MRI: Acquisition of a Parallel Configurable Computer for Research in Engineering and the Computational Sciences

该项目来自HBCU，使用配备了一组现场可编程门阵列的传统PC获得并行和可配置的计算框架，开发执行速度远远快于最佳单/并行PC实施的应用程序实施。可配置计算系统允许将相同的硬件用于不同的并行算法的不同应用。因此，可以将硬件架构的新版本加载到相同的可配置集成电路中，而不是替换昂贵的系统硬件，从而降低系统成本。开发该系统的一个主要目标是以透明的方式将算法映射到可配置的计算机上的过程自动化。使用该实现的项目包括：-并行可配置数字信号处理，-针对应用数学和计算科学中的问题的并行算法开发，-PET扫描仪数据的3D图像重建的并行实现，-用于手持电子鼻的可配置数字信号处理子系统的建模，-三次样条内插，以及-基于传感器的供应链。这些应用中的每一个都将被映射到并行可配置计算系统上，以执行与最佳单/并行PC实现的比较，并进行适当的优化以在该并行可配置计算机上更快地执行数量级。此外，该基础设施还支持以下研究活动：-使用三维PET图像重建检测癌症肿瘤，-控制航天器的机载处理和NASA科学家和工程师所需的其他科学数据处理，-高性能图像处理，-气味补救和评估，以及-基于传感器的供应链。自动识别中心估计，全球每年潜在的供应节省在1,500-3,000亿美元之间。通过部署RFID，仅沃尔玛一家每年就可以节省83.6亿美元，其中包括通过避免缺货而节省的6亿美元，以及避免盗窃的5.75亿美元。此外，为了使学生和教师的代表性不足，这项研究在一个主要的HBCU进行，并与其他大学和联邦机构合作。

This project from an HBCU, acquiring a parallel and configurable computing framework using conventional PCs equipped with a set of FPGAs, develops application implementations that execute much faster than the best single/parallel PC implementation. Configurable computing systems allow the same hardware to be used for various applications for distinct parallel algorithms. Hence, rather than replacing expensive system hardware, new versions of a hardware architecture can be loaded into the same configurable integrated circuit reducing the system costs. A major goal in developing this system consists in automating the process of mapping algorithms onto a configurable computer in a transparent fashion. Projects using this implementation include:-Parallel configurable digital signal processing,-Parallel algorithm development for problems in applied mathematics and the computational sciences,-Parallel implementation of 3D image reconstruction for PET scanner data,-Modeling of a configurable digital signal processing subsystem for handheld electronic nose,-Cubic spline interpolation, and-Sensor-based supply chains.Each of these applications will be mapped onto the parallel configurable computing system to perform comparisons with the best single/parallel PC implementation and to make the appropriate optimizations to execute orders of magnitude faster on this parallel configurable computer. Moreover, the infrastructure supports the following research activities:-Detection of cancerous tumors using 3-D PET image reconstruction,-Onboard processing for control of spacecraft and other science data processing required by NASA scientists and engineers,-High performance image processing,-Odor remediation and assessment, and-Sensor-based supply chains.Broader Impact: The work exhibits potential to bring serious savings to product tracking through their life cycle. The Auto-ID Center has estimated potential global supply savings between $150 to $300 billion a year. Through RFID deployment, Wal-Mart alone could save $8.36 Billion annually, including $600 million through avoiding stock-outs and $575 million avoiding theft. Furthermore, enabling under-represented students and faculty, this research takes place at a major HBCU with collaboration with other universities and federal agencies.