CAREER:Programming Models for Green Software

Energy efficiency is a critical goal for modern computing: power consumption represents a significant portion of operational cost for data centers and cloud computing providers; wireless sensor networks are only effective when individual nodes do not deplete their power quickly after deployment; increasing battery life is high on the wish list of millions of laptop and smartphone users. Prior research on energy-efficient computing is largely focused on innovations in VLSI, architectures, operating systems, and compiler optimizations. This project explores how innovations in programming models can contribute to energy-efficient computing. A focus on programming models has the distinctive benefits of promoting application-specific energy reduction, ensuring high portability in heterogeneous computing platforms, and facilitating the development of maintainable green software. The project centers around the design of a novel programming system where characteristics in energy-aware programming--such as power modes--can be declared or inferred as types, over which a type system can reason rigorously and compositionally. Types can further guide standard energy reduction techniques such as dynamic voltage and frequency scaling, and be exploited for certifying program energy characteristics in the Internet era. The outcome of the project includes the design of an energy-aware language and a compiler with novel features for energy reduction and certification. The project also brings the awareness of energy efficiency into computer programming classes and fosters next-generation green-conscious programmers--the contributors of a more sustainable society.