Detail

Pluggable aspect-oriented support for parallel computing

Current trend towards multi-core systems requires a massive migration of
software from traditional platforms to multi-processors systems. In
traditional environments, parallelization issues are mixed with domain
specific code and scattered throughout the system’s code. The resulting
lack of modularity and reusability significantly raises development and
maintenance costs.
This talk presents a new approach to develop parallel applications based on
Aspect-oriented programming (AOP). AOP can better localize parallelization
issues, improve reusability and attain (un)plugabble parallelizations.
This is a prerequisite to develop software that covers a wide range of
platforms, from multi-core systems to computational Grids. We propose an
incremental approach to develop aspect-oriented parallel applications, in
which features related to parallelization are incrementally developed and
composed, in order to build the combination of features suitable for each
specific platform. Our approach is supported by an AspectJ framework,
based on reusable aspects. In addition, an aspect-oriented domain specific
language is under development, whose aim is to simplify development of
applications based on the framework.