原位可视化研究与应用进展

Advances in Research and Application of In-situ Visualization

科学可视化是分析数值模拟数据的最有效手段。随着高性能计算机运算能力的不断提升,数值模拟问题的规模迅速增长,但IO带宽、存储系统容量和后处理的可视化模式很难满足超大规模并行数值模拟数据的可视化分析需求。原位可视化在运行数值模拟任务的高性能计算机系统上进行可视化计算和绘制,克服IO和存储容量的限制,实现了在线数据可视化。从数据组织管理、并行可视化算法、交互、集成框架与软件、应用案例等方面介绍了原位可视化近年研究和应用进展,对原位可视化面临的技术挑战以及未来发展趋势进行了分析。

Scientific-Visualization is the most effective means of analyzing numerical simulation data. With the continuous improvement of the high performance computing technology, numerical simulation scale is growing rapidly. It is very difficult to analyze large scale parallel numerical simulation data efficiently due to the limits of IO bandwidth, storage system capacity and post processing pattern of visualization. In-situ visualization utilized the same computer system that ran the simulation tasks to operate visual computing and rendering, so it could overcome the limits of IO and storage system. The advances of in-situ visualization in recent years were reviewed, including data management, large-scale parallel visualization algorithm, interaction, application framework, visualization software and some use cases. The challenges and trends of in-situ visualization were discussed.