IMEXStepper solves non-linear time dependent problems with user-defined size of the time step and using Newthon's method for the solution of the non-linear problem. More...

#include <imex_stepper.h>

Inheritance diagram for IMEXStepper< VEC >:

Public Member Functions
	IMEXStepper (std::string &name="")
	Constructor for the IMEXStepper class. More...

virtual void	declare_parameters (ParameterHandler &prm)
	Declare parameters for this class to function properly. More...

unsigned int	solve_dae (VEC &solution, VEC &solution_dot)
	Evolve. More...

void	compute_consistent_initial_conditions (const double &t, VEC &y, VEC &y_dot)
	compute_consistent_initial_conditions More...

double	get_alpha () const
	if initial time is different from final time (i.e., we are solving a time-dep problem and not a stationay one, return the inverse of dt. More...

void	set_initial_time (const double &t)
	Set initial time equal to `t` disregarding what is written in the parameter file. More...

Public Member Functions inherited from ParameterAcceptor
	ParameterAcceptor (const std::string section_name="")
	The constructor adds derived classes to the list of acceptors. More...

virtual	~ParameterAcceptor ()
	The destructor sets to zero the pointer relative to this index, so that it is safe to destroy the mother class. More...

virtual void	parse_parameters (ParameterHandler &prm)
	Parse the parameter file. More...

virtual void	parse_parameters_call_back ()
	Parse parameter call back. More...

std::string	get_section_name () const
	Return the section name of this class. More...

std::vector< std::string >	get_section_path () const
	Travers all registered classes, and figure out what subsections we need to enter. More...

template<class T >
void	add_parameter (ParameterHandler &prm, T *parameter, const std::string &entry, const std::string &default_value, const Patterns::PatternBase &pattern=Patterns::Anything(), const std::string &documentation=std::string())
	Add a parameter the given parameter list. More...

template<class T >
void	add_parameter (T &parameter, const std::string &entry, const std::string &documentation=std::string(), ParameterHandler &prm=ParameterAcceptor::prm)
	Add a parameter to the global parameter handler ParameterAcceptor::prm. More...

void	enter_my_subsection (ParameterHandler &prm)
	Make sure we enter the right subsection of the global parameter file. More...

void	leave_my_subsection (ParameterHandler &prm)
	This function undoes what the enter_my_subsection() function did. More...

template<>
std_cxx11::shared_ptr< Patterns::PatternBase >	to_pattern (const double &)
	double More...

template<>
std::string	to_string (const double &entry)

template<>
double	to_type (const std::string &parameter)

template<>
std_cxx11::shared_ptr< Patterns::PatternBase >	to_pattern (const int &)
	int More...

template<>
std::string	to_string (const int &entry)

template<>
int	to_type (const std::string &parameter)

template<>
std_cxx11::shared_ptr< Patterns::PatternBase >	to_pattern (const unsigned int &)
	unsigned int More...

template<>
std::string	to_string (const unsigned int &entry)

template<>
unsigned int	to_type (const std::string &parameter)

template<>
std_cxx11::shared_ptr< Patterns::PatternBase >	to_pattern (const bool &)
	bool More...

template<>
std::string	to_string (const bool &entry)

template<>
bool	to_type (const std::string &parameter)

Public Member Functions inherited from Subscriptor
	Subscriptor ()

	Subscriptor (const Subscriptor &)

	Subscriptor (Subscriptor &&)

virtual	~Subscriptor ()

Subscriptor &	operator= (const Subscriptor &)

Subscriptor &	operator= (Subscriptor &&)

void	subscribe (const char *identifier=0) const

void	unsubscribe (const char *identifier=0) const

unsigned int	n_subscriptions () const

void	list_subscribers () const

void	serialize (Archive &ar, const unsigned int version)

Public Attributes
std::function< shared_ptr< VEC >)>	create_new_vector
	Return a shared_ptr<VEC>. More...

std::function< int(const double t, const VEC &y, const VEC &y_dot, VEC &res)>	residual
	Compute residual. More...

std::function< int(const double t, const VEC &y, const VEC &y_dot, const double alpha)>	setup_jacobian
	Compute Jacobian. More...

std::function< int(const VEC &rhs, VEC &dst)>	solve_jacobian_system
	Solve linear system. More...

std::function< void(const double t, const VEC &sol, const VEC &sol_dot, const unsigned int step_number)>	output_step
	Store solutions to file. More...

std::function< bool(const double t, VEC &sol, VEC &sol_dot)>	solver_should_restart
	Evaluate wether the mesh should be refined or not. More...

std::function< VEC &()>	get_lumped_mass_matrix
	Return the lumped mass matrix vector. More...

std::function< int(const VEC &src, VEC &dst)>	jacobian_vmult
	Compute the matrix-vector product Jacobian times `src`, and the result is put in `dst`. More...

std::function< double(const VEC &vector)>	vector_norm
	Return the norm of `vector`. More...

Private Member Functions
void	compute_y_dot (const VEC &y, const VEC &prev, const double alpha, VEC &y_dot)

double	evaluate_step_size (const double &t)
	evaluate step size at time `t` according to the expression stored in _step_size More...

double	line_search_with_backtracking (const VEC &update, const VEC &prev_sol, const double &alpha, const double &t, VEC &sol, VEC &sol_dot, VEC &residual)
	line search algorithm with backtracking.The following sequence of Newton relaxation parameters is tested: 1, 1/2, 1/4,...,2^-i. More...

void	do_newton (const double t, const double alpha, const bool update_Jacobian, const VEC &previous_solution, VEC &solution, VEC &solution_dot)
	find solution applying the newton method with given More...

void	compute_previous_solution (const VEC &sol, const VEC &sol_dot, const double &alpha, VEC &prev)
	compute previous solution from given More...

void	set_functions_to_trigger_an_assert ()
	Set the std::functions above to trigger an assert if they are not implemented. More...

Private Attributes
KINSOLInterface< VEC >	kinsol
	kinsol solver More...

double	step_size
	Step size. More...

std::string	_step_size
	user defined step_size More...

double	initial_time
	Initial time for the ode. More...

double	final_time
	Final time. More...

double	abs_tol
	Absolute error tolerance for non linear iterations. More...

double	rel_tol
	Relative error tolerance for non linear iterations. More...

unsigned int	output_period
	Seconds between each output. More...

double	newton_alpha
	Alpha to use in Newton method for the update of the solution. More...

unsigned int	max_outer_non_linear_iterations
	Maximum number of outer iterations for Newton method. More...

unsigned int	max_inner_non_linear_iterations
	Maximum number of inner iterations for Newton method. More...

bool	update_jacobian_continuously
	Jacobian is updated at each outer iteration and time step. More...

bool	use_kinsol
	use kinsol solver true or false More...

ConditionalOStream	pcout
	Output stream. More...

bool	verbose
	print useful informations More...

unsigned int	n_max_backtracking
	i max More...

std::string	method
	method used for alpha selection More...

Additional Inherited Members
Static Public Member Functions inherited from ParameterAcceptor
static void	initialize (const std::string filename="", const std::string outfilename="")
	Call declare_all_parameters(), read filename (if it is present as input parameter) and parse_all_parameters() on the static member prm. More...

static void	clear ()
	Clear class list and global parameter file. More...

static void	parse_all_parameters (ParameterHandler &prm=ParameterAcceptor::prm)
	Parse the given ParameterHandler. More...

static void	log_info ()
	Print information about all stored classes. More...

static void	declare_all_parameters (ParameterHandler &prm=ParameterAcceptor::prm)
	Initialize the ParameterHandler with all derived classes parameters.This function enters the subsection returned by get_section_name() for each derived class, and declares all parameters that were added using add_parameter(). More...

template<class T >
static std_cxx11::shared_ptr< Patterns::PatternBase >	to_pattern (const T &)
	Given a class T, construct its default pattern to be used when declaring parameters. More...

template<class T >
static T	to_type (const std::string &)
	Given a string, fill the value of the given parameter. More...

template<class T >
static std::string	to_string (const T &)
	Given a parameter, return a string containing the given parameter. More...

Static Public Member Functions inherited from Subscriptor
static::ExceptionBase &	ExcInUse (int arg1, char *arg2, std::string &arg3)

static::ExceptionBase &	ExcNoSubscriber (char arg1, char arg2)

Static Public Attributes inherited from ParameterAcceptor
static ParameterHandler	prm
	Static parameter. More...

Protected Attributes inherited from ParameterAcceptor
const std::string	section_name
	The subsection name for this class. More...

Detailed Description

template<typename VEC = Vector<double>>
class IMEXStepper< VEC >

IMEXStepper solves non-linear time dependent problems with user-defined size of the time step and using Newthon's method for the solution of the non-linear problem.

It allows to use the Kinsol solver of SUNDIALS.

The user has to provide the following std::functions:

create_new_vector;
residual;
setup_jacobian;
solve_jacobian_system;
output_step;
solver_should_restart;
get_lumped_mass_matrix (only for kinsol);
jacobian_vmult (only for kinsol);
vector_norm (if kinsol is not used).

Definition at line 55 of file imex_stepper.h.

Constructor & Destructor Documentation

§ IMEXStepper()

template<typename VEC >

IMEXStepper< VEC >::IMEXStepper ( std::string & name = "" )

Constructor for the IMEXStepper class.

Takes a name for the section in parameter file.

Definition at line 63 of file imex_stepper.cc.

Member Function Documentation

§ compute_consistent_initial_conditions()

template<typename VEC >

void IMEXStepper< VEC >::compute_consistent_initial_conditions	(	const double &	t,
		VEC &	y,
		VEC &	y_dot
	)

compute_consistent_initial_conditions

Parameters

t

Definition at line 576 of file imex_stepper.cc.

§ compute_previous_solution()

template<typename VEC >

void IMEXStepper< VEC >::compute_previous_solution	(	const VEC &	sol,
		const VEC &	sol_dot,
		const double &	alpha,
		VEC &	prev
	)

private

compute previous solution from given

Parameters

sol
sol_dot
alpha

Definition at line 539 of file imex_stepper.cc.

§ compute_y_dot()

template<typename VEC >

void IMEXStepper< VEC >::compute_y_dot	(	const VEC &	y,
		const VEC &	prev,
		const double	alpha,
		VEC &	y_dot
	)

private

Definition at line 160 of file imex_stepper.cc.

§ declare_parameters()

template<typename VEC >

void IMEXStepper< VEC >::declare_parameters ( ParameterHandler & prm )

virtual

Declare parameters for this class to function properly.

Reimplemented from ParameterAcceptor.

Definition at line 91 of file imex_stepper.cc.

§ do_newton()

template<typename VEC >

void IMEXStepper< VEC >::do_newton	(	const double	t,
		const double	alpha,
		const bool	update_Jacobian,
		const VEC &	previous_solution,
		VEC &	solution,
		VEC &	solution_dot
	)

private

find solution applying the newton method with given

Parameters

t
alpha
update_Jacobian
previous_solution
solution_dot	at the end of the computation, the `solution_dot` is updated as well.

this function is called when KINSOL is NOT used

Definition at line 388 of file imex_stepper.cc.

§ evaluate_step_size()

template<typename VEC >

double IMEXStepper< VEC >::evaluate_step_size ( const double & t )

private

evaluate step size at time t according to the expression stored in _step_size

Definition at line 559 of file imex_stepper.cc.

§ get_alpha()

template<typename VEC >

double IMEXStepper< VEC >::get_alpha ( ) const

if initial time is different from final time (i.e., we are solving a time-dep problem and not a stationay one, return the inverse of dt.

If the problem is stationary, returns 0.

Returns

Definition at line 73 of file imex_stepper.cc.

§ line_search_with_backtracking()

template<typename VEC >

double IMEXStepper< VEC >::line_search_with_backtracking	(	const VEC &	update,
		const VEC &	prev_sol,
		const double &	alpha,
		const double &	t,
		VEC &	sol,
		VEC &	sol_dot,
		VEC &	residual
	)

private

line search algorithm with backtracking.The following sequence of Newton relaxation parameters is tested: 1, 1/2, 1/4,...,2^-i.

it returns the selected alpha and solution, solution_dot and residual are accordingly updated.

Definition at line 315 of file imex_stepper.cc.

§ set_functions_to_trigger_an_assert()

template<typename VEC >

void IMEXStepper< VEC >::set_functions_to_trigger_an_assert ( )

private

Set the std::functions above to trigger an assert if they are not implemented.

Definition at line 610 of file imex_stepper.cc.

§ set_initial_time()

template<typename VEC >

void IMEXStepper< VEC >::set_initial_time ( const double & t )

Set initial time equal to t disregarding what is written in the parameter file.

Definition at line 85 of file imex_stepper.cc.

§ solve_dae()

template<typename VEC >

unsigned int IMEXStepper< VEC >::solve_dae	(	VEC &	solution,
		VEC &	solution_dot
	)

Evolve.

This function returns the final number of steps.

Definition at line 168 of file imex_stepper.cc.