John Gustafson PhD
Chief Scientist of Vq Research and Visiting Scholar
Arizona State University
Prof. John L. Gustafson (www.johngustafson.net) is Chief Scientist of Vq Research and a Visiting Scholar at Arizona State University. Previously, he has been a Senior Fellow at AMD and a Director of Intel Labs. He is best known for his 1988 argument showing that parallel processing performance need not be limited by Amdahl’s law, now generally known as Gustafson’s law. He is a recipient of the inaugural Gordon Bell Prize and is a Golden Core member of IEEE.
Presentation Title:
A Revolution in Computing: Next-Generation Arithmetic
Presentation Abstract:
For 50 years, “floating-point operations per second” (FLOPS) has been the currency of technical computing performance. But the rise of AI and the end of Moore’s law have made us realize that IEEE standard floating-point (invented by Intel in 1977) is long overdue for replacement. Floating-point format is an imitation of a human-centric format, scientific notation. Just as decimal digits are a human-friendly but inefficient way to represent integers in electronic computers, floating-point format is a human-friendly but inefficient way to represent real numbers. I will present a new way to represent real numbers on computers that is both mathematically sound and follows engineering design goals. The elegant and simple approach can more than double speed and energy-efficiency for everything from Machine Learning to Computer Graphics to High-Performance Computing. This is a watershed, a revolution. And it is well underway.
