Special Session

15.1 - Memory Characterization for Mobile Workloads

10:30 AM–12:00 PM Jun 26, 2018 (America - Los Angeles)

Room 3024

workloads in recent decades. While traditional workloads have been analyzed rigorously, such analysis for mobile workloads has not been done yet. We provide a comprehensive workload characterization for Google consumer workloads. In mobile systems, energy consumption is one of the biggest concerns due to thermal budget and battery capacity. Thus, we focus on application’s memory behavior, which occupies the significant portion of total system energy. We found that on average more than 50 % of total energy is consumed by moving data back and forth between processors and memories. Therefore, we believe that any successful mobile system architecture should solve the energy efficiency problem in the future.