AdaptivePreconditioner a preconditioner designed for use with large mechanisms that leverages sparse solvers. More...

#include <AdaptivePreconditioner.h>

Inheritance diagram for AdaptivePreconditioner:

Detailed Description

AdaptivePreconditioner a preconditioner designed for use with large mechanisms that leverages sparse solvers.

It does this by pruning the preconditioner by a threshold value. It also neglects pressure dependence and third body contributions in its formation and has a finite difference approximation for temperature.

Definition at line 23 of file AdaptivePreconditioner.h.

Public Member Functions
void	initialize (size_t networkSize) override
	Called during setup for any processes that need to be completed prior to setup functions used in sundials.

const string	type () const override
	Derived type, corresponding to names registered with SystemJacobianFactory.

void	setup () override
	Perform preconditioner specific post-reactor setup operations such as factorize.

void	factorize () override
	Factorize the system matrix.

void	solve (const size_t stateSize, double rhs_vector, double output) override
	Solve a linear system using the system matrix M

void	stateAdjustment (vector< double > &state) override
	Adjust the state vector based on the preconditioner, e.g., Adaptive preconditioning uses a strictly positive composition when preconditioning which is handled by this function.

int	info () const override
	Get latest status of linear solver.

void	prunePreconditioner ()
	Prune preconditioner elements.

double	threshold ()
	Get the threshold value for setting elements.

double	ilutFillFactor ()
	Get ILUT fill factor.

double	ilutDropTol ()
	Get ILUT drop tolerance.

void	setThreshold (double threshold)
	Set the threshold value to compare elements against.

void	setIlutDropTol (double droptol)
	Set drop tolerance for ILUT.

void	setIlutFillFactor (int fillFactor)
	Set the fill factor for ILUT solver.

Public Member Functions inherited from EigenSparseJacobian
void	initialize (size_t networkSize) override
	Called during setup for any processes that need to be completed prior to setup functions used in sundials.

void	reset () override
	Reset parameters as needed.

void	setValue (size_t row, size_t col, double value) override
	Set a value at the specified row and column of the jacobian triplet vector.

void	updatePreconditioner () override
	Transform Jacobian vector and write into preconditioner, P = (I - gamma * J)

void	updateTransient (double rdt, int *mask) override
	Update the diagonal terms in the Jacobian by using the transient mask $\alpha$ .

Eigen::SparseMatrix< double >	jacobian ()
	Return underlying Jacobian matrix.

Eigen::SparseMatrix< double >	matrix ()
	Return the internal preconditioner matrix.

void	printPreconditioner () override
	Print preconditioner contents.

void	printJacobian ()
	Print jacobian contents.

Public Member Functions inherited from SystemJacobian
virtual const string	type () const =0
	Derived type, corresponding to names registered with SystemJacobianFactory.

virtual void	setValue (size_t row, size_t col, double value)
	Set a value at the specified row and column of the jacobian triplet vector.

virtual void	stateAdjustment (vector< double > &state)
	Adjust the state vector based on the preconditioner, e.g., Adaptive preconditioning uses a strictly positive composition when preconditioning which is handled by this function.

string	preconditionerSide () const
	Get preconditioner application side for CVODES.

virtual void	setPreconditionerSide (const string &preconSide)
	For iterative solvers, set the side where the preconditioner is applied.

virtual void	updatePreconditioner ()
	Transform Jacobian vector and write into preconditioner, P = (I - gamma * J)

virtual void	updateTransient (double rdt, int *mask)
	Update the diagonal terms in the Jacobian by using the transient mask $\alpha$ .

virtual void	solve (const size_t stateSize, double rhs_vector, double output)
	Solve a linear system using the system matrix M

virtual void	setup ()
	Perform preconditioner specific post-reactor setup operations such as factorize.

virtual void	reset ()
	Reset parameters as needed.

virtual void	initialize (size_t networkSize)
	Called during setup for any processes that need to be completed prior to setup functions used in sundials.

virtual void	setBandwidth (size_t bw)
	Used to provide system bandwidth for implementations that use banded matrix storage.

virtual void	printPreconditioner ()
	Print preconditioner contents.

virtual void	setGamma (double gamma)
	Set gamma used in preconditioning.

virtual double	gamma ()
	Get gamma used in preconditioning.

virtual void	setAbsoluteTolerance (double atol)
	Set the absolute tolerance in the solver outside of the network initialization.

virtual int	info () const
	Get latest status of linear solver.

double	elapsedTime () const
	Elapsed CPU time spent computing the Jacobian elements.

void	updateElapsed (double evalTime)
	Increase the elapsed time by the specified time.

int	nEvals () const
	Number of Jacobian evaluations.

void	incrementEvals ()
	Increment the number of times the Jacobian has been evaluated.

int	age () const
	Number of times 'incrementAge' has been called since the last evaluation.

void	incrementAge ()
	Increment the Jacobian age.

void	setAge (int age)
	Set the Jacobian age.

void	clearStats ()
	Clear collected stats about number of Jacobian evaluations, CPU time spent on Jacobian updates, and the number of Newton steps that have been taken since updating the Jacobian.

Protected Attributes
double	m_fill_factor = 0
	ILUT fill factor.

double	m_drop_tol = 0
	ILUT drop tolerance.

Eigen::IncompleteLUT< double >	m_solver
	Solver used in solving the linear system.

double	m_threshold = 0.0
	Minimum value a non-diagonal element must be to be included in the preconditioner.

double	m_prune_precon = true
	Bool set whether to prune the matrix or not.

Protected Attributes inherited from EigenSparseJacobian
vector< Eigen::Triplet< double > >	m_jac_trips
	Vector of triples representing the jacobian used in preconditioning.

Eigen::SparseMatrix< double >	m_identity
	Storage of appropriately sized identity matrix for making the preconditioner.

Eigen::SparseMatrix< double >	m_matrix
	Container that is the sparse preconditioner.

Protected Attributes inherited from SystemJacobian
size_t	m_dim
	Dimension of the system.

double	m_gamma = 1.0
	gamma value used in M = I - gamma*J

bool	m_init = false
	bool saying whether or not the system is initialized

double	m_atol = 0
	Absolute tolerance of the ODE solver.

double	m_elapsed = 0.0
	Elapsed CPU time taken to compute the Jacobian.

int	m_nevals = 0
	Number of Jacobian evaluations.

int	m_age = 100000
	Age of the Jacobian (times incrementAge() has been called)

string	m_precon_side = "none"
	For iterative solvers, side of the system to apply the preconditioner on.

Additional Inherited Members
virtual void	factorize ()
	Factorize the system matrix.

Constructor & Destructor Documentation

◆ AdaptivePreconditioner()

AdaptivePreconditioner ( )

Definition at line 12 of file AdaptivePreconditioner.cpp.

Member Function Documentation

◆ initialize()

void initialize ( size_t networkSize )

overridevirtual

Called during setup for any processes that need to be completed prior to setup functions used in sundials.

Parameters

networkSize the number of variables in the associated reactor network

Reimplemented from SystemJacobian.

Definition at line 24 of file AdaptivePreconditioner.cpp.

◆ type()

const string type ( ) const

inlineoverridevirtual

Derived type, corresponding to names registered with SystemJacobianFactory.

Implements SystemJacobian.

Definition at line 29 of file AdaptivePreconditioner.h.

◆ setup()

void setup ( )

overridevirtual

Perform preconditioner specific post-reactor setup operations such as factorize.

Deprecated:: To be removed after Cantera 3.2. Use updatePreconditioner() and factorize() instead.

Reimplemented from SystemJacobian.

Definition at line 40 of file AdaptivePreconditioner.cpp.

◆ factorize()

void factorize ( )

overridevirtual

Factorize the system matrix.

This method should be called at the end of implementations of updatePreconditioner() and updateTransient().

Reimplemented from SystemJacobian.

Definition at line 48 of file AdaptivePreconditioner.cpp.

◆ solve()

void solve	(	const size_t	stateSize,
		double *	rhs_vector,
		double *	output
	)

overridevirtual

Solve a linear system using the system matrix M

The matrix M can either be the preconditioner or the transient Jacobian matrix, depending on whether updateTransient() or updatePreconditioner() was called last.

Parameters

[in]	stateSize	length of the rhs and output vectors
[in]	rhs_vector	right hand side vector used in linear system
[out]	output	output vector for solution

Reimplemented from SystemJacobian.

Definition at line 76 of file AdaptivePreconditioner.cpp.

◆ stateAdjustment()

void stateAdjustment ( vector< double > & state )

overridevirtual

Adjust the state vector based on the preconditioner, e.g., Adaptive preconditioning uses a strictly positive composition when preconditioning which is handled by this function.

Parameters

state a vector containing the state to be updated

Reimplemented from SystemJacobian.

Definition at line 17 of file AdaptivePreconditioner.cpp.

◆ info()

int info ( ) const

inlineoverridevirtual

Get latest status of linear solver.

Zero indicates success. Meaning of non-zero values is solver dependent.

Reimplemented from SystemJacobian.

Definition at line 35 of file AdaptivePreconditioner.h.

◆ prunePreconditioner()

void prunePreconditioner ( )

Prune preconditioner elements.

Definition at line 64 of file AdaptivePreconditioner.cpp.

◆ threshold()

double threshold ( )

inline

Get the threshold value for setting elements.

Definition at line 43 of file AdaptivePreconditioner.h.

◆ ilutFillFactor()

double ilutFillFactor ( )

inline

Get ILUT fill factor.

Definition at line 46 of file AdaptivePreconditioner.h.

◆ ilutDropTol()

double ilutDropTol ( )

inline

Get ILUT drop tolerance.

Definition at line 49 of file AdaptivePreconditioner.h.

◆ setThreshold()

void setThreshold ( double threshold )

inline

Set the threshold value to compare elements against.

Parameters

threshold double value used in setting by threshold

Definition at line 53 of file AdaptivePreconditioner.h.

◆ setIlutDropTol()

void setIlutDropTol ( double droptol )

inline

Set drop tolerance for ILUT.

Parameters

droptol double value used in setting solver drop tolerance

Definition at line 60 of file AdaptivePreconditioner.h.

◆ setIlutFillFactor()

void setIlutFillFactor ( int fillFactor )

inline

Set the fill factor for ILUT solver.

Parameters

fillFactor fill in factor for ILUT solver

Definition at line 67 of file AdaptivePreconditioner.h.

Member Data Documentation

◆ m_fill_factor

double m_fill_factor = 0

protected

ILUT fill factor.

Definition at line 74 of file AdaptivePreconditioner.h.

◆ m_drop_tol

double m_drop_tol = 0

protected

ILUT drop tolerance.

Definition at line 77 of file AdaptivePreconditioner.h.

◆ m_solver

Eigen::IncompleteLUT<double> m_solver

protected

Solver used in solving the linear system.

Definition at line 80 of file AdaptivePreconditioner.h.

◆ m_threshold

double m_threshold = 0.0

protected

Minimum value a non-diagonal element must be to be included in the preconditioner.

Definition at line 84 of file AdaptivePreconditioner.h.

◆ m_prune_precon

double m_prune_precon = true

protected

Bool set whether to prune the matrix or not.

Definition at line 87 of file AdaptivePreconditioner.h.

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

Detailed Description

Public Member Functions

Protected Attributes

Additional Inherited Members

Constructor & Destructor Documentation

◆ AdaptivePreconditioner()

Member Function Documentation

◆ initialize()

◆ type()

◆ setup()

◆ factorize()

◆ solve()

◆ stateAdjustment()

◆ info()

◆ prunePreconditioner()

◆ threshold()

◆ ilutFillFactor()

◆ ilutDropTol()

◆ setThreshold()

◆ setIlutDropTol()

◆ setIlutFillFactor()

Member Data Documentation

◆ m_fill_factor

◆ m_drop_tol

◆ m_solver

◆ m_threshold

◆ m_prune_precon