Emulation platform for an adaptive NoC-based MPSoC architecture dedicated to spectral imaging for art authentication

Abstract : Real time spectral imaging applications are data and communication intensive algorithms. SoC (System On Chip) architecture using NoC (Network On Chip) communication is the most appropriate solution as it brings flexibility, performances and immediate reuse of IPs cores. Tuning the SoC-based NoC architecture for a dedicated algorithm is a complex task as the design space is huge and experimentations take time. In this paper we propose an emulation platform for the evaluation and exploration of performances of the architecture on FPGA. The emulation platform is based on the NoC and parameterized emulation blocks. Emulating the communication inside the FPGA gives precise timing performances compared to simulation. The experiments are fast and reliable without hardware requirements. The designer can evaluate the timing of the application with a restricted number of implementations in a short time to experimentation. The evaluation is illustrated for a spectral imaging algorithm dedicated to art authentication on the architecture containing MIPS processors and the Hermes NoC.