{"id":"https://openalex.org/W7163546258","doi":"https://doi.org/10.23919/date69613.2026.11539704","title":"Boosting the Performance of Tree-Based Speculative Decoding of LLMs on FPGAs","display_name":"Boosting the Performance of Tree-Based Speculative Decoding of LLMs on FPGAs","publication_year":2026,"publication_date":"2026-04-20","ids":{"openalex":"https://openalex.org/W7163546258","doi":"https://doi.org/10.23919/date69613.2026.11539704"},"language":null,"primary_location":{"id":"doi:10.23919/date69613.2026.11539704","is_oa":false,"landing_page_url":"https://doi.org/10.23919/date69613.2026.11539704","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2026 Design, Automation & Test in Europe Conference (DATE)","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/A5001331259","display_name":"Tielong Liu","orcid":"https://orcid.org/0009-0001-5500-4338"},"institutions":[{"id":"https://openalex.org/I4210094879","display_name":"Shandong Institute of Automation","ror":"https://ror.org/00qdtba35","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210094879","https://openalex.org/I4210142748"]},{"id":"https://openalex.org/I4210112150","display_name":"Institute of Automation","ror":"https://ror.org/022c3hy66","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Tielong Liu","raw_affiliation_strings":["Chinese Academy of Sciences,C2DL, Institute of Automation,Beijing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences,C2DL, Institute of Automation,Beijing,China","institution_ids":["https://openalex.org/I4210094879","https://openalex.org/I4210112150"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137820588","display_name":"Gang Li","orcid":null},"institutions":[{"id":"https://openalex.org/I4210094879","display_name":"Shandong Institute of Automation","ror":"https://ror.org/00qdtba35","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210094879","https://openalex.org/I4210142748"]},{"id":"https://openalex.org/I4210112150","display_name":"Institute of Automation","ror":"https://ror.org/022c3hy66","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Gang Li","raw_affiliation_strings":["Chinese Academy of Sciences,C2DL, Institute of Automation,Beijing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences,C2DL, Institute of Automation,Beijing,China","institution_ids":["https://openalex.org/I4210094879","https://openalex.org/I4210112150"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5012344182","display_name":"Zitao Mo","orcid":"https://orcid.org/0000-0001-8623-5465"},"institutions":[{"id":"https://openalex.org/I4210094879","display_name":"Shandong Institute of Automation","ror":"https://ror.org/00qdtba35","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210094879","https://openalex.org/I4210142748"]},{"id":"https://openalex.org/I4210112150","display_name":"Institute of Automation","ror":"https://ror.org/022c3hy66","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zitao Mo","raw_affiliation_strings":["Chinese Academy of Sciences,C2DL, Institute of Automation,Beijing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences,C2DL, Institute of Automation,Beijing,China","institution_ids":["https://openalex.org/I4210094879","https://openalex.org/I4210112150"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5003248892","display_name":"Z G Zhu","orcid":null},"institutions":[{"id":"https://openalex.org/I4210094879","display_name":"Shandong Institute of Automation","ror":"https://ror.org/00qdtba35","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210094879","https://openalex.org/I4210142748"]},{"id":"https://openalex.org/I4210112150","display_name":"Institute of Automation","ror":"https://ror.org/022c3hy66","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zeyu Zhu","raw_affiliation_strings":["Chinese Academy of Sciences,C2DL, Institute of Automation,Beijing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences,C2DL, Institute of Automation,Beijing,China","institution_ids":["https://openalex.org/I4210094879","https://openalex.org/I4210112150"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137913746","display_name":"Minnan Pei","orcid":null},"institutions":[{"id":"https://openalex.org/I4210094879","display_name":"Shandong Institute of Automation","ror":"https://ror.org/00qdtba35","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210094879","https://openalex.org/I4210142748"]},{"id":"https://openalex.org/I4210112150","display_name":"Institute of Automation","ror":"https://ror.org/022c3hy66","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Minnan Pei","raw_affiliation_strings":["Chinese Academy of Sciences,C2DL, Institute of Automation,Beijing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences,C2DL, Institute of Automation,Beijing,China","institution_ids":["https://openalex.org/I4210094879","https://openalex.org/I4210112150"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5137839075","display_name":"Jian Cheng","orcid":null},"institutions":[{"id":"https://openalex.org/I4210094879","display_name":"Shandong Institute of Automation","ror":"https://ror.org/00qdtba35","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210094879","https://openalex.org/I4210142748"]},{"id":"https://openalex.org/I4210112150","display_name":"Institute of Automation","ror":"https://ror.org/022c3hy66","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jian Cheng","raw_affiliation_strings":["Chinese Academy of Sciences,C2DL, Institute of Automation,Beijing,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences,C2DL, Institute of Automation,Beijing,China","institution_ids":["https://openalex.org/I4210094879","https://openalex.org/I4210112150"]}]}],"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":0.56568849,"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":"7"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12326","display_name":"Network Packet Processing and Optimization","score":0.1298000067472458,"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/T12326","display_name":"Network Packet Processing and Optimization","score":0.1298000067472458,"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/T11522","display_name":"VLSI and FPGA Design Techniques","score":0.12950000166893005,"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/T11697","display_name":"Numerical Methods and Algorithms","score":0.11069999635219574,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/boosting","display_name":"Boosting (machine learning)","score":0.8486999869346619},{"id":"https://openalex.org/keywords/decoding-methods","display_name":"Decoding methods","score":0.5753999948501587},{"id":"https://openalex.org/keywords/field-programmable-gate-array","display_name":"Field-programmable gate array","score":0.43389999866485596},{"id":"https://openalex.org/keywords/scheduling","display_name":"Scheduling (production processes)","score":0.2858999967575073}],"concepts":[{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.8486999869346619},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6575000286102295},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.5753999948501587},{"id":"https://openalex.org/C42935608","wikidata":"https://www.wikidata.org/wiki/Q190411","display_name":"Field-programmable gate array","level":2,"score":0.43389999866485596},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3068999946117401},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.29660001397132874},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.2858999967575073},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.26829999685287476},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.26809999346733093},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.24959999322891235}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.23919/date69613.2026.11539704","is_oa":false,"landing_page_url":"https://doi.org/10.23919/date69613.2026.11539704","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2026 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":0.4274941384792328}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":7,"referenced_works":["https://openalex.org/W2002555321","https://openalex.org/W2963204221","https://openalex.org/W3170490008","https://openalex.org/W3171639395","https://openalex.org/W4393578753","https://openalex.org/W4395112660","https://openalex.org/W4404782770"],"related_works":[],"abstract_inverted_index":{"As":[0],"an":[1,74],"efficient":[2],"alternative":[3],"to":[4,15,23,55,77,120,135],"autoregressive":[5,67,175],"decoding,":[6],"tree-based":[7,82],"speculative":[8,83],"decoding":[9,44,68,84,118],"(SD)":[10],"has":[11],"been":[12],"widely":[13],"adopted":[14],"accelerate":[16],"LLM":[17],"inference":[18],"on":[19,51,59,85,155],"GPUs.":[20],"However,":[21],"due":[22],"the":[24,65,79,109,116,137,149],"notable":[25],"disparity":[26],"in":[27],"compute":[28],"power":[29],"and":[30,61,113,165],"memory":[31],"bandwidth,":[32],"we":[33,72],"observe":[34],"that":[35],"a":[36,42,91,99,131,142,170],"specific":[37,100],"target-draft":[38,101],"model":[39,102],"pair":[40,103],"with":[41],"proper":[43],"configuration,":[45],"despite":[46],"demonstrating":[47],"significant":[48],"performance":[49,80,123],"gains":[50],"GPUs,":[52],"often":[53],"fails":[54],"maintain":[56],"its":[57],"efficacy":[58],"FPGAs,":[60],"may":[62],"even":[63],"underperform":[64],"standard":[66],"approachIn":[69],"this":[70],"paper,":[71],"propose":[73,130],"analytical":[75],"framework":[76],"revive":[78],"of":[81,145,172],"FPGAs.":[86],"We":[87,128,151],"introduce":[88],"effective":[89],"performance,":[90],"roofline-based":[92],"metric":[93],"designed":[94],"to:":[95],"1)":[96],"assess":[97],"whether":[98],"can":[104],"benefit":[105],"from":[106],"SD":[107,125,162],"for":[108,141],"given":[110],"FPGA":[111,157],"platform,":[112],"2)":[114],"determine":[115],"optimal":[117,138],"configuration":[119],"achieve":[121],"peak":[122],"when":[124],"is":[126],"applicable.":[127],"also":[129],"prior-score-based":[132],"search":[133],"strategy":[134],"identify":[136],"tree":[139],"structure":[140],"preset":[143],"number":[144],"nodes,":[146],"further":[147],"enhancing":[148],"performance.":[150],"evaluate":[152],"our":[153],"method":[154],"AMD":[156],"platforms":[158],"using":[159],"two":[160],"state-of-the-art":[161],"algorithms:":[163],"LongSpec":[164],"EAGLE-3.":[166],"Our":[167],"approach":[168],"demonstrates":[169],"speedup":[171],"2.54-3.89\u00d7":[173],"over":[174],"decoding.":[176]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-06-05T00:00:00"}