Accelerating datapath merging by task parallelisation on multicore systems

作     者：Fazlali, Mahmood Fallah, Mohammad K. Hosseinpour, Naemeh Katanforoush, Ali 

作者机构：Shahid Beheshti Univ Dept Comp Sci GC Tehran Iran Inst Res Fundamental Sci IPM Sch Comp Sci Tehran Iran 

出 版 物：《INTERNATIONAL JOURNAL OF PARALLEL EMERGENT AND DISTRIBUTED SYSTEMS》 (并行、紧急、分布式系统国际杂志)

年 卷 期：2019年第34卷第5期

页      面：615-628页

核心收录：

学科分类：08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：School of Computer Science  Institute for Research in Fundamental Sciences (IPM) [CS1395-4-670] 

主　　题：Datapath merging Branch and Bound algorithm task parallelisation shared memory paradigm 

摘      要：Datapath merging is an efficient approach to reduce hardware resources and configuration time in the synthesis of digital systems. In order to solve datapath merging, we have to find the maximum weighted clique, which is an NP-hard problem. So, datapath merging is a time-consuming process. In this article, we use OpenMP library to perform divide and conquer task parallelism to find the maximum weighted clique. Therefore, considerable reduction in the synthesis time and almost linear speedup has been achieved. The experimental results obtained from running this algorithm on different benchmarks represent speedup ranging from 1.2 times to 6.5 times for an 8-core system.