Lock-sensitive Interference Analysis for Java: Combining Program Dependence Graphs with Dynamic Pushdown Networks

Graf Jürgen, Hecker Martin, Mohr Martin, Nordhoff Benedikt

Abstract

We combine the static analysis techniques of Program Dependence Graphs (PDG) and Dynamic Pushdown Networks (DPN) to improve the precision of interference analysis for multithreaded Java programs. PDGs soundly approximate possible dependence between program points in sequential programs through data and control dependence edges. In a concurrent setting a third category of so-called interference edges captures the potential interferences between memory accesses in different threads. DPNs model concurrent programs with recursive procedures, dynamic thread creation and nested locking. We use a lock-sensitive analysis based on DPNs to remove spurious interference edges, and apply the results to information flow control.