I noticed a problem when running the viscoelastic bending beam benchmark, that the number of cells was not what I expected it to be. The attached picture shows Paraview with surface with edges visualization, which shows 13 groups of 4 cells, where 16 should exist. I simplified the parameter file significantly to the following, and still am able to reproduce the bug using deal.II 9.3.3(installed using Candi today) and ASPECT main (also from today). Most strange is so far I have only seen the bug when using 64 or more MPI processes. When I ran with 1, 4, 16, 32, or 63 it did not reproduce the error.
A minimal parameter file that still shows the bug
# Global parameters
set Dimension = 2
set Start time = 0
set Maximum time step = 250 # 1e3
set Pressure normalization = surface
set Surface pressure = 0.
set Output directory = test
set End time = 0e3
# Model geometry (7.5x5 km, 0.1 km spacing)
subsection Geometry model
set Model name = box
subsection Box
set X repetitions = 75
set Y repetitions = 50
set X extent = 7.5e3
set Y extent = 5e3
end
end
# Mesh refinement specifications
subsection Mesh refinement
set Initial adaptive refinement = 0
set Initial global refinement = 1
set Time steps between mesh refinement = 0
end
# Material model
subsection Material model
set Model name = simple
end
# Gravity model with zero gravity
subsection Gravity model
set Model name = vertical
subsection Vertical
set Magnitude = 0.
end
end
The output when running this parameter file. Note that there should be 15,000 cells from the model being 75\cdot50 refined once to be 150\cdot 100 cells, not 14,997.
-----------------------------------------------------------------------------
-- This is ASPECT, the Advanced Solver for Problems in Earth's ConvecTion.
-- . version 2.4.0-pre (main, 229edb808)
-- . using deal.II 9.3.3
-- . with 32 bit indices and vectorization level 2 (256 bits)
-- . using Trilinos 12.18.1
-- . using p4est 2.3.2
-- . running in DEBUG mode
-- . running with 64 MPI processes
-----------------------------------------------------------------------------
-----------------------------------------------------------------------------
-- For information on how to cite ASPECT, see:
-- https://aspect.geodynamics.org/citing.html?ver=2.4.0-pre&sha=229edb808&src=code
-----------------------------------------------------------------------------
Number of active cells: 14,997 (on 2 levels)
Number of degrees of freedom: 196,741 (120,994+15,250+60,497)
*** Timestep 0: t=0 years, dt=0 years
Skipping temperature solve because RHS is zero.
Rebuilding Stokes preconditioner...
Solving Stokes system... 0+0 iterations.
Postprocessing:
Termination requested by criterion: end time
+---------------------------------------------+------------+------------+
| Total wallclock time elapsed since start | 8.42s | |
| | | |
| Section | no. calls | wall time | % of total |
+---------------------------------+-----------+------------+------------+
| Assemble Stokes system | 1 | 0.907s | 11% |
| Assemble temperature system | 1 | 0.42s | 5% |
| Build Stokes preconditioner | 1 | 0.856s | 10% |
| Initialization | 1 | 2.36s | 28% |
| Postprocessing | 1 | 0.0116s | 0.14% |
| Setup dof systems | 1 | 0.554s | 6.6% |
| Setup initial conditions | 1 | 1.65s | 20% |
| Setup matrices | 1 | 0.201s | 2.4% |
| Solve Stokes system | 1 | 0.0899s | 1.1% |
+---------------------------------+-----------+------------+------------+
-- Total wallclock time elapsed including restarts:8s
-----------------------------------------------------------------------------
-- For information on how to cite ASPECT, see:
-- https://aspect.geodynamics.org/citing.html?ver=2.4.0-pre&sha=229edb808&src=code
-----------------------------------------------------------------------------
Please let me know if this is a known issue of some kind or if you have any suggestions for fixing this behavior.
Thank you,
Mack
