What's wrong with this kernel?

struct RGB16
{
  ushort R, G, B;
};

__kernel void scale2D_3u16(__global ushort* restrict in, __global ushort* restrict out)
{
  int xPos = get_global_id(0);
  int yPos = get_global_id(1);
  int p = yPos*get_global_size(0)+xPos;

  __global struct RGB16* pRGBIn = (__global struct RGB16*)in;
  __global struct RGB16* pRGBOut = (__global struct RGB16*)out;
  pRGBOut[p] = pRGBIn[p];
}

Input data (memory view) is:

0, 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4 ...

Using NVidia and AMD I get the same result as output. But using Intel OCL on HD4600 I get:

0, 0, 0, 0, 0, 1, 2, 2, 2, 0, 0, 3, 4, 4, 4, ...

I'm totally confused!