Re: [PATCH] GFP_THISNODE for the slab allocator

[Andrew Morton <akpm@osdl.org>]

On Sat, 16 Sep 2006 04:30:36 -0700
Paul Jackson <pj@sgi.com> wrote:

> Andrew wrote:
> > Why is it not sufficient to cache the most-recent zone*  in task_struct?
> 
> Because ...
> 
> pj - quoting himself:
> > Just one current_allocation_zone would not be enough.  Each node that
> > the cpuset allowed would require its own current_allocation_zone.  For
> > example, on a big honkin NUMA box with 2 CPUs per Node, tasks running
> > on CPU 32, Node 16, might be able to find free memory right on that
> > Node 16.  But another task in the same cpuset running on CPU 112, Node
> > 56 might have to scan past a dozen Nodes to Node 68 to find memory.
> 
> Extending the above example, the task on CPU 32 and the one on CPU
> 112 could be the same task, running in the same cpuset the whole time,
> after being rescheduled from one CPU to another.  The task would need
> not one cached most-recent zone*, but one for each node it might find
> itself on.
> 
> I'm pretty sure you don't want to put MAX_NUMNODES 'struct zone'
> pointers in each task struct.
> 

Could cache a zone* and the cpu number.  If the cpu number has changed
since last time, do another lookup.

--
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>