trying optimize 2mm

polybench-c-4.2.1-beta/linear-algebra/kernels/2mm/2mm_optimized.c

@@ -86,16 +86,20 @@ void kernel_2mm(int ni, int nj, int nk, int nl,
 
 #pragma scop
   /* D := alpha*A*B*C + beta*D */
-  for (i = 0; i < _PB_NI; i++)
-    for (j = 0; j < _PB_NL; j++)
+  for (i = 0; i < _PB_NI; i++){
+    for (j = 0; j < _PB_NJ; j++)
       {
 	tmp[i][j] = SCALAR_VAL(0.0);
 	for (k = 0; k < _PB_NK; ++k)
 	  tmp[i][j] += alpha * A[i][k] * B[k][j];
+      }
+    for (j = 0; j < _PB_NL; j++)
+      {
 	D[i][j] *= beta;
 	for (k = 0; k < _PB_NJ; ++k)
 	  D[i][j] += tmp[i][k] * C[k][j];
       }
+  } 
 #pragma endscop
 
 }