Performance evaluation of data migration methods between the host and the device in CUDA-based programming

Página do item simplificado
dc.contributor.authorSantos, Rafael Silva [UNESP]
dc.contributor.authorEler, Danilo Medeiros [UNESP]
dc.contributor.authorGarcia, Rogério Eduardo [UNESP]
dc.contributor.institutionUniversidade Estadual Paulista (Unesp)
dc.date.accessioned2018-12-11T17:27:47Z
dc.date.available2018-12-11T17:27:47Z
dc.date.issued2016-04-01
dc.description.abstractCUDA-based programming model is heterogeneous – composed of two components: host (CPU) and device (GPU). Both components have separated memory spaces and processing units. A great challenge to increase GPU-based application performance is the data migration between these memory spaces. Currently, the CUDA platform supports the following data migration methods: UMA, zero-copy, pageable and pinned memory. In this paper, we compare the zero-copy performance method with the other methods by considering the overall application runtime. Additionally, we investigated the aspects of data migration process to enunciate causes of the performance variations. The obtained results demonstrated in some cases the zero-copy memory can provide an average performance on 19% higher than the pinned memory transfer. In the studied situation, this method was the second most efficient. Finally, we present limitations of zero-copy memory as a resource for improving performance of CUDA applications.en
dc.description.affiliationUNESP - Univ Estadual Paulista
dc.description.affiliationUnespUNESP - Univ Estadual Paulista
dc.format.extent689-700
dc.identifierhttp://dx.doi.org/10.1007/978-3-319-32467-8_60
dc.identifier.citationAdvances in Intelligent Systems and Computing, v. 448, p. 689-700.
dc.identifier.doi10.1007/978-3-319-32467-8_60
dc.identifier.file2-s2.0-84962709054.pdf
dc.identifier.issn2194-5357
dc.identifier.lattes8031012573259361
dc.identifier.orcid0000-0003-1248-528X
dc.identifier.scopus2-s2.0-84962709054
dc.identifier.urihttp://hdl.handle.net/11449/177938
dc.language.isoeng
dc.relation.ispartofAdvances in Intelligent Systems and Computing
dc.rights.accessRightsAcesso aberto
dc.sourceScopus
dc.titlePerformance evaluation of data migration methods between the host and the device in CUDA-based programmingen
dc.typeCapítulo de livro
unesp.author.lattes8031012573259361[3]
unesp.author.orcid0000-0003-1248-528X[3]
unesp.departmentMatemática e Computação - FCTpt

Arquivos

Pacote Original
Agora exibindo 1 - 1 de 1
Imagem de Miniatura
Nome:
2-s2.0-84962709054.pdf
Tamanho:
472.45 KB
Formato:
Adobe Portable Document Format
Descrição:
Baixar

Coleções

Presidente Prudente - FCT - Faculdade de Ciências e Tecnologia