Compression and heuristic caching for GPU particle tracing in turbulent vector fields. Treib, M., Bürger, K., Wu, J., & Westermann, R. Communications in Computer and Information Science, 598:144-165, Springer, Cham, 2016. Website doi abstract bibtex © Springer International Publishing Switzerland 2016. Particle tracing in fully resolved turbulent vector fields is challenging due to their extreme resolution. Since particles can move along arbitrary paths through large parts of the domain, particle integration requires access to the entire field in an unpredictable order. Thus, techniques for particle tracing in such fields require a careful design to reduce performance constraints caused by memory and communication bandwidth. One possibility to achieve this is data compression, but so far it has been considered rather hesitantly due to supposed accuracy issues. We shed light on the use of data compression for turbulent vector fields, motivated by the observation that particle traces are always afflicted with inaccuracy. We quantitatively analyze the additional inaccuracies caused by lossy compression. We propose an adaptive data compression scheme using the discrete wavelet transform and integrate it into a block-based particle tracing approach. Furthermore, we present a priority-based GPU caching scheme to reduce memory access operations. In some experiments we confirm that the compression has only minor impact on the accuracy of the trajectories, and that on a desktop system our technique can achieve comparable performance to previous approaches on supercomputers.
@article{
title = {Compression and heuristic caching for GPU particle tracing in turbulent vector fields},
type = {article},
year = {2016},
keywords = {Data compression,Data streaming,Particle tracing,Turbulence,Vector fields},
pages = {144-165},
volume = {598},
websites = {http://link.springer.com/10.1007/978-3-319-29971-6_8},
publisher = {Springer, Cham},
id = {c967ab18-170c-3561-9b15-cf0969b9566f},
created = {2021-04-09T15:23:40.216Z},
file_attached = {false},
profile_id = {75799766-8e2d-3c98-81f9-e3efa41233d0},
group_id = {c9329632-2a50-3043-b803-cadc8dbdfc3f},
last_modified = {2021-04-09T15:23:40.216Z},
read = {false},
starred = {false},
authored = {false},
confirmed = {false},
hidden = {false},
source_type = {article},
private_publication = {false},
abstract = {© Springer International Publishing Switzerland 2016. Particle tracing in fully resolved turbulent vector fields is challenging due to their extreme resolution. Since particles can move along arbitrary paths through large parts of the domain, particle integration requires access to the entire field in an unpredictable order. Thus, techniques for particle tracing in such fields require a careful design to reduce performance constraints caused by memory and communication bandwidth. One possibility to achieve this is data compression, but so far it has been considered rather hesitantly due to supposed accuracy issues. We shed light on the use of data compression for turbulent vector fields, motivated by the observation that particle traces are always afflicted with inaccuracy. We quantitatively analyze the additional inaccuracies caused by lossy compression. We propose an adaptive data compression scheme using the discrete wavelet transform and integrate it into a block-based particle tracing approach. Furthermore, we present a priority-based GPU caching scheme to reduce memory access operations. In some experiments we confirm that the compression has only minor impact on the accuracy of the trajectories, and that on a desktop system our technique can achieve comparable performance to previous approaches on supercomputers.},
bibtype = {article},
author = {Treib, Marc and Bürger, Kai and Wu, Jun and Westermann, Rüdiger},
doi = {10.1007/978-3-319-29971-6_8},
journal = {Communications in Computer and Information Science}
}
Downloads: 0
{"_id":"AqcguRYvPWjXcrwoq","bibbaseid":"treib-brger-wu-westermann-compressionandheuristiccachingforgpuparticletracinginturbulentvectorfields-2016","authorIDs":[],"author_short":["Treib, M.","Bürger, K.","Wu, J.","Westermann, R."],"bibdata":{"title":"Compression and heuristic caching for GPU particle tracing in turbulent vector fields","type":"article","year":"2016","keywords":"Data compression,Data streaming,Particle tracing,Turbulence,Vector fields","pages":"144-165","volume":"598","websites":"http://link.springer.com/10.1007/978-3-319-29971-6_8","publisher":"Springer, Cham","id":"c967ab18-170c-3561-9b15-cf0969b9566f","created":"2021-04-09T15:23:40.216Z","file_attached":false,"profile_id":"75799766-8e2d-3c98-81f9-e3efa41233d0","group_id":"c9329632-2a50-3043-b803-cadc8dbdfc3f","last_modified":"2021-04-09T15:23:40.216Z","read":false,"starred":false,"authored":false,"confirmed":false,"hidden":false,"source_type":"article","private_publication":false,"abstract":"© Springer International Publishing Switzerland 2016. Particle tracing in fully resolved turbulent vector fields is challenging due to their extreme resolution. Since particles can move along arbitrary paths through large parts of the domain, particle integration requires access to the entire field in an unpredictable order. Thus, techniques for particle tracing in such fields require a careful design to reduce performance constraints caused by memory and communication bandwidth. One possibility to achieve this is data compression, but so far it has been considered rather hesitantly due to supposed accuracy issues. We shed light on the use of data compression for turbulent vector fields, motivated by the observation that particle traces are always afflicted with inaccuracy. We quantitatively analyze the additional inaccuracies caused by lossy compression. We propose an adaptive data compression scheme using the discrete wavelet transform and integrate it into a block-based particle tracing approach. Furthermore, we present a priority-based GPU caching scheme to reduce memory access operations. In some experiments we confirm that the compression has only minor impact on the accuracy of the trajectories, and that on a desktop system our technique can achieve comparable performance to previous approaches on supercomputers.","bibtype":"article","author":"Treib, Marc and Bürger, Kai and Wu, Jun and Westermann, Rüdiger","doi":"10.1007/978-3-319-29971-6_8","journal":"Communications in Computer and Information Science","bibtex":"@article{\n title = {Compression and heuristic caching for GPU particle tracing in turbulent vector fields},\n type = {article},\n year = {2016},\n keywords = {Data compression,Data streaming,Particle tracing,Turbulence,Vector fields},\n pages = {144-165},\n volume = {598},\n websites = {http://link.springer.com/10.1007/978-3-319-29971-6_8},\n publisher = {Springer, Cham},\n id = {c967ab18-170c-3561-9b15-cf0969b9566f},\n created = {2021-04-09T15:23:40.216Z},\n file_attached = {false},\n profile_id = {75799766-8e2d-3c98-81f9-e3efa41233d0},\n group_id = {c9329632-2a50-3043-b803-cadc8dbdfc3f},\n last_modified = {2021-04-09T15:23:40.216Z},\n read = {false},\n starred = {false},\n authored = {false},\n confirmed = {false},\n hidden = {false},\n source_type = {article},\n private_publication = {false},\n abstract = {© Springer International Publishing Switzerland 2016. Particle tracing in fully resolved turbulent vector fields is challenging due to their extreme resolution. Since particles can move along arbitrary paths through large parts of the domain, particle integration requires access to the entire field in an unpredictable order. Thus, techniques for particle tracing in such fields require a careful design to reduce performance constraints caused by memory and communication bandwidth. One possibility to achieve this is data compression, but so far it has been considered rather hesitantly due to supposed accuracy issues. We shed light on the use of data compression for turbulent vector fields, motivated by the observation that particle traces are always afflicted with inaccuracy. We quantitatively analyze the additional inaccuracies caused by lossy compression. We propose an adaptive data compression scheme using the discrete wavelet transform and integrate it into a block-based particle tracing approach. Furthermore, we present a priority-based GPU caching scheme to reduce memory access operations. In some experiments we confirm that the compression has only minor impact on the accuracy of the trajectories, and that on a desktop system our technique can achieve comparable performance to previous approaches on supercomputers.},\n bibtype = {article},\n author = {Treib, Marc and Bürger, Kai and Wu, Jun and Westermann, Rüdiger},\n doi = {10.1007/978-3-319-29971-6_8},\n journal = {Communications in Computer and Information Science}\n}","author_short":["Treib, M.","Bürger, K.","Wu, J.","Westermann, R."],"urls":{"Website":"http://link.springer.com/10.1007/978-3-319-29971-6_8"},"biburl":"https://bibbase.org/service/mendeley/75799766-8e2d-3c98-81f9-e3efa41233d0","bibbaseid":"treib-brger-wu-westermann-compressionandheuristiccachingforgpuparticletracinginturbulentvectorfields-2016","role":"author","keyword":["Data compression","Data streaming","Particle tracing","Turbulence","Vector fields"],"metadata":{"authorlinks":{}},"downloads":0},"bibtype":"article","biburl":"https://bibbase.org/service/mendeley/75799766-8e2d-3c98-81f9-e3efa41233d0","creationDate":"2019-09-10T01:57:27.606Z","downloads":0,"keywords":["data compression","data streaming","particle tracing","turbulence","vector fields"],"search_terms":["compression","heuristic","caching","gpu","particle","tracing","turbulent","vector","fields","treib","bürger","wu","westermann"],"title":"Compression and heuristic caching for GPU particle tracing in turbulent vector fields","year":2016,"dataSources":["hY2XmpNnxf2BGtC9Z","ya2CyA73rpZseyrZ8","pfsTJ6F3uc4Q2gYcz","bjcJTLfSPr782KPn4","wGLMZBQBxePE3WB2P","NnpNKRo5qgenEkteN","2252seNhipfTmjEBQ"]}