The processors along the last column(=2) send their data to the diagonal processors. The diagonal processors perform a column-wise broadcast. Each processor performs the local matrix multiplication.A sum-reduction is performed along the rows to add up the partial dot-products. The output vector C is stored in the last column of three processes.
#include
#include
#include
#include
#include
#include
main(int argc,char *argv[])
{
int i,j,k,row_offset,nlocal;
int *local_a,*local_b,*px,*product_matrix;
int p,rank,n,sender;
int ROW=0,COL=1;
int npes,dims[2],periods[2],keep_dims[2];
int other_rank,coords[2];
int last_column_coords;
int my2drank,mycoords[2];
int dim_size;
MPI_Comm comm_2d,comm_row,comm_col;
MPI_Status status;
MPI_Init(&argc,&argv);
MPI_Comm_size(MPI_COMM_WORLD,&p);
MPI_Comm_rank(MPI_COMM_WORLD,&rank);
n = 15;
nlocal = n/sqrt(p);
dim_size = sqrt(p);
local_a = (int *)malloc((nlocal*nlocal)*sizeof(int));
local_b = (int *)malloc((nlocal)*sizeof(int));
px = (int *)malloc((nlocal)*sizeof(int));
product_matrix = (int *)malloc((nlocal)*sizeof(int));
/*Compute the size of square grid*/
dims[ROW] = dims[COL] = sqrt(p);
/*Set up the catesian topology and get the rank and coordinates of the processin this topology*/
periods[ROW]=periods[COL]=1; /*set the periods for wrap-around connections*/
MPI_Cart_create(MPI_COMM_WORLD,2,dims,periods,1,&comm_2d);
MPI_Comm_rank(comm_2d,&my2drank); /*get the rank in new topology*/
MPI_Cart_coords(comm_2d,my2drank,2,mycoords); /*get the coordinates*/
/*Create row-based topology*/
keep_dims[ROW] = 0;
keep_dims[COL] = 1;
MPI_Cart_sub(comm_2d,keep_dims,&comm_row);
/*create column based topology*/
keep_dims[ROW] = 1;
keep_dims[COL] = 0;
MPI_Cart_sub(comm_2d,keep_dims,&comm_col);
/*************Generate partial matrix ai at each process*******************/
printf("Generating Local Matrix A at process %d\n",rank);
for(i=0;i
{
for(j=0;j
{
local_a[i*nlocal+j]=(mycoords[COL]*nlocal)+j+1;
printf("%d\t",local_a[i*nlocal+j]);
}
printf("\n");
}
/**********************************************************************/
/****Generate partial vector b at the processes in the last column****/
last_column_coords = (dim_size - 1);
if(mycoords[COL] == last_column_coords)
{
printf("Generating Vector B at process %d\n",rank);
for(i=0;i
{
local_b[i]=1;
printf("%d\n",local_b[i]);
}
}
/*Redistribute the vector b. */
/*the processors along the last column send their data to diagonal processors*/
if(mycoords[COL] == last_column_coords && mycoords[ROW] != last_column_coords)
{
coords[ROW] = mycoords[ROW];
coords[COL] = mycoords[ROW];
MPI_Cart_rank(comm_2d,coords,&other_rank);
printf("Distributing the partial vector B to Diagonal Process %d from Process %d\n",other_rank,rank);
MPI_Send(local_b,nlocal,MPI_INT,other_rank,1,comm_2d);
}
if(mycoords[ROW] == mycoords[COL] && mycoords[ROW] != last_column_coords)
{
coords[ROW]=mycoords[ROW];
coords[COL]= last_column_coords;
MPI_Cart_rank(comm_2d,coords,&other_rank);
printf("Receiving the partial vector B at Diagonal Process at %d from Process %d\n",rank,other_rank);
MPI_Recv(local_b,nlocal,MPI_INT,other_rank,1,comm_2d,&status);
}
/*the diagonal processors perform a column wise broadcast*/
coords[0] = mycoords[COL];
MPI_Cart_rank(comm_col,coords,&other_rank);
MPI_Bcast(local_b,nlocal,MPI_INT,other_rank,comm_col);
/*Main Computation Loop*/
for(i=0;i
{
px[i]=0;
for(j=0;j
px[i] += local_a[i*nlocal+j]*local_b[j];
}
/*perform the sum reduction along the rows to add up the partial dot-products and store the result in the last column of processes*/
coords[0] = last_column_coords;
MPI_Cart_rank(comm_row,coords,&other_rank);
MPI_Reduce(px,product_matrix,nlocal,MPI_INT,MPI_SUM,other_rank,comm_row);
if(mycoords[COL] == last_column_coords)
{
printf("Product Vector at process %d\n",rank);
for(i=0;i
{
printf("%d\n",product_matrix[i]);
}
}
MPI_Comm_free(&comm_2d);
MPI_Comm_free(&comm_row);
MPI_Comm_free(&comm_col);
free(px);
free(local_a);
free(local_b);
free(product_matrix);
MPI_Finalize();
}//main
