Template Numerical Library: TNL::Solvers::Linear::HypreBoomerAMG Class Reference

Wrapper for the BoomerAMG solver/preconditioner in Hypre. More...

#include <TNL/Solvers/Linear/Hypre.h>

Inheritance diagram for TNL::Solvers::Linear::HypreBoomerAMG:

Collaboration diagram for TNL::Solvers::Linear::HypreBoomerAMG:

[legend]

Public Member Functions
	HypreBoomerAMG (const Matrices::HypreParCSRMatrix &A)

HYPRE_Int	getNumIterations () const

void	setAdvectiveOptions (int restrict_type=0, int relax_order=1)
	Set parameters to use AIR AMG solver for advection-dominated problems.

void	setMaxIter (int max_iter)

void	setPrintLevel (int print_level)
	Set the print level: 0 = none, 1 = setup, 2 = solve, 3 = setup+solve.

void	setSystemsOptions (int dim, bool order_bynodes=false)
	More robust options for systems of PDEs.

void	setTol (double tol)

HYPRE_PtrToParSolverFcn	setupFcn () const override
	Hypre's internal setup function.

HYPRE_PtrToParSolverFcn	solveFcn () const override
	Hypre's internal solve function.

Public Member Functions inherited from TNL::Solvers::Linear::HypreSolver
	HypreSolver (const Matrices::HypreParCSRMatrix &A)

virtual	operator HYPRE_Solver () const
	Type-cast to HYPRE_Solver.

void	setErrorMode (ErrorMode err_mode) const
	Set the behavior for treating Hypre errors, see the ErrorMode enum. The default mode in the base class is ABORT_HYPRE_ERRORS.

virtual void	setMatrix (const Matrices::HypreParCSRMatrix &op, bool reuse_setup=false)
	Set the matrix of the linear system to be solved.

virtual void	setup (const Containers::HypreParVector &b, Containers::HypreParVector &x) const
	Setup the solver for solving the linear system Ax=b.

virtual void	solve (const Containers::HypreParVector &b, Containers::HypreParVector &x) const
	Solve the linear system Ax=b.

Protected Member Functions
void	postSolveHook () const override
	Hook function that is called at the end of solve.

Protected Member Functions inherited from TNL::Solvers::Linear::HypreSolver

Additional Inherited Members
Public Types inherited from TNL::Solvers::Linear::HypreSolver
enum	ErrorMode { IGNORE_HYPRE_ERRORS , WARN_HYPRE_ERRORS , ABORT_HYPRE_ERRORS }
	How to treat errors returned by Hypre function calls. More...

Protected Attributes inherited from TNL::Solvers::Linear::HypreSolver
const Matrices::HypreParCSRMatrix *	A = nullptr
	The linear system matrix.

ErrorMode	error_mode = ABORT_HYPRE_ERRORS
	How to treat Hypre errors.

bool	setup_called = false
	Indicates if Hypre's setup function was already called.

HYPRE_Solver	solver = nullptr
	Handle for the Hypre solver.

Detailed Description

Wrapper for the BoomerAMG solver/preconditioner in Hypre.

Note that the wrapper class sets its own default options that are different from Hypre. By default, the instance is set up for preconditioner use, i.e. with zero tolerance and one V cycle. At least these two parameters need to be changed when used as a solver.

Parameters can be set using native Hypre functions, e.g.

TNL::Solvers::Linear::HypreBoomerAMG solver;
HYPRE_BoomerAMGSetTol(solver, 1e-7);
HYPRE_BoomerAMGSetMaxIter(solver, 20);

See the Hypre Reference Manual for the available parameters.

Member Function Documentation

◆ postSolveHook()

void TNL::Solvers::Linear::HypreBoomerAMG::postSolveHook ( ) const

overrideprotectedvirtual

Hook function that is called at the end of solve.

Reimplemented from TNL::Solvers::Linear::HypreSolver.

◆ setAdvectiveOptions()

void TNL::Solvers::Linear::HypreBoomerAMG::setAdvectiveOptions	(	int	restrict_type = 0,
		int	relax_order = 1 )

Set parameters to use AIR AMG solver for advection-dominated problems.

See "Nonsymmetric Algebraic Multigrid Based on Local Approximate Ideal Restriction (AIR)," Manteuffel, Ruge, Southworth, SISC (2018), DOI:/10.1137/17M1144350.

Parameters

restrict_type	Defines which parallel restriction operator is used. There are the following options: 0: transpose of the interpolation operator 1: AIR-1 - Approximate Ideal Restriction (distance 1) 2: AIR-2 - Approximate Ideal Restriction (distance 2)
relax_order	Defines in which order the points are relaxed. There are the following options: 0: the points are relaxed in natural or lexicographic order on each processor 1: CF-relaxation is used

◆ setupFcn()

HYPRE_PtrToParSolverFcn TNL::Solvers::Linear::HypreBoomerAMG::setupFcn ( ) const

inlineoverridevirtual

Hypre's internal setup function.

Implements TNL::Solvers::Linear::HypreSolver.

◆ solveFcn()

HYPRE_PtrToParSolverFcn TNL::Solvers::Linear::HypreBoomerAMG::solveFcn ( ) const

inlineoverridevirtual

Hypre's internal solve function.

Implements TNL::Solvers::Linear::HypreSolver.

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

src/TNL/Solvers/Linear/Hypre.h
src/TNL/Solvers/Linear/Hypre.hpp

Public Member Functions