Abstract

Manufacturing process variability at low geometries and energy dissipation are the most challenging problems in the design of future computing systems. Currently, manufacturers go to great lengths to guarantee fault-free operation of their products by introducing redundancy in voltage margins, conservative layout rules, and extra protection circuitry. However, such design redundancy leading to significant energy overheads may not be really required, given that many modern workloads, such as multimedia, machine learning, visualization, etc. can tolerate a degree of imprecision in computations and data. In this talk, I will introduce SCoRPiO, an approach which seeks to exploit this observation and to relax reliability requirements for the hardware layer by allowing a controlled degree of imprecision to be introduced to computations. We show that one can gracefully trade-off output quality for energy savings, via the degree of unreliable task execution.

Biography

Nikolaos Bellas is Associate Professor at the Electrical and Computer Engineering Department of the University of Thessaly in Volos, Greece. His research interests include Approximate Computing, Low Power design, CAD tools for architectural synthesis, multimedia and image processing for parallel machines. He received his Diploma in Computer Engineering and Informatics from the University of Patras, Greece in 1992, and MSc and PhD in Electrical and Computer Engineering from the University of Illinois, Urbana-Champaign, in 1995 and 1998, respectively. From 1999 to 2007, he was a Principal Member of Technical Staff at Motorola Labs in the US. He was one of the architects of Falcon, an MPEG4 video decoding chip used by the first Motorola camera phone, and has worked extensively on architecting Systems On Chip for multimedia and imaging applications. He currently coordinates the FP7 FET-Open Project SCoRPiO on Significance-based computing. He holds 10 issued US patents.