AbstractCvodeSystem Class Reference

#include <AbstractCvodeSystem.hpp>

Inherits AbstractParameterisedSystem< N_Vector >.

Collaboration diagram for AbstractCvodeSystem:

Public Member Functions
	AbstractCvodeSystem (unsigned numberOfStateVariables)
virtual	~AbstractCvodeSystem ()
virtual void	EvaluateYDerivatives (realtype time, const N_Vector y, N_Vector ydot)=0
virtual void	EvaluateAnalyticJacobian (long int N, realtype time, N_Vector y, N_Vector ydot, CHASTE_CVODE_DENSE_MATRIX jacobian, N_Vector tmp1, N_Vector tmp2, N_Vector tmp3)
void	SetForceReset (bool autoReset)
void	SetMinimalReset (bool minimalReset)
void	ResetSolver ()
OdeSolution	Solve (realtype tStart, realtype tEnd, realtype maxDt, realtype tSamp)
void	Solve (realtype tStart, realtype tEnd, realtype maxDt)
void	SetMaxSteps (long int numSteps)
long int	GetMaxSteps ()
void	SetTolerances (double relTol=1e-5, double absTol=1e-7)
double	GetRelativeTolerance ()
double	GetAbsoluteTolerance ()
double	GetLastStepSize ()
bool	GetUseAnalyticJacobian () const
bool	HasAnalyticJacobian () const
void	ForceUseOfNumericalJacobian (bool useNumericalJacobian=true)
Protected Member Functions
void	Init ()
Protected Attributes
bool	mHasAnalyticJacobian
bool	mUseAnalyticJacobian
double	mRelTol
double	mAbsTol
void *	mpCvodeMem
long int	mMaxSteps
double	mLastInternalStepSize
Private Member Functions
template<class Archive >
void	save (Archive &archive, const unsigned int version) const
template<class Archive >
void	load (Archive &archive, const unsigned int version)
void	SetupCvode (N_Vector initialConditions, realtype tStart, realtype maxDt)
void	RecordStoppingPoint (double stopTime)
void	FreeCvodeMemory ()
void	CvodeError (int flag, const char *msg)
Private Attributes
N_Vector	mLastSolutionState
double	mLastSolutionTime
bool	mForceReset
bool	mForceMinimalReset
Friends
class	TestAbstractCvodeSystem
class	boost::serialization::access

Detailed Description

Abstract OdeSystem class for Cvode systems (N_Vector instead of std::vector)

Sets up variables and functions for a general CVODE system.

ODE systems are specified primarily by the EvaluateYDerivatives() method, which calculates the right-hand side of the system.

You can also specify an EvaluateAnalyticJacobian() method which will allow CVODE to evaluate the Jacobian matrix precisely instead of using multiple calls to EvaluateYDerivatives() to make an approximation to it. This generally provides extra accuracy and better convergence, and so gives a speed up for a given tolerance. Cardiac action potential model Jacobians are calculated automatically by PyCML (see python/pycml) via Maple for symbolic differentiation.

Instances can store their state internally in the mStateVariables vector in our base class AbstractParameterisedSystem (see also GetNumberOfStateVariables(), SetStateVariables() and rGetStateVariables()), although this is not essential - the vector may be empty, in which case AbstractIvpOdeSolver::SolveAndUpdateStateVariable may not be used to solve the system.

CVODE systems may also have a vector of parameters, which can be accessed through the GetParameter() and SetParameter() methods of our base class.

Information about what the parameters and state variables represent is provided by a subclass of AbstractOdeSystemInformation. Various wrapper methods (e.g. rGetStateVariableNames()) are provided in our base class to access this information.

Also, subclasses may define a condition at which ODE solvers should stop prematurely. For this class CVODE solvers are being used, so CalculateRootFunction() should be used to detect the stopping time.

Note that the default tolerances for the solver are set by SetTolerances(), these can make quite a difference to the time it takes to solve the ODE system.

Repeated calls to Solve will no longer set up and delete CVODE memory, unless the following method is called:

SetForceReset(true) - reset each time Solve() is called

default - reset if state variables change, or we ask to solve from a different time than the last solve call finished.

SetMinimalReset(true) - ignore changes in state vars and just reset if the time is inconsistent.

Definition at line 119 of file AbstractCvodeSystem.hpp.

Constructor & Destructor Documentation

AbstractCvodeSystem::AbstractCvodeSystem ( unsigned numberOfStateVariables )

Constructor.

Parameters:

numberOfStateVariables

the number of state variables in the ODE system

Definition at line 121 of file AbstractCvodeSystem.cpp.

References SetTolerances().

AbstractCvodeSystem::~AbstractCvodeSystem ( ) [virtual]

Virtual destructor since we have virtual methods.

Definition at line 154 of file AbstractCvodeSystem.cpp.

References DeleteVector(), FreeCvodeMemory(), mLastSolutionState, AbstractParameterisedSystem< N_Vector >::mParameters, and AbstractParameterisedSystem< N_Vector >::mStateVariables.

Member Function Documentation

void AbstractCvodeSystem::CvodeError	(	int	flag,
		const char *	msg
	)			`[private]`

Report an error from CVODE.

Parameters:

	flag	CVODE error code
	msg	Our description of the error

Definition at line 445 of file AbstractCvodeSystem.cpp.

References EXCEPTION, FreeCvodeMemory(), and mpCvodeMem.

Referenced by Solve().

virtual void AbstractCvodeSystem::EvaluateAnalyticJacobian	(	long int	N,
		realtype	time,
		N_Vector	y,
		N_Vector	ydot,
		CHASTE_CVODE_DENSE_MATRIX	jacobian,
		N_Vector	tmp1,
		N_Vector	tmp2,
		N_Vector	tmp3
	)			`[inline, virtual]`

This method is called by AbstractCvodeSystemJacAdaptor method in the .cpp file.

It provides an interface between the methods different versions of CVODE are expecting and the Jacobians provided by Chaste CVODE systems.

Parameters:

	N	the size of the ODE system
	time	the current time (used by ODE systems like y' = f(t,y) only I guess)
	y	the current state variables y for y' = f(t,y)
	ydot	the current set of derivatives y' = f(t,y)
	jacobian	a pointer to a jacobian, populated by this method.
	tmp1	working memory of the correct size provided by CVODE for temporary calculations
	tmp2	working memory of the correct size provided by CVODE for temporary calculations
	tmp3	working memory of the correct size provided by CVODE for temporary calculations

Definition at line 326 of file AbstractCvodeSystem.hpp.

References EXCEPTION.

virtual void AbstractCvodeSystem::EvaluateYDerivatives	(	realtype	time,
		const N_Vector	y,
		N_Vector	ydot
	)			`[pure virtual]`

Method to evaluate the derivatives of the system.

Parameters:

	time	the current time
	y	the current values of the state variables
	ydot	storage for the derivatives of the system; will be filled in on return

void AbstractCvodeSystem::ForceUseOfNumericalJacobian ( bool useNumericalJacobian = true )

Force the use of a numerical Jacobian, even if an analytic form is provided. This is needed for a handful of troublesome models.

Parameters:

useNumericalJacobian

Whether to use a numerical instead of the analytic Jacobian.

Definition at line 486 of file AbstractCvodeSystem.cpp.

References EXCEPTION, FreeCvodeMemory(), mHasAnalyticJacobian, and mUseAnalyticJacobian.

void AbstractCvodeSystem::FreeCvodeMemory ( ) [private]

Free CVODE memory when finished with.

Definition at line 435 of file AbstractCvodeSystem.cpp.

References mpCvodeMem.

Referenced by CvodeError(), ForceUseOfNumericalJacobian(), and ~AbstractCvodeSystem().

double AbstractCvodeSystem::GetAbsoluteTolerance ( )

Returns:: the absolute tolerance.

Definition at line 300 of file AbstractCvodeSystem.cpp.

References mAbsTol.

double AbstractCvodeSystem::GetLastStepSize ( )

Returns:: the last step size used internally by CVODE in the last Solve call.

Definition at line 305 of file AbstractCvodeSystem.cpp.

References mLastInternalStepSize.

long int AbstractCvodeSystem::GetMaxSteps ( )

Returns:: the maximum number of steps to be taken by the solver in its attempt to reach the next output time.

Definition at line 283 of file AbstractCvodeSystem.cpp.

References mMaxSteps.

double AbstractCvodeSystem::GetRelativeTolerance ( )

Returns:: the relative tolerance.

Definition at line 295 of file AbstractCvodeSystem.cpp.

References mRelTol.

bool AbstractCvodeSystem::GetUseAnalyticJacobian ( ) const

Returns:: whether an analytic Jacobian is used (mUseAnalyticJacobian).

Definition at line 480 of file AbstractCvodeSystem.cpp.

References mUseAnalyticJacobian.

bool AbstractCvodeSystem::HasAnalyticJacobian ( ) const

Returns:: whether the ODE system has an analytic Jacobian (mHasAnalyticJacobian).

Definition at line 475 of file AbstractCvodeSystem.cpp.

References mHasAnalyticJacobian.

void AbstractCvodeSystem::Init ( ) [protected]

Must be called by concrete subclass constructors to initialise the state variables, after setting mpSystemInfo.

Definition at line 142 of file AbstractCvodeSystem.cpp.

References DeleteVector(), AbstractParameterisedSystem< N_Vector >::GetInitialConditions(), AbstractParameterisedSystem< N_Vector >::mParameters, AbstractParameterisedSystem< N_Vector >::mStateVariables, and AbstractUntemplatedParameterisedSystem::rGetParameterNames().

template<class Archive >

void AbstractCvodeSystem::load	(	Archive &	archive,
		const unsigned int	version
	)			`[inline, private]`

Archive the member variables.

Parameters:

	archive	the archive
	version	the current version of this class

Definition at line 175 of file AbstractCvodeSystem.hpp.

References AbstractParameterisedSystem< N_Vector >::CheckParametersOnLoad(), CopyFromStdVector(), mAbsTol, mForceMinimalReset, mForceReset, mHasAnalyticJacobian, mLastInternalStepSize, mLastSolutionTime, mMaxSteps, AbstractUntemplatedParameterisedSystem::mNumberOfStateVariables, mRelTol, AbstractParameterisedSystem< N_Vector >::mStateVariables, and mUseAnalyticJacobian.

void AbstractCvodeSystem::RecordStoppingPoint ( double stopTime ) [private]

Record where the last solve got to so we know whether to re-initialise.

Parameters:

stopTime

the finishing time

Definition at line 414 of file AbstractCvodeSystem.cpp.

References CreateVectorIfEmpty(), AbstractUntemplatedParameterisedSystem::GetNumberOfStateVariables(), GetVectorComponent(), mForceReset, mLastSolutionState, mLastSolutionTime, AbstractParameterisedSystem< N_Vector >::mStateVariables, and SetVectorComponent().

Referenced by Solve().

void AbstractCvodeSystem::ResetSolver ( )

Successive calls to Solve will attempt to intelligently determine whether to re-initialise the internal CVODE solver, or whether we are simply extending the previous solution forward in time. This mechanism compares the state vector to its previous value, and the start time to the end of the last solve, which captures most cases where re-initialisation is required. However, changes to the RHS function can also require this, and cannot be automatically detected. In such cases users must call this function to force re-initialisation.

Definition at line 329 of file AbstractCvodeSystem.cpp.

References DeleteVector(), and mLastSolutionState.

Referenced by SetForceReset(), SetTolerances(), and AbstractCvodeCell::SetVoltageDerivativeToZero().

template<class Archive >

void AbstractCvodeSystem::save	(	Archive &	archive,
		const unsigned int	version
	)			const `[inline, private]`

Archive the member variables.

Parameters:

	archive	the archive
	version	the current version of this class

Definition at line 132 of file AbstractCvodeSystem.hpp.

References mAbsTol, MakeStdVec(), mForceMinimalReset, mForceReset, mHasAnalyticJacobian, mLastInternalStepSize, mLastSolutionTime, mMaxSteps, AbstractUntemplatedParameterisedSystem::mNumberOfStateVariables, AbstractParameterisedSystem< N_Vector >::mParameters, mRelTol, AbstractParameterisedSystem< N_Vector >::mStateVariables, mUseAnalyticJacobian, and AbstractUntemplatedParameterisedSystem::rGetParameterNames().

void AbstractCvodeSystem::SetForceReset ( bool autoReset )

Set whether to automatically re-initialise CVODE on every call to Solve, or whether to attempt to guess when re-initialisation is needed. For example it will re-initialise if the time changes, or any state variables change. You can safely change parameters between solve calls with or without resets.

See also ResetSolver and SetMinimalReset

Parameters:

autoReset

whether to reset on every Solve

Definition at line 310 of file AbstractCvodeSystem.cpp.

References mForceReset, and ResetSolver().

Referenced by SetMinimalReset().

void AbstractCvodeSystem::SetMaxSteps ( long int numSteps )

Change the maximum number of steps to be taken by the solver in its attempt to reach the next output time. Default is 500 (set by CVODE).

Parameters:

numSteps

new maximum

Definition at line 278 of file AbstractCvodeSystem.cpp.

References mMaxSteps.

void AbstractCvodeSystem::SetMinimalReset ( bool minimalReset )

Set whether to reduce the checking done when guessing when re-initialisation is needed, so it ignores changes in the state variables. If called with true argument, will call SetForceReset(false). You can safely change parameters between solve calls with or without resets.

Parameters:

minimalReset

whether to avoid checking for changes in state variables

Definition at line 319 of file AbstractCvodeSystem.cpp.

References mForceMinimalReset, and SetForceReset().

void AbstractCvodeSystem::SetTolerances	(	double	relTol = `1e-5`,
		double	absTol = `1e-7`
	)

Set relative and absolute tolerances; both scalars. If no parameters are given, tolerances will be reset to default values.

Parameters:

	relTol	the relative tolerance for the solver (defaults to 1e-5)
	absTol	the absolute tolerance for the solver (defaults to 1e-7)

Definition at line 288 of file AbstractCvodeSystem.cpp.

References mAbsTol, mRelTol, and ResetSolver().

Referenced by AbstractCvodeSystem().

void AbstractCvodeSystem::SetupCvode	(	N_Vector	initialConditions,
		realtype	tStart,
		realtype	maxDt
	)			`[private]`

Set up the CVODE data structures needed to solve the given system from a given point.

Parameters:

	initialConditions	initial conditions
	tStart	start time of simulation
	maxDt	maximum time step to take

Definition at line 335 of file AbstractCvodeSystem.cpp.

References EXCEPTION, AbstractUntemplatedParameterisedSystem::GetNumberOfStateVariables(), GetVectorComponent(), mAbsTol, mForceMinimalReset, mForceReset, mLastSolutionState, mLastSolutionTime, mMaxSteps, mpCvodeMem, mRelTol, AbstractParameterisedSystem< N_Vector >::mStateVariables, mUseAnalyticJacobian, and CompareDoubles::WithinAnyTolerance().

Referenced by Solve().

void AbstractCvodeSystem::Solve	(	realtype	tStart,
		realtype	tEnd,
		realtype	maxDt
	)

Simulate the cell, updating its internal state variables.

Uses the current values of the state variables at initial conditions. If the state variables have not been set (either by a prior solve, or a call to SetStateVariables) the initial conditions (given by GetInitialConditions) will be used.

Note:: See also the ResetSolver method.

Parameters:

	tStart	start time of simulation
	tEnd	end time of simulation
	maxDt	maximum time step to be taken by the adaptive solver (set this appropriately to avoid missing a stimulus)

Definition at line 225 of file AbstractCvodeSystem.cpp.

References CvodeError(), mLastInternalStepSize, mpCvodeMem, AbstractParameterisedSystem< N_Vector >::mStateVariables, RecordStoppingPoint(), SetupCvode(), UNUSED_OPT, and AbstractParameterisedSystem< N_Vector >::VerifyStateVariables().

OdeSolution AbstractCvodeSystem::Solve	(	realtype	tStart,
		realtype	tEnd,
		realtype	maxDt,
		realtype	tSamp
	)

Simulate the cell, returning a sampling of the state variables.

Uses the current values of the state variables at initial conditions. If the state variables have not been set (either by a prior solve, or a call to SetStateVariables) the initial conditions (given by GetInitialConditions) will be used.

The final values of the state variables will also be stored in this object.

Note:: See also the ResetSolver method.

Parameters:

	tStart	start time of simulation
	tEnd	end time of simulation
	maxDt	maximum time step to be taken by the adaptive solver (set this appropriately to avoid missing a stimulus)
	tSamp	sampling interval at which to store results