Refinement (dolfinx::refinement)

namespace dolfinx::refinement

Mesh refinement algorithms.

Methods for refining meshes uniformly, or with markers, using edge bisection.

Functions

mesh::Mesh refine(const mesh::Mesh &mesh, bool redistribute = true)

Create a uniformly refined mesh.

Parameters

  • mesh – [in] The mesh from which to build a refined Mesh

  • redistribute – [in] Optional argument to redistribute the refined mesh if mesh is a distributed mesh.

Returns

A refined mesh

mesh::Mesh refine(const mesh::Mesh &mesh, const xtl::span<const std::int32_t> &edges, bool redistribute = true)

Create a locally refined mesh.

Parameters

  • mesh – [in] The mesh from which to build a refined Mesh

  • edges – [in] Indices of the edges that should be split by this refinement. mesh::compute_incident_entities can be used to compute the edges that are incident to other entities, e.g. incident to cells.

  • redistribute – [in] Optional argument to redistribute the refined mesh if mesh is a distributed mesh.

Returns

A locally refined mesh

std::pair<MPI_Comm, std::map<std::int32_t, std::vector<int>>> compute_edge_sharing(const mesh::Mesh &mesh)

Compute the sharing of edges between processes.

The resulting MPI_Comm is over the neighborhood of shared edges, allowing direct communication between peers. The resulting map is from local edge index to the set of neighbors (within the comm) that share that edge.

Parameters

mesh – [in] Mesh

Returns

pair of comm and map

void update_logical_edgefunction(MPI_Comm neighbor_comm, const std::vector<std::vector<std::int32_t>> &marked_for_update, std::vector<std::int8_t> &marked_edges, const common::IndexMap &map_e)

Transfer marked edges between processes.

Parameters

  • neighbor_comm – [in] MPI Communicator for neighborhood

  • marked_for_update – [in] Lists of edges to be updates on each neighbor

  • marked_edges – [inout] Marked edges to be updated

  • map_e – [in] IndexMap for edges

std::pair<std::map<std::int32_t, std::int64_t>, xt::xtensor<double, 2>> create_new_vertices(MPI_Comm neighbor_comm, const std::map<std::int32_t, std::vector<std::int32_t>> &shared_edges, const mesh::Mesh &mesh, const std::vector<std::int8_t> &marked_edges)

Add new vertex for each marked edge, and create new_vertex_coordinates and global_edge->new_vertex map. Communicate new vertices with MPI to all affected processes.

Parameters

  • neighbor_comm – [in] MPI Communicator for neighborhood

  • shared_edges – [in]

  • mesh – [in] Existing mesh

  • marked_edges – [in]

Returns

edge_to_new_vertex map and geometry array

mesh::Mesh partition(const mesh::Mesh &old_mesh, const graph::AdjacencyList<std::int64_t> &cell_topology, const xt::xtensor<double, 2> &new_vertex_coordinates, bool redistribute, mesh::GhostMode ghost_mode)

Use vertex and topology data to partition new mesh across processes.

Parameters

  • old_mesh – [in]

  • cell_topology – [in] Topology of cells, (vertex indices)

  • new_vertex_coordinates – [in]

  • redistribute – [in] Call graph partitioner if true

  • ghost_mode – [in] None or shared_facet

Returns

New mesh

std::vector<std::int64_t> adjust_indices(const common::IndexMap &index_map, std::int32_t n)

brief description indices to account for extra n values on each process.

This is a utility to help add new topological vertices on each process into the space of the index map.

Parameters

  • index_map – [in] Index map for the current mesh vertices

  • n – [in] Number of new entries to be accommodated on this process

Returns

Global indices as if “n” extra values are appended on each process

mesh::MeshTags<std::int32_t> transfer_facet_meshtag(const mesh::MeshTags<std::int32_t> &parent_meshtag, const mesh::Mesh &refined_mesh, const std::vector<std::int32_t> &parent_cell, const std::vector<std::int8_t> &parent_facet)

Transfer facet MeshTags from coarse mesh to refined mesh.

Note

The refined mesh must not have been redistributed during refinement

Note

GhostMode must be GhostMode.none

Parameters

  • parent_meshtag – [in] Facet MeshTags on parent mesh

  • refined_mesh – [in] Refined mesh based on parent mesh

  • parent_cell – [in] Parent cell of each cell in refined mesh

  • parent_facet – [in] Local facets of parent in each cell in refined mesh

Returns

MeshTags on refined mesh, values copied over from coarse mesh

mesh::MeshTags<std::int32_t> transfer_cell_meshtag(const mesh::MeshTags<std::int32_t> &parent_meshtag, const mesh::Mesh &refined_mesh, const std::vector<std::int32_t> &parent_cell)

Transfer cell MeshTags from coarse mesh to refined mesh.

Note

The refined mesh must not have been redistributed during refinement

Note

GhostMode must be GhostMode.none

Parameters

  • parent_meshtag – [in] Cell MeshTags on parent mesh

  • refined_mesh – [in] Refined mesh based on parent mesh

  • parent_cell – [in] Parent cell of each cell in refined mesh

Returns

MeshTags on refined mesh, values copied over from coarse mesh

namespace plaza

Function in this namespace implement the refinement method described in Plaza and Carey “Local refinement of simplicial grids based on

the skeleton” (Applied Numerical Mathematics 32 (2000) 195-218).

Enums

enum class RefinementOptions : int

Selection of options when refining a Mesh. parent_cell will output a list containing the local parent cell index for each new cell, parent_facet will output a list of the cell-local facet indices in the parent cell of each facet in each new cell (or -1 if no match). parent_cell_and_facet will output both datasets.

Values:

enumerator none
enumerator parent_cell
enumerator parent_facet
enumerator parent_cell_and_facet

Functions

std::tuple<mesh::Mesh, std::vector<std::int32_t>, std::vector<std::int8_t>> refine(const mesh::Mesh &mesh, bool redistribute, RefinementOptions options)

Uniform refine, optionally redistributing and optionally calculating the parent-child relationships, selected by RefinementOptions.

Parameters

  • mesh – [in] Input mesh to be refined

  • redistribute – [in] Flag to call the mesh partitioner to redistribute after refinement

  • options – [in] RefinementOptions enum to choose the computation of parent facets, parent cells. If an option is unselected, an empty list is returned.

Returns

New Mesh and optional parent cell index, parent facet indices

std::tuple<mesh::Mesh, std::vector<std::int32_t>, std::vector<std::int8_t>> refine(const mesh::Mesh &mesh, const xtl::span<const std::int32_t> &edges, bool redistribute, RefinementOptions options)

Refine with markers, optionally redistributing, and optionally calculating the parent-child relationships, selected by RefinementOptions.

Parameters

  • mesh – [in] Input mesh to be refined

  • edges – [in] Indices of the edges that should be split by this refinement

  • redistribute – [in] Flag to call the Mesh Partitioner to redistribute after refinement

  • options – [in] RefinementOptions enum to choose the computation of parent facets, parent cells. If an option is unselected, an empty list is returned.

Returns

New Mesh and optional parent cell index, parent facet indices

std::tuple<graph::AdjacencyList<std::int64_t>, xt::xtensor<double, 2>, std::vector<std::int32_t>, std::vector<std::int8_t>> compute_refinement_data(const mesh::Mesh &mesh, RefinementOptions options)

Refine mesh returning new mesh data.

Parameters

  • mesh – [in] Input mesh to be refined

  • options – [in] RefinementOptions enum to choose the computation of parent facets, parent cells. If an option is unselected, an empty list is returned.

Returns

New mesh data: cell topology, vertex coordinates, and optional parent cell index, and parent facet indices.

std::tuple<graph::AdjacencyList<std::int64_t>, xt::xtensor<double, 2>, std::vector<std::int32_t>, std::vector<std::int8_t>> compute_refinement_data(const mesh::Mesh &mesh, const xtl::span<const std::int32_t> &edges, RefinementOptions options)

Refine with markers returning new mesh data.

Parameters

  • mesh – [in] Input mesh to be refined

  • edges – [in] Indices of the edges that should be split by this refinement

  • options – [in] RefinementOptions enum to choose the computation of parent facets, parent cells. If an option is unselected, an empty list is returned.

Returns

New mesh data: cell topology, vertex coordinates and parent cell index, and stored parent facet indices (if requested).