Iterative solvers

Classes and functions for iterative solution of linear equation systems. More...

Modules
	Multigrid
	Classes and functions for multigrid preconditioners and solvers.
	Preconditioners
	Preconditioners for iterative solvers, including smoothers for multigrid methods.
	QP Solvers
	Classes and methods for solving linear-quadratic programs.

Classes
class	Kaskade::PCGTerminationCriterion< R >
	Interface for IterateType::PCG termination criterion policy classes. More...
class	Kaskade::PCGCountingTerminationCriterion< R >
	PCG termination after a given number of iterations. More...
class	Kaskade::PCGEnergyErrorTerminationCriterion< R >
	TerminationCriterion based on an absolute energy error estimate. More...
class	Kaskade::Pcg< X, Xstar >
	preconditioned conjugate gradient method More...
class	Kaskade::ICC_0Preconditioner< Op >
	Incomplete Cholesky factorization by algorithm from book "Matrix Computations" by Gene Golub & Charles van Loan. More...

Functions
template<class Domain , class Range >
bool	Kaskade::requiresInitializedInput (Dune::Preconditioner< Domain, Range > const *p)
	Whether a preconditioner needs zero initialized result vector or not. More...

Detailed Description

Classes and functions for iterative solution of linear equation systems.

This includes on one hand actual solvers (Krylov methods such as Kaskade::Pcg) and related functions (such as termination criteria), and on the other hand preconditioners, both simple stationary iterations (such as Kaskade::JacobiPreconditioner) and multilevel preconditioners (such as Kaskade::AdditiveMultiGrid or Kaskade::MultiplicativeMultiGrid).

Function Documentation

requiresInitializedInput()

template<class Domain , class Range >

bool Kaskade::requiresInitializedInput

(

Dune::Preconditioner< Domain, Range > const *

p

)

Whether a preconditioner needs zero initialized result vector or not.

For a couple of known preconditioners, this function returns the actual information. For unknown preconditioners, it errs on the conservative side, i.e. reports true.

The function is moderately expensive, hence don't use it inside of loops.

Definition at line 46 of file linalg/apcg.hh.