面向自适应结构网格应用的并行可视分析.pdfVIP

面向自适应结构网格应用的并行可视分析 肖丽 艾志玮 王弘堃 曹小林 （北京应用物理与计算数学研究所，北京 100094 ） xiaoli@ Parallel Visual Analysis for Structured Adaptive Mesh Refinement Data Xiao Li, Ai Zhiwei , Wang Hongkun, Cao Xiaolin (Institute of Applied Physics and Computational Mathematics ,Beijing 100094,China) Abstract: In numerical simulations, SAMR(Structured Adaptive Mesh Refinement)-based techniques adaptively refine the domain space of a simulation, both spatially and temporally, into a hierarchy of nested, sequentially refined grids. These strategies are computationally efficient and provide significant storage benefits, such that SAMR is a highly effective simulation method for spanning a large range of spatiotemporal scales. But the dynamic aspects of the grid hierarchies pose significant challenges for visualization methods. Currently, the SAMR data produced by scientific simulations on hundreds of cores possess the following characteristics: their structure is very complex and the scale of the data is very large. For parallel visual analyzing such data, two key techniques are presented in this paper. First, a patch managing structure based on the lattice technology is designed in order to adapt to the complex SAMR data. Secondly, data communications are organized effectively based on adjacent relationships between mesh patches in order to parallel visualization of large scale data. By integrating the two key techniques into the large scale visualization system, the interactive parallel visual analysis for the data with hundreds of millions of mesh zones is realized with 72 cores in the visualization cluster, and the interaction time is less than 2 seconds. Key words: structured adaptive mesh refinement, parallel visual analysis, visualization cluster, lattice technology , communication algorithm 摘 要 自适应结构网格应用程序在数百上千核上模拟输出的结果具有数据