[cpp-threads] modes, pass 2

[Peter Dimov]

Doug Lea wrote:

[...]

>   fallible_compare_and_set_ordered(cmp, val) // LL/SC-friendly

[...]

> The "ordered" suffix needs a better name, but the intent
> is that it is the lightest barrier (often no-op) that
> prevents re-orderings of successive loads (for get_ordered)
> or stores (for set_ordered) or operations on the same
> variable (for CAS_ordered).

I don't understand fallible_*, and I don't understand *_ordered. Sorry. :-)

> For example, on x86 and sparc, only store_release needs
> an explict barrier (or locked xchg).

I think that all stores are _release on x86 and sparc TSO; no barrier is 
needed.

(As an aside, there are rumors that most/all existing Itaniums operate in a 
TSO-like mode, with all stores having release semantics. If true, this 
explains why the .acq versions of CAS and fetchadd are more expensive than 
the .rel versions.)