Efficient large-scale hybrid fluid simulation

  • Authors:
  • Abhinav Golas;Rahul Narain;Jason Sewall;Pavel Krajcevski;Ming Lin

  • Affiliations:
  • University of North Carolina at Chapel Hill;University of California, Berkeley;Intel Corporation;University of North Carolina at Chapel Hill;University of North Carolina at Chapel Hill

  • Venue:
  • ACM SIGGRAPH 2012 Talks
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

State-of-the-art methods for fluid simulation, including velocity-based grid methods and smoothed particle hydrodynamics [Bridson and Müller-Fischer 2007], require a detail vs. domain size tradeoff. As a result, scenes with large spatial scales are restricted to coarse detail under the restriction of limited computational resources. The elliptic problems solved for incompressibility projection in these simulations are bandwidth-bound, since domains of interest are not cache resident on current generation hardware. As a result, limited optimization is possible, and interactive performance is not possible for medium-large scenes.