Namespaces
namespace	detail

Classes
class	BlockDiagonalPreconditioner2x2
	Block-diagonal preconditioner for 2x2 block operators. More...

class	BlockTriangularPreconditioner2x2
	Static assertion: BlockDiagonalPreconditioner2x2 satisfies SolverLike concept. More...

class	Chebyshev
	Chebyshev accelerated Jacobi iterative solver for linear systems. More...

class	FGMRES
	Flexible GMRES (FGMRES) iterative solver for nonsymmetric linear systems. More...

struct	FGMRESOptions

class	GCAElementsCollector

class	IdentitySolver
	"Identity solver" for linear systems. More...

class	IterativeSolverParameters

class	IterativeSolverStatistics

class	Jacobi
	Jacobi iterative solver for linear systems. More...

class	LocalMatrixStorage

class	Multigrid
	Multigrid solver for linear systems. More...

class	PBiCGStab
	BiCGStab(l) iterative solver for general (possibly unsymmetric) linear systems. More...

class	PCG
	Preconditioned Conjugate Gradient (PCG) iterative solver for symmetric positive definite linear systems. More...

class	PMINRES
	Preconditioned MINRES (PMINRES) iterative solver for symmetric indefinite linear systems. More...

class	Richardson
	Richardson iterative solver for linear systems. Satisfies the SolverLike concept (see solver.hpp). Implements the update rule: More...

struct	SelectiveStorageKey

class	TwoGridGCA
	: Galerkin coarse approximation (GCA). TwoGridGCA takes a coarser and a finer operator. Each thread assembles a coarse-grid gca matrix in the coarser operator on a single hex. To do this, it loops the finer hexes of the coarse hex and its respective wedges. It computes the interpolation matrix P mapping from coarse wedge to the current fine wedge, computes the triple-product P^TAP with the fine-operator local matrix A and adds the resulting gca matrix up for all fine wedges comprising the coarse wedge. Finally, it stores the result in the wedge-wise matrix storage of the coarse operator. More...

Concepts
concept	SolverLike
	Concept for types that behave like linear solvers. Requires exposing OperatorType and a solve_impl method. See OperatorLike in operator.hpp for operator requirements.

Typedefs
template<SolverLike Solver>
using	SolutionOf = SrcOf< typename Solver::OperatorType >
	Alias for the solution vector type of a solver.

template<SolverLike Solver>
using	RHSOf = DstOf< typename Solver::OperatorType >
	Alias for the right-hand side vector type of a solver.

Enumerations
enum class	InterpolationMode { Constant , Linear }
	Modes for choosing interpolation weights. More...

Functions
int	map_to_coarse_element (const int fine_cell, const int level_range)

template<terra::linalg::OperatorLike OperatorT>
double	power_iteration (OperatorT &op, SrcOf< OperatorT > &tmpIt, SrcOf< OperatorT > &tmpAux, const int iterations)
	Power iteration to estimate the largest eigenvalue of a row-normalized operator D^{-1}A, where D^{-1} is the inverted diagonal of A.

template<SolverLike Solver, OperatorLike Operator, VectorLike SolutionVector, VectorLike RHSVector>
void	solve (Solver &solver, Operator &A, SolutionVector &x, const RHSVector &b)
	Solve a linear system using the given solver and operator. Calls the solver's solve_impl method.

Typedef Documentation

◆ RHSOf

template<SolverLike Solver>

using terra::linalg::solvers::RHSOf = typedef DstOf< typename Solver::OperatorType >

Alias for the right-hand side vector type of a solver.

◆ SolutionOf

template<SolverLike Solver>

using terra::linalg::solvers::SolutionOf = typedef SrcOf< typename Solver::OperatorType >

Alias for the solution vector type of a solver.

Enumeration Type Documentation

◆ InterpolationMode

enum class terra::linalg::solvers::InterpolationMode

strong

Modes for choosing interpolation weights.

Enumerator
Constant
Linear

Function Documentation

◆ map_to_coarse_element()

int terra::linalg::solvers::map_to_coarse_element	(	const int	fine_cell,
		const int	level_range
	)

◆ power_iteration()

template<terra::linalg::OperatorLike OperatorT>

double terra::linalg::solvers::power_iteration	(	OperatorT &	op,
		SrcOf< OperatorT > &	tmpIt,
		SrcOf< OperatorT > &	tmpAux,
		const int	iterations
	)

Power iteration to estimate the largest eigenvalue of a row-normalized operator D^{-1}A, where D^{-1} is the inverted diagonal of A.

Template Parameters

OperatorT Operator type (must satisfy OperatorLike).

◆ solve()

template<SolverLike Solver, OperatorLike Operator, VectorLike SolutionVector, VectorLike RHSVector>

void terra::linalg::solvers::solve	(	Solver &	solver,
		Operator &	A,
		SolutionVector &	x,
		const RHSVector &	b
	)

Solve a linear system using the given solver and operator. Calls the solver's solve_impl method.

Parameters

solver	The solver instance.
A	The operator (matrix).
x	Solution vector (output).
b	Right-hand side vector (input).

Namespaces

Classes

Concepts

Typedefs

Enumerations

Functions