amrex::JacobiSmoother< T > Class Template Reference

Jacobi smoother for AlgVector/SpMatrix linear systems. More...

#include <AMReX_Smoother_MV.H>

Public Member Functions
	JacobiSmoother (SpMatrix< T > const *a_A, bool a_l1=false)
	Construct a smoother that operates on a_A.
int	setNumIters (int a_niters)
	Update how many Jacobi sweeps to perform per apply.
void	setWeight (T w)
	Set the relaxation weight. Call this after setL1(), which resets it.
void	setL1 (bool b)
void	operator() (AlgVector< T > &xvec, AlgVector< T > const &bvec, bool with_initial_guess=false)
	Apply the smoother to solve approximately for xvec.
AlgVector< T > const &	scaling ()
	Diagonal or l1 scaling vector. The first call is collective.

Detailed Description

template<typename T>
class amrex::JacobiSmoother< T >

Jacobi smoother for AlgVector/SpMatrix linear systems.

Performs x += w * (b - A x) / d sweeps, where d is the diagonal of A (weighted Jacobi, weight 2/3 by default) or, with setL1(true), the l1 row norm sign(a_ii) * sum_j |a_ij| (l1-Jacobi, weight 1 by default), which converges for any symmetric positive definite matrix [Baker, Falgout, Kolev, Yang, SIAM J. Sci. Comput. 33 (2011), 2864-2887].

Constructor & Destructor Documentation

JacobiSmoother()

template<typename T >

amrex::JacobiSmoother< T >::JacobiSmoother ( SpMatrix< T > const *  a_A, bool  a_l1 = false  )

inlineexplicit

Construct a smoother that operates on a_A.

Parameters

a_A	Sparse matrix supplying diagonal entries and SpMV.
a_l1	Use l1-Jacobi scaling.

Member Function Documentation

operator()()

template<typename T >

void amrex::JacobiSmoother< T >::operator() ( AlgVector< T > &  xvec, AlgVector< T > const &  bvec, bool  with_initial_guess = false  )

inline

Apply the smoother to solve approximately for xvec.

Parameters

xvec	Solution vector updated in place.
bvec	RHS vector.
with_initial_guess	True if xvec already holds an initial guess.

scaling()

template<typename T >

AlgVector< T > const & amrex::JacobiSmoother< T >::scaling ( )

inline

Diagonal or l1 scaling vector. The first call is collective.

setL1()

template<typename T >

void amrex::JacobiSmoother< T >::setL1 ( bool  b )

inline

Switch between l1-Jacobi and weighted Jacobi, resetting the weight to the default of the new mode.

setNumIters()

template<typename T >

int amrex::JacobiSmoother< T >::setNumIters ( int  a_niters )

inline

Update how many Jacobi sweeps to perform per apply.

Parameters

a_niters

Desired number of smoothing iterations (4 by default).

Returns

Previous iteration count.

setWeight()

template<typename T >

void amrex::JacobiSmoother< T >::setWeight ( T  w )

inline

Set the relaxation weight. Call this after setL1(), which resets it.

---

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

• /home/runner/work/amrex/amrex/Src/LinearSolvers/AMReX_Smoother_MV.H