Messages in this thread | | | Date | Tue, 27 Mar 2012 13:19:45 +0200 | From | Frederic Weisbecker <> | Subject | Re: [PATCH 07/32] cpuset: Set up interface for nohz flag |
| |
On Wed, Mar 21, 2012 at 09:50:27AM -0500, Christoph Lameter wrote: > On Wed, 21 Mar 2012, Frederic Weisbecker wrote: > > > Prepare the interface to implement the nohz cpuset flag. > > This flag, once set, will tell the system to try to > > shutdown the periodic timer tick when possible. > > > > We use here a per cpu refcounter. As long as a CPU > > is contained into at least one cpuset that has the > > nohz flag set, it is part of the set of CPUs that > > run into adaptive nohz mode. > > What are the drawbacks for nohz? > > If there are none: Can we make nohz default behavior without relying on > cpusets?
I can't tell for now. I haven't yet covered everything the timer is handling. Until that happens I can't do measurements.
In theory this sounds like a win in every case. I just would like to test that in practice. This sets up hooks in kernel entry/exit. More IPIs here and there. May be this adds overhead on workloads involving a lot of syscalls or exceptions. I don't know.
Given this is not yet entirely clear, I think it may be better to keep this interface around until the patchset reaches a version that becomes mergeable. Then at this point we can get serious testing coverage to take the decision to drop the interface and make it unconditional on CONFIG_ADAPTIVE_NO_HZ.
| |