Ultra High Performance, Highly Reliable, Numeric Intensive Processors and Systems
Abstract
Arithmetic bandwidth remains one of the principal bottlenecks in real-time-high-end signal, image, and data processing applications. The problem is compounded when complex arithmetic is required. The problem, unfortunately, does not stop there. For military applications, size (volume) and power dissipation often are as important as bandwidth. Unfortunately speed and complexity (size-power) metrics are often in conflict. As a result, the defense signal processing systems designer finds that performance requirements often cannot be satisfied with contemporary hardware. As performance demands escalate, the gap between computational bandwidth and packaging will also increase. This will further stress a defense signal processing system. To achieve the numeric data processing speeds required in very high-end applications, a highly parallel arithmetic system called the residue number system (RNS) is considered. Recent results in this field have provided a framework in which very fast, reliable, and compact VLSI digital systems can be fabricated. It will be shown that the developed system exhibits a number of advantages which including speed and compactness. The reported technology is a central processing element for a signal processing system under development called the Gauss Machine.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1989
- Accession Number
- ADA214100
Entities
People
- Fred J. Taylor
Organizations
- University of Florida