{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,30]],"date-time":"2026-06-30T05:45:03Z","timestamp":1782798303418,"version":"3.54.5"},"reference-count":37,"publisher":"IEEE","license":[{"start":{"date-parts":[[2026,5,18]],"date-time":"2026-05-18T00:00:00Z","timestamp":1779062400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,5,18]],"date-time":"2026-05-18T00:00:00Z","timestamp":1779062400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100002701","name":"Ministry of Education","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100002701","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100002858","name":"China Postdoctoral Science Foundation","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100002858","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100012226","name":"Fundamental Research Funds for the Central Universities","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100012226","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100002766","name":"Beijing University of Posts and Telecommunications","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100002766","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,5,18]]},"DOI":"10.1109\/infocom59046.2026.11571693","type":"proceedings-article","created":{"date-parts":[[2026,6,29]],"date-time":"2026-06-29T19:38:15Z","timestamp":1782761895000},"page":"1-10","source":"Crossref","is-referenced-by-count":0,"title":["RecFlow: Unlocking GPU Efficiency for DLRM Inference via Fine-Grained Parallelism and Incremental Batching"],"prefix":"10.1109","author":[{"given":"Siheng","family":"Pan","sequence":"first","affiliation":[{"name":"Beijing University of Posts and Telecommunications,State Key Laboratory of Networking and Switching Technology,Beijing,China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shaolong","family":"Li","sequence":"additional","affiliation":[{"name":"Beijing University of Posts and Telecommunications,State Key Laboratory of Networking and Switching Technology,Beijing,China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Minwei","family":"Zhang","sequence":"additional","affiliation":[{"name":"Beijing University of Posts and Telecommunications,State Key Laboratory of Networking and Switching Technology,Beijing,China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shuxi","family":"Guo","sequence":"additional","affiliation":[{"name":"Beijing University of Posts and Telecommunications,State Key Laboratory of Networking and Switching Technology,Beijing,China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Haifeng","family":"Sun","sequence":"additional","affiliation":[{"name":"Beijing University of Posts and Telecommunications,State Key Laboratory of Networking and Switching Technology,Beijing,China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Qi","family":"Qi","sequence":"additional","affiliation":[{"name":"Beijing University of Posts and Telecommunications,State Key Laboratory of Networking and Switching Technology,Beijing,China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zirui","family":"Zhuang","sequence":"additional","affiliation":[{"name":"Beijing University of Posts and Telecommunications,State Key Laboratory of Networking and Switching Technology,Beijing,China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xiang","family":"Yang","sequence":"additional","affiliation":[{"name":"Meituan,Beijing,China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Chunyang","family":"Jiang","sequence":"additional","affiliation":[{"name":"Meituan,Beijing,China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jianxin","family":"Liao","sequence":"additional","affiliation":[{"name":"Beijing University of Posts and Telecommunications,State Key Laboratory of Networking and Switching Technology,Beijing,China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jingyu","family":"Wang","sequence":"additional","affiliation":[{"name":"Beijing University of Posts and Telecommunications,State Key Laboratory of Networking and Switching Technology,Beijing,China"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","author":"Naumov","year":"2019","journal-title":"Deep learning recommendation model for personalization and recommendation systems"},{"key":"ref2","doi-asserted-by":"crossref","DOI":"10.1145\/3604915.3608778","article-title":"Intune: Reinforcement learning-based data pipeline optimization for deep recommendation models","author":"Nagrecha","year":"2023"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1145\/3470496.3533727"},{"key":"ref4","first-page":"731","article-title":"{QuickUpdate}: a {Real-Time} personalization system for {Large-Scale} recommendation models","volume-title":"21st USENIX Symposium on Networked Systems Design and Implementation (NSDI 24)","author":"Matam"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1145\/3583780.3615208"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.14778\/3685800.3685832"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA45697.2020.00084"},{"key":"ref8","article-title":"Best practices for building and deploying recommender systems","volume-title":"NVIDIA, Tech. Rep.","year":"2022"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1145\/3627673.3680034"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1145\/3523227.3546765"},{"key":"ref11","author":"Agrawal","year":"2023","journal-title":"Sarathi: Efficient llm inference by piggybacking decodes with chunked prefills"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1109\/SC41406.2024.00047"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1145\/3623278.3624761"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1145\/3523227.3547387"},{"key":"ref15","first-page":"578","article-title":"Tvm: An automated end-to-end optimizing compiler for deep learning","volume-title":"13th USENIX Symposium on Operating Systems Design and Implementation (OSDI 18)","author":"Chen"},{"key":"ref16","article-title":"Cutlass: Cuda templates for linear algebra subroutines","year":"2025"},{"key":"ref17","article-title":"NVIDIA Ampere GPU Architecture: The World\u2019s Most Advanced Data Center GPU","year":"2020"},{"key":"ref18","article-title":"Towards latency-aware dnn optimization with gpu runtime analysis and tail effect elimination","author":"Yu","year":"2020"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1145\/3779212.3790135"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/TC.2024.3475589"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1145\/3689031.3696070"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1109\/ICCD53106.2021.00054"},{"key":"ref23","article-title":"Accelerating bandwidth-bound deep learning inference with main-memory accelerators","author":"Cho","year":"2020"},{"key":"ref24","article-title":"CUDA C++ Programming Guide","volume-title":"chapter 5: Streams and Concurrency","year":"2023"},{"key":"ref25","volume-title":"Multi-Process Service (MPS)","year":"2023"},{"key":"ref26","first-page":"749","article-title":"Nanoflow: Towards optimal large language model serving throughput","volume-title":"19th USENIX Symposium on Operating Systems Design and Implementation (OSDI 25)","author":"Zhu"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/ICCAD51958.2021.9643501"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1145\/3676536.3676718"},{"key":"ref29","article-title":"NVIDIA NSight Systems","year":"2023"},{"key":"ref30","article-title":"Regdem: Increasing gpu performance via shared memory register spilling","author":"Sakdhnagool","year":"2019"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1109\/ICPADS.2013.47"},{"key":"ref32","article-title":"Pytorch: An imperative style, high-performance deep learning library","volume":"32","author":"Paszke","year":"2019","journal-title":"Advances in neural information processing systems (NeurIPS \u201919)"},{"key":"ref33","article-title":"Fbgemm: Enabling high-performance low-precision deep learning inference","author":"Khudia","year":"2021"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1145\/3552326.3587445"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1145\/3638757"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1145\/3523227.3547405"},{"key":"ref37","doi-asserted-by":"crossref","DOI":"10.1145\/3466752.3480127","article-title":"Recpipe: Co-designing models and hardware to jointly optimize recommendation quality and performance","author":"Gupta","year":"2021"}],"event":{"name":"IEEE INFOCOM 2026 - IEEE Conference on Computer Communications","location":"Tokyo, Japan","start":{"date-parts":[[2026,5,18]]},"end":{"date-parts":[[2026,5,21]]}},"container-title":["IEEE INFOCOM 2026 - IEEE Conference on Computer Communications"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11571071\/11571169\/11571693.pdf?arnumber=11571693","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,30]],"date-time":"2026-06-30T05:19:11Z","timestamp":1782796751000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11571693\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,18]]},"references-count":37,"URL":"https:\/\/doi.org\/10.1109\/infocom59046.2026.11571693","relation":{},"subject":[],"published":{"date-parts":[[2026,5,18]]}}}