ANDRESTA: An Automated NoC-Based Design Flow for Real-Time Streaming Applications

摘要

Nowadays, there is a high demand for systems-on-chip to deploy real-time embedded systems, which remarkably improve the performance of a system by using multi-processor architectures. However, there are many issues related to this area including, the time-consuming development process, lack of detailed knowledge about hardware behavior and implementation, guarantee of real-time constraints, and most importantly, solving a large design space exploration problem efficiently. This paper proposes a fully automated NoC-based design flow for real-time streaming applications on an FPGA, where the system synthesis and platform tuning design spaces are explored simultaneously to find a global optimum, considering the real-time throughput constraint. Finally, a real-world case study is conducted to confirm that our design flow can optimally map a dataflow application to the target platform, while satisfying the real-time throughput constraint.