A collection of GPU-accelerated projects built on NVIDIA Warp — exploring computational physics and geometry processing through custom CUDA kernels authored in Python.
Each project lives in its own subfolder under projects/ with its own code, assets, and tests.
- Computational physics — fluid dynamics, continuum mechanics, thermodynamics, elasticity, wave propagation
- Geometry processing — mesh generation, remeshing, signed distance fields, Boolean operations, parametric surfaces
- Simulation kernels — custom Warp kernels for solvers that go beyond off-the-shelf physics engines
- 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 warp-lang numpyAdditional per-project dependencies are listed in each project's own README.md.
| Project | Domain | Description |
|---|---|---|
warpfluids |
Compressible CFD | 1-D Euler solver (WENO5-Z + HLLC + SSP-RK3), validated against JaxFluids. Phase 1 complete. |
Each project has its own test suite. To run all tests across every project:
python -m pytest projects/ -vOr for a specific project:
python -m pytest projects/<project_name>/tests/ -v- Create a subfolder:
projects/<your_project>/ - Add your Python files and a
tests/directory - Include a
README.mdinside the subfolder describing what the project demonstrates and how to run it - Keep all units SI (metres, radians, seconds) unless the problem domain dictates otherwise
Apache 2.0 — see individual file headers.