{"id":"https://openalex.org/W7171996367","doi":"https://doi.org/10.1145/3818671.3818672","title":"Sawtooth Wavefront Reordering: Enhanced CuTile FlashAttention on NVIDIA GB10","display_name":"Sawtooth Wavefront Reordering: Enhanced CuTile FlashAttention on NVIDIA GB10","publication_year":2026,"publication_date":"2026-03-22","ids":{"openalex":"https://openalex.org/W7171996367","doi":"https://doi.org/10.1145/3818671.3818672"},"language":null,"primary_location":{"id":"doi:10.1145/3818671.3818672","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3818671.3818672","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 18th Workshop on General Purpose Processing Using GPU","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1145/3818671.3818672","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5103007367","display_name":"Yifan Zhu","orcid":"https://orcid.org/0009-0000-6718-7787"},"institutions":[{"id":"https://openalex.org/I5388228","display_name":"University of Rochester","ror":"https://ror.org/022kthw22","country_code":"US","type":"education","lineage":["https://openalex.org/I5388228"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yifan Zhu","raw_affiliation_strings":["University of Rochester, Rochester, New York, USA"],"raw_orcid":"https://orcid.org/0009-0000-6718-7787","affiliations":[{"raw_affiliation_string":"University of Rochester, Rochester, New York, USA","institution_ids":["https://openalex.org/I5388228"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5026683719","display_name":"Yekai Pan","orcid":null},"institutions":[{"id":"https://openalex.org/I5388228","display_name":"University of Rochester","ror":"https://ror.org/022kthw22","country_code":"US","type":"education","lineage":["https://openalex.org/I5388228"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yekai Pan","raw_affiliation_strings":["University of Rochester, Rochester, New York, USA"],"raw_orcid":"https://orcid.org/0009-0001-7639-1488","affiliations":[{"raw_affiliation_string":"University of Rochester, Rochester, New York, USA","institution_ids":["https://openalex.org/I5388228"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5002777604","display_name":"Chen Ding","orcid":"https://orcid.org/0000-0003-4968-6659"},"institutions":[{"id":"https://openalex.org/I5388228","display_name":"University of Rochester","ror":"https://ror.org/022kthw22","country_code":"US","type":"education","lineage":["https://openalex.org/I5388228"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Chen Ding","raw_affiliation_strings":["University of Rochester, Rochester, New York, USA"],"raw_orcid":"https://orcid.org/0000-0003-4968-6659","affiliations":[{"raw_affiliation_string":"University of Rochester, Rochester, New York, USA","institution_ids":["https://openalex.org/I5388228"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I5388228"],"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":"1","last_page":"7"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":null,"topics":[],"keywords":[{"id":"https://openalex.org/keywords/sawtooth-wave","display_name":"Sawtooth wave","score":0.8166999816894531},{"id":"https://openalex.org/keywords/cache","display_name":"Cache","score":0.5336999893188477},{"id":"https://openalex.org/keywords/cuda","display_name":"CUDA","score":0.5289999842643738},{"id":"https://openalex.org/keywords/wavefront","display_name":"Wavefront","score":0.5273000001907349},{"id":"https://openalex.org/keywords/throughput","display_name":"Throughput","score":0.5144000053405762}],"concepts":[{"id":"https://openalex.org/C148837635","wikidata":"https://www.wikidata.org/wiki/Q1742397","display_name":"Sawtooth wave","level":2,"score":0.8166999816894531},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7775999903678894},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.6388000249862671},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.5336999893188477},{"id":"https://openalex.org/C2778119891","wikidata":"https://www.wikidata.org/wiki/Q477690","display_name":"CUDA","level":2,"score":0.5289999842643738},{"id":"https://openalex.org/C165699331","wikidata":"https://www.wikidata.org/wiki/Q461533","display_name":"Wavefront","level":2,"score":0.5273000001907349},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.5144000053405762},{"id":"https://openalex.org/C2777526259","wikidata":"https://www.wikidata.org/wiki/Q221836","display_name":"Flash (photography)","level":2,"score":0.4537999927997589},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.4424999952316284},{"id":"https://openalex.org/C459310","wikidata":"https://www.wikidata.org/wiki/Q117801","display_name":"Computational science","level":1,"score":0.31769999861717224},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2915000021457672},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.26339998841285706},{"id":"https://openalex.org/C53016008","wikidata":"https://www.wikidata.org/wiki/Q620167","display_name":"Front and back ends","level":2,"score":0.2547999918460846}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3818671.3818672","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3818671.3818672","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 18th Workshop on General Purpose Processing Using GPU","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3818671.3818672","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3818671.3818672","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 18th Workshop on General Purpose Processing Using GPU","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G2483447800","display_name":"Collaborative Research: SHF: Small: Programmable Hierarchical Caches: Design, Programming, and Prototyping","funder_award_id":"2114319","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G544617644","display_name":"SHF: Small: Data Movement Complexity: Theory and Optimization","funder_award_id":"2217395","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":8,"referenced_works":["https://openalex.org/W2047060659","https://openalex.org/W2079038734","https://openalex.org/W2164705534","https://openalex.org/W2605251767","https://openalex.org/W4237585060","https://openalex.org/W4405348495","https://openalex.org/W4406157566","https://openalex.org/W7133227460"],"related_works":[],"abstract_inverted_index":{"High-performance":[0],"attention":[1],"kernels":[2],"are":[3],"essential":[4],"for":[5],"Large":[6],"Language":[7],"Models.":[8],"This":[9],"paper":[10],"presents":[11],"analysis":[12,30],"of":[13,41],"CuTile-based":[14],"Flash":[15],"Attention":[16],"memory":[17],"behavior":[18],"and":[19,68,78],"a":[20,50],"technique":[21,53],"to":[22,80],"improve":[23],"its":[24],"cache":[25,43],"performance.":[26],"In":[27],"particular,":[28],"our":[29],"on":[31,85],"the":[32,38],"NVIDIA":[33],"GB10":[34],"(Grace":[35],"Blackwell)":[36],"identifies":[37],"main":[39],"cause":[40],"L2":[42,60,76],"miss.":[44],"Leveraging":[45],"this":[46],"insight,":[47],"we":[48],"introduce":[49],"new":[51],"programming":[52],"called":[54],"Sawtooth":[55],"Wavefront":[56],"Reordering":[57],"that":[58],"reduces":[59],"misses.":[61],"We":[62],"validate":[63],"it":[64],"in":[65,75,83],"both":[66],"CUDA":[67],"CuTile,":[69],"observing":[70],"50%":[71],"or":[72],"greater":[73],"reduction":[74],"misses":[77],"up":[79],"60%":[81],"increase":[82],"throughput":[84],"GB10.":[86]},"counts_by_year":[],"updated_date":"2026-08-06T08:24:18.245995","created_date":"2026-08-01T00:00:00"}