CFU-Playground

CFU Playground is a framework designed for the rapid development and evaluation of enhancements to FPGA-based "soft" processors specifically targeted at improving machine learning inference performance. It abstracts infrastructure details, enabling users to focus on developing new processor instructions, integrating them with TensorFlow Lite models, and measuring performance improvements through cycle counts. Notable features include the ability to use various supported FPGA boards, a reliance on open-source tools (except for Vivado), and support for rapid iteration on processor enhancements.