Today, mobile devices such as smart phones and tablets have become ubiquitous and are used everywhere. Millions of software applications can be purchased and installed on these devices, customizing them to personal interests and needs. However, the frequent use of mobile devices has let a new problem become omnipresent: their limited operation time, due to their limited energy capacities. Although energy consumption can be considered as being a hardware problem, the amount of energy required by today’s mobile devices highly depends on their current workloads, being highly influenced by the software running on them. Thus, although only hardware modules are consuming energy, operating systems, middleware services, and mobile applications highly influence the energy consumption of mobile devices, depending on how efficient they use and control hardware modules. Nevertheless, most of today’s mobile applications totally ignore their influence on the devices’ energy consumption, leading to energy wastes, shorter operation times, and thus, frustrated application users. A major reason for this energy-unawareness is the lack for appropriate tooling for the development of energy-aware mobile applications. As many mobile applications are today behaving energy-unaware and various mobile applications providing similar services exist, mobile application users aim to optimize their devices by installing applications being known as energy-saving or energy-aware; meaning that they consume less energy while providing the same services as their competitors. However, scarce information on the applications’ energy usage is available and, thus, users are forced to install and try many applications manually, before finding the applications fulfilling their personal functional, non-functional, and energy requirements. This thesis addresses the lack of tooling for the development of energy-aware mobile applications and the lack of comparability of mobile applications in terms of energy-awareness with the following two contributions: First, it proposes JouleUnit, an energy profiling and testing framework using unit-tests for the execution of application workloads while profiling their energy consumption in parallel. By extending a well-known testing concept and providing tooling integrated into the development environment Eclipse, JouleUnit requires a low learning curve for the integration into existing development and testing processes. Second, for the comparability of mobile applications in terms of energy efficiency, this thesis proposes an energy benchmarking and labeling service. Mobile applications belonging to the same usage domain are energy-profiled while executing a usage-domain specific benchmark in parallel. Thus, their energy consumption for specific use cases can be evaluated and compared afterwards. To abstract and summarize the profiling results, energy labels are derived that summarize the applications’ energy consumption over all evaluated use cases as a simple energy grade, ranging from A to G. Besides, users can decide how to weigh specific use cases for the computation of energy grades, as it is likely that different users use the same applications differently. The energy labeling service has been implemented for Android applications and evaluated for three different usage domains (being web browsers, email clients, and live wallpapers), showing that different mobile applications indeed differ in their energy consumption for the same services and, thus, their comparison is both possible and sensible. To the best of my knowledge, this is the first approach providing mobile application users comparable energy consumption information on mobile applications without installing and testing them on their own mobile devices.

Today, mobile devices such as smart phones and tablets have become ubiquitous and are used everywhere. Millions of software applications can be purchased and installed on these devices, customizing them to personal interests and needs. However, the frequent use of mobile devices has let a new problem become omnipresent: their limited operation time, due to their limited energy capacities. Although energy consumption can be considered as being a hardware problem, the amount of energy required by today’s mobile devices highly depends on their current workloads, being highly influenced by the software running on them. Thus, although only hardware modules are consuming energy, operating systems, middleware services, and mobile applications highly influence the energy consumption of mobile devices, depending on how efficient they use and control hardware modules. Nevertheless, most of today’s mobile applications totally ignore their influence on the devices’ energy consumption, leading to energy wastes, shorter operation times, and thus, frustrated application users. A major reason for this energy-unawareness is the lack for appropriate tooling for the development of energy-aware mobile applications. As many mobile applications are today behaving energy-unaware and various mobile applications providing similar services exist, mobile application users aim to optimize their devices by installing applications being known as energy-saving or energy-aware; meaning that they consume less energy while providing the same services as their competitors. However, scarce information on the applications’ energy usage is available and, thus, users are forced to install and try many applications manually, before finding the applications fulfilling their personal functional, non-functional, and energy requirements. This thesis addresses the lack of tooling for the development of energy-aware mobile applications and the lack of comparability of mobile applications in terms of energy-awareness with the following two contributions: First, it proposes JouleUnit, an energy profiling and testing framework using unit-tests for the execution of application workloads while profiling their energy consumption in parallel. By extending a well-known testing concept and providing tooling integrated into the development environment Eclipse, JouleUnit requires a low learning curve for the integration into existing development and testing processes. Second, for the comparability of mobile applications in terms of energy efficiency, this thesis proposes an energy benchmarking and labeling service. Mobile applications belonging to the same usage domain are energy-profiled while executing a usage-domain specific benchmark in parallel. Thus, their energy consumption for specific use cases can be evaluated and compared afterwards. To abstract and summarize the profiling results, energy labels are derived that summarize the applications’ energy consumption over all evaluated use cases as a simple energy grade, ranging from A to G. Besides, users can decide how to weigh specific use cases for the computation of energy grades, as it is likely that different users use the same applications differently. The energy labeling service has been implemented for Android applications and evaluated for three different usage domains (being web browsers, email clients, and live wallpapers), showing that different mobile applications indeed differ in their energy consumption for the same services and, thus, their comparison is both possible and sensible. To the best of my knowledge, this is the first approach providing mobile application users comparable energy consumption information on mobile applications without installing and testing them on their own mobile devices.

Neben einer kurzen Ausführungszeit rückt bei der Optimierung von Anwendungen und Algorithmen ein geringer Energieverbrauch der genutzten Rechenressourcen in den Fokus der aktuellen Forschung. Eine hohe Energie-Effizienz von Programmen wird dabei erreicht, indem der Energieverbrauch von Programmen und Technologien reduziert wird, ohne dafür die Ausführungszeit übermäßig zu erhöhen. Im parallelen wissenschaftlichen Rechnen ist der Bedarf an energie-effizienten Programmausführungen vor allem für Algorithmen der linearen Algebra gegeben, die als Unterfunktionen in einer Vielzahl von Anwendungen eingesetzt werden. Die Vektorisierung von Programmen durch die Prozessor- und Instruktionssatzerweiterung AVX zeigt Potenzial zur energie-effizienten Ausführung von Algorithmen der linearen Algebra, wobei die erzielte Energie-Effizienz von der Umsetzung der Implementierung abhängt. Für die gezeigten Untersuchungen werden drei repräsentativ ausgewählte Algorithmen der linearen Algebra für die Ausführung auf AVX-Vektoreinheiten genutzt. Bei der AVX-Vektorisierung der Algorithmen werden verschiedene Programmvarianten erstellt, mit denen Ausführungszeit und Energieverbrauch bei der Ausführung ermittelt werden. Die Programmvarianten unterscheiden sich dabei unter anderem in der Anwendung von Programmtransformationen, wie Loop Tiling oder einer veränderten Speicherzugriffsstruktur. Zusätzlich wird gezeigt, wie die Umsetzung verschiedener Programmieransätze, wie Autovektorisierung oder unterschiedlicher Instruktionssätze, sowie Implementierungsvarianten durch die Auswahl der verwendeten Instruktionen, die Ausführungszeit und den Energieverbrauch der Programmausführung beeinflussen. Die so erstellten Programmvarianten werden auf modernen Prozessoren verschiedener Architekturfamilien mit unterschiedlichen Ausführungsparametern, wie der eingestellten Prozessorfrequenz, ausgeführt. Die Untersuchungen zeigen, dass sich Ausführungszeit und Energieverbrauch von Programmen durch die Vektorisierung reduzieren lassen. Die Auswahl der Programmtransformationen, des Programmieransatzes und der Ausführungsparameter für die energie-effiziente Ausführung von vektorisierten Programmen kann dabei anwendungsspezifisch aufgrund der Eigenschaften des ausgewählten Algorithmus getroffen werden.