众核编程环境中多模型协同共存的系统化方法研究

项目名称： 众核编程环境中多模型协同共存的系统化方法研究

项目编号： No.U1304603

项目类型： 联合基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 曹仰杰

作者单位： 郑州大学

项目金额： 29万元

中文摘要： 众核处理器技术的不断发展为实现高效能计算提供了契机。如何有效提升应用软件并行执行能力，实现资源优化配置和性能可扩展性成为众核系统面临的关键科学问题。针对上述问题，课题着重研究众核编程环境中多模型协同共存的系统化方法；从低层异构资源统一抽象、资源动态优化分区、运行时优化控制以及任务高效调度等方面进行研究，旨在构建众核环境中支持多编程模型高效协同运行的统一技术框架，改善新型众核系统的整体效能和可扩展能力；研究成果对高效能计算机系统结构的基础理论，以及《国家中长期科学和技术发展规划》提出的先进计算平台建设具有重要意义。

中文关键词： 众核处理器；并行编程模型；自适应协同调度；在线竞争分析；运行时系统

英文摘要： The emergence of many-core architecture provides a big opportunity for future high-productivity computing systems. And how to improve the efficiency of running parallel applications, resource optimization, and performance scalability are becoming key scientific issues on modern many-core platforms. Aiming at addressing these problems mentioned above, we focus on studying the systematic approach to collaborative coexistence of multiple models in many-core programming environment. Specifically, unified memory-accessing abstraction, dynamic resource partitioning, run-time optimization, and efficient scheduling technologies are studied hierarchically in our research, which help to build a unified adaptive scheduling framework for supporting multiple programming models on novel many-core platforms. In summary, the results obtained in our research provide benefit to the fundamental theories of high-performance computing systems, as well as the construction of advanced computing infrastructure supported by National Science and Technology Develop Long-term Planning.

英文关键词： Many-core processor；Parallel programming model；Adaptive and collaborative scheduling；Online competitive analysis；Runtime system