{"id":"https://openalex.org/W4381233128","doi":"https://doi.org/10.1109/tvlsi.2023.3282046","title":"X-Former: In-Memory Acceleration of Transformers","display_name":"X-Former: In-Memory Acceleration of Transformers","publication_year":2023,"publication_date":"2023-06-19","ids":{"openalex":"https://openalex.org/W4381233128","doi":"https://doi.org/10.1109/tvlsi.2023.3282046"},"language":"en","primary_location":{"id":"doi:10.1109/tvlsi.2023.3282046","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tvlsi.2023.3282046","pdf_url":null,"source":{"id":"https://openalex.org/S37538908","display_name":"IEEE Transactions on Very Large Scale Integration (VLSI) Systems","issn_l":"1063-8210","issn":["1063-8210","1557-9999"],"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 Transactions on Very Large Scale Integration (VLSI) Systems","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/A5071950007","display_name":"Shrihari Sridharan","orcid":"https://orcid.org/0000-0002-4259-5263"},"institutions":[{"id":"https://openalex.org/I219193219","display_name":"Purdue University West Lafayette","ror":"https://ror.org/02dqehb95","country_code":"US","type":"education","lineage":["https://openalex.org/I219193219"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Shrihari Sridharan","raw_affiliation_strings":["School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USA"],"raw_orcid":"https://orcid.org/0000-0002-4259-5263","affiliations":[{"raw_affiliation_string":"School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USA","institution_ids":["https://openalex.org/I219193219"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5044634700","display_name":"Jacob R. Stevens","orcid":"https://orcid.org/0000-0002-7626-2846"},"institutions":[{"id":"https://openalex.org/I219193219","display_name":"Purdue University West Lafayette","ror":"https://ror.org/02dqehb95","country_code":"US","type":"education","lineage":["https://openalex.org/I219193219"]},{"id":"https://openalex.org/I4210114444","display_name":"Meta (United States)","ror":"https://ror.org/01zbnvs85","country_code":"US","type":"company","lineage":["https://openalex.org/I4210114444"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jacob R. Stevens","raw_affiliation_strings":["School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USA","Meta Platforms, Inc., Cambridge, MA, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USA","institution_ids":["https://openalex.org/I219193219"]},{"raw_affiliation_string":"Meta Platforms, Inc., Cambridge, MA, USA","institution_ids":["https://openalex.org/I4210114444"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5031161187","display_name":"Kaushik Roy","orcid":"https://orcid.org/0000-0002-0735-9695"},"institutions":[{"id":"https://openalex.org/I219193219","display_name":"Purdue University West Lafayette","ror":"https://ror.org/02dqehb95","country_code":"US","type":"education","lineage":["https://openalex.org/I219193219"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Kaushik Roy","raw_affiliation_strings":["School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USA"],"raw_orcid":"https://orcid.org/0000-0002-0735-9695","affiliations":[{"raw_affiliation_string":"School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USA","institution_ids":["https://openalex.org/I219193219"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5065766721","display_name":"Anand Raghunathan","orcid":"https://orcid.org/0000-0002-4624-564X"},"institutions":[{"id":"https://openalex.org/I219193219","display_name":"Purdue University West Lafayette","ror":"https://ror.org/02dqehb95","country_code":"US","type":"education","lineage":["https://openalex.org/I219193219"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Anand Raghunathan","raw_affiliation_strings":["School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USA","institution_ids":["https://openalex.org/I219193219"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":7.3079,"has_fulltext":false,"cited_by_count":60,"citation_normalized_percentile":{"value":0.9785456,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":90,"max":100},"biblio":{"volume":"31","issue":"8","first_page":"1223","last_page":"1233"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12808","display_name":"Ferroelectric and Negative Capacitance Devices","score":0.9997000098228455,"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"}},"topics":[{"id":"https://openalex.org/T12808","display_name":"Ferroelectric and Negative Capacitance Devices","score":0.9997000098228455,"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"}},{"id":"https://openalex.org/T10502","display_name":"Advanced Memory and Neural Computing","score":0.9987000226974487,"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"}},{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9976999759674072,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7894012928009033},{"id":"https://openalex.org/keywords/interleaved-memory","display_name":"Interleaved memory","score":0.4910762310028076},{"id":"https://openalex.org/keywords/computer-hardware","display_name":"Computer hardware","score":0.469635009765625},{"id":"https://openalex.org/keywords/operand","display_name":"Operand","score":0.4614068567752838},{"id":"https://openalex.org/keywords/dram","display_name":"Dram","score":0.4610980749130249},{"id":"https://openalex.org/keywords/massively-parallel","display_name":"Massively parallel","score":0.4525066316127777},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.45122307538986206},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.42042669653892517},{"id":"https://openalex.org/keywords/auxiliary-memory","display_name":"Auxiliary memory","score":0.4203472137451172},{"id":"https://openalex.org/keywords/content-addressable-memory","display_name":"Content-addressable memory","score":0.4104061722755432},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.4092920422554016},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.39895424246788025},{"id":"https://openalex.org/keywords/semiconductor-memory","display_name":"Semiconductor memory","score":0.28007185459136963},{"id":"https://openalex.org/keywords/computer-memory","display_name":"Computer memory","score":0.25081223249435425},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.16425171494483948},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.11216968297958374},{"id":"https://openalex.org/keywords/voltage","display_name":"Voltage","score":0.09485003352165222}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7894012928009033},{"id":"https://openalex.org/C63511323","wikidata":"https://www.wikidata.org/wiki/Q908936","display_name":"Interleaved memory","level":4,"score":0.4910762310028076},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.469635009765625},{"id":"https://openalex.org/C55526617","wikidata":"https://www.wikidata.org/wiki/Q719375","display_name":"Operand","level":2,"score":0.4614068567752838},{"id":"https://openalex.org/C7366592","wikidata":"https://www.wikidata.org/wiki/Q1255620","display_name":"Dram","level":2,"score":0.4610980749130249},{"id":"https://openalex.org/C190475519","wikidata":"https://www.wikidata.org/wiki/Q544384","display_name":"Massively parallel","level":2,"score":0.4525066316127777},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.45122307538986206},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.42042669653892517},{"id":"https://openalex.org/C82687282","wikidata":"https://www.wikidata.org/wiki/Q66221","display_name":"Auxiliary memory","level":2,"score":0.4203472137451172},{"id":"https://openalex.org/C53442348","wikidata":"https://www.wikidata.org/wiki/Q745101","display_name":"Content-addressable memory","level":3,"score":0.4104061722755432},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.4092920422554016},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.39895424246788025},{"id":"https://openalex.org/C98986596","wikidata":"https://www.wikidata.org/wiki/Q1143031","display_name":"Semiconductor memory","level":2,"score":0.28007185459136963},{"id":"https://openalex.org/C92855701","wikidata":"https://www.wikidata.org/wiki/Q5830907","display_name":"Computer memory","level":3,"score":0.25081223249435425},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.16425171494483948},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.11216968297958374},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.09485003352165222},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/tvlsi.2023.3282046","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tvlsi.2023.3282046","pdf_url":null,"source":{"id":"https://openalex.org/S37538908","display_name":"IEEE Transactions on Very Large Scale Integration (VLSI) Systems","issn_l":"1063-8210","issn":["1063-8210","1557-9999"],"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 Transactions on Very Large Scale Integration (VLSI) Systems","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[{"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4","score":0.5}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":43,"referenced_works":["https://openalex.org/W2033811947","https://openalex.org/W2105102111","https://openalex.org/W2442974303","https://openalex.org/W2518281301","https://openalex.org/W2545497739","https://openalex.org/W2606722458","https://openalex.org/W2613989746","https://openalex.org/W2626991402","https://openalex.org/W2896457183","https://openalex.org/W2913104037","https://openalex.org/W2923014074","https://openalex.org/W2963748441","https://openalex.org/W2969812992","https://openalex.org/W2979826702","https://openalex.org/W2983353765","https://openalex.org/W3003074159","https://openalex.org/W3017024317","https://openalex.org/W3018945530","https://openalex.org/W3031771821","https://openalex.org/W3035083896","https://openalex.org/W3037132819","https://openalex.org/W3042468132","https://openalex.org/W3091783050","https://openalex.org/W3100985894","https://openalex.org/W3111375540","https://openalex.org/W3141540147","https://openalex.org/W3159727696","https://openalex.org/W3165929947","https://openalex.org/W3176381399","https://openalex.org/W3185889905","https://openalex.org/W3200825586","https://openalex.org/W3207622241","https://openalex.org/W3211525823","https://openalex.org/W4287667694","https://openalex.org/W4292779060","https://openalex.org/W4385245566","https://openalex.org/W6739901393","https://openalex.org/W6755207826","https://openalex.org/W6773035625","https://openalex.org/W6778883912","https://openalex.org/W6780125659","https://openalex.org/W6783267081","https://openalex.org/W6783744057"],"related_works":["https://openalex.org/W2181404029","https://openalex.org/W2068291971","https://openalex.org/W1527643973","https://openalex.org/W4321517554","https://openalex.org/W2044064773","https://openalex.org/W2062140197","https://openalex.org/W1579430994","https://openalex.org/W4232365528","https://openalex.org/W2353759903","https://openalex.org/W1978318796"],"abstract_inverted_index":{"Transformers":[0,83,126,164],"have":[1],"achieved":[2],"great":[3],"success":[4],"in":[5,32,81,125,255,285],"a":[6,33,51,149,183,212,260,290],"wide":[7],"variety":[8],"of":[9,44,46,54,172,190,207,220],"natural":[10],"language":[11],"processing":[12,71,76,82,195,223],"(NLP)":[13],"tasks":[14],"due":[15],"to":[16,29,93,98,197,238,268],"the":[17,141,169,204,218,221],"self-attention":[18],"mechanism,":[19],"which":[20,121,216],"assigns":[21],"an":[22,95],"importance":[23],"score":[24,119],"for":[25,145,163],"every":[26],"word":[27],"relative":[28],"other":[30],"words":[31],"sequence.":[34],"However,":[35,117],"these":[36,178],"models":[37],"are":[38,122],"very":[39],"large,":[40],"often":[41],"reaching":[42],"hundreds":[43],"billions":[45],"parameters,":[47],"and":[48,74,132,158,165,193,225,246,257,266,276,287],"therefore":[49],"require":[50,137],"large":[52],"number":[53],"dynamic":[55],"random":[56],"access":[57],"memory":[58,90,115],"(DRAM)":[59],"accesses.":[60],"Hence,":[61],"traditional":[62],"deep":[63],"neural":[64,129,134],"network":[65,135],"(DNN)":[66],"accelerators":[67,86,153],"such":[68],"as":[69],"graphical":[70],"units":[72,77],"(GPUs)":[73],"tensor":[75],"(TPUs)":[78],"face":[79],"limitations":[80],"efficiently.":[84,201],"In-memory":[85],"based":[87],"on":[88],"nonvolatile":[89],"(NVM)":[91],"promise":[92],"be":[94],"effective":[96],"solution":[97],"this":[99],"challenge,":[100],"since":[101],"they":[102],"provide":[103],"high":[104,155],"storage":[105],"density":[106],"while":[107],"performing":[108],"massively":[109],"parallel":[110],"matrix\u2013vector":[111],"multiplications":[112],"(MVMs)":[113],"within":[114],"arrays.":[116],"attention":[118],"computations,":[120],"frequently":[123],"used":[124,162],"unlike":[127],"convolutional":[128],"networks":[130],"(CNNs)":[131],"recurrent":[133],"(RNNs),":[136],"MVMs":[138],"where":[139],"both":[140,191],"operands":[142],"change":[143],"dynamically":[144],"each":[146],"input.":[147],"As":[148],"result,":[150],"conventional":[151],"NVM-based":[152],"incur":[154],"write":[156,159],"latency":[157,256,286],"energy":[160,258,288],"when":[161],"further":[166],"suffer":[167],"from":[168],"low":[170],"endurance":[171],"most":[173],"NVM":[174,192,293],"technologies.":[175],"To":[176,202],"address":[177],"challenges,":[179],"we":[180,209,232],"present":[181],"X-Former,":[182,208],"hybrid":[184],"in-memory":[185,292],"hardware":[186,205],"accelerator":[187],"that":[188,234],"consists":[189],"CMOS":[194],"elements":[196,224],"execute":[198],"transformer":[199],"workloads":[200],"improve":[203],"utilization":[206],"also":[210],"propose":[211],"sequence":[213],"blocking":[214],"dataflow,":[215],"overlaps":[217],"computations":[219],"two":[222],"reduces":[226],"execution":[227],"time.":[228],"Across":[229],"several":[230],"benchmarks,":[231],"show":[233],"X-Former":[235],"achieves":[236],"up":[237,267],"<inline-formula":[239,247,269,277],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[240,248,270,278],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">":[241,249,271,279],"<tex-math":[242,250,272,280],"notation=\"LaTeX\">$69.8\\times":[243],"$":[244,252,274,282],"</tex-math></inline-formula>":[245,253,275,283],"notation=\"LaTeX\">$13\\times":[251],"improvements":[254,284],"over":[259,289],"NVIDIA":[261],"GeForce":[262],"GTX":[263],"1060":[264],"GPU":[265],"notation=\"LaTeX\">$24.1\\times":[273],"notation=\"LaTeX\">$7.95\\times":[281],"state-of-the-art":[291],"accelerator.":[294]},"counts_by_year":[{"year":2026,"cited_by_count":10},{"year":2025,"cited_by_count":36},{"year":2024,"cited_by_count":13},{"year":2023,"cited_by_count":1}],"updated_date":"2026-06-19T17:40:00.097472","created_date":"2025-10-10T00:00:00"}