A temporary data structure for matrix in COO format. More...

#include <lf/assemble/coomatrix.h>

Public Types
using	Scalar = SCALAR

using	Triplet = Eigen::Triplet<SCALAR>

using	TripletVec = std::vector<Triplet>

using	Index = Eigen::Index

Public Member Functions
	COOMatrix (size_type num_rows, size_type num_cols)

	COOMatrix (const COOMatrix &)=default

	COOMatrix (COOMatrix &&) noexcept=default

COOMatrix &	operator= (const COOMatrix &)=default

COOMatrix &	operator= (COOMatrix &&) noexcept=default

	~COOMatrix ()=default

Index	rows () const
	return number of rows

Index	cols () const
	return number of column

void	AddToEntry (gdof_idx_t i, gdof_idx_t j, SCALAR increment)
	Add a value to the specified entry.

void	setZero ()
	Erase all entries of the matrix.

template<typename PREDICATE >
void	setZero (PREDICATE &&pred)
	Erase specific entries of the COO matrix, that is, set them to zero.

TripletVec &	triplets ()
	Gives access to the vector of triplets.

const TripletVec &	triplets () const

template<typename VECTOR >
Eigen::Matrix< SCALAR, Eigen::Dynamic, 1 >	MatVecMult (SCALAR alpha, const VECTOR &vec) const
	Computes the product of a (scaled) vector with the matrix in COO format.

template<typename VECTOR , typename RESULTVECTOR >
void	MatVecMult (SCALAR alpha, const VECTOR &vec, RESULTVECTOR &resvec) const
	In-situ computes the product of a vector with the matrix in COO format.

Eigen::SparseMatrix< Scalar >	makeSparse () const
	Create an Eigen::SparseMatrix out of the COO format.

Eigen::Matrix< Scalar, Eigen::Dynamic, Eigen::Dynamic >	makeDense () const
	Create an Eigen::MatrixX from the COO format.

void	PrintInfo (std::ostream &o) const
	Ouput of triplet list describing COO matrix.

Private Attributes
size_type	rows_

size_type	cols_

TripletVec	triplets_ {}

Friends
template<typename SCALARTYPE >
std::ostream &	operator<< (std::ostream &o, const COOMatrix< SCALARTYPE > &mat)
	Output operator for COO matrix.

Detailed Description

template<typename SCALAR>
class lf::assemble::COOMatrix< SCALAR >

A temporary data structure for matrix in COO format.

Template Parameters

SCALAR basic scalar type for the matrix

This class provides a container for a matrix in triplet format, also known as COO format see Wikipedia. It essentially manages a vector of Eigen triplets_.

This matrix format allows efficient incremental matrix construction in the context of local assembly.

type requirements for template arguments

SCALAR must be a type that can serve as a scalar type for Eigen::Matrix.

Warning: Objects of this type are no matrix objects in the sense of Eigen. This means you cannot use them instead of an Eigen matrix type, but you have to convert them into a sparse Eigen matrix before by calling COOMatrix::makeSparse().

sample usage:

  // Obtain a shared pointed to a mesh
  std::shared_ptr<lf::mesh::Mesh> mesh_p{
      lf::mesh::test_utils::GenerateHybrid2DTestMesh(0)};
  // Initialization of index mapping for linear finite elements
  lf::assemble::UniformFEDofHandler loc_glob_map(
      mesh_p, {{lf::base::RefEl::kPoint(), 1}});
  // Initialize ENTITYMATRIXPROVIDER object for local computations
  lf::uscalfe::LinearFELaplaceElementMatrix loc_mat_laplace{};
  // Dimension of finite element space
  const lf::assemble::size_type N_dofs(loc_glob_map.NumDofs());
  // Matrix in triplet format holding Galerkin matrix
  lf::assemble::COOMatrix<double> mat(N_dofs, N_dofs);
  // Building the Galerkin matrix (trial space = test space)
  lf::assemble::AssembleMatrixLocally<lf::assemble::COOMatrix<double>>(
      0, loc_glob_map, loc_glob_map, loc_mat_laplace, mat);
  // the `mat` object now contains the Galerkin matrix in triplet/COO format
  // Initialize an Eigen sparse matrix object from `mat`
  Eigen::SparseMatrix<double> stiffness_matrix = {mat.makeSparse()};

Definition at line 52 of file coomatrix.h.

Member Typedef Documentation

◆ Index

template<typename SCALAR >

using lf::assemble::COOMatrix< SCALAR >::Index = Eigen::Index

Definition at line 57 of file coomatrix.h.

◆ Scalar

template<typename SCALAR >

using lf::assemble::COOMatrix< SCALAR >::Scalar = SCALAR

Definition at line 54 of file coomatrix.h.

◆ Triplet

template<typename SCALAR >

using lf::assemble::COOMatrix< SCALAR >::Triplet = Eigen::Triplet<SCALAR>

Definition at line 55 of file coomatrix.h.

◆ TripletVec

template<typename SCALAR >

using lf::assemble::COOMatrix< SCALAR >::TripletVec = std::vector<Triplet>

Definition at line 56 of file coomatrix.h.

Constructor & Destructor Documentation

◆ COOMatrix() [1/3]

template<typename SCALAR >

lf::assemble::COOMatrix< SCALAR >::COOMatrix	(	size_type	num_rows,
		size_type	num_cols )

inline

Set up zero matrix of a given size

Definition at line 60 of file coomatrix.h.

◆ COOMatrix() [2/3]

template<typename SCALAR >

lf::assemble::COOMatrix< SCALAR >::COOMatrix ( const COOMatrix< SCALAR > & )

default

◆ COOMatrix() [3/3]

template<typename SCALAR >

lf::assemble::COOMatrix< SCALAR >::COOMatrix ( COOMatrix< SCALAR > && )

defaultnoexcept

◆ ~COOMatrix()

template<typename SCALAR >

lf::assemble::COOMatrix< SCALAR >::~COOMatrix ( )

default

Member Function Documentation

◆ AddToEntry()

template<typename SCALAR >

void lf::assemble::COOMatrix< SCALAR >::AddToEntry	(	gdof_idx_t	i,
		gdof_idx_t	j,
		SCALAR	increment )

inline

Add a value to the specified entry.

Parameters

i	row index
j	column index
increment

Adds another index-value triplet.

Note: the size information for the matrix is adjusted to the passed indices. This ensures that the internally stored numbers of columns and rows are always bigger than the indices of any triplet. When manipulating the triplet information directly, one has to make sure that the size information remains valid.

Definition at line 87 of file coomatrix.h.

References lf::assemble::AssembleMatrixLocally(), lf::assemble::COOMatrix< SCALAR >::cols_, lf::assemble::COOMatrix< SCALAR >::rows_, and lf::assemble::COOMatrix< SCALAR >::triplets_.

◆ cols()

template<typename SCALAR >

Index lf::assemble::COOMatrix< SCALAR >::cols ( ) const

inline

return number of column

Definition at line 72 of file coomatrix.h.

References lf::assemble::COOMatrix< SCALAR >::cols_.

◆ makeDense()

template<typename SCALAR >

Eigen::Matrix< Scalar, Eigen::Dynamic, Eigen::Dynamic > lf::assemble::COOMatrix< SCALAR >::makeDense ( ) const

inline

Create an Eigen::MatrixX from the COO format.

Returns: A dense matrix representing the COO matrix

This method is mainly meant for debugging

Definition at line 188 of file coomatrix.h.

References lf::assemble::AssembleMatrixLocally(), lf::assemble::COOMatrix< SCALAR >::cols_, lf::assemble::COOMatrix< SCALAR >::rows_, and lf::assemble::COOMatrix< SCALAR >::triplets_.

◆ makeSparse()

template<typename SCALAR >

Eigen::SparseMatrix< Scalar > lf::assemble::COOMatrix< SCALAR >::makeSparse ( ) const

inline

Create an Eigen::SparseMatrix out of the COO format.

Returns: The created sparse matrix

Note: This method can be called multiple times and it does not modify the data stored in this COOMatrix.

Definition at line 172 of file coomatrix.h.

References lf::assemble::COOMatrix< SCALAR >::cols_, lf::assemble::COOMatrix< SCALAR >::rows_, and lf::assemble::COOMatrix< SCALAR >::triplets_.

◆ MatVecMult() [1/2]

template<typename SCALAR >

template<typename VECTOR >

Eigen::Matrix< SCALAR, Eigen::Dynamic, 1 > lf::assemble::COOMatrix< SCALAR >::MatVecMult	(	SCALAR	alpha,
		const VECTOR &	vec ) const

Computes the product of a (scaled) vector with the matrix in COO format.

Template Parameters

VECTOR a basic vector type for the argument vector

Parameters

alpha	scalar with which to multiply the argument vector before the matrix x vector multiplication.
vec	constant reference to a vector of type VECTOR

Returns: result vector, a dense vector of Eigen

Requirements for type VECTOR

An object of type VECTOR must provide a method Size() telling the length of the vector and operator [] for read access to vector entries.

Definition at line 236 of file coomatrix.h.

References lf::assemble::AssembleMatrixLocally().

◆ MatVecMult() [2/2]

template<typename SCALAR >

template<typename VECTOR , typename RESULTVECTOR >

void lf::assemble::COOMatrix< SCALAR >::MatVecMult	(	SCALAR	alpha,
		const VECTOR &	vec,
		RESULTVECTOR &	resvec ) const

In-situ computes the product of a vector with the matrix in COO format.

Template Parameters

VECTOR	a basic vector type for the argument vector
RESULTVECTOR	another vector type for returning the result

Parameters

alpha	scalar with which to multiply the argument vector before the matrix x vector multiplication.
vec	constant reference to a vector of type VECTOR
resvec	reference to an object of type RESULTVECTOR.

Note: the result of the matrix-vector product will be added to the entries of result!

Requirements for types VECTOR and RESULTVECTOR

An object of type VECTOR or RESULTVECTOR must provide a method Size() telling the length of the vector and operator [] for read/write access to vector entries.

Definition at line 250 of file coomatrix.h.

References lf::assemble::AssembleMatrixLocally().