A collection of demos, benchmarks, and extensions built on Newton — NVIDIA's open-source GPU-accelerated rigid-body physics engine.
Each project lives in its own subfolder under projects/ with its own code, assets, and tests.
- OS: Windows 10/11 or Linux
- GPU: NVIDIA GPU with CUDA Compute Capability ≥ 7.5 (Turing or newer)
- CUDA Toolkit: 12.x
- Driver: ≥ 525.60 (Windows) / ≥ 520.61 (Linux)
- Python: 3.10 or newer
pip install newton warp-lang usd-core numpy| Example | Description |
|---|---|
| rj45_hydro | RJ45 plug insertion with hydroelastic contacts and live pressure patch visualisation |
| spline_insertion | DIN 5480 16-tooth splined shaft kinematically driven into its mating hub; Newton mesh-SDF contacts visualised as arrows |
| compile_time_bench | Cold-start compile-time benchmark for Newton 1.2's Kamino, MuJoCo, and VBD solvers — randomized 3-iteration runs, phase + per-kernel breakdown, PDF report |
Each example has its own test suite. To run all tests across every example:
python -m pytest projects/ -vOr for a specific example:
python -m pytest projects/rj45_hydro/tests/ -v
python -m pytest projects/spline_insertion/tests/ -v- Create a subfolder:
projects/<your_example>/ - Add your Python files and a
tests/directory - Include a
README.mdinside the subfolder describing what the example demonstrates and how to run it
Apache 2.0 — see individual file headers.