A Self-Compilation Flow Demo on FOS – The FPGA Operating System

摘要

With the introduction of Zynq UltraScale+ MPSoCs equipped with powerful 64-bit ARM CPUs along with a 16nm UltraScale+ FPGA fabric in the same die, we can now build full hardware-software programmable systems and configure the FPGA with accelerators, as needed. Traditionally, accelerators had been developed offline using powerful servers running heavy lifting CAD toolchains. In this demo, we show a self-compilation system supporting a user-friendly Jupyter Notebook GUI and multi-tenancy use of the FPGA for educational purposes. From a user perspective, this system compiles accelerators at run-time directly on the ARM CPU without any involvement of the vendor tools. The final bitstreams then execute on the FPGA fabric using PR.