{"id":{"repo_id":"uiuc","oai_identifier":"oai:www.ideals.illinois.edu:2142/102488"},"canonical_url":"https://search.dev.ndltd.org/etd/uiuc/oai:www.ideals.illinois.edu:2142/102488","repository":{"repo_id":"uiuc","name":"University of Illinois - Urbana-Champaign","base_url":"https://www.ideals.illinois.edu/oai-pmh"},"display":{"title":"Techniques for optimizing dynamic parallelism on graphics processing units","abstract":"Dynamic parallelism is a feature of general purpose graphics processing units (GPUs) whereby threads running on a GPU can spawn other threads without CPU intervention. This feature is useful for programming applications with nested parallelism where threads executing in parallel may each identify additional work that can itself be parallelized. Unfortunately, current GPU microarchitectures do not efficiently support using dynamic parallelism for accelerating applications with nested parallelism due to the high overhead of grid launches, the limited number of grids that can execute simultaneously, and the limited supported depth of the dynamic call stack. The compiler techniques presented herein improve the performance of applications with nested parallelism that use dynamic parallelism by mitigating the aforementioned microarchitectural limitations. Horizontal aggregation fuses grids launched by threads in the same warp, block, or grid into a single aggregated grid, thereby reducing the total number of grids launched and increasing the amount of work per grid to improve occupancy. Vertical aggregation fuses grids down the call stack with their descendant grids, again reducing the total number of grids launched but also reducing the depth of the call stack and removing grid launches from the application's critical path. Evaluation of these compiler techniques shows that they result in substantial performance improvement over regular dynamic parallelism for benchmarks representing common nested parallelism patterns. This observation has held true for multiple architecture generations, showing the continued relevance of these techniques. This work shows that to make dynamic parallelism practical for accelerating applications with nested parallelism, compiler transformations can be used to aggregate dynamically launched grids, thereby amortizing their launch overhead and improving their occupancy, without the need for additional hardware support.","abstract_html":"Dynamic parallelism is a feature of general purpose graphics processing units (GPUs) whereby threads running on a GPU can spawn other threads without CPU intervention. This feature is useful for programming applications with nested parallelism where threads executing in parallel may each identify additional work that can itself be parallelized. Unfortunately, current GPU microarchitectures do not efficiently support using dynamic parallelism for accelerating applications with nested parallelism due to the high overhead of grid launches, the limited number of grids that can execute simultaneously, and the limited supported depth of the dynamic call stack. The compiler techniques presented herein improve the performance of applications with nested parallelism that use dynamic parallelism by mitigating the aforementioned microarchitectural limitations. Horizontal aggregation fuses grids launched by threads in the same warp, block, or grid into a single aggregated grid, thereby reducing the total number of grids launched and increasing the amount of work per grid to improve occupancy. Vertical aggregation fuses grids down the call stack with their descendant grids, again reducing the total number of grids launched but also reducing the depth of the call stack and removing grid launches from the application's critical path. Evaluation of these compiler techniques shows that they result in substantial performance improvement over regular dynamic parallelism for benchmarks representing common nested parallelism patterns. This observation has held true for multiple architecture generations, showing the continued relevance of these techniques. This work shows that to make dynamic parallelism practical for accelerating applications with nested parallelism, compiler transformations can be used to aggregate dynamically launched grids, thereby amortizing their launch overhead and improving their occupancy, without the need for additional hardware support.","abstract_has_math":false,"creators":["El Hajj, Izzat"],"institution":"University of Illinois at Urbana-Champaign","degree_name":"Ph.D.","degree_level":"Dissertation","degree_discipline":"Electrical & Computer Engr","degree_department":null,"school":null,"contributors":["Hwu, Wen-Mei W.","Chen, Deming","Lumetta, Steven S.","Milojicic, Dejan S"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2019,"date_issued":"2019-02-06T19:36:37Z","date_published":"2019-02-06T19:36:37Z","updated_at":"2026-07-22T22:24:42Z","subjects":["Graphics Processing Units","Dynamic Parallelism","Compilers","CUDA"],"languages":["en"],"rights":["Copyright 2018 Izzat El Hajj"],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://hdl.handle.net/2142/102488","outbound_label":"Handle","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Hwu, Wen-Mei W.","Chen, Deming","Lumetta, Steven S.","Milojicic, Dejan S"]},{"key":"dc:creator","label":"Author","values":["El Hajj, Izzat"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2019-02-06T19:36:37Z","2018-12-06","2018-12"]},{"key":"dc:type","label":"Dc Type","values":["text"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Electrical & Computer Engr"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Dissertation"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Ph.D."]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["University of Illinois at Urbana-Champaign"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Graphics Processing Units","Dynamic Parallelism","Compilers","CUDA"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["en"]},{"key":"dc:rights","label":"Dc Rights","values":["Copyright 2018 Izzat El Hajj"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["http://hdl.handle.net/2142/102488"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Dynamic parallelism is a feature of general purpose graphics processing units (GPUs) whereby threads running on a GPU can spawn other threads without CPU intervention. This feature is useful for programming applications with nested parallelism where threads executing in parallel may each identify additional work that can itself be parallelized. Unfortunately, current GPU microarchitectures do not efficiently support using dynamic parallelism for accelerating applications with nested parallelism due to the high overhead of grid launches, the limited number of grids that can execute simultaneously, and the limited supported depth of the dynamic call stack. The compiler techniques presented herein improve the performance of applications with nested parallelism that use dynamic parallelism by mitigating the aforementioned microarchitectural limitations. Horizontal aggregation fuses grids launched by threads in the same warp, block, or grid into a single aggregated grid, thereby reducing the total number of grids launched and increasing the amount of work per grid to improve occupancy. Vertical aggregation fuses grids down the call stack with their descendant grids, again reducing the total number of grids launched but also reducing the depth of the call stack and removing grid launches from the application's critical path. Evaluation of these compiler techniques shows that they result in substantial performance improvement over regular dynamic parallelism for benchmarks representing common nested parallelism patterns. This observation has held true for multiple architecture generations, showing the continued relevance of these techniques. This work shows that to make dynamic parallelism practical for accelerating applications with nested parallelism, compiler transformations can be used to aggregate dynamically launched grids, thereby amortizing their launch overhead and improving their occupancy, without the need for additional hardware support.","Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2019-02-05 without embargo terms","The student, Izzat El Hajj, accepted the attached license on 2018-12-06 at 09:42.","The student, Izzat El Hajj, submitted this Dissertation for approval on 2018-12-06 at 09:51.","This Dissertation was approved for publication on 2018-12-06 at 10:52.","DSpace SAF Submission Ingestion Package generated from Vireo submission #13234 on 2019-02-05 at 11:15:19","Made available in DSpace on 2019-02-06T19:36:37Z (GMT). No. of bitstreams: 3 ELHAJJ-DISSERTATION-2018.pdf: 2621361 bytes, checksum: a9d3334a4c7cf84d683791716905befa (MD5) LICENSE.txt: 4210 bytes, checksum: 253874311195eb5653bc8682bcdcb89a (MD5) PROQUEST_LICENSE.txt: 4556 bytes, checksum: d9887d645b76e9f515b40eb7446c9a18 (MD5) Previous issue date: 2018-12-06"]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Techniques for optimizing dynamic parallelism on graphics processing units"]}]}],"canonical_facts":{"dc:contributor":["Hwu, Wen-Mei W.","Chen, Deming","Lumetta, Steven S.","Milojicic, Dejan S"],"dc:creator":["El Hajj, Izzat"],"dc:date":["2019-02-06T19:36:37Z","2018-12-06","2018-12"],"dc:description":["Dynamic parallelism is a feature of general purpose graphics processing units (GPUs) whereby threads running on a GPU can spawn other threads without CPU intervention. This feature is useful for programming applications with nested parallelism where threads executing in parallel may each identify additional work that can itself be parallelized. Unfortunately, current GPU microarchitectures do not efficiently support using dynamic parallelism for accelerating applications with nested parallelism due to the high overhead of grid launches, the limited number of grids that can execute simultaneously, and the limited supported depth of the dynamic call stack. The compiler techniques presented herein improve the performance of applications with nested parallelism that use dynamic parallelism by mitigating the aforementioned microarchitectural limitations. Horizontal aggregation fuses grids launched by threads in the same warp, block, or grid into a single aggregated grid, thereby reducing the total number of grids launched and increasing the amount of work per grid to improve occupancy. Vertical aggregation fuses grids down the call stack with their descendant grids, again reducing the total number of grids launched but also reducing the depth of the call stack and removing grid launches from the application's critical path. Evaluation of these compiler techniques shows that they result in substantial performance improvement over regular dynamic parallelism for benchmarks representing common nested parallelism patterns. This observation has held true for multiple architecture generations, showing the continued relevance of these techniques. This work shows that to make dynamic parallelism practical for accelerating applications with nested parallelism, compiler transformations can be used to aggregate dynamically launched grids, thereby amortizing their launch overhead and improving their occupancy, without the need for additional hardware support.","Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2019-02-05 without embargo terms","The student, Izzat El Hajj, accepted the attached license on 2018-12-06 at 09:42.","The student, Izzat El Hajj, submitted this Dissertation for approval on 2018-12-06 at 09:51.","This Dissertation was approved for publication on 2018-12-06 at 10:52.","DSpace SAF Submission Ingestion Package generated from Vireo submission #13234 on 2019-02-05 at 11:15:19","Made available in DSpace on 2019-02-06T19:36:37Z (GMT). No. of bitstreams: 3 ELHAJJ-DISSERTATION-2018.pdf: 2621361 bytes, checksum: a9d3334a4c7cf84d683791716905befa (MD5) LICENSE.txt: 4210 bytes, checksum: 253874311195eb5653bc8682bcdcb89a (MD5) PROQUEST_LICENSE.txt: 4556 bytes, checksum: d9887d645b76e9f515b40eb7446c9a18 (MD5) Previous issue date: 2018-12-06"],"dc:format":["application/pdf"],"dc:identifier":["http://hdl.handle.net/2142/102488"],"dc:language":["en"],"dc:rights":["Copyright 2018 Izzat El Hajj"],"dc:subject":["Graphics Processing Units","Dynamic Parallelism","Compilers","CUDA"],"dc:title":["Techniques for optimizing dynamic parallelism on graphics processing units"],"dc:type":["text"],"thesis:degree_discipline":["Electrical & Computer Engr"],"thesis:degree_level":["Dissertation"],"thesis:degree_name":["Ph.D."],"thesis:institution_name":["University of Illinois at Urbana-Champaign"]},"updated_at":"2026-07-22T22:24:42Z"}