SAMRAI::solv::KINSOL_SAMRAIContext< DIM > Class Template Reference

Wraps the KINSOLSolver C++ wrapper class so that KINSOL may be used in applications that require a nonlinear solver. More...

#include <source/solvers/packages/sundials/kinsol/KINSOL_SAMRAIContext.h>

Inheritance diagram for SAMRAI::solv::KINSOL_SAMRAIContext< DIM >:

[legend]List of all members.

Public Member Functions
	KINSOL_SAMRAIContext (const std::string &object_name, tbox::Pointer< tbox::Database > input_db, KINSOLAbstractFunctions *my_functions)
	~KINSOL_SAMRAIContext ()
void	initialize (tbox::Pointer< SAMRAIVectorReal< DIM, double > > solution)
int	solve ()
KINSOLSolver *	getKINSOLSolver ()
void	getFromInput (tbox::Pointer< tbox::Database > db)
void	getFromRestart ()
void	putToDatabase (tbox::Pointer< tbox::Database > db)
virtual void	printClassData (std::ostream &os) const

---

Detailed Description

template<int DIM>
class SAMRAI::solv::KINSOL_SAMRAIContext< DIM >

Wraps the KINSOLSolver C++ wrapper class so that KINSOL may be used in applications that require a nonlinear solver.

Class KINSOL_SAMRAIContext<DIM> wraps the KINSOLSolver C++ wrapper class so that KINSOL may be used in applications that require a nonlinear solver. The class KINSOLSolver does not depend on SAMRAI. Making it derived from the SAMRAI nonlinear solver interface would require the KINSOL C++ wrapper to depend on SAMRAI.

Important note: This class can only create a KINSOL C++ wrapper instance, initialize it in a rudimentary way, and invoke the solution process. All other interaction with the nonlinear solver (i.e., setting parameters, retrieving solver statistics, etc.) must be done directly with the KINSOL wrapper accessible via the routine getKINSOLSolver(). Alternatively, solver parameters may be set up at initialization time using the SAMRAI input database.

If no parameters are read from input, KINSOL defaults are used. See KINSOL documentation for default information. Optional input keys and types are:

• residual_stop_tolerance double value for stopping tolerance on norm of scaled residual.

• max_nonlinear_iterations integer value for maximum number of nonlinear iterations (MXITER).

• max_krylov_dimension integer value for maximum dimension of Krylov space.

• global_newton_strategy integer flag for globalization strategy. Choices are "INEXACT_NEWTON" (default) and "LINESEARCH".

• max_newton_step double value for maximum allowable Newton step (MXNEWTONSTEP).

• nonlinear_step_tolerance double value for stopping tolerance on maximum entry in scaled Newton step.

• relative_function_error double value for relative error in function evaluation (RELFUNC).

• solution_update_constraint double value for constraint on relative change in solution (RELU).

• linear_convergence_test integer flag for linear solver convergence tolerance (ETACHOICE). Choices are "ETACONSTANT" (default), "ETACHOICE1", ETACHOICE2".

• max_sub_setup_calls number of nonlinear iterations between checks by the nonlinear residual monitoring algorithm (specifies lenght of subinterval) NOTE: should be a multiple of MaxStepsWithNoPrecondSetup

• residual_monitoring_params values of omega_min and omega_max scalars used by nonlinear residual monitoring algorithm. Default is [0.00001 and 0.9]

• residual_monitoring_constant constant value used by residual monitoring algorithm. If omega=0, then it is estimated using omega_min and omega_max. Default is 0.0.

• no_min_eps flag control whether or not the value * of eps is bounded below by 0.01*fnormtol. FALSE = "constrain value of eps by setting to the following: eps = MAX{0.01*fnormtol, eps}" TRUE = "do notconstrain value of eps". Default is FALSE

• eisenstat_walker_params array of two double values Eisenstat-Walker choice 2; i.e., the values are given as ETAALPHA, followed by ETAGAMMA. Note: the values only apply when linear convergence test is set to ETACHOICE2.

• linear_solver_relative_tolerance double value for constant linear solver relative tolerance (ETACONST). Note: value only apply when convergence test is set to ETACONSTANT.

• precond_setup_flag integer flag for preconditioner setup strategy (PRECOND_NO_INIT).

• max_solves_no_precond_setup integer value for number of nonlinear steps separating successive calls to preconditioner setup routine.

• max_linear_solve_restarts integer value for maximum number of linear solver restarts allowed.

• KINSOL_log_filename string value for name of KINSOL log file; default is "kinsol.log".

• KINSOL_print_flag integer flag for KINSOL log file print options (PRINTFL).

• uses_preconditioner boolean flag indicating whether a preconditioner is supplied. Default is false.

• uses_jac_times_vector boolean flag indicating whether an analytic Jacobian-vector product is supplied. Default is false.

Note that all input values may override values read in from restart. If no new input value is given, the restart value is used.

A sample input file entry might look like:

     residual_stop_tolerance  =  10.e-6
     max_nonlinear_iterations =  200
     max_newton_step          =  0.1
     KINSOL_log_filename      =  "mylogfile"
     KINSOL_print_flag        =  3   // print all output KINSOL has to offer

See also:

solv::NonlinearSolverStrategy

---

Constructor & Destructor Documentation

template<int DIM>

SAMRAI::solv::KINSOL_SAMRAIContext< DIM >::KINSOL_SAMRAIContext	(	const std::string &	object_name,
		tbox::Pointer< tbox::Database >	input_db,
		KINSOLAbstractFunctions *	my_functions
	)

Constructor for algs::KINSOL_SAMRAIContext allocates the KINSOL C++ wrapper object and initializes rudimentary state associated with user-supplied solver components. Then, it reads solver parameter from input and restart which may override default values.

When assertion checking is active, an unrecoverable exception will result if the name string is empty or the pointer to the user-defined KINSOL functions object is null.

template<int DIM>

SAMRAI::solv::KINSOL_SAMRAIContext< DIM >::~KINSOL_SAMRAIContext

(

)

Destructor for algs::KINSOL_SAMRAIContext destroys the KINSOL C++ wrapper object and the KINSOL solution vector wrapper.

---

Member Function Documentation

template<int DIM>

void SAMRAI::solv::KINSOL_SAMRAIContext< DIM >::initialize

(

tbox::Pointer< SAMRAIVectorReal< DIM, double > >

solution

)

[virtual]

Initialize the state of KINSOL based on vector argument representing the solution of the nonlinear system. In general, this routine must be called before the solve() routine is invoked.

Implements SAMRAI::solv::NonlinearSolverStrategy< DIM >.

template<int DIM>

int SAMRAI::solv::KINSOL_SAMRAIContext< DIM >::solve

(

)

[virtual]

Solve the nonlinear problem and return and integer value defined by KINSOL. A return value of 1 indicates success (i.e., KINSOL_SUCCESS). Consult the KINSOL documentation, KINSOL header file kinsol.h, or the header file for the class KINSOLSolver for more information about KINSOL return codes. In general, the initialize() routine must be called before this solve function to set up the solver.

Implements SAMRAI::solv::NonlinearSolverStrategy< DIM >.

template<int DIM>

KINSOLSolver * SAMRAI::solv::KINSOL_SAMRAIContext< DIM >::getKINSOLSolver

(

)

Return pointer to KINSOL solver C++ wrapper object.

template<int DIM>

void SAMRAI::solv::KINSOL_SAMRAIContext< DIM >::getFromInput

(

tbox::Pointer< tbox::Database >

db

)

Read input parameters from given database.

When assertion checking is active, an unrecoverable exception will result if the database pointer is null.

template<int DIM>

void SAMRAI::solv::KINSOL_SAMRAIContext< DIM >::getFromRestart

(

)

Retrieve solver parameters from restart database matching object name.

When assertion checking is active, an unrecoverable exception will result if a restart database matching the object name does not exist, or if the class version number does not match that in restart.

template<int DIM>

void SAMRAI::solv::KINSOL_SAMRAIContext< DIM >::putToDatabase

(

tbox::Pointer< tbox::Database >

db

)

[virtual]

Retrieve solver parameters from restart database matching object name.

When assertion checking is active, an unrecoverable exception will result if database pointer is null.

Implements SAMRAI::tbox::Serializable.

template<int DIM>

void SAMRAI::solv::KINSOL_SAMRAIContext< DIM >::printClassData

(

std::ostream &

os

)

const [virtual]

Print out all members of integrator instance to given output stream.

---

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

• source/solvers/packages/sundials/kinsol/KINSOL_SAMRAIContext.h
• source/solvers/packages/sundials/kinsol/KINSOL_SAMRAIContext.C

---