2014 | OriginalPaper | Buchkapitel
Fusion: Abstractions for Multicore/Manycore Heterogenous Parallel Programming Using GPUs
verfasst von : Anderson Boettge Pinheiro, Francisco Heron de Carvalho Junior, Neemias Gabriel Pena Batista Arruda, Tiago Carneiro
Erschienen in: Programming Languages
Verlag: Springer International Publishing
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Graphic processing units (GPU) have been consolidated as general-purpose computational devices that combine a challenging programming model with an impressive acceleration of HPC programs whose total execution time are dominated by performance-critical small sections of code. However, there is still a lack of high-level programming language abstractions for better specifying heterogenous parallel computations using these devices. This paper proposes Fusion, an extension of Java that introduces new abstractions for heterogenous multicore-GPU programming, taking advantage of new features introduced by the NVIDIA’s Kepler architecture, such as Hyper-Q and Dynamic Parallelism.