Parallel smoothers in hypre. More...

#include <hypre.hpp>

Inheritance diagram for mfem::HypreSmoother:

Collaboration diagram for mfem::HypreSmoother:

Public Types
enum	Type { Jacobi = 0, l1Jacobi = 1, l1GS = 2, l1GStr = 4, lumpedJacobi = 5, GS = 6, OPFS = 10, Chebyshev = 16, Taubin = 1001, FIR = 1002 }

Public Types inherited from mfem::Operator
enum	DiagonalPolicy { DIAG_ZERO, DIAG_ONE, DIAG_KEEP }
	Defines operator diagonal policy upon elimination of rows and/or columns. More...

enum	Type { ANY_TYPE, MFEM_SPARSEMAT, Hypre_ParCSR, PETSC_MATAIJ, PETSC_MATIS, PETSC_MATSHELL, PETSC_MATNEST, PETSC_MATHYPRE, PETSC_MATGENERIC, Complex_Operator, MFEM_ComplexSparseMat, Complex_Hypre_ParCSR, Complex_DenseMat, MFEM_Block_Matrix, MFEM_Block_Operator }
	Enumeration defining IDs for some classes derived from Operator. More...

Public Member Functions
	HypreSmoother ()

	HypreSmoother (const HypreParMatrix &A_, int type=default_type, int relax_times=1, double relax_weight=1.0, double omega=1.0, int poly_order=2, double poly_fraction=.3, int eig_est_cg_iter=10)

void	SetType (HypreSmoother::Type type, int relax_times=1)
	Set the relaxation type and number of sweeps. More...

void	SetSOROptions (double relax_weight, double omega)
	Set SOR-related parameters. More...

void	SetPolyOptions (int poly_order, double poly_fraction, int eig_est_cg_iter=10)
	Set parameters for polynomial smoothing. More...

void	SetTaubinOptions (double lambda, double mu, int iter)
	Set parameters for Taubin's lambda-mu method. More...

void	SetWindowByName (const char *window_name)
	Convenience function for setting canonical windowing parameters. More...

void	SetWindowParameters (double a, double b, double c)
	Set parameters for windowing function for FIR smoother. More...

void	SetFIRCoefficients (double max_eig)
	Compute window and Chebyshev coefficients for given polynomial order. More...

void	SetPositiveDiagonal (bool pos=true)
	After computing l1-norms, replace them with their absolute values. More...

void	SetOperatorSymmetry (bool is_sym)

virtual void	SetOperator (const Operator &op)

virtual void	Mult (const HypreParVector &b, HypreParVector &x) const
	Relax the linear system Ax=b. More...

virtual void	Mult (const Vector &b, Vector &x) const
	Operator application: `y=A(x)`. More...

virtual void	MultTranspose (const Vector &b, Vector &x) const
	Apply transpose of the smoother to relax the linear system Ax=b. More...

virtual	~HypreSmoother ()

virtual void	Mult (const Vector &x, Vector &y) const=0
	Operator application: `y=A(x)`. More...

Public Member Functions inherited from mfem::Solver
	Solver (int s=0, bool iter_mode=false)
	Initialize a square Solver with size s. More...

	Solver (int h, int w, bool iter_mode=false)
	Initialize a Solver with height h and width w. More...

Public Member Functions inherited from mfem::Operator
void	InitTVectors (const Operator Po, const Operator Ri, const Operator *Pi, Vector &x, Vector &b, Vector &X, Vector &B) const
	Initializes memory for true vectors of linear system. More...

	Operator (int s=0)
	Construct a square Operator with given size s (default 0). More...

	Operator (int h, int w)
	Construct an Operator with the given height (output size) and width (input size). More...

int	Height () const
	Get the height (size of output) of the Operator. Synonym with NumRows(). More...

int	NumRows () const
	Get the number of rows (size of output) of the Operator. Synonym with Height(). More...

int	Width () const
	Get the width (size of input) of the Operator. Synonym with NumCols(). More...

int	NumCols () const
	Get the number of columns (size of input) of the Operator. Synonym with Width(). More...

virtual MemoryClass	GetMemoryClass () const
	Return the MemoryClass preferred by the Operator. More...

virtual void	AddMult (const Vector &x, Vector &y, const double a=1.0) const
	Operator application: `y+=A(x)` (default) or `y+=a*A(x)`. More...

virtual void	AddMultTranspose (const Vector &x, Vector &y, const double a=1.0) const
	Operator transpose application: `y+=A^t(x)` (default) or `y+=a*A^t(x)`. More...

virtual void	ArrayMult (const Array< const Vector > &X, Array< Vector > &Y) const
	Operator application on a matrix: `Y=A(X)`. More...

virtual void	ArrayMultTranspose (const Array< const Vector > &X, Array< Vector > &Y) const
	Action of the transpose operator on a matrix: `Y=A^t(X)`. More...

virtual void	ArrayAddMult (const Array< const Vector > &X, Array< Vector > &Y, const double a=1.0) const
	Operator application on a matrix: `Y+=A(X)` (default) or `Y+=a*A(X)`. More...

virtual void	ArrayAddMultTranspose (const Array< const Vector > &X, Array< Vector > &Y, const double a=1.0) const
	Operator transpose application on a matrix: `Y+=A^t(X)` (default) or `Y+=a*A^t(X)`. More...

virtual Operator &	GetGradient (const Vector &x) const
	Evaluate the gradient operator at the point x. The default behavior in class Operator is to generate an error. More...

virtual void	AssembleDiagonal (Vector &diag) const
	Computes the diagonal entries into diag. Typically, this operation only makes sense for linear Operators. In some cases, only an approximation of the diagonal is computed. More...

virtual const Operator *	GetProlongation () const
	Prolongation operator from linear algebra (linear system) vectors, to input vectors for the operator. `NULL` means identity. More...

virtual const Operator *	GetRestriction () const
	Restriction operator from input vectors for the operator to linear algebra (linear system) vectors. `NULL` means identity. More...

virtual const Operator *	GetOutputProlongation () const
	Prolongation operator from linear algebra (linear system) vectors, to output vectors for the operator. `NULL` means identity. More...

virtual const Operator *	GetOutputRestrictionTranspose () const
	Transpose of GetOutputRestriction, directly available in this form to facilitate matrix-free RAP-type operators. More...

virtual const Operator *	GetOutputRestriction () const
	Restriction operator from output vectors for the operator to linear algebra (linear system) vectors. `NULL` means identity. More...

void	FormLinearSystem (const Array< int > &ess_tdof_list, Vector &x, Vector &b, Operator *&A, Vector &X, Vector &B, int copy_interior=0)
	Form a constrained linear system using a matrix-free approach. More...

void	FormRectangularLinearSystem (const Array< int > &trial_tdof_list, const Array< int > &test_tdof_list, Vector &x, Vector &b, Operator *&A, Vector &X, Vector &B)
	Form a column-constrained linear system using a matrix-free approach. More...

virtual void	RecoverFEMSolution (const Vector &X, const Vector &b, Vector &x)
	Reconstruct a solution vector x (e.g. a GridFunction) from the solution X of a constrained linear system obtained from Operator::FormLinearSystem() or Operator::FormRectangularLinearSystem(). More...

void	FormSystemOperator (const Array< int > &ess_tdof_list, Operator *&A)
	Return in A a parallel (on truedofs) version of this square operator. More...

void	FormRectangularSystemOperator (const Array< int > &trial_tdof_list, const Array< int > &test_tdof_list, Operator *&A)
	Return in A a parallel (on truedofs) version of this rectangular operator (including constraints). More...

void	FormDiscreteOperator (Operator *&A)
	Return in A a parallel (on truedofs) version of this rectangular operator. More...

void	PrintMatlab (std::ostream &out, int n, int m=0) const
	Prints operator with input size n and output size m in Matlab format. More...

virtual void	PrintMatlab (std::ostream &out) const
	Prints operator in Matlab format. More...

virtual	~Operator ()
	Virtual destructor. More...

Type	GetType () const
	Return the type ID of the Operator class. More...

Static Public Attributes
static constexpr Type	default_type = l1GS

Protected Attributes
HypreParMatrix *	A
	The linear system matrix. More...

HypreParVector *	B
	Right-hand side and solution vectors. More...

HypreParVector *	X

Memory< double >	auxB
	Auxiliary buffers for the case when the input or output arrays in methods like Mult(const Vector &, Vector &) need to be deep copied in order to be used by hypre. More...

Memory< double >	auxX

HypreParVector *	V
	Temporary vectors. More...

HypreParVector *	Z

HypreParVector *	X0
	FIR Filter Temporary Vectors. More...

HypreParVector *	X1

int	type

int	relax_times
	Number of relaxation sweeps. More...

double	relax_weight
	Damping coefficient (usually <= 1) More...

double	omega
	SOR parameter (usually in (0,2)) More...

int	poly_order
	Order of the smoothing polynomial. More...

double	poly_fraction
	Fraction of spectrum to smooth for polynomial relaxation. More...

int	poly_scale
	Apply the polynomial smoother to A or D^{-1/2} A D^{-1/2}. More...

double	lambda
	Taubin's lambda-mu method parameters. More...

double	mu

int	taubin_iter

double *	l1_norms
	l1 norms of the rows of A More...

bool	pos_l1_norms
	If set, take absolute values of the computed l1_norms. More...

int	eig_est_cg_iter
	Number of CG iterations to determine eigenvalue estimates. More...

double	max_eig_est
	Maximal eigenvalue estimate for polynomial smoothing. More...

double	min_eig_est
	Minimal eigenvalue estimate for polynomial smoothing. More...

double	window_params [3]
	Parameters for windowing function of FIR filter. More...

double *	fir_coeffs
	Combined coefficients for windowing and Chebyshev polynomials. More...

bool	A_is_symmetric
	A flag that indicates whether the linear system matrix A is symmetric. More...

Protected Attributes inherited from mfem::Operator
int	height
	Dimension of the output / number of rows in the matrix. More...

int	width
	Dimension of the input / number of columns in the matrix. More...

Additional Inherited Members
Public Attributes inherited from mfem::Solver
bool	iterative_mode
	If true, use the second argument of Mult() as an initial guess. More...

Protected Member Functions inherited from mfem::Operator
void	FormConstrainedSystemOperator (const Array< int > &ess_tdof_list, ConstrainedOperator *&Aout)
	see FormSystemOperator() More...

void	FormRectangularConstrainedSystemOperator (const Array< int > &trial_tdof_list, const Array< int > &test_tdof_list, RectangularConstrainedOperator *&Aout)
	see FormRectangularSystemOperator() More...

Operator *	SetupRAP (const Operator Pi, const Operator Po)
	Returns RAP Operator of this, using input/output Prolongation matrices Pi corresponds to "P", Po corresponds to "Rt". More...

Detailed Description

Parallel smoothers in hypre.

Definition at line 971 of file hypre.hpp.

Member Enumeration Documentation

◆ Type

enum mfem::HypreSmoother::Type

Hypre smoother types: 0 = Jacobi 1 = l1-scaled Jacobi 2 = l1-scaled block Gauss-Seidel/SSOR 4 = truncated l1-scaled block Gauss-Seidel/SSOR 5 = lumped Jacobi 6 = Gauss-Seidel 10 = On-processor forward solve for matrix w/ triangular structure 16 = Chebyshev 1001 = Taubin polynomial smoother 1002 = FIR polynomial smoother.

Enumerator
Jacobi
l1Jacobi
l1GS
l1GStr
lumpedJacobi
GS
OPFS
Chebyshev
Taubin
FIR

Definition at line 1039 of file hypre.hpp.

Constructor & Destructor Documentation

◆ HypreSmoother() [1/2]

mfem::HypreSmoother::HypreSmoother ( )

Definition at line 3383 of file hypre.cpp.

◆ HypreSmoother() [2/2]

mfem::HypreSmoother::HypreSmoother	(	const HypreParMatrix &	A_,
		int	type = `default_type`,
		int	relax_times = `1`,
		double	relax_weight = `1.0`,
		double	omega = `1.0`,
		int	poly_order = `2`,
		double	poly_fraction = `.3`,
		int	eig_est_cg_iter = `10`
	)

Definition at line 3405 of file hypre.cpp.

◆ ~HypreSmoother()

mfem::HypreSmoother::~HypreSmoother ( )

virtual

Definition at line 3782 of file hypre.cpp.

Member Function Documentation

◆ Mult() [1/3]

void mfem::HypreSmoother::Mult	(	const HypreParVector &	b,
		HypreParVector &	x
	)		const

virtual

Relax the linear system Ax=b.

Definition at line 3631 of file hypre.cpp.

◆ Mult() [2/3]

void mfem::HypreSmoother::Mult	(	const Vector &	x,
		Vector &	y
	)		const

virtual

Operator application: y=A(x).

Implements mfem::Operator.

Definition at line 3711 of file hypre.cpp.

◆ Mult() [3/3]

virtual void mfem::Operator::Mult

Operator application: y=A(x).

◆ MultTranspose()

void mfem::HypreSmoother::MultTranspose	(	const Vector &	b,
		Vector &	x
	)		const

virtual

Apply transpose of the smoother to relax the linear system Ax=b.

Reimplemented from mfem::Operator.

Definition at line 3772 of file hypre.cpp.

◆ SetFIRCoefficients()

void mfem::HypreSmoother::SetFIRCoefficients ( double max_eig )

Compute window and Chebyshev coefficients for given polynomial order.

Definition at line 3591 of file hypre.cpp.

◆ SetOperator()

void mfem::HypreSmoother::SetOperator ( const Operator & op )

virtual

Set/update the associated operator. Must be called after setting the HypreSmoother type and options.

Implements mfem::Solver.

Definition at line 3479 of file hypre.cpp.

◆ SetOperatorSymmetry()

void mfem::HypreSmoother::SetOperatorSymmetry ( bool is_sym )

inline

Explicitly indicate whether the linear system matrix A is symmetric. If A is symmetric, the smoother will also be symmetric. In this case, calling MultTranspose will be redirected to Mult. (This is also done if the smoother is diagonal.) By default, A is assumed to be nonsymmetric.

Definition at line 1083 of file hypre.hpp.

◆ SetPolyOptions()

void mfem::HypreSmoother::SetPolyOptions	(	int	poly_order,
		double	poly_fraction,
		int	eig_est_cg_iter = `10`
	)

Set parameters for polynomial smoothing.

By default, 10 iterations of CG are used to estimate the eigenvalues. Setting eig_est_cg_iter = 0 uses hypre's hypre_ParCSRMaxEigEstimate() instead.

Definition at line 3441 of file hypre.cpp.

◆ SetPositiveDiagonal()

void mfem::HypreSmoother::SetPositiveDiagonal ( bool pos = true )

inline

After computing l1-norms, replace them with their absolute values.

By default, the l1-norms take their sign from the corresponding diagonal entries in the associated matrix.

Definition at line 1077 of file hypre.hpp.

◆ SetSOROptions()

void mfem::HypreSmoother::SetSOROptions	(	double	relax_weight,
		double	omega
	)

Set SOR-related parameters.

Definition at line 3435 of file hypre.cpp.

◆ SetTaubinOptions()

void mfem::HypreSmoother::SetTaubinOptions	(	double	lambda,
		double	mu,
		int	iter
	)

Set parameters for Taubin's lambda-mu method.

Definition at line 3449 of file hypre.cpp.

◆ SetType()

void mfem::HypreSmoother::SetType	(	HypreSmoother::Type	type,
		int	relax_times = `1`
	)

Set the relaxation type and number of sweeps.

Definition at line 3429 of file hypre.cpp.

◆ SetWindowByName()

void mfem::HypreSmoother::SetWindowByName ( const char * window_name )

Convenience function for setting canonical windowing parameters.

Definition at line 3457 of file hypre.cpp.

◆ SetWindowParameters()

void mfem::HypreSmoother::SetWindowParameters	(	double	a,
		double	b,
		double	c
	)

Set parameters for windowing function for FIR smoother.

Definition at line 3472 of file hypre.cpp.

Member Data Documentation

◆ A

HypreParMatrix* mfem::HypreSmoother::A

protected

The linear system matrix.

Definition at line 975 of file hypre.hpp.

◆ A_is_symmetric

bool mfem::HypreSmoother::A_is_symmetric

protected

A flag that indicates whether the linear system matrix A is symmetric.

Definition at line 1025 of file hypre.hpp.

◆ auxB

Memory<double> mfem::HypreSmoother::auxB

mutableprotected

Auxiliary buffers for the case when the input or output arrays in methods like Mult(const Vector &, Vector &) need to be deep copied in order to be used by hypre.

Definition at line 981 of file hypre.hpp.

◆ auxX

Memory<double> mfem::HypreSmoother::auxX

mutableprotected

Definition at line 981 of file hypre.hpp.

◆ B

HypreParVector* mfem::HypreSmoother::B

mutableprotected

Right-hand side and solution vectors.

Definition at line 977 of file hypre.hpp.

◆ default_type

static constexpr Type mfem::HypreSmoother::default_type = l1GS

static

Definition at line 1043 of file hypre.hpp.

◆ eig_est_cg_iter

int mfem::HypreSmoother::eig_est_cg_iter

protected

Number of CG iterations to determine eigenvalue estimates.

Definition at line 1013 of file hypre.hpp.

◆ fir_coeffs

double* mfem::HypreSmoother::fir_coeffs

protected

Combined coefficients for windowing and Chebyshev polynomials.

Definition at line 1022 of file hypre.hpp.

◆ l1_norms

double* mfem::HypreSmoother::l1_norms

protected

l1 norms of the rows of A

Definition at line 1009 of file hypre.hpp.

◆ lambda

double mfem::HypreSmoother::lambda

protected

Taubin's lambda-mu method parameters.

Definition at line 1004 of file hypre.hpp.

◆ max_eig_est

double mfem::HypreSmoother::max_eig_est

protected

Maximal eigenvalue estimate for polynomial smoothing.

Definition at line 1015 of file hypre.hpp.

◆ min_eig_est

double mfem::HypreSmoother::min_eig_est

protected

Minimal eigenvalue estimate for polynomial smoothing.

Definition at line 1017 of file hypre.hpp.

◆ mu

double mfem::HypreSmoother::mu

protected

Definition at line 1005 of file hypre.hpp.

◆ omega

double mfem::HypreSmoother::omega

protected

SOR parameter (usually in (0,2))

Definition at line 995 of file hypre.hpp.

◆ poly_fraction

double mfem::HypreSmoother::poly_fraction

protected

Fraction of spectrum to smooth for polynomial relaxation.

Definition at line 999 of file hypre.hpp.

◆ poly_order

int mfem::HypreSmoother::poly_order

protected

Order of the smoothing polynomial.

Definition at line 997 of file hypre.hpp.

◆ poly_scale

int mfem::HypreSmoother::poly_scale

protected

Apply the polynomial smoother to A or D^{-1/2} A D^{-1/2}.

Definition at line 1001 of file hypre.hpp.

◆ pos_l1_norms

bool mfem::HypreSmoother::pos_l1_norms

protected

If set, take absolute values of the computed l1_norms.

Definition at line 1011 of file hypre.hpp.

◆ relax_times

int mfem::HypreSmoother::relax_times

protected

Number of relaxation sweeps.

Definition at line 991 of file hypre.hpp.

◆ relax_weight

double mfem::HypreSmoother::relax_weight

protected

Damping coefficient (usually <= 1)

Definition at line 993 of file hypre.hpp.

◆ taubin_iter

int mfem::HypreSmoother::taubin_iter

protected

Definition at line 1006 of file hypre.hpp.

◆ type

int mfem::HypreSmoother::type

protected

Smoother type from hypre_ParCSRRelax() in ams.c plus extensions, see the enumeration Type below.

Definition at line 989 of file hypre.hpp.

◆ V

HypreParVector* mfem::HypreSmoother::V

mutableprotected

Temporary vectors.

Definition at line 983 of file hypre.hpp.

◆ window_params

double mfem::HypreSmoother::window_params[3]

protected

Parameters for windowing function of FIR filter.

Definition at line 1019 of file hypre.hpp.

◆ X

HypreParVector * mfem::HypreSmoother::X

mutableprotected

Definition at line 977 of file hypre.hpp.

◆ X0

HypreParVector* mfem::HypreSmoother::X0

mutableprotected

FIR Filter Temporary Vectors.

Definition at line 985 of file hypre.hpp.

◆ X1

HypreParVector * mfem::HypreSmoother::X1

mutableprotected

Definition at line 985 of file hypre.hpp.

◆ Z

HypreParVector * mfem::HypreSmoother::Z

mutableprotected

Definition at line 983 of file hypre.hpp.

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

linalg/hypre.hpp
linalg/hypre.cpp

Public Types

Public Member Functions

Static Public Attributes

Protected Attributes

Additional Inherited Members

Detailed Description

Member Enumeration Documentation

◆ Type

Constructor & Destructor Documentation

◆ HypreSmoother() [1/2]

◆ HypreSmoother() [2/2]

◆ ~HypreSmoother()

Member Function Documentation

◆ Mult() [1/3]

◆ Mult() [2/3]

◆ Mult() [3/3]

◆ MultTranspose()

◆ SetFIRCoefficients()

◆ SetOperator()

◆ SetOperatorSymmetry()

◆ SetPolyOptions()

◆ SetPositiveDiagonal()

◆ SetSOROptions()

◆ SetTaubinOptions()

◆ SetType()

◆ SetWindowByName()

◆ SetWindowParameters()

Member Data Documentation

◆ A

◆ A_is_symmetric

◆ auxB

◆ auxX

◆ B

◆ default_type

◆ eig_est_cg_iter

◆ fir_coeffs

◆ l1_norms

◆ lambda

◆ max_eig_est

◆ min_eig_est

◆ mu

◆ omega

◆ poly_fraction

◆ poly_order

◆ poly_scale

◆ pos_l1_norms

◆ relax_times

◆ relax_weight

◆ taubin_iter

◆ type

◆ V

◆ window_params

◆ X

◆ X0

◆ X1

◆ Z