逻辑程序并行执行中的粒度分析方法

GRANULARITY ANALYSIS METHOD IN PARALLEL EXECUTION OF LAGIC PRAGRAMS

逻辑程序设计语言具有潜在丰富的并行性，但是由于任务并行执行时所需调度、通讯、同步等开销，使得一部分并行执行反而降低了执行效率。本文给出一种粒度分析方法。该方法在静态时通过分析估价出各子目标在单机上运行所需的ＣＰＵ时间ＴＣ，通过ＴＣ与基本粒度的比较产生该子句的最佳粒度执行流图ＢＥＧ（Ｎ’，Ｅ’），进而在ＢＥＧ（Ｎ’，Ｅ’）的基础上对子句ＣＧＥ进行优化。优化后的并行执行图表达式ＣＧＥ将具有最好的执行效率。

There is a lot of parallelisms in logic pragramming. But the cost of scheduling communication and synchronization of parallel execution of tasks may cut down the efficiency of multi--processor system. This article presents one granularity analysis method. By statically analyzing. the run-time of every goals running on single processon(TC) can be evaluated. Then the best granularity execution graph BEG (N'. E')the CGES can be producted.Based on the BEG(N. 'E')the CGES of every clause can be optimized. By granularity analysis, we can speed up the execution of LP programs actually.