Design And Implementation Of Accelerating Compression Subsystem Based On OpenACC

The processing and preservation of massive data is an unsolved problem when the era of big data is coming. In order to balance the performance and cost, the massive data must be compressed before it to be conserved. How to compress the massive data to decrease the time cost is the issue of the field of data management currently.This dissertation applied the GPU parallel computing technique. Considering the balance of algorithm complexity and compression speedup ratio, this dissertation chose the parallel which based on the Open ACC and speed up the original compression subsystem. Improving the processing speed of data compression, decreasing the compression time and improving the compression processing efficiency efficiently.This dissertation introduced the GPU programming technique and related programming construction firstly. Especially, this dissertation introduced the Open ACC technique detail.Including the implementation module and storage module of Open ACC, main computing component, guidance and clause of emphasis et al. Moreover, this dissertation analyzed the original compression subsystem detail which need to be improved. For the construction of system and compression algorithm, especially the source code of Huffman compression algorithm, this dissertation did a range of analysis and description. Based on these, Open ACC speeded up and optimized the Huffman compression in the first stage and the second stage.The progress of optimization and the improvement of code parallel optimization in every stage were introduced detail. Finally, the improved result had been compared in every stage through tests. This dissertation also compared and analyzed the influence of speedup by different hardware configuration.Experiments shows that:1) The higher configuration of graphic card GPU, the more obvious improvement of speedup.2) The configuration of CPU have an influence on speedup, the more strong CPU computing ability, the more obvious development of multinuclear and hyper-threading, the less obvious improvement of GPU speedup ratio.3) Since the compression algorithm of Huffman encode is sensitive to the file size,when the compression file become bigger, the speed of compression is related faster. And the data compression rate increased apparently.4) The improved Huffman compression algorithm based on Open Acc has no influence on data compression rate. Before and after compressed, the numbers of characters were not changed.5) The improved compression subsystem based on Open ACC has higher speedup ratio.The average compression speedup ratio can reach 8.58 X.In conclusion, related to original compression system, the parallel speedup optimization based on Open ACC can reach 8.58 X which is the average compression speedup ratio. It is a satisfied result. The improved compression subsystem can efficiently improve the compression efficiency which can fulfill the requirement of real time database processing system.Meanwhile, this dissertation has a big reference significance to related GPU parallel computing research.