核外算法在可视化中的应用

Application of Out-of-Core Algorithm on Visualization

大规模数据的可视化一直是科学计算可视化中极具挑战性的一个研究领域,因为生成的计算数据常常太大不能一次性完全输入到计算机的内存中,以至于数据在快速内存和相对慢速的外存(比如磁盘)之间的输入和输出成为计算的一个瓶颈。在没有大内存高性能并行机但内存相对太小的PC机不能一次性装载可视化数据的情况下,提出了核外算法和可视化技术有机结合在PC机上大量数据的可视化,具体实现了核外算法在PC上对大量数据的局部分块显示、全局粗糙显示、局部分块和全局粗糙相结合显示、数据点的查找。实验验证了核外算法在PC机上对大量不可一次性导入内存的数据显示和数据查找上是一种简便、省时、高效的方法。

Large- scale data visualization is always a challenge problem of the visualization in scientific computing. Due to the massive data produced by large- scale scientific and engineering computing, they are too large to fit into computer~ main memory at one time. As the result, the input/output communica- tion （I/O） between fast internal memory and slower external memory （ such as disk） can be a major per- formance bottleneck. In the condition without superpower computer of large memory, the memory of simple PC is too small to be used to load the visualization data. The combination of out- of- core algorithm and vi- sualization technique was developed in this paper, which realized the local partitioning visualization, over- all coarse visualization, combination of local partitioning and overall coarse visualization, and point see- king as well on simple PC. The experiments demonstrate that the out- of- core algorithm is an easy- used, time- saving and effective way to seek and visualize certain data among the large- scale data which could not be inputted into main memory at one time.