Introdução à Computação Heterogênea

Denise StringhiniRogério A. GonçalvesAlfredo Goldman

Several types of coprocessors have been used in order to accelerate the execution of onenode only applications. These coprocessors have been named accelerators and areresponsible for the heterogeneous nature of these computational nodes. This tutorialintroduces heterogeneous computing basic concepts from one the most popularaccelerators options that are the GPUs. First, a contextualization is presented where itsmain architecture components are described and related to the SIMD, SPMD and SIMTmodels, widely explored in heterogeneous computing. After that, the tutorial presentsthe alternatives in programming environments and tools for these accelerators alongwith heterogeneous programming examples that unify CPU and GPU. In this context,programming environments as CUDA, OpenCL, OpenMP and OpenACC are presented.They could be used in order to obtain parallelism in heterogeneous architecturescomposed by CPU and GPU.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: