Wednesday, August 25, 1999
Hewlett Packard Auditorium, 306 Soda Hall
Professor A. Richard Newton
The cost of designing and implementing a single-chip system is rising. Even if you know exactly what you want to build, the cost of the design tools, masks, engineering time, design verification, and test systems is likely to exceed $7M today for a high-end system-on-a-chip. Even if you can afford it, problems associated with deep sub-micron process technologies make getting the chip to work correctly the first time increasingly difficult. Even the design of the on-chip power grid can be a nightmare, especially for digital systems designers!
These factors, along with the economic requirements of consumer markets, are likely to lead us towards a large increase in the use of Application-Specific Programmable Platforms (ASPPs) rather than special-purpose system-on-a-chip designs. In such a world, tools for the optimal design and implementation of "programmable" parts become increasingly important. But in a world dominated by wiring delays and reliability issues, what are the appropriate programming models for such systems? And what should an on-chip runtime environment look like?
The majority of silicon in the future is likely to be consumed in distributed, embedded systems, often with real-time and reliability constraints. We are likely to see a revolution over the next few years in how we design and program such systems.
Professor Joseph Hellerstein
Berkeley has undoubtedly been the leading academic institution in information systems research. In this talk I will argue why information systems is the most critical research area in computing today, and give an overview of the highs and lows of the research area. I will also sketch Berkeley's contributions to the field, both intellectual and technological. Finally, I will give my view of where the research area is going
This is a repeat of an "area overview" given to the faculty at our annual retreat, and the first in a series of such talks in the EECS colloquium.