{"id":"https://openalex.org/W4407953632","doi":"https://doi.org/10.1145/3706628.3708857","title":"Neural Network Inference in High-Performance Computing: Closing the Gap for FINN based Reconfigurable Accelerators","display_name":"Neural Network Inference in High-Performance Computing: Closing the Gap for FINN based Reconfigurable Accelerators","publication_year":2025,"publication_date":"2025-02-26","ids":{"openalex":"https://openalex.org/W4407953632","doi":"https://doi.org/10.1145/3706628.3708857"},"language":"en","primary_location":{"id":"doi:10.1145/3706628.3708857","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3706628.3708857","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2025 ACM/SIGDA International Symposium on Field Programmable Gate Arrays","raw_type":"proceedings-article"},"type":"conference-abstract","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/A5094345769","display_name":"Linus Jungemann","orcid":"https://orcid.org/0009-0003-9757-988X"},"institutions":[{"id":"https://openalex.org/I206945453","display_name":"Paderborn University","ror":"https://ror.org/058kzsd48","country_code":"DE","type":"education","lineage":["https://openalex.org/I206945453"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Linus Jungemann","raw_affiliation_strings":["Paderborn University, Paderborn Center for Parallel Computing, Paderborn, Germany"],"raw_orcid":"https://orcid.org/0009-0003-9757-988X","affiliations":[{"raw_affiliation_string":"Paderborn University, Paderborn Center for Parallel Computing, Paderborn, Germany","institution_ids":["https://openalex.org/I206945453"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5116425625","display_name":"Bjarne Wintermann","orcid":null},"institutions":[{"id":"https://openalex.org/I206945453","display_name":"Paderborn University","ror":"https://ror.org/058kzsd48","country_code":"DE","type":"education","lineage":["https://openalex.org/I206945453"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Bjarne Wintermann","raw_affiliation_strings":["Paderborn University, Paderborn Center for Parallel Computing, Paderborn, Germany"],"raw_orcid":"https://orcid.org/0009-0000-0856-6250","affiliations":[{"raw_affiliation_string":"Paderborn University, Paderborn Center for Parallel Computing, Paderborn, Germany","institution_ids":["https://openalex.org/I206945453"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5062609300","display_name":"Heinrich Riebler","orcid":"https://orcid.org/0000-0002-3811-894X"},"institutions":[{"id":"https://openalex.org/I206945453","display_name":"Paderborn University","ror":"https://ror.org/058kzsd48","country_code":"DE","type":"education","lineage":["https://openalex.org/I206945453"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Heinrich Riebler","raw_affiliation_strings":["Paderborn University, Paderborn Center for Parallel Computing, Paderborn, Germany"],"raw_orcid":"https://orcid.org/0000-0002-3811-894X","affiliations":[{"raw_affiliation_string":"Paderborn University, Paderborn Center for Parallel Computing, Paderborn, Germany","institution_ids":["https://openalex.org/I206945453"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5039484700","display_name":"Christian Plessl","orcid":"https://orcid.org/0000-0001-5728-9982"},"institutions":[{"id":"https://openalex.org/I206945453","display_name":"Paderborn University","ror":"https://ror.org/058kzsd48","country_code":"DE","type":"education","lineage":["https://openalex.org/I206945453"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Christian Plessl","raw_affiliation_strings":["Paderborn University, Paderborn Center for Parallel Computing, Paderborn, Germany"],"raw_orcid":"https://orcid.org/0000-0001-5728-9982","affiliations":[{"raw_affiliation_string":"Paderborn University, Paderborn Center for Parallel Computing, Paderborn, Germany","institution_ids":["https://openalex.org/I206945453"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I206945453"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"52","last_page":"52"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10320","display_name":"Neural Networks and Applications","score":0.946399986743927,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"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/T10320","display_name":"Neural Networks and Applications","score":0.946399986743927,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"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/T11992","display_name":"CCD and CMOS Imaging Sensors","score":0.9138000011444092,"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/closing","display_name":"Closing (real estate)","score":0.8156376481056213},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6946917772293091},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6490646004676819},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.5423241257667542},{"id":"https://openalex.org/keywords/computer-architecture","display_name":"Computer architecture","score":0.37615349888801575},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3332594633102417}],"concepts":[{"id":"https://openalex.org/C2778775528","wikidata":"https://www.wikidata.org/wiki/Q5135432","display_name":"Closing (real estate)","level":2,"score":0.8156376481056213},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6946917772293091},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6490646004676819},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.5423241257667542},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.37615349888801575},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3332594633102417},{"id":"https://openalex.org/C17744445","wikidata":"https://www.wikidata.org/wiki/Q36442","display_name":"Political science","level":0,"score":0.0},{"id":"https://openalex.org/C199539241","wikidata":"https://www.wikidata.org/wiki/Q7748","display_name":"Law","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3706628.3708857","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3706628.3708857","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2025 ACM/SIGDA International Symposium on Field Programmable Gate Arrays","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2116422677","https://openalex.org/W2517007886","https://openalex.org/W2383994331","https://openalex.org/W1522019333","https://openalex.org/W278047738","https://openalex.org/W2076369646","https://openalex.org/W2373884197","https://openalex.org/W4244817184","https://openalex.org/W2961923709","https://openalex.org/W2516969888"],"abstract_inverted_index":{"In":[0],"recent":[1],"years,":[2],"Neural":[3],"Networks":[4],"(NNs)":[5],"have":[6],"become":[7,56],"one":[8],"of":[9,65,76,85,124,145,155,199,246],"the":[10,50,164,168,171,177,213,241,247,253,281,290],"most":[11],"prevailing":[12],"topics":[13],"in":[14,18,21,45,70,73,122,157,170,305],"computers":[15],"science,":[16],"both":[17,102,121],"research":[19],"and":[20,34,43,90,105,110,126,179,187,219,243,252,265,313],"industry.":[22],"NNs":[23,38,61],"are":[24,256],"used":[25,239],"for":[26,67,116,139,147,167,194,270],"data":[27],"analysis,":[28],"natural":[29],"language":[30],"processing,":[31],"autonomous":[32],"driving":[33],"more.":[35],"As":[36],"such,":[37],"also":[39,211],"see":[40],"more":[41],"application":[42],"use":[44,62],"High-Performance":[46,94],"Computing":[47,95],"(HPC).":[48],"At":[49],"same":[51],"time,":[52],"energy":[53,66,106,180,244,316],"efficiency":[54,181],"has":[55],"an":[57,117,134,152,224,235],"increasingly":[58],"critical":[59],"topic.":[60],"large":[63,74,262],"amounts":[64,75],"operation,":[68],"which":[69],"return":[71],"results":[72,286],"CO2":[77],"emissions.":[78],"This":[79],"work":[80,130],"presents":[81],"a":[82,99,206,266,274,296,309],"comprehensive":[83],"evaluation":[84],"current":[86,143],"NN":[87,108,136,192,231],"inference":[88,119,137,232,248],"soft-":[89],"hardware":[91],"configurations":[92],"within":[93],"(HPC)":[96],"environments,":[97],"with":[98,289],"focus":[100],"on":[101,132,215,223,234],"performance":[103,178,242],"metrics":[104],"consumption.":[107],"quantization":[109],"accelerators":[111,193,233],"such":[112],"as":[113],"FPGAs":[114],"allow":[115],"increased":[118],"efficiency,":[120],"terms":[123],"throughput":[125],"energy.":[127],"Therefore,":[128],"this":[129],"focuses":[131],"FINN,":[133,200],"efficient":[135],"framework":[138],"FPGAs,":[140],"highlighting":[141],"its":[142],"lack":[144],"support":[146],"HPC":[148,172],"systems.":[149],"We":[150,174],"provide":[151],"in-depth":[153],"analysis":[154],"FINN":[156],"order":[158],"to":[159,162,204,258,283,303],"implement":[160],"extensions":[161,198],"optimize":[163],"end-to-end":[165],"execution":[166],"usage":[169],"environment.":[173],"thoroughly":[175],"evaluate":[176],"gains":[182],"using":[183,277],"newly":[184,291],"implemented":[185],"optimizations":[186],"compare":[188],"it":[189],"against":[190],"existing":[191],"HPC.":[195],"With":[196],"our":[197],"we":[201,294],"were":[202],"able":[203],"achieve":[205],"1847\u00d7":[207],"higher":[208],"throughput,":[209,306],"while":[210,307],"decreasing":[212],"latency":[214,312],"average":[216],"by":[217,221,301],"0.9978\u00d7":[218],"EDP":[220],"0.9979\u00d7":[222],"Alveo":[225],"U55C":[226],"FPGA.":[227],"Data":[228],"flow":[229],"based":[230],"FPGA":[236],"should":[237],"be":[238],"if":[240],"footprint":[245],"process":[249],"is":[250,279],"crucial,":[251],"batch":[254,263],"sizes":[255,264],"small":[257],"medium.":[259],"For":[260],"extremely":[261],"very":[267],"limited":[268],"time":[269],"network-to-accelerator":[271],"(less":[272],"than":[273],"few":[275],"days),":[276],"GPUs":[278],"still":[280],"way":[282],"go.":[284],"Our":[285],"show":[287],"that":[288],"developed":[292],"driver,":[293],"outperform":[295],"high-end":[297],"Nvidia":[298],"A100":[299],"GPU":[300],"up":[302],"7.81x":[304],"having":[308],"0.87x":[310],"lower":[311,315],"0.88x":[314],"delay":[317],"product.":[318]},"counts_by_year":[],"updated_date":"2026-08-18T07:49:30.821534","created_date":"2025-10-10T00:00:00"}