HLS Flow for Alveo and MPSoC boards using Vivado HLS, Vivado and PYNQ

Table of contents

• Get Started with videos on YouTube
  • Download PYNQ-Z2 Board File
• Example structure
• Todos

Get Started with videos on YouTube

These two links are example videos which will teach you how to create an IP module (from zero to hero) and using it with PYNQ on both MPSoC and Alveo boards:

• MPSoC: https://www.youtube.com/watch?v=mkUhYrmlX0k
• ALVEO: https://www.youtube.com/watch?v=MuBzr28tUTM

The videos will show you how to create and build projects Vivado HLS, Vivado for MPSoC boards and how to export kernel from Vivado HLS for SDx or SDAccel environments in order to perform high level synthesis for Alveo boards.

Every example is created to be used with PYNQ (http://www.pynq.io/).

Tested on PYNQ Z2 (http://www.tul.com.tw/ProductsPYNQ-Z2.html) and Alveo U250 board but is compatible with every MPSoC board that support PYNQ framework (e.g. ZCU104, ZCU102, Ultra96 etc.).

Boot images are downloaded from the official site of the board (http://www.tul.com.tw/ProductsPYNQ-Z2.html)

Download PYNQ-Z2 Board File

Example structure

This is the structure of every example dir:

EXAMPLE_DIR
├── <project>_HLS
│   └── apc
│       ├── src
│       │   └── <file>.cpp
│       └── tb
│           └── main.cpp
└── jupyter notebooks
    └── test_<file>.ipynb

5 directories, 3 files

The directory of every example contais two sub-dirs:

• <project>_HLS: contains a sub-dir called apc, which includes two sud-dirs:
  • src: contains the file with the declaration of the function that is to be accelerated;
  • tb: contains a testbench file, used to test if the previous function is correct.
• jupyter notebooks: contains the notebook file format to test previous files.

The examples in this repo are:

• ARRAY INPUTS - DOT: example to accelerate a dot product between two matrices. Each entry in the product matrix is the dot product of a row in the first matrix and a column in the second matrix;
• ARRAY INPUTS - DOT - OPTIMIZATIONS: example to optimize the previous acceleration;
• ARRAY INPUTS - MUL: example to accelerate a multiplication between two matrices element by element. Each entry in the product matrix is the product of the first matrix element and the second matrix element;
• ARRAY INPUTS - MUL - OPTIMIZATIONS: example to optimize the previous acceleration;

Todos

• examples using stream bus (HLS and Vivado)
• examples for benchmark against numpy
• examples on Vitis

License

MIT

Copyright © 2020 MakarenaLabs