An Annotation-Based Framework for Parallel Computing

This paper presents a programming language for parallel computing based on code annotations. It has similar goals and philosophy as OpenMP but it is more tightly coupled to the object oriented paradigm. We include annotations for most common concurrency patterns and mechanisms, namely, one-way, futures, barriers, reads/writers and thread-local. Our current prototype is implemented using Java 5 annotations and AspectJ and provides a feasible and efficient alternative to the Java thread model