{"id":"https://openalex.org/W4413411747","doi":"https://doi.org/10.1145/3721145.3730430","title":"Fused3S: Fast Sparse Attention on Tensor Cores","display_name":"Fused3S: Fast Sparse Attention on Tensor Cores","publication_year":2025,"publication_date":"2025-06-08","ids":{"openalex":"https://openalex.org/W4413411747","doi":"https://doi.org/10.1145/3721145.3730430"},"language":"en","primary_location":{"id":"doi:10.1145/3721145.3730430","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3721145.3730430","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3721145.3730430","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 39th ACM International Conference on Supercomputing","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://dl.acm.org/doi/pdf/10.1145/3721145.3730430","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5076856296","display_name":"Zitong Li","orcid":"https://orcid.org/0000-0002-3029-2214"},"institutions":[{"id":"https://openalex.org/I204250578","display_name":"University of California, Irvine","ror":"https://ror.org/04gyf1771","country_code":"US","type":"education","lineage":["https://openalex.org/I204250578"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zitong Li","raw_affiliation_strings":["University of California, Irvine, Irvine, USA"],"raw_orcid":"https://orcid.org/0000-0002-3029-2214","affiliations":[{"raw_affiliation_string":"University of California, Irvine, Irvine, USA","institution_ids":["https://openalex.org/I204250578"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5074121652","display_name":"Aparna Chandramowlishwaran","orcid":"https://orcid.org/0000-0003-0840-4192"},"institutions":[{"id":"https://openalex.org/I204250578","display_name":"University of California, Irvine","ror":"https://ror.org/04gyf1771","country_code":"US","type":"education","lineage":["https://openalex.org/I204250578"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Aparna Chandramowlishwaran","raw_affiliation_strings":["University of California, Irvine, Irvine, USA"],"raw_orcid":"https://orcid.org/0000-0003-0840-4192","affiliations":[{"raw_affiliation_string":"University of California, Irvine, Irvine, USA","institution_ids":["https://openalex.org/I204250578"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.19989832,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"104","last_page":"118"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9933000206947327,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9933000206947327,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10036","display_name":"Advanced Neural Network Applications","score":0.9916999936103821,"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/T13553","display_name":"Age of Information Optimization","score":0.986299991607666,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/computer-science","display_name":"Computer science","score":0.6443973183631897},{"id":"https://openalex.org/keywords/tensor","display_name":"Tensor (intrinsic definition)","score":0.49956250190734863},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.11466154456138611}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6443973183631897},{"id":"https://openalex.org/C155281189","wikidata":"https://www.wikidata.org/wiki/Q3518150","display_name":"Tensor (intrinsic definition)","level":2,"score":0.49956250190734863},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.11466154456138611},{"id":"https://openalex.org/C202444582","wikidata":"https://www.wikidata.org/wiki/Q837863","display_name":"Pure mathematics","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3721145.3730430","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3721145.3730430","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3721145.3730430","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 39th ACM International Conference on Supercomputing","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3721145.3730430","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3721145.3730430","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3721145.3730430","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 39th ACM International Conference on Supercomputing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320332688","display_name":"University of California, Irvine","ror":"https://ror.org/04gyf1771"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4413411747.pdf","grobid_xml":"https://content.openalex.org/works/W4413411747.grobid-xml"},"referenced_works_count":13,"referenced_works":["https://openalex.org/W2130289795","https://openalex.org/W2294347342","https://openalex.org/W2734941459","https://openalex.org/W2765982206","https://openalex.org/W3173204484","https://openalex.org/W4296473245","https://openalex.org/W4381832157","https://openalex.org/W4391986945","https://openalex.org/W4395106472","https://openalex.org/W4401408710","https://openalex.org/W4401862084","https://openalex.org/W4403300952","https://openalex.org/W4405755887"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"Sparse":[0],"attention":[1],"is":[2],"a":[3,24],"core":[4,102],"building":[5],"block":[6],"in":[7],"many":[8],"leading":[9],"neural":[10],"network":[11],"models,":[12],"from":[13],"graph-structured":[14],"learning":[15],"to":[16,55],"sparse":[17,28,40,81],"sequence":[18,25],"modeling.It":[19],"can":[20],"be":[21],"decomposed":[22],"into":[23,128],"of":[26,74,87],"three":[27],"matrix":[29,34,41],"operations":[30,82],"(3S):":[31],"sampled":[32],"dense-dense":[33],"multiplication":[35,42],"(SDDMM),":[36],"softmax":[37],"normalization,":[38],"and":[39,62,69,104,115,123,147,150],"(SpMM).Efficiently":[43],"executing":[44],"the":[45,57,71,93],"3S":[46,96,141],"computational":[47],"pattern":[48],"on":[49,121],"modern":[50],"GPUs":[51],"remains":[52],"challenging":[53],"due":[54],"(a)":[56],"mismatch":[58],"between":[59],"unstructured":[60],"sparsity":[61],"tensor":[63,101],"cores":[64],"optimized":[65,79],"for":[66],"dense":[67],"operations,":[68],"(b)":[70],"high":[72],"cost":[73],"data":[75,106],"movement.Previous":[76],"works":[77],"have":[78],"these":[80,88],"individually":[83],"or":[84],"addressed":[85],"one":[86],"challenges.This":[89],"paper":[90],"introduces":[91],"Fused3S,":[92],"first":[94],"fused":[95],"algorithm":[97],"that":[98],"jointly":[99],"maximizes":[100],"utilization":[103],"minimizes":[105],"movement.Across":[107],"real-world":[108],"graph":[109],"datasets,":[110],"Fused3S":[111,127],"achieves":[112],"1.6":[113],"-16.3":[114],"1.5":[116],"-14":[117],"speedup":[118],"over":[119],"state-of-the-art":[120],"H100":[122],"A30":[124],"GPUs.Furthermore,":[125],"integrating":[126],"Graph":[129],"Transformer":[130],"inference":[131],"accelerates":[132],"end-to-end":[133],"performance":[134],"by":[135],"1.05":[136],"-5.36,":[137],"consistently":[138],"outperforming":[139],"all":[140],"baselines":[142],"across":[143],"diverse":[144],"datasets":[145],"(single":[146],"batched":[148],"graphs)":[149],"GPU":[151],"architectures.":[152]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}
