hepcat-ml-fpga-setup

0. Prereqs

Install some prerequisites.

sudo apt install libtinfo5 git curl build-essential

1. Vivado

Download Vivado 2020.1.

• Archive: https://www.xilinx.com/support/download/index.html/content/xilinx/en/downloadNav/vivado-design-tools/archive.html
• Direct download link: https://www.xilinx.com/member/forms/download/xef.html?filename=Xilinx_Unified_2020.1_0602_1208_Lin64.bin

Following instructions from https://danielmangum.com/posts/vivado-2020-x-ubuntu-20-04/, we need to "spoof" 18.04 by changing one line in /etc/os-version from

VERSION="20.04.6 LTS (Focal Fossa)"

to

VERSION="18.04.4 LTS (Bionic Beaver)"

This can be done as follows:

sudo cp /etc/os-version /etc/os-version.bak
sudo sed -i 's/VERSION="20.04.6 LTS (Focal Fossa)"/VERSION="18.04.4 LTS (Bionic Beaver)"/g' /etc/os-version

Then we can install Vitis/Vivado by opening a terminal and running

cd ~/Downloads
chmod +x Xilinx_Unified_2020.1_0602_1208_Lin64.bin && sudo ./Xilinx_Unified_2020.1_0602_1208_Lin64.bin

You will need to log in with a Xilinx username/password.

Keep all the default settings and press continue until it starts downloading/installing.

After installing Vivado, change /etc/os-version/ back to how it was!

sudo cp /etc/os-version.bak /etc/os-version

Patch Vivado for the y2k22 bug. Download y2k22_patch-1.2.zip from https://support.xilinx.com/s/article/76960?language=en_US. Then follow these instructions

sudo su
unzip y2k22_patch-1.2.zip -d /tools/Xilinx/ && rm y2k22_patch-1.2.zip
cd /tools/Xilinx
export LD_LIBRARY_PATH=$PWD/Vivado/2020.1/tps/lnx64/python-2.7.16/lib/
Vivado/2020.1/tps/lnx64/python-2.7.16/bin/python2.7 y2k22_patch/patch.py
exit

2. PYNQ-Z2 board files

Now install the PYNQ-Z2 board files.

git clone https://github.com/Xilinx/XilinxBoardStore/
sudo cp -r XilinxBoardStore/boards/TUL/pynq-z2 /tools/Xilinx/Vivado/2020.1/data/boards/board_files/
rm -rf XilinxBoardStore

3. Micromamba

Install micromamba following https://mamba.readthedocs.io/en/latest/installation.html

curl micro.mamba.pm/install.sh | bash

4. Python environment

git clone https://github.com/fastmachinelearning/hls4ml-tutorial
cd hls4ml-tutorial
micromamba create -f environment.yml

5. Set up (each time you log in)

source /tools/Xilinx/Vivado/2020.1/settings.sh

export XILINXD_LICENSE_FILE=2100@cselm2.ucsd.edu
export LM_LICENSE_FILE=2100@cselm2.ucsd.edu

micromamba activate hls4ml-tutorial

6. Set up the PYNQ-Z2 board

This can be done as part of notebook 7a or before starting the tutorial. Instructions are here: https://pynq.readthedocs.io/en/latest/getting_started/pynq_z2_setup.html. In particular, connect the PYNQ-Z2 board to a power source via the USB cable. Insert the SD card on the board. Make sure the jumpers are as indicated in the directions. Then power on the board. Wait for a minute or so until all the lights blink a few times.

Next, make sure the laptop can get on the WiFi network. Then, you can connect the PYNQ-Z2 board to the laptop via the ethernet cable. Set up a static IP address for the laptop's ethernet port in network settings. Use the IP address 192.168.2.1 and subnet mask 255.255.255.0. Note we may have to use different static IP addresses (1, 2, 3, 4, etc.) The PYNQ board, by default, will try to use 192.168.2.99.

Check everything is working by navigating your browser to http://192.168.2.99. You should see a JupyterHub running from the PYNQ-Z2 board. The username and password are both xilinx.

7. Run the tutorial

Run the hls4ml tutorial. Specifically, do parts 1, 4, 7a, 7b, and 7c.