{"id":"https://openalex.org/W4389491934","doi":"https://doi.org/10.1145/3613424.3614299","title":"Improving Data Reuse in NPU On-chip Memory with Interleaved Gradient Order for DNN Training","display_name":"Improving Data Reuse in NPU On-chip Memory with Interleaved Gradient Order for DNN Training","publication_year":2023,"publication_date":"2023-10-28","ids":{"openalex":"https://openalex.org/W4389491934","doi":"https://doi.org/10.1145/3613424.3614299"},"language":"en","primary_location":{"id":"doi:10.1145/3613424.3614299","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3613424.3614299","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3613424.3614299","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"56th Annual IEEE/ACM International Symposium on Microarchitecture","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/10.1145/3613424.3614299","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101538463","display_name":"Jung-Woo Kim","orcid":"https://orcid.org/0000-0003-4773-9251"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jungwoo Kim","raw_affiliation_strings":["KAIST, Republic of Korea"],"raw_orcid":"https://orcid.org/0000-0003-4773-9251","affiliations":[{"raw_affiliation_string":"KAIST, Republic of Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5025353753","display_name":"Seonjin Na","orcid":"https://orcid.org/0009-0009-0734-8126"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Seonjin Na","raw_affiliation_strings":["KAIST, Korea, South ? Republic of Korea"],"raw_orcid":"https://orcid.org/0009-0009-0734-8126","affiliations":[{"raw_affiliation_string":"KAIST, Korea, South ? Republic of Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100756237","display_name":"Sanghyeon Lee","orcid":"https://orcid.org/0009-0003-7060-6315"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Sanghyeon Lee","raw_affiliation_strings":["KAIST, Korea, South ? Republic of Korea","KAIST, Republic of Korea"],"raw_orcid":"https://orcid.org/0009-0003-7060-6315","affiliations":[{"raw_affiliation_string":"KAIST, Korea, South ? Republic of Korea","institution_ids":["https://openalex.org/I157485424"]},{"raw_affiliation_string":"KAIST, Republic of Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Sunho Lee","orcid":"https://orcid.org/0000-0003-4362-9565"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Sunho Lee","raw_affiliation_strings":["KAIST, Korea, South ? Republic of Korea","KAIST, Republic of Korea"],"raw_orcid":"https://orcid.org/0000-0003-4362-9565","affiliations":[{"raw_affiliation_string":"KAIST, Korea, South ? Republic of Korea","institution_ids":["https://openalex.org/I157485424"]},{"raw_affiliation_string":"KAIST, Republic of Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5047149607","display_name":"Jaehyuk Huh","orcid":"https://orcid.org/0000-0002-1742-047X"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jaehyuk Huh","raw_affiliation_strings":["KAIST, Korea, South ? Republic of Korea"],"raw_orcid":"https://orcid.org/0000-0002-1742-047X","affiliations":[{"raw_affiliation_string":"KAIST, Korea, South ? Republic of Korea","institution_ids":["https://openalex.org/I157485424"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":5,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":1.1533,"has_fulltext":true,"cited_by_count":4,"citation_normalized_percentile":{"value":0.78144487,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":90,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"438","last_page":"451"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9995999932289124,"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":0.9995999932289124,"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":0.9993000030517578,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.9983000159263611,"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":0.8238959908485413},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.6546640396118164},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5933833122253418},{"id":"https://openalex.org/keywords/interleaving","display_name":"Interleaving","score":0.5428943037986755},{"id":"https://openalex.org/keywords/reuse","display_name":"Reuse","score":0.49566757678985596},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.3976951241493225},{"id":"https://openalex.org/keywords/computer-hardware","display_name":"Computer hardware","score":0.3644660413265228},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.3299238681793213}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8238959908485413},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.6546640396118164},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5933833122253418},{"id":"https://openalex.org/C28034677","wikidata":"https://www.wikidata.org/wiki/Q17092530","display_name":"Interleaving","level":2,"score":0.5428943037986755},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.49566757678985596},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3976951241493225},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.3644660413265228},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3299238681793213},{"id":"https://openalex.org/C18903297","wikidata":"https://www.wikidata.org/wiki/Q7150","display_name":"Ecology","level":1,"score":0.0},{"id":"https://openalex.org/C86803240","wikidata":"https://www.wikidata.org/wiki/Q420","display_name":"Biology","level":0,"score":0.0},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3613424.3614299","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3613424.3614299","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3613424.3614299","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"56th Annual IEEE/ACM International Symposium on Microarchitecture","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3613424.3614299","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3613424.3614299","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3613424.3614299","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"56th Annual IEEE/ACM International Symposium on Microarchitecture","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G12967831","display_name":null,"funder_award_id":"IITP2017-0-00466","funder_id":"https://openalex.org/F4320335489","funder_display_name":"Institute for Information and Communications Technology Promotion"}],"funders":[{"id":"https://openalex.org/F4320328359","display_name":"Ministry of Science and ICT, South Korea","ror":"https://ror.org/01wpjm123"},{"id":"https://openalex.org/F4320335489","display_name":"Institute for Information and Communications Technology Promotion","ror":"https://ror.org/01g0hqq23"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4389491934.pdf","grobid_xml":"https://content.openalex.org/works/W4389491934.grobid-xml"},"referenced_works_count":39,"referenced_works":["https://openalex.org/W585658244","https://openalex.org/W1967134278","https://openalex.org/W2067523571","https://openalex.org/W2296514843","https://openalex.org/W2338908902","https://openalex.org/W2442974303","https://openalex.org/W2565305208","https://openalex.org/W2593390416","https://openalex.org/W2605347906","https://openalex.org/W2606722458","https://openalex.org/W2622263826","https://openalex.org/W2799042347","https://openalex.org/W2803002685","https://openalex.org/W2804032941","https://openalex.org/W2884361556","https://openalex.org/W2884711234","https://openalex.org/W2940862705","https://openalex.org/W3015636663","https://openalex.org/W3017521908","https://openalex.org/W3036878841","https://openalex.org/W3040850704","https://openalex.org/W3127736057","https://openalex.org/W3174529902","https://openalex.org/W3178941450","https://openalex.org/W3186289964","https://openalex.org/W3192708755","https://openalex.org/W3197788851","https://openalex.org/W4231063867","https://openalex.org/W4233429846","https://openalex.org/W4285113157","https://openalex.org/W4285264221","https://openalex.org/W4285789112","https://openalex.org/W4288083528","https://openalex.org/W4297736277","https://openalex.org/W4318541578","https://openalex.org/W6600768174","https://openalex.org/W6608548112","https://openalex.org/W6703652217","https://openalex.org/W6778883912"],"related_works":["https://openalex.org/W1655266410","https://openalex.org/W2389051085","https://openalex.org/W2330343234","https://openalex.org/W1901012776","https://openalex.org/W2463883322","https://openalex.org/W2229382548","https://openalex.org/W2391789612","https://openalex.org/W2814468324","https://openalex.org/W2389236462","https://openalex.org/W1614034078"],"abstract_inverted_index":{"During":[0],"training":[1],"tasks":[2],"for":[3,182,201,236,240,266],"machine":[4],"learning":[5],"models":[6],"with":[7,248,273],"neural":[8],"processing":[9,101],"units":[10],"(NPUs),":[11],"the":[12,17,33,44,55,64,78,92,105,121,143,153,178,183,189,195,211,217,230,256,278,282],"most":[13,212],"time-consuming":[14],"part":[15],"is":[16,198],"backward":[18,56,79,82,122],"pass,":[19,123],"which":[20],"incurs":[21],"significant":[22],"overheads":[23],"due":[24],"to":[25,31,62,77,141,168,187,209],"off-chip":[26,41],"memory":[27,48],"accesses.":[28,173],"For":[29],"NPUs,":[30],"mitigate":[32],"long":[34],"latency":[35],"and":[36,87,155,233,242,251,264,268],"limited":[37],"bandwidth":[38],"of":[39,66,117,138,145],"such":[40],"DRAM":[42],"accesses,":[43],"software-managed":[45],"on-chip":[46],"scratchpad":[47,147],"(SPM)":[49],"plays":[50],"a":[51,71,127,164,206,274],"crucial":[52],"role.":[53],"As":[54],"pass":[57,83],"computation":[58],"must":[59],"be":[60],"optimized":[61],"improve":[63,260],"effectiveness":[65],"SPM,":[67],"this":[68,114,124],"study":[69,125],"identifies":[70],"new":[72,115],"data":[73,108,118,129,191,225],"reuse":[74,109,119,226],"pattern":[75],"specific":[76],"computation.":[80],"The":[81,149,174,220,245],"includes":[84],"independent":[85],"input":[86,154],"weight":[88,156],"gradient":[89,95,135,157,172,185,238],"computations":[90,158,186,239],"sharing":[91],"same":[93],"output":[94,171],"in":[96,120],"each":[97],"layer.":[98],"Conventional":[99],"sequential":[100],"does":[102],"not":[103,199],"exploit":[104],"potential":[106,190],"inter-operation":[107],"opportunity":[110,116],"within":[111],"SPM.":[112],"With":[113],"proposes":[126],"novel":[128],"flow":[130],"transformation":[131],"scheme":[132,235],"called":[133],"interleaved":[134,184],"order,":[136],"consisting":[137],"three":[139],"techniques":[140,258,280],"enhance":[142],"utilization":[144],"NPU":[146],"memory.":[148],"first":[150],"technique":[151,176,222],"shuffles":[152],"by":[159,228,262,285],"interleaving":[160],"two":[161,237],"operations":[162],"into":[163],"single":[165],"fused":[166],"operation":[167],"reduce":[169,281],"redundant":[170],"second":[175],"adjusts":[177],"tile":[179],"access":[180],"order":[181,197,214],"maximize":[188],"locality.":[192],"However,":[193],"since":[194],"best":[196,231],"fixed":[200],"all":[202],"tensors,":[203],"we":[204],"propose":[205],"selection":[207],"algorithm":[208],"find":[210],"suitable":[213],"based":[215],"on":[216],"tensor":[218],"dimensions.":[219],"final":[221],"further":[223],"improves":[224],"chances":[227],"using":[229],"partitioning":[232],"mapping":[234],"single-core":[241,249],"multi-core":[243],"NPUs.":[244],"simulation-based":[246],"evaluation":[247],"edge":[250,267],"server":[252,269,276],"NPUs":[253,270],"shows":[254],"that":[255],"combined":[257],"can":[259],"performance":[261],"29.3%":[263],"14.5%":[265],"respectively.":[271],"Furthermore,":[272],"quad-core":[275],"NPU,":[277],"proposed":[279],"execution":[283],"time":[284],"23.7%.":[286]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":2},{"year":2023,"cited_by_count":1}],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}
