Technical Report: Resource Awareness for Efficiency in High-Level Programming Languages

Managing hardware resources is important to write efficient software, which conserves energy, time, and money. High-level programming languages by definition abstract from the underlying machine, such that efficiency should be recouped by compiler and runtime. Our Invasive Computing project especially targets multi-application scenarios, where resources must be dynamically reallocated for optimal efficiency. Integrating hardware and software developers into the process, we developed a framework within the high-level X10 language, which enables an application to be aware of different kinds of resources and their availability. We show how applications can adapt to different architectures and dynamically exchange resources to optimize the system as a whole.