CAREER: Fundamental Problems in Parallel Programming

职业：并行编程的基本问题

• 批准号: 9502889
• 负责人: Weijia Shang
• 金额: $ 13.33万
• 依托单位: Santa Clara University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 1995
• 资助国家: 美国
• 起止时间: 1995-07-01 至 1999-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9502889&HistoricalAwards=false
• 关键词: CAREER; Fundamental; Problems; Parallel; Programming

Parallel computers are difficult to program. Often, users have to specify the parallel execution, i.e., to specify data distributions to different local memory modules (as in High Performance Fortran), data communications between processors, and processor synchronizations. The insufficient programming support limits the applicability and further development of parallel systems. The objectives of this research are (1) to search and find solutions (optimal or feasible) for those fundamental problems in programming parallel systems such as how to distribute/replicate input data to different local memories in the system to minimize inter-processor communications, how to schedule and partition the computations to minimize the overall running time, how many processors should be used and how to choose the right message size for a message vector; (2) to develop new techniques for data alignment, scheduling partitioning and estimating the number of processors to be used for optimal execution of algorithms with nested loops; and (3) to implement these techniques and add them to some existing software tools such as PVM and MPI. Education: The education plans are as follows: (1) Undergraduate computer architecture lab at Santa Clara University has not been changed for six years and is obsolete. It is planned to update the lab and syllabi and redesign the students' projects to include recent advances in computer design, technology and debugging software to have students more motivated and prepared when they graduate. (2) Develop new parallel processing lab projects. One project is to have students run parallel programs in PVM environment so students can see and think those fundamental problems in parallel programming. (3) Be involved in mentoring women and underrepresented undergraduates and graduate students. (4) Do research projects with several senior undergraduate students, especially women engineering students to motivate them to pursue graduate studies in computer engineering at SCU

并行计算机很难编程。通常，用户必须指定并行执行，即指定数据分布到不同的本地内存模块（如在高性能Fortran中），处理器之间的数据通信和处理器同步。编程支持的不足限制了并行系统的适用性和进一步发展。本研究的目标是：(1)寻找并行系统编程中基本问题的解决方案（最优或可行），例如如何将输入数据分发/复制到系统的不同本地存储器中以减少处理器间的通信，如何调度和划分计算以减少总体运行时间，应该使用多少处理器以及如何为消息向量选择合适的消息大小；(2)开发数据对齐、调度分区和估计处理器数量的新技术，以优化嵌套循环算法的执行；(3)实现这些技术，并将其添加到一些现有的软件工具中，如PVM和MPI。教育：教育计划如下：(1)圣克拉拉大学本科计算机体系结构实验室已经六年没有变更，已经过时。计划更新实验室和教学大纲，重新设计学生的项目，以包括计算机设计、技术和调试软件的最新进展，使学生在毕业时更有动力和准备。(2)开发新的并行处理实验项目。其中一个项目是让学生在PVM环境中运行并行程序，这样学生就可以看到并思考并行编程中的基本问题。(3)参与指导女性和代表性不足的本科生和研究生。(4)与几名大四本科生，特别是工科女学生一起做研究项目，激励她们到四川大学攻读计算机工程专业的研究生