INMOST::Sparse::Matrix Class Reference

Class to store the distributed sparse matrix by compressed rows. More...

#include <inmost_sparse.h>

Public Types
typedef Rows::iterator	iterator
typedef Rows::const_iterator	const_iterator

Public Member Functions
	Matrix (std::string _name="", INMOST_DATA_ENUM_TYPE start=0, INMOST_DATA_ENUM_TYPE end=0, INMOST_MPI_Comm _comm=INMOST_MPI_COMM_WORLD)
	Main constructor of the Matrix class. More...
	Matrix (const Matrix &other)
Matrix &	operator= (Matrix const &other)
Row &	operator[] (INMOST_DATA_ENUM_TYPE i)
	Return reference to i-th Row of the matrix.
const Row &	operator[] (INMOST_DATA_ENUM_TYPE i) const
	Return reference to i-th Row of the matrix.
INMOST_DATA_ENUM_TYPE	Size () const
	Return the total number of rows in the matrix.
bool	Empty () const
iterator	Begin ()
iterator	End ()
const_iterator	Begin () const
const_iterator	End () const
void	SetInterval (INMOST_DATA_ENUM_TYPE start, INMOST_DATA_ENUM_TYPE end)
	Set the start and the end row numbers of the distributed matrix interval.
void	GetInterval (INMOST_DATA_ENUM_TYPE &start, INMOST_DATA_ENUM_TYPE &end) const
	Get the start and the end row numbers of the distributed matrix interval.
void	ShiftInterval (INMOST_DATA_ENUM_TYPE shift)
INMOST_DATA_ENUM_TYPE	GetFirstIndex () const
	Get the first row index of the distributed matrix interval.
INMOST_DATA_ENUM_TYPE	GetLastIndex () const
	Get the last row index of the distributed matrix interval.
INMOST_MPI_Comm	GetCommunicator () const
	Get the communicator which the matrix is associated with.
void	MoveRows (INMOST_DATA_ENUM_TYPE from, INMOST_DATA_ENUM_TYPE to, INMOST_DATA_ENUM_TYPE size)
void	Swap (Matrix &other)
void	MatVec (INMOST_DATA_REAL_TYPE alpha, Vector &x, INMOST_DATA_REAL_TYPE beta, Vector &y) const
	Matrix-vector product of the form: y = alpha*A*x + beta * y. More...
void	MatVecTranspose (INMOST_DATA_REAL_TYPE alpha, Vector &x, INMOST_DATA_REAL_TYPE beta, Vector &y) const
	Matrix-vector product with transposed matrix of the form: y = alpha*A^T*x + beta * y. More...
void	Clear ()
	Clear all data of the matrix.
INMOST_DATA_ENUM_TYPE	Nonzeros ()
	Count number of nonzeros in matrix.
void	Load (std::string file, INMOST_DATA_ENUM_TYPE beg=ENUMUNDEF, INMOST_DATA_ENUM_TYPE end=ENUMUNDEF, std::string file_ord="")
	Load the matrix from a single data file in MTX format using the specified interval. More...
void	Save (std::string file, const AnnotationService *annotation=NULL)
	Save the distributed matrix to a single data file in MTX format using parallel MPI I/O. More...
bool &	isParallel ()
	Check that matrix is in parallel state.
const bool &	isParallel () const
std::string	GetName () const
	Get the matrix name specified in the main constructor.
void	Sort ()
	Sort rows.
INMOST_DATA_REAL_TYPE	Residual (Sparse::Vector &RHS, Sparse::Vector &SOL)
	Calculate the real residual. More...

Static Public Member Functions
static void	ZAXPBY (INMOST_DATA_REAL_TYPE alpha, const Sparse::Matrix &X, INMOST_DATA_REAL_TYPE beta, const Sparse::Matrix &Y, Sparse::Matrix &Z)
	ZAXPBY operation for sparse matrices Z = alpha * X + beta * Y.

Detailed Description

Class to store the distributed sparse matrix by compressed rows.

The format used to store sparse matrix is analogous to Compressed Row Storage format (CRS).

See also

http://netlib.org/linalg/html_templates/node91.html

Definition at line 474 of file inmost_sparse.h.

Constructor & Destructor Documentation

Matrix()

INMOST::Sparse::Matrix::Matrix	(	std::string	_name = "",
		INMOST_DATA_ENUM_TYPE	start = 0,
		INMOST_DATA_ENUM_TYPE	end = 0,
		INMOST_MPI_Comm	_comm = INMOST_MPI_COMM_WORLD
	)

Main constructor of the Matrix class.

Parameters

_name	Name of the matrix, empty string by default.
start	Start of the local data interval.
end	End of the local data interval.
_comm	Communicator for parallel data exchanges, MPI_COMM_WORLD by default.

Member Function Documentation

Load()

void INMOST::Sparse::Matrix::Load	(	std::string	file,
		INMOST_DATA_ENUM_TYPE	beg = ENUMUNDEF,
		INMOST_DATA_ENUM_TYPE	end = ENUMUNDEF,
		std::string	file_ord = ""
	)

Load the matrix from a single data file in MTX format using the specified interval.

If interval is not specified, then it will be automatically constructed, with the about equal block size (the last block may has larger dimension).

MatVec()

void INMOST::Sparse::Matrix::MatVec	(	INMOST_DATA_REAL_TYPE	alpha,
		Vector &	x,
		INMOST_DATA_REAL_TYPE	beta,
		Vector &	y
	)		const

Matrix-vector product of the form: y = alpha*A*x + beta * y.

Parameters

y	Input/output vector.

MatVecTranspose()

void INMOST::Sparse::Matrix::MatVecTranspose	(	INMOST_DATA_REAL_TYPE	alpha,
		Vector &	x,
		INMOST_DATA_REAL_TYPE	beta,
		Vector &	y
	)		const

Matrix-vector product with transposed matrix of the form: y = alpha*A^T*x + beta * y.

Parameters

y	Input/output vector.

Residual()

INMOST_DATA_REAL_TYPE INMOST::Sparse::Matrix::Residual	(	Sparse::Vector &	RHS,
		Sparse::Vector &	SOL
	)

Calculate the real residual.

Parameters

RHS	The right-hand side Vector b.
SOL	The initial guess to the solution on input and the solution Vector x on return.

Returns

||A*x-b||

It is assumed that the coefficient matrix A have been set and the preconditioner have been already constructed.

Save()

void INMOST::Sparse::Matrix::Save	(	std::string	file,
		const AnnotationService *	annotation = NULL
	)

Save the distributed matrix to a single data file in MTX format using parallel MPI I/O.

See also

http://math.nist.gov/MatrixMarket/formats.html

---

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

• inmost_sparse.h