gridpack::math::MatrixT< T, I > Class Template Reference

A parallel or serial matrix of real values. More...

#include <matrix.hpp>

Inheritance diagram for gridpack::math::MatrixT< T, I >:

[legend]

Collaboration diagram for gridpack::math::MatrixT< T, I >:

[legend]

List of all members.

Public Types
typedef BaseMatrixInterface< T, I >::TheType	TheType
typedef BaseMatrixInterface< T, I >::IdxType	IdxType
Public Member Functions
	MatrixT (const parallel::Communicator &dist, const int &local_rows, const int &local_cols, const MatrixStorageType &storage_type=Sparse)
	Constructor.
	MatrixT (const parallel::Communicator &dist, const int &local_rows, const int &local_cols, const int &max_nz_per_row)
	Sparse matrix constructor with maximum number of nonzeros in a row.
	MatrixT (const parallel::Communicator &dist, const int &local_rows, const int &local_cols, const int *nz_by_row)
	Sparse matrix constructor with number of nonzeros for each row.
	MatrixT (MatrixImplementation< T, I > *impl)
	Construct with an existing (allocated) implementation.
	~MatrixT (void)
	Destructor.
MatrixStorageType	storageType (void) const
	Get the storage type of this matrix.
MatrixT *	clone (void) const
MatrixT *	localClone (void) const
void	equate (const MatrixT &A)
	Get a part of the matrix (new instance allocated).
void	multiplyDiagonal (const VectorT< T, I > &x)
	Multiply this matrix diagonal by the specified vector.
void	addDiagonalVector (const VectorT< T, I > &x)
	Add the specified vector to the diagonal of this matrix.
void	add (const MatrixT &A)
	Add another matrix to this one, in place.
Static Public Member Functions
static MatrixT *	createDense (const parallel::Communicator &comm, const int &global_rows, const int &global_cols, const int &local_rows, const int &local_cols)
	Create a Dense Matrix instance with more/less specific ownership.

---

Detailed Description

template<typename T, typename I = int>
class gridpack::math::MatrixT< T, I >

A parallel or serial matrix of real values.

This class uses the Pimpl idiom for implementation in order so the interface is completely free of the underlying library. If constructed with a parallel environment with only one process, a serial storage scheme is created, otherwise it's parallel.

---

Member Typedef Documentation

template<typename T, typename I = int>

typedef BaseMatrixInterface<T, I>::IdxType gridpack::math::MatrixT< T, I >::IdxType

The size/index type used.

Reimplemented from gridpack::math::BaseMatrixInterface< T, I >.

template<typename T, typename I = int>

typedef BaseMatrixInterface<T, I>::TheType gridpack::math::MatrixT< T, I >::TheType

The numeric type used.

Reimplemented from gridpack::math::BaseMatrixInterface< T, I >.

---

Constructor & Destructor Documentation

template<typename T, typename I = int>

gridpack::math::MatrixT< T, I >::MatrixT	(	const parallel::Communicator &	dist,
		const int &	local_rows,
		const int &	local_cols,
		const MatrixStorageType &	storage_type = Sparse
	)

Constructor.

A Matrix must be instantiated simulutaneously on all processes involved in the specified communicator. Each process in the communicator will own the number of rows requested.

Parameters:

	dist	parallel environment
	local_rows	matrix rows to be owned by the local process
	local_cols	matrix columns to be owned by the local process
	storage_type	specify dense or sparse storage

Returns:

template<typename T, typename I = int>

gridpack::math::MatrixT< T, I >::MatrixT	(	const parallel::Communicator &	dist,
		const int &	local_rows,
		const int &	local_cols,
		const int &	max_nz_per_row
	)

Sparse matrix constructor with maximum number of nonzeros in a row.

If the underlying math implementation supports it, this constructs a sparse matrix and pre-allocates it to allow max_nz_per_row nonzeros in each row. If the underlying math library supports it, max_nz_per_row does not need to be the same on all processors.

Parameters:

	dist	parallel environment
	local_rows	matrix rows to be owned by the local process
	local_cols	matrix columns to be owned by the local process
	max_nz_per_row	maximum number of nonzeros in a row

Returns:

new MatrixT

template<typename T, typename I = int>

gridpack::math::MatrixT< T, I >::MatrixT	(	const parallel::Communicator &	dist,
		const int &	local_rows,
		const int &	local_cols,
		const int *	nz_by_row
	)

Sparse matrix constructor with number of nonzeros for each row.

Parameters:

	dist	parallel environment
	local_rows	matrix rows to be owned by the local process
	local_cols	matrix columns to be owned by the local process
	nz_by_row

Returns:

template<typename T, typename I = int>

gridpack::math::MatrixT< T, I >::MatrixT

(

MatrixImplementation< T, I > *

impl

)

Construct with an existing (allocated) implementation.

For internal use only.

Parameters:

impl

Returns:

References gridpack::parallel::WrappedDistributed::WrappedDistributed().

template<typename T, typename I = int>

gridpack::math::MatrixT< T, I >::~MatrixT

(

void

)

Destructor.

A matrix must be destroyed simulutaneously on all processes involved in the communicator used to instantiate it.

---

Member Function Documentation

template<typename T, typename I = int>

void gridpack::math::MatrixT< T, I >::add

(

const MatrixT< T, I > &

A

)

Add another matrix to this one, in place.

Collective.

The specified matrix A must be the same global size (rows and columns) as this instance, but local ownership is not important, nor are any differences in nonzero entry patterns.

Parameters:

A

matrix to add to this instance

template<typename T, typename I = int>

void gridpack::math::MatrixT< T, I >::addDiagonalVector

(

const VectorT< T, I > &

x

)

Add the specified vector to the diagonal of this matrix.

Collective.

Parameters:

x

template<typename T, typename I = int>

MatrixT* gridpack::math::MatrixT< T, I >::clone

(

void

)

const

Make an exact replica of this instance Collective.

Returns:

pointer to new (allocated) matrix instance

References gridpack::math::MatrixT< T, I >::MatrixT().

template<typename T, typename I = int>

static MatrixT* gridpack::math::MatrixT< T, I >::createDense	(	const parallel::Communicator &	comm,
		const int &	global_rows,
		const int &	global_cols,
		const int &	local_rows,
		const int &	local_cols
	)			[static]

Create a Dense Matrix instance with more/less specific ownership.

Parameters:

	comm
	stype
	global_rows
	global_cols
	local_rows
	local_cols

Returns:

template<typename T, typename I = int>

void gridpack::math::MatrixT< T, I >::equate

(

const MatrixT< T, I > &

A

)

Get a part of the matrix (new instance allocated).

Make this Matrix equal to another Collective.

Parameters:

A

template<typename T, typename I = int>

MatrixT* gridpack::math::MatrixT< T, I >::localClone

(

void

)

const

References gridpack::math::MatrixT< T, I >::MatrixT().

template<typename T, typename I = int>

void gridpack::math::MatrixT< T, I >::multiplyDiagonal

(

const VectorT< T, I > &

x

)

Multiply this matrix diagonal by the specified vector.

Collective.

This is element by element multiplication

Parameters:

x

factor by which all diagonal elements in the matrix are multiplied

template<typename T, typename I = int>

MatrixStorageType gridpack::math::MatrixT< T, I >::storageType

(

void

)

const

Get the storage type of this matrix.

---

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

• matrix.hpp