On Tue, 18 Sep 2007 11:47:44 +1000 Nick Piggin <nickpiggin@yahoo.com.au> wrote:

quoted text
> On Wednesday 19 September 2007 03:44, Andrew Morton wrote:
> > On Tue, 18 Sep 2007 11:29:50 +1000 Nick Piggin <nickpiggin@yahoo.com.au> 
> wrote:
> > > It would be interesting to test -mm kernels. They have a patch which
> > > reduces zone lock contention quite a lot.
> >
> > They do?  Which patch?
> 
> Hmm... mm-buffered-write-cleanup.patch.

hmm.

quoted text
> 
> > > I think your patch is a nice idea, and with less zone lock contention in
> > > other areas, it is possible that it might produce a relatively larger
> > > improvement.
> >
> > I'm a bit wobbly about this patch - it adds additional single-cpu overhead
> > to reduce multiple-cpu overhead and latency.
> 
> Yeah, that's true. Although maybe it gets significantly more after the
> patch in -mm.
> 
> Possibly other page batching sites have similar issues on UP... I wonder
> if a type of pagevec that turns into a noop on UP would be interesting...
> probably totally unmeasurable and not worth the cost of code
> maintenance ;)

Yes, I wonder that.  Some of the additional overhead will come from
the additional get_page/put_page which is needed for pagevec ownership.
-
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/