Approximate Hardware Techniques for Energy-Quality Scaling Across the System

Abstract

For error-resilient applications, such as machine learning and signal processing, a significant improvement in energy efficiency can be achieved by relaxing exactness constraint on output quality. This paper presents a taxonomy of hardware techniques to exploit the trade-off between energy efficiency and quality in various computer subsystems. We classify approximate hardware techniques according to target subsystem and support for dynamic energy-quality scaling.

Publication
In International Conference on Electronics, Information, and Communication
Di Wu
Di Wu
PhD student

A Wisconsin Badger in Computer Architecture!

Related