parallel::distributed::SolutionTransfer< dim, VectorType, spacedim > Class Template Reference

#include <deal.II/distributed/solution_transfer.h>

Public Member Functions
	SolutionTransfer (const DoFHandler< dim, spacedim > &dof_handler, const bool average_values=false)
	~SolutionTransfer ()=default
void	prepare_for_coarsening_and_refinement (const std::vector< const VectorType * > &all_in)
void	prepare_for_coarsening_and_refinement (const VectorType &in)
void	interpolate (std::vector< VectorType * > &all_out)
void	interpolate (VectorType &out)
void	prepare_for_serialization (const VectorType &in)
void	prepare_for_serialization (const std::vector< const VectorType * > &all_in)
void	deserialize (VectorType &in)
void	deserialize (std::vector< VectorType * > &all_in)

Private Member Functions
std::vector< char >	pack_callback (const typename Triangulation< dim, spacedim >::cell_iterator &cell, const CellStatus status)
void	unpack_callback (const typename Triangulation< dim, spacedim >::cell_iterator &cell, const CellStatus status, const boost::iterator_range< std::vector< char >::const_iterator > &data_range, std::vector< VectorType * > &all_out, VectorType &valence)
void	register_data_attach ()

Private Attributes
SmartPointer< const DoFHandler< dim, spacedim >, SolutionTransfer< dim, VectorType, spacedim > >	dof_handler
const bool	average_values
std::vector< const VectorType * >	input_vectors
unsigned int	handle

Detailed Description

template<int dim, typename VectorType, int spacedim = dim>
class parallel::distributed::SolutionTransfer< dim, VectorType, spacedim >

Transfer a discrete FE function (like a solution vector) by interpolation while refining and/or coarsening a distributed grid and handles the necessary communication.

Note

It is important to note, that if you use more than one SolutionTransfer object at the same time, that the calls to prepare_*() and interpolate()/deserialize() need to be in the same order.

Note on ghost elements

In a parallel computation PETSc or Trilinos vector may contain ghost elements or not. For reading in information with prepare_for_coarsening_and_refinement() or prepare_for_serialization() you need to supply vectors with ghost elements, so that all locally_active elements can be read. On the other hand, ghosted vectors are generally not writable, so for calls to interpolate() or deserialize() you need to supply distributed vectors without ghost elements. More precisely, during interpolation the current algorithm writes into all locally active degrees of freedom.

Transferring a solution

Here VectorType is your favorite vector type, e.g. PETScWrappers::MPI::Vector, TrilinosWrappers::MPI::Vector, or corresponding block vectors.

parallel::distributed::SolutionTransfer<dim, VectorType>
  soltrans(dof_handler);
// flag some cells for refinement and coarsening, e.g.
parallel::distributed::GridRefinement::
  refine_and_coarsen_fixed_fraction(tria,
                                    error_indicators,
                                    0.3,
                                    0.05);
 
// prepare the triangulation,
tria.prepare_coarsening_and_refinement();
 
// prepare the SolutionTransfer object for coarsening and refinement
// and give the solution vector that we intend to interpolate later,
soltrans.prepare_for_coarsening_and_refinement(solution);
 
// actually execute the refinement,
tria.execute_coarsening_and_refinement();
 
// redistribute dofs,
dof_handler.distribute_dofs(fe);
 
// and interpolate the solution
VectorType interpolated_solution;
 
//create VectorType in the right size here
soltrans.interpolate(interpolated_solution);

As the grid is distributed, it is important to note that the old solution(s) must be copied to one that also provides access to the locally relevant DoF values (these values required for the interpolation process):

// Create initial indexsets pertaining to the grid before refinement
const IndexSet &locally_owned_dofs    = dof_handler.locally_owned_dofs();
const IndexSet  locally_relevant_dofs =
  DoFTools::extract_locally_relevant_dofs(dof_handler);
 
// The solution vector only knows about locally owned DoFs
TrilinosWrappers::MPI::Vector solution;
solution.reinit(locally_owned_dofs,
                mpi_communicator);
...
// Transfer solution to vector that provides access to
// locally relevant DoFs
TrilinosWrappers::MPI::Vector old_solution;
old_solution.reinit(locally_owned_dofs,
                    locally_relevant_dofs,
                    mpi_communicator);
old_solution = solution;
 
// Initialize SolutionTransfer object
parallel::distributed::SolutionTransfer<dim, VectorType>
  soltrans(dof_handler);
soltrans.prepare_for_coarsening_and_refinement(old_solution);
...
// Refine grid
// Recreate locally_owned_dofs and locally_relevant_dofs index sets
...
solution.reinit(locally_owned_dofs, mpi_communicator);
soltrans.interpolate(solution);

Different from PETSc and Trilinos vectors, LinearAlgebra::distributed::Vector allows writing into ghost elements. For a ghosted vector the interpolation step can be accomplished via

interpolated_solution.zero_out_ghost_values();
soltrans.interpolate(interpolated_solution);
interpolated_solution.update_ghost_values();

Use for Serialization

This class can be used to serialize and later deserialize a distributed mesh with solution vectors to a file. If you use more than one DoFHandler and therefore more than one SolutionTransfer object, they need to be serialized and deserialized in the same order.

If vector has the locally relevant DoFs, serialization works as follows:

parallel::distributed::SolutionTransfer<dim, VectorType>
  sol_trans(dof_handler);
sol_trans.prepare_for_serialization(vector);
 
triangulation.save(filename);

For deserialization the vector needs to be a distributed vector (without ghost elements):

//[create coarse mesh...]
triangulation.load(filename);
 
parallel::distributed::SolutionTransfer<dim, VectorType>
  sol_trans(dof_handler);
sol_trans.deserialize(distributed_vector);

Note on usage with DoFHandler with hp-capabilities

Since data on DoFHandler objects with hp-capabilities is associated with many different FiniteElement objects, each cell's data has to be processed with its corresponding future_fe_index. Further, if refinement is involved, data will be packed on the parent cell with its future_fe_index and unpacked later with the same index on its children. If cells get coarsened into one, data will be packed on the children with the least dominant finite element of their common subspace, and unpacked on the parent with this particular finite element (consult hp::FECollection::find_dominated_fe_extended() for more information).

Transferring a solution across refinement works exactly like in the non-hp-case. However, when considering serialization, we also have to store the active FE indices in an additional step. A code snippet demonstrating serialization with the parallel::distributed::SolutionTransfer class with DoFHandler objects with hp-capabilities is provided in the following. Here VectorType is your favorite vector type, e.g. PETScWrappers::MPI::Vector, TrilinosWrappers::MPI::Vector, or corresponding block vectors.

If vector has the locally relevant DoFs, serialization works as follows:

parallel::distributed::
  SolutionTransfer<dim, VectorType, DoFHandler<dim,spacedim>>
    sol_trans(hp_dof_handler);
 
hp_dof_handler.prepare_for_serialization_of_active_fe_indices();
sol_trans.prepare_for_serialization(vector);
 
triangulation.save(filename);

For deserialization the vector needs to be a distributed vector (without ghost elements):

//[create coarse mesh...]
triangulation.load(filename);
 
hp::FECollection<dim,spacedim> fe_collection;
//[prepare identical fe_collection...]
 
DoFHandler<dim,spacedim> hp_dof_handler(triangulation);
// We need to introduce our dof_handler to the fe_collection
// before setting all active FE indices.
hp_dof_handler.deserialize_active_fe_indices();
hp_dof_handler.distribute_dofs(fe_collection);
 
parallel::distributed::
  SolutionTransfer<dim,VectorType,DoFHandler<dim,spacedim>>
    sol_trans(hp_dof_handler);
sol_trans.deserialize(distributed_vector);

Interaction with hanging nodes

In essence, this class implements the same steps as does SolutionTransfer (though the implementation is entirely separate). Consequently, the same issue with hanging nodes and coarsening can happen with this class as happens with SolutionTransfer. See there for an extended discussion.

Definition at line 223 of file solution_transfer.h.

Constructor & Destructor Documentation

SolutionTransfer()

template<int dim, typename VectorType , int spacedim>

SolutionTransfer< dim, VectorType, spacedim >::SolutionTransfer	(	const DoFHandler< dim, spacedim > &	dof_handler,
		const bool	average_values = false )

Constructor.

Parameters

[in]	dof_handler	The DoFHandler on which all operations will happen. At the time when this constructor is called, the DoFHandler still points to the Triangulation before the refinement in question happens.
[in]	average_values	Average the contribututions to the same DoF coming from different cells. Note: averaging requires an additional communication step, since the valence of the DoF has to be determined.

Definition at line 115 of file solution_transfer.cc.

~SolutionTransfer()

template<int dim, typename VectorType , int spacedim = dim>

parallel::distributed::SolutionTransfer< dim, VectorType, spacedim >::~SolutionTransfer ( )

default

Destructor.

Member Function Documentation

prepare_for_coarsening_and_refinement() [1/2]

template<int dim, typename VectorType , int spacedim>

void SolutionTransfer< dim, VectorType, spacedim >::prepare_for_coarsening_and_refinement

(

const std::vector< const VectorType * > &

all_in

)

Prepare the current object for coarsening and refinement. It stores the dof indices of each cell and stores the dof values of the vectors in all_in in each cell that'll be coarsened. all_in includes all vectors that are to be interpolated onto the new (refined and/or coarsened) grid.

Definition at line 134 of file solution_transfer.cc.

prepare_for_coarsening_and_refinement() [2/2]

template<int dim, typename VectorType , int spacedim>

void SolutionTransfer< dim, VectorType, spacedim >::prepare_for_coarsening_and_refinement

(

const VectorType &

in

)

Same as the previous function but for only one discrete function to be interpolated.

Definition at line 176 of file solution_transfer.cc.

interpolate() [1/2]

template<int dim, typename VectorType , int spacedim>

void SolutionTransfer< dim, VectorType, spacedim >::interpolate

(

std::vector< VectorType * > &

all_out

)

Interpolate the data previously stored in this object before the mesh was refined or coarsened onto the current set of cells. Do so for each of the vectors provided to prepare_for_coarsening_and_refinement() and write the result into the given set of vectors.

Definition at line 232 of file solution_transfer.cc.

interpolate() [2/2]

template<int dim, typename VectorType , int spacedim>

void SolutionTransfer< dim, VectorType, spacedim >::interpolate

(

VectorType &

out

)

Same as the previous function. It interpolates only one function. It assumes the vectors having the right sizes (i.e. in.size()==n_dofs_old, out.size()==n_dofs_refined)

Multiple calling of this function is NOT allowed. Interpolating several functions can be performed in one step by using prepare_for_coarsening_and_refinement(all_in) and interpolate(all_out).

Definition at line 300 of file solution_transfer.cc.

prepare_for_serialization() [1/2]

template<int dim, typename VectorType , int spacedim>

void SolutionTransfer< dim, VectorType, spacedim >::prepare_for_serialization

(

const VectorType &

in

)

Prepare the serialization of the given vector. The serialization is done by Triangulation::save(). The given vector needs all information on the locally active DoFs (it must be ghosted). See documentation of this class for more information.

Definition at line 187 of file solution_transfer.cc.

prepare_for_serialization() [2/2]

template<int dim, typename VectorType , int spacedim>

void SolutionTransfer< dim, VectorType, spacedim >::prepare_for_serialization

(

const std::vector< const VectorType * > &

all_in

)

Same as the function above, only for a list of vectors.

Definition at line 198 of file solution_transfer.cc.

deserialize() [1/2]

template<int dim, typename VectorType , int spacedim>

void SolutionTransfer< dim, VectorType, spacedim >::deserialize

(

VectorType &

in

)

Execute the deserialization of the given vector. This needs to be done after calling Triangulation::load(). The given vector must be a fully distributed vector without ghost elements. See documentation of this class for more information.

Definition at line 208 of file solution_transfer.cc.

deserialize() [2/2]

template<int dim, typename VectorType , int spacedim>

void SolutionTransfer< dim, VectorType, spacedim >::deserialize

(

std::vector< VectorType * > &

all_in

)

Same as the function above, only for a list of vectors.

Definition at line 218 of file solution_transfer.cc.

pack_callback()

template<int dim, typename VectorType , int spacedim>

std::vector< char > SolutionTransfer< dim, VectorType, spacedim >::pack_callback ( const typename Triangulation< dim, spacedim >::cell_iterator & cell, const CellStatus status )

private

A callback function used to pack the data on the current mesh into objects that can later be retrieved after refinement, coarsening and repartitioning.

Definition at line 310 of file solution_transfer.cc.

unpack_callback()

template<int dim, typename VectorType , int spacedim>

void SolutionTransfer< dim, VectorType, spacedim >::unpack_callback ( const typename Triangulation< dim, spacedim >::cell_iterator & cell, const CellStatus status, const boost::iterator_range< std::vector< char >::const_iterator > & data_range, std::vector< VectorType * > & all_out, VectorType & valence )

private

A callback function used to unpack the data on the current mesh that has been packed up previously on the mesh before refinement, coarsening and repartitioning.

Definition at line 378 of file solution_transfer.cc.

register_data_attach()

template<int dim, typename VectorType , int spacedim>

void SolutionTransfer< dim, VectorType, spacedim >::register_data_attach ( )

private

Registers the pack_callback() function to the parallel::distributed::Triangulation that has been assigned to the DoFHandler class member and stores the returning handle.

Definition at line 150 of file solution_transfer.cc.

Member Data Documentation

dof_handler

template<int dim, typename VectorType , int spacedim = dim>

SmartPointer<const DoFHandler<dim, spacedim>, SolutionTransfer<dim, VectorType, spacedim> > parallel::distributed::SolutionTransfer< dim, VectorType, spacedim >::dof_handler

private

Pointer to the degree of freedom handler to work with.

Definition at line 324 of file solution_transfer.h.

average_values

template<int dim, typename VectorType , int spacedim = dim>

const bool parallel::distributed::SolutionTransfer< dim, VectorType, spacedim >::average_values

private

Flag indicating if averaging should be performed.

Definition at line 329 of file solution_transfer.h.

input_vectors

template<int dim, typename VectorType , int spacedim = dim>

std::vector<const VectorType *> parallel::distributed::SolutionTransfer< dim, VectorType, spacedim >::input_vectors

private

A vector that stores pointers to all the vectors we are supposed to copy over from the old to the new mesh.

Definition at line 335 of file solution_transfer.h.

handle

template<int dim, typename VectorType , int spacedim = dim>

unsigned int parallel::distributed::SolutionTransfer< dim, VectorType, spacedim >::handle

private

The handle that the Triangulation has assigned to this object with which we can access our memory offset and our pack function.

Definition at line 341 of file solution_transfer.h.

---

The documentation for this class was generated from the following files:

• include/deal.II/distributed/solution_transfer.h
• source/distributed/solution_transfer.cc