Memory characterization to analyze and predict multimedia performance and power in embedded systems

Main memory performance and power are becoming increasingly critical for embedded and general purpose computing systems. That is the primary motivation for us to focus on analyzing and modeling memory utilization for multimedia applications in embedded systems like cell-phones and PDAs with limited hardware resources and battery supply. In addition, the paper presents scenarios of utilizing the memory characteristic model to predict multimedia application performance, optimize embedded software, and adapt hardware resources to save system power while guaranteeing adequate performance. We present a simple, but effective indicator to identify the application´s dynamic memory and computational composition. Using this, we can easily and accurately forecast application performance and help determine optimal resources needed by the multimedia applications.