{"id":"https://openalex.org/W4410553236","doi":"https://doi.org/10.23919/date64628.2025.10993079","title":"LightMamba: Efficient Mamba Acceleration on FPGA with Quantization and Hardware Co-design","display_name":"LightMamba: Efficient Mamba Acceleration on FPGA with Quantization and Hardware Co-design","publication_year":2025,"publication_date":"2025-03-31","ids":{"openalex":"https://openalex.org/W4410553236","doi":"https://doi.org/10.23919/date64628.2025.10993079"},"language":"en","primary_location":{"id":"doi:10.23919/date64628.2025.10993079","is_oa":false,"landing_page_url":"https://doi.org/10.23919/date64628.2025.10993079","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 Design, Automation & 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/A5102763687","display_name":"Renjie Wei","orcid":"https://orcid.org/0000-0001-6378-5763"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Renjie Wei","raw_affiliation_strings":["Institute for Artificial Intelligence"],"affiliations":[{"raw_affiliation_string":"Institute for Artificial Intelligence","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113389019","display_name":"Songqiang Xu","orcid":"https://orcid.org/0009-0001-8373-0732"},"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":"Songqiang Xu","raw_affiliation_strings":["School of Software and Microelectronics, Peking University,Beijing,China"],"affiliations":[{"raw_affiliation_string":"School of Software and Microelectronics, Peking University,Beijing,China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5026294096","display_name":"Linfeng Zhong","orcid":"https://orcid.org/0000-0001-8277-4277"},"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":"Linfeng Zhong","raw_affiliation_strings":["School of Electronic and Computer Engineering, Peking University,Shenzhen,China"],"affiliations":[{"raw_affiliation_string":"School of Electronic and Computer Engineering, Peking University,Shenzhen,China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5002987540","display_name":"Zebin Yang","orcid":"https://orcid.org/0000-0002-3271-883X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zebin Yang","raw_affiliation_strings":["Institute for Artificial Intelligence"],"affiliations":[{"raw_affiliation_string":"Institute for Artificial Intelligence","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039548811","display_name":"Qingyu Guo","orcid":"https://orcid.org/0000-0002-8509-3933"},"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":"Qingyu Guo","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/A5002438548","display_name":"Yuan Wang","orcid":"https://orcid.org/0000-0002-4951-4286"},"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":"Yuan Wang","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/A5002760019","display_name":"Runsheng Wang","orcid":"https://orcid.org/0000-0002-7514-0767"},"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":"Runsheng Wang","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/A5092072535","display_name":"Meng Li","orcid":"https://orcid.org/0009-0007-8356-4432"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Meng Li","raw_affiliation_strings":["Institute for Artificial Intelligence"],"affiliations":[{"raw_affiliation_string":"Institute for Artificial Intelligence","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":8,"corresponding_author_ids":["https://openalex.org/A5102763687"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":12.2638,"has_fulltext":false,"cited_by_count":5,"citation_normalized_percentile":{"value":0.98569277,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":97,"max":98},"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/T10904","display_name":"Embedded Systems Design Techniques","score":0.945900022983551,"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/T10904","display_name":"Embedded Systems Design Techniques","score":0.945900022983551,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/field-programmable-gate-array","display_name":"Field-programmable gate array","score":0.8180081844329834},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6646740436553955},{"id":"https://openalex.org/keywords/hardware-acceleration","display_name":"Hardware acceleration","score":0.647229015827179},{"id":"https://openalex.org/keywords/acceleration","display_name":"Acceleration","score":0.643287718296051},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.529671311378479},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.5261903405189514},{"id":"https://openalex.org/keywords/computer-hardware","display_name":"Computer hardware","score":0.5073413252830505},{"id":"https://openalex.org/keywords/computer-architecture","display_name":"Computer architecture","score":0.3306691646575928},{"id":"https://openalex.org/keywords/physics","display_name":"Physics","score":0.08539626002311707},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.05318272113800049}],"concepts":[{"id":"https://openalex.org/C42935608","wikidata":"https://www.wikidata.org/wiki/Q190411","display_name":"Field-programmable gate array","level":2,"score":0.8180081844329834},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6646740436553955},{"id":"https://openalex.org/C13164978","wikidata":"https://www.wikidata.org/wiki/Q600158","display_name":"Hardware acceleration","level":3,"score":0.647229015827179},{"id":"https://openalex.org/C117896860","wikidata":"https://www.wikidata.org/wiki/Q11376","display_name":"Acceleration","level":2,"score":0.643287718296051},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.529671311378479},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.5261903405189514},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.5073413252830505},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.3306691646575928},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.08539626002311707},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.05318272113800049},{"id":"https://openalex.org/C74650414","wikidata":"https://www.wikidata.org/wiki/Q11397","display_name":"Classical mechanics","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.23919/date64628.2025.10993079","is_oa":false,"landing_page_url":"https://doi.org/10.23919/date64628.2025.10993079","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 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","display_name":"Affordable and clean energy","score":0.8100000023841858}],"awards":[{"id":"https://openalex.org/G4372204237","display_name":null,"funder_award_id":"62495102,92464104","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":28,"referenced_works":["https://openalex.org/W1779006005","https://openalex.org/W2890894339","https://openalex.org/W2998617917","https://openalex.org/W3047848469","https://openalex.org/W3194676777","https://openalex.org/W4308083513","https://openalex.org/W4308083526","https://openalex.org/W4393578753","https://openalex.org/W4393949386","https://openalex.org/W4399487406","https://openalex.org/W4409282403","https://openalex.org/W4409282814","https://openalex.org/W6727099177","https://openalex.org/W6749838110","https://openalex.org/W6763018175","https://openalex.org/W6842758154","https://openalex.org/W6847478871","https://openalex.org/W6850625674","https://openalex.org/W6850936240","https://openalex.org/W6852337893","https://openalex.org/W6854866820","https://openalex.org/W6859298233","https://openalex.org/W6860710830","https://openalex.org/W6862020115","https://openalex.org/W6862909604","https://openalex.org/W6868720932","https://openalex.org/W6868789466","https://openalex.org/W6872309109"],"related_works":["https://openalex.org/W2111241003","https://openalex.org/W2355315220","https://openalex.org/W4200391368","https://openalex.org/W1967938402","https://openalex.org/W2386041993","https://openalex.org/W1608572506","https://openalex.org/W2146872326","https://openalex.org/W2518118925","https://openalex.org/W3158825072","https://openalex.org/W3159273459"],"abstract_inverted_index":{"State":[0],"space":[1],"models":[2,16],"(SSMs)":[3],"like":[4],"Mamba":[5,18,32,70,106],"have":[6],"recently":[7],"attracted":[8],"much":[9],"attention.":[10],"Compared":[11],"to":[12,35,38,88,94,108],"Transformer-based":[13],"large":[14],"language":[15],"(LLMs),":[17],"achieves":[19],"linear":[20],"computation":[21,46,93,107,116],"complexity":[22],"with":[23],"the":[24,39,44,61,90,105,110,125,132,153,172],"sequence":[25],"length":[26],"and":[27,43,64,84,112,123,128,145],"demonstrates":[28],"superior":[29],"performance.":[30],"However,":[31],"is":[33,167],"hard":[34],"accelerate":[36],"due":[37],"scattered":[40],"activation":[41],"outliers":[42],"complex":[45],"dependency,":[47],"rendering":[48],"existing":[49],"LLM":[50],"accelerators":[51],"inefficient.":[52],"In":[53],"this":[54],"paper,":[55],"we":[56],"propose":[57,74],"LightMamba":[58,139,162],"that":[59,80,102,170],"co-designs":[60],"quantization":[62,78,83,87],"algorithm":[63,79],"FPGA":[65,100,144],"accelerator":[66,101,133],"architecture":[67],"for":[68],"efficient":[69],"inference.":[71],"We":[72,96,137],"first":[73],"an":[75,99],"FPGA-friendly":[76],"post-training":[77],"features":[81],"rotation-assisted":[82],"power-of-two":[85],"SSM":[86],"reduce":[89],"majority":[91],"of":[92,131,171],"4-bit.":[95],"further":[97],"design":[98],"partially":[103],"unrolls":[104],"balance":[109],"efficiency":[111,130,151],"hardware":[113,126],"costs.":[114],"Through":[115],"reordering":[117],"as":[118,120],"well":[119],"fine-grained":[121],"tiling":[122],"fusion,":[124],"utilization":[127],"memory":[129],"get":[134],"drastically":[135],"improved.":[136],"implement":[138],"on":[140,158],"Xilinx":[141],"Versal":[142],"VCK190":[143],"achieve":[146],"4.65~6.06":[147],"x":[148,169],"higher":[149],"energy":[150],"over":[152],"GPU":[154,173],"baseline.":[155,174],"When":[156],"evaluated":[157],"Alveo":[159],"U280":[160],"FPGA,":[161],"reaches":[163],"93":[164],"tokens/s,":[165],"which":[166],"1.43":[168]},"counts_by_year":[{"year":2025,"cited_by_count":5}],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}