SF::abstract_nonlinear_solver< T, S > Struct Template Reference

NOTE: A LOT OF THESE THINGS NEED TO BE CHANGED AT A LATER STAGE. THEY ARE CURRENTLY COPIED FROM THE ABSTRACT_LINEAR_SOLVER. More...

#include <SF_abstract_lin_solver.h>

Collaboration diagram for SF::abstract_nonlinear_solver< T, S >:

Public Types
enum	norm_t {   absPreResidual , absUnpreResidual , relResidual , absPreRelResidual ,   norm_unset }

Public Member Functions
virtual void	operator() (abstract_vector< T, S > &x, const abstract_vector< T, S > &b)=0
virtual	~abstract_nonlinear_solver ()=default
virtual void	setup_solver (abstract_vector< T, S > &residuum, abstract_matrix< T, S > &mat, double tol, int max_it, short norm, const char *name, bool has_nullspace, void *logger, const char *solver_opts_file, const char *default_opts, PetscErrorCode function(SNES, Vec, Vec, void *), PetscErrorCode jacobian(SNES, Vec, Mat, Mat, void *), void *solver_pointer=NULL)=0

Public Attributes
const char *	name
	the solver name More...
const char *	options_file
	the file containing the solver options More...
norm_t	norm = norm_unset
double	final_residual = -1.0
	Holds the residual after convergence. More...
int	niter = -1
	number of iterations More...
int	max_it = 0
	max number of iterations More...
int	reason = 0
	number of iterations More...
double	time = 0.0
	solver runtime More...

Protected Member Functions
virtual void	set_stopping_criterion (norm_t normtype, double tol, int max_it, bool verbose, void *logger)=0
norm_t	convert_param_norm_type (short param_norm_type) const

Detailed Description

template<class T, class S>
struct SF::abstract_nonlinear_solver< T, S >

NOTE: A LOT OF THESE THINGS NEED TO BE CHANGED AT A LATER STAGE. THEY ARE CURRENTLY COPIED FROM THE ABSTRACT_LINEAR_SOLVER.

Simple struct holding informations and operations related to nonlinear solvers.

Template Parameters

T	Integer type (indices).
S	Floating point type (values).

See also

ginkgo_solver

petsc_solver

Note

The initialization of solvers varies hugely, so we provide a generic setup function containing redundant information for some solvers.

Definition at line 138 of file SF_abstract_lin_solver.h.

Member Enumeration Documentation

norm_t

template<class T , class S >

enum SF::abstract_nonlinear_solver::norm_t

Enum representing the residual norm stopping criterion type.

Enumerator
absPreResidual
absUnpreResidual
relResidual
absPreRelResidual
norm_unset

Definition at line 144 of file SF_abstract_lin_solver.h.

Constructor & Destructor Documentation

~abstract_nonlinear_solver()

template<class T , class S >

virtual SF::abstract_nonlinear_solver< T, S >::~abstract_nonlinear_solver ( )

virtualdefault

Default destructor.

Member Function Documentation

convert_param_norm_type()

template<class T , class S >

norm_t SF::abstract_nonlinear_solver< T, S >::convert_param_norm_type ( short  param_norm_type ) const

inlineprotected

Simple helper function converting a short (number) to norm_t

Parameters

param_norm_type

a short number representing the norm type

Returns

the norm type as a norm_t enum

Definition at line 207 of file SF_abstract_lin_solver.h.

operator()()

template<class T , class S >

virtual void SF::abstract_nonlinear_solver< T, S >::operator() ( abstract_vector< T, S > &  x, const abstract_vector< T, S > &  b  )

pure virtual

Solve the system Ax=b.

Parameters

x	The solution vector
b	The right hand side

set_stopping_criterion()

template<class T , class S >

virtual void SF::abstract_nonlinear_solver< T, S >::set_stopping_criterion ( norm_t  normtype, double  tol, int  max_it, bool  verbose, void *  logger  )

protectedpure virtual

Set the stopping criterion parameters for the solver.

Parameters

normtype	The type of residual norm.
tol	The tolerance to which the solver should converge.
mat_it	The maximum number of iterations for the solver.
verbose	Whether to print detailed informations.
logger	A generic pointer to the logger object.

setup_solver()

template<class T , class S >

virtual void SF::abstract_nonlinear_solver< T, S >::setup_solver ( abstract_vector< T, S > &  residuum, abstract_matrix< T, S > &  mat, double  tol, int  max_it, short  norm, const char *  name, bool  has_nullspace, void *  logger, const char *  solver_opts_file, const char *  default_opts, PetscErrorCode   functionSNES, Vec, Vec, void *, PetscErrorCode   jacobianSNES, Vec, Mat, Mat, void *, void *  solver_pointer = NULL  )

pure virtual

Setup the solver characteristics.

Parameters

residuum	The residual vector on the rhs of the nonlinear equation
mat	The system matrix.
tol	The tolerance to which the solver should converge.
mat_it	The maximum number of iterations for the solver.
norm
name	The solver name.
has_nullspace	Whether there is a constant nullspace or not.
logger	A generic pointer to the logger object.
solver_opts_file	File where the solver options are stored (not implemented yet)
default_opts	A string representing default PETSc options. Used by petsc only.
function	Function to be evaluated in the SNES solver scheme. Needs to be a PetscErrorCode with input parameters of type (SNES, Vec, Vec, void *)

Member Data Documentation

final_residual

template<class T , class S >

double SF::abstract_nonlinear_solver< T, S >::final_residual = -1.0

Holds the residual after convergence.

Definition at line 148 of file SF_abstract_lin_solver.h.

max_it

template<class T , class S >

int SF::abstract_nonlinear_solver< T, S >::max_it = 0

max number of iterations

Definition at line 150 of file SF_abstract_lin_solver.h.

name

template<class T , class S >

const char* SF::abstract_nonlinear_solver< T, S >::name

the solver name

Definition at line 140 of file SF_abstract_lin_solver.h.

niter

template<class T , class S >

int SF::abstract_nonlinear_solver< T, S >::niter = -1

number of iterations

Definition at line 149 of file SF_abstract_lin_solver.h.

norm

template<class T , class S >

norm_t SF::abstract_nonlinear_solver< T, S >::norm = norm_unset

Definition at line 145 of file SF_abstract_lin_solver.h.

options_file

template<class T , class S >

const char* SF::abstract_nonlinear_solver< T, S >::options_file

the file containing the solver options

Definition at line 141 of file SF_abstract_lin_solver.h.

reason

template<class T , class S >

int SF::abstract_nonlinear_solver< T, S >::reason = 0

number of iterations

Definition at line 151 of file SF_abstract_lin_solver.h.

time

template<class T , class S >

double SF::abstract_nonlinear_solver< T, S >::time = 0.0

solver runtime

Definition at line 152 of file SF_abstract_lin_solver.h.

---

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

• fem/slimfem/src/SF_abstract_lin_solver.h