monolish::equation::SOR< MATRIX, Float > Class Template Reference

SOR solver class. More...

#include <monolish_equation.hpp>

Inheritance diagram for monolish::equation::SOR< MATRIX, Float >:

Collaboration diagram for monolish::equation::SOR< MATRIX, Float >:

Public Member Functions
void	apply_precond (const vector< Float > &r, vector< Float > &z)
void	create_precond (MATRIX &A)
std::string	name () const
	get solver name "monolish::equation::SOR" More...
int	solve (MATRIX &A, vector< Float > &x, vector< Float > &b)
	solve Ax = b by SOR method(lib=0: monolish) More...
std::string	solver_name () const
	get solver name "SOR" More...
Public Member Functions inherited from monolish::solver::solver< MATRIX, Float >
size_t	get_final_iter ()
double	get_final_residual ()
initvec_scheme	get_initvec_scheme () const
	get handling scheme of initial vector handling More...
int	get_lib () const
	get library option More...
size_t	get_maxiter () const
	get maxiter More...
size_t	get_miniter () const
	get miniter More...
Float	get_omega ()
	get the relaxation coefficient omega for SOR method (Default: 1.9) More...
bool	get_print_rhistory () const
	get print rhistory status More...
int	get_reorder ()
	0: no ordering 1: symrcm, 2: symamd, 3: csrmetisnd is used to reduce zero fill-in. More...
size_t	get_residual_method () const
	get residual method(default=0) More...
int	get_singularity ()
	-1 if A is symmetric positive definite. default reorder algorithm is csrmetisnd More...
double	get_tol () const
	get tolerance More...
template<class PRECOND >
void	set_apply_precond (PRECOND &p)
	set precondition apply function More...
template<class PRECOND >
void	set_create_precond (PRECOND &p)
	set precondition create function More...
void	set_initvec_scheme (initvec_scheme scheme)
	set how to handle initial vector More...
void	set_lib (int l)
	set library option (to change library, monolish, cusolver, etc.) More...
void	set_maxiter (size_t max)
	set max iter. (default = SIZE_MAX) More...
void	set_miniter (size_t min)
	set min iter. (default = 0) More...
void	set_omega (Float w)
	set the relaxation coefficient omega for SOR method (0 < w < 2, Default: 1.9) More...
void	set_print_rhistory (bool flag)
	print rhistory to standart out true/false. (default = false) More...
void	set_reorder (int r)
	0: no ordering 1: symrcm, 2: symamd, 3: csrmetisnd is used to reduce zero fill-in. More...
void	set_residual_method (size_t r)
	set residual method (default=0) More...
void	set_rhistory_filename (std::string file)
	rhistory filename More...
void	set_tol (double t)
	set tolerance (default:1.0e-8) More...
	solver ()
	create solver class More...
	~solver ()
	delete solver class More...

Private Member Functions
int	monolish_SOR (MATRIX &A, vector< Float > &x, vector< Float > &b)

Additional Inherited Members
Protected Member Functions inherited from monolish::solver::solver< MATRIX, Float >
Float	get_residual (vector< Float > &x)
Protected Attributes inherited from monolish::solver::solver< MATRIX, Float >
size_t	final_iter = 0
double	final_resid = 0
initvec_scheme	initvecscheme = initvec_scheme::RANDOM
int	lib = 0
size_t	maxiter = SIZE_MAX
size_t	miniter = 0
Float	omega = 1.9
precondition< MATRIX, Float >	precond
bool	print_rhistory = false
int	reorder = 3
size_t	resid_method = 0
std::string	rhistory_file
std::ostream *	rhistory_stream
int	singularity
double	tol = 1.0e-8

Detailed Description

template<typename MATRIX, typename Float>
class monolish::equation::SOR< MATRIX, Float >

SOR solver class.

Note

attribute:

• solver : true
• preconditioner : true

input / architecture

• Dense / Intel : true
• Dense / NVIDIA : true
• Dense / OSS : true
• Sparse / Intel : true
• Sparse / NVIDIA : true
• Sparse / OSS : true

Warning

SOR is not completely parallelized. The part of solving the lower triangular matrix is performed sequentially. On the GPU, one vector is received before the lower triangular matrix solving process, and one vector is sent after the process.

Definition at line 286 of file monolish_equation.hpp.

Member Function Documentation

apply_precond()

template<typename MATRIX , typename Float >

void monolish::equation::SOR< MATRIX, Float >::apply_precond	(	const vector< Float > &	r,
		vector< Float > &	z
	)

create_precond()

template<typename MATRIX , typename Float >

void monolish::equation::SOR< MATRIX, Float >::create_precond

(

MATRIX &

A

)

monolish_SOR()

template<typename MATRIX , typename Float >

int monolish::equation::SOR< MATRIX, Float >::monolish_SOR ( MATRIX &  A, vector< Float > &  x, vector< Float > &  b  )

private

name()

template<typename MATRIX , typename Float >

std::string monolish::equation::SOR< MATRIX, Float >::name ( ) const

inline

get solver name "monolish::equation::SOR"

Note

• # of computation: 1
• Multi-threading: false
• GPU acceleration: false

Definition at line 315 of file monolish_equation.hpp.

solve()

template<typename MATRIX , typename Float >

int monolish::equation::SOR< MATRIX, Float >::solve	(	MATRIX &	A,
		vector< Float > &	x,
		vector< Float > &	b
	)

solve Ax = b by SOR method(lib=0: monolish)

Parameters

[in]	A	CRS format Matrix
[in]	x	solution vector
[in]	b	right hand vector

Returns

error code, see https://github.com/ricosjp/monolish/blob/master/include/common/monolish_common.hpp

Warning

SOR is not completely parallelized. The part of solving the lower triangular matrix is performed sequentially. On the GPU, one vector is received before the lower triangular matrix solving process, and one vector is sent after the process.

solver_name()

template<typename MATRIX , typename Float >

std::string monolish::equation::SOR< MATRIX, Float >::solver_name ( ) const

inline

get solver name "SOR"

Note

• # of computation: 1
• Multi-threading: false
• GPU acceleration: false

Definition at line 324 of file monolish_equation.hpp.

---

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

• monolish_equation.hpp