{"id":"https://openalex.org/W4404400671","doi":"https://doi.org/10.1145/3694715.3695961","title":"Uncovering Nested Data Parallelism and Data Reuse in DNN Computation with FractalTensor","display_name":"Uncovering Nested Data Parallelism and Data Reuse in DNN Computation with FractalTensor","publication_year":2024,"publication_date":"2024-11-04","ids":{"openalex":"https://openalex.org/W4404400671","doi":"https://doi.org/10.1145/3694715.3695961"},"language":"en","primary_location":{"id":"doi:10.1145/3694715.3695961","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3694715.3695961","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3694715.3695961","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM SIGOPS 30th Symposium on Operating Systems Principles","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://dl.acm.org/doi/pdf/10.1145/3694715.3695961","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5011115760","display_name":"Siran Liu","orcid":"https://orcid.org/0009-0004-9727-5023"},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Siran Liu","raw_affiliation_strings":["Peking University, Beijing, CN"],"raw_orcid":"https://orcid.org/0009-0004-9727-5023","affiliations":[{"raw_affiliation_string":"Peking University, Beijing, CN","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114091299","display_name":"C. Qi","orcid":null},"institutions":[{"id":"https://openalex.org/I4210165038","display_name":"University of Chinese Academy of Sciences","ror":"https://ror.org/05qbk4x57","country_code":"CN","type":"education","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210165038"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chengxiang Qi","raw_affiliation_strings":["University of Chinese Academy of Sciences, Beijing, CN"],"raw_orcid":"https://orcid.org/0009-0008-0881-9500","affiliations":[{"raw_affiliation_string":"University of Chinese Academy of Sciences, Beijing, CN","institution_ids":["https://openalex.org/I4210165038"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101444118","display_name":"Ying Cao","orcid":"https://orcid.org/0000-0003-2454-9092"},"institutions":[{"id":"https://openalex.org/I4210113369","display_name":"Microsoft Research Asia (China)","ror":"https://ror.org/0300m5276","country_code":"CN","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210113369"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ying Cao","raw_affiliation_strings":["Microsoft Research Asia, Beijing, CN"],"raw_orcid":"https://orcid.org/0000-0003-2454-9092","affiliations":[{"raw_affiliation_string":"Microsoft Research Asia, Beijing, CN","institution_ids":["https://openalex.org/I4210113369"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5005718995","display_name":"Chao Yang","orcid":"https://orcid.org/0000-0001-7426-6248"},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chao Yang","raw_affiliation_strings":["Peking University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-7426-6248","affiliations":[{"raw_affiliation_string":"Peking University, Beijing, China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102115175","display_name":"Weifang Hu","orcid":"https://orcid.org/0000-0002-4855-995X"},"institutions":[{"id":"https://openalex.org/I47720641","display_name":"Huazhong University of Science and Technology","ror":"https://ror.org/00p991c53","country_code":"CN","type":"education","lineage":["https://openalex.org/I47720641"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Weifang Hu","raw_affiliation_strings":["Huazhong University of Science and Technology, Wuhan, China"],"raw_orcid":"https://orcid.org/0000-0002-4855-995X","affiliations":[{"raw_affiliation_string":"Huazhong University of Science and Technology, Wuhan, China","institution_ids":["https://openalex.org/I47720641"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5016486553","display_name":"Xuanhua Shi","orcid":"https://orcid.org/0000-0001-8451-8656"},"institutions":[{"id":"https://openalex.org/I47720641","display_name":"Huazhong University of Science and Technology","ror":"https://ror.org/00p991c53","country_code":"CN","type":"education","lineage":["https://openalex.org/I47720641"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xuanhua Shi","raw_affiliation_strings":["Huazhong University of Science and Technology, Wuhan, CN"],"raw_orcid":"https://orcid.org/0000-0001-8451-8656","affiliations":[{"raw_affiliation_string":"Huazhong University of Science and Technology, Wuhan, CN","institution_ids":["https://openalex.org/I47720641"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100346602","display_name":"Fan Yang","orcid":"https://orcid.org/0000-0002-0378-060X"},"institutions":[{"id":"https://openalex.org/I4210113369","display_name":"Microsoft Research Asia (China)","ror":"https://ror.org/0300m5276","country_code":"CN","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210113369"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Fan Yang","raw_affiliation_strings":["Microsoft Research Asia, Beijing, CN"],"raw_orcid":"https://orcid.org/0000-0002-0378-060X","affiliations":[{"raw_affiliation_string":"Microsoft Research Asia, Beijing, CN","institution_ids":["https://openalex.org/I4210113369"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100438310","display_name":"Mao Yang","orcid":"https://orcid.org/0009-0009-6455-3898"},"institutions":[{"id":"https://openalex.org/I4210113369","display_name":"Microsoft Research Asia (China)","ror":"https://ror.org/0300m5276","country_code":"CN","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210113369"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Mao Yang","raw_affiliation_strings":["Microsoft Research Asia, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0009-6455-3898","affiliations":[{"raw_affiliation_string":"Microsoft Research Asia, Beijing, China","institution_ids":["https://openalex.org/I4210113369"]}]}],"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":1,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"160","last_page":"177"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.994700014591217,"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":0.994700014591217,"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/T10320","display_name":"Neural Networks and Applications","score":0.993399977684021,"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"}},{"id":"https://openalex.org/T10502","display_name":"Advanced Memory and Neural Computing","score":0.9925000071525574,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.8424399495124817},{"id":"https://openalex.org/keywords/parallelism","display_name":"Parallelism (grammar)","score":0.7808944582939148},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.7447088956832886},{"id":"https://openalex.org/keywords/reuse","display_name":"Reuse","score":0.6260980367660522},{"id":"https://openalex.org/keywords/data-parallelism","display_name":"Data parallelism","score":0.5945891737937927},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5880528092384338},{"id":"https://openalex.org/keywords/computer-architecture","display_name":"Computer architecture","score":0.4241289496421814},{"id":"https://openalex.org/keywords/programming-language","display_name":"Programming language","score":0.41057419776916504}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8424399495124817},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.7808944582939148},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.7447088956832886},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.6260980367660522},{"id":"https://openalex.org/C61483411","wikidata":"https://www.wikidata.org/wiki/Q3124522","display_name":"Data parallelism","level":3,"score":0.5945891737937927},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5880528092384338},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.4241289496421814},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.41057419776916504},{"id":"https://openalex.org/C86803240","wikidata":"https://www.wikidata.org/wiki/Q420","display_name":"Biology","level":0,"score":0.0},{"id":"https://openalex.org/C18903297","wikidata":"https://www.wikidata.org/wiki/Q7150","display_name":"Ecology","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3694715.3695961","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3694715.3695961","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3694715.3695961","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM SIGOPS 30th Symposium on Operating Systems Principles","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3694715.3695961","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3694715.3695961","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3694715.3695961","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM SIGOPS 30th Symposium on Operating Systems Principles","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":false},"content_urls":{"pdf":"https://content.openalex.org/works/W4404400671.pdf"},"referenced_works_count":20,"referenced_works":["https://openalex.org/W1997744477","https://openalex.org/W2097699872","https://openalex.org/W2147193503","https://openalex.org/W2152506070","https://openalex.org/W2164890169","https://openalex.org/W2644985445","https://openalex.org/W2798725115","https://openalex.org/W2944745903","https://openalex.org/W2949967139","https://openalex.org/W2954698171","https://openalex.org/W2979365412","https://openalex.org/W3103594582","https://openalex.org/W3177452048","https://openalex.org/W4212986322","https://openalex.org/W4281710230","https://openalex.org/W4384648484","https://openalex.org/W4387302777","https://openalex.org/W4389326242","https://openalex.org/W4389524555","https://openalex.org/W4400611982"],"related_works":["https://openalex.org/W2950520577","https://openalex.org/W1501159154","https://openalex.org/W1554644772","https://openalex.org/W2003935582","https://openalex.org/W2494130044","https://openalex.org/W3170887803","https://openalex.org/W74409296","https://openalex.org/W3209384898","https://openalex.org/W4230999561","https://openalex.org/W1595672120"],"abstract_inverted_index":{"To":[0,116],"speed":[1],"up":[2,184],"computation,":[3],"deep":[4],"neural":[5],"networks":[6],"(DNNs)":[7],"usually":[8],"rely":[9],"on":[10,89,167,175,189],"highly":[11],"optimized":[12],"tensor":[13,18,63],"operators.":[14],"Despite":[15],"the":[16,30,47,121,124],"effectiveness,":[17],"operators":[19,80,86],"are":[20,73],"often":[21],"defined":[22],"empirically":[23],"with":[24,64],"ad":[25],"hoc":[26],"semantics.":[27],"This":[28,91],"hinders":[29],"analysis":[31,113],"and":[32,83,102,114,163,173,187],"optimization":[33],"across":[34,146],"operator":[35],"boundaries.":[36],"FractalTensor":[37,49,69],"is":[38,50,61,151],"a":[39,51,62,127,140,192],"programming":[40],"framework":[41],"that":[42,179],"addresses":[43],"this":[44],"challenge.":[45],"At":[46],"core,":[48],"nested":[52,99,128],"list-based":[53],"abstract":[54],"data":[55,100,104,144,161],"type":[56],"(ADT),":[57],"where":[58],"each":[59],"element":[60],"static":[65],"shape":[66],"or":[67],"another":[68],"(i.e.,":[70],"nested).":[71],"DNNs":[72,170],"then":[74,152],"de-fined":[75],"by":[76,183],"high-order":[77],"array":[78,84],"compute":[79],"like":[81,87,171],"map/reduce/scan":[82],"access":[85,105,164],"window/stride":[88],"FractalTensor.":[90],"new":[92,108],"way":[93],"of":[94,143],"DNN":[95],"definition":[96],"explicitly":[97],"exposes":[98],"parallelism":[101],"fine-grained":[103],"patterns,":[106],"opening":[107],"opportunities":[109],"for":[110,195],"whole":[111],"program":[112],"optimization.":[115],"exploit":[117],"these":[118],"opportunities,":[119],"from":[120],"FractalTensor-based":[122],"code":[123],"compiler":[125],"extracts":[126],"multi-dimensional":[129],"dataflow":[130],"graph":[131,159],"called":[132],"Extended":[133],"Task":[134],"Dependence":[135],"Graph":[136],"(ETDG),":[137],"which":[138],"provides":[139],"holistic":[141],"view":[142],"dependency":[145],"different":[147],"granularity.":[148],"The":[149],"ETDG":[150],"transformed":[153],"into":[154],"an":[155],"efficient":[156],"implementation":[157],"through":[158,191],"coarsening,":[160],"reordering,":[162],"materialization.":[165],"Evaluation":[166],"six":[168],"representative":[169],"RNN":[172],"FlashAttention":[174],"NVIDIA":[176],"A100":[177],"shows":[178],"Fractal-Tensor":[180],"achieves":[181],"speedup":[182],"to":[185],"5.45x":[186],"2.14x":[188],"average":[190],"unified":[193],"solution":[194],"diverse":[196],"optimizations.":[197]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}