We present recent progress in real-time water simulation on the GPU. First, we will discuss a hybrid water simulation method that combines height grid and particles. The grid-based shallow water solver can handle arbitrary underlying terrain slope and water depth. Breaking waves, foams, and splashes are represented by particles. Small scale waves are added procedurally to enhance surface details. The method is demonstrated in various scenes including a river flowing along a valley with big rocks, steep cliffs, and waterfalls. We will then discuss a new 3D Eulerian liquid simulation method that can generate even more realistic results albeit at a higher cost. We introduce a specialized grid representation composed of regular cubic cells on top of tall cells. With this layout, we can represent water without consuming excessive amounts of memory and computing power, and focus efforts on the area near the surface where it matters most. Moreover, we propose a specialized multi-grid algorithm for solving the Poisson equation efficiently. We demonstrate the method in various scenarios such as a scene with a big wave crashing into a lighthouse. We then conclude the session with our thoughts on how to combine the two approaches for level-of-details.

Nuttapong Chentanez got his PhD in computer science from the University of California, Berkeley. His dissertation was about interactive simulation of surgical needle insertion and steering. He is now a research consultant for the physics research group of NVIDIA. His main research interests include liquid and deformable body simulation both for real-time and off-line applications.