Parallel LZMA Compression on the GPU

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering

Final Year Thesis Oral Defense

Title: "Parallel LZMA Compression on the GPU"

by

LIN Geng

Abstract:

7-Zip is a popular compression program. Its core algorithm, LZMA 
(Lempel-Ziv-Markov chain Algorithm), is the most time-consuming component 
in the computation. To speed up 7-Zip, we propose to utilize the GPU in 
LZMA. Specifically, we design a modularized pipeline to separate the two 
components of LZMA - the match finder and the range coder, to allow 
parallelization between the two components on the CPU and the GPU 
respectively. Furthermore, we parallelize the match finder algorithm on 
the GPU and integrate the entire pipeline into 7-Zip. Our evaluation on 
our implementation in comparison with the original 7-Zip software has 
shown promising results on both compression speed and ratio.

		
Date            : 27 April 2018 (Friday)

Time            : 11:00 - 11:40

Venue           : Room 1505 (near lifts 25/26), HKUST

Advisor         : Dr. LUO Qiong

2nd Reader      : Dr. WANG Wei