Abstract

Comprehending distributed systems is a challenging task because of interdependency and non-determinability that exist in distributed systems. Program slicing, as a well-know decomposition and reduction technique, has been extended to assist during the comprehension of distributed application source code. Dynamic predicate slicing is a relatively new slicing technique that adopts the notion of global predicates as slicing criteria for distributed message passing programs. Dynamic predicate slicing focuses on identifying all these states during an execution of a message passing program, in which a particular predicate might be changed. In this research, a Coarse-Grained dynamic predicate slicing algorithm is implemented by using instrumentation techniques to insert extra instructions into applications for collecting and analyzing run-time information during executions. Dynamic predicated slicing is accomplished by multi-thread parallel computation utilizing both static information and dynamic information. An initial case study is presented to validate the applicability of the approach and to explore the overhead associated with collecting the dynamic information and the slice computation.