Package org.apache.sysds.runtime.data

Class DenseBlock

java.lang.Object

org.apache.sysds.runtime.data.DenseBlock

All Implemented Interfaces:

Serializable, Block

Direct Known Subclasses:

DenseBlockDRB, DenseBlockLDRB

public abstract class DenseBlock
extends Object
implements Serializable, Block

This DenseBlock is an abstraction for different dense, row-major matrix formats. For efficient dense operations, this API does not expose a row but a row-block abstraction, where a block can contain one or many contiguous rows.

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	DenseBlock.Type

Method Summary

Modifier and Type	Method	Description
abstract int	blockSize()	Get the number of rows per block, except last one.
abstract int	blockSize(int bix)	Get the number of rows of the given block.
abstract long	capacity()	Get the total length of allocated blocks.
boolean	contains(double pattern)	Checks if the block contains at least one value of the given pattern.
abstract long	countNonZeros()	Compute the number of non-zero values, which potentially makes a full pass over the underlying blocks.
abstract int	countNonZeros(int r)	Compute the number of non-zero values for the given row, which potentially makes a full pass over the underlying row.
abstract long	countNonZeros(int rl, int ru, int cl, int cu)	Compute the number of non-zero values, which potentially makes a full pass over the underlying blocks in the row range.
boolean	equals(Object o)
boolean	equals(DenseBlock o, double eps)	Verify if the values in this dense block is equivalent to that dense block, not taking into account the dimensions of the contained values.
static double	estimateMemory(long nrows, long ncols)
abstract double	get(int[] ix)	Get the value of a given cell
abstract double	get(int r, int c)	Get the value for a given row and column.
int	getCumODims(int i)	Get the ith cumulative dimensions size of the dense block, without row.
int	getDim(int i)	Get the ith dimensions size of the dense block.
abstract long	getLong(int[] ix)	Get the value of a given cell as long
void	getNextIndexes(int[] ix)	Calculates the next index array.
abstract String	getString(int[] ix)	Get the value of a given cell as a String
abstract void	incr(int r, int c)	Increments the given value for a given row and column.
abstract void	incr(int r, int c, double delta)	Increments the given value for a given row and column by delta.
abstract int	index(int r)	Get the block index for a given row.
abstract boolean	isContiguous()	Indicates if the dense block has a single underlying block, i.e., if numBlocks==1.
abstract boolean	isContiguous(int rl, int ru)	Indicates if the dense block has a single underlying block for the given row range.
abstract boolean	isNumeric()	Indicates if the dense block is numeric.
abstract boolean	isNumeric(Types.ValueType vt)	Indicates if the dense block is a specific numeric value type.
abstract int	numBlocks()	Get the number of allocated blocks.
int	numCols()	Get the number of columns / first dimension
int	numDims()	Get the number of dimensions.
int	numRows()	Get the number of rows.
abstract int	pos(int r)	Get the position for a given row within its associated block.
abstract int	pos(int[] ix)	Get the position for a given cell within the associated block.
abstract int	pos(int r, int c)	Get the position for a given row and column within the associated block.
void	reset()	Resets the dense block by deleting non-zero values.
void	reset(int[] dims)	Resets the dense block by deleting non-zero values.
void	reset(int[] dims, double v)	Resets the dense block by deleting non-zeros.
void	reset(int rlen, int clen)	Resets the dense block by deleting non-zeros.
void	reset(int rlen, int[] odims)	Resets the dense block by deleting non-zeros.
abstract void	reset(int rlen, int[] odims, double v)	Resets the dense block by setting the given value.
void	reset(int rlen, int clen, double v)	Resets the dense block by setting the given value.
abstract DenseBlock	set(double v)	Set the given value for the entire dense block (fill).
abstract DenseBlock	set(int[] ix, double v)	Set the specified cell to the given value.
abstract DenseBlock	set(int[] ix, long v)	Set the specified cell to the given value.
abstract DenseBlock	set(int[] ix, String v)	Set the specified cell to the given value.
abstract DenseBlock	set(int r, double[] v)	Copy the given vector into the given row.
abstract DenseBlock	set(int r, int c, double v)	Set the given value for a given row and column.
abstract DenseBlock	set(int rl, int ru, int cl, int cu, double v)	Set the given value for an entire index range of the dense block (fill).
DenseBlock	set(int rl, int ru, int cl, int cu, DenseBlock db)	Copy the given dense block into the specified index range.
DenseBlock	set(int r, KahanObject kbuff)	Copy the given kahan object sum and correction into the given row.
DenseBlock	set(String s)	Set the given string for the entire dense block (fill).
abstract DenseBlock	set(DenseBlock db)	Copy the given dense block.
DenseBlock	set(KahanObject kbuff)	Copy the given kahan object sum and correction.
void	setDims(int[] dims)	Set the dimensions of the dense MatrixBlock.
long	size()	Get the length of the dense block as the product of all dimensions.
abstract int	size(int bix)	Get the length of the given block.
String	toString()
abstract double[]	values(int r)	Get the allocated block for the given row.
abstract double[]	valuesAt(int bix)	Get an allocated block.

Methods inherited from class java.lang.Object

getClass, hashCode, notify, notifyAll, wait, wait, wait

Method Detail

getDim

public final int getDim(int i)

Get the ith dimensions size of the dense block.

Parameters:

i - the number of dimension to get

Returns:

the size of the dimension

getCumODims

public final int getCumODims(int i)

Get the ith cumulative dimensions size of the dense block, without row.

Parameters:

i - the number of the cumulative dimension to get (0 equals the second dimension!)

Returns:

the size of the dimension cumulative with all following dimensions

reset

public final void reset()

Resets the dense block by deleting non-zero values. After this call all countNonZeros() calls are guaranteed to return 0.

reset

public final void reset(int[] dims)

Resets the dense block by deleting non-zero values. After this call all countNonZeros() calls are guaranteed to return 0. If the new dimensions exceed the current capacity, the underlying storage is extended accordingly.

Parameters:

dims - length and size of dimensions.

reset

public final void reset(int[] dims,
                        double v)

Resets the dense block by deleting non-zeros.

Parameters:

dims - lenth and size of dimensions

v - value

reset

public final void reset(int rlen,
                        int clen)

Resets the dense block by deleting non-zeros.

Parameters:

rlen - number of rows

clen - number of columns

reset

public final void reset(int rlen,
                        int[] odims)

Resets the dense block by deleting non-zeros.

Parameters:

rlen - number of rows

odims - offsets of other dimensions

reset

public final void reset(int rlen,
                        int clen,
                        double v)

Resets the dense block by setting the given value.

Parameters:

rlen - number of rows

clen - number of columns

v - value

reset

public abstract void reset(int rlen,
                           int[] odims,
                           double v)

Resets the dense block by setting the given value.

Parameters:

rlen - number of rows

odims - other dimensions

v - value

estimateMemory

public static double estimateMemory(long nrows,
                                    long ncols)

setDims

public void setDims(int[] dims)

Set the dimensions of the dense MatrixBlock.

Parameters:

dims - The dimensions to set, first dimension is rows, second cols.

numRows

public final int numRows()

Get the number of rows.

Returns:

number of rows

numCols

public final int numCols()

Get the number of columns / first dimension

Returns:

number of columns

numDims

public final int numDims()

Get the number of dimensions.

Returns:

number of dimensions, min 2

numBlocks

public abstract int numBlocks()

Get the number of allocated blocks.

Returns:

number of blocks

blockSize

public abstract int blockSize()

Get the number of rows per block, except last one.

Returns:

number of rows in block

blockSize

public abstract int blockSize(int bix)

Get the number of rows of the given block.

Parameters:

bix - block index

Returns:

number of rows in block

isNumeric

public abstract boolean isNumeric()

Indicates if the dense block is numeric.

Returns:

true if numeric (FP, INT, BOOLEAN)

isNumeric

public abstract boolean isNumeric(Types.ValueType vt)

Indicates if the dense block is a specific numeric value type.

Parameters:

vt - value type to check

Returns:

true if numeric and of value type vt

isContiguous

public abstract boolean isContiguous()

Indicates if the dense block has a single underlying block, i.e., if numBlocks==1.

Returns:

true if single block

isContiguous

public abstract boolean isContiguous(int rl,
                                     int ru)

Indicates if the dense block has a single underlying block for the given row range.

Parameters:

rl - row lower index

ru - row upper index (inclusive)

Returns:

true if single block in row range

size

public final long size()

Get the length of the dense block as the product of all dimensions.

Returns:

length

size

public abstract int size(int bix)

Get the length of the given block.

Parameters:

bix - block index

Returns:

length

capacity

public abstract long capacity()

Get the total length of allocated blocks.

Returns:

capacity

countNonZeros

public abstract long countNonZeros()

Compute the number of non-zero values, which potentially makes a full pass over the underlying blocks.

Returns:

number of non-zeros

countNonZeros

public abstract int countNonZeros(int r)

Compute the number of non-zero values for the given row, which potentially makes a full pass over the underlying row.

Parameters:

r - row index

Returns:

number of non-zeros

countNonZeros

public abstract long countNonZeros(int rl,
                                   int ru,
                                   int cl,
                                   int cu)

Compute the number of non-zero values, which potentially makes a full pass over the underlying blocks in the row range.

Parameters:

rl - row lower index

ru - row upper index (exclusive)

cl - column lower index

cu - column upper index (exclusive)

Returns:

number of non-zeros

values

public abstract double[] values(int r)

Get the allocated block for the given row. This call is equivalent to valuesAt(indexes(r)).

Parameters:

r - row index

Returns:

block

valuesAt

public abstract double[] valuesAt(int bix)

Get an allocated block.

Parameters:

bix - block index

Returns:

block

index

public abstract int index(int r)

Get the block index for a given row.

Parameters:

r - row index

Returns:

block index

pos

public abstract int pos(int r)

Get the position for a given row within its associated block.

Parameters:

r - row index

Returns:

block position

pos

public abstract int pos(int r,
                        int c)

Get the position for a given row and column within the associated block.

Parameters:

r - row index

c - column index

Returns:

block position

pos

public abstract int pos(int[] ix)

Get the position for a given cell within the associated block.

Parameters:

ix - cell indexes

Returns:

block position

incr

public abstract void incr(int r,
                          int c)

Increments the given value for a given row and column.

Parameters:

r - row index

c - column index

incr

public abstract void incr(int r,
                          int c,
                          double delta)

Increments the given value for a given row and column by delta.

Parameters:

r - row index

c - column index

delta - increment value

set

public abstract DenseBlock set(double v)

Set the given value for the entire dense block (fill).

Parameters:

v - value

Returns:

self

set

public DenseBlock set(String s)

Set the given string for the entire dense block (fill). Generally the string will be parsed, except for string DenseBlock.

Parameters:

s - string

Returns:

self

set

public abstract DenseBlock set(int rl,
                               int ru,
                               int cl,
                               int cu,
                               double v)

Set the given value for an entire index range of the dense block (fill).

Parameters:

rl - row lower index

ru - row upper index (exclusive)

cl - column lower index

cu - column upper index (exclusive)

v - value

Returns:

self

set

public abstract DenseBlock set(int r,
                               int c,
                               double v)

Set the given value for a given row and column.

Parameters:

r - row index

c - column index

v - value

Returns:

self

set

public abstract DenseBlock set(int r,
                               double[] v)

Copy the given vector into the given row.

Parameters:

r - row index

v - value vector

Returns:

self

set

public abstract DenseBlock set(DenseBlock db)

Copy the given dense block.

Parameters:

db - dense block

Returns:

self

set

public DenseBlock set(int rl,
                      int ru,
                      int cl,
                      int cu,
                      DenseBlock db)

Copy the given dense block into the specified index range.

Parameters:

rl - row lower index

ru - row upper index (exclusive)

cl - column lower index

cu - column upper index (exclusive)

db - dense block

Returns:

self

getNextIndexes

public void getNextIndexes(int[] ix)

Calculates the next index array. Note that if the given index array was the last element, the next index will be outside of range.

Parameters:

ix - the index array which will be incremented to the next index array

set

public DenseBlock set(KahanObject kbuff)

Copy the given kahan object sum and correction.

Parameters:

kbuff - kahan object

Returns:

self

set

public abstract DenseBlock set(int[] ix,
                               double v)

Set the specified cell to the given value.

Parameters:

ix - cell indexes

v - value

Returns:

self

set

public abstract DenseBlock set(int[] ix,
                               long v)

Set the specified cell to the given value.

Parameters:

ix - cell indexes

v - value

Returns:

self

set

public abstract DenseBlock set(int[] ix,
                               String v)

Set the specified cell to the given value.

Parameters:

ix - cell indexes

v - value as String

Returns:

self

set

public DenseBlock set(int r,
                      KahanObject kbuff)

Copy the given kahan object sum and correction into the given row.

Parameters:

r - row index

kbuff - kahan object

Returns:

self

get

public abstract double get(int r,
                           int c)

Get the value for a given row and column.

Specified by:

get in interface Block

Parameters:

r - row index

c - column index

Returns:

value

get

public abstract double get(int[] ix)

Get the value of a given cell

Parameters:

ix - cell indexes

Returns:

value

getString

public abstract String getString(int[] ix)

Get the value of a given cell as a String

Parameters:

ix - cell indexes

Returns:

value as String

getLong

public abstract long getLong(int[] ix)

Get the value of a given cell as long

Parameters:

ix - cell indexes

Returns:

value as long

contains

public boolean contains(double pattern)

Checks if the block contains at least one value of the given pattern. Implementations need to handle NaN patterns as well (note that NaN==NaN yields false).

Parameters:

pattern - checked pattern

Returns:

true if pattern appears at least once, otherwise false

equals

public boolean equals(Object o)

Overrides:

equals in class Object

equals

public boolean equals(DenseBlock o,
                      double eps)

Verify if the values in this dense block is equivalent to that dense block, not taking into account the dimensions of the contained values. Note in some cases one or the other block is allocated bigger than the other, so the values compared is only the values of the smaller block.

Parameters:

o - Other block

eps - Epsilon allowed

Returns:

If the blocs are equivalent.

toString

public String toString()

Overrides:

toString in class Object