{"id":"https://openalex.org/W3048263222","doi":"https://doi.org/10.1145/3404397.3404407","title":"Extremely Low-bit Convolution Optimization for Quantized Neural Network on Modern Computer Architectures","display_name":"Extremely Low-bit Convolution Optimization for Quantized Neural Network on Modern Computer Architectures","publication_year":2020,"publication_date":"2020-08-09","ids":{"openalex":"https://openalex.org/W3048263222","doi":"https://doi.org/10.1145/3404397.3404407","mag":"3048263222"},"language":"en","primary_location":{"id":"doi:10.1145/3404397.3404407","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3404397.3404407","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"49th International Conference on Parallel Processing - ICPP","raw_type":"proceedings-article"},"type":"conference-paper","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/A5088518659","display_name":"Qingchang Han","orcid":null},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Qingchang Han","raw_affiliation_strings":["Beihang University SenseTime Research, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beihang University SenseTime Research, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5011523353","display_name":"Yongmin Hu","orcid":null},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yongmin Hu","raw_affiliation_strings":["Beihang University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beihang University","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102785611","display_name":"Fengwei Yu","orcid":"https://orcid.org/0000-0003-0352-8700"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fengwei Yu","raw_affiliation_strings":["SenseTime Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SenseTime Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5018705589","display_name":"Hailong Yang","orcid":"https://orcid.org/0000-0003-1101-7927"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hailong Yang","raw_affiliation_strings":["Beihang University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beihang University","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100339921","display_name":"Bing Liu","orcid":"https://orcid.org/0000-0002-2365-6606"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bing Liu","raw_affiliation_strings":["SenseTime Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SenseTime Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5082233466","display_name":"Peng Hu","orcid":"https://orcid.org/0000-0003-3868-3997"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Peng Hu","raw_affiliation_strings":["SenseTime Research Beihang University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SenseTime Research Beihang University","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048414655","display_name":"Ruihao Gong","orcid":"https://orcid.org/0000-0002-6024-7086"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ruihao Gong","raw_affiliation_strings":["Beihang University SenseTime Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beihang University SenseTime Research","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100359580","display_name":"Yanfei Wang","orcid":"https://orcid.org/0000-0001-7992-3999"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yanfei Wang","raw_affiliation_strings":["SenseTime Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SenseTime Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100431210","display_name":"Rui Wang","orcid":"https://orcid.org/0000-0002-2045-4582"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Rui Wang","raw_affiliation_strings":["Beihang University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beihang University","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5074183877","display_name":"Zhongzhi Luan","orcid":"https://orcid.org/0000-0002-7186-0556"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhongzhi Luan","raw_affiliation_strings":["Beihang University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beihang University","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5079362609","display_name":"Depei Qian","orcid":"https://orcid.org/0000-0002-5382-1473"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Depei Qian","raw_affiliation_strings":["Beihang University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beihang University","institution_ids":["https://openalex.org/I82880672"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":1.1106,"has_fulltext":false,"cited_by_count":20,"citation_normalized_percentile":{"value":0.85037928,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":90,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"12"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9998999834060669,"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"}},"topics":[{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9998999834060669,"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"}},{"id":"https://openalex.org/T10627","display_name":"Advanced Image and Video Retrieval Techniques","score":0.9991000294685364,"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"}},{"id":"https://openalex.org/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.9980999827384949,"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":0.8516359329223633},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.743408203125},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.5792644023895264},{"id":"https://openalex.org/keywords/convolution","display_name":"Convolution (computer science)","score":0.5430192351341248},{"id":"https://openalex.org/keywords/memory-hierarchy","display_name":"Memory hierarchy","score":0.5387799143791199},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5295929312705994},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.38724786043167114},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.34594571590423584},{"id":"https://openalex.org/keywords/computer-hardware","display_name":"Computer hardware","score":0.3344459533691406},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.23208266496658325},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.11304989457130432}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8516359329223633},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.743408203125},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.5792644023895264},{"id":"https://openalex.org/C45347329","wikidata":"https://www.wikidata.org/wiki/Q5166604","display_name":"Convolution (computer science)","level":3,"score":0.5430192351341248},{"id":"https://openalex.org/C2778100165","wikidata":"https://www.wikidata.org/wiki/Q1589327","display_name":"Memory hierarchy","level":3,"score":0.5387799143791199},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5295929312705994},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.38724786043167114},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.34594571590423584},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.3344459533691406},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.23208266496658325},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.11304989457130432},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3404397.3404407","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3404397.3404407","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"49th International Conference on Parallel Processing - ICPP","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.5199999809265137,"display_name":"Affordable and clean energy","id":"https://metadata.un.org/sdg/7"}],"awards":[{"id":"https://openalex.org/G7784462196","display_name":null,"funder_award_id":"61502019, 61732002","funder_id":"https://openalex.org/F4320327720","funder_display_name":"Foundation for Innovative Research Groups of the National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320327720","display_name":"Foundation for Innovative Research Groups of the National Natural Science Foundation of China","ror":null}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":25,"referenced_works":["https://openalex.org/W1861492603","https://openalex.org/W2002257715","https://openalex.org/W2108598243","https://openalex.org/W2125389748","https://openalex.org/W2172654076","https://openalex.org/W2194775991","https://openalex.org/W2565516711","https://openalex.org/W2626696598","https://openalex.org/W2736230459","https://openalex.org/W2766205627","https://openalex.org/W2771253733","https://openalex.org/W2784372305","https://openalex.org/W2789545641","https://openalex.org/W2791673912","https://openalex.org/W2804032941","https://openalex.org/W2914209329","https://openalex.org/W2952122856","https://openalex.org/W2962298324","https://openalex.org/W2963446712","https://openalex.org/W2963674932","https://openalex.org/W2980186997","https://openalex.org/W3004061291","https://openalex.org/W3008102594","https://openalex.org/W3035364348","https://openalex.org/W3101543398"],"related_works":["https://openalex.org/W2979160909","https://openalex.org/W2114837856","https://openalex.org/W4387838477","https://openalex.org/W2359364609","https://openalex.org/W2322476848","https://openalex.org/W3204400881","https://openalex.org/W3214410901","https://openalex.org/W3204296682","https://openalex.org/W3183118997","https://openalex.org/W2917767146"],"abstract_inverted_index":{"With":[0],"the":[1,11,21,32,51,87,124,170,203,214,221],"continuous":[2],"demand":[3],"for":[4,64,106,138],"higher":[5,151],"accuracy":[6,37],"of":[7,20,53,91,197,216,228],"deep":[8],"neural":[9],"networks,":[10],"model":[12,25,33],"size":[13,34],"has":[14],"increased":[15],"significantly.":[16],"Quantization":[17],"is":[18,220],"one":[19],"most":[22],"widely":[23],"used":[24],"compression":[26],"methods,":[27],"which":[28],"can":[29],"effectively":[30],"reduce":[31],"without":[35],"severe":[36],"loss.":[38],"Modern":[39],"processors":[40],"such":[41,208],"as":[42,209],"ARM":[43,71,99,237],"CPU":[44,72,238],"and":[45,61,77,82,110,130,162,175,206,211,239],"NVIDIA":[46,78,154,242],"GPU":[47,79,173],"have":[48],"already":[49],"provided":[50],"support":[52],"low-bit":[54,69,93,199,230],"arithmetic":[55],"instructions.":[56],"However,":[57],"there":[58],"lack":[59],"efficient":[60,226],"practical":[62],"optimizations":[63],"convolution":[65,94,114,139,200],"computation":[66,126,171],"towards":[67],"extremely":[68,92,198,229],"on":[70,95,236,241],"(e.g.,":[73,80,145],"2":[74,107,233],"\u223c":[75,108,112,147,234],"8-bit)":[76],"4-bit":[81],"8-bit).":[83],"This":[84],"paper":[85],"explores":[86],"performance":[88,196],"optimization":[89],"methods":[90],"diverse":[96],"architectures.":[97],"On":[98,153],"CPU,":[100],"we":[101,122,156],"propose":[102,157,180],"two":[103],"instruction":[104],"schemes":[105],"3-bit":[109],"4":[111,146],"8-bit":[113,235],"with":[115,127,140],"corresponding":[116],"register":[117],"allocation":[118],"methods.":[119],"In":[120],"addition,":[121],"re-design":[123],"GEMM":[125],"data":[128,159,186],"padding":[129],"packing":[131],"optimizations.":[132],"We":[133,178],"also":[134,179],"implement":[135],"winograd":[136],"algorithm":[137],"some":[141],"specific":[142],"bit":[143],"width":[144],"6-bit)":[148],"to":[149,167,172,183,202],"achieve":[150,194],"performance.":[152],"GPU,":[155],"a":[158],"partition":[160],"mechanism":[161],"multi-level":[163],"memory":[164,176],"access":[165],"optimizations,":[166],"better":[168,195],"adapt":[169],"thread":[174],"hierarchy.":[177],"quantization":[181],"fusion":[182],"eliminate":[184],"unnecessary":[185],"access.":[187],"The":[188],"experiment":[189],"results":[190],"demonstrate":[191],"our":[192,217],"implementations":[193,227],"compared":[201],"state-of-the-art":[204],"frameworks":[205],"libraries":[207],"ncnn":[210],"cuDNN.":[212],"To":[213],"best":[215],"knowledge,":[218],"this":[219],"first":[222],"work":[223],"that":[224],"provides":[225],"convolutions":[231],"covering":[232],"4-bit/8-bit":[240],"GPU.":[243]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":2},{"year":2024,"cited_by_count":4},{"year":2023,"cited_by_count":1},{"year":2022,"cited_by_count":2},{"year":2021,"cited_by_count":10}],"updated_date":"2026-07-30T17:31:21.811387","created_date":"2025-10-10T00:00:00"}