tr1::shared_ptr and v3's atomic ops

[Nathan Myers]

On Thu, Mar 24, 2005 at 10:31:59AM -0800, Richard Henderson wrote:
> [*] Well, those that can perform them at all without resorting to
> spinlocks.  So true 80386 is out for many of them, as is Sparc prior
> to v9.  I'm sure there are other small architectures that just can't
> hack it either.

I haven't studied the ia64 atomicity design, but it seems to me we 
would best distinguish atomicity guarantees on systems with multiple 
processors from that on just a single processor.  This translates
to bus-transaction vs. instruction-interrupt atomicity.  For 386 and 
similarly obscure targets, instruction-interrupt atomicity probably 
suffices, and is far better than falling back to mutex (never mind
spinlock!), which probably is, itself, just as fraught on those 
targets.

Nathan Myers
ncm@cantrip.org