Computing That Serves

GPU-Based Scientific Visualization


Thursday, September 23, 2004 - 11:00am


Claudio Silva, University of Utah

The advent of modern Graphics Processing Units (GPUs) with their high degree of parallelism is causing major shifts in the efficiency of algorithms in computer graphics and scientific visualization. Techniques that were previously seen as too slow for mainstream use can now be made practical. GPU-based implementation of these algorithms is often non-trivial, because the same features that make the GPUs so fast also cause them to be hard to target for algorithms. Trying to exploit the high levels of bandwidth and floating-point performance available turns out to be quite hard, since the general "flow" of the algorithms has to be substantially modified to enable efficient execution.

In this talk, I will describe recently developed GPU-based algorithms for several important scientific visualization tasks, including isosurface generation, and volume rendering of unstructured grids. I will also discuss the application of these techniques on the visualization of spatial and temporal data from a complex environmental observation and forecasting system for the Columbia River.


Claudio T. Silva is an Associate Professor of Computer Science at the University of Utah. Before joining Utah in 2003, he worked in industry (IBM and AT&T), government (Sandia and LLNL), and academia (Stony Brook and OGI). He received his Ph.D. in Computer Science from the State University of New York at Stony Brook in 1996, and his B.S. in Mathematics from the Federal University of Ceara (Brazil) in 1990.  He co-authored over 60 technical papers and 4 patents. He serves on the editorial board of the IEEE Transactions on Visualization and Computer Graphics, and on numerous program committees. He was co-chair of IEEE Symposium on Volume Visualization and Graphics 2004 and the IEEE Symposium on Parallel and Large-Data Visualization and Graphics 2003. He will be papers co-chair of the IEEE Visualization 2005.