Re: [PATCH 1/3] slub: set a criteria for slub node partial adding

[Shaohua Li]

On Sat, 2011-12-03 at 04:02 +0800, Christoph Lameter wrote:
> On Fri, 2 Dec 2011, Eric Dumazet wrote:
> 
> > netperf (loopback or ethernet) is a known stress test for slub, and your
> > patch removes code that might hurt netperf, but benefit real workload.
> >
> > Have you tried instead this far less intrusive solution ?
> >
> > if (tail == DEACTIVATE_TO_TAIL ||
> >     page->inuse > page->objects / 4)
> >          list_add_tail(&page->lru, &n->partial);
> > else
> >          list_add(&page->lru, &n->partial);
> 
> One could also move this logic to reside outside of the call to
> add_partial(). This is called mostly from __slab_free() so the logic could
> be put in there.
I'm wondering where the improvement comes from. The new added partial
page almost always has few free objects (the inuse < objects/4 isn't
popular I thought), that's why we add it to list tail.

--
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/