SHF: Small: New Frontiers in Constraint-Based Program Analysis

SHF：小型：基于约束的程序分析的新领域

• 批准号: 1526270
• 负责人: Mayur Naik
• 金额: $ 45万
• 依托单位: Georgia Tech Research Corporation
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2015
• 资助国家: 美国
• 起止时间: 2015-09-01 至 2017-04-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1526270&HistoricalAwards=false
• 关键词: SHF; Small; New; Frontiers; Constraint

Title: SHF:Small:New Frontiers in Constraint-Based Program AnalysisConstraint-based analysis is a popular approach to program analysis: it allows to separate analysis specification from analysis implementation, it enables sophisticated implementations by leveraging advances in off-the-shelf solvers, and it provides natural program specifications as constraints. This project proposes Dominoes, a framework that extends the benefits of constraint-based analysis by enabling automatic synthesis of common and emerging use-cases of program analyses, such as finding good abstractions, analyzing incomplete programs, and incorporating user feedback. The intellectual merit of this project is to fundamentally advance demand-driven, compositional, and learning-based analysis techniques. By automatically synthesizing use-cases once and for all, Dominoes amplifies the traditional benefits of constraint-based analysis, liberating analysis designers from having to re-implement those use-cases for their analyses. The project's broader significance and importance lies in enhancing the applicability and usefulness of program analyses by making them more automated, scalable, and flexible. Artifacts embodying these analyses will improve software quality in aspects of reliability, security, performance, and energy efficiency. Dominoes will also improve the productivity of analysis users by allowing them to adapt analyses to their feedback.Dominoes automatically synthesizes implementations of use-cases for any program analysis expressed in Datalog, a popular declarative logic programming language. Existing constraint-based analysis frameworks predominantly focus on solving hard constraints, whereas Dominoes also accommodates soft constraints that arise naturally in diverse use-cases of program analysis, e.g., to model various tradeoffs, intuitions of analysis users, and missing program specifications. The versatility of Dominoes is demonstrated by applying it to three important use-cases: client-driven analysis, summary-based analysis, and user-guided analysis. Despite their diversity, all three use-cases entail solving instances of the maximum satisfiability (MaxSAT) problem, which consists of a combination of hard (inviolable) constraints and soft (violable) constraints. Solving such mixed constraints is not only computationally hard but also poses the problem of specifying weights or confidences of soft constraints. Dominoes develops MaxSAT optimizations comprising demand-driven, compositional, and learning-based methods that are general and independent of any analysis, use-case, or solver, and aim to scale to instances well beyond the reach of existing MaxSAT solvers.

标题：SHF：Small：基于约束的程序分析的新前沿基于约束的分析是一种流行的程序分析方法：它允许将分析规范与分析实现分开，它通过利用现成求解器的进步来实现复杂的实现，并且它提供了自然的程序规范作为约束。 该项目提出了多米诺骨牌，这是一个框架，通过自动合成程序分析的常见和新兴用例（例如寻找良好的抽象、分析不完整的程序和合并用户反馈）来扩展基于约束的分析的优势。 该项目的智力价值是从根本上推进需求驱动、组合和基于学习的分析技术。通过一劳永逸地自动综合用例，Dominoes 放大了基于约束的分析的传统优势，使分析设计者不必为他们的分析重新实现这些用例。 该项目更广泛的意义和重要性在于通过使程序分析更加自动化、可扩展和灵活来增强程序分析的适用性和有用性。 体现这些分析的工件将在可靠性、安全性、性能和能源效率方面提高软件质量。 Dominoes 还将允许分析用户根据反馈调整分析，从而提高他们的工作效率。Dominoes 会自动综合以 Datalog（一种流行的声明性逻辑编程语言）表达的任何程序分析的用例实现。 现有的基于约束的分析框架主要侧重于解决硬约束，而多米诺骨牌还适应程序分析的不同用例中自然出现的软约束，例如，对各种权衡、分析用户的直觉和缺失的程序规范进行建模。 多米诺骨牌的多功能性通过将其应用于三个重要用例来证明：客户端驱动的分析、基于摘要的分析和用户引导的分析。 尽管存在多样性，所有三个用例都需要解决最大可满足性 (MaxSAT) 问题的实例，该问题由硬（不可违反）约束和软（可违反）约束的组合组成。 解决这种混合约束不仅计算困难，而且还带来了指定软约束的权重或置信度的问题。 Dominoes 开发的 MaxSAT 优化包括需求驱动、组合和基于学习的方法，这些方法是通用的且独立于任何分析、用例或求解器，旨在扩展到远远超出现有 MaxSAT 求解器范围的实例。