[CIG-SHORT] two pylith problems
Matthew Knepley
knepley at gmail.com
Wed Dec 5 17:14:52 PST 2007
On Dec 5, 2007 7:02 PM, Brad Aagaard <baagaard at usgs.gov> wrote:
> Chris-
>
> The two-cell examples don't use the asm preconditioner for a reason- it isn't
> appropriate for those problems because they are not fully constrained (in
> many cases rotations are possible). Try it on examples/3d/hex8 or
> examples/3d/tet4. You may not see much difference on these small problems,
> but we found a significant reduction in the number of iterations required for
> convergence for the strike-slip benchmark.
Brad is correct. However
1) Always give -petsc.ksp_monitor -petsc.ksp_view to determine what
is going on.
It is also possible to get different answers if you do not converge.
2) If you run with a dislocation and use ILU, you might get an error
with zero pivot
(but I can't tell unless you send the error message). These can
be worked around.
3) That problems looks like a problem with cohesive cells, but the
fixes should be
in the latest build. However, I cannot tell from the message.
It could also be a
mesh problem since it is very hard to check that the mesh is
valid. When I get
assertions and C++ exceptions, I see the error message. I am
not sure why your
output does not have it.
Matt
> Brad
>
>
>
> On Wednesday 05 December 2007, Christopher DiCaprio wrote:
> > I've encountered two problems in running the new pylith linux binary.
> >
> > Following the suggestion to use the asm preconditioner I tried it out
> > on one of the included examples. When I run examples/twocells/
> > twotet4/axialdisp.cfg I get different answers depending on the
> > preconditioner. Here are the relevant parts of the vtk files:
> >
> > pc_type = asm
> > -----------------------------------------------------------------
> > VECTORS displacements_t0.000000 double
> > -0.01 0 0
> > 2.6718e-19 0.0025 -0.00840477
> > 2.6718e-19 -0.00840477 -0.0025
> > 2.6718e-19 -0.0025 0.00840477
> > 0.01 0 0
> >
> > pc_type = jacobi
> > -----------------------------------------------------------------
> > VECTORS displacements_t0.000000 double
> > -0.01 0 0
> > -2.86102e-19 0.0025 1.70977e-17
> > 3.57628e-19 2.91236e-18 -0.0025
> > -2.86102e-19 -0.0025 1.34636e-17
> > 0.01 0 0
> >
> > Note the difference in the y-displacement.
> >
> > When I run examples/twocells/twotet4/displacement.cfg I run into
> > petsc errors when using pc_type = asm. See the attached file
> > dislocation.asm.log. There are no errors when using the jacobi
> > preconditioner.
> >
> >
> >
> > My other problem concerns a large mesh. I reported a problem using
> > this mesh to the list before and Brad thought it was a bug in the
> >
> > cohesive cells. With the new binary I get a different error:
> > >> /home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> >
> > python2.5/site-packages/pylith/meshio/MeshIO.py:45:read
> > -- meshiolagrit(info)
> > -- Reading finite-element mesh
> > Traceback (most recent call last):
> > File "<string>", line 1, in <module>
> > File "/home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> > python2.5/site-packages/pythia-0.8.1.4-py2.5.egg/mpi/__init__.py",
> > line 42, in mpistart
> > File "/home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> > python2.5/site-packages/pythia-0.8.1.4-py2.5.egg/pyre/applications/
> > __init__.py", line 41, in start
> > File "/home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> > python2.5/site-packages/pythia-0.8.1.4-py2.5.egg/pyre/applications/
> > Shell.py", line 139, in run
> > File "/home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> > python2.5/site-packages/pythia-0.8.1.4-py2.5.egg/pyre/applications/
> > SuperScript.py", line 49, in execute
> > File "/home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> > python2.5/site-packages/pythia-0.8.1.4-py2.5.egg/pyre/applications/
> > SuperScript.py", line 58, in main
> > File "/home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> > python2.5/site-packages/pythia-0.8.1.4-py2.5.egg/pyre/applications/
> > SuperScript.py", line 62, in runSubscript
> > File "/home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> > python2.5/site-packages/pythia-0.8.1.4-py2.5.egg/pyre/applications/
> > Application.py", line 42, in run
> > File "/home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> > python2.5/site-packages/pythia-0.8.1.4-py2.5.egg/pyre/applications/
> > Shell.py", line 139, in run
> > File "/home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> > python2.5/site-packages/pythia-0.8.1.4-py2.5.egg/mpi/Application.py",
> > line 118, in _onComputeNodes
> > File "/home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> > python2.5/site-packages/pythia-0.8.1.4-py2.5.egg/mpi/Application.py",
> > line 124, in onComputeNodes
> > File "/home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> > python2.5/site-packages/pylith/PyLithApp.py", line 90, in main
> > File "/home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> > python2.5/site-packages/pylith/topology/MeshImporter.py", line 81, in
> > create
> > File "/home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> > python2.5/site-packages/pylith/topology/MeshGenerator.py", line 96,
> > in _adjustTopology
> > File "/home/buildbot/install/x86_linux_single_nosched_binbot/lib/
> > python2.5/site-packages/pylith/faults/Fault.py", line 132, in
> > adjustTopology
> > File "faults.pyx", line 108, in faults.Fault.adjustTopology
> > RuntimeError: Caught unknown C++ exception.
> > --pyre-start: mpiexec: exit 1
> > /home/dicaprio/local/pylith-1.0.2-linux-i686/bin/pylith: /home/
> > dicaprio/local/pylith-1.0.2-linux-i686/bin/nemesis: exit 1
> >
> >
> >
> > Is this a problem with my mesh or a bug?
> >
> > Thanks,
> > Chris
>
>
> _______________________________________________
> CIG-SHORT mailing list
> CIG-SHORT at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-short
>
--
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which
their experiments lead.
-- Norbert Wiener
More information about the CIG-SHORT
mailing list