Language Support for Parallel Scientific Computation on Distributed Memory Multiprocessors

该奖项支持对分布式内存多处理器（MIMD）上并行科学计算的编程语言支持的持续研究。越来越清楚的是，这类并行计算机将在未来的大规模科学计算中发挥重要作用。然而，目前这种计算机很难编程。这项研究的目标是在不牺牲结果代码的运行时效率的情况下，使这些计算机编程更容易、更自然地进行大规模并行科学计算。第二个目标是使这些程序在不同的分布式内存多处理器之间可移植。主要问题是对使用动态数据和过程结构的并行数值算法的高级语言支持。研究人员将与这些计算机（潜在）用户的科学家进行广泛的互动。这种互动的目的是增加对强调科学家并行算法的范式的理解，将这些范式整合到编程语言研究中，并评估这项研究如何满足他们的需求。//

This award supports continued research into programming language support for parallel scientific computation on (MIMD) distributed memory multiprocessors. It is becoming clear that this class of parallel computers will play an important role in the future of large scale scientific computation. Currently, however, such computers are very difficult to program. The goal of this research is to make it considerably easier and more natural to program these computers for large scale parallel scientific computation, without sacrificing the run-time efficiency of the resultant code. A second goal is to make such programs portable between different distributed memory multiprocessors. The main issues are high-level language support for parallel numerical algorithms that use dynamic data and process structures. The researchers will interact extensively with scientists who are (potential) users of these computers. This interaction is aimed at increasing understanding of the paradigms that underline the scientists' parallel algorithms, integrating these paradigms into programming language research, and assessing how well this research is meeting their needs.//