Abstract
In this paper we present a new parallelization scheme for the FMM near-field. The parallelization is based on the Global Arrays Toolkit and uses one-sided communication with overlapping. It employs a purely static load-balancing approach to minimize the number of communication steps and benefits from a maximum utilization of data locality. In contrast to other implementations the communication is initiated by the process owning the data via a put call, not the process receiving the data (via a get call). © Springer-Verlag Berlin Heidelberg 2007.
Cite
CITATION STYLE
Kabadshow, I., & Lang, B. (2007). Latency-optimized parallelization of the FMM near-field computations. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 4487 LNCS, pp. 716–722). Springer Verlag. https://doi.org/10.1007/978-3-540-72584-8_95
Register to see more suggestions
Mendeley helps you to discover research relevant for your work.
