{"id":"https://openalex.org/W4401454756","doi":"https://doi.org/10.1145/3677333.3678153","title":"Fusing Depthwise and Pointwise Convolutions for Efficient Inference on GPUs","display_name":"Fusing Depthwise and Pointwise Convolutions for Efficient Inference on GPUs","publication_year":2024,"publication_date":"2024-08-09","ids":{"openalex":"https://openalex.org/W4401454756","doi":"https://doi.org/10.1145/3677333.3678153"},"language":"en","primary_location":{"id":"doi:10.1145/3677333.3678153","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3677333.3678153","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3677333.3678153","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"The 53rd International Conference on Parallel Processing Workshops","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/10.1145/3677333.3678153","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5057731855","display_name":"Fareed Qararyah","orcid":"https://orcid.org/0000-0002-3955-2836"},"institutions":[{"id":"https://openalex.org/I66862912","display_name":"Chalmers University of Technology","ror":"https://ror.org/040wg7k59","country_code":"SE","type":"education","lineage":["https://openalex.org/I66862912"]}],"countries":["SE"],"is_corresponding":false,"raw_author_name":"Fareed Qararyah","raw_affiliation_strings":["Chalmers University of Technology, Sweden"],"raw_orcid":"https://orcid.org/0000-0002-3955-2836","affiliations":[{"raw_affiliation_string":"Chalmers University of Technology, Sweden","institution_ids":["https://openalex.org/I66862912"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5008522376","display_name":"Muhammad Waqar Azhar","orcid":"https://orcid.org/0000-0003-0477-4540"},"institutions":[{"id":"https://openalex.org/I66862912","display_name":"Chalmers University of Technology","ror":"https://ror.org/040wg7k59","country_code":"SE","type":"education","lineage":["https://openalex.org/I66862912"]}],"countries":["SE"],"is_corresponding":false,"raw_author_name":"Muhammad Waqar Azhar","raw_affiliation_strings":["Chalmers University of Technology, Sweden"],"raw_orcid":"https://orcid.org/0000-0003-0477-4540","affiliations":[{"raw_affiliation_string":"Chalmers University of Technology, Sweden","institution_ids":["https://openalex.org/I66862912"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5052702825","display_name":"Mohammad Ali Maleki","orcid":"https://orcid.org/0000-0002-9019-3605"},"institutions":[{"id":"https://openalex.org/I66862912","display_name":"Chalmers University of Technology","ror":"https://ror.org/040wg7k59","country_code":"SE","type":"education","lineage":["https://openalex.org/I66862912"]}],"countries":["SE"],"is_corresponding":false,"raw_author_name":"Mohammad Ali Maleki","raw_affiliation_strings":["Chalmers University of Technology, Sweden"],"raw_orcid":"https://orcid.org/0000-0002-9019-3605","affiliations":[{"raw_affiliation_string":"Chalmers University of Technology, Sweden","institution_ids":["https://openalex.org/I66862912"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5043427541","display_name":"Pedro Trancoso","orcid":"https://orcid.org/0000-0002-2776-9253"},"institutions":[{"id":"https://openalex.org/I66862912","display_name":"Chalmers University of Technology","ror":"https://ror.org/040wg7k59","country_code":"SE","type":"education","lineage":["https://openalex.org/I66862912"]}],"countries":["SE"],"is_corresponding":false,"raw_author_name":"Pedro Trancoso","raw_affiliation_strings":["Chalmers University of Technology, Sweden"],"raw_orcid":"https://orcid.org/0000-0002-2776-9253","affiliations":[{"raw_affiliation_string":"Chalmers University of Technology, Sweden","institution_ids":["https://openalex.org/I66862912"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":1.0845,"has_fulltext":true,"cited_by_count":5,"citation_normalized_percentile":{"value":0.78324505,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":95,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"58","last_page":"67"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9994999766349792,"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":0.9994999766349792,"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/T10627","display_name":"Advanced Image and Video Retrieval Techniques","score":0.9993000030517578,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.9988999962806702,"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/pointwise","display_name":"Pointwise","score":0.8363621234893799},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.7840203046798706},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.703952431678772},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.43434953689575195},{"id":"https://openalex.org/keywords/convolution","display_name":"Convolution (computer science)","score":0.4280489385128021},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.20372161269187927},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.07646864652633667}],"concepts":[{"id":"https://openalex.org/C2777984123","wikidata":"https://www.wikidata.org/wiki/Q9248237","display_name":"Pointwise","level":2,"score":0.8363621234893799},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7840203046798706},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.703952431678772},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.43434953689575195},{"id":"https://openalex.org/C45347329","wikidata":"https://www.wikidata.org/wiki/Q5166604","display_name":"Convolution (computer science)","level":3,"score":0.4280489385128021},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.20372161269187927},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.07646864652633667},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.0}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.1145/3677333.3678153","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3677333.3678153","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3677333.3678153","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"The 53rd International Conference on Parallel Processing Workshops","raw_type":"proceedings-article"},{"id":"pmh:oai:research.chalmers.se:542723","is_oa":true,"landing_page_url":"https://research.chalmers.se/en/publication/542723","pdf_url":"https://research.chalmers.se/publication/543131/file/543131_Fulltext.pdf","source":{"id":"https://openalex.org/S4306402469","display_name":"Chalmers Research (Chalmers University of Technology)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I66862912","host_organization_name":"Chalmers University of Technology","host_organization_lineage":["https://openalex.org/I66862912"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":""},{"id":"pmh:oai:research.chalmers.se:543131","is_oa":false,"landing_page_url":"https://research.chalmers.se/en/publication/2ae70c70-160a-475d-97f3-cb8f00b389de","pdf_url":null,"source":{"id":"https://openalex.org/S4306402469","display_name":"Chalmers Research (Chalmers University of Technology)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I66862912","host_organization_name":"Chalmers University of Technology","host_organization_lineage":["https://openalex.org/I66862912"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":""}],"best_oa_location":{"id":"doi:10.1145/3677333.3678153","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3677333.3678153","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3677333.3678153","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"The 53rd International Conference on Parallel Processing Workshops","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G4255628366","display_name":null,"funder_award_id":"957197","funder_id":"https://openalex.org/F4320320940","funder_display_name":"Stiftelsen f\u00f6r\u00a0Strategisk Forskning"},{"id":"https://openalex.org/G5185518776","display_name":null,"funder_award_id":"957197","funder_id":"https://openalex.org/F4320320300","funder_display_name":"European Commission"},{"id":"https://openalex.org/G7842005466","display_name":null,"funder_award_id":"Horizon 2020","funder_id":"https://openalex.org/F4320320300","funder_display_name":"European Commission"},{"id":"https://openalex.org/G7875605236","display_name":null,"funder_award_id":"101036168","funder_id":"https://openalex.org/F4320320300","funder_display_name":"European Commission"},{"id":"https://openalex.org/G979276111","display_name":null,"funder_award_id":"CHI19-0048","funder_id":"https://openalex.org/F4320320940","funder_display_name":"Stiftelsen f\u00f6r\u00a0Strategisk Forskning"}],"funders":[{"id":"https://openalex.org/F4320320300","display_name":"European Commission","ror":"https://ror.org/00k4n6c32"},{"id":"https://openalex.org/F4320320940","display_name":"Stiftelsen f\u00f6r\u00a0Strategisk Forskning","ror":"https://ror.org/044wr7g58"}],"has_content":{"grobid_xml":false,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4401454756.pdf"},"referenced_works_count":40,"referenced_works":["https://openalex.org/W2015861736","https://openalex.org/W2028499920","https://openalex.org/W2155893237","https://openalex.org/W2172654076","https://openalex.org/W2194775991","https://openalex.org/W2510697685","https://openalex.org/W2531409750","https://openalex.org/W2565125333","https://openalex.org/W2604314403","https://openalex.org/W2618530766","https://openalex.org/W2627042741","https://openalex.org/W2796013597","https://openalex.org/W2804032941","https://openalex.org/W2891946740","https://openalex.org/W2900082550","https://openalex.org/W2935331687","https://openalex.org/W2963125010","https://openalex.org/W2963163009","https://openalex.org/W2963701245","https://openalex.org/W2970971581","https://openalex.org/W2985331920","https://openalex.org/W2997828269","https://openalex.org/W2997929983","https://openalex.org/W3008378296","https://openalex.org/W3090495278","https://openalex.org/W3136416617","https://openalex.org/W3166036439","https://openalex.org/W3167597877","https://openalex.org/W3169997198","https://openalex.org/W3174885745","https://openalex.org/W3190452774","https://openalex.org/W4205146836","https://openalex.org/W4213249026","https://openalex.org/W4214493665","https://openalex.org/W4249932213","https://openalex.org/W4312950730","https://openalex.org/W4321496239","https://openalex.org/W4390658705","https://openalex.org/W6691839725","https://openalex.org/W6795901243"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2748952813","https://openalex.org/W143502885","https://openalex.org/W42113618","https://openalex.org/W2103468410","https://openalex.org/W2480115405","https://openalex.org/W3197542402","https://openalex.org/W1856228368","https://openalex.org/W4361003569","https://openalex.org/W2981421796"],"abstract_inverted_index":{"Depthwise":[0],"and":[1,7,31,58,127,135,143,153,161,182,196,209,236,245],"pointwise":[2,59,128,134],"convolutions":[3,60,137,156],"have":[4,18,37],"fewer":[5,9],"parameters":[6],"perform":[8],"operations":[10],"than":[11,42],"standard":[12,43],"convolutions.":[13,113],"As":[14],"a":[15,38,93,121],"result,":[16],"they":[17,36],"become":[19],"increasingly":[20],"used":[21],"in":[22],"various":[23,223],"compact":[24],"DNNs,":[25],"including":[26],"convolutional":[27],"neural":[28],"networks":[29],"(CNNs)":[30],"vision":[32],"transformers":[33],"(ViTs).":[34],"However,":[35],"lower":[39],"compute-to-memory-access":[40],"ratio":[41],"convolutions,":[44],"making":[45],"their":[46],"memory":[47,64,107,138,177,207],"accesses":[48,178,208],"often":[49],"the":[50,63,88,148,162,176,206,242],"performance":[51],"bottleneck.":[52,66],"This":[53,114],"paper":[54,115],"explores":[55],"fusing":[56,71,91],"depthwise":[57,112,126,136],"to":[61,159,174,203,214,217,234,239],"overcome":[62],"access":[65],"The":[67,76],"focus":[68],"is":[69],"on":[70,74,79,190],"these":[72],"operators":[73],"GPUs.":[75],"prior":[77],"art":[78],"GPU-based":[80],"fusion":[81,152],"suffers":[82],"from":[83],"one":[84],"or":[85,98],"more":[86],"of":[87,123,171,179,205,212,222,232,241],"following:":[89],"(1)":[90],"either":[92],"convolution":[94],"with":[95,151],"an":[96],"element-wise":[97],"multiple":[99],"non-convolutional":[100],"operators,":[101],"(2)":[102],"not":[103,110],"explicitly":[104],"optimizing":[105],"for":[106],"accesses,":[108,139],"(3)":[109],"supporting":[111],"proposes":[116],"Fused":[117],"Convolutional":[118],"Modules":[119],"(FCMs),":[120],"set":[122],"novel":[124],"fused":[125],"GPU":[129,186],"kernels.":[130],"FCMs":[131,200],"significantly":[132],"reduce":[133],"improving":[140],"execution":[141],"time":[142],"energy":[144],"efficiency.":[145],"To":[146],"evaluate":[147],"trade-offs":[149],"associated":[150],"determine":[154],"which":[155],"are":[157,248],"beneficial":[158],"fuse":[160],"optimal":[163],"FCM":[164,183,244],"parameters,":[165],"we":[166],"propose":[167],"FusePlanner.":[168],"FusePlanner":[169,246],"consists":[170],"cost":[172],"models":[173],"estimate":[175],"depthwise,":[180],"pointwise,":[181],"kernels":[184],"given":[185],"characteristics.":[187],"Our":[188],"experiments":[189],"three":[191],"GPUs":[192],"using":[193,225],"representative":[194],"CNNs":[195,224],"ViTs":[197],"demonstrate":[198],"that":[199],"save":[201],"up":[202,213,233,238],"83%":[204],"achieve":[210],"speedups":[211,231],"3.7x":[215],"compared":[216],"cuDNN.":[218],"Complete":[219],"model":[220],"implementations":[221,247],"our":[226],"modules":[227],"outperform":[228],"TVMs\u2019":[229],"achieving":[230],"1.8x":[235],"saving":[237],"two-thirds":[240],"energy.":[243],"open":[249],"source:":[250],"https://github.com/fqararyah/Fusing_DW_and_PW_on_GPUs":[251]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":4}],"updated_date":"2026-06-16T09:24:06.705377","created_date":"2024-08-10T00:00:00"}