{"id":"https://openalex.org/W4408182386","doi":"https://doi.org/10.1109/tcsi.2025.3546256","title":"EdgeLLM: A Highly Efficient CPU-FPGA Heterogeneous Edge Accelerator for Large Language Models","display_name":"EdgeLLM: A Highly Efficient CPU-FPGA Heterogeneous Edge Accelerator for Large Language Models","publication_year":2025,"publication_date":"2025-03-06","ids":{"openalex":"https://openalex.org/W4408182386","doi":"https://doi.org/10.1109/tcsi.2025.3546256"},"language":"en","primary_location":{"id":"doi:10.1109/tcsi.2025.3546256","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tcsi.2025.3546256","pdf_url":null,"source":{"id":"https://openalex.org/S116977442","display_name":"IEEE Transactions on Circuits and Systems I Regular Papers","issn_l":"1549-8328","issn":["1549-8328","1558-0806"],"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 Circuits and Systems I: Regular Papers","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/A5038495391","display_name":"Mingqiang Huang","orcid":"https://orcid.org/0000-0002-7794-3985"},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"id":"https://openalex.org/I4210145761","display_name":"Shenzhen Institutes of Advanced Technology","ror":"https://ror.org/04gh4er46","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210145761"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Mingqiang Huang","raw_affiliation_strings":["Chinese Academy of Sciences, Shenzhen Institute of Advanced Technology, Shenzhen, China","Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China"],"raw_orcid":"https://orcid.org/0000-0002-7794-3985","affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences, Shenzhen Institute of Advanced Technology, Shenzhen, China","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210145761"]},{"raw_affiliation_string":"Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210145761"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5055425079","display_name":"Ao Shen","orcid":"https://orcid.org/0000-0003-4722-2367"},"institutions":[{"id":"https://openalex.org/I3045169105","display_name":"Southern University of Science and Technology","ror":"https://ror.org/049tv2d57","country_code":"CN","type":"education","lineage":["https://openalex.org/I3045169105"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ao Shen","raw_affiliation_strings":["School of Microelectronics, Southern University of Science and Technology, Shenzhen, China"],"raw_orcid":"https://orcid.org/0000-0003-4722-2367","affiliations":[{"raw_affiliation_string":"School of Microelectronics, Southern University of Science and Technology, Shenzhen, China","institution_ids":["https://openalex.org/I3045169105"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5090865912","display_name":"Kai Li","orcid":"https://orcid.org/0000-0003-3251-931X"},"institutions":[{"id":"https://openalex.org/I3045169105","display_name":"Southern University of Science and Technology","ror":"https://ror.org/049tv2d57","country_code":"CN","type":"education","lineage":["https://openalex.org/I3045169105"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Kai Li","raw_affiliation_strings":["School of Microelectronics, Southern University of Science and Technology, Shenzhen, China"],"raw_orcid":"https://orcid.org/0000-0003-3251-931X","affiliations":[{"raw_affiliation_string":"School of Microelectronics, Southern University of Science and Technology, Shenzhen, China","institution_ids":["https://openalex.org/I3045169105"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109758412","display_name":"Haoxiang Peng","orcid":null},"institutions":[{"id":"https://openalex.org/I3045169105","display_name":"Southern University of Science and Technology","ror":"https://ror.org/049tv2d57","country_code":"CN","type":"education","lineage":["https://openalex.org/I3045169105"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Haoxiang Peng","raw_affiliation_strings":["School of Microelectronics, Southern University of Science and Technology, Shenzhen, China"],"raw_orcid":"https://orcid.org/0009-0007-8154-8463","affiliations":[{"raw_affiliation_string":"School of Microelectronics, Southern University of Science and Technology, Shenzhen, China","institution_ids":["https://openalex.org/I3045169105"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100772375","display_name":"Boyu Li","orcid":"https://orcid.org/0000-0001-9709-9673"},"institutions":[{"id":"https://openalex.org/I3045169105","display_name":"Southern University of Science and Technology","ror":"https://ror.org/049tv2d57","country_code":"CN","type":"education","lineage":["https://openalex.org/I3045169105"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Boyu Li","raw_affiliation_strings":["School of Microelectronics, Southern University of Science and Technology, Shenzhen, China"],"raw_orcid":"https://orcid.org/0000-0001-9709-9673","affiliations":[{"raw_affiliation_string":"School of Microelectronics, Southern University of Science and Technology, Shenzhen, China","institution_ids":["https://openalex.org/I3045169105"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101309752","display_name":"Yupeng Su","orcid":"https://orcid.org/0009-0007-0575-4038"},"institutions":[{"id":"https://openalex.org/I3045169105","display_name":"Southern University of Science and Technology","ror":"https://ror.org/049tv2d57","country_code":"CN","type":"education","lineage":["https://openalex.org/I3045169105"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yupeng Su","raw_affiliation_strings":["School of Microelectronics, Southern University of Science and Technology, Shenzhen, China"],"raw_orcid":"https://orcid.org/0009-0007-0575-4038","affiliations":[{"raw_affiliation_string":"School of Microelectronics, Southern University of Science and Technology, Shenzhen, China","institution_ids":["https://openalex.org/I3045169105"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5034853402","display_name":"Hao Yu","orcid":"https://orcid.org/0000-0002-2674-4118"},"institutions":[{"id":"https://openalex.org/I3045169105","display_name":"Southern University of Science and Technology","ror":"https://ror.org/049tv2d57","country_code":"CN","type":"education","lineage":["https://openalex.org/I3045169105"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hao Yu","raw_affiliation_strings":["School of Microelectronics, Southern University of Science and Technology, Shenzhen, China"],"raw_orcid":"https://orcid.org/0000-0002-2674-4118","affiliations":[{"raw_affiliation_string":"School of Microelectronics, Southern University of Science and Technology, Shenzhen, China","institution_ids":["https://openalex.org/I3045169105"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":51.187,"has_fulltext":false,"cited_by_count":50,"citation_normalized_percentile":{"value":0.9997791,"is_in_top_1_percent":true,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":99,"max":100},"biblio":{"volume":"72","issue":"7","first_page":"3352","last_page":"3365"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.829200029373169,"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":0.829200029373169,"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/T10320","display_name":"Neural Networks and Applications","score":0.7871999740600586,"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/T10028","display_name":"Topic Modeling","score":0.7093999981880188,"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/field-programmable-gate-array","display_name":"Field-programmable gate array","score":0.6913948059082031},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6553599238395691},{"id":"https://openalex.org/keywords/enhanced-data-rates-for-gsm-evolution","display_name":"Enhanced Data Rates for GSM Evolution","score":0.5205137133598328},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.41660889983177185},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.368672251701355},{"id":"https://openalex.org/keywords/computer-architecture","display_name":"Computer architecture","score":0.32897645235061646},{"id":"https://openalex.org/keywords/telecommunications","display_name":"Telecommunications","score":0.16796448826789856}],"concepts":[{"id":"https://openalex.org/C42935608","wikidata":"https://www.wikidata.org/wiki/Q190411","display_name":"Field-programmable gate array","level":2,"score":0.6913948059082031},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6553599238395691},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.5205137133598328},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.41660889983177185},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.368672251701355},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.32897645235061646},{"id":"https://openalex.org/C76155785","wikidata":"https://www.wikidata.org/wiki/Q418","display_name":"Telecommunications","level":1,"score":0.16796448826789856}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/tcsi.2025.3546256","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tcsi.2025.3546256","pdf_url":null,"source":{"id":"https://openalex.org/S116977442","display_name":"IEEE Transactions on Circuits and Systems I Regular Papers","issn_l":"1549-8328","issn":["1549-8328","1558-0806"],"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 Circuits and Systems I: Regular Papers","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G1286478985","display_name":null,"funder_award_id":"JCYJ20200109115210307","funder_id":"https://openalex.org/F4320336740","funder_display_name":"Chongqing Municipality Key Research and Development Program of China"},{"id":"https://openalex.org/G1335581868","display_name":null,"funder_award_id":"2022ZD0210600","funder_id":"https://openalex.org/F4320336612","funder_display_name":"Science and Technology Innovation Foundation of Harbin"},{"id":"https://openalex.org/G2781527179","display_name":null,"funder_award_id":"62034007","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G3688050962","display_name":null,"funder_award_id":"2023B1515020051","funder_id":"https://openalex.org/F4320321921","funder_display_name":"Natural Science Foundation of Guangdong Province"},{"id":"https://openalex.org/G662046889","display_name":null,"funder_award_id":"92464102","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320321921","display_name":"Natural Science Foundation of Guangdong Province","ror":null},{"id":"https://openalex.org/F4320336612","display_name":"Science and Technology Innovation Foundation of Harbin","ror":null},{"id":"https://openalex.org/F4320336740","display_name":"Chongqing Municipality Key Research and Development Program of China","ror":null}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":32,"referenced_works":["https://openalex.org/W1861492603","https://openalex.org/W2094756095","https://openalex.org/W2108598243","https://openalex.org/W2194775991","https://openalex.org/W2257979135","https://openalex.org/W2606722458","https://openalex.org/W3047848469","https://openalex.org/W3162542754","https://openalex.org/W3185702163","https://openalex.org/W4285250061","https://openalex.org/W4308479898","https://openalex.org/W4323655724","https://openalex.org/W4380874786","https://openalex.org/W4384561707","https://openalex.org/W4385245566","https://openalex.org/W4387617694","https://openalex.org/W4387789586","https://openalex.org/W4393578753","https://openalex.org/W6637373629","https://openalex.org/W6755207826","https://openalex.org/W6769627184","https://openalex.org/W6784577980","https://openalex.org/W6800751262","https://openalex.org/W6810081322","https://openalex.org/W6839366308","https://openalex.org/W6848451824","https://openalex.org/W6852408377","https://openalex.org/W6852800892","https://openalex.org/W6857551316","https://openalex.org/W6869164663","https://openalex.org/W6872893770","https://openalex.org/W6877022392"],"related_works":["https://openalex.org/W2111241003","https://openalex.org/W2355315220","https://openalex.org/W4200391368","https://openalex.org/W2096844293","https://openalex.org/W2363944576","https://openalex.org/W2351041855","https://openalex.org/W2570254841","https://openalex.org/W1967938402","https://openalex.org/W2386041993","https://openalex.org/W1608572506"],"abstract_inverted_index":{"The":[0,185],"rapid":[1],"advancements":[2],"in":[3,52,213],"artificial":[4],"intelligence":[5],"(AI),":[6],"particularly":[7],"the":[8,40,62,68,89,98,117,122,129,146,152,173,195],"Large":[9],"Language":[10],"Models":[11],"(LLMs),":[12],"have":[13],"profoundly":[14],"affected":[15],"our":[16],"daily":[17],"work":[18],"and":[19,50,94,124,135,148,221],"communication":[20],"forms.":[21],"However,":[22],"it":[23,208],"is":[24],"still":[25],"a":[26,137],"challenge":[27],"to":[28,39,60,114,120,157,171],"deploy":[29],"LLMs":[30],"on":[31,67,106,177],"resource-constrained":[32],"edge":[33],"devices":[34],"(such":[35],"as":[36],"robots),":[37],"due":[38],"intensive":[41],"computation":[42],"requirements,":[43],"heavy":[44],"memory":[45],"access,":[46],"diverse":[47],"operator":[48],"types":[49],"difficulties":[51],"compilation.":[53],"In":[54],"this":[55],"work,":[56],"we":[57,70,127,166],"proposed":[58,167],"EdgeLLM":[59],"address":[61,121],"above":[63],"issues.":[64],"Firstly,":[65],"focusing":[66],"computation,":[69],"designed":[71],"mix-precision":[72],"processing":[73],"element":[74],"array":[75],"together":[76],"with":[77,202],"group":[78],"systolic":[79],"architecture,":[80],"that":[81],"can":[82],"efficiently":[83],"support":[84],"both":[85],"FP$16\\ast":[86,95],"$FP16":[87],"for":[88,97],"MHA":[90],"block":[91],"(Multi-Head":[92],"Attention)":[93],"$INT4":[96],"FFN":[99],"layer":[100],"(Feed-Forward":[101],"Network).":[102],"Meanwhile":[103],"specific":[104],"optimization":[105],"log-scale":[107],"structured":[108],"weight":[109],"sparsity,":[110],"has":[111],"been":[112],"used":[113],"further":[115],"increase":[116],"efficiency.":[118],"Secondly,":[119],"compilation":[123],"deployment":[125],"issue,":[126],"analyzed":[128],"whole":[130,174],"operators":[131,160],"within":[132],"LLM":[133,175,222],"models":[134],"developed":[136],"universal":[138],"data":[139,154,163],"parallelism":[140],"scheme,":[141],"by":[142],"which":[143],"all":[144],"of":[145,206,215],"input":[147],"output":[149],"features":[150],"maintain":[151],"same":[153],"shape,":[155],"enabling":[156],"process":[158],"different":[159],"without":[161],"any":[162],"rearrangement.":[164],"Then":[165],"an":[168],"end-to-end":[169],"compiler":[170],"map":[172],"model":[176],"CPU-FPGA":[178],"heterogeneous":[179],"system":[180],"(AMD":[181],"Xilinx":[182],"VCU128":[183],"FPGA).":[184],"accelerator":[186,205],"achieves$1.91\\times":[187],"$higher":[188,191],"throughput":[189],"and$7.55\\times":[190],"energy":[192,219],"efficiency":[193,220],"than":[194],"commercial":[196],"GPU":[197],"(NVIDIA":[198],"A100-SXM4-80G).":[199],"When":[200],"compared":[201],"state-of-the-art":[203],"FPGA":[204],"FlightLLM,":[207],"shows":[209],"10-24%":[210],"better":[211],"performance":[212],"terms":[214],"HBM":[216],"bandwidth":[217],"utilization,":[218],"throughput.":[223]},"counts_by_year":[{"year":2026,"cited_by_count":27},{"year":2025,"cited_by_count":23}],"updated_date":"2026-07-31T08:31:51.225901","created_date":"2025-10-10T00:00:00"}
