Scheduling strategies for multiprocessor real-time DSP
Scheduling strategies for multiprocessor real-time DSP
复制标题
DOI：
10.1109/glocom.1989.64160
复制
发表时间：
1989-11
期刊：
影响因子：
--
通讯作者：
Edward A. Lee;S. Ha
中科院分区：
文献类型：
--
作者：
Edward A. Lee;S. Ha
作者探讨的可能性，导致在低实施成本，并可以针对广泛的一类DSP（数字信号处理）的应用程序的自动编译器。他们定义了四类调度策略：（1）完全动态;（2）静态分配;（3）自定时;（4）完全静态。从（1）到（4），更多的调度活动在编译时执行，而在运行时执行的较少。作者认为，对于大多数DSP应用程序，自定时调度是最有吸引力的策略。在这种策略中，演员到处理器的分配和演员的激发顺序由编译器确定。在运行时只确定触发的定时。这种确定的运行时成本是最小的。此外，适合这种模型的自动调度技术在通用性和效率方面都在增长。
The authors explore the possibilities for automatic schedulers that result in low implementation cost and can target a broad class of DSP (digital signal processing) applications. They define four classes of scheduling strategies: (1) fully dynamic; (2) static assignment; (3) self-timed; and (4) fully static. Moving from (1) to (4), more scheduling activity is performed at compile time and less at run time. The authors argue that for most DSP applications, self-timed scheduling is the most attractive strategy. In this strategy, the assignment of actors to processors and the ordering of the firing of actors are determined by a compiler. Only the timing of the firing is determined at run time. The run-time cost of this determination is minimal. Furthermore, automatic scheduling techniques that fit this model are growing in generality and efficiency.>