Package org.flag4j.operations

Class RealDenseMatrixMultiplyDispatcher

java.lang.Object

org.flag4j.operations.RealDenseMatrixMultiplyDispatcher

public final class RealDenseMatrixMultiplyDispatcher
extends Object

Singleton class which stores a map of all viable real dense matrix multiply algorithms and uses that map to dispatch a real dense matrix multiply problem to the appropriate algorithm.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
protected static enum	RealDenseMatrixMultiplyDispatcher.AlgorithmNames	Simple enum class containing all possible choices of real dense matrix multiply algorithms.

Field Summary

Fields

Modifier and Type	Field	Description
private final Map<RealDenseMatrixMultiplyDispatcher.AlgorithmNames,RealDenseTensorBinaryOperation>	algorithmMap	lookup table for all real dense matrix multiply algorithms.
private static final int	CONCURRENT_SWAPPED_THRESHOLD	Threshold for matrices to use the concurrent ikj algorithm.
private static final int	SEQUENTIAL_SWAPPED_THRESHOLD	Threshold for small matrices which should be multiplied using the standard ikj algorithm.
private static RealDenseMatrixMultiplyDispatcher	singletonInstance	Singleton instance of this class.
private static final double	SQUARENESS_RATIO	Ration measuring squareness.

Constructor Summary

Constructors

Modifier	Constructor	Description
private	RealDenseMatrixMultiplyDispatcher()	Creates an instance containing a map of all viable real dense matrix multiply algorithms.

Method Summary

Modifier and Type	Method	Description
static double[]	dispatch(double[] src1, Shape shape1, double[] src2, Shape shape2)	Dispatches a matrix multiply problem to the appropriate algorithm based on the size of the matrices.
static double[]	dispatch(Matrix A, Matrix B)	Dispatches a matrix multiply problem to the appropriate algorithm based on the size of the matrices.
static double[]	dispatchTranspose(Matrix A, Matrix B)	Dispatches a matrix multiply-transpose problem equivalent to A.mult(B.T()) to the appropriate algorithm based on the size of the matrices.
static RealDenseMatrixMultiplyDispatcher	getInstance()	Gets the singleton instance of this class.
private static double	getRatio(Shape shape)	Computes the squareness ratio of a matrix.
(package private) static RealDenseMatrixMultiplyDispatcher.AlgorithmNames	selectAlgorithm(Shape shape1, Shape shape2)	Dynamically chooses matrix multiply algorithm based on the shapes of the two matrices to multiply.
(package private) static RealDenseMatrixMultiplyDispatcher.AlgorithmNames	selectAlgorithmTranspose(Shape shape)	Selects the matrix multiplication-transpose algorithm to use based on the size of the first matrix.

Methods inherited from class java.lang.Object

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

Field Details

singletonInstance

private static RealDenseMatrixMultiplyDispatcher singletonInstance

Singleton instance of this class.

algorithmMap

private final Map<RealDenseMatrixMultiplyDispatcher.AlgorithmNames,RealDenseTensorBinaryOperation> algorithmMap

lookup table for all real dense matrix multiply algorithms.

SQUARENESS_RATIO

private static final double SQUARENESS_RATIO

Ration measuring squareness. the closer to one, the more square the matrix is.

See Also:

• Constant Field Values

SEQUENTIAL_SWAPPED_THRESHOLD

private static final int SEQUENTIAL_SWAPPED_THRESHOLD

Threshold for small matrices which should be multiplied using the standard ikj algorithm.

See Also:

• Constant Field Values

CONCURRENT_SWAPPED_THRESHOLD

private static final int CONCURRENT_SWAPPED_THRESHOLD

Threshold for matrices to use the concurrent ikj algorithm.

See Also:

• Constant Field Values

Constructor Details

RealDenseMatrixMultiplyDispatcher

private RealDenseMatrixMultiplyDispatcher()

Creates an instance containing a map of all viable real dense matrix multiply algorithms.

Method Details

getInstance

public static RealDenseMatrixMultiplyDispatcher getInstance()

Gets the singleton instance of this class. If this class has not been instanced, a new instance will be created.

Returns:

The singleton instance of this class.

dispatch

public static double[] dispatch(Matrix A,
 Matrix B)

Dispatches a matrix multiply problem to the appropriate algorithm based on the size of the matrices.

Parameters:

A - First matrix in the multiplication.

B - Second matrix in the multiplication.

Returns:

The result of the matrix multiplication.

dispatch

public static double[] dispatch(double[] src1,
 Shape shape1,
 double[] src2,
 Shape shape2)

Dispatches a matrix multiply problem to the appropriate algorithm based on the size of the matrices.

Parameters:

src1 - Entries of the first matrix.

shape1 - Shape of the first matrix.

src2 - Entries of the second matrix.

shape2 - Shape of the second matrix.

Returns:

The result of the matrix multiplication.

dispatchTranspose

public static double[] dispatchTranspose(Matrix A,
 Matrix B)

Dispatches a matrix multiply-transpose problem equivalent to A.mult(B.T()) to the appropriate algorithm based on the size of the matrices.

Parameters:

A - First matrix in the multiplication.

B - Second matrix in the multiplication and the matrix to transpose.

Returns:

The matrix multiply-transpose result of A and B.

selectAlgorithm

static RealDenseMatrixMultiplyDispatcher.AlgorithmNames selectAlgorithm(Shape shape1,
 Shape shape2)

Dynamically chooses matrix multiply algorithm based on the shapes of the two matrices to multiply.

Parameters:

shape1 - The shape of the first matrix.

shape2 - The shape fo the second matrix.

Returns:

The algorithm to use in the matrix multiplication.

selectAlgorithmTranspose

static RealDenseMatrixMultiplyDispatcher.AlgorithmNames selectAlgorithmTranspose(Shape shape)

Selects the matrix multiplication-transpose algorithm to use based on the size of the first matrix.

Parameters:

shape - Shape of the first matrix.

Returns:

The algorithm to use to compute the matrix multiplication-transpose.

getRatio

private static double getRatio(Shape shape)

Computes the squareness ratio of a matrix. This is a value between 0 and 1 with 1 being perfectly square and 0 being a row/column vector.

Parameters:

shape - Shape of the matrix to compute the squareness ratio of.

Returns:

The squareness ratio for the specified shape.