XPS: CLCCA (XPS: DSD) Future Extreme Scale Frameworks using DSL and ERTS

XPS：CLCCA（XPS：DSD）使用 DSL 和 ERTS 的未来极端规模框架

• 批准号: 1337145
• 负责人: Martin Berzins
• 金额: $ 70万
• 依托单位: University of Utah
• 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: 2013-09-15 至 2017-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1337145&HistoricalAwards=false
• 关键词: XPS; CLCCA; DSD; Future; Extreme

This project will lay the foundations for solving large multi-scale multi-physics engineering problems using new computational frameworks on the next decade's exascale computers. These frameworks will anticipate trends in proposed future computer hardware by being adaptive, asynchronous, fault-tolerant, and energy-aware and will address the possible billion-way parallelism of exascale systems by both generating efficient specific code through a domain specific language (DSL) and by efficiently scheduling that code through an Exascale Run Time System, ERTS.The project's prototype computational framework will use Directed Acyclic Graphs (DAGs) to generate code in the DSL that is specialized for multicore nodes and/or accelerators and it will also use DAGS at the runtime system level in the ERTS. At a nodal multicore or accelerator level the DSL will be a declarative, high-level, type-safe domain-specific language for multi-scale, multi-physics simulations that will improve productivity by automating the writing of optimized code that is executed by the ERTS. In order to ensure that these activities produce relevant solutions for important engineering applications with high impact such as, for example, the design of new batteries and fuel cells or clean coal boilers, the project will employ a multi-disciplinary approach that couples computer science advances to the solution of such meaningful engineering problems.

该项目将为在未来十年的亿级计算机上使用新的计算框架来解决大型多尺度多物理工程问题奠定基础。这些框架将通过自适应、异步、容错和能源感知来预测拟议的未来计算机硬件的趋势，并将通过通过域特定语言(DSL)生成高效的特定代码和通过Exascale运行时系统ERTS有效地调度该代码来解决亿级系统可能的亿级并行性。该项目的原型计算框架将使用有向无环图(DAG)来生成DSL中专门用于多核节点和/或加速器的代码，它还将在ERTS中的运行时系统级别使用DAG。在节点多核或加速器级别，DSL将是一种声明性的、高级的、特定于类型安全的域特定语言，用于多尺度、多物理模拟，它将通过自动编写由ERTS执行的优化代码来提高生产率。为了确保这些活动为具有高度影响力的重要工程应用提供相关的解决方案，例如，新电池和燃料电池的设计或洁净煤锅炉的设计，该项目将采用多学科方法，将计算机科学的进步与解决这些有意义的工程问题结合起来。