CSR-AES: Collaborative Research: Behavior-Based Speculative Parallelization and Optimization on Desktop Multiprocessors

CSR-AES：协作研究：桌面多处理器上基于行为的推测并行化和优化

• 批准号: 0720499
• 负责人: Xipeng Shen
• 金额: $ 15万
• 依托单位: College of William and Mary
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2007
• 资助国家: 美国
• 起止时间: 2007-09-01 至 2010-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0720499&HistoricalAwards=false
• 关键词: CSR; AES; Collaborative; Research; Behavior

The on-going shift in processor technology to favor multicoremultiprocessors is opening new opportunities for software speculation,where program code is speculatively executed to improve speed at thecost of having to monitor for errors and the risk of having tore-execute code when an error happens.This project develops a new, behavior-based approach, which allows auser or a profiling tool to parallelize or optimize a program based onpartial information about the program code and the input. It mainlydevelops two programming techniques: Behavior-oriented parallelization(BOP), which speculatively executes possibly parallel regions, andFast track, which uses software speculation to support the use ofunsafely optimized code.The exponential increase in microprocessor performance over the past30 years has had incalculable impact on science, commerce, government,and quality of life. Continuing this revolution through the comingdecade will depend on a large degree of processor-level parallelism.Behavior-based oftware speculation promises to improve the performanceof existing, sequential software and of new software that reuses anexisting code base. It simplifies parallelization and should thusimprove the productivity of future software development. The outcomeof this proposal will be a suite of techniques for general-purposesoftware, and new tools that will be transitioned to industrialpartners wherever possible, and will be used in both undergraduate andgraduate courses.

处理器技术正在向多核多处理器倾斜，这为软件推测打开了新的机会，在这种情况下，程序代码被投机性地执行以提高速度，代价是必须监控错误并在错误发生时重新执行代码的风险。这个项目开发了一种新的基于行为的方法，它允许用户或剖析工具根据关于程序代码和输入的部分信息来并行化或优化程序。它主要开发了两种编程技术：面向行为的并行化(BOP)和快速跟踪(Fast Track)，前者推测地执行可能的并行区域，后者使用软件推测来支持安全优化代码的使用。在过去30年中，微处理器性能的指数级增长对科学、商业、政府和生活质量产生了不可估量的影响。在接下来的十年中继续这场革命将在很大程度上依赖于处理器级的并行性。基于行为的软件推测承诺改善现有的顺序软件和重用现有代码库的新软件的性能。它简化了并行化，应该可以提高未来软件开发的生产率。这一提议的结果将是一套通用软件的技术，以及尽可能过渡到工业合作伙伴的新工具，并将在本科生和研究生课程中使用。