Module org.apache.lucene.core

Package org.apache.lucene.util.packed

Class BulkOperation

java.lang.Object

org.apache.lucene.util.packed.BulkOperation

All Implemented Interfaces:

PackedInts.Decoder, PackedInts.Encoder

Direct Known Subclasses:

BulkOperationPacked, BulkOperationPackedSingleBlock

abstract class BulkOperation
extends Object
implements PackedInts.Decoder, PackedInts.Encoder

Efficient sequential read/write of packed integers.

Field Summary

Fields

Modifier and Type	Field	Description
private static final BulkOperation[]	packedBulkOps
private static final BulkOperation[]	packedSingleBlockBulkOps

Constructor Summary

Constructors

Constructor	Description
BulkOperation()

Method Summary

Modifier and Type	Method	Description
final int	computeIterations(int valueCount, int ramBudget)	For every number of bits per value, there is a minimum number of blocks (b) / values (v) you need to write in order to reach the next block boundary:
static BulkOperation	of(PackedInts.Format format, int bitsPerValue)
protected int	writeLong(long block, byte[] blocks, int blocksOffset)

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.apache.lucene.util.packed.PackedInts.Decoder

byteBlockCount, byteValueCount, decode, decode, decode, decode, longBlockCount, longValueCount

Methods inherited from interface org.apache.lucene.util.packed.PackedInts.Encoder

byteBlockCount, byteValueCount, encode, encode, encode, encode, longBlockCount, longValueCount

Field Details

packedBulkOps

private static final BulkOperation[] packedBulkOps

packedSingleBlockBulkOps

private static final BulkOperation[] packedSingleBlockBulkOps

Constructor Details

BulkOperation

BulkOperation()

Method Details

of

public static BulkOperation of(PackedInts.Format format,
 int bitsPerValue)

writeLong

protected int writeLong(long block,
 byte[] blocks,
 int blocksOffset)

computeIterations

public final int computeIterations(int valueCount,
 int ramBudget)

For every number of bits per value, there is a minimum number of blocks (b) / values (v) you need to write in order to reach the next block boundary:

  - 16 bits per value -> b=2, v=1
  - 24 bits per value -> b=3, v=1
  - 50 bits per value -> b=25, v=4
  - 63 bits per value -> b=63, v=8
  - ...
 

A bulk read consists in copying iterations*v values that are contained in iterations*b blocks into a long[] (higher values of iterations are likely to yield a better throughput): this requires n * (b + 8v) bytes of memory.

This method computes iterations as ramBudget / (b + 8v) (since a long is 8 bytes).