The paper combines advances in computational fluid dynamics with advances in computer architectures and programming paradigms to extend the state-of-the-art in large-scale computations of complex industrial flows. The present work combines traditional distributed memory programming with shared memory programming, and hence has the advantage of imposed data locality of the former as well as the ease and flexibility of the latter. Conceptually, it manifests the architectural features of Distributed Shared Memory (DSM) and complements the concept of hierarchical domain decomposition for cache friendly algorithms on microprocessor based parallel systmes. We demonstrate this concept through three-dimensional simulations of developing flow and heat transfer in a multilouvered fin array.
展开▼
