{"id":"https://openalex.org/W7131072561","doi":"https://doi.org/10.1109/cgo68049.2026.11395227","title":"Progressive Low-Precision Approximation of Tensor Operators on GPUs: Enabling Greater Trade-Offs between Performance and Accuracy","display_name":"Progressive Low-Precision Approximation of Tensor Operators on GPUs: Enabling Greater Trade-Offs between Performance and Accuracy","publication_year":2026,"publication_date":"2026-01-31","ids":{"openalex":"https://openalex.org/W7131072561","doi":"https://doi.org/10.1109/cgo68049.2026.11395227"},"language":null,"primary_location":{"id":"doi:10.1109/cgo68049.2026.11395227","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cgo68049.2026.11395227","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":"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/A5044452017","display_name":"Fan Luo","orcid":"https://orcid.org/0000-0002-7770-1417"},"institutions":[{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Fan Luo","raw_affiliation_strings":["Chinese Academy of Sciences,State Key Lab of Processors, Institute of Computing Technology,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences,State Key Lab of Processors, Institute of Computing Technology,China","institution_ids":["https://openalex.org/I4210090176"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126622002","display_name":"Guangli Li","orcid":null},"institutions":[{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Guangli Li","raw_affiliation_strings":["Chinese Academy of Sciences,State Key Lab of Processors, Institute of Computing Technology,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences,State Key Lab of Processors, Institute of Computing Technology,China","institution_ids":["https://openalex.org/I4210090176"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126585790","display_name":"Zhaoyang Hao","orcid":null},"institutions":[{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhaoyang Hao","raw_affiliation_strings":["Chinese Academy of Sciences,State Key Lab of Processors, Institute of Computing Technology,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences,State Key Lab of Processors, Institute of Computing Technology,China","institution_ids":["https://openalex.org/I4210090176"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126619375","display_name":"Xueying Wang","orcid":null},"institutions":[{"id":"https://openalex.org/I139759216","display_name":"Beijing University of Posts and Telecommunications","ror":"https://ror.org/04w9fbh59","country_code":"CN","type":"education","lineage":["https://openalex.org/I139759216"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xueying Wang","raw_affiliation_strings":["Beijing University of Posts and Telecommunications,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beijing University of Posts and Telecommunications,China","institution_ids":["https://openalex.org/I139759216"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Xiaobing Feng","orcid":null},"institutions":[{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xiaobing Feng","raw_affiliation_strings":["Chinese Academy of Sciences,State Key Lab of Processors, Institute of Computing Technology,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences,State Key Lab of Processors, Institute of Computing Technology,China","institution_ids":["https://openalex.org/I4210090176"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5086633294","display_name":"Huimin Cui","orcid":"https://orcid.org/0000-0002-2491-7679"},"institutions":[{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Huimin Cui","raw_affiliation_strings":["Chinese Academy of Sciences,State Key Lab of Processors, Institute of Computing Technology,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences,State Key Lab of Processors, Institute of Computing Technology,China","institution_ids":["https://openalex.org/I4210090176"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5126636239","display_name":"Jingling Xue","orcid":null},"institutions":[{"id":"https://openalex.org/I31746571","display_name":"UNSW Sydney","ror":"https://ror.org/03r8z3t63","country_code":"AU","type":"education","lineage":["https://openalex.org/I31746571"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Jingling Xue","raw_affiliation_strings":["University of New South Wales,School of Computer Science and Engineering,Australia"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of New South Wales,School of Computer Science and Engineering,Australia","institution_ids":["https://openalex.org/I31746571"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":7,"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":0.28037383,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"670","last_page":"682"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12303","display_name":"Tensor decomposition and applications","score":0.46950000524520874,"subfield":{"id":"https://openalex.org/subfields/2605","display_name":"Computational Mathematics"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T12303","display_name":"Tensor decomposition and applications","score":0.46950000524520874,"subfield":{"id":"https://openalex.org/subfields/2605","display_name":"Computational Mathematics"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.12999999523162842,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.04569999873638153,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"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/exploit","display_name":"Exploit","score":0.6298999786376953},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5906000137329102},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.5472999811172485},{"id":"https://openalex.org/keywords/tensor","display_name":"Tensor (intrinsic definition)","score":0.5149000287055969},{"id":"https://openalex.org/keywords/space","display_name":"Space (punctuation)","score":0.44839999079704285},{"id":"https://openalex.org/keywords/evolutionary-algorithm","display_name":"Evolutionary algorithm","score":0.3125999867916107},{"id":"https://openalex.org/keywords/operator","display_name":"Operator (biology)","score":0.3061000108718872}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6751000285148621},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.6298999786376953},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5906000137329102},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5472999811172485},{"id":"https://openalex.org/C155281189","wikidata":"https://www.wikidata.org/wiki/Q3518150","display_name":"Tensor (intrinsic definition)","level":2,"score":0.5149000287055969},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4717000126838684},{"id":"https://openalex.org/C2778572836","wikidata":"https://www.wikidata.org/wiki/Q380933","display_name":"Space (punctuation)","level":2,"score":0.44839999079704285},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.4011000096797943},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3874000012874603},{"id":"https://openalex.org/C459310","wikidata":"https://www.wikidata.org/wiki/Q117801","display_name":"Computational science","level":1,"score":0.3797999918460846},{"id":"https://openalex.org/C159149176","wikidata":"https://www.wikidata.org/wiki/Q14489129","display_name":"Evolutionary algorithm","level":2,"score":0.3125999867916107},{"id":"https://openalex.org/C17020691","wikidata":"https://www.wikidata.org/wiki/Q139677","display_name":"Operator (biology)","level":5,"score":0.3061000108718872},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.2924000024795532},{"id":"https://openalex.org/C148764684","wikidata":"https://www.wikidata.org/wiki/Q621751","display_name":"Approximation algorithm","level":2,"score":0.29100000858306885},{"id":"https://openalex.org/C51255310","wikidata":"https://www.wikidata.org/wiki/Q1163016","display_name":"Tensor product","level":2,"score":0.2874000072479248},{"id":"https://openalex.org/C83581934","wikidata":"https://www.wikidata.org/wiki/Q527381","display_name":"Arbitrary-precision arithmetic","level":2,"score":0.2791999876499176},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.27459999918937683},{"id":"https://openalex.org/C2989134064","wikidata":"https://www.wikidata.org/wiki/Q288510","display_name":"Execution time","level":2,"score":0.2743000090122223},{"id":"https://openalex.org/C105902424","wikidata":"https://www.wikidata.org/wiki/Q1197129","display_name":"Evolutionary computation","level":2,"score":0.271699994802475},{"id":"https://openalex.org/C145242015","wikidata":"https://www.wikidata.org/wiki/Q774123","display_name":"Approximation theory","level":2,"score":0.2712000012397766},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.2590999901294708}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/cgo68049.2026.11395227","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cgo68049.2026.11395227","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":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320316083","display_name":"Tencent","ror":"https://ror.org/00hhjss72"},{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320321543","display_name":"China Postdoctoral Science Foundation","ror":"https://ror.org/0426zh255"},{"id":"https://openalex.org/F4320334704","display_name":"Australian Research Council","ror":"https://ror.org/05mmh0f86"},{"id":"https://openalex.org/F4320335777","display_name":"National Key Research and Development Program of China","ror":null}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":37,"referenced_works":["https://openalex.org/W2018136118","https://openalex.org/W2081368694","https://openalex.org/W2094281253","https://openalex.org/W2111593426","https://openalex.org/W2143283746","https://openalex.org/W2159937165","https://openalex.org/W2365032803","https://openalex.org/W2566629575","https://openalex.org/W2748818695","https://openalex.org/W2770653135","https://openalex.org/W2775811337","https://openalex.org/W2791673912","https://openalex.org/W2838756705","https://openalex.org/W2895305554","https://openalex.org/W2902494873","https://openalex.org/W2948348275","https://openalex.org/W2952851876","https://openalex.org/W2954390672","https://openalex.org/W2963015836","https://openalex.org/W2963521335","https://openalex.org/W2963928453","https://openalex.org/W2982479999","https://openalex.org/W3047017957","https://openalex.org/W3126212707","https://openalex.org/W3127794782","https://openalex.org/W3139307480","https://openalex.org/W3159717682","https://openalex.org/W3164441165","https://openalex.org/W3170647102","https://openalex.org/W4229070237","https://openalex.org/W4237680980","https://openalex.org/W4249492500","https://openalex.org/W4306178406","https://openalex.org/W4311277703","https://openalex.org/W4391952402","https://openalex.org/W4393147284","https://openalex.org/W4399851697"],"related_works":[],"abstract_inverted_index":{"Recent":[0],"GPUs":[1,64,135],"integrate":[2],"specialized":[3],"hardware":[4],"for":[5,14],"low-precision":[6,41,131],"arithmetic":[7,132],"(e.g.,":[8],"FP16,":[9],"INT8),":[10],"offering":[11],"substantial":[12],"speedups":[13],"tensor":[15],"operations.":[16],"However,":[17],"existing":[18],"methods":[19],"typically":[20],"rely":[21],"on":[22,108,133],"coarse,":[23],"operator-level":[24,138],"trial-and-error":[25],"tuning,":[26],"which":[27],"restricts":[28],"the":[29,59,122],"performance\u2013accuracy":[30],"trade-off":[31,47],"space":[32,48],"and":[33,70,111,116,127],"limits":[34],"achievable":[35],"gains.We":[36],"present":[37],"Platensor,":[38],"a":[39,76,79],"progressive":[40],"approximation":[42],"framework":[43],"that":[44,83,93,102,118],"expands":[45],"this":[46],"through":[49],"ne-grained,":[50],"tile-level":[51,86],"strategies.":[52],"The":[53],"key":[54],"idea":[55],"is":[56],"to":[57,65,137],"exploit":[58],"tiled":[60],"computation":[61],"patterns":[62],"of":[63,105],"enable":[66],"flexible":[67],"precision":[68],"control":[69],"richer":[71],"optimization":[72],"opportunities.":[73],"Platensor":[74,119],"performs":[75],"two-phase":[77],"exploration:":[78],"fast":[80],"rule-based":[81],"pass":[82],"selects":[84],"promising":[85],"configurations,":[87],"followed":[88],"by":[89],"an":[90],"evolutionary":[91],"search":[92],"refines":[94],"them.":[95],"It":[96],"then":[97],"automatically":[98],"generates":[99],"optimized":[100],"kernels":[101],"combine":[103],"tiles":[104],"different":[106],"precisions.Experiments":[107],"GEMM":[109],"operators":[110],"representative":[112],"applications\u2014including":[113],"kNN,":[114],"LLMs,":[115],"HPL-MxP\u2014show":[117],"significantly":[120],"broadens":[121],"attainable":[123],"performance\u2013":[124],"accuracy":[125],"trade-offs":[126],"more":[128],"fully":[129],"leverages":[130],"modern":[134],"compared":[136],"tuning.":[139]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-02-24T00:00:00"}