Dune::UGGrid< dim > Class Template Reference

Front-end for the grid manager of the finite element toolbox UG3. More...

#include <dune/grid/uggrid.hh>

Public Types
typedef UGGridFamily< dim >	GridFamily
	type of the used GridFamily for this grid
typedef UG::DOUBLE	ctype
	The type used to store coordinates.
typedef unsigned int	Rank
	The type used for process ranks.
Exported constants
enum
	A constant that exports the template parameter dim.
enum
	A constant that exports the template parameter dimworld.
Exported types
typedef GridFamily::Traits::LeafGridView	LeafGridView
	type of view for leaf grid
typedef GridFamily::Traits::LevelGridView	LevelGridView
	type of view for level grid
typedef GridFamily::Traits::LeafIntersection	LeafIntersection
	A type that is a model of Dune::Intersection, an intersections of two codimension 1 of two codimension 0 entities in the leaf view.
typedef GridFamily::Traits::LevelIntersection	LevelIntersection
	A type that is a model of Dune::Intersection, an intersections of two codimension 1 of two codimension 0 entities in a level view.
typedef GridFamily::Traits::LeafIntersectionIterator	LeafIntersectionIterator
	A type that is a model of Dune::IntersectionIterator which is an iterator that allows to examine, but not to modify, the intersections of codimension 1 of an leaf element (entity of codimension 0) with other leaf elements.
typedef GridFamily::Traits::LevelIntersectionIterator	LevelIntersectionIterator
	A type that is a model of Dune::IntersectionIterator which is an iterator that allows to examine, but not to modify, the intersections of codimension 1 of an element (entity of codimension 0) with other elements on the same level.
typedef GridFamily::Traits::HierarchicIterator	HierarchicIterator
	A type that is a model of Dune::HierarchicIterator A type of iterator that allows to examine, but not to modify, entities of codimension 0 that result from refinement of an entity of codimension 0.
typedef GridFamily::Traits::LevelIndexSet	LevelIndexSet
	A type that is a model of Dune::IndexSet which provides a consecutive, but non persistent, numbering for entities on a grid level.
typedef GridFamily::Traits::LeafIndexSet	LeafIndexSet
	A type that is a model of Dune::IndexSet which provides a consecutive, but non persistent, numbering for entities in the leaf grid.
typedef GridFamily::Traits::GlobalIdSet	GlobalIdSet
	A type that is a model of Dune::IdSet which provides a unique and persistent numbering for all entities in the grid. The numbering is unique over all processes over which the grid is partitioned. The numbering is not necessarily consecutive.
typedef GridFamily::Traits::LocalIdSet	LocalIdSet
	A type that is a model of Dune::IdSet which provides a unique and persistent numbering for all entities in the grid. The numbering is only unique in a single process and it is not necessarily consecutive.
typedef GridFamily::Traits::CollectiveCommunication	CollectiveCommunication
	A type that is a model of Dune::CollectiveCommunication. It provides a portable way for collective communication on the set of processes used by the grid.

Public Member Functions
	UGGrid (UGCollectiveCommunication comm={})
	Default constructor.
	~UGGrid () noexcept(false)
	Destructor.
int	maxLevel () const
template<typename Seed >
Traits::template Codim< Seed::codimension >::Entity	entity (const Seed &seed) const
	Create an Entity from an EntitySeed.
int	size (int level, int codim) const
	Number of grid entities per level and codim.
int	size (int codim) const
	number of leaf entities per codim in this process
int	size (int level, GeometryType type) const
	number of entities per level and geometry type in this process
int	size (GeometryType type) const
	number of leaf entities per geometry type in this process
size_t	numBoundarySegments () const
	Return the number of boundary segments.
const Traits::GlobalIdSet &	globalIdSet () const
	Access to the GlobalIdSet.
const Traits::LocalIdSet &	localIdSet () const
	Access to the LocalIdSet.
const Traits::LevelIndexSet &	levelIndexSet (int level) const
	Access to the LevelIndexSets.
const Traits::LeafIndexSet &	leafIndexSet () const
	Access to the LeafIndexSet.
Traits::LevelGridView	levelGridView (int level) const
	View for a grid level for All_Partition.
Traits::LeafGridView	leafGridView () const
	View for the leaf grid for All_Partition.
bool	mark (int refCount, const typename Traits ::template Codim< 0 >::Entity &e)
	Marks an entity to be refined/coarsened in a subsequent adapt. More...
bool	loadBalance ()
	default implementation of load balance does nothing and returns false
Adaptivity and grid refinement
bool	mark (int refCount, const typename Codim< 0 >::Entity &e)
	Marks an entity to be refined/coarsened in a subsequent adapt. More...
int	getMark (const typename Codim< 0 >::Entity &e) const
	returns adaptation mark for given entity More...

Grid Refinement Methods
enum	RefinementType { LOCAL , COPY }
	The different forms of grid refinement that UG supports. More...
enum	ClosureType { GREEN , NONE }
	Decide whether to add a green closure to locally refined grid sections or not. More...
bool	mark (int refCount, const typename Traits::template Codim< 0 >::Entity &e)
	Mark element for refinement. More...
bool	mark (const typename Traits::template Codim< 0 >::Entity &e, typename UG_NS< dim >::RefinementRule rule, int side=0)
	Mark method accepting a UG refinement rule. More...
int	getMark (const typename Traits::template Codim< 0 >::Entity &e) const
	Query whether element is marked for refinement.
bool	preAdapt ()
	returns true, if some elements might be coarsend during grid adaption, here always returns true
bool	adapt ()
	Triggers the grid refinement process.
void	postAdapt ()
	Clean up refinement markers.
template<class DataHandle >
bool	loadBalance (DataHandle &dataHandle)
	Distributes the grid and some data over the available nodes in a distributed machine. More...
bool	loadBalance (int minlevel=0)
	Distributes this grid over the available nodes in a distributed machine. More...
bool	loadBalance (const std::vector< Rank > &targetProcessors, unsigned int fromLevel)
	Distribute this grid over a distributed machine. More...
template<class DataHandle >
bool	loadBalance (const std::vector< Rank > &targetProcessors, unsigned int fromLevel, DataHandle &dataHandle)
	Distributes the grid over the processes of a parallel machine, and sends data along with it. More...
const UGCollectiveCommunication &	comm () const
void	getChildrenOfSubface (const typename Traits::template Codim< 0 >::Entity &e, int elementSide, int maxl, std::vector< typename Traits::template Codim< 0 >::Entity > &childElements, std::vector< unsigned char > &childElementSides) const
	Rudimentary substitute for a hierarchic iterator on faces. More...
void	setRefinementType (RefinementType type)
	Sets the type of grid refinement.
void	setClosureType (ClosureType type)
	Sets the type of grid refinement closure.
void	setPosition (const typename Traits::template Codim< dim >::Entity &e, const FieldVector< double, dim > &pos)
	Sets a vertex to a new position. More...
void	globalRefine (int n)
	Does uniform refinement. More...
void	saveState (const std::string &filename) const
	Save entire grid hierarchy to disk. More...
void	loadState (const std::string &filename)
	Read entire grid hierarchy from disk. More...

Parallel data distribution and communication
Codim< EntitySeed::codimension >::Entity	entity (const EntitySeed &seed) const
	obtain Entity from EntitySeed.
GridImp &	asImp ()
	Barton-Nackman trick.
const GridImp &	asImp () const
	Barton-Nackman trick.

Detailed Description

template<int dim>
class Dune::UGGrid< dim >

Front-end for the grid manager of the finite element toolbox UG3.

This is the implementation of the grid interface using the UG3 grid management system. It is best described in this paper. To our knowledge, the original code is not available anymore, but the relevant parts have been forked into the Dune module dune-uggrid, available from .

UGGrid provides conforming grids in two and three space dimensions. The grids can be mixed, i.e. 2d grids can contain triangles and quadrilaterals and 3d grids can contain tetrahedra and hexahedra and also pyramids and prisms. The grid refinement rules are very flexible. Local adaptive red/green refinement is the default, but a special method in the UGGrid class allows you to directly access a number of anisotropic refinements rules. Last but not least, the UG grid manager is completely parallelized, and you can use boundaries parametrized by either analytical expressions or high-resolution piecewise linear surfaces.

In your Dune application, you can instantiate objects of the type UGGrid<2> or UGGrid<3>. You can have more than one, if you choose. It is even possible to have 2d and 3d grids at the same time, even though the original UG system never intended to support this!

See the documentation for the factory class GridFactory<UGGrid<dimworld> > to learn how to create UGGrid objects.

Member Enumeration Documentation

ClosureType

template<int dim>

enum Dune::UGGrid::ClosureType

Decide whether to add a green closure to locally refined grid sections or not.

Enumerator
GREEN	Standard red/green refinement.
NONE	No closure, results in nonconforming meshes.

RefinementType

template<int dim>

enum Dune::UGGrid::RefinementType

The different forms of grid refinement that UG supports.

Enumerator
LOCAL	New level consists only of the refined elements and the closure.
COPY	New level consists of the refined elements and the unrefined ones, too.

Member Function Documentation

comm()

template<int dim>

const UGCollectiveCommunication & Dune::UGGrid< dim >::comm ( ) const

inline

the collective communication

getChildrenOfSubface()

template<int dim>

void Dune::UGGrid< dim >::getChildrenOfSubface	(	const typename Traits::template Codim< 0 >::Entity &	e,
		int	elementSide,
		int	maxl,
		std::vector< typename Traits::template Codim< 0 >::Entity > &	childElements,
		std::vector< unsigned char > &	childElementSides
	)		const

Rudimentary substitute for a hierarchic iterator on faces.

Parameters

	e,elementSide	Grid face specified by an element and one of its sides
	maxl	The finest level that should be traversed by the iterator
[out]	childElements	For each subface: element index, elementSide, and level
[out]	childElementSides	Indices for transformation because Dune numbers the faces of several elements differently than UG

getMark()

int Dune::Grid< dim, dimworld, double , UGGridFamily< dim > >::getMark ( const typename Codim< 0 >::Entity &  e ) const

inlineinherited

returns adaptation mark for given entity

Parameters

[in]

e

Entity for which adaptation mark should be determined

Returns

int adaptation mark currently set for given Entity e

globalRefine()

template<int dim>

void Dune::UGGrid< dim >::globalRefine

(

int

n

)

Does uniform refinement.

Parameters

n	Number of uniform refinement steps

loadBalance() [1/4]

template<int dim>

bool Dune::UGGrid< dim >::loadBalance	(	const std::vector< Rank > &	targetProcessors,
		unsigned int	fromLevel
	)

Distribute this grid over a distributed machine.

Parameters

[in]	targetProcessors	For each leaf element the rank of the process the element shall be sent to
[in]	fromLevel	The lowest level that gets redistributed (set to 0 when in doubt)

This method allows to (re-)distribute the grid controlled by an external grid repartitioning library. You need to get that library to assign a target rank to each interior element in the leaf grid. With this information in a std::vector, call this method, and UG will do the actual repartitioning. Each leaf element will be sent to the assigned target rank. For all other elements we look at where there children are being sent to. The parent is then sent to where most of its children are (Familienzusammenfuehrung).

The size of the input array targetProcessors is expected to be equal to the number of elements in the 'all'-partition, i.e., the number Interior elements plus the number of Ghost elements. To get the array entry corresponding to an Interior element, a MultipleCodimMultipleGeomTypeMapper with layout class MCMGElementLayout is used. Array entries corresponding to Ghost elements are ignored.

In some cases you may also want to leave the lowest levels on one process, to have them all together for multigrid coarse grid corrections. In that case, use the fromLevel parameter with a value other than zero, to redistribute only elements above a certain level.

The fromLevel argument is also needed to allow the compiler to distinguish this method from the loadBalance method with a single template DataHandle argument.

Note

In theory you can assign a target rank to any element on any level, and UG will magically transfer the element to that rank and make everything come out right. This is not supported by the UGGrid interface, because I didn't see a use case for it. If you do need it please ask on the Dune mailing list.

Returns

true

loadBalance() [2/4]

template<int dim>

template<class DataHandle >

bool Dune::UGGrid< dim >::loadBalance ( const std::vector< Rank > &  targetProcessors, unsigned int  fromLevel, DataHandle &  dataHandle  )

inline

Distributes the grid over the processes of a parallel machine, and sends data along with it.

Parameters

[in]	targetProcessors	For each leaf element the rank of the process the element shall be sent to
[in]	fromLevel	The lowest level that gets redistributed (set to 0 when in doubt)
[in,out]	dataHandle	A data handle object that does the gathering and scattering of data

Template Parameters

DataHandle

works like the data handle for the communicate methods.

Returns

true

References Dune::GridDefaultImplementation< dim, dim, double, UGGridFamily< dim > >::leafGridView(), and Dune::GridDefaultImplementation< dim, dim, double, UGGridFamily< dim > >::loadBalance().

loadBalance() [3/4]

template<int dim>

template<class DataHandle >

bool Dune::UGGrid< dim >::loadBalance ( DataHandle &  dataHandle )

inline

Distributes the grid and some data over the available nodes in a distributed machine.

Template Parameters

DataHandle

works like the data handle for the communicate methods.

Returns

True, if grid has changed, false otherwise

References Dune::GridDefaultImplementation< dim, dim, double, UGGridFamily< dim > >::leafGridView(), and Dune::GridDefaultImplementation< dim, dim, double, UGGridFamily< dim > >::loadBalance().

loadBalance() [4/4]

template<int dim>

bool Dune::UGGrid< dim >::loadBalance

(

int

minlevel = 0

)

Distributes this grid over the available nodes in a distributed machine.

Bug:

The return value is always 'true'

Parameters

minlevel

The coarsest grid level that gets distributed

loadState()

template<int dim>

void Dune::UGGrid< dim >::loadState

(

const std::string &

filename

)

Read entire grid hierarchy from disk.

Test implementation – not working!

mark() [1/4]

template<int dim>

bool Dune::UGGrid< dim >::mark	(	const typename Traits::template Codim< 0 >::Entity &	e,
		typename UG_NS< dim >::RefinementRule	rule,
		int	side = 0
	)

Mark method accepting a UG refinement rule.

Parameters

e	element to be marked for refinement
rule	One of the UG refinement rules
side	If rule==UG::D2::BLUE (one quadrilateral is split into two rectangles) you can choose the orientation of the cut by setting side==0 or side==1

The available values for RefinementRule are: (see the RefinementRule enum in ug/gm/gm.h)

2D

• NO_REFINEMENT
• COPY
• RED
• BLUE
• COARSE
• BISECTION_1
• BISECTION_2_Q
• BISECTION_2_T1
• BISECTION_2_T2
• BISECTION_3

3D

• NO_REFINEMENT
• COPY
• RED
• COARSE
• TETRA_RED_HEX
• PRISM_BISECT_1_2
• PRISM_QUADSECT
• PRISM_BISECT_HEX0
• PRISM_BISECT_HEX1
• PRISM_BISECT_HEX2
• PRISM_ROTATE_LEFT
• PRISM_ROTATE_RGHT
• PRISM_QUADSECT_HEXPRI0
• PRISM_RED_HEX
• PRISM_BISECT_0_1
• PRISM_BISECT_0_2
• PRISM_BISECT_0_3
• HEX_BISECT_0_1
• HEX_BISECT_0_2
• HEX_BISECT_0_3
• HEX_TRISECT_0
• HEX_TRISECT_5
• HEX_QUADSECT_0
• HEX_QUADSECT_1
• HEX_QUADSECT_2
• HEX_BISECT_HEXPRI0
• HEX_BISECT_HEXPRI1

mark() [2/4]

bool Dune::Grid< dim, dimworld, double , UGGridFamily< dim > >::mark ( int  refCount, const typename Codim< 0 >::Entity &  e  )

inlineinherited

Marks an entity to be refined/coarsened in a subsequent adapt.

Parameters

[in]	refCount	Number of subdivisions that should be applied. Negative value means coarsening.
[in]	e	Entity that should be marked

Returns

true if Entity was marked, false otherwise.

mark() [3/4]

bool Dune::GridDefaultImplementation< dim, dimworld, double , UGGridFamily< dim > >::mark ( int  refCount, const typename Traits ::template Codim< 0 >::Entity &  e  )

inlineinherited

Marks an entity to be refined/coarsened in a subsequent adapt.

Parameters

[in]	refCount	Number of subdivisions that should be applied. Negative value means coarsening.
[in]	e	Entity to Entity that should be refined

Returns

true if Entity was marked, false otherwise.

Note

• default implementation is: return false; for grids with no adaptation.
• for the grid programmer: this method is implemented as a template method, because the Entity type is not defined when the class is instantiated You won't need this trick in the implementation. In your implementation you should use it as

  bool mark( int refCount,
             typename Traits::template Codim<0>::Entity & e ).

  This template method will vanish due to the inheritance rules.

mark() [4/4]

template<int dim>

bool Dune::UGGrid< dim >::mark	(	int	refCount,
		const typename Traits::template Codim< 0 >::Entity &	e
	)

Mark element for refinement.

Parameters

refCount	1: mark for red refinement -1: mark for coarsening 0: delete a possible refinement mark
e	Element to be marked

Returns

• true, if element was marked
• false, if nothing changed

maxLevel()

template<int dim>

int Dune::UGGrid< dim >::maxLevel

(

)

const

Return maximum level defined in this grid. Levels are numbered 0 ... maxlevel with 0 the coarsest level.

Referenced by Dune::UGGrid< dim >::levelIndexSet().

saveState()

template<int dim>

void Dune::UGGrid< dim >::saveState

(

const std::string &

filename

)

const

Save entire grid hierarchy to disk.

Test implementation – not working!

setPosition()

template<int dim>

void Dune::UGGrid< dim >::setPosition	(	const typename Traits::template Codim< dim >::Entity &	e,
		const FieldVector< double, dim > &	pos
	)

Sets a vertex to a new position.

Changing a vertex' position changes its position on all grid levels!

---

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

• dune/grid/uggrid.hh