Messages in this thread | ![/](/images/icornerl.gif) | | Date | Thu, 10 Oct 2013 09:45:16 +0200 | From | Ingo Molnar <> | Subject | Re: [PATCH 0/6] Optimize the cpu hotplug locking -v2 |
| |
* Andrew Morton <akpm@linux-foundation.org> wrote:
> On Thu, 10 Oct 2013 09:27:57 +0200 Ingo Molnar <mingo@kernel.org> wrote: > > > > > Should be fairly straightforward to test: the sys_sched_getaffinity() > > > > and sys_sched_setaffinity() syscalls both make use of > > > > get_online_cpus()/put_online_cpus(), so a testcase frobbing affinities > > > > on N CPUs in parallel ought to demonstrate scalability improvements > > > > pretty nicely. > > > > > > Well, an in-kernel microbenchmark which camps in a loop doing get/put > > > would measure this as well. > > > > > > But neither approach answers the question "how useful is this patchset". > > > > Even ignoring all the other reasons cited, sys_sched_getaffinity() / > > sys_sched_setaffinity() are prime time system calls, and as long as > > the patches are correct, speeding them up is worthwhile. > > That I would not have guessed. What's the use case for calling > get/set_affinity at high frequency?
I don't think high-freq usage is common (at all).
It could happen in AIM7-like workloads that start up a ton of binaries in parallel, which can easily create parallel sched_getaffinity() calls during process startup?
Thanks,
Ingo
| ![\](/images/icornerr.gif) |