{"id":"https://openalex.org/W4360606017","doi":"https://doi.org/10.1109/isscc42615.2023.10067615","title":"A 127.8TOPS/W Arbitrarily Quantized 1-to-8b Scalable-Precision Accelerator for General-Purpose Deep Learning with Reduction of Storage, Logic and Latency Waste","display_name":"A 127.8TOPS/W Arbitrarily Quantized 1-to-8b Scalable-Precision Accelerator for General-Purpose Deep Learning with Reduction of Storage, Logic and Latency Waste","publication_year":2023,"publication_date":"2023-02-19","ids":{"openalex":"https://openalex.org/W4360606017","doi":"https://doi.org/10.1109/isscc42615.2023.10067615"},"language":"en","primary_location":{"id":"doi:10.1109/isscc42615.2023.10067615","is_oa":false,"landing_page_url":"https://doi.org/10.1109/isscc42615.2023.10067615","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2023 IEEE International Solid- State Circuits Conference (ISSCC)","raw_type":"proceedings-article"},"type":"article","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/A5042758402","display_name":"Seunghyun Moon","orcid":"https://orcid.org/0000-0003-0027-2666"},"institutions":[{"id":"https://openalex.org/I123900574","display_name":"Pohang University of Science and Technology","ror":"https://ror.org/04xysgw12","country_code":"KR","type":"education","lineage":["https://openalex.org/I123900574"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Seunghyun Moon","raw_affiliation_strings":["Pohang University of Science and Technology,Pohang,Korea","Pohang University of Science and Technology, Pohang, Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Pohang University of Science and Technology,Pohang,Korea","institution_ids":["https://openalex.org/I123900574"]},{"raw_affiliation_string":"Pohang University of Science and Technology, Pohang, Korea","institution_ids":["https://openalex.org/I123900574"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133767555","display_name":"Han-Gyeol Mun","orcid":"https://orcid.org/0000-0002-8532-1385"},"institutions":[{"id":"https://openalex.org/I123900574","display_name":"Pohang University of Science and Technology","ror":"https://ror.org/04xysgw12","country_code":"KR","type":"education","lineage":["https://openalex.org/I123900574"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Han-Gyeol Mun","raw_affiliation_strings":["Pohang University of Science and Technology,Pohang,Korea","Pohang University of Science and Technology, Pohang, Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Pohang University of Science and Technology,Pohang,Korea","institution_ids":["https://openalex.org/I123900574"]},{"raw_affiliation_string":"Pohang University of Science and Technology, Pohang, Korea","institution_ids":["https://openalex.org/I123900574"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5027088400","display_name":"Hyunwoo Son","orcid":"https://orcid.org/0000-0002-9833-1718"},"institutions":[{"id":"https://openalex.org/I189442560","display_name":"Gyeongsang National University","ror":"https://ror.org/00saywf64","country_code":"KR","type":"education","lineage":["https://openalex.org/I189442560"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Hyunwoo Son","raw_affiliation_strings":["Gyeongsang National University,Jinju,Korea","Gyeongsang National University, Jinju, Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Gyeongsang National University,Jinju,Korea","institution_ids":["https://openalex.org/I189442560"]},{"raw_affiliation_string":"Gyeongsang National University, Jinju, Korea","institution_ids":["https://openalex.org/I189442560"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5033088278","display_name":"Jae\u2010Yoon Sim","orcid":"https://orcid.org/0000-0003-1814-6211"},"institutions":[{"id":"https://openalex.org/I123900574","display_name":"Pohang University of Science and Technology","ror":"https://ror.org/04xysgw12","country_code":"KR","type":"education","lineage":["https://openalex.org/I123900574"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jae-Yoon Sim","raw_affiliation_strings":["Pohang University of Science and Technology,Pohang,Korea","Pohang University of Science and Technology, Pohang, Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Pohang University of Science and Technology,Pohang,Korea","institution_ids":["https://openalex.org/I123900574"]},{"raw_affiliation_string":"Pohang University of Science and Technology, Pohang, Korea","institution_ids":["https://openalex.org/I123900574"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":1.8082,"has_fulltext":false,"cited_by_count":17,"citation_normalized_percentile":{"value":0.8724713,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":94,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"21","last_page":"23"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9997000098228455,"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":0.9997000098228455,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.9961000084877014,"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":0.9950000047683716,"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":0.791319727897644},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.734711766242981},{"id":"https://openalex.org/keywords/bottleneck","display_name":"Bottleneck","score":0.6977139115333557},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.607614278793335},{"id":"https://openalex.org/keywords/compressed-sensing","display_name":"Compressed sensing","score":0.5843939185142517},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.5338937044143677},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5197417736053467},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.5048754811286926},{"id":"https://openalex.org/keywords/computer-hardware","display_name":"Computer hardware","score":0.3702460825443268},{"id":"https://openalex.org/keywords/computer-architecture","display_name":"Computer architecture","score":0.35229188203811646},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.33386218547821045},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.3022245466709137},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.20953476428985596}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.791319727897644},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.734711766242981},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.6977139115333557},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.607614278793335},{"id":"https://openalex.org/C124851039","wikidata":"https://www.wikidata.org/wiki/Q2665459","display_name":"Compressed sensing","level":2,"score":0.5843939185142517},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.5338937044143677},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5197417736053467},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.5048754811286926},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.3702460825443268},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.35229188203811646},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.33386218547821045},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3022245466709137},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.20953476428985596},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/isscc42615.2023.10067615","is_oa":false,"landing_page_url":"https://doi.org/10.1109/isscc42615.2023.10067615","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2023 IEEE International Solid- State Circuits Conference (ISSCC)","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":12,"referenced_works":["https://openalex.org/W2786771851","https://openalex.org/W2792607576","https://openalex.org/W2963480671","https://openalex.org/W2964133305","https://openalex.org/W3171513315","https://openalex.org/W4220686330","https://openalex.org/W4220764408","https://openalex.org/W4220865834","https://openalex.org/W4285255298","https://openalex.org/W4286571858","https://openalex.org/W6791603202","https://openalex.org/W6809349333"],"related_works":["https://openalex.org/W1657880117","https://openalex.org/W2595172197","https://openalex.org/W2127970246","https://openalex.org/W2084856301","https://openalex.org/W1001352512","https://openalex.org/W3183118997","https://openalex.org/W3214410901","https://openalex.org/W3204400881","https://openalex.org/W3204296682","https://openalex.org/W2917767146"],"abstract_inverted_index":{"Research":[0],"on":[1,7],"deep":[2,439],"learning":[3,440],"accelerators":[4,396],"has":[5,234,284,297],"focused":[6],"inference":[8],"tasks":[9],"to":[10,55,74,78,110,245,333,364,388,393,408,424,442,478,483],"improve":[11,484],"performance":[12],"by":[13,165,186,269,314,384,453],"means":[14],"of":[15,30,33,144,159,190,239,241,272,305,331,411,427],"maximally":[16],"utilizing":[17],"sparsity":[18,41,60,101,114,154,189],"and":[19,42,61,102,206,228,278,372,390,399,404,449,455,476,482,496],"quantization.":[20],"Unlike":[21],"CNN-only":[22],"networks,":[23],"however,":[24],"recent":[25,148],"state-of-the-art":[26],"(SOTA)":[27],"models":[28,69,77,149],"consist":[29],"multiple":[31,86],"blocks":[32],"various":[34,68],"layers":[35],"with":[36,126,327,359,401,447,471],"different":[37],"layer-by-layer":[38,128],"characteristics":[39],"in":[40,49,80,92,122,134,153,172,197,207,257],"required":[43,103,106,368],"precision.":[44,104],"This":[45,116,414],"trend":[46],"presents":[47],"challenges":[48],"building":[50],"a":[51,127,135,156,169,177,222,236,328,338,353,360,402,417,435,465],"general":[52],"accelerator":[53,179,441,459],"architecture":[54,356],"maximize":[56],"the":[57,75,81,90,98,113,118,124,141,183,188,191,201,214,225,229,312,322,367,376,409,422],"benefits":[58],"from":[59,71],"quantization,":[62],"while":[63],"supporting":[64],"efficient":[65,303],"processing":[66,304,500],"for":[67,120,302,357,370,421],"ranging":[70],"traditional":[72],"CNNs":[73],"new":[76],"come":[79],"future.":[82],"First,":[83],"there":[84,220,296],"are":[85,267,374],"considerations":[87],"that":[88,180,469],"include":[89],"bottleneck":[91],"data":[93,133,160,173,184,473],"bandwidth,":[94,486],"as":[95,97,112,249,316],"well":[96],"trade-off":[99],"between":[100,224],"The":[105,307,458],"precision":[107,227,330,362,406],"is":[108,195,204,217,221,335,382],"likely":[109],"increase":[111],"increases.":[115],"underpins":[117],"need":[119],"flexibility":[121],"setting":[123],"quantization":[125,230,233,247,281,291],"configuration.":[129],"In":[130],"addition,":[131],"storing":[132],"unified":[136],"format":[137,185,199,209],"can":[138],"also":[139],"prohibit":[140],"maximum":[142],"utilization":[143],"hardware":[145,300,345,491],"resources.":[146],"Since":[147],"have":[150,252],"large":[151],"variations":[152],"[11],":[155],"major":[157],"portion":[158],"movement":[161],"might":[162],"be":[163],"taken":[164],"sending":[166],"zeros,":[167],"causing":[168],"severe":[170],"waste":[171,410,426],"bandwidth.":[174],"We":[175],"propose":[176,352,434],"sparsity-aware":[178],"adaptively":[181],"changes":[182],"detecting":[187],"given":[192,268],"task.":[193],"Data":[194],"stored":[196],"raw":[198,475],"when":[200,213],"sparse":[202,215],"rate":[203,216],"low":[205],"compressed":[208],"(run-length":[210],"coding,":[211],"RLC)":[212],"high.":[218],"Second,":[219],"correlation":[223],"effective":[226,485],"policy.":[231],"Arbitrary":[232],"demonstrated":[235],"higher":[237],"level":[238],"quality":[240],"result":[242],"(QoR)":[243],"compared":[244],"linear":[246,270],"(denoted":[248],"INT).":[250],"There":[251],"been":[253,298],"two":[254,472],"representative":[255],"approaches":[256],"nonlinear":[258],"quantization:":[259],"1)":[260,464],"arbitrary":[261,280,285],"basis":[262],"(AB)":[263],"where":[264],"quantized":[265,287,452],"values":[266],"combinations":[271],"<tex":[273],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[274],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">$n$</tex>":[275],"independent":[276],"bases,":[277],"2)":[279,487],"(AQ)":[282],"which":[283],"2":[286],"values.":[288],"Though":[289],"these":[290],"schemes":[292],"achieve":[293],"good":[294],"accuracy,":[295],"no":[299],"implementation":[301],"AQ.":[306],"conventional":[308],"INT":[309],"multiplication":[310,420],"increases":[311],"complexity":[313],"4x":[315],"both":[317],"input":[318,448],"precisions":[319,369],"double.":[320],"On":[321],"other":[323],"hand,":[324],"if":[325],"AQ":[326,358,454,489],"scalable":[329,361],"up":[332,363],"8b":[334],"implemented":[336],"using":[337],"look-up-table":[339],"(LUT)":[340],"approach,":[341],"it":[342],"would":[343],"explode":[344],"complexity.":[346],"To":[347],"resolve":[348],"this":[349],"problem,":[350],"we":[351,433],"hierarchical":[354],"decoding":[355],"8b.":[365],"Finally,":[366],"inputs":[371,389,398],"weights":[373,400,423],"not":[375],"same":[377],"[4],":[378],"[10].":[379],"Good":[380],"QoR":[381],"realized":[383],"assigning":[385],"more":[386],"bits":[387,392],"fewer":[391],"weights.":[394],"Previous":[395],"handle":[397],"fixed":[403],"equal":[405],"leading":[407],"computational":[412],"energy.":[413,428],"work":[415],"employs":[416],"dynamic-precision":[418],"bit-serial":[419,498],"minimize":[425],"Putting":[429],"them":[430],"all":[431],"together,":[432],"1-to-8b":[436],"scalable-precision":[437],"general-purpose":[438],"support":[443],"multiply-and-accumulate":[444],"(MAC)":[445],"operations":[446],"weight":[450],"vectors":[451],"AB,":[456],"respectively.":[457],"includes":[460],"three":[461],"main":[462],"features:":[463],"zero":[466],"elimination":[467],"scheme":[468],"works":[470],"formats,":[474],"RLC,":[477],"save":[479],"storage":[480],"cost":[481],"extended-precision":[488],"computing":[490],"without":[492,501],"exploding":[493],"logic":[494],"complexity,":[495],"3)":[497],"AB":[499],"unnecessary":[502],"computations.":[503]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":9},{"year":2024,"cited_by_count":3},{"year":2023,"cited_by_count":4}],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}