On Wed, Feb 18, 2009 at 02:44:16PM -0800, David Brownell wrote:
> On Wednesday 18 February 2009, Mark A. Greer wrote:
> > +DAVINCI_SOC_START(dm355)
> > +       .cpu_clks       = dm355_clks,
> > +       .mux_pins       = dm355_pins,
> > +       .mux_pins_num   = ARRAY_SIZE(dm355_pins),
> > +DAVINCI_SOC_END
> 
> That counts as a gratuitous/needless use of macros.
> Just use normal struct decls for this stuff, it'll
> be a lot more clear what's going on.

Personally, I think its kinda cool (and not unprecedented)
but I don't really care so, okay.

> Also, correct me if I'm wrong but the EDMA init
> will need to grow up a bit too ... e.g. register
> a SOC-specific platform_device in the soc-specific
> init logic, grow platform_data to describe how
> many TCs it's got (dm646x has more than two) and
> how many hardware channels it's got (OMAP-L1xx
> has just 32, not 64), use ioremap(), etc.

Yep...I'm nowhere near done.

> (That's sort of the converse of this patch ...
> removing some pseudo-commonality.)

I don't see how its converse to what I'm doing.  I'm trying to lay the
groudwork to enable that.  Encapsulate SoC specifics in a structure
then make common code use the info to do whatever.  To add a new SoC,
just fill out a new struct with the right info.  Beats sprinkling
cpu_is_* calls all over the place, IMHO.  Hopefully, cpu_is_* disappears
when this is all over.

Mark
--

_______________________________________________
Davinci-linux-open-source mailing list
[email protected]
http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source

Reply via email to