CEPS: AbstractOdeSolver Class Reference

Detailed Description

Solves ODE systems .

For several numerical schemes, the evolution function is split between a linear and a non linear part.

$y'=f(t,y) = f_{NL}(t,y) + f_{L}(t,)y$

Class usage:

1./ If used with "free" non-member functions (than can exist on their own), or static member functions:

mySolver = ceps::getNew<DerivatedFromAbstractOdeSolver>(...);
mySolver->setEvolutionFunction(myNonLinearPartFunction,myLinearPartFunction);
# or
mySolver->setEvolutionFunction(myNonLinearPartFunction);
mySolver->initialize(sizeOfInitialCondition,initialCondition);
for(CepsInt i=0; i<nSteps; i++,t+=dt)
  mySolver->takeOneStep(t,dt);
CepsReal* sol = mySolver->getSolution();

2./ If used with AbstractIonicModel for cardiac problems, the solver is either for Gate or Misc variables, so the interaction with ionic model and the numerical scheme itself may change. This is why you need to select the right initialization method.

See also: ionicModelSolver

Definition at line 68 of file AbstractOdeSolver.hpp.

#include <AbstractOdeSolver.hpp>

Inheritance diagram for AbstractOdeSolver:

[legend]

Public Types
using	RateFunction = ceps::Function< void(CepsReal, CepsReal , CepsReal , CepsUInt)>
	Typedef for function f(t,y) More...

Public Member Functions
	AbstractOdeSolver ()
	Constructor. More...

virtual	~AbstractOdeSolver ()
	Destructor. More...

virtual void	initialize (CepsReal y0, CepsUInt n, RateFunction nonlinearFunc, RateFunction *linearFunc=nullptr, CepsBool copy=false)
	Allocate data arrays and sets initial condition with given data. More...

CepsUInt	getNbMultiSteps () const
	Tells how many time steps the method covers. More...

CepsUInt	getNbSubSteps () const
	Tells if several intermediate steps are needed to get to $y^{n+1}$ . More...

CepsReal *	getSolution () const
	Returns a pointer to the solution. Size is the same as y0 argument in initialize. More...

virtual void	takeOneStep (CepsReal t, CepsReal dt)=0
	Make a full ode step : must be defined in every child class. More...

virtual void	takeOneSubStep (CepsReal t, CepsReal dt)
	Progress by a substep, for RK methods. Does nothing by default. More...

Protected Member Functions
virtual void	shiftDataArraysForNextStep ()
	Rotates the pointers on data such that yN becomes yN-1, etc... More...

void	evaluateDerivative (CepsReal t, CepsReal y, CepsReal dty, CepsUInt n, CepsBool nLinPart=true)
	Selects and computes depending on how it was passed to the class. More...

CepsReal	phi1 (CepsReal x) const
	Function $\varphi_1$ used in exponential schemes. More...

Protected Attributes
CepsUInt	m_ySize
	Function $\varphi_n$ used in exponential schemes. Return all phi_n until n, starting at phi0. More...

CepsReal *	m_yNp1
	Solution at time $t^{n+1}$ . More...

CepsVector< CepsReal * >	m_yNm
	Solution at time and before. More...

CepsUInt	m_nbMultiSteps
	Number of steps for multi step solvers. More...

CepsUInt	m_nbSubSteps
	Number of sub steps (eg RK solvers) More...

CepsUInt	m_currentStep
	Ongoing step. More...

RateFunction *	m_fL
	Linear function to call in evaluateDerivative. More...

RateFunction *	m_fNL
	Non-linear function to call in evaluateDerivative. More...

CepsBool	m_ownedSolData
	Flag for destructor. More...

CepsBool	m_initialized
	Check at deallocation. More...

Member Typedef Documentation

◆ RateFunction

using AbstractOdeSolver::RateFunction = ceps::Function<void(CepsReal,CepsReal*,CepsReal*,CepsUInt)>

Typedef for function f(t,y)

Definition at line 74 of file AbstractOdeSolver.hpp.

Constructor & Destructor Documentation

◆ AbstractOdeSolver()

AbstractOdeSolver::AbstractOdeSolver ( )

Constructor.

Definition at line 32 of file AbstractOdeSolver.cpp.

◆ ~AbstractOdeSolver()

AbstractOdeSolver::~AbstractOdeSolver ( )

virtual

Destructor.

Definition at line 45 of file AbstractOdeSolver.cpp.

Member Function Documentation

◆ evaluateDerivative()

void AbstractOdeSolver::evaluateDerivative	(	CepsReal	t,
		CepsReal *	y,
		CepsReal *	dty,
		CepsUInt	n,
		CepsBool	nLinPart = `true`
	)

protected

Selects and computes depending on how it was passed to the class.

Either through classes like AbstractIonicModel, or manually via setEvolutionFunction(...)

Parameters

t	time t^n
y	values at t^n
dty	result array
n	size of arrays
nLinPart	true to compute the non-linear part of the evolution function, false for the linear part

Definition at line 136 of file AbstractOdeSolver.cpp.

◆ getNbMultiSteps()

CepsUInt AbstractOdeSolver::getNbMultiSteps ( ) const

Tells how many time steps the method covers.

Definition at line 98 of file AbstractOdeSolver.cpp.

◆ getNbSubSteps()

CepsUInt AbstractOdeSolver::getNbSubSteps ( ) const

Tells if several intermediate steps are needed to get to $y^{n+1}$ .

Definition at line 104 of file AbstractOdeSolver.cpp.

◆ getSolution()

CepsReal * AbstractOdeSolver::getSolution ( ) const

Returns a pointer to the solution. Size is the same as y0 argument in initialize.

Definition at line 110 of file AbstractOdeSolver.cpp.

◆ initialize()

void AbstractOdeSolver::initialize	(	CepsReal *	y0,
		CepsUInt	n,
		AbstractOdeSolver::RateFunction *	nonLinearFunc,
		AbstractOdeSolver::RateFunction *	linearFunc = `nullptr`,
		CepsBool	copy = `false`
	)

virtual

Allocate data arrays and sets initial condition with given data.

This method assumes a single step ODE solver, and should be overloaded for multi-step solvers.

Parameters

[in]	y0	initial data
[in]	n	allocation size for each array
[in]	nonLinearFunc	Pointer to ceps::Function f'(t,y)

See also: MemberFunction

Parameters

[in] linearFunc Pointer to ceps::Function f'(t,y)

See also: MemberFunction

Parameters

[in] copy if false, the array y0 will be used to store yn (thus overwritten) be careful with the delete statement! There will be arrays for times and $t^{n+1}$ , and, if needed, other arrays for multi-step methods

Reimplemented in SBDFOdeSolver, RushLarsenOdeSolver, RungeKuttaOdeSolver, FBEOdeSolver, and ExponentialAdamsBashforthOdeSolver.

Definition at line 57 of file AbstractOdeSolver.cpp.

◆ phi1()

CepsReal AbstractOdeSolver::phi1 ( CepsReal x ) const

protected

Function $\varphi_1$ used in exponential schemes.

$\varphi_1(x) = \displaystyle\frac{\mathrm{e}^x-1}{x}$

Definition at line 152 of file AbstractOdeSolver.cpp.

◆ shiftDataArraysForNextStep()

void AbstractOdeSolver::shiftDataArraysForNextStep ( )

protectedvirtual

Rotates the pointers on data such that yN becomes yN-1, etc...

Reimplemented in ExponentialAdamsBashforthOdeSolver, SBDFOdeSolver, and RushLarsenOdeSolver.

Definition at line 124 of file AbstractOdeSolver.cpp.

◆ takeOneStep()

virtual void AbstractOdeSolver::takeOneStep	(	CepsReal	t,
		CepsReal	dt
	)

pure virtual

Make a full ode step : must be defined in every child class.

Parameters

[in]	t	time
[in]	dt	ode time step

Implemented in SBDFOdeSolver, RushLarsenOdeSolver, RungeKuttaOdeSolver, FBEOdeSolver, and ExponentialAdamsBashforthOdeSolver.

◆ takeOneSubStep()

void AbstractOdeSolver::takeOneSubStep	(	CepsReal	t,
		CepsReal	dt
	)

virtual

Progress by a substep, for RK methods. Does nothing by default.

Reimplemented in RungeKuttaOdeSolver.

Definition at line 116 of file AbstractOdeSolver.cpp.

Field Documentation

◆ m_currentStep

CepsUInt AbstractOdeSolver::m_currentStep

protected

Ongoing step.

Definition at line 176 of file AbstractOdeSolver.hpp.

◆ m_fL

RateFunction* AbstractOdeSolver::m_fL

protected

Linear function to call in evaluateDerivative.

Definition at line 178 of file AbstractOdeSolver.hpp.

◆ m_fNL

RateFunction* AbstractOdeSolver::m_fNL

protected

Non-linear function to call in evaluateDerivative.

Definition at line 179 of file AbstractOdeSolver.hpp.

◆ m_initialized

CepsBool AbstractOdeSolver::m_initialized

protected

Check at deallocation.

Definition at line 182 of file AbstractOdeSolver.hpp.

◆ m_nbMultiSteps

CepsUInt AbstractOdeSolver::m_nbMultiSteps

protected

Number of steps for multi step solvers.

Definition at line 174 of file AbstractOdeSolver.hpp.

◆ m_nbSubSteps

CepsUInt AbstractOdeSolver::m_nbSubSteps

protected

Number of sub steps (eg RK solvers)

Definition at line 175 of file AbstractOdeSolver.hpp.

◆ m_ownedSolData

CepsBool AbstractOdeSolver::m_ownedSolData

protected

Flag for destructor.

Definition at line 181 of file AbstractOdeSolver.hpp.

◆ m_yNm

CepsVector<CepsReal*> AbstractOdeSolver::m_yNm

protected

Solution at time and before.

Definition at line 172 of file AbstractOdeSolver.hpp.

◆ m_yNp1

CepsReal* AbstractOdeSolver::m_yNp1

protected

Solution at time $t^{n+1}$ .

Definition at line 171 of file AbstractOdeSolver.hpp.

◆ m_ySize

CepsUInt AbstractOdeSolver::m_ySize

protected

Function $\varphi_n$ used in exponential schemes. Return all phi_n until n, starting at phi0.

$\varphi_n(x) = \displaystyle\frac{\varphi_{n-1}(x)-1}{x}$ Size of solution vector

Definition at line 170 of file AbstractOdeSolver.hpp.

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

/builds/xxjFYZX3/0/carmen/ceps/src/ode/solver/AbstractOdeSolver.hpp
/builds/xxjFYZX3/0/carmen/ceps/src/ode/solver/AbstractOdeSolver.cpp

Detailed Description

Public Types

Public Member Functions

Protected Member Functions

Protected Attributes

Member Typedef Documentation

◆ RateFunction

Constructor & Destructor Documentation

◆ AbstractOdeSolver()

◆ ~AbstractOdeSolver()

Member Function Documentation

◆ evaluateDerivative()

◆ getNbMultiSteps()

◆ getNbSubSteps()

◆ getSolution()

◆ initialize()

◆ phi1()

◆ shiftDataArraysForNextStep()

◆ takeOneStep()

◆ takeOneSubStep()

Field Documentation

◆ m_currentStep

◆ m_fL

◆ m_fNL

◆ m_initialized

◆ m_nbMultiSteps

◆ m_nbSubSteps

◆ m_ownedSolData

◆ m_yNm

◆ m_yNp1

◆ m_ySize