On Mar 26, 2005, at 3:56 PM, Manuel Bouyer wrote:
BTW, shouldn't the bus_space functions have such barriers too ? The esiop(4) scheduler could be affected by this, for example.
In the case of the bus_space functions, they have the appropriate clobbers for the compiler's benefit. Not sure how x86 handles insn reordering as it relates to memory operations.
-- thorpej