Skip to content
System Verilog code describing a fully combinational binarized neural network.
SystemVerilog Other
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

Combinational Binarized Neural Network (BNN)

This repository contains the System-Verilog source files describing a HW fully-combinational BNN model.

Folders Content

  • src\ includes the HDL description of basic BNN building blocks (spatial convolutional and fully-connected layers)
  • test\ contains multiple example of network models
  • src\ includes network-specific HDL descriptions. Among them, contains the networks parameters.
  • tb\ inlcudes the testbench files for running simulation.
  • sim\ used for simulation purpose with ModelSim (tested with version 10.6b)

Try the code

Simply navigate to the sim\ folder of the desidered netowork example and type

make clean build run

for sythesis and simulation. Stimulus is defined in tb/


Sythensis results with GF22 SOI technology are reported in the paper. Please, acknowledge it if you make use of our code.

Rusci, Manuele, Lukas Cavigelli, and Luca Benini. 
"Design Automation for Binarized  Neural Networks: A Quantum Leap Opportunity?." 
Circuits and Systems (ISCAS), 2018 IEEE International Symposium on. IEEE, 2018.

You can’t perform that action at this time.