Store and solve a local sparse linear problem. More...

#include <Ifpack2_SparseContainer_decl.hpp>

Inheritance diagram for Ifpack2::SparseContainer< MatrixType, InverseType >:

Public Types
Public typedefs
typedef MatrixType	matrix_type
	The first template parameter of this class.
typedef InverseType	inverse_type
	The second template parameter of this class.
typedef MatrixType::scalar_type	scalar_type
	The type of entries in the input (global) matrix.
typedef MatrixType::local_ordinal_type	local_ordinal_type
	The type of local indices in the input (global) matrix.
typedef MatrixType::global_ordinal_type	global_ordinal_type
	The type of global indices in the input (global) matrix.
typedef MatrixType::node_type	node_type
	The Kokkos Node type of the input (global) matrix.
typedef Container< MatrixType > ::row_matrix_type	row_matrix_type
	The (base class) type of the input matrix.
Public Member Functions
Constructor and destructor
	SparseContainer (const Teuchos::RCP< const row_matrix_type > &matrix, const Teuchos::ArrayView< const local_ordinal_type > &localRows)
	Constructor.
virtual	~SparseContainer ()
	Destructor (declared virtual for memory safety of derived classes).
Get and set methods
virtual size_t	getNumRows () const
	The number of rows in the local matrix on the calling process.
virtual bool	isInitialized () const
	Whether the container has been successfully initialized.
virtual bool	isComputed () const
	Whether the container has been successfully computed.
virtual void	setParameters (const Teuchos::ParameterList &List)
	Set all necessary parameters.
Mathematical functions
virtual void	initialize ()
	Do all set-up operations that only require matrix structure.
virtual void	compute ()
	Extract the local diagonal block and prepare the solver.
virtual void	apply (const Tpetra::MultiVector< scalar_type, local_ordinal_type, global_ordinal_type, node_type > &X, Tpetra::MultiVector< scalar_type, local_ordinal_type, global_ordinal_type, node_type > &Y, Teuchos::ETransp mode=Teuchos::NO_TRANS, scalar_type alpha=Teuchos::ScalarTraits< scalar_type >::one(), scalar_type beta=Teuchos::ScalarTraits< scalar_type >::zero()) const
	Compute `Y := alpha * M^{-1} X + beta*Y`.
virtual void	weightedApply (const Tpetra::MultiVector< scalar_type, local_ordinal_type, global_ordinal_type, node_type > &X, Tpetra::MultiVector< scalar_type, local_ordinal_type, global_ordinal_type, node_type > &Y, const Tpetra::Vector< scalar_type, local_ordinal_type, global_ordinal_type, node_type > &D, Teuchos::ETransp mode=Teuchos::NO_TRANS, scalar_type alpha=Teuchos::ScalarTraits< scalar_type >::one(), scalar_type beta=Teuchos::ScalarTraits< scalar_type >::zero()) const
	Compute `Y := alpha * diag(D) * M^{-1} (diag(D) * X) + beta*Y`.
Miscellaneous methods
virtual std::ostream &	print (std::ostream &os) const
	Print information about this object to the given output stream.
Implementation of Teuchos::Describable
virtual std::string	description () const
	A one-line description of this object.
virtual void	describe (Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel verbLevel=Teuchos::Describable::verbLevel_default) const
	Print the object with some verbosity level to the given FancyOStream.

Detailed Description

template<typename MatrixType, typename InverseType>
class Ifpack2::SparseContainer< MatrixType, InverseType >

Store and solve a local sparse linear problem.

Please refer to the documentation of the Container interface. Currently, Containers are used by BlockRelaxation. Block relaxations need to be able to do two things:

Store the diagonal blocks
Solve linear systems with each diagonal block

These correspond to the two template parameters:

MatrixType, which stores a sparse matrix
InverseType, which solves linear systems with that matrix

This class stores each block as a sparse matrix. Thus, MatrixType must be a specialization of Tpetra::RowMatrix or of its subclass Tpetra::CrsMatrix. Using a sparse matrix for each block is a good idea when the blocks are large and sparse. For small and / or dense blocks, it would probably be better to use an implementation of Container that stores the blocks densely.

The InverseType template parameter represents the class to use for solving linear systems with a block. In SparseContainer, this template parameter must be a specialization of Preconditioner. Specifically, InverseType must implement the following methods:

A constructor that takes an RCP<const MatrixType>
setParameters(Teuchos::ParameterList&)
initialize()
compute()
apply (const MV& X, MV& Y, ...), where MV is the appropriate specialization of Tpetra::MultiVector

We also assume that InverseType has the following typedefs:

scalar_type
local_ordinal_type
global_ordinal_type
node_type

MatrixType and InverseType may store values of different types, and may have different template parameters (e.g., local or global ordinal types). You may mix and match so long as implicit conversions are available. The most obvious use case for this are:

MatrixType::global_ordinal_type=long long and InverseType::global_ordinal_type=short
MatrixType::scalar_type=float and InverseType::scalar_type=double

SparseContainer currently assumes the following about the column and row Maps of the input matrix:

On all processes, the column and row Maps begin with the same set of on-process entries, in the same order. That is, on-process row and column indices are the same.
On all processes, all off-process indices in the column Map of the input matrix occur after that initial set.

These assumptions may be violated if MatrixType is a Tpetra::CrsMatrix specialization and was constructed with a user-provided column Map. The assumptions are not mathematically necessary and could be relaxed at any time. Implementers who wish to do so will need to modify the extract() method, so that it translates explicitly between local row and column indices, instead of just assuming that they are the same.

Member Typedef Documentation

template<typename MatrixType, typename InverseType>

typedef MatrixType Ifpack2::SparseContainer< MatrixType, InverseType >::matrix_type

The first template parameter of this class.

This must be either a Tpetra::RowMatrix specialization or a Tpetra::CrsMatrix specialization. It may have entirely different template parameters (e.g., scalar_type) than InverseType.

template<typename MatrixType, typename InverseType>

typedef InverseType Ifpack2::SparseContainer< MatrixType, InverseType >::inverse_type

The second template parameter of this class.

This must be a specialization of Ifpack2::Preconditioner or one of its subclasses. It may have entirely different template parameters (e.g., scalar_type) than MatrixType.

template<typename MatrixType, typename InverseType>

typedef MatrixType::scalar_type Ifpack2::SparseContainer< MatrixType, InverseType >::scalar_type

The type of entries in the input (global) matrix.

Reimplemented from Ifpack2::Container< MatrixType >.

template<typename MatrixType, typename InverseType>

typedef MatrixType::local_ordinal_type Ifpack2::SparseContainer< MatrixType, InverseType >::local_ordinal_type

The type of local indices in the input (global) matrix.

Reimplemented from Ifpack2::Container< MatrixType >.

template<typename MatrixType, typename InverseType>

typedef MatrixType::global_ordinal_type Ifpack2::SparseContainer< MatrixType, InverseType >::global_ordinal_type

The type of global indices in the input (global) matrix.

Reimplemented from Ifpack2::Container< MatrixType >.

template<typename MatrixType, typename InverseType>

typedef MatrixType::node_type Ifpack2::SparseContainer< MatrixType, InverseType >::node_type

The Kokkos Node type of the input (global) matrix.

Reimplemented from Ifpack2::Container< MatrixType >.

template<typename MatrixType, typename InverseType>

typedef Container<MatrixType>::row_matrix_type Ifpack2::SparseContainer< MatrixType, InverseType >::row_matrix_type

The (base class) type of the input matrix.

The input matrix to the constructor may be either a Tpetra::RowMatrix specialization or a Tpetra::CrsMatrix specialization. However, we want to make the constructor as general as possible, so we always accept the matrix as a Tpetra::RowMatrix. This typedef is the appropriate specialization of Tpetra::RowMatrix.

Reimplemented from Ifpack2::Container< MatrixType >.