deepware-ai/bnna

bnn accelerator

BNN Accelerator

Requirements

• GHDL
• GTKWave
• VHDL

Quick Instruction

cloning repository

  $ git clone https://github.com/bnna-project/bnna.git  
  $ cd bnna  

compiling VHDL code and looking on wave diagrams in GTKWave

  $ ghdl -s test_file.vhdl                 #Syntax Check  
  $ ghdl -a test_file.vhdl                 #Analyse  
  $ ghdl -e test_file.vhdl                 #Build   
  $ ghdl -r test_file --vcd=testbench.vcd  #VCD-Dump  
  $ gtkwave testbench.vcd                  #Start GTKWave  

Also you can compile and look on wave diagrams in GTKWave with command

  $ bash script.sh test_file.vhdl test_file_testbench.vhdl  

Last file must be testbench !!!

auto compiling the project

In the compile folder is a script to compile all the components automatically. You can use it like this:

  $ cd compile
  $ bash auto_compile.sh

After this you just have to compile the test benches you want to use. For example like this (you should still be in the compile folder):

  $ bash script.sh ../BNNA/tb_presentation.vhdl

References

1. FP-BNN: Binarized neural network on FPGA
   https://www.sciencedirect.com/science/article/abs/pii/S0925231217315655

2. Stochastic Computing for Hardware Implementationof Binarized Neural Networks
   https://arxiv.org/abs/1906.00915

3. FINN: A Framework for Fast, Scalable Binarized Neural Network Inference
   https://arxiv.org/abs/1612.07119