Efficient Flow-Sensitive Pointer Analysis on Full-Sparse Memory SSA
Year: 2015
Type: dissertation
Abstract: Pointer analysis is a fundamental research topic in computer science. It statically determines the potential runtime targets of pointers. Many clients benefit from this information, including compiler optimization, bug detection, security analysis and change impact analysis, etc. As a key dimension in pointer analysis, flow-sensitivity improves its precision by considering program execution order. Ideally, flow-sensitive pointer analysis should be performed by analyzing each program path independently. However, even ignoring the branch conditions, such solution remains intractable and extremely expensive for whole program analysis due to potentially unbounded program paths. Finding the right balance between precision and efficiency in flow-sensitive pointer analysis lies at the heart of pointer analysis. In this thesis, we first introduce an efficient inter-procedural full-sparse memory SSA construction algorithm. Then we improve flow-sensitive pointer analysis based on the memory SSA with the following two contributions: First, Selfs, a region-based selective flow-sensitive pointer analysis, is proposed to allow precision and efficiency trade-offs to be made according to region partitioning. By maintaining flow-sensitivity between regions instead of statements, Selfs is faster than the state-of-the-art full-sparse flow-sensitive analysis while achieving the same precision when used for alias queries. Second, Spas explores the intra-procedural path correlations on top of sparse flow-sensitive and context-sensitive pointer analysis. By using binary decision diagrams to represent the compact path conditions, Spas improves the precision of pointer analysis while only introducing a small overhead.
Source: UNSWorks (UNSW Sydney)
Author: Ye, Sen
Language: English
Subfield: Software
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6927294225