Until recently, performance has been the single most important issue in the design of hardware and software. However, with the proliferation of battery-powered embedded and portable devices and the ever-increasing clock frequencies of general-purpose processors, the problem of effective energy optimization has become very important for a wide variety of architectures. The goal of this project is to design and implement an energy-aware optimizing compiler framework that take as input a source code written in a high-level language, energy and performance constraints, architectural description and technology related parameters, and generates as output an energy/performance optimized executable and an estimation of energy consumption. Some key contributions include (i) high-level energy models that can be used by a compiler, (ii) a simulator infrastructure to validate compiler output, (iii) new, energy-oriented compiler optimizations, and (iv) investigating the necessary os and architectural support for effective cooperation between the compiler, os, and architecture so as to address the growing energy problem in a unified manner. The proposed energy/performance-aware approach represents a radical shift from pure performance-oriented strategies used so far in optimizing compilers, and will open new avenues of research in computing in a world that is increasingly becoming more energy-conscious.
|Effective start/end date||8/15/01 → 9/30/07|
- National Science Foundation: $250,000.00