#include <deal.II/multigrid/mg_smoother.h>

Inheritance diagram for mg::SmootherRelaxation< RelaxationType, VectorType >:

Public Member Functions
	SmootherRelaxation (const unsigned int steps=1, const bool variable=false, const bool symmetric=false, const bool transpose=false)

template<typename MatrixType2 >
void	initialize (const MGLevelObject< MatrixType2 > &matrices, const typename RelaxationType::AdditionalData &additional_data=typename RelaxationType::AdditionalData())

template<typename MatrixType2 , class DATA >
void	initialize (const MGLevelObject< MatrixType2 > &matrices, const MGLevelObject< DATA > &additional_data)

void	clear () override

virtual void	smooth (const unsigned int level, VectorType &u, const VectorType &rhs) const override

virtual void	apply (const unsigned int level, VectorType &u, const VectorType &rhs) const override

std::size_t	memory_consumption () const

RelaxationType &	operator[] (const unsigned int level)

const RelaxationType &	operator[] (const unsigned int level) const

const RelaxationType &	back () const

void	resize (const unsigned int new_minlevel, const unsigned int new_maxlevel, Args &&...args)

void	clear_elements ()

unsigned int	min_level () const

unsigned int	max_level () const

unsigned int	n_levels () const

void	apply (ActionFunctionObjectType action)

void	set_steps (const unsigned int)

void	set_variable (const bool)

void	set_symmetric (const bool)

void	set_transpose (const bool)

void	set_debug (const unsigned int level)

Protected Attributes
GrowingVectorMemory< VectorType >	vector_memory

unsigned int	steps

bool	variable

bool	symmetric

bool	transpose

unsigned int	debug

Private Attributes
unsigned int	minlevel

std::vector< std::shared_ptr< RelaxationType > >	objects

Subscriptor functionality
Classes derived from Subscriptor provide a facility to subscribe to this object. This is mostly used by the SmartPointer class.
void	subscribe (std::atomic< bool > *const validity, const std::string &identifier="") const

void	unsubscribe (std::atomic< bool > *const validity, const std::string &identifier="") const

unsigned int	n_subscriptions () const

template<typename StreamType >
void	list_subscribers (StreamType &stream) const

void	list_subscribers () const

template<class Archive >
void	serialize (Archive &ar, const unsigned int version)

void	subscribe (std::atomic< bool > *const validity, const std::string &identifier="") const

void	unsubscribe (std::atomic< bool > *const validity, const std::string &identifier="") const

unsigned int	n_subscriptions () const

template<typename StreamType >
void	list_subscribers (StreamType &stream) const

void	list_subscribers () const

template<class Archive >
void	serialize (Archive &ar, const unsigned int version)

std::atomic< unsigned int >	counter

std::map< std::string, unsigned int >	counter_map

std::vector< std::atomic< bool > * >	validity_pointers

const std::type_info *	object_info

using	map_value_type = decltype(counter_map)::value_type

using	map_iterator = decltype(counter_map)::iterator

static std::mutex	mutex

static ::ExceptionBase &	ExcInUse (int arg1, std::string arg2, std::string arg3)

static ::ExceptionBase &	ExcNoSubscriber (std::string arg1, std::string arg2)

static ::ExceptionBase &	ExcInUse (int arg1, std::string arg2, std::string arg3)

static ::ExceptionBase &	ExcNoSubscriber (std::string arg1, std::string arg2)

void	check_no_subscribers () const noexcept

Detailed Description

template<class RelaxationType, typename VectorType>
class mg::SmootherRelaxation< RelaxationType, VectorType >

Smoother using relaxation classes.

A relaxation class is an object that satisfies the relaxation concept.

This class performs smoothing on each level. The operation can be controlled by several parameters. First, the relaxation parameter omega is used in the underlying relaxation method. steps is the number of relaxation steps on the finest level (on all levels if variable is off). If variable is true, the number of smoothing steps is doubled on each coarser level. This results in a method having the complexity of the W-cycle, but saving grid transfers. This is the method proposed by Bramble at al.

The option symmetric switches on alternating between the smoother and its transpose in each step as proposed by Bramble.

transpose uses the transposed smoothing operation using Tstep instead of the regular step of the relaxation scheme.

If you are using block matrices, the second initialize function offers the possibility to extract a single block for smoothing. In this case, the multigrid method must be used only with the vector associated to that single block.

Definition at line 183 of file mg_smoother.h.

Constructor & Destructor Documentation

◆ SmootherRelaxation()

template<class RelaxationType , typename VectorType >

mg::SmootherRelaxation< RelaxationType, VectorType >::SmootherRelaxation	(	const unsigned int	steps = `1`,
		const bool	variable = `false`,
		const bool	symmetric = `false`,
		const bool	transpose = `false`
	)

Constructor. Sets smoothing parameters.

Member Function Documentation

◆ initialize() [1/2]

template<class RelaxationType , typename VectorType >

template<typename MatrixType2 >

void mg::SmootherRelaxation< RelaxationType, VectorType >::initialize	(	const MGLevelObject< MatrixType2 > &	matrices,
		const typename RelaxationType::AdditionalData &	additional_data = `typename RelaxationType::AdditionalData()`
	)

Initialize for matrices. This function initializes the smoothing operator with the same smoother for each level.

additional_data is an object of type RelaxationType::AdditionalData and is handed to the initialization function of the relaxation method.

◆ initialize() [2/2]

template<class RelaxationType , typename VectorType >

template<typename MatrixType2 , class DATA >

void mg::SmootherRelaxation< RelaxationType, VectorType >::initialize	(	const MGLevelObject< MatrixType2 > &	matrices,
		const MGLevelObject< DATA > &	additional_data
	)

Initialize matrices and additional data for each level.

If minimal or maximal level of the two objects differ, the greatest common range is utilized. This way, smoothing can be restricted to certain levels even if the matrix was generated for all levels.

◆ clear()

template<class RelaxationType , typename VectorType >

void mg::SmootherRelaxation< RelaxationType, VectorType >::clear ( )

overridevirtual

Empty all vectors.

Implements MGSmootherBase< VectorType >.

◆ smooth()

template<class RelaxationType , typename VectorType >

virtual void mg::SmootherRelaxation< RelaxationType, VectorType >::smooth	(	const unsigned int	level,
		VectorType &	u,
		const VectorType &	rhs
	)		const

overridevirtual

The actual smoothing method.

Implements MGSmootherBase< VectorType >.

◆ apply() [1/2]

template<class RelaxationType , typename VectorType >

virtual void mg::SmootherRelaxation< RelaxationType, VectorType >::apply	(	const unsigned int	level,
		VectorType &	u,
		const VectorType &	rhs
	)		const

overridevirtual

The apply variant of smoothing, setting the vector u to zero before calling the smooth function. This function is equivalent to the following code

u = 0;

smooth(level, u, rhs);

mg::SmootherRelaxation::smooth

virtual void smooth(const unsigned int level, VectorType &u, const VectorType &rhs) const override

level

unsigned int level

Definition: grid_out.cc:4606

In the multigrid preconditioner interfaces, the apply() method is used for the pre-smoothing operation because the previous content in the solution vector needs to be overwritten for a new incoming residual. On the other hand, all subsequent operations need to smooth the content already present in the vector u given the right hand side, which is done by smooth().

Reimplemented from MGSmootherBase< VectorType >.

◆ memory_consumption()

template<class RelaxationType , typename VectorType >

std::size_t mg::SmootherRelaxation< RelaxationType, VectorType >::memory_consumption ( ) const

Memory used by this object.

◆ operator[]() [1/2]

RelaxationType & MGLevelObject< RelaxationType >::operator[] ( const unsigned int level )

inherited

Access object on level level.

Definition at line 89 of file mg_level_object.h.

◆ operator[]() [2/2]

const RelaxationType & MGLevelObject< RelaxationType >::operator[] ( const unsigned int level ) const

inherited

Access object on level level.

This function can be called on a const object, and consequently returns a const reference.

Definition at line 98 of file mg_level_object.h.

◆ back()

const RelaxationType & MGLevelObject< RelaxationType >::back

inherited

Return object on level max.

Definition at line 104 of file mg_level_object.h.

◆ resize()

void MGLevelObject< RelaxationType >::resize	(	const unsigned int	new_minlevel,
		const unsigned int	new_maxlevel,
		Args &&...	args
	)

inherited

Delete all previous contents of this object and reset its size according to the values of new_minlevel and new_maxlevel.

Parameters

[in]	new_minlevel	The lowest level for which to provision memory for level objects.
[in]	new_maxlevel	The highest level for which to provision memory for level objects.
[in]	args	Optional arguments passed to the constructor of the underlying object.

Precondition: minlevel <= maxlevel

Definition at line 121 of file mg_level_object.h.

◆ clear_elements()

void MGLevelObject< RelaxationType >::clear_elements

inherited

Call clear on all objects stored by this object. This function is only implemented for some Object classes, e.g., matrix types or the PreconditionBlockSOR and similar classes. Using this function will fail with a compiler error if the Object template type to this class does not provide a clear() member function.

Definition at line 143 of file mg_level_object.h.

◆ min_level()

unsigned int MGLevelObject< RelaxationType >::min_level

inherited

The coarsest level for which this class stores a level object.

Definition at line 149 of file mg_level_object.h.

◆ max_level()

unsigned int MGLevelObject< RelaxationType >::max_level

inherited

The highest level for which this class stores a level object.

Definition at line 155 of file mg_level_object.h.

◆ n_levels()

unsigned int MGLevelObject< RelaxationType >::n_levels

inherited

Number of levels, i.e., max_level()-min_level()+1.

Definition at line 161 of file mg_level_object.h.

◆ apply() [2/2]

void MGLevelObject< RelaxationType >::apply ( ActionFunctionObjectType action )

inherited

Apply the action action to every object stored in here. The parameter action is expected to be a function object that accepts the syntax action(const unsigned int level, Object &object); This means this function can accept a lambda, a std::function, or a plain function pointer.

Definition at line 175 of file mg_level_object.h.

◆ set_steps()

template<typename VectorType >

void MGSmoother< VectorType >::set_steps ( const unsigned int )

inherited

Modify the number of smoothing steps on finest level.

◆ set_variable()

template<typename VectorType >

void MGSmoother< VectorType >::set_variable ( const bool )

inherited

Switch on/off variable smoothing.

◆ set_symmetric()

template<typename VectorType >

void MGSmoother< VectorType >::set_symmetric ( const bool )

inherited

Switch on/off symmetric smoothing.

◆ set_transpose()

template<typename VectorType >

void MGSmoother< VectorType >::set_transpose ( const bool )

inherited

Switch on/off transposed smoothing. The effect is overridden by set_symmetric().

◆ set_debug()

template<typename VectorType >

void MGSmoother< VectorType >::set_debug ( const unsigned int level )

inherited

Set debug to a nonzero value to get debug information logged to deallog. Increase to get more information

Member Data Documentation

◆ minlevel

unsigned int MGLevelObject< RelaxationType >::minlevel

privateinherited

Level of first component.

Definition at line 187 of file mg_level_object.h.

◆ objects

std::vector<std::shared_ptr<RelaxationType > > MGLevelObject< RelaxationType >::objects

privateinherited

Array of the objects to be held.

Definition at line 192 of file mg_level_object.h.

◆ vector_memory

template<typename VectorType >

GrowingVectorMemory<VectorType> MGSmoother< VectorType >::vector_memory

mutableprotectedinherited

A memory object to be used for temporary vectors.

The object is marked as mutable since we will need to use it to allocate temporary vectors also in functions that are const.

Definition at line 98 of file mg_smoother.h.

◆ steps

template<typename VectorType >

unsigned int MGSmoother< VectorType >::steps

protectedinherited

Number of smoothing steps on the finest level. If no variable smoothing is chosen, this is the number of steps on all levels.

Definition at line 104 of file mg_smoother.h.

◆ variable

template<typename VectorType >

bool MGSmoother< VectorType >::variable

protectedinherited

Variable smoothing: double the number of smoothing steps whenever going to the next coarser level

Definition at line 110 of file mg_smoother.h.

◆ symmetric

template<typename VectorType >

bool MGSmoother< VectorType >::symmetric

protectedinherited

Symmetric smoothing: in the smoothing iteration, alternate between the relaxation method and its transpose.

Definition at line 116 of file mg_smoother.h.

◆ transpose

template<typename VectorType >

bool MGSmoother< VectorType >::transpose

protectedinherited

Use the transpose of the relaxation method instead of the method itself. This has no effect if symmetric smoothing is chosen.

Definition at line 122 of file mg_smoother.h.

◆ debug

template<typename VectorType >

unsigned int MGSmoother< VectorType >::debug

protectedinherited

Output debugging information to deallog if this is nonzero.

Definition at line 127 of file mg_smoother.h.

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

include/deal.II/multigrid/mg_smoother.h

Public Member Functions

Protected Attributes

Private Attributes

Subscriptor functionality

Detailed Description

Constructor & Destructor Documentation

◆ SmootherRelaxation()

Member Function Documentation

◆ initialize() [1/2]

◆ initialize() [2/2]

◆ clear()

◆ smooth()

◆ apply() [1/2]

◆ memory_consumption()

◆ operator[]() [1/2]

◆ operator[]() [2/2]

◆ back()

◆ resize()

◆ clear_elements()

◆ min_level()

◆ max_level()

◆ n_levels()

◆ apply() [2/2]

◆ set_steps()

◆ set_variable()

◆ set_symmetric()

◆ set_transpose()

◆ set_debug()

Member Data Documentation

◆ minlevel

◆ objects

◆ vector_memory

◆ steps

◆ variable

◆ symmetric

◆ transpose

◆ debug