A Framework for Estimating and Minimizing Energy Dissipation of Embedded HW/SW Systems  Permission to make digital/hard copy of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commerdal advantage, the copyright notice, the title of the publication and its date appear, and notice is given that copying is by permission of ACM, Inc. To copy otherwise, to republish, to post on servers or to redistribute to lists, requi

Abstract

Embedded system design is one of the most challenging tasks in VLSI CAD because of the vast amount of system parameters to fix and the great variety of constraints to meet. In this paper we focus on the constraint of low energy dissipation, an indispensable peculiarity of embedded mobile computing systems. We present the first comprehensive framework that simultaneously evaluates the tradeoffs of energy dissipations of software and hardware such as caches and main memory. Unlike previous work in low power research which focused only on software or hardware, our framework optimizes system parameters to minimize energy dissipation of the overall system. The trade-off between system performance and energy dissipation is also explored. Experimental results show that our Avalanche framework can drastically reduce system energy dissipation. IntroductionThe design of embedded systems is a challenging task for today's VLSI CAD environments. As opposed to a general purpose computing system, an embedded system performs just one particular application that is known a priori. Therefore, the system can be designed with respect to the particular application to have lower cost, higher performance, or be more energy-efficient. Energy efficiency is a hot topic in embedded system design. As mobile computing systems (e.g. cellular phones, laptop computers, video cams, etc.) become more popular, how to length the battery life of these systems becomes a critical issue.From the design process point of view, many of the embedded systems can be integrated on just one chip (systems on a chip) using core based design techniques. Previous work in core-based system design has mainly focused on performance and cost constraints. Some recent work has been presented in co-synthesis for low power [1,2]. However, the trade-off in energy dissipation among software 1 , memory and hardware has not yet been explored. This is a challenging and indispensable task for the design of low power embedded systems. Consider for example, that the use of a bigger cache can reduce the number of cache misses and speed up the software execution, which may cause less energy dissipation on the processor. On the other hand, a larger cache size also causes bigger switching capacitance for cache accesses and therefore increases the cache energy dissipation per access.In this paper we present our framework Avalanche, the first framework that explores the design space of hardware/software systems in terms of overall system energy dissipation. Since embedded system design usually has multiple constraints such as performance and power, our framework 1 We use the term software energy dissipation for the energy that is dissipated within a processor core. Permission to make digital/hard copy of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage, the copyright notice, the title of the publication and its date appear, and notice is given that copying is by pe...