[LAU] ALSA_OUT hack

[drew Roberts]

On Wednesday 10 November 2010 05:08:00 torbenh wrote:
> On Tue, Nov 09, 2010 at 11:43:09PM -0800, Ken Restivo wrote:
> > Question about the alsa_out tool in the JACK github.
> >
> > How critical is the libresample portion of the code? What would happen if
> > I removed it?
> >
> > i.e. if I were sending, say, one stereo pair to one amp, and another
> > stereo pair to a different amp, and didn't care about the word clocks
> > being in sync, do I really need that libresample stuff sucking up
> > valuable CPU?  Or could I whack it? What would be the difficulties
> > created by doing that? Again, this'd be only for output channels and I
> > wouldn't need the word clocks to be in sync.
>
> you are confusing things here.
> its not about wordclocks being in sync. if your wordclocks are in sync.
> you dont need alsa_out.
>
> but most likely your wordclocks are not in sync. this means your
> soundcards run at different drifing samplerates, and one soundcard will
> consume a different amount of samples per jack cycle. but jack is
> feeding period samples each cycle.
>
> first try using -q option set to 0 because that will use linear resampling.
> reducing the value of the -s option (default 256) will also improve cpu
> usage. but might yield worse performance of the algorithm. it depends a bit
> on the soundcards you are using.

Didn't he mention having one physical soundcard with 4 outs that only 
presented 2 outs to jack? If so, do any such single physical soundcards 
actually have two clocks? The alsa-out trick would just get the other two 
outs on the same card to show in jack. Or is my memory acting up this 
morning?

all the best,

drew