{"id":"https://openalex.org/W4410583349","doi":"https://doi.org/10.23919/date64628.2025.10993092","title":"DOTS: DRAM-PIM Optimization for Tall and Skinny GEMM Operations in LLM Inference","display_name":"DOTS: DRAM-PIM Optimization for Tall and Skinny GEMM Operations in LLM Inference","publication_year":2025,"publication_date":"2025-03-31","ids":{"openalex":"https://openalex.org/W4410583349","doi":"https://doi.org/10.23919/date64628.2025.10993092"},"language":"en","primary_location":{"id":"doi:10.23919/date64628.2025.10993092","is_oa":false,"landing_page_url":"https://doi.org/10.23919/date64628.2025.10993092","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 Design, Automation &amp;amp; Test in Europe Conference (DATE)","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/A5086975094","display_name":"Gyeonghwan Park","orcid":null},"institutions":[{"id":"https://openalex.org/I139264467","display_name":"Seoul National University","ror":"https://ror.org/04h9pn542","country_code":"KR","type":"education","lineage":["https://openalex.org/I139264467"]}],"countries":["KR"],"is_corresponding":true,"raw_author_name":"Gyeonghwan Park","raw_affiliation_strings":["Seoul National University,Electical and Computer Engineering,Korea"],"affiliations":[{"raw_affiliation_string":"Seoul National University,Electical and Computer Engineering,Korea","institution_ids":["https://openalex.org/I139264467"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5029049879","display_name":"SangHyeok Han","orcid":"https://orcid.org/0000-0002-5914-5883"},"institutions":[{"id":"https://openalex.org/I139264467","display_name":"Seoul National University","ror":"https://ror.org/04h9pn542","country_code":"KR","type":"education","lineage":["https://openalex.org/I139264467"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Sanghyeok Han","raw_affiliation_strings":["Seoul National University,Electical and Computer Engineering,Korea"],"affiliations":[{"raw_affiliation_string":"Seoul National University,Electical and Computer Engineering,Korea","institution_ids":["https://openalex.org/I139264467"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102023303","display_name":"Boram Yoon","orcid":"https://orcid.org/0000-0003-0069-6514"},"institutions":[{"id":"https://openalex.org/I139264467","display_name":"Seoul National University","ror":"https://ror.org/04h9pn542","country_code":"KR","type":"education","lineage":["https://openalex.org/I139264467"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Byungkuk Yoon","raw_affiliation_strings":["Seoul National University,Electical and Computer Engineering,Korea"],"affiliations":[{"raw_affiliation_string":"Seoul National University,Electical and Computer Engineering,Korea","institution_ids":["https://openalex.org/I139264467"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5003219699","display_name":"Jae\u2010Joon Kim","orcid":"https://orcid.org/0000-0001-5175-8258"},"institutions":[{"id":"https://openalex.org/I139264467","display_name":"Seoul National University","ror":"https://ror.org/04h9pn542","country_code":"KR","type":"education","lineage":["https://openalex.org/I139264467"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jae-Joon Kim","raw_affiliation_strings":["Seoul National University,Electical and Computer Engineering,Korea"],"affiliations":[{"raw_affiliation_string":"Seoul National University,Electical and Computer Engineering,Korea","institution_ids":["https://openalex.org/I139264467"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5086975094"],"corresponding_institution_ids":["https://openalex.org/I139264467"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.06183827,"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":"2"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10346","display_name":"Magnetic confinement fusion research","score":0.9812999963760376,"subfield":{"id":"https://openalex.org/subfields/3106","display_name":"Nuclear and High Energy Physics"},"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/T10346","display_name":"Magnetic confinement fusion research","score":0.9812999963760376,"subfield":{"id":"https://openalex.org/subfields/3106","display_name":"Nuclear and High Energy Physics"},"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/T12077","display_name":"Vacuum and Plasma Arcs","score":0.9713000059127808,"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/T11367","display_name":"Particle accelerators and beam dynamics","score":0.9609000086784363,"subfield":{"id":"https://openalex.org/subfields/2202","display_name":"Aerospace 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/dram","display_name":"Dram","score":0.8314651250839233},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6747733354568481},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.5420939922332764},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.4909534752368927},{"id":"https://openalex.org/keywords/computer-architecture","display_name":"Computer architecture","score":0.4762963354587555},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.21363642811775208},{"id":"https://openalex.org/keywords/computer-hardware","display_name":"Computer hardware","score":0.1531878411769867}],"concepts":[{"id":"https://openalex.org/C7366592","wikidata":"https://www.wikidata.org/wiki/Q1255620","display_name":"Dram","level":2,"score":0.8314651250839233},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6747733354568481},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.5420939922332764},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.4909534752368927},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.4762963354587555},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.21363642811775208},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.1531878411769867}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.23919/date64628.2025.10993092","is_oa":false,"landing_page_url":"https://doi.org/10.23919/date64628.2025.10993092","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 Design, Automation &amp;amp; Test in Europe Conference (DATE)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":5,"referenced_works":["https://openalex.org/W3100710793","https://openalex.org/W3189166979","https://openalex.org/W3190062760","https://openalex.org/W4221001402","https://openalex.org/W4394998968"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W3120961607","https://openalex.org/W4401568740","https://openalex.org/W3148568549","https://openalex.org/W2098207691","https://openalex.org/W1648516568","https://openalex.org/W361036515","https://openalex.org/W2161286015"],"abstract_inverted_index":{"For":[0],"large":[1,112],"language":[2],"models":[3,110],"(LLMs),":[4],"increasing":[5],"token":[6],"lengths":[7],"require":[8],"smaller":[9],"batch":[10,59],"sizes":[11,60],"due":[12],"to":[13,22],"increase":[14],"in":[15,31,55,78],"memory":[16,28],"requirement":[17],"for":[18,109],"KV":[19],"caching,":[20],"leading":[21],"under-utilization":[23],"of":[24,96],"processing":[25],"units":[26],"and":[27,49,69,99,104],"bandwidth":[29],"bottleneck":[30],"NPUs.":[32],"To":[33],"address":[34],"the":[35],"challenge,":[36],"we":[37],"propose":[38],"DOTS,":[39],"a":[40],"new":[41],"DRAM-PIM":[42,65],"architecture":[43],"that":[44,89],"can":[45],"handle":[46],"both":[47],"GEMV":[48],"GEMM":[50,56],"efficiently,":[51],"even":[52],"outperforming":[53],"NPUs":[54],"operations":[57],"when":[58],"are":[61],"small.":[62],"The":[63],"proposed":[64,91],"reduces":[66],"power":[67],"consumption":[68],"latency":[70],"caused":[71],"by":[72],"frequent":[73],"DRAM":[74,80],"row":[75],"activation":[76],"switching":[77],"conventional":[79],"PIMs":[81],"with":[82],"negligible":[83],"hardware":[84],"overhead.":[85],"Simulation":[86],"results":[87],"show":[88],"our":[90],"design":[92],"achieves":[93],"throughput":[94],"improvements":[95],"1.83x,":[97],"1.92x,":[98],"1.7x":[100],"over":[101],"GPU,":[102],"NPU,":[103],"heterogeneous":[105],"NPU/PIM":[106],"systems,":[107],"respectively,":[108],"as":[111,113],"or":[114],"larger":[115],"than":[116],"OPT-175B.":[117]},"counts_by_year":[],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}
