{"id":"https://openalex.org/W7084126764","doi":"https://doi.org/10.1109/newcas64648.2025.11107035","title":"High Utilization Vision Transformer Accelerator with Flexible Dataflow","display_name":"High Utilization Vision Transformer Accelerator with Flexible Dataflow","publication_year":2025,"publication_date":"2025-06-22","ids":{"openalex":"https://openalex.org/W7084126764","doi":"https://doi.org/10.1109/newcas64648.2025.11107035"},"language":"en","primary_location":{"id":"doi:10.1109/newcas64648.2025.11107035","is_oa":false,"landing_page_url":"https://doi.org/10.1109/newcas64648.2025.11107035","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 23rd IEEE Interregional NEWCAS Conference (NEWCAS)","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":null,"display_name":"Ting-Chun Lin","orcid":null},"institutions":[{"id":"https://openalex.org/I25846049","display_name":"National Tsing Hua University","ror":"https://ror.org/00zdnkx70","country_code":"TW","type":"education","lineage":["https://openalex.org/I25846049"]}],"countries":["TW"],"is_corresponding":false,"raw_author_name":"Ting-Chun Lin","raw_affiliation_strings":["Institute of Communications Engineering, National Tsing Hua University,Hsinchu,Taiwan"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Communications Engineering, National Tsing Hua University,Hsinchu,Taiwan","institution_ids":["https://openalex.org/I25846049"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Ching-Te Chiu","orcid":null},"institutions":[{"id":"https://openalex.org/I25846049","display_name":"National Tsing Hua University","ror":"https://ror.org/00zdnkx70","country_code":"TW","type":"education","lineage":["https://openalex.org/I25846049"]}],"countries":["TW"],"is_corresponding":false,"raw_author_name":"Ching-Te Chiu","raw_affiliation_strings":["National Tsing Hua University,Department of Computer Science,Hsinchu,Taiwan"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National Tsing Hua University,Department of Computer Science,Hsinchu,Taiwan","institution_ids":["https://openalex.org/I25846049"]}]},{"author_position":"last","author":{"id":null,"display_name":"Jhao-Wei Chen","orcid":null},"institutions":[{"id":"https://openalex.org/I25846049","display_name":"National Tsing Hua University","ror":"https://ror.org/00zdnkx70","country_code":"TW","type":"education","lineage":["https://openalex.org/I25846049"]}],"countries":["TW"],"is_corresponding":false,"raw_author_name":"Jhao-Wei Chen","raw_affiliation_strings":["National Tsing Hua University,Department of Computer Science,Hsinchu,Taiwan"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National Tsing Hua University,Department of Computer Science,Hsinchu,Taiwan","institution_ids":["https://openalex.org/I25846049"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I25846049"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"475","last_page":"479"},"is_retracted":false,"is_paratext":false,"is_xpac":true,"primary_topic":{"id":"https://openalex.org/T11992","display_name":"CCD and CMOS Imaging Sensors","score":0.5799999833106995,"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"}},"topics":[{"id":"https://openalex.org/T11992","display_name":"CCD and CMOS Imaging Sensors","score":0.5799999833106995,"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"}},{"id":"https://openalex.org/T10502","display_name":"Advanced Memory and Neural Computing","score":0.13420000672340393,"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"}},{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.08510000258684158,"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/transformer","display_name":"Transformer","score":0.569100022315979},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.5652999877929688},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5536999702453613},{"id":"https://openalex.org/keywords/dataflow","display_name":"Dataflow","score":0.5318999886512756},{"id":"https://openalex.org/keywords/reuse","display_name":"Reuse","score":0.5260000228881836},{"id":"https://openalex.org/keywords/dram","display_name":"Dram","score":0.5016000270843506},{"id":"https://openalex.org/keywords/software-deployment","display_name":"Software deployment","score":0.4740999937057495},{"id":"https://openalex.org/keywords/overlay","display_name":"Overlay","score":0.4244999885559082},{"id":"https://openalex.org/keywords/stencil","display_name":"Stencil","score":0.382099986076355}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7307000160217285},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.569100022315979},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.5652999877929688},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5536999702453613},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.5389999747276306},{"id":"https://openalex.org/C96324660","wikidata":"https://www.wikidata.org/wiki/Q205446","display_name":"Dataflow","level":2,"score":0.5318999886512756},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.5260000228881836},{"id":"https://openalex.org/C7366592","wikidata":"https://www.wikidata.org/wiki/Q1255620","display_name":"Dram","level":2,"score":0.5016000270843506},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.4740999937057495},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.46549999713897705},{"id":"https://openalex.org/C136085584","wikidata":"https://www.wikidata.org/wiki/Q910289","display_name":"Overlay","level":2,"score":0.4244999885559082},{"id":"https://openalex.org/C76752949","wikidata":"https://www.wikidata.org/wiki/Q7607499","display_name":"Stencil","level":2,"score":0.382099986076355},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.3718000054359436},{"id":"https://openalex.org/C177774035","wikidata":"https://www.wikidata.org/wiki/Q1246948","display_name":"Granularity","level":2,"score":0.3702000081539154},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.3691999912261963},{"id":"https://openalex.org/C13164978","wikidata":"https://www.wikidata.org/wiki/Q600158","display_name":"Hardware acceleration","level":3,"score":0.35499998927116394},{"id":"https://openalex.org/C58013763","wikidata":"https://www.wikidata.org/wiki/Q5754574","display_name":"High-level synthesis","level":3,"score":0.34790000319480896},{"id":"https://openalex.org/C2776221188","wikidata":"https://www.wikidata.org/wiki/Q21072556","display_name":"Design space exploration","level":2,"score":0.3359000086784363},{"id":"https://openalex.org/C42935608","wikidata":"https://www.wikidata.org/wiki/Q190411","display_name":"Field-programmable gate array","level":2,"score":0.3181999921798706},{"id":"https://openalex.org/C47487241","wikidata":"https://www.wikidata.org/wiki/Q5227230","display_name":"Data access","level":2,"score":0.313400000333786},{"id":"https://openalex.org/C2781198647","wikidata":"https://www.wikidata.org/wiki/Q1633673","display_name":"Datapath","level":2,"score":0.30309998989105225},{"id":"https://openalex.org/C55526617","wikidata":"https://www.wikidata.org/wiki/Q719375","display_name":"Operand","level":2,"score":0.2985000014305115},{"id":"https://openalex.org/C2778456923","wikidata":"https://www.wikidata.org/wiki/Q5337692","display_name":"Edge computing","level":3,"score":0.2955000102519989},{"id":"https://openalex.org/C31395832","wikidata":"https://www.wikidata.org/wiki/Q1318674","display_name":"Testbed","level":2,"score":0.28110000491142273},{"id":"https://openalex.org/C142962650","wikidata":"https://www.wikidata.org/wiki/Q240838","display_name":"Reconfigurable computing","level":3,"score":0.2791999876499176},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.26499998569488525},{"id":"https://openalex.org/C83283714","wikidata":"https://www.wikidata.org/wiki/Q121117","display_name":"Supercomputer","level":2,"score":0.26170000433921814},{"id":"https://openalex.org/C205711294","wikidata":"https://www.wikidata.org/wiki/Q176953","display_name":"Rendering (computer graphics)","level":2,"score":0.2549999952316284},{"id":"https://openalex.org/C79581498","wikidata":"https://www.wikidata.org/wiki/Q1367530","display_name":"Suite","level":2,"score":0.2513999938964844}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/newcas64648.2025.11107035","is_oa":false,"landing_page_url":"https://doi.org/10.1109/newcas64648.2025.11107035","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 23rd IEEE Interregional NEWCAS Conference (NEWCAS)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":19,"referenced_works":["https://openalex.org/W3017024317","https://openalex.org/W3034312118","https://openalex.org/W3035022492","https://openalex.org/W3094502228","https://openalex.org/W3096609285","https://openalex.org/W3171125843","https://openalex.org/W3192708755","https://openalex.org/W3206453033","https://openalex.org/W4205481600","https://openalex.org/W4220865834","https://openalex.org/W4280595844","https://openalex.org/W4286300805","https://openalex.org/W4293025165","https://openalex.org/W4360831795","https://openalex.org/W4380881077","https://openalex.org/W4387789586","https://openalex.org/W4390873361","https://openalex.org/W4400811522","https://openalex.org/W4401568518"],"related_works":[],"abstract_inverted_index":{"Vision":[0],"Transformers":[1],"(ViTs)":[2],"have":[3],"significantly":[4,59],"advanced":[5],"computer":[6],"vision":[7,69],"tasks":[8],"by":[9],"utilizing":[10],"self-attention":[11,34],"mechanisms.":[12],"However,":[13],"their":[14,56],"high":[15],"memory":[16],"and":[17,46,91,134,177],"computational":[18],"demands":[19,40],"pose":[20],"challenges":[21],"for":[22,77,125],"deployment":[23],"on":[24,159],"edge":[25],"devices.":[26],"While":[27],"most":[28],"transformer":[29,70,122],"accelerators":[30],"prioritize":[31],"optimizing":[32],"the":[33,36,78,96],"module,":[35],"feed-forward":[37],"network":[38],"also":[39],"considerable":[41],"computation.":[42],"Additionally,":[43],"hardware":[44,101,142,157],"utilization":[45,158],"data":[47,115],"reuse":[48,138],"across":[49,109,188],"dataflows":[50,130],"in":[51,81],"transformers":[52],"remain":[53],"underexplored,":[54],"despite":[55],"potential":[57],"to":[58,99,113,140],"boost":[60],"performance.":[61],"In":[62],"this":[63],"paper,":[64],"we":[65],"present":[66],"a":[67,86,105,136,150,172,189],"high-utilization":[68],"accelerator":[71,148],"with":[72,104,155],"flexible":[73,129],"dataflow,":[74],"optimized":[75],"specifically":[76],"linear":[79,132],"layers":[80],"ViTs.":[82],"Our":[83],"approach":[84],"introduces":[85],"tile":[87],"size":[88],"search":[89],"strategy":[90,139],"implements":[92],"intra-layer":[93],"parallelization":[94],"at":[95,163],"tiling":[97],"level":[98],"maximize":[100],"utilization,":[102],"along":[103],"DRAM":[106],"access":[107],"analysis":[108],"various":[110],"computation":[111],"orders":[112],"enhance":[114],"reuse.":[116],"The":[117,147],"design":[118],"features":[119],"three":[120],"pipelined":[121],"GEMM":[123],"engines":[124],"head-level":[126],"parallelism,":[127],"supporting":[128],"during":[131,144],"operations,":[133],"incorporates":[135],"resource":[137],"minimize":[141],"overhead":[143],"non-linear":[145],"operations.":[146],"achieves":[149],"2.77x":[151],"speedup":[152],"over":[153],"[1]":[154],"100%":[156],"DeiT-Tiny.":[160],"It":[161],"operates":[162],"1000":[164],"MHz":[165],"using":[166],"TSMC":[167],"40":[168],"nm":[169],"technology,":[170],"delivering":[171],"throughput":[173],"of":[174,181],"384":[175],"GOPS":[176],"an":[178],"area":[179],"efficiency":[180],"<tex":[182,190],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[183,191],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">$\\text{116.79":[184],"GOPS}":[185],"/":[186],"\\text{mm}^{2}$</tex>":[187,193],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">$\\text{3.29}\\,":[192],"area.":[194]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}