A Hardware Architecture for Dynamic Performance and Energy
January 29, 2002 Tuesday
Hamerschlag Hall 1112
Carnegie Mellon University
With the increasing gap between processor and memory performance, many
applications are experiencing a mismatch between time spent performing
useful computation versus idling while a memory access completes. This
talk will present preliminary work on a hardware architecture to take
advantage of this mismatch to reduce energy consumption of applications,
while incurring a bounded degradation in performance.
Phillip Stanley-Marbell is a first year PhD student
in the department of Electrical and Computer Engineering at Carnegie Mellon,
working under Prof. Diana Marculescu. His research interests include microarhitectures
for power-aware systems, virtual machines and mobile code.