Python-OptiX is a Python wrapper for the NVIDIA OptiX ray tracing engine, allowing for GPU-accelerated ray tracing applications in Python.
This package aims to provide a more pythonic, object-oriented interface to OptiX by wrapping the original C-like API using Cython. It does so by primarily relying on the CuPy package.
Only Linux is officially supported at the moment. Experimental windows support is available.
Python-OptiX always supports the most recent version of the OptiX SDK. The current version supports OptiX 9.1.0 (ABI 118).
Requirements:
- NVIDIA driver R590 or later
- CUDA Toolkit 12.0 or later
- CuPy
Install a recent version of the CUDA Toolkit (12.0+) and make sure the CUDA header files are installed as well.
Note, that for some variants of the CUDA Toolkit,
like the one installed by the conda package manager, all required headers are not installed by default.
conda-environments require installation of the additional cudatoolkit-dev package in
order for this package to work.
The OptiX SDK itself does not need to be installed: the OptiX headers matching the supported version are
downloaded automatically from the NVIDIA/optix-dev repository
(using curl), both during the build and on first use at runtime, where they are cached in
~/.cache/python-optix/optix-dev-9.1.0. On machines without network access, extract the
v9.1.0 release of that repository to this location beforehand.
python-optix requires the CUDA include path during setup as well as at runtime to compile the CUDA kernels.
If CUDA is not found automatically, set the CUDA_PATH variable to the location of the CUDA Toolkit.
export CUDA_PATH=/path/to/cuda_toolkit # if not found automatically
python -m pip install python-optix[cupy-cuda12x]
git clone https://github.com/mortacious/python-optix.git
cd python-optix
export CUDA_PATH=/path/to/cuda_toolkit # if not found automatically
python -m pip install [-e] .
The examples directory contains Python ports of several OptiX SDK samples
(hello, triangle, spheres, opacity micromaps, dynamic geometry and materials, the denoiser and
task-based module compilation). Their additional dependencies can be installed with
python -m pip install python-optix[examples]
- With driver 595.71.05,
DeviceContext.cancel_creations(optixDeviceContextCancelCreations) crashes inside the driver once the context has actually compiled a module. UseModule.cancel_creationfor individual modules instead.
Contributions to Python Optix are welcome! If you find a bug or have a feature request, please open an issue on GitHub. If you want to contribute code, please fork the repository and submit a pull request.
Python Optix is licensed under the MIT License. See LICENSE for details.