Rapid, Hardware Accelerated, Large Data Visualization

Scientists and engineers use computer simulation to model the behavior of physical phenomena. These simulation techniques often produce prodigious amounts of output data. In order to view results, interactive visualization is used since it provides the ability to visually inspect data and interactively query particular values. However, the size of data is becoming large enough to challenge the ability of current visualization systems to produce results at interactive rates, greatly reducing the effectiveness of the visualization process. We propose to increase the performance of visualization by developing a computational framework that leverages a relatively untapped part of most computer systems: the graphics hardware (GPU). We will use GPU processing in a parallel computing environment to accelerate the rendering and processing of results data. In Phase I we will demonstrate this framework by integrating the technology into the open-source VTK/ParaView parallel visualization systems, and deploying the software on a supercomputing cluster.