| 摘要: |
| 分布式计算环境中基于消息传递机制的分布式共享缓冲区中，Ｃａｃｈｅ效率是算法性能的“瓶颈”．本文在分布式共享缓冲区上实现了一个并行体绘制算法．在数据空间，八叉树快速分类改善了Ｃａｃｈｅ的空间相关性；在图象空间，Ｈｉｌｂｅｒｔ象素遍历方式改善了Ｃａｃｈｅ的时间相关性．在曙光１０００和ＳＧＩ工作站网络上的实验结果都表明，算法的网络数据传送量大大减少，Ｃａｃｈｅ效率明显提高，绘制时间大大缩短． |
| 关键词: 并行体绘制 八叉树 象素遍历 光线追踪 |
| DOI： |
| 分类号: |
| 基金项目:本文研究得到国家自然科学基金资助. |
|
| PARALLEL VOLUME RENDERING ALGORITHM FOR DISTRIBUTED COMPUTING ENVIRONMENT |
|
Yu Shengming,Li Hua,Liu Shenquan
|
| Abstract: |
| This paper presents a distributed shared buffer paradigm based on message passing in distributed computing environment. According to this paradigm, this paper implements a parallel volume rendering algorithm. In order to improve the efficiency of the cache, the parallel algorithm take advantage of octree classification to increase local coherence of the cache in data space. In image space, the parallel algorithm use the Hilbert pixel traversal order to increase the temporal coherence of the cache. The experiments demonstrate good speedup and salability of this approach. |
| Key words: Parallel volume rendering octree pixel traversal ray tracing. |
