Skip to content

Latest commit

 

History

History
130 lines (102 loc) · 4.5 KB

README.md

File metadata and controls

130 lines (102 loc) · 4.5 KB

Library for Integrating FPGA to TensorFlow

This instruction describes how to build, install and integrate FlexCNN, developed by UCLA VAST LAB, to TensorFlow. This implementation is adapted from the tensorflow custom operation interface.

Content

  1. Hardware and Operating System
  2. Requirements and Dependencies
  3. Build The Library From The Source
  4. API and Usage
  5. Project File Tree
  6. Acknowledgement

HardWare and Operating System

Development

For development, the OS should be Ubuntu 18.04 LTS

Testing and Deployment

For testing and depolyment, despite the os requirement above, the server/PC should also equips with Xilinx Virtex UltraScale+ FPGA VCU1525 Development Kit

Requirements and Dependencies

Requirements

The Xilinx Runtime v2018.3 should be installed.

If also want to compile the library from source, the Xilinx Deployment Shell v2018.3, Xilinx Development Shell and SDAccel Design Environment v2018.3 should also be installed. You can find them through this link.

Dependencies

This library uses the cmake (version >= 3.0) as the building system and the googletest as the test framework. It also depends on the tensorflow.

Build the Library From The Source

  1. Pull the source code from this repo 
    git clone [URL of this repo]
  2. In ./src/libsacc.cpp change the following line (give your project path) 
    std::string Sacc::lib_base_path = "/path/to/libsacc/";
  3. Setting environments 
    cd libsacc
source env.sh
  4. Installing packages 
    python3.6 -m venv venv
source venv/bin/activate
sudo apt-get install build-essential libcap-dev
sudo apt install opencl-headers
pip3 install cython
pip3 install numpy
pip3 install -r requirements.txt
  5. Build the library via (execute in the root of the library: "libsacc" directory) 
    mkdir build
cd build
cmake ..
make all

Now, the FlexCNN library is successfully installed in your system.

Use it with Your Own Application

  1. Make sure you have installed the library successfully (refer to previous section).

  2. You first should build the bitstream. To do that, follow the instructions here.

  3. Now, that you have the required files, copy the bitstream (.xclbin file), host executable (.exe file) and instructions (.insts file) to ./config.

  4. Replace your TensorFlow project with ../tf_DSA. For that, follow the instructions here

API and Usage

API

This library follows the tensorflow custom operation interface. You can find details here.

Usage

Setting up the environment variable XILINX_XRT correspondingly (just source the env.sh in the root of this project) and then you can call this operation in python as below:

import tensorflow as tf

class SaccTest(tf.test.TestCase):
  def SaccTest(self):
    sacc_module = tf.load_op_library('/path/to/this/lib/libsacc.so')
    with tf.Session(''):
      result = sacc_module.sacc([[1, 2], [3, 4]])
      result.eval()

if __name__ == "__main__":
  tf.test.main()

Project File Tree

The project file structure is shown below,

.
+-- CMakeList.txt
+-- src
|   +-- libsacc.cpp
|   +-- xcl2.cpp # xilinx opencl utils
+-- inc
|   +-- CMakeLists.txt
|   +-- sa_params.h
|   +-- sa_types.h
|   +-- sacc.hpp
|   +-- xcl2.hpp # xilinx runtime support
+-- test
|   +-- CMakeLists.txt
|   +-- test.cpp
+-- utils
|   +-- gen_params.sh
+-- docker
|   +-- Dockerfile 
|   +-- run.sh
+-- data
+-- config

The src and inc contain source files and header files.

The data folder contains bias and weight information used by the accelerator and the config contains the instructions and the FPGA binary.

Acknowledgement

The first version of this library was developed by Tong He.