Output details
11 - Computer Science and Informatics
University of Edinburgh
Design-Space Exploration of Resource-Sharing Solutions for Custom Instruction Set Extensions
<02> Originality: This was the first paper to present a parametric heuristic that exposed the trade-offs between logic latency and design complexity.
Significance: Resource sharing is key to reducing hardware cost. The new paradigm presented in this paper was subsequently extended by the authors (published in DATE 2012) to show how trade-offs in complex design spaces can be learned automatically. This led to an EPSRC-funded research project (EP/I013539/1), investigating customized hardware to support visible light communications.
Rigour: Previous approaches typically searched for a single solution; this paper parameterised the design space and searched it efficiently.