Re: [PATCH 3/7] barrier: a scalable synchonisation barrier

[Paul E. McKenney]

On Sat, Feb 03, 2007 at 10:24:04PM -0500, Alan Stern wrote:
> On Sat, 3 Feb 2007, Paul E. McKenney wrote:
> 
> > > And another note: this all assumes that STORE-MB-LOAD works "correctly", yes?
> > > We have other code which relies on that, should not be a problem.
> > 
> > We have been working with Doug Lea of SUNY Oswego, Sebatian Burckhardt of
> > University of Pennsylvania, and Vijay Saraswat of IBM Research towards
> > a "universal memory model" that accommodates all machines.  Currently,
> > it does in fact handle store-mb-load the way we all want, thankfully!
> 
> We should add that many places in the kernel do depend on proper behavior 
> for this data access pattern.  So whatever "universal memory model" we end 
> up with, it had better handle the pattern correctly if Linux is to support 
> it.

Agreed!

> It's interesting to note, however, that this does exclude simple MESI.

Yep!  And also a number of compiler optimizations, as it turns out.  ;-)

There is a tension between nice-to-software memory-barrier properties
on the one hand and easily understood code on the other.  But I guess
that this is true of pretty much any software tool.

							Thanx, Paul
-
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  http://www.tux.org/lkml/