A false-sharing-eliminable parallel tasks scheduling algorithm based on DAG
Li‐Sheng Wang, Kete Wang, Xixi Li · 2010
Since there exists the possibility for the parallel tasks in the structure of graph DAG to be executed sequentially due to the problem of Cache's False Sharing in muti-core CPU environment, this paper presents parallel task scheduling algorithm for DAG which could eliminate some of the occurrences of Cache's False Sharing. This algorithm which is an expansion of usual DAG task scheduling algorithm based on Wave Font Method is designed for decreasing the occurrences of Cache's false sharing. Also an experiment has been done to simulate the performance of this algorithm and the result shows that compared with WFM, the improved DAG task scheduling algorithm does decrease the occurrences number of tasks' sequential execution and reduce the total execution time of parallel tasks by 13%, and remain the muti-core CPU load-balanced.