{"id":"https://openalex.org/W4414499857","doi":"https://doi.org/10.1109/aicas64808.2025.11173155","title":"Optimizing Inference Performance for Large Language Models on ARMv9 Architecture","display_name":"Optimizing Inference Performance for Large Language Models on ARMv9 Architecture","publication_year":2025,"publication_date":"2025-04-28","ids":{"openalex":"https://openalex.org/W4414499857","doi":"https://doi.org/10.1109/aicas64808.2025.11173155"},"language":"en","primary_location":{"id":"doi:10.1109/aicas64808.2025.11173155","is_oa":false,"landing_page_url":"https://doi.org/10.1109/aicas64808.2025.11173155","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE 7th International Conference on Artificial Intelligence Circuits and Systems (AICAS)","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/A5110668160","display_name":"Longhao Chen","orcid":"https://orcid.org/0009-0008-7803-8208"},"institutions":[{"id":"https://openalex.org/I50760025","display_name":"Hangzhou Dianzi University","ror":"https://ror.org/0576gt767","country_code":"CN","type":"education","lineage":["https://openalex.org/I50760025"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Longhao Chen","raw_affiliation_strings":["Hangzhou Dianzi University,Zhuoyue Honors College,Hangzhou,China"],"affiliations":[{"raw_affiliation_string":"Hangzhou Dianzi University,Zhuoyue Honors College,Hangzhou,China","institution_ids":["https://openalex.org/I50760025"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100439613","display_name":"Cheng Zhang","orcid":"https://orcid.org/0000-0002-0059-5644"},"institutions":[{"id":"https://openalex.org/I149594827","display_name":"Xidian University","ror":"https://ror.org/05s92vm98","country_code":"CN","type":"education","lineage":["https://openalex.org/I149594827"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Cheng Zhang","raw_affiliation_strings":["Xidian University,Hangzhou Institute of Technology,Hangzhou,China"],"affiliations":[{"raw_affiliation_string":"Xidian University,Hangzhou Institute of Technology,Hangzhou,China","institution_ids":["https://openalex.org/I149594827"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5001660190","display_name":"Huiyuan Zhang","orcid":"https://orcid.org/0000-0002-1532-2462"},"institutions":[{"id":"https://openalex.org/I50760025","display_name":"Hangzhou Dianzi University","ror":"https://ror.org/0576gt767","country_code":"CN","type":"education","lineage":["https://openalex.org/I50760025"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Huiyuan Zhang","raw_affiliation_strings":["Hangzhou Dianzi University,Zhuoyue Honors College,Hangzhou,China"],"affiliations":[{"raw_affiliation_string":"Hangzhou Dianzi University,Zhuoyue Honors College,Hangzhou,China","institution_ids":["https://openalex.org/I50760025"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100342218","display_name":"Chi Wang","orcid":"https://orcid.org/0000-0003-0627-2341"},"institutions":[{"id":"https://openalex.org/I50760025","display_name":"Hangzhou Dianzi University","ror":"https://ror.org/0576gt767","country_code":"CN","type":"education","lineage":["https://openalex.org/I50760025"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chi Wang","raw_affiliation_strings":["Hangzhou Dianzi University,School of Automation,Hangzhou,China"],"affiliations":[{"raw_affiliation_string":"Hangzhou Dianzi University,School of Automation,Hangzhou,China","institution_ids":["https://openalex.org/I50760025"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102397055","display_name":"Yina Zhao","orcid":null},"institutions":[{"id":"https://openalex.org/I37461747","display_name":"Wuhan University","ror":"https://ror.org/033vjfk17","country_code":"CN","type":"education","lineage":["https://openalex.org/I37461747"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yina Zhao","raw_affiliation_strings":["Wuhan University,School of Geodesy and Geomatics,Wuhan,China"],"affiliations":[{"raw_affiliation_string":"Wuhan University,School of Geodesy and Geomatics,Wuhan,China","institution_ids":["https://openalex.org/I37461747"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100667924","display_name":"Xiaoxi Li","orcid":"https://orcid.org/0000-0001-5597-7968"},"institutions":[{"id":"https://openalex.org/I149594827","display_name":"Xidian University","ror":"https://ror.org/05s92vm98","country_code":"CN","type":"education","lineage":["https://openalex.org/I149594827"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xiaoxi Li","raw_affiliation_strings":["Xidian University,Hangzhou Institute of Technology,Hangzhou,China"],"affiliations":[{"raw_affiliation_string":"Xidian University,Hangzhou Institute of Technology,Hangzhou,China","institution_ids":["https://openalex.org/I149594827"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5058519191","display_name":"Xiguang Wu","orcid":"https://orcid.org/0009-0002-8945-8247"},"institutions":[{"id":"https://openalex.org/I149594827","display_name":"Xidian University","ror":"https://ror.org/05s92vm98","country_code":"CN","type":"education","lineage":["https://openalex.org/I149594827"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xiguang Wu","raw_affiliation_strings":["Xidian University,Hangzhou Institute of Technology,Hangzhou,China"],"affiliations":[{"raw_affiliation_string":"Xidian University,Hangzhou Institute of Technology,Hangzhou,China","institution_ids":["https://openalex.org/I149594827"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5062269789","display_name":"Wei Mao","orcid":"https://orcid.org/0000-0003-2527-6778"},"institutions":[{"id":"https://openalex.org/I149594827","display_name":"Xidian University","ror":"https://ror.org/05s92vm98","country_code":"CN","type":"education","lineage":["https://openalex.org/I149594827"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wei Mao","raw_affiliation_strings":["Xidian University,Hangzhou Institute of Technology,Hangzhou,China"],"affiliations":[{"raw_affiliation_string":"Xidian University,Hangzhou Institute of Technology,Hangzhou,China","institution_ids":["https://openalex.org/I149594827"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101777261","display_name":"Genquan Han","orcid":null},"institutions":[{"id":"https://openalex.org/I149594827","display_name":"Xidian University","ror":"https://ror.org/05s92vm98","country_code":"CN","type":"education","lineage":["https://openalex.org/I149594827"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Genquan Han","raw_affiliation_strings":["Xidian University,Hangzhou Institute of Technology,Hangzhou,China"],"affiliations":[{"raw_affiliation_string":"Xidian University,Hangzhou Institute of Technology,Hangzhou,China","institution_ids":["https://openalex.org/I149594827"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":9,"corresponding_author_ids":["https://openalex.org/A5110668160"],"corresponding_institution_ids":["https://openalex.org/I50760025"],"apc_list":null,"apc_paid":null,"fwci":2.26,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.90860296,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":96,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"5"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9162999987602234,"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":0.9162999987602234,"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/inference","display_name":"Inference","score":0.7340999841690063},{"id":"https://openalex.org/keywords/rewriting","display_name":"Rewriting","score":0.5763999819755554},{"id":"https://openalex.org/keywords/software-deployment","display_name":"Software deployment","score":0.5687000155448914},{"id":"https://openalex.org/keywords/variety","display_name":"Variety (cybernetics)","score":0.42089998722076416},{"id":"https://openalex.org/keywords/decoding-methods","display_name":"Decoding methods","score":0.40220001339912415},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.3912000060081482},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.37940001487731934},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.3578000068664551}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.805400013923645},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7340999841690063},{"id":"https://openalex.org/C154690210","wikidata":"https://www.wikidata.org/wiki/Q1668499","display_name":"Rewriting","level":2,"score":0.5763999819755554},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.5687000155448914},{"id":"https://openalex.org/C136197465","wikidata":"https://www.wikidata.org/wiki/Q1729295","display_name":"Variety (cybernetics)","level":2,"score":0.42089998722076416},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.40220001339912415},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3912000060081482},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.37940001487731934},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3752000033855438},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3578000068664551},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.3474000096321106},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.34130001068115234},{"id":"https://openalex.org/C2778915421","wikidata":"https://www.wikidata.org/wiki/Q3643177","display_name":"Performance improvement","level":2,"score":0.3391000032424927},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.3301999866962433},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.31299999356269836},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.310699999332428},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3050999939441681},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.3043999969959259},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.2818000018596649},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.27810001373291016},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.273499995470047},{"id":"https://openalex.org/C176649486","wikidata":"https://www.wikidata.org/wiki/Q2308807","display_name":"Memory management","level":3,"score":0.26669999957084656},{"id":"https://openalex.org/C139571649","wikidata":"https://www.wikidata.org/wiki/Q1156793","display_name":"Program optimization","level":3,"score":0.26570001244544983},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.2624000012874603},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.25999999046325684},{"id":"https://openalex.org/C152752567","wikidata":"https://www.wikidata.org/wiki/Q116877","display_name":"Code refactoring","level":3,"score":0.25200000405311584},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.2515999972820282}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/aicas64808.2025.11173155","is_oa":false,"landing_page_url":"https://doi.org/10.1109/aicas64808.2025.11173155","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE 7th International Conference on Artificial Intelligence Circuits and Systems (AICAS)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320323230","display_name":"Xidian University","ror":"https://ror.org/05s92vm98"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":4,"referenced_works":["https://openalex.org/W2998617917","https://openalex.org/W3019166713","https://openalex.org/W4310936523","https://openalex.org/W4322766882"],"related_works":[],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"have":[4,47],"attracted":[5],"extensive":[6],"attetion":[7],"for":[8,29,63,94],"their":[9],"remarkable":[10],"performance":[11,33,85,135],"across":[12],"a":[13,155],"variety":[14],"of":[15,75,128,137],"tasks.":[16],"However,":[17],"the":[18,50,73,87,129,151,161],"considerable":[19],"computational":[20,51],"and":[21,46,52,103,118,126,144],"memory":[22,53,104],"demands":[23],"associated":[24],"with":[25,153],"LLMs":[26],"present":[27],"challenges":[28],"ensuring":[30],"inference":[31,99,134],"speed":[32],"when":[34],"deployed":[35],"in":[36,115,141,147,158,164],"resource-constrained":[37],"scenarios.":[38],"Current":[39],"researches":[40],"widely":[41],"adopt":[42],"model":[43,131],"compression":[44],"techniques":[45],"greatly":[48],"reduced":[49,107],"overhead.":[54],"Meanwhile,":[55],"some":[56],"frameworks":[57],"that":[58],"serve":[59],"as":[60],"deployment":[61,69],"tools":[62],"large":[64],"models":[65],"can":[66],"provide":[67],"optimized":[68],"strategies":[70],"based":[71],"on":[72,86,132],"characteristics":[74],"hardware":[76],"architectures.":[77],"In":[78],"this":[79,165],"work,":[80],"we":[81],"explored":[82],"LLM":[83],"optimization":[84,122],"ARMv9":[88],"computing":[89],"platform,":[90],"which":[91],"is":[92,167],"mainstream":[93],"intelligent":[95],"edge":[96],"usage.":[97],"The":[98],"efficiency":[100],"was":[101,106,139],"enhanced":[102],"usage":[105],"by":[108],"lowering":[109],"dequantization":[110],"cost,":[111],"rewriting":[112],"specific":[113],"functions":[114],"LLaMA.cpp":[116],"framework":[117],"using":[119],"advanced":[120],"compilation":[121],"techniques.":[123],"Through":[124],"optimizating":[125],"evaluating":[127],"Qwen-1.8B":[130],"ARMv9,":[133],"improvement":[136],"23.36\u00d7":[138],"achieved":[140],"decoding":[142],"stage":[143,149],"1.68":[145],"\u00d7":[146],"prefilling":[148],"over":[150],"baseline":[152],"only":[154],"1.04%":[156],"loss":[157],"accuracy.":[159],"All":[160],"developed":[162],"code":[163],"work":[166],"open":[168],"source":[169],"at":[170],"https://github.com/CEATRG/LLaMA.cpp-arm.":[171]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-04-17T18:11:37.981687","created_date":"2025-10-10T00:00:00"}