EAGER: Productive and Scalable Utilization of Manycores in Big Data Applications

EAGER：在大数据应用中高效且可扩展地利用多核

• 批准号: 1346812
• 负责人: Soheil Ghiasi
• 金额: $ 12.75万
• 依托单位: University of California-Davis
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2013
• 资助国家: 美国
• 起止时间: 2013-09-01 至 2018-02-28
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1346812&HistoricalAwards=false
• 关键词: EAGER; Productive; Scalable; Utilization; Manycores

Many important electronics applications, including medical imaging, cyber security and digital entertainment, require processing massive amount of data under limited time, power and price budgets. Modern many-core processor chips offer a promising platform for computing such data-intensive workloads, however, their utilization is hindered due to the inherent difficulty and unproductivity of parallel software development at scale. The project aims to address this critical problem by developing methodologies and tools for productive and scalable development of software that would efficiently run on thousands of processor cores. Central to the approach is development of "functionally-consistent structurally-malleable streaming specification," which enables design-time exploration of a rich software implementation space, under the constraints of hardware platform, while maintaining the application's end-to-end functional behavior. The project results are expected to improve performance, energy consumption, and application development cost in a large number of data-driven disciplines, ranging from medical imaging and cyber security to multimedia and beyond.

许多重要的电子应用程序，包括医学成像，网络安全和数字娱乐，都需要在有限的时间，电力和价格预算下处理大量数据。现代多核处理器芯片为计算此类数据密集型工作负载提供了有前途的平台，但是由于并行软件开发的固有难度和非生产力，因此其利用受到了阻碍。该项目旨在通过开发方法和工具来解决这个关键问题，以实现有效地在数千个处理器内核上运行的软件的生产和可扩展性开发。该方法的核心是开发“功能上一致的结构可容纳的流媒体规范”，该规范可以在硬件平台的约束下对丰富的软件实现空间进行设计时间探索，同时维护应用程序的端到端功能行为。预计该项目的结果将改善大量数据驱动的学科的性能，能耗和应用开发成本，从医学成像和网络安全到多媒体及其他地区。