Re: [PATCH 3/3] ARM: S3C64XX: Implement basic power domain support

["Tomasz Figa" <tomasz.figa@gmail.com>]

Hi,

W dniu 2 grudnia 2011 01:56 użytkownik Mark Brown
<broonie@opensource.wolfsonmicro.com> napisał:
> On Fri, Dec 02, 2011 at 09:35:44AM +0900, Kyungmin Park wrote:
>
>> I'm not sure what's the next step at s3c64xx for generic power domain.
>> Related with exysno4 series, it's helpful to read following threads.
>> http://68.183.106.108/lists/linux-pm/msg26291.html
>
>> "I don't think we should control/gate the clocks with regarding power
>> domain" from Mr. Kim
>
> I tend to agree with him that gating the clocks automatically using
> runtime PM would be nice as if nothing else it saves code in drivers.
> I'm not sure how easy it's going to be to transition to doing that in
> one jump, though - there's a lot of SoCs sharing IP right back to
> S3C24xx.
>
> Of course it should be fairly easy for most of the domains on s3c64xx
> right now as there's only one actual driver involved in mainline, we
> should just be gating almost all of the clocks concerned immediately
> after we boot (assuming they aren't gated by default, I'd need to check)
> and never touching them again anyway.  It'd only be the framebuffer
> clocks that'd need some work and for a first pass we could just mark the
> framebuffer domain as always on and get most of the win.  Sometimes lack
> of driver support is a good thing :)

Please do not forget that there might be some drivers not yet submited
to mainline and mainline should not break them with an assumption that
there are no such drivers.

For example, there is an on-going work on an open source OpenGL
implementation for the GPU in S3C6410, known as OpenFIMG. Currently it
uses a little kernel module for low level hardware management
(interrupts, contexts, locking, power management), involving clock
gating and runtime power management, but a DRM driver is planned.

Best regards,
Tom