Barrier / atomic operations on multiprocessors.

[Ray Whitmer <ray at personallegal dot net>]

Is there a generic way to access efficient platform-appropriate memory
barriers (also compare and swap) in any arbitrary C or C++ program
compiled under gcc (not just a compiler ordering directive, but
extending to hardware guarantees)?

If it is documented somewhere, I apologize.

It would be critical to many sorts of new projects attempting lock-free
operations for multi-processor machines.

Every significant project I know winds up redefining their own, and it
would seem to me that the GCC compiler is a very appropriate place to
have these generically defined.

I am just trying to make a lock-free hashtable.  I do not see a C++
standard emerging for years.


Thanks,

Ray Whitmer