{"id":"https://openalex.org/W4409427297","doi":"https://doi.org/10.1109/jiot.2025.3560382","title":"Latency-Critical Quantized Inference With Transformer Decoders on ARM and RISC-V CPUs","display_name":"Latency-Critical Quantized Inference With Transformer Decoders on ARM and RISC-V CPUs","publication_year":2025,"publication_date":"2025-04-14","ids":{"openalex":"https://openalex.org/W4409427297","doi":"https://doi.org/10.1109/jiot.2025.3560382"},"language":"en","primary_location":{"id":"doi:10.1109/jiot.2025.3560382","is_oa":true,"landing_page_url":"https://doi.org/10.1109/jiot.2025.3560382","pdf_url":null,"source":{"id":"https://openalex.org/S2480266640","display_name":"IEEE Internet of Things Journal","issn_l":"2327-4662","issn":["2327-4662","2372-2541"],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Internet of Things Journal","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://doi.org/10.1109/jiot.2025.3560382","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5018719028","display_name":"H\u00e8ctor Mart\u00ednez","orcid":"https://orcid.org/0000-0001-5891-4479"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"H\u00e9ctor Mart\u00ednez","raw_affiliation_strings":["Electronic and Computer Engineering, Universidad de C&#x00F3;rdoba, C&#x00F3;rdoba, Spain","Universidad de C&#x00F3;rdoba, Spain"],"raw_orcid":"https://orcid.org/0000-0001-5891-4479","affiliations":[{"raw_affiliation_string":"Electronic and Computer Engineering, Universidad de C&#x00F3;rdoba, C&#x00F3;rdoba, Spain","institution_ids":[]},{"raw_affiliation_string":"Universidad de C&#x00F3;rdoba, Spain","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5005202822","display_name":"Sandra Catal\u00e1n","orcid":"https://orcid.org/0000-0002-9321-2728"},"institutions":[{"id":"https://openalex.org/I10902133","display_name":"Universitat Jaume I","ror":"https://ror.org/02ws1xc11","country_code":"ES","type":"education","lineage":["https://openalex.org/I10902133"]}],"countries":["ES"],"is_corresponding":false,"raw_author_name":"Sandra Catal\u00e1n","raw_affiliation_strings":["Computer Science and Egineering, Universitat Jaume I de Castell&#x00F3;n, Castell&#x00F3;n de la Plana, Spain","Universitat Jaume I de Castell&#x00F3;n, Spain"],"raw_orcid":"https://orcid.org/0000-0002-9321-2728","affiliations":[{"raw_affiliation_string":"Computer Science and Egineering, Universitat Jaume I de Castell&#x00F3;n, Castell&#x00F3;n de la Plana, Spain","institution_ids":["https://openalex.org/I10902133"]},{"raw_affiliation_string":"Universitat Jaume I de Castell&#x00F3;n, Spain","institution_ids":["https://openalex.org/I10902133"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5033947083","display_name":"Adri\u00e1n Castell\u00f3","orcid":"https://orcid.org/0000-0002-8576-8451"},"institutions":[{"id":"https://openalex.org/I1317635190","display_name":"Valve (United States)","ror":"https://ror.org/020esap58","country_code":"US","type":"company","lineage":["https://openalex.org/I1317635190"]},{"id":"https://openalex.org/I9617848","display_name":"Universitat Polit\u00e8cnica de Catalunya","ror":"https://ror.org/03mb6wj31","country_code":"ES","type":"education","lineage":["https://openalex.org/I9617848"]}],"countries":["ES","US"],"is_corresponding":false,"raw_author_name":"Adri\u00e1n Castell\u00f3","raw_affiliation_strings":["DISCA, Universitat Polit&#x00E8;cnica de Val&#x00E8;ncia, Val&#x00E8;ncia, Spain","Universitat Polit&#x00E8;cnica de Val&#x00E8;ncia, Spain"],"raw_orcid":"https://orcid.org/0000-0002-8576-8451","affiliations":[{"raw_affiliation_string":"DISCA, Universitat Polit&#x00E8;cnica de Val&#x00E8;ncia, Val&#x00E8;ncia, Spain","institution_ids":["https://openalex.org/I1317635190"]},{"raw_affiliation_string":"Universitat Polit&#x00E8;cnica de Val&#x00E8;ncia, Spain","institution_ids":["https://openalex.org/I9617848"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5007319120","display_name":"Jose I. Mestre","orcid":"https://orcid.org/0000-0001-5954-4131"},"institutions":[{"id":"https://openalex.org/I10902133","display_name":"Universitat Jaume I","ror":"https://ror.org/02ws1xc11","country_code":"ES","type":"education","lineage":["https://openalex.org/I10902133"]}],"countries":["ES"],"is_corresponding":false,"raw_author_name":"Jose I. Mestre","raw_affiliation_strings":["Computer Science and Egineering, Universitat Jaume I de Castell&#x00F3;n, Castell&#x00F3;n de la Plana, Spain","Universitat Jaume I de Castell&#x00F3;n, Spain"],"raw_orcid":"https://orcid.org/0000-0001-5954-4131","affiliations":[{"raw_affiliation_string":"Computer Science and Egineering, Universitat Jaume I de Castell&#x00F3;n, Castell&#x00F3;n de la Plana, Spain","institution_ids":["https://openalex.org/I10902133"]},{"raw_affiliation_string":"Universitat Jaume I de Castell&#x00F3;n, Spain","institution_ids":["https://openalex.org/I10902133"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5012806004","display_name":"Enrique S. Quintana\u2013Ort\u0301\u0131","orcid":"https://orcid.org/0000-0002-5454-165X"},"institutions":[{"id":"https://openalex.org/I1317635190","display_name":"Valve (United States)","ror":"https://ror.org/020esap58","country_code":"US","type":"company","lineage":["https://openalex.org/I1317635190"]},{"id":"https://openalex.org/I9617848","display_name":"Universitat Polit\u00e8cnica de Catalunya","ror":"https://ror.org/03mb6wj31","country_code":"ES","type":"education","lineage":["https://openalex.org/I9617848"]}],"countries":["ES","US"],"is_corresponding":false,"raw_author_name":"Enrique S. Quintana-Ort\u00ed","raw_affiliation_strings":["DISCA, Universitat Polit&#x00E8;cnica de Val&#x00E8;ncia, Val&#x00E8;ncia, Spain","Universitat Polit&#x00E8;cnica de Val&#x00E8;ncia, Spain"],"raw_orcid":"https://orcid.org/0000-0002-5454-165X","affiliations":[{"raw_affiliation_string":"DISCA, Universitat Polit&#x00E8;cnica de Val&#x00E8;ncia, Val&#x00E8;ncia, Spain","institution_ids":["https://openalex.org/I1317635190"]},{"raw_affiliation_string":"Universitat Polit&#x00E8;cnica de Val&#x00E8;ncia, Spain","institution_ids":["https://openalex.org/I9617848"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":4.188,"has_fulltext":false,"cited_by_count":3,"citation_normalized_percentile":{"value":0.93491639,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":94,"max":97},"biblio":{"volume":"12","issue":"13","first_page":"25676","last_page":"25690"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10320","display_name":"Neural Networks and Applications","score":0.9657999873161316,"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/T10320","display_name":"Neural Networks and Applications","score":0.9657999873161316,"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/T10271","display_name":"Seismic Imaging and Inversion Techniques","score":0.9485999941825867,"subfield":{"id":"https://openalex.org/subfields/1908","display_name":"Geophysics"},"field":{"id":"https://openalex.org/fields/19","display_name":"Earth and Planetary Sciences"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11609","display_name":"Geophysical Methods and Applications","score":0.9315000176429749,"subfield":{"id":"https://openalex.org/subfields/2212","display_name":"Ocean 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/computer-science","display_name":"Computer science","score":0.7530222535133362},{"id":"https://openalex.org/keywords/reduced-instruction-set-computing","display_name":"Reduced instruction set computing","score":0.5764592885971069},{"id":"https://openalex.org/keywords/latency","display_name":"Latency (audio)","score":0.5675839185714722},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.49843740463256836},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.47760745882987976},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.4589484930038452},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.45223870873451233},{"id":"https://openalex.org/keywords/instruction-set","display_name":"Instruction set","score":0.2757566571235657},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.14336541295051575},{"id":"https://openalex.org/keywords/voltage","display_name":"Voltage","score":0.13876217603683472},{"id":"https://openalex.org/keywords/telecommunications","display_name":"Telecommunications","score":0.10286650061607361},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.08963173627853394}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7530222535133362},{"id":"https://openalex.org/C126298526","wikidata":"https://www.wikidata.org/wiki/Q189376","display_name":"Reduced instruction set computing","level":3,"score":0.5764592885971069},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.5675839185714722},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.49843740463256836},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.47760745882987976},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.4589484930038452},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.45223870873451233},{"id":"https://openalex.org/C202491316","wikidata":"https://www.wikidata.org/wiki/Q272683","display_name":"Instruction set","level":2,"score":0.2757566571235657},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.14336541295051575},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.13876217603683472},{"id":"https://openalex.org/C76155785","wikidata":"https://www.wikidata.org/wiki/Q418","display_name":"Telecommunications","level":1,"score":0.10286650061607361},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.08963173627853394},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/jiot.2025.3560382","is_oa":true,"landing_page_url":"https://doi.org/10.1109/jiot.2025.3560382","pdf_url":null,"source":{"id":"https://openalex.org/S2480266640","display_name":"IEEE Internet of Things Journal","issn_l":"2327-4662","issn":["2327-4662","2372-2541"],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Internet of Things Journal","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1109/jiot.2025.3560382","is_oa":true,"landing_page_url":"https://doi.org/10.1109/jiot.2025.3560382","pdf_url":null,"source":{"id":"https://openalex.org/S2480266640","display_name":"IEEE Internet of Things Journal","issn_l":"2327-4662","issn":["2327-4662","2372-2541"],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Internet of Things Journal","raw_type":"journal-article"},"sustainable_development_goals":[{"score":0.7699999809265137,"display_name":"Affordable and clean energy","id":"https://metadata.un.org/sdg/7"}],"awards":[{"id":"https://openalex.org/G2247491333","display_name":null,"funder_award_id":"UJI-2023-04","funder_id":"https://openalex.org/F4320322926","funder_display_name":"Universitat Jaume I"},{"id":"https://openalex.org/G4695567461","display_name":null,"funder_award_id":"POSTDOC2100025","funder_id":"https://openalex.org/F4320326754","funder_display_name":"Junta de Andaluc\u00eda"},{"id":"https://openalex.org/G5712359147","display_name":null,"funder_award_id":"CIAPOS/2023/431","funder_id":"https://openalex.org/F4320321864","funder_display_name":"Generalitat Valenciana"},{"id":"https://openalex.org/G6440010706","display_name":null,"funder_award_id":"RYC2021-033973-I","funder_id":"https://openalex.org/F4320322930","funder_display_name":"Ministerio de Ciencia e Innovaci\u00f3n"},{"id":"https://openalex.org/G6443677161","display_name":null,"funder_award_id":"PID2023-146569NB-C2","funder_id":"https://openalex.org/F4320322930","funder_display_name":"Ministerio de Ciencia e Innovaci\u00f3n"}],"funders":[{"id":"https://openalex.org/F4320321864","display_name":"Generalitat Valenciana","ror":"https://ror.org/0097mvx21"},{"id":"https://openalex.org/F4320322926","display_name":"Universitat Jaume I","ror":"https://ror.org/02ws1xc11"},{"id":"https://openalex.org/F4320322930","display_name":"Ministerio de Ciencia e Innovaci\u00f3n","ror":"https://ror.org/034900433"},{"id":"https://openalex.org/F4320326754","display_name":"Junta de Andaluc\u00eda","ror":"https://ror.org/01jem9c82"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":22,"referenced_works":["https://openalex.org/W2043275593","https://openalex.org/W2073061372","https://openalex.org/W2084379367","https://openalex.org/W2252007067","https://openalex.org/W2412152731","https://openalex.org/W2516525699","https://openalex.org/W4229666556","https://openalex.org/W4308090443","https://openalex.org/W4378697133","https://openalex.org/W4385245566","https://openalex.org/W4386826409","https://openalex.org/W4390224392","https://openalex.org/W4401861582","https://openalex.org/W4405390006","https://openalex.org/W6727099177","https://openalex.org/W6727208969","https://openalex.org/W6788001715","https://openalex.org/W6790838377","https://openalex.org/W6842258392","https://openalex.org/W6847478871","https://openalex.org/W6850162387","https://openalex.org/W6851775633"],"related_works":["https://openalex.org/W2055243143","https://openalex.org/W1986418932","https://openalex.org/W2357796999","https://openalex.org/W4321636575","https://openalex.org/W3128807919","https://openalex.org/W3176411177","https://openalex.org/W4206178588","https://openalex.org/W3094491777","https://openalex.org/W3214715529","https://openalex.org/W4287635093"],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"are":[3,54,69],"transforming":[4],"industries":[5],"but":[6],"face":[7],"challenges":[8],"due":[9,73],"to":[10,50,74],"their":[11,75],"high":[12],"computational":[13,123],"and":[14,31,41,62,79,98,105,132,151],"energy":[15,149],"demands.":[16],"Model":[17],"compression":[18],"via":[19],"quantization":[20,48,144],"mitigates":[21],"these":[22],"barriers":[23],"by":[24,71],"reducing":[25],"the":[26,114,119,138,141],"bit":[27],"precision":[28,152],"of":[29,113,121,140,143],"parameters":[30],"arithmetic":[32,81],"operations,":[33],"enabling":[34],"deployment":[35],"on":[36,47,88,145],"resource-constrained":[37],"devices":[38],"like":[39],"smartphones":[40],"edge":[42,154],"platforms.":[43],"This":[44,134],"paper":[45],"focuses":[46],"applied":[49],"transformer":[51,146],"decoders,":[52],"which":[53],"critical":[55],"for":[56],"tasks":[57],"such":[58,128],"as":[59,129],"text":[60],"generation":[61],"conversational":[63],"artificial":[64],"intelligence.":[65],"Unlike":[66],"encoders,":[67],"decoders":[68],"constrained":[70],"memory":[72],"sequential":[76],"processing":[77],"nature":[78],"low":[80],"intensity.":[82],"We":[83],"propose":[84],"optimizations":[85],"targeting":[86],"inference":[87],"low-power":[89],"CPUs,":[90],"emphasizing":[91],"efficient":[92],"linear":[93],"layers":[94],"with":[95],"quantized":[96],"data/arithmetic":[97],"cache":[99],"optimization.":[100],"Using":[101],"two":[102],"representative":[103],"ARM":[104],"RISC-V":[106],"platforms,":[107],"we":[108],"present":[109],"optimized":[110],"mixed-precision":[111],"implementations":[112],"matrix":[115],"multiplication":[116],"that":[117,122],"outperform":[118],"instance":[120],"kernel":[124],"in":[125,153],"popular":[126],"libraries":[127],"BLIS,":[130],"XNNPACK":[131],"ARMCL.":[133],"work":[135],"thus":[136],"advances":[137],"understanding":[139],"impact":[142],"decoder":[147],"efficiency,":[148],"consumption":[150],"environments.":[155]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":2}],"updated_date":"2026-03-27T05:58:40.876381","created_date":"2025-10-10T00:00:00"}
