power management

[Jim Gettys]

On Wed, 2007-03-28 at 21:49 -0500, Gopi Palaniappan wrote:
> Hi:
> 
>     I have the same problem. There seems to be lot of information on
> the olpc website but not easy to get to without spending time
> searching for it. After a lot of searching and forgetting to bookmark
> and researching, I decided to add the links of interest to me in my
> blog: 
> http://tags07.blogspot.com/2007/03/olpc-links.html
> Feel free to refer to it.  
> 
> On a different note, I'd like to work on PM user interface and perhaps
> also on the kernel side. 
> I havn't dived into the hardware specs yet for OLPC. It would be nice
> to have a wiki page with some basics to help ramp up the learning
> curve.

http://wiki.laptop.org/go/Hardware_specification

> Some qns I have:
> 1. Does the hardware support Voltage/Frequency scaling? 

No.

> 2. What kind of sleep modes does it support?

Suspend to RAM.

But we can leave the display image intact on the screen, and the
wireless interface forwarding packets.  So the system can appear to be
operating even with the processor suspended.

> 3. Is there an infrastructure for power measurements? 

Being set up; not fully ready yet.

> Do we have any current drain measurements/readings?

Just some very early ones.

> 4. Where are you experiencing the instability..is it at the hardware
> layer or the syspend-resume mechanism? 
> 5. Recently we did a lot of cleanup in the kernel to short-circuit the
> default pm_suspend process, which has unnecessary layers that trumps
> the suspends... the biggest one that comes to my mind is the dealing
> with UNINTERRUPTIBLE SLEEP processes. I am not sure if you are
> experiencing similar issues? 
> 6. Is there a need for periodic wakeup? on mobile devices its required
> to wakeup almost every minute to update clock display, battery level,
> rf levels.it makes sense to sync up all these activities during the
> wakeup period.I am not sure if a laptop has such a requirement..

Yes, we can anticipate probably doing this; but we have an embedded
controller which will reduce the need.

> 7. Usually display/backlight is the biggest sucker of power.. do we
> use any special displays that support partial refresh?

We have a novel display.

In reflective mode, at 25hz, it consumes 100mw.  In backlight color
mode, it consumes a maximum of 1 watt, using LED backlights.

> 8. Suspend to RAM is good..it will give an "Instant-On" experience..is
> this in the main kernel now or does it need special patches? If so
> where can i find these patches? 
> 
> I'd also like to discuss user interface APIs and useful features for a
> power controller module in userspace. But i understand a stable
> suspend and resume of devices should be the priority.
> 
> 
> thanks,
> Gopi
> 
> 
> 
> 
> 
> 
> 
> On 3/28/07, Jordan Crouse <jordan.crouse at amd.com> wrote:
>         On 28/03/07 19:42 +0200, Jens Axboe wrote:
>         > Hi,
>         >
>         > I'd love to help out with the power management and
>         suspend/resume parts,
>         > but so far I'm having trouble finding out what the status of
>         it is, 
>         > goals, directions, etc. Did I miss a wiki page on this? If
>         not, anyone
>         > care to outline the current problems that need tackling?
>         
>         Once we get the core drivers working happily (graphics, DCON,
>         USB, and 
>         CaFE NAND), then we'll be ready to be more organized with wiki
>         pages, and
>         images and such.  Following working core drivers, our emphasis
>         will shift
>         to the fringe drivers.
>         
>         My personal target is to survive 1000 consecutive
>         suspend/resume cycles 
>         with a full system running a nominal load (sugar, nominal
>         CPU load + network).  If we accomplish that, then I think
>         we'll be ready for
>         anything they throw at us.
>         
>         After that we'll shift our attention to optimization, but I
>         would 
>         like to stress that at this point, stability trumps speed any
>         day of the
>         week and twice on Sunday.
>         
>         Jordan
>         
>         --
>         Jordan Crouse
>         Senior Linux Engineer
>         Advanced Micro Devices, Inc.
>         <www.amd.com/embeddedprocessors>
>         
>         
>         _______________________________________________
>         Devel mailing list
>         Devel at laptop.org
>         http://mailman.laptop.org/mailman/listinfo/devel
> 
> _______________________________________________
> Devel mailing list
> Devel at laptop.org
> http://mailman.laptop.org/mailman/listinfo/devel
-- 
Jim Gettys
One Laptop Per Child