pynq-image

This repo is used to build custom SD card images for PYNQ based boards PYNQ-Z2, Ultra96v2, and ZC706.

Prerequisites

Ubuntu 18.04

This following packages are needed.

sudo dpkg --add-architecture i386
sudo apt-get install lib32stdc++6 libgtk2.0-0:i386 libfontconfig1:i386 libx11-6:i386 libxext6:i386 libxrender1:i386 libsm6:i386 libqtgui4:i386 libxtst6:i386 libxtst6
sudo apt-get install sudo gcc python python3 libncurses5-dev libncursesw5-dev zlib1g:i386 pv wget gawk less git
sudo apt-get install net-tools xterm autoconf libtool texinfo zlib1g-dev gcc-multilib build-essential

Xilinx tools

This flow is based on the v2.6.0 of PYNQ that needs Vivado, Vitis, and PetaLinux version 2020.1 for compatiblity.

PetaLinux cache

During the build PetaLinux needs to download data in the form of the architecure-dependent sstate-cache and the architecture-agnostic downloads directories. To avoid redoing this every time we run the build we can pre-download the directories from here. The README file in the Xilinx download page describes how to extract the files and where to put them in the PetaLinux install directory.

Prebuilt board-agnostic Linux image

To save time on the build process, we need to download the prebuilt Linux image v2.6(here) for the board's ARM architecture(arm for {PYNQ-Z2, ZC706.} and aarch64 for Ultra96v2). Using the PREBUILT flag to point to this image will skip the the board-agnostic stage of the build.

PYNQ source distribution

To avoid rebuilding the PYNQ source distribution package, and consequently bypass the need to build bitstreams and MicroBlazes’ bsps and binaries, we need to download a prebuilt PYNQ sdist tarball v2.6 from the PYNQ releases section. The PYNQ_SDIST flag is used during build to point to this tarball.

Building custom BSPs

On some boards, we have enabled a anti-freezing mechanism using OS based watchdog timers, so before building the final image we need to build its custom BSP file.

Ultra96v2 BSP

To do this go into the Ultra96-PYNQ submodule and follow the instructions in README on "Creating your own BSP".

ZC706 BSP

To do this go into the ZC706-PYNQ submodule and follow the instructions in README.

Build

Before building the image make sure all the prerequisites are present and make sure you have a few hours, 50-100 GBs of free disk space, and enough RAM and CPU.

Setup the environment

Last thing before build is setup the build environment by running the setup_host.sh script at PYNQ/sdbuild/scripts. On some systems you have to run sudo update-binfmts --display to make sure that qemu for your ARM architecure is marked as enabled. If not, manually enable it by running for example sudo update-binfmts --enable qemu-arm.

PYNQ-Z2

To build the image for PYNQ-Z2 go into PYNQ/sdbuild and run:

make BOARDS=Pynq-Z2 PREBUILT=<path to bionic.arm.<version>.img> PYNQ_SDIST=<path to pynq-<version>.tar.gz>

Ultra96v2

To build the image for Ultra96v2 go into PYNQ/sdbuild and run:

make BOARDDIR=<path to Ultra96-PYNQ submodule> PREBUILT=<path to bionic.aarch64.<version>.img> PYNQ_SDIST=<pynq-<version>.tar.gz>

ZC706

To build the image for ZC706 go into PYNQ/sdbuild and run:

make BOARDDIR=<path to ZC706-PYNQ submodule> PREBUILT=<path to bionic.arm.<version>.img> PYNQ_SDIST=<path to pynq-<version>.tar.gz>

Changing CMA size

If you need to change the CMA memory size for allocating larger memories during benchmarks runs, you can change it at PYNQ/sdbuild/boot/meta-pynq/recipes-kernel/linux/linux-xlnx/pynq.cfg before running the build. The current CMA size is set to 256MB.