Embedded Systems Code Optimization and Power Consumption

In a growing number of complex heterogeneous embedded systems, the relevance of software components is rapidly increasing. Issues such as development time, flexibility, and reusability are, in fact, better addressed by software-based solutions. Due to the processing regularity of multimedia and DSP applications, statically scheduled devices such as VLIW processors are viable options over dynamically scheduled processors such as state-of-the-art superscalar GPPs. The programs that run on a particular architecture will significantly affect the energy usage of a processor. The manner in which a program exercises certain parts of a processor will vary the contributions of individual structures to the total energy consumption of the processor. Minimizing power dissipation may be handled by hardware or software optimizations; in hardware through circuit design, and in software through compile-time analysis and code reshaping. While hardware optimization techniques have been the foci of several studies and are fairly mature, software approaches to optimizing power are relatively new. contents