GPU Scheduler Modeling for Early Power-Performance Estimation of Mobile Applications

User experience and battery life are key concerns for smartphone makers. Given the growing trend of graphic-rich applications on mobile devices, embedded Graphics Processing Units (GPUs) are increasingly being incorporated in smartphone hardware platforms. In this project the intern will develop fast, early and accurate models of embedded GPUs, before the GPU hardware is available. These models will be useful for early optimization of smartphone applications by determining the appropriate application code to execute on the GPU and by using the appropriate policy to schedule jobs on the GPU across multiple applications, while minimizing energy consumption.

Umair Aftab
Faculty Supervisor: 
Dr. Samar Abdi