A Distributed Dynamic Aspect Machine for Scientific Software Development

This paper proposes the use of an event-based dynamic AOP machine as an infrastructure for interactive development of high performance scientific software. Advice codes in the proposed approach are similar to mobile agents that execute on distributed computational nodes. The key ideas underlying this approach are multi-level separation of parallelisation concerns and event-driven dynamic join points. The primary aim of the research is to use the AOP paradigm to improve productivity for scientific software development; the dynamic AOP machine is also expected to be further developed as an interactive computational grid.