The Internet of Things (IoT), contains a huge number of nodes and this amount is rapidly increasing. This makes it worthwhile to invest in energy saving for the edge nodes even when they are not battery powered. Using cheap and energy efficient ocessor based hardware for the edge nodes yields a significant contribution to green computing. However, this hardware has major restriction on the processing power and memory available. These restrictions impose substantial constraints on the software executable on such nodes. Task-oriented programming (TOP) offers a convenient way to program entire applications from a single source at a high level of abstraction. A tailor-made domain-specific dialect of TOP combines the advantages of this paradigm with the constraints of ocessors. In this paper we show that we can reduce the energy consumption of the edge nodes significantly in TOP. The system switches automatically to a low-power sleep mode when the current state of all tasks in the system allow a nap. In addition, we introduce a high-level way to deal with interrupts. This can replace the power hungry polling of sensors by much greener waiting for an event in a low-power mode of the system.