Aidge CUDA library

You can find in this folder the library that implements the CUDA operators. [TOC]

Installation

Dependencies

• CUDA ToolKit 12.4
• CUDnn9 make sure to install the CUDA 12 compatible version
• GCC
• Make/Ninja
• CMake
• Python (optional, if you have no intend to use this library in python with pybind)

Aidge dependencies

• aidge_core
• aidge_backend_cpu

Pip installation

pip install . -v

TIPS: Use environment variables to change compilation options:

• AIDGE_INSTALL: to set the installation folder. Defaults to /usr/local/lib. ⚠️ This path must be identical to aidge_core install path.
• AIDGE_PYTHON_BUILD_TYPE: to set the compilation mode to Debug or Release
• AIDGE_BUILD_GEN: to set the build backend with

Standard C++ Compilation

You will need to compile first the Core library before compiling the CUDA one. The makefile is designed to do it for you.