(679e) A Novel Computational Framework for Reactive Flow and Multiphysics Simulations

Reactive flow simulations represent a challenging endeavour in computational science. The underlying coupling between fluid transport and chemical kinetics necessitates careful attention to algorithm design thus reducing code modularity and increasing software complexity. We present a framework that facilitates the implementation of transport equations using modern programming techniques. This framework consists of two C++ libraries that allow the abstraction of a differential equation using discrete operators and generic expressions. Then, by utilizing graph theory, the expressions are mapped into a graph and a solution algorithm is automatically generated. This functionality grants developers the capability to rapidly prototype new ideas and generate reliable simulation software without attention to algorithmic details. Furthermore, our approach exposes additional levels of parallelism such as algorithm decomposition and vectorized parallelism on field and stencil operations. The result is a transparent and easy to use tool that allows end users to develop scalable and highly efficient simulation code without specialized hardware knowledge.