Context Sensitivity without Contexts: A Cut-Shortcut Approach to Fast and Precise(…)
summary
Proposes Cut-Shortcut that cuts PFG edges causing merged flows and adds direct shortcut edges to simulate context sensitivity without cloning; implements two versions (Datalog for Doop, Java for Tai-e), encodes three program patterns (field access, container access, local flow), and runs on 10+10 large Java apps from prior literature. Evaluation shows scaling on all tested targets where selective CS and some context-sensitive variants time out, running faster than context-insensitive on most benchmarks and delivering precision comparable to scalable context sensitivity.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:05Pointer analysis basics
- 2:48Contact vs context sensitivity
- 4:14Selective limits
- 5:17Cut-Shortcut approach
- 6:46Formalizing Cut-Shortcut
- 7:53Nebulous paths
- 9:20Implementation challenges
- 10:45Program patterns
- 11:29Evaluation setup
- 12:56Recall and scalability
- 13:38Performance vs precision
- 14:43Conclusion
- 15:27Q&A
related talks
09mDynamic Dispatch of Context-Sensitive Optimizations
29mA Flow-Insensitive-Complete Program Representation
18mBoosting the Performance of Alias-Aware IFDS Analysis with CFL-based Environment(…)
01h 05mControl Flow Obfuscation for Featherweight Java + Efficient Fair Conjunction
14mDataflow-Based Pruning for Speeding up Superoptimization
01h 02mLogic, Optimization, and Constraint Programming: A Fruitful Collaboration
13mHash Consed Points-To Sets
21mFlow-Analysis-Based Closure Optimization