Figure 5
From: gEMpicker: a highly parallel GPU-accelerated particle picking tool for cryo-electron microscopy
The reduction time on (a) Griffon (64-node cluster, 20Gb/s InfiniBand), and (b) Adonis (8-node GPU cluster, 40Gb/s InfiniBand) using different inter-process communication strategies (i.e. the MPI_Reduce or MPI_Send and MPI_Recv functions), and different methods for calculating the reduce operation on each node (i.e. with or without using OpenMP parallelisation). Note that the time axis in (b) is 10 × smaller than in (a).