{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,9,21]],"date-time":"2025-09-21T07:15:15Z","timestamp":1758438915494,"version":"3.44.0"},"reference-count":34,"publisher":"Association for Computing Machinery (ACM)","issue":"3","funder":[{"DOI":"10.13039\/501100001809","name":"NSFC","doi-asserted-by":"crossref","award":["62472435, 62025208, 62172430"],"award-info":[{"award-number":["62472435, 62025208, 62172430"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"crossref"}]},{"name":"STIP of Hunan Province","award":["2022RC3065"],"award-info":[{"award-number":["2022RC3065"]}]},{"name":"Foundation of PDL","award":["2023-JKWPDL-02"],"award-info":[{"award-number":["2023-JKWPDL-02"]}]},{"name":"Key Laboratory of Advanced Microprocessor Chips and Systems"}],"content-domain":{"domain":["dl.acm.org"],"crossmark-restriction":true},"short-container-title":["ACM Trans. Archit. Code Optim."],"published-print":{"date-parts":[[2025,9,30]]},"abstract":"<jats:p>Deep learning becomes increasingly popular, and its main workload is Sparse-Sparse Matrix Multiplication (SpMSpM). Most SpMSpM accelerators usually only support a single dataflow. Different dataflows have different performance in different computing environments. Therefore, the single-dataflow accelerator cannot maintain the highest performance in all environments. Compared with single-dataflow accelerators, multi-dataflow accelerators provide flexible options for different workloads and improve the overall performance. Flexagon, Sparm, and SPADA are state-of-the-art multi-dataflow accelerators. However, the computation process of Flexagon and Sparm is not fully pipelined, and SPADA cannot support inner product dataflow. Additionally, Flexagon, Sparm, and SPADA cannot switch dataflows quickly and accurately. Inspired by these observations, we present SpMARD, a SpMSpM accelerator with reconfigurable dataflow. The computation process of SpMARD is fully pipelined, and SpMARD can support six dataflow variants simultaneously. Through the design of a Two-stage Pipeline Adder Network (TPAN) and a Position-based Psum Array (PPA), SpMARD can execute element-level merging, which can hide the merging overhead. Through the quantitative analysis of dataflows, we implement a Dataflow Switcher (DSwitcher), which can switch dataflows more efficiently. For the SpMSpM workload, the performance (GOPS) of the SpMARD we proposed is 1.27 times that of Flexagon, 1.18 times that of Sparm, and 1.22 times that of SPADA.<\/jats:p>","DOI":"10.1145\/3747847","type":"journal-article","created":{"date-parts":[[2025,8,4]],"date-time":"2025-08-04T11:08:34Z","timestamp":1754305714000},"page":"1-23","update-policy":"https:\/\/doi.org\/10.1145\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["SpMARD: A Sparse-Sparse Matrix Multiplication Accelerator with Reconfigurable Dataflow for DNN Workloads"],"prefix":"10.1145","volume":"22","author":[{"ORCID":"https:\/\/orcid.org\/0009-0004-9441-0509","authenticated-orcid":false,"given":"Bo","family":"Wang","sequence":"first","affiliation":[{"name":"College of Computer Science and Technology, National University of Defense Technology","place":["Changsha, China"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1710-4060","authenticated-orcid":false,"given":"Sheng","family":"Ma","sequence":"additional","affiliation":[{"name":"College of Computer Science and Technology, National University of Defense Technology","place":["Changsha, China"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5600-3740","authenticated-orcid":false,"given":"Yunping","family":"Zhao","sequence":"additional","affiliation":[{"name":"College of Computer Science and Technology, National University of Defense Technology","place":["changsha, China"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0009-0007-5551-2897","authenticated-orcid":false,"given":"Shengbai","family":"Luo","sequence":"additional","affiliation":[{"name":"College of Computer Science and Technology, National University of Defense Technology","place":["Changsha, China"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4439-7436","authenticated-orcid":false,"given":"Lizhou","family":"Wu","sequence":"additional","affiliation":[{"name":"College of Computer Science and Technology, National University of Defense Technology","place":["Changsha, China"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1008-4805","authenticated-orcid":false,"given":"Jianmin","family":"Zhang","sequence":"additional","affiliation":[{"name":"College of Computer Science and Technology, National University of Defense Technology","place":["Changsha, China"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9743-2034","authenticated-orcid":false,"given":"Dongsheng","family":"Li","sequence":"additional","affiliation":[{"name":"College of Computer Science and Technology, National University of Defense Technology","place":["Changsha, China"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1509-1761","authenticated-orcid":false,"given":"Tiejun","family":"Li","sequence":"additional","affiliation":[{"name":"College of Computer Science and Technology, National University of Defense Technology","place":["Changsha, China"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0431-8852","authenticated-orcid":false,"given":"Zhuojun","family":"Chen","sequence":"additional","affiliation":[{"name":"School of Physics & Electronics, Hunan University","place":["Changsha, China"]}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"320","published-online":{"date-parts":[[2025,9,19]]},"reference":[{"key":"e_1_3_1_2_2","article-title":"GitHub - meta-llama\/llama: Inference code for Llama models","author":"AI Meta","year":"2023","unstructured":"Meta AI. 2023. GitHub - meta-llama\/llama: Inference code for Llama models. Retrieved January 14, 2025 from https:\/\/github.com\/meta-llama\/llama. (2023).","journal-title":"https:\/\/github.com\/meta-llama\/llama"},{"key":"e_1_3_1_3_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.01147"},{"key":"e_1_3_1_4_2","unstructured":"Liang-Chieh Chen George Papandreou Florian Schroff and Hartwig Adam. 2017. Rethinking atrous convolution for semantic image segmentation. arXiv:1706.05587. Retrieved from https:\/\/arxiv.org\/abs\/1706.05587"},{"key":"e_1_3_1_5_2","doi-asserted-by":"publisher","DOI":"10.1109\/JPROC.2021.3098483"},{"key":"e_1_3_1_6_2","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA57654.2024.00081"},{"key":"e_1_3_1_7_2","doi-asserted-by":"publisher","DOI":"10.1007\/s11704-024-40149-8"},{"key":"e_1_3_1_8_2","doi-asserted-by":"publisher","DOI":"10.1109\/PACT.2019.00009"},{"key":"e_1_3_1_9_2","article-title":"Learning both weights and connections for efficient neural network","author":"Han Song","year":"2015","unstructured":"Song Han, Jeff Pool, John Tran, and William Dally. 2015. Learning both weights and connections for efficient neural network. In Proceedings of the 29th International Conference on Neural Information Processing Systems.","journal-title":"Proceedings of the 29th International Conference on Neural Information Processing Systems"},{"key":"e_1_3_1_10_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"e_1_3_1_11_2","doi-asserted-by":"publisher","DOI":"10.1145\/3352460.3358275"},{"key":"e_1_3_1_12_2","doi-asserted-by":"publisher","DOI":"10.1145\/3065386"},{"key":"e_1_3_1_13_2","doi-asserted-by":"publisher","DOI":"10.1145\/3296957.3173176"},{"key":"e_1_3_1_14_2","doi-asserted-by":"publisher","DOI":"10.1145\/3575693.3575706"},{"key":"e_1_3_1_15_2","doi-asserted-by":"publisher","DOI":"10.1109\/ASAP61560.2024.00034"},{"key":"e_1_3_1_16_2","unstructured":"Sachin Mehta and Mohammad Rastegari. 2021. Mobilevit: Light-weight general-purpose and mobile-friendly vision transformer. arXiv:2110.02178. Retrieved from https:\/\/arxiv.org\/abs\/2110.02178"},{"key":"e_1_3_1_17_2","doi-asserted-by":"publisher","DOI":"10.1109\/IISWC53511.2021.00028"},{"key":"e_1_3_1_18_2","doi-asserted-by":"publisher","DOI":"10.1145\/3582016.3582069"},{"key":"e_1_3_1_19_2","article-title":"SuiteSparse Matrix Collection","author":"Florida The University of","unstructured":"The University of Florida. 2019. SuiteSparse Matrix Collection. Retrieved January 14, 2025 from https:\/\/sparse.tamu.edu\/","journal-title":"https:\/\/sparse.tamu.edu\/"},{"key":"e_1_3_1_20_2","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA.2018.00067"},{"key":"e_1_3_1_21_2","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA47549.2020.00015"},{"key":"e_1_3_1_22_2","doi-asserted-by":"publisher","DOI":"10.1145\/3489517.3530506"},{"key":"e_1_3_1_23_2","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA53966.2022.00068"},{"key":"e_1_3_1_24_2","doi-asserted-by":"publisher","DOI":"10.5555\/2627435.2670313"},{"key":"e_1_3_1_25_2","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO50266.2020.00068"},{"key":"e_1_3_1_26_2","doi-asserted-by":"publisher","DOI":"10.1145\/3634703"},{"key":"e_1_3_1_27_2","doi-asserted-by":"publisher","DOI":"10.1145\/3649329.3657362"},{"key":"e_1_3_1_28_2","doi-asserted-by":"publisher","DOI":"10.1145\/3604802"},{"key":"e_1_3_1_29_2","doi-asserted-by":"publisher","DOI":"10.1145\/3460776"},{"key":"e_1_3_1_30_2","doi-asserted-by":"publisher","DOI":"10.1109\/DAC56929.2023.10247897"},{"key":"e_1_3_1_31_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICTAI.2019.00197"},{"key":"e_1_3_1_32_2","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA59077.2024.00072"},{"key":"e_1_3_1_33_2","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO61859.2024.00108"},{"key":"e_1_3_1_34_2","doi-asserted-by":"publisher","DOI":"10.1145\/3445814.3446702"},{"key":"e_1_3_1_35_2","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA47549.2020.00030"}],"container-title":["ACM Transactions on Architecture and Code Optimization"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/3747847","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,9,20]],"date-time":"2025-09-20T00:49:31Z","timestamp":1758329371000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3747847"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,9,19]]},"references-count":34,"journal-issue":{"issue":"3","published-print":{"date-parts":[[2025,9,30]]}},"alternative-id":["10.1145\/3747847"],"URL":"https:\/\/doi.org\/10.1145\/3747847","relation":{},"ISSN":["1544-3566","1544-3973"],"issn-type":[{"type":"print","value":"1544-3566"},{"type":"electronic","value":"1544-3973"}],"subject":[],"published":{"date-parts":[[2025,9,19]]},"assertion":[{"value":"2025-02-06","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2025-06-23","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2025-09-19","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}