{"id":"https://openalex.org/W2535374105","doi":"https://doi.org/10.1145/2966884.2966912","title":"Efficient Large Message Broadcast using NCCL and CUDA-Aware MPI for Deep Learning","display_name":"Efficient Large Message Broadcast using NCCL and CUDA-Aware MPI for Deep Learning","publication_year":2016,"publication_date":"2016-09-25","ids":{"openalex":"https://openalex.org/W2535374105","doi":"https://doi.org/10.1145/2966884.2966912","mag":"2535374105"},"language":"en","primary_location":{"id":"doi:10.1145/2966884.2966912","is_oa":false,"landing_page_url":"https://doi.org/10.1145/2966884.2966912","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 23rd European MPI Users' Group Meeting","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/A5004330728","display_name":"Ammar Ahmad Awan","orcid":"https://orcid.org/0000-0002-6272-3760"},"institutions":[{"id":"https://openalex.org/I52357470","display_name":"The Ohio State University","ror":"https://ror.org/00rs6vg23","country_code":"US","type":"education","lineage":["https://openalex.org/I52357470"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"A. A. Awan","raw_affiliation_strings":["Dept of Computer Science and Engineering, The Ohio State University"],"affiliations":[{"raw_affiliation_string":"Dept of Computer Science and Engineering, The Ohio State University","institution_ids":["https://openalex.org/I52357470"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048039700","display_name":"Khaled Hamidouche","orcid":"https://orcid.org/0000-0003-4836-5335"},"institutions":[{"id":"https://openalex.org/I52357470","display_name":"The Ohio State University","ror":"https://ror.org/00rs6vg23","country_code":"US","type":"education","lineage":["https://openalex.org/I52357470"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"K. Hamidouche","raw_affiliation_strings":["Dept of Computer Science and Engineering, The Ohio State University"],"affiliations":[{"raw_affiliation_string":"Dept of Computer Science and Engineering, The Ohio State University","institution_ids":["https://openalex.org/I52357470"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5112002767","display_name":"Akshay Venkatesh","orcid":null},"institutions":[{"id":"https://openalex.org/I52357470","display_name":"The Ohio State University","ror":"https://ror.org/00rs6vg23","country_code":"US","type":"education","lineage":["https://openalex.org/I52357470"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"A. Venkatesh","raw_affiliation_strings":["Dept of Computer Science and Engineering, The Ohio State University"],"affiliations":[{"raw_affiliation_string":"Dept of Computer Science and Engineering, The Ohio State University","institution_ids":["https://openalex.org/I52357470"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5112459975","display_name":"D.K. Panda","orcid":null},"institutions":[{"id":"https://openalex.org/I52357470","display_name":"The Ohio State University","ror":"https://ror.org/00rs6vg23","country_code":"US","type":"education","lineage":["https://openalex.org/I52357470"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"D. K. Panda","raw_affiliation_strings":["Dept of Computer Science and Engineering, The Ohio State University"],"affiliations":[{"raw_affiliation_string":"Dept of Computer Science and Engineering, The Ohio State University","institution_ids":["https://openalex.org/I52357470"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5004330728"],"corresponding_institution_ids":["https://openalex.org/I52357470"],"apc_list":null,"apc_paid":null,"fwci":3.0446,"has_fulltext":false,"cited_by_count":51,"citation_normalized_percentile":{"value":0.94553471,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":90,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"15","last_page":"22"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.998199999332428,"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":0.998199999332428,"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.998199999332428,"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/T11181","display_name":"Advanced Data Storage Technologies","score":0.9980000257492065,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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":0.9009594321250916},{"id":"https://openalex.org/keywords/cuda","display_name":"CUDA","score":0.7873253226280212},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.645907461643219},{"id":"https://openalex.org/keywords/latency","display_name":"Latency (audio)","score":0.5766467452049255},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.5410010814666748},{"id":"https://openalex.org/keywords/general-purpose-computing-on-graphics-processing-units","display_name":"General-purpose computing on graphics processing units","score":0.48932236433029175},{"id":"https://openalex.org/keywords/message-passing","display_name":"Message passing","score":0.47714635729789734},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.46580085158348083},{"id":"https://openalex.org/keywords/remote-direct-memory-access","display_name":"Remote direct memory access","score":0.46146658062934875},{"id":"https://openalex.org/keywords/computer-architecture","display_name":"Computer architecture","score":0.459451287984848},{"id":"https://openalex.org/keywords/low-latency","display_name":"Low latency (capital markets)","score":0.43948280811309814},{"id":"https://openalex.org/keywords/gpu-cluster","display_name":"GPU cluster","score":0.4253781735897064},{"id":"https://openalex.org/keywords/operating-system","display_name":"Operating system","score":0.16282489895820618},{"id":"https://openalex.org/keywords/computer-network","display_name":"Computer network","score":0.15979132056236267},{"id":"https://openalex.org/keywords/telecommunications","display_name":"Telecommunications","score":0.08612290024757385}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.9009594321250916},{"id":"https://openalex.org/C2778119891","wikidata":"https://www.wikidata.org/wiki/Q477690","display_name":"CUDA","level":2,"score":0.7873253226280212},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.645907461643219},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.5766467452049255},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.5410010814666748},{"id":"https://openalex.org/C50630238","wikidata":"https://www.wikidata.org/wiki/Q971505","display_name":"General-purpose computing on graphics processing units","level":3,"score":0.48932236433029175},{"id":"https://openalex.org/C854659","wikidata":"https://www.wikidata.org/wiki/Q1859284","display_name":"Message passing","level":2,"score":0.47714635729789734},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.46580085158348083},{"id":"https://openalex.org/C130795937","wikidata":"https://www.wikidata.org/wiki/Q2561570","display_name":"Remote direct memory access","level":2,"score":0.46146658062934875},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.459451287984848},{"id":"https://openalex.org/C46637626","wikidata":"https://www.wikidata.org/wiki/Q6693015","display_name":"Low latency (capital markets)","level":2,"score":0.43948280811309814},{"id":"https://openalex.org/C2781335571","wikidata":"https://www.wikidata.org/wiki/Q2633544","display_name":"GPU cluster","level":3,"score":0.4253781735897064},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.16282489895820618},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.15979132056236267},{"id":"https://openalex.org/C76155785","wikidata":"https://www.wikidata.org/wiki/Q418","display_name":"Telecommunications","level":1,"score":0.08612290024757385},{"id":"https://openalex.org/C86803240","wikidata":"https://www.wikidata.org/wiki/Q420","display_name":"Biology","level":0,"score":0.0},{"id":"https://openalex.org/C205649164","wikidata":"https://www.wikidata.org/wiki/Q1071","display_name":"Geography","level":0,"score":0.0},{"id":"https://openalex.org/C151730666","wikidata":"https://www.wikidata.org/wiki/Q7205","display_name":"Paleontology","level":1,"score":0.0},{"id":"https://openalex.org/C13280743","wikidata":"https://www.wikidata.org/wiki/Q131089","display_name":"Geodesy","level":1,"score":0.0},{"id":"https://openalex.org/C21442007","wikidata":"https://www.wikidata.org/wiki/Q1027879","display_name":"Graphics","level":2,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/2966884.2966912","is_oa":false,"landing_page_url":"https://doi.org/10.1145/2966884.2966912","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 23rd European MPI Users' Group Meeting","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":17,"referenced_works":["https://openalex.org/W110182411","https://openalex.org/W1637731592","https://openalex.org/W1686810756","https://openalex.org/W1965548909","https://openalex.org/W2042670967","https://openalex.org/W2105549957","https://openalex.org/W2108598243","https://openalex.org/W2131940306","https://openalex.org/W2146375161","https://openalex.org/W2155893237","https://openalex.org/W2163605009","https://openalex.org/W2170135819","https://openalex.org/W2294872507","https://openalex.org/W2604272474","https://openalex.org/W2618530766","https://openalex.org/W2962911728","https://openalex.org/W3105781365"],"related_works":["https://openalex.org/W2099148634","https://openalex.org/W2751263050","https://openalex.org/W1502869929","https://openalex.org/W2049347805","https://openalex.org/W2104094072","https://openalex.org/W4311118888","https://openalex.org/W1531092195","https://openalex.org/W3105798355","https://openalex.org/W4391791482","https://openalex.org/W2917381094"],"abstract_inverted_index":{"Emerging":[0],"paradigms":[1,25],"like":[2,149],"High":[3],"Performance":[4],"Data":[5],"Analytics":[6],"(HPDA)":[7],"and":[8,46,83,151,216,229],"Deep":[9],"Learning":[10],"(DL)":[11],"pose":[12],"at":[13],"least":[14],"two":[15],"new":[16,137,167,186],"design":[17,205,215],"challenges":[18,138,206],"for":[19,30,75,86,193,207,247],"existing":[20,145],"MPI":[21,62,95,147],"runtimes.":[22],"First,":[23],"these":[24,136,185],"require":[26],"an":[27],"efficient":[28],"support":[29],"communicating":[31],"unusually":[32],"large":[33,125],"messages":[34],"across":[35],"processes.":[36],"And":[37],"second,":[38],"the":[39,91,119,141,161,173,180,204,212,250],"communication":[40,73,127,158,168,194],"buffers":[41,190],"used":[42,192],"by":[43,139,164],"HPDA":[44],"applications":[45],"DL":[47,236],"frameworks":[48],"generally":[49],"reside":[50],"on":[51],"a":[52,97,166,222,227,230],"GPU's":[53],"memory.":[54],"In":[55,131,218],"this":[56,132,178],"context,":[57],"we":[58,134,220],"observe":[59],"that":[60,84,121,183],"conventional":[61],"runtimes":[63,148],"have":[64],"been":[65,102],"optimized":[66],"over":[67],"decades":[68],"to":[69,104,156,241],"achieve":[70],"lowest":[71],"possible":[72],"latency":[74,159],"relatively":[76],"smaller":[77],"message":[78,126,196],"sizes":[79,197],"(up-to":[80],"1":[81],"Megabyte)":[82],"too":[85],"CPU":[87],"memory":[88],"buffers.":[89,130],"With":[90],"advent":[92],"of":[93,99,107,118,128,160,175,200,214,233],"CUDA-Aware":[94,146,231],"runtimes,":[96],"lot":[98],"research":[100],"has":[101],"conducted":[103],"improve":[105,157],"performance":[106,142,224],"GPU":[108,129,189],"buffer":[109],"based":[110],"communication.":[111],"However,":[112],"little":[113],"exists":[114],"in":[115,144,244],"current":[116],"state":[117],"art":[120],"deals":[122],"with":[123,195,211],"very":[124],"paper,":[133],"investigate":[135],"analyzing":[140],"bottlenecks":[143],"MVAPICH2-GDR,":[150],"propose":[152],"hierarchical":[153,252],"collective":[154],"designs":[155],"MPI_Bcast":[162,253],"primitive":[163],"exploiting":[165],"library":[169],"called":[170],"NCCL.":[171],"To":[172],"best":[174],"our":[176,208],"knowledge,":[177],"is":[179],"first":[181],"work":[182,209],"addresses":[184],"requirements":[187],"where":[188],"are":[191],"surpassing":[198],"hundreds":[199],"megabytes.":[201],"We":[202,238],"highlight":[203],"along":[210],"details":[213],"implementation.":[217],"addition,":[219],"provide":[221],"comprehensive":[223],"evaluation":[225],"using":[226,249],"Micro-benchmark":[228],"adaptation":[232],"Microsoft":[234],"CNTK":[235,248],"framework.":[237],"report":[239],"up":[240],"47%":[242],"improvement":[243],"training":[245],"time":[246],"proposed":[251],"design.":[254]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":7},{"year":2024,"cited_by_count":1},{"year":2023,"cited_by_count":6},{"year":2022,"cited_by_count":4},{"year":2021,"cited_by_count":7},{"year":2020,"cited_by_count":7},{"year":2019,"cited_by_count":2},{"year":2018,"cited_by_count":11},{"year":2017,"cited_by_count":4},{"year":2016,"cited_by_count":1}],"updated_date":"2026-03-06T13:50:29.536080","created_date":"2025-10-10T00:00:00"}