[alsa-devel] [PATCHv2 - CA0132 HDA Codec 1/2] ALSA: Update Creative CA0132 codec to add DSP features.

Takashi Iwai tiwai at suse.de
Wed Aug 1 07:48:02 CEST 2012

At Tue, 31 Jul 2012 19:38:05 -0700,
Ian Minett wrote:
> >>>The only question is where to implement the DSP loader stuff and how.
> >>
> >> Ok, we understand. Given that the codec cannot call into the controller's
> >> PCM interface to load the DSP image, is there an acceptable way you could
> >> recommend for us to add this functionality to the module, without breaking
> >> the architecture?
> >
> >Well, as mentioned earlier, we need to push it up to the controller
> >layer, i.e. hda_intel.c.  For example, add a new op for firmware DMA
> >transfer in hda_bus_ops.
> >
> >Could you describe briefly the procedure for the DMA transfer of
> >CA0132?  The code looks too complex to follow.
> Sure - the DSP code download procedure is:
> 1.  The CA0132 driver extracts DSP data segments that need to download to
>       the DSP chip.
> 2.  Each of these data segments will be downloaded to the DSP chip via
>       HDA-link, like audio data.
> 3.  The HDA controller and CA0132 are programmed to transfer the DSP data
>       segments.

So, the data transfer here is just like the normal PCM streaming?
For example, if you have an op like

	int (*load_dsp)(struct hda_bus *bus, void *buffer, int size);

and calling this for each segment would work?
(In addition, we'd need to give some way to determine the stop
 condition in the codec side.)

> 4.  The HDA controller FIFO must be flushed on every data segment to ensure
>       that there is no previous residual data.

Actually FIFO flush might be needed for the audio transfer as well.



> These are the controller ops that the CA0132 driver used for the DMA transfer:
> #define azx_pcm_open(a)       (a->ops->open(a))
> #define azx_pcm_close(a)      (a->ops->close(a))
> #define azx_pcm_prepare(a)    (a->ops->prepare(a))
> #define azx_pcm_trigger(a, b) (a->ops->trigger(a, b))
> #define azx_pcm_hw_free(a)    (a->ops->hw_free(a))
> Thanks,
> Ian

More information about the Alsa-devel mailing list