CoRiBoDynamics::ConjugateGradientSolver Class Reference

#include <ConjugateGradient.h>

Collaboration diagram for CoRiBoDynamics::ConjugateGradientSolver:

Classes
class	VectorTask

Public Member Functions
	ConjugateGradientSolver (LinearOperator *A)
void	SetThreadPool (CoreBoundThreadPool *ThreadPool)
void	solve (const Vector &B, Vector &X, double &tolerance, int &iterations)
void	ParallellizeVectorOperations (bool b)

Protected Member Functions
double	CG_squared_norm (const Vector &a)
void	CG_flip_and_add (const Vector &a, Vector &b)
void	CG_scale_and_add (const Vector &a, double beta, Vector &b)
double	MultiOP (Vector &p, Vector &tmp, double alpha, Vector &x, Vector &r)

Protected Attributes
LinearOperator *	m_A
Vector	m_residual
Vector	m_p
Vector	m_z
Vector	m_tmp
CoreBoundThreadPool *	m_ThreadPool
bool	m_parallelize_vector_operations
std::vector< VectorTask >	m_vector_task

Detailed Description

Solves for the linear equation A*X=B, for a given tolerance. Main purpose is to keep working memory persistent to minimize allocation overhead, and manage parallellism of vector operations

Interface for an general linear (matrix) operation. Important functions are to apply the linear operator to a vector, and to compute an approximation of the operator inverse.

Member Function Documentation

solve()

void CoRiBoDynamics::ConjugateGradientSolver::solve	(	const Vector &	B,
		Vector &	X,
		double &	tolerance,
		int &	iterations
	)

Solves for the linear equation A*X=B, for a given 'tolerance'. Uses the given X as starting iteration value; 'iterations' specifies max iteration number, and will be set to actual iteration count

---

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

• reloadrepos/fhsim_coribo/src/Utilities/ConjugateGradient.h