{"id":"https://openalex.org/W7163552274","doi":"https://doi.org/10.23919/date69613.2026.11539190","title":"Fast and Energy-Efficient Support for Low-Precision LLMs on PIM","display_name":"Fast and Energy-Efficient Support for Low-Precision LLMs on PIM","publication_year":2026,"publication_date":"2026-04-20","ids":{"openalex":"https://openalex.org/W7163552274","doi":"https://doi.org/10.23919/date69613.2026.11539190"},"language":null,"primary_location":{"id":"doi:10.23919/date69613.2026.11539190","is_oa":false,"landing_page_url":"https://doi.org/10.23919/date69613.2026.11539190","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2026 Design, Automation & Test in Europe Conference (DATE)","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/A5040024119","display_name":"B H Kim","orcid":null},"institutions":[{"id":"https://openalex.org/I123900574","display_name":"Pohang University of Science and Technology","ror":"https://ror.org/04xysgw12","country_code":"KR","type":"education","lineage":["https://openalex.org/I123900574"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Byeori Kim","raw_affiliation_strings":["Pohang University of Science and Technology (POSTECH),Department of Computer Science and Engineering,Pohang,South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Pohang University of Science and Technology (POSTECH),Department of Computer Science and Engineering,Pohang,South Korea","institution_ids":["https://openalex.org/I123900574"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100712411","display_name":"Sangjun Lee","orcid":"https://orcid.org/0009-0009-9595-1821"},"institutions":[{"id":"https://openalex.org/I123900574","display_name":"Pohang University of Science and Technology","ror":"https://ror.org/04xysgw12","country_code":"KR","type":"education","lineage":["https://openalex.org/I123900574"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Sangjun Lee","raw_affiliation_strings":["Pohang University of Science and Technology (POSTECH),Graduate School of Artificial Intelligence,Pohang,South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Pohang University of Science and Technology (POSTECH),Graduate School of Artificial Intelligence,Pohang,South Korea","institution_ids":["https://openalex.org/I123900574"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5064884832","display_name":"Eunhyeok Park","orcid":"https://orcid.org/0000-0002-7331-9819"},"institutions":[{"id":"https://openalex.org/I123900574","display_name":"Pohang University of Science and Technology","ror":"https://ror.org/04xysgw12","country_code":"KR","type":"education","lineage":["https://openalex.org/I123900574"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Eunhyeok Park","raw_affiliation_strings":["Pohang University of Science and Technology (POSTECH),Graduate School of Artificial Intelligence,Pohang,South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Pohang University of Science and Technology (POSTECH),Graduate School of Artificial Intelligence,Pohang,South Korea","institution_ids":["https://openalex.org/I123900574"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I123900574"],"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":"1","last_page":"7"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10049","display_name":"Magnetic properties of thin films","score":0.09809999912977219,"subfield":{"id":"https://openalex.org/subfields/3107","display_name":"Atomic and Molecular Physics, and Optics"},"field":{"id":"https://openalex.org/fields/31","display_name":"Physics and Astronomy"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10049","display_name":"Magnetic properties of thin films","score":0.09809999912977219,"subfield":{"id":"https://openalex.org/subfields/3107","display_name":"Atomic and Molecular Physics, and Optics"},"field":{"id":"https://openalex.org/fields/31","display_name":"Physics and Astronomy"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10558","display_name":"Advancements in Semiconductor Devices and Circuit Design","score":0.05139999836683273,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"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/T11263","display_name":"Electromagnetic Simulation and Numerical Methods","score":0.03579999879002571,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/government","display_name":"Government (linguistics)","score":0.26089999079704285},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.2290000021457672},{"id":"https://openalex.org/keywords/disease","display_name":"Disease","score":0.22450000047683716},{"id":"https://openalex.org/keywords/work","display_name":"Work (physics)","score":0.22360000014305115}],"concepts":[{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.31299999356269836},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.2992999851703644},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.28279998898506165},{"id":"https://openalex.org/C71924100","wikidata":"https://www.wikidata.org/wiki/Q11190","display_name":"Medicine","level":0,"score":0.2786000072956085},{"id":"https://openalex.org/C2778137410","wikidata":"https://www.wikidata.org/wiki/Q2732820","display_name":"Government (linguistics)","level":2,"score":0.26089999079704285},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2290000021457672},{"id":"https://openalex.org/C2779134260","wikidata":"https://www.wikidata.org/wiki/Q12136","display_name":"Disease","level":2,"score":0.22450000047683716},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.22360000014305115},{"id":"https://openalex.org/C162324750","wikidata":"https://www.wikidata.org/wiki/Q8134","display_name":"Economics","level":0,"score":0.22339999675750732},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.22059999406337738}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.23919/date69613.2026.11539190","is_oa":false,"landing_page_url":"https://doi.org/10.23919/date69613.2026.11539190","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2026 Design, Automation & Test in Europe Conference (DATE)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/7","score":0.8606730699539185,"display_name":"Affordable and clean energy"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":14,"referenced_works":["https://openalex.org/W2162639668","https://openalex.org/W2725159389","https://openalex.org/W2896090304","https://openalex.org/W3006586535","https://openalex.org/W3100710793","https://openalex.org/W3189166979","https://openalex.org/W3210580311","https://openalex.org/W4297097348","https://openalex.org/W4392427708","https://openalex.org/W4393147284","https://openalex.org/W4394998968","https://openalex.org/W4395073431","https://openalex.org/W4406650295","https://openalex.org/W4407466309"],"related_works":[],"abstract_inverted_index":{"Processing-in-Memory":[0],"(PIM)":[1],"has":[2],"gained":[3],"momentum":[4],"as":[5],"a":[6,36,108,149,157,167,176,198],"promising":[7],"approach":[8],"for":[9,35,110],"mitigating":[10],"memory":[11,60],"bottlenecks,":[12],"and":[13,91,119,125,179,201],"it":[14,49,164,189],"is":[15,50,95,162],"particularly":[16],"well":[17],"suited":[18],"to":[19,43,52,112,143,184],"autoregressive":[20],"decoding":[21],"in":[22,55,69,83,117,148,170,194],"large":[23,37,45],"language":[24],"models":[25],"(LLMs),":[26],"where":[27],"memory-bound":[28],"General":[29],"Matrix-Vector":[30],"Multiplication":[31],"(GEMV)":[32],"operations":[33,116],"account":[34],"portion":[38],"of":[39,47,140,160],"the":[40,44,81,130,133,144],"workload.":[41],"Due":[42],"size":[46,75,159],"LLMs,":[48,70],"challenging":[51],"support":[53,114],"them":[54],"PIM":[56,145],"environments":[57],"with":[58,123,156,175,197],"limited":[59],"capacity.":[61],"Applying":[62],"group-wise":[63],"weight-only":[64],"quantization":[65],"(GWQ),":[66],"widely":[67],"used":[68,147],"can":[71,98],"effectively":[72],"reduce":[73],"model":[74],"while":[76],"minimizing":[77],"accuracy":[78],"degradation.":[79],"However,":[80],"weights":[82],"GWQ-applied":[84],"LLMs":[85,122],"are":[86],"typically":[87],"dequantized":[88],"using":[89],"scales":[90],"zero-points":[92],"before":[93],"GEMV":[94,115],"performed,":[96],"which":[97],"introduce":[99],"non-trivial":[100],"latency":[101],"overhead.":[102],"In":[103],"this":[104],"paper,":[105],"we":[106],"propose":[107],"method":[109,135],"DRAM-PIM":[111],"efficiently":[113],"symmetric":[118],"asymmetric":[120,154],"GWQ-based":[121],"INT2":[124],"INT4":[126],"precision.":[127],"Based":[128],"on":[129],"Newton":[131],"scheme,":[132],"proposed":[134],"incurs":[136],"an":[137,191],"area":[138],"overhead":[139],"20.6%":[141],"compared":[142,183],"units":[146],"16-bank,":[150],"single-channel":[151],"system.":[152],"When":[153],"GWQ":[155],"group":[158],"128":[161],"applied,":[163],"achieves":[165,190],"approximately":[166],"4":[168],"reduction":[169,193],"storage":[171],"at":[172],"INT4,":[173],"along":[174,196],"1.16\u00d7":[177],"speedup":[178,200],"1.41\u00d7":[180],"energy":[181,203],"efficiency":[182],"FP16":[185],"GEMV.":[186],"At":[187],"INT2,":[188],"8\u00d7":[192],"storage,":[195],"1.27\u00d7":[199],"1.57\u00d7":[202],"efficiency.":[204]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2026-06-05T00:00:00"}