Computer Scientists Win Best Paper Award at HPCA

Computational sprinting Enlarge
Computational sprinting

Computer science researchers at the University of Michigan and the University of Pennsylvania, including undergraduate engineering student Yixin Luo, doctoral student Anuj Chandawalla, Professor Thomas F. Wenisch, and Professor and CSE Chair Marios Papaefthymiou, have won the Best Paper Award at the 18th International Symposium on High Performance Computer Architecture (HPCA), which took place February 25 – 29, 2012 in New Orleans.

The paper, entitled “Computational Sprinting,” proposes a means of scaling performance in mobile devices with limited cooling options. Current mobile architectures have reached a utilization wall in which sustained chip performance is limited primarily by the heat produced by multiple core chips within the devices. Recognizing that many mobile applications do not demand sustained levels of high performance, the researchers have proposed a method and supporting architecture for producing very short bursts of intensive computation in response to sporadic user activity.

The paper explores activating otherwise powered-down cores for sub-second bursts of intense parallel computation. The approach exploits the concept of computational sprinting, in which a chip temporarily exceeds its sustainable thermal power budget to provide instantaneous throughput, after which the chip must return to nominal operation to cool down. To demonstrate the feasibility of this approach, the authors analyze the thermal and electrical characteristics of a smart-phone-like system that nominally operates a single core (1W peak), but can sprint with up to 16 cores for hundreds of milliseconds. They describe a thermal design that incorporates phase-change materials to provide thermal capacitance to enable such sprints. They analyze image recognition kernels to show that parallel sprinting has the potential to achieve the task response time of a 16W chip within the thermal constraints of a 1W mobile platform.

Prof. Wenisch received his PhD in electrical and computer engineering from Carnegie Mellon University in 2007 and joined the faculty at Michigan that year. His research is focused on computer architecture with particular emphasis on multiprocessor and multicore systems, multicore programmability, data center architecture, and performance evaluation methodology. Prof. Wenisch was the recipient of an NSF CAREER award in 2009 and was named a Morris Wellman Faculty Development Assistant Professor of EECS in 2011. He is inventor on one patent and co-inventor on four patents.

Prof. Papaefthymiou is the Chair of Computer Science and Engineering at the U-M, and is a co-founder and Chief Scientist of Cyclos Semiconductor, a startup company specializing in energy-efficient chips for power-critical applications. His research addresses a broad spectrum of problems in computer design with an emphasis on architectures and design methodologies for energy-efficient high-performance computers. He is also active in the field of parallel and distributed computing. Among other distinctions, Prof. Papaefthymiou has received faculty recognition awards from Yale College, the EECS Department and Graduate School at U-M, a Young Investigator Award from ARO, CAREER and ITR Awards from NSF, and several IBM Partnership Awards.

More info:

“Computational Sprinting” Pushes Smartphones Till They’re Tired