On November 8, 2004 Department of Computer Sciences (UTCS) faculty, graduate students, FoCS members and other guests attended the Fall 2004 Visions of Computer Sciences Lectures.
For the Fall 2004 Visions lectures, UTCS celebrated professors Stephen W. Keckler and David Zuckerman for their recent awards and honors. Stephen W. Keckler received the 2003 Grace Murray Hopper Award and Professor David Zuckerman received a 2004 Guggenheim Fellowship. Each honoree presented a talk on the topic of their choosing.
Dr. Keckler spoke on "A Rise From the Ashes - Dataflow Architectures," addressing the history and the potential future of data flow architectures.
Dr. Zuckerman spoke on "Extracting Randomness: Past and Future," which discussed the problems and research on reandomness and computation.
- 5 - 6 p.m. Lecture Program, Avaya Auditorium (ACES 2.302)
- Dr. J Strother Moore, Chair -- Remarks and introductions
- Dr. Steven Keckler -- Lecture
- Dr. David Zuckerman -- Lecture
- 6 - 7 p.m. Invitation Only Reception, ACES 6.102
Stephen W. Keckler will speak on "A Rise From the Ashes - Dataflow Architectures"
Abstract: Dataflow architectures were once thought to be capable of attaining the holy grail of computer architecture - easily programmable massive parallelism. While many prototypes were built in the 70's and 80's, architects were never able to overcome the overheads inherent in dataflow instruction execution and the inability to efficiently provide common idioms such as iteration and rewritable memory semantics. In the 90's however, microprocessor architects found use for "dataflow in the small" as a means of implementing out-of-order instruction execution. While effective, these techniques require substantial hardware to reconstruct the dataflow graphs dynamically, and are reaching their complexity and power scaling limits. This lecture will outline a renaissance in dataflow architectures, called Explicit Datagraph Execution or (EDGE), enabled by nanoscale technology as well as instruction set and compiler innovations. These EDGE architectures provide scalability, efficiency, and programmability not achievable von Neumann or traditional dataflow models.
David Zuckerman will speak on "Extracting Randomness: Past and Future"
Abstract: Randomness is extremely useful in computer science, but obtaining truly random bits is expensive. Is true randomness really necessary? Randomness extractors are a tool to attack this question. An extractor is an algorithm which extracts randomness from a low-quality random source, using some additional truly random bits. This lecture will survey extractors and their applications to seemingly unrelated areas, and discuss where research in this area is heading.