An example of cuda ray tracing in pure python syntax. It reproduces the result of Ray Tracing in One Weekend
The original cpp code takes around 4 hours to complete, while this python CUDA implementation takes less than 1 minute.
NOTE: The code is not guaranteed to be fully optimized, but I intend to provide some insights on efficient python ray tracing implementation.
ATTENTION: Currently it consumes ~16G of memory, so be aware if your PC doesn't have enough memory. You can decrease SAMPLES_PER_RAY to reduce memory consumption.
scene has 485 objs, takes 0.0107 s to construct
rendering time: 47.7054 s
saving to test.png ...
Ubuntu with NVIDIA GPU (tested on RTX 2080Ti)
conda install numba matplotlib
python demo.py