Abstract
Intel's Xeon Phi coprocessor presents a manycore architecture that is superficially similar to a standard multicore SMP. Xeon Phi can be programmed using the OpenMP standard for shared-memory parallelism. We investigate the performance and optimisation of two real-world scientific codes, parallelised with OpenMP and accelerated on Xeon Phi, and compare with a conventional CPU architecture. We conclude that Xeon Phi offers the potential of significant speedup compared to conventional CPU architectures, much of which is attainable through the use of OpenMP. © 2013 Springer-Verlag.
Cite
CITATION STYLE
Barker, J., & Bowden, J. (2013). Manycore parallelism through OpenMP: High-performance scientific computing with Xeon Phi. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 8122 LNCS, pp. 45–57). https://doi.org/10.1007/978-3-642-40698-0_4
Register to see more suggestions
Mendeley helps you to discover research relevant for your work.
