LAMA
lama::IterativeSolver Class Reference

Uses iterative methods to solve the equation system. More...

#include <IterativeSolver.hpp>

Inheritance diagram for lama::IterativeSolver:

Data Structures

struct  IterativeSolverRuntime

Public Member Functions

 IterativeSolver (const std::string &id)
 Creates a solver with a given ID.
 IterativeSolver (const std::string &id, LoggerPtr logger)
 Constructs a new solver with the given id and logger.
 IterativeSolver (const IterativeSolver &other)
 Copy constructor that copies the status independent solver information.
virtual ~IterativeSolver ()
 Destructor for the IterativeSolver class.
virtual void initialize (const Matrix &coefficients)
 Used to initialize a solver with a certain matrix A from A*u=f.
virtual void solveImpl ()
 Solves the equation system.
void setStoppingCriterion (const CriterionPtr criterion)
 set a new StoppingCriterion to the solver.
void setPreconditioner (SolverPtr const conditioner)
 Sets the preconditioner of this solver.
const SolverPtr getPreconditioner () const
 returns the preconditioner of this solver
int getIterationCount () const
 returns the number of iterations, this solver has done so far.
virtual SolverPtr copy ()=0
 Copies the status indepedent solver informations to create a new instance of the same type.
virtual IterativeSolverRuntimegetRuntime ()=0
 Returns the complete configuration of the derived class.
virtual const
IterativeSolverRuntime
getConstRuntime () const =0
 Returns the complete const configuration of the derived class.
virtual void solve (Vector &solution, const Vector &rhs)
 Solves the equation system based on the given rhs.
virtual void solveInit (Vector &solution, const Vector &rhs)
 Initializes the solver with rhs and solution.
virtual void solveFinalize ()
 Finalizes the solving process.
const std::string & getId () const
 Returns the ID of this solver.
const VectorgetResidual () const
 Contingently calculates the current residual based on the coefficients, rhs and solution currently associated with this.
const MatrixgetCoefficients () const
 Gets the matrix A from A*u=f.
void setLogger (LoggerPtr logger)
 Redefines mLogger.
void setLogLevel (LogLevel::LogLevel level)
 Switches the loglevel of mLogger.
virtual void setContext (ContextPtr context)
 Sets the context where this solver should be executed.

Protected Member Functions

bool criteriaAreSatisfied () const
 Checks if all of the stopping criteria are satisfied.
virtual void iterate ()=0
 Represents one iteration step of the solver.
void logStartSolve ()
 Logging methods to maintain code-readability.
void logEndSolve ()
void logIterationEndAndResidual ()
void logIterationStart ()
 LAMA_LOG_DECL_STATIC_LOGGER (logger)
virtual void writeAt (std::ostream &stream) const
 Overrides Printable::writeAt with more useful stuff.

Protected Attributes

SolverPtr mPreconditioner
 The preconditioner of this solver.
CriterionPtr mCriterionRootComponent
 The root stopping criterion evaluated every iteration in the solve method.
std::string mId
 The ID of this solver.
LoggerPtr mLogger
 The solver logger.
ContextPtr mContext
 For forcing the context solver dependent.

Detailed Description

Uses iterative methods to solve the equation system.


Constructor & Destructor Documentation

lama::IterativeSolver::IterativeSolver ( const std::string &  id)

Creates a solver with a given ID.

Parameters:
idThe ID for the solver.
lama::IterativeSolver::IterativeSolver ( const std::string &  id,
LoggerPtr  logger 
)

Constructs a new solver with the given id and logger.

Parameters:
idThe name or id of the solver.
loggerThe logger which the solver shall use

Copy constructor that copies the status independent solver information.

References getPreconditioner(), mCriterionRootComponent, and mPreconditioner.

Destructor for the IterativeSolver class.


Member Function Documentation

virtual SolverPtr lama::IterativeSolver::copy ( ) [pure virtual]

Copies the status indepedent solver informations to create a new instance of the same type.

Returns:
shared pointer of the copied solver

Implements lama::Solver.

Implemented in lama::OmegaSolver, lama::SimpleAMG, lama::SpecializedJacobi, lama::DefaultJacobi, lama::SOR, lama::GMRES, and lama::CG.

bool lama::IterativeSolver::criteriaAreSatisfied ( ) const [protected]

Checks if all of the stopping criteria are satisfied.

return Whether the criteria are satisfied or not

References lama::Solver::getId(), LAMA_THROWEXCEPTION, and mCriterionRootComponent.

Referenced by solveImpl().

const Matrix & lama::Solver::getCoefficients ( ) const [inherited]

Gets the matrix A from A*u=f.

Returns:
The coefficient matrix A.

References lama::Solver::getConstRuntime(), LAMA_ASSERT_DEBUG, and lama::Solver::SolverRuntime::mCoefficients.

Referenced by lama::InverseSolver::invert(), and lama::LUSolver::piamax_own().

virtual const IterativeSolverRuntime& lama::IterativeSolver::getConstRuntime ( ) const [pure virtual]

Returns the complete const configuration of the derived class.

Implements lama::Solver.

Implemented in lama::OmegaSolver, lama::GMRES, lama::SimpleAMG, lama::DefaultJacobi, lama::CG, lama::SpecializedJacobi, and lama::SOR.

Referenced by getIterationCount(), and solveImpl().

const std::string & lama::Solver::getId ( ) const [inherited]

Returns the ID of this solver.

Returns:
The ID of this solver.

References lama::Solver::mId.

Referenced by criteriaAreSatisfied(), solveImpl(), and lama::SolverConfigGrammar::SolverConfigGrammar().

returns the preconditioner of this solver

Returns:
the preconditioner

References mPreconditioner.

Referenced by lama::MetaSolver::initializePreconditioner(), and IterativeSolver().

Returns the complete configuration of the derived class.

Implements lama::Solver.

Implemented in lama::OmegaSolver, lama::GMRES, lama::SimpleAMG, lama::DefaultJacobi, lama::CG, lama::SpecializedJacobi, and lama::SOR.

Referenced by logEndSolve(), logIterationEndAndResidual(), logIterationStart(), and solveImpl().

void lama::IterativeSolver::initialize ( const Matrix coefficients) [virtual]

Used to initialize a solver with a certain matrix A from A*u=f.

This method initializes a solver with a certain coefficient-matrix. The only thing it does is storing the matrix pointer as a member for derived solver classes to use it. The caller delegates the property of the pointer to the Solver instance.

This method may be overwritten by derived classes which desire more complex initialization.

Parameters:
coefficientsThe matrix A from A*u=f.

Reimplemented from lama::Solver.

Reimplemented in lama::DefaultJacobi, lama::OmegaSolver, lama::GMRES, lama::DefaultJacobi, lama::CG, lama::SOR, lama::SimpleAMG, and lama::SpecializedJacobi.

References lama::Solver::mLogger, mPreconditioner, and lama::LogLevel::solverInformation.

virtual void lama::IterativeSolver::iterate ( ) [protected, pure virtual]

Represents one iteration step of the solver.

Use this abstract method to implement new iterative solvers (e.g. Jacobi)

This method represents one solver iteration. Idealically a derived iterative solver only overwrites this method. It is only used by the solve() method of this class. Use the member variables m_rhs, mSolution and m_coefficients during the iteration process.

Implemented in lama::OmegaSolver, lama::DefaultJacobi, lama::GMRES, lama::DefaultJacobi, lama::CG, lama::SpecializedJacobi, lama::SOR, and lama::SimpleAMG.

Referenced by solveImpl().

Logging methods to maintain code-readability.

References lama::LogLevel::convergenceHistory, lama::l2Norm(), lama::Solver::mLogger, and lama::LogLevel::solverInformation.

Referenced by solveImpl().

void lama::Solver::setContext ( ContextPtr  context) [virtual, inherited]

Sets the context where this solver should be executed.

Sets the context where this solver should be executed. Caution: This overrides the context of the coefficients matrix A from A * u = f used to inializ this solver.

Parameters:
[in]contextthe context where this solver should be executed.

Reimplemented in lama::InverseSolver.

References lama::Solver::mContext.

Referenced by lama::GMRES::initialize().

void lama::Solver::setLogger ( LoggerPtr  logger) [inherited]

Redefines mLogger.

References lama::Solver::mLogger.

void lama::Solver::setLogLevel ( LogLevel::LogLevel  level) [inherited]

Switches the loglevel of mLogger.

Returns:

References lama::Solver::mLogger.

Sets the preconditioner of this solver.

Preconditioner should be set before initializing the this solver, otherwise the preconditioner initialization should be executed manually

Parameters:
conditionerThe preconditioner

References mPreconditioner.

set a new StoppingCriterion to the solver.

Parameters:
[in]criterionthe new criterion.

References LAMA_ASSERT_ERROR, and mCriterionRootComponent.

Referenced by lama::SingleGridSetup::initialize().

void lama::Solver::solve ( Vector solution,
const Vector rhs 
) [virtual, inherited]

Solves the equation system based on the given rhs.

The solver needs to be initialized first with the matrix from the equation to solve, e.g. A from A*u=f (call solver::initialize(A) for example) This method is abstract. It has to be implemented by a class which inherits from this class.

Parameters:
rhsThe right hand side of A*u=f.
solutionThe solution from A*u=f. Mostly used as starting solution for an IterativeSolver.

Reimplemented in lama::LUSolver, lama::DefaultJacobi, and lama::SpecializedJacobi.

References lama::Solver::getConstRuntime(), LAMA_REGION, lama::Solver::solveFinalize(), lama::Solver::solveImpl(), and lama::Solver::solveInit().

Referenced by lama::SimpleAMG::cycle().

void lama::Solver::solveFinalize ( ) [virtual, inherited]

Finalizes the solving process.

Reimplemented in lama::DefaultJacobi, and lama::SpecializedJacobi.

Referenced by lama::Solver::solve().

Solves the equation system.

Rhs and starting solution have to be initialized first! (call solveInit( rhs, solution ) ). The solver needs to be initialized first with the matrix from the equation to solve, e.g. A from A*u=f (call solver::initialize(A) for example)

This method solves the equation system by using the given rhs and solution. For most iterative solvers the solution-vector is used as a starting solution for the solve process. This class does not take responsibility for deleting the vectors after the solver! Make sure you do not delete the vectors during the solver process.

Implements lama::Solver.

References criteriaAreSatisfied(), getConstRuntime(), lama::Solver::getId(), getRuntime(), iterate(), LAMA_THROWEXCEPTION, logEndSolve(), logIterationEndAndResidual(), logIterationStart(), logStartSolve(), and lama::IterativeSolver::IterativeSolverRuntime::mIterations.

Referenced by lama::SpecializedJacobi::solve(), and lama::DefaultJacobi::solve().

void lama::Solver::solveInit ( Vector solution,
const Vector rhs 
) [virtual, inherited]
void lama::Solver::writeAt ( std::ostream &  stream) const [protected, virtual, inherited]

Overrides Printable::writeAt with more useful stuff.

Reimplemented from Printable.

References lama::Solver::mId.


Field Documentation

ContextPtr lama::Solver::mContext [protected, inherited]

For forcing the context solver dependent.

If the context for a solver is set, the context of the input matrix will be ignored

Referenced by lama::SpecializedJacobi::iterateTyped(), lama::InverseSolver::setContext(), and lama::Solver::setContext().

The root stopping criterion evaluated every iteration in the solve method.

Referenced by criteriaAreSatisfied(), IterativeSolver(), and setStoppingCriterion().


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