MRI: Acquisition of a Computational Environment for Scientific Computing

Sameh， Ahmed Purdue University MRI：获取用于科学计算的计算环境该项目提议获取一个多用途计算平台，作为高性能计算、几何建模和可视化以及并行计算机系统软件支持的前沿研究的联合顶级资源。提议的基础设施包括32个处理器SGI Origin 2000 （195 MHz R10000处理器可升级到250 MHz R10000，随后升级到300 MHz R12000处理器）和Infinite Reality Graphics（两个管道）。该基础设施是支持各种研究活动的理想平台。它为系统软件设计人员提供了一个内存组织，在这个组织上，他们可以测试和开发能够为深度缓存层次结构优化应用程序的软件。分布式共享内存为应用程序开发人员提供了使用线程和基于指令的api从串行程序到并行程序的轻松转换，同时仍然支持MPI等消息传递库。无限现实引擎为支持计算机密集型设计和仿真应用程序提供了强大的图形和可视化功能。随着R12000 cpu的升级，该基础设施在未来几年内可能仍将是最先进的计算设施。

EIA-9871053 Sameh, Ahmed Purdue University MRI: Acquisition of a Computational Environment for Scientific Computing This project proposes the acquisition of a multi-use computing platform to serve as a joint, top-of-the-line resource of cutting-edge research in high performance computing, geometric modeling and visualization and system software support for parallel computers. The proposed infrastructure consists of a 32 processor SGI Origin 2000 (with 195 MHz R10000 processors upgradable to 250 MHz R10000, and subsequently to 300 MHz R12000 processors) and Infinite Reality Graphics (two pipes). This infrastructure is an ideal platform for supporting various research activities. It provides system software designers with a memory organization on which they can test and develop software capable of optimizing applications for deep cache hierarchies. The distributed shared memory provides application developers an easy transition from serial to parallel programs using threaded and directive-based APIs while still supporting message passing libraries such as MPI. The infinite reality engines provide powerful graphics and visualization capabilities for supporting computer-intensive design and simulation applications. With the upgrade path to R12000 CPUs, the infrastructure is likely to remain a state-of-the-are computing facility for the next several years.