Re: About snd_soc_dai_set_fmt()

[Mark Brown <broonie@xxxxxxxxxx>]

On Thu, Jun 06, 2024 at 06:55:29PM +0800, Shengjiu Wang wrote:
> On Wed, Jun 5, 2024 at 7:40 PM Mark Brown <broonie@xxxxxxxxxx> wrote:

> > Do you have an example of such a use case?  If you can do that it feels
> > more like one end has two single direction DAIs (ISTR some of the
> > Freescale platforms were built like that?) and it might be better to do
> > that and make it easier to combine two DAIs.  That would also work for
> > things like combining dedicated DACs and ADCs.

> one of case is:
> 
> +-------+                +-----------------+
> | SAI1  |   --TX-->  |                     |
> |           |   <--RX-- |                     |
> +-------+                |                     |
>                              | AUDIOMIX  |
> +-------+                |                     |
> | SAI2  |   --TX-->  |                    |
> +-------+                +----------------+

> AUDIOMIX mixes the output of SAI1 and SAI2,  AUDIOMIX's output
> is sent to SAI1.

So that's a DPCM case but hitting the DPCM scaling limits.  One of these
days someone will actually have a use case that generates enough pull to
get digital routing done...

> Use multi-DAI may be an option,  but for fsl_sai driver, in some cases it
> is used as single DAI.  So how to support single DAI and multi-DAI in one
> driver?  Should we define a new dts property to distinguish these two cases
> in order to register different component + dai driver?

Some drivers have bodged it by adding 3 DAIs, one bidirectional and the
other two unidirectional and relying on users to DTRT.  Nothing is super
great there.

Attachment: signature.asc
Description: PGP signature