Seamless Integration of FPGA-based MPSoCs into Computing Continuum Infrastructures

The rapid evolution of Edge Processing Systems (EPSs) and the Computing Continuum (CC) has led to the widespread use of heterogeneous computing platforms. In particular, Multi-Processor System on Chip (MPSoC)-based devices that combine general-purpose processors and Field Programmable Gate Array (FPGA) fabric have emerged as a powerful solution for handling both software-defined and computationally-intensive workloads. This paper presents a novel design flow that integrates existing tools and supports commercially available FPGA-based MPSoCs to enable the orchestration of acceleration-oriented applications at the edge within a CC environment. The proposed design flow leverages modular components to build custom Linux distributions, including necessary drivers and packages for containerized workload management. By replacing vendor-specific solutions (i.e., AMD’s Ubuntu and PetaLinux) with the design of a lightweight, purpose-built Linux distribution, our approach provides a unified and flexible cloud-to-edge development environment. Experimental results show that such a customized distribution not only maintains timing and power/energy consumption comparable to vendor-optimized setups, but can also reduce CPU and RAM usage, SD memory footprint, and container size.