Hi, Jaya,

I'm confused when you say "cpu timing" could be
adjusted by user through fb_var_screeninfo. Could you elaborate on
that?

my saying is the case that display controller creates cpu timing.
display controller with cpu mode would create cpu timing signals
according to cpu timing variables.

the reason, using cpu interface, is because it could reduce power consumption.
in case of rgb interface, screen data with 60fps are transferred to lcd panel automatically
according to rgb signals.
but cpu interface is transferred only when triggerred, of course the way of triggering could
be different according to platform. (also adjusting cpu timing delay)
ex) if there is stop screen, it doesn't need to trigger and
      if slow moving, it could increase cpu timing delay.

as I said to Andraw before, cpu timing variables could be used gernerically.
because lcd panel includes cpu mode also.
but linux framebuffer framework didn't consider cpu mode.

if there is some part I don't know, please give me some advices.
I could modify my patchs anytime.

thank you.

------- Original Message -------
Sender : Jaya Kumar[off-list ref] 
Date   : 2010-07-02 08:47 (GMT+09:00)
Title  : Re: [PATCH 1/3] FB: Add some members for CPU Interface.

Hi InKi,

Thanks for your reply.

On Wed, Jun 30, 2010 at 12:36 PM, InKi Dae [off-list ref] wrote:

Hi Jaya,

2010/6/30 Jaya Kumar [off-list ref]:

quoted

2010/6/29 InKi Dae [off-list ref]:

quoted

CPU interface needs cs, wr setup, wr act and hold delay.
I added some members for them to common framework.

Hi InKi Dae,

The patch looks interesting. Could you help us understand more about
it from a big picture perspective? ie: how is this "cpu interface"
used? I think fb_var_screeninfo is intended to be a very generic data
structure and since it is exposed to userspace we should be cautious
about what we add to it. I didn't understand the purpose of exposing
cs, wr setup, wr act and hold delay to userspace.

in case of lcd panel with cpu interface, it could get screen data
through arm core
or display controller supporting cpu interface mode.
display controller of s5pv210 chip can create cpu interface signal according to
cs, wr setup, wr act and hold time setting.
the reason I added cpu timing variables to fb_var_screeninfo is for using common
framework when setting them to the display controller as cpu timing info.
please, see [PATCH 2/3], s3c_fb_set_timing function part.

and cpu timing could be adjusted by user through fb_var_screeninfo,
for this, suitable checking should be considered at machine specific
fb_check_var func.
also you can see that through [PATCH 2/3], s3c_fb_check_var.

I'm still having difficulty understanding this. I guess what I am
asking is why do these new variables "cs, wr setup, wr act and hold
time setting", have to be in fb_var_screeninfo which is a generic
platform/implementation independent structure exposed to userspace? If
I've understood things correctly, the variables themselves, eg: wr
setup seem to be specific to this implementation rather than being
generic like the other members of fb_var_sreeninfo. Have you
considered an alternative way by which you can achieve the desired
functionality. Also, I'm confused when you say "cpu timing" could be
adjusted by user through fb_var_screeninfo. Could you elaborate on
that?

I haven't read through the remaining part of the message about
mipi-dsi issues, I'm hoping developers who have implemented the other
MIPI-DSI implementations in fbdev could chime in.

Thanks,
jaya