Conjunct application of model-based and code-based optimization to derive efficient parallel implementations from SDL

Protocol implementations automatically derived from formal descriptions often do not fulfill the performance requirements of high speed applications. On the other hand, automated implementation techniques may bring remarkable benefits for the quality of implementations. We present an approach to derive efficient implementations from SDL specifications. Our approach is based on two concepts, the activity thread model and the early and systematic integration of performance aspects. First, we analyze in which cases activity threads can be derived from SDL specifications and how this can be done. Second, we describe how performance aspects of the application can be integrated in the development process for high speed protocols. Thus, implementation strategies adapted to the specific needs of the application can be applied