Efficient parallelization of MLFMA for electromagnetic scattering analysis of large-scale thin dielectric-coated object using MT3000 heterogeneous processor

IF 1.6 4区地球科学 Q3 ASTRONOMY & ASTROPHYSICS

Radio Science Pub Date : 2026-09-01 Epub Date: 2026-10-06 DOI:10.1029/2026RS008617

Zhaoyuan Wang;Jihong Gu;Jingtao Ye;Xin Chen;Mengmeng Li;Dazhi Ding;Guoqing Xiao

{"title":"Efficient parallelization of MLFMA for electromagnetic scattering analysis of large-scale thin dielectric-coated object using MT3000 heterogeneous processor","authors":"Zhaoyuan Wang;Jihong Gu;Jingtao Ye;Xin Chen;Mengmeng Li;Dazhi Ding;Guoqing Xiao","doi":"10.1029/2026RS008617","DOIUrl":null,"url":null,"abstract":"This paper proposes a parallel multilevel fast multipole algorithm (MLFMA) combined with the thin dielectric sheet (TDS) approximation-based method of moments, termed as TDS-MLFMA. Implemented on the homegrown MT3000 heterogeneous processor, this algorithm efficiently analyzes electromagnetic scattering characteristics of electrically large object coated with thin dielectric materials. A three-level parallel strategy tailored to the octree structure of TDS-MLFMA is adopted, employing boxes partitioning for low layers and plane waves partitioning for high layers, with a transition layer bridging high and low layers. The finest-layer aggregation is accelerated via heterogeneous computation by MT3000, and double-buffering mechanism is utilized to overlap computation and data transmission. Numerical examples verify the accuracy and efficiency of the proposed method. It achieves 1.5 speedup over CPU-only computation, and maintains above 50% parallel efficiency with processes ranging from 20 to 200.","PeriodicalId":49638,"journal":{"name":"Radio Science","volume":"61 9","pages":"1-11"},"PeriodicalIF":1.6000,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Radio Science","FirstCategoryId":"94","ListUrlMain":"https://ieeexplore.ieee.org/document/11721489/","RegionNum":4,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2026/10/6 0:00:00","PubModel":"Epub","JCR":"Q3","JCRName":"ASTRONOMY & ASTROPHYSICS","Score":null,"Total":0}

引用次数: 0

Abstract

This paper proposes a parallel multilevel fast multipole algorithm (MLFMA) combined with the thin dielectric sheet (TDS) approximation-based method of moments, termed as TDS-MLFMA. Implemented on the homegrown MT3000 heterogeneous processor, this algorithm efficiently analyzes electromagnetic scattering characteristics of electrically large object coated with thin dielectric materials. A three-level parallel strategy tailored to the octree structure of TDS-MLFMA is adopted, employing boxes partitioning for low layers and plane waves partitioning for high layers, with a transition layer bridging high and low layers. The finest-layer aggregation is accelerated via heterogeneous computation by MT3000, and double-buffering mechanism is utilized to overlap computation and data transmission. Numerical examples verify the accuracy and efficiency of the proposed method. It achieves 1.5 speedup over CPU-only computation, and maintains above 50% parallel efficiency with processes ranging from 20 to 200.

查看原文本刊更多论文

基于MT3000异构处理器的MLFMA高效并行化大规模介质薄膜物体电磁散射分析

本文提出了一种并行多电平快速多极子算法（MLFMA），并将其与基于薄介质片（TDS）近似的矩量方法相结合，称为TDS-MLFMA。该算法在国产MT3000异构处理器上实现，有效地分析了涂有薄介质材料的电大物体的电磁散射特性。采用针对TDS-MLFMA八叉树结构量身定制的三层并行策略，低层采用方框划分，高层采用平面波划分，并用过渡层桥接高低层。MT3000通过异构计算加速了最优层聚合，利用双缓冲机制实现了计算和数据传输的重叠。数值算例验证了该方法的准确性和有效性。与仅使用cpu的计算相比，它实现了1.5倍的加速，并且在进程从20到200的情况下保持了50%以上的并行效率。

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

Radio Science 工程技术-地球化学与地球物理

CiteScore

3.30

自引率

12.50%

发文量

112

审稿时长

1 months

期刊介绍： Radio Science (RDS) publishes original scientific contributions on radio-frequency electromagnetic-propagation and its applications. Contributions covering measurement, modelling, prediction and forecasting techniques pertinent to fields and waves - including antennas, signals and systems, the terrestrial and space environment and radio propagation problems in radio astronomy - are welcome. Contributions may address propagation through, interaction with, and remote sensing of structures, geophysical media, plasmas, and materials, as well as the application of radio frequency electromagnetic techniques to remote sensing of the Earth and other bodies in the solar system.