Abstract-Embedded systems execute applications with varying performance requirements. These applications exercise the hardware differently depending on the computation task, generating varying workloads with time. Energy minimization with such workload and performance variations within (intra) and across (inter) applications is particularly challenging. To address this challenge we propose an online approach, capable of minimizing energy through adaptation to these variations. At the core of this approach is a reinforcement learning algorithm that suitably selects the appropriate voltage/frequency scaling (VFS) based on workload predictions to meet the applications' performance requirements. The adaptation is then facilitated and expedited through learning transfer, which uses the interaction between the application, runtime and hardware layers to adjust the VFS. The proposed approach is implemented as a power governor in Linux and extensively validated on an ARM Cortex-A8 running different benchmark applications. We show that with intraand inter-application variations, our proposed approach can effectively minimize energy consumption by up to 33% compared to the existing approaches. Scaling the approach to multi-core systems, we also demonstrate that it can minimize energy by up to 18% with 2X reduction in the learning time when compared with an existing approach.

Run-Time Management (RTM) systems are used in embedded systems to dynamically adapt hardware performance to minimise energy consumption. A significant challenge is that RTM software can require laborious manual adjustment across different hardware platforms due to the diversity of architecture characteristics. Model-driven development offers the potential to simplify the management of platform diversity by shifting the focus away from handwritten platform-specific code to platform-independent models from which platform-specific implementations are automatically generated. Furthermore, the use of formal verification provides the means to ensure that implementations are correct-by-construction. In this paper, we present a framework for automatic generation of RTM implementations from platformindependent formal models. The methodology in designing the RTM systems uses a high-level mathematical language, Event-B, which can describe systems at different abstraction levels. A code generation tool is used to translate platform-independent Event-B RTM models to platform-specific

Abstract-Run-Time Management (RTM) systems are used to control energy hooks at run-time to minimise the energy consumption of embedded systems with single and many-core processors. Typically, such RTM systems are aware of application requirements and utilise workload prediction and machine learning algorithms to estimate the optimal configuration. An RTM mechanism should not compromise the reliability or performance of the platform it is managing. Because of the potential complexity and interaction with the platform and its applications, we are using rigorous design methods that allow us to master the complexity and verify the correctness of our designs in a formal way. The formal RTM design can be verified earlier in the development process before implementation, which early verification can reduce the cost of fixing potential failures which can be very demanding in testing the system after implementation. In addition, the formal model of a RTM system can be automatically translated into executable code to be executed on the hardware. Automatic code generation reduces the efforts of hand-coded implementation and is portable across different architectures and Operating Systems (OSs). In this paper we propose a formal approach toward automatic generation of RTM system code, for a video decoder application, from a verified formal model of a RTM. The formal model of the RTM system is developed using the Event-B formal modelling language and is verified using theorem proving and model checking. The automatically generated RTM system has been integrated in an embedded platform as a Linux governor, and provides up to 4% improvement over Linux's default Ondemand governor.