Class of parallel CSR matrices. More...

#include <parallel_csr_matrix.hpp>

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

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

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

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

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

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

int	Setup (int nrow_local, int ncol_local, parallel_log &parlog)
	Free the current parallel matrix, and set up new paralle matrix. More...

int	Setup (int nrow_local, long int nrow_start, long int nrow_global, int ncol_local, long int ncol_start, long int ncol_global, parallel_log &parlog)
	Free the current parallel matrix, and set up new paralle matrix. The default nnz is 0. More...

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

long int	GetNumRowsGlobal () const
	Get the global number of rows of the matrix. More...

long int	GetNumColsGlobal () const
	Get the global number of columns of the matrix. More...

long int	GetRowStartGlobal () const
	Get the global number of the first row of the matrix. More...

long int	GetColStartGlobal () const
	Get the global number of the first column of the matrix. More...

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

int	GetGraphArrays (vector_long &vtxdist, vector_long &xadj, vector_long &adjncy)
	Get the vector arrays for the graph, used as inputs for ParMETIS. Diagonal entry is removed. More...

CsrMatrixClass< T > &	GetDiagMat ()
	Get the diagonal matrix. More...

CsrMatrixClass< T > &	GetOffdMat ()
	Get the off-diagonal matrix. More...

vector_long &	GetOffdMap ()
	Get the off-diagonal column map vector. More...

bool &	IsOffdMapSorted ()
	Tell if the off-diagonal map vector is sorted. More...

int	SetOffdMatSorted (bool is_offd_map_sorted)
	Set the off-diagonal map vector to sorted/unsorted. More...

int	SortOffdMap ()
	Sort the off-diagonal map vector. More...

int	SetupVectorPtrStr (ParallelVectorClass< T > &vec)
	Update the structure of a parallel vector to have same distribution as the row distribution of this matrix. More...

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

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

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

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

int	GetDiagScale (T &scale)
	Get the value leads to max diagonal no. More...

int	Transpose ()
	Transpost the current matrix. More...

int	Transpose (ParallelCsrMatrixClass< T > &AT)
	Transpost the current matrix. More...

int	SetupMatvec ()
	Setup the matvec_helper for parallel matvec. More...

int	SetupMatvecStart ()
	Setup the matvec_helper for parallel matvec. More...

int	SetupMatvecOver ()
	Finishing up the setup phase of matvec_helper for parallel matvec. More...

virtual int	MatVec (char trans, const T &alpha, const VectorClass< T > &x, const T &beta, VectorClass< T > &y)
	In place parallel csr 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. Note that z should not equal to y. More...

int	MatVecOffd (char trans, const T &alpha, const VectorClass< T > &x, const T &beta, VectorClass< T > &y)
	In place parallel csr Matrix-Vector product with offdiagonal block only ==> y := alphaAx + betay, or y := alphaA'x + betay. More...

int	MatMat (const T &alpha, const ParallelCsrMatrixClass< T > &A, char transa, const ParallelCsrMatrixClass< T > &B, char transb, const T &beta)
	Dense Matrix-Matrix multiplication, C = alphatransa(A)transb(B)+beta*C where C is this matrix. More...

int	SubMatrix (vector_long &rows, vector_long &cols, int location, ParallelCsrMatrixClass< T > &parcsrmat_out)
	Extract parallel submatrix from parallel CSR matrix. More...

int	Laplacian (int nx, int ny, int nz, int dx, int dy, int dz, T alphax, T alphay, T alphaz, T shift, parallel_log &parlog, bool rand_perturb=false)
	Generate Laplacian matrix on the host memory, 5-pt for 2D problem and 7-pt for 3D problem. More...

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

int	LaplacianWithPartition (int nx, int ny, int nz, int dx, int dy, int dz, int d2x, int d2y, int d2z, T alphax, T alphay, T alphaz, T shift, parallel_log &parlog, bool rand_perturb=false)
	Generate Laplacian matrix on the host memory, 5-pt for 2D problem and 7-pt for 3D problem. More...

int	HelmholtzWithPartition (int n, int dx, int dy, int dz, int d2x, int d2y, int d2z, T w, parallel_log &parlog)
	Generate 3D Helmholtz matrix with 7-pt on [0,1]^3. -\Delta u - w^2 u = f. More...

int	ReadFromSingleMMFile (const char *matfile, int idxin, parallel_log &parlog)
	Read matrix on the host memory, matrix market format. More...

int	PlotPatternGnuPlot (const char *datafilename, int pttype=0)
	Plot the pattern of the parallel csr matrix using gnuplot. Similar to spy in the MATLAB. Function for testing purpose. More...

template<typename T1 >
int	SetComplexShift (T1 diagonal_shift)
	Set the diagonal complex shift for some preconditioner options. More...

template<typename T1 >
int	GetComplexShift (T1 &diagonal_shift)
	Get the diagonal complex shift for some preconditioner options. More...

int &	GetSeparatorNumSubdomains ()
	Get number of subdomains on the top level if setted in advange. More...

vector_int &	GetSeparatorDomi ()
	Number of subdomain for each node, -1 is the separator. More...

Public Attributes
std::conditional< PargemslrIsDoublePrecision< T >::value, double, float >::type	_diagonal_shift
	The diagonal complex shift in some preconditioners. The preconditioner would be built on A+_diagonal_shift*I. More...

CommunicationHelperClass	_comm_helper
	The data struxture for the parallel matvec. Stores communication inforamtion, and working buffers. More...

CommunicationHelperClass	_trans_comm_helper
	The data struxture for the parallel matvec of A transpose. Stores communication inforamtion, and working buffers. More...

Detailed Description

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

Class of parallel CSR matrices.

Constructor & Destructor Documentation

◆ ParallelCsrMatrixClass() [1/3]

template<typename T >

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

The constructor of ParallelCsrMatrixClass. The default memory location is the host memory.

◆ ParallelCsrMatrixClass() [2/3]

template<typename T >

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

The copy constructor of ParallelCsrMatrixClass.

Parameters

[in] mat The target matrix.

◆ ParallelCsrMatrixClass() [3/3]

template<typename T >

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

The move constructor of ParallelCsrMatrixClass.

Parameters

[in] mat The target matrix.

◆ ~ParallelCsrMatrixClass()

template<typename T >

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

virtual

The destructor of ParallelCsrMatrixClass. Simply a call to the free function.

Member Function Documentation

◆ Clear()

template<typename T >

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

virtual

Free the current matrix.

Returns: Return error message.

◆ Eye()

template<typename T >

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

virtual

Create an indentity matrix. Should setup the diag_mat and offd_mat first.

Returns: Return error message.

◆ Fill()

template<typename T >

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

virtual

Fill the matrix with constant value.

Parameters

[in] v The value to be filled.

Returns: Return error message.

◆ GetColStartGlobal()

template<typename T >

template long int pargemslr::ParallelCsrMatrixClass< T >::GetColStartGlobal ( ) const

Get the global number of the first column of the matrix.

Returns: Return the global number of the first column of the matrix.

◆ GetComplexShift()

template<typename T >

template<typename T1 >

int pargemslr::ParallelCsrMatrixClass< T >::GetComplexShift ( T1 & diagonal_shift )

inline

Get the diagonal complex shift for some preconditioner options.

Parameters

[in] diagonal_shift The complex shift will be 0+shifti.

Returns: Return error message.

◆ GetDataLocation()

template<typename T >

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

virtual

Get the data location of the matrix.

Returns: Return the locatino of the matrix.

◆ GetDiagMat()

template<typename T >

template CsrMatrixClass< complexd > & pargemslr::ParallelCsrMatrixClass< T >::GetDiagMat ( )

Get the diagonal matrix.

Returns: Return the diagonal matrix.

◆ GetDiagScale()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::GetDiagScale ( T & scale )

Scale the current csr matrix.

Parameters

[in] scale The scale.

Returns: Return error message.

◆ GetGraphArrays()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::GetGraphArrays	(	vector_long &	vtxdist,
		vector_long &	xadj,
		vector_long &	adjncy
	)

Get the vector arrays for the graph, used as inputs for ParMETIS. Diagonal entry is removed.

Parameters

[in]	vtxdist	The vertex distribution of size np+1.
[in]	xadj	The local I in CSR format.
[in]	adjncy	The local J in CSR format.

Returns: Return error message.

◆ GetNumColsGlobal()

template<typename T >

template long int pargemslr::ParallelCsrMatrixClass< T >::GetNumColsGlobal ( ) const

Get the global number of columns of the matrix.

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

◆ GetNumColsLocal()

template<typename T >

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

virtual

Get the local number of columns of the matrix.

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

◆ GetNumNonzeros()

template<typename T >

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

virtual

Get the number of nonzeros in this matrix.

Returns: Return the number of nonzeros of the matrix.

◆ GetNumRowsGlobal()

template<typename T >

template long int pargemslr::ParallelCsrMatrixClass< T >::GetNumRowsGlobal ( ) const

Get the global number of rows of the matrix.

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

◆ GetNumRowsLocal()

template<typename T >

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

virtual

Get the local number of rows of the matrix.

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

◆ GetOffdMap()

template<typename T >

template vector_long & pargemslr::ParallelCsrMatrixClass< T >::GetOffdMap ( )

Get the off-diagonal column map vector.

Returns: Return the off-diagonal column map vector.

◆ GetOffdMat()

template<typename T >

template CsrMatrixClass< complexd > & pargemslr::ParallelCsrMatrixClass< T >::GetOffdMat ( )

Get the off-diagonal matrix.

Returns: Return the off-diagonal matrix.

◆ GetRowStartGlobal()

template<typename T >

template long int pargemslr::ParallelCsrMatrixClass< T >::GetRowStartGlobal ( ) const

Get the global number of the first row of the matrix.

Returns: Return the global number of the first row of the matrix.

◆ GetSeparatorDomi()

template<typename T >

vector_int& pargemslr::ParallelCsrMatrixClass< T >::GetSeparatorDomi ( )

inline

Number of subdomain for each node, -1 is the separator.

Returns: Return the number of subdomain for each node, -1 is the separator.

◆ GetSeparatorNumSubdomains()

template<typename T >

int& pargemslr::ParallelCsrMatrixClass< T >::GetSeparatorNumSubdomains ( )

inline

Get number of subdomains on the top level if setted in advange.

Returns: Return the number of subdomains.

◆ Helmholtz()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::Helmholtz	(	int	n,
		int	dx,
		int	dy,
		int	dz,
		T	w,
		parallel_log &	parlog
	)

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 on each dimension
[in]	dx
[in]	dy
[in]	dz	the decomposition of the problem on the z dimension. Should have np == dxdydz.
[in]	w	The w value.
[in]	parlog	The Parallel log data structure, if parallel_log._comm == NULL, will use the global one.

Returns: Return error message.

◆ HelmholtzWithPartition()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::HelmholtzWithPartition	(	int	n,
		int	dx,
		int	dy,
		int	dz,
		int	d2x,
		int	d2y,
		int	d2z,
		T	w,
		parallel_log &	parlog
	)

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 on each dimension
[in]	dx
[in]	dy
[in]	dz	the decomposition of the problem on each dimension. Should have np == dxdydz.
[in]	d2x
[in]	d2y
[in]	d2z	the decomposition of the problem on each dimension. Should have d2xd2yd2z % np == 0.
[in]	w	The w value.
[in]	parlog	The Parallel log data structure, if parallel_log._comm == NULL, will use the global one.

Returns: Return error message.

◆ IsOffdMapSorted()

template<typename T >

template bool & pargemslr::ParallelCsrMatrixClass< T >::IsOffdMapSorted ( )

Tell if the off-diagonal map vector is sorted.

Returns: Return true if off-diagonal map is sorted.

◆ Laplacian()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::Laplacian	(	int	nx,
		int	ny,
		int	nz,
		int	dx,
		int	dy,
		int	dz,
		T	alphax,
		T	alphay,
		T	alphaz,
		T	shift,
		parallel_log &	parlog,
		bool	rand_perturb = `false`
	)

Generate Laplacian matrix on the host memory, 5-pt for 2D problem and 7-pt for 3D problem.

Parameters

[in]	nx
[in]	ny
[in]	nz	Problem size on each dimension
[in]	dx
[in]	dy
[in]	dz	the decomposition of the problem on each dimension. Should have np == dxdydz.
[in]	alphax
[in]	alphay
[in]	alphaz	Alpha value on each direction
[in]	shift	The shift value on diagonal
[in]	parlog	The Parallel log data structure, if parallel_log._comm == NULL, will use the global one
[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.

◆ LaplacianWithPartition()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::LaplacianWithPartition	(	int	nx,
		int	ny,
		int	nz,
		int	dx,
		int	dy,
		int	dz,
		int	d2x,
		int	d2y,
		int	d2z,
		T	alphax,
		T	alphay,
		T	alphaz,
		T	shift,
		parallel_log &	parlog,
		bool	rand_perturb = `false`
	)

Generate Laplacian matrix on the host memory, 5-pt for 2D problem and 7-pt for 3D problem.
Meth partition is created in this case based on infor in d2x/y/z.

Parameters

[in]	nx
[in]	ny
[in]	nz	Problem size on each dimension
[in]	dx
[in]	dy
[in]	dz	the decomposition of the processors on each dimension. Should have dxdydz == np.
[in]	d2x
[in]	d2y
[in]	d2z	the decomposition of the problem on each dimension. Should have d2xd2yd2z % np == 0.
[in]	alphax
[in]	alphay
[in]	alphaz	Alpha value on each direction
[in]	shift	The shift value on diagonal
[in]	parlog	The Parallel log data structure, if parallel_log._comm == NULL, will use the global one.
[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.

◆ MatMat()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::MatMat	(	const T &	alpha,
		const ParallelCsrMatrixClass< T > &	A,
		char	transa,
		const ParallelCsrMatrixClass< T > &	B,
		char	transb,
		const T &	beta
	)

Dense Matrix-Matrix multiplication, C = alpha*transa(A)*transb(B)+beta*C where C is this matrix.
If C is not pre-allocated, C will be allocated such that it has the same location as A.

Note: TODO: currently not supported.

Parameters

[in]	alpha	The alpha value.
[in]	A	The first csr matrix.
[in]	transa	Whether we transpose A or not.
[in]	B	The second csr matrix.
[in]	transb	Whether we transpose B or not.
[in]	beta	The beta value.

Returns: Return error message.

◆ MatVec() [1/2]

template<typename T >

int pargemslr::ParallelCsrMatrixClass< T >::MatVec	(	char	trans,
		const T &	alpha,
		const VectorClass< T > &	x,
		const T &	beta,
		const VectorClass< T > &	y,
		VectorClass< T > &	z
	)

virtual

In place Matrix-Vector product ==> z := alpha*A*x + beta*y, or z := alpha*A'*x + beta*y. Note that z should not equal to y.

Note: TODO: Currently A' is not supported.

Parameters

[in]	trans	Whether or not transpose matrix A.
[in]	alpha	The alpha value.
[in]	x	The first vector.
[in]	beta	The beta value.
[in]	y	The second vector.
[out]	z	The output vector.

Returns: Return error message.

◆ MatVec() [2/2]

template<typename T >

int pargemslr::ParallelCsrMatrixClass< T >::MatVec	(	char	trans,
		const T &	alpha,
		const VectorClass< T > &	x,
		const T &	beta,
		VectorClass< T > &	y
	)

virtual

In place parallel csr Matrix-Vector product ==> y := alpha*A*x + beta*y, or y := alpha*A'*x + beta*y.

Note: TODO: Currently A' is not supported.

Parameters

[in]	trans	Whether or not transpose matrix A.
[in]	alpha	The alpha value.
[in]	x	The left vector.
[in]	beta	The beta value.
[in,out]	y	The product vector.

Returns: Return error message.

◆ MatVecOffd()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::MatVecOffd	(	char	trans,
		const T &	alpha,
		const VectorClass< T > &	x,
		const T &	beta,
		VectorClass< T > &	y
	)

In place parallel csr Matrix-Vector product with offdiagonal block only ==> y := alpha*A*x + beta*y, or y := alpha*A'*x + beta*y.

Parameters

[in]	trans	Whether or not transpose matrix A.
[in]	alpha	The alpha value.
[in]	x	The left vector.
[in]	beta	The beta value.
[in,out]	y	The product vector.

Returns: Return error message.

◆ MoveData()

template<typename T >

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

virtual

Move the data to another memory location.

Parameters

[in] location The location move to.

Returns: Return error message.

◆ operator=() [1/2]

template<typename T >

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

The = operator of ParallelCsrMatrixClass.

Parameters

[in] mat The target matrix.

Returns: Return the matrix.

◆ operator=() [2/2]

template<typename T >

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

The = operator of ParallelCsrMatrixClass.

Parameters

[in] mat The target matrix.

Returns: Return the matrix.

◆ PlotPatternGnuPlot()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::PlotPatternGnuPlot	(	const char *	datafilename,
		int	pttype = `0`
	)

Plot the pattern of the parallel 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.

Returns: Return error message.

◆ ReadFromSingleMMFile()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::ReadFromSingleMMFile	(	const char *	matfile,
		int	idxin,
		parallel_log &	parlog
	)

Read matrix on the host memory, matrix market format.

Note: The current version doesn't support long int as input.

Parameters

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

Returns: Return error message.

◆ Scale()

template<typename T >

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

virtual

Scale the current csr matrix.

Parameters

[in] alpha The scale.

Returns: Return error message.

◆ SetComplexShift()

template<typename T >

template<typename T1 >

int pargemslr::ParallelCsrMatrixClass< T >::SetComplexShift ( T1 diagonal_shift )

inline

Set the diagonal complex shift for some preconditioner options.

Parameters

[in] diagonal_shift The complex shift will be 0+shifti.

Returns: Return error message.

◆ SetOffdMatSorted()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::SetOffdMatSorted ( bool is_offd_map_sorted )

Set the off-diagonal map vector to sorted/unsorted.

Returns: Return error message.

◆ Setup() [1/2]

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::Setup	(	int	nrow_local,
		int	ncol_local,
		parallel_log &	parlog
	)

Free the current parallel matrix, and set up new paralle matrix.

Note: The diag_mat and offd_mat need to be initialized separately.

Parameters

[in]	nrow_local	The local number of rows.
[in]	ncol_local	The local number of cols.
[in]	parlog	The Parallel log data structure, if parallel_log._comm == NULL, will use the global one.

Returns: Return error message.

◆ Setup() [2/2]

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::Setup	(	int	nrow_local,
		long int	nrow_start,
		long int	nrow_global,
		int	ncol_local,
		long int	ncol_start,
		long int	ncol_global,
		parallel_log &	parlog
	)

Free the current parallel matrix, and set up new paralle matrix. The default nnz is 0.

Note: The diag_mat and offd_mat need to be initialized separately.

Parameters

[in]	nrow_local	The local number of rows.
[in]	nrow_start	The start index of the rows.
[in]	nrow_global	The global length of the rows.
[in]	ncol_local	The local number of cols.
[in]	ncol_start	The start index of the columns.
[in]	ncol_global	The global length of the columns.
[in]	parlog	The Parallel log data structure, if parallel_log._comm == NULL, will use the global one.

Returns: Return error message.

◆ SetupMatvec()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::SetupMatvec ( )

Setup the matvec_helper for parallel matvec.

Returns: Return error message.

◆ SetupMatvecOver()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::SetupMatvecOver ( )

Finishing up the setup phase of matvec_helper for parallel matvec.

Returns: Return error message.

◆ SetupMatvecStart()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::SetupMatvecStart ( )

Setup the matvec_helper for parallel matvec.
Non-blocking communication, need to call SetupMatvecOver function to finish up.

Returns: Return error message.

◆ SetupVectorPtrStr()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::SetupVectorPtrStr ( ParallelVectorClass< T > & vec )

Update the structure of a parallel vector to have same distribution as the row distribution of this matrix.

Parameters

[in,out] The target vector.

Returns: Return error message.

◆ SortOffdMap()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::SortOffdMap ( )

Sort the off-diagonal map vector.

Returns: Return error message.

◆ SubMatrix()

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::SubMatrix	(	vector_long &	rows,
		vector_long &	cols,
		int	location,
		ParallelCsrMatrixClass< T > &	parcsrmat_out
	)

Extract parallel submatrix from parallel CSR matrix.
Each processor prvide several local rows/cols to keep (in global indices). No exchange of global information.

Parameters

[in]	rows	The rows to keep. The permutation in rows also applied.
[in]	cols	The columns to keep. The permutation in cols also applied.
[in]	location	The location of the output matrix.
[out]	parcsrmat_out	The output matrix.

Returns: Return error message.

◆ Transpose() [1/2]

template<typename T >

template int pargemslr::ParallelCsrMatrixClass< T >::Transpose ( )

Transpost the current matrix.

Returns: Return error message.

◆ Transpose() [2/2]

template<typename T >

int pargemslr::ParallelCsrMatrixClass< T >::Transpose ( ParallelCsrMatrixClass< T > & AT )

Transpost the current matrix.

Parameters

[out] AT The output matrix.

Returns: Return error message.

Member Data Documentation

◆ _comm_helper

template<typename T >

CommunicationHelperClass pargemslr::ParallelCsrMatrixClass< T >::_comm_helper

The data struxture for the parallel matvec. Stores communication inforamtion, and working buffers.

◆ _diagonal_shift

template<typename T >

std::conditional<PargemslrIsDoublePrecision<T>::value, double, float>::type pargemslr::ParallelCsrMatrixClass< T >::_diagonal_shift

The diagonal complex shift in some preconditioners. The preconditioner would be built on A+_diagonal_shift*I.

Note: TODO: if useful, change to private.

◆ _trans_comm_helper

template<typename T >

CommunicationHelperClass pargemslr::ParallelCsrMatrixClass< T >::_trans_comm_helper

The data struxture for the parallel matvec of A transpose. Stores communication inforamtion, and working buffers.

Note: Not yet in use.

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

SRC/utils/structs.hpp
SRC/matrices/parallel_csr_matrix.hpp
SRC/matrices/parallel_csr_matrix.cpp

Public Member Functions

Public Attributes

Detailed Description

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

Constructor & Destructor Documentation

◆ ParallelCsrMatrixClass() [1/3]

◆ ParallelCsrMatrixClass() [2/3]

◆ ParallelCsrMatrixClass() [3/3]

◆ ~ParallelCsrMatrixClass()

Member Function Documentation

◆ Clear()

◆ Eye()

◆ Fill()

◆ GetColStartGlobal()

◆ GetComplexShift()

◆ GetDataLocation()

◆ GetDiagMat()

◆ GetDiagScale()

◆ GetGraphArrays()

◆ GetNumColsGlobal()

◆ GetNumColsLocal()

◆ GetNumNonzeros()

◆ GetNumRowsGlobal()

◆ GetNumRowsLocal()

◆ GetOffdMap()

◆ GetOffdMat()

◆ GetRowStartGlobal()

◆ GetSeparatorDomi()

◆ GetSeparatorNumSubdomains()

◆ Helmholtz()

◆ HelmholtzWithPartition()

◆ IsOffdMapSorted()

◆ Laplacian()

◆ LaplacianWithPartition()

◆ MatMat()

◆ MatVec() [1/2]

◆ MatVec() [2/2]

◆ MatVecOffd()

◆ MoveData()

◆ operator=() [1/2]

◆ operator=() [2/2]

◆ PlotPatternGnuPlot()

◆ ReadFromSingleMMFile()

◆ Scale()

◆ SetComplexShift()

◆ SetOffdMatSorted()

◆ Setup() [1/2]

◆ Setup() [2/2]

◆ SetupMatvec()

◆ SetupMatvecOver()

◆ SetupMatvecStart()

◆ SetupVectorPtrStr()

◆ SortOffdMap()

◆ SubMatrix()

◆ Transpose() [1/2]

◆ Transpose() [2/2]

Member Data Documentation

◆ _comm_helper

◆ _diagonal_shift

◆ _trans_comm_helper

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