jlabeit/parallel-divsufsort

Parallel Lightweight Suffix Array Algorithm.

Description

Parallel-DivSufSort is a parallel lightweight suffix array construction algorithm for byte alphabets written in C++. It is a implementation based on:

• divsufsort implementation of induced sorting by Yuta Mori.
• parallel-range-lite implementation of prefix doubling for integer alphabets.

A detailed description and benchmarks of the algorithm can be found in the following work.

Julian Labeit, Julian Shun, and Guy E. Blelloch. Parallel Lightweight Wavelet Tree, Suffix Array and FM-Index Construction. DCC 2015.

Installation

The following steps have been tested on Ubuntu 14.04 with gcc 5.3.0 and cmake 2.8.12.

git clone https://github.com/jlabeit/parallel-divsufsort.git
cd parallel-divsufsort
mkdir build
cd build
cmake ..
make
make install

Note that in the default version the cilkplus implementation by gcc is used for parallelization. To change this setting edit parallelization settings in the CMakeLists.txt file.

Getting Started

An example application can be found in examples/main.cpp. The library provides two basic functions to build the suffix array over a text.

// 32 bit version.
uint8_t divsufsort(const uint8_t *T, int32_t *SA, int32_t n);
// 64 bit version.
uint8_t divsufsort(const uint8_t *T, int64_t *SA, int64_t n);

To use the library include the header divsufsort.h, link against the library divsufsort and libprange.

Benchmarks

See Benchmarks page for details.