Dynamic Algorithm Switching in Parallel Simulations using AOP

We present a modular approach to implementing dynamic algorithm switching for parallel scientific simulations. Our approach leverages modern software engineering techniques to implement fine-grained control of algorithmic behavior in scientific simulations as well as to improve modularity in realizing the algorithm switching functionality onto existing application source code. Through fine-grained control of functional behavior in an application, our approach enables design and implementation of application-specific dynamic algorithm switching scenarios. To ensure modularity, our approach considers dynamic algorithm switching as a separate concern with regard to a given application and encourages separate development and transparent integration of the switching functionality without directly modifying the original application code. By applying and evaluating our approach with a real-world scientific application to switch its simulation algorithms dynamically, we demonstrate the applicability and effectiveness of our approach to constructing efficient parallel simulations.