LinearSolver< T > Class Template Reference

#include <LinearSolver.H>

Inheritance diagram for LinearSolver< T >:

Public Member Functions
virtual	~LinearSolver ()
virtual void	setHomogeneous (bool a_homogeneous)=0
virtual void	define (LinearOp< T > *a_operator, bool a_homogeneous=false)=0
virtual void	solve (T &a_phi, const T &a_rhs)=0
virtual void	setConvergenceMetrics (Real a_metric, Real a_tolerance)
	Set a convergence metric, along with solver tolerance, if desired. More...

Detailed Description

template<class T>
class LinearSolver< T >

Generic linear solver template. BiCGStab and others are built on top of this.

Constructor & Destructor Documentation

~LinearSolver()

template<class T>

virtual LinearSolver< T >::~LinearSolver ( )

inlinevirtual

Member Function Documentation

setHomogeneous()

template<class T>

virtual void LinearSolver< T >::setHomogeneous ( bool  a_homogeneous )

pure virtual

reset whether the solver is homogeneous.

Implemented in PetscSolver< T >, PetscSolver< LevelData< FArrayBox > >, MergeSolver< T >, GMRESSolver< T >, RelaxSolver< T >, BiCGStabSolver< T >, BiCGStabSolver< LevelData< FArrayBox > >, BiCGStabSolver< LevelData< T > >, NoOpSolver< T >, NoOpSolver< LevelDataType >, NoOpSolver< LevelData< EBCellFAB > >, NoOpSolver< LevelData< FArrayBox > >, NoOpSolver< LevelData< T > >, NoOpSolver< LevelData< MFCellFAB > >, MFSimpleSolver, and EBSimpleSolver.

define()

template<class T>

virtual void LinearSolver< T >::define ( LinearOp< T > *  a_operator, bool  a_homogeneous = false  )

pure virtual

Define the operator and whether it is a homogeneous solver or not. The LinearSolver does not take over ownership of this a_operator object. It does not call delete on it when the LinearSolver is deleted. It is meant to be like a late-binding reference. If you created a_operator with new, you should call delete on it after LinearSolver is deleted if you want to avoid memory leaks.

Implemented in PetscSolverPoisson< T >, PetscSolver< T >, PetscSolver< LevelData< FArrayBox > >, GMRESSolver< T >, MergeSolver< T >, RelaxSolver< T >, BiCGStabSolver< T >, BiCGStabSolver< LevelData< FArrayBox > >, BiCGStabSolver< LevelData< T > >, NoOpSolver< T >, NoOpSolver< LevelDataType >, NoOpSolver< LevelData< EBCellFAB > >, NoOpSolver< LevelData< FArrayBox > >, NoOpSolver< LevelData< T > >, NoOpSolver< LevelData< MFCellFAB > >, MFSimpleSolver, and EBSimpleSolver.

solve()

template<class T>

virtual void LinearSolver< T >::solve ( T &  a_phi, const T &  a_rhs  )

pure virtual

Solve L(phi) = rhs (phi = L^-1 (rhs)).

Implemented in PetscSolver< T >, PetscSolver< LevelData< FArrayBox > >, MergeSolver< T >, GMRESSolver< T >, RelaxSolver< T >, BiCGStabSolver< T >, BiCGStabSolver< LevelData< FArrayBox > >, BiCGStabSolver< LevelData< T > >, NoOpSolver< T >, NoOpSolver< LevelDataType >, NoOpSolver< LevelData< EBCellFAB > >, NoOpSolver< LevelData< FArrayBox > >, NoOpSolver< LevelData< T > >, NoOpSolver< LevelData< MFCellFAB > >, MFSimpleSolver, and EBSimpleSolver.

setConvergenceMetrics()

template<class T>

virtual void LinearSolver< T >::setConvergenceMetrics ( Real  a_metric, Real  a_tolerance  )

inlinevirtual

Set a convergence metric, along with solver tolerance, if desired.

Default implementation does nothing, since there are probably cases (liked direct solves), where this has no real meaning.

Reimplemented in MergeSolver< T >, BiCGStabSolver< T >, BiCGStabSolver< LevelData< FArrayBox > >, BiCGStabSolver< LevelData< T > >, GMRESSolver< T >, NoOpSolver< T >, NoOpSolver< LevelDataType >, NoOpSolver< LevelData< EBCellFAB > >, NoOpSolver< LevelData< FArrayBox > >, NoOpSolver< LevelData< T > >, and NoOpSolver< LevelData< MFCellFAB > >.

---

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

• LinearSolver.H