Incorporating energy efficient data structures into modular software implementations for internet-based embedded systems

TL;DR: This paper introduces tradeoffs between energy consumption and on-chip memory space consumption during the implementation phase of such a system and shows that non-trivial data structure implementations can lead to a better matching of the software onto the platform.

Abstract: In current day software design of Internet applications, a lot of attention is paid to the reusability and extensibility of the design. However, due to an increased participation of embedded, hand-held devices in Internet activity, migrating code on the Internet has to be designed and implemented well in terms of energy consumption, execution speed, and on-chip memory space consumption. These parameters should be treated equally with respect to other more common design criteria. This paper introduces tradeoffs between energy consumption and on-chip memory space consumption during the implementation phase of such a system. The content-aware knowledge of the designer is exploited during the implementation so that energy efficient data structures are implemented in a relatively easy way. We also show that optimizing for energy consumption is not necessarily the same as optimizing for execution speed (or vice versa).Gains in energy consumption and/or execution speed are presented for only a relatively small overhead in on-chip memory space consumption. Depending on the behaviour of the adjacent software modules (of the module under investigation) on the one hand, and some general platform related criteria on the other hand, we show that non-trivial data structure implementations can lead to a better matching of the software onto the platform.