{"id":"https://openalex.org/W4410492390","doi":"https://doi.org/10.1109/cicc63670.2025.10982870","title":"A 52.03TOPS/W DCIM-Based Accelerator with FlashAttention and Sparsity-Aware Alignment for LLMs","display_name":"A 52.03TOPS/W DCIM-Based Accelerator with FlashAttention and Sparsity-Aware Alignment for LLMs","publication_year":2025,"publication_date":"2025-04-13","ids":{"openalex":"https://openalex.org/W4410492390","doi":"https://doi.org/10.1109/cicc63670.2025.10982870"},"language":"en","primary_location":{"id":"doi:10.1109/cicc63670.2025.10982870","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cicc63670.2025.10982870","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE Custom Integrated Circuits Conference (CICC)","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/A5100461621","display_name":"Bo Liu","orcid":"https://orcid.org/0000-0002-0894-1054"},"institutions":[{"id":"https://openalex.org/I76569877","display_name":"Southeast University","ror":"https://ror.org/04ct4d772","country_code":"CN","type":"education","lineage":["https://openalex.org/I76569877"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Bo Liu","raw_affiliation_strings":["School of Integrated Circuits, Southeast University,Nanjing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Southeast University,Nanjing,China","institution_ids":["https://openalex.org/I76569877"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5088334968","display_name":"Xingyu Xu","orcid":"https://orcid.org/0000-0002-7015-9717"},"institutions":[{"id":"https://openalex.org/I76569877","display_name":"Southeast University","ror":"https://ror.org/04ct4d772","country_code":"CN","type":"education","lineage":["https://openalex.org/I76569877"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xingyu Xu","raw_affiliation_strings":["School of Integrated Circuits, Southeast University,Nanjing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Southeast University,Nanjing,China","institution_ids":["https://openalex.org/I76569877"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5088633393","display_name":"Yang Zhang","orcid":"https://orcid.org/0009-0005-1455-5749"},"institutions":[{"id":"https://openalex.org/I76569877","display_name":"Southeast University","ror":"https://ror.org/04ct4d772","country_code":"CN","type":"education","lineage":["https://openalex.org/I76569877"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yang Zhang","raw_affiliation_strings":["School of Integrated Circuits, Southeast University,Nanjing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Southeast University,Nanjing,China","institution_ids":["https://openalex.org/I76569877"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109689226","display_name":"Xilong Kang","orcid":null},"institutions":[{"id":"https://openalex.org/I76569877","display_name":"Southeast University","ror":"https://ror.org/04ct4d772","country_code":"CN","type":"education","lineage":["https://openalex.org/I76569877"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xilong Kang","raw_affiliation_strings":["School of Integrated Circuits, Southeast University,Nanjing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Southeast University,Nanjing,China","institution_ids":["https://openalex.org/I76569877"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039733006","display_name":"Qingwen Wei","orcid":"https://orcid.org/0009-0009-7047-7182"},"institutions":[{"id":"https://openalex.org/I76569877","display_name":"Southeast University","ror":"https://ror.org/04ct4d772","country_code":"CN","type":"education","lineage":["https://openalex.org/I76569877"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Qingwen Wei","raw_affiliation_strings":["School of Integrated Circuits, Southeast University,Nanjing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Southeast University,Nanjing,China","institution_ids":["https://openalex.org/I76569877"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5062140266","display_name":"Zihan Zou","orcid":"https://orcid.org/0009-0002-1510-1341"},"institutions":[{"id":"https://openalex.org/I76569877","display_name":"Southeast University","ror":"https://ror.org/04ct4d772","country_code":"CN","type":"education","lineage":["https://openalex.org/I76569877"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zihan Zou","raw_affiliation_strings":["School of Integrated Circuits, Southeast University,Nanjing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Southeast University,Nanjing,China","institution_ids":["https://openalex.org/I76569877"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5042474689","display_name":"Jun Yang","orcid":"https://orcid.org/0000-0002-9401-3528"},"institutions":[{"id":"https://openalex.org/I76569877","display_name":"Southeast University","ror":"https://ror.org/04ct4d772","country_code":"CN","type":"education","lineage":["https://openalex.org/I76569877"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jun Yang","raw_affiliation_strings":["School of Integrated Circuits, Southeast University,Nanjing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Southeast University,Nanjing,China","institution_ids":["https://openalex.org/I76569877"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5089840020","display_name":"Hao Cai","orcid":"https://orcid.org/0000-0002-0566-9411"},"institutions":[{"id":"https://openalex.org/I76569877","display_name":"Southeast University","ror":"https://ror.org/04ct4d772","country_code":"CN","type":"education","lineage":["https://openalex.org/I76569877"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hao Cai","raw_affiliation_strings":["School of Integrated Circuits, Southeast University,Nanjing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Southeast University,Nanjing,China","institution_ids":["https://openalex.org/I76569877"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5113009142","display_name":"Xin Si","orcid":null},"institutions":[{"id":"https://openalex.org/I76569877","display_name":"Southeast University","ror":"https://ror.org/04ct4d772","country_code":"CN","type":"education","lineage":["https://openalex.org/I76569877"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xin Si","raw_affiliation_strings":["School of Integrated Circuits, Southeast University,Nanjing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Integrated Circuits, Southeast University,Nanjing,China","institution_ids":["https://openalex.org/I76569877"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I76569877"],"apc_list":null,"apc_paid":null,"fwci":5.8643,"has_fulltext":false,"cited_by_count":2,"citation_normalized_percentile":{"value":0.94636834,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":91,"max":97},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"3"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11338","display_name":"Advancements in Photolithography Techniques","score":0.9916999936103821,"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"}},"topics":[{"id":"https://openalex.org/T11338","display_name":"Advancements in Photolithography Techniques","score":0.9916999936103821,"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/T10412","display_name":"Microfluidic and Capillary Electrophoresis Applications","score":0.9760000109672546,"subfield":{"id":"https://openalex.org/subfields/2204","display_name":"Biomedical 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/T10781","display_name":"Plasma Diagnostics and Applications","score":0.9743000268936157,"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/computer-science","display_name":"Computer science","score":0.550133228302002},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.3622557520866394}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.550133228302002},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.3622557520866394}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/cicc63670.2025.10982870","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cicc63670.2025.10982870","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE Custom Integrated Circuits Conference (CICC)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.41999998688697815,"id":"https://metadata.un.org/sdg/13","display_name":"Climate action"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":14,"referenced_works":["https://openalex.org/W2950656546","https://openalex.org/W3134304371","https://openalex.org/W3134526034","https://openalex.org/W4220865834","https://openalex.org/W4221086307","https://openalex.org/W4360605703","https://openalex.org/W4360606782","https://openalex.org/W4393406875","https://openalex.org/W4401247357","https://openalex.org/W6838322825","https://openalex.org/W6848451824","https://openalex.org/W6851775633","https://openalex.org/W6854094408","https://openalex.org/W6854308872"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"Decoder-based":[0],"large":[1],"language":[2],"models":[3],"(LLMs)":[4],"are":[5],"designed":[6],"specifically":[7],"for":[8,51,118],"text":[9],"generation,":[10],"consisting":[11],"primarily":[12],"of":[13,33,38,210],"multiple":[14],"decoder":[15],"layers":[16],"that":[17],"utilize":[18],"unidirectional":[19],"attention":[20,40],"mechanism":[21,175],"[1],":[22],"as":[23,61,83,154,156],"shown":[24],"in":[25,49,181,187,194],"Fig.":[26],"1.":[27],"Due":[28],"to":[29,69],"the":[30,36,39,71,96,113,128,139,163,172],"enormous":[31],"number":[32],"parameters":[34],"and":[35,46,56,64,98,105,150,207],"complexity":[37,170],"mechanism,":[41],"LLMs":[42,120],"require":[43],"substantial":[44],"computational":[45],"memory":[47,99,152,159],"resources":[48],"deployment":[50],"resource-constrained":[52],"edge":[53],"devices.":[54],"Algorithms":[55],"model":[57,202],"compression":[58],"methods":[59],"such":[60,82],"KV":[62,133],"Cache":[63,134],"pruning":[65,203],"have":[66],"been":[67],"proposed":[68],"reduce":[70,95,168],"workload,":[72],"achieving":[73],"remarkable":[74],"success":[75],"[2],":[76],"[3].":[77],"However,":[78],"high-efficient":[79],"hardware":[80,215],"implementation":[81],"sparsity":[84,199],"exploitation":[85],"[14]":[86],"still":[87],"remains":[88],"unexplored.":[89],"Meanwhile,":[90],"Digital":[91],"Computing-in-Memory":[92],"(DCIM)":[93],"can":[94,167],"computation":[97,169,190],"demands":[100],"by":[101,201],"minimizing":[102],"data":[103],"movement":[104],"enabling":[106],"more":[107],"efficient":[108],"parallel":[109],"processing":[110],"directly":[111],"within":[112],"memory,":[114],"making":[115],"it":[116],"suitable":[117],"memory-bound":[119],"[6\u20139,":[121],"11\u201313].":[122],"Designing":[123],"DCIM-based":[124],"LLM":[125],"accelerators":[126],"faces":[127],"following":[129],"challenges:":[130],"1)":[131],"The":[132],"size":[135,166,180],"scales":[136],"linearly":[137],"with":[138],"context":[140],"length,":[141],"which":[142,213],"induces":[143],"a":[144],"mismatch":[145],"between":[146],"high":[147],"storage":[148],"demand":[149],"on-chip":[151],"size,":[153],"well":[155],"frequent":[157],"external":[158],"access.":[160],"2)":[161],"Scaling":[162],"CIM":[164,188,211],"Macro":[165,189],"from":[171],"FlashAttention":[173],"(FA)":[174],"[4],":[176],"[5],":[177],"but":[178],"tiling":[179],"FA":[182],"introduces":[183],"imbalanced":[184],"workload":[185],"distribution":[186],"stages,":[191],"exacerbating":[192],"inefficiencies":[193],"resource":[195],"utilization.":[196],"3)":[197],"Unstructured":[198],"brought":[200],"causes":[204],"low":[205],"spatial":[206],"temporal":[208],"utilization":[209],"Macro,":[212],"compromises":[214],"efficiency.":[216]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":1}],"updated_date":"2026-08-11T07:18:39.950985","created_date":"2025-10-10T00:00:00"}
