A Survey of Open-source Tools for FPGA-based Inference of Artificial Neural Networks

Abstract

During the recent years artificial neural networks have become a great part of everyday life. One of the big problems in AI is acceleration of neural network inference using different hardware: from CPUs and GPUs to FPGAs and ASICs. Many open-source tools have been proposed for this purpose. This article contains a review of a range of open-source tools for neural network optimization, acceleration and hardware synthesis. Tools of three types have been chosen for evaluation: 1) translating neural network models into synthesizable C; 2) accelerating neural network models using custom hardware accelerators; 3) synthesizing Verilog from neural network models. Some of the tools have been tested using five simple neural network examples. Intel CPU, NVIDIA GPU and Cyclone V FPGA hardware platforms have been used for evaluation. Results show that the tested tools can successfully process neural network models and optimize them for CPU and GPU execution, whereas FPGA execution results are controversial.