OpenCLTemplate.LinearAlgebra.SparseLinalg Class Reference

OpenCL linear algebra functions. More...

Classes
class	CLImgSparseMatrix
	Builds a new OpenCL Image2D containig a sparse matrix. More...
class	CLImgVector
	Builds a new OpenCL Image2D containing vector data. More...
class	CLLinalgSrc
	OpenCL dot product source.
Public Member Functions
	SparseLinalg ()
	Constructor.
float	DotProduct (CLImgVector v1, CLImgVector v2)
	Computes dot product of 2 vectors using their OpenCL images. Assumes data has been inserted to VectorData and WriteToDevice() has been called.
float	DotProductNoCL (CLImgVector v1, CLImgVector v2)
	Computes dot product of 2 vectors without OpenCL.
double	ExactDotProductNoCL (CLImgVector v1, CLImgVector v2)
	Computes dot product of 2 vectors without OpenCL, in double precision.
void	Multiply (CLImgSparseMatrix M, CLImgVector x, CLImgVector y)
	Computes M*x and stores the result in y. Does not automatically read result from device memory.
double[]	ExactMultiply (CLImgSparseMatrix M, CLImgVector x)
	Computes exact double-precision product of sparse matrix M and vector x.
float[]	MultiplyNoCL (CLImgSparseMatrix M, CLImgVector x)
	Computes product of sparse matrix M and vector x.
float[]	LinSolveNoCL (CLImgSparseMatrix M, CLImgVector b, float tol)
	Solves linear system Mx = b using conjugate gradient method.
float[]	LinSolveCL (CLImgSparseMatrix M, CLImgVector b, float tol)
	Solves linear system Mx = b using conjugate gradient method. Writes variables to Device memory. Improves solution if accuracy is low.
void	LinSolveCLStep (CLImgSparseMatrix M, CLImgVector b, float tol, ref CLImgVector x)
	Solves linear system Mx = b using conjugate gradient method. Doesn't try to improve the solution obtained.
	SparseLinalg ()
	Constructor.
float	DotProduct (CLImgVector v1, CLImgVector v2)
	Computes dot product of 2 vectors using their OpenCL images. Assumes data has been inserted to VectorData and WriteToDevice() has been called.
float	DotProductNoCL (CLImgVector v1, CLImgVector v2)
	Computes dot product of 2 vectors without OpenCL.
double	ExactDotProductNoCL (CLImgVector v1, CLImgVector v2)
	Computes dot product of 2 vectors without OpenCL, in double precision.
void	Multiply (CLImgSparseMatrix M, CLImgVector x, CLImgVector y)
	Computes M*x and stores the result in y. Does not automatically read result from device memory.
double[]	ExactMultiply (CLImgSparseMatrix M, CLImgVector x)
	Computes exact double-precision product of sparse matrix M and vector x.
float[]	MultiplyNoCL (CLImgSparseMatrix M, CLImgVector x)
	Computes product of sparse matrix M and vector x.
float[]	LinSolveNoCL (CLImgSparseMatrix M, CLImgVector b, float tol)
	Solves linear system Mx = b using conjugate gradient method.
float[]	LinSolveCL (CLImgSparseMatrix M, CLImgVector b, float tol)
	Solves linear system Mx = b using conjugate gradient method. Writes variables to Device memory. Improves solution if accuracy is low.
void	LinSolveCLStep (CLImgSparseMatrix M, CLImgVector b, float tol)
	Solves linear system Mx = b using conjugate gradient method. Doesn't try to improve the solution obtained.
	SparseLinalg ()
	Constructor.
float	DotProduct (CLImgVector v1, CLImgVector v2)
	Computes dot product of 2 vectors using their OpenCL images. Assumes data has been inserted to VectorData and WriteToDevice() has been called.
float	DotProductNoCL (CLImgVector v1, CLImgVector v2)
	Computes dot product of 2 vectors without OpenCL.
double	ExactDotProductNoCL (CLImgVector v1, CLImgVector v2)
	Computes dot product of 2 vectors without OpenCL, in double precision.
void	Multiply (CLImgSparseMatrix M, CLImgVector x, CLImgVector y)
	Computes M*x and stores the result in y. Does not automatically read result from device memory.
double[]	ExactMultiply (CLImgSparseMatrix M, CLImgVector x)
	Computes exact double-precision product of sparse matrix M and vector x.
float[]	MultiplyNoCL (CLImgSparseMatrix M, CLImgVector x)
	Computes product of sparse matrix M and vector x.
float[]	LinSolveNoCL (CLImgSparseMatrix M, CLImgVector b, float tol)
	Solves linear system Mx = b using conjugate gradient method.
float[]	LinSolveCL (CLImgSparseMatrix M, CLImgVector b, float tol)
	Solves linear system Mx = b using conjugate gradient method. Writes variables to Device memory. Improves solution if accuracy is low.
void	LinSolveCLStep (CLImgSparseMatrix M, CLImgVector b, float tol, ref CLImgVector x)
	Solves linear system Mx = b using conjugate gradient method. Doesn't try to improve the solution obtained.
Public Attributes
const int	IMGWIDTH = 4096
	Width to be used in float4's.
const int	GLOBALWORKSIZE = 2048
	Total number of workers.
CLImgVector	x
	Initial guess/answer holder.
Static Public Attributes
static int	MAXITER = 8
	Maximum number of iterations.

Detailed Description

OpenCL linear algebra functions.

Constructor & Destructor Documentation

OpenCLTemplate.LinearAlgebra.SparseLinalg.SparseLinalg ( )

Constructor.

OpenCLTemplate.LinearAlgebra.SparseLinalg.SparseLinalg ( )

Constructor.

OpenCLTemplate.LinearAlgebra.SparseLinalg.SparseLinalg ( )

Constructor.

Member Function Documentation

float OpenCLTemplate.LinearAlgebra.SparseLinalg.DotProduct	(	CLImgVector	v1,
		CLImgVector	v2
	)

Computes dot product of 2 vectors using their OpenCL images. Assumes data has been inserted to VectorData and WriteToDevice() has been called.

Parameters:

	v1	First vector
	v2	Second vector

float OpenCLTemplate.LinearAlgebra.SparseLinalg.DotProduct	(	CLImgVector	v1,
		CLImgVector	v2
	)

Computes dot product of 2 vectors using their OpenCL images. Assumes data has been inserted to VectorData and WriteToDevice() has been called.

Parameters:

	v1	First vector
	v2	Second vector

float OpenCLTemplate.LinearAlgebra.SparseLinalg.DotProduct	(	CLImgVector	v1,
		CLImgVector	v2
	)

Computes dot product of 2 vectors using their OpenCL images. Assumes data has been inserted to VectorData and WriteToDevice() has been called.

Parameters:

	v1	First vector
	v2	Second vector

float OpenCLTemplate.LinearAlgebra.SparseLinalg.DotProductNoCL	(	CLImgVector	v1,
		CLImgVector	v2
	)

Computes dot product of 2 vectors without OpenCL.

float OpenCLTemplate.LinearAlgebra.SparseLinalg.DotProductNoCL	(	CLImgVector	v1,
		CLImgVector	v2
	)

Computes dot product of 2 vectors without OpenCL.

float OpenCLTemplate.LinearAlgebra.SparseLinalg.DotProductNoCL	(	CLImgVector	v1,
		CLImgVector	v2
	)

Computes dot product of 2 vectors without OpenCL.

double OpenCLTemplate.LinearAlgebra.SparseLinalg.ExactDotProductNoCL	(	CLImgVector	v1,
		CLImgVector	v2
	)

Computes dot product of 2 vectors without OpenCL, in double precision.

double OpenCLTemplate.LinearAlgebra.SparseLinalg.ExactDotProductNoCL	(	CLImgVector	v1,
		CLImgVector	v2
	)

Computes dot product of 2 vectors without OpenCL, in double precision.

double OpenCLTemplate.LinearAlgebra.SparseLinalg.ExactDotProductNoCL	(	CLImgVector	v1,
		CLImgVector	v2
	)

Computes dot product of 2 vectors without OpenCL, in double precision.

double [] OpenCLTemplate.LinearAlgebra.SparseLinalg.ExactMultiply	(	CLImgSparseMatrix	M,
		CLImgVector	x
	)

Computes exact double-precision product of sparse matrix M and vector x.

Parameters:

	M	Sparse matrix
	x	Vector to be multiplied

double [] OpenCLTemplate.LinearAlgebra.SparseLinalg.ExactMultiply	(	CLImgSparseMatrix	M,
		CLImgVector	x
	)

Computes exact double-precision product of sparse matrix M and vector x.

Parameters:

	M	Sparse matrix
	x	Vector to be multiplied

double [] OpenCLTemplate.LinearAlgebra.SparseLinalg.ExactMultiply	(	CLImgSparseMatrix	M,
		CLImgVector	x
	)

Computes exact double-precision product of sparse matrix M and vector x.

Parameters:

	M	Sparse matrix
	x	Vector to be multiplied

float [] OpenCLTemplate.LinearAlgebra.SparseLinalg.LinSolveCL	(	CLImgSparseMatrix	M,
		CLImgVector	b,
		float	tol
	)

Solves linear system Mx = b using conjugate gradient method. Writes variables to Device memory. Improves solution if accuracy is low.

Parameters:

	M	Matrix M
	b	Vector b
	tol	Error tolerance

float [] OpenCLTemplate.LinearAlgebra.SparseLinalg.LinSolveCL	(	CLImgSparseMatrix	M,
		CLImgVector	b,
		float	tol
	)

Solves linear system Mx = b using conjugate gradient method. Writes variables to Device memory. Improves solution if accuracy is low.

Parameters:

	M	Matrix M
	b	Vector b
	tol	Error tolerance

float [] OpenCLTemplate.LinearAlgebra.SparseLinalg.LinSolveCL	(	CLImgSparseMatrix	M,
		CLImgVector	b,
		float	tol
	)

Solves linear system Mx = b using conjugate gradient method. Writes variables to Device memory. Improves solution if accuracy is low.

Parameters:

	M	Matrix M
	b	Vector b
	tol	Error tolerance

void OpenCLTemplate.LinearAlgebra.SparseLinalg.LinSolveCLStep	(	CLImgSparseMatrix	M,
		CLImgVector	b,
		float	tol,
		ref CLImgVector	x
	)

Solves linear system Mx = b using conjugate gradient method. Doesn't try to improve the solution obtained.

Parameters:

	M	Matrix M
	b	Vector b
	tol	Error tolerance
	x	Initial guess

void OpenCLTemplate.LinearAlgebra.SparseLinalg.LinSolveCLStep	(	CLImgSparseMatrix	M,
		CLImgVector	b,
		float	tol
	)

Solves linear system Mx = b using conjugate gradient method. Doesn't try to improve the solution obtained.

Parameters:

	M	Matrix M
	b	Vector b
	tol	Error tolerance

void OpenCLTemplate.LinearAlgebra.SparseLinalg.LinSolveCLStep	(	CLImgSparseMatrix	M,
		CLImgVector	b,
		float	tol,
		ref CLImgVector	x
	)

Solves linear system Mx = b using conjugate gradient method. Doesn't try to improve the solution obtained.

Parameters:

	M	Matrix M
	b	Vector b
	tol	Error tolerance
	x	Initial guess

float [] OpenCLTemplate.LinearAlgebra.SparseLinalg.LinSolveNoCL	(	CLImgSparseMatrix	M,
		CLImgVector	b,
		float	tol
	)

Solves linear system Mx = b using conjugate gradient method.

Parameters:

	M	Matrix M
	b	Vector b
	tol	Error tolerance

float [] OpenCLTemplate.LinearAlgebra.SparseLinalg.LinSolveNoCL	(	CLImgSparseMatrix	M,
		CLImgVector	b,
		float	tol
	)

Solves linear system Mx = b using conjugate gradient method.

Parameters:

	M	Matrix M
	b	Vector b
	tol	Error tolerance

float [] OpenCLTemplate.LinearAlgebra.SparseLinalg.LinSolveNoCL	(	CLImgSparseMatrix	M,
		CLImgVector	b,
		float	tol
	)

Solves linear system Mx = b using conjugate gradient method.

Parameters:

	M	Matrix M
	b	Vector b
	tol	Error tolerance

void OpenCLTemplate.LinearAlgebra.SparseLinalg.Multiply	(	CLImgSparseMatrix	M,
		CLImgVector	x,
		CLImgVector	y
	)

Computes M*x and stores the result in y. Does not automatically read result from device memory.

Parameters:

	M	Sparse matrix
	x	Vector to be multiplied
	y	Result

void OpenCLTemplate.LinearAlgebra.SparseLinalg.Multiply	(	CLImgSparseMatrix	M,
		CLImgVector	x,
		CLImgVector	y
	)

Computes M*x and stores the result in y. Does not automatically read result from device memory.

Parameters:

	M	Sparse matrix
	x	Vector to be multiplied
	y	Result

void OpenCLTemplate.LinearAlgebra.SparseLinalg.Multiply	(	CLImgSparseMatrix	M,
		CLImgVector	x,
		CLImgVector	y
	)

Computes M*x and stores the result in y. Does not automatically read result from device memory.

Parameters:

	M	Sparse matrix
	x	Vector to be multiplied
	y	Result

float [] OpenCLTemplate.LinearAlgebra.SparseLinalg.MultiplyNoCL	(	CLImgSparseMatrix	M,
		CLImgVector	x
	)

Computes product of sparse matrix M and vector x.

Parameters:

	M	Sparse matrix
	x	Vector to be multiplied

float [] OpenCLTemplate.LinearAlgebra.SparseLinalg.MultiplyNoCL	(	CLImgSparseMatrix	M,
		CLImgVector	x
	)

Computes product of sparse matrix M and vector x.

Parameters:

	M	Sparse matrix
	x	Vector to be multiplied

float [] OpenCLTemplate.LinearAlgebra.SparseLinalg.MultiplyNoCL	(	CLImgSparseMatrix	M,
		CLImgVector	x
	)

Computes product of sparse matrix M and vector x.

Parameters:

	M	Sparse matrix
	x	Vector to be multiplied

Member Data Documentation

const int OpenCLTemplate::LinearAlgebra.SparseLinalg::GLOBALWORKSIZE = 2048

Total number of workers.

const int OpenCLTemplate::LinearAlgebra.SparseLinalg::IMGWIDTH = 4096

Width to be used in float4's.

static int OpenCLTemplate::LinearAlgebra.SparseLinalg::MAXITER = 8 [static]

Maximum number of iterations.

CLImgVector OpenCLTemplate::LinearAlgebra.SparseLinalg::x

Initial guess/answer holder.

OpenCLTemplate.LinearAlgebra.SparseLinalg Class Reference

Classes

Public Member Functions

Public Attributes

Static Public Attributes

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation