Ariadne is a tool for reachability analysis and model checking of hybrid systems. Additionally, it is a framework for rigorous computation featuring arithmetic, linear algebra, calculus, geometry, algebraic and differential equations, and optimization solvers.
The command-line installation instructions are presented for Debian Linux systems and derivatives (using apt) and macOS systems (using Homebrew). However, openSUSE and Fedora are known to be working when using their own package managers. Windows installations are not supported yet.
Official packages are available for Ubuntu LTS (currently 20.04) and derivatives and macOS 11, but if your architecture or particular setup necessarily requires compilation from sources, instructions are provided. The build system used is CMake. The library is tested for compilation using gcc (minimum required: 10.2) and clang (minimum required: 11.0).
Supplied packages are published using the official Launchpad platform for Ubuntu, and a custom Homebrew tap repository for macOS. Packages simply require all the dependencies, namely: MPFR, Cairo, Gnuplot and Python 3.
In order to install the apt package, first you need to import the ppa repository to the ppa list:
sudo add-apt-repository ppa:ariadne-cps/ariadne
Then you can install Ariadne (along with any missing dependencies):
sudo apt-get install ariadne
which will be updated with the latest release as with other apt packages.
In order to install the Homebrew package, you just need to
brew install ariadne-cps/tap/ariadne
which in one line both sets up the "tap" for Ariadne and installs the package along with any missing dependencies. The package will be upgraded with any new versions after a brew upgrade is issued. The package currently supports x86-64 architectures on macOS Big Sur. Other configurations trigger an automatic build, therefore you should not need to deal with sources any time.
If installed from sources, the only required library dependency is MPFR. To enable the graphical output you will require either Cairo or Gnuplot in order to save into png files. Finally, the Python bindings require the Python headers (version 3 is only supported, since version 2 is discontinued). In particular for Python, there is an internal Git submodule dependency on the header-only pybind11 library. In order to build the all the submodules of the library, Git must be installed even if Ariadne has been downloaded as an archive. Download of the dependencies is automatic though.
Finally, if you want to build the documentation, you need Doxygen and a working Latex distribution (including the Math packages).
Please note that adding new library dependencies after preparing the build environment requires to re-run the CMake command.
Specific instructions for Ubuntu and macOS follow, starting from installation from pre-compiled packages.
apt packages: cmake pkg-config git libmpfr-dev libcairo2-dev gnuplot and either clang-11 or g++-10 for the compiler toolchain.
Additional package required for the Python interface: python3-dev.
Additional packages required for documentation: doxygen doxygen-latex
Homebrew packages: cmake git mpfr cairo gnuplot and gcc@10 if using GCC.
For Cairo support, you may need to set up a permanent variable for the path of pkgconfig by adding the following line in your ~\.bash_profile:
export PKG_CONFIG_PATH=/usr/local/opt/libffi/lib/pkgconfig
To allow building the documentation: brew install --cask mactex-no-gui and brew install doxygen.
A pre-packaged source zip/tar.gz archive is always available in the releases section of Ariadne's GitHub space. Still, it is usually preferable to clone the repository using Git, in order to keep the distribution updated as soon as a new release is available. To do that, you shall issue
git clone https://github.com/ariadne-cps/ariadne
which creates an ariadne directory under the present working directory. Let's switch into that directory.
To build the library from sources in a clean way, it is preferable that you set up a build subdirectory, say:
$ mkdir build && cd build
Then you can prepare the build environment, choosing a Release build for maximum performance:
$ cmake .. -DCMAKE_BUILD_TYPE=Release
At this point, if no error arises, you can build the C++ library and its Python bindings with:
$ cmake --build .
This build does not include the examples, tutorials and tests targets. You can build those by supplying targets with the following:
$ cmake --build . --target <TARGET>
or by using the everything target to include all code targets.
To build the doc target for documentation, explicitly use:
$ cmake --build . --target doc
then you can access the built documentation from the docs/html/index.html file in the build directory.
To install the library globally from built sources, you must do
$ cmake --build . --target install
using sudo if you require administrator privileges for a Linux installation. Please note that the installation will build the whole distribution beforehand, hence it is preferable that you first build the other targets without administrator privileges, build the install target.
To find the installed library under Ubuntu, you may need to set the LD_LIBRARY_PATH in the .bashrc file of your home directory:
export LD_LIBRARY_PATH=/usr/local/lib
The tutorials directory contains two CMake projects that rely on a correct installation of Ariadne, either by using a package or by building the sources. You can copy a project directory in any place on your file system and follow the instructions on the README file inside to check that your installation was successful.
Due to limitations of the C++ standard library on macOS since C++11, you won't be able to build an executable with GCC if the Ariadne library has been built using Clang, and viceversa. Hence on macOS you shall use the same compiler for both Ariadne and any projects that depend on it. If Ariadne comes from the Homebrew package, then it has been built using g++ 10.
If you would like to contribute to Ariadne, please contact the developers. We are especially interested to hear how the documentation and user interface could be improved.
- Luca Geretti [email protected]
- Pieter Collins [email protected]