Fig. 5

Parallelized 3D diffusion performance of pSpatiocyte. Diffusion model is the same as in Fig. 3 but with 30% of the total voxels occupied by diffusing molecules. a Speedup ratios relative to 64 cores. Red, green, and blue lines represent lattices with 512³ (coarse), 1024³ (intermediate), and 2048³ (fine) voxels, respectively. b Efficiency of strong scaling on increasing number of CPU cores. The resolutions of lattice are the same as in (a). c Components of simulation elapsed times. The average times taken for initialization and computation are given by init and calc, respectively. For inter-process communications, the pack and unpack times are the durations required to manage data before sending and after receiving it, respectively. The time spent on the MPI Sendrecv function is denoted by mpi. The duration to duplicate communicator objects is indicated by dup. d Average elapsed times per voxel per simulation step. Voxel sizes from smallest to largest denote 32³, 64³, 128³, 256³, and 512³ voxels per process, respectively. e Efficiency of weak scaling on increasing number of CPU cores. Colors have the same representation as in (d)