Opm::LinearSolverFactory Class Reference

#include <LinearSolverFactory.hpp>

Inheritance diagram for Opm::LinearSolverFactory:

Public Member Functions
	LinearSolverFactory ()
	Default constructor. More...
	LinearSolverFactory (const parameter::ParameterGroup &param)
virtual	~LinearSolverFactory ()
	Destructor. More...
virtual LinearSolverReport	solve (const int size, const int nonzeros, const int *ia, const int *ja, const double *sa, const double *rhs, double *solution, const boost::any &add=boost::any()) const
virtual void	setTolerance (const double tol)
virtual double	getTolerance () const
LinearSolverReport	solve (const CSRMatrix *A, const double *rhs, double *solution) const

Detailed Description

Concrete class encapsulating any available linear solver. For the moment, this means UMFPACK and dune-istl. Since both are optional dependencies, either or both may be unavailable, depending on configuration.

Constructor & Destructor Documentation

Opm::LinearSolverFactory::LinearSolverFactory

(

)

Default constructor.

Opm::LinearSolverFactory::LinearSolverFactory

(

const parameter::ParameterGroup &

param

)

Construct from parameters. The accepted parameters are (default) (allowed values): linsolver ("umfpack") ("umfpack", "istl", "petsc") For the umfpack solver to be available, this class must be compiled with UMFPACK support, as indicated by the variable HAVE_SUITESPARSE_UMFPACK_H in config.h. For the istl solver to be available, this class must be compiled with dune-istl support, as indicated by the variable HAVE_DUNE_ISTL in config.h. For the petsc solver to be available, this class must be compiled with petsc support, as indicated by the variable HAVE_PETSC in config.h. Any further parameters are passed on to the constructors of the actual solver used, see LinearSolverUmfpack, LinearSolverIstl and LinearSolverPetsc for details.

virtual Opm::LinearSolverFactory::~LinearSolverFactory ( )

virtual

Destructor.

Member Function Documentation

virtual double Opm::LinearSolverFactory::getTolerance ( ) const

virtual

Get tolerance for the linear solver.

Parameters

[out]

tolerance

value Not used for LinearSolverFactory. Returns -1.

Implements Opm::LinearSolverInterface.

virtual void Opm::LinearSolverFactory::setTolerance ( const double  tol )

virtual

Set tolerance for the linear solver.

Parameters

[in]

tol

tolerance value Not used for LinearSolverFactory

Implements Opm::LinearSolverInterface.

LinearSolverReport Opm::LinearSolverInterface::solve ( const CSRMatrix *  A, const double *  rhs, double *  solution  ) const

inherited

Solve a linear system, with a matrix given in compressed sparse row format.

Parameters

[in]	A	matrix in CSR format
[in]	rhs	array of length A->m containing the right hand side
[in,out]	solution	array of length A->m to which the solution will be written, may also be used as initial guess by iterative solvers. Note: this method is a convenience method that calls the virtual solve() method.

virtual LinearSolverReport Opm::LinearSolverFactory::solve ( const int  size, const int  nonzeros, const int *  ia, const int *  ja, const double *  sa, const double *  rhs, double *  solution, const boost::any &  add = boost::any()  ) const

virtual

Solve a linear system, with a matrix given in compressed sparse row format.

Parameters

[in]	size	# of rows in matrix
[in]	nonzeros	# of nonzeros elements in matrix
[in]	ia	array of length (size + 1) containing start and end indices for each row
[in]	ja	array of length nonzeros containing column numbers for the nonzero elements
[in]	sa	array of length nonzeros containing the values of the nonzero elements
[in]	rhs	array of length size containing the right hand side
[in,out]	solution	array of length size to which the solution will be written, may also be used as initial guess by iterative solvers.

Implements Opm::LinearSolverInterface.

---

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

• LinearSolverFactory.hpp