{"id":"https://openalex.org/W7125932074","doi":"https://doi.org/10.1145/3774934.3786434","title":"Accelerating Sparse Transformer Inference on GPU","display_name":"Accelerating Sparse Transformer Inference on GPU","publication_year":2026,"publication_date":"2026-01-28","ids":{"openalex":"https://openalex.org/W7125932074","doi":"https://doi.org/10.1145/3774934.3786434"},"language":null,"primary_location":{"id":"doi:10.1145/3774934.3786434","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3774934.3786434","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 31st ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5124137329","display_name":"Wenhao Dai","orcid":null},"institutions":[{"id":"https://openalex.org/I204553293","display_name":"China University of Petroleum, Beijing","ror":"https://ror.org/041qf4r12","country_code":"CN","type":"education","lineage":["https://openalex.org/I204553293"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wenhao Dai","raw_affiliation_strings":["China University of Petroleum-Beijing, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0000-5551-7603","affiliations":[{"raw_affiliation_string":"China University of Petroleum-Beijing, Beijing, China","institution_ids":["https://openalex.org/I204553293"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124075277","display_name":"Haodong Deng","orcid":null},"institutions":[{"id":"https://openalex.org/I204553293","display_name":"China University of Petroleum, Beijing","ror":"https://ror.org/041qf4r12","country_code":"CN","type":"education","lineage":["https://openalex.org/I204553293"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Haodong Deng","raw_affiliation_strings":["China University of Petroleum-Beijing, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0008-5627-6918","affiliations":[{"raw_affiliation_string":"China University of Petroleum-Beijing, Beijing, China","institution_ids":["https://openalex.org/I204553293"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5001245929","display_name":"Mengfei Rong","orcid":null},"institutions":[{"id":"https://openalex.org/I204553293","display_name":"China University of Petroleum, Beijing","ror":"https://ror.org/041qf4r12","country_code":"CN","type":"education","lineage":["https://openalex.org/I204553293"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Mengfei Rong","raw_affiliation_strings":["China University of Petroleum-Beijing, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0009-2202-2327","affiliations":[{"raw_affiliation_string":"China University of Petroleum-Beijing, Beijing, China","institution_ids":["https://openalex.org/I204553293"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Xinyu Yang","orcid":"https://orcid.org/0009-0006-9931-2486"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xinyu Yang","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0006-9931-2486","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124059107","display_name":"Hongyu Liu","orcid":null},"institutions":[{"id":"https://openalex.org/I98301712","display_name":"Baidu (China)","ror":"https://ror.org/03vs3wt56","country_code":"CN","type":"company","lineage":["https://openalex.org/I98301712"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hongyu Liu","raw_affiliation_strings":["Baidu, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0005-6128-3619","affiliations":[{"raw_affiliation_string":"Baidu, Beijing, China","institution_ids":["https://openalex.org/I98301712"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5017670541","display_name":"Fangxin Liu","orcid":"https://orcid.org/0000-0002-8769-293X"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Fangxin Liu","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China"],"raw_orcid":"https://orcid.org/0000-0002-8769-293X","affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Hailong Yang","orcid":"https://orcid.org/0000-0003-1101-7927"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hailong Yang","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0003-1101-7927","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5022081570","display_name":"Qianwen Cao","orcid":"https://orcid.org/0000-0001-7491-5201"},"institutions":[{"id":"https://openalex.org/I204553293","display_name":"China University of Petroleum, Beijing","ror":"https://ror.org/041qf4r12","country_code":"CN","type":"education","lineage":["https://openalex.org/I204553293"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Qianwen Cao","raw_affiliation_strings":["China University of Petroleum-Beijing, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-7491-5201","affiliations":[{"raw_affiliation_string":"China University of Petroleum-Beijing, Beijing, China","institution_ids":["https://openalex.org/I204553293"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5060999547","display_name":"Qingxiao Sun","orcid":"https://orcid.org/0000-0003-2927-362X"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Qingxiao Sun","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0003-2927-362X","affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"620","last_page":"634"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.35760000348091125,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.35760000348091125,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.2045000046491623,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11273","display_name":"Advanced Graph Neural Networks","score":0.039799999445676804,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.704800009727478},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.6105999946594238},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5270000100135803},{"id":"https://openalex.org/keywords/fusion","display_name":"Fusion","score":0.487199991941452},{"id":"https://openalex.org/keywords/multi-core-processor","display_name":"Multi-core processor","score":0.44679999351501465},{"id":"https://openalex.org/keywords/sparse-matrix","display_name":"Sparse matrix","score":0.42010000348091125},{"id":"https://openalex.org/keywords/template","display_name":"Template","score":0.3774999976158142}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7551000118255615},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.704800009727478},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.6105999946594238},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5270000100135803},{"id":"https://openalex.org/C158525013","wikidata":"https://www.wikidata.org/wiki/Q2593739","display_name":"Fusion","level":2,"score":0.487199991941452},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.4546999931335449},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.45239999890327454},{"id":"https://openalex.org/C78766204","wikidata":"https://www.wikidata.org/wiki/Q555032","display_name":"Multi-core processor","level":2,"score":0.44679999351501465},{"id":"https://openalex.org/C56372850","wikidata":"https://www.wikidata.org/wiki/Q1050404","display_name":"Sparse matrix","level":3,"score":0.42010000348091125},{"id":"https://openalex.org/C82714645","wikidata":"https://www.wikidata.org/wiki/Q438331","display_name":"Template","level":2,"score":0.3774999976158142},{"id":"https://openalex.org/C177774035","wikidata":"https://www.wikidata.org/wiki/Q1246948","display_name":"Granularity","level":2,"score":0.3334999978542328},{"id":"https://openalex.org/C2777402240","wikidata":"https://www.wikidata.org/wiki/Q6783436","display_name":"Masking (illustration)","level":2,"score":0.3278999924659729},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.31869998574256897},{"id":"https://openalex.org/C17020691","wikidata":"https://www.wikidata.org/wiki/Q139677","display_name":"Operator (biology)","level":5,"score":0.3140999972820282},{"id":"https://openalex.org/C46743427","wikidata":"https://www.wikidata.org/wiki/Q1341685","display_name":"Inference engine","level":3,"score":0.3107999861240387},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.30959999561309814},{"id":"https://openalex.org/C134765980","wikidata":"https://www.wikidata.org/wiki/Q879126","display_name":"Bitwise operation","level":2,"score":0.3028999865055084},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.27079999446868896},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.2639999985694885},{"id":"https://openalex.org/C459310","wikidata":"https://www.wikidata.org/wiki/Q117801","display_name":"Computational science","level":1,"score":0.26179999113082886},{"id":"https://openalex.org/C2778119891","wikidata":"https://www.wikidata.org/wiki/Q477690","display_name":"CUDA","level":2,"score":0.25859999656677246}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3774934.3786434","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3774934.3786434","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 31st ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":11,"referenced_works":["https://openalex.org/W2893007245","https://openalex.org/W3131922516","https://openalex.org/W4251637954","https://openalex.org/W4280496502","https://openalex.org/W4387757648","https://openalex.org/W4391136507","https://openalex.org/W4392450088","https://openalex.org/W4394841768","https://openalex.org/W4405756380","https://openalex.org/W4408891420","https://openalex.org/W7106707867"],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"(LLMs)":[3],"are":[4],"popular":[5],"around":[6],"the":[7,16,47,69,99,119,127,141],"world":[8],"due":[9],"to":[10,38,61,63,101,111,122,140],"their":[11],"powerful":[12],"understanding":[13],"capabilities.":[14],"As":[15],"core":[17],"component":[18],"of":[19,50,148],"LLMs,":[20],"accelerating":[21],"Transformer":[22,37,82],"through":[23,131],"parallelization":[24],"has":[25],"gradually":[26],"become":[27],"a":[28,75],"hot":[29],"research":[30],"topic.":[31],"Mask":[32],"layers":[33],"introduce":[34],"sparsity":[35],"into":[36],"reduce":[39],"calculations.":[40],"However,":[41],"previous":[42],"works":[43],"rarely":[44],"focus":[45],"on":[46,90],"performance":[48],"optimization":[49],"sparse":[51],"Transformer.":[52],"In":[53],"addition,":[54],"current":[55],"static":[56],"operator":[57],"fusion":[58,120],"schemes":[59],"fail":[60],"adapt":[62],"diverse":[64],"application":[65],"scenarios.":[66],"To":[67],"address":[68],"above":[70],"problems,":[71],"we":[72],"propose":[73],"STOF,":[74],"framework":[76],"that":[77,83,138],"incorporates":[78],"optimizations":[79],"for":[80],"Sparse":[81],"enables":[84],"flexible":[85],"masking":[86],"and":[87,125,153],"Operator":[88],"Fusion":[89],"GPU.":[91],"For":[92,114],"multi-head":[93],"attention":[94],"(MHA)":[95],"structure,":[96],"STOF":[97,117,144],"maps":[98,118],"computation":[100,152],"row-wise":[102],"or":[103],"block-wise":[104],"kernels":[105],"with":[106],"unique":[107],"storage":[108],"formats":[109],"according":[110],"analytical":[112],"modeling.":[113],"downstream":[115],"operators,":[116],"scheme":[121],"compilation":[123],"templates":[124],"determines":[126],"optimal":[128],"running":[129],"configuration":[130],"two-stage":[132],"searching.":[133],"The":[134],"experimental":[135],"results":[136],"show":[137],"compared":[139],"state-of-the-art":[142],"work,":[143],"achieves":[145],"maximum":[146],"speedups":[147],"1.6\u00d7":[149],"in":[150,155],"MHA":[151],"1.4\u00d7":[154],"end-to-end":[156],"inference.":[157]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2026-01-29T00:00:00"}