cusp::relaxation::sor< ValueType, MemorySpace > Class Template Reference

Detailed description

template<typename ValueType, typename MemorySpace>
class cusp::relaxation::sor< ValueType, MemorySpace >

Represents a Successive Over-Relaxation relaxation scheme.

Template Parameters

ValueType	value_type of the array
MemorySpace	memory space of the array (cusp::host_memory or cusp::device_memory)

Overview

Computes vertex coloring and performs indexed Successive Over-Relaxation relaxation

Example

#include <cusp/array1d.h>
#include <cusp/csr_matrix.h>
#include <cusp/monitor.h>

#include <cusp/blas/blas.h>
#include <cusp/linear_operator.h>
#include <cusp/gallery/poisson.h>

// include cusp sor header file
#include <cusp/relaxation/sor.h>

int main()
{
   // Construct 5-pt Poisson example
   cusp::csr_matrix<int, float, cusp::device_memory> A;
   cusp::gallery::poisson5pt(A, 5, 5);

   // Initialize data
   cusp::array1d<float, cusp::device_memory> x(A.num_rows, 0);
   cusp::array1d<float, cusp::device_memory> b(A.num_rows, 1);

   // Allocate temporaries
   cusp::array1d<float, cusp::device_memory> r(A.num_rows);

   // Construct sor relaxation class
   cusp::relaxation::sor<float, cusp::device_memory> M(A);

   // Compute initial residual
   cusp::multiply(A, x, r);
   cusp::blas::axpy(b, r, float(-1));

   // Construct monitor with stopping criteria of 100 iterations or 1e-4 residual error
   cusp::monitor<float> monitor(b, 100, 1e-4, 0, true);

   // Iteratively solve system
   while (!monitor.finished(r))
   {
       M(A, b, x);
       cusp::multiply(A, x, r);
       cusp::blas::axpy(b, r, float(-1));
       ++monitor;
   }
 }

Definition at line 98 of file sor.h.

#include <sor.h>

Inheritance diagram for cusp::relaxation::sor< ValueType, MemorySpace >:

Public Methods
	sor (void)
template<typename MatrixType >
	sor (const MatrixType &A, const ValueType omega, sweep default_direction=SYMMETRIC)
template<typename MemorySpace2 >
	sor (const sor< ValueType, MemorySpace2 > &A)
template<typename MatrixType , typename VectorType1 , typename VectorType2 >
void	operator() (const MatrixType &A, const VectorType1 &b, VectorType2 &x)
template<typename MatrixType , typename VectorType1 , typename VectorType2 >
void	operator() (const MatrixType &A, const VectorType1 &b, VectorType2 &x, const ValueType omega, sweep direction)
Public Methods inherited from cusp::linear_operator<ValueType, MemorySpace >
	linear_operator (void)
	linear_operator (intnum_rows, intnum_cols)
	linear_operator (intnum_rows, intnum_cols, intnum_entries)

Constructor & Destructor Documentation

template<typename ValueType, typename MemorySpace>

cusp::relaxation::sor< ValueType, MemorySpace >::sor ( void  )

inline

This constructor creates an empty sor smoother.

Definition at line 110 of file sor.h.

template<typename ValueType, typename MemorySpace>

template<typename MatrixType >

cusp::relaxation::sor< ValueType, MemorySpace >::sor ( const MatrixType &  A, const ValueType  omega, sweep  default_direction = SYMMETRIC  )

inline

This constructor creates a sor smoother using a given matrix and sweeping strategy (FORWARD, BACKWARD, SYMMETRIC).

Template Parameters

MatrixType

Type of input matrix used to create this sor smoother.

Parameters

A	Input matrix used to create smoother.
omega	Damping factor used in SOR smoother.
default_direction	Sweep strategy used to perform Gauss-Seidel smoothing.

Definition at line 124 of file sor.h.

template<typename ValueType, typename MemorySpace>

template<typename MemorySpace2 >

cusp::relaxation::sor< ValueType, MemorySpace >::sor ( const sor< ValueType, MemorySpace2 > &  A )

inline

Copy constructor for sor smoother.

Template Parameters

MemorySpace2

Memory space of input sor smoother.

Parameters

A	Input sor smoother.

Definition at line 134 of file sor.h.

Member Function Documentation

template<typename ValueType, typename MemorySpace>

template<typename MatrixType , typename VectorType1 , typename VectorType2 >

void cusp::relaxation::sor< ValueType, MemorySpace >::operator()	(	const MatrixType &	A,
		const VectorType1 &	b,
		VectorType2 &	x
	)

Perform SOR relaxation using default omega damping factor specified during construction of this sor smoother

Template Parameters

MatrixType	Type of input matrix.
VectorType1	Type of input right-hand side vector.
VectorType2	Type of input approximate solution vector.

Parameters

A	matrix of the linear system
x	approximate solution of the linear system
b	right-hand side of the linear system

template<typename ValueType, typename MemorySpace>

template<typename MatrixType , typename VectorType1 , typename VectorType2 >

void cusp::relaxation::sor< ValueType, MemorySpace >::operator()	(	const MatrixType &	A,
		const VectorType1 &	b,
		VectorType2 &	x,
		const ValueType	omega,
		sweep	direction
	)

Perform SOR relaxation using specified omega damping factor.

Template Parameters

MatrixType	Type of input matrix.
VectorType1	Type of input right-hand side vector.
VectorType2	Type of input approximate solution vector.

Parameters

A	matrix of the linear system
x	approximate solution of the linear system
b	right-hand side of the linear system
omega	Damping factor used in SOR smoother.
direction	sweeping strategy for this gauss_seidel smoother (FORWARD, BACKWARD, SYMMETRIC).