{"id":"https://openalex.org/W4386191499","doi":"https://doi.org/10.1145/3617689","title":"Improving Computation and Memory Efficiency for Real-world Transformer Inference on GPUs","display_name":"Improving Computation and Memory Efficiency for Real-world Transformer Inference on GPUs","publication_year":2023,"publication_date":"2023-08-26","ids":{"openalex":"https://openalex.org/W4386191499","doi":"https://doi.org/10.1145/3617689"},"language":"en","primary_location":{"id":"doi:10.1145/3617689","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3617689","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3617689","source":{"id":"https://openalex.org/S26056741","display_name":"ACM Transactions on Architecture and Code Optimization","issn_l":"1544-3566","issn":["1544-3566","1544-3973"],"is_oa":true,"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 Architecture and Code Optimization","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://dl.acm.org/doi/pdf/10.1145/3617689","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5024085713","display_name":"Jiangsu Du","orcid":"https://orcid.org/0000-0003-4707-9492"},"institutions":[{"id":"https://openalex.org/I157773358","display_name":"Sun Yat-sen University","ror":"https://ror.org/0064kty71","country_code":"CN","type":"education","lineage":["https://openalex.org/I157773358"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Jiangsu Du","raw_affiliation_strings":["School of Computer Science and Engineering, Sun Yat-sen University, China"],"raw_orcid":"https://orcid.org/0000-0003-4707-9492","affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Sun Yat-sen University, China","institution_ids":["https://openalex.org/I157773358"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048148542","display_name":"Jiazhi Jiang","orcid":"https://orcid.org/0000-0002-1417-3012"},"institutions":[{"id":"https://openalex.org/I157773358","display_name":"Sun Yat-sen University","ror":"https://ror.org/0064kty71","country_code":"CN","type":"education","lineage":["https://openalex.org/I157773358"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jiazhi Jiang","raw_affiliation_strings":["School of Computer Science and Engineering, Sun Yat-sen University, China"],"raw_orcid":"https://orcid.org/0000-0002-1417-3012","affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Sun Yat-sen University, China","institution_ids":["https://openalex.org/I157773358"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101733890","display_name":"Zheng Jiang","orcid":"https://orcid.org/0009-0009-4718-2933"},"institutions":[{"id":"https://openalex.org/I157773358","display_name":"Sun Yat-sen University","ror":"https://ror.org/0064kty71","country_code":"CN","type":"education","lineage":["https://openalex.org/I157773358"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jiang Zheng","raw_affiliation_strings":["School of Computer Science and Engineering, Sun Yat-sen University, China"],"raw_orcid":"https://orcid.org/0009-0009-4718-2933","affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Sun Yat-sen University, China","institution_ids":["https://openalex.org/I157773358"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100442157","display_name":"Hongbin Zhang","orcid":"https://orcid.org/0009-0007-9145-3919"},"institutions":[{"id":"https://openalex.org/I157773358","display_name":"Sun Yat-sen University","ror":"https://ror.org/0064kty71","country_code":"CN","type":"education","lineage":["https://openalex.org/I157773358"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hongbin Zhang","raw_affiliation_strings":["School of Computer Science and Engineering, Sun Yat-sen University, China"],"raw_orcid":"https://orcid.org/0009-0007-9145-3919","affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Sun Yat-sen University, China","institution_ids":["https://openalex.org/I157773358"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5041534890","display_name":"Dan Huang","orcid":"https://orcid.org/0000-0001-5582-1031"},"institutions":[{"id":"https://openalex.org/I157773358","display_name":"Sun Yat-sen University","ror":"https://ror.org/0064kty71","country_code":"CN","type":"education","lineage":["https://openalex.org/I157773358"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Dan Huang","raw_affiliation_strings":["School of Computer Science and Engineering, Sun Yat-sen University, China"],"raw_orcid":"https://orcid.org/0000-0001-5582-1031","affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Sun Yat-sen University, China","institution_ids":["https://openalex.org/I157773358"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101633465","display_name":"Yutong Lu","orcid":"https://orcid.org/0000-0001-5315-3375"},"institutions":[{"id":"https://openalex.org/I157773358","display_name":"Sun Yat-sen University","ror":"https://ror.org/0064kty71","country_code":"CN","type":"education","lineage":["https://openalex.org/I157773358"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yutong Lu","raw_affiliation_strings":["School of Computer Science and Engineering, Sun Yat-sen University, China"],"raw_orcid":"https://orcid.org/0000-0001-5315-3375","affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Sun Yat-sen University, China","institution_ids":["https://openalex.org/I157773358"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5024085713"],"corresponding_institution_ids":["https://openalex.org/I157773358"],"apc_list":null,"apc_paid":null,"fwci":1.2521,"has_fulltext":true,"cited_by_count":11,"citation_normalized_percentile":{"value":0.82140855,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":94,"max":99},"biblio":{"volume":"20","issue":"4","first_page":"1","last_page":"22"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9987999796867371,"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":0.9987999796867371,"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/T10028","display_name":"Topic Modeling","score":0.9962000250816345,"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/T10627","display_name":"Advanced Image and Video Retrieval Techniques","score":0.9930999875068665,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.846305251121521},{"id":"https://openalex.org/keywords/memory-footprint","display_name":"Memory footprint","score":0.7901532649993896},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.6462857723236084},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5162113308906555},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.5160105228424072},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.48260748386383057},{"id":"https://openalex.org/keywords/granularity","display_name":"Granularity","score":0.4817750155925751},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.4151463806629181},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.2913670539855957},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.22236156463623047},{"id":"https://openalex.org/keywords/programming-language","display_name":"Programming language","score":0.1330450475215912}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.846305251121521},{"id":"https://openalex.org/C74912251","wikidata":"https://www.wikidata.org/wiki/Q6815727","display_name":"Memory footprint","level":2,"score":0.7901532649993896},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.6462857723236084},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5162113308906555},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.5160105228424072},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.48260748386383057},{"id":"https://openalex.org/C177774035","wikidata":"https://www.wikidata.org/wiki/Q1246948","display_name":"Granularity","level":2,"score":0.4817750155925751},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.4151463806629181},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2913670539855957},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.22236156463623047},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.1330450475215912},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.0},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3617689","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3617689","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3617689","source":{"id":"https://openalex.org/S26056741","display_name":"ACM Transactions on Architecture and Code Optimization","issn_l":"1544-3566","issn":["1544-3566","1544-3973"],"is_oa":true,"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 Architecture and Code Optimization","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1145/3617689","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3617689","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3617689","source":{"id":"https://openalex.org/S26056741","display_name":"ACM Transactions on Architecture and Code Optimization","issn_l":"1544-3566","issn":["1544-3566","1544-3973"],"is_oa":true,"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 Architecture and Code Optimization","raw_type":"journal-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G611074686","display_name":null,"funder_award_id":"2021YFB0301300","funder_id":"https://openalex.org/F4320335777","funder_display_name":"National Key Research and Development Program of China"},{"id":"https://openalex.org/G6186175735","display_name":null,"funder_award_id":"2019B030302002","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/F4320335777","display_name":"National Key Research and Development Program of China","ror":null}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4386191499.pdf","grobid_xml":"https://content.openalex.org/works/W4386191499.grobid-xml"},"referenced_works_count":14,"referenced_works":["https://openalex.org/W2913707927","https://openalex.org/W2923014074","https://openalex.org/W2970971581","https://openalex.org/W3014367186","https://openalex.org/W3093348380","https://openalex.org/W3130716829","https://openalex.org/W4226064176","https://openalex.org/W4281758439","https://openalex.org/W4282962596","https://openalex.org/W4287639545","https://openalex.org/W4295312788","https://openalex.org/W4320067900","https://openalex.org/W4385245566","https://openalex.org/W6739901393"],"related_works":["https://openalex.org/W2931688134","https://openalex.org/W2377919138","https://openalex.org/W2378857091","https://openalex.org/W4206178588","https://openalex.org/W4287635093","https://openalex.org/W3094491777","https://openalex.org/W3214715529","https://openalex.org/W3094340691","https://openalex.org/W2929170389","https://openalex.org/W2748428003"],"abstract_inverted_index":{"Transformer":[0],"models":[1],"have":[2,28],"emerged":[3],"as":[4],"a":[5,30,83,110,115,127,193,211],"leading":[6],"approach":[7,118,144,156,190],"in":[8,21,70,74,126,170],"the":[9,34,41,51,64,94,103,123,135,142,154,158,163,167,177,188,219,225],"field":[10],"of":[11,67,93,166,213,232],"natural":[12],"language":[13],"processing":[14,25,43],"(NLP)":[15],"and":[16,37,59,90,106,157,183,199,228],"are":[17],"increasingly":[18],"being":[19],"deployed":[20],"production":[22],"environments.":[23],"Graphic":[24],"units":[26],"(GPUs)":[27],"become":[29],"popular":[31],"choice":[32],"for":[33,54,86],"transformer":[35,55,96,111,221],"deployment":[36],"often":[38],"rely":[39],"on":[40,98,218,224],"batch":[42],"technique":[44],"to":[45,63,121,140,145,151,191],"ensure":[46],"high":[47],"hardware":[48],"performance.":[49,77],"Nonetheless,":[50],"current":[52],"practice":[53],"inference":[56,97],"encounters":[57],"computational":[58],"memory":[60,91,107,179,197,230],"redundancy":[61],"due":[62],"heavy-tailed":[65],"distribution":[66],"sequence":[68],"lengths":[69],"NLP":[71],"scenarios,":[72],"resulting":[73],"low":[75],"practical":[76],"In":[78],"this":[79],"article,":[80],"we":[81,186],"propose":[82,187],"unified":[84,208],"solution":[85,101,209],"improving":[87],"both":[88],"computation":[89,105,117],"efficiency":[92],"real-world":[95],"GPUs.":[99],"The":[100],"eliminates":[102],"redundant":[104,132,148],"footprint":[108,198,231],"across":[109],"model.":[112],"At":[113],"first,":[114],"GPU-oriented":[116],"is":[119],"proposed":[120],"process":[122],"self-attention":[124,168,226],"module":[125,138,169],"fine-grained":[128,155],"manner,":[129],"eliminating":[130],"its":[131,147],"computation.":[133,149],"Next,":[134],"multi-layer":[136],"perceptron":[137],"continues":[139],"use":[141],"word-accumulation":[143,159],"eliminate":[146],"Then,":[150],"better":[152,194],"unify":[153],"approach,":[160],"it":[161],"organizes":[162],"data":[164],"layout":[165],"block":[171],"granularity.":[172],"Since":[173],"aforementioned":[174],"approaches":[175],"make":[176],"required":[178],"size":[180],"largely":[181],"reduce":[182],"constantly":[184],"fluctuate,":[185],"chunk-based":[189],"enable":[192],"balance":[195],"between":[196],"allocation/free":[200],"efficiency.":[201],"Our":[202],"experimental":[203],"results":[204,234],"show":[205],"that":[206],"our":[207],"achieves":[210],"decrease":[212],"average":[214],"latency":[215],"by":[216,235],"28%":[217],"entire":[220],"model,":[222],"63.8%":[223],"module,":[227],"reduces":[229],"intermediate":[233],"7.8\u00d7,":[236],"compared":[237],"with":[238],"prevailing":[239],"frameworks.":[240]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":8},{"year":2024,"cited_by_count":2}],"updated_date":"2026-05-21T06:26:12.895304","created_date":"2025-10-10T00:00:00"}