Distributed Hunt test over Ha=2000 with mumps #30
Description
Launching the Hunt driver in parallel for Ha > 2000 (approx) gives me the following error:
errorHunt_Ha5000_N2n20.log
outputHunt_Ha5000_N2n20.log
In this case I was launching a Ha=5000 case with 40 processes. The problem can be solved with the same mesh in serial, using the julia solver in approximately 30 min.
Solution
The Hunt funciton was launched with the following settings:
mpiexec -n ${SLURM_NPROCS} julia --project=.. -e
'
using GridapMHD: hunt
hunt(
nc=(60,60),
np=(5,8),
backend=:mpi,
L=1.0,
B=(0.,2000.,0.),
nsums = 1000,
debug=false,
vtk=true,
title="hunt_2000_mumps",
mesh = false,
BL_adapted = true,
solver=:petsc,
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",
)'