Creates matrices by computing low rank approximations of dense sub matrices, e.g. via ACA or SVD.

#include <TMatBuilder.hh>

Inheritance diagram for TDenseMatBuilder< T >:

Classes
struct	stat_t
	statistical data of computations More...

Public Member Functions
	TDenseMatBuilder (const coeff_fn_t coeff_fn, const TLowRankApx lr_apx, const bool recompress=CFG::Arith::recompress, const bool cvt_to_dense=CFG::Arith::convert_to_dense, stat_t *stat=nullptr)

void	set_recompression (const bool b)
	(de-) activate recompression of low-rank blocks

void	set_stat (stat_t *stat)
	assign statistics object

virtual std::unique_ptr< TMatrix >	build_leaf (const TBlockCluster *bc, const matform_t matformat, const TTruncAcc &acc) const
	construct matrices for leaves in the block cluster tree

virtual std::unique_ptr< TMatrix >	build_ghost (const TBlockCluster *bc) const
	build placeholder matrix for remote blocks

virtual matform_t	matrix_format () const
	return matrix format

Public Member Functions inherited from TMatBuilder
	TMatBuilder (const bool coarsening=CFG::Arith::coarsen_build)

void	set_coarsening (const bool b)
	(de-) activate coarsening with standard accuracy (default: off)

void	set_coarsening (const TTruncAcc &acc)
	activate coarsening with accuracy acc

void	set_build_ghosts (const bool b)

virtual std::unique_ptr< TMatrix >	build (const TBlockClusterTree bct, const TTruncAcc &acc, TProgressBar progress=nullptr) const
	build the H-matrix with block-wise accuracy defined by acc

virtual std::unique_ptr< TMatrix >	build (const TBlockCluster bc, const TTruncAcc &acc, TProgressBar progress=nullptr) const

virtual std::unique_ptr< TMatrix >	build (const TBlockClusterTree cluster, const matform_t matformat, const TTruncAcc &acc, TProgressBar progress=nullptr) const
	same as More...

virtual std::unique_ptr< TMatrix >	build (const TBlockCluster cluster, const matform_t matformat, const TTruncAcc &acc, TProgressBar progress=nullptr) const
	same as More...

virtual std::unique_ptr< TBlockMatrix >	build_blocked (const TBlockCluster *bc) const
	build blocked matrix

const stat_t &	statistics () const
	return statistics information

Protected Member Functions
std::unique_ptr< TMatrix >	build_dense (const TBlockCluster *bc, const matform_t matformat, const TTruncAcc &acc) const

std::unique_ptr< TMatrix >	build_rank (const TBlockCluster *bc, const TTruncAcc &acc) const

Protected Member Functions inherited from TMatBuilder
virtual std::unique_ptr< TMatrix >	thr_build (const TBlockCluster bc, const matform_t matformat, const TTruncAcc &acc, TProgressBar progress) const
	threaded building process

Protected Attributes
const coeff_fn_t *	_coeff_fn
	function for computing matrix entries

const TLowRankApx *	_lr_apx
	low-rank approximation algorithm

bool	_recompress
	if true, use recompression for low-rank blocks

const bool	_cvt_to_dense
	if true, convert low-rank to dense if rank too high

stat_t *	_stat
	statistics object

Constructor & Destructor Documentation

TDenseMatBuilder	(	const coeff_fn_t *	coeff_fn,
		const TLowRankApx *	lr_apx,
		const bool	recompress = `CFG::Arith::recompress`,
		const bool	cvt_to_dense = `CFG::Arith::convert_to_dense`,
		stat_t *	stat = `nullptr`
	)

Construct matrix construction object with given coefficient function and low-rank approximation object.

Member Function Documentation

std::unique_ptr< TMatrix > build_dense	(	const TBlockCluster *	bc,
		const matform_t	matformat,
		const TTruncAcc &	acc
	)		const

protected

build dense matrix for given block cluster by directly evaluating the matrix coefficient via the coefficient function

std::unique_ptr< TMatrix > build_rank	(	const TBlockCluster *	bc,
		const TTruncAcc &	acc
	)		const

protected

build low-rank matrix for given block cluster by using internal low-rank approximation object

Classes

Public Member Functions

Protected Member Functions

Protected Attributes

Constructor & Destructor Documentation

Member Function Documentation