org.apache.sysds.runtime.matrix.data

Class LibMatrixOuterAgg

java.lang.Object

org.apache.sysds.runtime.matrix.data.LibMatrixOuterAgg

public class LibMatrixOuterAgg
extends Object

ACS: Purpose of this library is to make some of the unary outer aggregate operator more efficient. Today these operators are being handled through common operations. This library will expand per need and priority to include these operators through this support. To begin with, first operator being handled is unary aggregate for less than (<), rowsum operation. Other list will be added soon are rowsum on >, <=, >=, ==, and != operation.

Method Summary

Modifier and Type	Method and Description
static void	adjustRowIndicesMax(int[] vix, double[] vmb, BinaryOperator bOp) This function adjusts indices to be leveraged in uarimaxXX functions.
static void	adjustRowIndicesMin(int[] vix, double[] vmb, BinaryOperator bOp) This function adjusts indices to be leveraged in uariminXX functions.
static void	aggregateMatrix(MatrixBlock in1Val, MatrixBlock outVal, double[] bv, int[] bvi, BinaryOperator bOp, AggregateUnaryOperator uaggOp)
static boolean	isCompareOperator(BinaryOperator bOp) This will return if uaggOp is of type RowIndexMin
static boolean	isRowIndexMax(AggregateUnaryOperator uaggOp) This will return if uaggOp is of type RowIndexMax
static boolean	isRowIndexMin(AggregateUnaryOperator uaggOp) This will return if uaggOp is of type RowIndexMin
static boolean	isSupportedUaggOp(AggregateUnaryOperator uaggOp, BinaryOperator bOp)
static int[]	prepareRowIndices(int iCols, double[] vmb, BinaryOperator bOp, AggregateUnaryOperator uaggOp)
static int[]	prepareRowIndicesMax(int iCols, double[] vmb, BinaryOperator bOp) This function will return max indices, based on column vector data.
static int[]	prepareRowIndicesMin(int iCols, double[] vmb, BinaryOperator bOp) This function will return min indices, based on column vector data.
static void	resetOutputMatrix(MatrixIndexes in1Ix, MatrixBlock in1Val, MatrixIndexes outIx, MatrixBlock outVal, AggregateUnaryOperator uaggOp) ReSet output matrix
static void	setMaxIndexInPartition(int[] vix, double[] vmb) This function will set maximum index in the partition to all cells in partition.
static void	setMinIndexInPartition(int[] vix, double[] vmb) This function will set minimum index in the partition to all cells in partition.
static void	shiftLeft(int[] vix, double[] vmb) This function will shift indices from one partition to next in left direction.
static void	shiftRight(int[] vix, double[] vmb) This function will shift indices from one partition to next in right direction.

Methods inherited from class java.lang.Object

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

Method Detail

isRowIndexMax

public static boolean isRowIndexMax(AggregateUnaryOperator uaggOp)

This will return if uaggOp is of type RowIndexMax

Parameters:

uaggOp - aggregate unary operator

Returns:

true if aggregate unary operator is of type rowIndexMax

isRowIndexMin

public static boolean isRowIndexMin(AggregateUnaryOperator uaggOp)

This will return if uaggOp is of type RowIndexMin

Parameters:

uaggOp - aggregate unary operator

Returns:

true if aggregate unary operator is of type rowIndexMin

isCompareOperator

public static boolean isCompareOperator(BinaryOperator bOp)

This will return if uaggOp is of type RowIndexMin

Parameters:

bOp - binary operator

Returns:

true/false, based on if its one of the six operators (<, <=, >, >=, == and !=)

isSupportedUaggOp

public static boolean isSupportedUaggOp(AggregateUnaryOperator uaggOp,
                                        BinaryOperator bOp)

prepareRowIndices

public static int[] prepareRowIndices(int iCols,
                                      double[] vmb,
                                      BinaryOperator bOp,
                                      AggregateUnaryOperator uaggOp)

prepareRowIndicesMax

public static int[] prepareRowIndicesMax(int iCols,
                                         double[] vmb,
                                         BinaryOperator bOp)

This function will return max indices, based on column vector data. This indices will be computed based on operator. These indices can be used to compute max index for a given input value in subsequent operation. e.g. Right Vector has data (V1) : 6 3 9 7 2 4 4 3 Original indices for this data will be (I1): 1 2 3 4 5 6 7 8 Sorting this data based on value will be (V2): 2 3 3 4 4 6 7 9 Then indices will be ordered as (I2): 5 2 8 6 7 1 4 3 CumMax of I2 will be A: (CumMin(I2)) 5 5 8 8 8 8 8 8 CumMax of I2 in reverse order be B: 8 8 8 7 7 4 4 3 Values from vector A is used to compute RowIndexMax for > & >= operators Values from vector B is used to compute RowIndexMax for < & <= operators Values from I2 is used to compute RowIndexMax for == operator. Original values are directly used to compute RowIndexMax for != operator Shifting values from vector A or B is required to compute final indices. Once indices are shifted from vector A or B, their cell value corresponding to input data will be used.

Parameters:

iCols - ?

vmb - ?

bOp - binary operator

Returns:

array of maximum row indices

prepareRowIndicesMin

public static int[] prepareRowIndicesMin(int iCols,
                                         double[] vmb,
                                         BinaryOperator bOp)

This function will return min indices, based on column vector data. This indices will be computed based on operator. These indices can be used to compute min index for a given input value in subsequent operation. e.g. Right Vector has data (V1) : 6 3 9 7 2 4 4 3 Original indices for this data will be (I1): 1 2 3 4 5 6 7 8 Sorting this data based on value will be (V2): 2 3 3 4 4 6 7 9 Then indices will be ordered as (I2): 5 2 8 6 7 1 4 3 CumMin of I2 will be A: (CumMin(I2)) 5 2 2 2 2 1 1 1 CumMin of I2 in reverse order be B: 1 1 1 1 1 1 3 3 Values from vector A is used to compute RowIndexMin for > operator Values from vector B is used to compute RowIndexMin for <, <= and >= operators Values from I2 is used to compute RowIndexMax for == operator. Original values are directly used to compute RowIndexMax for != operator Shifting values from vector A or B is required to compute final indices. Once indices are shifted from vector A or B, their cell value corresponding to input data will be used.

Parameters:

iCols - ?

vmb - ?

bOp - binary operator

Returns:

array of minimum row indices

resetOutputMatrix

public static void resetOutputMatrix(MatrixIndexes in1Ix,
                                     MatrixBlock in1Val,
                                     MatrixIndexes outIx,
                                     MatrixBlock outVal,
                                     AggregateUnaryOperator uaggOp)

ReSet output matrix

Parameters:

in1Ix - input matrix indexes

in1Val - input matrix block

outIx - output matrix indexes

outVal - output matrix block

uaggOp - aggregate unary operator

aggregateMatrix

public static void aggregateMatrix(MatrixBlock in1Val,
                                   MatrixBlock outVal,
                                   double[] bv,
                                   int[] bvi,
                                   BinaryOperator bOp,
                                   AggregateUnaryOperator uaggOp)

adjustRowIndicesMax

public static void adjustRowIndicesMax(int[] vix,
                                       double[] vmb,
                                       BinaryOperator bOp)

This function adjusts indices to be leveraged in uarimaxXX functions. Initially vector containing indices are sorted based on value and then CumMax/CumMin per need for <, <=, >, >= operator, where as just sorted indices based on value for ==, and != operators. There is need to shift these indices for different operators, which is handled through this function.

Parameters:

vix - ?

vmb - ?

bOp - binary operator

adjustRowIndicesMin

public static void adjustRowIndicesMin(int[] vix,
                                       double[] vmb,
                                       BinaryOperator bOp)

This function adjusts indices to be leveraged in uariminXX functions. Initially vector containing indices are sorted based on value and then CumMin per need for <, <=, >, >= operator, where as just sorted indices based on value for ==, and != operators. There is need to shift these indices for different operators, which is handled through this function.

Parameters:

vix - ?

vmb - ?

bOp - binary operator

shiftRight

public static void shiftRight(int[] vix,
                              double[] vmb)

This function will shift indices from one partition to next in right direction. For an example, if there are two sorted vector based on value like following, where V2 is sorted data, and I2 are its corresponding indices. Then this function will shift indices to right by one partition I2". Left most partition remained untouched. Sorting this data based on value will be (V2): 2 3 3 4 4 6 7 9 Then indices will be ordered as (I2): 5 2 8 6 7 1 4 3 Shift Right by one partition (I2") 5 5 2 8 6 7 1 4

Parameters:

vix - ?

vmb - ?

shiftLeft

public static void shiftLeft(int[] vix,
                             double[] vmb)

This function will shift indices from one partition to next in left direction. For an example, if there are two sorted vector based on value like following, where V2 is sorted data, and I2 are its corresponding indices. Then this function will shift indices to right by one partition I2". Right most partition remained untouched. Sorting this data based on value will be (V2): 2 3 3 4 4 6 7 9 Then indices will be ordered as (I2): 5 2 8 6 7 1 4 3 Shift Left by one partition (I2") 2 8 6 7 1 4 3 3

Parameters:

vix - ?

vmb - ?

setMinIndexInPartition

public static void setMinIndexInPartition(int[] vix,
                                          double[] vmb)

This function will set minimum index in the partition to all cells in partition. For an example, if there are two sorted vector based on value like following, where V2 is sorted data, and I2 are its corresponding indices. In this case, for partition with value = 4, has two different indices -- 6, and 7. This function will set indices to both cells to minimum value of 6. Sorting this data based on value will be (V2): 2 3 3 4 4 6 7 9 Then indices will be ordered as (I2): 5 2 8 6 7 1 4 3 Minimum indices set in the partition (I2") 5 2 8 6 6 1 4 3

Parameters:

vix - ?

vmb - ?

setMaxIndexInPartition

public static void setMaxIndexInPartition(int[] vix,
                                          double[] vmb)

This function will set maximum index in the partition to all cells in partition. For an example, if there are two sorted vector based on value like following, where V2 is sorted data, and I2 are its corresponding indices. In this case, for partition with value = 4, has two different indices -- 6, and 7. This function will set indices to both cells to maximum value of 7. Sorting this data based on value will be (V2): 2 3 3 4 4 6 7 9 Then indices will be ordered as (I2): 5 2 8 6 7 1 4 3 Maximum indices set in the partition (I2") 5 2 8 7 7 1 4 3

Parameters:

vix - ?

vmb - ?