Model & Software Development
Our overarching goal is to enable the rapid development of robust tools for large-scale, parallel simulation science and provide scalable visualization capability for large-scale datasets. Additionally, we have the strategic goal of supporting massively parallel scalability for forthcoming exascale architectures. Efforts will focus on exposing parallelism at three primary levels: domain decomposition, algorithm decomposition (concurrency in algorithms), and vector parallelism (concurrency in atomic operations such as array operations). Over the last fifteen years, as part of the ASCII C-SAFE program at the University of Utah, the Uintah Computational Framework (UCF) has been developed to support MPI parallelism based on domain decomposition. The UCF will serve as the starting point for the framework development in this project, but will be modified to support multilevel parallelism and provide a migratory path to exascale.
Model & Software Development Project Sites: