Class CoarsenSchedule<DIM> performs the communication operations to coarsen data from a finer level to a coarser level.
More...

#include <ibtk/SideDataSynchronization.h>

Inheritance diagram for SAMRAI::xfer::CoarsenSchedule< DIM >:

[legend]

Public Member Functions
	CoarsenSchedule (tbox::Pointer< hier::PatchLevel< DIM > > crse_level, tbox::Pointer< hier::PatchLevel< DIM > > fine_level, const tbox::Pointer< xfer::CoarsenClasses< DIM > > coarsen_classes, tbox::Pointer< xfer::CoarsenTransactionFactory< DIM > > transaction_factory, xfer::CoarsenPatchStrategy< DIM > *patch_strategy, bool fill_coarse_data)

virtual	~CoarsenSchedule ()

void	reset (const tbox::Pointer< xfer::CoarsenClasses< DIM > > coarsen_classes)

void	coarsenData () const

const tbox::Pointer< xfer::CoarsenClasses< DIM > > &	getEquivalenceClasses () const

virtual void	printClassData (std::ostream &stream) const

Static Public Member Functions
static void	setScheduleGenerationMethod (const std::string &method)

Private Member Functions
	CoarsenSchedule (const CoarsenSchedule< DIM > &)

void	operator= (const CoarsenSchedule< DIM > &)

void	generateSchedule ()
	Main schedule generation routine which passes control to one of the algorithmic variations based on value of s_schedule_generation_method. More...

void	generateScheduleNSquared ()
	This version of the schedule generation procedure uses N^2 algorithms to determine box intersections; i.e., the original SAMRAI implementation which checks every box against every other. More...

void	generateScheduleBoxGraph ()
	This version of the schedule generation procedure uses a bipartite graph algorithm to determine which source patches contribute data to each destination patch. More...

void	generateScheduleBoxTree ()
	This version of the schedule generation procedure uses a recursive binary box tree algorithm to determine which source patches contribute data to each destination patch and to compute unfilled_boxes. More...

void	generateTemporaryLevel ()
	Generate a temporary coarse level by coarsening the fine level. Note that this function does not allocate patch data storage. More...

void	setupRefineAlgorithm ()
	Set up refine algorithms to transfer coarsened data and to fill temporary coarse level before coarsening operations, if needed. More...

void	coarsenSourceData (xfer::CoarsenPatchStrategy< DIM > *patch_strategy) const
	Coarsen source patch data from the fine patch level into the source patch data on the coarse temporary patch level. More...

hier::IntVector< DIM >	getMaxGhostsToGrow () const
	Calculate the maximum ghost cell width to grow boxes to check for overlaps. More...

void	constructScheduleTransactions (tbox::Pointer< hier::PatchLevel< DIM > > dst_level, int dst_patch_id, tbox::Pointer< hier::PatchLevel< DIM > > src_level, int src_patch_id)
	Function that constructs schedule transactions that move data from source patch on source level to destination patch on destination level. More...

void	setCoarsenItems (const tbox::Pointer< xfer::CoarsenClasses< DIM > > coarsen_classes)
	Utility function to set up local copies of patch data source, destination, etc. indices and necessary data coarsening information stored in the coarsen classes object generated by the coarsen algorithm. More...

void	clearCoarsenItems ()
	Utility function to clear local copies of coarsen items. More...

void	initialCheckCoarsenClassItems () const
	Utility function to check coarsen items to see whether source and destination patch data components have sufficient ghost cell widths to satisfy the "ghost width to coarsen" functionality described in the CoarsenAlgorithm<DIM> class header. More...

Private Attributes
tbox::Pointer< xfer::CoarsenClasses< DIM > >	d_coarsen_classes

int	d_number_coarsen_items

const xfer::CoarsenClasses< DIM >::Data **	d_coarsen_items

tbox::Pointer< hier::PatchLevel< DIM > >	d_crse_level

tbox::Pointer< hier::PatchLevel< DIM > >	d_fine_level

tbox::Pointer< hier::PatchLevel< DIM > >	d_temp_crse_level

xfer::CoarsenPatchStrategy< DIM > *	d_coarsen_patch_strategy

tbox::Pointer< xfer::CoarsenTransactionFactory< DIM > >	d_transaction_factory

hier::IntVector< DIM >	d_ratio_between_levels

hier::ComponentSelector	d_sources

tbox::Pointer< tbox::Schedule >	d_schedule

bool	d_fill_coarse_data

tbox::Pointer< RefineAlgorithm< DIM > >	d_precoarsen_refine_algorithm

tbox::Pointer< RefineSchedule< DIM > >	d_precoarsen_refine_schedule

tbox::Pointer< tbox::Timer >	t_coarsen_data

tbox::Pointer< tbox::Timer >	t_gen_sched_n_squared

tbox::Pointer< tbox::Timer >	t_gen_sched_box_graph

tbox::Pointer< tbox::Timer >	t_gen_sched_box_tree

Static Private Attributes
static const hier::IntVector< DIM >	s_constant_zero_intvector

static const hier::IntVector< DIM >	s_constant_one_intvector

static std::string	s_schedule_generation_method

Detailed Description

template<int DIM>
class SAMRAI::xfer::CoarsenSchedule< DIM >

Typically, data is coarsened from the interiors of source patch components on the source patch level into interiors of destination patch components on
the destination level. However, variations are possible for special situations; see the CoarsenAlgorithm<DIM> class header for more information. Generally, the source patch data must contain sufficient ghost cells to satisfy the coarsening operators involved. If a coarsen operator has a non-zero ghost cell width, then the source ghost cells must be filled before the coarsen schedule is executed. The communication schedule is executed by calling member function coarsenData().

Each schedule object is typically created by a coarsen algorithm and represents communication dependencies for a particular configuration of the AMR hierarchy. The communication schedule is only valid for that particular configuration and must be regenerated when the AMR patch hierarchy changes. However, as long as the patch levels involved in the creation of the schedule remain unchanged, the schedule may be used for multiple communication cycles. For more information about creating refine schedules, see the CoarsenAlgorithm<DIM> header file.

NOTE: Algorithmic variations are available by calling the static method CoarsenSchedule<DIM>::setScheduleGenerationMethod(), which sets the option for all instances of the class.

See also: xfer::CoarsenAlgorithm; xfer::CoarsenPatchStrategy; xfer::CoarsenClasses

Constructor & Destructor Documentation

◆ CoarsenSchedule() [1/2]

template<int DIM>

SAMRAI::xfer::CoarsenSchedule< DIM >::CoarsenSchedule	(	tbox::Pointer< hier::PatchLevel< DIM > >	crse_level,
		tbox::Pointer< hier::PatchLevel< DIM > >	fine_level,
		const tbox::Pointer< xfer::CoarsenClasses< DIM > >	coarsen_classes,
		tbox::Pointer< xfer::CoarsenTransactionFactory< DIM > >	transaction_factory,
		xfer::CoarsenPatchStrategy< DIM > *	patch_strategy,
		bool	fill_coarse_data
	)

Constructor to create a coarsen schedule that coarsens data from source patch data components on the fine level into the destination patch data components on the coarse level. In general, this constructor is called by a CoarsenAlgorithm<DIM> object. For possible variations on data coarsening, see the CoarsenAlgorithm<DIM> class header information.

If the coarsening operators require data from ghost cells, then the associated source patch data components must have a sufficient ghost cell width and and they must be filled with valid data before calling coarsenData().

Parameters

crse_level	Pointer to coarse (destination) patch level.
fine_level	Pointer to fine (source) patch level.
coarsen_classes	Pointer to structure containing patch data and operator information. In general, this is constructed by the calling CoarsenAlgorithm<DIM> object.
transaction_factory	Pointer to a factory object that will create data transactions.
patch_strategy	Pointer to a coarsen patch strategy object that provides user-defined coarsen operations. This ponter may be null, in which case no user-defined coarsen operations will be performed.
fill_coarse_data	Boolean indicating whether coarse data should be filled before coarsening operations are done.

When assertion checking is active, unrecoverable assertions will result if either patch level pointer, the refine classes pointer, or the transaction factory pointer, is null.

◆ ~CoarsenSchedule()

template<int DIM>

virtual SAMRAI::xfer::CoarsenSchedule< DIM >::~CoarsenSchedule ( )

virtual

The virtual destructor for the schedule releases all internal storage.

◆ CoarsenSchedule() [2/2]

template<int DIM>

SAMRAI::xfer::CoarsenSchedule< DIM >::CoarsenSchedule ( const CoarsenSchedule< DIM > & )

private

Member Function Documentation

◆ setScheduleGenerationMethod()

template<int DIM>

static void SAMRAI::xfer::CoarsenSchedule< DIM >::setScheduleGenerationMethod ( const std::string & method )

static

Static function to set box intersection algorithm to use during schedule construction for all CoarsenSchedule objects.
If this method is not called, the default will be used.

Parameters

method string identifying box intersection method. Valid choices are: "BOX_TREE" (default case), "BOX_GRAPH", and "ORIG_NSQUARED". More details can be found below in the comments for the generateSchedule() routine.

If an invalid string is passed, an unrecoverable error will result.

◆ reset()

template<int DIM>

void SAMRAI::xfer::CoarsenSchedule< DIM >::reset ( const tbox::Pointer< xfer::CoarsenClasses< DIM > > coarsen_classes )

Reset this coarsen schedule to perform data transfers asssociated with coarsen class items in function argument. In general, this function is called by a CoarsenAlgorithm<DIM> object.

Parameters

coarsen_classes Pointer to structure containing patch data and operator information. In general, this is constructed by the calling CoarsenAlgorithm<DIM> object. This pointer must be non-null.

◆ coarsenData()

template<int DIM>

void SAMRAI::xfer::CoarsenSchedule< DIM >::coarsenData ( ) const

Execute the stored communication schedule and perform the data movement.

◆ getEquivalenceClasses()

template<int DIM>

const tbox::Pointer< xfer::CoarsenClasses<DIM> >& SAMRAI::xfer::CoarsenSchedule< DIM >::getEquivalenceClasses ( ) const

Return const reference to the pointer to coarsen equivalence classes used in schedule.

◆ printClassData()

template<int DIM>

virtual void SAMRAI::xfer::CoarsenSchedule< DIM >::printClassData ( std::ostream & stream ) const

virtual

Print the coarsen schedule state to the specified data stream.

Parameters

stream Output data stream.

◆ operator=()

template<int DIM>

void SAMRAI::xfer::CoarsenSchedule< DIM >::operator= ( const CoarsenSchedule< DIM > & )

private

◆ generateSchedule()

template<int DIM>

void SAMRAI::xfer::CoarsenSchedule< DIM >::generateSchedule ( )

private

The resulting communication schedule will move source patch data from a temporary coarse level (i.e., coarsened version of fine level) into the destination patch data of the destination (coarse) level.

The generateSchedule() routine invokes various versions of the schedule generation process implemented in the similarly named routines below based on the chosen schedule generation method. The different options will not change the result of the application but may improve its performance, especially for large numbers of processors. Note that the algorithm choice may be changed by calling the setScheduleGenerationMethod() routine.

The possibilities are as follows:

if setScheduleGenerationMethod("BOX_TREE") is called use generateScheduleBoxTree() to generate the schedule. NOTE: THIS IS THE DEFAULT OPTION.
if setScheduleGenerationMethod("BOX_GRAPH") is called use generateScheduleBoxGraph() to generate the schedule.
if setScheduleGenerationMethod("ORIG_NSQUARED") is called use generateScheduleNSquared() to generate the schedule.

◆ generateScheduleNSquared()

template<int DIM>

void SAMRAI::xfer::CoarsenSchedule< DIM >::generateScheduleNSquared ( )

private

◆ generateScheduleBoxGraph()

template<int DIM>

void SAMRAI::xfer::CoarsenSchedule< DIM >::generateScheduleBoxGraph ( )

private

◆ generateScheduleBoxTree()

template<int DIM>

void SAMRAI::xfer::CoarsenSchedule< DIM >::generateScheduleBoxTree ( )

private

◆ generateTemporaryLevel()

template<int DIM>

void SAMRAI::xfer::CoarsenSchedule< DIM >::generateTemporaryLevel ( )

private

◆ setupRefineAlgorithm()

template<int DIM>

void SAMRAI::xfer::CoarsenSchedule< DIM >::setupRefineAlgorithm ( )

private

The associated schedules are set in the generateSchedule() routine.

◆ coarsenSourceData()

template<int DIM>

void SAMRAI::xfer::CoarsenSchedule< DIM >::coarsenSourceData ( xfer::CoarsenPatchStrategy< DIM > * patch_strategy ) const

private

◆ getMaxGhostsToGrow()

template<int DIM>

hier::IntVector<DIM> SAMRAI::xfer::CoarsenSchedule< DIM >::getMaxGhostsToGrow ( ) const

private

◆ constructScheduleTransactions()

template<int DIM>

void SAMRAI::xfer::CoarsenSchedule< DIM >::constructScheduleTransactions	(	tbox::Pointer< hier::PatchLevel< DIM > >	dst_level,
		int	dst_patch_id,
		tbox::Pointer< hier::PatchLevel< DIM > >	src_level,
		int	src_patch_id
	)

private

◆ setCoarsenItems()

template<int DIM>

void SAMRAI::xfer::CoarsenSchedule< DIM >::setCoarsenItems ( const tbox::Pointer< xfer::CoarsenClasses< DIM > > coarsen_classes )

private

An array of coarsen data items is stored locally here to facilitate interaction with transations.

◆ clearCoarsenItems()

template<int DIM>

void SAMRAI::xfer::CoarsenSchedule< DIM >::clearCoarsenItems ( )

private

◆ initialCheckCoarsenClassItems()

template<int DIM>

void SAMRAI::xfer::CoarsenSchedule< DIM >::initialCheckCoarsenClassItems ( ) const

private

Specifically, the destination data must have a ghost cell width at least as large as the ghost cell width to coarsen. The source data must have a ghost cell width at least as large as the ghost cell width to coarsen refined to the source (finer) level index space. Although it is redundant if the coarsen algorithm created the coarsen classes, the routine CoarsenClasses<DIM>::checkCoarsenItem() is also called.

If any entries are erroneous an assertion is thrown with a descriptive error message and program halts.

Member Data Documentation

◆ s_constant_zero_intvector

template<int DIM>

const hier::IntVector<DIM> SAMRAI::xfer::CoarsenSchedule< DIM >::s_constant_zero_intvector

staticprivate

Constant int vectors used to avoid recreating these vectors in loops.

◆ s_constant_one_intvector

template<int DIM>

const hier::IntVector<DIM> SAMRAI::xfer::CoarsenSchedule< DIM >::s_constant_one_intvector

staticprivate

◆ s_schedule_generation_method

template<int DIM>

std::string SAMRAI::xfer::CoarsenSchedule< DIM >::s_schedule_generation_method

staticprivate

Selects algorithm used to generate communication schedule.

◆ d_coarsen_classes

template<int DIM>

tbox::Pointer< xfer::CoarsenClasses<DIM> > SAMRAI::xfer::CoarsenSchedule< DIM >::d_coarsen_classes

private

Structures that store coarsen data items.

◆ d_number_coarsen_items

template<int DIM>

int SAMRAI::xfer::CoarsenSchedule< DIM >::d_number_coarsen_items

private

◆ d_coarsen_items

template<int DIM>

const xfer::CoarsenClasses<DIM>::Data** SAMRAI::xfer::CoarsenSchedule< DIM >::d_coarsen_items

private

◆ d_crse_level

template<int DIM>

tbox::Pointer< hier::PatchLevel<DIM> > SAMRAI::xfer::CoarsenSchedule< DIM >::d_crse_level

private

Cached pointers to the coarse, fine, and temporary patch levels.

◆ d_fine_level

template<int DIM>

tbox::Pointer< hier::PatchLevel<DIM> > SAMRAI::xfer::CoarsenSchedule< DIM >::d_fine_level

private

◆ d_temp_crse_level

template<int DIM>

tbox::Pointer< hier::PatchLevel<DIM> > SAMRAI::xfer::CoarsenSchedule< DIM >::d_temp_crse_level

private

◆ d_coarsen_patch_strategy

template<int DIM>

xfer::CoarsenPatchStrategy<DIM>* SAMRAI::xfer::CoarsenSchedule< DIM >::d_coarsen_patch_strategy

private

Object supporting interface to user-defined spatial data coarsening operations.

◆ d_transaction_factory

template<int DIM>

tbox::Pointer< xfer::CoarsenTransactionFactory<DIM> > SAMRAI::xfer::CoarsenSchedule< DIM >::d_transaction_factory

private

Factory object used to create data transactions when schedule is constructed.

◆ d_ratio_between_levels

template<int DIM>

hier::IntVector<DIM> SAMRAI::xfer::CoarsenSchedule< DIM >::d_ratio_between_levels

private

Cached ratio between source (fine) level and destination (coarse) level.

◆ d_sources

template<int DIM>

hier::ComponentSelector SAMRAI::xfer::CoarsenSchedule< DIM >::d_sources

private

Stored source patch data indices for rapid data allocation/deallocation.

◆ d_schedule

template<int DIM>

tbox::Pointer<tbox::Schedule> SAMRAI::xfer::CoarsenSchedule< DIM >::d_schedule

private

Level-to-level communication schedule between the temporary coarse level and (actual) destination level.

◆ d_fill_coarse_data

template<int DIM>

bool SAMRAI::xfer::CoarsenSchedule< DIM >::d_fill_coarse_data

private

Boolean indicating whether source data on the coarse temporary level must be filled before coarsening operations (see comments for class constructor in header file), and refine algorithm and schedule needed to peform up these fill operations.

◆ d_precoarsen_refine_algorithm

template<int DIM>

tbox::Pointer< RefineAlgorithm<DIM> > SAMRAI::xfer::CoarsenSchedule< DIM >::d_precoarsen_refine_algorithm

private

◆ d_precoarsen_refine_schedule

template<int DIM>

tbox::Pointer< RefineSchedule<DIM> > SAMRAI::xfer::CoarsenSchedule< DIM >::d_precoarsen_refine_schedule

private

◆ t_coarsen_data

template<int DIM>

tbox::Pointer<tbox::Timer> SAMRAI::xfer::CoarsenSchedule< DIM >::t_coarsen_data

private

Timer objects for performance measurement.

◆ t_gen_sched_n_squared

template<int DIM>

tbox::Pointer<tbox::Timer> SAMRAI::xfer::CoarsenSchedule< DIM >::t_gen_sched_n_squared

private

◆ t_gen_sched_box_graph

template<int DIM>

tbox::Pointer<tbox::Timer> SAMRAI::xfer::CoarsenSchedule< DIM >::t_gen_sched_box_graph

private

◆ t_gen_sched_box_tree

template<int DIM>

tbox::Pointer<tbox::Timer> SAMRAI::xfer::CoarsenSchedule< DIM >::t_gen_sched_box_tree

private

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

/samrai/include/CoarsenSchedule.h

Public Member Functions

Static Public Member Functions

Private Member Functions

Private Attributes

Static Private Attributes

Detailed Description

template<int DIM> class SAMRAI::xfer::CoarsenSchedule< DIM >

Constructor & Destructor Documentation

◆ CoarsenSchedule() [1/2]

◆ ~CoarsenSchedule()

◆ CoarsenSchedule() [2/2]

Member Function Documentation

◆ setScheduleGenerationMethod()

◆ reset()

◆ coarsenData()

◆ getEquivalenceClasses()

◆ printClassData()

◆ operator=()

◆ generateSchedule()

◆ generateScheduleNSquared()

◆ generateScheduleBoxGraph()

◆ generateScheduleBoxTree()

◆ generateTemporaryLevel()

◆ setupRefineAlgorithm()

◆ coarsenSourceData()

◆ getMaxGhostsToGrow()

◆ constructScheduleTransactions()

◆ setCoarsenItems()

◆ clearCoarsenItems()

◆ initialCheckCoarsenClassItems()

Member Data Documentation

◆ s_constant_zero_intvector

◆ s_constant_one_intvector

◆ s_schedule_generation_method

◆ d_coarsen_classes

◆ d_number_coarsen_items

◆ d_coarsen_items

◆ d_crse_level

◆ d_fine_level

◆ d_temp_crse_level

◆ d_coarsen_patch_strategy

◆ d_transaction_factory

◆ d_ratio_between_levels

◆ d_sources

◆ d_schedule

◆ d_fill_coarse_data

◆ d_precoarsen_refine_algorithm

◆ d_precoarsen_refine_schedule

◆ t_coarsen_data

◆ t_gen_sched_n_squared

◆ t_gen_sched_box_graph

◆ t_gen_sched_box_tree

template<int DIM>
class SAMRAI::xfer::CoarsenSchedule< DIM >