Center for Research on Parallel Computation

莱斯大学与加州理工学院(Caltech)、阿贡国家实验室和洛斯阿拉莫斯国家实验室合作提出的这项提案要求提供资金，以建立一个并行计算研究科学技术中心。中心主任将由莱斯大学的肯·肯尼迪教授担任。该中心的主要目标将是使并行计算系统更易于使用。为了实现这一目标，该中心将利用每个参与机构的个人的独特技能，在多个层面上解决这一问题。将在使用/应用并行系统的所有方面发挥作用的补充领域是计算机体系结构、编程系统、计算数学和科学应用。特别是，该中心将在所有级别上解决并行编程问题：应用程序、算法、编译器和机器。必须通过设计和构建新的应用程序来在并行机器上解决重要的科学和工程问题，从而获得并行编程经验。必须为计算数学的基本问题开发新的、本质上是并行的算法。必须开发编程系统，使计算机用户能够在高抽象程度上解决问题，而不会因为科学家参与系统的体系结构细节而影响效率。最后，并行体系结构中的技术基础必须通过在利用底层组件的设计、简化生成高效代码的任务的设计以及合并在其中的算法之间的工程精明妥协来继续。这一努力将涉及来自许多不同领域的研究人员和科学家，导致这个真正多学科中心的所有参与者的交叉受精。

This proposal from Rice University, in collaboration with the California Institute of Technology (Caltech), Argonne National Laboratory, and Los Alamos National Laboratory requests funds to establish a Science and Technology Center for Research on Parallel Computation. The Director of the Center will be Professor Ken Kennedy of Rice University. The principal goal of the Center will be to make parallel computing systems easier to use. To achieve this goal, the Center will attack this problem on many levels by taking advantage of the unique skills of individuals from each of the participating institutions. The complementary areas that will be brought to bear on all aspects of use/application of parallel systems are computer architecture, programming systems, computational mathematics, and scientific applications. In particular, the Center will attack the problem of parallel programming at all levels: application, algorithm, compiler, and machine. Experience must be gained in parallel programming by designing and building new application programs that solve important scientific and engineering problems on parallel machines. New, inherently parallel algorithms must be developed for the fundamental problems of computational mathematics. Programming systems must be developed that permit the computer user to solve problems at a high level of abstraction without compromising efficiency by involvement of the scientist in the architectural details of the system. Finally the technological base in parallel architectures must be continued by engineering shrewd compromises between designs that make use of the underlying components, designs that simplify the task of generating efficient code, and the algorithms that are incorporated within them. This effort will involve researchers and scientists from many different areas resulting in the cross- fertilization of all involved in this truly multidisciplinary Center.