TrilinosWrappers::PreconditionAMG Class Reference

#include <deal.II/lac/trilinos_precondition.h>

Inheritance diagram for TrilinosWrappers::PreconditionAMG:

Classes
struct	AdditionalData

Public Types
using	size_type = ::types::global_dof_index

Public Member Functions
	~PreconditionAMG () override
void	initialize (const SparseMatrix &matrix, const AdditionalData &additional_data=AdditionalData())
void	initialize (const Epetra_RowMatrix &matrix, const AdditionalData &additional_data=AdditionalData())
void	initialize (const SparseMatrix &matrix, const Teuchos::ParameterList &ml_parameters)
void	initialize (const Epetra_RowMatrix &matrix, const Teuchos::ParameterList &ml_parameters)
template<typename number >
void	initialize (const ::SparseMatrix< number > &deal_ii_sparse_matrix, const AdditionalData &additional_data=AdditionalData(), const double drop_tolerance=1e-13, const ::SparsityPattern *use_this_sparsity=nullptr)
void	reinit ()
void	clear ()
size_type	memory_consumption () const
MPI_Comm	get_mpi_communicator () const
void	transpose ()
virtual void	vmult (MPI::Vector &dst, const MPI::Vector &src) const
virtual void	vmult (::Vector< double > &dst, const ::Vector< double > &src) const
virtual void	vmult (::LinearAlgebra::distributed::Vector< double > &dst, const ::LinearAlgebra::distributed::Vector< double > &src) const
virtual void	Tvmult (MPI::Vector &dst, const MPI::Vector &src) const
virtual void	Tvmult (::Vector< double > &dst, const ::Vector< double > &src) const
virtual void	Tvmult (::LinearAlgebra::distributed::Vector< double > &dst, const ::LinearAlgebra::distributed::Vector< double > &src) const
template<class Archive >
void	serialize (Archive &ar, const unsigned int version)
Access to underlying Trilinos data
Epetra_Operator &	trilinos_operator () const
Partitioners
IndexSet	locally_owned_domain_indices () const
IndexSet	locally_owned_range_indices () const
Subscriptor functionality
Classes derived from Subscriptor provide a facility to subscribe to this object. This is mostly used by the SmartPointer class.
void	subscribe (std::atomic< bool > *const validity, const std::string &identifier="") const
void	unsubscribe (std::atomic< bool > *const validity, const std::string &identifier="") const
unsigned int	n_subscriptions () const
template<typename StreamType >
void	list_subscribers (StreamType &stream) const
void	list_subscribers () const

Static Public Member Functions
static ::ExceptionBase &	ExcNonMatchingMaps (std::string arg1)
static ::ExceptionBase &	ExcInUse (int arg1, std::string arg2, std::string arg3)
static ::ExceptionBase &	ExcNoSubscriber (std::string arg1, std::string arg2)

Protected Attributes
Teuchos::RCP< Epetra_Operator >	preconditioner
Epetra_MpiComm	communicator

Private Types
using	map_value_type = decltype(counter_map)::value_type
using	map_iterator = decltype(counter_map)::iterator

Private Member Functions
void	check_no_subscribers () const noexcept

Private Attributes
std::shared_ptr< SparseMatrix >	trilinos_matrix
std::atomic< unsigned int >	counter
std::map< std::string, unsigned int >	counter_map
std::vector< std::atomic< bool > * >	validity_pointers
const std::type_info *	object_info

Static Private Attributes
static std::mutex	mutex

Detailed Description

This class implements an algebraic multigrid (AMG) preconditioner based on the Trilinos ML implementation, which is a black-box preconditioner that works well for many PDE-based linear problems. What this class does is twofold. When the initialize() function is invoked, a ML preconditioner object is created based on the matrix that we want the preconditioner to be based on. A call of the respective vmult function does call the respective operation in the Trilinos package, where it is called ApplyInverse. Use of this class is explained in the step-31 tutorial program.

Since the Trilinos objects we want to use are heavily dependent on Epetra objects, we recommend using this class in conjunction with Trilinos (Epetra) sparse matrices and vectors. There is support for use with matrices of the SparseMatrix class and corresponding vectors, too, but this requires generating a copy of the matrix, which is slower and takes (much) more memory. When doing such a copy operation, we can still profit from the fact that some of the entries in the preconditioner matrix are zero and hence can be neglected.

The implementation is able to distinguish between matrices from elliptic problems and convection dominated problems. We use the standard options provided by Trilinos ML for elliptic problems, except that we use a Chebyshev smoother instead of a symmetric Gauss-Seidel smoother. For most elliptic problems, Chebyshev provides a better damping of high frequencies (in the algebraic sense) than Gauss-Seidel (SSOR), and is faster (Chebyshev requires only some matrix-vector products, whereas SSOR requires substitutions which are more expensive). Moreover, Chebyshev is perfectly parallel in the sense that it does not degenerate when used on many processors. SSOR, on the other hand, gets more Jacobi-like on many processors.

For proper functionality of this class we recommend using Trilinos v9.0 and higher. Older versions may have problems with generating the coarse-matrix structure when using matrices with many nonzero entries per row (i.e., matrices stemming from higher order finite element discretizations).

Definition at line 1320 of file trilinos_precondition.h.

Member Typedef Documentation

size_type

using TrilinosWrappers::PreconditionBase::size_type = ::types::global_dof_index

inherited

Declare the type for container size.

Definition at line 82 of file trilinos_precondition.h.

map_value_type

using Subscriptor::map_value_type = decltype(counter_map)::value_type

privateinherited

The data type used in counter_map.

Definition at line 229 of file subscriptor.h.

map_iterator

using Subscriptor::map_iterator = decltype(counter_map)::iterator

privateinherited

The iterator type used in counter_map.

Definition at line 234 of file subscriptor.h.

Constructor & Destructor Documentation

~PreconditionAMG()

TrilinosWrappers::PreconditionAMG::~PreconditionAMG ( )

override

Destructor.

Definition at line 212 of file trilinos_precondition_ml.cc.

Member Function Documentation

initialize() [1/5]

void TrilinosWrappers::PreconditionAMG::initialize	(	const SparseMatrix &	matrix,
		const AdditionalData &	additional_data = AdditionalData() )

Let Trilinos compute a multilevel hierarchy for the solution of a linear system with the given matrix. The function uses the matrix format specified in TrilinosWrappers::SparseMatrix.

Definition at line 221 of file trilinos_precondition_ml.cc.

initialize() [2/5]

void TrilinosWrappers::PreconditionAMG::initialize	(	const Epetra_RowMatrix &	matrix,
		const AdditionalData &	additional_data = AdditionalData() )

Let Trilinos compute a multilevel hierarchy for the solution of a linear system with the given matrix. As opposed to the other initialize function above, this function uses an abstract interface to an object of type Epetra_RowMatrix which allows a user to pass quite general objects to the ML preconditioner.

This initialization routine is useful in cases where the operator to be preconditioned is not a TrilinosWrappers::SparseMatrix object but still allows getting a copy of the entries in each of the locally owned matrix rows (method ExtractMyRowCopy) and implements a matrix-vector product (methods Multiply or Apply). An example are operators which provide faster matrix-vector multiplications than possible with matrix entries (matrix-free methods). These implementations can be beneficially combined with Chebyshev smoothers that only perform matrix-vector products. The interface class Epetra_RowMatrix is very flexible to enable this kind of implementation.

Definition at line 230 of file trilinos_precondition_ml.cc.

initialize() [3/5]

void TrilinosWrappers::PreconditionAMG::initialize	(	const SparseMatrix &	matrix,
		const Teuchos::ParameterList &	ml_parameters )

Let Trilinos compute a multilevel hierarchy for the solution of a linear system with the given matrix. The function uses the matrix format specified in TrilinosWrappers::SparseMatrix.

This function is similar to the one above, but allows the user to set all the options of the Trilinos ML preconditioner. In order to find out about all the options for ML, we refer to the ML documentation page. In particular, users need to follow the ML instructions in case a vector-valued problem ought to be solved.

Definition at line 256 of file trilinos_precondition_ml.cc.

initialize() [4/5]

void TrilinosWrappers::PreconditionAMG::initialize	(	const Epetra_RowMatrix &	matrix,
		const Teuchos::ParameterList &	ml_parameters )

Let Trilinos compute a multilevel hierarchy for the solution of a linear system with the given matrix. As opposed to the other initialize function above, this function uses an abstract interface to an object of type Epetra_RowMatrix which allows a user to pass quite general objects to the ML preconditioner.

Definition at line 265 of file trilinos_precondition_ml.cc.

initialize() [5/5]

template<typename number >

void TrilinosWrappers::PreconditionAMG::initialize	(	const ::SparseMatrix< number > &	deal_ii_sparse_matrix,
		const AdditionalData &	additional_data = AdditionalData(),
		const double	drop_tolerance = 1e-13,
		const ::SparsityPattern *	use_this_sparsity = nullptr )

Let Trilinos compute a multilevel hierarchy for the solution of a linear system with the given matrix. This function takes a deal.II matrix and copies the content into a Trilinos matrix, so the function can be considered rather inefficient.

Definition at line 276 of file trilinos_precondition_ml.cc.

reinit()

void TrilinosWrappers::PreconditionAMG::reinit

(

)

This function can be used for a faster recalculation of the preconditioner construction when the matrix entries underlying the preconditioner have changed, but the matrix sparsity pattern has remained the same. What this function does is taking the already generated coarsening structure, computing the AMG prolongation and restriction according to a smoothed aggregation strategy and then building the whole multilevel hierarchy. This function can be considerably faster than the initialize function, since the coarsening pattern is usually the most difficult thing to do when setting up the AMG ML preconditioner.

Definition at line 310 of file trilinos_precondition_ml.cc.

clear()

void TrilinosWrappers::PreconditionAMG::clear

(

)

Destroys the preconditioner, leaving an object like just after having called the constructor.

Definition at line 320 of file trilinos_precondition_ml.cc.

memory_consumption()

PreconditionAMG::size_type TrilinosWrappers::PreconditionAMG::memory_consumption

(

)

const

Prints an estimate of the memory consumption of this class.

Definition at line 329 of file trilinos_precondition_ml.cc.

get_mpi_communicator()

MPI_Comm TrilinosWrappers::PreconditionBase::get_mpi_communicator ( ) const

inherited

Return the underlying MPI communicator.

Definition at line 47 of file trilinos_precondition.cc.

transpose()

void TrilinosWrappers::PreconditionBase::transpose ( )

inherited

Sets an internal flag so that all operations performed by the matrix, i.e., multiplications, are done in transposed order. However, this does not reshape the matrix to transposed form directly, so care should be taken when using this flag.

Note

Calling this function any even number of times in succession will return the object to its original state.

vmult() [1/3]

virtual void TrilinosWrappers::PreconditionBase::vmult ( MPI::Vector & dst, const MPI::Vector & src ) const

virtualinherited

Apply the preconditioner.

Reimplemented in TrilinosWrappers::PreconditionIdentity.

vmult() [2/3]

virtual void TrilinosWrappers::PreconditionBase::vmult ( ::Vector< double > & dst, const ::Vector< double > & src ) const

virtualinherited

Apply the preconditioner on deal.II data structures instead of the ones provided in the Trilinos wrapper class.

Reimplemented in TrilinosWrappers::PreconditionIdentity.

vmult() [3/3]

virtual void TrilinosWrappers::PreconditionBase::vmult ( ::LinearAlgebra::distributed::Vector< double > & dst, const ::LinearAlgebra::distributed::Vector< double > & src ) const

virtualinherited

Apply the preconditioner on deal.II parallel data structures instead of the ones provided in the Trilinos wrapper class.

Tvmult() [1/3]

virtual void TrilinosWrappers::PreconditionBase::Tvmult ( MPI::Vector & dst, const MPI::Vector & src ) const

virtualinherited

Apply the transpose preconditioner.

Reimplemented in TrilinosWrappers::PreconditionIdentity.

Tvmult() [2/3]

virtual void TrilinosWrappers::PreconditionBase::Tvmult ( ::Vector< double > & dst, const ::Vector< double > & src ) const

virtualinherited

Apply the transpose preconditioner on deal.II data structures instead of the ones provided in the Trilinos wrapper class.

Reimplemented in TrilinosWrappers::PreconditionIdentity.

Tvmult() [3/3]

virtual void TrilinosWrappers::PreconditionBase::Tvmult ( ::LinearAlgebra::distributed::Vector< double > & dst, const ::LinearAlgebra::distributed::Vector< double > & src ) const

virtualinherited

Apply the transpose preconditioner on deal.II parallel data structures instead of the ones provided in the Trilinos wrapper class.

trilinos_operator()

Epetra_Operator & TrilinosWrappers::PreconditionBase::trilinos_operator ( ) const

inherited

Calling this function from an uninitialized object will cause an exception.

Definition at line 54 of file trilinos_precondition.cc.

locally_owned_domain_indices()

IndexSet TrilinosWrappers::PreconditionBase::locally_owned_domain_indices ( ) const

inherited

Return the partitioning of the domain space of this matrix, i.e., the partitioning of the vectors this matrix has to be multiplied with.

Definition at line 63 of file trilinos_precondition.cc.

locally_owned_range_indices()

IndexSet TrilinosWrappers::PreconditionBase::locally_owned_range_indices ( ) const

inherited

Return the partitioning of the range space of this matrix, i.e., the partitioning of the vectors that are result from matrix-vector products.

Definition at line 70 of file trilinos_precondition.cc.

subscribe()

void Subscriptor::subscribe ( std::atomic< bool > *const validity, const std::string & identifier = "" ) const

inherited

Subscribes a user of the object by storing the pointer validity. The subscriber may be identified by text supplied as identifier.

Definition at line 135 of file subscriptor.cc.

unsubscribe()

void Subscriptor::unsubscribe ( std::atomic< bool > *const validity, const std::string & identifier = "" ) const

inherited

Unsubscribes a user from the object.

Note

The identifier and the validity pointer must be the same as the one supplied to subscribe().

Definition at line 155 of file subscriptor.cc.

n_subscriptions()

unsigned int Subscriptor::n_subscriptions ( ) const

inlineinherited

Return the present number of subscriptions to this object. This allows to use this class for reference counted lifetime determination where the last one to unsubscribe also deletes the object.

Definition at line 300 of file subscriptor.h.

list_subscribers() [1/2]

template<typename StreamType >

void Subscriptor::list_subscribers ( StreamType & stream ) const

inlineinherited

List the subscribers to the input stream.

Definition at line 317 of file subscriptor.h.

list_subscribers() [2/2]

void Subscriptor::list_subscribers ( ) const

inherited

List the subscribers to deallog.

Definition at line 203 of file subscriptor.cc.

serialize()

template<class Archive >

void Subscriptor::serialize ( Archive & ar, const unsigned int version )

inlineinherited

Read or write the data of this object to or from a stream for the purpose of serialization using the BOOST serialization library.

This function does not actually serialize any of the member variables of this class. The reason is that what this class stores is only who subscribes to this object, but who does so at the time of storing the contents of this object does not necessarily have anything to do with who subscribes to the object when it is restored. Consequently, we do not want to overwrite the subscribers at the time of restoring, and then there is no reason to write the subscribers out in the first place.

Definition at line 309 of file subscriptor.h.

check_no_subscribers()

void Subscriptor::check_no_subscribers ( ) const

privatenoexceptinherited

Check that there are no objects subscribing to this object. If this check passes then it is safe to destroy the current object. It this check fails then this function will either abort or print an error message to deallog (by using the AssertNothrow mechanism), but will not throw an exception.

Note

Since this function is just a consistency check it does nothing in release mode.

If this function is called when there is an uncaught exception then, rather than aborting, this function prints an error message to the standard error stream and returns.

Definition at line 52 of file subscriptor.cc.

Member Data Documentation

trilinos_matrix

std::shared_ptr<SparseMatrix> TrilinosWrappers::PreconditionAMG::trilinos_matrix

private

A copy of the deal.II matrix into Trilinos format.

Definition at line 1681 of file trilinos_precondition.h.

preconditioner

Teuchos::RCP<Epetra_Operator> TrilinosWrappers::PreconditionBase::preconditioner

protectedinherited

This is a pointer to the preconditioner object that is used when applying the preconditioner.

Definition at line 222 of file trilinos_precondition.h.

communicator

Epetra_MpiComm TrilinosWrappers::PreconditionBase::communicator

protectedinherited

Internal communication pattern in case the matrix needs to be copied from deal.II format.

Definition at line 228 of file trilinos_precondition.h.

counter

std::atomic<unsigned int> Subscriptor::counter

mutableprivateinherited

Store the number of objects which subscribed to this object. Initially, this number is zero, and upon destruction it shall be zero again (i.e. all objects which subscribed should have unsubscribed again).

The creator (and owner) of an object is counted in the map below if HE manages to supply identification.

We use the mutable keyword in order to allow subscription to constant objects also.

This counter may be read from and written to concurrently in multithreaded code: hence we use the std::atomic class template.

Definition at line 218 of file subscriptor.h.

counter_map

std::map<std::string, unsigned int> Subscriptor::counter_map

mutableprivateinherited

In this map, we count subscriptions for each different identification string supplied to subscribe().

Definition at line 224 of file subscriptor.h.

validity_pointers

std::vector<std::atomic<bool> *> Subscriptor::validity_pointers

mutableprivateinherited

In this vector, we store pointers to the validity bool in the SmartPointer objects that subscribe to this class.

Definition at line 240 of file subscriptor.h.

object_info

const std::type_info* Subscriptor::object_info

mutableprivateinherited

Pointer to the typeinfo object of this object, from which we can later deduce the class name. Since this information on the derived class is neither available in the destructor, nor in the constructor, we obtain it in between and store it here.

Definition at line 248 of file subscriptor.h.

mutex

std::mutex Subscriptor::mutex

staticprivateinherited

A mutex used to ensure data consistency when accessing the mutable members of this class. This lock is used in the subscribe() and unsubscribe() functions, as well as in list_subscribers().

Definition at line 271 of file subscriptor.h.

---

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

• include/deal.II/lac/trilinos_precondition.h
• source/lac/trilinos_precondition_ml.cc