Research Internship / Forschungspraxis / IDP
Feel free to contact me for related topics.
The GPU is an inherent part of modern processor architectures. It is used for applications which require heavy graphical computations, for example games. Consequently, the GPU turns out to be a large power consumer in such applications. Especially on mobile devices like phones or tablets, the reduction of power consumption prolongs the life time of the battery.
In this work, we are looking for the connection between CPU and GPU workload. Therefore, we want to have a closer look at OpenGL calls. We want to identify patterns between the power consumption and the workload on the CPU, the GPU and the OpenGL calls made. Based on the results, we want to derive algorithms to predict the power consumption of applications – here games – based on the OpenGL calls made.
The development platform we use has the same processor architecture as the Samsung Galaxy S5 phone where an Android Kitkat distribution is running on. The platform has an integrated power measurement interface.
The following tasks have to be done:
- Log OpenGL calls and power consumption of GPU and CPU
- Analyze the power consumption related to different calls
- Characterize the OpenGL calls based on their power consumption
- Optional: Come up with a mechanism to scale CPU and GPU frequency based on OpenGL calls
- Profound Knowledge of C
- Knowledge of the Linux/Android kernel
- Helpful: Experience with OpenGL
- Helpful: Knowledge of the structure of the Android operating system
- Diligent, independent and well-organized work performance