{"id":"https://openalex.org/W4395106447","doi":"https://doi.org/10.1145/3620666.3651357","title":"PrimePar: Efficient Spatial-temporal Tensor Partitioning for Large Transformer Model Training","display_name":"PrimePar: Efficient Spatial-temporal Tensor Partitioning for Large Transformer Model Training","publication_year":2024,"publication_date":"2024-04-24","ids":{"openalex":"https://openalex.org/W4395106447","doi":"https://doi.org/10.1145/3620666.3651357"},"language":"en","primary_location":{"id":"doi:10.1145/3620666.3651357","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3620666.3651357","pdf_url":null,"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 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 3","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1145/3620666.3651357","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5037211561","display_name":"Haoran Wang","orcid":"https://orcid.org/0000-0002-5968-8845"},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"funder","lineage":["https://openalex.org/I19820366"]},{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Haoran Wang","raw_affiliation_strings":["Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210090176","https://openalex.org/I19820366"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5005666540","display_name":"Lei Wang","orcid":"https://orcid.org/0009-0006-2313-5348"},"institutions":[{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]},{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"funder","lineage":["https://openalex.org/I19820366"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Lei Wang","raw_affiliation_strings":["Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210090176","https://openalex.org/I19820366"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5044810210","display_name":"Haobo Xu","orcid":"https://orcid.org/0000-0002-0243-6516"},"institutions":[{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]},{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"funder","lineage":["https://openalex.org/I19820366"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Haobo Xu","raw_affiliation_strings":["Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210090176","https://openalex.org/I19820366"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100346965","display_name":"Ying Wang","orcid":"https://orcid.org/0000-0001-5172-4736"},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"funder","lineage":["https://openalex.org/I19820366"]},{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ying Wang","raw_affiliation_strings":["Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210090176","https://openalex.org/I19820366"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5015161455","display_name":"Yuming Li","orcid":"https://orcid.org/0009-0004-4414-6682"},"institutions":[{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]},{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"funder","lineage":["https://openalex.org/I19820366"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yuming Li","raw_affiliation_strings":["Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210090176","https://openalex.org/I19820366"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101512833","display_name":"Yinhe Han","orcid":"https://orcid.org/0000-0001-5113-8067"},"institutions":[{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]},{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"funder","lineage":["https://openalex.org/I19820366"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yinhe Han","raw_affiliation_strings":["Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China"],"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210090176","https://openalex.org/I19820366"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5037211561"],"corresponding_institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210090176"],"apc_list":null,"apc_paid":null,"fwci":2.0509,"has_fulltext":false,"cited_by_count":6,"citation_normalized_percentile":{"value":0.85913853,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":90,"max":100},"biblio":{"volume":null,"issue":null,"first_page":"801","last_page":"817"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12303","display_name":"Tensor decomposition and applications","score":0.9970999956130981,"subfield":{"id":"https://openalex.org/subfields/2605","display_name":"Computational Mathematics"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T12303","display_name":"Tensor decomposition and applications","score":0.9970999956130981,"subfield":{"id":"https://openalex.org/subfields/2605","display_name":"Computational Mathematics"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.993399977684021,"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/T11269","display_name":"Algorithms and Data Compression","score":0.9589999914169312,"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/computer-science","display_name":"Computer science","score":0.7741988301277161},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.6677596569061279},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.6476051807403564},{"id":"https://openalex.org/keywords/partition","display_name":"Partition (number theory)","score":0.5157577991485596},{"id":"https://openalex.org/keywords/tensor","display_name":"Tensor (intrinsic definition)","score":0.4788232147693634},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.4736907482147217},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.46707776188850403},{"id":"https://openalex.org/keywords/space-partitioning","display_name":"Space partitioning","score":0.4620899558067322},{"id":"https://openalex.org/keywords/theoretical-computer-science","display_name":"Theoretical computer science","score":0.4437350630760193},{"id":"https://openalex.org/keywords/overlay","display_name":"Overlay","score":0.4180137515068054},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.3588571548461914},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.3295295238494873},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.12933072447776794}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7741988301277161},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.6677596569061279},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.6476051807403564},{"id":"https://openalex.org/C42812","wikidata":"https://www.wikidata.org/wiki/Q1082910","display_name":"Partition (number theory)","level":2,"score":0.5157577991485596},{"id":"https://openalex.org/C155281189","wikidata":"https://www.wikidata.org/wiki/Q3518150","display_name":"Tensor (intrinsic definition)","level":2,"score":0.4788232147693634},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.4736907482147217},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.46707776188850403},{"id":"https://openalex.org/C13670688","wikidata":"https://www.wikidata.org/wiki/Q3500548","display_name":"Space partitioning","level":2,"score":0.4620899558067322},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.4437350630760193},{"id":"https://openalex.org/C136085584","wikidata":"https://www.wikidata.org/wiki/Q910289","display_name":"Overlay","level":2,"score":0.4180137515068054},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3588571548461914},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3295295238494873},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.12933072447776794},{"id":"https://openalex.org/C202444582","wikidata":"https://www.wikidata.org/wiki/Q837863","display_name":"Pure mathematics","level":1,"score":0.0},{"id":"https://openalex.org/C114614502","wikidata":"https://www.wikidata.org/wiki/Q76592","display_name":"Combinatorics","level":1,"score":0.0},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.0},{"id":"https://openalex.org/C2524010","wikidata":"https://www.wikidata.org/wiki/Q8087","display_name":"Geometry","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3620666.3651357","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3620666.3651357","pdf_url":null,"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 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 3","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3620666.3651357","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3620666.3651357","pdf_url":null,"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 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 3","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":56,"referenced_works":["https://openalex.org/W1442374986","https://openalex.org/W1833917188","https://openalex.org/W1969805974","https://openalex.org/W1980723535","https://openalex.org/W1997835767","https://openalex.org/W2024804689","https://openalex.org/W2053546108","https://openalex.org/W2060393849","https://openalex.org/W2090457779","https://openalex.org/W2102056800","https://openalex.org/W2115295617","https://openalex.org/W2138503535","https://openalex.org/W2140639562","https://openalex.org/W2339765813","https://openalex.org/W2433451779","https://openalex.org/W2605347906","https://openalex.org/W2612076670","https://openalex.org/W2734941459","https://openalex.org/W2786066748","https://openalex.org/W2883830791","https://openalex.org/W2884700152","https://openalex.org/W2901299405","https://openalex.org/W2950813464","https://openalex.org/W2963358710","https://openalex.org/W2969388332","https://openalex.org/W2970971581","https://openalex.org/W2983101505","https://openalex.org/W3016430712","https://openalex.org/W3030163527","https://openalex.org/W3036879053","https://openalex.org/W3037847693","https://openalex.org/W3040573126","https://openalex.org/W3101104221","https://openalex.org/W3115267280","https://openalex.org/W3129831491","https://openalex.org/W3130395060","https://openalex.org/W3134495297","https://openalex.org/W3204998121","https://openalex.org/W3205803342","https://openalex.org/W4224308101","https://openalex.org/W4226479682","https://openalex.org/W4239722617","https://openalex.org/W4249516033","https://openalex.org/W4280611847","https://openalex.org/W4287748830","https://openalex.org/W4292779060","https://openalex.org/W4308760226","https://openalex.org/W4318541692","https://openalex.org/W4322718191","https://openalex.org/W4380874786","https://openalex.org/W4384918448","https://openalex.org/W6638803421","https://openalex.org/W6683722107","https://openalex.org/W6684859321","https://openalex.org/W6739901393","https://openalex.org/W6811340617"],"related_works":["https://openalex.org/W2058965144","https://openalex.org/W2164382479","https://openalex.org/W2146343568","https://openalex.org/W3008857645","https://openalex.org/W3037605386","https://openalex.org/W2073947114","https://openalex.org/W2061980994","https://openalex.org/W2090617251","https://openalex.org/W4389304954","https://openalex.org/W1540431531"],"abstract_inverted_index":{"With":[0],"the":[1,58,66,73,117,227],"rapid":[2],"up-scaling":[3],"of":[4,60,79],"transformer-based":[5],"large":[6],"language":[7],"models":[8,12],"(LLM),":[9],"training":[10,20,42,48,81,149,167,206,217,220,244],"these":[11],"is":[13,24,63,232],"becoming":[14],"increasingly":[15],"demanding":[16],"on":[17,40,52],"novel":[18],"parallel":[19,28,47,144,164,181],"techniques.":[21,159],"Tensor":[22],"partitioning":[23,55,78,200],"an":[25,174],"extensively":[26],"researched":[27],"technique,":[29],"encompassing":[30],"data":[31],"and":[32,35,95,119,126,141],"model":[33,166],"parallelism,":[34,239],"has":[36],"a":[37,107,137],"significant":[38],"influence":[39],"LLM":[41,80],"performance.":[43],"However,":[44],"existing":[45],"state-of-the-art":[46,158,215],"systems":[49,218],"are":[50],"based":[51],"incomplete":[53],"tensor":[54,77,109,186,199],"space,":[56],"where":[57],"distribution":[59],"partitioned":[61],"sub-operators":[62,114],"limited":[64],"to":[65,86,122,143,157,168,203,214,224,241],"spatial":[67,118],"dimension.":[68],"We":[69],"discover":[70],"that":[71,112,146,177,196],"introducing":[72],"temporal":[74,120],"dimension":[75],"into":[76],"instance":[82],"provides":[83],"extra":[84],"opportunities":[85],"avoid":[87],"collective":[88],"communication":[89,99,125],"across":[90,230],"devices,":[91,170],"saving":[92],"memory":[93,127,154,211],"space":[94,140,188],"also":[96],"overlapping":[97],"device-to-device":[98],"with":[100,151,189,208],"computation.":[101],"In":[102],"this":[103],"paper,":[104],"we":[105,171],"propose":[106],"new":[108,134],"partition":[110,187],"primitive":[111,135],"distributes":[113],"along":[115],"both":[116],"dimensions":[121],"further":[123,172],"explore":[124],"overhead":[128],"reduction":[129],"over":[130],"current":[131],"solutions.":[132],"This":[133],"creates":[136],"broader":[138],"parallelization":[139],"leads":[142],"solutions":[145,182],"achieve":[147],"better":[148],"throughput":[150,207,245],"lower":[152],"peak":[153,210],"occupancy":[155,212],"compared":[156,213],"To":[160],"efficiently":[161],"deploy":[162],"optimized":[163,198],"transformer":[165],"multiple":[169],"present":[173],"optimization":[175],"algorithm":[176],"can":[178,246],"find":[179],"optimal":[180],"from":[183],"our":[184,197],"spatial-temporal":[185],"acceptable":[190],"search":[191],"time.":[192],"Our":[193],"evaluation":[194],"shows":[195],"achieves":[201],"up":[202,240],"1.68":[204],"\u00d7":[205,243],"69%":[209],"distributed":[216],"when":[219],"LLMs.":[221],"Upon":[222],"scaling":[223],"32":[225],"GPUs,":[226],"geo-mean":[228],"speedup":[229],"benchmarks":[231],"1.30":[233],"\u00d7.":[234],"When":[235],"applied":[236],"in":[237],"3D":[238],"1.46":[242],"be":[247],"achieved.":[248]},"counts_by_year":[{"year":2026,"cited_by_count":2},{"year":2025,"cited_by_count":3},{"year":2024,"cited_by_count":1}],"updated_date":"2026-03-06T13:50:29.536080","created_date":"2025-10-10T00:00:00"}