{"id":"https://openalex.org/W4416429558","doi":"https://doi.org/10.1109/iccad66269.2025.11240889","title":"LP-Spec: Leveraging LPDDR PIM for Efficient LLM Mobile Speculative Inference with Architecture-Dataflow Co-Optimization","display_name":"LP-Spec: Leveraging LPDDR PIM for Efficient LLM Mobile Speculative Inference with Architecture-Dataflow Co-Optimization","publication_year":2025,"publication_date":"2025-10-26","ids":{"openalex":"https://openalex.org/W4416429558","doi":"https://doi.org/10.1109/iccad66269.2025.11240889"},"language":null,"primary_location":{"id":"doi:10.1109/iccad66269.2025.11240889","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iccad66269.2025.11240889","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/ACM International Conference On Computer Aided Design (ICCAD)","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/A5101930482","display_name":"Siyuan He","orcid":"https://orcid.org/0000-0001-8992-9550"},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Siyuan He","raw_affiliation_strings":["School of Integrated Circuits Peking University,Beijing,China"],"affiliations":[{"raw_affiliation_string":"School of Integrated Circuits Peking University,Beijing,China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5023075240","display_name":"Zhiguang Zhu","orcid":"https://orcid.org/0009-0008-7859-9734"},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhantong Zhu","raw_affiliation_strings":["School of Integrated Circuits Peking University,Beijing,China"],"affiliations":[{"raw_affiliation_string":"School of Integrated Circuits Peking University,Beijing,China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5081603207","display_name":"Yandong He","orcid":"https://orcid.org/0000-0003-3465-4718"},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yandong He","raw_affiliation_strings":["School of Integrated Circuits Peking University,Beijing,China"],"affiliations":[{"raw_affiliation_string":"School of Integrated Circuits Peking University,Beijing,China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5088551028","display_name":"Tianyu Jia","orcid":"https://orcid.org/0000-0002-4570-4613"},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Tianyu Jia","raw_affiliation_strings":["School of Integrated Circuits Peking University,Beijing,China"],"affiliations":[{"raw_affiliation_string":"School of Integrated Circuits Peking University,Beijing,China","institution_ids":["https://openalex.org/I20231570"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5101930482"],"corresponding_institution_ids":["https://openalex.org/I20231570"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.41060236,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"9"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.35910001397132874,"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":0.35910001397132874,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.16670000553131104,"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/T14347","display_name":"Big Data and Digital Economy","score":0.14090000092983246,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/security-token","display_name":"Security token","score":0.6416000127792358},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.6348000168800354},{"id":"https://openalex.org/keywords/energy-consumption","display_name":"Energy consumption","score":0.5253999829292297},{"id":"https://openalex.org/keywords/dram","display_name":"Dram","score":0.5245000123977661},{"id":"https://openalex.org/keywords/mobile-device","display_name":"Mobile device","score":0.5200999975204468},{"id":"https://openalex.org/keywords/memory-bandwidth","display_name":"Memory bandwidth","score":0.5120000243186951},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5040000081062317},{"id":"https://openalex.org/keywords/scheduling","display_name":"Scheduling (production processes)","score":0.44029998779296875}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8737000226974487},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.6416000127792358},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.6348000168800354},{"id":"https://openalex.org/C2780165032","wikidata":"https://www.wikidata.org/wiki/Q16869822","display_name":"Energy consumption","level":2,"score":0.5253999829292297},{"id":"https://openalex.org/C7366592","wikidata":"https://www.wikidata.org/wiki/Q1255620","display_name":"Dram","level":2,"score":0.5245000123977661},{"id":"https://openalex.org/C186967261","wikidata":"https://www.wikidata.org/wiki/Q5082128","display_name":"Mobile device","level":2,"score":0.5200999975204468},{"id":"https://openalex.org/C188045654","wikidata":"https://www.wikidata.org/wiki/Q17148339","display_name":"Memory bandwidth","level":2,"score":0.5120000243186951},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5040000081062317},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.48969998955726624},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.44029998779296875},{"id":"https://openalex.org/C2778476105","wikidata":"https://www.wikidata.org/wiki/Q628539","display_name":"Workload","level":2,"score":0.40149998664855957},{"id":"https://openalex.org/C144543869","wikidata":"https://www.wikidata.org/wiki/Q2738570","display_name":"Mobile computing","level":2,"score":0.36410000920295715},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.33559998869895935},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.335099995136261},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.3312999904155731},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.32839998602867126},{"id":"https://openalex.org/C176649486","wikidata":"https://www.wikidata.org/wiki/Q2308807","display_name":"Memory management","level":3,"score":0.3199999928474426},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.3077000081539154},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.3009999990463257},{"id":"https://openalex.org/C46743427","wikidata":"https://www.wikidata.org/wiki/Q1341685","display_name":"Inference engine","level":3,"score":0.29899999499320984},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.2831999957561493},{"id":"https://openalex.org/C100800780","wikidata":"https://www.wikidata.org/wiki/Q1175867","display_name":"Memory controller","level":3,"score":0.28200000524520874},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.28049999475479126},{"id":"https://openalex.org/C150202949","wikidata":"https://www.wikidata.org/wiki/Q107602","display_name":"Pointer (user interface)","level":2,"score":0.2777000069618225},{"id":"https://openalex.org/C2776257435","wikidata":"https://www.wikidata.org/wiki/Q1576430","display_name":"Bandwidth (computing)","level":2,"score":0.26739999651908875},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.2644999921321869}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/iccad66269.2025.11240889","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iccad66269.2025.11240889","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/ACM International Conference On Computer Aided Design (ICCAD)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":19,"referenced_works":["https://openalex.org/W2346205343","https://openalex.org/W2896090304","https://openalex.org/W2982008795","https://openalex.org/W3044083334","https://openalex.org/W3134519092","https://openalex.org/W3189166979","https://openalex.org/W4220686330","https://openalex.org/W4221001402","https://openalex.org/W4226126604","https://openalex.org/W4360605920","https://openalex.org/W4387064011","https://openalex.org/W4389519819","https://openalex.org/W4392427708","https://openalex.org/W4394998968","https://openalex.org/W4395106409","https://openalex.org/W4395112660","https://openalex.org/W4401211642","https://openalex.org/W4408925909","https://openalex.org/W4409282855"],"related_works":[],"abstract_inverted_index":{"LLM":[0,101,146],"inference":[1,21,37,147],"on":[2,125,148],"mobile":[3,150,154],"devices":[4],"faces":[5],"extraneous":[6],"challenges":[7],"due":[8],"to":[9,72,99,110,133,144],"limited":[10],"memory":[11,106],"bandwidth":[12],"and":[13,22,32,95,116,137,166,173,181,189],"computational":[14],"resources.":[15],"To":[16],"address":[17],"these":[18],"issues,":[19],"speculative":[20,36,65,102],"processing-in-memory":[23],"(PIM)":[24],"techniques":[25],"have":[26],"been":[27],"explored":[28],"at":[29],"the":[30,126],"algorithmic":[31],"hardware":[33],"levels.":[34],"However,":[35],"results":[38],"in":[39,63,169],"more":[40],"compute-intensive":[41],"GEMM":[42],"operations,":[43],"creating":[44],"new":[45],"design":[46],"trade-offs":[47],"for":[48],"existing":[49],"GEMV-accelerated":[50,157],"PIM":[51,89,117,180],"architectures.":[52],"Furthermore,":[53,119],"there":[54],"exists":[55],"a":[56,120],"significant":[57],"amount":[58],"of":[59],"redundant":[60],"draft":[61,92,128],"tokens":[62],"tree-based":[64],"inference,":[66],"necessitating":[67],"efficient":[68],"token":[69,93,129],"management":[70],"schemes":[71],"minimize":[73,134],"energy":[74,135,171],"consumption.":[75],"In":[76],"this":[77],"work,":[78],"we":[79],"present":[80],"LP-Spec,":[81],"an":[82],"architecture-dataflow":[83],"co-design":[84],"leveraging":[85],"hybrid":[86],"LPDDR5":[87],"performance-enhanced":[88],"architecture":[90],"with":[91,177],"pruning":[94],"dynamic":[96],"workload":[97],"scheduling":[98],"accelerate":[100],"inference.":[103],"A":[104],"near-data":[105],"controller":[107],"is":[108,131],"proposed":[109],"enable":[111],"data":[112,121],"reallocation":[113],"between":[114],"DRAM":[115],"banks.":[118],"allocation":[122],"unit":[123],"based":[124],"hardware-aware":[127],"pruner":[130],"developed":[132],"consumption":[136],"fully":[138],"exploit":[139],"parallel":[140],"execution":[141],"opportunities.":[142],"Compared":[143,176],"end-to-end":[145],"other":[149],"solutions":[151],"such":[152],"as":[153],"NPUs":[155],"or":[156],"PIMs,":[158],"our":[159],"LP-Spec":[160,185],"achieves":[161],"13.21\u00d7":[162],",":[163],"7.56":[164],"\u00d7,":[165],"99.87\u00d7":[167],"improvements":[168],"performance,":[170],"efficiency,":[172],"energy-delay-product":[174],"(EDP).":[175],"prior":[178],"AttAcc":[179],"RTX":[182],"3090":[183],"GPU,":[184],"can":[186],"obtain":[187],"12.83\u00d7":[188],"415.31\u00d7":[190],"EDP":[191],"reduction":[192],"benefits.":[193]},"counts_by_year":[],"updated_date":"2026-03-07T16:01:11.037858","created_date":"2025-11-20T00:00:00"}