CsrSparseMatrix< T > Class Template Reference

Concrete implementation of a matrix class representing a Compressed Sparse Rows (CSR) matrix. This class provides basic element-wise arithmetic operations and access functionality. It also includes a build-up feature that allows elements to be registered using the Coordinate (COO) format prior to finalization. Note: Adding elements after the initial build-up is not supported, as it is inefficient and unnecessary for static mesh use cases. More...

#include <CsrMatrix.h>

Inheritance diagram for CsrSparseMatrix< T >:

Collaboration diagram for CsrSparseMatrix< T >:

Public Member Functions
	CsrSparseMatrix ()
	Constructor for an empty CsrSparseMatrix object. Leaves all storage arrays empty but sets the flag to allow dynamic creation.
	CsrSparseMatrix (const std::valarray< T > &values, const std::valarray< int > &column_indices, const std::valarray< int > &row_pointers, int rows, int cols)
	Constructor for an CsrSparseMatrix object from given CSR arrays and row and column numbers. Disables dynamic creation.
void	registerElement (T value, int rows, int cols) override
	Registers the element and it's initial value within a COO style vector for dynamic creation.
void	buildAtRuntime ()
	Builds the CSR matrix from the registered elements. It sorts all elements of the COO format vector by rows (ascending) and within one row by columns (ascending) and marks the beginning of the rows while it progresses through the column sorting. Afterwards it simply writes the values and column indices into the corresponding layer. The sorting of the column indices it done to fully comply with the CSR format and allow for O( log(m) ) random access time (m = elements within a row). The flag m_dynamic_creation is always true, when using the default constructor and not passing already prebuild CSR arrays to the constructor.
void	buildAtRuntime (int rows, int cols) override
	Function overload for enforcing matrix boundaries. Builds the CSR matrix from the registered elements. It sorts all elements of the COO format vector by rows (ascending) and within one row by columns (ascending) and marks the beginning of the rows while it progresses through the column sorting. Afterwards it simply writes the values and column indices into the corresponding layer. The sorting of the column indices it done to fully comply with the CSR format and allow for O( log(m) ) random access time (m = elements within a row).
T &	operator() (int rows, int cols) override
	Parentheses operator to allow access to the non-zero matrix values as Matrix(row,column). The static CSR matrix only supports writing to predefined non-zero positions. Right now throws an error (int) when trying to access a zero element. Uses divide and conquer to allow access in O( log(m) ) time (m = elements within a row).
T	operator() (int rows, int cols) const override
	Parentheses operator to allow access to the non-zero matrix values as Matrix(row,column) in const contexts. Uses divide and conquer to allow access in O( log(m) ) time (m = elements within a row).
std::valarray< T >	scalarAdd (T value) const override
	Addition of a scalar to the matrix values element-wise.
std::valarray< T >	scalarSubtract (T value) const override
	Subtraction of a scalar from the matrix values element-wise.
std::valarray< T >	scalarMultiply (T value) const override
	Multiplication of a scalar to the matrix values element-wise.
std::valarray< T >	scalarDivide (T value) const override
	Division of a scalar from the matrix values element-wise.
std::valarray< T >	vectorElemAdd (const std::valarray< T > &value) const override
	Addition of a vector to the matrix values element-wise.
std::valarray< T >	vectorElemSubtract (const std::valarray< T > &value) const override
	Subtraction of a vector from the matrix values element-wise.
std::valarray< T >	vectorElemMultiply (const std::valarray< T > &value) const override
	Multiplication of a vector to the matrix values element-wise.
std::valarray< T >	vectorElemDivide (const std::valarray< T > &value) const override
	Division of a vector from the matrix values element-wise.
std::valarray< T >	matrixVectorMultiplication (const std::valarray< T > &value) const override
	Matrix-vector-multiplication for sparse matrices.
const MatrixType	getMatrixType () override
void	printMatrixInDenseFormat ()
	Prints the entire matrix with zero values.
const std::vector< std::pair< int, int > > &	getNonZeroElements () override
int	getNonZeroElementNumber () const
const std::valarray< int > &	getColumnIndices ()
const std::valarray< int > &	getRowPointers ()
const std::valarray< T > &	getMatrixElements ()
void	updateNonZeroElements () override
	Sets the m_non_zero_elememts parameter of the parent class.
	~CsrSparseMatrix () override=default
Public Member Functions inherited from GenericMatrix< T >
	GenericMatrix ()
	Constructor for an empty GenericMatrix object. This should only be used for dynamically created matrices.
	GenericMatrix (int rows, int cols)
	Constructor for an empty GenericMatrix object whose value array holds rows*cols values.
	GenericMatrix (int rows, int cols, const std::valarray< T > &values, bool allow_mismatched_size)
	Constructor for a GenericMatrix object whose value array holds rows*cols values given by the user. Can throw an std::ivalid_argument if provided with a value array of different size than rows*cols.
	GenericMatrix (const GenericMatrix &)=default
GenericMatrix &	operator= (const GenericMatrix &)=default
	GenericMatrix (GenericMatrix &&)=default
GenericMatrix &	operator= (GenericMatrix &&)=default
std::pair< int, int >	size () const
	Function that returns the size of the matrix as a pair in (rows, columns) format.
void	zeroMatrix ()
	Function that zeros out the matrix.
void	overwriteMatrixValues (const std::valarray< T > &value)
	Overwrites the matrix values with a new valarray. This is used for optimized overwrite when arithmetic operations on multiple matrices are used.
virtual	~GenericMatrix ()=default

Additional Inherited Members
Protected Attributes inherited from GenericMatrix< T >
int	m_rows {0}
int	m_cols {0}
std::valarray< T >	m_matrix {}
std::vector< std::pair< int, int > >	m_non_zero_elements

Detailed Description

template<typename T>
class CsrSparseMatrix< T >

Concrete implementation of a matrix class representing a Compressed Sparse Rows (CSR) matrix. This class provides basic element-wise arithmetic operations and access functionality. It also includes a build-up feature that allows elements to be registered using the Coordinate (COO) format prior to finalization. Note: Adding elements after the initial build-up is not supported, as it is inefficient and unnecessary for static mesh use cases.

Constructor & Destructor Documentation

CsrSparseMatrix() [1/2]

template<typename T >

CsrSparseMatrix< T >::CsrSparseMatrix

(

)

Constructor for an empty CsrSparseMatrix object. Leaves all storage arrays empty but sets the flag to allow dynamic creation.

CsrSparseMatrix() [2/2]

template<typename T >

CsrSparseMatrix< T >::CsrSparseMatrix	(	const std::valarray< T > &	values,
		const std::valarray< int > &	column_indices,
		const std::valarray< int > &	row_pointers,
		int	rows,
		int	cols
	)

Constructor for an CsrSparseMatrix object from given CSR arrays and row and column numbers. Disables dynamic creation.

Parameters

values	The value array of the CSR format with the non-zero values of the sparse matrix.
column_indices	The array holding all the column index values of the CSR format.
row_pointers	The array with the indices to the elements that mark the beginning of a new row of the CSR format.
rows	The number of rows of the matrix.
cols	The number of columns of the matrix.

~CsrSparseMatrix()

template<typename T >

CsrSparseMatrix< T >::~CsrSparseMatrix ( )

overridedefault

Member Function Documentation

buildAtRuntime() [1/2]

template<typename T >

void CsrSparseMatrix< T >::buildAtRuntime

(

)

Builds the CSR matrix from the registered elements. It sorts all elements of the COO format vector by rows (ascending) and within one row by columns (ascending) and marks the beginning of the rows while it progresses through the column sorting. Afterwards it simply writes the values and column indices into the corresponding layer. The sorting of the column indices it done to fully comply with the CSR format and allow for O( log(m) ) random access time (m = elements within a row). The flag m_dynamic_creation is always true, when using the default constructor and not passing already prebuild CSR arrays to the constructor.

buildAtRuntime() [2/2]

template<typename T >

void CsrSparseMatrix< T >::buildAtRuntime ( int  rows, int  cols  )

overridevirtual

Function overload for enforcing matrix boundaries. Builds the CSR matrix from the registered elements. It sorts all elements of the COO format vector by rows (ascending) and within one row by columns (ascending) and marks the beginning of the rows while it progresses through the column sorting. Afterwards it simply writes the values and column indices into the corresponding layer. The sorting of the column indices it done to fully comply with the CSR format and allow for O( log(m) ) random access time (m = elements within a row).

Parameters

rows	the row to access.
cols	the column to access.

Implements GenericMatrix< T >.

getColumnIndices()

template<typename T >

const std::valarray< int > & CsrSparseMatrix< T >::getColumnIndices ( )

inline

getMatrixElements()

template<typename T >

const std::valarray< T > & CsrSparseMatrix< T >::getMatrixElements ( )

inline

getMatrixType()

template<typename T >

const MatrixType CsrSparseMatrix< T >::getMatrixType ( )

inlineoverridevirtual

Implements GenericMatrix< T >.

getNonZeroElementNumber()

template<typename T >

int CsrSparseMatrix< T >::getNonZeroElementNumber ( ) const

inline

getNonZeroElements()

template<typename T >

const std::vector< std::pair< int, int > > & CsrSparseMatrix< T >::getNonZeroElements ( )

inlineoverridevirtual

Implements GenericMatrix< T >.

getRowPointers()

template<typename T >

const std::valarray< int > & CsrSparseMatrix< T >::getRowPointers ( )

inline

matrixVectorMultiplication()

template<typename T >

std::valarray< T > CsrSparseMatrix< T >::matrixVectorMultiplication ( const std::valarray< T > &  value ) const

overridevirtual

Matrix-vector-multiplication for sparse matrices.

Template Parameters

T	Numeric type used in the simulation (e.g., float, double)

Parameters

value

The vector with the values to multiply to the matrix from the right.

Returns

std::valarray<T> The resulting vector.

Implements GenericMatrix< T >.

operator()() [1/2]

template<typename T >

T CsrSparseMatrix< T >::operator() ( int  rows, int  cols  ) const

overridevirtual

Parentheses operator to allow access to the non-zero matrix values as Matrix(row,column) in const contexts. Uses divide and conquer to allow access in O( log(m) ) time (m = elements within a row).

Parameters

rows	the row to access.
cols	the column to access.

Implements GenericMatrix< T >.

operator()() [2/2]

template<typename T >

T & CsrSparseMatrix< T >::operator() ( int  rows, int  cols  )

overridevirtual

Parentheses operator to allow access to the non-zero matrix values as Matrix(row,column). The static CSR matrix only supports writing to predefined non-zero positions. Right now throws an error (int) when trying to access a zero element. Uses divide and conquer to allow access in O( log(m) ) time (m = elements within a row).

Parameters

rows	the row to access.
cols	the column to access.

Implements GenericMatrix< T >.

printMatrixInDenseFormat()

template<typename T >

void CsrSparseMatrix< T >::printMatrixInDenseFormat

(

)

Prints the entire matrix with zero values.

registerElement()

template<typename T >

void CsrSparseMatrix< T >::registerElement ( T  value, int  rows, int  cols  )

overridevirtual

Registers the element and it's initial value within a COO style vector for dynamic creation.

Parameters

value	The initial value of the element.
rows	The row index of that element within the matrix.
cols	The column index of that element within the matrix.

Reimplemented from GenericMatrix< T >.

scalarAdd()

template<typename T >

std::valarray< T > CsrSparseMatrix< T >::scalarAdd ( T  value ) const

overridevirtual

Addition of a scalar to the matrix values element-wise.

Parameters

value

The scalar value to be added to the calling object.

Implements GenericMatrix< T >.

scalarDivide()

template<typename T >

std::valarray< T > CsrSparseMatrix< T >::scalarDivide ( T  value ) const

overridevirtual

Division of a scalar from the matrix values element-wise.

Parameters

value

The scalar value to divide to the calling object by.

Implements GenericMatrix< T >.

scalarMultiply()

template<typename T >

std::valarray< T > CsrSparseMatrix< T >::scalarMultiply ( T  value ) const

overridevirtual

Multiplication of a scalar to the matrix values element-wise.

Parameters

value

The scalar value to be multiplied to the calling object.

Implements GenericMatrix< T >.

scalarSubtract()

template<typename T >

std::valarray< T > CsrSparseMatrix< T >::scalarSubtract ( T  value ) const

overridevirtual

Subtraction of a scalar from the matrix values element-wise.

Parameters

value

The scalar value to be subtracted from the calling object.

Implements GenericMatrix< T >.

updateNonZeroElements()

template<typename T >

void CsrSparseMatrix< T >::updateNonZeroElements ( )

overridevirtual

Sets the m_non_zero_elememts parameter of the parent class.

Template Parameters

T	Numeric type used in the simulation (e.g., float, double)

Implements GenericMatrix< T >.

vectorElemAdd()

template<typename T >

std::valarray< T > CsrSparseMatrix< T >::vectorElemAdd ( const std::valarray< T > &  value ) const

overridevirtual

Addition of a vector to the matrix values element-wise.

Parameters

value

The vector values to be added to the calling object.

Implements GenericMatrix< T >.

vectorElemDivide()

template<typename T >

std::valarray< T > CsrSparseMatrix< T >::vectorElemDivide ( const std::valarray< T > &  value ) const

overridevirtual

Division of a vector from the matrix values element-wise.

Parameters

value

The vector value to divide to the calling object by.

Implements GenericMatrix< T >.

vectorElemMultiply()

template<typename T >

std::valarray< T > CsrSparseMatrix< T >::vectorElemMultiply ( const std::valarray< T > &  value ) const

overridevirtual

Multiplication of a vector to the matrix values element-wise.

Parameters

value

The vector value to be multiplied to the calling object.

Implements GenericMatrix< T >.

vectorElemSubtract()

template<typename T >

std::valarray< T > CsrSparseMatrix< T >::vectorElemSubtract ( const std::valarray< T > &  value ) const

overridevirtual

Subtraction of a vector from the matrix values element-wise.

Parameters

value

The vector value to be subtracted from the calling object.

Implements GenericMatrix< T >.

---

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

• src/CsrMatrix.h