Re: [PATCH v2] cpuidle: poll_state: Add time limit to poll_idle()

["Rafael J. Wysocki" <rafael@kernel.org>]

On Tue, Mar 27, 2018 at 8:02 PM, Rik van Riel <riel@surriel.com> wrote:
> On Tue, 2018-03-27 at 18:42 +0200, Rafael J. Wysocki wrote:
>> On Thu, Mar 22, 2018 at 5:32 PM, Rik van Riel <riel@surriel.com>
>> wrote:
>> > On Wed, 2018-03-14 at 13:04 +0100, Peter Zijlstra wrote:
>> >
>> > > On x86 we don't have to use that time_check_counter thing,
>> > > sched_clock()
>> > > is really cheap, not sure if it makes sense on other platforms.
>> >
>> > Are you sure? I saw a 5-10% increase in CPU use,
>> > for a constant query rate to a memcache style
>> > workload, with v3 of this patch.
>>
>> I think I know what's going on.
>
> I ran my tests wrong, and the script never propagated
> errors back to me. Sigh.

Again, no worries. :-)

> However, the poll_idle() that reads the TSC at a
> reduced rate seems to perform better than the one
> that reads the TSC every time it goes around the
> loop.

OK

> The size of the idle loop seems to make a slight
> difference, too. Having just one cpu_relax() in
> the entire loop seems to be better than having
> them all over the place.

OK

Thanks for the above observations, they match my understanding of
what's happening.

I'll resend the patch to read the TSC at a reduced rate with a proper
changelog and I'll tentatively add it to my pm-cpuidle branch.

Thanks!