CAREER: Advanced System-Level Support for Hybrid Multi-Tasking Computing

职业：混合多任务计算的高级系统级支持

• 批准号: 0952425
• 负责人: Katherine Morrow
• 金额: $ 40.71万
• 依托单位: University of Wisconsin-Madison
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2010
• 资助国家: 美国
• 起止时间: 2010-03-01 至 2016-02-29
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0952425&HistoricalAwards=false
• 关键词: CAREER; Advanced; System; Level; Support

This work investigates new methods to provide operating system support for hybrid computing systems. A hybrid system includes one or more instruction-based processors coupled with accelerator logic that exploits application parallelism to increase performance. The work focuses on accelerators constructed from reconfigurable hardware, such as field-programmable gate array (FPGA) logic. Reconfigurable hardware also reduces energy by directly implementing the needed computation instead of fetching, decoding, and executing instructions. Future work will extend the investigation to other types of accelerators, such as general-purpose graphics processing units (GPGPUs). In particular, the research studies the interdependence of accelerator management and software thread scheduling in multi-tasking hybrid systems, and proposes integrating these two system-level functions to increase the overall performance and energy-efficiency. This will encourage industry to further embrace hybrid computing in the development of new products, and thus, enable more capable and energy-efficient mobile devices; quiet multi-function household information appliances; portable medical devices; and more flexible, capable, and robust infrastructure systems in transportation, healthcare, and other key industries. This project also targets to improve teaching, learning, and diversity in Computer Engineering. Activities include developing curriculum and teaching aids, outreach activities, undergraduate research, and involvement in education-related committees. As a part of this project, participants will create new instructional materials for Computer Engineering topics at a variety of levels of understanding.

本文研究了为混合计算系统提供操作系统支持的新方法。混合系统包括一个或多个基于指令的处理器，该处理器与利用应用程序并行性来提高性能的加速器逻辑相耦合。这项工作的重点是由可重构硬件构建的加速器，例如现场可编程门阵列(FPGA)逻辑。可重构硬件还通过直接实现所需的计算而不是获取、解码和执行指令来降低能耗。未来的工作将把研究扩展到其他类型的加速器，如通用图形处理单元(GPGPU)。特别是，研究了多任务混合系统中加速器管理和软件线程调度的相互依赖关系，并提出将这两个系统级功能集成起来，以提高系统的整体性能和能效。这将鼓励行业在新产品开发中进一步拥抱混合计算，从而使更强大、更节能的移动设备、安静的多功能家用电器、便携式医疗设备以及交通、医疗等关键行业的更灵活、更强大的基础设施系统。该项目还旨在改善计算机工程的教、学和多样性。活动包括开发课程和教具、外展活动、本科生研究和参与与教育有关的委员会。作为这个项目的一部分，参与者将在不同的理解水平上为计算机工程主题创建新的教学材料。