src/lib_test_gpu.cu~

   1 void verif_cumuls(int ** img_in, uint64 ** d_img_1, uint64 ** d_img_x, uint64 ** d_img_x2){
   2 //allocation memoire CPU
   3   uint64  img_1[H][L];
   4   uint64  img_x[H][L];
   5   uint64  img_x2[H][L];
   6
   7   /*pour test comparaison*/
   8   uint64  img_1b[H][L];
   9   uint64  img_xb[H][L];
  10   uint64  img_x2b[H][L];
  11   cudaMemcpy( img_1b, *d_img_1, taille*sizeof(uint64), cudaMemcpyDeviceToHost );
  12   cudaMemcpy( img_xb, *d_img_x, taille*sizeof(uint64), cudaMemcpyDeviceToHost);
  13   cudaMemcpy( img_x2b, *d_img_x2, taille*sizeof(uint64), cudaMemcpyDeviceToHost);
  14
  15   for (int i=0; i<H; i++)
  16         {
  17           img_1[i][0] = 1 ;
  18           img_x[i][0] = img_in[i][0] ;
  19           img_x2[i][0]= img_in[i][0]*img_in[i][0] ;
  20           for (int j=1; j<L; j++)
  21                 {
  22                   img_1[i][j]  = img_1[i][j-1]  + 1 ;
  23                   img_x[i][j]  = img_x[i][j-1]  + img_in[i][j] ;
  24                   img_x2[i][j] = img_x2[i][j-1] + img_in[i][j]*img_in[i][j] ;
  25                 }
  26         }
  27
  28   int cpt = 0;
  29   int cpt_err = 0;
  30   for (int i=0; i<H; i++){
  31         for (int j=0; j<L; j++){
  32           if( (img_1[i][j] !=  img_1b[i][j]) ||  (img_x[i][j] !=  img_xb[i][j]) ||  (img_x2[i][j] !=  img_x2b[i][j]) )
  33                 {
  34                   if (!(j%BS)) printf("\n");
  35 #ifdef DISPLAY_ERR_IMG_CUMUL
  36                   printf("(%d,%d)CPU:%lu GPU:%lu\n",i, j, img_x2[i][j], img_x2b[i][j]);
  37                   printf("(%d,%d):CPU=%lu  GPU=%lu\n",i,j,img_x[i][j], img_xb[i][j]);
  38 #endif
  39                   cpt_err++;
  40                 }
  41           cpt++;
  42         }
  43   }
  44   printf("%d erreurs / %d points\n", cpt_err, cpt );
  45 }