Project Details

Program Slicing is a technique of identifying parts (slice) of a program that may influence or be influenced by a given set of program points (slice criteria). This technique has been extensively applied to program understanding. In the Bandera project [7], the application of program slicing was considered to reduce the size of the model extracted from concurrent Java programs for the purpose of program verification via model checking. However, the absence of a configurable and modular program slicer for Java prompted us to build one from scratch, and I volunteered to build the slicer. Most of the proposed slicing techniques [28, 13] were geared towards sequential programs written in procedural languages such as C. In the process of implementing the slicer, I (like many others) found that these techniques either fail or lead to imprecise slices due to aliasing in concurrent programs written in object-oriented languages. Imprecise interference dependence was one of the main reason. Interference dependence [16, 11] is a form of data dependence stemming from reads and writes on same objects that occur in different threads. As the data flow between variables in different threads is dependent on the possibly exponential number of interleavings between threads, mere extensions of solutions to data-dependence calculation prove to be costly. However, as nonescaping objects cannot contribute to interference dependence, information from escape analysis [1, 6] can be used to prune spurious interference dependences, hence, improve the accuracy of slicing. As an experiment, I adapted Ruf’s escape analysis [25] to prune interference dependences. Based on the positive results, I extended the adapted analysis to symbolically track read, write, and lock actions on the objects. This extension pruned more dependences and led to smaller slices. Similarly, Ruf’s analysis was extended to improve the precision of ready dependences, a form on control dependence that arises due to synchronization constructs (alike enter-monitor/exitmonitor/wait/notify in Java) in concurrent systems. This work along with the results was published at the International Conference on Compiler Construction (CC) 2004 [23]. During the development of the slicer, non-terminating systems and modern program structures (exceptions) exposed some “shortcomings” in control dependence as prescribed in earlier literature. I devised alternative definitions to address these “shortcomings” and, as part of the slicer, I implemented the algorithms to calculate control dependence according to alternate definitions. The definitions were unique as they could be applied directly to both reducible and irreducible control