Collection of matrix-matrix operations. More...

#include <EpetraExt_MatrixMatrix.h>

Public Member Functions
virtual	~MatrixMatrix ()
	destructor More...

Static Public Member Functions
static int	Multiply (const Epetra_CrsMatrix &A, bool transposeA, const Epetra_CrsMatrix &B, bool transposeB, Epetra_CrsMatrix &C, bool call_FillComplete_on_result=true, bool keep_all_hard_zeros=false)
	Given Epetra_CrsMatrix objects A, B and C, form the product C = A*B. More...

static int	Add (const Epetra_CrsMatrix &A, bool transposeA, double scalarA, Epetra_CrsMatrix &B, double scalarB)
	Given Epetra_CrsMatrix objects A and B, form the sum B = aA + bB. More...

static int	Add (const Epetra_CrsMatrix &A, bool transposeA, double scalarA, const Epetra_CrsMatrix &B, bool transposeB, double scalarB, Epetra_CrsMatrix *&C)
	Given Epetra_CrsMatrix objects A and B, form the sum C = aA + bB. More...

static int	Jacobi (double omega, const Epetra_Vector &Dinv, const Epetra_CrsMatrix &A, const Epetra_CrsMatrix &B, Epetra_CrsMatrix &C, bool call_FillComplete_on_result=true)
	Given Epetra_CrsMatrix objects A, B and C, and Epetra_Vector Dinv, form the product C = (I-omega * Dinv A)*B In a parallel setting, A and B need not have matching distributions, but C needs to have the same row-map as A. More...

Detailed Description

Collection of matrix-matrix operations.

This class basically functions as a namespace, containing only static methods. See the program epetraext/test/MatrixMatrix/cxx_main.cpp for a usage example.

Definition at line 69 of file EpetraExt_MatrixMatrix.h.

Constructor & Destructor Documentation

virtual EpetraExt::MatrixMatrix::~MatrixMatrix ( )

inlinevirtual

destructor

Definition at line 73 of file EpetraExt_MatrixMatrix.h.

Member Function Documentation

int EpetraExt::MatrixMatrix::Multiply	(	const Epetra_CrsMatrix &	A,
		bool	transposeA,
		const Epetra_CrsMatrix &	B,
		bool	transposeB,
		Epetra_CrsMatrix &	C,
		bool	call_FillComplete_on_result = `true`,
		bool	keep_all_hard_zeros = `false`
	)

static

Given Epetra_CrsMatrix objects A, B and C, form the product C = A*B.

In a parallel setting, A and B need not have matching distributions, but C needs to have the same row-map as A.

Parameters

A	Input, must already have had 'FillComplete()' called.
transposeA	Input, whether to use transpose of matrix A.
B	Input, must already have had 'FillComplete()' called.
transposeB	Input, whether to use transpose of matrix B.
C	Result. On entry to this method, it doesn't matter whether FillComplete() has already been called on C or not. If it has, then C's graph must already contain all nonzero locations that will be produced when forming the product A*B. On exit, C.FillComplete() will have been called, unless the last argument to this function is specified to be false.
call_FillComplete_on_result	Optional argument, defaults to true. Power users may specify this argument to be false if they DON'T want this function to call C.FillComplete. (It is often useful to allow this function to call C.FillComplete, in cases where one or both of the input matrices are rectangular and it is not trivial to know which maps to use for the domain- and range-maps.)
keep_all_hard_zeros	Optional argument, defaults to false. If true, Multiply, keeps all entries in C corresponding to hard zeros. If false, the following happens by case: AB^T, A^TB^T - Does not store entries caused by hard zeros in C. A^TB (unoptimized) - Hard zeros are always stored (this option has no effect) AB, A^T*B (optimized) - Hard zeros in corresponding to hard zeros in A are not stored, There are certain cases involving reuse of C, where this can be useful.

Returns: error-code, 0 if successful. non-zero returns may result if A or B are not already Filled, or if errors occur in putting values into C, etc.

Definition at line 1304 of file EpetraExt_MatrixMatrix.cpp.

int EpetraExt::MatrixMatrix::Add	(	const Epetra_CrsMatrix &	A,
		bool	transposeA,
		double	scalarA,
		Epetra_CrsMatrix &	B,
		double	scalarB
	)

static

Given Epetra_CrsMatrix objects A and B, form the sum B = a*A + b*B.

Parameters

A	Input, must already have had 'FillComplete()' called.
transposeA	Input, whether to use transpose of matrix A.
scalarA	Input, scalar multiplier for matrix A.
B	Result. On entry to this method, it doesn't matter whether FillComplete() has already been called on B or not. If it has, then B's graph must already contain all nonzero locations that will be produced when forming the sum.
scalarB	Input, scalar multiplier for matrix B.

Returns: error-code, 0 if successful. non-zero returns may result if A is not already Filled, or if errors occur in putting values into B, etc.

Definition at line 1423 of file EpetraExt_MatrixMatrix.cpp.

int EpetraExt::MatrixMatrix::Add	(	const Epetra_CrsMatrix &	A,
		bool	transposeA,
		double	scalarA,
		const Epetra_CrsMatrix &	B,
		bool	transposeB,
		double	scalarB,
		Epetra_CrsMatrix *&	C
	)

static

Given Epetra_CrsMatrix objects A and B, form the sum C = a*A + b*B.

Parameters

A	Input, must already have had 'FillComplete()' called.
transposeA	Input, whether to use transpose of matrix A.
scalarA	Input, scalar multiplier for matrix A.
B	Input, must already have had 'FillComplete()' called.
transposeB	Input, whether to use transpose of matrix B.
scalarB	Input, scalar multiplier for matrix B.
C	Result. On entry to this method, C can be NULL or a pointer to an unfilled or filled matrix. If C is NULL then a new object is allocated and must be deleted by the user. If C is not NULL and FillComplete has already been called then the sparsity pattern is assumed to be fixed and compatible with the sparsity of A+B. If FillComplete has not been called then the sum is completed and the function returns without calling FillComplete on C.

Returns: error-code, 0 if successful. non-zero returns may result if A or is not already Filled, or if errors occur in putting values into C, etc.

Definition at line 1532 of file EpetraExt_MatrixMatrix.cpp.

int EpetraExt::MatrixMatrix::Jacobi	(	double	omega,
		const Epetra_Vector &	Dinv,
		const Epetra_CrsMatrix &	A,
		const Epetra_CrsMatrix &	B,
		Epetra_CrsMatrix &	C,
		bool	call_FillComplete_on_result = `true`
	)

static

Given Epetra_CrsMatrix objects A, B and C, and Epetra_Vector Dinv, form the product C = (I-omega * Dinv A)*B

In a parallel setting, A and B need not have matching distributions, but C needs to have the same row-map as A.

Parameters

omega	Input, scalar multiplier for Dinverse A
Dinv	Input, Epetra_Vector representing a diagonal matrix, must match A's RowMap
A	Input, must already have had 'FillComplete()' called.
B	Input, must already have had 'FillComplete()' called.
C	Result. On entry to this method, it doesn't matter whether FillComplete() has already been called on C or not. If it has, then C's graph must already contain all nonzero locations that will be produced when forming the product A*B. On exit, C.FillComplete() will have been called, unless the last argument to this function is specified to be false.
call_FillComplete_on_result	Optional argument, defaults to true. Power users may specify this argument to be false if they DON'T want this function to call C.FillComplete. (It is often useful to allow this function to call C.FillComplete, in cases where one or both of the input matrices are rectangular and it is not trivial to know which maps to use for the domain- and range-maps.)

Returns: error-code, 0 if successful. non-zero returns may result if A or B are not already Filled, or if errors occur in putting values into C, etc.

Definition at line 1695 of file EpetraExt_MatrixMatrix.cpp.

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

Public Member Functions

Static Public Member Functions

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation