dune-grid  2.7.0
Public Types | Public Member Functions | Protected Member Functions | List of all members
Dune::UGGrid< dim > Class Template Reference

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

#include <dune/grid/uggrid.hh>

Inheritance diagram for Dune::UGGrid< dim >:
Inheritance graph

Public Types

typedef UGGridFamily< dim > GridFamily
 type of the used GridFamily for this grid More...
 
typedef UGGridFamily< dim >::Traits Traits
 
typedef UG::DOUBLE ctype
 The type used to store coordinates. More...
 
typedef unsigned int Rank
 The type used for process ranks. More...
 

Public Member Functions

 UGGrid (UGCollectiveCommunication comm={})
 Default constructor. More...
 
 ~UGGrid () noexcept(false)
 Destructor. More...
 
int maxLevel () const
 
template<typename Seed >
Traits::template Codim< Seed::codimension >::Entity entity (const Seed &seed) const
 Create an Entity from an EntitySeed. More...
 
int size (int level, int codim) const
 Number of grid entities per level and codim. More...
 
int size (int codim) const
 number of leaf entities per codim in this process More...
 
int size (int level, GeometryType type) const
 number of entities per level and geometry type in this process More...
 
int size (GeometryType type) const
 number of leaf entities per geometry type in this process More...
 
size_t numBoundarySegments () const
 Return the number of boundary segments. More...
 
const Traits::GlobalIdSetglobalIdSet () const
 Access to the GlobalIdSet. More...
 
const Traits::LocalIdSetlocalIdSet () const
 Access to the LocalIdSet. More...
 
const Traits::LevelIndexSetlevelIndexSet (int level) const
 Access to the LevelIndexSets. More...
 
const Traits::LeafIndexSetleafIndexSet () const
 Access to the LeafIndexSet. More...
 
Traits::LevelGridView levelGridView (int level) const
 View for a grid level for All_Partition. More...
 
Traits::LeafGridView leafGridView () const
 View for the leaf grid for All_Partition. More...
 
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 More...
 

Protected Member Functions

 DUNE_DEPRECATED_MSG ("use the facade class' `impl()` method instead") static typename std
 return real implementation of interface class 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. More...
 
bool preAdapt ()
 returns true, if some elements might be coarsend during grid adaption, here always returns true More...
 
bool adapt ()
 Triggers the grid refinement process. More...
 
void postAdapt ()
 Clean up refinement markers. More...
 
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 UGCollectiveCommunicationcomm () 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. More...
 
void setClosureType (ClosureType type)
 Sets the type of grid refinement closure. More...
 
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...
 
static void setDefaultHeapSize (unsigned size) DUNE_DEPRECATED_MSG("Do not set the UGGrid default heap size---it is ignored anyway!")
 Sets the default heap size. More...
 

Detailed Description

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

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

This is the implementation of the grid interface using the UG grid management system (http://www.iwr.uni-heidelberg.de/frame/iwrwikiequipment/software/ug). UG 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.

To use this module you need the UG library. See the DUNE installation notes on how to obtain and install it.

In your Dune application, you can now 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.

Please send any questions, suggestions, or bug reports to the Dune mailing list dune@.nosp@m.list.nosp@m.s.dun.nosp@m.e-pr.nosp@m.oject.nosp@m..org

For installation instructions see http://www.dune-project.org/external_libraries/install_ug.html .

Member Typedef Documentation

◆ ctype

template<int dim>
typedef UG::DOUBLE Dune::UGGrid< dim >::ctype

The type used to store coordinates.

◆ GridFamily

template<int dim>
typedef UGGridFamily<dim> Dune::UGGrid< dim >::GridFamily

type of the used GridFamily for this grid

◆ Rank

template<int dim>
typedef unsigned int Dune::UGGrid< dim >::Rank

The type used for process ranks.

◆ Traits

template<int dim>
typedef UGGridFamily<dim>::Traits Dune::UGGrid< dim >::Traits

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.

Constructor & Destructor Documentation

◆ UGGrid()

template<int dim>
Dune::UGGrid< dim >::UGGrid ( UGCollectiveCommunication  comm = {})

Default constructor.

◆ ~UGGrid()

template<int dim>
Dune::UGGrid< dim >::~UGGrid ( )
noexcept

Destructor.

Member Function Documentation

◆ adapt()

template<int dim>
bool Dune::UGGrid< dim >::adapt ( )

Triggers the grid refinement process.

◆ comm()

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

the collective communication

◆ DUNE_DEPRECATED_MSG()

Dune::GridDefaultImplementation< dim, dimworld, double , UGGridFamily< dim > >::DUNE_DEPRECATED_MSG ( "use the facade class' `impl()` method instead"  )
inlineprotectedinherited

return real implementation of interface class

◆ entity()

template<int dim>
template<typename Seed >
Traits::template Codim<Seed::codimension>::Entity Dune::UGGrid< dim >::entity ( const Seed &  seed) const
inline

Create an Entity from an EntitySeed.

◆ 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,elementSideGrid face specified by an element and one of its sides
maxlThe finest level that should be traversed by the iterator
[out]childElementsFor each subface: element index, elementSide, and level
[out]childElementSidesIndices for transformation because Dune numbers the faces of several elements differently than UG

◆ getMark()

template<int dim>
int Dune::UGGrid< dim >::getMark ( const typename Traits::template Codim< 0 >::Entity e) const

Query whether element is marked for refinement.

◆ globalIdSet()

template<int dim>
const Traits::GlobalIdSet& Dune::UGGrid< dim >::globalIdSet ( ) const
inline

Access to the GlobalIdSet.

◆ globalRefine()

template<int dim>
void Dune::UGGrid< dim >::globalRefine ( int  n)

Does uniform refinement.

Parameters
nNumber of uniform refinement steps

◆ leafGridView()

Traits::LeafGridView Dune::GridDefaultImplementation< dim, dimworld, double , UGGridFamily< dim > >::leafGridView
inlineinherited

View for the leaf grid for All_Partition.

◆ leafIndexSet()

template<int dim>
const Traits::LeafIndexSet& Dune::UGGrid< dim >::leafIndexSet ( ) const
inline

Access to the LeafIndexSet.

◆ levelGridView()

Traits::LevelGridView Dune::GridDefaultImplementation< dim, dimworld, double , UGGridFamily< dim > >::levelGridView ( int  level) const
inlineinherited

View for a grid level for All_Partition.

◆ levelIndexSet()

template<int dim>
const Traits::LevelIndexSet& Dune::UGGrid< dim >::levelIndexSet ( int  level) const
inline

Access to the LevelIndexSets.

◆ loadBalance() [1/5]

bool Dune::GridDefaultImplementation< dim, dimworld, double , UGGridFamily< dim > >::loadBalance
inlineinherited

default implementation of load balance does nothing and returns false

◆ loadBalance() [2/5]

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]targetProcessorsFor each leaf element the rank of the process the element shall be sent to
[in]fromLevelThe 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() [3/5]

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]targetProcessorsFor each leaf element the rank of the process the element shall be sent to
[in]fromLevelThe lowest level that gets redistributed (set to 0 when in doubt)
[in,out]dataHandleA data handle object that does the gathering and scattering of data
Template Parameters
DataHandleworks like the data handle for the communicate methods.
Returns
true

◆ loadBalance() [4/5]

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
DataHandleworks like the data handle for the communicate methods.
Returns
True, if grid has changed, false otherwise

◆ loadBalance() [5/5]

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
minlevelThe 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!

◆ localIdSet()

template<int dim>
const Traits::LocalIdSet& Dune::UGGrid< dim >::localIdSet ( ) const
inline

Access to the LocalIdSet.

◆ mark() [1/3]

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
eelement to be marked for refinement
ruleOne of the UG refinement rules
sideIf 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() [2/3]

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]refCountNumber of subdivisions that should be applied. Negative value means coarsening.
[in]eEntity 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() [3/3]

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
eElement 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.

◆ numBoundarySegments()

template<int dim>
size_t Dune::UGGrid< dim >::numBoundarySegments ( ) const
inline

Return the number of boundary segments.

◆ postAdapt()

template<int dim>
void Dune::UGGrid< dim >::postAdapt ( )

Clean up refinement markers.

◆ preAdapt()

template<int dim>
bool Dune::UGGrid< dim >::preAdapt ( )

returns true, if some elements might be coarsend during grid adaption, here always returns true

◆ saveState()

template<int dim>
void Dune::UGGrid< dim >::saveState ( const std::string &  filename) const

Save entire grid hierarchy to disk.

Test implementation – not working!

◆ setClosureType()

template<int dim>
void Dune::UGGrid< dim >::setClosureType ( ClosureType  type)
inline

Sets the type of grid refinement closure.

◆ setDefaultHeapSize()

template<int dim>
static void Dune::UGGrid< dim >::setDefaultHeapSize ( unsigned  size)
inlinestatic

Sets the default heap size.

UGGrid keeps an internal heap to allocate memory from, which must be specified on grid creation (at the latest). This sets the default heap size, which is used when no heap size is given to the constructor.

◆ 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!

◆ setRefinementType()

template<int dim>
void Dune::UGGrid< dim >::setRefinementType ( RefinementType  type)
inline

Sets the type of grid refinement.

◆ size() [1/4]

template<int dim>
int Dune::UGGrid< dim >::size ( GeometryType  type) const
inline

number of leaf entities per geometry type in this process

◆ size() [2/4]

template<int dim>
int Dune::UGGrid< dim >::size ( int  codim) const
inline

number of leaf entities per codim in this process

◆ size() [3/4]

template<int dim>
int Dune::UGGrid< dim >::size ( int  level,
GeometryType  type 
) const
inline

number of entities per level and geometry type in this process

◆ size() [4/4]

template<int dim>
int Dune::UGGrid< dim >::size ( int  level,
int  codim 
) const

Number of grid entities per level and codim.


The documentation for this class was generated from the following file:
Dune::UGGrid::mark
bool mark(int refCount, const typename Traits::template Codim< 0 >::Entity &e)
Mark element for refinement.