Newton method to solve nonlinear equation of form: F(U) = U - b - R(U) = 0. U is the solution vector, b is a constant, and R(U) is some nonlinear function of U, which is computed in the ComputeRHS() Ops function. More...

#include <NewtonSolver.H>

Inheritance diagram for NewtonSolver< Vec, Ops >:

Public Member Functions
	NewtonSolver ()=default

	~NewtonSolver () override=default

	NewtonSolver (const NewtonSolver &)=delete

NewtonSolver &	operator= (const NewtonSolver &)=delete

	NewtonSolver (NewtonSolver &&) noexcept=delete

NewtonSolver &	operator= (NewtonSolver &&) noexcept=delete

void	Define (const Vec &a_U, Ops *a_ops) override
	Read user-provided parameters that control the nonlinear solver. Allocate intermediate data containers needed by the solver. For Newton, setup the linear solver for computing the Newton step. More...

void	Solve (Vec &a_U, const Vec &a_b, amrex::Real a_time, amrex::Real a_dt) const override
	Solve the specified nonlinear equation for U. Picard: U = b + R(U). Newton: F(U) = U - b - R(U) = 0. More...

void	GetSolverParams (amrex::Real &a_rtol, amrex::Real &a_atol, int &a_maxits) override
	Return the convergence parameters used by the nonlinear solver. More...

void	CurTime (amrex::Real a_time) const

void	CurTimeStep (amrex::Real a_dt) const

void	PrintParams () const override
	Print parameters used by the nonlinear solver. More...

Public Member Functions inherited from NonlinearSolver< Vec, Ops >
	NonlinearSolver ()=default

virtual	~NonlinearSolver ()=default

	NonlinearSolver (const NonlinearSolver &)=delete

NonlinearSolver &	operator= (const NonlinearSolver &)=delete

	NonlinearSolver (NonlinearSolver &&) noexcept=delete

NonlinearSolver &	operator= (NonlinearSolver &&) noexcept=delete

bool	IsDefined () const
	Check if the nonlinear solver has been defined. More...

void	Verbose (bool a_verbose)
	Allow caller to dynamically change the verbosity flag. For example, one may want to only print solver information every 100 steps. More...

Private Member Functions
void	ParseParameters ()

void	EvalResidual (Vec &a_F, const Vec &a_U, const Vec &a_b, amrex::Real a_time, amrex::Real a_dt, int a_iter) const
	Compute the nonlinear residual: F(U) = U - b - R(U). More...

Private Attributes
Vec	m_dU
	Intermediate Vec containers used by the solver. More...

Vec	m_F

Vec	m_R

Ops *	m_ops = nullptr
	Pointer to Ops class. More...

bool	m_require_convergence = true
	Flag to determine whether convergence is required. More...

amrex::Real	m_rtol = 1.0e-6
	Relative tolerance for the Newton solver. More...

amrex::Real	m_atol = 0.
	Absolute tolerance for the Newton solver. More...

int	m_maxits = 100
	Maximum iterations for the Newton solver. More...

amrex::Real	m_gmres_rtol = 1.0e-4
	Relative tolerance for GMRES. More...

amrex::Real	m_gmres_atol = 0.
	Absolute tolerance for GMRES. More...

int	m_gmres_maxits = 1000
	Maximum iterations for GMRES. More...

int	m_gmres_verbose_int = 2
	Verbosity flag for GMRES. More...

int	m_gmres_restart_length = 30
	Restart iteration for GMRES. More...

amrex::Real	m_cur_time

amrex::Real	m_dt

bool	m_update_pc = false

bool	m_update_pc_init = false

std::unique_ptr< JacobianFunctionMF< Vec, Ops > >	m_linear_function
	The linear function used by GMRES to compute A*v. In the contect of JFNK, A = dF/dU (i.e., system Jacobian) More...

std::unique_ptr< amrex::GMRES< Vec, JacobianFunctionMF< Vec, Ops > > >	m_linear_solver
	The linear solver (GMRES) object. More...

Additional Inherited Members
Protected Attributes inherited from NonlinearSolver< Vec, Ops >
bool	m_is_defined = false

bool	m_verbose = true

Detailed Description

template<class Vec, class Ops>
class NewtonSolver< Vec, Ops >

Newton method to solve nonlinear equation of form: F(U) = U - b - R(U) = 0. U is the solution vector, b is a constant, and R(U) is some nonlinear function of U, which is computed in the ComputeRHS() Ops function.

Constructor & Destructor Documentation

◆ NewtonSolver() [1/3]

template<class Vec , class Ops >

NewtonSolver< Vec, Ops >::NewtonSolver ( )

default

◆ ~NewtonSolver()

template<class Vec , class Ops >

NewtonSolver< Vec, Ops >::~NewtonSolver ( )

overridedefault

◆ NewtonSolver() [2/3]

template<class Vec , class Ops >

NewtonSolver< Vec, Ops >::NewtonSolver ( const NewtonSolver< Vec, Ops > & )

delete

◆ NewtonSolver() [3/3]

template<class Vec , class Ops >

NewtonSolver< Vec, Ops >::NewtonSolver ( NewtonSolver< Vec, Ops > && )

deletenoexcept

Member Function Documentation

◆ CurTime()

template<class Vec , class Ops >

void NewtonSolver< Vec, Ops >::CurTime ( amrex::Real a_time ) const

inline

◆ CurTimeStep()

template<class Vec , class Ops >

void NewtonSolver< Vec, Ops >::CurTimeStep ( amrex::Real a_dt ) const

inline

◆ Define()

template<class Vec , class Ops >

void NewtonSolver< Vec, Ops >::Define	(	const Vec &	,
		Ops *
	)

overridevirtual

Read user-provided parameters that control the nonlinear solver. Allocate intermediate data containers needed by the solver. For Newton, setup the linear solver for computing the Newton step.

Implements NonlinearSolver< Vec, Ops >.

◆ EvalResidual()

template<class Vec , class Ops >

void NewtonSolver< Vec, Ops >::EvalResidual	(	Vec &	a_F,
		const Vec &	a_U,
		const Vec &	a_b,
		amrex::Real	a_time,
		amrex::Real	a_dt,
		int	a_iter
	)		const

private

Compute the nonlinear residual: F(U) = U - b - R(U).

◆ GetSolverParams()

template<class Vec , class Ops >

void NewtonSolver< Vec, Ops >::GetSolverParams	(	amrex::Real &	,
		amrex::Real &	,
		int &
	)

inlineoverridevirtual

Return the convergence parameters used by the nonlinear solver.

Implements NonlinearSolver< Vec, Ops >.

◆ operator=() [1/2]

template<class Vec , class Ops >

NewtonSolver& NewtonSolver< Vec, Ops >::operator= ( const NewtonSolver< Vec, Ops > & )

delete

◆ operator=() [2/2]

template<class Vec , class Ops >

NewtonSolver& NewtonSolver< Vec, Ops >::operator= ( NewtonSolver< Vec, Ops > && )

deletenoexcept

◆ ParseParameters()

template<class Vec , class Ops >

void NewtonSolver< Vec, Ops >::ParseParameters

private

◆ PrintParams()

template<class Vec , class Ops >

void NewtonSolver< Vec, Ops >::PrintParams ( ) const

inlineoverridevirtual

Print parameters used by the nonlinear solver.

Implements NonlinearSolver< Vec, Ops >.

◆ Solve()

template<class Vec , class Ops >

void NewtonSolver< Vec, Ops >::Solve	(	Vec &	,
		const Vec &	,
		amrex::Real	,
		amrex::Real
	)		const

overridevirtual

Solve the specified nonlinear equation for U. Picard: U = b + R(U). Newton: F(U) = U - b - R(U) = 0.

Implements NonlinearSolver< Vec, Ops >.

Member Data Documentation

◆ m_atol

template<class Vec , class Ops >

amrex::Real NewtonSolver< Vec, Ops >::m_atol = 0.

private

Absolute tolerance for the Newton solver.

◆ m_cur_time

template<class Vec , class Ops >

amrex::Real NewtonSolver< Vec, Ops >::m_cur_time

mutableprivate

◆ m_dt

template<class Vec , class Ops >

amrex::Real NewtonSolver< Vec, Ops >::m_dt

private

◆ m_dU

template<class Vec , class Ops >

Vec NewtonSolver< Vec, Ops >::m_dU

mutableprivate

Intermediate Vec containers used by the solver.

◆ m_F

template<class Vec , class Ops >

Vec NewtonSolver< Vec, Ops >::m_F

private

◆ m_gmres_atol

template<class Vec , class Ops >

amrex::Real NewtonSolver< Vec, Ops >::m_gmres_atol = 0.

private

Absolute tolerance for GMRES.

◆ m_gmres_maxits

template<class Vec , class Ops >

int NewtonSolver< Vec, Ops >::m_gmres_maxits = 1000

private

Maximum iterations for GMRES.

◆ m_gmres_restart_length

template<class Vec , class Ops >

int NewtonSolver< Vec, Ops >::m_gmres_restart_length = 30

private

Restart iteration for GMRES.

◆ m_gmres_rtol

template<class Vec , class Ops >

amrex::Real NewtonSolver< Vec, Ops >::m_gmres_rtol = 1.0e-4

private

Relative tolerance for GMRES.

◆ m_gmres_verbose_int

template<class Vec , class Ops >

int NewtonSolver< Vec, Ops >::m_gmres_verbose_int = 2

private

Verbosity flag for GMRES.

◆ m_linear_function

template<class Vec , class Ops >

std::unique_ptr<JacobianFunctionMF<Vec,Ops> > NewtonSolver< Vec, Ops >::m_linear_function

private

The linear function used by GMRES to compute A*v. In the contect of JFNK, A = dF/dU (i.e., system Jacobian)

◆ m_linear_solver

template<class Vec , class Ops >

std::unique_ptr<amrex::GMRES<Vec,JacobianFunctionMF<Vec,Ops> > > NewtonSolver< Vec, Ops >::m_linear_solver

private

The linear solver (GMRES) object.

◆ m_maxits

template<class Vec , class Ops >

int NewtonSolver< Vec, Ops >::m_maxits = 100

private

Maximum iterations for the Newton solver.

◆ m_ops

template<class Vec , class Ops >

Ops* NewtonSolver< Vec, Ops >::m_ops = nullptr

private

Pointer to Ops class.

◆ m_R

template<class Vec , class Ops >

Vec NewtonSolver< Vec, Ops >::m_R

private

◆ m_require_convergence

template<class Vec , class Ops >

bool NewtonSolver< Vec, Ops >::m_require_convergence = true

private

Flag to determine whether convergence is required.

◆ m_rtol

template<class Vec , class Ops >

amrex::Real NewtonSolver< Vec, Ops >::m_rtol = 1.0e-6

private

Relative tolerance for the Newton solver.

◆ m_update_pc

template<class Vec , class Ops >

bool NewtonSolver< Vec, Ops >::m_update_pc = false

mutableprivate

◆ m_update_pc_init

template<class Vec , class Ops >

bool NewtonSolver< Vec, Ops >::m_update_pc_init = false

mutableprivate

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

/home/docs/checkouts/readthedocs.org/user_builds/warpx/checkouts/24.08/Source/NonlinearSolvers/NewtonSolver.H

Public Member Functions

Private Member Functions

Private Attributes

Additional Inherited Members

Detailed Description

template<class Vec, class Ops> class NewtonSolver< Vec, Ops >

Constructor & Destructor Documentation

◆ NewtonSolver() [1/3]

◆ ~NewtonSolver()

◆ NewtonSolver() [2/3]

◆ NewtonSolver() [3/3]

Member Function Documentation

◆ CurTime()

◆ CurTimeStep()

◆ Define()

◆ EvalResidual()

◆ GetSolverParams()

◆ operator=() [1/2]

◆ operator=() [2/2]

◆ ParseParameters()

◆ PrintParams()

◆ Solve()

Member Data Documentation

◆ m_atol

◆ m_cur_time

◆ m_dt

◆ m_dU

◆ m_F

◆ m_gmres_atol

◆ m_gmres_maxits

◆ m_gmres_restart_length

◆ m_gmres_rtol

◆ m_gmres_verbose_int

◆ m_linear_function

◆ m_linear_solver

◆ m_maxits

◆ m_ops

◆ m_R

◆ m_require_convergence

◆ m_rtol

◆ m_update_pc

◆ m_update_pc_init

template<class Vec, class Ops>
class NewtonSolver< Vec, Ops >