{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,1,21]],"date-time":"2026-01-21T22:11:40Z","timestamp":1769033500414,"version":"3.49.0"},"reference-count":16,"publisher":"IEEE","license":[{"start":{"date-parts":[[2025,11,14]],"date-time":"2025-11-14T00:00:00Z","timestamp":1763078400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,11,14]],"date-time":"2025-11-14T00:00:00Z","timestamp":1763078400000},"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":[[2025,11,14]]},"DOI":"10.1109\/cloudcom67567.2025.11331396","type":"proceedings-article","created":{"date-parts":[[2026,1,20]],"date-time":"2026-01-20T20:37:16Z","timestamp":1768941436000},"page":"1-7","source":"Crossref","is-referenced-by-count":0,"title":["Vectorization of Narrow Matrix Multiplication for Ascend AI Inference Acceleration"],"prefix":"10.1109","author":[{"given":"Shurygin","family":"Anton","sequence":"first","affiliation":[{"name":"Moscow Institute of Physics and Technology,Moscow,Russia"}]},{"given":"Frolov","family":"Aleksandr","sequence":"additional","affiliation":[{"name":"Moscow Institute of Physics and Technology,Moscow,Russia"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1145\/3412382.3458272"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1145\/3140659.3080246"},{"key":"ref3","article-title":"CUTLASS: Fast linear algebra in CUDA C++","volume-title":"Nvidia Developer Blog","author":"Kerr","year":"2017"},{"key":"ref4","article-title":"Towards efficient multi-scale deformable attention on NPU","volume-title":"arXiv preprint","author":"Huang","year":"2025"},{"key":"ref5","first-page":"33","article-title":"Accelerating sparse matrix-matrix multiplication with the Ascend AI core","volume-title":"Proc. Workshop Accelerated Mach. Learn","author":"Moustafa","year":"2023"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA51647.2021.00071"},{"key":"ref7","volume-title":"samples","year":"2025"},{"key":"ref8","volume-title":"cann-community","year":"2025"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1016\/c2020-0-00270-7"},{"key":"ref10","volume-title":"cann-ops-adv","year":"2025"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1706.03762"},{"key":"ref12","first-page":"16344","article-title":"Flashattention: Fast and memory-efficient exact attention with io-awareness","volume":"35","author":"Dao","year":"2022","journal-title":"Advances in neural information processing systems"},{"key":"ref13","article-title":"Large language model inference acceleration: A comprehensive hardware perspective","author":"Li","year":"2024","journal-title":"arXiv preprint"},{"key":"ref14","article-title":"Deepseek-v3 technical report","author":"Liu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref15","article-title":"Understanding the capabilities, limitations, and societal impact of large language models","author":"Tamkin","year":"2021","journal-title":"arXiv preprint"},{"key":"ref16","article-title":"RoFormer: Enhanced transformer with rotary position embedding","volume-title":"arXiv preprint","author":"Su","year":"2021"}],"event":{"name":"2025 lEEE International Conference on Cloud Computing Technology and Science (CloudCom)","location":"Shenzhen, China","start":{"date-parts":[[2025,11,14]]},"end":{"date-parts":[[2025,11,16]]}},"container-title":["2025 lEEE International Conference on Cloud Computing Technology and Science (CloudCom)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11330195\/11331311\/11331396.pdf?arnumber=11331396","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,1,21]],"date-time":"2026-01-21T07:18:39Z","timestamp":1768979919000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11331396\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,11,14]]},"references-count":16,"URL":"https:\/\/doi.org\/10.1109\/cloudcom67567.2025.11331396","relation":{},"subject":[],"published":{"date-parts":[[2025,11,14]]}}}