{"id":"https://openalex.org/W4413039156","doi":"https://doi.org/10.1145/3759458","title":"HMSA: High-Performance Heterogeneous Mixed-Precision CNN Systolic Array Accelerator on FPGA","display_name":"HMSA: High-Performance Heterogeneous Mixed-Precision CNN Systolic Array Accelerator on FPGA","publication_year":2025,"publication_date":"2025-08-07","ids":{"openalex":"https://openalex.org/W4413039156","doi":"https://doi.org/10.1145/3759458"},"language":"en","primary_location":{"id":"doi:10.1145/3759458","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3759458","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3759458","source":{"id":"https://openalex.org/S136160450","display_name":"ACM Transactions on Embedded Computing Systems","issn_l":"1539-9087","issn":["1539-9087","1558-3465"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319798","host_organization_name":"Association for Computing Machinery","host_organization_lineage":["https://openalex.org/P4310319798"],"host_organization_lineage_names":["Association for Computing Machinery"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ACM Transactions on Embedded Computing Systems","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"bronze","oa_url":"https://dl.acm.org/doi/pdf/10.1145/3759458","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5068652901","display_name":"Yongxiang Cao","orcid":"https://orcid.org/0009-0000-2020-5116"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yongxiang Cao","raw_affiliation_strings":["School of computer science and engineering, Beihang University","School of computer science and engineering, Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0000-2020-5116","affiliations":[{"raw_affiliation_string":"School of computer science and engineering, Beihang University","institution_ids":["https://openalex.org/I82880672"]},{"raw_affiliation_string":"School of computer science and engineering, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048928060","display_name":"Hongxu Jiang","orcid":"https://orcid.org/0000-0001-6633-6605"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hongxu Jiang","raw_affiliation_strings":["School of computer science and engineering, Beihang University","School of computer science and engineering, Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-6633-6605","affiliations":[{"raw_affiliation_string":"School of computer science and engineering, Beihang University","institution_ids":["https://openalex.org/I82880672"]},{"raw_affiliation_string":"School of computer science and engineering, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5082656879","display_name":"Huiyong Li","orcid":"https://orcid.org/0000-0003-0518-7313"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Huiyong Li","raw_affiliation_strings":["School of Artificial Intelligence, Beihang University","School of Artificial Intelligence, Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0003-0518-7313","affiliations":[{"raw_affiliation_string":"School of Artificial Intelligence, Beihang University","institution_ids":["https://openalex.org/I82880672"]},{"raw_affiliation_string":"School of Artificial Intelligence, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Yu Tang","orcid":"https://orcid.org/0009-0009-8186-4567"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yu Tang","raw_affiliation_strings":["School of computer science and engineering, Beihang University","School of computer science and engineering, Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0009-8186-4567","affiliations":[{"raw_affiliation_string":"School of computer science and engineering, Beihang University","institution_ids":["https://openalex.org/I82880672"]},{"raw_affiliation_string":"School of computer science and engineering, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Dongcheng Shi","orcid":"https://orcid.org/0009-0008-1905-3870"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Dongcheng Shi","raw_affiliation_strings":["School of computer science and engineering, Beihang University","School of computer science and engineering, Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0008-1905-3870","affiliations":[{"raw_affiliation_string":"School of computer science and engineering, Beihang University","institution_ids":["https://openalex.org/I82880672"]},{"raw_affiliation_string":"School of computer science and engineering, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5001952974","display_name":"Guocheng Zhao","orcid":null},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Guocheng Zhao","raw_affiliation_strings":["School of computer science and engineering, Beihang University","School of computer science and engineering, Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0001-5477-3020","affiliations":[{"raw_affiliation_string":"School of computer science and engineering, Beihang University","institution_ids":["https://openalex.org/I82880672"]},{"raw_affiliation_string":"School of computer science and engineering, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I82880672"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.15396235,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"24","issue":"5s","first_page":"1","last_page":"27"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10502","display_name":"Advanced Memory and Neural Computing","score":0.9990000128746033,"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":0.9990000128746033,"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":0.9975000023841858,"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"}},{"id":"https://openalex.org/T11992","display_name":"CCD and CMOS Imaging Sensors","score":0.995199978351593,"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/field-programmable-gate-array","display_name":"Field-programmable gate array","score":0.802705705165863},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.787268340587616},{"id":"https://openalex.org/keywords/systolic-array","display_name":"Systolic array","score":0.5737700462341309},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.5583215951919556},{"id":"https://openalex.org/keywords/leverage","display_name":"Leverage (statistics)","score":0.5341479182243347},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.5215601921081543},{"id":"https://openalex.org/keywords/hardware-acceleration","display_name":"Hardware acceleration","score":0.49901437759399414},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.47433915734291077},{"id":"https://openalex.org/keywords/reconfigurability","display_name":"Reconfigurability","score":0.4614744186401367},{"id":"https://openalex.org/keywords/convolutional-neural-network","display_name":"Convolutional neural network","score":0.4606615900993347},{"id":"https://openalex.org/keywords/digital-signal-processing","display_name":"Digital signal processing","score":0.4574216902256012},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.4570811092853546},{"id":"https://openalex.org/keywords/lookup-table","display_name":"Lookup table","score":0.4448254704475403},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.438265323638916},{"id":"https://openalex.org/keywords/gate-array","display_name":"Gate array","score":0.41078975796699524},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.3674507141113281},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.3668696880340576},{"id":"https://openalex.org/keywords/computer-architecture","display_name":"Computer architecture","score":0.3562019169330597},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.30433541536331177},{"id":"https://openalex.org/keywords/computer-hardware","display_name":"Computer hardware","score":0.2980935573577881},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.1536608636379242}],"concepts":[{"id":"https://openalex.org/C42935608","wikidata":"https://www.wikidata.org/wiki/Q190411","display_name":"Field-programmable gate array","level":2,"score":0.802705705165863},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.787268340587616},{"id":"https://openalex.org/C150741067","wikidata":"https://www.wikidata.org/wiki/Q2377218","display_name":"Systolic array","level":3,"score":0.5737700462341309},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.5583215951919556},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.5341479182243347},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.5215601921081543},{"id":"https://openalex.org/C13164978","wikidata":"https://www.wikidata.org/wiki/Q600158","display_name":"Hardware acceleration","level":3,"score":0.49901437759399414},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.47433915734291077},{"id":"https://openalex.org/C2780149590","wikidata":"https://www.wikidata.org/wiki/Q7302742","display_name":"Reconfigurability","level":2,"score":0.4614744186401367},{"id":"https://openalex.org/C81363708","wikidata":"https://www.wikidata.org/wiki/Q17084460","display_name":"Convolutional neural network","level":2,"score":0.4606615900993347},{"id":"https://openalex.org/C84462506","wikidata":"https://www.wikidata.org/wiki/Q173142","display_name":"Digital signal processing","level":2,"score":0.4574216902256012},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.4570811092853546},{"id":"https://openalex.org/C134835016","wikidata":"https://www.wikidata.org/wiki/Q690265","display_name":"Lookup table","level":2,"score":0.4448254704475403},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.438265323638916},{"id":"https://openalex.org/C114237110","wikidata":"https://www.wikidata.org/wiki/Q114901","display_name":"Gate array","level":3,"score":0.41078975796699524},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.3674507141113281},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.3668696880340576},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.3562019169330597},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.30433541536331177},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.2980935573577881},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.1536608636379242},{"id":"https://openalex.org/C76155785","wikidata":"https://www.wikidata.org/wiki/Q418","display_name":"Telecommunications","level":1,"score":0.0},{"id":"https://openalex.org/C14580979","wikidata":"https://www.wikidata.org/wiki/Q876049","display_name":"Very-large-scale integration","level":2,"score":0.0},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.0},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3759458","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3759458","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3759458","source":{"id":"https://openalex.org/S136160450","display_name":"ACM Transactions on Embedded Computing Systems","issn_l":"1539-9087","issn":["1539-9087","1558-3465"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319798","host_organization_name":"Association for Computing Machinery","host_organization_lineage":["https://openalex.org/P4310319798"],"host_organization_lineage_names":["Association for Computing Machinery"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ACM Transactions on Embedded Computing Systems","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1145/3759458","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3759458","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3759458","source":{"id":"https://openalex.org/S136160450","display_name":"ACM Transactions on Embedded Computing Systems","issn_l":"1539-9087","issn":["1539-9087","1558-3465"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319798","host_organization_name":"Association for Computing Machinery","host_organization_lineage":["https://openalex.org/P4310319798"],"host_organization_lineage_names":["Association for Computing Machinery"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ACM Transactions on Embedded Computing Systems","raw_type":"journal-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4413039156.pdf","grobid_xml":"https://content.openalex.org/works/W4413039156.grobid-xml"},"referenced_works_count":35,"referenced_works":["https://openalex.org/W2009295221","https://openalex.org/W2018758602","https://openalex.org/W2276486856","https://openalex.org/W2594492285","https://openalex.org/W2794141774","https://openalex.org/W2794986474","https://openalex.org/W2903735800","https://openalex.org/W2952857977","https://openalex.org/W2963367920","https://openalex.org/W2982479999","https://openalex.org/W2987956272","https://openalex.org/W3158020960","https://openalex.org/W3206749822","https://openalex.org/W4206556022","https://openalex.org/W4211085526","https://openalex.org/W4211090839","https://openalex.org/W4211095909","https://openalex.org/W4290802892","https://openalex.org/W4293149165","https://openalex.org/W4321637080","https://openalex.org/W4366821684","https://openalex.org/W4383749366","https://openalex.org/W4387526806","https://openalex.org/W4390431261","https://openalex.org/W4391429102","https://openalex.org/W4393318691","https://openalex.org/W4396781678","https://openalex.org/W4398249887","https://openalex.org/W4400233374","https://openalex.org/W4400314919","https://openalex.org/W4401508067","https://openalex.org/W4402742487","https://openalex.org/W4402835945","https://openalex.org/W4405520952","https://openalex.org/W4408903096"],"related_works":["https://openalex.org/W2078268640","https://openalex.org/W4319439883","https://openalex.org/W4285429137","https://openalex.org/W4293243607","https://openalex.org/W4364860222","https://openalex.org/W369161638","https://openalex.org/W2039824819","https://openalex.org/W2065192738","https://openalex.org/W1547795075","https://openalex.org/W2952433095"],"abstract_inverted_index":{"In":[0,225],"power-constrained":[1],"and":[2,25,37,97,110,281,292,299],"real-time-demanding":[3],"embedded":[4,179],"scenarios,":[5],"Field-Programmable":[6],"Gate":[7],"Arrays":[8],"(FPGAs)":[9],"emerge":[10],"as":[11],"ideal":[12],"options":[13],"for":[14,235],"accelerating":[15],"neural":[16,101,176,275],"network":[17,102],"inference,":[18],"owing":[19],"to":[20,49,60,92,129,143,165,244,288],"the":[21,52,61,69,105,121,133,171,197,228,257,296,304,312],"reconfigurability,":[22],"high":[23,215],"reliability,":[24],"flexibility":[26],"of":[27,55,175,201,220,227,265],"FPGAs.":[28,180],"Mixed":[29],"precision":[30,274],"quantization":[31,135,168,302],"technology":[32],"significantly":[33],"reduces":[34,311],"computational":[35,83],"complexity":[36],"bandwidth":[38],"requirements":[39],"while":[40],"preserving":[41],"model":[42,145,313],"accuracy.":[43],"However,":[44,104],"existing":[45,73],"FPGA":[46,240,260],"accelerators":[47],"fail":[48],"fully":[50,127],"leverage":[51,166],"parallel":[53,95,198],"advantages":[54],"mixed":[56,273,301],"precision,":[57],"which":[58,195],"leads":[59,142],"actual":[62,285],"inference":[63,173,250],"speedup":[64],"being":[65],"markedly":[66],"lower":[67],"than":[68],"theoretical":[70],"prediction.":[71],"Reviewing":[72],"methods,":[74],"we":[75,152,230],"found":[76],"three":[77],"main":[78],"drawbacks":[79],"in":[80,100,284],"enhancing":[81,170],"practical":[82,172],"performance.":[84],"Firstly,":[85],"FPGAs":[86],"primarily":[87],"rely":[88],"on":[89,178,211,239,256],"DSP":[90,107,185],"slices":[91],"achieve":[93],"high-performance":[94,155],"multiplication":[96],"accumulation":[98],"(MAC)":[99],"inference.":[103],"current":[106],"PE":[108],"design":[109],"data":[111,186],"packing":[112,187],"methods":[113],"are":[114,125],"not":[115,126],"compatible":[116],"with":[117,214,295],"mixed-precision":[118,134,157,167],"models.":[119],"Secondly,":[120],"remaining":[122],"logic":[123],"resources":[124],"utilized":[128],"accelerate":[130],"computations.":[131],"Thirdly,":[132],"bit-width":[136,236,307],"selection":[137,237,308],"method":[138,188,234,309],"without":[139,248],"hardware-guided":[140],"guidance":[141],"additional":[144],"accuracy":[146,246,314],"loss.":[147,315],"To":[148],"address":[149],"these":[150],"challenges,":[151],"propose":[153,182,204,231],"a":[154,191,205,262],"heterogeneous":[156,206,222],"systolic":[158],"array":[159],"(SA)":[160],"accelerator,":[161],"HMSA.":[162],"It":[163],"aims":[164],"fully,":[169],"efficiency":[174],"networks":[177],"We":[181,203],"an":[183,232],"optimized":[184,306],"guided":[189],"by":[190],"resource-performance":[192],"cost":[193],"model,":[194],"enhances":[196],"computing":[199,223],"performance":[200],"accelerators.":[202],"convolutional":[207],"acceleration":[208],"architecture":[209,213],"based":[210,238],"SA":[212],"scalability,":[216],"enabling":[217],"efficient":[218],"utilization":[219],"FPGA\u2019s":[221],"resources.":[224],"terms":[226],"algorithm,":[229],"optimization":[233],"hardware":[241],"architecture,":[242],"aiming":[243],"avoid":[245],"degradation":[247],"improving":[249],"speed.":[251],"Experiments":[252],"confirm":[253],"that":[254],"HMSA":[255,277],"Xilinx":[258],"XC7VX690T":[259],"reaches":[261],"peak":[263],"throughput":[264],"6.385":[266],"TOP/s":[267],"at":[268],"W1A8":[269],"precision.":[270],"When":[271],"inferring":[272],"networks,":[276],"achieves":[278],"3.53\u00d7,":[279],"5.46\u00d7,":[280],"1.58\u00d7":[282],"improvements":[283],"throughput/DSP":[286],"compared":[287],"state-of-the-art":[289,297],"MPA,":[290],"MSD,":[291],"MP-OPU.":[293],"Compared":[294],"MBFQuant":[298],"Edge-MPQ":[300],"algorithms,":[303],"proposed":[305],"effectively":[310]},"counts_by_year":[],"updated_date":"2026-05-21T06:26:12.895304","created_date":"2025-10-10T00:00:00"}