On Tue, Feb 01, 2011 at 12:43:56PM +1000, Linus Torvalds wrote:

On Tue, Feb 1, 2011 at 12:18 PM, Paul E. McKenney
[off-list ref] wrote:

quoted

Documentation/memory-barriers.txt specifies this.

Good, so it really is documented, with both cases explicitly mentioned.

That said, I do think that if your memory ordering is much weaker than
x86, you are going to see bugs that most testers don't see, and it
simply might not be worth it.

IBM's CPUs do split the difference, with s390 having somewhat stronger
ordering than x86, and with powerpc being rather weaker (along with arm,
ia64, and some flavors of mips, but stronger than alpha).  But yes, this
does mean that there are bugs that don't show up on x86 and s390, but
that do on powerpc, arm, ia64, and some mips, to say nothing of alpha.
Similarly, there are bugs that show up on x86 due to unsynchronized
timestamp counters that powerpc and s390 avoid due to the guaranteed
synchronization on those platforms.

Whether the weaker ordering provides worthwhile benefits is an interesting
debate that I do not believe we will be able to resolve here.  ;-)

							Thanx, Paul