Architecture, Algorithms, and Circuits for an Energy-Efficient Single-Chip Tera-Op/sec Digital Signal Processor

高能效单芯片万亿次运算/秒数字信号处理器的架构、算法和电路

• 批准号: 0430090
• 负责人: Bevan Baas
• 金额: $ 15万
• 依托单位: University of California-Davis
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2004
• 资助国家: 美国
• 起止时间: 2004-09-15 至 2008-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0430090&HistoricalAwards=false
• 关键词: Architecture; Algorithms; Circuits; Energy; Efficient

Architecture, Algorithms, and Circuits for an Energy-Efficient Single-Chip Tera-Op/sec Digital Signal ProcessorAbstractDigital Signal Processing (DSP) workloads are typically numerically intensive, parallelizable, and increasingly required in a number of applications. The project will develop the architecture, algorithms, circuit, chip layout, and software for a single-chip processing system that is capable of calculating DSP workloads with high performance and high energy efficiency. Processing chips contain a very large number of small asynchronously clocked programmable processors connected by a reconfigurable 2-dimensional grid network. The research will result in the design, fabrication, test, and characterization of a prototype processing chip, including the development of software for a small number of complex multi-algorithm applications. The project will evaluate the merits of the proposed architecture for these types of DSP workloads. The research may lead to new capabilities that were previously constrained by maximum levels of power dissipation or throughput, in numerous applications such as life-extending medical devices, high volume consumer products, and very high throughput radar and image processors.

数字信号处理（DSP）工作负载通常是数字密集型的、可并行化的，并且要求越来越高 在许多应用中。该项目将为单芯片处理系统开发架构、算法、电路、芯片布局和软件，该系统能够以高性能和高能效计算DSP工作负载。 处理芯片包含大量的小型异步时钟可编程处理器，这些处理器通过可重新配置的二维网格网络连接。 该研究将导致原型处理芯片的设计，制造，测试和表征，包括为少量复杂的多算法应用程序开发软件。 该项目将评估这些类型的DSP工作负载的建议架构的优点。这项研究可能会带来新的功能，这些功能以前受到最大功耗或吞吐量的限制，适用于许多应用，如延长寿命的医疗设备，大批量消费产品以及超高吞吐量雷达和图像处理器。