Preconditioners and linear solvers

MOOSE allows users to utilise the full power of the PETSc preconditioners and linear solvers. The following choices have been found to be effective for various types of PorousFlow simulations.

  • pc_type = lu, pc_factor_mat_solver_package = mumps (the most robust choice)

  • pc_type = bjacobi, ksp_type = bcgs

  • pc_type = bjacobi, ksp_type = gmres

  • pc_type = asm, pc_asm_overlap = 2, sub_pc_type = lu, sub_pc_factor_shift_type = NONZERO, ksp_type = gmres

  • pc_type = asm, pc_asm_overlap = 2, sub_pc_type = lu, sub_pc_factor_shift_type = NONZERO, ksp_type = gmres along with the following options -ksp_diagonal_scale -ksp_diagonal_scale_fix -ksp_gmres_modifiedgramschmidt

These options can be set in the Preconditioning block in the input file

Example of using solver options

[Preconditioning]
  [./andy]
    type = SMP
    full = true
    petsc_options = '-snes_converged_reason'
    petsc_options_iname = '-ksp_type -pc_type -sub_pc_type -snes_max_it -sub_pc_factor_shift_type -pc_asm_overlap -snes_atol -snes_rtol '
    petsc_options_value = 'gmres asm lu 100 NONZERO 2 1E-14 1E-12'
  [../]
[]
(modules/porous_flow/test/tests/newton_cooling/nc04.i)

The "mumps" method is usually superior to the "asm + LU" method, but isn't always installed on all computer systems. Increasing the pc_asm_overlap improves the strength of the preconditioner, but uses a huge amount of memory. The NONZERO shifting helps reduce problems when the diagonal of the Jacobian contains zeroes.