SPH Taichi

A high-performance implementation of Smooth Particle Hydrodynamics (SPH) simulator in Taichi. (working in progress)

Examples

• Dragon Bath (~420 K particles, ~280 FPS on RTX 3090 GPU, with timestep 4e-4)

• Armadillo Bath (~1.74 M particles, ~80 FPS on RTX 3090 GPU, with timestep 4e-4)

Features

Currently, the following features have been implemented:

• Cross-platform: Windows, Linux
• Support massively parallel GPU computing
• Weakly Compressible SPH (WCSPH)[1]
• One-way/two-way fluid-solid coupling[2]
• Shape-matching based rigid-body simulator[3]
• Neighborhood search accelerated by GPU parallel prefix sum + counting sort

Note

The GPU parallel prefix sum is only supported by cuda/vulkan backend currently.

Install

python -m pip install -r requirements.txt

To reproduce the demos show above:

python run_simulation.py --scene_file ./data/scenes/dragon_bath.json

python run_simulation.py --scene_file ./data/scenes/armadillo_bath_dynamic.json

Reference

1. M. Becker and M. Teschner (2007). "Weakly compressible SPH for free surface flows". In:Proceedings of the 2007 ACM SIGGRAPH/Eurographics symposium on Computer animation. Eurographics Association, pp. 209–217.
2. N. Akinci, M. Ihmsen, G. Akinci, B. Solenthaler, and M. Teschner. 2012. Versatile rigid-fluid coupling for incompressible SPH. ACM Transactions on Graphics 31, 4 (2012), 62:1–62:8.
3. Miles Macklin, Matthias Müller, Nuttapong Chentanez, and Tae-Yong Kim. 2014. Unified particle physics for real-time applications. ACM Trans. Graph. 33, 4, Article 153 (July 2014), 12 pages.

Acknowledgement

Implementation is largely inspired by SPlisHSPlasH.

Diff-Taichi with DFSPH

Try

python run_simulation_diff_forward_checker.py --scene_file ./data/scenes/armadillo_bath_small_rigid_box.json

to enable GPU scene preview.

Try

python run_simulation_diff_forward.py --scene_file ./data/scenes/armadillo_bath_small_rigid_box.json

to calculate gradient in forward mode autodiff