Write an MPI program to simulate Conway's Game of Life cellular automaton. The program should take as input two values, m (the grid dimension, use an mxm grid) and k (the number of time steps). The...

Write an MPI program to simulate Conway's Game of Life cellular automaton. The program should take as input two values, m (the grid dimension, use an mxm grid) and k (the number of time steps). The initial state of the system could be defined in the code. The m x m grid must be partitioned across multiple processes (i.e., each process is responsible for updating m x m/p grid values). Use MPI_Sendrecv for inter-process communication of boundary cells. Vary the number of MPI processes, and determine running time for different values of m (e.g., m = 5000, m = 10,000) and k (maybe 10, 100). Also, determine the time spent in inter-process communication.
What fraction of the overall running time is spent in inter-process communication? How does the code scale with increasing grid size and increasing MPI process concurrencies?


#include #include #include #include #include static double get_walltime() { struct timeval tp; gettimeofday(&tp, NULL); return ((double) (tp.tv_sec) + 1e-6 * tp.tv_usec); } int main(int argc, char **argv) { if (argc != 3) { printf("%s \n", argv[0]); printf("Program for parallel Game of Life\n"); printf("with 1D grid partitioning\n"); printf(": grid dimension (an mxm grid is created)\n"); printf(": number of time steps\n"); printf("(initial pattern specified inside code)\n"); exit(1); } unsigned long m, k; m = atol(argv[1]); k = atol(argv[2]); int *grid_current; int *grid_next; grid_current = (int *) malloc(m * m * sizeof(int)); if (grid_current == NULL) { printf("Error allocating memory for grid_current!\n"); exit(1); } grid_next = (int *) malloc(m * m * sizeof(int)); if (grid_next == NULL) { printf("Error allocating memory for grid_next!\n"); exit(1); } int i, j, t; /* static initalization, so that we can verify output */ /* using very simple initialization right now */ /* this isn't a good check for parallel debugging */ for (i=0; i