ptems::ILUPreconditioner< T > Class Template Reference

Defines a preconditioner using incomplete LU (ILU) factorisation. More...

#include <ptems/solvers/preconditioners/ilu.hpp>

Inheritance diagram for ptems::ILUPreconditioner< T >:

Collaboration diagram for ptems::ILUPreconditioner< T >:

Public Types
typedef T	value_type
	The entry types of the matrix to precondition. More...

Public Member Functions
	ILUPreconditioner (const SparsityPattern &sparsity, PreconditionerType type=PreconditionerType::Right)
	Creates an incomplete LU factorisation preconditioner with the given sparsity pattern. More...
void	Initialise (const ParallelComm &comm, const std::vector< SparseMatrixEntry< T >> &mat, std::size_t firstRow, std::size_t size, const std::vector< std::size_t > rowOffset) override
	Initialise the preconditioner for the specified matrix. More...
virtual void	ApplyLeft (const ParallelComm &comm, std::vector< T > &rhs) override
	Apply the left preconditioner to the specified vector overriding the vector with the result. More...
virtual void	ApplyRight (const ParallelComm &comm, std::vector< T > &rhs) override
	Apply the right preconditioner to the specified vector overriding the vector with the result. More...
virtual PreconditionerType	Type () const override
	Gets the type of the preconditioner. More...

Detailed Description

template<typename T>
class ptems::ILUPreconditioner< T >

Defines a preconditioner using incomplete LU (ILU) factorisation.

An ILU factorisation performs LU factorisation computes SPARSE lower ( \(L\)) and upper ( \(U\)) triangular matrices such that the residual \(LU-A\), where \(A\) is the matrix being factorised has a given sparsity pattern, and, ideally, such that \(LU \approx A\).

Template Parameters

The	type of the matrix entries the pre-conditioner is for

Member Typedef Documentation

value_type

template<typename T >

typedef T ptems::Preconditioner< T >::value_type

inherited

The entry types of the matrix to precondition.

Constructor & Destructor Documentation

ILUPreconditioner()

template<typename T >

ptems::ILUPreconditioner< T >::ILUPreconditioner ( const SparsityPattern &  sparsity, PreconditionerType  type = PreconditionerType::Right  )

inline

Creates an incomplete LU factorisation preconditioner with the given sparsity pattern.

Note

If the sparsity pattern is the same as the matrix being factorised then this gives an ILU(0) preconditioner.

Parameters

sparsity	Sparsity pattern of L+U
type	Specifies whether this preconditioner should be applied as a left, right, or split preconditioner 9default=Right)

Member Function Documentation

ApplyLeft()

template<typename T >

virtual void ptems::ILUPreconditioner< T >::ApplyLeft ( const ParallelComm &  comm, std::vector< T > &  rhs  )

inlineoverridevirtual

Apply the left preconditioner to the specified vector overriding the vector with the result.

Attention

If the preconditioner is a right preconditioner this method must perform a no-op (return vector unchanged)

Parameters

comm	MPI communicator for parallel preconditioning
rhs	Vector to precondition (contains complete vector NOT slice). Update with the preconditioned vector

Implements ptems::Preconditioner< T >.

ApplyRight()

template<typename T >

virtual void ptems::ILUPreconditioner< T >::ApplyRight ( const ParallelComm &  comm, std::vector< T > &  rhs  )

inlineoverridevirtual

Apply the right preconditioner to the specified vector overriding the vector with the result.

Attention

If the preconditioner is a left preconditioner this method must perform a no-op (return vector unchanged)

Parameters

comm	MPI communicator for parallel preconditioning
rhs	Vector to precondition (contains complete vector NOT slice). Update with the preconditioned vector

Implements ptems::Preconditioner< T >.

Initialise()

template<typename T >

void ptems::ILUPreconditioner< T >::Initialise ( const ParallelComm &  comm, const std::vector< SparseMatrixEntry< T >> &  mat, std::size_t  firstRow, std::size_t  size, const std::vector< std::size_t >  rowOffset  )

inlineoverridevirtual

Initialise the preconditioner for the specified matrix.

The preconditioner can be called with a "slice" if using MPI (different processors get different slices)

Attention

If the linear solver is reused (for example in non-linear solvers) this method may be called multiply times with different matrices. The underlying implementation must, therefore, treat each call as a complete new initialisation (essentially construction) of the preconditioner.

Parameters

comm	MPI communicator for parallel preconditioning
mat	The slice of the matrix for the current process as a sparse matrix. Entries are ordered by row, then column
firstRow	The row index of the first row in the slice
size	The number of rows (and columns) of the FULL matrix
rowOffset	The offset into mat for each row in the slice (the offset with index 0 is for firstRow (and will always by 0), 1 for firstRow+1 etc.)

Implements ptems::Preconditioner< T >.

Type()

template<typename T >

virtual PreconditionerType ptems::ILUPreconditioner< T >::Type ( ) const

inlineoverridevirtual

Gets the type of the preconditioner.

Implements ptems::Preconditioner< T >.

---

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

• ptems/solvers/preconditioners/ilu.hpp