[aspect-devel] Floating Point Exception error for simple setup

FELIPE ORELLANA ROVIROSA f_orellana at berkeley.edu
Thu Mar 10 17:44:37 PST 2016


 Hi,

     Thanks for the suggestions.

     I preferred to recreate the 'build' directory (which is the last one
to be created before compilation) from scratch to make things clean. I
succeeded in creating the Aspect setup files again. Then I executed the
command

     cmake -D ASPECT_USE_FP_EXCEPTIONS=OFF .

  and then I compiled ASPECT

   make -j8            (recommended for Aspect on Stampede)

  It all compiled successfully (100%, successful linkages, etc).

  I rerun the 3D convection in a cube example, I did it twice, now I got
these errors (always after like 50 seconds of run):

  ibrun /home1/04020/unfelipe/packages/ASPECT/build/aspect examAsp.prm
TACC: Starting up job 6710481
TACC: Setting up parallel environment for MVAPICH2+mpispawn.
TACC: Starting parallel tasks...
-----------------------------------------------------------------------------
-- This is ASPECT, the Advanced Solver for Problems in Earth's ConvecTion.
--     . version 1.4.0-pre
--     . running in DEBUG mode
--     . running with 256 MPI processes
--     . using Trilinos
-----------------------------------------------------------------------------


-----------------------------------------------------------------------------
The output directory </home1/04020/unfelipe/exAou/> provided in the input
file appears not to exist.
ASPECT will create it for you.
-----------------------------------------------------------------------------


Number of active cells: 512 (on 4 levels)
Number of degrees of freedom: 20,381 (14,739+729+4,913)

*** Timestep 0:  t=0 seconds
   Solving temperature system... 0 iterations.
   Rebuilding Stokes preconditioner...
   Solving Stokes system... 27 iterations.


--------------------------------------------------------
An error occurred in line <291> of file
</home1/04020/unfelipe/packages/ASPECT/source/simulator/simulator_access.cc>
in function
    const dealii::FiniteElement<dim>&
aspect::SimulatorAccess<dim>::get_fe() const [with int dim = 3]
The violated condition was:
    simulator->dof_handler.n_locally_owned_dofs() != 0
The name and call sequence of the exception was:
    ExcMessage("You are trying to access the FiniteElement before the DOFs
have been " "initialized. This may happen when accessing the Simulator from
a plugin " "that gets
 executed early in some cases (like material models) or from " "an early
point in the core code.")
Additional Information:
You are trying to access the FiniteElement before the DOFs have been
initialized. This may happen when accessing the Simulator from a plugin
that gets executed early in som
e cases (like material models) or from an early point in the core code.

Stacktrace:
-----------
#0  /home1/04020/unfelipe/packages/ASPECT/build/aspect:
aspect::SimulatorAccess<3>::get_fe() const
#1  /home1/04020/unfelipe/packages/ASPECT/build/aspect:
aspect::MeshRefinement::ThermalEnergyDensity<3>::execute(dealii::Vector<float>&)
const
#2  /home1/04020/unfelipe/packages/ASPECT/build/aspect:
aspect::MeshRefinement::Manager<3>::execute(dealii::Vector<float>&) const
#3  /home1/04020/unfelipe/packages/ASPECT/build/aspect:
aspect::Simulator<3>::refine_mesh(unsigned int)
#4  /home1/04020/unfelipe/packages/ASPECT/build/aspect:
aspect::Simulator<3>::run()
#5  /home1/04020/unfelipe/packages/ASPECT/build/aspect: main
--------------------------------------------------------

--------------------------------------------------------

...

Hope this can clarify the work flow of Aspect 1.4.0-pre, so you can find
the solution,

i greatly appreciate your help,

Felipe

On Thu, Mar 10, 2016 at 1:51 PM, Rene Gassmoeller <
rene.gassmoeller at mailbox.org> wrote:

> Hi Felipe,
> 'cmake' is only a helper program that sets up a proper Makefile for
> compiling aspect. To actually change your executable you will always need
> to recompile aspect after changing things with cmake (see
> https://cmake.org/overview/ for a more lengthy explanation). Try typing
> 'make' in your build folder (/home1/04020/unfelipe/packages/ASPECT/build/)
> and start the model again. This should help.
>
> @Timo: I know the FP_Exceptions are very helpful in debugging, but it
> seems the test for their availability is really fragile and breaks on many
> clusters. This is basically not our fault, but I have the feeling half of
> the crashes currently reported on github or the mailing list are caused by
> them either being falsely activated by the build system, or manually
> activated although they are not available on the system. Do you think it
> would help to switch them off by default and hide their cmake option behind
> an 'Advanced' switch? It would save us a lot of meaningless crashes and
> mails ;-). I have started to simply disable them manually on every new
> system, since I only need them on the ones that I use for developing.
>
> Best,
> Rene
>
>
>
> On 03/10/2016 03:11 PM, FELIPE ORELLANA ROVIROSA wrote:
>
>
> Hi Timo,
>
>     Thanks for having replied.
>
>     Here dealing with the FPE thing..
>
> I executed the command you advised me:
>
> login1.stampede(37)$ cmake -D ASPECT_USE_FP_EXCEPTIONS=OFF .
>
> and it worked to this:
> -- ====================================================
> -- ============ Configuring ASPECT ====================
> -- ====================================================
> -- Found deal.II version 8.5.0-pre at
> '/home1/04020/unfelipe/packages/deal.II/lib/cmake/deal.II'
> -- Using PETSc = 'OFF'
> -- Added 1 tests.
> -- Enabling dynamic loading of plugins from the input file
> -- Enabling checking of compatible deal.II library when loading plugins
> -- Autopilot invoked
> -- Run   $ make info  to print a detailed help message
> -- Writing config into detailed.log...
> -- Configuring done
> WARNING: Target "aspect" requests linking to directory
> "/opt/apps/intel/13/composer_xe_2013.2.146/mkl/lib/intel64".  Targets may
> link only to libraries.  CMake is dropping the item.
> WARNING: Target "quick_mpi" requests linking to directory
> "/opt/apps/intel/13/composer_xe_2013.2.146/mkl/lib/intel64".  Targets may
> link only to libraries.  CMake is dropping the item.
> -- Generating done
> -- Build files have been written to:
> /home1/04020/unfelipe/packages/ASPECT/build
>
> Then, I tried running both the 3D and the 2D examples in parallel. For
> this later one this is the error i got in the output file:
>
> ibrun /home1/04020/unfelipe/packages/ASPECT/build/aspect c2Dbox.prm
> TACC: Starting up job 6708828
> TACC: Setting up parallel environment for MVAPICH2+mpispawn.
> TACC: Starting parallel tasks...
>
> -----------------------------------------------------------------------------
> -- This is ASPECT, the Advanced Solver for Problems in Earth's ConvecTion.
> --     . version 1.4.0-pre
> --     . running in DEBUG mode
> --     . running with 256 MPI processes
> --     . using Trilinos
>
> -----------------------------------------------------------------------------
>
> [c558-501.stampede.tacc.utexas.edu:mpi_rank_1][error_sighandler] Caught
> error: Floating point exception (signal 8)
> [c558-501.stampede.tacc.utexas.edu:mpi_rank_5][error_sighandler] Caught
> error: Floating point exception (signal 8)
> [c558-501.stampede.tacc.utexas.edu:mpi_rank_2][error_sighandler] Caught
> error: Floating point exception (signal 8)
> [c558-501.stampede.tacc.utexas.edu:mpi_rank_6][error_sighandler] Caught
> error: Floating point exception (signal 8)
> [c558-501.stampede.tacc.utexas.edu:mpi_rank_13][error_sighandler] Caught
> error: Floating point exception (signal 8)
> [c558-501.stampede.tacc.utexas.edu:mpi_rank_9][error_sighandler] Caught
> error: Floating point exception (signal 8)
> [c558-501.stampede.tacc.utexas.edu:mpi_rank_10][error_sighandler] Caught
> error: Floating point exception (signal 8)
> [c558-501.stampede.tacc.utexas.edu:mpi_rank_14][error_sighandler] Caught
> error: Floating point exception (signal 8)
> [c558-501.stampede.tacc.utexas.edu:mpi_rank_0][error_sighandler] Caught
> error: Floating point exception (signal 8)
> ...
>
> **************************************
>  This is my module list:
>       Currently Loaded Modules:
>   1) xalt/0.6   2) TACC   3) git/2.7.0   4) gcc/4.7.1   5)
> mvapich2/1.9a2   6) mkl/13.0.2.146   7) cmake/3.1.0
>
>    After having executed the command for FPE OFF you advised me, the only
> things that changed in the build directory were: tests dir, CMake dir and
> Makefile. The ASPECT executable remains nevertheless unchanged..  I wonder
> if this is right..
>
> hope you can find out what my problem is,
> hope you can help me,
> thanks,
>
> Felipe
>
> On Wed, Mar 9, 2016 at 5:49 AM, Timo Heister <heister at clemson.edu> wrote:
>
>> Hey Felipe,
>>
>> the floating point exception feature is still a somewhat fragile
>> option. You can turn it off with
>>
>>  cmake -D ASPECT_USE_FP_EXCEPTIONS=OFF .
>>
>> in your build directory.
>>
>> To improve this feature, can you tell me what compiler versions you
>> are using? Do you see this problem also when you run the 2d
>> convection-box on one processor?
>>
>> Thanks,
>> Timo
>>
>>
>>
>> On Wed, Mar 9, 2016 at 3:21 AM, FELIPE ORELLANA ROVIROSA
>> <f_orellana at berkeley.edu> wrote:
>> >
>> >  Hi all,
>> >
>> >        I am trying to run Aspect in parallel for 3D computations, and I
>> am
>> > getting errors related to 'floating point exception'. I know other
>> people in
>> > the community has suffered this same problem, in different situations,
>> > versions and machines.
>> >
>> >       For instance, I am using version 1.4.0-pre on TACC Stampede. When
>> > trying to run the 3D convection example, after short running times I
>> get the
>> > error:
>> >
>> >
>> >
>> -----------------------------------------------------------------------------
>> > -- This is ASPECT, the Advanced Solver for Problems in Earth's
>> ConvecTion.
>> > --     . version 1.4.0-pre
>> > --     . running in DEBUG mode
>> > --     . running with 256 MPI processes
>> > --     . using Trilinos
>> >
>> -----------------------------------------------------------------------------
>> >
>> > [c557-404.stampede.tacc.utexas.edu:mpi_rank_2][error_sighandler] Caught
>> > error: Floating point exception (signal 8)
>> > [c557-404.stampede.tacc.utexas.edu:mpi_rank_5][error_sighandler] Caught
>> > error: Floating point exception (signal 8)
>> > [c557-404.stampede.tacc.utexas.edu:mpi_rank_8][error_sighandler] Caught
>> > error: Floating point exception (signal 8)
>> > [c557-404.stampede.tacc.utexas.edu:mpi_rank_10][error_sighandler]
>> Caught
>> > error: Floating point exception (signal 8)
>> > [c557-404.stampede.tacc.utexas.edu:mpi_rank_12][error_sighandler]
>> Caught
>> > error: Floating point exception (signal 8)
>> > [c557-404.stampede.tacc.utexas.edu:mpi_rank_9][error_sighandler] Caught
>> > error: Floating point exception (signal 8)
>> > [c557-404.stampede.tacc.utexas.edu:mpi_rank_1][error_sighandler] Caught
>> > error: Floating point exception (signal 8)
>> > [c557-404.stampede.tacc.utexas.edu:mpi_rank_6][error_sighandler] Caught
>> > error: Floating point exception (signal 8)
>> > [c557-404.stampede.tacc.utexas.edu:mpi_rank_15][error_sighandler]
>> Caught
>> > error: Floating point exception (signal 8)
>> > [c557-404.stampede.tacc.utexas.edu:mpi_rank_4][error_sighandler] Caught
>> > error: Floating point exception (signal 8)
>> > [c557-404.stampede.tacc.utexas.edu:mpi_rank_7][error_sighandler] Caught
>> > error: Floating point exception (signal 8)
>> > ...
>> >
>> >     Is there a vulnerability with the Aspect version I am using?
>> >   or the problem lies on another issue?
>> >
>> >     What would be the recommendation?
>> >
>> > I greatly appreciate your help,
>> > cheers,
>> >
>> > Felipe
>> >
>> > _______________________________________________
>> > 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 listAspect-devel at geodynamics.orghttp://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/20160310/6c0e48d4/attachment-0001.html>


More information about the Aspect-devel mailing list