{"id":"https://openalex.org/W4409475001","doi":"https://doi.org/10.1109/lca.2025.3560786","title":"MixDiT: Accelerating Image Diffusion Transformer Inference With Mixed-Precision MX Quantization","display_name":"MixDiT: Accelerating Image Diffusion Transformer Inference With Mixed-Precision MX Quantization","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4409475001","doi":"https://doi.org/10.1109/lca.2025.3560786"},"language":"en","primary_location":{"id":"doi:10.1109/lca.2025.3560786","is_oa":false,"landing_page_url":"https://doi.org/10.1109/lca.2025.3560786","pdf_url":null,"source":{"id":"https://openalex.org/S17643076","display_name":"IEEE Computer Architecture Letters","issn_l":"1556-6056","issn":["1556-6056","1556-6064","2473-2575"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Computer Architecture Letters","raw_type":"journal-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/A5107947006","display_name":"D. Kim","orcid":null},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":true,"raw_author_name":"Daeun Kim","raw_affiliation_strings":["School of Computing, Graduate School of AI Semiconductor, Korea Advanced Institute of Science and Technology, Daejeon, South Korea"],"affiliations":[{"raw_affiliation_string":"School of Computing, Graduate School of AI Semiconductor, Korea Advanced Institute of Science and Technology, Daejeon, South Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101320389","display_name":"Jinwoo Hwang","orcid":null},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jinwoo Hwang","raw_affiliation_strings":["School of Computing, Graduate School of AI Semiconductor, Korea Advanced Institute of Science and Technology, Daejeon, South Korea"],"affiliations":[{"raw_affiliation_string":"School of Computing, Graduate School of AI Semiconductor, Korea Advanced Institute of Science and Technology, Daejeon, South Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5022686032","display_name":"Changhun Oh","orcid":"https://orcid.org/0000-0003-2002-1928"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Changhun Oh","raw_affiliation_strings":["School of Computing, Graduate School of AI Semiconductor, Korea Advanced Institute of Science and Technology, Daejeon, South Korea"],"affiliations":[{"raw_affiliation_string":"School of Computing, Graduate School of AI Semiconductor, Korea Advanced Institute of Science and Technology, Daejeon, South Korea","institution_ids":["https://openalex.org/I157485424"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5037553165","display_name":"Jongse Park","orcid":"https://orcid.org/0000-0002-6629-449X"},"institutions":[{"id":"https://openalex.org/I157485424","display_name":"Korea Advanced Institute of Science and Technology","ror":"https://ror.org/05apxxy63","country_code":"KR","type":"education","lineage":["https://openalex.org/I157485424"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jongse Park","raw_affiliation_strings":["School of Computing, Graduate School of AI Semiconductor, Korea Advanced Institute of Science and Technology, Daejeon, South Korea"],"affiliations":[{"raw_affiliation_string":"School of Computing, Graduate School of AI Semiconductor, Korea Advanced Institute of Science and Technology, Daejeon, South Korea","institution_ids":["https://openalex.org/I157485424"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5107947006"],"corresponding_institution_ids":["https://openalex.org/I157485424"],"apc_list":null,"apc_paid":null,"fwci":1.2784,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.78712113,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":96,"max":99},"biblio":{"volume":"24","issue":"1","first_page":"141","last_page":"144"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10688","display_name":"Image and Signal Denoising Methods","score":0.9779000282287598,"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/T10688","display_name":"Image and Signal Denoising Methods","score":0.9779000282287598,"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/T13114","display_name":"Image Processing Techniques and Applications","score":0.9726999998092651,"subfield":{"id":"https://openalex.org/subfields/2214","display_name":"Media Technology"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10052","display_name":"Medical Image Segmentation Techniques","score":0.9348999857902527,"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/computer-science","display_name":"Computer science","score":0.6707535982131958},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.6232129335403442},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6200273036956787},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.5490128397941589},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.4286920726299286},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.36974167823791504},{"id":"https://openalex.org/keywords/computer-vision","display_name":"Computer vision","score":0.3656069040298462},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.24755015969276428},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.11689284443855286},{"id":"https://openalex.org/keywords/voltage","display_name":"Voltage","score":0.09471717476844788}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6707535982131958},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.6232129335403442},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6200273036956787},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.5490128397941589},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4286920726299286},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.36974167823791504},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.3656069040298462},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.24755015969276428},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.11689284443855286},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.09471717476844788}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/lca.2025.3560786","is_oa":false,"landing_page_url":"https://doi.org/10.1109/lca.2025.3560786","pdf_url":null,"source":{"id":"https://openalex.org/S17643076","display_name":"IEEE Computer Architecture Letters","issn_l":"1556-6056","issn":["1556-6056","1556-6064","2473-2575"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Computer Architecture Letters","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":10,"referenced_works":["https://openalex.org/W4380874652","https://openalex.org/W4390872297","https://openalex.org/W4401211881","https://openalex.org/W4404625217","https://openalex.org/W4405003117","https://openalex.org/W4407953477","https://openalex.org/W6862995742","https://openalex.org/W6868797372","https://openalex.org/W6869514076","https://openalex.org/W6870184360"],"related_works":["https://openalex.org/W2051487156","https://openalex.org/W2073681303","https://openalex.org/W2053286651","https://openalex.org/W2055243143","https://openalex.org/W2181743346","https://openalex.org/W2187401768","https://openalex.org/W4206178588","https://openalex.org/W3094491777","https://openalex.org/W3214715529","https://openalex.org/W4287635093"],"abstract_inverted_index":{"<underline":[0,3],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[1,4],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">Di</u>ffusion":[2],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">T</u>ransformer":[5],"(DiT)":[6],"has":[7,52],"driven":[8],"significant":[9],"progress":[10],"in":[11,151],"image":[12],"generation":[13],"tasks.":[14],"However,":[15],"DiT":[16,60,93,98],"inferencing":[17,61],"is":[18],"notoriously":[19],"compute-intensive":[20],"and":[21,35,66,130],"incurs":[22],"long":[23],"latency":[24],"even":[25],"on":[26,38],"datacenter-scale":[27],"GPUs,":[28],"primarily":[29],"due":[30],"to":[31,42,91,106],"its":[32,43],"iterative":[33],"nature":[34],"heavy":[36],"reliance":[37],"GEMM":[39,112],"operations":[40],"inherent":[41],"encoder-based":[44],"structure.":[45],"To":[46,73,114],"address":[47],"the":[48,97,119],"challenge,":[49],"prior":[50],"work":[51],"explored":[53],"quantization,":[54],"but":[55],"achieving":[56],"low-precision":[57],"quantization":[58],"for":[59],"with":[62,148],"both":[63],"high":[64],"accuracy":[65],"substantial":[67],"speedup":[68,117,142],"remains":[69],"an":[70,80],"open":[71],"problem.":[72],"this":[74,76],"end,":[75],"paper":[77],"proposes":[78],"MixDiT,":[79],"algorithm-hardware":[81],"co-designed":[82],"acceleration":[83],"solution":[84],"that":[85,126,139],"exploits":[86],"mixed":[87],"Microscaling":[88],"(MX)":[89],"formats":[90],"quantize":[92],"activation":[94,99],"values.":[95],"MixDiTquantizes":[96],"tensors":[100],"by":[101],"selectively":[102],"applying":[103],"higher":[104],"precision":[105,133],"magnitude-based":[107],"outliers,":[108],"which":[109],"produce":[110],"mixed-precision":[111,120],"operations.":[113],"achieve":[115],"tangible":[116],"from":[118],"arithmetic,":[121],"we":[122],"design":[123],"a":[124,141],"MixDiTaccelerator":[125],"enables":[127],"precision-flexible":[128],"multiplications":[129],"efficient":[131],"MX":[132],"conversions.":[134],"Our":[135],"experimental":[136],"results":[137],"show":[138],"MixDiTdelivers":[140],"of":[143],"2.10\u20135.32\u00d7":[144],"over":[145],"RTX":[146],"3090,":[147],"no":[149],"loss":[150],"FID.":[152]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-03-27T05:58:40.876381","created_date":"2025-10-10T00:00:00"}