[alsa-devel] ASoC v2 on PowerPC

[Jon Smirl]

On 10/5/07, Timur Tabi <timur at freescale.com> wrote:
> Jon Smirl wrote:
> > On 10/5/07, Liam Girdwood <lg at opensource.wolfsonmicro.com> wrote:
> >> A good place to start would be looking at the other supported
> >> architectures and codecs and see what fits in best with your target SoC
> >> CPU. Fwiw, Timur Tabi has some working PPC code, although I'm not sure
> >> if it will work on Efika.
> >
> > Efika is STAC9766, ac97 with no i2c control. I dug into the imx ac97
> > driver and discovered that it isn't finished.
>
> A lot of the i.MX code just sits in BSPs and is never forwarded to the main
> repositories.

Aren't you fixing that?

> > Timur, we may need to get some Bestcomm tasks written. I haven't
> > explored this yet but I've been told that no one can get anything
> > except the general DMA task working for audio. This mean all audio
> > input has to be 32b. Isn't Bestcomm capable of expanding 16/24 bit
> > data? Our target hardware is i2s, not ac97, so I need to get i2s
> > running too.
>
> I'm working on an audio driver for the 8610, but I'm going to target ASoC V1
> first, just so that I can get a better feel for how V2 is different (a sort of
> "trial by fire").

I started looking at v2 because of the posts about modularizing it and
using open firmware to identify and load the codecs. There doesn't
seem to be that much difference.

> I don't know much about Bestcomm, but the DMA engine on the the 8610 is much
> simpler, from what I've heard.  I only work on 83xx, 85xx, and 86xx, so I
> really can't answer any Bestcomm questions.

We are on 5200 because it is cheaper and 5200 has all the peripherals
we need. Now that I am writing code for it I see that BestComm adds a
lot of complexity. Hopefully I can get i2s working on it, I'd hate to
switch CPUs after doing all this work.

> As for requiring all audio to be 32b, that's not such a bad thing.  Any
> application that uses the standard ALSA interface will have format-conversion
> plug-ins automatically invoked.  The overhead will be minor, since it has to
> do a byte-copy anyway.
>
> --
> Timur Tabi
> Linux Kernel Developer @ Freescale
>


-- 
Jon Smirl
jonsmirl at gmail.com