{"id":"https://openalex.org/W4413967157","doi":"https://doi.org/10.1109/jssc.2025.3603187","title":"T-REX: Hardware\u2013Software Co-Optimized Transformer Accelerator With Reduced External Memory Access and Enhanced Hardware Utilization","display_name":"T-REX: Hardware\u2013Software Co-Optimized Transformer Accelerator With Reduced External Memory Access and Enhanced Hardware Utilization","publication_year":2025,"publication_date":"2025-09-03","ids":{"openalex":"https://openalex.org/W4413967157","doi":"https://doi.org/10.1109/jssc.2025.3603187"},"language":"en","primary_location":{"id":"doi:10.1109/jssc.2025.3603187","is_oa":false,"landing_page_url":"https://doi.org/10.1109/jssc.2025.3603187","pdf_url":null,"source":{"id":"https://openalex.org/S83637746","display_name":"IEEE Journal of Solid-State Circuits","issn_l":"0018-9200","issn":["0018-9200","1558-173X"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Journal of Solid-State Circuits","raw_type":"journal-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/I24062138","display_name":"Konkuk University","ror":"https://ror.org/025h1m602","country_code":"KR","type":"education","lineage":["https://openalex.org/I24062138"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Seunghyun Moon","raw_affiliation_strings":["Konkuk University, Seoul, Republic of Korea"],"raw_orcid":"https://orcid.org/0000-0003-0027-2666","affiliations":[{"raw_affiliation_string":"Konkuk University, Seoul, Republic of Korea","institution_ids":["https://openalex.org/I24062138"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100786430","display_name":"Mao Li","orcid":"https://orcid.org/0009-0009-5968-7250"},"institutions":[{"id":"https://openalex.org/I78577930","display_name":"Columbia University","ror":"https://ror.org/00hj8s172","country_code":"US","type":"education","lineage":["https://openalex.org/I78577930"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Mao Li","raw_affiliation_strings":["Columbia University, New York, NY, USA"],"raw_orcid":"https://orcid.org/0009-0009-5968-7250","affiliations":[{"raw_affiliation_string":"Columbia University, New York, NY, USA","institution_ids":["https://openalex.org/I78577930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5014493157","display_name":"Gregory K. Chen","orcid":"https://orcid.org/0000-0002-4813-3844"},"institutions":[{"id":"https://openalex.org/I1343180700","display_name":"Intel (United States)","ror":"https://ror.org/01ek73717","country_code":"US","type":"company","lineage":["https://openalex.org/I1343180700"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Gregory K. Chen","raw_affiliation_strings":["Intel Corporation, Hillsboro, OR, USA"],"raw_orcid":"https://orcid.org/0000-0002-4813-3844","affiliations":[{"raw_affiliation_string":"Intel Corporation, Hillsboro, OR, USA","institution_ids":["https://openalex.org/I1343180700"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5030456056","display_name":"Phil Knag","orcid":"https://orcid.org/0000-0001-6794-8806"},"institutions":[{"id":"https://openalex.org/I1343180700","display_name":"Intel (United States)","ror":"https://ror.org/01ek73717","country_code":"US","type":"company","lineage":["https://openalex.org/I1343180700"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Phil C. Knag","raw_affiliation_strings":["Intel Corporation, Hillsboro, OR, USA"],"raw_orcid":"https://orcid.org/0000-0001-6794-8806","affiliations":[{"raw_affiliation_string":"Intel Corporation, Hillsboro, OR, USA","institution_ids":["https://openalex.org/I1343180700"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5074107306","display_name":"Ram Krishnamurthy","orcid":"https://orcid.org/0000-0002-2428-7099"},"institutions":[{"id":"https://openalex.org/I1343180700","display_name":"Intel (United States)","ror":"https://ror.org/01ek73717","country_code":"US","type":"company","lineage":["https://openalex.org/I1343180700"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ram K. Krishnamurthy","raw_affiliation_strings":["Intel Corporation, Hillsboro, OR, USA"],"raw_orcid":"https://orcid.org/0000-0002-2428-7099","affiliations":[{"raw_affiliation_string":"Intel Corporation, Hillsboro, OR, USA","institution_ids":["https://openalex.org/I1343180700"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5011887658","display_name":"Mingoo Seok","orcid":"https://orcid.org/0000-0002-9722-0979"},"institutions":[{"id":"https://openalex.org/I78577930","display_name":"Columbia University","ror":"https://ror.org/00hj8s172","country_code":"US","type":"education","lineage":["https://openalex.org/I78577930"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Mingoo Seok","raw_affiliation_strings":["Columbia University, New York, NY, USA"],"raw_orcid":"https://orcid.org/0000-0002-9722-0979","affiliations":[{"raw_affiliation_string":"Columbia University, New York, NY, USA","institution_ids":["https://openalex.org/I78577930"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.2253,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.49210684,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":91,"max":95},"biblio":{"volume":"61","issue":"1","first_page":"154","last_page":"167"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10468","display_name":"Photovoltaic System Optimization Techniques","score":0.9890999794006348,"subfield":{"id":"https://openalex.org/subfields/2105","display_name":"Renewable Energy, Sustainability and the Environment"},"field":{"id":"https://openalex.org/fields/21","display_name":"Energy"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10468","display_name":"Photovoltaic System Optimization Techniques","score":0.9890999794006348,"subfield":{"id":"https://openalex.org/subfields/2105","display_name":"Renewable Energy, Sustainability and the Environment"},"field":{"id":"https://openalex.org/fields/21","display_name":"Energy"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11222","display_name":"Magnetic Properties and Applications","score":0.9828000068664551,"subfield":{"id":"https://openalex.org/subfields/2504","display_name":"Electronic, Optical and Magnetic Materials"},"field":{"id":"https://openalex.org/fields/25","display_name":"Materials Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11343","display_name":"Power Transformer Diagnostics and Insulation","score":0.972100019454956,"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/software","display_name":"Software","score":0.6284668445587158},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.5633226037025452},{"id":"https://openalex.org/keywords/hardware-acceleration","display_name":"Hardware acceleration","score":0.5598280429840088},{"id":"https://openalex.org/keywords/computer-hardware","display_name":"Computer hardware","score":0.5482750535011292},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5212091207504272},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.48344120383262634},{"id":"https://openalex.org/keywords/operating-system","display_name":"Operating system","score":0.35585665702819824},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.2814025282859802},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.25060996413230896},{"id":"https://openalex.org/keywords/voltage","display_name":"Voltage","score":0.087675541639328}],"concepts":[{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.6284668445587158},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.5633226037025452},{"id":"https://openalex.org/C13164978","wikidata":"https://www.wikidata.org/wiki/Q600158","display_name":"Hardware acceleration","level":3,"score":0.5598280429840088},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.5482750535011292},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5212091207504272},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.48344120383262634},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.35585665702819824},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.2814025282859802},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.25060996413230896},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.087675541639328}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/jssc.2025.3603187","is_oa":false,"landing_page_url":"https://doi.org/10.1109/jssc.2025.3603187","pdf_url":null,"source":{"id":"https://openalex.org/S83637746","display_name":"IEEE Journal of Solid-State Circuits","issn_l":"0018-9200","issn":["0018-9200","1558-173X"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Journal of Solid-State Circuits","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320306087","display_name":"Semiconductor Research Corporation","ror":"https://ror.org/047z4n946"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":32,"referenced_works":["https://openalex.org/W2027886943","https://openalex.org/W2794952988","https://openalex.org/W2897083155","https://openalex.org/W3092424727","https://openalex.org/W3103168911","https://openalex.org/W3182465333","https://openalex.org/W4205825342","https://openalex.org/W4220865834","https://openalex.org/W4221086307","https://openalex.org/W4285255298","https://openalex.org/W4285286496","https://openalex.org/W4360605545","https://openalex.org/W4360606017","https://openalex.org/W4360606423","https://openalex.org/W4360831960","https://openalex.org/W4379116138","https://openalex.org/W4380881077","https://openalex.org/W4385216221","https://openalex.org/W4385216645","https://openalex.org/W4386763875","https://openalex.org/W4387042316","https://openalex.org/W4387250389","https://openalex.org/W4389166736","https://openalex.org/W4392739437","https://openalex.org/W4396918368","https://openalex.org/W4401211806","https://openalex.org/W4401568324","https://openalex.org/W4401880601","https://openalex.org/W4404954358","https://openalex.org/W4404955778","https://openalex.org/W4408174495","https://openalex.org/W4412945527"],"related_works":["https://openalex.org/W2418291489","https://openalex.org/W3096519538","https://openalex.org/W2744747300","https://openalex.org/W4241166160","https://openalex.org/W2068121105","https://openalex.org/W2384826897","https://openalex.org/W1973516247","https://openalex.org/W1997466117","https://openalex.org/W3152699334","https://openalex.org/W2390254310"],"abstract_inverted_index":{"Transformers":[0],"have":[1],"emerged":[2],"as":[3],"the":[4,29,56,70,123,177],"dominant":[5],"neural":[6],"network":[7],"architecture":[8],"thanks":[9],"to":[10,68,83,106,126],"their":[11],"versatility-":[12],"and":[13,87,122,129,162],"human-level":[14],"performance.":[15],"However,":[16],"a":[17,21,37,50,63,77,94,116,140],"transformer":[18,141],"model":[19],"contains":[20],"very":[22],"large":[23],"number":[24],"of":[25,45,58,79,173],"parameters,":[26],"which":[27,98],"makes":[28],"overhead":[30],"associated":[31],"with":[32,55],"external":[33],"memory":[34],"access":[35],"(EMA),":[36],"critical":[38],"bottleneck":[39],"in":[40,104,132,139,149],"computing":[41],"systems.":[42],"In":[43],"light":[44],"this,":[46],"we":[47,92,114],"develop":[48,62,93,115],"T-REX,":[49],"hardware":[51,130],"accelerator":[52],"for":[53],"transformers":[54],"goal":[57],"minimizing":[59],"EMA.":[60],"We":[61,74],"novel":[64],"factorizing":[65],"training":[66],"method":[67],"reduce":[69],"parameter":[71,89,108],"count":[72],"substantially.":[73],"also":[75],"combine":[76],"range":[78],"data":[80],"compression":[81],"techniques":[82],"effectively":[84],"compress":[85],"dense":[86],"sparse":[88],"matrices.":[90],"Furthermore,":[91],"dynamic":[95],"batching":[96],"technique,":[97],"can":[99],"process":[100],"multiple":[101],"short":[102],"inputs":[103],"parallel":[105],"increase":[107],"reuse.":[109],"Last":[110],"but":[111],"not":[112],"least,":[113],"two-directional":[117],"accessible":[118],"register":[119],"file":[120],"(TRF)":[121],"optimized":[124],"dataflow":[125],"improve":[127],"throughput":[128],"utilization":[131],"many":[133],"back-to-back":[134],"matrix":[135],"multiplications":[136],"(MMs)":[137],"commonly":[138],"model.":[142],"The":[143],"T-REX":[144],"test":[145],"chip":[146],"is":[147],"prototyped":[148],"Intel":[150],"16-nm":[151],"FinFET.":[152],"It":[153],"achieves":[154],"68\u2013<inline-formula":[155],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[156,164],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">":[157,165],"<tex-math":[158,166],"notation=\"LaTeX\">$567~\\mu":[159],"$</tex-math>":[160,168],"</inline-formula>s/token":[161],"0.41\u2013<inline-formula":[163],"notation=\"LaTeX\">$3.95~\\mu":[167],"</inline-formula>J/token,":[169],"demonstrating":[170],"1\u20132":[171],"orders":[172],"magnitude":[174],"improvement":[175],"over":[176],"prior":[178],"work.":[179]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2025-12-30T23:08:21.542490","created_date":"2025-10-10T00:00:00"}
