Namespace FastBWT

public class BWT_Medium_Sorter

BWT Sorter for normal-sized blocks.

Constructors

public BWT_Small_Sorter(int BlockSize): Create new BWT_Small_Sorter that supports blocks of up to BlockSize length.

Properties

public List<StatisticStep> Statistics { get; private set; }: Statistics of the last sorting. This collection will be overwritten in Sort methods.

Methods

public abstract void GenerateVector(byte[] SourceBlock, ICollection Index, long DataSize = 0): Generate the sorting vector for source bytes. The actual type of Array shall be uint[].
public abstract long Sort(byte[] SourceBlock, byte[] TargetBlock, long DataSize = 0): Sort SourceBlock to TargetBlock. This method internally generates the sorting vector and fills the TargetBlock.

Remarks

This sorter supports blocks of up to about 2G and is designed for "normal-sized" blocks. Consider to use this sorter if the size of your blocks is 200k - 50M.

It uses uint[] as a type for sorting vector. Big regions in statistics are 64k or longer.

This sorter uses the same sorting algorithm as BWT_Small_Sorter but with some optimizations:

Initialization is multithreaded and uses 2-bytes bucket sorting reducing the first iteration. For larger blocks this may spare up to 20% of time.
Scheduling of big regions is more aggressive. Each worker sorts always the largest region first. This reduces the penalty for sorting of very large region.
Region shall be much larger to be considered as a big one.
Small regions are sorted in larger batches to reduce synchronization.