Class of COO matrices. Note that most coo matrix functions are not yet on the device. More...

#include <coo_matrix.hpp>

Inheritance diagram for pargemslr::CooMatrixClass< T >:

Public Member Functions
	CooMatrixClass ()
	The constructor of CooMatrixClass. More...

	CooMatrixClass (const CooMatrixClass< T > &mat)
	The copy constructor of CooMatrixClass. More...

	CooMatrixClass (CooMatrixClass< T > &&mat)
	The move constructor of CooMatrixClass. More...

CooMatrixClass< T > &	operator= (const CooMatrixClass< T > &mat)
	The = operator of CooMatrixClass. More...

CooMatrixClass< T > &	operator= (CooMatrixClass< T > &&mat)
	The = operator of CooMatrixClass. More...

virtual int	Clear ()
	Free the current matrix. More...

virtual	~CooMatrixClass ()
	The destructor of CooMatrixClass. More...

int	Setup (int nrows, int ncols)
	Free the current matrix, and create the new matrix using the default reserved memory size on the host. More...

int	Setup (int nrows, int ncols, int reserve)
	Free the current matrix, and create the new matrix and reserve some space in the host memory. More...

int	Setup (int nrows, int ncols, int reserve, int location)
	Free the current matrix, and create the new matrix and reserve some space in the memory. More...

virtual int	GetDataLocation () const
	Get the data location of the matrix. More...

virtual int	GetNumRowsLocal () const
	Get the local number of rows of the matrix. More...

virtual int	GetNumColsLocal () const
	Get the local number of columns of the matrix. More...

virtual long int	GetNumNonzeros () const
	Get the number of nonzeros in this matrix. More...

int	SetNumCols (int cols)
	Set the number of cols in this matrix. More...

int	SetNumNonzeros ()
	Set the number of nonzeros in this matrix. More...

virtual int	Eye ()
	Create an indentity matrix. More...

virtual int	Fill (const T &v)
	Fill the matrix pattern with constant value. More...

virtual int	Scale (const T &alpha)
	Scale the matrix. More...

virtual int	MoveData (const int &location)
	Move the data to another memory location. More...

int	PushBack (int row, int col, T v)
	Insert value to the COO matrix. More...

int *	GetI () const
	Get the I pointer of the matrix. More...

int *	GetJ () const
	Get the J pointer of the matrix. More...

T *	GetData () const
	Get the data pointer of the matrix. More...

IntVectorClass< int > &	GetIVector ()
	Get the reference to the I vector. More...

const IntVectorClass< int > &	GetIVector () const
	Get the reference to the I vector. More...

IntVectorClass< int > &	GetJVector ()
	Get the reference to the J vector. More...

const IntVectorClass< int > &	GetJVector () const
	Get the reference to the J vector. More...

SequentialVectorClass< T > &	GetDataVector ()
	Get the reference to the data vector. More...

const SequentialVectorClass< T > &	GetDataVector () const
	Get the reference to the data vector. More...

int	ToCsr (int location, CsrMatrixClass< T > &csrmat_out)
	Convert the matrix to csr. More...

int	Laplacian (int nx, int ny, int nz, T alphax, T alphay, T alphaz, T shift, bool rand_perturb=false)
	Generate Laplacian matrix, 5-pt for 2D problem and 7-pt for 3D problem. More...

int	Helmholtz (int n, T w)
	Generate 3D Helmholtz matrix with 7-pt on [0,1]^3. -\Delta u - w^2 u = f. More...

int	ReadFromMMFile (const char *matfile, int idxin)
	Read matrix on the host memory, matrix market format. More...

int	PlotPatternGnuPlot (const char datafilename, int rperm, int *cperm, int conditiona, int conditionb)
	Plot the pattern of the csr matrix using gnuplot. Similar to spy in the MATLAB. Function for testing purpose. More...

Public Member Functions inherited from pargemslr::MatrixClass< T >
	MatrixClass ()
	The constructor of MatrixClass. More...

	MatrixClass (const MatrixClass< T > &mat)
	The copy constructor of MatrixClass. More...

	MatrixClass (MatrixClass< T > &&mat)
	The move constructor of MatrixClass. More...

virtual	~MatrixClass ()
	The destructor of MatrixClass. More...

virtual int	MatVec (char trans, const T &alpha, const VectorClass< T > &x, const T &beta, VectorClass< T > &y)
	In place Matrix-Vector product ==> y := alphaAx + betay, or y := alphaA'x + betay. More...

virtual int	MatVec (char trans, const T &alpha, const VectorClass< T > &x, const T &beta, const VectorClass< T > &y, VectorClass< T > &z)
	In place Matrix-Vector product ==> z := alphaAx + betay, or z := alphaA'x + betay. More...

PrecisionEnum	GetPrecision () const
	Return the precision of the matrix. More...

Detailed Description

template<typename T>
class pargemslr::CooMatrixClass< T >

Class of COO matrices. Note that most coo matrix functions are not yet on the device.

Constructor & Destructor Documentation

◆ CooMatrixClass() [1/3]

template<typename T >

template pargemslr::CooMatrixClass< T >::CooMatrixClass ( )

The constructor of CooMatrixClass.

◆ CooMatrixClass() [2/3]

template<typename T >

pargemslr::CooMatrixClass< T >::CooMatrixClass ( const CooMatrixClass< T > & mat )

The copy constructor of CooMatrixClass.

Parameters

[in] mat The other matrix.

◆ CooMatrixClass() [3/3]

template<typename T >

pargemslr::CooMatrixClass< T >::CooMatrixClass ( CooMatrixClass< T > && mat )

The move constructor of CooMatrixClass.

Parameters

[in] mat The other matrix.

◆ ~CooMatrixClass()

template<typename T >

template pargemslr::CooMatrixClass< T >::~CooMatrixClass ( )

virtual

The destructor of CooMatrixClass.

Member Function Documentation

◆ Clear()

template<typename T >

template int pargemslr::CooMatrixClass< T >::Clear ( )

virtual

Free the current matrix.

Returns: Return error message.

Reimplemented from pargemslr::MatrixClass< T >.

◆ Eye()

template<typename T >

template int pargemslr::CooMatrixClass< T >::Eye ( )

virtual

Create an indentity matrix.

Returns: Return error message.

Reimplemented from pargemslr::MatrixClass< T >.

◆ Fill()

template<typename T >

template int pargemslr::CooMatrixClass< T >::Fill ( const T & v )

virtual

Fill the matrix pattern with constant value.

Parameters

[in] v The value to be filled.

Returns: Return error message.

Reimplemented from pargemslr::MatrixClass< T >.

◆ GetData()

template<typename T >

template complexd * pargemslr::CooMatrixClass< T >::GetData ( ) const

Get the data pointer of the matrix.

Returns: Return the data pointer.

◆ GetDataLocation()

template<typename T >

template int pargemslr::CooMatrixClass< T >::GetDataLocation ( ) const

virtual

Get the data location of the matrix.

Returns: Return the data location of the matrix.

Get the data location of the matrix.

Returns: Return the length of the vector.

Reimplemented from pargemslr::MatrixClass< T >.

◆ GetDataVector() [1/2]

template<typename T >

template SequentialVectorClass< complexd > & pargemslr::CooMatrixClass< T >::GetDataVector ( )

Get the reference to the data vector.

Returns: Return the reference to the data vector.

◆ GetDataVector() [2/2]

template<typename T >

template const SequentialVectorClass< complexd > & pargemslr::CooMatrixClass< T >::GetDataVector ( ) const

Get the reference to the data vector.

Returns: Return the reference to the data vector.

◆ GetI()

template<typename T >

template int * pargemslr::CooMatrixClass< T >::GetI ( ) const

Get the I pointer of the matrix.

Returns: Return the data pointer.

◆ GetIVector() [1/2]

template<typename T >

template IntVectorClass< int > & pargemslr::CooMatrixClass< T >::GetIVector ( )

Get the reference to the I vector.

Returns: Return the reference to the I vector.

◆ GetIVector() [2/2]

template<typename T >

template const IntVectorClass< int > & pargemslr::CooMatrixClass< T >::GetIVector ( ) const

Get the reference to the I vector.

Returns: Return the reference to the I vector.

◆ GetJ()

template<typename T >

template int * pargemslr::CooMatrixClass< T >::GetJ ( ) const

Get the J pointer of the matrix.

Returns: Return the data pointer.

◆ GetJVector() [1/2]

template<typename T >

template IntVectorClass< int > & pargemslr::CooMatrixClass< T >::GetJVector ( )

Get the reference to the J vector.

Returns: Return the reference to the J vector.

◆ GetJVector() [2/2]

template<typename T >

template const IntVectorClass< int > & pargemslr::CooMatrixClass< T >::GetJVector ( ) const

Get the reference to the J vector.

Returns: Return the reference to the J vector.

◆ GetNumColsLocal()

template<typename T >

template int pargemslr::CooMatrixClass< T >::GetNumColsLocal ( ) const

virtual

Get the local number of columns of the matrix.

Returns: Return the local number of columns of the matrix.

Get the local number of columns of the matrix.

Returns: Return the length of the matrix.

Implements pargemslr::MatrixClass< T >.

◆ GetNumNonzeros()

template<typename T >

template long int pargemslr::CooMatrixClass< T >::GetNumNonzeros ( ) const

virtual

Get the number of nonzeros in this matrix.

Returns: Return the number of nonzeros in this matrix.

Reimplemented from pargemslr::MatrixClass< T >.

◆ GetNumRowsLocal()

template<typename T >

template int pargemslr::CooMatrixClass< T >::GetNumRowsLocal ( ) const

virtual

Get the local number of rows of the matrix.

Returns: Return the local number of rows of the matrix.

Get the local number of rows of the matrix.

Returns: Return the length of the matrix.

Implements pargemslr::MatrixClass< T >.

◆ Helmholtz()

template<typename T >

template int pargemslr::CooMatrixClass< T >::Helmholtz	(	int	n,
		T	w
	)

Generate 3D Helmholtz matrix with 7-pt on [0,1]^3. -\Delta u - w^2 u = f. Robin boundary condition u'-iwu = 0 is used.

Parameters

[in]	n	Problem size.
[in]	w	The w value.

Returns: Return error message.

◆ Laplacian()

template<typename T >

template int pargemslr::CooMatrixClass< T >::Laplacian	(	int	nx,
		int	ny,
		int	nz,
		T	alphax,
		T	alphay,
		T	alphaz,
		T	shift,
		bool	rand_perturb = `false`
	)

Generate Laplacian matrix, 5-pt for 2D problem and 7-pt for 3D problem.

Parameters

[in]	nx
[in]	ny
[in]	nz	Problem size on each dimension
[in]	alphax
[in]	alphay
[in]	alphaz	Alpha value on each direction
[in]	shift	The shift value on diagonal.
[in]	rand_perturb	Set to true to add extra random real pertubation to diagonal entries as abs(shift)*Rand(). We only add shift that might makes the problem harder.

Returns: Return error message.

◆ MoveData()

template<typename T >

template int pargemslr::CooMatrixClass< T >::MoveData ( const int & location )

virtual

Move the data to another memory location.

Parameters

[in] location The location move to.

Returns: Return error message.

Reimplemented from pargemslr::MatrixClass< T >.

◆ operator=() [1/2]

template<typename T >

CooMatrixClass< T > & pargemslr::CooMatrixClass< T >::operator= ( const CooMatrixClass< T > & mat )

The = operator of CooMatrixClass.

Parameters

[in] mat The other matrix.

Returns: Return this matrix.

◆ operator=() [2/2]

template<typename T >

CooMatrixClass< T > & pargemslr::CooMatrixClass< T >::operator= ( CooMatrixClass< T > && mat )

The = operator of CooMatrixClass.

Parameters

[in] mat The other matrix.

Returns: Return this matrix.

◆ PlotPatternGnuPlot()

template<typename T >

template int pargemslr::CooMatrixClass< T >::PlotPatternGnuPlot	(	const char *	datafilename,
		int *	rperm,
		int *	cperm,
		int	conditiona,
		int	conditionb
	)

Plot the pattern of the csr matrix using gnuplot. Similar to spy in the MATLAB. Function for testing purpose.

Parameters

[in]	datafilename	The filename of the temp file holding the data.
[in]	rperm	The row permutation vector.
[in]	cperm	The column permutation vector. Set either rperm or cperm to NULL to spy(A), otherwise spy(A(rperm,cperm)).
[in]	conditiona	First condition.
[in]	conditionb	Secend condition, only plot when conditiona == conditionb.

Returns: Return error message.

◆ PushBack()

template<typename T >

template int pargemslr::CooMatrixClass< T >::PushBack	(	int	row,
		int	col,
		T	v
	)

Insert value to the COO matrix.

Parameters

[in]	row.
[in]	col	The col/row to be inserted into the csr/csc.
[in]	v	The value to be inserted.

Returns: Return error message.

◆ ReadFromMMFile()

template<typename T >

template int pargemslr::CooMatrixClass< T >::ReadFromMMFile	(	const char *	matfile,
		int	idxin
	)

Read matrix on the host memory, matrix market format.

Parameters

[in]	matfile	The file.
[in]	idxin	The index base of the input file. 0-based or 1-based.

Returns: Return error message.

◆ Scale()

template<typename T >

template int pargemslr::CooMatrixClass< T >::Scale ( const T & alpha )

virtual

Scale the matrix.

Parameters

[in] alpha The scale.

Returns: Return error message.

Reimplemented from pargemslr::MatrixClass< T >.

◆ SetNumCols()

template<typename T >

template int pargemslr::CooMatrixClass< T >::SetNumCols ( int cols )

Set the number of cols in this matrix.

Returns: Return error message.

◆ SetNumNonzeros()

template<typename T >

template int pargemslr::CooMatrixClass< T >::SetNumNonzeros ( )

Set the number of nonzeros in this matrix.

Returns: Return error message.

◆ Setup() [1/3]

template<typename T >

template int pargemslr::CooMatrixClass< T >::Setup	(	int	nrows,
		int	ncols
	)

Free the current matrix, and create the new matrix. The default reserved memory size on the host is MAX(nrow, ncol)*_coo_reserve_fact.

Parameters

[in]	nrows	The number of rows of the new matrix.
[in]	ncols	The number of cols of the new matrix.

Returns: Return the length of the matrix.

◆ Setup() [2/3]

template<typename T >

template int pargemslr::CooMatrixClass< T >::Setup	(	int	nrows,
		int	ncols,
		int	reserve
	)

Free the current matrix, and create the new matrix and reserve some space in the host memory.

Parameters

[in]	nrows	The number of rows of the new matrix.
[in]	ncols	The number of cols of the new matrix.
[in]	reserve	The reserved vector length. A good estimation can avoid to many re-allocation when inserting values.

Returns: Return the length of the matrix.

◆ Setup() [3/3]

template<typename T >

template int pargemslr::CooMatrixClass< T >::Setup	(	int	nrows,
		int	ncols,
		int	reserve,
		int	location
	)

Free the current matrix, and create the new matrix and reserve some space in the memory.

Parameters

[in]	nrows	The number of rows of the new matrix.
[in]	ncols	The number of cols of the new matrix.
[in]	reserve	The reserved vector length. A good estimation can avoid to many re-allocation when inserting values.
[in]	location	The location of the memory.

Returns: Return the length of the matrix.

◆ ToCsr()

template<typename T >

template int pargemslr::CooMatrixClass< T >::ToCsr	(	int	location,
		CsrMatrixClass< T > &	csrmat_out
	)

Convert the matrix to csr.

Parameters

[in]	location	The location of the output csr matrix.
[out]	csrmat_out	The output csr matrix.

Returns: Return error message.

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

SRC/matrices/coo_matrix.hpp
SRC/matrices/coo_matrix.cpp

Public Member Functions

Detailed Description

template<typename T> class pargemslr::CooMatrixClass< T >

Constructor & Destructor Documentation

◆ CooMatrixClass() [1/3]

◆ CooMatrixClass() [2/3]

◆ CooMatrixClass() [3/3]

◆ ~CooMatrixClass()

Member Function Documentation

◆ Clear()

◆ Eye()

◆ Fill()

◆ GetData()

◆ GetDataLocation()

◆ GetDataVector() [1/2]

◆ GetDataVector() [2/2]

◆ GetI()

◆ GetIVector() [1/2]

◆ GetIVector() [2/2]

◆ GetJ()

◆ GetJVector() [1/2]

◆ GetJVector() [2/2]

◆ GetNumColsLocal()

◆ GetNumNonzeros()

◆ GetNumRowsLocal()

◆ Helmholtz()

◆ Laplacian()

◆ MoveData()

◆ operator=() [1/2]

◆ operator=() [2/2]

◆ PlotPatternGnuPlot()

◆ PushBack()

◆ ReadFromMMFile()

◆ Scale()

◆ SetNumCols()

◆ SetNumNonzeros()

◆ Setup() [1/3]

◆ Setup() [2/3]

◆ Setup() [3/3]

◆ ToCsr()

template<typename T>
class pargemslr::CooMatrixClass< T >