GridPartitioner

Create a uniform grid that overlays the mesh to be partitioned. Assign all elements within each cell of the grid to the same processor.

Description

Partitions the mesh by creating a grid in x,y,z and assigning each element of the mesh to a separate "cell" within the grid. This is useful when you have simple cartesian meshes and you just want to specify partitioning fairly directly. Sometimes a human is the best partitioner!

This is an example of a 2x2 grid partitioning for use on 4 processors.

2x2 GridPartitioner Example

warning

The number of cells (nx*ny*nz) MUST be equal to the number of MPI processes you're attempting to use!

How it Works

The GridPartitioner works by creating a GeneratedMesh for the "grid". That's the reason why this object takes similar input file parameters to a GeneratedMesh. The GeneratedMesh created by GridPartitioner is guaranteed to contain the original domain within it.

To assign the processor IDs the centroid of each element of the mesh to be partitioned is searched for in the GeneratedMesh. The ID of the element of the GeneratedMesh that it lies within is then assigned as the processor_id.

Input Parameters

allow_renumberingTrueIf allow_renumbering=false, node and element numbers are kept fixed until deletion
Default:True
C++ Type:bool
Options:
Description:If allow_renumbering=false, node and element numbers are kept fixed until deletion
bias_x1The amount by which to grow (or shrink) the cells in the x-direction.
Default:1
C++ Type:double
Options:
Description:The amount by which to grow (or shrink) the cells in the x-direction.
bias_y1The amount by which to grow (or shrink) the cells in the y-direction.
Default:1
C++ Type:double
Options:
Description:The amount by which to grow (or shrink) the cells in the y-direction.
bias_z1The amount by which to grow (or shrink) the cells in the z-direction.
Default:1
C++ Type:double
Options:
Description:The amount by which to grow (or shrink) the cells in the z-direction.
gauss_lobatto_gridFalseGrade mesh into boundaries according to Gauss-Lobatto quadrature spacing.
Default:False
C++ Type:bool
Options:
Description:Grade mesh into boundaries according to Gauss-Lobatto quadrature spacing.
ghosting_patch_sizeThe number of nearest neighbors considered for ghosting purposes when 'iteration' patch update strategy is used. Default is 5 * patch_size.
C++ Type:unsigned int
Options:
Description:The number of nearest neighbors considered for ghosting purposes when 'iteration' patch update strategy is used. Default is 5 * patch_size.
max_leaf_size10The maximum number of points in each leaf of the KDTree used in the nearest neighbor search. As the leaf size becomes larger,KDTree construction becomes faster but the nearest neighbor searchbecomes slower.
Default:10
C++ Type:unsigned int
Options:
Description:The maximum number of points in each leaf of the KDTree used in the nearest neighbor search. As the leaf size becomes larger,KDTree construction becomes faster but the nearest neighbor searchbecomes slower.
nx1Number of elements in the X direction
Default:1
C++ Type:unsigned int
Options:
Description:Number of elements in the X direction
ny1Number of elements in the Y direction
Default:1
C++ Type:unsigned int
Options:
Description:Number of elements in the Y direction
nz1Number of elements in the Z direction
Default:1
C++ Type:unsigned int
Options:
Description:Number of elements in the Z direction
parallel_typeREPLICATEDDEFAULT: Use libMesh::ReplicatedMesh unless --distributed-mesh is specified on the command line REPLICATED: Always use libMesh::ReplicatedMesh DISTRIBUTED: Always use libMesh::DistributedMesh
Default:REPLICATED
C++ Type:MooseEnum
Options:DEFAULT REPLICATED DISTRIBUTED
Description:DEFAULT: Use libMesh::ReplicatedMesh unless --distributed-mesh is specified on the command line REPLICATED: Always use libMesh::ReplicatedMesh DISTRIBUTED: Always use libMesh::DistributedMesh

Optional Parameters

centroid_partitioner_directionSpecifies the sort direction if using the centroid partitioner. Available options: x, y, z, radial
C++ Type:MooseEnum
Options:x y z radial
Description:Specifies the sort direction if using the centroid partitioner. Available options: x, y, z, radial
partitionerlinearSpecifies a mesh partitioner to use when splitting the mesh for a parallel computation.
Default:linear
C++ Type:MooseEnum
Options:default metis parmetis linear centroid hilbert_sfc morton_sfc
Description:Specifies a mesh partitioner to use when splitting the mesh for a parallel computation.

Partitioning Parameters

construct_node_list_from_side_listTrueWhether or not to generate nodesets from the sidesets (usually a good idea).
Default:True
C++ Type:bool
Options:
Description:Whether or not to generate nodesets from the sidesets (usually a good idea).
control_tagsAdds user-defined labels for accessing object parameters via control logic.
C++ Type:std::vector
Options:
Description:Adds user-defined labels for accessing object parameters via control logic.
enableTrueSet the enabled status of the MooseObject.
Default:True
C++ Type:bool
Options:
Description:Set the enabled status of the MooseObject.
nemesisFalseIf nemesis=true and file=foo.e, actually reads foo.e.N.0, foo.e.N.1, ... foo.e.N.N-1, where N = # CPUs, with NemesisIO.
Default:False
C++ Type:bool
Options:
Description:If nemesis=true and file=foo.e, actually reads foo.e.N.0, foo.e.N.1, ... foo.e.N.N-1, where N = # CPUs, with NemesisIO.
patch_size40The number of nodes to consider in the NearestNode neighborhood.
Default:40
C++ Type:unsigned int
Options:
Description:The number of nodes to consider in the NearestNode neighborhood.
patch_update_strategyneverHow often to update the geometric search 'patch'. The default is to never update it (which is the most efficient but could be a problem with lots of relative motion). 'always' will update the patch for all slave nodes at the beginning of every timestep which might be time consuming. 'auto' will attempt to determine at the start of which timesteps the patch for all slave nodes needs to be updated automatically.'iteration' updates the patch at every nonlinear iteration for a subset of slave nodes for which penetration is not detected. If there can be substantial relative motion between the master and slave surfaces during the nonlinear iterations within a timestep, it is advisable to use 'iteration' option to ensure accurate contact detection.
Default:never
C++ Type:MooseEnum
Options:never always auto iteration
Description:How often to update the geometric search 'patch'. The default is to never update it (which is the most efficient but could be a problem with lots of relative motion). 'always' will update the patch for all slave nodes at the beginning of every timestep which might be time consuming. 'auto' will attempt to determine at the start of which timesteps the patch for all slave nodes needs to be updated automatically.'iteration' updates the patch at every nonlinear iteration for a subset of slave nodes for which penetration is not detected. If there can be substantial relative motion between the master and slave surfaces during the nonlinear iterations within a timestep, it is advisable to use 'iteration' option to ensure accurate contact detection.

Advanced Parameters

Input Files

test/tests/partitioners/grid_partitioner/grid_partitioner.i
test/tests/vectorpostprocessors/parallel_consistency/parallel_consistency.i
test/tests/auxkernels/ghosting_aux/no_ghosting.i
test/tests/auxkernels/ghosting_aux/ghosting_aux.i

test/tests/partitioners/grid_partitioner/grid_partitioner.i

[Mesh]
  type = GeneratedMesh
  dim = 2
  nx = 10
  ny = 10
  [Partitioner]
    type = GridPartitioner
    nx = 2
    ny = 2
    nz = 1
  []
[]

[Variables]
  [u]
  []
[]

[Kernels]
  [diff]
    type = Diffusion
    variable = u
  []
[]

[BCs]
  [left]
    type = DirichletBC
    variable = u
    boundary = 'left'
    value = 0
  []
  [right]
    type = DirichletBC
    variable = u
    boundary = 'right'
    value = 1
  []
[]

[Executioner]
  type = Steady
  solve_type = PJFNK
  petsc_options_iname = '-pc_type -pc_hypre_type'
  petsc_options_value = 'hypre boomeramg'
[]

[Outputs]
  exodus = true
[]

[AuxVariables]
  [pid]
    family = MONOMIAL
    order = CONSTANT
  []
[]

[AuxKernels]
  [pid_aux]
    type = ProcessorIDAux
    variable = pid
    execute_on = 'INITIAL'
  []
[]

test/tests/vectorpostprocessors/parallel_consistency/parallel_consistency.i

[Mesh]
  type = GeneratedMesh
  dim = 2
  nx = 10
  ny = 10

  # To make this deterministic
  [Partitioner]
    type = GridPartitioner
    nx = 2
    ny = 1
    nz = 1
  []
[]

[Variables]
  [u]
  []
[]

[Kernels]
  [diff]
    type = Diffusion
    variable = u
  []
[]

[BCs]
  [left]
    type = DirichletBC
    variable = u
    boundary = 'left'
    value = 0
  []
  [right]
    type = DirichletBC
    variable = u
    boundary = 'right'
    value = 1
  []
[]

[VectorPostprocessors]
  [constant]
    type = ConstantVectorPostprocessor
    value = '3 4'
    execute_on = 'TIMESTEP_END'
  []
[]

[Executioner]
  type = Steady
  solve_type = PJFNK
  petsc_options_iname = '-pc_type -pc_hypre_type'
  petsc_options_value = 'hypre boomeramg'
[]

[Outputs]
  execute_on = 'timestep_end'
  exodus = true
[]

[AuxVariables]
  [scattered]
    family = MONOMIAL
    order = CONSTANT
  []
[]

[AuxKernels]
  [viewit]
    type = VectorPostprocessorVisualizationAux
    vpp = 'constant'
    vector_name = value
    variable = scattered
    execute_on = 'TIMESTEP_END'
  []
[]

test/tests/auxkernels/ghosting_aux/no_ghosting.i

[Mesh]
  type = GeneratedMesh
  dim = 2
  nx = 10
  ny = 10
 [Partitioner]
    type = GridPartitioner
    nx = 2
    ny = 2
  []

  output_ghosting = true
[]

[Variables]
  [u]
  []
[]

[Kernels]
  [diff]
    type = Diffusion
    variable = u
  []
[]

[BCs]
  [left]
    type = DirichletBC
    variable = u
    boundary = 'left'
    value = 0
  []
  [right]
    type = DirichletBC
    variable = u
    boundary = 'right'
    value = 1
  []
[]

[Executioner]
  type = Steady
  solve_type = PJFNK
  petsc_options_iname = '-pc_type -pc_hypre_type'
  petsc_options_value = 'hypre boomeramg'
[]

[Outputs]
  exodus = true
[]

[AuxVariables]
  [pid]
    family = MONOMIAL
    order = CONSTANT
  []
[]

[AuxKernels]
  [pid]
    type = ProcessorIDAux
    variable = pid
  []
[]

[Problem]
  default_ghosting = false
[]

test/tests/auxkernels/ghosting_aux/ghosting_aux.i

[Mesh]
  type = GeneratedMesh
  dim = 2
  nx = 10
  ny = 10
  [Partitioner]
    type = GridPartitioner
    nx = 2
    ny = 2
  []

  output_ghosting = true
[]

[Variables]
  [u]
  []
[]

[Kernels]
  [diff]
    type = Diffusion
    variable = u
  []
[]

[AuxVariables]
  [pid]
    family = MONOMIAL
    order = CONSTANT
  []
[]

[AuxKernels]
  [pid]
    type = ProcessorIDAux
    variable = pid
  []
[]

[BCs]
  [left]
    type = DirichletBC
    variable = u
    boundary = 'left'
    value = 0
  []
  [right]
    type = DirichletBC
    variable = u
    boundary = 'right'
    value = 1
  []
[]

[Executioner]
  type = Steady
  solve_type = PJFNK
  petsc_options_iname = '-pc_type -pc_hypre_type'
  petsc_options_value = 'hypre boomeramg'
[]

[Outputs]
  exodus = true
[]

[Problem]
  default_ghosting = true
[]

References

Description
How it Works
Input Parameters
Input Files
References

Application Usage

Physics and Syntax

Application Development

Framework Development

MOOSEDocs

Infrastructure

Questions

Information and Tools