基于多核的粗粒度2.5维电磁场正演并行算法

吉林大学学报(理学版)

基于多核的粗粒度2.5维电磁场正演并行算法

白洪涛^1,2, 李昂^2,3, 欧阳丹彤^2,3, 邢书豪^2,4, 刘雪飞^2,4

1. 吉林大学公共计算机教学与研究中心, 长春 130012;2. 吉林大学符号计算与知识工程教育部重点实验室, 长春 130012;3. 吉林大学计算机科学与技术学院, 长春 130012; 4. 吉林大学软件学院, 长春 130012

收稿日期:2013-05-17 出版日期:2014-05-26 发布日期:2014-08-27
通讯作者: 欧阳丹彤 E-mail:ouyangdantong@163.com

Dimensional CSAMT Forward CoarseGrainedParallel Algorithms Based on Multicore

BAI Hongtao^1,2, LI Ang^2,3, OUYANG Dantong^2,3 , XING Shuhao^2,4, LIU Xuefei^2,4

1. Center for Computer Fundamental Education, Jilin University, Changchun 130012, China; 2. Key Laboratory of Symbolic Computation and Knowledge
Engineering ofMinistry of Education, Jilin University, Changchun 130012, China;3. College of Computer Science and Technology, Jilin University, Changchun 130012, China; 4. College of Software, Jilin University, Changchun 130012, China

Received:2013-05-17 Online:2014-05-26 Published:2014-08-27
Contact: OUYANG Dantong E-mail:ouyangdantong@163.com

摘要/Abstract

摘要：

基于CPU多核处理器实现了粗粒度2.5维电磁场并行正演算法, 使用OpenMP对串行算法的频率域进行粗粒度分解, 主线程进行任务分配, 从线程在CPU多核上并行计算各频率域电磁场值, 并分析了OpenMP并行效率与模型及核心数目的关系. 多组不同模型测试结果表明, 本文并行方法与串行方法得到的数值结果相同, 并获得了接近CPU逻辑核心数目的性能提升.

关键词: 2.5维正演, 频率分解, OpenMP, 粗粒度并行

Abstract:

A 2.5dimension controlled source audiofrequency magnetotelluric coarsegrained forward parallel computing algorithm based on multicore was implemented. This algorithm decomposes the different frequencies to each thread of the CPU core, while the master thread executed the allocation of parallel tasks and slave threads computed the electromagnetic field values of each frequency in parallel. The relationship between the number of cores and parallel efficiency was analyzed. The experiments demonstrate that our parallel algorithm can gain a nearlinear speedup of logical cores of CPU, and at the same time, acquire the same data accuracy.

Key words: 2.5D forward, frequency decomposition, OpenMP, coarsegrained parallel

中图分类号:

TP391

白洪涛, 李昂, 欧阳丹彤, 邢书豪, 刘雪飞. 基于多核的粗粒度2.5维电磁场正演并行算法[J]. 吉林大学学报(理学版), 2014, 52(03): 509-514.

BAI Hongtao, LI Ang, OUYANG Dantong, XING Shuhao, LIU Xuefei. Dimensional CSAMT Forward CoarseGrainedParallel Algorithms Based on Multicore[J]. Journal of Jilin University Science Edition, 2014, 52(03): 509-514.