Towards Integrated Visualization and Computation of Fluid Flows on Modern Graphics Hardware

Modern graphics hardware has emerged as a lowcost massively parallel computing platform, allowing scientists and engineers to tackle computationally “big” problems on desktop computers. NVIDIA has introduced the Compute Unified Device Architecture (CUDA) to ease the programming of massively parallel graphics cards. CUDA provides orders of magnitude acceleration of numerical calculations on the graphics processing units (GPU). Our research is motivated by these recent breakthroughs and focuses on integrating computation and visualization of fluid flow simulations on modern GPUs. In particular, we adopt the CUDA programming model for fast numerical computations and the Open Graphics Library (OpenGL) for visualization of the time-dependent flow physics. Numerical simulations enable one to quickly explore structures and patterns in fluid flow systems and make predictions for design improvements in engineering systems. Our integrated simulation approach on the GPU has the potential to shorten engineering analysis of fluid flow systems substantially