On Mon, Feb 2, 2015 at 7:34 AM, Lisa Nguyen <lisa.ngu...@linaro.org> wrote:
> Set the sampling_rate to zero if the sampling_rate attribute cannot
> be found in either directories listed in the if condition. This will
> also resolve minor bugs where the sampling_rate is missing.
>
> Signed-off-by: Lisa Nguyen <lisa.ngu...@linaro.org>
> ---
> include/functions.sh | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/include/functions.sh b/include/functions.sh
> index e3eb140..4486da8 100644
> --- a/include/functions.sh
> +++ b/include/functions.sh
> @@ -164,8 +164,10 @@ wait_latency() {
> # consider per-policy governor case
> if [ -e $CPU_PATH/$wait_latency_cpu/cpufreq/$gov ]; then
> sampling_rate=$(cat
> $CPU_PATH/$wait_latency_cpu/cpufreq/$gov/sampling_rate)
> - else
> + elif [ -e $CPU_PATH/cpufreq/$gov/sampling_rate ]; then
> sampling_rate=$(cat $CPU_PATH/cpufreq/$gov/sampling_rate)
> + else
> + sampling_rate=0
> fi
> sampling_rate=$((sampling_rate * 1000)) # unit nsec
>
Hmm, let us think about this a bit. In what situation will be not find
a sampling rate for the governor? In the case where we are NOT using
ondemand governor. This is expected on Android which uses the
interactive governor.
So the entire wait_latency() function is going to return somewhat
incorrect values when run on Android.
I suggest the following removing this patch from the series and
properly fixing this problem by doing tests on ubuntu and android with
ondemand and interactive respectively to figure out what variables
should be read to get the sampling rate.
_______________________________________________
linaro-dev mailing list
linaro-dev@lists.linaro.org
http://lists.linaro.org/mailman/listinfo/linaro-dev
