{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T15:43:28Z","timestamp":1782920608651,"version":"3.54.5"},"reference-count":29,"publisher":"IEEE","license":[{"start":{"date-parts":[[2026,5,25]],"date-time":"2026-05-25T00:00:00Z","timestamp":1779667200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,5,25]],"date-time":"2026-05-25T00:00:00Z","timestamp":1779667200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,5,25]]},"DOI":"10.1109\/ipdps65963.2026.00107","type":"proceedings-article","created":{"date-parts":[[2026,6,30]],"date-time":"2026-06-30T20:55:16Z","timestamp":1782852916000},"page":"1301-1313","source":"Crossref","is-referenced-by-count":0,"title":["Design and Implementation of Casting Compression for GPU-Aware MPI Collectives"],"prefix":"10.1109","author":[{"given":"Chen-Chun","family":"Chen","sequence":"first","affiliation":[{"name":"The Ohio State University,Department of Computer Science and Engineering,Columbus,Ohio"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nicholas","family":"Contini","sequence":"additional","affiliation":[{"name":"The Ohio State University,Department of Computer Science and Engineering,Columbus,Ohio"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lang","family":"Xu","sequence":"additional","affiliation":[{"name":"The Ohio State University,Department of Computer Science and Engineering,Columbus,Ohio"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jake","family":"Queiser","sequence":"additional","affiliation":[{"name":"The Ohio State University,Department of Computer Science and Engineering,Columbus,Ohio"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hari","family":"Subramoni","sequence":"additional","affiliation":[{"name":"The Ohio State University,Department of Computer Science and Engineering,Columbus,Ohio"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Dhabaleswar K.","family":"Panda","sequence":"additional","affiliation":[{"name":"The Ohio State University,Department of Computer Science and Engineering,Columbus,Ohio"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","article-title":"TOP 500 Supercomputer Sites","author":"Strohmaier","year":"1993"},{"key":"ref2","article-title":"Olcf frontier supercomputer 2023-04-29 hpl power data used for top500\/green500 submission","author":"Atchley","year":"2023"},{"key":"ref3","article-title":"Lawrence livermore national laboratory\u2019s el capitan verified as world\u2019s fastest supercomputer","author":"Thomas","year":"2024"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-50371-0_19"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1021\/acs.jcim.3c01153"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1109\/TVCG.2014.2346458"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1007\/s00450-011-0171-3"},{"key":"ref8","article-title":"PTX ISA 8.7 documentation \u2014 docs.nvidia.com"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1109\/IPDPS64566.2025.00088"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-642-33518-1_16"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1109\/ICC45855.2022.9838754"},{"key":"ref12","article-title":"Horovod: fast and easy distributed deep learning in tensorflow","author":"Sergeev","year":"2018"},{"key":"ref13","article-title":"Pytorch: An imperative style, high-performance deep learning library","author":"Paszke","year":"2019"},{"key":"ref14","article-title":"Very deep convolutional networks for large-scale image recognition","author":"Simonyan","year":"2015"},{"key":"ref15","article-title":"Wide residual networks","author":"Zagoruyko","year":"2017"},{"key":"ref16","doi-asserted-by":"crossref","DOI":"10.1109\/CVPR.2017.634","article-title":"Aggregated residual transformations for deep neural networks","author":"Xie","year":"2017"},{"key":"ref17","article-title":"Efficientnet: Rethinking model scaling for convolutional neural networks","author":"Tan","year":"2020"},{"key":"ref18","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/2022.bigscience-1.9","article-title":"Gpt-neox-20b: An open-source autoregressive language model","author":"Black","year":"2022"},{"key":"ref19","article-title":"Megatron-lm: Training multi-billion parameter language models using model parallelism","author":"Shoeybi","year":"2020"},{"key":"ref20","doi-asserted-by":"crossref","DOI":"10.1109\/SC41405.2020.00024","article-title":"Zero: Memory optimizations toward training trillion parameter models","author":"Rajbhandari","year":"2020"},{"key":"ref21","article-title":"Language models are few-shot learners","author":"Brown","year":"2020"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1109\/CLUSTER.2015.59"},{"key":"ref23","article-title":"Quantization and training of neural networks for efficient integer-arithmetic-only inference","author":"Jacob","year":"2017"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1109\/IPDPS49936.2021.00053"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.23919\/ISC.2024.10528931"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-07312-0_1"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/HiPC56025.2022.00016"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/IPDPS54959.2023.00023"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/ccgrid59990.2024.00031"}],"event":{"name":"2026 IEEE International Parallel and Distributed Processing Symposium (IPDPS)","location":"New Orleans, LA, USA","start":{"date-parts":[[2026,5,25]]},"end":{"date-parts":[[2026,5,29]]}},"container-title":["2026 IEEE International Parallel and Distributed Processing Symposium (IPDPS)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11575315\/11575316\/11575350.pdf?arnumber=11575350","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T14:44:23Z","timestamp":1782917063000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11575350\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,25]]},"references-count":29,"URL":"https:\/\/doi.org\/10.1109\/ipdps65963.2026.00107","relation":{},"subject":[],"published":{"date-parts":[[2026,5,25]]}}}