Analysis of energy dissipation in the memory hierarchy of embedded systems: a case study

This paper presents an experimental study of the energy dissipation in the memory hierarchy of an embedded system during the execution of a specific application. Because of its ubiquity in many applications in different fields, we have chosen the Discrete Cosine Transform (DCT) as a benchmark for our study. The experiments consist of evaluating the energy dissipation of the memory hierarchy of the system, by comparing different software implementations of the DCT. An instruction-level simulator and a cache simulator have been used to perform energy estimation. The experiments show that the fastest (i.e., more hardware- or performance-efficient) implementation is not always the most beneficial from the memory energy stand-point