________________________
> > Aspect-devel mailing list
> > Aspect-devel at geodynamics.org
> > http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
> >
>
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/aspect-devel/attachments/20141001/365e1e97/attachment.html>