{"id":"https://openalex.org/W7163674489","doi":"https://doi.org/10.48550/arxiv.2606.05362","title":"MOSAIC: A Workload-Driven Simulation and Design-Space Exploration Framework for Heterogeneous NPUs","display_name":"MOSAIC: A Workload-Driven Simulation and Design-Space Exploration Framework for Heterogeneous NPUs","publication_year":2026,"publication_date":"2026-06-03","ids":{"openalex":"https://openalex.org/W7163674489","doi":"https://doi.org/10.48550/arxiv.2606.05362"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.05362","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05362","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.05362","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5102891312","display_name":"Arghadip Das","orcid":"https://orcid.org/0000-0001-6043-4685"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Das, Arghadip","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137963735","display_name":"Hoseok Kim","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kim, Hoseok","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137929945","display_name":"Soomin Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Soomin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066464351","display_name":"Arnab Raha","orcid":"https://orcid.org/0000-0002-8848-1069"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Raha, Arnab","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5052021148","display_name":"Deepak A. Mathaikutty","orcid":"https://orcid.org/0009-0004-3768-1337"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mathaikutty, Deepak A","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5102009759","display_name":"Vijay Raghunathan","orcid":"https://orcid.org/0000-0003-4713-5386"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Raghunathan, Vijay","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"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":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10502","display_name":"Advanced Memory and Neural Computing","score":0.258899986743927,"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":0.258899986743927,"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/T11948","display_name":"Machine Learning in Materials Science","score":0.13439999520778656,"subfield":{"id":"https://openalex.org/subfields/2505","display_name":"Materials Chemistry"},"field":{"id":"https://openalex.org/fields/25","display_name":"Materials Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.10220000147819519,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.713699996471405},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5005999803543091},{"id":"https://openalex.org/keywords/matrix-multiplication","display_name":"Matrix multiplication","score":0.4925999939441681},{"id":"https://openalex.org/keywords/tile","display_name":"Tile","score":0.47290000319480896},{"id":"https://openalex.org/keywords/homogeneous","display_name":"Homogeneous","score":0.45559999346733093},{"id":"https://openalex.org/keywords/base","display_name":"Base (topology)","score":0.38850000500679016},{"id":"https://openalex.org/keywords/matrix","display_name":"Matrix (chemical analysis)","score":0.38350000977516174},{"id":"https://openalex.org/keywords/convolution","display_name":"Convolution (computer science)","score":0.3711000084877014},{"id":"https://openalex.org/keywords/transformation-matrix","display_name":"Transformation matrix","score":0.37040001153945923}],"concepts":[{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.713699996471405},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6901000142097473},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5005999803543091},{"id":"https://openalex.org/C17349429","wikidata":"https://www.wikidata.org/wiki/Q1049914","display_name":"Matrix multiplication","level":3,"score":0.4925999939441681},{"id":"https://openalex.org/C2780728851","wikidata":"https://www.wikidata.org/wiki/Q468402","display_name":"Tile","level":2,"score":0.47290000319480896},{"id":"https://openalex.org/C66882249","wikidata":"https://www.wikidata.org/wiki/Q169336","display_name":"Homogeneous","level":2,"score":0.45559999346733093},{"id":"https://openalex.org/C459310","wikidata":"https://www.wikidata.org/wiki/Q117801","display_name":"Computational science","level":1,"score":0.39340001344680786},{"id":"https://openalex.org/C42058472","wikidata":"https://www.wikidata.org/wiki/Q810214","display_name":"Base (topology)","level":2,"score":0.38850000500679016},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3855000138282776},{"id":"https://openalex.org/C106487976","wikidata":"https://www.wikidata.org/wiki/Q685816","display_name":"Matrix (chemical analysis)","level":2,"score":0.38350000977516174},{"id":"https://openalex.org/C45347329","wikidata":"https://www.wikidata.org/wiki/Q5166604","display_name":"Convolution (computer science)","level":3,"score":0.3711000084877014},{"id":"https://openalex.org/C165443888","wikidata":"https://www.wikidata.org/wiki/Q1482183","display_name":"Transformation matrix","level":3,"score":0.37040001153945923},{"id":"https://openalex.org/C75172450","wikidata":"https://www.wikidata.org/wiki/Q623950","display_name":"Fast Fourier transform","level":2,"score":0.3626999855041504},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.35370001196861267},{"id":"https://openalex.org/C62611344","wikidata":"https://www.wikidata.org/wiki/Q1062658","display_name":"Node (physics)","level":2,"score":0.3400999903678894},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.3343999981880188},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3343000113964081},{"id":"https://openalex.org/C111030470","wikidata":"https://www.wikidata.org/wiki/Q1430460","display_name":"Curse of dimensionality","level":2,"score":0.33250001072883606},{"id":"https://openalex.org/C57733114","wikidata":"https://www.wikidata.org/wiki/Q1006032","display_name":"Discrete Fourier transform (general)","level":5,"score":0.32109999656677246},{"id":"https://openalex.org/C2780595030","wikidata":"https://www.wikidata.org/wiki/Q3860309","display_name":"Multiplication (music)","level":2,"score":0.3208000063896179},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.3206999897956848},{"id":"https://openalex.org/C13336665","wikidata":"https://www.wikidata.org/wiki/Q125977","display_name":"Vector space","level":2,"score":0.2978000044822693},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.29649999737739563},{"id":"https://openalex.org/C28719098","wikidata":"https://www.wikidata.org/wiki/Q44946","display_name":"Point (geometry)","level":2,"score":0.28610000014305115},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.28279998898506165},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.28209999203681946},{"id":"https://openalex.org/C81363708","wikidata":"https://www.wikidata.org/wiki/Q17084460","display_name":"Convolutional neural network","level":2,"score":0.2718999981880188},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.26489999890327454},{"id":"https://openalex.org/C102519508","wikidata":"https://www.wikidata.org/wiki/Q6520159","display_name":"Fourier transform","level":2,"score":0.26170000433921814},{"id":"https://openalex.org/C172430144","wikidata":"https://www.wikidata.org/wiki/Q17111997","display_name":"Symmetric multiprocessor system","level":2,"score":0.25940001010894775},{"id":"https://openalex.org/C21308566","wikidata":"https://www.wikidata.org/wiki/Q7169365","display_name":"Permutation (music)","level":2,"score":0.2540999948978424},{"id":"https://openalex.org/C104267543","wikidata":"https://www.wikidata.org/wiki/Q208163","display_name":"Signal processing","level":3,"score":0.2522999942302704},{"id":"https://openalex.org/C141353440","wikidata":"https://www.wikidata.org/wiki/Q182221","display_name":"Fuse (electrical)","level":2,"score":0.25110000371932983}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.05362","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05362","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.05362","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05362","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"score":0.42767223715782166,"display_name":"Affordable and clean energy","id":"https://metadata.un.org/sdg/7"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"AI":[0],"model":[1,163],"architectures":[2,16],"are":[3,32,84],"diversifying":[4],"rapidly.":[5],"Although":[6],"dense":[7],"matrix":[8],"multiplication":[9],"underlies":[10],"today's":[11],"CNNs":[12],"and":[13,29,45,83,111,130,140,150,178,181,215],"transformers,":[14],"emerging":[15],"(state-space":[17],"models,":[18],"long":[19],"convolutions":[20],"via":[21],"the":[22,121,227,244],"fast":[23],"Fourier":[24],"transform":[25],"(FFT),":[26],"Kolmogorov-Arnold":[27],"networks,":[28],"spiking":[30],"networks)":[31],"not":[33],"multiply-accumulate":[34],"(MAC)":[35],"dominated;":[36],"they":[37],"spend":[38],"much":[39],"of":[40,124,187],"their":[41,174],"computation":[42],"on":[43,86],"vector":[44],"non-MAC":[46,141,154,171],"primitives":[47],"that":[48,162],"homogeneous,":[49],"MAC-centric":[50],"neural":[51],"processing":[52],"units":[53,152],"(NPUs)":[54],"serve":[55],"poorly.":[56],"This":[57],"has":[58],"motivated":[59],"heterogeneous":[60,68],"NPUs":[61],"(HPUs)":[62],"built":[63],"from":[64],"non-identical":[65],"tiles.":[66],"Prior":[67],"designs":[69],"vary":[70],"only":[71],"one":[72],"or":[73,80],"two":[74],"coarse":[75],"knobs":[76],"(typically":[77],"MAC":[78,147],"precision":[79],"array":[81,128],"size)":[82],"evaluated":[85],"narrow":[87],"workloads;":[88],"no":[89],"existing":[90],"framework":[91,114],"supports":[92],"fine-grained":[93],"HPU":[94,116,230],"design,":[95],"where":[96],"tiles":[97,172,188],"differ":[98],"across":[99,184],"many":[100],"architectural":[101],"dimensions":[102],"at":[103],"once.":[104],"We":[105],"present":[106],"MOSAIC,":[107],"an":[108],"analytical":[109],"simulator":[110],"design-space-exploration":[112],"(DSE)":[113],"for":[115,153],"microarchitecture":[117],"design.":[118],"MOSAIC":[119,169,233],"searches":[120],"joint":[122],"space":[123],"tile-level":[125],"heterogeneity:":[126],"beyond":[127],"size":[129],"precision,":[131],"it":[132],"varies":[133],"tile-type":[134],"composition":[135],"(large":[136],"Big,":[137],"small":[138],"Little,":[139],"Special-Function":[142],"tiles),":[143],"dataflow,":[144],"sparsity":[145],"mode,":[146],"engine":[148],"type,":[149,168],"special-function":[151],"operators":[155,183],"(FFT,":[156],"spiking-integrate,":[157],"polynomial).":[158],"Unlike":[159],"prior":[160],"simulators":[161],"a":[164,185,190,197,211,224],"single":[165],"homogeneous":[166,247],"tile":[167],"models":[170,180,208],"with":[173,189,200,206],"own":[175],"energy,":[176],"area,":[177],"timing":[179],"maps":[182],"mix":[186],"heterogeneity-aware":[191],"compiler.":[192],"A":[193],"multi-seed":[194],"pipeline":[195],"pairing":[196],"stratified":[198],"sweep":[199],"genetic-algorithm":[201],"refinement":[202],"returns":[203],"Pareto-optimal":[204],"designs,":[205],"cost":[207],"calibrated":[209],"to":[210],"7":[212],"nm":[213],"node":[214],"cross-validated":[216],"against":[217],"NVIDIA's":[218],"Deep":[219],"Learning":[220],"Accelerator":[221],"(NVDLA).":[222],"Across":[223],"20-workload":[225],"suite,":[226],"best":[228,245],"general-purpose":[229],"found":[231],"by":[232],"(~200":[234],"mm^2":[235],"Big+Little+Special-Function)":[236],"achieves":[237],"+46.91%":[238],"mean":[239],"iso-area":[240,246],"energy":[241],"savings":[242],"over":[243],"baseline.":[248]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-06T00:00:00"}