Parallel-LZ77

Parallel-LZ77 copied to clipboard

This code is part of the paper "Practical Parallel Lempel-Ziv Factorization", by Julian Shun and Fuyao Zhao, presented at Data Compression Conference (DCC) 2013.

The code can be compiled with the following compilers by defining the respective environment variables. Make sure at most one of the following variables are set.

The Cilk Plus compiler in g++ version 4.8.0 (define CILK). The Cilk Plus compiler in icpc (define CILKP). OpenMP (define OPENMP). Sequential g++ (define none of the above variables).

To compile, simply run "make". See the Makefile for more details.

Our parallel code can be run by executing plz77_1 with the following arguments: -p : Number of processors to use (default is 1) -i : Input file name -f : The variant of the parallel algorithm to use. "0" corresponds to PLZ1, "1" correponds to PLZ2, and "2" corresponds to PLZ3. PLZ3 is the fastest of the three; see the paaper for more details.

Our sequential code can be run by executing lz77_1. The arguments are the same as above, except you do not need to set the number of processors.

The algorithm is run for 4 iterations by default (the last 3 iterations are usually faster due to cache effects). This can be modified in test.h.