Fluidity

ASCII GMSH files broken on CX1

Bug #1173592 reported by Jon Hill
14
This bug affects 2 people
Affects Status Importance Assigned to Milestone
Fluidity
Fix Released
Critical
Jon Hill

Bug Description

The intel compiler seems to struggle with ASCII GMSH files. At the last dev meeting we agreed the following:
 - Make all GMSH files in tests binary
 - set up a new test for ASCII GMSH that will pass on CX1 *if* an error message is displayed and pass on other machines if the file is read successfully.
 - #IFDEF around ASCII reader for intel compilers to display and error message that binary should be used and exit

Everyone happy with that?

Tags: cx1 gmsh

Related branches

lp://staging/~fluidity-core/fluidity/gmsh_intel
Tim Greaves: Approve
Diff: 4085 lines (+1935/-1527)
60 files modified
femtools/Read_GMSH.F90 (+8/-0)
tests/1mat-shocktube-gmsh/Makefile (+1/-1)
tests/3material-droplet-p1dgp2-test-cty-cv-rhop1dg-3d/Makefile (+1/-1)
tests/3material-droplet-p1dgp2-test-cty-cv-rhop1dg/Makefile (+1/-1)
tests/Helmholtz-anisotropic-scalar-mms-p1p1cg-structured/Helmholtz-anisotropic-scalar-mms-p1p1cg-structured.xml (+1/-1)
tests/Helmholtz-anisotropic-scalar-mms-p1p1cg-structured/Makefile (+4/-4)
tests/Helmholtz-anisotropic-vector-mms-p1p1cg-unstructured/Helmholtz-anisotropic-vector-mms-p1p1cg-unstructured.xml (+13/-13)
tests/Helmholtz-anisotropic-vector-mms-p1p1cg-unstructured/Makefile (+4/-4)
tests/Stokes_may_sinker_stepxy/Makefile (+1/-1)
tests/Stokes_may_solcx_stepx/Makefile (+1/-1)
tests/Stokes_may_solky_exponential/Makefile (+1/-1)
tests/bed_shear_from_grad_u_CG/Makefile (+1/-1)
tests/bed_shear_from_grad_u_DG/Makefile (+1/-1)
tests/darcy_p0p1_test_cty_cv_pressBCinlet/Makefile (+2/-2)
tests/darcy_p0p1_test_cty_cv_pressBCinlet_2phase_constant_volf_and_abs/Makefile (+2/-2)
tests/darcy_p0p1_test_cty_cv_pressBCinlet_advect_tracer_p0_steptwochannel/Makefile (+1/-1)
tests/darcy_p0p1_test_cty_cv_pressBCinlet_advect_tracer_p1cg/Makefile (+2/-2)
tests/darcy_p0p1_test_cty_cv_pressBCinlet_advect_tracer_p1cv/Makefile (+2/-2)
tests/darcy_p0p1_test_cty_cv_pressBCinlet_advect_tracer_p1cv_steptwochannel/Makefile (+1/-1)
tests/darcy_p0p1_test_cty_cv_pressBCinlet_advect_tracer_p1cv_steptwochannel_adaptive/Makefile (+1/-1)
tests/darcy_p0p1_test_cty_cv_pressBCinlet_advect_tracer_p1dg/Makefile (+2/-2)
tests/darcy_p0p1_test_cty_cv_pressBCinlet_rotated_domain/Makefile (+1/-1)
tests/darcy_p0p1_test_cty_cv_velBCinlet/Makefile (+2/-2)
tests/darcy_p0p1cv_pressBCinlet/Makefile (+2/-2)
tests/darcy_p0p1cv_pressBCinlet_1solidphase/Makefile (+2/-2)
tests/darcy_p0p1cv_pressBCinlet_rotated_domain/Makefile (+1/-1)
tests/darcy_p0p1cv_velBCinlet/Makefile (+2/-2)
tests/darcy_p0p1cv_velBCinlet_1solidphase/Makefile (+2/-2)
tests/darcy_p0p1cv_velBCinlet_rotated_domain/Makefile (+1/-1)
tests/darcy_p1dgp2_pressBCinlet/Makefile (+2/-2)
tests/darcy_p1dgp2_test_cty_cv_pressBCinlet/Makefile (+2/-2)
tests/darcy_p1dgp2_velBCinlet/Makefile (+2/-2)
tests/detectors_parallel_adaptivity/src/square.msh (+1659/-1431)
tests/divergence_free_velocity/Makefile (+1/-1)
tests/divergence_free_velocity_press_cg_test_cty_cv/Makefile (+1/-1)
tests/gravity-col-press_cg_test_cty_cv/Makefile (+1/-1)
tests/mmat-balancepressure-p1dgp2-test-cty-cv-rhop1dg-3d/Makefile (+1/-1)
tests/mmat-balancepressure-p1dgp2-test-cty-cv-rhop1dg/Makefile (+1/-1)
tests/mmat-gravity-col-press_cg_test_cty_cv/Makefile (+1/-1)
tests/mms_les_second_order_p2p1/Makefile (+3/-6)
tests/mms_tracer_cv_3d_adv_excludemass/Makefile (+4/-4)
tests/mms_tracer_p2cv_3d_adv_excludemass/Makefile (+4/-4)
tests/pressure_fs_topbottom/Makefile (+1/-1)
tests/pydiag_material_phase_support_multiple_velocity_children/Makefile (+1/-1)
tests/simple_vector_null_space/Makefile (+1/-1)
tests/sinking_velocity/Makefile (+1/-1)
tests/sinking_velocity_cv/Makefile (+1/-1)
tests/sloshing_tank/Makefile (+1/-1)
tests/temporal_max_vector/Makefile (+1/-1)
tests/test_gmsh_intel/Makefile (+4/-0)
tests/test_gmsh_intel/boring.flml (+112/-0)
tests/test_gmsh_intel/surface.msh (+19/-0)
tests/test_gmsh_intel/test_gmsh_intel.xml (+40/-0)
tests/test_vorticity_2d/Makefile (+1/-1)
tests/viscous_fs_simpletopbottom/Makefile (+1/-1)
tests/viscous_fs_simpletopbottom_explicit/Makefile (+1/-1)
tests/viscous_fs_simpletopbottom_explicit_varrho/Makefile (+1/-1)
tests/viscous_fs_simpletopbottom_varrho/Makefile (+1/-1)
tests/viscous_fs_top/Makefile (+1/-1)
tests/viscous_fs_top_explicit/Makefile (+1/-1)
Revision history for this message
Jon Hill (jon-hill) wrote :

There are 188 flml files that use the gmsh format and 23 Makefiles that already use binary gmsh, leaving some 150-ish tests to edit...

Revision history for this message
Jon Hill (jon-hill) wrote :

List of FLMLs that use GMSH format. These tests will need their Makefiles checking for the -bin flag.

Revision history for this message
Frank Milthaler (f-milthaler) wrote :

Alternatively, you could run the meshconverter on all those gmsh files, with input and output format being 'gmsh'. It will read in the mesh file and dump it as a binary gmsh file.
This approach also prevents minor differences in the mesh wich might occur when trying to re-generate the mesh files, due using a newer gmsh version now.
As the meshconverter will then overwrite the input mesh file with the output mesh file (since the mesh format for input and output are the same), it might be worth creating a backup of all relevant tests first.

Revision history for this message
Simon Mouradian (mouradian) wrote :

gmsh -convert also converts to binary format.

Revision history for this message
Jon Hill (jon-hill) wrote :

Thanks both - useful advice

Revision history for this message
Jon Hill (jon-hill) wrote :

I seemed to have let this one lapse...will try and get it done this week!

Jon Hill (jon-hill)
Changed in fluidity:
status: Confirmed → In Progress
Revision history for this message
Jon Hill (jon-hill) wrote :

I have a test, but CX1 is down it appears, so I can't test my test

Revision history for this message
Jon Hill (jon-hill) wrote :

OK, my test works on CX1 (i.e. the run fails gracefully) and on local machines (works on local machines). Now to transfer all ASCII GMSH files to binary.
test_gmsh_intel: Running
test_gmsh_intel: Calling 'make input':
rm -f ldd_* ./src/*.msh *.ele *.edge *.node *.face *.halo *.vtu *pvtu *.stat *.log-* *.err-* matrixdump*
test_gmsh_intel: ldd ../../bin/fluidity > ldd_output; fluidity -v1 -l boring.flml
test_gmsh_intel: Log: fluidity.err-0
test_gmsh_intel: There are issues with the Intel compiler and GMSH ASCII files
test_gmsh_intel: Please use binary GMSH or convert your mesh to triangle
test_gmsh_intel: *** ERROR ***
test_gmsh_intel: Error message: Exiting
test_gmsh_intel: [unset]: aborting job:
test_gmsh_intel: application called MPI_Abort(MPI_COMM_WORLD, 15) - process 0
test_gmsh_intel:
test_gmsh_intel: Assigning variables:
test_gmsh_intel: Assigning solvers_converged = True
test_gmsh_intel: Running failure tests:
test_gmsh_intel: Running Solvers converged:
test_gmsh_intel: success.
test_gmsh_intel: Running Correct error message:
test_gmsh_intel: success.
test_gmsh_intel: PP
Passes: 2
Failures: 0
Warnings: 0

Revision history for this message
Jon Hill (jon-hill) wrote :

Done, merge requested

Jon Hill (jon-hill)
Changed in fluidity:
status: In Progress → Fix Committed
Tim Greaves (tim-greaves)
Changed in fluidity:
status: Fix Committed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Duplicates of this bug

Subscribing...

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.