Hi Daniel,
Thank  you for your feedback. The problem I mentioned is produced under Android environment. You can compile the latest powertop on android5.0 and run them by 'powertop --csv'. You will find all pstat shows 100% on idle.
Besides, I also find bellowing inquiry on 01.org which ask the same question.

https://lists.01.org/pipermail/powertop/2015-June/001801.html

BRs
zhaoyang

On 7 July 2015 at 05:46, Daniel Leung <daniel.leung@linux.intel.com> wrote:
On Wed,  1 Jul 2015 19:34:40 +0800
Zhaoyang Huang <zhaoyang.huang@linaro.org> wrote:

> all children cpu's pstat data will be cleared in process_cpu_data,which
> cause the pstat shows abnormal, we have to comment this line of code. However,
> it looks too rough.
>
> Signed-off-by: Zhaoyang Huang <zhaoyang.huang@linaro.org>
> ---
>  src/main.cpp |    7 +++++++
>  1 file changed, 7 insertions(+)
>
> diff --git a/src/main.cpp b/src/main.cpp
> index 1e56af1..9526084 100644
> --- a/src/main.cpp
> +++ b/src/main.cpp
> @@ -221,7 +221,14 @@ void one_measurement(int seconds, char *workload)
>       devices_end_measurement();
>       end_power_measurement();
>
> +     /*
> +     FIX ME:all children cpu's pstat data will be cleared in process_cpu_data,which
> +     cause the pstat shows abnormal, we have to comment this line of code. However,
> +     it looks to rough
> +     */
> +     /*
>       process_cpu_data();
> +     */
>       process_process_data();
>
>       /* output stats */
> --
> 1.7.9.5
>

Did you see the issue on Android? Could you give me more details? Maybe I can try to re-produce it on my end.

--
Daniel Leung <daniel.leung@linux.intel.com>