Return to search

Constraint-Based Thread-Modular Abstract Interpretation

In this dissertation, I present a set of novel constraint-based thread-modular abstract-interpretation techniques for static analysis of concurrent programs. Specifically, I integrate a lightweight constraint solver into a thread-modular abstract interpreter to reason about inter-thread interference more accurately. Then, I show how to extend the new analyzer from programs running on sequentially consistent memory to programs running on weak memory. Finally, I show how to perform incremental abstract interpretation, with and without the previously mentioned constraint solver, by analyzing only regions of the program impacted by a program modification. I also demonstrate, through experiments, that these new constraint-based static analyzers are significantly more accurate than prior abstract interpretation-based static analyzers, with lower runtime overhead, and that the incremental technique can drastically reduce runtime overhead in the presence of small program modifications. / Ph. D.
Date25 July 2018
CreatorsKusano, Markus Jan Urban
ContributorsElectrical and Computer Engineering, Wang, Chao, Hsiao, Michael S., Zeng, Haibo, Lee, Dongyoon, Schaumont, Patrick R.
PublisherVirginia Tech
Source SetsVirginia Tech Theses and Dissertation
Detected LanguageEnglish
FormatETD, application/pdf
RightsIn Copyright,

Page generated in 0.0023 seconds