Wrappers for the function evaluators for Nonlinear solvers and Time steppers. More...

#include <ResidJacEval.h>

Inheritance diagram for ResidJacEval:

Collaboration diagram for ResidJacEval:

Public Member Functions
	ResidJacEval (doublereal atol=1.0e-13)
	Default constructor. More...

	ResidJacEval (const ResidJacEval &right)
	Copy Constructor. More...

ResidJacEval &	operator= (const ResidJacEval &right)
	Assignment operator. More...

virtual ResidJacEval *	duplMyselfAsResidJacEval () const
	Duplication routine for objects derived from residJacEval. More...

virtual int	nEquations () const
	Return the number of equations in the equation system. More...

virtual int	evalResidNJ (const doublereal t, const doublereal delta_t, const doublereal const y, const doublereal const ydot, doublereal *const resid, const ResidEval_Type_Enum evalType=Base_ResidEval, const int id_x=-1, const doublereal delta_x=0.0)
	Evaluate the residual function. More...

virtual int	eval (const doublereal t, const doublereal const y, const doublereal const ydot, doublereal *const r)
	Evaluate the residual function. More...

virtual int	getInitialConditions (const doublereal t0, doublereal const y, doublereal const ydot)
	Fill in the initial conditions. More...

virtual doublereal	filterNewStep (const doublereal t, const doublereal const ybase, doublereal const step)
	Filter the solution predictions. More...

virtual doublereal	filterSolnPrediction (const doublereal t, doublereal *const y)
	Filter the solution predictions. More...

void	setAtol (doublereal atol)
	Set a global value of the absolute tolerance. More...

virtual int	evalTimeTrackingEqns (const doublereal t, const doublereal delta_t, const doublereal const y, const doublereal const ydot)
	Evaluate the time tracking equations, if any. More...

virtual bool	evalStoppingCritera (const doublereal t, const doublereal delta_t, const doublereal const y, const doublereal const ydot)
	Evaluate any stopping criteria other than a final time limit. More...

virtual int	calcDeltaSolnVariables (const doublereal t, const doublereal const y, const doublereal const ydot, doublereal const delta_y, const doublereal const solnWeights=0)
	Return a vector of delta y's for calculation of the numerical Jacobian. More...

virtual void	calcSolnScales (const doublereal t, const doublereal const y, const doublereal const y_old, doublereal *const yScales)
	Returns a vector of column scale factors that can be used to column scale Jacobians. More...

virtual void	user_out2 (const int ifunc, const doublereal t, const doublereal delta_t, const doublereal const y, const doublereal const ydot)
	This function may be used to create output at various points in the execution of an application. More...

virtual void	user_out (const int ifunc, const doublereal t, const doublereal y, const doublereal ydot)
	This function may be used to create output at various points in the execution of an application. More...

virtual int	matrixConditioning (doublereal const matrix, const int nrows, doublereal const rhs)
	Multiply the matrix by another matrix that leads to better conditioning. More...

virtual int	evalJacobian (const doublereal t, const doublereal delta_t, doublereal cj, const doublereal const y, const doublereal const ydot, GeneralMatrix &J, doublereal *const resid)
	Calculate an analytical jacobian and the residual at the current time and values. More...

virtual int	evalJacobianDP (const doublereal t, const doublereal delta_t, doublereal cj, const doublereal const y, const doublereal const ydot, doublereal const jacobianColPts, doublereal *const resid)
	Calculate an analytical jacobian and the residual at the current time and values. More...

Public Member Functions inherited from ResidEval
virtual void	constrain (const int k, const int flag)
	Constrain solution component k. More...

int	constraint (const int k) const

virtual void	initSizes ()
	Initialization function. More...

virtual void	setAlgebraic (const int k)
	Specify that solution component k is purely algebraic - that is, the derivative of this component does not appear in the residual function. More...

virtual bool	isAlgebraic (const int k)

virtual int	evalSS (const doublereal t, const doublereal const y, doublereal const r)

virtual int	evalSimpleTD (const doublereal t, const doublereal const y, const doublereal const yold, doublereal deltaT, doublereal *const r)

virtual void	writeSolution (int ievent, const double time, const double deltaT, const int time_step_num, const double y, const double ydot)
	Write out to a file or to standard output the current solution. More...

int	nparams () const
	Return the number of parameters in the calculation. More...

Protected Attributes
doublereal	m_atol
	constant value of atol More...

int	neq_
	Number of equations. More...

Protected Attributes inherited from ResidEval
std::vector< int >	m_alg
	Mapping vector that stores whether a degree of freedom is a DAE or not. More...

std::map< int, int >	m_constrain

Detailed Description

Wrappers for the function evaluators for Nonlinear solvers and Time steppers.

A class for full (non-sparse dense matrices with Fortran-compatible data storage. The class adds support for identifying what types of calls are made to the residual evaluator by adding the ResidEval_Type_Enum class.

Definition at line 51 of file ResidJacEval.h.

Constructor & Destructor Documentation

ResidJacEval ( doublereal atol = 1.0e-13 )

Default constructor.

Parameters

atol	Initial value of the global tolerance (defaults to 1.0E-13)

Definition at line 20 of file ResidJacEval.cpp.

Referenced by ResidJacEval::duplMyselfAsResidJacEval().

ResidJacEval ( const ResidJacEval & right )

Copy Constructor.

Definition at line 26 of file ResidJacEval.cpp.

References ResidJacEval::operator=().

Member Function Documentation

ResidJacEval & operator= ( const ResidJacEval & right )

Assignment operator.

Definition at line 32 of file ResidJacEval.cpp.

References ResidJacEval::m_atol, and ResidJacEval::neq_.

Referenced by ResidJacEval::ResidJacEval().

ResidJacEval * duplMyselfAsResidJacEval ( ) const

virtual

Duplication routine for objects derived from residJacEval.

This virtual routine can be used to duplicate objects which inherit from ResidJacEval even if the application only has a pointer to ResidJacEval to work with.

These routines are basically wrappers around the derived copy constructor.

Definition at line 46 of file ResidJacEval.cpp.

References ResidJacEval::ResidJacEval().

Referenced by NonlinearSolver::operator=().

int nEquations ( ) const

virtual

Return the number of equations in the equation system.

Implements ResidEval.

Definition at line 51 of file ResidJacEval.cpp.

References ResidJacEval::neq_.

Referenced by BEulerInt::doNewtonSolve(), BEulerInt::initializeRJE(), DAE_Solver::nEquations(), and NonlinearSolver::NonlinearSolver().

int evalResidNJ	(	const doublereal	t,
		const doublereal	delta_t,
		const doublereal *const	y,
		const doublereal *const	ydot,
		doublereal *const	resid,
		const ResidEval_Type_Enum	evalType = `Base_ResidEval`,
		const int	id_x = `-1`,
		const doublereal	delta_x = `0.0`
	)

virtual

Evaluate the residual function.

Parameters

t	Time (input)
delta_t	The current value of the time step (input)
y	Solution vector (input, do not modify)
ydot	Rate of change of solution vector. (input, do not modify)
resid	Value of the residual that is computed (output)
evalType	Type of the residual being computed (defaults to Base_ResidEval)
id_x	Index of the variable that is being numerically differenced to find the jacobian (defaults to -1, which indicates that no variable is being differenced or that the residual doesn't take this issue into account)
delta_x	Value of the delta used in the numerical differencing

Returns: Returns a flag to indicate that operation is successful. 1 Means a successful operation -0 or neg value Means an unsuccessful operation

Definition at line 155 of file ResidJacEval.cpp.

Referenced by BEulerInt::beuler_jac(), NonlinearSolver::beuler_jac(), BEulerInt::doNewtonSolve(), NonlinearSolver::doResidualCalc(), ResidJacEval::eval(), and ida_resid().

int eval	(	const doublereal	t,
		const doublereal *const	y,
		const doublereal *const	ydot,
		doublereal *const	r
	)

virtual

Evaluate the residual function.

Called by the integrator.

Parameters

t	time. (input)
y	solution vector. (input)
ydot	rate of change of solution vector. (input)
r	residual vector (output)

Reimplemented from ResidEval.

Definition at line 163 of file ResidJacEval.cpp.

References ResidJacEval::evalResidNJ().

int getInitialConditions	(	const doublereal	t0,
		doublereal *const	y,
		doublereal *const	ydot
	)

virtual

Fill in the initial conditions.

Values for both the solution and the value of ydot may be provided.

Parameters

[in]	t0	Time
[out]	y	Solution vector
[out]	ydot	Rate of change of solution vector.

Returns: Returns a flag to indicate that operation is successful. 1 Means a successful operation -0 or neg value Means an unsuccessful operation

Reimplemented from ResidEval.

Definition at line 66 of file ResidJacEval.cpp.

References ResidJacEval::neq_.

Referenced by BEulerInt::initializeRJE().

doublereal filterNewStep	(	const doublereal	t,
		const doublereal *const	ybase,
		doublereal *const	step
	)

virtual

Filter the solution predictions.

Codes might provide a predicted step change. This routine filters the predicted solution vector eliminating illegal directions.

Parameters

t	Time (input)
ybase	Solution vector (input, output)
step	Proposed step in the solution that will be cropped

Returns: Return the norm of the amount of filtering

Definition at line 129 of file ResidJacEval.cpp.

Referenced by NonlinearSolver::filterNewStep().

doublereal filterSolnPrediction	(	const doublereal	t,
		doublereal *const	y
	)

virtual

Filter the solution predictions.

Codes might provide a predicted solution vector. This routine filters the predicted solution vector.

Parameters

t	Time (input)
y	Solution vector (input, output)

Returns: Return the norm of the amount of filtering

Definition at line 134 of file ResidJacEval.cpp.

Referenced by BEulerInt::calc_y_pred(), and NonlinearSolver::filterNewSolution().

void setAtol ( doublereal atol )

Set a global value of the absolute tolerance.

Parameters

atol	Value of atol

Definition at line 56 of file ResidJacEval.cpp.

References ResidJacEval::m_atol.

int evalTimeTrackingEqns	(	const doublereal	t,
		const doublereal	delta_t,
		const doublereal *const	y,
		const doublereal *const	ydot
	)

virtual

Evaluate the time tracking equations, if any.

Evaluate time integrated quantities that are calculated at the end of every successful time step. This call is made once at the end of every successful time step that advances the time. It's also made once at the start of the time stepping.

Parameters

t	Time (input)
delta_t	The current value of the time step (input)
y	Solution vector (input, do not modify)
ydot	Rate of change of solution vector. (input, do not modify)

Returns: Returns a flag to indicate that operation is successful. 1 Means a successful operation -0 or neg value Means an unsuccessful operation

Definition at line 93 of file ResidJacEval.cpp.

bool evalStoppingCritera	(	const doublereal	t,
		const doublereal	delta_t,
		const doublereal *const	y,
		const doublereal *const	ydot
	)

virtual

Evaluate any stopping criteria other than a final time limit.

If we are to stop the time integration for any reason other than reaching a final time limit, tout, provide a test here. This call is made at the end of every successful time step iteration

Returns: If true, the the time stepping is stopped. If false, then time stepping is stopped if t >= tout Defaults to false.

Parameters

t	Time (input)
delta_t	The current value of the time step (input)
y	Solution vector (input, do not modify)
ydot	Rate of change of solution vector. (input, do not modify)

Definition at line 140 of file ResidJacEval.cpp.

int calcDeltaSolnVariables	(	const doublereal	t,
		const doublereal *const	y,
		const doublereal *const	ydot,
		doublereal *const	delta_y,
		const doublereal *const	solnWeights = `0`
	)

virtual

Return a vector of delta y's for calculation of the numerical Jacobian.

There is a default algorithm provided.

 delta_y[i] = atol[i] + 1.0E-6 ysoln[i]
 delta_y[i] = atol[i] + MAX(1.0E-6 ysoln[i] * 0.01 * solnWeights[i])

Parameters

t	Time (input)
y	Solution vector (input, do not modify)
ydot	Rate of change of solution vector. (input, do not modify)
delta_y	Value of the delta to be used in calculating the numerical jacobian
solnWeights	Value of the solution weights that are used in determining convergence (default = 0)

Returns: Returns a flag to indicate that operation is successful. 1 Means a successful operation -0 or neg value Means an unsuccessful operation

Definition at line 100 of file ResidJacEval.cpp.

References ResidJacEval::m_atol, and ResidJacEval::neq_.

Referenced by BEulerInt::beuler_jac(), and NonlinearSolver::beuler_jac().

void calcSolnScales	(	const doublereal	t,
		const doublereal *const	y,
		const doublereal *const	y_old,
		doublereal *const	yScales
	)

virtual

Returns a vector of column scale factors that can be used to column scale Jacobians.

Default to yScales[] = 1.0

Parameters

t	Time (input)
y	Solution vector (input, do not modify)
y_old	Old Solution vector (input, do not modify)
yScales	Value of the column scales

Definition at line 117 of file ResidJacEval.cpp.

References ResidJacEval::neq_.

Referenced by NonlinearSolver::calcColumnScales(), and BEulerInt::setColumnScales().

void user_out2	(	const int	ifunc,
		const doublereal	t,
		const doublereal	delta_t,
		const doublereal *const	y,
		const doublereal *const	ydot
	)

virtual

This function may be used to create output at various points in the execution of an application.

Parameters

ifunc	identity of the call 0 Initial call 1 Called at the end of every successful time step -1 Called at the end of every unsuccessful time step 2 Called at the end of every call to integrateRJE()
t	Time (input)
delta_t	The current value of the time step (input)
y	Solution vector (input, do not modify)
ydot	Rate of change of solution vector. (input)

Definition at line 80 of file ResidJacEval.cpp.

Referenced by ResidJacEval::user_out().

void user_out	(	const int	ifunc,
		const doublereal	t,
		const doublereal *	y,
		const doublereal *	ydot
	)

virtual

This function may be used to create output at various points in the execution of an application.

This routine calls user_out2().

Parameters

ifunc	identity of the call
t	Time (input)
y	Solution vector (input, do not modify)
ydot	Rate of change of solution vector. (input)

Definition at line 86 of file ResidJacEval.cpp.

References ResidJacEval::user_out2().

int matrixConditioning	(	doublereal *const	matrix,
		const int	nrows,
		doublereal *const	rhs
	)

virtual

Multiply the matrix by another matrix that leads to better conditioning.

Provide a left sided matrix that will multiply the current jacobian, after scaling and lead to a better conditioned system. This routine is called just before the matrix is factored.

Original Problem: J delta_x = - Resid

New problem: M (J delta_x) = - M Resid

Parameters

matrix	Pointer to the current jacobian (if zero, it's already been factored)
nrows	offsets for the matrix
rhs	residual vector. This also needs to be lhs multiplied by M

Returns: Returns a flag to indicate that operation is successful. 1 Means a successful operation -0 or neg value Means an unsuccessful operation

Definition at line 149 of file ResidJacEval.cpp.

Referenced by BEulerInt::doNewtonSolve().

int evalJacobian	(	const doublereal	t,
		const doublereal	delta_t,
		doublereal	cj,
		const doublereal *const	y,
		const doublereal *const	ydot,
		GeneralMatrix &	J,
		doublereal *const	resid
	)

virtual

Calculate an analytical jacobian and the residual at the current time and values.

Only called if the jacFormation method is set to analytical

Parameters

t	Time (input)
delta_t	The current value of the time step (input)
cj	Coefficient of yprime used in the evaluation of the jacobian
y	Solution vector (input, do not modify)
ydot	Rate of change of solution vector. (input, do not modify)
J	Reference to the SquareMatrix object to be calculated (output)
resid	Value of the residual that is computed (output)

Returns: Returns a flag to indicate that operation is successful. 1 Means a successful operation -0 or neg value Means an unsuccessful operation

Definition at line 171 of file ResidJacEval.cpp.

References GeneralMatrix::colPts(), and ResidJacEval::evalJacobianDP().

Referenced by BEulerInt::beuler_jac(), and NonlinearSolver::beuler_jac().

int evalJacobianDP	(	const doublereal	t,
		const doublereal	delta_t,
		doublereal	cj,
		const doublereal *const	y,
		const doublereal *const	ydot,
		doublereal const	jacobianColPts,
		doublereal *const	resid
	)

virtual

Calculate an analytical jacobian and the residual at the current time and values.

Only called if the jacFormation method is set to analytical

Parameters

t	Time (input)
delta_t	The current value of the time step (input)
cj	Coefficient of yprime used in the evaluation of the jacobian
y	Solution vector (input, do not modify)
ydot	Rate of change of solution vector. (input, do not modify)
jacobianColPts	Pointer to the vector of pts to columns of the SquareMatrix object to be calculated (output)
resid	Value of the residual that is computed (output)