A demand-driven solver for constraint-based control flow analysis

This thesis develops a demand driven solver for constraint based control flow analysis. Our approach is modular, flow-sensitive and scaling. It allows to efficiently construct the interprocedural control flow graph (ICFG) for object-oriented languages. The analysis is based on the formal semantics of a Java-like language. It is proven to be correct with respect to this semantics. The base algorithms are given and we evaluate the applicability of our approach to real world programs. Construction of the ICFG is a key problem for the translation and optimization of object-oriented languages. The more accurate these graphs are, the more applicable, precise and faster are these analyses. While most present techniques are flow-insensitive, we present a flow-sensitive approach that is scalable. The analysis result is twofold. On the one hand, it allows to identify and delete uncallable methods, thus minimizing the program’s footprint. This is especially important in the setting of embedded systems, where usually memory resources are quite expensive. On the other hand, the interprocedural control flow graph generated is much more precise than those generated with present techniques. This allows for increased accuracy when performing data flow analyses. Also this aspect is important for embedded systems, as more precise analyses allow the compiler to apply better optimizations, resulting in smaller and/or faster programs. Experimental results are given that demonstrate the applicability and scalability of the analysis.