MAIN PROGRAM : matrices multiplication:  C=A*B
       here we'd to keep C matrix "row by row" !!!

#include <stdio.h>
#include <iostream.h>
#include <mpi.h>
#define N 400                // matrix dimension
#define comm MPI_COMM_WORLD  // standard MPI communicator

static MPI_Status status;    // current status of MPI operation
static int IP,KP;            // current process number and total number of processes
static int tag=0;
static double A[N][N], B[N][N], C[N][N];  // all the 3 needed matrices

main (int argc, char **argv)
{ int i,j,k,nc1,nc,rep;  double t,e;
  char pname[MPI_MAX_PROCESSOR_NAME];

MPI_Init(&argc, &argv);
MPI_Comm_rank(comm,&IP);          // who I am ?
MPI_Comm_size(comm,&KP);          // how manu of us are there in a team ?
MPI_Get_processor_name(pname,&i); // what are processor names ?

 nc=N/KP; j=N%KP;                 // we consider N > KP !!!
 (IP<j) ? nc++,nc1=IP*nc : nc1=j*(nc+1)+nc*(IP-j); //distribute columns
 cout << "Process " << IP << " of " << KP << " started on " << pname <<
"  with columns from " << nc1 << " to " << nc1+nc-1 << endl;
 if(IP==0)   // process 0 sends the whole matrices A and B to others
   for(k=1;k<KP;k++)
     { MPI_Send(&A,N*N,MPI_DOUBLE,k,tag,comm);
       MPI_Send(&B,N*N,MPI_DOUBLE,k,tag,comm); }
 else       // other processes receives the source matrices
   { MPI_Recv(&A,N*N,MPI_DOUBLE,0,tag,comm,&status);
     MPI_Recv(&B,N*N,MPI_DOUBLE,0,tag,comm,&status); }
// ------------------ start multiplication
 for(i=0;i<N;i++)  for(j=0;j<nc;j++) 
  { t=0; for(k=0; k<N; k++) t+=(A[i][k])*(B[k][j+nc1]); C[j+nc1][i]=t; }
// ------------------ finish multiplication
 if(IP==0)   // process 0 receives the results
  for(k=1;k<KP;k++)
   {  nc=N/KP; j=N%KP; (k<j) ? nc++,nc1=k*nc : nc1=j*(nc+1)+nc*(k-j);
      MPI_Recv(&(C[nc1][0]),N*nc,MPI_DOUBLE,k,tag,comm,&status); }
 else       // others send their results (they know nc1 and nc!)
   MPI_Send(&(C[nc1][0]),N*nc,MPI_DOUBLE,0,tag,comm);
MPI_Finalize();
}