Memory-aware system scenario approach energy impact

Abstract

System scenario methodologies propose the use of different scenarios, e.g., different platform configurations, in order to exploit variations in computational and memory needs during the lifetime of an application. In this paper several extensions are proposed for a system scenario based methodology with a focus on improving memory organisation. The conventional methodology targets mostly execution time while this work aims at including memory costs into the exploration. The effectiveness of the proposed extensions is demonstrated and tested using two real applications, which are dynamic and suitable for execution on modern embedded systems. Reductions in memory energy consumption of 40 to 70% is shown.