Advanced School for Computing and Imaging (ASCI)

ASCI office
Delft University of Technology
Building 28, room 04.E120
Van Mourik Broekmanweg 6
2628 XE – DELFT, The Netherlands

P: +31 15 27 88032

Visiting hours office
Monday, Tuesday, Thursday: 10:00 – 15:00


The ASCI office is located at the Delft University of Technology campus.  It is easily accessible by bicycle, public transport and car. The numbers of buildings can help you find your way around the campus. Make sure you remember the name and building number of your destination.

Contact us at +31 15 278 8032 or send us an email at

Energy- and time-aware scheduling for heterogeneous high-performance embedded systems

Energy- and time-aware scheduling for heterogeneous high-performance embedded systems

Author : Julius Philipp Roeder
Promotor(s) : Prof.dr. Andy Pimentel / Prof.dr.Clemens Grelck / Prof.dr. Sebastian Altmeyer
University : University of Amsterdam
Year of publication : 2023
Link to repository : Link to thesis


This thesis discusses the topic of high-performance embedded systems, their widespread usage, and the need to optimise their hardware utilisation concerning energy consumption and time. Throughout the thesis, various techniques are suggested to execute problems on less powerful hardware, thus saving resources and expenses. Alternatively, these techniques can be utilised to execute more complex problems using the same hardware.

First, the measurement setup used throughout the dissertation is detailed, as well as a set of experiments that determine the importance of the sampling rate for accurate energy measurement of high-performance embedded systems. Second, a novel energy model, system model, and energy-aware scheduler are introduced. They outperform existing schedulers. Third, various ranking algorithms are explored and compared. Fourth, a new task model is presented, which divides tasks into multiple phases, allowing for a more fine-grained separation of workloads. Lastly, a ranking-independent Reinforcement-Learning(RL)-based scheduler is presented. While performing similarly to a greedy heuristic, experiments show that including Graph Convolutional Neural Networks in the RL-scheduler improves the final schedules significantly.