Goal-Driven Reconfiguration Of Polymorphous Architectures

Polymorphous computing architectures refer to computing platforms whose computation and communication structures can be changed over time. The objective of such platforms is to use the underlying reconfigurable components and attributes to adapt to dynamically changing constraints, objectives, and characteristics in the applications that execute on them. In this paper, we present a model for executing applications with non-deterministic execution times and time-varying performance requirements on a polymorphous architecture. We analyze the complexity of various issues related to the model, and identify some broadly-applicable conditions under which the complexity of these issues can be reduced. We develop a heuristic framework for guiding the runtime configuration adaptation process, and show through simulation experiments that this approach can efficiently handle both dynamics in performance requirements and in task execution times.