{"id":"https://openalex.org/W7124982014","doi":"https://doi.org/10.1109/asicon66040.2025.11326063","title":"Optimizing LLM inference for FPGAs","display_name":"Optimizing LLM inference for FPGAs","publication_year":2025,"publication_date":"2025-10-21","ids":{"openalex":"https://openalex.org/W7124982014","doi":"https://doi.org/10.1109/asicon66040.2025.11326063"},"language":"en","primary_location":{"id":"doi:10.1109/asicon66040.2025.11326063","doi":"https://doi.org/10.1109/asicon66040.2025.11326063","is_oa":false,"landing_page_url":"https://doi.org/10.1109/asicon66040.2025.11326063","pdf_url":null,"source":{"id":"https://openalex.org/S4363607945","display_name":"IEEE International Conference on ASIC (ASICON)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE 16th International Conference on ASIC (ASICON)","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","is_traditional":true,"indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://hdl.handle.net/10044/1/124656","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5123384550","display_name":"Jorge R De Freitas","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I47508984","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10","country_code":"GB","type":"education","lineage":["https://openalex.org/I47508984"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Jorge R De Freitas","raw_affiliation_strings":["Imperial College London,UK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Imperial College London,UK","institution_ids":["https://openalex.org/I47508984"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123417783","display_name":"Jose G. F. Coutinho","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I47508984","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10","country_code":"GB","type":"education","lineage":["https://openalex.org/I47508984"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Jose G. F. Coutinho","raw_affiliation_strings":["Imperial College London,UK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Imperial College London,UK","institution_ids":["https://openalex.org/I47508984"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5083303676","display_name":"Ce Guo","orcid":"https://orcid.org/0000-0002-0272-9175","observed_orcids":["https://orcid.org/0000-0002-0272-9175","https://orcid.org/0009-0007-0554-8322","https://orcid.org/0000-0002-4472-4604"]},"institutions":[{"id":"https://openalex.org/I47508984","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10","country_code":"GB","type":"education","lineage":["https://openalex.org/I47508984"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Ce Guo","raw_affiliation_strings":["Imperial College London,UK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Imperial College London,UK","institution_ids":["https://openalex.org/I47508984"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5058306867","display_name":"S\u00fcleyman S\u0131rr\u0131 Demirsoy","orcid":"https://orcid.org/0009-0006-5881-7883","observed_orcids":["https://orcid.org/0009-0006-5881-7883"]},"institutions":[{"id":"https://openalex.org/I4210094575","display_name":"Altera (United Kingdom)","ror":"https://ror.org/00m96gg93","country_code":"GB","type":"company","lineage":["https://openalex.org/I4210094575"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Suleyman Demirsoy","raw_affiliation_strings":["Altera,UK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Altera,UK","institution_ids":["https://openalex.org/I4210094575"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5110068140","display_name":"Wayne Luk","orcid":"https://orcid.org/0000-0002-6750-927X","observed_orcids":["https://orcid.org/0000-0002-6750-927X"]},"institutions":[{"id":"https://openalex.org/I47508984","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10","country_code":"GB","type":"education","lineage":["https://openalex.org/I47508984"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Wayne Luk","raw_affiliation_strings":["Imperial College London,UK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Imperial College London,UK","institution_ids":["https://openalex.org/I47508984"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5054475218","display_name":"Zhiqiang Que","orcid":"https://orcid.org/0000-0002-9263-6529","observed_orcids":["https://orcid.org/0000-0002-9263-6529"]},"institutions":[{"id":"https://openalex.org/I47508984","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10","country_code":"GB","type":"education","lineage":["https://openalex.org/I47508984"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Zhiqiang Que","raw_affiliation_strings":["Imperial College London,UK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Imperial College London,UK","institution_ids":["https://openalex.org/I47508984"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.9226,"has_fulltext":false,"cited_by_count":2,"citation_normalized_percentile":{"value":0.72919367,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":89,"max":95},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"4"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10904","display_name":"Embedded Systems Design Techniques","score":0.9945999979972839,"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":0.9945999979972839,"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":0.00139999995008111,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.000699999975040555,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/large-language-model-inference","display_name":"large language model inference","score":0.8920000195503235},{"id":"https://openalex.org/keywords/fpga-accelerator","display_name":"FPGA accelerator","score":0.8579999804496765},{"id":"https://openalex.org/keywords/oneapi","display_name":"oneAPI","score":0.7649999856948853},{"id":"https://openalex.org/keywords/hardware-acceleration","display_name":"hardware acceleration","score":0.32100000977516174},{"id":"https://openalex.org/keywords/pipelined-hardware-architecture","display_name":"pipelined hardware architecture","score":0.7179999947547913}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8030999898910522},{"id":"https://openalex.org/C42935608","wikidata":"https://www.wikidata.org/wiki/Q190411","display_name":"Field-programmable gate array","level":2,"score":0.7894999980926514},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.5888000130653381},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5618000030517578},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.5325000286102295},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5271000266075134},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.4932999908924103},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.46619999408721924},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.43549999594688416},{"id":"https://openalex.org/C142962650","wikidata":"https://www.wikidata.org/wiki/Q240838","display_name":"Reconfigurable computing","level":3,"score":0.42089998722076416},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.3878999948501587},{"id":"https://openalex.org/C34165917","wikidata":"https://www.wikidata.org/wiki/Q188267","display_name":"Programming paradigm","level":2,"score":0.36480000615119934},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.36010000109672546},{"id":"https://openalex.org/C206345919","wikidata":"https://www.wikidata.org/wiki/Q20380951","display_name":"Resource (disambiguation)","level":2,"score":0.3409999907016754},{"id":"https://openalex.org/C133162039","wikidata":"https://www.wikidata.org/wiki/Q1061077","display_name":"Code generation","level":3,"score":0.32919999957084656},{"id":"https://openalex.org/C107598950","wikidata":"https://www.wikidata.org/wiki/Q259864","display_name":"Microarchitecture","level":2,"score":0.32690000534057617},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.3165000081062317},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3109000027179718},{"id":"https://openalex.org/C184596265","wikidata":"https://www.wikidata.org/wiki/Q2651576","display_name":"Model of computation","level":3,"score":0.3102000057697296},{"id":"https://openalex.org/C2778119891","wikidata":"https://www.wikidata.org/wiki/Q477690","display_name":"CUDA","level":2,"score":0.3043000102043152},{"id":"https://openalex.org/C49154492","wikidata":"https://www.wikidata.org/wiki/Q5300","display_name":"Central processing unit","level":2,"score":0.3034999966621399},{"id":"https://openalex.org/C2776834041","wikidata":"https://www.wikidata.org/wiki/Q25346349","display_name":"Execution model","level":2,"score":0.29809999465942383},{"id":"https://openalex.org/C58013763","wikidata":"https://www.wikidata.org/wiki/Q5754574","display_name":"High-level synthesis","level":3,"score":0.29580000042915344},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.28679999709129333},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.2840999960899353},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.2653999924659729},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.2581999897956848},{"id":"https://openalex.org/C36503486","wikidata":"https://www.wikidata.org/wiki/Q11235244","display_name":"Domain (mathematical analysis)","level":2,"score":0.2574000060558319},{"id":"https://openalex.org/C2776221188","wikidata":"https://www.wikidata.org/wiki/Q21072556","display_name":"Design space exploration","level":2,"score":0.25619998574256897},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.2540999948978424},{"id":"https://openalex.org/C2777062904","wikidata":"https://www.wikidata.org/wiki/Q545406","display_name":"Toolchain","level":3,"score":0.2529999911785126}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.1109/asicon66040.2025.11326063","doi":"https://doi.org/10.1109/asicon66040.2025.11326063","is_oa":false,"landing_page_url":"https://doi.org/10.1109/asicon66040.2025.11326063","pdf_url":null,"source":{"id":"https://openalex.org/S4363607945","display_name":"IEEE International Conference on ASIC (ASICON)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE 16th International Conference on ASIC (ASICON)","raw_type":"proceedings-article"},{"id":"pmh:oai:research-information.bris.ac.uk:openaire_cris_publications/fd6291a3-514f-4d9b-869b-a3043d5852d1","doi":null,"is_oa":false,"landing_page_url":"https://hdl.handle.net/1983/fd6291a3-514f-4d9b-869b-a3043d5852d1","pdf_url":null,"source":{"id":"https://openalex.org/S4306400895","display_name":"Bristol Research (University of Bristol)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I36234482","host_organization_name":"University of Bristol","host_organization_lineage":["https://openalex.org/I36234482"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"De Freitas, J R, Coutinho, J G F, Guo, C, Demirsoy, S, Luk, W & Que, Z 2026, Optimizing LLM inference for FPGAs. in 2025 IEEE 16th International Conference on ASIC (ASICON). Proceedings of International Conference on ASIC, IEEE Computer Society. https://doi.org/10.1109/ASICON66040.2025.11326063","raw_type":"info:eu-repo/semantics/conferenceObject"},{"id":"pmh:oai:spiral.imperial.ac.uk:10044/1/124656","doi":null,"is_oa":true,"landing_page_url":"https://hdl.handle.net/10044/1/124656","pdf_url":null,"source":{"id":"https://openalex.org/S4306401396","display_name":"Spiral (Imperial College London)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I47508984","host_organization_name":"Imperial College London","host_organization_lineage":["https://openalex.org/I47508984"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"2025 IEEE 16th International Conference on ASIC","raw_type":"Conference Paper"}],"best_oa_location":{"id":"pmh:oai:spiral.imperial.ac.uk:10044/1/124656","doi":null,"is_oa":true,"landing_page_url":"https://hdl.handle.net/10044/1/124656","pdf_url":null,"source":{"id":"https://openalex.org/S4306401396","display_name":"Spiral (Imperial College London)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I47508984","host_organization_name":"Imperial College London","host_organization_lineage":["https://openalex.org/I47508984"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"2025 IEEE 16th International Conference on ASIC","raw_type":"Conference Paper"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":12,"referenced_works":["https://openalex.org/W3006881356","https://openalex.org/W4200157451","https://openalex.org/W4319918345","https://openalex.org/W4390880528","https://openalex.org/W4393578753","https://openalex.org/W4402727233","https://openalex.org/W4405909140","https://openalex.org/W4408182386","https://openalex.org/W4409132079","https://openalex.org/W4409282707","https://openalex.org/W4410553409","https://openalex.org/W7124249593"],"related_works":["https://openalex.org/W4414620442","https://openalex.org/W4405909140","https://openalex.org/W4408860991","https://openalex.org/W4393578753","https://openalex.org/W7164188446","https://openalex.org/W4416157303","https://openalex.org/W4403577857","https://openalex.org/W4393578447","https://openalex.org/W4393949386","https://openalex.org/W7127922577"],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"deliver":[4],"state-of-the-art":[5],"performance":[6],"but":[7],"demand":[8],"high":[9],"computation":[10],"and":[11,25,65,101],"memory,":[12],"making":[13],"deployment":[14],"in":[15],"resource-limited":[16],"settings":[17],"challenging.":[18],"Field-Programmable":[19],"Gate":[20],"Arrays":[21],"(FPGAs)":[22],"offer":[23],"parallelism":[24],"efficiency,":[26],"yet":[27],"most":[28],"prior":[29],"FPGA":[30,118],"accelerators":[31],"rely":[32],"on":[33,86],"low-level,":[34],"platform-specific":[35],"flows":[36],"that":[37],"hinder":[38],"portability.":[39],"This":[40],"work":[41],"presents":[42],"oneLLM,":[43],"to":[44],"our":[45],"knowledge,":[46],"the":[47,114],"first":[48],"FPGA-based":[49],"LLM":[50,121],"inference":[51],"design":[52],"using":[53],"Intel\u2019s":[54],"oneAPI,":[55],"enabling":[56],"a":[57,98,106],"unified":[58],"high-level":[59],"programming":[60],"model":[61],"across":[62],"CPUs,":[63],"GPUs,":[64],"FPGAs.":[66],"Our":[67],"deeply":[68],"pipelined,":[69],"multi-kernel":[70],"hardware":[71],"architecture":[72],"connects":[73],"specialized":[74],"kernels":[75],"via":[76],"oneAPI":[77],"pipes":[78],"for":[79,120],"on-chip":[80],"streaming,":[81],"reducing":[82],"host\u2013device":[83],"communication.":[84],"Implemented":[85],"an":[87],"Intel":[88],"Agilex":[89],"7":[90],"FPGA,":[91],"it":[92],"achieves":[93],"3":[94],"times":[95,103],"faster":[96,104],"than":[97,105],"CPU":[99],"implementation,":[100],"8.8":[102],"non-pipelined":[107],"baseline":[108],"while":[109],"meeting":[110],"resource":[111],"constraints,":[112],"demonstrating":[113],"potential":[115],"of":[116],"portable":[117],"development":[119],"acceleration.":[122],"Code":[123],"available":[124],"at":[125],"https://github.com/custom-computing-ic/llm-oneapi-fpga.":[126]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":1}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-01-21T00:00:00"}