The overlapping_layout class contains the algorithms related to managing overlapping parallel index sets, such as the indexing of the nodes, faces or edges of a distributed mesh. More...

#include <SF_parallel_layout.h>

Inheritance diagram for SF::overlapping_layout< T >:

Collaboration diagram for SF::overlapping_layout< T >:

Public Member Functions
	overlapping_layout ()
	Non-parameterized constructor. Use assign() to initialize later. More...

void	assign (const vector< T > &idx, MPI_Comm comm)
	Initialization function. More...

const vector< T > &	interface () const
	Retrieve the local indices of the subdomain interfaces. More...

const vector< T > &	algebraic_nodes () const
	Getter function for the local indices forming the local algebraic node set. More...

void	permute_algebraic_nodes (const vector< T > &perm_nod)
	Reorder the local algebraic node set. More...

const vector< T > &	algebraic_layout () const
	Getter function for the global algebraic node layout. More...

const vector< T > &	layout () const
	Return the the overlapping layout. More...

T	localize_algebraic (const T global_idx, const vector< T > &global_alg_nbr, const int rank)
	map a global (REF_NBR, reference numbering) index to local algebraic (non-overlapping) indexing w.r.t. the provided global_alg_nbr numbering (usually PETSC_NBR, PETSc numbering) More...

size_t	num_global_idx () const
	Retrieve the global number of indices. More...

size_t	num_local_idx () const
	Retrieve the local number of indices. More...

size_t	num_algebraic_idx () const
	Retrieve the number of local algebraic indices. More...

template<class V >
void	reduce (vector< V > &ndat, const char *op) const
	Compute a reduction on overlapping data. More...

Public Member Functions inherited from SF::parallel_layout< T >
void	globalize (vector< T > &lvec) const
	Globalize local indices. More...

T	globalize (const T lidx) const
	Globalize local indices. More...

void	localize (vector< T > &gvec) const
	Localize global indices. More...

template<class V >
void	localize (vector< T > &gidx, vector< V > &gdat) const
	Localize global indices and associated data. More...

T	localize (T gidx) const

void	assign (const vector< T > &idx)
	Assign a parallel distributed index set that defines the parallel layout. More...

Additional Inherited Members
Protected Attributes inherited from SF::parallel_layout< T >
vector< T >	_l2g
	The global indices of the local DD domain. Also serves as the local-to-global map. More...

hashmap::unordered_map< T, T >	_g2l
	The global-to-local map for the DD domain. More...

Detailed Description

template<class T>
class SF::overlapping_layout< T >

The overlapping_layout class contains the algorithms related to managing overlapping parallel index sets, such as the indexing of the nodes, faces or edges of a distributed mesh.

Definition at line 363 of file SF_parallel_layout.h.

Constructor & Destructor Documentation

◆ overlapping_layout()

template<class T >

SF::overlapping_layout< T >::overlapping_layout ( )

inline

Non-parameterized constructor. Use assign() to initialize later.

Definition at line 607 of file SF_parallel_layout.h.

Member Function Documentation

◆ algebraic_layout()

template<class T >

const vector<T>& SF::overlapping_layout< T >::algebraic_layout ( ) const

inline

Getter function for the global algebraic node layout.

Returns: A vector holding the displacements between the local algebraic sizes.

Definition at line 685 of file SF_parallel_layout.h.

Here is the caller graph for this function:

◆ algebraic_nodes()

template<class T >

const vector<T>& SF::overlapping_layout< T >::algebraic_nodes ( ) const

inline

Getter function for the local indices forming the local algebraic node set.

The list is ascending by default, but permute_algebraic_nodes() may reorder it to express an alternative algebraic numbering, so callers must not rely on it being sorted. What always holds is that position i corresponds to the i-th local slot of the PETSc vectors built from this layout.

Returns: A vector holding local indices w.r.t. the DD domain.

Definition at line 659 of file SF_parallel_layout.h.

Here is the caller graph for this function:

◆ assign()

template<class T >

void SF::overlapping_layout< T >::assign	(	const vector< T > &	idx,
		MPI_Comm	comm
	)

inline

Initialization function.

It generates the parallel layout from the overlapping sets global node indices forming the different local domains.

Parameters

idx	Vector of global indices forming the local DD domain. Usually the reference numbering should be used.
comm	The MPI communicator for parallel communication.

Definition at line 619 of file SF_parallel_layout.h.

Here is the caller graph for this function:

◆ interface()

template<class T >

const vector<T>& SF::overlapping_layout< T >::interface ( ) const

inline

Retrieve the local indices of the subdomain interfaces.

The local indices are with respect to the global index set "nod" used during initialization.

Returns: The interface indices.

Definition at line 644 of file SF_parallel_layout.h.

◆ layout()

template<class T >

const vector<T>& SF::overlapping_layout< T >::layout ( ) const

inline

Return the the overlapping layout.

Returns: A vector holding the displacements between the local overlapping sizes.

Definition at line 694 of file SF_parallel_layout.h.

◆ localize_algebraic()

template<class T >

T SF::overlapping_layout< T >::localize_algebraic	(	const T	global_idx,
		const vector< T > &	global_alg_nbr,
		const int	rank
	)

inline

map a global (REF_NBR, reference numbering) index to local algebraic (non-overlapping) indexing w.r.t. the provided global_alg_nbr numbering (usually PETSC_NBR, PETSc numbering)

Parameters

global_idx	The global index to localize
global_alg_nbr	The numbering w.r.t. we localize
rank	Process rank

Returns: Either the local index, or -1 if the global index is not on the local partition

Definition at line 709 of file SF_parallel_layout.h.

◆ num_algebraic_idx()

template<class T >

size_t SF::overlapping_layout< T >::num_algebraic_idx ( ) const

inline

Retrieve the number of local algebraic indices.

Definition at line 735 of file SF_parallel_layout.h.

◆ num_global_idx()

template<class T >

size_t SF::overlapping_layout< T >::num_global_idx ( ) const

inline

Retrieve the global number of indices.

Definition at line 725 of file SF_parallel_layout.h.

◆ num_local_idx()

template<class T >

size_t SF::overlapping_layout< T >::num_local_idx ( ) const

inline

Retrieve the local number of indices.

Definition at line 730 of file SF_parallel_layout.h.

Here is the caller graph for this function:

◆ permute_algebraic_nodes()

template<class T >

void SF::overlapping_layout< T >::permute_algebraic_nodes ( const vector< T > & perm_nod )

inline

Reorder the local algebraic node set.

An alternative algebraic ordering is expressed by permuting this list rather than by handing out permuted indices, because the position a node has here is what addresses the local PETSc arrays. Keeping the two in step is what lets a renumbering stay invisible to everything that consumes algebraic_nodes().

Parameters

perm_nod The same node set in the desired order.

Definition at line 674 of file SF_parallel_layout.h.

Here is the caller graph for this function:

◆ reduce()

template<class T >

template<class V >

void SF::overlapping_layout< T >::reduce	(	vector< V > &	ndat,
		const char *	op
	)		const

inline

Compute a reduction on overlapping data.

Template Parameters

V	Value type. Usually int, double or float.

Parameters

ndat	Vector of parallel overlapping data.
op	Reduction operation string. Must be either "min", "max" or "sum".

Definition at line 749 of file SF_parallel_layout.h.

Here is the caller graph for this function:

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

fem/slimfem/src/SF_container.h
fem/slimfem/src/SF_parallel_layout.h

Public Member Functions

Additional Inherited Members

Detailed Description

template<class T> class SF::overlapping_layout< T >

Constructor & Destructor Documentation

◆ overlapping_layout()

Member Function Documentation

◆ algebraic_layout()

◆ algebraic_nodes()

◆ assign()

◆ interface()

◆ layout()

◆ localize_algebraic()

◆ num_algebraic_idx()

◆ num_global_idx()

◆ num_local_idx()

◆ permute_algebraic_nodes()

◆ reduce()

template<class T>
class SF::overlapping_layout< T >