Study on Massively Parallel Simulations and Visualizations

大规模并行仿真与可视化研究

• 批准号: 07308063
• 负责人: HORIGUCHI Susumu
• 金额: $ 0.9万
• 依托单位: JAPAN ADVANCED INSTITUTE OF SCIENCE AND TECHNOLOGY,Hokuriku
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (B)
• 财政年份: 1995
• 资助国家: 日本
• 起止时间: 1995 至 1996
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-07308063/
• 关键词: Massively parallel Computer; Numerical Simulations; 3D Image System; Visualization; High-Speed Simulations; 並列計算機シミュレーション

Computer simulations have been important and are frequently used in advanced science and technology. Advanced simulations require large computation times and huge memory capacities of computers. Massively parallel computers have been attractive for advanced simulations instead of supercomputers in this decade. Parallelized simulations, however, have not been studied sufficiently and the visualizations of simulation results also have been more important recently.The study on massively parallel simulations and visualizations focuses on the advanced scientific field such as physics, Chemistry, Material Science, Fluid Dynamics and Neural Networks, We implemented many parallelized simulations on massively parallel computers : CM-5, ncube2 and Parsytec GC and demonstrated visualization of simulation results by Computer Graphics.To achieve high-speed parallel simulations, we developed the dynamics load balancing method for Molecular Dynamics on CM-5, the global communication method for room illlumination simulation by Radiosity on Parsytec GC,and the packed message passing method for neural network on ncube2. We also snow the effective visualizations of huge simulation data using 3D computer graphics.

计算机模拟一直是重要的，并经常用于先进的科学和技术。高级模拟需要大量的计算时间和巨大的计算机内存容量。在这十年中，大规模并行计算机已经成为高级模拟而不是超级计算机的吸引力。然而，并行化仿真的研究还不够充分，仿真结果的可视化也越来越受到重视。大规模并行仿真和可视化的研究集中在物理、化学、材料科学、流体力学和神经网络等前沿科学领域，我们在大规模并行计算机CM-5、ncube2和Parsytec GC上实现了许多并行仿真，并通过计算机图形学展示了仿真结果的可视化。为了实现高速并行模拟，我们在CM-5上开发了分子动力学的动态负载平衡方法，在Parsytec GC上开发了Radiosity用于室内照明模拟的全局通信方法，在ncube2上开发了神经网络的打包消息传递方法。我们还使用3D计算机图形学对大量仿真数据进行了有效的可视化。

项目成果

K.Shida, K.Ohno, M.Kimira, Y.Kawazoe: ""Parallelized Simulation of Complicated Polymer Structures and Its Efficiency"" IEICE Trans.of Information and Systems. (to be published). (1997)

K.Shida、K.Ohno、M.Kimira、Y.Kawazoe：“复杂聚合物结构的并行化模拟及其效率”IEICE Trans.of 信息和系统。

K.Yamamori, T.Abe, S.Horiguchi: ""Parallel Learning Schemes of Neural Networks on Massively Parallel Computer"" Trans.of IEICE D-II. Vol.J80-D-II,No.1. 350-353 (1997)

K.Yamamori、T.Abe、S.Horiguchi：“大规模并行计算机上神经网络的并行学习方案”Trans.of IEICE D-II。

H.Iwata, J.H.Kim and S.Horiguchi: ""The Automatic Texture Mapping Technique for 3-Dimensional Visualization"" Proceeding of International Conference on Multimedia Japan'96 Yokohama, Japan. 94-101 (1996)

H.Iwata、J.H.Kim 和 S.Horiguchi：“3 维可视化的自动纹理映射技术”日本多媒体国际会议 96 横滨，日本。

加藤聡,堀口進: "″下部側頭葉皮質IT野における神経細胞結合モデル″," JAIST Research Report. IS-RR-96-34P. 1-20 (1996)

Satoshi Kato，Susumu Horiguchi：“下颞叶皮层 IT 区域的神经元连接模型”，JAIST 研究报告 1-20 (1996)。

R.Hayashi and S.Horiguchi: ""Parallelized Simulation of Molecular Dynamics by Domain Decomposition Strategy"" submitted for parallel and Distributed Simulations at World Congress on System Simulations. (1997)

R.Hayashi 和 S.Horiguchi：“通过域分解策略对分子动力学进行并行化模拟”，提交给世界系统模拟大会上的并行和分布式模拟。