cs_saddle_itsol.c File Reference

In-house iterative solvers defined by blocks and associated to CDO discretizations. More...

#include "cs_defs.h"
#include <assert.h>
#include <float.h>
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <bft_mem.h>
#include "cs_array.h"
#include "cs_blas.h"
#include "cs_cdo_solve.h"
#include "cs_log.h"
#include "cs_parall.h"
#include "cs_parameters.h"
#include "cs_saddle_itsol.h"

Include dependency graph for cs_saddle_itsol.c:

Functions
cs_saddle_block_precond_t *	cs_saddle_block_precond_create (cs_param_precond_block_t block_type, cs_param_schur_approx_t schur_type, cs_param_sles_t *m11_slesp, cs_sles_t *m11_sles)
	Create and initialize a cs_saddle_block_precond_t structure. More...
void	cs_saddle_block_precond_free (cs_saddle_block_precond_t **p_sbp)
	Free a cs_saddle_block_precond_t structure. More...
void	cs_saddle_minres (cs_saddle_system_t *ssys, cs_saddle_block_precond_t *sbp, cs_real_t *x1, cs_real_t *x2, cs_iter_algo_t *algo)
	Apply the MINRES algorithm to a saddle point problem (the system is stored in a hybrid way). Please refer to cs_saddle_system_t structure definition. The stride is equal to 1 for the matrix (db_size[3] = 1) and the vector. More...
void	cs_saddle_gcr (int restart, cs_saddle_system_t *ssys, cs_saddle_block_precond_t *sbp, cs_real_t *x1, cs_real_t *x2, cs_iter_algo_t *algo)
	Apply the GCR algorithm to a saddle point problem (the system is stored in a hybrid way). Please refer to cs_saddle_system_t structure definition. The stride is equal to 1 for the matrix (db_size[3] = 1) and the vector. This algorithm is taken from 2010 Notay's paper: "An aggregation-based algebraic multigrid method" ETNA (vol. 37) More...
void	cs_matrix_vector_multiply_gs_allocated (const cs_range_set_t *rset, const cs_matrix_t *mat, cs_real_t *vec, cs_real_t *matvec)
	Perform a matrix-vector multiplication in case of scatter-view array as input parameter. Thus, one performs a scatter --> gather (before the multiplication) and a gather --> scatter operation after the multiplication. One assumes that matvec is allocated to the right size. No check is done. More...
void	cs_matrix_vector_multiply_gs (const cs_range_set_t *rset, const cs_matrix_t *mat, cs_lnum_t vec_len, cs_real_t *vec, cs_real_t **p_matvec)
	Perform a matrix-vector multiplication in case of scatter-view array as input parameter. Thus, one performs a scatter --> gather (before the multiplication) and a gather --> scatter operation after the multiplication. The output parameter matvec is not allocated. A check on the size is done for the input array. More...

Detailed Description

In-house iterative solvers defined by blocks and associated to CDO discretizations.

Function Documentation

cs_matrix_vector_multiply_gs()

void cs_matrix_vector_multiply_gs	(	const cs_range_set_t *	rset,
		const cs_matrix_t *	mat,
		cs_lnum_t	vec_len,
		cs_real_t *	vec,
		cs_real_t **	p_matvec
	)

Perform a matrix-vector multiplication in case of scatter-view array as input parameter. Thus, one performs a scatter --> gather (before the multiplication) and a gather --> scatter operation after the multiplication. The output parameter matvec is not allocated. A check on the size is done for the input array.

The stride is equal to 1 for the matrix (db_size[3] = 1) and the vector

Parameters

[in]	rset	pointer to a cs_range_set_t structure
[in]	mat	matrix
[in]	vec_len	size of vec
[in,out]	vec	vector of real numbers
[out]	p_matvec	resulting vector for the matrix-vector product

cs_matrix_vector_multiply_gs_allocated()

void cs_matrix_vector_multiply_gs_allocated	(	const cs_range_set_t *	rset,
		const cs_matrix_t *	mat,
		cs_real_t *	vec,
		cs_real_t *	matvec
	)

Perform a matrix-vector multiplication in case of scatter-view array as input parameter. Thus, one performs a scatter --> gather (before the multiplication) and a gather --> scatter operation after the multiplication. One assumes that matvec is allocated to the right size. No check is done.

The stride is equal to 1 for the matrix (db_size[3] = 1) and the vector

Parameters

[in]	rset	pointer to a cs_range_set_t structure
[in]	mat	matrix
[in,out]	vec	vector
[in,out]	matvec	resulting vector for the matrix-vector product

cs_saddle_block_precond_create()

cs_saddle_block_precond_t* cs_saddle_block_precond_create	(	cs_param_precond_block_t	block_type,
		cs_param_schur_approx_t	schur_type,
		cs_param_sles_t *	m11_slesp,
		cs_sles_t *	m11_sles
	)

Create and initialize a cs_saddle_block_precond_t structure.

Parameters

[in]	block_type	type of block preconditioner
[in]	schur_type	type of Schur approximation
[in]	m11_slesp	pointer to the settings for the M11 block
[in]	m11_sles	pointer to the cs_sles_t struct. for the M11 block

Returns

a pointer to the new allocated strcuture

cs_saddle_block_precond_free()

void cs_saddle_block_precond_free

(

cs_saddle_block_precond_t **

p_sbp

)

Free a cs_saddle_block_precond_t structure.

Parameters

[in,out]

p_sbp

double pointer to the structure to free

cs_saddle_gcr()

void cs_saddle_gcr	(	int	restart,
		cs_saddle_system_t *	ssys,
		cs_saddle_block_precond_t *	sbp,
		cs_real_t *	x1,
		cs_real_t *	x2,
		cs_iter_algo_t *	algo
	)

Apply the GCR algorithm to a saddle point problem (the system is stored in a hybrid way). Please refer to cs_saddle_system_t structure definition. The stride is equal to 1 for the matrix (db_size[3] = 1) and the vector. This algorithm is taken from 2010 Notay's paper: "An aggregation-based algebraic multigrid method" ETNA (vol. 37)

Parameters

[in]	restart	number of iterations before restarting
[in]	ssys	pointer to a cs_saddle_system_t structure
[in]	sbp	block-preconditioner for the Saddle-point problem
[in,out]	x1	array for the first part
[in,out]	x2	array for the second part
[in,out]	algo	pointer to a cs_iter_algo_t structure

cs_saddle_minres()

void cs_saddle_minres	(	cs_saddle_system_t *	ssys,
		cs_saddle_block_precond_t *	sbp,
		cs_real_t *	x1,
		cs_real_t *	x2,
		cs_iter_algo_t *	algo
	)

Apply the MINRES algorithm to a saddle point problem (the system is stored in a hybrid way). Please refer to cs_saddle_system_t structure definition. The stride is equal to 1 for the matrix (db_size[3] = 1) and the vector.

Parameters

[in]	ssys	pointer to a cs_saddle_system_t structure
[in]	sbp	Block-preconditioner for the Saddle-point problem
[in,out]	x1	array for the first part
[in,out]	x2	array for the second part
[in,out]	algo	pointer to a cs_iter_algo_t structure