System-level Integration of Optics into Multiprocessor Interconnect Architecture

将光学器件系统级集成到多处理器互连架构中

• 批准号: 9411587
• 负责人: Timothy Pinkston
• 金额: $ 27万
• 依托单位: University of Southern California
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1994
• 资助国家: 美国
• 起止时间: 1994-09-01 至 1998-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9411587&HistoricalAwards=false
• 关键词: System; level; Integration; Optics; into

9411587 Pinkston Communication complexity and latency is a critical problem in multiprocessor systems. A major portion of communication latency is associated with the interconnect network. Optics has many advantages for achieving low latency, scalable interprocessor communication. Recently, many optical interconnects have been designed; however, little regard for system-level influences on the interconnect architecture were taken into account in these designs. The proposed research is to investigate an integrated design solution for low latency multiprocessor communication. Program behavior, architecture, and optical technology considerations are factored into the overall interconnect design. This will lead to a design capable of increased network throughput and optimal processor-memory connectivity. Engineering design principles will be developed for applying high bandwidth, reconfigurable optical interconnects to electronic digital multiprocessor systems. An optical interconnect strategy that incorporates latency reduction and hiding techniques will be developed. An interconnect simulator will be written to assist in the design and performance evaluation efforts. A small scale optical interconnect demonstrator unit will be built to show feasibility. Hence, an outgrowth of this research is an optical design framework (design methodology, simulation tools and prototype) that will build infrastructure and allow for empirical evaluation of a large design space. ***

9411587 Pinkston 通信复杂性和延迟是多处理器系统中的一个关键问题。 通信延迟的主要部分与互连网络相关。 光学器件在实现低延迟、可扩展的处理器间通信方面具有许多优势。 最近，已经设计了许多光互连；然而，这些设计中很少考虑系统级对互连架构的影响。 拟议的研究旨在研究低延迟多处理器通信的集成设计解决方案。 程序行为、架构和光学技术考虑因素均纳入整体互连设计中。 这将导致设计能够提高网络吞吐量和优化处理器内存连接。 将开发工程设计原理，将高带宽、可重新配置的光学互连应用于电子数字多处理器系统。 将开发一种结合了延迟减少和隐藏技术的光学互连策略。 将编写互连模拟器来协助设计和性能评估工作。 将建造一个小型光互连演示单元以展示可行性。 因此，这项研究的成果是一个光学设计框架（设计方法、模拟工具和原型），它将构建基础设施并允许对大型设计空间进行实证评估。 ***