|
|
On Fri, 31 Mar 2006, Andi Kleen wrote:
> On Friday 31 March 2006 18:22, Hans Boehm wrote:
>
> > My impression is that approach (1) tends not to stick, since it involves
> > a substantial performance hit on architectures on which the fence is
> > not implicitly included in atomic operations. Those include Itanium and
> > PowerPC.
>
> At least the PPC people are eating the overhead because back when they
> didn't they had a long string of subtle powerpc only bugs caused by that
PPC has barriers for both smb_mb_before/after cases. IMHO we should do the
same for ia64 and not fuzz around.
> It's a stability/maintainability vs performance issue. I doubt the
> performance advantage would be worth the additional work. I guess
> with the engineering time you would need to spend getting all this right
> you could do much more fruitful optimizations.
Agreed.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at www.tux.org/lkml/">http://www.tux.org/lkml/
|
|