Deriving High-Performance Real-Time Multicore Systems Based on Simulink Applications

MATLAB/Simulink is commonly used for designing model-based dynamic embedded systems. Throughout Real-Time Workshop toolkits, it can generate C or C++ programs for various target platforms, which is useful to develop embedded systems. However, the current toolkits generate only single programs, so that it does not leverage multicore technology for performance improvement. In this paper, we provide a new automatic code generation scheme for multicore real-time systems by inserting user-defined S-Functions for Simulink applications. The proposed scheme uses mailboxes for synchronization among threads in order to reduce the overhead. Users can easily develop multiple subtasks of a Simulink application on multicore systems. We develop the automatic code generation for RTAI real-time systems and evaluate the performance throughout experiments.