Messages in this thread | | | Date | Thu, 5 Dec 2013 17:44:41 +0900 | From | Joonsoo Kim <> | Subject | Re: [patch 2/2] fs: buffer: move allocation failure loop into the allocator |
| |
On Wed, Dec 04, 2013 at 04:33:43PM +0000, Christoph Lameter wrote: > On Thu, 5 Dec 2013, Joonsoo Kim wrote: > > > Now we have cpu partial slabs facility, so I think that slowpath isn't really > > slow. And it doesn't much increase the management overhead in the node > > partial lists, because of cpu partial slabs. > > Well yes that may address some of the issues here. > > > And larger frame may cause more slab_lock contention or cmpxchg contention > > if there are parallel freeings. > > > > But, I don't know which one is better. Is larger frame still better? :) > > Could you run some tests to figure this one out? There are also > some situations in which we disable the per cpu partial pages though. > F.e. for low latency/realtime. I posted in kernel synthetic > benchmarks for slab a while back. That maybe something to start with.
I could try. But my trial would not figure this out, since my machine has just 4 cores which normally cannot produce heavy contention. Anyway, could you tell me where I can find your synthetic benchmarks for slab?
Thanks.
| |