Tiling seems to be one of the hardest parts to grasp for newcomers to GPU programming. We have explained the motivation (potential additional performance through the use of tile_static memory) and mechanics of tiling (use the tiled overload of parallel_for_each passing a tiled_extent and receiving a tiled_index in the lambda) in previous blog posts and many of our samples also use tiling, so you can learn from real examples.
In my humble opinion, the best way to learn about tiling is to read the second of the C++ AMP msdn magazine articles. To accompany that article, I have just published a screencast that uses one of the simplistic examples in the article to introduce you to the tiling constructs:
If you enjoy learning through screencasts and videos, stay tuned on the C++ AMP channel9 page.