Software for Transient Parallel Adaptive Finite Element Computation

多维瞬态偏微分系统的计算要求很高，例如，可能涉及事故模拟、天气预报、化学和材料制造、湍流和燃烧。在并行计算机上平衡计算负载必须是动态的，因为h-或p-细化的自适应将破坏初始平衡。迄今为止的努力主要集中在h-细化上，而不是可能更有效的p-细化。本研究旨在开发必要的技术，以支持异构并行计算环境下瞬态偏微分系统的有效和可靠的自适应计算。这些技术将被整合到一个可扩展的软件框架中，并使用涉及流体流动、材料加工、化学反应和晶体生长的应用程序进行评估。

Multi-dimensional transient partial differential systems are computationally demanding and may, for example, involve accident simulation, weather forecasting, chemical and material fabrication, turbulence, and combustion. Balancing computational loading on a parallel computer must be dynamic, since adaptivity by either h- or p-refinement will upset an initial balance. Efforts to date have concentrated on h-refinement rather than the potentially more efficient p-refinement. This research seeks to develop the technologies necessary to support effective and reliable adaptive computation for transient partial differential systems on heterogeneous parallel computing environments. These technologies will be incorporated into an extensible software framework and evaluated using applications involving fluid flow, material processing, chemical reaction, and crystal growth.