[alsa-devel] How to propagate parameters from FE to BE with compressed audio interfaces

[Pierre-Louis Bossart]

Hi Daniel,

On 11/14/18 9:35 AM, Daniel Baluta wrote:
> Hello,
>
> soc_compr_set_params_fe calls dpcm_be_dai_hw_params in order to setup
> BE hw_params.
>
> Is there a way to transfer parameters like for example the rate from FE
> to BE?
>
> In soc_compr_set_params_fe there is this code:
>
> *
> * Create an empty hw_params for the BE as the machine driver must
> * fix this up to match DSP decoder and ASRC configuration.
> * I.e. machine driver fixup for compressed BE is mandatory.
> */
> memset(&fe->dpcm[fe_substream->stream].hw_params, 0,
>         sizeof(struct snd_pcm_hw_params));
>
> So, information about parameters is just ignored.
>
> The only way we can setup hw_params for BE is via
> be_hw_params_fixup() but this function doesn't have any information
> from FE.
>
> My use case is something like this:
>
> FE:
>      - cpu: FSL SAI DAI interface
>      - platform: DSP
>      - codec: dummy codec
> BE:
>      - cpu: dummy
>      - platform: dummy
>      - codec: cs42888.
>
> codec could handle 44100Hz and 48000Hz streams, but the BE doesn't have
> any way to know what is the rate cplay sets to FE.

I don't think we planned for the case you mention, but IIRC it was 
intentional.

In our implementations and the non-Intel ones I am familiar with, the BE 
typically handles data that is mixed between PCM channels (for 
tone/alerts/low latency) and decompressed data. So the rate adaptation 
of the decoder output is handled inside the DSP and the BE is configured 
at a set rate. I also don't remember when was the last time we supported 
44.1 kHz, maybe 2011?

You *could* add support for dynamic BE reconfigurations but then it 
creates a headache for the mixer usages. Not to mention that decoders 
support rates that the BE may not support so SRC is likely required 
anyways. And I forget schemes like AAC+ implicit signaling where you 
first have to decode before figuring out if you have valid SBR data 
which will result in a 2x upsampling of the output - in other words you 
may not necessarily know what the output frequency of the decoder is 
just by looking up the file so doing a configuration at the DPCM level 
is complicated at best.

-Pierre