{"id":"https://openalex.org/W4318541578","doi":"https://doi.org/10.1145/3575693.3575747","title":"FLAT: An Optimized Dataflow for Mitigating Attention Bottlenecks","display_name":"FLAT: An Optimized Dataflow for Mitigating Attention Bottlenecks","publication_year":2023,"publication_date":"2023-01-27","ids":{"openalex":"https://openalex.org/W4318541578","doi":"https://doi.org/10.1145/3575693.3575747"},"language":"en","primary_location":{"id":"doi:10.1145/3575693.3575747","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3575693.3575747","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3575693.3575747","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://dl.acm.org/doi/pdf/10.1145/3575693.3575747","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5035130804","display_name":"Sheng-Chun Kao","orcid":"https://orcid.org/0000-0001-7928-9027"},"institutions":[{"id":"https://openalex.org/I130701444","display_name":"Georgia Institute of Technology","ror":"https://ror.org/01zkghx44","country_code":"US","type":"education","lineage":["https://openalex.org/I130701444"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Sheng-Chun Kao","raw_affiliation_strings":["Georgia Institute of Technology, USA"],"affiliations":[{"raw_affiliation_string":"Georgia Institute of Technology, USA","institution_ids":["https://openalex.org/I130701444"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103255393","display_name":"Suvinay Subramanian","orcid":"https://orcid.org/0000-0002-8715-8964"},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Suvinay Subramanian","raw_affiliation_strings":["Google, USA"],"affiliations":[{"raw_affiliation_string":"Google, USA","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5026344507","display_name":"Gaurav Agrawal","orcid":null},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Gaurav Agrawal","raw_affiliation_strings":["Microsoft, USA"],"affiliations":[{"raw_affiliation_string":"Microsoft, USA","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5070172290","display_name":"Amir Yazdanbakhsh","orcid":"https://orcid.org/0000-0001-8199-7671"},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Amir Yazdanbakhsh","raw_affiliation_strings":["Google, USA"],"affiliations":[{"raw_affiliation_string":"Google, USA","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5034089074","display_name":"Tushar Krishna","orcid":"https://orcid.org/0000-0001-5738-6942"},"institutions":[{"id":"https://openalex.org/I130701444","display_name":"Georgia Institute of Technology","ror":"https://ror.org/01zkghx44","country_code":"US","type":"education","lineage":["https://openalex.org/I130701444"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Tushar Krishna","raw_affiliation_strings":["Georgia Institute of Technology, USA"],"affiliations":[{"raw_affiliation_string":"Georgia Institute of Technology, USA","institution_ids":["https://openalex.org/I130701444"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":5,"corresponding_author_ids":["https://openalex.org/A5035130804"],"corresponding_institution_ids":["https://openalex.org/I130701444"],"apc_list":null,"apc_paid":null,"fwci":6.8057,"has_fulltext":false,"cited_by_count":57,"citation_normalized_percentile":{"value":0.97857068,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":96,"max":100},"biblio":{"volume":null,"issue":null,"first_page":"295","last_page":"310"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9998000264167786,"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":0.9998000264167786,"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9987000226974487,"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/T10502","display_name":"Advanced Memory and Neural Computing","score":0.9987000226974487,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic 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/dataflow","display_name":"Dataflow","score":0.9136664867401123},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.8236532211303711},{"id":"https://openalex.org/keywords/memory-footprint","display_name":"Memory footprint","score":0.679153323173523},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.5762951374053955},{"id":"https://openalex.org/keywords/bottleneck","display_name":"Bottleneck","score":0.5448051691055298},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.5369738340377808},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.5326359272003174},{"id":"https://openalex.org/keywords/distributed-computing","display_name":"Distributed computing","score":0.4499850273132324},{"id":"https://openalex.org/keywords/dataflow-architecture","display_name":"Dataflow architecture","score":0.44327986240386963},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.4122087061405182},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.39208561182022095},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.26564544439315796},{"id":"https://openalex.org/keywords/computer-network","display_name":"Computer network","score":0.1355108916759491}],"concepts":[{"id":"https://openalex.org/C96324660","wikidata":"https://www.wikidata.org/wiki/Q205446","display_name":"Dataflow","level":2,"score":0.9136664867401123},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8236532211303711},{"id":"https://openalex.org/C74912251","wikidata":"https://www.wikidata.org/wiki/Q6815727","display_name":"Memory footprint","level":2,"score":0.679153323173523},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.5762951374053955},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.5448051691055298},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.5369738340377808},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.5326359272003174},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.4499850273132324},{"id":"https://openalex.org/C176727019","wikidata":"https://www.wikidata.org/wiki/Q1172415","display_name":"Dataflow architecture","level":3,"score":0.44327986240386963},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.4122087061405182},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.39208561182022095},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.26564544439315796},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.1355108916759491},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.0},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3575693.3575747","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3575693.3575747","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3575693.3575747","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3575693.3575747","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3575693.3575747","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3575693.3575747","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2","raw_type":"proceedings-article"},"sustainable_development_goals":[{"score":0.8600000143051147,"display_name":"Affordable and clean energy","id":"https://metadata.un.org/sdg/7"}],"awards":[{"id":"https://openalex.org/G5529443368","display_name":null,"funder_award_id":"1909900","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4318541578.pdf","grobid_xml":"https://content.openalex.org/works/W4318541578.grobid-xml"},"referenced_works_count":52,"referenced_works":["https://openalex.org/W2055312318","https://openalex.org/W2067523571","https://openalex.org/W2080422774","https://openalex.org/W2094756095","https://openalex.org/W2194775991","https://openalex.org/W2289252105","https://openalex.org/W2590246587","https://openalex.org/W2605347906","https://openalex.org/W2625954420","https://openalex.org/W2806666179","https://openalex.org/W2900082550","https://openalex.org/W2919624000","https://openalex.org/W2940862705","https://openalex.org/W2945146780","https://openalex.org/W2949967139","https://openalex.org/W2963358710","https://openalex.org/W2981680926","https://openalex.org/W2998183051","https://openalex.org/W2998732502","https://openalex.org/W3016542674","https://openalex.org/W3017024317","https://openalex.org/W3034573343","https://openalex.org/W3036878841","https://openalex.org/W3038827379","https://openalex.org/W3092462694","https://openalex.org/W3112293503","https://openalex.org/W3131920484","https://openalex.org/W3131922516","https://openalex.org/W3157609068","https://openalex.org/W3159727696","https://openalex.org/W3174529902","https://openalex.org/W3181262653","https://openalex.org/W3189877953","https://openalex.org/W3206453033","https://openalex.org/W3208253409","https://openalex.org/W4214493665","https://openalex.org/W4214588794","https://openalex.org/W4224267386","https://openalex.org/W4249932213","https://openalex.org/W4255852734","https://openalex.org/W4288083528","https://openalex.org/W4293118024","https://openalex.org/W4300171661","https://openalex.org/W4308083827","https://openalex.org/W6633126254","https://openalex.org/W6685052161","https://openalex.org/W6739901393","https://openalex.org/W6752057402","https://openalex.org/W6763509872","https://openalex.org/W6779879114","https://openalex.org/W6780775112","https://openalex.org/W6785783668"],"related_works":["https://openalex.org/W2564598376","https://openalex.org/W1484403103","https://openalex.org/W2584408851","https://openalex.org/W2115158825","https://openalex.org/W2101960124","https://openalex.org/W2783505431","https://openalex.org/W2187181201","https://openalex.org/W2521947294","https://openalex.org/W2070353846","https://openalex.org/W4236419692"],"abstract_inverted_index":{"Attention":[0],"mechanisms,":[1],"primarily":[2],"designed":[3],"to":[4,52,65,111,130,167,209,219,241],"capture":[5],"pairwise":[6],"correlations":[7],"between":[8],"words,":[9],"have":[10],"become":[11],"the":[12,31,106,118,132,142,150,168,223],"backbone":[13],"of":[14,33,45,71,121,163,198],"machine":[15],"learning,":[16],"expanding":[17],"beyond":[18],"natural":[19],"language":[20],"processing":[21],"into":[22],"other":[23],"domains.":[24],"This":[25,48,74,94],"growth":[26,60,110],"in":[27,61],"adaptation":[28],"comes":[29],"at":[30,42],"cost":[32],"prohibitively":[34],"large":[35],"memory":[36,62,107,152],"requirements":[37],"and":[38,58,68,159,161],"computational":[39],"complexity,":[40],"especially":[41],"higher":[43],"number":[44],"input":[46,72,203],"elements.":[47,73,230,243],"limitation":[49],"is":[50],"due":[51],"inherently":[53],"limited":[54,69],"data":[55,133],"reuse":[56,134],"opportunities":[57],"quadratic":[59,109],"footprints,":[63],"leading":[64],"severe":[66],"memory-boundedness":[67],"scalability":[70],"work":[75],"addresses":[76],"these":[77],"challenges":[78],"by":[79],"devising":[80],"a":[81,101,113,127,195],"tailored":[82],"dataflow":[83,95,176,217],"optimization,":[84],"called":[85],"FLAT,":[86],"for":[87,226,237],"attention":[88,98,136,220],"mechanisms":[89],"without":[90],"altering":[91],"their":[92],"functionality.":[93],"processes":[96],"costly":[97],"operations":[99,221],"through":[100],"unique":[102],"fusion":[103],"mechanism,":[104,124],"transforming":[105],"footprint":[108],"merely":[112],"linear":[114],"one.":[115],"To":[116],"realize":[117],"full":[119],"potential":[120],"this":[122],"bespoke":[123],"we":[125],"propose":[126],"tiling":[128],"approach":[129],"enhance":[131],"across":[135,194],"operations.":[137],"Our":[138,211],"method":[139],"both":[140],"mitigates":[141],"off-chip":[143],"bandwidth":[144],"bottleneck":[145],"as":[146,148],"well":[147],"reduces":[149],"on-chip":[151,179],"requirement.":[153],"FLAT":[154,186,233],"delivers":[155],"1.94x":[156],"(1.76x)":[157],"speedup":[158],"49%":[160],"(42%)":[162],"energy":[164],"savings":[165],"compared":[166],"state-of-the-art":[169,215],"Edge":[170],"(Cloud)":[171],"accelerators":[172],"with":[173,202,239],"no":[174],"customized":[175],"optimization.":[177],"When":[178],"resources":[180],"are":[181],"scarce":[182],"(20":[183],"KB-200":[184],"KB),":[185],"yields,":[187],"on":[188],"average,":[189],"1.5x":[190],"end-to-end":[191],"latency":[192],"reduction":[193],"diverse":[196],"range":[197],"conventional":[199],"attention-based":[200],"models":[201,236],"sequence":[204],"lengths":[205],"ranging":[206],"from":[207],"512-token":[208],"64K-token.":[210],"evaluations":[212],"demonstrate":[213],"that":[214],"DNN":[216],"applied":[218],"reach":[222],"efficiency":[224],"limit":[225],"inputs":[227,238],"above":[228],"512":[229],"In":[231],"contrast,":[232],"unblocks":[234],"transformer":[235],"up":[240],"64K":[242]},"counts_by_year":[{"year":2026,"cited_by_count":6},{"year":2025,"cited_by_count":27},{"year":2024,"cited_by_count":21},{"year":2023,"cited_by_count":3}],"updated_date":"2026-04-10T15:06:20.359241","created_date":"2025-10-10T00:00:00"}