|
Impact of Future
Technologies (a.k.a Effect of the End of Moore's
Law) Moderator: Trevor Mudge (University of Michigan) Panelists: Fred Chong
(University of Chicago), Igor Markov (University of Michigan) Moderator’s
Introduction TBA Fred
Chong’s position statement Emerging technologies will play a key
role in addressing the gap between the needs of Big Data and Moore's Law
improvements to CMOS. At one extreme, Quantum Computation offers both the
largest gains and the most risk.
Quantum devices, however, also offer promising improvements in
classical computations. In general,
systems and architectures will need to adapt to the unique properties of new
devices. Moreover, a paradigm shift
will be needed at the user and compiler level, with some explicit tolerance
to variation and errors at the device level. Igor
Markov’s position statement The death march of Moore's law motivates
us to discuss obstacles to computing and ways to circumvent them. As a
result, we are looking at a variety of emerging technologies and some exotic
possibilities, such as quantum computing, 3D circuits, carbon nanotubes, etc.
Revolutionary new opportunities are often offset by grave limitations. The
promise of asymptotic runtime improvements may run against heavy
implementation constants and very narrow markets. Yet, more modest steps may
offer a viable path to higher-performance computing. New breakthroughs will
require concerted improvements in switching elements, memories, interconnect,
full-chip optimization, architecture platforms, compilers and runtime
support. One thing is clear - the free ride on the back of Moore's law is
over, and researchers must now distinguish worthwhile directions from dead ends. |