cern.colt.matrix.tfloat.algo.solver.preconditioner

Class FloatSSOR

java.lang.Object

cern.colt.matrix.tfloat.algo.solver.preconditioner.FloatSSOR

All Implemented Interfaces:

FloatPreconditioner

public class FloatSSOR
extends java.lang.Object
implements FloatPreconditioner

SSOR preconditioner. Uses symmetrical sucessive overrelaxation as a preconditioner. Meant for symmetrical, positive definite matrices. For best performance, omega must be carefully chosen (between 0 and 2).

Constructor Summary

Constructors

Constructor and Description
FloatSSOR(int n) Constructor for SSOR.
FloatSSOR(int n, boolean reverse, float omegaF, float omegaR) Constructor for SSOR

Method Summary

Modifier and Type	Method and Description
FloatMatrix1D	apply(FloatMatrix1D b, FloatMatrix1D x) Solves the approximate problem with the given right hand side.
void	setMatrix(FloatMatrix2D A) Sets the operator matrix for the preconditioner.
void	setOmega(float omegaF, float omegaR) Sets the overrelaxation parameters
FloatMatrix1D	transApply(FloatMatrix1D b, FloatMatrix1D x) Solves the approximate transpose problem with the given right hand side.

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

FloatSSOR

public FloatSSOR(int n,
                 boolean reverse,
                 float omegaF,
                 float omegaR)

Constructor for SSOR

Parameters:

n - Problem size (number of rows)

reverse - True to perform a reverse sweep as well as the forward sweep. If false, this preconditioner becomes the SOR method instead

omegaF - Overrelaxation parameter for the forward sweep. Between 0 and 2.

omegaR - Overrelaxation parameter for the backwards sweep. Between 0 and 2.

FloatSSOR

public FloatSSOR(int n)

Constructor for SSOR. Uses omega=1 with a backwards sweep

Parameters:

n - Problem size (number of rows)

Method Detail

setOmega

public void setOmega(float omegaF,
                     float omegaR)

Sets the overrelaxation parameters

Parameters:

omegaF - Overrelaxation parameter for the forward sweep. Between 0 and 2.

omegaR - Overrelaxation parameter for the backwards sweep. Between 0 and 2.

setMatrix

public void setMatrix(FloatMatrix2D A)

Description copied from interface: FloatPreconditioner

Sets the operator matrix for the preconditioner. This method must be called before a preconditioner is used by an iterative solver

Specified by:

setMatrix in interface FloatPreconditioner

Parameters:

A - Matrix to setup the preconditioner for. Not modified

apply

public FloatMatrix1D apply(FloatMatrix1D b,
                           FloatMatrix1D x)

Description copied from interface: FloatPreconditioner

Solves the approximate problem with the given right hand side. Result is stored in given solution vector

Specified by:

apply in interface FloatPreconditioner

Parameters:

b - Right hand side of problem

x - Result is stored here

Returns:

x

transApply

public FloatMatrix1D transApply(FloatMatrix1D b,
                                FloatMatrix1D x)

Description copied from interface: FloatPreconditioner

Solves the approximate transpose problem with the given right hand side. Result is stored in given solution vector

Specified by:

transApply in interface FloatPreconditioner

Parameters:

b - Right hand side of problem

x - Result is stored here

Returns:

x