Messages in this thread |  | | Date | Thu, 14 Apr 2011 04:07:47 +0900 | From | Tejun Heo <> | Subject | Re: [patch v2 4/4] percpu_counter: use atomic64 for counter |
| |
On Wed, Apr 13, 2011 at 03:57:19PM +0800, shaohua.li@intel.com wrote: > This can improve some workloads with percpu_counter->lock heavily contented. > For example, vm_committed_as sometimes causes the contention. We should tune > the batch count, but if we can make percpu_counter better, why not? In a 24 > CPUs system, 24 processes run stress mmap()/mmunmap(), the atomic method > gives 50x faster.
Christoph already raised the issue but I'd also love to know a bit more detail on the test than "50x faster".
Thanks.
-- tejun
|  |