{"id":"https://openalex.org/W7235877320","doi":"https://doi.org/10.5281/zenodo.24331218","title":"Tensor Core Matrix Multiplication on the NVIDIA T4: A WMMA FP16 GEMM with Double Buffering, Compared Against cuBLAS","display_name":"Tensor Core Matrix Multiplication on the NVIDIA T4: A WMMA FP16 GEMM with Double Buffering, Compared Against cuBLAS","alternate_titles":[],"publication_year":2026,"publication_date":"2026-11-18","ids":{"openalex":"https://openalex.org/W7235877320","doi":"https://doi.org/10.5281/zenodo.24331218"},"language":"en","primary_location":{"id":"doi:10.5281/zenodo.24331218","doi":"https://doi.org/10.5281/zenodo.24331218","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24331218","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article-journal"},"parent_works":[],"lineage":["https://openalex.org/W7235877320"],"child_works_count":0,"type":"article","is_traditional":true,"indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.5281/zenodo.24331218","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137803823","display_name":"Syed Tarteel Ejaz","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":true,"contributor_roles":[],"raw_author_name":"EJAZ, SYED TARTEEL","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5137803823"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"reviewed_by_count":0,"citation_normalized_percentile":{"value":0.0,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_pulled":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9990000128746033,"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":0.9990000128746033,"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/T11697","display_name":"Numerical Methods and Algorithms","score":0.0003000000142492354,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/T10792","display_name":"Matrix Theory and Algorithms","score":0.00019999999494757503,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/tensor-cores","display_name":"tensor cores","score":0.6470000147819519},{"id":"https://openalex.org/keywords/matrix-multiplication","display_name":"matrix multiplication","score":0.9269999861717224},{"id":"https://openalex.org/keywords/gemm","display_name":"GEMM","score":0.9340000152587891},{"id":"https://openalex.org/keywords/double-buffering","display_name":"double buffering","score":0.7200000286102295},{"id":"https://openalex.org/keywords/fp16","display_name":"FP16","score":0.699999988079071},{"id":"https://openalex.org/keywords/cublas","display_name":"cuBLAS","score":0.7559999823570251},{"id":"https://openalex.org/keywords/cuda","display_name":"CUDA","score":0.6069999933242798},{"id":"https://openalex.org/keywords/deep-learning-acceleration","display_name":"deep learning acceleration","score":0.6129999756813049}],"concepts":[{"id":"https://openalex.org/C17349429","wikidata":"https://www.wikidata.org/wiki/Q1049914","display_name":"Matrix multiplication","level":3,"score":0.6237999796867371},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5932999849319458},{"id":"https://openalex.org/C2778119891","wikidata":"https://www.wikidata.org/wiki/Q477690","display_name":"CUDA","level":2,"score":0.5741999745368958},{"id":"https://openalex.org/C2780595030","wikidata":"https://www.wikidata.org/wiki/Q3860309","display_name":"Multiplication (music)","level":2,"score":0.5734000205993652},{"id":"https://openalex.org/C459310","wikidata":"https://www.wikidata.org/wiki/Q117801","display_name":"Computational science","level":1,"score":0.5566999912261963},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.5392000079154968},{"id":"https://openalex.org/C155281189","wikidata":"https://www.wikidata.org/wiki/Q3518150","display_name":"Tensor (intrinsic definition)","level":2,"score":0.5231999754905701},{"id":"https://openalex.org/C106487976","wikidata":"https://www.wikidata.org/wiki/Q685816","display_name":"Matrix (chemical analysis)","level":2,"score":0.4878999888896942},{"id":"https://openalex.org/C151319957","wikidata":"https://www.wikidata.org/wiki/Q752739","display_name":"Asynchronous communication","level":2,"score":0.4302999973297119},{"id":"https://openalex.org/C78766204","wikidata":"https://www.wikidata.org/wiki/Q555032","display_name":"Multi-core processor","level":2,"score":0.4131999909877777},{"id":"https://openalex.org/C35912277","wikidata":"https://www.wikidata.org/wiki/Q1243369","display_name":"Double-precision floating-point format","level":3,"score":0.39750000834465027},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.367000013589859},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.3517000079154968},{"id":"https://openalex.org/C94375191","wikidata":"https://www.wikidata.org/wiki/Q11205","display_name":"Arithmetic","level":1,"score":0.3075999915599823},{"id":"https://openalex.org/C2164484","wikidata":"https://www.wikidata.org/wiki/Q5170150","display_name":"Core (optical fiber)","level":2,"score":0.29350000619888306},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.2851000130176544},{"id":"https://openalex.org/C2988995629","wikidata":"https://www.wikidata.org/wiki/Q2915729","display_name":"Matrix algebra","level":3,"score":0.28060001134872437},{"id":"https://openalex.org/C2780365336","wikidata":"https://www.wikidata.org/wiki/Q25047934","display_name":"Single-core","level":2,"score":0.2711000144481659},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.25859999656677246}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.5281/zenodo.24331218","doi":"https://doi.org/10.5281/zenodo.24331218","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24331218","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article-journal"},{"id":"doi:10.5281/zenodo.24331219","doi":"https://doi.org/10.5281/zenodo.24331219","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24331219","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article-journal"}],"best_oa_location":{"id":"doi:10.5281/zenodo.24331218","doi":"https://doi.org/10.5281/zenodo.24331218","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24331218","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article-journal"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7168398349","https://openalex.org/W3034986667","https://openalex.org/W4319917399","https://openalex.org/W3043303806","https://openalex.org/W4283812690","https://openalex.org/W2964017942","https://openalex.org/W7165868854","https://openalex.org/W4389386944","https://openalex.org/W2091371833","https://openalex.org/W4309672181"],"abstract_inverted_index":{"Matrix":[0],"multiplication":[1],"(GEMM)":[2],"is":[3],"the":[4,14,32,57,62,103,109,113,126],"main":[5],"cost":[6],"in":[7,41,78],"deep-learning":[8],"workloads,":[9],"and":[10,38,50,70,129],"Tensor":[11],"Cores":[12],"are":[13],"hardware":[15,117],"NVIDIA":[16],"GPUs":[17],"provide":[18],"to":[19],"speed":[20],"it":[21],"up.":[22],"This":[23],"report":[24,107],"describes":[25],"a":[26,44,51,73,122],"from-scratch":[27],"CUDA":[28],"GEMM":[29],"that":[30,54,112],"uses":[31],"WMMA":[33],"API":[34],"with":[35,46,60],"FP16":[36],"inputs":[37],"FP32":[39],"accumulation,":[40],"two":[42],"forms:":[43],"baseline":[45],"one":[47],"shared-memory":[48],"buffer":[49],"double-buffered":[52],"version":[53],"overlaps":[55],"loading":[56],"next":[58],"tile":[59],"computing":[61],"current":[63],"one.":[64],"Both":[65],"were":[66],"verified":[67],"against":[68],"cuBLAS":[69,91],"benchmarked":[71],"on":[72],"Tesla":[74],"T4":[75,114],"(Turing,":[76],"sm_75)":[77],"Google":[79],"Colab":[80],"at":[81],"three":[82],"matrix":[83],"sizes.":[84],"The":[85,106],"kernels":[86],"reach":[87],"about":[88],"63-67%":[89],"of":[90],"throughput.":[92],"On":[93],"this":[94],"GPU,":[95],"double":[96],"buffering":[97],"gave":[98],"no":[99,116],"measurable":[100],"gain":[101],"over":[102],"single-buffer":[104],"baseline.":[105],"explains":[108],"design,":[110],"states":[111],"has":[115],"asynchronous":[118],"copy":[119],"(cp.async),":[120],"gives":[121],"likely":[123],"reason":[124],"for":[125],"flat":[127],"result,":[128],"lists":[130],"what":[131],"would":[132],"be":[133],"tried":[134],"next.":[135],"Source":[136],"code:":[137],"https://github.com/19YearsOlds/61.-Tensor-Core-GEMM-WMMA-FP16-on-NVIDIA-T4":[138]},"cited_by_api_url":"https://api.openalex.org/works?filter=cites:W7235877320","counts_by_year":[],"updated_date":"2026-11-20T13:11:24.684918","created_date":"2026-11-19T00:00:00"}
