An Efficient Ambient Noise Cross-Correlation Algorithm on Heterogeneous CPU-GPU Cluster

2022 IEEE 13th International Symposium on Parallel Architectures, Algorithms and Programming (PAAP) Pub Date : 2022-11-25 DOI:10.1109/PAAP56126.2022.10010612

Chao Wu, Xing Tan, Huikun Li, Guangzhong Sun

{"title":"An Efficient Ambient Noise Cross-Correlation Algorithm on Heterogeneous CPU-GPU Cluster","authors":"Chao Wu, Xing Tan, Huikun Li, Guangzhong Sun","doi":"10.1109/PAAP56126.2022.10010612","DOIUrl":null,"url":null,"abstract":"Calculation of noise cross-correlation functions (NCF) plays a vital role in ambient noise seismology. However high computation and storage requirements of NCF become huge obstacles in development of ambient noise seismology. Nowadays heterogeneous clusters show great impact in scientific computing. In this paper, we propose a parallel NCF algorithm based on heterogeneous cluster. Firstly, input SAC files are partitioned by dimension of dates and distributed to different nodes of heterogeneous cluster. Secondly, during NCF computation in each node, workloads are divided by station pairs, and calculated in different graphics processing units (GPUs) embedded in computing node. Finally, all NCF results of different dates are gathered to one computing node and stacked using all central processing units (CPUs) of that node. Experimental results demonstrate that the parallel NCF computing can be accomplished within much less time than that of the CPU counterpart, and the speedup remains almost linear when increasing number of computing nodes.","PeriodicalId":336339,"journal":{"name":"2022 IEEE 13th International Symposium on Parallel Architectures, Algorithms and Programming (PAAP)","volume":"152 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-11-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 IEEE 13th International Symposium on Parallel Architectures, Algorithms and Programming (PAAP)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/PAAP56126.2022.10010612","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 1

Abstract

Calculation of noise cross-correlation functions (NCF) plays a vital role in ambient noise seismology. However high computation and storage requirements of NCF become huge obstacles in development of ambient noise seismology. Nowadays heterogeneous clusters show great impact in scientific computing. In this paper, we propose a parallel NCF algorithm based on heterogeneous cluster. Firstly, input SAC files are partitioned by dimension of dates and distributed to different nodes of heterogeneous cluster. Secondly, during NCF computation in each node, workloads are divided by station pairs, and calculated in different graphics processing units (GPUs) embedded in computing node. Finally, all NCF results of different dates are gathered to one computing node and stacked using all central processing units (CPUs) of that node. Experimental results demonstrate that the parallel NCF computing can be accomplished within much less time than that of the CPU counterpart, and the speedup remains almost linear when increasing number of computing nodes.

查看原文本刊更多论文

一种高效的异构CPU-GPU集群环境噪声互相关算法

噪声互相关函数的计算在环境噪声地震学中起着至关重要的作用。然而，NCF的高计算和存储要求成为环境噪声地震学发展的巨大障碍。目前，异构集群在科学计算中发挥着重要的作用。本文提出了一种基于异构聚类的并行NCF算法。首先，将输入的SAC文件按日期维度进行分区，并分布到异构集群的不同节点;其次，在每个节点的NCF计算过程中，将工作负载按工位对划分，并在计算节点内嵌入的不同图形处理单元(gpu)中进行计算。最后，将不同日期的所有NCF结果收集到一个计算节点，并使用该节点的所有中央处理单元(cpu)进行堆叠。实验结果表明，与CPU计算相比，并行NCF计算可以在更短的时间内完成，并且随着计算节点数量的增加，加速几乎保持线性。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

2022 IEEE 13th International Symposium on Parallel Architectures, Algorithms and Programming (PAAP)

自引率

0.00%

发文量