Kevin Kiningham
These directions have only been tested on Ubuntu 16.04 with SDK version 18.0.
Download and install the Intel FPGA SDK For OpenCL.
Install Eigen
sudo apt install libeigen3-dev
Install OpenCL headers (note that you may have to install additional drivers depending on your GPU).
sudo apt install opencl-headers clinfo
Install CMake
sudo apt get install cmake
Install Pipenv
pip install pipenv --user
Install all python dependencies by running pipenv inside of the project directory
pipenv install
This project is distributed as a C++ python extension which can be built using the provided setup.py.
pipenv run python setup.py build
import numpy as np
from accel import accel
L = np.load('KKT_Factorized_L.npz')
D = np.load('KKT_Factorized_D.npz')
runner = accel.Kernelrunner()
# Run the solver with a random initial trajectory for 35 iterations
x, w = runner.run(L, D, 35)