warpSPH 0.3.0

GPU-accelerated SPH neighbor search and operators built with NVIDIA Warp and PyTorch

compressibleSPH

A GPU-oriented Python package for compressible Smoothed Particle Hydrodynamics (SPH) experiments and benchmarks.

The project is organized around reusable builders for simulation configuration, interchangeable SPH schemes, adaptive time integration, case setup helpers, and export/import utilities for trajectories and postprocessing.

What This Package Implements

Core SPH Schemes

The package currently exposes three compressible SPH schemes via CompressibleSPHScheme:

• Monaghan
• CompSPH
• CRKSPH

Scheme dispatch is handled by buildScheme(...) in src/compressibleSPH/schemes/builder.py.

Kernels

Kernel selection is part of SimulationConfig.kernel (type KernelFunctions, imported from sphWarpCore).

The parser helper parseKernelFunctions(...) accepts any kernel enum name exposed by your installed sphWarpCore version.

Integrators

Time integration is selected with SimulationConfig.integrationScheme (type IntegrationSchemeType, from integrators).

The helper buildConfig(...) returns both:

• a configured SimulationConfig
• the integrator object from getIntegrator(...)

Spatial Operators and Support Modes

Simulation configuration includes selectable operator and support modes:

• supportMode (SupportScheme)
• gradientMode (GradientScheme)
• laplacianMode (LaplacianScheme)
• samplingScheme (SamplingScheme)

Viscosity Switches

ViscositySwitch includes:

• Balsara1995
• Colagrossi2004
• CullenDehnen2010
• CullenHopkins
• MorrisMonaghan1997
• Rosswog2000
• NoneSwitch

Adaptive Support Schemes

AdaptiveSupportScheme includes:

• NoScheme
• Monaghan
• Owen

Energy Schemes

EnergyScheme includes:

• equalWork
• PdV
• diminishing
• monotonic
• hybrid
• CRK

I/O and Reproducibility Utilities

The package provides trajectory/state serialization utilities:

• prepExport(...)
• exportSimulationSystem(...)
• importSimulationSystem(...)
• importConfigs(...)

These write/read HDF5 state data and JSON config metadata for reproducible runs.

Package Layout

• src/compressibleSPH/configurations/: simulation and scheme configuration dataclasses/builders
• src/compressibleSPH/schemes/: scheme implementations and scheme builder
• src/compressibleSPH/systems/: state/system containers
• src/compressibleSPH/modules/: runtime modules (for example timestep support)
• src/compressibleSPH/sample/: sampling utilities
• src/compressibleSPH/caseUtils/: case setup helpers used by examples
• src/compressibleSPH/io.py: export/import and parsing helpers
• examples/compressible/: benchmark notebooks and generated media

Environment Setup

Recommended setup uses a dedicated conda environment and editable installs for backend repositories.

1) Create and prepare conda environment

conda create -n 'warp' python=3.13
conda activate warp
conda install -c anaconda ipykernel -y
conda install nvidia/label/cuda-13.0.0::cuda-toolkit cudnn
pip install toml scipy numba tqdm h5py matplotlib ipywidgets ipympl imageio scikit-image imageio_ffmpeg seaborn pandas
pip install warp-lang torch

2) Clone backend repositories

git clone https://github.com/wi-re/integrators
git clone https://github.com/wi-re/sphPlotting
git clone https://github.com/wi-re/warpSPH

3) Install backends in editable mode

cd integrators && pip install -e . && cd ..
cd sphPlotting && pip install -e . && cd ..
cd warpSPH && pip install -e . && cd ..

Current status: cloning and editable installs are the accepted setup path for now.

Quick Start

If you are running directly from this repository, ensure src is on your Python path:

export PYTHONPATH="$PWD/src:$PYTHONPATH"

Minimal usage pattern:

import torch
from compressibleSPH import *

# Build global simulation config + integrator
config, integrator = buildConfig(
    dim=1,
    dt=1e-3,
    adaptiveDt=True,
    cflFactor=0.3,
)

# Pick an SPH scheme
scheme = CompressibleSPHScheme.CRKSPH
SimulationSystem, SimulationState, SimulationConfig, SimulationUpdate, fn, export_fn, import_fn = buildScheme(scheme)

# Build scheme-specific config
schemeConfig = SimulationConfig()
schemeConfig.gamma = 5.0 / 3.0

# Case setup is typically done with helpers in compressibleSPH.caseUtils

Compressible Examples

A full gallery page with previews and embedded videos is available at:

• examples/compressible/EXAMPLES_SUMMARY.md

Table of Cases

Case	Notebook	Preview (PNG)	Video (MP4)
01. Sod Shock Tube (1D)	01-Sod_Shock_Tube_1D.ipynb		MP4
02. Linear Wave	02-Linear_Wave.ipynb		MP4
03. Kidder Isentropic Compression	03-Kidder_Isentropic_Compression.ipynb		MP4
04. Noh Implosion	04-Noh_Implosion.ipynb		MP4
05. Woodward-Colella Double Blastwave	05-Woodward_Colella.ipynb		MP4
06. Sedov-Taylor Blastwave (1D)	06-Sedov_Taylor_Blastwave_1D.ipynb		MP4
07. Sedov-Taylor Blastwave (2D)	07-Sedov_Taylor_Blastwave_2D.ipynb		MP4
08. Hydrostatic	08-Hydrostatic.ipynb		MP4
09. Gresho-Chan Vortex	09-Gresho_Chan_Vortex.ipynb		MP4
10. Yee Vortex	10-Yee_Vortex.ipynb		MP4
11. Shearing Noh Implosion (2D)	11-Shearing_Noh_Implosion_2D.ipynb		MP4
12. Kelvin-Helmholtz	12-Kelvin-Helmholtz.ipynb		MP4
13. Rayleigh-Taylor	13-Rayleigh_Taylor.ipynb		MP4
14. Triple Point (Equal Resolution)	14-Triple_point.ipynb		MP4
15. Triple Point (Equal Mass)	15-Triple_point_equalMass.ipynb

Precision Note (Notebook Workflows)

In the example notebooks, floating-point precision is configured early in the import/setup cell. If precision is changed (for example float32 to float64), restart the Jupyter kernel before rerunning.