{"id":"https://openalex.org/W7135018308","doi":"https://doi.org/10.1109/tc.2026.3672901","title":"FireFly-T: High-Throughput Sparsity Exploitation for Spiking Transformer Acceleration With Dual-Engine Overlay Architecture","display_name":"FireFly-T: High-Throughput Sparsity Exploitation for Spiking Transformer Acceleration With Dual-Engine Overlay Architecture","publication_year":2026,"publication_date":"2026-03-12","ids":{"openalex":"https://openalex.org/W7135018308","doi":"https://doi.org/10.1109/tc.2026.3672901"},"language":null,"primary_location":{"id":"doi:10.1109/tc.2026.3672901","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tc.2026.3672901","pdf_url":null,"source":{"id":"https://openalex.org/S157670870","display_name":"IEEE Transactions on Computers","issn_l":"0018-9340","issn":["0018-9340","0016-9340","1557-9956","2326-3814"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Computers","raw_type":"journal-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/A5128909257","display_name":"Tenglong Li","orcid":null},"institutions":[{"id":"https://openalex.org/I4210100255","display_name":"Beijing Academy of Artificial Intelligence","ror":"https://ror.org/016a74861","country_code":"CN","type":"other","lineage":["https://openalex.org/I4210100255"]},{"id":"https://openalex.org/I4210165038","display_name":"University of Chinese Academy of Sciences","ror":"https://ror.org/05qbk4x57","country_code":"CN","type":"education","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210165038"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Tenglong Li","raw_affiliation_strings":["School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0007-3266-2075","affiliations":[{"raw_affiliation_string":"School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210100255","https://openalex.org/I4210165038"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124318978","display_name":"Jindong Li","orcid":null},"institutions":[{"id":"https://openalex.org/I4210100255","display_name":"Beijing Academy of Artificial Intelligence","ror":"https://ror.org/016a74861","country_code":"CN","type":"other","lineage":["https://openalex.org/I4210100255"]},{"id":"https://openalex.org/I4210165038","display_name":"University of Chinese Academy of Sciences","ror":"https://ror.org/05qbk4x57","country_code":"CN","type":"education","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210165038"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jindong Li","raw_affiliation_strings":["School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-4009-916X","affiliations":[{"raw_affiliation_string":"School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210100255","https://openalex.org/I4210165038"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103702012","display_name":"Ge Shen","orcid":null},"institutions":[{"id":"https://openalex.org/I4210165038","display_name":"University of Chinese Academy of Sciences","ror":"https://ror.org/05qbk4x57","country_code":"CN","type":"education","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210165038"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Guobin Shen","raw_affiliation_strings":["School of Future Technology, University of Chinese Academy of Sciences, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-4069-2107","affiliations":[{"raw_affiliation_string":"School of Future Technology, University of Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210165038"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128814758","display_name":"Dongcheng Zhao","orcid":null},"institutions":[{"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":"Dongcheng Zhao","raw_affiliation_strings":["Brain-inspired Cognitive AI Lab, Institute of Automation, Chinese Academy of Sciences, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-0593-8650","affiliations":[{"raw_affiliation_string":"Brain-inspired Cognitive AI Lab, Institute of Automation, Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210112150"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Qian Zhang","orcid":"https://orcid.org/0000-0001-5314-4233"},"institutions":[{"id":"https://openalex.org/I4210100255","display_name":"Beijing Academy of Artificial Intelligence","ror":"https://ror.org/016a74861","country_code":"CN","type":"other","lineage":["https://openalex.org/I4210100255"]},{"id":"https://openalex.org/I4210165038","display_name":"University of Chinese Academy of Sciences","ror":"https://ror.org/05qbk4x57","country_code":"CN","type":"education","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210165038"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Qian Zhang","raw_affiliation_strings":["School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-5314-4233","affiliations":[{"raw_affiliation_string":"School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210100255","https://openalex.org/I4210165038"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5128907089","display_name":"Yi Zeng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yi Zeng","raw_affiliation_strings":["Center for Long-term AI, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-9595-9091","affiliations":[{"raw_affiliation_string":"Center for Long-term AI, Beijing, China","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":4.828,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.93475986,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":94,"max":97},"biblio":{"volume":"75","issue":"6","first_page":"2185","last_page":"2199"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10502","display_name":"Advanced Memory and Neural Computing","score":0.11289999634027481,"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"}},"topics":[{"id":"https://openalex.org/T10502","display_name":"Advanced Memory and Neural Computing","score":0.11289999634027481,"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/T10623","display_name":"Thin-Film Transistor Technologies","score":0.10570000112056732,"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/T12510","display_name":"Magneto-Optical Properties and Applications","score":0.08990000188350677,"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/acceleration","display_name":"Acceleration","score":0.5407000184059143},{"id":"https://openalex.org/keywords/architecture","display_name":"Architecture","score":0.5335000157356262},{"id":"https://openalex.org/keywords/overlay","display_name":"Overlay","score":0.5170999765396118},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.4521999955177307},{"id":"https://openalex.org/keywords/parallel-architecture","display_name":"Parallel architecture","score":0.44200000166893005}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7702000141143799},{"id":"https://openalex.org/C117896860","wikidata":"https://www.wikidata.org/wiki/Q11376","display_name":"Acceleration","level":2,"score":0.5407000184059143},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.5335000157356262},{"id":"https://openalex.org/C136085584","wikidata":"https://www.wikidata.org/wiki/Q910289","display_name":"Overlay","level":2,"score":0.5170999765396118},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.4521999955177307},{"id":"https://openalex.org/C2985918086","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel architecture","level":3,"score":0.44200000166893005},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.35370001196861267},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.35109999775886536},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.34880000352859497},{"id":"https://openalex.org/C24326235","wikidata":"https://www.wikidata.org/wiki/Q126095","display_name":"Electronic engineering","level":1,"score":0.3409999907016754},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.31520000100135803},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.3125999867916107},{"id":"https://openalex.org/C104267543","wikidata":"https://www.wikidata.org/wiki/Q208163","display_name":"Signal processing","level":3,"score":0.29330000281333923},{"id":"https://openalex.org/C98025372","wikidata":"https://www.wikidata.org/wiki/Q477538","display_name":"Systems architecture","level":3,"score":0.25189998745918274}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/tc.2026.3672901","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tc.2026.3672901","pdf_url":null,"source":{"id":"https://openalex.org/S157670870","display_name":"IEEE Transactions on Computers","issn_l":"0018-9340","issn":["0018-9340","0016-9340","1557-9956","2326-3814"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Computers","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.7018916606903076,"id":"https://metadata.un.org/sdg/7","display_name":"Affordable and clean energy"}],"awards":[{"id":"https://openalex.org/G2927787986","display_name":null,"funder_award_id":"E411230101","funder_id":"https://openalex.org/F4320335639","funder_display_name":"Institute of Automation, Chinese Academy of Sciences"}],"funders":[{"id":"https://openalex.org/F4320335639","display_name":"Institute of Automation, Chinese Academy of Sciences","ror":null}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Spiking":[0,14],"transformers":[1],"are":[2],"emerging":[3],"as":[4,212,214,235],"a":[5,56,62,69,81,100,156],"promising":[6],"architecture":[7,59],"that":[8,60,85,178,202],"combines":[9],"the":[10,19,76,120,125,134,148,226],"energy":[11,210],"efficiency":[12],"of":[13,23,129],"Neural":[15],"Networks":[16],"(SNNs)":[17],"with":[18,43,105,141,183],"powerful":[20],"attention":[21,140,154],"mechanisms":[22],"transformers.":[24,242],"However,":[25],"existing":[26],"hardware":[27,238],"accelerators":[28],"lack":[29],"support":[30,115],"for":[31,65,72,138,240],"spiking":[32,73,139,153,241],"attention,":[33],"exhibit":[34],"limited":[35],"throughput":[36],"when":[37],"exploiting":[38],"fine-grained":[39,87],"sparsity,":[40],"and":[41,68,108,162,195,207,216,225],"struggle":[42],"scalable":[44,101,116],"parallelism":[45],"in":[46,152],"sparse":[47,63,77,83],"computation.":[48],"To":[49,95],"address":[50],"these":[51],"challenges,":[52],"we":[53,79,98,123],"propose":[54],"FireFly-T,":[55],"dual-engine":[57],"overlay":[58,171],"integrates":[61],"engine":[64,71],"activation":[66],"sparsity":[67,88],"binary":[70,121],"attention.":[74],"In":[75,119],"engine,":[78,122],"present":[80],"high-throughput":[82],"decoder":[84],"exploits":[86],"by":[89],"concurrently":[90],"extracting":[91],"multiple":[92],"non-zero":[93],"spikes.":[94],"complement":[96],"this,":[97],"introduce":[99],"load":[102],"balancing":[103],"mechanism":[104],"weight":[106],"dispatch":[107],"out-of-order":[109],"execution,":[110],"eliminating":[111],"bank":[112],"conflicts":[113],"to":[114,131,186,222],"multidimensional":[117],"parallelism.":[118],"leverage":[124],"byte-level":[126],"write":[127],"capability":[128],"SRAMs":[130],"efficiently":[132],"manipulate":[133],"3D":[135],"dataflows":[136,182],"required":[137],"minimal":[142],"resource":[143,193],"overhead.":[144],"We":[145],"also":[146],"optimize":[147],"core":[149],"AND-PopCount":[150],"operation":[151],"through":[155],"LUT6-based":[157],"implementation,":[158],"improving":[159],"timing":[160],"closure":[161],"reducing":[163],"LUT":[164],"utilization":[165,194],"on":[166],"Xilinx":[167],"FPGAs.":[168],"As":[169],"an":[170,176,236],"architecture,":[172],"FireFly-T":[173],"further":[174],"incorporates":[175],"orchestrator":[177],"dynamically":[179],"manipulates":[180],"input":[181],"flexible":[184],"adaptation":[185],"diverse":[187],"network":[188],"topologies,":[189],"while":[190],"ensuring":[191],"efficient":[192,237],"maintaining":[196],"high":[197],"throughput.":[198],"Experimental":[199],"results":[200,231],"demonstrate":[201],"our":[203],"accelerator":[204],"achieves":[205],"1.39\u00d7":[206],"2.40\u00d7":[208],"higher":[209],"efficiency,":[211,220],"well":[213],"4.21\u00d7":[215],"7.10\u00d7":[217],"greater":[218],"DSP":[219],"compared":[221],"FireFly":[223],"v2":[224],"transformer-enabled":[227],"SpikeTA,":[228],"respectively.":[229],"These":[230],"highlight":[232],"its":[233],"potential":[234],"platform":[239]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-05-12T06:07:45.972803","created_date":"2026-03-13T00:00:00"}