NonlinearSolverSelector< VectorType >::AdditionalData Class Reference

#include <deal.II/numerics/nonlinear.h>

Public Types
enum	SolutionStrategy { newton , linesearch , picard }
enum	SolverType { automatic , kinsol , nox , petsc_snes }

Public Member Functions
	AdditionalData (const SolverType &solver_type=automatic, const SolutionStrategy &strategy=linesearch, const unsigned int maximum_non_linear_iterations=200, const double function_tolerance=1e-8, const double relative_tolerance=1e-5, const double step_tolerance=0.0, const unsigned int anderson_subspace_size=0)

Public Attributes
SolverType	solver_type
SolutionStrategy	strategy
unsigned int	maximum_non_linear_iterations
double	function_tolerance
const double	relative_tolerance
double	step_tolerance
unsigned int	anderson_subspace_size

Detailed Description

template<typename VectorType = Vector<double>>
class NonlinearSolverSelector< VectorType >::AdditionalData

Additional data that will be sent through the NonlinearSolverSelector class, then into the specific nonlinear solver classes themselves. Many of these parameters can also be foud in the documentation for KINSOL and NOX.

Definition at line 85 of file nonlinear.h.

Member Enumeration Documentation

SolutionStrategy

template<typename VectorType = Vector<double>>

enum NonlinearSolverSelector::AdditionalData::SolutionStrategy

NonlinearSolverSelector solution strategy. The solver types included in this class both use a Newton-Krylov solver with a line search and even Picard iterations (KINSOL).

Enumerator
newton	Standard Newton iteration.
linesearch	Newton iteration with line search.
picard	Picard iteration.

Definition at line 93 of file nonlinear.h.

SolverType

template<typename VectorType = Vector<double>>

enum NonlinearSolverSelector::AdditionalData::SolverType

Enumerator
automatic	Default parameter, will use whatever solver is available with KINSOL as the priority.
kinsol	KINSOL nonlinear solver, part of the SUNDIALS package.
nox	NOX nonlinear solver, part of the TRILINOS package.
petsc_snes

Definition at line 109 of file nonlinear.h.

Constructor & Destructor Documentation

AdditionalData()

template<typename VectorType >

NonlinearSolverSelector< VectorType >::AdditionalData::AdditionalData	(	const SolverType &	solver_type = automatic,
		const SolutionStrategy &	strategy = linesearch,
		const unsigned int	maximum_non_linear_iterations = 200,
		const double	function_tolerance = 1e-8,
		const double	relative_tolerance = 1e-5,
		const double	step_tolerance = 0.0,
		const unsigned int	anderson_subspace_size = 0
	)

Initialization parameters for NonlinearSolverSelector.

Parameters

solver_type	Nonlinear solver type.
strategy	Method of solving the nonlinear problem.
maximum_non_linear_iterations	Maximum number of nonlinear iterations.
function_tolerance	Absolute stopping tolerance for the norm of the residual \(F(u)\).
relative_tolerance	Relative stopping tolerance.
step_tolerance	Tolerance for minimum scaled step length
anderson_subspace_size	Size of the Anderson acceleration subspace, use 0 to disable.

Definition at line 523 of file nonlinear.h.

Member Data Documentation

solver_type

template<typename VectorType = Vector<double>>

SolverType NonlinearSolverSelector< VectorType >::AdditionalData::solver_type

The type of nonlinear solver to use. If the default 'automatic' is used, it will choose the first available package in the order KINSOL, NOX, or SNES.

Definition at line 157 of file nonlinear.h.

strategy

template<typename VectorType = Vector<double>>

SolutionStrategy NonlinearSolverSelector< VectorType >::AdditionalData::strategy

The solution strategy to use. For this class, you can choose from SolutionStrategy::newton, SolutionStrategy::linesearch, or SolutionStrategy::picard. More details on this can be found on the KINSOL documentation.

Definition at line 165 of file nonlinear.h.

maximum_non_linear_iterations

template<typename VectorType = Vector<double>>

unsigned int NonlinearSolverSelector< VectorType >::AdditionalData::maximum_non_linear_iterations

Maximum number of nonlinear iterations allowed.

Definition at line 170 of file nonlinear.h.

function_tolerance

template<typename VectorType = Vector<double>>

double NonlinearSolverSelector< VectorType >::AdditionalData::function_tolerance

A scalar used as a stopping tolerance on the scaled maximum norm of the system function \(F(u)\) or \(G(u)\).

If set to zero, default values will be used.

Definition at line 178 of file nonlinear.h.

relative_tolerance

template<typename VectorType = Vector<double>>

const double NonlinearSolverSelector< VectorType >::AdditionalData::relative_tolerance

Relative \(l_2\) tolerance of the residual to be reached.

Note

Solver terminates successfully if either the function tolerance or the relative tolerance has been reached.

Definition at line 186 of file nonlinear.h.

step_tolerance

template<typename VectorType = Vector<double>>

double NonlinearSolverSelector< VectorType >::AdditionalData::step_tolerance

A scalar used as a stopping tolerance on the minimum scaled step length.

If set to zero, default values will be used.

Definition at line 194 of file nonlinear.h.

anderson_subspace_size

template<typename VectorType = Vector<double>>

unsigned int NonlinearSolverSelector< VectorType >::AdditionalData::anderson_subspace_size

The size of the subspace used with Anderson acceleration in conjunction with Picard iteration.

If you set this to 0, no acceleration is used.

Definition at line 202 of file nonlinear.h.

---

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

• include/deal.II/numerics/nonlinear.h