Node in the asynchronous Berger-Rigoutsos (BR) dendogram. More...

#include <AsyncBergerRigoutsosNode.h>

Inheritance diagram for SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >:

Classes
struct	CommonParams
	Parameters shared among all dendogram nodes in an dendogram and collectively managed by those nodes. More...

Public Types
enum	OwnerMode { SINGLE_OWNER = 0, MOST_OVERLAP = 1, FEWEST_OWNED = 2, LEAST_ACTIVE = 3 }

typedef hier::LayerNodeSet< DIM >::Node	GraphNode
	Shorthand for the box-graph node corresponding to boxes. More...

typedef hier::LayerNodeSet< DIM >::NodeContainer	GraphNodeContainer
	Shorthand for a container of graph-nodes. More...

typedef hier::LayerEdgeSet< DIM >::NabrContainer	GraphNabrContainer
	Shorthand for a container of neighbor graph-nodes. More...

typedef hier::LayerEdgeSet< DIM >::Connectivity	Connectivity
	Shortthand for the connectivity between two sets of graph nodes. More...

typedef std::set< int >	IntSet
	Shorthand for a sorted, possibly incontiguous, set of integers. More...

enum	JobState { COMMUNICATION_WAIT = 0, NONCOMMUNICATION_WAIT = 1, JOB_IS_COMPLETED = 2 }

Public Member Functions
	AsyncBergerRigoutsosNode (CommonParams common_params, const hier::Box< DIM > bound_box=NULL, mesh::AsyncBergerRigoutsosNode< DIM > *parent=NULL, const int child_number=1)
	Construct a node of a BR dendogram. More...

	~AsyncBergerRigoutsosNode (void)
	Destructor. More...

Algorithm mode settings
void	setMaxGhostCellWidth (const hier::IntVector< DIM > &max_gcw)
	Set the maximum ghost cell width used for checking overlaps. More...

void	setAlgorithmAdvanceMode (const std::string &algo_advance_mode)
	Set the mode for advancing the asynchronous implementation. More...

void	setOwnerMode (const std::string &mode)
	Set the method for choosing the owner. Choices: More...

void	setUseLevelBoxes (bool flag)
	Switch on or off the use of global level boxes. More...

void	setComputeEdges (int compute_edges)
	Edge computation flag. More...

void	runAlgorithm ()
	Run the BR algorithm to find boxes. More...

Access to outputs
const hier::LayerNodeSet< DIM > &	getNewNodes () const
	Get the output boxes in a hier::LayerNodeSet<DIM> form. More...

const Connectivity &	getNewCnect () const
	Get the connectivity between input and output graph nodes (between the tagged and new graph nodes). More...

Developer's methods for analysis and debugging this class.
virtual void	printClassData (std::ostream &os, int detail_level=0) const

int	getMaxNodes () const
	Max number of local nodes for dendogram. More...

int	getMaxGeneration () const
	max generation count for the local nodes in the dendogram. More...

int	getMaxOwnership () const
	Max number of locally owned nodes in the dendogram. More...

double	getAvgNumberOfCont () const
	Average number of continuations for local nodes in dendogram. More...

int	getMaxNumberOfCont () const
	Max number of continuations for local nodes in dendogram. More...

int	getNumBoxesGenerated () const
	Number of boxes generated (but not necessarily owned) on the local process. More...

void	setLogNodeHistory (bool flag)
	Set whether to log dendogram node action history (useful for debugging). More...

Private Member Functions
Delegated tasks for various phases of running algorithm.
void	makeLocalTagHistogram ()

void	reduceHistogram_start ()

bool	reduceHistogram_check ()

void	computeMinimalBoundingBoxForTags ()

void	acceptOrSplitBox ()

void	broadcastAcceptability_start ()

bool	broadcastAcceptability_check ()

void	countOverlapWithLocalPatches ()

void	gatherGroupingCriteria_start ()

bool	gatherGroupingCriteria_check ()

void	formChildGroups ()
	Form child groups from gathered overlap counts. More...

void	formChildGroupsUsingLevelBoxes ()
	Form child groups from local copy of all level boxes. More...

void	broadcastChildGroups_start ()

bool	broadcastChildGroups_check ()

void	runChildren_start ()

bool	runChildren_check ()

void	broadcastToDropouts_start ()

bool	broadcastToDropouts_check ()

void	createLayerNode ()

void	eraseLayerNode ()

void	computeNewGraphEdges ()
	Compute new graph edges touching local tag nodes. More...

void	shareNewEdgesWithOwners ()
	Participants send new edge data to graph node owners. More...

Utilities for implementing algorithm
int	findOwnerInGroup (int owner, const tbox::Array< int > &group) const
	Find the index of the owner in the group. More...

void	claimMPITag ()
	Claim a unique tag from process's available tag pool. More...

int	computeCommunicationTreeDegree (int group_size) const
	Heuristically determine "best" tree degree for communication group size. More...

bool	findZeroCutPoint (int &cut_pt, const int dim)

bool	findZeroCutSwath (int &cut_lo, int &cut_hi, const int dim)

void	cutAtLaplacian (int &cut_pt, const int dim, const int lo, const int hi, const int min_size)

int	getHistogramBufferSize (const hier::Box< DIM > &box) const

int *	putHistogramToBuffer (int *buffer)

int *	getHistogramFromBuffer (int *buffer)

int *	putBoxToBuffer (const hier::Box< DIM > &box, int *buffer) const

int *	getBoxFromBuffer (hier::Box< DIM > &box, int *buffer) const

void	computeDropoutGroup (const tbox::Array< int > &main_group, const tbox::Array< int > &sub_group, tbox::Array< int > &dropouts, const int add_group) const
	Compute list of non-participating processes. More...

BoxAcceptance	intToBoxAcceptance (int i) const

bool	boxAccepted () const

bool	boxRejected () const

bool	boxHasNoTag () const

Private Attributes
Tree-related data
AsyncBergerRigoutsosNode *	d_parent
	Parent node (or NULL for the root node). More...

AsyncBergerRigoutsosNode *	d_lft_child
	Left child. More...

AsyncBergerRigoutsosNode *	d_rht_child
	Right child. More...

Data for one recursion of the BR algorithm
hier::Box< DIM >	d_box

int	d_owner

Id of participating processes.
tbox::Array< int >	d_group

int	d_mpi_tag
	MPI tag for message within a dendogram node. More...

int	d_overlap
	Overlap count with d_box. More...

BoxAcceptance	d_box_acceptance
	Whether and how box is accepted. More...

tbox::Array< int >	d_histogram [DIM]
	Histogram for all dimensions of box d_box. More...

GraphNode	d_node
	Distributed graph node corresponding to an accepted box. More...

GraphNodeContainer::iterator	d_node_iterator
	Distributed graph node iterator corresponding to an accepted box on the owner. More...

WaitPhase	d_wait_phase
	Name of wait phase when continueAlgorithm() exits before completion. More...

Lower-level parameters for communication.
tbox::Array< int >	d_send_msg
	Buffer for organizing outgoing data. More...

tbox::Array< int >	d_recv_msg
	Buffer for organizing incoming data. More...

tbox::AsyncCommGroup *	d_comm_group

Deubgging aid
const int	d_generation
	Generation number. More...

int	d_n_cont
	Number of times continueAlgorithm was called. More...

Implementations for tbox::RelaunchableJob
enum	WaitPhase { for_data_only, to_be_launched, reduce_histogram, bcast_acceptability, gather_grouping_criteria, bcast_child_groups, run_children, bcast_to_dropouts, completed, deallocated }
	Names of algorithmic phases while outside of continueAlgorithm(). More...

enum	PhaseTag { reduce_histogram_tag = 0, bcast_acceptability_tag = 0, gather_grouping_criteria_tag = 0, bcast_child_groups_tag = 0, bcast_to_dropouts_tag = 0, total_phase_tags = 1 }
	MPI tags identifying messages. More...

enum	BoxAcceptance { undetermined = -2, hasnotag_by_owner = -1, rejected_by_calculation = 0, accepted_by_calculation = 1, rejected_by_owner = 2, accepted_by_owner = 3, rejected_by_recombination = 4, accepted_by_recombination = 5, rejected_by_dropout_bcast = 6, accepted_by_dropout_bcast = 7 }
	Candidate box acceptance state. More...

virtual void	continueJob ()

virtual JobState	getJobState ()
	Get the current state of the job, as a JobState enumeration. More...

virtual tbox::AsyncCommGroup *	getCommunicationGroup ()
	Get the communication group object that the job uses for communications. More...

WaitPhase	continueAlgorithm ()
	Continue the the BR algorithm. More...

Utilities to help analysis and debugging
const int	d_pos
	Unique id in the binary dendogram. More...

CommonParams *	d_common
	Common parameters shared with descendents and ancestors. More...

tbox::List< tbox::RelaunchableJob * >::Iterator	inRelaunchQueue (tbox::RelaunchableJob *node_ptr) const

bool	inGroup (tbox::Array< int > &group, int rank=-1) const

Detailed Description

template<int DIM>
class SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >

In mesh generation, the BR algorithm can be used to cluster tagged cells into boxes. This algorithm is described in Berger and Rigoutsos, IEEE Trans. on Sys, Man, and Cyber (21)5:1278-1286.

This class implements the BR algorithm to execute in a non-recursive way, in order to improve parallel efficiency over recursive implementations. To facilitate a non-recursive implementation, data in the recursive tree is maintained in a "BR dendogram", nodes of which are instances of this class.

Clarification on the uses of the word "node":

Dendogram node: Node in the BR dendogram (this class).
Graph node: Node in a box graph. The box graph is the form of the outputs of this class. Each output graph node corresponds to a box generated by the BR algorithm.
Processor: MPI process id. This is called a node in some context. For clarity, we avoid this use of "node".

Each dendogram node is associated with a candidate box, an owner process coordinating distributed computations on the box and a group of processors participating in those computations. Should the candidate box be one of the final output boxes, the owner also owns the graph node associated with the box.

To use this class:

Construct an object of type CommonParams.
Construct the root node of the dendogram using the CommonParams object.
Finetune the algorithm settings using the methods under "Algorithm settings".
If needed, various non-algorithmic flags using set...() methods under "Methods for analysis and debugging".
Initiate the algorithm using runAlgorithm() in the job_relauncher parameter in the CommonParams object.
Get the output graph nodes and edges using methods under "Access to outputs".

This class creates its output in a distributed nested-level box-graph (DNBG) format. The output is distributed over all processes running the algorithm, with each process owning a subset of the DNBG. The 2 primary outputs of this implementation are:

Set of graph nodes containing input tags. Each node corresponds to an output box. See getNewNodes() and hier::LayerNodeSet<DIM>.
Connectivity between graph nodes on the tagged level and the new graph nodes. See getNewCnect() and hier::LayerEdgeSet<DIM>.

Member Typedef Documentation

◆ GraphNode

template<int DIM>

typedef hier::LayerNodeSet<DIM>::Node SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::GraphNode

◆ GraphNodeContainer

template<int DIM>

typedef hier::LayerNodeSet<DIM>::NodeContainer SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::GraphNodeContainer

◆ GraphNabrContainer

template<int DIM>

typedef hier::LayerEdgeSet<DIM>::NabrContainer SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::GraphNabrContainer

◆ Connectivity

template<int DIM>

typedef hier::LayerEdgeSet<DIM>::Connectivity SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::Connectivity

◆ IntSet

template<int DIM>

typedef std::set<int> SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::IntSet

Member Enumeration Documentation

◆ OwnerMode

template<int DIM>

enum SAMRAI::mesh::AsyncBergerRigoutsosNode::OwnerMode

Enumerator
SINGLE_OWNER
MOST_OVERLAP
FEWEST_OWNED
LEAST_ACTIVE

◆ WaitPhase

template<int DIM>

enum SAMRAI::mesh::AsyncBergerRigoutsosNode::WaitPhase

private

"For_data_only" phase is when the dendogram node is only used to store data. If the node is to be executed, it enters the "to_be_launched" phase.

All names beginning with "reduce", "gather" or "bcast" refer to communication phases, where control is returned before the algorithm completes.

The "children" phase does not explicitly contain communication, but the children may perform communication.

The "completed" phase is when the algorithm has run to completion. This is where the recursive implementation would return.

The "deallocated" phase is for debugging. This phase is set by the destructor, just to help find dendogram nodes that are deallocated but somehow was referenced.

Enumerator
for_data_only
to_be_launched
reduce_histogram
bcast_acceptability
gather_grouping_criteria
bcast_child_groups
run_children
bcast_to_dropouts
completed
deallocated

◆ PhaseTag

template<int DIM>

enum SAMRAI::mesh::AsyncBergerRigoutsosNode::PhaseTag

private

Each message tag is the d_mpi_tag plus a PhaseTag. Originally, there were different tags for different communication phases, determined by d_mpi_tag plus a PhaseTag. But this is not really needed, so all phases use the tag d_mpi_tag. The PhaseTag type is just here in case we have to go back to using them.

Enumerator
reduce_histogram_tag
bcast_acceptability_tag
gather_grouping_criteria_tag
bcast_child_groups_tag
bcast_to_dropouts_tag
total_phase_tags

◆ BoxAcceptance

template<int DIM>

enum SAMRAI::mesh::AsyncBergerRigoutsosNode::BoxAcceptance

private

Note that accepted values are odd and rejected and undetermined values are even! See boxAccepted(), boxRejected() and boxHasNoTag().

It is not critical to have all values shown, but the values help in debugging.

Meaning of values:

"hasnotag_by_owner": histogram is truly empty (after sum reduction). We don't accept the box, but we don't split it either. (This can only happen at the root dendogram node, as child boxes are guaranteed to have tags.)
"(rejected|accepted)_by_calculation": decision by calculation on the owner process.
"(rejected|accepted)_by_owner": decision by owner process, broadcast to participants.
"(rejected|accepted)_by_recombination": decision by recombination on local process.
"(rejected|accepted)_by_dropout_bcast": decision by participant group, broadcast to the dropout group.

Enumerator
undetermined
hasnotag_by_owner
rejected_by_calculation
accepted_by_calculation
rejected_by_owner
accepted_by_owner
rejected_by_recombination
accepted_by_recombination
rejected_by_dropout_bcast
accepted_by_dropout_bcast

◆ JobState

enum SAMRAI::tbox::RelaunchableJob::JobState

inherited

Enumerator
COMMUNICATION_WAIT
NONCOMMUNICATION_WAIT
JOB_IS_COMPLETED

Constructor & Destructor Documentation

◆ AsyncBergerRigoutsosNode()

template<int DIM>

SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::AsyncBergerRigoutsosNode	(	CommonParams *	common_params,
		const hier::Box< DIM > *	bound_box = `NULL`,
		mesh::AsyncBergerRigoutsosNode< DIM > *	parent = `NULL`,
		const int	child_number = `1`
	)

Construct a node node of a BR dendogram, which you can use to run the BR algorithm.

Parameters

common_params	The common parameter object you plan to use to run the ABR algorithm.
bound_box	Bounding box for tagged cells.
parent	Parent node of the node being constructed. Set to NULL if you are constructing the root node of the BR dendogram.
child_number	0 if constructing left child and 1 if constructing right child. Set to 1 if you are constructing the root node of the BR dendogram.

◆ ~AsyncBergerRigoutsosNode()

template<int DIM>

SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::~AsyncBergerRigoutsosNode ( void )

Deallocate internal data.

Member Function Documentation

◆ setMaxGhostCellWidth()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::setMaxGhostCellWidth ( const hier::IntVector< DIM > & max_gcw )

Overlap checking is done to determine nearest-neighbor relationships when generating connectivity to new graph nodes. If a box grown by this ammount intersects another box, the two boxes are considered neighbors.

By default the max ghost cell width is one in each direction.

◆ setAlgorithmAdvanceMode()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::setAlgorithmAdvanceMode ( const std::string & algo_advance_mode )

Choices are:

"SYNCHRONOUS" --> wait for each communication stage to complete before moving on, thus resulting in synchronous execution.
"ROUND_ROBIN" --> check for completed communication stages in round-robin fashion instead of waiting for a specific one.
"ADVANCE_ANY" --> advance an dendogram node through its communication stage by using tbox::AsyncCommStage::advanceAny().
"ADVANCE_SOME" --> advance an dendogram node through its communication stage by using tbox::AsyncCommStage::advanceSome().

The default is "ADVANCE_SOME".

Asynchronous modes are NOT guaranteed to compute the output graph nodes in any particular order. The order depends on the ordering of message completion, which is not deterministic. If you require consistent outputs, we suggest you have a scheme for reordering the output boxes.

◆ setOwnerMode()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::setOwnerMode ( const std::string & mode )

"MOST_OVERLAP" Ownership is given to the processor with the most overlap on the candidate box. Default.
"SINGLE_OWNER" In single-owner mode, the initial owner (process 0) always participates and owns all dendogram nodes.
"FEWEST_OWNED" Choose the processor that owns the fewest dendogram nodes when the choice is made. This is meant to relieve bottle-necks caused by excessive ownership.
"LEAST_ACTIVE" Choose the processor that participates in the fewest number of dendogram nodes when the choice is made. This is meant to relieve bottle-necks caused by excessive participation.

Experiments show that "MOST_OVERLAP" gives the best clustering speed, while "SINGLE_OWNER" may give a faster output globalization (since you don't need an all-gather).

◆ setUseLevelBoxes()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::setUseLevelBoxes ( bool flag )

If off, the global level boxes will neither be used nor be generated. This feature is in anticipation of future support for the distributed nested-level box graph in SAMRAI.

◆ setComputeEdges()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::setComputeEdges ( int compute_edges )

Valid values to set are:

0 = No edge computation.
1 = Compute directed edges from input to output graph nodes. With this option, it is possible to determine output nodes neighboring any input nodes, but not possible to determine input nodes neighboring a specific output node.
2 = Compute directed edges from input to output graph nodes as well as the reverse. With this option, it is possible to determine output nodes neighboring any input nodes, as well as input nodes neighboring any output node. This is accomplished using an additional edge-sharing communication after all graph nodes have been created.

By default, the value is 2.

◆ runAlgorithm()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::runAlgorithm ( )

◆ getNewNodes()

template<int DIM>

const hier::LayerNodeSet<DIM>& SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::getNewNodes ( ) const

◆ getNewCnect()

template<int DIM>

const Connectivity& SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::getNewCnect ( ) const

The connectivity data generated depend on the flag set using setComputeEdges().

◆ printClassData()

template<int DIM>

virtual void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::printClassData	(	std::ostream &	os,
		int	detail_level = `0`
	)		const

virtual

◆ getMaxNodes()

template<int DIM>

int SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::getMaxNodes ( ) const

◆ getMaxGeneration()

template<int DIM>

int SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::getMaxGeneration ( ) const

◆ getMaxOwnership()

template<int DIM>

int SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::getMaxOwnership ( ) const

◆ getAvgNumberOfCont()

template<int DIM>

double SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::getAvgNumberOfCont ( ) const

◆ getMaxNumberOfCont()

template<int DIM>

int SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::getMaxNumberOfCont ( ) const

◆ getNumBoxesGenerated()

template<int DIM>

int SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::getNumBoxesGenerated ( ) const

◆ setLogNodeHistory()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::setLogNodeHistory ( bool flag )

◆ continueJob()

template<int DIM>

virtual void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::continueJob ( )

privatevirtual

Continue the user-defined job.

Returns: whether job finished.

Implements SAMRAI::tbox::RelaunchableJob.

◆ getJobState()

template<int DIM>

virtual JobState SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::getJobState ( )

privatevirtual

Implements SAMRAI::tbox::RelaunchableJob.

◆ getCommunicationGroup()

template<int DIM>

virtual tbox::AsyncCommGroup* SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::getCommunicationGroup ( )

privatevirtual

Implements SAMRAI::tbox::RelaunchableJob.

◆ continueAlgorithm()

template<int DIM>

WaitPhase SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::continueAlgorithm ( )

private

Parameters for finding boxes are internal. They should be set in the constructor.

In parallel, this the method may return before algorithm is completed. In serial, no communication is done, so the algorithm IS completed when this method returns. The method is completed if it returns WaitPhase::completed. This method may and should be called multiple times as long as the algorithm has not completed.

If this method returns before the algorithm is complete, this object will have put itself on the leaf queue to be checked for completion later.

Returns: The communication phase currently running.

◆ makeLocalTagHistogram()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::makeLocalTagHistogram ( )

private

◆ reduceHistogram_start()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::reduceHistogram_start ( )

private

◆ reduceHistogram_check()

template<int DIM>

bool SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::reduceHistogram_check ( )

private

◆ computeMinimalBoundingBoxForTags()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::computeMinimalBoundingBoxForTags ( )

private

◆ acceptOrSplitBox()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::acceptOrSplitBox ( )

private

◆ broadcastAcceptability_start()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::broadcastAcceptability_start ( )

private

◆ broadcastAcceptability_check()

template<int DIM>

bool SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::broadcastAcceptability_check ( )

private

◆ countOverlapWithLocalPatches()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::countOverlapWithLocalPatches ( )

private

◆ gatherGroupingCriteria_start()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::gatherGroupingCriteria_start ( )

private

◆ gatherGroupingCriteria_check()

template<int DIM>

bool SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::gatherGroupingCriteria_check ( )

private

◆ formChildGroups()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::formChildGroups ( )

private

◆ formChildGroupsUsingLevelBoxes()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::formChildGroupsUsingLevelBoxes ( )

private

◆ broadcastChildGroups_start()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::broadcastChildGroups_start ( )

private

◆ broadcastChildGroups_check()

template<int DIM>

bool SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::broadcastChildGroups_check ( )

private

◆ runChildren_start()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::runChildren_start ( )

private

◆ runChildren_check()

template<int DIM>

bool SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::runChildren_check ( )

private

◆ broadcastToDropouts_start()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::broadcastToDropouts_start ( )

private

◆ broadcastToDropouts_check()

template<int DIM>

bool SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::broadcastToDropouts_check ( )

private

◆ createLayerNode()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::createLayerNode ( )

private

◆ eraseLayerNode()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::eraseLayerNode ( )

private

◆ computeNewGraphEdges()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::computeNewGraphEdges ( )

private

◆ shareNewEdgesWithOwners()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::shareNewEdgesWithOwners ( )

private

◆ findOwnerInGroup()

template<int DIM>

int SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::findOwnerInGroup	(	int	owner,
		const tbox::Array< int > &	group
	)		const

private

◆ claimMPITag()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::claimMPITag ( )

private

◆ computeCommunicationTreeDegree()

template<int DIM>

int SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::computeCommunicationTreeDegree ( int group_size ) const

private

◆ findZeroCutPoint()

template<int DIM>

bool SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::findZeroCutPoint	(	int &	cut_pt,
		const int	dim
	)

private

◆ findZeroCutSwath()

template<int DIM>

bool SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::findZeroCutSwath	(	int &	cut_lo,
		int &	cut_hi,
		const int	dim
	)

private

◆ cutAtLaplacian()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::cutAtLaplacian	(	int &	cut_pt,
		const int	dim,
		const int	lo,
		const int	hi,
		const int	min_size
	)

private

◆ getHistogramBufferSize()

template<int DIM>

int SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::getHistogramBufferSize ( const hier::Box< DIM > & box ) const

private

◆ putHistogramToBuffer()

template<int DIM>

int* SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::putHistogramToBuffer ( int * buffer )

private

◆ getHistogramFromBuffer()

template<int DIM>

int* SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::getHistogramFromBuffer ( int * buffer )

private

◆ putBoxToBuffer()

template<int DIM>

int* SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::putBoxToBuffer	(	const hier::Box< DIM > &	box,
		int *	buffer
	)		const

private

◆ getBoxFromBuffer()

template<int DIM>

int* SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::getBoxFromBuffer	(	hier::Box< DIM > &	box,
		int *	buffer
	)		const

private

◆ computeDropoutGroup()

template<int DIM>

void SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::computeDropoutGroup	(	const tbox::Array< int > &	main_group,
		const tbox::Array< int > &	sub_group,
		tbox::Array< int > &	dropouts,
		const int	add_group
	)		const

private

◆ intToBoxAcceptance()

template<int DIM>

BoxAcceptance SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::intToBoxAcceptance ( int i ) const

private

◆ boxAccepted()

template<int DIM>

bool SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::boxAccepted ( ) const

inlineprivate

◆ boxRejected()

template<int DIM>

bool SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::boxRejected ( ) const

inlineprivate

◆ boxHasNoTag()

template<int DIM>

bool SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::boxHasNoTag ( ) const

inlineprivate

◆ inRelaunchQueue()

template<int DIM>

tbox::List<tbox::RelaunchableJob*>::Iterator SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::inRelaunchQueue ( tbox::RelaunchableJob * node_ptr ) const

private

◆ inGroup()

template<int DIM>

bool SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::inGroup	(	tbox::Array< int > &	group,
		int	rank = `-1`
	)		const

private

Member Data Documentation

◆ d_pos

template<int DIM>

const int SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_pos

private

To have succinct formula, the root dendogram node has d_pos of 1.
Parent id is d_pos/2
Left child id is 2*d_pos
Right child id is 2*d_pos+1
Generation number is ln(d_pos)

This parameter is only used for debugging.

The id of a node grows exponentially with each generation. If the position in the binary tree is too big to be represented by an integer, d_pos is set to -1 for a left child and -2 for a right child.

◆ d_common

template<int DIM>

CommonParams* SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_common

private

Only the root of the tree allocates the common parameters. For all others, this pointer is set by the parent.

◆ d_parent

template<int DIM>

AsyncBergerRigoutsosNode* SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_parent

private

◆ d_lft_child

template<int DIM>

AsyncBergerRigoutsosNode* SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_lft_child

private

◆ d_rht_child

template<int DIM>

AsyncBergerRigoutsosNode* SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_rht_child

private

◆ d_box

template<int DIM>

hier::Box<DIM> SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_box

private

◆ d_owner

template<int DIM>

int SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_owner

private

◆ d_group

template<int DIM>

tbox::Array<int> SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_group

private

◆ d_mpi_tag

template<int DIM>

int SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_mpi_tag

private

The tag is determined by on the process that owns the parent when the parent decides to split its box. The tags are broadcasted along with the children boxes.

◆ d_overlap

template<int DIM>

int SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_overlap

private

◆ d_box_acceptance

template<int DIM>

BoxAcceptance SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_box_acceptance

private

See also: BoxAcceptance.

◆ d_histogram

template<int DIM>

tbox::Array<int> SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_histogram[DIM]

private

If local process is d_owner, this is initially the local histogram, then later, the reduced histogram. If not, it is just the local histogram.

◆ d_node

template<int DIM>

GraphNode SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_node

private

On the owner process, this belongs in a hier::LayerNodeSet<DIM> object. On contributor nodes, this is used to identify the layer node id assigned by the owner. The layer node id is important for computing neighbor data.

◆ d_node_iterator

template<int DIM>

GraphNodeContainer::iterator SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_node_iterator

private

This is relevant only on the owner, where the d_node is in a container. On contributors, the graph node is non-local and stands alone.

◆ d_wait_phase

template<int DIM>

WaitPhase SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_wait_phase

private

◆ d_send_msg

template<int DIM>

tbox::Array<int> SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_send_msg

private

◆ d_recv_msg

template<int DIM>

tbox::Array<int> SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_recv_msg

private

◆ d_comm_group

template<int DIM>

tbox::AsyncCommGroup* SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_comm_group

private

◆ d_generation

template<int DIM>

const int SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_generation

private

The generation number is the parent's generation number plus 1. The root has generation number 1.

◆ d_n_cont

template<int DIM>

int SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >::d_n_cont

private

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

/samrai/include/AsyncBergerRigoutsosNode.h

Classes

Public Types

Public Member Functions

Private Member Functions

Private Attributes

Implementations for tbox::RelaunchableJob

Utilities to help analysis and debugging

Detailed Description

template<int DIM> class SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >

Member Typedef Documentation

◆ GraphNode

◆ GraphNodeContainer

◆ GraphNabrContainer

◆ Connectivity

◆ IntSet

Member Enumeration Documentation

◆ OwnerMode

◆ WaitPhase

◆ PhaseTag

◆ BoxAcceptance

◆ JobState

Constructor & Destructor Documentation

◆ AsyncBergerRigoutsosNode()

◆ ~AsyncBergerRigoutsosNode()

Member Function Documentation

◆ setMaxGhostCellWidth()

◆ setAlgorithmAdvanceMode()

◆ setOwnerMode()

◆ setUseLevelBoxes()

◆ setComputeEdges()

◆ runAlgorithm()

◆ getNewNodes()

◆ getNewCnect()

◆ printClassData()

◆ getMaxNodes()

◆ getMaxGeneration()

◆ getMaxOwnership()

◆ getAvgNumberOfCont()

◆ getMaxNumberOfCont()

◆ getNumBoxesGenerated()

◆ setLogNodeHistory()

◆ continueJob()

◆ getJobState()

◆ getCommunicationGroup()

◆ continueAlgorithm()

◆ makeLocalTagHistogram()

◆ reduceHistogram_start()

◆ reduceHistogram_check()

◆ computeMinimalBoundingBoxForTags()

◆ acceptOrSplitBox()

◆ broadcastAcceptability_start()

◆ broadcastAcceptability_check()

◆ countOverlapWithLocalPatches()

◆ gatherGroupingCriteria_start()

◆ gatherGroupingCriteria_check()

◆ formChildGroups()

◆ formChildGroupsUsingLevelBoxes()

◆ broadcastChildGroups_start()

◆ broadcastChildGroups_check()

◆ runChildren_start()

◆ runChildren_check()

◆ broadcastToDropouts_start()

◆ broadcastToDropouts_check()

◆ createLayerNode()

◆ eraseLayerNode()

◆ computeNewGraphEdges()

◆ shareNewEdgesWithOwners()

◆ findOwnerInGroup()

◆ claimMPITag()

◆ computeCommunicationTreeDegree()

◆ findZeroCutPoint()

◆ findZeroCutSwath()

◆ cutAtLaplacian()

◆ getHistogramBufferSize()

◆ putHistogramToBuffer()

◆ getHistogramFromBuffer()

◆ putBoxToBuffer()

◆ getBoxFromBuffer()

◆ computeDropoutGroup()

◆ intToBoxAcceptance()

template<int DIM>
class SAMRAI::mesh::AsyncBergerRigoutsosNode< DIM >