org.ejml.alg.dense.decomposition.chol

Class CholeskyDecompositionBlock_D64

java.lang.Object

org.ejml.alg.dense.decomposition.chol.CholeskyDecompositionCommon_D64

org.ejml.alg.dense.decomposition.chol.CholeskyDecompositionBlock_D64

All Implemented Interfaces:

CholeskyDecomposition<DenseMatrix64F>, DecompositionInterface<DenseMatrix64F>

public class CholeskyDecompositionBlock_D64
extends CholeskyDecompositionCommon_D64

This is an implementation of Cholesky that processes internal submatrices as blocks. This is done to reduce the number of cache issues.

Constructor Summary

Constructors

Constructor and Description
CholeskyDecompositionBlock_D64(int blockWidth) Creates a CholeksyDecomposition capable of decomposing a matrix that is n by n, where n is the width.

Method Summary

Modifier and Type	Method and Description
void	setExpectedMaxSize(int numRows, int numCols) Declares additional internal data structures.
static void	solveL_special(double[] L, DenseMatrix64F b_src, int indexSrc, int indexDst, DenseMatrix64F B) This is a variation on the TriangularSolver.solveL(double[], double[], int) function.
static void	symmRankTranA_sub(DenseMatrix64F a, DenseMatrix64F c, int startIndexC) Performs this operation: c = c - aTa where c is a submatrix.

Methods inherited from class org.ejml.alg.dense.decomposition.chol.CholeskyDecompositionCommon_D64

_getVV, computeDeterminant, decompose, getT, getT, inputModified, isLower

Methods inherited from class java.lang.Object

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

Constructor Detail

CholeskyDecompositionBlock_D64

public CholeskyDecompositionBlock_D64(int blockWidth)

Creates a CholeksyDecomposition capable of decomposing a matrix that is n by n, where n is the width.

Parameters:

blockWidth - The width of a block.

Method Detail

setExpectedMaxSize

public void setExpectedMaxSize(int numRows,
                               int numCols)

Declares additional internal data structures.

Overrides:

setExpectedMaxSize in class CholeskyDecompositionCommon_D64

solveL_special

public static void solveL_special(double[] L,
                                  DenseMatrix64F b_src,
                                  int indexSrc,
                                  int indexDst,
                                  DenseMatrix64F B)

This is a variation on the TriangularSolver.solveL(double[], double[], int) function. It grabs the input from the top right row rectangle of the source matrix then writes the results to the lower bottom column rectangle. The rectangle matrices just matrices are submatrices of the matrix that is being decomposed. The results are also written to B.

Parameters:

L - A lower triangular matrix.

b_src - matrix with the vectors that are to be solved for

indexSrc - First index of the submatrix where the inputs are coming from.

indexDst - First index of the submatrix where the results are going to.

B -

symmRankTranA_sub

public static void symmRankTranA_sub(DenseMatrix64F a,
                                     DenseMatrix64F c,
                                     int startIndexC)

Performs this operation:

c = c - a^Ta
where c is a submatrix.

Only the upper triangle is updated.

Parameters:

a - A matrix.

c - A matrix.

startIndexC - start of the submatrix in c.