Multiple GPU Fluid Simulation

  • Authors: Donghao Ren, Junsheng Guo
  • Instructor: Prof. John R. Gilbert, TA: Veronika Strnadova
  • Course: Winter 2014 CS240A - Applied Parallel Computing

Introduction

We implemented a Multiple GPU fluid simulation program that runs on Triton or Lonestar machines. We parallelized the Smooth Particle Hydrodynamics algorithm first with a single GPU and then with multiple GPUs. The program is written in CUDA and MPI.

The source code is available on Github now!

Documents

If you are interested in this implementation, please read our final project report here:

Demo

Dam-break: 4 GPUs, 8,000,000 particles, 3000 frames (in-progress), 1000ms/frame simulation, 4000ms/frame rendering.

References

  • Fluid Simulation, SIGGRAPH 2007 Course Notes, R. Bridson et al.
  • Optimization strategies for CPU and GPU implementations of a smoothed particle hydrodynamics method, Computer Physics Communications, J. M. Dominguz et al.
  • Interactive SPH Simulation and Rendering on the GPU, Eurographics/ACM SIGGRAPH 2010, P. Goswami et al.
  • http://fluids3.com/