int timestep(const t_param params, t_speed* cells, t_speed* tmp_cells, int* obstacles)
{
accelerate_flow(params,cells,obstacles);
propagate(params,cells,tmp_cells);
rebound_or_collision(params,cells,tmp_cells,obstacles);
return EXIT_SUCCESS;
}
int main(int argc, char* argv[])
{
param_t params; /* struct to hold parameter values */
speed_t* cells = NULL; /* grid containing fluid densities */
speed_t* tmp_cells = NULL; /* scratch space */
int* obstacles = NULL; /* grid indicating which cells are blocked */
int* rowsSetup = NULL;
int* accelgrid = NULL;
float* av_vels = NULL; /* a record of the av. velocity computed for each timestep */
int ii, rank, size, tag=0, jj; /* generic counter */
struct timeval timstr; /* structure to hold elapsed time */
struct rusage ru; /* structure to hold CPU time--system and user */
double tic,toc; /* floating point numbers to calculate elapsed wallclock time */
double usrtim; /* floating point number to record elapsed user CPU time */
double systim; /* floating point number to record elapsed system CPU time */
int halorow;
int buff;
int extra;
int start_row;
int end_row;
MPI_Status status;
accel_area_t accel_area;
/* initialise our data structures and load values from file */
initialise(argv[1], &accel_area, ¶ms, &cells, &tmp_cells, &obstacles, &av_vels, &accelgrid);
// Initialize MPI environment.
MPI_Init(&argc, &argv);
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Comm_size(MPI_COMM_WORLD, &size);
extra = params.ny%size;
halorow = (rank<extra) ? (params.ny/size + 1) * params.nx : (params.ny/size) * params.nx;
calc_row_setup(&rowsSetup, rank, halorow, extra, params);
buff = rowsSetup[0];
start_row = rowsSetup[1];
end_row = rowsSetup[2];
/* iterate for max_iters timesteps */
gettimeofday(&timstr,NULL);
tic=timstr.tv_sec+(timstr.tv_usec/1000000.0);
for(ii=0; ii<params.max_iters; ii++) {
accelerate_flow(params,accel_area,cells,start_row, end_row, accelgrid);
lattice(params,cells,tmp_cells,obstacles, av_vels, start_row, end_row, ii);
}
gettimeofday(&timstr,NULL);
toc=timstr.tv_sec+(timstr.tv_usec/1000000.0);
getrusage(RUSAGE_SELF, &ru);
timstr=ru.ru_utime;
usrtim=timstr.tv_sec+(timstr.tv_usec/1000000.0);
timstr=ru.ru_stime;
systim=timstr.tv_sec+(timstr.tv_usec/1000000.0);
float* buffer = malloc(buff * 9 * sizeof(float));
if(rank != 0) {
for(ii=0; ii<halorow; ii++) {
buffer[9*ii] = cells[start_row+ii].speeds[0];
buffer[9*ii+1] = cells[start_row+ii].speeds[1];
buffer[9*ii+2] = cells[start_row+ii].speeds[2];
buffer[9*ii+3] = cells[start_row+ii].speeds[3];
buffer[9*ii+4] = cells[start_row+ii].speeds[4];
buffer[9*ii+5] = cells[start_row+ii].speeds[5];
buffer[9*ii+6] = cells[start_row+ii].speeds[6];
buffer[9*ii+7] = cells[start_row+ii].speeds[7];
buffer[9*ii+8] = cells[start_row+ii].speeds[8];
}
MPI_Send(buffer, 9*buff, MPI_FLOAT, 0, tag, MPI_COMM_WORLD);
}
else {
if(extra == 0) {
for(ii=1; ii<size; ii++) {
MPI_Recv(buffer, 9*buff, MPI_FLOAT, ii, tag, MPI_COMM_WORLD, &status);
for(jj=0; jj<halorow; jj++) {
cells[halorow*ii+jj].speeds[0] = buffer[9*jj];
cells[halorow*ii+jj].speeds[1] = buffer[9*jj+1];
cells[halorow*ii+jj].speeds[2] = buffer[9*jj+2];
cells[halorow*ii+jj].speeds[3] = buffer[9*jj+3];
cells[halorow*ii+jj].speeds[4] = buffer[9*jj+4];
cells[halorow*ii+jj].speeds[5] = buffer[9*jj+5];
cells[halorow*ii+jj].speeds[6] = buffer[9*jj+6];
cells[halorow*ii+jj].speeds[7] = buffer[9*jj+7];
cells[halorow*ii+jj].speeds[8] = buffer[9*jj+8];
}
}
} else {
for(ii=1; ii<extra; ii++) {
MPI_Recv(buffer, 9*buff, MPI_FLOAT, ii, tag, MPI_COMM_WORLD, &status);
for(jj=0; jj<halorow; jj++) {
cells[halorow*ii+jj].speeds[0] = buffer[9*jj];
cells[halorow*ii+jj].speeds[1] = buffer[9*jj+1];
cells[halorow*ii+jj].speeds[2] = buffer[9*jj+2];
cells[halorow*ii+jj].speeds[3] = buffer[9*jj+3];
cells[halorow*ii+jj].speeds[4] = buffer[9*jj+4];
cells[halorow*ii+jj].speeds[5] = buffer[9*jj+5];
cells[halorow*ii+jj].speeds[6] = buffer[9*jj+6];
cells[halorow*ii+jj].speeds[7] = buffer[9*jj+7];
cells[halorow*ii+jj].speeds[8] = buffer[9*jj+8];
}
}
for(ii=extra; ii<size; ii++) {
MPI_Recv(buffer, 9*buff, MPI_FLOAT, ii, tag, MPI_COMM_WORLD, &status);
for(jj=0; jj<(params.ny/size) * params.nx; jj++) {
int local_extra = halorow * extra + (halorow-params.nx)* (ii-extra);
cells[local_extra + jj].speeds[0] = buffer[9*jj];
cells[local_extra + jj].speeds[1] = buffer[9*jj+1];
cells[local_extra + jj].speeds[2] = buffer[9*jj+2];
cells[local_extra + jj].speeds[3] = buffer[9*jj+3];
cells[local_extra + jj].speeds[4] = buffer[9*jj+4];
cells[local_extra + jj].speeds[5] = buffer[9*jj+5];
cells[local_extra + jj].speeds[6] = buffer[9*jj+6];
cells[local_extra + jj].speeds[7] = buffer[9*jj+7];
cells[local_extra + jj].speeds[8] = buffer[9*jj+8];
}
}
}
free(buffer);
buffer = NULL;
}
MPI_Finalize(); /*Finilize MPI*/
if(rank == 0) {
printf("==done==\n");
printf("Reynolds number:\t\t%.12E\n",calc_reynolds(params,cells,obstacles,av_vels[params.max_iters-1]));
printf("Elapsed time:\t\t\t%.6lf (s)\n", toc-tic);
printf("Elapsed user CPU time:\t\t%.6lf (s)\n", usrtim);
printf("Elapsed system CPU time:\t%.6lf (s)\n", systim);
write_values(params,cells,obstacles,av_vels);
finalise(¶ms, &cells, &tmp_cells, &obstacles, &av_vels, &accelgrid, &rowsSetup);
}
return EXIT_SUCCESS;
}
