LinAlgebra Class Reference

#include <linear_algebra.h>

Inheritance diagram for LinAlgebra:

Public Member Functions
	LinAlgebra (Settings &s, Mesh::mesh &my_msh)
algebra::MatrixShape	build_shape ()
void	prepareElements (Eigen::Vector3d const &Hext, timing const &t_prm)
void	prepareElements (double const A_Hext, timing const &t_prm)
void	buildInitGuess (std::vector< double > &G) const
bool	solve (timing const &t_prm)
void	set_DW_vz (double vz)
double	get_v_max (void)
void	setExtSpaceField (Settings &s)
void	base_projection ()
Public Member Functions inherited from solver< DIM_PB_MAG >
	solver (Mesh::mesh &_msh, std::vector< Tetra::prm > &_pTetra, std::vector< Facette::prm > &_pFac, const std::string name, const double _tol, const bool v, const int max_iter)

Private Attributes
Nodes::index	idx_dir
algebra::r_sparseMat	K
std::vector< double >	L_rhs
std::vector< double >	Xw
const int	verbose
double	DW_vz
double	v_max
std::vector< Eigen::Matrix< double, Nodes::DIM, Tetra::NPI > >	extSpaceField
std::vector< int >	lvd

Additional Inherited Members
Protected Member Functions inherited from solver< DIM_PB_MAG >
void	buildMat (std::vector< int > &ind, Eigen::Matrix< double, DIM_PROBLEM *N, DIM_PROBLEM *N > &Ke, algebra::w_sparseMat &K)
void	buildVect (std::vector< int > &ind, std::vector< double > &Le, std::vector< double > &L)
Protected Attributes inherited from solver< DIM_PB_MAG >
Mesh::mesh *	msh
const int	NOD
const std::vector< Tetra::prm > &	paramTet
const std::vector< Facette::prm > &	paramFac
algebra::iteration< double >	iter
Static Protected Attributes inherited from solver< DIM_PB_MAG >
static const int	DIM_PB

Detailed Description

convenient class to grab altogether some part of the calculations involved using algebra::bicg solver at each timestep. The solver is handled by solver method, and is using algebra sparse matrices(Row major). The write sparse matrix is prepared in 'batch mode', to add all non zero coefficients with a '+=' logic. Then it is turned into a read sparse matrix before being used by bicg algo. Be aware of time units: when entering solver method, division by gamma0 and multiplication by gamma0 when ending are mandatory. The bicg algorithm is monitored by iter object. When debugging it might be usefull to set iter verbosity differently from LinAlgebra Solver (see constructor list initialization)

Constructor & Destructor Documentation

LinAlgebra()

LinAlgebra::LinAlgebra ( Settings &  s, Mesh::mesh &  my_msh  )

inline

constructor When debugging it might be usefull to set iter verbosity differently

Parameters

[in]	s
[in]	my_msh

Member Function Documentation

base_projection()

void LinAlgebra::base_projection

(

)

computes local vector basis {ep,eq} in the tangeant plane for projection on the elements

build_shape()

algebra::MatrixShape LinAlgebra::build_shape ( )

inline

build a matrix shape suitable for our mesh

buildInitGuess()

void LinAlgebra::buildInitGuess

(

std::vector< double > &

G

)

const

build init guess for bicg solver

Parameters

[out]

G

get_v_max()

double LinAlgebra::get_v_max ( void  )

inline

getter for v_max

prepareElements() [1/2]

void LinAlgebra::prepareElements	(	double const	A_Hext,
		timing const &	t_prm
	)

computes inner data structures of tetraedrons and triangular facettes (K matrices and L vectors) this member function is overloaded to fit to two different situations, either if std::function passed to element = tetra is corresponding to the simple case of constant external field applied to the magnetic region or space dependant. Here is the variable space applied field.

Parameters

[in]	A_Hext	amplitude applied field (might be time dependant)
[in]	t_prm

prepareElements() [2/2]

void LinAlgebra::prepareElements	(	Eigen::Vector3d const &	Hext,
		timing const &	t_prm
	)

computes inner data structures of tetraedrons and triangular facettes (K matrices and L vectors) this member function is overloaded to fit to two different situations, either if std::function passed to element = tetra is corresponding to the simple case of constant external field applied to the magnetic region or space dependant. Here is the constant space applied field.

Parameters

[in]	Hext	applied field
[in]	t_prm

set_DW_vz()

void LinAlgebra::set_DW_vz ( double  vz )

inline

setter for DW_dz

Parameters

[in]

vz

setExtSpaceField()

void LinAlgebra::setExtSpaceField

(

Settings &

s

)

when external applied field is of field_type R4toR3 values of field_space are stored in spaceField

Parameters

[in]

s

solve()

bool LinAlgebra::solve

(

timing const &

t_prm

)

call the solver, uses stabilized biconjugate gradient solver (bicg) with diagonal preconditionner, sparse matrix and vector are filled with multiThreading. Sparse matrix is row major.

Parameters

[in]

t_prm

Member Data Documentation

DW_vz

double LinAlgebra::DW_vz

private

speed of the domain wall

extSpaceField

std::vector< Eigen::Matrix<double,Nodes::DIM,Tetra::NPI> > LinAlgebra::extSpaceField

private

external applied space field, values on gauss points, size is number of tetraedrons

idx_dir

Nodes::index LinAlgebra::idx_dir

private

recentering index direction if any

K

algebra::r_sparseMat LinAlgebra::K

private

matrix of the system to solve

L_rhs

std::vector<double> LinAlgebra::L_rhs

private

RHS vector of the system to solve

lvd

std::vector<int> LinAlgebra::lvd

private

list of the Dirichlet indices where (vp,vq) are zero, initialized by constructor

v_max

double LinAlgebra::v_max

private

maximum speed of the magnetization in the whole physical object

verbose

const int LinAlgebra::verbose

private

verbosity

Xw

std::vector<double> LinAlgebra::Xw

private

solution of the system to solve

---

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

• linear_algebra.h
• linear_algebra.cpp
• solver.cpp