Jonathan Cohen is a research scientist with NVIDIA, where he focuses on physical simulation and parallel software architecture. Prior to joining NVIDIA, he held technology R&D positions in the feature film visual effects industry. His work can be seen on films such as Spider-Man 3, The Chonicles of Narnia: The Lion, the Witch, and the Warddrobe, The Cat in the Hat, and the upcoming John Hancock. In 2007, he was awarded an Academy Award for Technical Achievement for his work on fluid simulation and volumetric modeling. He has held positions with Sony Pictures Entertainment, Rhythm and Hues Studios, The University of Southern California, and Microsoft. He received a Bachelors of Science from Brown University.
Visual simulation of turbulent fluids requires faithfully capturing characteristic properties of turbulence, specifically the turbulent energy spectrum. This talk will provide a quick introduction to computational fluid dynamics techniques, with an emphasis on the numerical requirements for simulating turbulence. Turbulence is difficult to achieve numerically because it requires solving a Poisson Equation over a large computational grid. Therefore, we present an implementation of a fast GPU-based multigrid Poisson solver written in CUDA. The GPU-based solver is over an order of magnitude faster than an equivalent CPU-based solver.