Visualization, virtual reality, and animation within the data flow model of computing

Introduction This paper presents our perspective on the utility of data-flow programming in the field of scientific visualization. The Cornell Theory Center (C'I'C) is an established center of visualizetion production. Scientists from across the country use the center's resources co explore the data collected from their research. The CTC uses IBM's Visualization Data Explorer TM to do most of its visualization, and maintains a repository of DX extensions that are available free to the public. The discussion in this paper primarily focuses on the flexibility and speed of development afforded by the use of modular programming. In particular, DX is shown to provide sufficient flexibility to be useful in settings ranging from animation production to Cornell computer science education. Examples from actual work in progress are used in this paper to underpin our advocacy OF modular data-flow programming. We be~n by examining how data from a specialized application such as chemistry can easily fit within the mathematical model of scientific data representation provided by DX. Included as examples of the modular extensibil i ty of DX are a description OF The link between DX and The Electronic Visualization Laboratory's CAVE virtual reality environment and an explanation OF the CTC-developed DX interface to RenderHan. TM Finally, as an example of the quick learning curve associated with DX, sections of the curriculum developed for Cornell computer science classes on graphics (C5417 and CS418) are presented. Researchers not only need to combine existing tools in novel and innovative ways, but need the flexibility to add new reals and new interfaces. Virtual reality, for example, is only beginning to be used in actual research, so existing codes may not yet address the needs of a specific community like chemists. Researchers also need to present their data in a clear and polished Form which not only involves higher-quality, more photorealistic rendering but animation as well. With the exception of initial design, these applications are less interactive and more batch-oriented. Education on the other hand, has its own special requirements. In the sections that follow, we address each of these needs, discussing how the data-flow paradigm fits into rhe application area and how it can be interfaced with existing programs in r.he discipline.