/Rainborg

GPU-accelerated Position-based Fluid Simulation.

Primary LanguageC++

Rainborg

Position-based Fluid Simulation.

Screencap from dambreak.xml

Dependencies

This project has a few dependencies, which CMake will try to find:

Compilation

Once the dependencies are installed, simply run CMake on the build directly, like so:

> mkdir build
> cd build
> cmake ..

You can also configure CMake flags GPU_ENABLED and USE_PNG to enable CUDA compilation and PNG output respectively. For PNG output, make sure you have an empty directly pngs in your build directory.

> ccmake ..
# configure variables...

Once you've configured everything, run the generated Makefile:

> make

This should generate the executable ./testflow.

Running our code

Type ./testflow -h for a list of commands and their descriptions. Note that a scene file (such as the ones included in /assets) is required for the simulation to run.

Structure

  • Main:
    • Parse command line arguments
    • Windows and display loop
    • Initialize things??
      • Load Files
      • Create scenes + data structures
  • Parser:
    • Scene Parser
      • XML nodes:
      • fluid block positions, rest densities, colors, velocity, mass
        • fluid constants - viscosity, etc
      • fluid boundaries
      • camera position, rotation
      • timestep, timeframe
    • Mesh Parser
      • obj file, ply (or whatever)
      • translate to mesh boundary object triangles
  • Scene:
    • Boundaries -> Mesh objects
    • FluidSim
      • Fluid Particles
        • mass, rest density (shared)
        • Position
        • Velocity
        • Color
      • Bounding box
      • FluidSim.step(dt, scene)
    • Forces (gravity)

Todo

  • Rendering
    • Spheres (with lighting)
    • Optional import mesh, boundary
  • Simulation
    • Do the thing
    • Go through FOSSSim, stepper
    • Boundary/collision detection handling (static rigid objects)
    • Serial
      • external forces (gravity)
      • predict position
      • find neighbors
      • jacobi iterator
        • constraint, lambda
        • s-correction
        • calculate delta-p
        • collision detection response
        • update positions
      • vorticity confinement
      • position update
    • Parallel
      • Buckets, arbitrary acceleration data structures
      • same thing...
    • Future

References

The algorithms in this project follow the following papers:

Miles Macklin, Matthias Muller, Position Based Fluids

Takahiro Harada, Seiichi Koshizuka, Yoichiro Kawaguchi, Smoothed Particle Hydrodynamics on GPUs

Techniques for CUDA-OpenGL interop functions are used from here: http://www.nvidia.com/content/gtc/documents/1055_gtc09.pdf

Additionally, an C++ OpenGL framework by Daniel Chappuis is modified for use in this project: https://github.com/DanielChappuis/opengl-framework

Finally, some XML parsing and Math utility functions are taken from FOSSSim, the codebase from Computer Animation W4167.