AffineMatrixFunction needs to compute the maximum eigenvalue of an affine matrix function $F(y)=C+\sum y_iA_i$ . This class prepares $F(y)$ in useful form for iterative eigenvalue solvers. More...

#include <Bigmatrix.hxx>

Inheritance diagram for ConicBundle::Bigmatrix:

Public Member Functions
void	clear ()
	clear all data

CH_Matrix_Classes::Integer	get_nmult () const
	return number of matrix vector products

void	reset_nmult ()
	reset the counter of matrix vector products to zero

int	init (const CH_Matrix_Classes::Matrix &yi, CH_Matrix_Classes::Integer indim, const CoeffmatPointer C, const SparseCoeffmatVector *Ai, const bool dense=false)
	compute a good representation of $F(y)=C+\sum y_iA_i$ (if dense==true directly a dense one)

void	set_tol (CH_Matrix_Classes::Real t)
	sets the tolerance for considering computed values as zeros

int	get_dense () const
	return true if a dense representation was computed

virtual CH_Matrix_Classes::Integer	lanczosdim () const
	the order of the matrix as required by CH_Matrix_Classes::Lanczosmatrix

virtual CH_Matrix_Classes::Integer	lanczosflops () const
	the estimated flop count of a matrix vector multiplication as required by CH_Matrix_Classes::Lanczosmatrix

virtual int	lanczosmult (const CH_Matrix_Classes::Matrix &A, CH_Matrix_Classes::Matrix &B) const
	computes B = bigMatrix * A (A and B must not refer to the same memory) as required by CH_Matrix_Classes::Lanczosmatrix

int	make_symmatrix (CH_Matrix_Classes::Symmatrix &S) const
	write a dense version of the current matrix to S

const CH_Matrix_Classes::Symmatrix &	get_symrep () const
	compute the dense version locally and return a reference to it

Private Types
typedef std::map< CH_Matrix_Classes::Integer, CH_Matrix_Classes::Real >	Indexhash
	for collecting the indices of sparse data this seems useful

Private Attributes
CH_Matrix_Classes::Real	tol
	tolerance for considering an element to be zero

CH_Matrix_Classes::Integer	dim
	size of the matrix

CH_Matrix_Classes::Integer	nz
	total number of nonzeros (except diagonal)

CH_Matrix_Classes::Indexmatrix	colnz
	nonzeros of each column

std::vector< CH_Matrix_Classes::Indexmatrix >	rowind
	nonzero row indices in each column

std::vector< CH_Matrix_Classes::Matrix >	rowval
	corresponding nonzero row values

CH_Matrix_Classes::Matrix	di
	diagonal elements

std::vector< Indexhash >	rowhash
	for finding nonzero indices

std::vector< CoeffmatPointer >	mcp
	pointers to dense matrices

CH_Matrix_Classes::Matrix	mcv
	multiplier values for the dense matrices

CH_Matrix_Classes::Integer	nmult
	counts the number of Mat*vec operations

bool	use_dense
	usually false, set to true if matrix is sufficiently dense

CH_Matrix_Classes::Symmatrix	symrep
	if use_dense is true, the matrix is stored here

bool	symrep_init
	flag whether symrep is computed already

CH_Matrix_Classes::Matrix	At
	auxilliary variable used in matrix vector products

CH_Matrix_Classes::Matrix	Bt
	auxilliary variable used in matrix vector products

CH_Matrix_Classes::Indexmatrix	collecti
	for collecting the sparse part

CH_Matrix_Classes::Indexmatrix	collectj
	for collecting the sparse part

CH_Matrix_Classes::Matrix	collectval
	for collecting the sparse part

Friends
std::ostream &	operator<< (std::ostream &out, const Bigmatrix &)
	output information

Additional Inherited Members
Protected Member Functions inherited from CH_Matrix_Classes::Memarrayuser
	Memarrayuser ()
	if memarray is NULL, then a new Memarray is generated. In any case the number of users of the Memarray is incremented

virtual	~Memarrayuser ()
	the number of users is decremented and the Memarray memory manager is destructed, if the number is zero.

Static Protected Attributes inherited from CH_Matrix_Classes::Memarrayuser
static Memarray *	memarray
	pointer to common memory manager for all Memarrayusers, instantiated in memarray.cxx

Detailed Description

AffineMatrixFunction needs to compute the maximum eigenvalue of an affine matrix function $F(y)=C+\sum y_iA_i$ . This class prepares $F(y)$ in useful form for iterative eigenvalue solvers.

In AffineMatrixFunction the matrices $C$ and $A_i$ are of type Coeffmat, the information provided by these matrices is used to decide which parts should be represented in a sparse form and which parts in a dense form. The intention is that in the end matrix vector products can be computed efficiently, so that e.g. a Lanczcos algorithm can be used. For this purpose Bigmatrix is a publically derived CH_Matrix_Classes::Lanczosmatrix.

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

Bigmatrix.hxx

Public Member Functions

Private Types

Private Attributes

Friends

Additional Inherited Members

Detailed Description