Hi Joan,
sorry, this got delayed since you posted as a non-member to the mailing list.
I'm glad to hear that things seem to work out by now.
> => Where can I find documentation on the command line arguments passed via
> the CMakeLists.txt
> file(dumux/test/porousmediumflow/richards/implicit/lens/CMakeLists.txt) ? I
> am thinking about " --script fuzzy" "--zeroThreshold {"process rank":100}"
> etc.
Documentation for the dune_add_test ,acro is here:
https://dune-project.org/sphinx/content/sphinx/core-2.6/commands/dune_add_test.html
Concerning runtest.py, you can execute
python runtest.py -h
in dumux/bin/testing or have a look at the file itself at
https://git.iws.uni-stuttgart.de/dumux-repositories/dumux/blob/master/bin/testing/runtest.py
> => Is the Grid.Overlap=0 parameter only needed for the YaspGrid type ? More
> generally, where can I find documentation about the different grids and their
> use ?
Yes to your first question, where the improtant thing is that you have to set 0
overlap for Box and 1 for cell-centered discretizations. Concerning your second
question, the best source would be the Dune documentation, a good starting
point might be
https://dune-project.org/doc/grids/
It might also be helpful to look at Section 6.6 of our handbook
https://dumux.org/documents/dumux-handbook-master.pdf
Kind regards
Bernd
--
_______________________________________________________________
Bernd Flemisch phone: +49 711 685 69162
IWS, Universität Stuttgart fax: +49 711 685 60430
Pfaffenwaldring 61 email: be...@iws.uni-stuttgart.de
D-70569 Stuttgart url: www.hydrosys.uni-stuttgart.de
_______________________________________________________________
________________________________
Von: Dumux <dumux-boun...@listserv.uni-stuttgart.de> im Auftrag von Joan Delort
Ylla <delo...@erdw.ethz.ch>
Gesendet: Montag, 20. Mai 2019 16:07:29
An: dumux@listserv.uni-stuttgart.de
Betreff: Re: [DuMuX] [CCS] Parallel computation with Box Discretization method
Dear Simon, DuMuX users,
Here's the testing procedure (gcc/6.3.0 --- cmake/3.5.2 --- open-mpi/1.5.6 ---
dumux/3.0 --- dune/2.6).
I "build" the overall dumux installation repo (the one containing also the dune
modules folders) as following (cmake.opts is the classic one) :
$ ./dune-common/bin/dunecontrol --opts=dumux/cmake.opts all
Then I go into the dumux folder and run cmake :
$ cd dumux
$ mkdir build-cmake && cd build-cmake
$ cmake ..
Then I go into the test folder of interest
(dumux/test/porousmediumflow/richards/implicit/lens) and try to compile the
tests that can be found in the CMakeLists.txt file :
$ make test_richards_lens_box
---> compiles fine
$ make test_richards_lens_box_parallel_ug
---> compiles fine
$ make test_richards_lens_box_parallel_alu
---> compiles fine
I try then to run those tests on the cluster by running the following command :
$ submit "mpirun -n 4 /path/to/problem/file /path/to/input/file"
Reassuringly, the tests seem to run fine on the cluster, however, some
questions :
=> Where can I find documentation on the command line arguments passed via the
CMakeLists.txt file
(dumux/test/porousmediumflow/richards/implicit/lens/CMakeLists.txt) ? I am
thinking about " --script fuzzy" "--zeroThreshold {"process rank":100}" etc.
=> Is the Grid.Overlap=0 parameter only needed for the YaspGrid type ? More
generally, where can I find documentation about the different grids and their
use ?
Thanks for your help,
Joan
On 26/04/2019 15:12, Simon Scholz wrote:
Dear Joan,
I am no expert in running things in parallel, but your setup seems fine to me.
What DuMuX version are you using? I am assuming it is release/3.0 with dune
2.6, correct?
In general it is possible to run a simulation using the box-method in parallel
with DuMuX. We test this e.g. with the richards tests located in
/test/porousmediumflow/richards/implicit/lens
Maybe you can try and build those (parallel) tests and see if they run with the
setup on you cluster. If this works then they can serve as a guideline. If this
does not work my last idea would be to check the grid that you are using. Maybe
you can try to change to using ALU or UG and see if the problem persists.
My best,
Simon
On 25.04.19 20:07, Joan Delort Ylla wrote:
Dear DuMuX users,
I am using a CCS model that includes hydrogen impurities and biofilm growth and
transport.
I inherit from the 2pncNIMin problem.
To be able to easily compute pressure gradients (needed for biofilm
attachment/detachment) I switched from the CCTpfa method to the Box method.
My application compiles, links and executes fine (please note that until now a
succesful simulation would take about 2h of clock time on 40 cores).
I run my simulations on a cluster in parallel using open-mpi/1.5.6, a typical
job submission being :
$ submit "mpirun -n 40 /path/to/problem/file /path/to/input/file"
At first I got the following error message (at runtime):
Dune reported error: Dune::InvalidStateException
[BoxFVGridGeometry:/path/to/dumux/root/dumux/dumux/discretization/box/fvgridgeometry.hh:125]:
The box discretization method only works with zero overlap for parallel
computations. Set the parameter "Grid.Overlap" in the input file. ---> Abort!
Which I solved by adding the following to my input file:
[Grid]
#other parameters
Overlap = 0
However, after doing so, the first time step of the simulation will run forever
and output the following error:
Assemble: r(x^k) = dS/dt + div F - q; M = grad r^MSolve: M deltax^k =
r^[[KCould not build any aggregates. Probably no connected nodes.
So here is my question :
Is it possible to run a simulation in parallel with the Box discretization
method ? And if so, where should I look for a way to do it ?
Thanks for your help, and I'll be happy to provide you with any extra
information you might need.
Joan
_______________________________________________
Dumux mailing list
Dumux@listserv.uni-stuttgart.de<mailto:Dumux@listserv.uni-stuttgart.de>
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
--
_______________________________________________________________________
Simon Scholz
Department of Hydromechanics and Modelling of Hydrosystems (LH2)
Simon Scholz phone: +49 711 685 67015
IWS, University of Stuttgart fax: +49 711 685 60430
Pfaffenwaldring 61 email:
simon.sch...@iws.uni-stuttgart.de<mailto:simon.sch...@iws.uni-stuttgart.de>
D-70569 Stuttgart url:
www.hydrosys.uni-stuttgart.de<http://www.hydrosys.uni-stuttgart.de>
_______________________________________________________________________
_______________________________________________
Dumux mailing list
Dumux@listserv.uni-stuttgart.de
https://listserv.uni-stuttgart.de/mailman/listinfo/dumux
