PRISM: A Laboratory for Research in Future High-Performance Parallel Computing

这个基础设施奖是为了奖励获得一台SIMD和一台MIMD并行计算机，以支持并行编程工具、编程语言、数据库、并行优化算法、科学计算、计算机体系结构、人工智能和并行计算模型等领域的研究。这项研究将涉及开发专门针对SIMD和MIMD结构的算法，并将这些计算机用作部门网络上的计算服务器。高性能计算的下一步发展将是高速并行处理器。目前这些处理器主要有两类--单指令多数据(SIMD)处理器和多指令多数据(MIMD)处理器。SIMD计算机在其所有处理元件上执行单个指令序列。每个处理单元可以具有其正在处理的不同数据。MIMD计算机可以具有在其每个处理元件上运行的不同程序，该处理元件可以具有其自己的数据。为了开发使用这些计算机所需的软件，必须具有支持仪器、程序开发和资源分配的软件工具。这类工具的开发得到了这一基础设施的支持。确定哪些问题在哪种类型的处理器上运行得最好也很重要。因此，这些计算机将用于人工智能、数据库和计算机体系结构的研究，以确定这些问题领域的有效算法。最后，这些计算机是高效的数字处理器，将用于科学计算和数值优化问题。

This infrastructure award is for the acquisition of a SIMD and a MIMD parallel computer for the support of research in the areas of parallel programming tools, programming languages, databases, parallel optimization algorithms, scientific computing, computer architecture, artificial intelligence, and models of parallel computation. The research will involve both the development of algorithms specific to the SIMD and MIMD architectures and the use of these computers as compute servers over the departmental network. The next development in high performance computing will be high speed parallel processors. There are currently two main classifications of these processors-Single Instruction Multiple Data (SIMD) and Multiple Instruction Multiple Data (MIMD) processors. SIMD computers execute a single sequence of instructions on all of its processing elements. Each processing element may have different data that it is processing. A MIMD computer may have a different program running on each of its processing elements that may have their own data. In order to develop the software necessary to utilize these computers it is necessary to have in place software tools that support instrumentation, program development, and resource allocation. The development of such tools is supported by this infrastructure. It is also important to determine which problems run best on which type of processor. Thus these computers will be used in research in artificial intelligence, databases, and computer architecture in determining efficient algorithms for these problem domains. Finally, these computers are efficient numeric processors and will be used for both scientific computing and numeric optimization problems.