We present a fast method for physically-based animation of fluids on adaptive, unstructured meshes. Our algorithm is capable of correctly handling large-scale fluid forces, as well as their interaction with elastic objects. Our adaptive mesh representation can resolve boundary conditions accurately while maintaining a high level of efficiency.
|Scene||# of cells||secs/frame|
We have implemented our method and tested it on several demonstrative scenes; skyscrapers buffeted by swirling winds, an iconic bridge rocked by gusts of air, and a space station deforming under a flow of fluid-like particles.
Our method is capable of exploiting the parallelism of modern multi-core processors. We have parallelized our RDS fluid solver using OpenMP; this figure demonstrates the near-linear scaling we have achieved. This measurement was performed on an SGI Altix machine.