On Modeling the RapidIO communication link using the AAA Methodology
Résumé
Embedded real-time applications in communication systems have significant time constraints, thus requiring high processing power. One time manual scheduling developed for single-processor applications is not suited for multiprocessor architectures: manual data transfers and synchronizations quickly become very complex, leading to wasted time and potential deadlocks. We aim to develop a fast and automatic prototyping process optimized for parallel architectures particularly suited to static executives. The process is based on AAA methodology which improves algorithm implementation on multiprocessor architectures by finding the best match between algorithms and architectures. This paper presents an automatic code generation approach for architectures using the RapidIO media. A new communication model which optimizes data transfers in the generated code is depicted. Finally, we demonstrate the effectiveness of the approach with a Locally Adaptive Resolution (LAR) coding (PC-driven) application deployed on a multi-DSP platform.