Re: [PATCH] GFP_THISNODE for the slab allocator

[Andrew Morton <akpm@osdl.org>]

On Sat, 16 Sep 2006 21:37:04 -0700 (PDT)
Christoph Lameter <clameter@sgi.com> wrote:

> On Sat, 16 Sep 2006, Andrew Morton wrote:
> 
> > I don't see how any of this could help.  If one has a memory container
> > which is constructed from 50 zones, that linear search is just going to do
> > a lot of linear searching when the container approaches anything like
> > fullness.
> 
> One would not construct a memory container from 50 zones but build a 
> single zone as a memory container of that size.
> 
> This could work by creating a new fake node and allocating a certain 
> amount of  memory from the old zone for the fake node. Then one would have 
> a zone that is the container and not a container that consists of 
> gazillions of fake nodes.

Well yes, there are various things one could do if one wanted to make lots
of kernel changes.  I believe Magnus posted some patches along these lines
a while back.

But it's not clear that we _need_ to make such changes. 
nodes-as-containers works OK out-of-the-box.  Apart from the fact that
get_page_from_freelist() sucks.  And speeding that up will speed up other
workloads.

Would prefer to make the kernel faster, rather than more complex...

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>