// compute correctly:
reduce(hh, bufferA);
reduce(bufferA, hh);
reduce(hh, bufferA);


// and error in this case.
      for(int i = 0; i < 3; i++)
      {
        if(i == 0)
        {
          reduce(hh, bufferA);
        }
        else if(i == 1)
        {
          reduce(bufferA, hh);
        }
        else
        {
          reduce(hh, bufferA);
        }
      }