Optimization at runtime on a nanoprocessor architecture

Abstract

Our research addresses the need to efficiently execute and control the increasingly demanding and diverse nature of applications running on embedded systems. We propose optimizing an application at runtime, sharing execution resources between the runtime optimizations and the application. We use the TRIPS processor (developed by the University of Texas at Austin) to demonstrate runtime optimization using speculative slice execution on a nanoprocessor architecture (NA). Preliminary results are promising. Despite the current implementations limitations, we show speedups of 7%/25% (whole application/single task), with larger speedups are possible for future implementations.