Install Numba on Python3 Edit on GitHub

Install on Pop!_OS

Prerequisits

These instructions assume that the CUDA toolkit is already set up and working on your installation of Pop_OS.

Install The latest Numba library

To install Numba, please run this command:

sudo apt install python3-numba

Environmental Variable setup

To make changes on the system level, add the below in /etc/environment . To make changes on user level (assuming you have the default bash shell), add in ~/.bashrc. Note the system level approach is not dependendent on your choice of shell.

# path to CUDA driver shared library file.
export NUMBAPRO_CUDA_DRIVER=/usr/lib/x86_64-linux-gnu/libcuda.so
# path to CUDA libNVVM shared library file
export NUMBAPRO_NVVM=/usr/lib/cuda/nvvm/lib64/libnvvm.so
# path to CUDA libNVVM libdevice directory which contains .bc files.
export NUMBAPRO_LIBDEVICE=/usr/lib/cuda/nvvm/libdevice
# path to CUDA libraries
export NUMBAPRO_CUDALIB=/usr/lib/cuda/lib64

Finally, restart your computer to complete setup.

Test installation

To test your new Numba installation, from the commandline type:

numba -s

Details of the installation should be returned.

Further information

For further details on Numba, please see: http://numba.pydata.org/