A newton solver, using local error estimates for each vector component. More...

#include <newton.hh>

Public Types
using	VectorType = VectorType_

Public Member Functions
	Newton (double abstol_=1e-10, double reltol_=1e-6, double assemble_threshold_=0., uint max_steps_=21, uint n_stepsize_iterations_=21)

void	reinit (const VectorType &u)
	Should it be necessary, one can force a recalculation of the jacobian hereby.

void	operator() (VectorType &iterate)
	Perform a newton iteration on the input vector.

double	threshold (const double thr)
	Set the jacobian recalculation threshold, i.e. below which reduction we force a jacobian update.

const VectorType &	get_residual () const

const VectorType &	get_initial () const

const VectorType &	get_iterate () const

double	average_time_newton () const

uint	num_newton_calls () const

void	set_ignore_nonconv (bool x)
	Make the newton algorithm return a success even if the evolution did not fulfill the convergence criterion. The iteration always fails if the solution contains NaNs or infinities.

double	get_error ()
	Get the latest error.

uint	get_step ()
	Get the number of steps taken in the latest iteration.

uint	get_jacobians ()
	Get the number of jacobians created in the latest iteration.

Public Attributes
std::function< void(VectorType &, const VectorType &)>	residual

std::function< void(VectorType &, const VectorType &)>	lin_solve

std::function< void(const VectorType &)>	update_jacobian

Private Member Functions
double	get_EEst ()
	Calculate the current error estimate.

bool	check ()
	Check if the iteration should go on.

Private Attributes
VectorType	residual_vector

VectorType	initial_vector

VectorType	iterate_vector

double	abstol

double	reltol

double	assemble_threshold

uint	max_steps

uint	n_stepsize_iterations

bool	converged

bool	ignore_nonconv

double	resnorm

uint	step

uint	jacobians

std::vector< double >	timings_newton

Detailed Description

template<typename VectorType_>
class DiFfRG::Newton< VectorType_ >

A newton solver, using local error estimates for each vector component.

Template Parameters

VectorType_ Vector type used by the simulation (i.e. any dealii vector type)

Member Typedef Documentation

◆ VectorType

template<typename VectorType_ >

using DiFfRG::Newton< VectorType_ >::VectorType = VectorType_

Constructor & Destructor Documentation

◆ Newton()

template<typename VectorType_ >

DiFfRG::Newton< VectorType_ >::Newton	(	double	abstol_ = 1e-10,
		double	reltol_ = 1e-6,
		double	assemble_threshold_ = 0.,
		uint	max_steps_ = 21,
		uint	n_stepsize_iterations_ = 21 )

inline

Member Function Documentation

◆ average_time_newton()

template<typename VectorType_ >

double DiFfRG::Newton< VectorType_ >::average_time_newton ( ) const

inline

◆ check()

template<typename VectorType_ >

bool DiFfRG::Newton< VectorType_ >::check ( )

inlineprivate

Check if the iteration should go on.

Returns: true; false

◆ get_EEst()

template<typename VectorType_ >

double DiFfRG::Newton< VectorType_ >::get_EEst ( )

inlineprivate

Calculate the current error estimate.

Returns: double

◆ get_error()

template<typename VectorType_ >

double DiFfRG::Newton< VectorType_ >::get_error ( )

inline

Get the latest error.

◆ get_initial()

template<typename VectorType_ >

const VectorType & DiFfRG::Newton< VectorType_ >::get_initial ( ) const

inline

◆ get_iterate()

template<typename VectorType_ >

const VectorType & DiFfRG::Newton< VectorType_ >::get_iterate ( ) const

inline

◆ get_jacobians()

template<typename VectorType_ >

uint DiFfRG::Newton< VectorType_ >::get_jacobians ( )

inline

Get the number of jacobians created in the latest iteration.

◆ get_residual()

template<typename VectorType_ >

const VectorType & DiFfRG::Newton< VectorType_ >::get_residual ( ) const

inline

◆ get_step()

template<typename VectorType_ >

uint DiFfRG::Newton< VectorType_ >::get_step ( )

inline

Get the number of steps taken in the latest iteration.

◆ num_newton_calls()

template<typename VectorType_ >

uint DiFfRG::Newton< VectorType_ >::num_newton_calls ( ) const

inline

◆ operator()()

template<typename VectorType_ >

void DiFfRG::Newton< VectorType_ >::operator() ( VectorType & iterate )

inline

Perform a newton iteration on the input vector.

Parameters

iterate the input vector which contains the initial guess and later the solution.

◆ reinit()

template<typename VectorType_ >

void DiFfRG::Newton< VectorType_ >::reinit ( const VectorType & u )

inline

Should it be necessary, one can force a recalculation of the jacobian hereby.

◆ set_ignore_nonconv()

template<typename VectorType_ >

void DiFfRG::Newton< VectorType_ >::set_ignore_nonconv ( bool x )

inline

Make the newton algorithm return a success even if the evolution did not fulfill the convergence criterion. The iteration always fails if the solution contains NaNs or infinities.

Parameters

x

◆ threshold()

template<typename VectorType_ >

double DiFfRG::Newton< VectorType_ >::threshold ( const double thr )

inline

Set the jacobian recalculation threshold, i.e. below which reduction we force a jacobian update.

Returns: double the old threshold

Member Data Documentation

◆ abstol

template<typename VectorType_ >

double DiFfRG::Newton< VectorType_ >::abstol

private

◆ assemble_threshold

template<typename VectorType_ >

double DiFfRG::Newton< VectorType_ >::assemble_threshold

private

◆ converged

template<typename VectorType_ >

bool DiFfRG::Newton< VectorType_ >::converged

private

◆ ignore_nonconv

template<typename VectorType_ >

bool DiFfRG::Newton< VectorType_ >::ignore_nonconv

private

◆ initial_vector

template<typename VectorType_ >

VectorType DiFfRG::Newton< VectorType_ >::initial_vector

private

◆ iterate_vector

template<typename VectorType_ >

VectorType DiFfRG::Newton< VectorType_ >::iterate_vector

private

◆ jacobians

template<typename VectorType_ >

uint DiFfRG::Newton< VectorType_ >::jacobians

private

◆ lin_solve

template<typename VectorType_ >

std::function<void(VectorType &, const VectorType &)> DiFfRG::Newton< VectorType_ >::lin_solve

◆ max_steps

template<typename VectorType_ >

uint DiFfRG::Newton< VectorType_ >::max_steps

private

◆ n_stepsize_iterations

template<typename VectorType_ >

uint DiFfRG::Newton< VectorType_ >::n_stepsize_iterations

private

◆ reltol

template<typename VectorType_ >

double DiFfRG::Newton< VectorType_ >::reltol

private

◆ residual

template<typename VectorType_ >

std::function<void(VectorType &, const VectorType &)> DiFfRG::Newton< VectorType_ >::residual

◆ residual_vector

template<typename VectorType_ >

VectorType DiFfRG::Newton< VectorType_ >::residual_vector

private

◆ resnorm

template<typename VectorType_ >

double DiFfRG::Newton< VectorType_ >::resnorm

private

◆ step

template<typename VectorType_ >

uint DiFfRG::Newton< VectorType_ >::step

private

◆ timings_newton

template<typename VectorType_ >

std::vector<double> DiFfRG::Newton< VectorType_ >::timings_newton

private

◆ update_jacobian

template<typename VectorType_ >

std::function<void(const VectorType &)> DiFfRG::Newton< VectorType_ >::update_jacobian

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

/home/runner/work/DiFfRG/DiFfRG/DiFfRG/include/DiFfRG/timestepping/solver/newton.hh

Public Types

Public Member Functions

Public Attributes

Private Member Functions

Private Attributes

Detailed Description

Member Typedef Documentation

◆ VectorType

Constructor & Destructor Documentation

◆ Newton()

Member Function Documentation

◆ average_time_newton()

◆ check()

◆ get_EEst()

◆ get_error()

◆ get_initial()

◆ get_iterate()

◆ get_jacobians()

◆ get_residual()

◆ get_step()

◆ num_newton_calls()

◆ operator()()

◆ reinit()

◆ set_ignore_nonconv()

◆ threshold()

Member Data Documentation

◆ abstol

◆ assemble_threshold

◆ converged

◆ ignore_nonconv

◆ initial_vector

◆ iterate_vector

◆ jacobians

◆ lin_solve

◆ max_steps

◆ n_stepsize_iterations

◆ reltol

◆ residual

◆ residual_vector

◆ resnorm

◆ step

◆ timings_newton

◆ update_jacobian