Acquisition of a Multi-processor Computing Facility for Nonlinear Mathematical Field Problems

宾夕法尼亚州立大学理学院和工程学院的一个由教职员工和学生组成的团队提议购买一台多处理器计算机设备，以解决数学场方程。教师来自数学系、天文与天体物理系、航空航天工程系和机械工程系。该计划的目标是：在并行计算机体系结构上对科学和工程中的非定常、三维、非线性现象进行数值模拟；比较和分享应用于类似数学问题的数值方法；并将并行计算机上的数值解整合到研究生课程中。这些目标将通过以下方式实现：(1)建立一个多处理器计算机设施，供教职员工和学生团队使用；(2)对具有共同主题和挑战的数学问题进行研究；(3)开发用于研究生课程的模型问题和解决方案；(4)通过万维网传播研究和教学活动的结果。研究将在以下领域进行：黑洞碰撞；非定常流场与声学；电磁学;magneto-thermocapillary对流;地球物理流体流动。不同的数值方法将用于解决这些领域的问题，包括：自适应有限元；高阶有限差分;还有光谱技术。虽然研究课题涵盖了许多不同的现象，但它们都有共同的数学和计算挑战。它们都涉及三维，非定常，非线性，偏微分方程的解。它们都是涉及到很长时间尺度的问题。此外，它们都需要分析与时间相关的大型解决方案数据集。最后，在多处理器上高效运行并最大限度地利用硬件架构的数值方法对所有仿真都至关重要。这里提出的工具不仅具有先进的体系结构，而且还具有用于调试、分析和调优程序的开发工具。计算机还允许所有三种形式的并行处理：共享内存；消息传递（MPI和PVM）；和数据并行（Fortran 90和高性能Fortran）。本提案中的活动得到了Silicon Graphics Inc.和IBM的支持和合作。为了促进更多的合作和传播，研究小组将设立一个定期讲习班/研讨会系列。这项活动的目标是分享并行计算技术的知识，了解软件和算法的最新发展，并使设施用户了解其他小组成员的详细活动。除了研究活动中的共同线索外，拟议的研究将与我们的研究生教育活动有广泛的联系。宾夕法尼亚州立大学正在建立一个高性能计算中心。这部分是由美国国家科学基金会在其研究-课程发展（CRCD）联合计划下的拨款支持的。该中心将结合我们的研究活动和我们的研究生教育计划，包括辅修高性能计算。目前已经开设了五门新的应用并行处理研究生和本科生课程。学生访问拟议的研究设施，以及接触研究活动将为教学计划提供宝贵的增强

A team of faculty, staff and students in the College of Science and the College of Engineering at the Pennsylvania State University propose to acquire a multi-processor computer facility to enable the solution of mathematical field equations. The faculty are drawn from the Department of Mathematics, Astronomy and Astrophysics, Aerospace Engineering, and Mechanical Engineering. The goals of the program are: to develop numerical simulations on parallel computer architectures of unsteady, three-dimensional, non-linear phenomena in science and engineering; to compare and share numerical methods applied to similar mathematical problems; and to integrate numerical solutions on parallel computers into the gradua te curriculum. These goals will be achieved by (1) the establishment of a multi-processor computer facility accessible to a team of faculty, staff, and students, (2) the conduct of research on mathematical problems with common themes and challenges, (3) the development of model problems and solutions for use in graduate coursework and, ( 4) the dissemination of the results of the research and instructional activities through the World Wide Web. Research will be conducted in the areas of : black hole collisions; unsteady flow and acoustics; electromagnetics; magneto-thermocapillary convection; and geophysical fluid flows. Different numerical methods will be used to solve problems in these areas including: adaptive finite-elements; high-order finite-differences; and spectral techniques. Though the research topics cover many different phenomena, they all have common mathematical and computational challenges. They all involve the solution of three-dimensional, unsteady, non-linear, partial differential equations. They are all problems that involve a wide range of length and time scales. Also, they all require the analysis of large time-dependent solution data sets. Finally, numerical approaches that run efficiently on multiple processors and take maximum advantage of the hardware architecture are of vital importance to all the simulations. The facility proposed here has not only advanced architecture but also development tools for debugging, profiling, and tuning the programs. The computer also allows all three forms of parallel processing: shared memory; message-passing (MPI and PVM); and data parallel (Fortran 90 and High Performance Fortran). The activities in this proposal have the support of and cooperation with both Silicon Graphics Inc. and IBM. To promote additional cooperation and dissemination the research team will establish a regular workshop/seminar series. This activity will have the goals, to share knowledge on techniques of parallel computation, to keep abreast of new developments in software and algorithms and, to keep facility users informed in detail of the activities of the other team members. In addition to the common threads in the research activities, the proposed research will have extensive links to our graduate education activities. Penn State is establishing a Center for High Performance Computation. This is partially supported by a grant from the National Science Foundation under its Combined Research-Curriculum Development (CRCD) Program. The Center will combine our research activities and our graduate education program, including a minor in High Performance Computation. Five new graduate and undergraduate courses in applied parallel processing are already being offered. Access by students to the proposed research facility, as well as exposure to the research activities will provide an invaluable enhancement to the instructional program