Timing Performances of Automatically Generated Code Using MDA Approaches

Abstract

To improve usual software development in the C programming language, a model-driven approach seems very promising. Writing generic models, independent of any platform, is attractive and will reduce the work to implement software on multiple platforms. Another very attractive feature of the MDA approach is complete and automatic code generation. Within the very concrete context of an industrial example, this article aims to answer two questions: “To what extent models written during development are platform independent?” and “To what extent does automatic code generation reduce timing performances of applications?”. To answer these questions, the study focuses on the sequence of models needed to transform a system specification into an implementation, and then on the timing performances of the code generated by CASE tools supplying different level of services.

This work was partly supported by Schneider Electric under grant 008686.