[pulseaudio-discuss] Pulseaudio Mixing & DSP Mixing

Marc-André Lureau marcandre.lureau at gmail.com
Wed Dec 23 07:06:49 PST 2009


On Wed, Dec 23, 2009 at 3:49 PM, Lennart Poettering
<lennart at poettering.net> wrote:
> On Wed, 23.12.09 14:04, tieg (tie.ge at access-company.com) wrote:
> That said, I am quite sure that hw mixing is not particularly useful
> on desktops. It has some validity however in embedded environments.

On the N900, mixing almost never happen. Due to policy, only one
stream is playing at a time usually. Only event sounds gets mixed for
a few milliseconds - only if you have them enabled (most avanced users
tend to disable them).

The complexity and roundtrip required to make use of the DSP on omap3
is not worth. Afaik, the TI dsp bridge is really not that performant,
and does not deal well with task/resource management,... Better be
careful with what you do with it, like knowing well your set of tasks,
their resources...

However, if your use case is to do audio mixing, for some specific
music/audio device or application, then that could make sense. It
would need to be verified though, because the DSP could sleep and live
resampling + mixing with NEON is still quite cheap on CPU side only,
it might not make you jump c/p-states for instance.


Marc-André Lureau
Sent from Barcelona, Spain

More information about the pulseaudio-discuss mailing list