Messages in this thread | | | Date | Mon, 26 Sep 2011 16:49:06 +0300 | Subject | Re: [PATCH 5/5] slub: Only IPI CPUs that have per cpu obj to flush | From | Gilad Ben-Yossef <> |
| |
On Mon, Sep 26, 2011 at 3:05 PM, Gilad Ben-Yossef <gilad@benyossef.com> wrote:
>> The problem with a per-cpu cpumask is that you need to disable >> preemption over the whole for_each_online_cpu() scan and that's not >> really sane on very large machines as that can easily take a very long >> time indeed. > > hmm... I might be thick, but why disable the preemption with the > per-cpu cpumask at all? ... > > Does that makes sense or have I've gone over board with this concept? :-)
Scratch that. The cpumask must be per cache or the patch doesn't make sense at all. So sadly the only sane place to put it is in struct kmem_cache.
I think we can still update the cpumask field without caring about preemption for the reasons stated above, but the per cache memory overhead is still there I'm afraid.
Gilad
-- Gilad Ben-Yossef Chief Coffee Drinker gilad@benyossef.com Israel Cell: +972-52-8260388 US Cell: +1-973-8260388 http://benyossef.com
"I've seen things you people wouldn't believe. Goto statements used to implement co-routines. I watched C structures being stored in registers. All those moments will be lost in time... like tears in rain... Time to die. "
| |