{"id":"https://openalex.org/W7131104738","doi":"https://doi.org/10.1109/cgo68049.2026.11394849","title":"Tawa: Automatic Warp Specialization for Modern GPUs with Asynchronous References","display_name":"Tawa: Automatic Warp Specialization for Modern GPUs with Asynchronous References","publication_year":2026,"publication_date":"2026-01-31","ids":{"openalex":"https://openalex.org/W7131104738","doi":"https://doi.org/10.1109/cgo68049.2026.11394849"},"language":null,"primary_location":{"id":"doi:10.1109/cgo68049.2026.11394849","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cgo68049.2026.11394849","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2026 IEEE/ACM International Symposium on Code Generation and Optimization (CGO)","raw_type":"proceedings-article"},"type":"conference-paper","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/A5008694159","display_name":"Hongzheng Chen","orcid":"https://orcid.org/0000-0002-6617-0075"},"institutions":[{"id":"https://openalex.org/I205783295","display_name":"Cornell University","ror":"https://ror.org/05bnh6r87","country_code":"US","type":"education","lineage":["https://openalex.org/I205783295"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hongzheng Chen","raw_affiliation_strings":["Cornell University,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Cornell University,USA","institution_ids":["https://openalex.org/I205783295"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126590839","display_name":"Bin Fan","orcid":null},"institutions":[{"id":"https://openalex.org/I4210127875","display_name":"Nvidia (United States)","ror":"https://ror.org/03jdj4y14","country_code":"US","type":"company","lineage":["https://openalex.org/I4210127875"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Bin Fan","raw_affiliation_strings":["NVIDIA,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA,USA","institution_ids":["https://openalex.org/I4210127875"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5005200478","display_name":"Alexander J. Collins","orcid":null},"institutions":[{"id":"https://openalex.org/I1304085615","display_name":"Nvidia (United Kingdom)","ror":"https://ror.org/02kr42612","country_code":"GB","type":"company","lineage":["https://openalex.org/I1304085615","https://openalex.org/I4210127875"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Alexander Collins","raw_affiliation_strings":["NVIDIA,United Kingdom"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA,United Kingdom","institution_ids":["https://openalex.org/I1304085615"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5075117421","display_name":"Bastian Hagedorn","orcid":"https://orcid.org/0000-0003-1833-5514"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bastian Hagedorn","raw_affiliation_strings":["NVIDIA,Germany"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA,Germany","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5090222029","display_name":"Evghenii Gaburov","orcid":null},"institutions":[{"id":"https://openalex.org/I4210127875","display_name":"Nvidia (United States)","ror":"https://ror.org/03jdj4y14","country_code":"US","type":"company","lineage":["https://openalex.org/I4210127875"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Evghenii Gaburov","raw_affiliation_strings":["NVIDIA,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA,USA","institution_ids":["https://openalex.org/I4210127875"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5046294658","display_name":"Masahiro Masuda","orcid":"https://orcid.org/0009-0006-1054-3836"},"institutions":[{"id":"https://openalex.org/I4210127875","display_name":"Nvidia (United States)","ror":"https://ror.org/03jdj4y14","country_code":"US","type":"company","lineage":["https://openalex.org/I4210127875"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Masahiro Masuda","raw_affiliation_strings":["NVIDIA,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA,USA","institution_ids":["https://openalex.org/I4210127875"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5065900862","display_name":"Matthew Brookhart","orcid":null},"institutions":[{"id":"https://openalex.org/I4210127875","display_name":"Nvidia (United States)","ror":"https://ror.org/03jdj4y14","country_code":"US","type":"company","lineage":["https://openalex.org/I4210127875"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Matthew Brookhart","raw_affiliation_strings":["NVIDIA,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA,USA","institution_ids":["https://openalex.org/I4210127875"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126622376","display_name":"Chris Sullivan","orcid":null},"institutions":[{"id":"https://openalex.org/I4210127875","display_name":"Nvidia (United States)","ror":"https://ror.org/03jdj4y14","country_code":"US","type":"company","lineage":["https://openalex.org/I4210127875"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Chris Sullivan","raw_affiliation_strings":["NVIDIA,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA,USA","institution_ids":["https://openalex.org/I4210127875"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5016696594","display_name":"Jason Knight","orcid":null},"institutions":[{"id":"https://openalex.org/I4210127875","display_name":"Nvidia (United States)","ror":"https://ror.org/03jdj4y14","country_code":"US","type":"company","lineage":["https://openalex.org/I4210127875"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jason Knight","raw_affiliation_strings":["NVIDIA,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA,USA","institution_ids":["https://openalex.org/I4210127875"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126584068","display_name":"Zhiru Zhang","orcid":null},"institutions":[{"id":"https://openalex.org/I205783295","display_name":"Cornell University","ror":"https://ror.org/05bnh6r87","country_code":"US","type":"education","lineage":["https://openalex.org/I205783295"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zhiru Zhang","raw_affiliation_strings":["Cornell University,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Cornell University,USA","institution_ids":["https://openalex.org/I205783295"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5104909714","display_name":"Vinod K. Grover","orcid":"https://orcid.org/0000-0003-0115-3896"},"institutions":[{"id":"https://openalex.org/I4210127875","display_name":"Nvidia (United States)","ror":"https://ror.org/03jdj4y14","country_code":"US","type":"company","lineage":["https://openalex.org/I4210127875"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Vinod Grover","raw_affiliation_strings":["NVIDIA,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"NVIDIA,USA","institution_ids":["https://openalex.org/I4210127875"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"255","last_page":"267"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.5246999859809875,"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":0.5246999859809875,"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/T12292","display_name":"Graph Theory and Algorithms","score":0.10490000247955322,"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/T10481","display_name":"Computer Graphics and Visualization Techniques","score":0.05119999870657921,"subfield":{"id":"https://openalex.org/subfields/1704","display_name":"Computer Graphics and Computer-Aided Design"},"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/dataflow","display_name":"Dataflow","score":0.857200026512146},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.8374000191688538},{"id":"https://openalex.org/keywords/asynchronous-communication","display_name":"Asynchronous communication","score":0.7657999992370605},{"id":"https://openalex.org/keywords/compiler","display_name":"Compiler","score":0.6736999750137329},{"id":"https://openalex.org/keywords/kernel","display_name":"Kernel (algebra)","score":0.6328999996185303},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.5325000286102295},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.4708000123500824},{"id":"https://openalex.org/keywords/software","display_name":"Software","score":0.45170000195503235},{"id":"https://openalex.org/keywords/programming-paradigm","display_name":"Programming paradigm","score":0.4207000136375427}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8858000040054321},{"id":"https://openalex.org/C96324660","wikidata":"https://www.wikidata.org/wiki/Q205446","display_name":"Dataflow","level":2,"score":0.857200026512146},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.8374000191688538},{"id":"https://openalex.org/C151319957","wikidata":"https://www.wikidata.org/wiki/Q752739","display_name":"Asynchronous communication","level":2,"score":0.7657999992370605},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.6858999729156494},{"id":"https://openalex.org/C169590947","wikidata":"https://www.wikidata.org/wiki/Q47506","display_name":"Compiler","level":2,"score":0.6736999750137329},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.6328999996185303},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5325000286102295},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.4708000123500824},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.45170000195503235},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.44589999318122864},{"id":"https://openalex.org/C34165917","wikidata":"https://www.wikidata.org/wiki/Q188267","display_name":"Programming paradigm","level":2,"score":0.4207000136375427},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.37630000710487366},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.3709000051021576},{"id":"https://openalex.org/C2778119891","wikidata":"https://www.wikidata.org/wiki/Q477690","display_name":"CUDA","level":2,"score":0.36320000886917114},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.3321000039577484},{"id":"https://openalex.org/C133162039","wikidata":"https://www.wikidata.org/wiki/Q1061077","display_name":"Code generation","level":3,"score":0.32499998807907104},{"id":"https://openalex.org/C78766204","wikidata":"https://www.wikidata.org/wiki/Q555032","display_name":"Multi-core processor","level":2,"score":0.3163999915122986},{"id":"https://openalex.org/C137364921","wikidata":"https://www.wikidata.org/wiki/Q27929394","display_name":"Parallel programming model","level":3,"score":0.2732999920845032},{"id":"https://openalex.org/C202491316","wikidata":"https://www.wikidata.org/wiki/Q272683","display_name":"Instruction set","level":2,"score":0.25529998540878296},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.2551000118255615},{"id":"https://openalex.org/C195701839","wikidata":"https://www.wikidata.org/wiki/Q4008398","display_name":"Partial evaluation","level":2,"score":0.2547000050544739},{"id":"https://openalex.org/C2776834041","wikidata":"https://www.wikidata.org/wiki/Q25346349","display_name":"Execution model","level":2,"score":0.25429999828338623},{"id":"https://openalex.org/C19024347","wikidata":"https://www.wikidata.org/wiki/Q211496","display_name":"High-level programming language","level":3,"score":0.2517000138759613},{"id":"https://openalex.org/C76518257","wikidata":"https://www.wikidata.org/wiki/Q271680","display_name":"Software framework","level":5,"score":0.2515999972820282}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/cgo68049.2026.11394849","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cgo68049.2026.11394849","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2026 IEEE/ACM International Symposium on Code Generation and Optimization (CGO)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"display_name":"Affordable and clean energy","score":0.6930969953536987,"id":"https://metadata.un.org/sdg/7"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":21,"referenced_works":["https://openalex.org/W2049875313","https://openalex.org/W2171399035","https://openalex.org/W2804032941","https://openalex.org/W2954698171","https://openalex.org/W3090487264","https://openalex.org/W3122286897","https://openalex.org/W3138798301","https://openalex.org/W4211118386","https://openalex.org/W4321782107","https://openalex.org/W4327930469","https://openalex.org/W4385245566","https://openalex.org/W4393145114","https://openalex.org/W4393407047","https://openalex.org/W4393949386","https://openalex.org/W4394661856","https://openalex.org/W4394998532","https://openalex.org/W4394998694","https://openalex.org/W4408903581","https://openalex.org/W4411260261","https://openalex.org/W4415797154","https://openalex.org/W7125944635"],"related_works":[],"abstract_inverted_index":{"Modern":[0],"GPUs":[1,128],"feature":[2],"specialized":[3],"hardware":[4,100,138],"units":[5],"that":[6,54,70,134],"enable":[7],"high-performance,":[8,73],"asynchronous":[9,90],"dataflow":[10,116],"execution.":[11],"However,":[12],"the":[13,35,114,163,166],"conventional":[14],"SIMT":[15],"programming":[16,175],"model":[17],"is":[18,34,55,85],"fundamentally":[19],"misaligned":[20],"with":[21,172],"this":[22,62,103],"task-parallel":[23],"hardware,":[24],"creating":[25],"a":[26,77,86],"significant":[27],"programmability":[28],"gap.":[29],"While":[30],"hardware-level":[31],"warp":[32],"specialization":[33],"key":[36],"to":[37,44,82,142],"unlocking":[38],"peak":[39],"performance,":[40],"it":[41],"forces":[42],"developers":[43,119],"manually":[45],"orchestrate":[46],"complex,":[47],"low-level":[48,99],"communication":[49,96],"and":[50,58,112,161],"software":[51],"pipelines\u2014a":[52],"process":[53],"labor-intensive,":[56],"error-prone,":[57],"unsustainable.":[59],"To":[60],"address":[61],"challenge,":[63],"we":[64],"present":[65],"Tawa,":[66],"an":[67],"automated":[68],"compiler":[69],"systematically":[71],"generates":[72],"warp-specialized":[74],"code":[75],"from":[76],"high-level,":[78],"tile-based":[79],"program.":[80],"Central":[81],"our":[83],"approach":[84],"novel":[87],"IR":[88],"abstraction,":[89,104],"references":[91],"(aref),":[92],"which":[93],"expresses":[94],"warp-level":[95],"without":[97],"exposing":[98],"details.":[101],"Using":[102],"Tawa":[105,135,155],"automatically":[106],"partitions":[107],"programs":[108],"into":[109],"producer-consumer":[110],"roles":[111],"manages":[113],"intricate":[115],"pipeline,":[117],"relieving":[118],"of":[120,165],"invasive":[121],"kernel":[122,171],"rewriting.":[123],"Evaluation":[124],"on":[125],"NVIDIA":[126],"H100":[127],"across":[129],"representative":[130],"LLM":[131],"kernels":[132],"shows":[133],"delivers":[136],"high":[137],"utilization,":[139],"achieving":[140],"up":[141],"1.1":[143],"\u00d7":[144],"speedup":[145,158],"over":[146,159],"highly":[147],"optimized":[148],"cuBLAS":[149],"GEMM":[150],"kernels.":[151],"For":[152],"attention":[153],"workloads,":[154],"attains":[156],"1.2\u00d7":[157],"Triton":[160],"matches":[162],"performance":[164],"hand-optimized":[167],"CUTLASS":[168],"C++":[169],"FlashAttention-3":[170],"far":[173],"less":[174],"effort.":[176]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2026-02-24T00:00:00"}