FDP

Finite difference preconditioner (FDP) builds a numerical Jacobian for preconditioning, only use for testing and verification.

Overview

The Finite Difference Preconditioner (FDP) forms a "Numerical Jacobian" by doing direct finite differences of residual statements. This is extremely slow and inefficient, but is a great debugging tool because it allows you to form a nearly perfect preconditioner. FDP contains the same options for specifying off-diagonal blocks as SMP. Since FDP builds the perfect approximate Jacobian it can be useful to use it directly to solve instead of using JFNK. The finite differencing is sensitive to the differencing parameter which can be specified using:


petsc_options_iname = '-mat_fd_coloring_err -mat_fd_type'
petsc_options_value = '1e-6                 ds'

Example Input File Syntax

[Preconditioning]
  active = 'FDP_jfnk'

  [./FDP_jfnk]
    type = FDP

    off_diag_row = 'forced'
    off_diag_column = 'diffused'

    #Preconditioned JFNK (default)
    solve_type = 'PJFNK'

    petsc_options_iname = '-pc_type -mat_fd_coloring_err -mat_fd_type'
    petsc_options_value = 'lu       1e-6                 ds'
  [../]

  [./FDP_n]
    type = FDP

    off_diag_row = 'forced'
    off_diag_column = 'diffused'

    solve_type = 'NEWTON'

    petsc_options_iname = '-pc_type -mat_fd_coloring_err -mat_fd_type'
    petsc_options_value = 'lu       1e-6                 ds'
  [../]

  [./FDP_n_full]
    type = FDP

    full = true

    solve_type = 'NEWTON'

    petsc_options_iname = '-pc_type -mat_fd_coloring_err -mat_fd_type'
    petsc_options_value = 'lu       1e-6                 ds'
  [../]
[]

Input Parameters

finite_difference_typecoloringstandard: standard finite differencecoloring: finite difference based on coloring
Default:coloring
C++ Type:MooseEnum
Unit:(no unit assumed)
Options:standard, coloring
Controllable:No
Description:standard: standard finite differencecoloring: finite difference based on coloring
fullFalseSet to true if you want the full set of couplings between variables simply for convenience so you don't have to set every off_diag_row and off_diag_column combination.
Default:False
C++ Type:bool
Unit:(no unit assumed)
Controllable:No
Description:Set to true if you want the full set of couplings between variables simply for convenience so you don't have to set every off_diag_row and off_diag_column combination.
implicit_geometric_couplingFalseSet to true if you want to add entries into the matrix for degrees of freedom that might be coupled by inspection of the geometric search objects.
Default:False
C++ Type:bool
Unit:(no unit assumed)
Controllable:No
Description:Set to true if you want to add entries into the matrix for degrees of freedom that might be coupled by inspection of the geometric search objects.
ksp_normunpreconditionedSets the norm that is used for convergence testing
Default:unpreconditioned
C++ Type:MooseEnum
Unit:(no unit assumed)
Options:none, preconditioned, unpreconditioned, natural, default
Controllable:No
Description:Sets the norm that is used for convergence testing
nl_sysThe nonlinear system whose linearization this preconditioner should be applied to.
C++ Type:NonlinearSystemName
Unit:(no unit assumed)
Controllable:No
Description:The nonlinear system whose linearization this preconditioner should be applied to.
off_diag_columnThe variable names for the off-diagonal columns you want to add into the matrix; they will be associated with an off-diagonal row from the same position in off_diag_row.
C++ Type:std::vector<NonlinearVariableName>
Unit:(no unit assumed)
Controllable:No
Description:The variable names for the off-diagonal columns you want to add into the matrix; they will be associated with an off-diagonal row from the same position in off_diag_row.
off_diag_rowThe variable names for the off-diagonal rows you want to add into the matrix; they will be associated with an off-diagonal column from the same position in off_diag_column.
C++ Type:std::vector<NonlinearVariableName>
Unit:(no unit assumed)
Controllable:No
Description:The variable names for the off-diagonal rows you want to add into the matrix; they will be associated with an off-diagonal column from the same position in off_diag_column.
pc_sidedefaultPreconditioning side
Default:default
C++ Type:MooseEnum
Unit:(no unit assumed)
Options:left, right, symmetric, default
Controllable:No
Description:Preconditioning side

Optional Parameters

control_tagsAdds user-defined labels for accessing object parameters via control logic.
C++ Type:std::vector<std::string>
Unit:(no unit assumed)
Controllable:No
Description:Adds user-defined labels for accessing object parameters via control logic.
enableTrueSet the enabled status of the MooseObject.
Default:True
C++ Type:bool
Unit:(no unit assumed)
Controllable:No
Description:Set the enabled status of the MooseObject.

Advanced Parameters

mffd_typewpSpecifies the finite differencing type for Jacobian-free solve types. Note that the default is wp (for Walker and Pernice).
Default:wp
C++ Type:MooseEnum
Unit:(no unit assumed)
Options:wp, ds
Controllable:No
Description:Specifies the finite differencing type for Jacobian-free solve types. Note that the default is wp (for Walker and Pernice).
petsc_optionsSingleton PETSc options
C++ Type:MultiMooseEnum
Unit:(no unit assumed)
Options:-dm_moose_print_embedding, -dm_view, -ksp_converged_reason, -ksp_gmres_modifiedgramschmidt, -ksp_monitor, -ksp_monitor_snes_lg-snes_ksp_ew, -ksp_snes_ew, -snes_converged_reason, -snes_ksp, -snes_ksp_ew, -snes_linesearch_monitor, -snes_mf, -snes_mf_operator, -snes_monitor, -snes_test_display, -snes_view
Controllable:No
Description:Singleton PETSc options
petsc_options_inameNames of PETSc name/value pairs
C++ Type:MultiMooseEnum
Unit:(no unit assumed)
Options:-ksp_atol, -ksp_gmres_restart, -ksp_max_it, -ksp_pc_side, -ksp_rtol, -ksp_type, -mat_fd_coloring_err, -mat_fd_type, -mat_mffd_type, -pc_asm_overlap, -pc_factor_levels, -pc_factor_mat_ordering_type, -pc_hypre_boomeramg_grid_sweeps_all, -pc_hypre_boomeramg_max_iter, -pc_hypre_boomeramg_strong_threshold, -pc_hypre_type, -pc_type, -snes_atol, -snes_linesearch_type, -snes_ls, -snes_max_it, -snes_rtol, -snes_divergence_tolerance, -snes_type, -sub_ksp_type, -sub_pc_type
Controllable:No
Description:Names of PETSc name/value pairs
petsc_options_valueValues of PETSc name/value pairs (must correspond with "petsc_options_iname"
C++ Type:std::vector<std::string>
Unit:(no unit assumed)
Controllable:No
Description:Values of PETSc name/value pairs (must correspond with "petsc_options_iname"
solve_typePJFNK: Preconditioned Jacobian-Free Newton Krylov JFNK: Jacobian-Free Newton Krylov NEWTON: Full Newton Solve FD: Use finite differences to compute Jacobian LINEAR: Solving a linear problem
C++ Type:MooseEnum
Unit:(no unit assumed)
Options:PJFNK, JFNK, NEWTON, FD, LINEAR
Controllable:No
Description:PJFNK: Preconditioned Jacobian-Free Newton Krylov JFNK: Jacobian-Free Newton Krylov NEWTON: Full Newton Solve FD: Use finite differences to compute Jacobian LINEAR: Solving a linear problem

Petsc Parameters

(moose/examples/ex11_prec/fdp.i)

[Mesh]
  file = square.e
[]

[Variables]
  [./diffused]
    order = FIRST
    family = LAGRANGE
  [../]

  [./forced]
    order = FIRST
    family = LAGRANGE
  [../]
[]

# The Preconditioning block
[Preconditioning]
  active = 'FDP_jfnk'

  [./FDP_jfnk]
    type = FDP

    off_diag_row    = 'forced'
    off_diag_column = 'diffused'


  #Preconditioned JFNK (default)
  solve_type = 'PJFNK'


    petsc_options_iname = '-pc_type -mat_fd_coloring_err -mat_fd_type'
    petsc_options_value = 'lu       1e-6                 ds'
  [../]

  [./FDP_n]
    type = FDP

    off_diag_row    = 'forced'
    off_diag_column = 'diffused'


    solve_type = 'NEWTON'

    petsc_options_iname = '-pc_type -mat_fd_coloring_err -mat_fd_type'
    petsc_options_value = 'lu       1e-6                 ds'
  [../]

  [./FDP_n_full]
    type = FDP

    full = true


    solve_type = 'NEWTON'

    petsc_options_iname = '-pc_type -mat_fd_coloring_err -mat_fd_type'
    petsc_options_value = 'lu       1e-6                 ds'
  [../]
[]

[Kernels]
  [./diff_diffused]
    type = Diffusion
    variable = diffused
  [../]

  [./conv_forced]
    type = CoupledForce
    variable = forced
    v = diffused
  [../]

  [./diff_forced]
    type = Diffusion
    variable = forced
  [../]
[]

[BCs]
  #Note we have active on, and neglect the right_forced BC
  active = 'left_diffused right_diffused left_forced'

  [./left_diffused]
    type = DirichletBC
    variable = diffused
    boundary = 'left'
    value = 0
  [../]

  [./right_diffused]
    type = DirichletBC
    variable = diffused
    boundary = 'right'
    value = 100
  [../]

  [./left_forced]
    type = DirichletBC
    variable = forced
    boundary = 'left'
    value = 0
  [../]

  [./right_forced]
    type = DirichletBC
    variable = forced
    boundary = 'right'
    value = 0
  [../]
[]

[Executioner]
  type = Steady



[]

[Outputs]
  exodus = true
[]

Overview
Example Input File Syntax
Input Parameters