The following packages are recommended to run bempp-cl
numpy, scipy, numba, pytest, pyopencl, meshio, plotly, jupyter.
The easiest installation is through conda. The bempp directory contains a file environment.yaml. Simply run
conda env create -f environment.yml
and a complete conda environment called bempp is setup. Once it is created change into the environment with
conda activate bempp
You can now install the bempp-cl package with
pip install git+git://github.com/bempp/bempp-cl.git
By default bempp-cl uses the pocl (portablecl.org) OpenCL CPU library. Other OpenCL capable devices can be used if a corresponding ICD file is available. These need to be copied into the ~/.conda/envs/bempp/conda/etc/OpenCL/vendors directory.