{"id":"https://openalex.org/W4407953375","doi":"https://doi.org/10.1145/3706628.3708831","title":"DPUV4E: High-Throughput DPU Architecture Design for CNN on Versal ACAP","display_name":"DPUV4E: High-Throughput DPU Architecture Design for CNN on Versal ACAP","publication_year":2025,"publication_date":"2025-02-26","ids":{"openalex":"https://openalex.org/W4407953375","doi":"https://doi.org/10.1145/3706628.3708831"},"language":"en","primary_location":{"id":"doi:10.1145/3706628.3708831","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3706628.3708831","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2025 ACM/SIGDA International Symposium on Field Programmable Gate Arrays","raw_type":"proceedings-article"},"type":"conference-abstract","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":"Guoyu Li","orcid":"https://orcid.org/0009-0008-5914-323X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guoyu Li","raw_affiliation_strings":["Advanced Micro Devices, Inc., Beijing, China"],"raw_orcid":"https://orcid.org/0009-0008-5914-323X","affiliations":[{"raw_affiliation_string":"Advanced Micro Devices, Inc., Beijing, China","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5116425539","display_name":"Pengbo Zheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pengbo Zheng","raw_affiliation_strings":["Advanced Micro Devices, Inc., Beijing, China"],"raw_orcid":"https://orcid.org/0009-0008-1046-4702","affiliations":[{"raw_affiliation_string":"Advanced Micro Devices, Inc., Beijing, China","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101960234","display_name":"Jian Weng","orcid":"https://orcid.org/0009-0004-3415-2077"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jian Weng","raw_affiliation_strings":["Advanced Micro Devices, Inc., Beijing, China"],"raw_orcid":"https://orcid.org/0009-0004-3415-2077","affiliations":[{"raw_affiliation_string":"Advanced Micro Devices, Inc., Beijing, China","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5047913839","display_name":"Enshan Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Enshan Yang","raw_affiliation_strings":["Advanced Micro Devices, Inc., Beijing, China"],"raw_orcid":"https://orcid.org/0009-0000-4953-3025","affiliations":[{"raw_affiliation_string":"Advanced Micro Devices, Inc., Beijing, China","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"45","last_page":"45"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9970999956130981,"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":0.9970999956130981,"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/T12702","display_name":"Brain Tumor Detection and Classification","score":0.9962000250816345,"subfield":{"id":"https://openalex.org/subfields/2808","display_name":"Neurology"},"field":{"id":"https://openalex.org/fields/28","display_name":"Neuroscience"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}},{"id":"https://openalex.org/T10522","display_name":"Medical Imaging Techniques and Applications","score":0.9937999844551086,"subfield":{"id":"https://openalex.org/subfields/2741","display_name":"Radiology, Nuclear Medicine and Imaging"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/throughput","display_name":"Throughput","score":0.7355726361274719},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7080883979797363},{"id":"https://openalex.org/keywords/architecture","display_name":"Architecture","score":0.6302989721298218},{"id":"https://openalex.org/keywords/computer-architecture","display_name":"Computer architecture","score":0.5797719955444336},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.3987804651260376},{"id":"https://openalex.org/keywords/operating-system","display_name":"Operating system","score":0.3572142422199249}],"concepts":[{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.7355726361274719},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7080883979797363},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.6302989721298218},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.5797719955444336},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.3987804651260376},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.3572142422199249},{"id":"https://openalex.org/C555944384","wikidata":"https://www.wikidata.org/wiki/Q249","display_name":"Wireless","level":2,"score":0.0},{"id":"https://openalex.org/C142362112","wikidata":"https://www.wikidata.org/wiki/Q735","display_name":"Art","level":0,"score":0.0},{"id":"https://openalex.org/C153349607","wikidata":"https://www.wikidata.org/wiki/Q36649","display_name":"Visual arts","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3706628.3708831","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3706628.3708831","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2025 ACM/SIGDA International Symposium on Field Programmable Gate Arrays","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"display_name":"Industry, innovation and infrastructure","id":"https://metadata.un.org/sdg/9","score":0.41999998688697815}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2136583354","https://openalex.org/W2111238207","https://openalex.org/W2760721665","https://openalex.org/W2418291489","https://openalex.org/W2107954672","https://openalex.org/W330130819","https://openalex.org/W2288610023","https://openalex.org/W2112044895","https://openalex.org/W3121416282","https://openalex.org/W2038503502"],"abstract_inverted_index":{"Convolutional":[0],"Neural":[1],"Networks":[2],"(CNNs)":[3],"remain":[4],"prevalent":[5],"in":[6,23,32],"computer":[7],"vision":[8],"applications,":[9,49],"and":[10,16,35,104,176,198],"FPGAs,":[11],"known":[12],"for":[13,47,82,138,194,202],"their":[14],"flexibility":[15],"energy":[17],"efficiency,":[18],"have":[19],"become":[20],"essential":[21],"components":[22],"heterogeneous":[24],"acceleration":[25],"systems.":[26],"However,":[27],"traditional":[28,162],"FPGAs":[29],"face":[30],"challenges":[31],"balancing":[33],"performance":[34],"versatility":[36],"due":[37],"to":[38,54,93,107,123,135,152,178,192,200],"limited":[39],"on-chip":[40],"resources.":[41],"AMD's":[42],"Versal":[43,84],"ACAP":[44],"architecture,":[45,85],"tailored":[46],"AI":[48,51],"incorporates":[50],"Engines":[52],"(AIEs)":[53],"deliver":[55],"high":[56],"computational":[57,110],"power.":[58],"Nevertheless,":[59],"the":[60,68,72,83,119,130,159],"platform":[61],"suffers":[62],"from":[63,89],"insufficient":[64],"memory":[65],"bandwidth,":[66],"hindering":[67],"full":[69],"utilization":[70],"of":[71,132,161],"AIEs'":[73],"theoretical":[74],"performance.":[75],"In":[76],"this":[77],"paper,":[78],"we":[79,128],"present":[80],"DPUV4E":[81],"providing":[86],"configurations":[87],"ranging":[88],"2PE":[90],"(32.6":[91],"TOPS)":[92],"8PE":[94],"(131.0":[95],"TOPS).":[96],"We":[97],"design":[98,156,187],"two":[99],"computation":[100,113],"units,":[101],"Conv":[102],"PE":[103,134],"DWC":[105],"PE,":[106],"support":[108],"different":[109],"patterns.":[111],"Each":[112],"unit's":[114],"data":[115,120],"flow":[116],"efficiently":[117],"utilizes":[118],"reuse":[121],"opportunities":[122],"mitigate":[124],"bandwidth":[125],"bottlenecks.":[126],"Additionally,":[127],"extend":[129],"functionality":[131],"each":[133],"utilize":[136],"AIEs":[137],"non-convolutional":[139],"operations,":[140],"reducing":[141,167],"resource":[142],"overhead.":[143],"Experiments":[144],"on":[145],"over":[146],"50":[147],"models":[148,197],"show":[149],"that":[150],"compared":[151],"previous":[153],"designs,":[154],"our":[155,186],"provides":[157],"8.6\u00d7":[158],"TOPS/W":[160],"FPGA-based":[163],"DPU":[164],"designs":[165],"while":[166],"DSP":[168],"usage":[169,173],"by":[170,174,190],"95.8%,":[171],"LUT":[172],"44.7%,":[175],"latency":[177],"68.5%":[179],"under":[180],"single-batch":[181],"conditions.":[182],"For":[183],"end-to-end":[184],"inference,":[185],"improves":[188],"throughput":[189],"up":[191,199],"2.1\u00d7":[193],"depth-wise":[195],"convolution":[196],"1.7\u00d7":[201],"standard":[203],"models.":[204]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2026-08-18T07:49:30.821534","created_date":"2025-10-10T00:00:00"}