Laboratory for Parallel Processing

基础设施奖是为了获得开发并行处理实验室所需的设备。主要收购的是处理器MIMD(多指令、多数据)并行计算机和工作站。将在软件工程、并行算法和数据结构、网络和模拟、计算机视觉和数据库管理系统方面进行研究。新体系结构并行计算机的出现开辟了与有效使用这些计算机相关的新的研究领域。MIMD计算机的工作原理是允许单独的处理器对独立的数据流进行操作。发现有效使用这种类型的体系结构的算法是一个跨越几个学科的主要研究问题。佛罗里达大学的并行处理实验室正在调查从实验算法研究到计算机视觉应用的几个研究课题。特别是：1.软件工程小组正在开发并行程序设计语言环境，以便与基于知识的开发和维护环境一起用于并行程序的开发。2.并行算法小组正专注于通过实验确定处理器通信和输入/输出设备的瓶颈。实验结果将与理论性能模型进行比较。3.网络和仿真小组将并行计算机用作负载均衡和其他性能增强算法的实验试验台，并作为执行复杂仿真的计算资源。4.计算机视觉小组正在开发利用并行性的算法。在许多视觉应用和MIMD架构之间有着天然的匹配。5.数据库管理小组正在设计算法，以便在MIMD体系结构的计算机上实现一个面向对象的语义数据库模型。//

The infrastructure award is for the acquisition of equipment to develop a laboratory for parallel processing. The major acquisition is a 64 processor MIMD (Multiple Instruction, Multiple Data) parallel computer and workstations. Research will be performed in software engineering, parallel algorithms and data structures, networks and simulation, computer vision, and database management systems. The advent of new architecture parallel computers has opened new areas of research related to the effective use of these computers. The MIMD computers work by allowing individual processors to operate on independent data streams. The discovery of algorithms to effectively use this type of architecture is a major research question crossing several disciplines. The parallel processing laboratory at the University of Florida is investigating several research topics ranging from experimental algorithm studies to computer vision applications. In particular: 1. The software engineering group is developing a parallel programming language environment to be used in the development of parallel programs in conjunction with a knowledge based development and maintenance environment. 2. The parallel algorithms group is concentrating on experimentally determining bottlenecks with both processor communication and input/output devices. The experimental results will be compared with theoretical models of performance. 3. The network and simulation group is using the parrallel computer as both an experimental testbed for load balancing and other performance enhancing algorithms and as a computational resource to perform complex simulations. 4. The computer vision group is developing algorithms to utilize parallelism. There is a natural match between many vision applications and the MIMD architecture. 5. The database management group is designing algorithms to implement an object oriented semantic database model on a MIMD architecture computer. //