Class containing several options for the HSS code and data-structures. More...

#include <HSSOptions.hpp>

Inheritance diagram for strumpack::HSS::HSSOptions< scalar_t >:

Collaboration diagram for strumpack::HSS::HSSOptions< scalar_t >:

Public Types
using	real_t = typename RealType< scalar_t >::value_type

Public Member Functions
	HSSOptions (const structured::StructuredOptions< scalar_t > &sopts)

void	set_d0 (int d0)

void	set_dd (int dd)

void	set_p (int p)

void	set_nnz0 (int nnz0)

void	set_nnz (int nnz)

void	set_random_engine (random::RandomEngine random_engine)

void	set_random_distribution (random::RandomDistribution random_distribution)

void	set_compression_algorithm (CompressionAlgorithm a)

void	set_compression_sketch (CompressionSketch a)

void	set_SJLT_algo (SJLTAlgo a)

void	set_clustering_algorithm (ClusteringAlgorithm a)

void	set_approximate_neighbors (int neighbors)

void	set_ann_iterations (int iters)

void	set_user_defined_random (bool user_defined_random)

void	set_synchronized_compression (bool sync)

void	set_log_ranks (bool log_ranks)

int	d0 () const

int	nnz0 () const

int	nnz () const

int	dd () const

int	p () const

random::RandomEngine	random_engine () const

random::RandomDistribution	random_distribution () const

CompressionAlgorithm	compression_algorithm () const

CompressionSketch	compression_sketch () const

SJLTAlgo	SJLT_algo () const

ClusteringAlgorithm	clustering_algorithm () const

int	approximate_neighbors () const

int	ann_iterations () const

bool	user_defined_random () const

bool	synchronized_compression () const

bool	log_ranks () const

void	set_from_command_line (int argc, const char const cargv) override

void	describe_options () const override

Public Member Functions inherited from strumpack::structured::StructuredOptions< scalar_t >
	StructuredOptions (Type type)

void	set_rel_tol (real_t rel_tol)

void	set_abs_tol (real_t abs_tol)

void	set_leaf_size (int leaf_size)

void	set_max_rank (int max_rank)

void	set_type (Type a)

void	set_verbose (bool verbose)

real_t	rel_tol () const

real_t	abs_tol () const

int	leaf_size () const

int	max_rank () const

Type	type () const

bool	verbose () const

virtual void	set_from_command_line (int argc, const char const cargv)

virtual void	describe_options () const

Detailed Description

template<typename scalar_t>
class strumpack::HSS::HSSOptions< scalar_t >

Class containing several options for the HSS code and data-structures.

Template Parameters

scalar_t scalar type, can be float, double, std::complex<float> or std::complex<double>. This is used here mainly because tolerances might depend on the precision.

Member Typedef Documentation

◆ real_t

template<typename scalar_t >

using strumpack::HSS::HSSOptions< scalar_t >::real_t = typename RealType<scalar_t>::value_type

real_t is the real type corresponding to the (possibly complex) scalar_t template parameter

Member Function Documentation

◆ ann_iterations()

template<typename scalar_t >

int strumpack::HSS::HSSOptions< scalar_t >::ann_iterations ( ) const

inline

Get the number of iterations used in the approximate nearest neighbors search algorithm used in the HSS compression algorithm for kernel matrices. The algorithm will build iters number of random trees.

Returns: Number of random trees to be build

◆ approximate_neighbors()

template<typename scalar_t >

int strumpack::HSS::HSSOptions< scalar_t >::approximate_neighbors ( ) const

inline

Get the number of approximate nearest neighbors used in the HSS compression algorithm for kernel matrices.

Returns: Number of approximate neighbors

◆ clustering_algorithm()

template<typename scalar_t >

ClusteringAlgorithm strumpack::HSS::HSSOptions< scalar_t >::clustering_algorithm ( ) const

inline

Get the clustering algorithm to be used. This is used when constructing an HSS approximation of a kernel matrix.

Returns: clustering algorithm

See also: set_clustering_algorithm

◆ compression_algorithm()

template<typename scalar_t >

CompressionAlgorithm strumpack::HSS::HSSOptions< scalar_t >::compression_algorithm ( ) const

inline

Return which variant of the compression algorithm to use.

Returns: Variant of HSS compression algorithm

See also: set_compression_algorithm

◆ d0()

template<typename scalar_t >

int strumpack::HSS::HSSOptions< scalar_t >::d0 ( ) const

inline

Get the initial number of random vector that will be used in adaptive randomized HSS compression. See the manual for more info.

Returns: initial guess for the rank, used in adaptive compression

See also: set_d0(), set_dd(), set_p()

◆ dd()

template<typename scalar_t >

int strumpack::HSS::HSSOptions< scalar_t >::dd ( ) const

inline

Increment for the number of random vectors during adaptive compression.

Returns: amount with which the number of random vectors will be incremented

See also: set_d0(), set_dd()

◆ describe_options()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::describe_options ( ) const

overridevirtual

Print an overview of the available command line options and their current values.

Reimplemented from strumpack::structured::StructuredOptions< scalar_t >.

◆ log_ranks()

template<typename scalar_t >

bool strumpack::HSS::HSSOptions< scalar_t >::log_ranks ( ) const

inline

Check if the ranks should be printed to a log file. NOT supported currently

Returns: True is the ranks should be printed to a log file, else False.

See also: set_log_ranks

◆ nnz()

template<typename scalar_t >

int strumpack::HSS::HSSOptions< scalar_t >::nnz ( ) const

inline

Get the increment number of nonzero entries per row in the SJLT sketch when incrementing by size dd

Returns: additional number of nonzeros per row, added in adaptive compression step with SJLT

See also: set_nnz0(), set_nnz()

◆ nnz0()

template<typename scalar_t >

int strumpack::HSS::HSSOptions< scalar_t >::nnz0 ( ) const

inline

Get the initial number of nonzero entries per row in the SJLT sketch.

Returns: initial number of nonzeros per row, used in adaptive compression with SJLT

See also: set_nnz0(), set_nnz()

◆ p()

template<typename scalar_t >

int strumpack::HSS::HSSOptions< scalar_t >::p ( ) const

inline

Get the current value of the oversampling parameter.

Returns: the oversampling parameter.

See also: set_p()

◆ random_distribution()

template<typename scalar_t >

random::RandomDistribution strumpack::HSS::HSSOptions< scalar_t >::random_distribution ( ) const

inline

Return the type of random distribution to use in the random sampling HSS construction.

Returns: random distribution

See also: set_random_distribution

◆ random_engine()

template<typename scalar_t >

random::RandomEngine strumpack::HSS::HSSOptions< scalar_t >::random_engine ( ) const

inline

Return the type of random engine to use.

Returns: random engine

See also: set_random_engine

◆ set_ann_iterations()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_ann_iterations ( int iters )

inline

Set the number of iterations used in the approximate nearest neighbors search algorithm used in the HSS compression algorithm for kernel matrices. The algorithm will build iters number of random trees.

Parameters

iters Number of random trees to be build

◆ set_approximate_neighbors()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_approximate_neighbors ( int neighbors )

inline

Set the number of approximate nearest neighbors used in the HSS compression algorithm for kernel matrices.

Parameters

neighbors Number of approximate neighbors

◆ set_clustering_algorithm()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_clustering_algorithm ( ClusteringAlgorithm a )

inline

Specify the clustering algorithm. This is used when constructing a kernel matrix approximation.

Parameters

a	Clustering algorithm.

◆ set_compression_algorithm()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_compression_algorithm ( CompressionAlgorithm a )

inline

Specify the variant of the adaptive compression algorithm. See the manual for more information.

Parameters

a	Type of (adaptive) compression scheme

◆ set_compression_sketch()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_compression_sketch ( CompressionSketch a )

inline

Specify the variant of the sketching matrix.

Parameters

a	Type of sketching matrix

◆ set_d0()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_d0 ( int d0 )

inline

Set the initial number of random samples to be used in the random sampling HSS construction algorithm. See the manual for more information on the randomized compression algorithm.

◆ set_dd()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_dd ( int dd )

inline

Set the number of random to be used to increment the random samples vectors in the adaptive randomized HSS compression algorithm. This is only used when compression_algorithm() == CompressionAlgorithm::STABLE. See the manual for more information on the randomized compression algorithm.

◆ set_from_command_line()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_from_command_line	(	int	argc,
		const char const	cargv
	)

overridevirtual

Parse the command line options given by argc and argv. The options will not be modified. Run with –help to see an overview of available options, or call describe_options().

Parameters

argc	Number of elements in argv
argv	Array with options

Reimplemented from strumpack::structured::StructuredOptions< scalar_t >.

◆ set_log_ranks()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_log_ranks ( bool log_ranks )

inline

Log the HSS ranks to a file. TODO is this currently supported??

◆ set_nnz()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_nnz ( int nnz )

inline

Set the number of nonzeros per row added each time dd columns are added to the SJLT sketch.

◆ set_nnz0()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_nnz0 ( int nnz0 )

inline

Set the initial number of nonzeros per row to be used in SJLT sketch.

◆ set_p()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_p ( int p )

inline

Oversampling parameter. Used in adaptive compression, to check stopping criterion.

◆ set_random_distribution()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_random_distribution ( random::RandomDistribution random_distribution )

inline

Set the random distribution, used in randomized compression.

See also: RandomEngine, RandomDistribution, set_random_engine()

◆ set_random_engine()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_random_engine ( random::RandomEngine random_engine )

inline

Set the random engine, used in randomized compression.

See also: RandomEngine, RandomDistribution, set_random_distribution()

◆ set_SJLT_algo()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_SJLT_algo ( SJLTAlgo a )

inline

Specify the variant of the sketching matrix.

Parameters

a	Type of sketching matrix

◆ set_synchronized_compression()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_synchronized_compression ( bool sync )

inline

Set this to true if you require communication in the element extraction routine, since in that case the element extraction is collective, and has to be synchronized.

◆ set_user_defined_random()

template<typename scalar_t >

void strumpack::HSS::HSSOptions< scalar_t >::set_user_defined_random ( bool user_defined_random )

inline

Set this to true if you want to manually fill the random sample vectors with random values.

◆ synchronized_compression()

template<typename scalar_t >

bool strumpack::HSS::HSSOptions< scalar_t >::synchronized_compression ( ) const

inline

Whether or not the synchronize the element extraction routine.

Returns: True if synchronization is required in the element extraction routine, else False.

See also: set_synchromized_compression

◆ user_defined_random()

template<typename scalar_t >

bool strumpack::HSS::HSSOptions< scalar_t >::user_defined_random ( ) const

inline

Will the user define its own random matrices?

Returns: True if the user will fill up the random matrices, used in the random sampling based HSS construcion, him/her-self.

See also: set_user_defined_random

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

/home/pieterg/LBL/STRUMPACK/src/HSS/HSSOptions.hpp

Public Types

Public Member Functions

Detailed Description

Member Typedef Documentation

◆ real_t

Member Function Documentation

◆ ann_iterations()

◆ approximate_neighbors()

◆ clustering_algorithm()

◆ compression_algorithm()

◆ d0()

◆ dd()

◆ describe_options()

◆ log_ranks()

◆ nnz()

◆ nnz0()

◆ p()

◆ random_distribution()

◆ random_engine()

◆ set_ann_iterations()

◆ set_approximate_neighbors()

◆ set_clustering_algorithm()

◆ set_compression_algorithm()

◆ set_compression_sketch()

◆ set_d0()

◆ set_dd()

◆ set_from_command_line()

◆ set_log_ranks()

◆ set_nnz()

◆ set_nnz0()

◆ set_p()

◆ set_random_distribution()

◆ set_random_engine()

◆ set_SJLT_algo()

◆ set_synchronized_compression()

◆ set_user_defined_random()

◆ synchronized_compression()

◆ user_defined_random()