Reference documentation for deal.II version 8.4.1
Public Types | Public Member Functions | Public Attributes | Private Attributes | List of all members

#include <deal.II/lac/matrix_block.h>

Inheritance diagram for MatrixBlock< MatrixType >:
[legend]

Public Types

typedef types::global_dof_index size_type
 
typedef MatrixType::value_type value_type
 

Public Member Functions

 MatrixBlock ()
 
 MatrixBlock (const MatrixBlock< MatrixType > &M)
 
 MatrixBlock (size_type i, size_type j)
 
void reinit (const BlockSparsityPattern &sparsity)
 
void add (const size_type i, const size_type j, const typename MatrixType::value_type value)
 
template<typename number >
void add (const std::vector< size_type > &indices, const FullMatrix< number > &full_matrix, const bool elide_zero_values=true)
 
template<typename number >
void add (const std::vector< size_type > &row_indices, const std::vector< size_type > &col_indices, const FullMatrix< number > &full_matrix, const bool elide_zero_values=true)
 
template<typename number >
void add (const size_type row_index, const std::vector< size_type > &col_indices, const std::vector< number > &values, const bool elide_zero_values=true)
 
template<typename number >
void add (const size_type row, const size_type n_cols, const size_type *col_indices, const number *values, const bool elide_zero_values=true, const bool col_indices_are_sorted=false)
 
template<class VectorType >
void vmult (VectorType &w, const VectorType &v) const
 
template<class VectorType >
void vmult_add (VectorType &w, const VectorType &v) const
 
template<class VectorType >
void Tvmult (VectorType &w, const VectorType &v) const
 
template<class VectorType >
void Tvmult_add (VectorType &w, const VectorType &v) const
 
std::size_t memory_consumption () const
 
 DeclException2 (ExcBlockIndexMismatch, size_type, size_type,<< "Block index "<< arg1<< " does not match "<< arg2)
 
- Public Member Functions inherited from Subscriptor
 Subscriptor ()
 
 Subscriptor (const Subscriptor &)
 
virtual ~Subscriptor ()
 
Subscriptoroperator= (const Subscriptor &)
 
void subscribe (const char *identifier=0) const
 
void unsubscribe (const char *identifier=0) const
 
unsigned int n_subscriptions () const
 
void list_subscribers () const
 
 DeclException3 (ExcInUse, int, char *, std::string &,<< "Object of class "<< arg2<< " is still used by "<< arg1<< " other objects."<< "\n\n"<< "(Additional information: "<< arg3<< ")\n\n"<< "See the entry in the Frequently Asked Questions of "<< "deal.II (linked to from http://www.dealii.org/) for "<< "a lot more information on what this error means and "<< "how to fix programs in which it happens.")
 
 DeclException2 (ExcNoSubscriber, char *, char *,<< "No subscriber with identifier <"<< arg2<< "> subscribes to this object of class "<< arg1<< ". Consequently, it cannot be unsubscribed.")
 
template<class Archive >
void serialize (Archive &ar, const unsigned int version)
 

Public Attributes

size_type row
 
size_type column
 
MatrixType matrix
 

Private Attributes

BlockIndices row_indices
 
BlockIndices column_indices
 

Detailed Description

template<typename MatrixType>
class MatrixBlock< MatrixType >

A wrapper around a matrix object, storing the coordinates in a block matrix as well.

This class is an alternative to BlockMatrixBase, if you only want to generate a single block of the system, not the whole system. Using the add() functions of this class, it is possible to use the standard assembling functions used for block matrices, but only enter in one of the blocks and still avoiding the index computations involved. The reason for this class is, that we may need a different number of matrices for different blocks in a block system. For example, a preconditioner for the Oseen system can be built as a block system, where the pressure block is of the form M-1FA-1 with M the pressure mass matrix, A the pressure Laplacian and F the advection diffusion operator applied to the pressure space. Since only a single matrix is needed for the other blocks, using BlockSparseMatrix or similar would be a waste of memory.

While the add() functions make a MatrixBlock appear like a block matrix for assembling, the functions vmult(), Tvmult(), vmult_add(), and Tvmult_add() make it behave like a MatrixType, when it comes to applying it to a vector. This behavior allows us to store MatrixBlock objects in vectors, for instance in MGLevelObject without extracting the matrix first.

MatrixBlock comes handy when using BlockMatrixArray. Once the MatrixBlock has been properly initialized and filled, it can be used in the simplest case as:

...
BlockMatrixArray matrix (n_blocks, n_blocks);
for (size_type i=0;i<blocks.size;++i)
matrix.enter(blocks.block(i).row, blocks.block(i).column, blocks.matrix(i));

Here, we have not gained very much, except that we do not need to set up empty blocks in the block system.

Note
This class expects, that the row and column BlockIndices objects for the system are equal. If they are not, some functions will throw ExcNotImplemented.
Todo:
Example for the product preconditioner of the pressure Schur complement.
See also
Block (linear algebra)
Author
Guido Kanschat, 2006

Definition at line 33 of file matrix_block.h.

Member Typedef Documentation

template<typename MatrixType>
typedef types::global_dof_index MatrixBlock< MatrixType >::size_type

Declare type for container size.

Definition at line 111 of file matrix_block.h.

template<typename MatrixType>
typedef MatrixType::value_type MatrixBlock< MatrixType >::value_type

Declare a type for matrix entries.

Definition at line 116 of file matrix_block.h.

Constructor & Destructor Documentation

template<typename MatrixType >
MatrixBlock< MatrixType >::MatrixBlock ( )
inline

Constructor rendering an uninitialized object.

Definition at line 594 of file matrix_block.h.

template<typename MatrixType>
MatrixBlock< MatrixType >::MatrixBlock ( const MatrixBlock< MatrixType > &  M)
inline

Copy constructor.

Definition at line 603 of file matrix_block.h.

template<typename MatrixType>
MatrixBlock< MatrixType >::MatrixBlock ( size_type  i,
size_type  j 
)
inline

Constructor setting block coordinates, but not initializing the matrix.

Definition at line 616 of file matrix_block.h.

Member Function Documentation

template<typename MatrixType >
void MatrixBlock< MatrixType >::reinit ( const BlockSparsityPattern sparsity)
inline

Reinitialize the matrix for a new BlockSparsityPattern. This adjusts the matrix as well as the row_indices and column_indices.

Note
The row and column block structure of the sparsity pattern must be equal.

Definition at line 625 of file matrix_block.h.

template<typename MatrixType>
void MatrixBlock< MatrixType >::add ( const size_type  i,
const size_type  j,
const typename MatrixType::value_type  value 
)
inline

Add value to the element (i,j). Throws an error if the entry does not exist or if it is in a different block.

Definition at line 649 of file matrix_block.h.

template<typename MatrixType >
template<typename number >
void MatrixBlock< MatrixType >::add ( const std::vector< size_type > &  indices,
const FullMatrix< number > &  full_matrix,
const bool  elide_zero_values = true 
)
inline

Add all elements in a FullMatrix into sparse matrix locations given by indices. This function assumes a quadratic sparse matrix and a quadratic full_matrix. The global locations are translated into locations in this block and ExcBlockIndexMismatch is thrown, if the global index does not point into the block referred to by row and column.

Todo:
elide_zero_values is currently ignored.

The optional parameter elide_zero_values can be used to specify whether zero values should be added anyway or these should be filtered away and only non-zero data is added. The default value is true, i.e., zero values won't be added into the matrix.

Definition at line 733 of file matrix_block.h.

template<typename MatrixType >
template<typename number >
void MatrixBlock< MatrixType >::add ( const std::vector< size_type > &  row_indices,
const std::vector< size_type > &  col_indices,
const FullMatrix< number > &  full_matrix,
const bool  elide_zero_values = true 
)
inline

Add all elements in a FullMatrix into global locations given by row_indices and col_indices, respectively. The global locations are translated into locations in this block and ExcBlockIndexMismatch is thrown, if the global index does not point into the block referred to by row and column.

Todo:
elide_zero_values is currently ignored.

The optional parameter elide_zero_values can be used to specify whether zero values should be added anyway or these should be filtered away and only non-zero data is added. The default value is true, i.e., zero values won't be added into the matrix.

Definition at line 672 of file matrix_block.h.

template<typename MatrixType >
template<typename number >
void MatrixBlock< MatrixType >::add ( const size_type  row_index,
const std::vector< size_type > &  col_indices,
const std::vector< number > &  values,
const bool  elide_zero_values = true 
)
inline

Set several elements in the specified row of the matrix with column indices as given by col_indices to the respective value. This is the function doing the actual work for the ones adding full matrices. The global locations row_index and col_indices are translated into locations in this block and ExcBlockIndexMismatch is thrown, if the global index does not point into the block referred to by row and column.

Todo:
elide_zero_values is currently ignored.

The optional parameter elide_zero_values can be used to specify whether zero values should be added anyway or these should be filtered away and only non-zero data is added. The default value is true, i.e., zero values won't be added into the matrix.

Definition at line 754 of file matrix_block.h.

template<typename MatrixType >
template<typename number >
void MatrixBlock< MatrixType >::add ( const size_type  row,
const size_type  n_cols,
const size_type col_indices,
const number *  values,
const bool  elide_zero_values = true,
const bool  col_indices_are_sorted = false 
)
inline

Add an array of values given by values in the given global matrix row at columns specified by col_indices in the sparse matrix.

The optional parameter elide_zero_values can be used to specify whether zero values should be added anyway or these should be filtered away and only non-zero data is added. The default value is true, i.e., zero values won't be added into the matrix.

Definition at line 693 of file matrix_block.h.

template<typename MatrixType >
template<class VectorType >
void MatrixBlock< MatrixType >::vmult ( VectorType &  w,
const VectorType &  v 
) const
inline

Matrix-vector-multiplication, forwarding to the same function in MatrixType. No index computations are done, thus, the vectors need to have sizes matching matrix.

Definition at line 772 of file matrix_block.h.

template<typename MatrixType >
template<class VectorType >
void MatrixBlock< MatrixType >::vmult_add ( VectorType &  w,
const VectorType &  v 
) const
inline

Matrix-vector-multiplication, forwarding to the same function in MatrixType. No index computations are done, thus, the vectors need to have sizes matching matrix.

Definition at line 782 of file matrix_block.h.

template<typename MatrixType >
template<class VectorType >
void MatrixBlock< MatrixType >::Tvmult ( VectorType &  w,
const VectorType &  v 
) const
inline

Matrix-vector-multiplication, forwarding to the same function in MatrixType. No index computations are done, thus, the vectors need to have sizes matching matrix.

Definition at line 792 of file matrix_block.h.

template<typename MatrixType >
template<class VectorType >
void MatrixBlock< MatrixType >::Tvmult_add ( VectorType &  w,
const VectorType &  v 
) const
inline

Matrix-vector-multiplication, forwarding to the same function in MatrixType. No index computations are done, thus, the vectors need to have sizes matching matrix.

Definition at line 802 of file matrix_block.h.

template<typename MatrixType >
std::size_t MatrixBlock< MatrixType >::memory_consumption ( ) const
inline

The memory used by this object.

Definition at line 811 of file matrix_block.h.

template<typename MatrixType>
MatrixBlock< MatrixType >::DeclException2 ( ExcBlockIndexMismatch  ,
size_type  ,
size_type  ,
<< "Block index "<< arg1<< " does not match "<<  arg2 
)

The block number computed from an index by using BlockIndices does not match the block coordinates stored in this object.

Member Data Documentation

template<typename MatrixType>
size_type MatrixBlock< MatrixType >::row

Row coordinate. This is the position of the data member matrix on the global matrix.

Definition at line 281 of file matrix_block.h.

template<typename MatrixType>
size_type MatrixBlock< MatrixType >::column

Column coordinate. This is the position of the data member matrix on the global matrix.

Definition at line 286 of file matrix_block.h.

template<typename MatrixType>
MatrixType MatrixBlock< MatrixType >::matrix

The matrix itself

Definition at line 291 of file matrix_block.h.

template<typename MatrixType>
BlockIndices MatrixBlock< MatrixType >::row_indices
private

The row BlockIndices of the whole system. Using row(), this allows us to find the index of the first row degree of freedom for this block.

Definition at line 298 of file matrix_block.h.

template<typename MatrixType>
BlockIndices MatrixBlock< MatrixType >::column_indices
private

The column BlockIndices of the whole system. Using column(), this allows us to find the index of the first column degree of freedom for this block.

Definition at line 304 of file matrix_block.h.


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