顺序程序任务并行性的识别

探讨了任务并行性识别中任务执行条件的导出，利用任务给点间的相关性（包括数据相关和控制相关）就可求出并行执行条件．提出了层次任务图模型下直接生成优化任务执行条件的算法，省去了后控树及控制相关图（ＣＤＧ）的产生，其过程简单且比较高效．

The Identification of Task-level Parallelism in a Sequential Program

The identification of parallelism at the loop level has all along been the main target ofparallelizing compilers. The variety of new parallel architecture and recent progress in inter-procedural dependence analysis suggests new areas in the development of task-level paral-lelism of parallel compilation. The identification of task-level parallellism is studied with thefocus on the derivation of the execution conditions of tasks using the data dependence andcontrol dependence between tasks. A new algorithm for the formation of the optimized exe-cution conditon of tasks based on the hierarchy task graph model is proposed. The construc-tion of the post-dominate tree and control dependence graph(CDG)is not required here. Iftask data depend only on one other task, the task execution condition generated by this algo-rithm will have no redundancy. Otherwise,there may be redundancy.