Messages in this thread | | | Date | Mon, 21 Sep 2009 19:07:39 +0100 | From | Mel Gorman <> | Subject | Re: [RFC PATCH 0/3] Fix SLQB on memoryless configurations V2 |
| |
On Mon, Sep 21, 2009 at 01:54:12PM -0400, Christoph Lameter wrote: > Lets just keep SLQB back until the basic issues with memoryless nodes are > resolved.
It's not even super-clear that the memoryless nodes issues are entirely related to SLQB. Sachin for example says that there was a stall issue with memoryless nodes that could be triggered without SLQB. Sachin, is that still accurate?
If so, it's possible that SLQB somehow exasperates the problem in some unknown fashion.
> There does not seem to be an easy way to deal with this. Some > thought needs to go into how memoryless node handling relates to per cpu > lists and locking. List handling issues need to be addressed before SLQB. > can work reliably. The same issues can surface on x86 platforms with weird > NUMA memory setups. >
Can you spot if there is something fundamentally wrong with patch 2? I.e. what is wrong with treating the closest node as local instead of only the closest node?
> Or just allow SLQB for !NUMA configurations and merge it now. >
Forcing SLQB !NUMA will not rattle out any existing list issues unfortunately :(.
-- Mel Gorman Part-time Phd Student Linux Technology Center University of Limerick IBM Dublin Software Lab
| |