Installation
sisl is easy to install using any of your preferred methods.
Required dependencies
Running sisl requires these dependencies:
Optional dependencies:
pytest (for running the test suite)
pathos (for parallel BrillouinZone calculations)
tqdm (for displaying progress-bars)
plotly (for advanced visualization)
sisl implements certain methods in Cython which speeds up the execution. Cython is required if one wishes to re-generate the C-sources with a different Cython version. Note that this is not a necessary step and should typically only be considered by developers of Cython modules.
Installation of the stable sisl releases can be done by following the common conventions
using pip
or conda
methods:
pip will install from pre-build wheels when they are found, if not it will try and install from a source-distribution.
python3 -m pip install sisl
# for better analysis
python3 -m pip install sisl[analysis]
# for advanced plotting functionality
python3 -m pip install sisl[viz]
Conda enviroments are clever, but fragile. It is recommended to contain the sisl installation in a separate environment to decouple it from other fragile components. Their inter-dependencies may result in problematic installations.
conda create -n sisl
conda activate sisl
conda config --add channels conda-forge
conda install -c conda-forge python sisl
This is equivalent to a development installation which requires a C and fortran compiler as well as some other packages:
python3 -m pip install setuptools_scm "scikit-build-core[pyproject]" Cython
python3 -m pip install git+https://github.com/zerothi/sisl.git --prefix <prefix>
The remaining dependencies should automatically be installed.
Using conda as development environment can be done, but may be a bit more cumbersome to work with. To install sisl from sources one needs a conda environment with the following content:
conda create -n sisl
conda activate sisl
conda config --add channels conda-forge
conda install -c conda-forge fortran-compiler c-compiler python scikit-build-core pyproject-metadata
conda install -c conda-forge cython scipy netcdf4 cftime plotly matplotlib
subsequent installations of sisl should follow dev|pip
tab
Editable installs are currently not fully supported by scikit-build-core
and
is considered experimental. One may get it to work by doing:
git clone git+https://github.com/zerothi/sisl.git
cd sisl
python3 -m pip install -e .
Windows
Note
Currently compiling sisl on Windows is not tested in CI, any contributions to get sisl up and running on Windows would be greatly appreciated. Please help us out by opening an issue.
To install sisl
on Windows one will require a specification of
the compilers used. Typically one may do
python3 -m pip install . --prefix=<prefix>
but sometimes setuptools
does not intercept the flags in the build process.
Since 3.12 distutils
has been deprecated and one needs to pass explicit linker flags to the CMake environment.
Testing your installation
After installation (by either of the above listed methods) you are encouraged to perform the shipped tests to ensure everything got installed correctly.
Note that pytest needs to be installed to run the tests. Testing the installation may be done by:
pytest --pyargs sisl
The above will run the default test-suite which covers most of sisl
.
Additional tests may be runned by cloning the stripped branch of
sisl-files
and setting the environment variable SISL_FILES_TESTS to the path of the cloned repository.
A basic procedure would be:
git clone https://github.com/zerothi/sisl-files.git
SISL_FILES_TESTS=$(pwd)/sisl-files pytest --pyargs sisl
Compile time options
By default sisl enables everything that is possible, i.e. the compilation flags listed here are primarily intended for debugging, performance analysis/regressions and should typically not be touched.
Warning
It is not recommended to use these flags for production runs.
Passing options to the build-system through pip
should de done with
the following convention
python3 -m pip install --config-settings=cmake.define.CMAKE_BUILD_PARALLEL_LEVEL=5 ...
python3 -m pip install --global-option=cmake.define.CMAKE_BUILD_PARALLEL_LEVEL=5 ...
In the above case the compilation of the C/Fortran sources are compiled in parallel using 5 cores. This may greatly reduce compilation times.
There exists a set of compile time definitions that may be handy for developers. These are all CMake definitions and can be added like this:
python3 -m pip install --config-settings=cmake.define.WITH_FORTRAN=YES .
python3 -m pip install --global-option=cmake.define.WITH_FORTRAN=YES .
The options are:
Option |
Default |
Description |
---|---|---|
|
|
If OFF, no fortran sources will be compiled, this may be useful in debug situations. For full support this should be kept ON. |
|
|
Other name of |
|
|
If ON, warning messages will be printed when arrays are copied upon fortran routine calls. |
|
|
Minimum number of elements before
|
|
|
Add line-directives when Cythonizing sources. |
|
|
Add information to the GDB debugger. |
|
|
Add annotated output (html) when Cythonizing sources. |