Approach: The Remote Interactive Visualization and Analysis (RIVA) system has been developed. The RIVA system consists of a parallel 3-D perspective terrain renderer running on the Cray T3D or Intel Paragon, an a 3-D Navigator GUI, Flexible Flyer, running on the SGI workstation, and an HiPPI interface between the parallel renderer and a remote framebuffer. The parallel renderer is capable of rendering multi-spectral, multi-resolution, and multi-phenomenology datasets at the same time. For multi-resolution and multi-phenomenology dataset rendering, the final image is composited by overlaying one dataset on top of another using an adjustable opacity. For multi-spectral dataset rendering, the coloration of the final image can be composited by any transformation and combination of the individual spectral bands.
A parallel volume renderer has been designed based on a ``splatting'' algorithm with object space decomposition. A rendering API is being implemented on the Cray T3D that has a wide range of support for data distribution, classification, shading and lighting, viewing matrix manipulation and data compositing. The renderer is capable of handling time-variant, 3-D spatial data with multiple variables. An X-window based GUI is also being developed to drive an interactive volume renderer for post-processing data analysis.
Accomplishments: The development of the RIVA system was completed in FY'95. Version 1.1 of the RIVA system was delivered to the HPCC Software Exchange in February 1995. It includes source code for the parallel renderer and supporting utility programs, the executables for the GUI and a small Mars test data set. The delivered RIVA renderer works on the Paragon and the T3D. The interactive exploration and visualization of the Global Mars using RIVA was demonstrated at Supercomputing `94. A technical paper entitled ``Remote Interactive Visualization and Analysis (RIVA) Using Parallel Supercomputers'' will be published in the Proceedings of the 1995 Parallel Rendering Symposium. The design of the parallel volume renderer API is complete. A specification and a design document was prepared. The parallel part of the API is being implemented on the Cray T3D. An X-window based GUI is being implemented using TCL/TK language.
Significance: The demand for parallel supercomputer in interactive scientific visualization is increasing as the ability of the machines to produce large output datasets has dramatically increased. Development of a parallel renderer API enables scientists to visualize their results on-the-fly, enabling computational steering and thus improving the model development and processing time.
Status/Plans: The RIVA system with full multi-datasets rendering capability is only available on the Cray T3D. We plan to convert it into a portable version using MPI and deliver it to the HPCC Software Exchange if enough resource is provided. The parallel rendering effort is still in its early stage. We need to do performance tuning add additional features for handling time-variant data, off-the-core rendering, arbitrary clipping planes, and compositing of volume images with 2-D terrain image.
Point of Contact:
Peggy Li
Jet Propulsion Laboratory
(818)354-1341
peggy@sun11.jpl.nasa.gov