Python Codegen packages

coverage report

Overview

This repository contains three python packages which build on top of each other

• pystencils: code generation for arbitrary stencil codes
• lbmpy: lattice Boltzmann code generation (build with pystencils)
• pygrandchem: special phase field model for ternary solidification based on pystencils

Documentation

The documentation can be accessed here.

Quick installation with package manager

pystencils requires Python>=3.5. If you have a recent distribution you can install the requirements with your package manager. If you use a older system or have no root access, read the next section.

For example on a recent Ubuntu you can install pystencils with:

# Get dependencies
apt install build-essential git python3-pip python3-numpy python3-appdirs python3-joblib python3-jinja2
# Clone repository
git clone https://i10git.cs.fau.de/software/pystencils.git && cd pystencils
python3 install_all.py

This gives you a minimal installation that can be used e.g. to create walberla kernels. If you want to setup, run, post-process and analyse your simulations directly from Python additional packages are helpful.

Full installation with conda

apt install graphviz ffmpeg # optional packages for visualization that don't come with conda

# Clone repository
git clone https://i10git.cs.fau.de/software/pystencils.git && cd pystencils

# Create environment with all required (and optional) packages
conda env create -f conda_environment_user.yml  # for development use environment_dev.yml
source activate pystencils

python3 install_all.py

# Optional, if you have a NVidia GPU and CUDA is installed
pip install pycuda

# Open Jupyter notebook to view tutorials
jupyter notebook

Then navigate to pystencils/doc/notebooks or lbmpy/doc/notebooks and look at the IPython notebooks as examples to get started.