Jesse Lu;David Qu;Jim Qu;Ryan Fong;Geun Ho Ahn;Jelena Vučković
{"title":"Overcoming Memory Bandwidth Limitations In GPU-Accelerated Finite-Difference Time-Domain Simulations: A systolic update scheme","authors":"Jesse Lu;David Qu;Jim Qu;Ryan Fong;Geun Ho Ahn;Jelena Vučković","doi":"10.1109/MAP.2026.3668263","DOIUrl":"https://doi.org/10.1109/MAP.2026.3668263","url":null,"abstract":"The exponential growth of artificial intelligence (AI) has fueled the development of high-bandwidth photonic interconnect fabrics as a critical component of modern AI supercomputers <xref>[1]</xref>. As the demand for ever-increasing AI compute and connectivity continues to grow, the need for high-throughput photonic simulation engines to accelerate and even revolutionize photonic design and verification workflows will become an increasingly indispensable capability for the integrated photonics industry. Unfortunately, the mainstay and workhorse of photonic simulation algorithms, the finite-difference time-domain (FDTD) method <xref>[10]</xref>, is a memory-intensive and computationally lightweight algorithm that is fundamentally misaligned with modern computational platforms, which are equipped to deal with compute-heavy, memory-light workloads instead. This article proposes a systolic update scheme for the FDTD method, which circumvents this mismatch by reducing the need for global synchronization as well as minimizing the amount of global memory access needed per cell update by aggressively reusing existing field values already present within the GPU cache. We present an initial prototype of our scheme on a full 3D FDTD algorithm that achieves a performance of <inline-formula><tex-math>$sim{0.15}$</tex-math></inline-formula> trillion cell updates per second (TCUPS) on a single Nvidia H100 GPU. Our work paves the way for the increasingly efficient, cost-effective, and high-throughput photonic simulation engines needed to continue powering the AI era.","PeriodicalId":13090,"journal":{"name":"IEEE Antennas and Propagation Magazine","volume":"68 3","pages":"25-31"},"PeriodicalIF":5.7,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11435533","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148171853","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Remembering Fred E. Gardiol [In Memoriam]","authors":"Juan R. Mosig;Michel Ney","doi":"10.1109/MAP.2026.3681507","DOIUrl":"https://doi.org/10.1109/MAP.2026.3681507","url":null,"abstract":"Recounts the career and contributions of Fred E. Gardiol.","PeriodicalId":13090,"journal":{"name":"IEEE Antennas and Propagation Magazine","volume":"68 3","pages":"70-71"},"PeriodicalIF":5.7,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11552438","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148171744","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
L. Baumann;J. A. Hawkins;H. M. Aktulga;D. Dault;B. Shanker
{"title":"Multilevel Fast Multipole Algorithm (MLFMA)-Augmented Decoupled Field and Potential Formulations for Analysis of Scattering From Dielectric Objects: Accelerated analysis via decoupled potential and decoupled field formulations","authors":"L. Baumann;J. A. Hawkins;H. M. Aktulga;D. Dault;B. Shanker","doi":"10.1109/MAP.2026.3686370","DOIUrl":"https://doi.org/10.1109/MAP.2026.3686370","url":null,"abstract":"Recently, multiple promising formulations have been developed that appear to address some of the oldest and most challenging problems in computational electromagnetics, specifically formulations that overcome both low-frequency and dense-mesh breakdowns. Two of these equations are called the decoupled potential integral equation (DPIE) and decoupled field integral equation (DFIE) and are intimately related to each other. The efficacy of both, in terms of condition number across a broad range of frequencies, has been demonstrated via canonical basis functions on a sphere. Only recently has the DPIE been implemented for piecewise tessellations on electrically small objects, while the DFIE has been implemented using the Nyström discretization. One of the challenges in implementing both integral equations is the number of operators involved; a straightforward remedy to amortize costs is to use the multilevel fast multipole algorithm (MLFMA) scheme such that the requisite matrix vector products are effected via the minimum number of operations. Thus, the main goal of this article is three-fold: 1) review the construction of the DPIE and DFIE, 2) discuss the implementation of the DFIE on tessellated surfaces, and 3) present an efficient implementation of the MLFMA for these operators. We also present a range of results that demonstrate the computational efficiency and efficacy on electrically large multiscale structures. In addition, we will present some analytical results that shed light on the behavior of these equations in a wide frequency band.","PeriodicalId":13090,"journal":{"name":"IEEE Antennas and Propagation Magazine","volume":"68 3","pages":"42-52"},"PeriodicalIF":5.7,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148171820","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Call for Ideas","authors":"","doi":"10.1109/MAP.2026.3686084","DOIUrl":"https://doi.org/10.1109/MAP.2026.3686084","url":null,"abstract":"","PeriodicalId":13090,"journal":{"name":"IEEE Antennas and Propagation Magazine","volume":"68 3","pages":"3-3"},"PeriodicalIF":5.7,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11552426","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148171887","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Short Courses","authors":"Raymond P. Wasky","doi":"10.1109/MAP.2026.3684934","DOIUrl":"https://doi.org/10.1109/MAP.2026.3684934","url":null,"abstract":"","PeriodicalId":13090,"journal":{"name":"IEEE Antennas and Propagation Magazine","volume":"68 3","pages":"54-54"},"PeriodicalIF":5.7,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11552435","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148171877","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}