SIAM-CSE19-MS2-Coutinho.pdf

CSE simulations in high-performance computers are still costly. They often involve the selection of many computational parameters and options. The set-up of such parameters is usually a trial-and-error process even for experienced users. In this talk, we will show how to extend in-situ visualization techniques with in-transit data analysis to provide information to help control the simulations at runtime. Often, by only observing a region of interest, an experienced analyst can infer that something is not going well, deciding to stop it or change parameters. However, to do that, visual information should be complemented with information regarding the evolution of quantities of interest. We use for the simulations the libMesh library, which provides a platform for parallel, adaptive, multiphysics finite element computations. We discuss the integration of libMesh with in-situ visualization and in-transit data analysis tools. We present a parallel performance analysis showing that the overhead for both in-situ visualization and in-transit data analysis is negligible. Our tools enable monitoring the quantities of interest at runtime and steer the simulation based on the solver convergence or other data and visual information. The data analysis tool registers the provenance of the simulation data for reproducibility, including registering the runtime changes on simulation parameters.在高性能计算机上开展的计算科学与工程（CSE）模拟依然成本高昂。此类模拟通常涉及大量计算参数与配置项的选取，即便对于经验丰富的使用者而言，参数设置往往也需要反复试错。本报告将介绍如何结合传输中数据分析（in-transit data analysis）技术拓展原位可视化（in-situ visualization）方法，以在模拟运行过程中提供辅助信息，助力实时管控模拟流程。通常情况下，经验丰富的分析师仅通过观测感兴趣区域，即可推断模拟流程是否出现异常，并据此决定终止模拟或调整参数；然而要实现这一点，视觉信息需要辅以与关注量演化相关的补充信息。本次研究采用libMesh库开展模拟，该库为并行、自适应多物理场有限元计算提供了支撑平台。我们探讨了将libMesh与原位可视化、传输中数据分析工具的集成方案，并开展了并行性能分析，结果显示原位可视化与传输中数据分析的系统开销均可忽略不计。我们开发的工具可在模拟运行过程中实时监测关注量，并基于求解器收敛性或其他数据与视觉信息对模拟流程进行实时管控。该数据分析工具可记录模拟数据的溯源信息以保障可复现性，其中包括模拟运行过程中参数变更的相关记录。