BinDRAM: Binary Neural Network on Unmodified Commodity DRAM

Prerequisite

Our real DRAM chip characterization is based on the open-source FPGA-based DRAM characterization infrastructure DRAM Bender. Please check out and follow the installation instructions of DRAM Bender.

The software dependencies for the characterization are:

• GNU Make, CMake 3.10+
• c++-17 build toolchain (tested with gcc-9)
• Python 3.9+
• pip packages pandas, scipy, matplotlib, and seaborn

Hardware Setup

Our real DRAM chip characterization infrastructure consists of the following components:

• A host x86 machine with a PCIe 3.0 x16 slot
• An FPGA board with a DIMM/SODIMM slot supported by DRAM Bender (e.g., Xilinx Alveo U200)

Directory Structure

DRAM-Bender           # A fork of DRAM Bender that contains the characterization program
  └ sources           
    └ apps           
      └ BNN    # Source code of the BNN program    

Step 1

Clone the repo in your home directory

  $ git clone https://github.com/WolFtSoN/BinDRAM

Go to the program folder

  $ cd /home/<your_user_name>/DRAM-Bender/sources/apps/BNN/

Step 2

Run bnn.py to run the BNN model. When the script successfully finishes, a message "DONE!!" in terminal should appear.

  $ python3 bnn.py