{"id":"https://openalex.org/W3008515469","doi":"https://doi.org/10.1145/3373087.3375301","title":"When Massive GPU Parallelism Ain't Enough","display_name":"When Massive GPU Parallelism Ain't Enough","publication_year":2020,"publication_date":"2020-02-23","ids":{"openalex":"https://openalex.org/W3008515469","doi":"https://doi.org/10.1145/3373087.3375301","mag":"3008515469"},"language":"en","primary_location":{"id":"doi:10.1145/3373087.3375301","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3373087.3375301","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2020 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays","raw_type":"proceedings-article"},"type":"article","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/A5026761218","display_name":"Vladimir Rybalkin","orcid":"https://orcid.org/0000-0002-0926-6062"},"institutions":[{"id":"https://openalex.org/I153267046","display_name":"University of Kaiserslautern","ror":"https://ror.org/04zrf7b53","country_code":"DE","type":"education","lineage":["https://openalex.org/I153267046"]}],"countries":["DE"],"is_corresponding":true,"raw_author_name":"Vladimir Rybalkin","raw_affiliation_strings":["University of Kaiserslautern, Kaiserslautern, Germany"],"affiliations":[{"raw_affiliation_string":"University of Kaiserslautern, Kaiserslautern, Germany","institution_ids":["https://openalex.org/I153267046"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5059285190","display_name":"Norbert Wehn","orcid":"https://orcid.org/0000-0002-9010-086X"},"institutions":[{"id":"https://openalex.org/I153267046","display_name":"University of Kaiserslautern","ror":"https://ror.org/04zrf7b53","country_code":"DE","type":"education","lineage":["https://openalex.org/I153267046"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Norbert Wehn","raw_affiliation_strings":["University of Kaiserslautern, Kaiserslautern, Germany"],"affiliations":[{"raw_affiliation_string":"University of Kaiserslautern, Kaiserslautern, Germany","institution_ids":["https://openalex.org/I153267046"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":["https://openalex.org/A5026761218"],"corresponding_institution_ids":["https://openalex.org/I153267046"],"apc_list":null,"apc_paid":null,"fwci":1.1724,"has_fulltext":false,"cited_by_count":13,"citation_normalized_percentile":{"value":0.80617732,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":89,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"111","last_page":"121"},"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/T10601","display_name":"Handwritten Text Recognition Techniques","score":0.9995999932289124,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9972000122070312,"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":0.8914196491241455},{"id":"https://openalex.org/keywords/mnist-database","display_name":"MNIST database","score":0.6998279094696045},{"id":"https://openalex.org/keywords/field-programmable-gate-array","display_name":"Field-programmable gate array","score":0.5692185759544373},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.5380153059959412},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.5370646119117737},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.46346205472946167},{"id":"https://openalex.org/keywords/computer-architecture","display_name":"Computer architecture","score":0.4385952055454254},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.4353744387626648},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.43139681220054626},{"id":"https://openalex.org/keywords/hardware-acceleration","display_name":"Hardware acceleration","score":0.4268464744091034},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.4235474467277527},{"id":"https://openalex.org/keywords/data-parallelism","display_name":"Data parallelism","score":0.415479838848114},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.3492732644081116},{"id":"https://openalex.org/keywords/parallelism","display_name":"Parallelism (grammar)","score":0.22869518399238586},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.17197838425636292}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8914196491241455},{"id":"https://openalex.org/C190502265","wikidata":"https://www.wikidata.org/wiki/Q17069496","display_name":"MNIST database","level":3,"score":0.6998279094696045},{"id":"https://openalex.org/C42935608","wikidata":"https://www.wikidata.org/wiki/Q190411","display_name":"Field-programmable gate array","level":2,"score":0.5692185759544373},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.5380153059959412},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5370646119117737},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.46346205472946167},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.4385952055454254},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.4353744387626648},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.43139681220054626},{"id":"https://openalex.org/C13164978","wikidata":"https://www.wikidata.org/wiki/Q600158","display_name":"Hardware acceleration","level":3,"score":0.4268464744091034},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4235474467277527},{"id":"https://openalex.org/C61483411","wikidata":"https://www.wikidata.org/wiki/Q3124522","display_name":"Data parallelism","level":3,"score":0.415479838848114},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3492732644081116},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.22869518399238586},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.17197838425636292}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3373087.3375301","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3373087.3375301","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2020 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/7","display_name":"Affordable and clean energy","score":0.8899999856948853}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":28,"referenced_works":["https://openalex.org/W855255571","https://openalex.org/W1771459135","https://openalex.org/W1895577753","https://openalex.org/W1901129140","https://openalex.org/W1909234690","https://openalex.org/W2112796928","https://openalex.org/W2300242332","https://openalex.org/W2317505761","https://openalex.org/W2327501763","https://openalex.org/W2469490737","https://openalex.org/W2565125333","https://openalex.org/W2573601984","https://openalex.org/W2585720638","https://openalex.org/W2765235648","https://openalex.org/W2786974559","https://openalex.org/W2802022891","https://openalex.org/W2803263632","https://openalex.org/W2827334269","https://openalex.org/W2902759232","https://openalex.org/W2916222163","https://openalex.org/W2950894517","https://openalex.org/W2953318193","https://openalex.org/W2963526839","https://openalex.org/W2964325005","https://openalex.org/W2964333985","https://openalex.org/W3013596877","https://openalex.org/W3102058420","https://openalex.org/W3102169921"],"related_works":["https://openalex.org/W3102660566","https://openalex.org/W2177838837","https://openalex.org/W2963831937","https://openalex.org/W2748614007","https://openalex.org/W4285144618","https://openalex.org/W2080419754","https://openalex.org/W2532502681","https://openalex.org/W4285104150","https://openalex.org/W4389319487","https://openalex.org/W4319952061"],"abstract_inverted_index":{"Multidimensional":[0],"Long":[1],"Short-Term":[2],"Memory":[3],"(MD-LSTM)":[4],"neural":[5,58],"network":[6],"is":[7,61,89,118],"an":[8,119],"extension":[9],"of":[10,85,95,109,136,164,207,238],"one-dimensional":[11],"LSTM":[12],"for":[13,93,102,122,157,172,186,241],"data":[14],"with":[15,234],"more":[16],"than":[17],"one":[18],"dimension":[19],"that":[20,47,65,105,116,125,199],"allows":[21],"MD-LSTM":[22,70],"to":[23,56,90,98,211,218],"show":[24],"state-of-the-art":[25,235],"results":[26],"in":[27,71,78,130,205,232],"various":[28],"applications":[29],"including":[30],"handwritten":[31,187],"text":[32],"recognition,":[33],"medical":[34],"imaging,":[35],"and":[36,53,80,181,213,229],"many":[37],"more.":[38],"However,":[39],"efficient":[40,103],"implementation":[41,204],"suffers":[42],"from":[43,177],"very":[44],"sequential":[45],"execution":[46],"tremendously":[48],"slows":[49],"down":[50],"both":[51],"training":[52,104],"inference":[54,94],"compared":[55],"other":[57],"networks.":[59],"This":[60],"the":[62,72,86,141,146,153,221],"primary":[63],"reason":[64],"prevents":[66],"intensive":[67],"research":[68,88,113],"involving":[69],"recent":[73],"years,":[74],"despite":[75],"large":[76],"progress":[77],"microelectronics":[79],"architectures.":[81],"The":[82],"main":[83],"goal":[84],"current":[87],"provide":[91,140],"acceleration":[92],"MD-LSTM,":[96],"so":[97],"open":[99],"a":[100,128,133,161],"door":[101],"can":[106,126],"boost":[107],"application":[108],"MD-LSTM.":[110,158],"By":[111],"this":[112,149],"we":[114,151,195],"advocate":[115],"FPGA":[117],"alternative":[120],"platform":[121],"deep":[123],"learning":[124],"offer":[127],"solution":[129],"cases":[131],"when":[132],"massive":[134],"parallelism":[135],"GPUs":[137],"does":[138],"not":[139],"necessary":[142],"performance":[143],"required":[144],"by":[145,209,216],"application.":[147],"In":[148],"paper,":[150],"present":[152],"first":[154],"hardware":[155],"architecture":[156,194],"We":[159],"conduct":[160],"systematic":[162],"exploration":[163],"precision":[165],"vs.":[166],"accuracy":[167,228],"trade-off":[168],"using":[169],"challenging":[170],"dataset":[171,185],"historical":[173],"document":[174],"image":[175],"binarization":[176],"DIBCO":[178],"2017":[179],"contest,":[180],"well":[182],"known":[183],"MNIST":[184,242],"digits":[188],"recognition.":[189],"Based":[190],"on":[191],"our":[192,224],"new":[193],"implement":[196],"FPGA-based":[197,236],"accelerator":[198,225],"outperforms":[200],"NVIDIA":[201],"K80":[202],"GPU":[203],"terms":[206],"runtime":[208],"up":[210,217],"50x":[212],"energy":[214],"efficiency":[215],"746x.":[219],"At":[220],"same":[222],"time,":[223],"demonstrates":[226],"higher":[227],"comparable":[230],"throughput":[231],"comparison":[233],"implementations":[237],"multilayer":[239],"perceptron":[240],"dataset.":[243]},"counts_by_year":[{"year":2025,"cited_by_count":1},{"year":2023,"cited_by_count":1},{"year":2022,"cited_by_count":1},{"year":2021,"cited_by_count":8},{"year":2020,"cited_by_count":2}],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}