Wrapper class to SUNDIALS CVODE. More...

#include <CVode.h>

Public Types
enum	LMM { cv_bdf = CV_BDF, cv_adams = CV_ADAMS }

enum	ITER { cv_functional = CV_FUNCTIONAL, cv_newton = CV_NEWTON }

Public Member Functions
	CVode (LMM cv_lmm, ITER cv_iter)

virtual	~CVode ()
	Destructor.

void	init (std::shared_ptr< GenericVector > u0, double atol, double rtol, long int mxsteps=0)

double	step (double dt)

double	get_time () const

void	set_time (double t0)

virtual void	derivs (double t, std::shared_ptr< GenericVector > u, std::shared_ptr< GenericVector > udot)

virtual int	jacobian (std::shared_ptr< const GenericVector > v, std::shared_ptr< GenericVector > Jv, double t, std::shared_ptr< const GenericVector > y, std::shared_ptr< const GenericVector > fy)

virtual int	jacobian_setup (double t, std::shared_ptr< GenericVector > Jv, std::shared_ptr< GenericVector > y)

virtual int	psolve (double tn, std::shared_ptr< const GenericVector >y, std::shared_ptr< const GenericVector > fy, std::shared_ptr< const GenericVector > r, std::shared_ptr< GenericVector > z, double gamma, double delta, int lr)

std::map< std::string, double >	statistics ()

Detailed Description

Wrapper class to SUNDIALS CVODE.

Constructor & Destructor Documentation

CVode::CVode	(	LMM	cv_lmm,
		ITER	cv_iter
	)

Constructor

Parameters

cv_lmm	linear multistep method
cv_iter	iteration type

Member Function Documentation

void CVode::derivs	(	double	t,
		std::shared_ptr< GenericVector >	u,
		std::shared_ptr< GenericVector >	udot
	)

virtual

Overloaded function for time derivatives of u at time t. Given the vector u, at time t, provide the time derivative udot.

Parameters

t	time
u	input vector of values u
udot	output vector containing computed derivative of u at time t

double CVode::get_time ( ) const

Get current time

Returns: double current time

void CVode::init	(	std::shared_ptr< GenericVector >	u0,
		double	atol,
		double	rtol,
		long int	mxsteps = `0`
	)

Initialise CVode

Parameters

u0	Input vector
atol	absolute tolerance
rtol	relative tolerance
mxsteps	maximum number of steps

int CVode::jacobian	(	std::shared_ptr< const GenericVector >	v,
		std::shared_ptr< GenericVector >	Jv,
		double	t,
		std::shared_ptr< const GenericVector >	y,
		std::shared_ptr< const GenericVector >	fy
	)

virtual

Given the values (t, y, fy, v), compute Jv = (df/dy)v

Parameters

v	vector to be multiplied by the Jacobian df/dy
Jv	output vector of (df/dy)*v
t	current value of the independent variable.
y	current value of the ependent variable.
fy	vector f(t,y)

Returns: success flag, 0 if successful

int CVode::jacobian_setup	(	double	t,
		std::shared_ptr< GenericVector >	Jv,
		std::shared_ptr< GenericVector >	y
	)

virtual

User-defined setup function called once per Newton iteration. Data structures for usage by the Jacobian function can be setup here

Parameters

t	current value of the independent variable
Jv	current value of the dependent variable vector, namely the predicted value of y(t).
y	vector f(t,y).

Returns: success flag, 0 if success

int CVode::psolve	(	double	tn,
		std::shared_ptr< const GenericVector >	y,
		std::shared_ptr< const GenericVector >	fy,
		std::shared_ptr< const GenericVector >	r,
		std::shared_ptr< GenericVector >	z,
		double	gamma,
		double	delta,
		int	lr
	)

virtual

Overloaded preconditioner solver function

Parameters

tn	current value of the independent variable.
y	current value of the dependent variable vector.
fy	vector f(t,y)
r	right-hand side vector of the linear system.
z	output vector computed by PrecSolve.
gamma	scalar appearing in the Newton matrix.
delta	input tolerance if an iterative method is used.
lr	input flag indicating whether to use left or right preconditioner.

Returns: success flag, 0 if success

Overloaded preconditioner solver function

void CVode::set_time ( double t0 )

Set the current time

Parameters

t0	current time

std::map< std::string, double > CVode::statistics ( )

Return statistics

Returns: map structure containing information stored in the CVode structure, ie. number of solver steps, RHS evaluations, current time

double CVode::step ( double dt )

Advance time by timestep dt

Parameters

dt timestep

Returns: CVODE return flag

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

/home/fenics/shared/dolfin/ts/CVode.h
/home/fenics/shared/dolfin/ts/CVode.cpp

Public Types

Public Member Functions

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation