{"id":"https://openalex.org/W7164563682","doi":"https://doi.org/10.1109/lca.2026.3702973","title":"NDPool: Correctness-Preserving Shared Execution for Efficient LLM Inference on CXL-NDP Systems","display_name":"NDPool: Correctness-Preserving Shared Execution for Efficient LLM Inference on CXL-NDP Systems","publication_year":2026,"publication_date":"2026-06-12","ids":{"openalex":"https://openalex.org/W7164563682","doi":"https://doi.org/10.1109/lca.2026.3702973"},"language":null,"primary_location":{"id":"doi:10.1109/lca.2026.3702973","is_oa":false,"landing_page_url":"https://doi.org/10.1109/lca.2026.3702973","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/A5088901467","display_name":"Yeji Jung","orcid":"https://orcid.org/0009-0008-7840-4669"},"institutions":[{"id":"https://openalex.org/I193352282","display_name":"Daegu Gyeongbuk Institute of Science and Technology","ror":"https://ror.org/03frjya69","country_code":"KR","type":"education","lineage":["https://openalex.org/I193352282"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Yeji Jung","raw_affiliation_strings":["Daegu-Gyeongbuk Institute of Science and Technology, Daegu, Republic of Korea"],"raw_orcid":"https://orcid.org/0009-0008-7840-4669","affiliations":[{"raw_affiliation_string":"Daegu-Gyeongbuk Institute of Science and Technology, Daegu, Republic of Korea","institution_ids":["https://openalex.org/I193352282"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5035850620","display_name":"Hwanjun Lee","orcid":"https://orcid.org/0009-0001-0513-2112"},"institutions":[{"id":"https://openalex.org/I193352282","display_name":"Daegu Gyeongbuk Institute of Science and Technology","ror":"https://ror.org/03frjya69","country_code":"KR","type":"education","lineage":["https://openalex.org/I193352282"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Hwanjun Lee","raw_affiliation_strings":["Daegu-Gyeongbuk Institute of Science and Technology, Daegu, Republic of Korea"],"raw_orcid":"https://orcid.org/0009-0001-0513-2112","affiliations":[{"raw_affiliation_string":"Daegu-Gyeongbuk Institute of Science and Technology, Daegu, Republic of Korea","institution_ids":["https://openalex.org/I193352282"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130576788","display_name":"Sungju Kim","orcid":null},"institutions":[{"id":"https://openalex.org/I193352282","display_name":"Daegu Gyeongbuk Institute of Science and Technology","ror":"https://ror.org/03frjya69","country_code":"KR","type":"education","lineage":["https://openalex.org/I193352282"]},{"id":"https://openalex.org/I193775966","display_name":"Yonsei University","ror":"https://ror.org/01wjejq96","country_code":"KR","type":"education","lineage":["https://openalex.org/I193775966"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Sungju Kim","raw_affiliation_strings":["Daegu-Gyeongbuk Institute of Science and Technology, Daegu, Republic of Korea","Yonsei University, Seoul, Republic of Korea"],"raw_orcid":"https://orcid.org/0009-0006-7869-7145","affiliations":[{"raw_affiliation_string":"Daegu-Gyeongbuk Institute of Science and Technology, Daegu, Republic of Korea","institution_ids":["https://openalex.org/I193352282"]},{"raw_affiliation_string":"Yonsei University, Seoul, Republic of Korea","institution_ids":["https://openalex.org/I193775966"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101513594","display_name":"Seulki Kim","orcid":"https://orcid.org/0000-0001-5505-1022"},"institutions":[{"id":"https://openalex.org/I193352282","display_name":"Daegu Gyeongbuk Institute of Science and Technology","ror":"https://ror.org/03frjya69","country_code":"KR","type":"education","lineage":["https://openalex.org/I193352282"]},{"id":"https://openalex.org/I193775966","display_name":"Yonsei University","ror":"https://ror.org/01wjejq96","country_code":"KR","type":"education","lineage":["https://openalex.org/I193775966"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Seulki Kim","raw_affiliation_strings":["Daegu-Gyeongbuk Institute of Science and Technology, Daegu, Republic of Korea","Yonsei University, Seoul, Republic of Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Daegu-Gyeongbuk Institute of Science and Technology, Daegu, Republic of Korea","institution_ids":["https://openalex.org/I193352282"]},{"raw_affiliation_string":"Yonsei University, Seoul, Republic of Korea","institution_ids":["https://openalex.org/I193775966"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113430170","display_name":"Yunhyeong Jeon","orcid":null},"institutions":[{"id":"https://openalex.org/I193352282","display_name":"Daegu Gyeongbuk Institute of Science and Technology","ror":"https://ror.org/03frjya69","country_code":"KR","type":"education","lineage":["https://openalex.org/I193352282"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Yunhyeong Jeon","raw_affiliation_strings":["Daegu-Gyeongbuk Institute of Science and Technology, Daegu, Republic of Korea"],"raw_orcid":"https://orcid.org/0009-0005-2890-3134","affiliations":[{"raw_affiliation_string":"Daegu-Gyeongbuk Institute of Science and Technology, Daegu, Republic of Korea","institution_ids":["https://openalex.org/I193352282"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101885459","display_name":"Daehoon Kim","orcid":"https://orcid.org/0000-0003-0837-0877"},"institutions":[{"id":"https://openalex.org/I193775966","display_name":"Yonsei University","ror":"https://ror.org/01wjejq96","country_code":"KR","type":"education","lineage":["https://openalex.org/I193775966"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Daehoon Kim","raw_affiliation_strings":["Yonsei University, Seoul, Republic of Korea"],"raw_orcid":"https://orcid.org/0000-0003-0837-0877","affiliations":[{"raw_affiliation_string":"Yonsei University, Seoul, Republic of Korea","institution_ids":["https://openalex.org/I193775966"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"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":0.75511082,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"25","issue":"2","first_page":"243","last_page":"246"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11424","display_name":"Security and Verification in Computing","score":0.11010000109672546,"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"}},"topics":[{"id":"https://openalex.org/T11424","display_name":"Security and Verification in Computing","score":0.11010000109672546,"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"}},{"id":"https://openalex.org/T11321","display_name":"Error Correcting Code Techniques","score":0.08699999749660492,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T11181","display_name":"Advanced Data Storage Technologies","score":0.07720000296831131,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/inference","display_name":"Inference","score":0.5013999938964844},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.30630001425743103},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.2971999943256378},{"id":"https://openalex.org/keywords/algorithm-design","display_name":"Algorithm design","score":0.2515000104904175}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8357999920845032},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5013999938964844},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.4503999948501587},{"id":"https://openalex.org/C115903868","wikidata":"https://www.wikidata.org/wiki/Q80993","display_name":"Software engineering","level":1,"score":0.3467000126838684},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.33660000562667847},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3174999952316284},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.30630001425743103},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2971999943256378},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2816999852657318},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.28060001134872437},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2605000138282776},{"id":"https://openalex.org/C106516650","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm design","level":2,"score":0.2515000104904175}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/lca.2026.3702973","is_oa":false,"landing_page_url":"https://doi.org/10.1109/lca.2026.3702973","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":[{"id":"https://openalex.org/G5170403967","display_name":null,"funder_award_id":"IO260311-15806-01","funder_id":"https://openalex.org/F4320332195","funder_display_name":"Samsung"}],"funders":[{"id":"https://openalex.org/F4320332195","display_name":"Samsung","ror":"https://ror.org/04w3jy968"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"inference":[1],"performance":[2],"of":[3,94],"Large":[4],"Language":[5],"Models":[6],"(LLMs)":[7],"is":[8,105],"increasingly":[9],"constrained":[10],"by":[11,177,182],"memory":[12,36],"capacity":[13],"and":[14,115,161,181],"bandwidth":[15],"rather":[16],"than":[17],"raw":[18],"computation,":[19],"particularly":[20],"for":[21,34,68,90,144,150],"long-context":[22],"workloads.":[23],"CXL-based":[24],"Near-Data":[25],"Processing":[26],"(CXL-NDP)":[27],"has":[28],"emerged":[29],"as":[30],"a":[31,78,83,91,130,185],"promising":[32],"solution":[33],"expanding":[35],"capacity,":[37],"but":[38],"existing":[39],"systems":[40],"suffer":[41],"from":[42],"severe":[43],"resource":[44],"under-utilization":[45],"due":[46],"to":[47,65,157,179],"exclusive":[48],"device":[49,103],"allocation":[50],"under":[51,169],"hybrid":[52,193],"parallel":[53,194],"execution.":[54],"To":[55,124],"minimize":[56],"per-request":[57],"latency,":[58],"prior":[59],"designs":[60],"dedicate":[61],"multiple":[62],"NDP":[63],"devices":[64,87,140],"tensor":[66],"parallelism":[67],"FC":[69],"layers,":[70],"while":[71],"non-FC":[72],"layers":[73],"are":[74],"executed":[75],"only":[76],"on":[77,184],"single":[79],"master":[80],"device.":[81],"As":[82],"result,":[84],"most":[85],"auxiliary":[86],"remain":[88],"idle":[89,138],"large":[92],"fraction":[93],"execution":[95,133,159],"time,":[96],"fundamentally":[97],"limiting":[98],"overall":[99],"system":[100,175],"throughput.":[101],"Safe":[102],"sharing":[104],"challenging":[106],"because":[107],"out-of-order":[108],"peer-to-peer":[109],"communication":[110],"can":[111],"break":[112],"request":[113],"isolation":[114],"computational":[116],"correctness":[117],"at":[118],"the":[119,190],"shared":[120,132],"CXL":[121],"receive":[122,148],"path.":[123],"address":[125],"this":[126],"limitation,":[127],"we":[128],"proposeNDPool,":[129],"correctness-preserving":[131],"framework":[134],"that":[135],"dynamically":[136],"reuses":[137],"TP":[139],"across":[141],"concurrent":[142],"requests":[143],"FC-layer":[145],"execution.NDPoolcombines":[146],"ordered":[147],"control":[149],"safe":[151],"data":[152],"delivery":[153],"with":[154],"adaptive":[155],"scheduling":[156],"reduce":[158],"bubbles":[160],"maximize":[162],"shared-device":[163],"utilization.":[164],"Experimental":[165],"results":[166],"show":[167],"that,":[168],"equivalent":[170],"service-level":[171],"objective":[172],"(SLO)":[173],"constraints,NDPoolimproves":[174],"throughput":[176],"up":[178],"3.7\u00d7":[180],"1.5\u00d7":[183],"geometric":[186],"mean":[187],"basis":[188],"over":[189],"state-of-the-art":[191],"CXL-NDP":[192],"design.":[195]},"counts_by_year":[],"updated_date":"2026-06-26T06:17:10.115597","created_date":"2026-06-13T00:00:00"}