Installation

Installation via pip or conda

Stable versions of dynasor are provided via PyPI and as part of conda-forge. This implies that dynasor can be installed using pip:

pip install dynasor

or using conda:

conda install -c conda-forge dynasor

Installation via setup.py

If you want to use the most recent (development) version you can clone the repository:

git clone https://gitlab.com/materials-modeling/dynasor.git

and subsequently install using the setup.py script while standing in the dynasor directory as follows:

pip install --user .

or:

python3 setup.py install --user

Requirements

dynasor requires Python 3.8+ and depends on the following libraries:

  • ASE (trajectory reading)

  • MDAnalysis (trajectory reading)

  • NumPy (numerical linear algebra)

  • Numba (computational efficiency)

  • pandas (data handling)

numba and icc_rt

dynasor employs numba for the efficient calculation of correlation functions. To get the full benefit of using numba (see here) you need to install icc_rt, which can be installed using, e.g., conda or pip. icc_rt can lead to certain use cases running 5 to 10 times faster.

Note that there is an existing bug when installing icc_rt with pip (see this github issue). If you run numba -s the ideal output is:

__SVML Information__
SVML State, config.USING_SVML                 : True
SVML Library Loaded                           : True
llvmlite Using SVML Patched LLVM              : True
SVML Operational                              : True

If any of these are False you might have to add the python library directory to your LD_LIBRARY_PATH. For example if you installed numba in your local user-owned python directotry, the following line could work:

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HOME/.local/lib

If this does not resolve the issue or if want to read up on the background you may want to consult the aforementioned github issue.