On Thu, Jan 23, 2025 at 11:38:33AM +0200, Daniel Baluta wrote:
> On Wed, Jan 22, 2025 at 9:55 PM Frank Li <Frank.li@xxxxxxx> wrote:
> >
> > On Wed, Jan 22, 2025 at 06:35:42PM +0200, Daniel Baluta wrote:
> > > After commit 2b9cdef13648 ("ASoC: SOF: imx: Add devicetree support
> > > to select topologies") we select topology to be used by the board
> > > compatible string in the dts.
> >
> > I am confused. why not use "sof-imx8-cs42888.tplg" in imx8's dts instead
> > use board's compatible string "fsl,imx8qxp-mek-bb".
> >
> > More and more boards will be added in future. This file will become bigger
> > and bigger!
>
> Hi Frank,
>
> So you mean create a dts property named e.g sof-tplg-name and read the
> topology name from dts?
>
> dsp {
> firmware-name="sof-imx8.ri";
> tplg-name = "sof-imx8-cs428888.tplg";
> };
dsp {
compatible = "imx8-dsp-xxxx";
...
}
in dsp driver, map "imx8-dsp-xxxx" to "sof-imx8-cs428888.tplg" and
"sof-imx8.ri".
>
>
> This was our first approach but then we aligned with the community on
> using compatible to match
> for properties like default fw path, default tplg path, fw name, tplg name.
What's origial email thread. I think "compatible" refer to dsp node, not
to boards level compatible string.
>
> Now dts should somehow describe the hardware not the software (e.g
> tplg binary) used on that platform.
> The growing number of boards will not be a problem, each one will have
> a compatible in the dts.
Yes, but your C files will fill up meansless map code when more board
added.
>
> But the concern here is how we describe a hardware setup where we have
> a CPU board + a baseboard
> like it is the case for cs42888 codec?
>
> Our approach was that in our dts or overlay dts that describes SOF
> with cs428888 codec to put the exact
> compatible "fsl,imx8qxp-mek-bb" and match it with static vector in SOF
> driver that you see in this patch.
>
> SOF core relies on the compatible string in order to identify the
> correct tplg_filename for example.
Any application ctrl tools can link these nodes?
Frank
>
> thanks,
> Daniel.
