MRI: Acquisition of a Graphics Supercomputer for Synthetic Environments Serving Science and Engineering

Principal Investigator: Henry Fuchs and Frederick P. Brooks Jr. 
Funding Agency: National Science Foundation
Agency Number: EIA-9871222

Abstract
The Department of Computer Science at UNC has since 1965 made a strategic emphasis on graphics and imaging. Over one-third of the teaching faculty (7 of 19) and one-half of the total faculty (21 of 42) work in the graphics and imaging cluster. We concentrate on experimental systems: a majority of our work is in interactive graphics, where the demand is inherently and necessarily for high performance. A scientist and an interactive system work together in discovery or analysis, in contrast to off-line animation, chiefly used for illustration.

The proposed instrument, is a state-of-the-art real-time graphics computer, a Silicon Graphics RealityMonster. In spite of its bombastic name, this is a very serious machine, intimately combining eight high-performance graphics rendering pipelines with 32 conventional RISC processors. This instrument will accelerate research on real-time synthetic-environments applications in science and engineering.

UNC has made real-time synthetic environments and advanced image processing the central foci of our graphics research since 1970. Our research includes both applications and technology. Our department currently has active collaborative projects developing graphics intelligence-amplifying systems with eight science and engineering disciplines: surgeons, diagnostic radiologists, radiation oncologists, gene therapists, protein biochemists, condensed-matter physicists, material scientists working at nanometer-scale, and structural designers. These driving problems have guided our across-the-board attack on all of the key technological problems limiting synthetic environment systems.

The principal projects set forth in this proposal are those for which both the exceptional graphics power and the intimately coupled high-performance computing power are crucial for enabling otherwise impossible research advances in the next three years.

• Rendering images of the world at nanometer scales as data comes from atomic-force microscopes.
• Augmented reality for applications such as giving the surgeon a view through the body by using images from inside the body, through a laparoscope.
• Rendering massive models of complex man-made structures, such as ships and factories, for walkthroughs.
• Rendering from reprojected images of the real world, as well as traditional geometric models.
• Rapidly rendering to large, multi-megapixel displays either head-mounted or covering the walls.
• Tracking outdoors for augmented reality.

Beside these principal projects we briefly describe other projects that will be substantially advanced by one or another capability of the proposed instrument:

• interactive virtual prototyping,
• molecular graphics,
• figural models for use in synthetic environments, and
• scientist-steered molecular dynamics.

Several investigators on this proposal are participating on one of two other proposals to the MRI program. In the next section, we discuss the differences in research goals and equipment requests.

The timely acquisition of the RealityMonster instrument is key to advancing our support of science and engineering, to maintaining the momentum of many research groups, and to attracting the ablest faculty and students.