E4S-Project / testsuite

E4S test suite with validation tests
MIT License
19 stars 31 forks source link

omega-h test fails on perlmutter #49

Closed wspear closed 1 year ago

wspear commented 1 year ago

@cwsmith

The test defined here: https://github.com/E4S-Project/testsuite/tree/master/validation_tests/omega-h

Fails against the variant of omega-h installed by e4s 22.11 on perlmutter:

-- linux-sles15-zen3 / gcc@11.2.0 -------------------------------
omega-h@9.34.13~cuda~examples~ipo+mpi+optimize+shared+symbols~throw+trilinos~warnings+zlib build_system=cmake build_type=RelWithDebInfo
==> 1 installed package

With the following console output:


++ srun -n 2 ./castle
[0] number of vertices 8
[0]: Local vertex number 0 has global index 0
[0]: Local vertex number 1 has global index 2
[0]: Local vertex number 2 has global index 3
[0]: Local vertex number 3 has global index 6
[0]: Local vertex number 4 has global index 7
[0]: Local vertex number 5 has global index 8
[0]: Local vertex number 6 has global index 9
[0]: Local vertex number 7 has global index 10
[0]: local  0 4 0 6
[0]: global 1 7 0 9
[0]: local  1 1 3 7
[0]: global 3 2 6 10
[0]: local  2 3 2 5
[0]: global 6 6 3 8
[0]: local  3 2 4 5
[0]: global 7 3 7 8
[0]: local  4 5 4 6
[0]: global 9 8 7 9
[0]: local  5 5 6 7
[0]: global 10 8 9 10
[0]: local  6 3 5 7
[0]: global 13 6 8 10
[1] number of vertices 8
[1]: Local vertex number 0 has global index 0
[1]: Local vertex number 1 has global index 1
[1]: Local vertex number 2 has global index 2
[1]: Local vertex number 3 has global index 4
[1]: Local vertex number 4 has global index 5
[1]: Local vertex number 5 has global index 9
[1]: Local vertex number 6 has global index 10
[1]: Local vertex number 7 has global index 11
[1]: local  0 4 2 6
[1]: global 0 5 2 10
[1]: local  1 0 3 5
[1]: global 2 0 4 9
[1]: local  2 3 1 7
[1]: global 4 4 1 11
[1]: local  3 1 4 7
[1]: global 5 1 5 11
[1]: local  4 4 6 7
[1]: global 8 5 10 11
[1]: local  5 5 3 7
[1]: global 11 9 4 11
[1]: local  6 6 5 7
[1]: global 12 10 9 11
assertion array.size() == nents_[ent_dim] * ncomps failed at /tmp/lpeyrala/spack-stage/spack-stage-omega-h-9.34.13-rihdjj4qcpstotswnpa5ulldhcgseej2/spack-src/src/Omega_h_mesh.cpp +146
assertion array.size() == nents_[ent_dim] * ncomps failed at /tmp/lpeyrala/spack-stage/spack-stage-omega-h-9.34.13-rihdjj4qcpstotswnpa5ulldhcgseej2/spack-src/src/Omega_h_mesh.cpp +146
srun: error: nid001032: task 0: Aborted
srun: launch/slurm: _step_signal: Terminating StepId=3727136.74
slurmstepd: error: *** STEP 3727136.74 ON nid001032 CANCELLED AT 2022-11-21T20:23:15 ***
srun: error: nid001032: task 1: Aborted
cwsmith commented 1 year ago

Hi @wspear.

The castle test should be disabled. AFAIK, that code/example is no longer maintained. Is that the only test that fails?

wspear commented 1 year ago

@cwsmith We only had the castle test as part of the testsuite. Which example would be appropriate for sanity-checking purposes? fieldOnSquare requires a 'square.msh' input file which isn't provided with the source as far as I can tell. Is it also not maintained?

wspear commented 1 year ago

I generated square.msh for fieldOnSquare, added that to the test repo and switched to that test which runs successfully. This test needs to run on an installed omega-h so the install/test dependencies (gmsh) can't be relied on.

cwsmith commented 1 year ago

@wspear Sounds good. Thank you. I took a look at the field_on_square example and it is much simpler that will be easy to maintain.

Would you like a version of it that does not rely on Gmsh?