In recent years, manufacturers of processors are focusing on parallel architectures in order to increase performance.This shift in hardware evolution is provoking a fundamental turn towards concurrency in software development.Unfortunately, developing concurrent programs which are correct and efficient is hard, as the underlying programming model is much more complex than it is for simple sequential programs.The goal of this research is to study and to develop program analysis to support and improve concurrent software development in declarative languages. The characteristics of these languages offer opportunities, as they are good candidates for building concurrent applications while their simple and uniform data representation, together with a small and formally defined semantics makes them well-adapted to automatic program analysis techniques. In our work, we focus primarily on developing static analysis techniques for detecting race conditions at the application level in Mercury and Prolog programs. A further step is to derive (semi-) automatically the location and the granularity of the critical sections using a data-centric approach.
展开▼
