This is just a comment for the "community" at large and is merely my observation.
This solution presents a double-edged sword. On the one hand, it forces the programmer to work closer to the lower (hardware) level. For an embedded programmer like me this is a "good thing". The other edge of the sword, though, is that it removes the level of encapsulation that was present in the previous methodology and thus, by requiring more "knowledge" by the programmer, it reduces the pool of programmers capable of handling the task. And, this is a "bad thing."
There will be some that will say that things like processor expert, etc. will handle this. But, I see two additional effects of this move. The first is a further extension of the idea that only a select few need to understand the inner machinations of the processor. The other is that in my experience the "processor expert" implementation, although great in concept, almost always falls short of delivering what it promises.
Well, back to the appropriate Family Reference Manuals...
Thanks, Xiangjun, for the pointer.