[pulseaudio-discuss] [RFC] Dynamic reconfiguration of sampling rate

[Mark Brown]

On Tue, Jan 18, 2011 at 12:16:33PM +0000, Colin Guthrie wrote:
> 'Twas brillig, and Mark Brown at 18/01/11 11:51 did gyre and gimble:

> > If you play a new message notification over a movie then if the
> > notification is an MP3 at 44.1kHz and the movie is at 48kHz you need SRC
> > as well as mixing, and even on a phone you may want to route things to
> > different sets of outputs simultaneously (eg, music to headphones and
> > ringtone to headphones plus speaker).  There's also the simple argument
> > that if your hardware is capable of offloading some of the DSP work from
> > the CPU more efficiently than the CPU can do it itself that's a win.

> Without wanting to fork this thread, I think the routing is separate
> from the SRC stuff (unless the hardware is suitably outsmarting us here!)

There ends up being overlap in terms of implementation underneath - to
a large extent what you're doing is breaking the signal path in software
at some point prior to spitting out the final mixed stream and stuffing
it into hardware.  It's all done under software direction still, it's
just the data path is getting offloaded.