process - Matrices multiply, Cannon algorithm implementation using MPI -

first of all, of course saw similar questions , solutions, implementation little bit different. main problem that, code works 1 process, doesn't work more processes. don't know cause of this... in communication between processes can't figure out ;/

#include <mpi.h> #include <stdio.h> #include <math.h> #include <iostream> using namespace std; int main(int argc, char **argv) {     int x = 0;     double kk;     int proces;     int numprocs;     int right_neigh, left_neigh, up_neigh, down_neigh;     int tag = 99;      static const int n = 6; //size of matrices      int psa[n][n]; //nxn     int psb[n][n];     int pra[n][n];     int prb[n][n];     int c[n][n];      (int = 0; < n; i++) { //let's make fist matrix         (int j = 0; j < n; j++) {             psa[i][j] = (int)rand() % 100 + 1;             psb[i][j] = (int)rand() % 100 + 1;             c[i][j] = 0;         }     }      (int = 0; < n; i++) { //an 2nd 1         (int j = 0; j < n; j++) {             pra[i][j] = psa[i][j];             prb[i][j] = psb[i][j];          }     }     mpi_status statrecv[2];     mpi_request reqsend[2], reqrecv[2];      mpi_init(&argc, &argv);     mpi_comm_rank(mpi_comm_world, &proces);     mpi_comm_size(mpi_comm_world, &numprocs);       int pp = numprocs;     double np = numprocs;     kk = sqrt(np);     int k = (int)kk;     if (proces < k) // below neighbour set     {         left_neigh = (proces + k - 1) % k;         right_neigh = (proces + k + 1) % k;         up_neigh = ((k - 1)*k) + proces;     }     if (proces == k)     {         left_neigh = ((proces + k - 1) % k) + k;         right_neigh = ((proces + k + 1) % k) + k;         up_neigh = proces - k;     }     if (proces > k)     {         x = proces / k;         left_neigh = ((proces + k - 1) % k) + x * k;         right_neigh = ((proces + k + 1) % k) + x * k;         up_neigh = proces - k;     }     if (proces == 0 || (proces / k) < (k - 1))     {         down_neigh = proces + k;     }     if ((proces / k) == (k - 1))     {         down_neigh = proces - ((k - 1)*k);     }     x = 0;      for(int kk = 0; kk < pp; kk++) //algorithm     {         (int = 0; < n / pp; i++)         {             (int j = 0; j < n / pp; j++)             {                 (int k = 0; k < n / pp; k++)                 {                     c[i][j] += psa[i][k] * psb[k][j];                 }             }         }         mpi_irecv(pra, n*n / pp / pp,mpi_float,left_neigh, tag,mpi_comm_world, reqrecv);         mpi_irecv(prb, n*n / pp / pp,mpi_float,down_neigh,tag,mpi_comm_world,&reqrecv[1]);         mpi_isend(psa, n*n / pp / pp,mpi_float,right_neigh,tag,mpi_comm_world, reqsend);         mpi_isend(psb, n*n / pp / pp,mpi_float,up_neigh,tag,mpi_comm_world,&reqsend[1]);         mpi_wait(reqrecv, statrecv);      }      cout << "a" << endl; //show result     (int = 0; < n; i++) {         (int j = 0; j < n; j++) {             cout << pra[i][j] << " ";         }         cout << endl;     }     cout << "b" << endl;     (int = 0; < n; i++) {         (int j = 0; j < n; j++) {             cout << prb[i][j] << " ";         }         cout << endl;     }     cout << "c" << endl;     (int = 0; < n; i++) {         (int j = 0; j < n; j++) {             cout << c[i][j] << " ";         }         cout << endl;     }       mpi_finalize();      return 0; } 

ok made it. cool, friend helped me out. admin please not remove it, can helpful someone.

#include <mpi.h> #include <stdio.h> #include <math.h> #include <iostream> using namespace std; int main(int argc, char **argv) {  int x = 0;  double kk;  int proces;     int numprocs;   int prawy_sasiad, lewy_sasiad, gorny_sasiad, dolny_sasiad;  int tag = 99;  static const int n = 4; //rozmiar tablic const int pp = 2; // pierwiastek z liczby procesow int a[n][n] = {}, b[n][n] = {}; (int = 0; < n; i++) {//inicjalizacja macierzy glownych (int j = 0; j < n; j++) { a[i][j] = (int)rand() % 100 + 1; b[i][j] = (int)rand() % 100 + 1; } }  /* int val = 1; (int = 0; < n; i++) { //inicjalizacja macierzy glownych     (int j = 0; j < n; j++) {         a[i][j] = val;         b[i][j] = val;         val++;     } } */ mpi_status  statrecv2; mpi_request reqsend2, reqrecv2; mpi_status  statrecv[2]; mpi_request reqsend[2], reqrecv[2]; mpi_init(0, 0); mpi_comm_rank(mpi_comm_world, &proces); mpi_comm_size(mpi_comm_world, &numprocs); int pra[n / pp][n / pp] = {}, psa[n / pp][n / pp] = {};// podmacierze int prb[n / pp][n / pp] = {}, psb[n / pp][n / pp] = {}; //int c[n / pp][n / pp] = {};//wynikowa int c[n][n] = {};//wynikowa //cout << proces << endl; (int = 0; < n / pp; i++)//podzielenie macierzy glownej na podmacierze, kazdy proces otrzymuje inna podmacierz {     (int j = 0; j < n / pp; j++)     {         psa[i][j] = a[proces / pp*(n / pp) + i][proces%pp*(n / pp) + j];         psb[i][j] = b[proces / pp*(n / pp) + i][proces%pp*(n / pp) + j];         //cout << a[proces / pp*(n / pp) + i][proces%pp*(n / pp) + j] << " ";     }     //cout << endl; }  double np = numprocs; kk = sqrt(np); int k = (int)kk;   if (proces < k) // ustawienie sasiadow {     lewy_sasiad = (proces + k - 1) % k;     prawy_sasiad = (proces + k + 1) % k;     gorny_sasiad = ((k - 1)*k) + proces; } if (proces == k) {     lewy_sasiad = ((proces + k - 1) % k) + k;     prawy_sasiad = ((proces + k + 1) % k) + k;     gorny_sasiad = proces - k; } if (proces > k) {     x = proces / k;     lewy_sasiad = ((proces + k - 1) % k) + x * k;     prawy_sasiad = ((proces + k + 1) % k) + x * k;     gorny_sasiad = proces - k; } if (proces == 0 || (proces / k) < (k - 1)) {     dolny_sasiad = proces + k; } if ((proces / k) == (k - 1)) {     dolny_sasiad = proces - ((k - 1)*k); } x = 0; int p = 0; do{ //przesuniecia     if (p < proces / pp)// w wierszu     {          mpi_irecv(pra, n*n / pp / pp, mpi_float, prawy_sasiad, tag, mpi_comm_world, &reqrecv2);         mpi_isend(psa, n*n / pp / pp, mpi_float, lewy_sasiad, tag, mpi_comm_world, &reqsend2);         mpi_wait(&reqrecv2, &statrecv2);         (int = 0; < n / pp; i++)         {             (int j = 0; j < n / pp; j++)             {                 psa[i][j] = pra[i][j];             }         } }     mpi_barrier(mpi_comm_world); if (p < proces % pp)// w kolumnie {      mpi_irecv(prb, n*n / pp / pp, mpi_float, dolny_sasiad, tag, mpi_comm_world, &reqrecv2);     mpi_isend(psb, n*n / pp / pp, mpi_float, gorny_sasiad, tag, mpi_comm_world, &reqsend2);     mpi_wait(&reqrecv2, &statrecv2);     (int = 0; < n / pp; i++)     {         (int j = 0; j < n / pp; j++)         {             psb[i][j] = prb[i][j];         }     }  } mpi_barrier(mpi_comm_world); p++; } while (p < n); //mpi_barrier(mpi_comm_world);   (int kkk = 0; kkk < pp; kkk++) //algorytm {     (int = 0; < n / pp; i++)     {         (int j = 0; j < n / pp; j++)         {             (int k = 0; k < n / pp; k++)             {                 c[i][j] += psa[i][k] * psb[k][j];             }         }     }       mpi_irecv(pra, n*n / pp / pp, mpi_float, prawy_sasiad, tag, mpi_comm_world, reqrecv);     mpi_irecv(prb, n*n / pp / pp, mpi_float, dolny_sasiad, tag, mpi_comm_world, &reqrecv[1]);     mpi_isend(psa, n*n / pp / pp, mpi_float, lewy_sasiad, tag, mpi_comm_world, reqsend);     mpi_isend(psb, n*n / pp / pp, mpi_float, gorny_sasiad, tag, mpi_comm_world, &reqsend[1]);     mpi_wait(reqrecv, statrecv);     mpi_barrier(mpi_comm_world);      (int = 0; < n / pp; i++)     {         (int j = 0; j < n / pp; j++)         {             psa[i][j] = pra[i][j];         }     }       (int = 0; < n / pp; i++)     {         (int j = 0; j < n / pp; j++)         {             psb[i][j] = prb[i][j];         }     }   }   cout << "proces: " << proces << " "; (int = 0; < n / pp; i++) {     (int j = 0; j < n / pp; j++)     {         cout << c[i][j] << " ";     } }  mpi_finalize();  return 0; }