[Qemu-devel] gsoc idea

[Qemu-devel] gsoc idea

[Kővágó Zoltán]

hi,

i'm thinking about doing a gsoc this year, and i'm interested if you 
have any opinions on my idea. (i hope i'm not late for this...)

currently the audio system in qemu is limited to 2 channels (stereo) 
sound. also it mixes all sound (if there's more than one sound card) 
into a single stream, you can't send them to different physical sound 
cards for example. if you need something like that, you'll have to 
either pass-through a sound card (but then you can't use it from the 
host), or use pulseaudio/jack/etc to send the audio over the network 
(which is problematic in case of a windows guest)

imho the audio subsystem should be rewritten to support any number of 
audio channels. also, the current global mix everything together into a 
single output method (along with it's environment variable based 
configuration, since it only allows global options, not per-device) 
should be replaced, with a system similar to how network or drives works 
(i.e. each virtual sound card would be connected to an audio device, 
with their own options)

is it a viable idea, or i should scrap it altogether?

Re: [Qemu-devel] gsoc idea

[Gerd Hoffmann]

On Sa, 2015-03-07 at 03:39 +0100, Kővágó Zoltán wrote:
> hi,
> 
> i'm thinking about doing a gsoc this year, and i'm interested if you 
> have any opinions on my idea. (i hope i'm not late for this...)

Should not be too late for this, deadline is a week out.

> currently the audio system in qemu is limited to 2 channels (stereo) 
> sound. also it mixes all sound (if there's more than one sound card) 
> into a single stream, you can't send them to different physical sound 
> cards for example. if you need something like that, you'll have to 
> either pass-through a sound card (but then you can't use it from the 
> host), or use pulseaudio/jack/etc to send the audio over the network 
> (which is problematic in case of a windows guest)

> imho the audio subsystem should be rewritten to support any number of 
> audio channels. also, the current global mix everything together into a 
> single output method 

I don't think you have to put *everything* upside down.  As far I know
the mixing and the two channels are not designed into the audio core
code.

> (along with it's environment variable based 
> configuration, since it only allows global options, not per-device) 
> should be replaced, with a system similar to how network or drives works 
> (i.e. each virtual sound card would be connected to an audio device, 
> with their own options)

For the audio backend configuration a serious makeover would be very
good indeed.

> is it a viable idea, or i should scrap it altogether?

Sounds good, go ahead, I can mentor this.

cheers,
  Gerd

Re: [Qemu-devel] gsoc idea

[Gerd Hoffmann]

  Hi,

> > is it a viable idea, or i should scrap it altogether?
> 
> Sounds good, go ahead, I can mentor this.

Draft added to wiki page:

http://qemu-project.org/Google_Summer_of_Code_2015#QEMU_audio_backend

Drop me a note when you think this needs changes.

cheers,
  Gerd