{"id":"https://openalex.org/W4393949386","doi":"https://doi.org/10.1145/3656177","title":"Understanding the Potential of FPGA-based Spatial Acceleration for Large Language Model Inference","display_name":"Understanding the Potential of FPGA-based Spatial Acceleration for Large Language Model Inference","publication_year":2024,"publication_date":"2024-04-04","ids":{"openalex":"https://openalex.org/W4393949386","doi":"https://doi.org/10.1145/3656177"},"language":"en","primary_location":{"id":"doi:10.1145/3656177","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3656177","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3656177","source":{"id":"https://openalex.org/S112809824","display_name":"ACM Transactions on Reconfigurable Technology and Systems","issn_l":"1936-7406","issn":["1936-7406","1936-7414"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319798","host_organization_name":"Association for Computing Machinery","host_organization_lineage":["https://openalex.org/P4310319798"],"host_organization_lineage_names":["Association for Computing Machinery"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ACM Transactions on Reconfigurable Technology and Systems","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"bronze","oa_url":"https://dl.acm.org/doi/pdf/10.1145/3656177","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5008694159","display_name":"Hongzheng Chen","orcid":"https://orcid.org/0000-0002-6617-0075"},"institutions":[{"id":"https://openalex.org/I205783295","display_name":"Cornell University","ror":"https://ror.org/05bnh6r87","country_code":"US","type":"education","lineage":["https://openalex.org/I205783295"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hongzheng Chen","raw_affiliation_strings":["Cornell University, Ithaca, United States","Cornell University, Ithaca, USA"],"raw_orcid":"https://orcid.org/0000-0002-6617-0075","affiliations":[{"raw_affiliation_string":"Cornell University, Ithaca, United States","institution_ids":["https://openalex.org/I205783295"]},{"raw_affiliation_string":"Cornell University, Ithaca, USA","institution_ids":["https://openalex.org/I205783295"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100445508","display_name":"Jiahao Zhang","orcid":"https://orcid.org/0009-0000-8379-7489"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jiahao Zhang","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0000-8379-7489","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039033697","display_name":"Yixiao Du","orcid":"https://orcid.org/0000-0002-6106-1283"},"institutions":[{"id":"https://openalex.org/I205783295","display_name":"Cornell University","ror":"https://ror.org/05bnh6r87","country_code":"US","type":"education","lineage":["https://openalex.org/I205783295"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yixiao Du","raw_affiliation_strings":["Cornell University, Ithaca, United States","Cornell University, Ithaca, USA"],"raw_orcid":"https://orcid.org/0000-0002-6106-1283","affiliations":[{"raw_affiliation_string":"Cornell University, Ithaca, United States","institution_ids":["https://openalex.org/I205783295"]},{"raw_affiliation_string":"Cornell University, Ithaca, USA","institution_ids":["https://openalex.org/I205783295"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5032824854","display_name":"Shaojie Xiang","orcid":"https://orcid.org/0000-0002-6901-8837"},"institutions":[{"id":"https://openalex.org/I205783295","display_name":"Cornell University","ror":"https://ror.org/05bnh6r87","country_code":"US","type":"education","lineage":["https://openalex.org/I205783295"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Shaojie Xiang","raw_affiliation_strings":["Cornell University, Ithaca, United States","Cornell University, Ithaca, USA"],"raw_orcid":"https://orcid.org/0000-0002-6901-8837","affiliations":[{"raw_affiliation_string":"Cornell University, Ithaca, United States","institution_ids":["https://openalex.org/I205783295"]},{"raw_affiliation_string":"Cornell University, Ithaca, USA","institution_ids":["https://openalex.org/I205783295"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5029785504","display_name":"Zichao Yue","orcid":"https://orcid.org/0009-0003-8585-5947"},"institutions":[{"id":"https://openalex.org/I205783295","display_name":"Cornell University","ror":"https://ror.org/05bnh6r87","country_code":"US","type":"education","lineage":["https://openalex.org/I205783295"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zichao Yue","raw_affiliation_strings":["Cornell University, Ithaca, United States","Cornell University, Ithaca, USA"],"raw_orcid":"https://orcid.org/0009-0003-8585-5947","affiliations":[{"raw_affiliation_string":"Cornell University, Ithaca, United States","institution_ids":["https://openalex.org/I205783295"]},{"raw_affiliation_string":"Cornell University, Ithaca, USA","institution_ids":["https://openalex.org/I205783295"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5059293268","display_name":"Niansong Zhang","orcid":"https://orcid.org/0000-0002-2850-0176"},"institutions":[{"id":"https://openalex.org/I205783295","display_name":"Cornell University","ror":"https://ror.org/05bnh6r87","country_code":"US","type":"education","lineage":["https://openalex.org/I205783295"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Niansong Zhang","raw_affiliation_strings":["Cornell University, Ithaca, United States","Cornell University, Ithaca, USA"],"raw_orcid":"https://orcid.org/0000-0002-2850-0176","affiliations":[{"raw_affiliation_string":"Cornell University, Ithaca, United States","institution_ids":["https://openalex.org/I205783295"]},{"raw_affiliation_string":"Cornell University, Ithaca, USA","institution_ids":["https://openalex.org/I205783295"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5050643298","display_name":"Yaohui Cai","orcid":"https://orcid.org/0000-0003-3785-3413"},"institutions":[{"id":"https://openalex.org/I205783295","display_name":"Cornell University","ror":"https://ror.org/05bnh6r87","country_code":"US","type":"education","lineage":["https://openalex.org/I205783295"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yaohui Cai","raw_affiliation_strings":["Cornell University, Ithaca, United States","Cornell University, Ithaca, USA"],"raw_orcid":"https://orcid.org/0000-0003-3785-3413","affiliations":[{"raw_affiliation_string":"Cornell University, Ithaca, United States","institution_ids":["https://openalex.org/I205783295"]},{"raw_affiliation_string":"Cornell University, Ithaca, USA","institution_ids":["https://openalex.org/I205783295"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5037210004","display_name":"Zhiru Zhang","orcid":"https://orcid.org/0000-0002-0778-0308"},"institutions":[{"id":"https://openalex.org/I205783295","display_name":"Cornell University","ror":"https://ror.org/05bnh6r87","country_code":"US","type":"education","lineage":["https://openalex.org/I205783295"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zhiru Zhang","raw_affiliation_strings":["Cornell University, Ithaca, United States","Cornell University, Ithaca, USA"],"raw_orcid":"https://orcid.org/0000-0002-0778-0308","affiliations":[{"raw_affiliation_string":"Cornell University, Ithaca, United States","institution_ids":["https://openalex.org/I205783295"]},{"raw_affiliation_string":"Cornell University, Ithaca, USA","institution_ids":["https://openalex.org/I205783295"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":8,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":18.5988,"has_fulltext":true,"cited_by_count":61,"citation_normalized_percentile":{"value":0.99452583,"is_in_top_1_percent":true,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":98,"max":100},"biblio":{"volume":"18","issue":"1","first_page":"1","last_page":"29"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9983999729156494,"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"}},"topics":[{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9983999729156494,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.9977999925613403,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.9750999808311462,"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":0.8628547787666321},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.7506406307220459},{"id":"https://openalex.org/keywords/acceleration","display_name":"Acceleration","score":0.6335927248001099},{"id":"https://openalex.org/keywords/field-programmable-gate-array","display_name":"Field-programmable gate array","score":0.618813693523407},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.39195629954338074},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.18833139538764954}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8628547787666321},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7506406307220459},{"id":"https://openalex.org/C117896860","wikidata":"https://www.wikidata.org/wiki/Q11376","display_name":"Acceleration","level":2,"score":0.6335927248001099},{"id":"https://openalex.org/C42935608","wikidata":"https://www.wikidata.org/wiki/Q190411","display_name":"Field-programmable gate array","level":2,"score":0.618813693523407},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.39195629954338074},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.18833139538764954},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0},{"id":"https://openalex.org/C74650414","wikidata":"https://www.wikidata.org/wiki/Q11397","display_name":"Classical mechanics","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3656177","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3656177","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3656177","source":{"id":"https://openalex.org/S112809824","display_name":"ACM Transactions on Reconfigurable Technology and Systems","issn_l":"1936-7406","issn":["1936-7406","1936-7414"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319798","host_organization_name":"Association for Computing Machinery","host_organization_lineage":["https://openalex.org/P4310319798"],"host_organization_lineage_names":["Association for Computing Machinery"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ACM Transactions on Reconfigurable Technology and Systems","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1145/3656177","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3656177","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3656177","source":{"id":"https://openalex.org/S112809824","display_name":"ACM Transactions on Reconfigurable Technology and Systems","issn_l":"1936-7406","issn":["1936-7406","1936-7414"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319798","host_organization_name":"Association for Computing Machinery","host_organization_lineage":["https://openalex.org/P4310319798"],"host_organization_lineage_names":["Association for Computing Machinery"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ACM Transactions on Reconfigurable Technology and Systems","raw_type":"journal-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/7","display_name":"Affordable and clean energy","score":0.8999999761581421}],"awards":[{"id":"https://openalex.org/G1452086972","display_name":null,"funder_award_id":"JUMP 2.0","funder_id":"https://openalex.org/F4320306087","funder_display_name":"Semiconductor Research Corporation"},{"id":"https://openalex.org/G199662392","display_name":null,"funder_award_id":"2007832","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G2929905519","display_name":"Collaborative Research: FMitF: Track I: DeepSmith: Scheduling with Quality Guarantees for Efficient DNN Model Execution","funder_award_id":"2019306","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G7564846446","display_name":null,"funder_award_id":"2118709","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320306087","display_name":"Semiconductor Research Corporation","ror":"https://ror.org/047z4n946"},{"id":"https://openalex.org/F4320332180","display_name":"Defense Advanced Research Projects Agency","ror":"https://ror.org/02caytj08"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4393949386.pdf","grobid_xml":"https://content.openalex.org/works/W4393949386.grobid-xml"},"referenced_works_count":87,"referenced_works":["https://openalex.org/W2047674856","https://openalex.org/W2094756095","https://openalex.org/W2322411027","https://openalex.org/W2473344385","https://openalex.org/W2560674852","https://openalex.org/W2565125333","https://openalex.org/W2767510344","https://openalex.org/W2798956872","https://openalex.org/W2891946740","https://openalex.org/W2896457183","https://openalex.org/W2899915146","https://openalex.org/W2906737788","https://openalex.org/W2912918068","https://openalex.org/W2913668833","https://openalex.org/W2913707927","https://openalex.org/W2913954081","https://openalex.org/W2963015836","https://openalex.org/W2965373594","https://openalex.org/W2969388332","https://openalex.org/W2973727699","https://openalex.org/W2980282514","https://openalex.org/W2998183051","https://openalex.org/W3047848469","https://openalex.org/W3086105743","https://openalex.org/W3112948415","https://openalex.org/W3115388607","https://openalex.org/W3117511472","https://openalex.org/W3129734321","https://openalex.org/W3129831491","https://openalex.org/W3130240120","https://openalex.org/W3130920634","https://openalex.org/W3133395503","https://openalex.org/W3161542527","https://openalex.org/W3162542754","https://openalex.org/W3176468986","https://openalex.org/W3184454880","https://openalex.org/W3199934250","https://openalex.org/W3204998121","https://openalex.org/W3206837665","https://openalex.org/W3210312974","https://openalex.org/W3211485653","https://openalex.org/W4200208024","https://openalex.org/W4205983429","https://openalex.org/W4206230517","https://openalex.org/W4206557440","https://openalex.org/W4211095909","https://openalex.org/W4211118386","https://openalex.org/W4211147898","https://openalex.org/W4213153339","https://openalex.org/W4226064176","https://openalex.org/W4230315356","https://openalex.org/W4239385313","https://openalex.org/W4241618768","https://openalex.org/W4253739894","https://openalex.org/W4254620533","https://openalex.org/W4280611847","https://openalex.org/W4281651027","https://openalex.org/W4281758439","https://openalex.org/W4285056663","https://openalex.org/W4293023328","https://openalex.org/W4293024053","https://openalex.org/W4293025835","https://openalex.org/W4299994276","https://openalex.org/W4300865759","https://openalex.org/W4308083513","https://openalex.org/W4310282800","https://openalex.org/W4311887664","https://openalex.org/W4312037452","https://openalex.org/W4312060029","https://openalex.org/W4319166707","https://openalex.org/W4321636575","https://openalex.org/W4321637273","https://openalex.org/W4322718191","https://openalex.org/W4322718253","https://openalex.org/W4362598949","https://openalex.org/W4379260375","https://openalex.org/W4383749446","https://openalex.org/W4384918448","https://openalex.org/W4385326807","https://openalex.org/W4387321091","https://openalex.org/W4388093177","https://openalex.org/W4389162879","https://openalex.org/W4394998694","https://openalex.org/W6798182279","https://openalex.org/W6800751262","https://openalex.org/W6838461927","https://openalex.org/W6849805803"],"related_works":["https://openalex.org/W2748952813","https://openalex.org/W2111241003","https://openalex.org/W2390279801","https://openalex.org/W4200391368","https://openalex.org/W2358668433","https://openalex.org/W2355315220","https://openalex.org/W2210979487","https://openalex.org/W2376932109","https://openalex.org/W2074043759","https://openalex.org/W2001405890"],"abstract_inverted_index":{"Recent":[0],"advancements":[1],"in":[2,19,57,174,290,303,312],"large":[3],"language":[4],"models":[5,27],"(LLMs)":[6],"boasting":[7],"billions":[8],"of":[9,34,74,91,124,189,200,222],"parameters":[10],"have":[11,28,231],"generated":[12],"a":[13,106,116,125,198,279,296,300],"significant":[14],"demand":[15],"for":[16,25,45,78,95,120,150,166,269,285],"efficient":[17],"deployment":[18],"inference":[20,80],"workloads.":[21],"While":[22],"hardware":[23,43,93],"accelerators":[24,268],"Transformer-based":[26],"been":[29],"extensively":[30],"studied,":[31],"the":[32,70,89,122,132,159,167,172,220,270,291,308,313],"majority":[33],"existing":[35],"approaches":[36],"rely":[37],"on":[38,81,139,193,244],"temporal":[39],"architectures":[40],"that":[41,205],"reuse":[42],"units":[44,94],"different":[46],"network":[47],"layers":[48],"and":[49,72,135,163,208,227,239,299],"operators.":[50],"However,":[51],"these":[52],"methods":[53],"often":[54],"encounter":[55],"challenges":[56],"achieving":[58,295],"low":[59],"latency":[60],"due":[61],"to":[62,147,260,265,283,307],"considerable":[63],"memory":[64,112,136],"access":[65],"overhead.":[66],"This":[67,142,210],"article":[68],"investigates":[69],"feasibility":[71],"potential":[73],"model-specific":[75],"spatial":[76,126,177],"acceleration":[77,178],"LLM":[79,127,191],"field-programmable":[82],"gate":[83],"arrays":[84],"(FPGAs).":[85],"Our":[86],"approach":[87,256],"involves":[88],"specialization":[90],"distinct":[92],"specific":[96],"operators":[97],"or":[98],"layers,":[99],"facilitating":[100],"direct":[101],"communication":[102],"between":[103],"them":[104],"through":[105],"dataflow":[107],"architecture":[108],"while":[109,294],"minimizing":[110],"off-chip":[111],"accesses.":[113],"We":[114],"introduce":[115],"comprehensive":[117],"analytical":[118,225],"model":[119,143,192,226],"estimating":[121],"performance":[123],"accelerator,":[128],"taking":[129],"into":[130],"account":[131],"on-chip":[133],"compute":[134],"resources":[137],"available":[138,215],"an":[140,190,245,287],"FPGA.":[141],"can":[144,157,179,257],"be":[145,213],"extended":[146],"multi-FPGA":[148],"settings":[149],"distributed":[151],"inference.":[152],"Through":[153],"our":[154,224,255],"analysis,":[155],"we":[156,195,230,277],"identify":[158],"most":[160],"effective":[161],"parallelization":[162],"buffering":[164],"schemes":[165],"accelerator":[168],"and,":[169],"crucially,":[170],"determine":[171],"scenarios":[173],"which":[175],"FPGA-based":[176,267],"outperform":[180],"its":[181],"GPU-based":[182],"counterpart.":[183],"To":[184,218],"enable":[185],"more":[186],"productive":[187],"implementations":[188],"FPGAs,":[194],"further":[196],"provide":[197],"library":[199,211],"high-level":[201],"synthesis":[202],"(HLS)":[203],"kernels":[204],"are":[206],"composable":[207],"reusable.":[209],"will":[212],"made":[214],"as":[216],"open-source.":[217],"validate":[219],"effectiveness":[221],"both":[223],"HLS":[228],"library,":[229],"implemented":[232],"Bidirectional":[233],"Encoder":[234],"Representations":[235],"from":[236],"Transformers":[237,242],"(BERT)":[238],"Generative":[240],"Pre-trained":[241],"(GPT2)":[243],"AMD":[246],"Xilinx":[247],"Alveo":[248],"U280":[249],"FPGA":[250,288],"device.":[251],"Experimental":[252],"results":[253],"demonstrate":[254],"achieve":[258],"up":[259],"13.4\u00d7":[261],"speedup":[262,281,298],"when":[263],"compared":[264,282,306],"previous":[266],"BERT":[271],"model.":[272],"For":[273],"GPT":[274],"generative":[275],"inference,":[276],"attain":[278],"2.2\u00d7":[280],"Design":[284],"Excellence,":[286],"overlay,":[289],"prefill":[292],"stage,":[293],"1.9\u00d7":[297],"5.7\u00d7":[301],"improvement":[302],"energy":[304],"efficiency":[305],"NVIDIA":[309],"A100":[310],"GPU":[311],"decode":[314],"stage.":[315]},"counts_by_year":[{"year":2026,"cited_by_count":12},{"year":2025,"cited_by_count":44},{"year":2024,"cited_by_count":5}],"updated_date":"2026-06-12T08:23:45.883708","created_date":"2024-04-05T00:00:00"}