Research Initiation Award: Massively Parallel Processing Based on the Decoupled Synchronization Computation Model of Execution

研究启动奖：基于执行解耦同步计算模型的大规模并行处理

• 批准号: 9309572
• 负责人: Paraskevas Evripidou
• 金额: $ 9.63万
• 依托单位: Southern Methodist University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1993
• 资助国家: 美国
• 起止时间: 1993-08-15 至 1995-03-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9309572&HistoricalAwards=false
• 关键词: Research; Initiation; Award; Massively; Parallel

9309572 Evripidou The Decoupled Data-Driven (D3) machine separates the synchronization and computation portions of a program and executes them asynchronously. The compiler generates threads that utilize the processor pipeline efficiently by reducing pipeline stalls within a thread. Priority is given to threads which have their data in close proximity (i.e. in the cache or registers). Decoupling synchronization and computation ensures the execution pipeline is not interrupted by long synchronization events. The overall goal of this project is the implementation of the computation model into an abstract machine that will execute on existing large scale shared memory multiprocessors. Achieving this goal requires the development of: a microtasking environment that supports the abstract machine on commercial machines, a multi-language compiler that generates code for the abstract machine, and a hardware thread synchronization unit. ***

9309572 Evripidou解耦数据驱动（D3）机器分离程序的同步和计算部分，并异步执行它们。 编译器生成的线程通过减少线程内的流水线停顿来有效地利用处理器流水线。 优先级被给予那些数据非常接近的线程（即在该高速缓存或寄存器中）。 解耦同步和计算确保执行管道不会被长同步事件中断。 这个项目的总体目标是实现的计算模型到一个抽象的机器，将执行现有的大规模共享内存多处理器。 实现这一目标需要开发：一个支持商业机器上的抽象机器的微任务环境，一个为抽象机器生成代码的多语言编译器，以及一个硬件线程同步单元。 ***