[binarycoffee]

I have always wondered about that.

If optimizing repeated atomic loads is indeed allowed, waiting for a signal by spinning on an atomic load could loop forever. Yet I have the feeling most people consider such code to be valid. Are they wrong?

[gpderetta]

Some optimizations are allowed within the rules of the memory model. For example, for an atomic x:

   x=1
   x=2

Can be changed to just x=2. Still forward progress and eventual visibility must be guaranteed.

[manwe150]

Acquire loads (and stronger) are required not to loop forever, if they could be observable from another thread. It is mentioned in the linked paper.