aidge-core 0.3.0

Core algorithms for operators and graph of the AIDGE framework

Aidge Core library

You can find here the C++ code of the Core library of Aidge.

[TOC]

Pip installation

To install aidge_core using pip, run the following command in your python environnement :

pip install . -v

TIPS : Use environment variables to change compilation options:

• AIDGE_INSTALL : to set the installation folder. Defaults to <python_prefix>/lib/libAidge
• AIDGE_PYTHON_BUILD_TYPE : to set the compilation mode to Debug or Release or "" (for default flags). Defaults to Release.
• AIDGE_BUILD_GEN : to set the build backend (for development mode) or "" for the cmake default. Default to "".
• AIDGE_BUILD_TEST : to build the C++ unit tests. Set to "ON" or "OFF". Default to "OFF".

Pip installation for development

To setup aidge_core using pip in development (or editable mode), use the --no-build-isolation -e options to pip.

For instance run the following command in your python environnement for a typical setup :

export AIDGE_BUILD_TEST=ON              # enable C++ unit tests
export AIDGE_PYTHON_BUILD_TYPE=         # default flags (no debug info but fastest build time)
export AIDGE_PYTHON_BUILD_TYPE=Debug    # or if one really need to debug the C++ code
pip install -U pip setuptools setuptools_scm[toml] cmake   # Pre-install build requirements (refer to the pyproject.toml [build-system] section)
pip install -v --no-build-isolation -e .

In this configuration python files can be modified directly without re-installation.

The C++ build dir will be created in build/ and recompilation and install of python bindings can be done directly with:

make -C build install -j $(nproc)
# or with cmake
cmake --build build -j $(nproc) && cmake --install build

One can also use an alternate cmake build backend such as ninja which can be installed easily though pip, for instance :

pip install -U ninja
export AIDGE_BUILD_GEN=Ninja
pip install -v --no-build-isolation -e .

In this case ninja is used instead of make as build backend, and recompilation when needed is done with:

ninja -C build install  # note that by default ninja use available parallelism, no need for -j option
# or with cmake
cmake --build build && cmake --install build

Note that python development (or editable mode) is not always robust to changes in the python package setup, or when changing the build backend with AIDGE_BUILD_GEN. In order to re-install when the build breaks, re-execute the commands:

rm -rf *-egg-info build/
pip install -v --no-build-isolation -e .

Standard C++ Compilation

Create two directories build and ìnstall.

Then inside build :

cmake -DCMAKE_INSTALL_PREFIX:PATH=$(path_to_install_folder) $(CMAKE PARAMETERS) $(projet_root)

make all install

Compilation options

Option	Value type	Description
-DCMAKE_INSTALL_PREFIX:PATH	str	Path to the install folder
-DCMAKE_BUILD_TYPE	str	If Debug, compile in debug mode, Release compile with highest optimisations or "" (empty) , default= Release
-DWERROR	bool	If ON show warning as error during compilation phase, default=OFF
-DTEST	bool	If ON build C++ unit tests, default=ON
-DPYBIND	bool	If ON activate python binding, default=OFF
-DPYBIND_INSTALL_PREFIX:PATH	str	Path to the python module install folder when -DPYBIND=ON, defaults to $CMAKE_INSTALL_PREFIX/python_packages/<module>

If one compiles with -DPYBIND=ON, -DPYBIND_INSTALL_PREFIX:PATH can be used to install the python module directly in the python sources tree (for instance $PWD/aidge_core). setup.py takes care of this and installs the module at the right place.

Run tests

CPP

Inside of the build file run:

ctest --output-on-failure

Python