org.apache.sysds.runtime.compress

Class CompressedMatrixBlockFactory

java.lang.Object

org.apache.sysds.runtime.compress.CompressedMatrixBlockFactory

public class CompressedMatrixBlockFactory
extends Object

Factory pattern to construct a CompressedMatrixBlock.

Method Summary

Modifier and Type	Method and Description
static org.apache.commons.lang3.tuple.Pair<MatrixBlock,CompressionStatistics>	compress(MatrixBlock mb) Default sequential compression with no parallelization
static org.apache.commons.lang3.tuple.Pair<MatrixBlock,CompressionStatistics>	compress(MatrixBlock mb, CompressionSettings customSettings)
static org.apache.commons.lang3.tuple.Pair<MatrixBlock,CompressionStatistics>	compress(MatrixBlock mb, int k)
static org.apache.commons.lang3.tuple.Pair<MatrixBlock,CompressionStatistics>	compress(MatrixBlock mb, int k, CompressionSettings compSettings) The main method for compressing the input matrix.
static CompressedMatrixBlock	createConstant(int numRows, int numCols, double value) Method for constructing a compressed matrix out of an constant input.
double	getLastTimePhase()
void	setNextTimePhase(double time)

Methods inherited from class java.lang.Object

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

Method Detail

compress

public static org.apache.commons.lang3.tuple.Pair<MatrixBlock,CompressionStatistics> compress(MatrixBlock mb)

Default sequential compression with no parallelization

Parameters:

mb - The matrixBlock to compress

Returns:

A Pair of a Matrix Block and Compression Statistics.

compress

public static org.apache.commons.lang3.tuple.Pair<MatrixBlock,CompressionStatistics> compress(MatrixBlock mb,
                                                                                              CompressionSettings customSettings)

compress

public static org.apache.commons.lang3.tuple.Pair<MatrixBlock,CompressionStatistics> compress(MatrixBlock mb,
                                                                                              int k)

compress

public static org.apache.commons.lang3.tuple.Pair<MatrixBlock,CompressionStatistics> compress(MatrixBlock mb,
                                                                                              int k,
                                                                                              CompressionSettings compSettings)

The main method for compressing the input matrix. SAMPLE-BASED DECISIONS: Decisions such as testing if a column is amenable to bitmap compression or evaluating co-coding potentials are made based on a subset of the rows. For large data sets, sampling might take a significant amount of time. So, we generate only one sample and use it for the entire compression process. Once the compression plan is selected based on sampling, the plan is verified and decisions are overwritten by full estimates.

Parameters:

mb - The matrix block to compress

k - The number of threads used to execute the compression

compSettings - The Compression settings used

Returns:

A pair of an possibly compressed matrix block and compression statistics.

createConstant

public static CompressedMatrixBlock createConstant(int numRows,
                                                   int numCols,
                                                   double value)

Method for constructing a compressed matrix out of an constant input. Since the input is a constant value it is trivially compressable, therefore we skip the entire compression planning and directly return a compressed constant matrix

Parameters:

numRows - The number of Rows in the matrix

numCols - The number of Columns in the matrix

value - The value contained in the matrix

Returns:

The Compressed Constant matrix.

setNextTimePhase

public void setNextTimePhase(double time)

getLastTimePhase

public double getLastTimePhase()