AuPertProc

Using graphical tools for model-driven design of embedded control software, in  general, leads to executable code that is not optimal w.r.t. performance and processor use, because structuring models while editing is done from a modeling point of view.  This research uses graph-theoretical approaches to design algorithms, which optimize a structured model towards optimal execution for real-time control. These algorithms will be implemented as part of the compiler of the graphical tool chain for embedded control software development, TERRA and LUNA.

Results 2013

Theory has been developed to proof that a performance gain can be achieved by combining parallel processes. Performance and memory occupation are the main themes here. Graphically formulated embedded control software processes consist of many short processes, whereby context switching between execution of processes takes a considerable amount of processor time. The transformation process towards a more effective graph may not cost more memory than available. Using the developed graph-theoretical concepts, we showed that the amount of context switches can be reduced by combining synchronizing processes. The next step in the work is to further develop these transformations, and produce a first prototype implementation as one of the stages of the compiler.

Project lead

no picture available
dr.ir. Jan Broenink