Obscure error in hunt driver with solid coupling #44
Comments
|
Thanks for reporting @guillermogf , you forgot to mention that the error did not reproduce in 9ea6729 for the following configuration. However, this example is too large for debugging. using GridapMHD: hunt
using SparseMatricesCSR
using GridapPETSc
solver = Dict(
:solver => :petsc,
:matrix_type => SparseMatrixCSR{0,PetscScalar,PetscInt},
:vector_type => Vector{PetscScalar},
:petsc_options => "-snes_monitor -ksp_error_if_not_converged true \
-ksp_converged_reason -ksp_type preonly -pc_type lu \
-pc_factor_mat_solver_type mumps -mat_mumps_icntl_28 1 \
-mat_mumps_icntl_29 2 -mat_mumps_icntl_4 3 \
-mat_mumps_cntl_1 0.001",
:niter => 100,
:rtol => 1e-5,
)
hunt(;
title="hunt_solidcoupling_Ha1000_120x120",
backend=:mpi,
np=(4,4),
nc=(120,120),
B=(0.0,1000,0.0),
L=1.0,
σw1=0.0001,
σw2=1000.0,
tw=0.1,
nsums=100,
vtk=true,
debug=false,
BL_adapted=true,
solver=solver,
path="../data/"
) |
|
FYI @principejavier |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Hi,
I've detected an error when running the hunt driver with solid coupling (tw != 0).
The error cause has proved difficult to isolate as small variations in the model seem to suppress it or trigger it. Most notably, running the same problem setup with a different number of parallel processes can trigger it. The error has been observed at different mesh sizes, Hartmann numbers and tw values. It has been reproduced not only with the mpi backend and petsc solver but also with the julia solver and sequential backend (as reported by @pmartorell).
Here is a running script with which I can reproduce the issue:
Notably either changing np=(4,4) to np=(2,2), or tw=0.5 to tw=0.1 suppresses the error. Toggling the
BL_adaptedflag also seems to affect whether the error is triggered.Here follows the abriged output produced. Please find here the full output file. It's too long and a bit messy as it collects the error from several of the 16 mpi processes.
Abriged output
Environment used:
I will report back should I find simpler or further examples to reproduce the issue.
The text was updated successfully, but these errors were encountered: