关键词: Chromium, ZLib, 哈夫曼, 几何数据压缩, 分布式并行绘制系统

Abstract: The deficient network bandwidth has severely curtailed the rendering speed of large geometric scenes in parallel graphics rendering systems such as Chromium. Through lossless compression of geometric data in the network transmission, an efficient method which alleviates the network burden is proposed. This method can easily be used to implement different algorithms for a specific geometry data compression. We have used ZLib and Huffman compression algorithms in Chromium, and tested the speedup and compression ratios of 10 kinds of OpenGL applications, and the parallel performance in the four configurations. The speed of the tested programs that uses ZLib has been improved to varying degrees and the maximum raising is 3 times; the average ratio of data compression is above 5, up to the maximum 30; the speedup ratio of parallel rendering on a single server is the highest. The overall performance of the ZLib algorithm is good, which can effectively reduce the network traffic.

Key words: chromium, zlib, huffman, geometric data compression, distributed parallel rendering system