Efficient Data Compression using CUDA programming in GPU

Report Number:  TR-GWU-CS-2010-003
Submit Date:  2010-11-21
Author:  Ganapathy Mani

The aim of this project is to explore and test the new potential performance improvement that can be achieved by the use of GPU processing architecture for several different types of compression algorithms. The compression algorithms are the choice of focus on the data parallelism on the GPU device. The specific algorithms ported to the CUDA architecture included lossless data compression techniques, Vorbis Audio Encoding and JPEG. The lossless data compression has been chosen as being distinctly unique from the lossy compression techniques like Vorbis and libjpeg. JPEG was considered to be different from audio in that the encoding concepts and significantly enough to contain divergent opportunities for exploitation of vector processing and data level parallelism.