-
Notifications
You must be signed in to change notification settings - Fork 11
Installation instructions
- NVIDIA drivers
- NVIDIA CUDA development toolkit
- OpenMP binary and development package
- Python 2 and/or Python 3
- Python - matplotlib
- OpenMPI binary and development package
- Python - mpi4py (for using MPI from Python)
On a terminal, cd to the folder where you unpacked the tarball, and type:
autoreconf -i
./configure
make
sudo make install
For support: [email protected]
To install all the necessary packages, open a terminal and type:
sudo apt update
sudo apt install nvidia-cuda-toolkit openmpi-bin libopenmpi-dev libomp-dev python python-matplotlib python-mpi4py
Then cd to the folder where you unpacked the tarball, and type:
autoreconf -i
./configure
make
sudo make install
If you have not installed it yet, you must install clang. Open a terminal and type:
xcode-select --install
If you have not installed it yet, install Homebrew with the command:
/usr/bin/ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"
Install the NVIDIA Web Drivers appropriate for your MacOS version/build number from:
Install NVIDIA CUDA development toolkit following the instructions in:
https://docs.nvidia.com/cuda/cuda-installation-guide-mac-os-x/index.html
To verify that the CUDA Drivers and the compiler are installed and loaded correctly, in a terminal window use the commands:
kextstat | grep -i cuda
nvcc -V
Do not forget to setup CUDA development PATH, by adding the following lines to your .bash_profile file
export PATH=/Developer/NVIDIA/CUDA-10.2/bin${PATH:+:${PATH}}
export DYLD_LIBRARY_PATH=/Developer/NVIDIA/CUDA-10.2/lib\
${DYLD_LIBRARY_PATH:+:${DYLD_LIBRARY_PATH}}
Install necessary packages:
brew install openmpi mpi4py libomp
sudo easy_install pip
pip install matplotlib numpy
Then cd to the folder where you unpacked the tarball, and type:
autoreconf -i
./configure
make
sudo make install