Predictable Application Mapping for Manycore Real-Time and Cyber-Physical Systems

Cyber Physical Systems (CPS) are typically implemented on multicore platforms to handle computational pressure and performance requirements. Complex data flows in those applications cause contention among networked processors, resulting in unpredictable performance. In this work, we explore the impact of application mapping on network contention and predictability. A mapping algorithm focusing on minimizing the number of shared paths while not worsening the total path count is proposed. This algorithm is verified against other recent works over synthetic and realistic traffic patterns. The proposed algorithm considerably improved predictability with 0% shared paths and reasonable latency constraints compared to the other recently proposed algorithms.