分布式并行绘制系统中几何指令流压缩的研究与实现

Research and Implementation of Geometry Instruction Stream Compression in Distributed Parallel Rendering System

对分布式并行绘制系统的几何指令流进行压缩能缓解网络带宽瓶颈 .对操作码使用 L ZW算法 ,对法向量使用球面对称网格剖分算法 ,对颜色和位置数据使用 DPCM型预测编码算法 ,根据位置数据的特殊性 ,使用了 4类预测器和自适应量化算法 .对几何指令流组合使用多种压缩算法取得了良好的效果 ,在几何模型质量基本没有损失的情况下 ,指令平均长度压缩到原来的 1/ 3左右 ,执行速度达到了 4 0 0指令 /

Compression of geometry instruction stream in distributed parallel rendering systems can alleviate network bandwidth bottleneck. LZW method is used in op codes compression. A sphere symmetric grid method is used for normal compression. Prediction delta coding with DPCM model is used in position and color compression. 4 kinds of predictions and adaptive quantization are used in position compression. Multiple approaches work well in geometry instruction stream compression. The average length of instructions is compressed to 1/3 of the original one with almost no loss of quality and the executing speed is better than 400 instructions/ms.