Re: PROBLEM: cpuidle: Quickly notice prediction failure for repeat mode

[Daniel Lezcano <daniel.lezcano@linaro.org>]

On 12/19/2013 08:51 PM, Mark Oteiza wrote:
> Hi,
>
> Commit 1485191 causes a regression on my hardware (laptop) manifesting
> itself as increased power consumption when playing music on mpd using
> pulseaudio.
>
> I bisected to find 1485191.  I tested the commit it reverts (69a37bea),
> and the commit prior 69a37bea, e34a00d6.  Here's a summary:
>
>    status      sha   power  coretemp
>    ---------------------------------
>    BAD    14851912     14W       54C
>    GOOD   228b3023      9W       47C
>    GOOD   69a37bea   ~8.8W       47C
>    GOOD   e45a00d6   ~9.3W       47C

Hi Mark,

thanks for the measurements.

Is it possible you do some perf benchmarking. IIRC, this patch was 
reverted because it introduced some perf regression.

   -- Daniel

>
> I thought it was strange that it happens only with that combination of
> audio software, so I was happy to patch out that changeset which was bad
> for me.  Past 3.12 (now on 3.12.5), the issue persists and I'm not sure
> I can easily merge anymore.
>
> I attached some info I hope will be helpful.
>
> Thanks,
>


-- 
  <http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog