Research And Implementation Of A Lossless Compression Algorithm Based On Gpu

With the rapid development of information technology, data is growing rapidly. It is increasingly urgent for effective information storage and transmission due to limited storage and network bandwidth. Data compression is one of the solutions. However, data compression often requires additional expenses, especially the huge amounts of data. Therefore, new high-performance technology is needed to design more efficient and higher quality algorithms to achieve the massive data compression. How to reduce compression time is very important.In recent years, high-performance general computing of GPUs is developing quickly. GPU is far beyond the CPU in floating-point computation, parallel computation and memory bandwidth. Heterogeneous processing platforms of the CPU+GPU have become mainstream parallel solution. This study is to design and optimization of parallel lossless compression algorithm to enhance the compression performance of the algorithm using high parallel computing power of CPU+GPU.In this study, the architecture of the GPU was discussed with detailed description of hardware architecture and the software of the CUDA programming model and the optimization strategy for CUDA program. Then, the base of the compression algorithm was described. Finally, two lossless compression algorithm of GPU was realized:the LZ77and bzip2algorithms. The algorithm could improve the computational throughput by loading compute-intensive part of the compression algorithm on GPU from CPU. And then the optimized design was performed combining the characteristics of parallel architecture of CUDA and algorithm. A detailed description of parallel design and implementation of the CUDA platform based on the lossless compression algorithm was showed. The results showed that the parallel compression was able to achieve good acceleration compared with the performance on CPU.