EAGER: Improving Robustness of Numerical Software

EAGER：提高数值软件的鲁棒性

• 批准号: 1349666
• 负责人: Tao Xie
• 金额: $ 10万
• 依托单位: University of Illinois at Urbana-Champaign
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2013
• 资助国家: 美国
• 起止时间: 2013-10-01 至 2016-03-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1349666&HistoricalAwards=false
• 关键词: EAGER; Improving; Robustness; Numerical; Software

Numerical software is increasingly playing a critical role in society, and thus numerical errors in software can have disastrous consequences. However, it is very difficult to test numerical software, and numerical errors are hard to detect because they may not necessarily result in system crashes. There is a strong need for effective and practical techniques or tools to detect and prevent such errors and improve robustness of numerical software.This project explores novel and practical techniques for testing and analyzing numerical software to detect numerical errors in order to improve its robustness. In particular, this project carries out a set of preliminary research tasks to demonstrate the feasibility of the techniques, including developing an initial public repository of numeric constraints and exploring techniques to guide, optimize, and use parallel path exploration in symbolic execution. The research in this project enhances the infrastructure for teaching and research by providing open source tools and data sets for use by students and practitioners, and for enhancement by other researchers.

数值软件在社会中扮演着越来越重要的角色，因此软件中的数值错误可能会造成灾难性的后果。然而，测试数值软件是非常困难的，并且数值错误很难检测到，因为它们不一定会导致系统崩溃。迫切需要有效和实用的技术或工具来检测和防止此类错误，并提高数值软件的鲁棒性。本项目探索新的和实用的技术来测试和分析数值软件，以检测数值误差，以提高其稳健性。特别是，该项目开展了一系列初步研究任务，以证明这些技术的可行性，包括开发一个数字约束的初始公共存储库，以及探索在符号执行中指导、优化和使用并行路径探索的技术。该项目的研究通过为学生和从业者提供开源工具和数据集，并为其他研究人员提供改进，从而增强了教学和研究的基础设施。