Package jdistlib.accelerator

Class SparseCholeskyPlan

java.lang.Object

jdistlib.accelerator.SparseCholeskyPlan

public final class SparseCholeskyPlan
extends Object

Provider-facing symbolic sparse-Cholesky plan.

The plan owns a fill-reducing symmetric permutation and the zero-based CSR pattern of the resulting lower factor. Native providers use factorValues(CsrMatrix) or factorValues(FloatCsrMatrix) to scatter a new authoritative input triangle into that fixed pattern before a numerical refactorization. Symbolic analysis therefore remains reusable without exposing provider-specific factor storage in the public factor API.

Method Summary

Modifier and Type	Method	Description
static SparseCholeskyPlan	analyze(CsrMatrix matrix, MatrixTriangle triangle, SparseOrdering ordering)	Analyzes an FP64 sparse SPD structure without performing numeric factorization.
static SparseCholeskyPlan	analyze(FloatCsrMatrix matrix, MatrixTriangle triangle, SparseOrdering ordering)	Analyzes an FP32 sparse SPD structure without performing numeric factorization.
int	dimension()
int[]	factorColumnIndices()	Returns zero-based lower-factor CSR column indices.
int	factorNonzeroCount()
int[]	factorRowStarts()	Returns zero-based lower-factor CSR row starts.
double[]	factorValues(CsrMatrix matrix)	Scatters an FP64 authoritative triangle into the fixed factor pattern.
float[]	factorValues(FloatCsrMatrix matrix)	Scatters an FP32 authoritative triangle into the fixed factor pattern.
int[]	permutation()	Returns the new-to-original symmetric permutation.
int	structuralNonzeroCount()

Methods inherited from class java.lang.Object

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

Method Details

analyze

public static SparseCholeskyPlan analyze(CsrMatrix matrix,
 MatrixTriangle triangle,
 SparseOrdering ordering)

Analyzes an FP64 sparse SPD structure without performing numeric factorization.

analyze

public static SparseCholeskyPlan analyze(FloatCsrMatrix matrix,
 MatrixTriangle triangle,
 SparseOrdering ordering)

Analyzes an FP32 sparse SPD structure without performing numeric factorization.

dimension

public int dimension()

structuralNonzeroCount

public int structuralNonzeroCount()

factorNonzeroCount

public int factorNonzeroCount()

permutation

public int[] permutation()

Returns the new-to-original symmetric permutation.

factorColumnIndices

public int[] factorColumnIndices()

Returns zero-based lower-factor CSR column indices.

factorRowStarts

public int[] factorRowStarts()

Returns zero-based lower-factor CSR row starts.

factorValues

public double[] factorValues(CsrMatrix matrix)

Scatters an FP64 authoritative triangle into the fixed factor pattern.

factorValues

public float[] factorValues(FloatCsrMatrix matrix)

Scatters an FP32 authoritative triangle into the fixed factor pattern.