Re: Polypaudio for Gnome 2.10, the next steps

[Lennart Poettering <mztabzr 0pointer de>]

On Thu, 25.11.04 16:02, Mike Hearn (mike navi cx) wrote:

> [context is: dmix and polypaudio, from the gnome desktop-devel-list, see
> lists.gnome.org]
> 
> On Thu, 25 Nov 2004 03:06:48 +0100, Lennart Poettering wrote:
> > You're right. The mixing is done in the client. However this is no
> > win. The client select()s on a unix socket for being notified when to
> > write the audio data to the shared memory segment. i.e. the extra
> > context switch is always there. 
> 
> OK. I didn't get that far ;) I'll take your word for it. What is the
> semaphore used for then?

For the locking of the sysv shared memory segment. The PF_UNIX fd is not
used for locking and not used for passing the actual audio data
around. It is used for buffer change notification via select() or
poll(). It's the fd that is exported through the ALSA API.

> I got curious and CCd Jaroslav, maybe he can explain how dmix is better
> than a sound server (if it really is). Presumably he must think this
> otherwise why implement it?

I don't see how he should implement this otherwise. He needs a file
descriptor for change notification because he needs something to
export to the ALSA client. On the other side of that unix socket must be
another process/thread, so he has to fork on off.

Lennart

-- 
name { Lennart Poettering } loc { Hamburg - Germany }
mail { mzft (at) 0pointer (dot) de } gpg { 1A015CC4 }  
www { http://0pointer.de/lennart/ } icq# { 11060553 }