cs_cdocb_scaleq_sles.cpp File Reference

Functions dedicated to to the linear algebra settings and operations in case of CDO cell-based schemes with a scalar-valued equations. More...

#include "cs_defs.h"
#include <assert.h>
#include <string.h>
#include <bft_mem.h>
#include "cs_array.h"
#include "cs_blas.h"
#include "cs_cdo_blas.h"
#include "cs_cdo_solve.h"
#include "cs_equation.h"
#include "cs_cdocb_scaleq.h"
#include "cs_fp_exception.h"
#include "cs_matrix_default.h"
#include "cs_parall.h"
#include "cs_saddle_solver.h"
#include "cs_saddle_system.h"
#include "cs_timer.h"
#include "cs_cdocb_scaleq_sles.h"

Include dependency graph for cs_cdocb_scaleq_sles.cpp:

Functions
void	cs_cdocb_scaleq_sles_init_sharing (const cs_mesh_t *mesh, const cs_cdo_connect_t *connect, const cs_cdo_quantities_t *quant)
	Set pointers to shared structures. More...
void	cs_cdocb_scaleq_sles_init_system_helper (const cs_param_saddle_t *saddlep, cs_cdocb_scaleq_t *eqc)
	Define the system helper for a CDO-Cb scheme solving a scalar-valued equation (saddle-point system) More...
void	cs_cdocb_scaleq_sles_init_solver (const cs_equation_param_t *eqp, const cs_param_saddle_t *saddlep, cs_cdocb_scaleq_t *eqc)
	Define the saddle solver and its context for a CDO-Cb scheme solving a scalar-valued equation (saddle-point problem) More...
int	cs_cdocb_scaleq_sles_alu (cs_saddle_solver_t *solver, cs_real_t *flux, cs_real_t *pot)
	Solve a linear system arising from the discretization of a scalar-valed equation stemming from CDO cell-based schemes. Solve the saddle-point system using the Augmented Lagrangian-Uzawa algorithm. More...
int	cs_cdocb_scaleq_sles_block_krylov (cs_saddle_solver_t *solver, cs_real_t *flux, cs_real_t *pot)
	Solve a linear system arising from the discretization of a scalar-valed equation stemming from CDO cell-based schemes. The system is split into a flux block and the (unassembled) divergence operator. Block preconditioning using a Schur approximation on a Krylov solver such as the GCR or MINRES is available. More...
int	cs_cdocb_scaleq_sles_full_system (cs_saddle_solver_t *solver, cs_real_t *flux, cs_real_t *pot)
	Solve the saddle-point linear system arising from the discretization of the scalar-valued CDO cell-based scheme. The full system is treated as one block and solved as it is. In this situation, PETSc or MUMPS are usually considered. More...
int	cs_cdocb_scaleq_sles_gkb_inhouse (cs_saddle_solver_t *solver, cs_real_t *flux, cs_real_t *pot)
	Solve a linear system arising from the discretization of a scalar-valed equation stemming from CDO cell-based schemes. Solve this system using the Golub-Kahan Bidiagonalization algorithm. In-house implementation. The PETSc implementation is also available but appears less efficient in our tests. More...
int	cs_cdocb_scaleq_sles_notay (cs_saddle_solver_t *solver, cs_real_t *flux, cs_real_t *pot)
	Solve a linear system arising from the discretization of a scalar-valed equation stemming from CDO cell-based schemes. Solve this system using the Notay's algebraic transformation. The full system is treated as one block and solved as it is. In this situation, PETSc or MUMPS are usually considered. More...
int	cs_cdocb_scaleq_sles_uzawa_cg (cs_saddle_solver_t *solver, cs_real_t *flux, cs_real_t *pot)
	Solve a linear system arising from the discretization of a scalar-valed equation stemming from CDO cell-based schemes. Solve this system using the Uzawa-CG algorithm. More...

Detailed Description

Functions dedicated to to the linear algebra settings and operations in case of CDO cell-based schemes with a scalar-valued equations.

Function Documentation

cs_cdocb_scaleq_sles_alu()

int cs_cdocb_scaleq_sles_alu	(	cs_saddle_solver_t *	solver,
		cs_real_t *	flux,
		cs_real_t *	pot
	)

Solve a linear system arising from the discretization of a scalar-valed equation stemming from CDO cell-based schemes. Solve the saddle-point system using the Augmented Lagrangian-Uzawa algorithm.

Parameters

[in,out]	solver	pointer to a cs_saddle_solver_t structure
[in,out]	flux	values of the flux at faces (scalar)
[in,out]	pot	values of the potential in cells

Returns

the (cumulated) number of iterations of the solver

cs_cdocb_scaleq_sles_block_krylov()

int cs_cdocb_scaleq_sles_block_krylov	(	cs_saddle_solver_t *	solver,
		cs_real_t *	flux,
		cs_real_t *	pot
	)

Solve a linear system arising from the discretization of a scalar-valed equation stemming from CDO cell-based schemes. The system is split into a flux block and the (unassembled) divergence operator. Block preconditioning using a Schur approximation on a Krylov solver such as the GCR or MINRES is available.

Parameters

[in,out]	solver	pointer to a cs_saddle_solver_t structure
[in,out]	flux	values of the flux at faces (scalar)
[in,out]	pot	values of the potential in cells

Returns

the (cumulated) number of iterations of the solver

cs_cdocb_scaleq_sles_full_system()

int cs_cdocb_scaleq_sles_full_system	(	cs_saddle_solver_t *	solver,
		cs_real_t *	flux,
		cs_real_t *	pot
	)

Solve the saddle-point linear system arising from the discretization of the scalar-valued CDO cell-based scheme. The full system is treated as one block and solved as it is. In this situation, PETSc or MUMPS are usually considered.

Parameters

[in,out]	solver	pointer to a saddle-point solver
[in,out]	flux	values of the flux at faces (scalar)
[in,out]	pot	values of the potential in cells

Returns

the (cumulated) number of iterations of the solver

cs_cdocb_scaleq_sles_gkb_inhouse()

int cs_cdocb_scaleq_sles_gkb_inhouse	(	cs_saddle_solver_t *	solver,
		cs_real_t *	flux,
		cs_real_t *	pot
	)

Solve a linear system arising from the discretization of a scalar-valed equation stemming from CDO cell-based schemes. Solve this system using the Golub-Kahan Bidiagonalization algorithm. In-house implementation. The PETSc implementation is also available but appears less efficient in our tests.

Parameters

[in,out]	solver	pointer to a cs_saddle_solver_t structure
[in,out]	flux	values of the flux at faces (scalar)
[in,out]	pot	values of the potential in cells

Returns

the (cumulated) number of iterations of the solver

cs_cdocb_scaleq_sles_init_sharing()

void cs_cdocb_scaleq_sles_init_sharing	(	const cs_mesh_t *	mesh,
		const cs_cdo_connect_t *	connect,
		const cs_cdo_quantities_t *	quant
	)

Set pointers to shared structures.

Parameters

[in]	mesh	pointer to the mesh structure
[in]	connect	pointer to additional CDO connectivities
[in]	quant	pointer to additional CDO mesh quantities

cs_cdocb_scaleq_sles_init_solver()

void cs_cdocb_scaleq_sles_init_solver	(	const cs_equation_param_t *	eqp,
		const cs_param_saddle_t *	saddlep,
		cs_cdocb_scaleq_t *	eqc
	)

Define the saddle solver and its context for a CDO-Cb scheme solving a scalar-valued equation (saddle-point problem)

Parameters

[in]	eqp	set of equation parameters
[in]	saddlep	parameters for solving a saddle-point problem
[in,out]	eqc	pointer to a context structure cast on-the-fly

cs_cdocb_scaleq_sles_init_system_helper()

void cs_cdocb_scaleq_sles_init_system_helper	(	const cs_param_saddle_t *	saddlep,
		cs_cdocb_scaleq_t *	eqc
	)

Define the system helper for a CDO-Cb scheme solving a scalar-valued equation (saddle-point system)

Parameters

[in]	saddlep	parameters for solving a saddle-point problem
[in,out]	eqc	pointer to a context structure cast on-the-fly

cs_cdocb_scaleq_sles_notay()

int cs_cdocb_scaleq_sles_notay	(	cs_saddle_solver_t *	solver,
		cs_real_t *	flux,
		cs_real_t *	pot
	)

Solve a linear system arising from the discretization of a scalar-valed equation stemming from CDO cell-based schemes. Solve this system using the Notay's algebraic transformation. The full system is treated as one block and solved as it is. In this situation, PETSc or MUMPS are usually considered.

Parameters

[in,out]	solver	pointer to a cs_saddle_solver_t structure
[in,out]	flux	values of the flux at faces (scalar)
[in,out]	pot	values of the potential in cells

Returns

the (cumulated) number of iterations of the solver

cs_cdocb_scaleq_sles_uzawa_cg()

int cs_cdocb_scaleq_sles_uzawa_cg	(	cs_saddle_solver_t *	solver,
		cs_real_t *	flux,
		cs_real_t *	pot
	)

Solve a linear system arising from the discretization of a scalar-valed equation stemming from CDO cell-based schemes. Solve this system using the Uzawa-CG algorithm.

Parameters

[in,out]	solver	pointer to a cs_saddle_solver_t structure
[in,out]	flux	values of the flux at faces (scalar)
[in,out]	pot	values of the potential in cells

Returns

the (cumulated) number of iterations of the solver