ITR: Adaptive Information Processing through Precomputation

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

Future information appliances are expected to become increasingly mobile and operate under highly fluctuating workloads, thus imposing stringent design constraints on their energy dissipation, size, cost, performance, and quality of service. This project explores algorithmic and design automatin technologies for creating computing systems that dynamically adapt their operation to their workloads with the objective to maximize thieir energy efficiency and resource utilization. In the area of algorithms, research is conducted on the design and analysis of precomputation schemes that can be used to predict the load of a computation in the future and adapt system resources accordingly. In the area of design automation, the project is concerned with behavioral-level transformations and scheduling algorithms that rely on precomputation to generate energy-efficient hardware or software implementations of general computations.