{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,8]],"date-time":"2026-10-08T06:52:12Z","timestamp":1791442332099,"version":"4.3.3"},"reference-count":42,"publisher":"IEEE","funder":[{"DOI":"10.13039\/100000185","name":"DARPA","doi-asserted-by":"publisher","award":["HR0011-25-3-0242"],"award-info":[{"award-number":["HR0011-25-3-0242"]}],"id":[{"id":"10.13039\/100000185","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100014036","name":"MURI","doi-asserted-by":"publisher","award":["FA9550-24-1-0327"],"award-info":[{"award-number":["FA9550-24-1-0327"]}],"id":[{"id":"10.13039\/100014036","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000015","name":"DOE","doi-asserted-by":"publisher","award":["DE-SC0023171"],"award-info":[{"award-number":["DE-SC0023171"]}],"id":[{"id":"10.13039\/100000015","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100006227","name":"Lawrence Livermore National Laboratory","doi-asserted-by":"publisher","award":["DE-AC52-07NA27344 (LLNL-PROC-2014338)"],"award-info":[{"award-number":["DE-AC52-07NA27344 (LLNL-PROC-2014338)"]}],"id":[{"id":"10.13039\/100006227","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,6]]},"DOI":"10.23919\/isc.2026.11520474","type":"proceedings-article","created":{"date-parts":[[2026,6,8]],"date-time":"2026-06-08T19:48:54Z","timestamp":1780948134000},"page":"1-12","source":"Crossref","is-referenced-by-count":0,"title":["Accelerating High-Order Finite Element Simulations at Extreme Scale with FP64 Tensor Cores"],"prefix":"10.23919","author":[{"given":"Jiqun","family":"Tu","sequence":"first","affiliation":[{"name":"NVIDIA Corporation,USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ian","family":"Karlin","sequence":"additional","affiliation":[{"name":"Queen's University,Canada"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"John","family":"Camier","sequence":"additional","affiliation":[{"name":"Lawrence Livermore National Laboratory,USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Veselin","family":"Dobrev","sequence":"additional","affiliation":[{"name":"Lawrence Livermore National Laboratory,USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Tzanio","family":"Kolev","sequence":"additional","affiliation":[{"name":"Lawrence Livermore National Laboratory,USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Stefan","family":"Henneking","sequence":"additional","affiliation":[{"name":"The University of Texas at Austin,USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Omar","family":"Ghattas","sequence":"additional","affiliation":[{"name":"The University of Texas at Austin,USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","volume-title":"NVIDIA Tesla V100 GPU Architecture","year":"2017"},{"key":"ref2","volume-title":"NVIDIA V100"},{"key":"ref3","volume-title":"NVIDIA GB200"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1017\/s0962492921000064"},{"key":"ref5","volume-title":"NVIDIA A100"},{"key":"ref6","volume-title":"NVIDIA GH200"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1021\/acs.jctc.2c00241"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1016\/j.cpc.2011.12.006"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1137\/1.9781611971811"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1109\/ipdps49936.2021.00114"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1145\/3774934.3786414"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1145\/3695466"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1145\/3712285.3771787"},{"key":"ref14","volume-title":"MFEM: Modular Finite Element Methods [Software]"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1016\/j.camwa.2020.06.009"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1007\/3-540-47789-6_66"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1145\/1089014.1089020"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.2172\/1968587"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1177\/10943420241261981"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1007\/978-1-4757-3951-0"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1177\/10943420231175462"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1016\/j.jcp.2026.114682"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1137\/24m1683172"},{"key":"ref24","article-title":"Real-time probabilistic tsunami forecasting in Cascadia from sparse offshore pressure observations","author":"Henneking","year":"2026","journal-title":"arXiv"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1007\/s10596-015-9472-0"},{"key":"ref26","volume-title":"SIMT architecture"},{"key":"ref27","volume-title":"Shared memory"},{"key":"ref28","volume-title":"Matrix fragments for mma.m8n8k4 with. f64 floating point type"},{"key":"ref29","volume-title":"Multiply-and-accumulate instruction: mma"},{"key":"ref30","volume-title":"CUTLASS"},{"key":"ref31","volume-title":"CUBLAS"},{"key":"ref32","volume-title":"CUBLASDx"},{"key":"ref33","volume-title":"TOP500"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1109\/ipdpsw.2018.00091"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1177\/10943420241239588"},{"key":"ref36","article-title":"Ozaki Scheme II: A GEMM-oriented emulation of floating-point matrix multiplication using an integer modular technique","author":"Ozaki","year":"2025","journal-title":"arXiv"},{"key":"ref37","volume-title":"ECP Milestone Report CEED-MS40: Support ECP applications in their exascale challenge problem runs","author":"Kolev","year":"2023"},{"key":"ref38","volume-title":"CEED: Center for Efficient Exascale Discretizations, Exascale Computing Project (ECP)"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-86169-7_33"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1145\/3394277.3401860"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1016\/j.jocs.2025.102659"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-63759-9_31"}],"event":{"name":"ISC High Performance 2026 Research Paper Proceedings (41st International Conference)","location":"Hamburg, Germany","start":{"date-parts":[[2026,6,22]]},"end":{"date-parts":[[2026,6,26]]}},"container-title":["ISC High Performance 2026 Research Paper Proceedings (41st International Conference)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11520448\/11520449\/11520474.pdf?arnumber=11520474","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,8]],"date-time":"2026-10-08T06:18:58Z","timestamp":1791440338000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11520474\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6]]},"references-count":42,"URL":"https:\/\/doi.org\/10.23919\/isc.2026.11520474","relation":{},"subject":[],"published":{"date-parts":[[2026,6]]}}}