{"id":"https://openalex.org/W4416873868","doi":"https://doi.org/10.1109/sbac-padw69789.2025.00024","title":"Energy-Aware Deep Learning on GPUs through Parameter Sharing and Mixed Precision Training","display_name":"Energy-Aware Deep Learning on GPUs through Parameter Sharing and Mixed Precision Training","publication_year":2025,"publication_date":"2025-10-28","ids":{"openalex":"https://openalex.org/W4416873868","doi":"https://doi.org/10.1109/sbac-padw69789.2025.00024"},"language":"en","primary_location":{"id":"doi:10.1109/sbac-padw69789.2025.00024","is_oa":false,"landing_page_url":"https://doi.org/10.1109/sbac-padw69789.2025.00024","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/SBC 37th International Symposium on Computer Architecture and High Performance Computing Workshops (SBAC-PADW)","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/A5092356330","display_name":"Roblex Nana Tchakoute","orcid":"https://orcid.org/0009-0009-8636-8466"},"institutions":[{"id":"https://openalex.org/I4210115519","display_name":"Centre de Recherche en Informatique","ror":"https://ror.org/020cdve92","country_code":"FR","type":"facility","lineage":["https://openalex.org/I190752583","https://openalex.org/I2746051580","https://openalex.org/I4210091621","https://openalex.org/I4210115519","https://openalex.org/I70768539"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Roblex Nana Tchakoute","raw_affiliation_strings":["Mines Paris - PSL,Centre de Recherche en Informatique (CRI),Fontainebleau,France"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Mines Paris - PSL,Centre de Recherche en Informatique (CRI),Fontainebleau,France","institution_ids":["https://openalex.org/I4210115519"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5057912061","display_name":"Claude Tadonki","orcid":"https://orcid.org/0000-0003-1194-6400"},"institutions":[{"id":"https://openalex.org/I4210115519","display_name":"Centre de Recherche en Informatique","ror":"https://ror.org/020cdve92","country_code":"FR","type":"facility","lineage":["https://openalex.org/I190752583","https://openalex.org/I2746051580","https://openalex.org/I4210091621","https://openalex.org/I4210115519","https://openalex.org/I70768539"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Claude Tadonki","raw_affiliation_strings":["Mines Paris - PSL,Centre de Recherche en Informatique (CRI),Fontainebleau,France"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Mines Paris - PSL,Centre de Recherche en Informatique (CRI),Fontainebleau,France","institution_ids":["https://openalex.org/I4210115519"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.3694967,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"116","last_page":"123"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.5497000217437744,"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"}},"topics":[{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.5497000217437744,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.22619999945163727,"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/T14347","display_name":"Big Data and Digital Economy","score":0.07289999723434448,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/memory-footprint","display_name":"Memory footprint","score":0.7983999848365784},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.7200000286102295},{"id":"https://openalex.org/keywords/suite","display_name":"Suite","score":0.7067999839782715},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.7032999992370605},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.5418000221252441},{"id":"https://openalex.org/keywords/memory-model","display_name":"Memory model","score":0.4869999885559082},{"id":"https://openalex.org/keywords/efficient-energy-use","display_name":"Efficient energy use","score":0.48100000619888306},{"id":"https://openalex.org/keywords/energy","display_name":"Energy (signal processing)","score":0.428600013256073}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8284000158309937},{"id":"https://openalex.org/C74912251","wikidata":"https://www.wikidata.org/wiki/Q6815727","display_name":"Memory footprint","level":2,"score":0.7983999848365784},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.7200000286102295},{"id":"https://openalex.org/C79581498","wikidata":"https://www.wikidata.org/wiki/Q1367530","display_name":"Suite","level":2,"score":0.7067999839782715},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.7032999992370605},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5723999738693237},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5418000221252441},{"id":"https://openalex.org/C12186640","wikidata":"https://www.wikidata.org/wiki/Q6815743","display_name":"Memory model","level":3,"score":0.4869999885559082},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.48100000619888306},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4659000039100647},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.428600013256073},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.4203000068664551},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.4050000011920929},{"id":"https://openalex.org/C117896860","wikidata":"https://www.wikidata.org/wiki/Q11376","display_name":"Acceleration","level":2,"score":0.3930000066757202},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.31859999895095825},{"id":"https://openalex.org/C2780165032","wikidata":"https://www.wikidata.org/wiki/Q16869822","display_name":"Energy consumption","level":2,"score":0.3149999976158142},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.3019999861717224},{"id":"https://openalex.org/C2778565505","wikidata":"https://www.wikidata.org/wiki/Q2207566","display_name":"Spec#","level":2,"score":0.29760000109672546},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.289900004863739},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.2766000032424927},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.27390000224113464},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.2721000015735626},{"id":"https://openalex.org/C76970557","wikidata":"https://www.wikidata.org/wiki/Q1869750","display_name":"Loop unrolling","level":3,"score":0.26489999890327454},{"id":"https://openalex.org/C132943942","wikidata":"https://www.wikidata.org/wiki/Q2562511","display_name":"Footprint","level":2,"score":0.2603999972343445}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/sbac-padw69789.2025.00024","is_oa":false,"landing_page_url":"https://doi.org/10.1109/sbac-padw69789.2025.00024","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/SBC 37th International Symposium on Computer Architecture and High Performance Computing Workshops (SBAC-PADW)","raw_type":"proceedings-article"},{"id":"pmh:oai:HAL:hal-05393922v1","is_oa":false,"landing_page_url":"https://minesparis-psl.hal.science/hal-05393922","pdf_url":null,"source":{"id":"https://openalex.org/S4306402512","display_name":"HAL (Le Centre pour la Communication Scientifique Directe)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I1294671590","host_organization_name":"Centre National de la Recherche Scientifique","host_organization_lineage":["https://openalex.org/I1294671590"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"2025 IEEE/SBC 37th International Symposium on Computer Architecture and High Performance Computing Workshops (SBAC-PADW), Oct 2025, Bonito, Brazil. pp.116-123, ⟨10.1109/SBAC-PADW69789.2025.00024⟩","raw_type":"Conference papers"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":17,"referenced_works":["https://openalex.org/W1972384243","https://openalex.org/W1976412024","https://openalex.org/W2003519051","https://openalex.org/W2194775991","https://openalex.org/W2251939518","https://openalex.org/W2566079294","https://openalex.org/W2606722458","https://openalex.org/W2617091226","https://openalex.org/W2923014074","https://openalex.org/W2963809228","https://openalex.org/W3105645800","https://openalex.org/W3213111742","https://openalex.org/W4289827994","https://openalex.org/W4385570274","https://openalex.org/W4401326396","https://openalex.org/W4402666098","https://openalex.org/W4404740455"],"related_works":[],"abstract_inverted_index":{"The":[0],"design":[1],"of":[2,52,66,73,156],"Deep":[3],"Learning":[4],"models":[5,175],"increasingly":[6],"relies":[7],"on":[8,32,79],"advanced":[9],"techniques":[10,55],"such":[11],"as":[12],"parameter":[13,103],"sharing":[14],"and":[15,22,37,43,76,90,112,140],"mixed":[16,91,148],"precision":[17,150],"training":[18,154],"to":[19,100,109,115,129,136,190],"handle":[20],"computational":[21],"memory":[23,38,120,138],"costs.":[24],"Although":[25],"effective":[26],"in":[27,194],"theory,":[28],"their":[29,195],"practical":[30],"impact":[31],"system-level":[33],"performance,":[34],"energy":[35,180],"consumption,":[36],"subsystem":[39],"behavior":[40],"is":[41,106],"complex":[42],"interdependent.":[44],"This":[45],"paper":[46],"presents":[47],"a":[48,63,70,80],"performance/energy":[49],"trade-off":[50],"analysis":[51,86],"the":[53,159,199],"aforementioned":[54],"through":[56],"an":[57,124,143,152,165],"empirical":[58,166],"case":[59],"study.":[60],"We":[61],"benchmark":[62],"diverse":[64],"suite":[65],"six":[67],"models,":[68],"including":[69],"direct":[71],"comparison":[72],"DistilBERT":[74],"(conventional)":[75],"ALBERT":[77,105],"(parameter-sharing)":[78],"multi-GPU":[81],"NVIDIA":[82],"A100":[83],"platform.":[84],"Our":[85],"across":[87],"FP32,":[88],"TF32,":[89],"BF16":[92,149],"precisions":[93],"reveals":[94],"two":[95],"key":[96],"findings.":[97],"First,":[98],"contrary":[99],"its":[101,130],"smaller":[102],"count,":[104],"empirically":[107],"up":[108,114],"2.2\u00d7":[110],"slower":[111],"triggers":[113],"~":[116,157],"3\u00d7":[117],"higher":[118],"GPU":[119],"footprint":[121],"than":[122],"DistilBERT;":[123],"effect":[125],"that":[126,134,173],"we":[127,171,188],"attribute":[128],"deep":[131],"conceptual":[132],"unrolling":[133],"led":[135],"many":[137],"activations":[139],"thereby":[141],"incurs":[142],"important":[144],"overhead.":[145],"Second,":[146],"while":[147],"provides":[151],"average":[153],"speed-up":[155],"2.1\u00d7,":[158],"benefits":[160],"are":[161,178],"strongly":[162],"model-dependent.":[163],"Using":[164],"Throughput-per-Watt":[167],"(Samples/Joule)":[168],"efficiency":[169],"metric,":[170],"show":[172,184],"compute-bound":[174],"like":[176],"TinyLlama":[177],"more":[179],"efficient,":[181],"whereas":[182],"CNNs":[183],"marginal":[185],"improvements,":[186],"which":[187],"link":[189],"implicit":[191],"TF32":[192],"acceleration":[193],"FP32":[196],"baseline":[197],"via":[198],"cuDNN":[200],"library.":[201]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-12-01T00:00:00"}