Free Access

Fig. 18

image

Computing time per time step (open circles) as a function of the number of CPU cores. Two MPI tasks, each spawning ten OpenMP threads that are assigned to the ten cores of a CPU socket were used on each of the two-socket nodes, i.e., the total number of MPI tasks is ten times smaller than the number of CPU cores given on the abscissa. Different colors correspond to different sizes of the numerical grid. Dashed inclined lines indicate ideal strong scalability for a given grid size. Two sets of measurements in which both the number of grid points and the number of processor cores was increased by a factor of 23 from left to right are marked by filled circles. The horizontal dotted lines are a reference for ideal weak scalability. The diamond symbols correspond to additional runs that employed a plain MPI parallelization (OpenMP switched off), i.e., the number of MPI tasks equals the number of cores.

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.