SHF: Small: Reducing the Cost of Computation in CMPs

University Of California-San Diego, La Jolla CA

Investigators

Abstract

"This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5)." Computer processor industry has moved fully into the multi-core era to enable continual scaling of performance, but at the cost of increased energy consumption and increased cooling costs due to higher temperatures and thermal gradients. This proposal describes three major research thrusts that address these costs in multiple ways: (1) New modeling and simulation tools: We will integrate performance, power, temperature, reliability and cooling estimation, so that the designers will be able to analyze the impact of design choices and runtime decisions over significant time spans. (2) Runtime thread scheduling policies: We will identify and demonstrate power and thermal scheduling mechanisms that maintain performance, reduce the total energy consumption, and eliminate or reduce hot spots, but also maximize processor lifetime. The policies will use the data from thermal sensors and performance counters to proactively drive the management decisions. (3) New cooling strategies: Our goal is to create thermal management and cooling control algorithms that work in tandem to reduce the overall energy consumption. The proposed research forms the basis for discovery and learning in the areas of multi core processors, and, more generally, system design and management. Graduate and undergraduate students will be involved in various parts of the proposed research and help in connecting this work with other NSF sponsored projects. The results of research, tools and coursework materials developed will be freely and easily distributed to engineering community at large.