Re: [Qemu-devel] [PATCH 0/2] Make mixer emulation configurable at runtime

[Markus Armbruster]

Gerd Hoffmann <address@hidden> writes:

>   Hi,
>
>>   Only pulse and the spice audio backends set VOICE_VOLUME_CAP.
>> 
>>   Impact on users isn't obvious to me.
>
> It offloads volume control to the audio backend, i.e. it will be done by
> pulseaudio/spice-client instead of mixemu.
>
>> Questions (not just for you, Bandan):
>> 
>> 1. Why is CONFIG_MIXEMU off by default?
>
> malc wants it this way because mixemu eats cpu time.
>
>> 2. Why do we bother providing these devices when CONFIG_MIXEMU off?
>> 
>>    Why would anyone want hda audio devices without a mixer?  Why
>>    wouldn't anyone who wants hda audio devices also want CONFIG_MIXEMU
>>    enabled?
>
> It'll actually work just fine with HDA.  The guest figures there is no
> volume control and will fallback to do volume control in software then.
> Net effect is that the audio sample processing to change volume is done
> by the guest instead of mixemu, so you don't actually save cpu time by
> turning off mixemu ...

I don't doubt it works, I doubt anybody would want the software mixing
done in the guest instead of mixemu so badly that he wants to use a
special device that makes that possible.

> I think AC97 has broken volume control with CONFIG_MIXEMU=n, because the
> guest tries to use the non-functional volume controls.
>
> So, yes, for serious sound support you want CONFIG_MIXEMU=y.

Which brings me back to question 1: why is non-serious sound support the
default?