Skip to content

TheRealJohnCollier/PYNQ_Composable_Pipeline

 
 

Repository files navigation

license python docs

Composable Pipeline

The Composable pipeline is an overlay with a novel and clever architecture that allow us to adapt how the data flows between a series of IP cores.

The key characteristic of a composable overlay is an AXI4-Stream Switch, which plays the same role as an old telephone switchboard. The AXI4-Stream Switch provides the runtime routing of how data flow from one IP core to another.

On top of that, the composable overlay includes Dynamic Function eXchange DFX regions that brings new functionality to the design at runtime.

The combination of these two technologies plus a pythonic API, built on top of pynq, provide an unprecedented flexibility.

The composable overlay architecture of a composable video overlay for the PYNQ-Z2 is shown in the image below

Supported Boards

Note: PYNQ-Z1 is supported with the same composable overlay as PYNQ-Z2

To rebuild the composable pipeline you need Vitis and Vivado 2020.2. Navigate to one of the supported boards folder and run make, only steps for Linux are provided.

Clone this repository

git clone https://github.com/Xilinx/PYNQ_Composable_Pipeline.git --recursive

Note that this project depends on the Vitis Accelerated Libraries and PYNQ

Install composable pipeline on your board

git clone https://github.com/Xilinx/PYNQ_Composable_Pipeline
python3 -m pip install PYNQ_Composable_Pipeline/ --no-build-isolation
pynq-get-notebooks pynq_composable -p $PYNQ_JUPYTER_NOTEBOOKS

The notebooks will be delivered into the folder /home/xilinx/jupyter_notebooks

Documentation

Find out more documentation on Read The Docs.

Contributing

We welcome contributions, please review the contributing guidelines to contribute.

Licenses

Copyright (C) 2021-2022 Xilinx, Inc

SPDX-License-Identifier: BSD-3-Clause

Third-party license

Binary Files License

Pre-compiled binary files are not provided under an OSI-approved open source license, because Xilinx is incapable of providing 100% corresponding sources.

Binary files are provided under the following license


Copyright© 2021-2022 Xilinx

Packages

No packages published

Languages

  • Tcl 51.9%
  • Jupyter Notebook 35.8%
  • Python 8.4%
  • Verilog 1.8%
  • C++ 1.7%
  • Makefile 0.4%