[arch-general] [arch-announce] GNOME3 in [testing]

[Oon-Ee Ng]

On Sat, Apr 9, 2011 at 8:21 AM, Matthew Monaco <dgbaley27 at verizon.net> wrote:
> On 04/08/2011 02:11 AM, Tom Gundersen wrote:
>>
>> On Fri, Apr 8, 2011 at 10:38 AM, Matthew Monaco<dgbaley27 at verizon.net>
>>  wrote:
>>>
>>> I like having volume control for HDMI out. And I like that this
>>> mysterious
>>> starving audio, playing silence, glitch has disappeared. But I don't like
>>> that pulse uses so much CPU and i think it's a major reason why my
>>> computer
>>> is runn 5-8 degrees C hotter with gnome3
>>
>> FWIW, this is almost certainly a bug that can be fixed. In my
>> experience PA allows huge power savings compared to ALSA (down from
>> more than 100 wakeups/sec to less than 10 wakeups/sec).
>>
>> Try debugging a bit to see what is going on powertop2 should be very
>> helpful.
>>
>> Cheers,
>>
>> Tom
>>
>
> I don't see how your CPU can be _less_ active with PA. I thought PA added a
> layer that didn't exist before and didn't actually remove anything.
>
> I'll say though, that changing VLC from 'default' or explicitly ALSA to
> PulseAudio dropped CPU usage of PA from ~16% to 3%. (It's mysterious to me
> why that 3% remains when the media is paused though).
>
> I guess I need to go through gstreamer and other app-specific settings and
> explicitly point to PA.
>
PA buffers better than ALSA, or is supposed to, in any case. Of
course, if you're using its ALSA-emulation that's a moot point. AFAICR
(haven't had this problem for ages), CPU usage from PA is basically
caused by the resampling, which you can just modify to a simpler
method if you don't need the higher quality of the default (speex, I
believe).

So yes, PA would save CPU cycles and battery (its meant for use on
cellphones as well, after all, the nokia people were involved at some
point).