Compressed sparse row format. More...

#include <spm.h>

Inheritance diagram for gridripper::math::CsrSpM< TD, TI >:

Public Member Functions
	CsrSpM (int nrows, int ncols, int nonzero)
	Creates a sparse matrix with the specified number of rows and (nonzero) diagonals.
void	realloc (int nrows, int ncols, int nonzero)
	Reallocates memory (if needed).
void	setRowPtr (int row, TI p)
	Sets the data array index of a row.
void	set (int row, int col, TD v, int i)
	Sets a matrix element.
const TD *	internalArray () const
	Gets the data array.
TD *	internalArray ()
	Gets the data array.
int	arrayLength () const
	Gets the length of the internal data array.
const TI *	internalRowPtrArray () const
	Gets the array containing the indices of the first nonzero row elements.
TI *	internalRowPtrArray ()
	Gets the array containing the indices of the first nonzero row elements.
const TI *	internalColumnIndexArray () const
	Gets the column indices array.
TI *	internalColumnIndexArray ()
	Gets the column indices array.
int	countNonzerosInRow (int row) const
	Gets the number of nonzeros in the specified row.
void	gmv (TD y, TD alpha, const TD x, TD beta, const TD *b) const
	General matrix-vector operation.
int	getNonzeroCount () const
	Gets the number of nonzero elements.
void	toDenseMatrix (TD *mat) const
	Converts the sparse matrix to a dense matrix.

Detailed Description

template<class TD, class TI>
class gridripper::math::CsrSpM< TD, TI >

Compressed sparse row format.

See N. Bell and M. Garland, "Efficient Sparse Matrix-Vector Multiplication on CUDA", NVIDIA Technical Report NVR-2008-004.

Version:: 07/03/2009

Since:: 06/30/2009

Author:: Peter Csizmadia

Constructor & Destructor Documentation

template<class TD , class TI >

gridripper::math::CsrSpM< TD, TI >::CsrSpM	(	int	nrows,
		int	ncols,
		int	nonzero
	)			`[inline]`

Creates a sparse matrix with the specified number of rows and (nonzero) diagonals.

Parameters:

	nrows	the number of rows
	ncols	the number of columns
	nonzero	number of nonzero elements

Member Function Documentation

template<class TD , class TI >

int gridripper::math::CsrSpM< TD, TI >::arrayLength ( ) const [inline]

Gets the length of the internal data array.

Returns:: array length

template<class TD , class TI >

int gridripper::math::CsrSpM< TD, TI >::countNonzerosInRow ( int row ) const [inline, virtual]

Gets the number of nonzeros in the specified row.

Parameters:

row

the row index

Returns:: the number of nonzeros

Implements gridripper::math::SpM< TD >.

template<class TD , class TI >

int gridripper::math::CsrSpM< TD, TI >::getNonzeroCount ( ) const [inline]

Gets the number of nonzero elements.

Returns:: the number of nonzero elements

template<class TD , class TI >

void gridripper::math::CsrSpM< TD, TI >::gmv	(	TD *	y,
		TD	alpha,
		const TD *	x,
		TD	beta,
		const TD *	b
	)			const `[inline, virtual]`

General matrix-vector operation.

Computes y = alpha*A*x + beta*b.

Parameters:

	y	the output vector
	alpha	scalar multiplier applied to A
	x	the x vector
	beta	scalar multiplier applied to b
	b	the b vector or NULL

Implements gridripper::math::SpM< TD >.

template<class TD , class TI >

TD* gridripper::math::CsrSpM< TD, TI >::internalArray ( ) [inline]

Gets the data array.

Returns:: the data array

template<class TD , class TI >

const TD* gridripper::math::CsrSpM< TD, TI >::internalArray ( ) const [inline]

Gets the data array.

Returns:: the data array

template<class TD , class TI >

const TI* gridripper::math::CsrSpM< TD, TI >::internalColumnIndexArray ( ) const [inline]

Gets the column indices array.

Returns:: the indices

template<class TD , class TI >

TI* gridripper::math::CsrSpM< TD, TI >::internalColumnIndexArray ( ) [inline]

Gets the column indices array.

Returns:: the indices

template<class TD , class TI >

TI* gridripper::math::CsrSpM< TD, TI >::internalRowPtrArray ( ) [inline]

Gets the array containing the indices of the first nonzero row elements.

Returns:: array of row pointers

template<class TD , class TI >

const TI* gridripper::math::CsrSpM< TD, TI >::internalRowPtrArray ( ) const [inline]

Gets the array containing the indices of the first nonzero row elements.

Returns:: array of row pointers

template<class TD , class TI >

void gridripper::math::CsrSpM< TD, TI >::realloc	(	int	nrows,
		int	ncols,
		int	nonzero
	)			`[inline]`

Reallocates memory (if needed).

Parameters:

	nrows	new number of rows
	ncols	new number of columns
	nonzero	maximum number of nonzero elements

template<class TD , class TI >

void gridripper::math::CsrSpM< TD, TI >::set	(	int	row,
		int	col,
		TD	v,
		int	i
	)			`[inline]`

Sets a matrix element.

Parameters:

	row	the row index
	col	the column
	v	the value
	i	index

template<class TD , class TI >

void gridripper::math::CsrSpM< TD, TI >::setRowPtr	(	int	row,
		TI	p
	)			`[inline]`

Sets the data array index of a row.

Parameters:

	row	the row index
	p	index of first nonzero element in the data array

template<class TD , class TI >

void gridripper::math::CsrSpM< TD, TI >::toDenseMatrix ( TD * mat ) const [inline, virtual]

Converts the sparse matrix to a dense matrix.

Parameters:

mat

output array for the dense matrix

Implements gridripper::math::SpM< TD >.

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

/home/laszloa/SVN/GR/gridripper/include/gridripper/math/spm.h

gridripper::math::CsrSpM< TD, TI > Class Template Reference

Public Member Functions

Detailed Description

template<class TD, class TI> class gridripper::math::CsrSpM< TD, TI >

Constructor & Destructor Documentation

Member Function Documentation

template<class TD, class TI>
class gridripper::math::CsrSpM< TD, TI >