RE: DPM

Hi,
I feel big pleasure reading the DPM document - finally, we will have OS
features in order to dynamically support applications. Great job is going
on.

I have been working on practical implementation of power management into
products (cell phones. The modern architecture solutions for portable
devices (PDA, cell phones,...) are multi cores at first, including a lot of
smart peripherals, HW accelerators and interfaces. At second, it includes
already multiple RF communication channels, such as BT, WLAN, which are
pretty complicate systems itself.  Not only LCD control, but other
peripherals and RF channels has multiple operating points, requiring also
dynamic power management for optimized power control.

1. 	In order to efficiently support  applications under OS, it should
include complicate multiple power managements architecture, covering all HW
resources, not only Core. And we do have it as addition to OS as
application, which is not efficient way due to lower performance as result.

I have reviewed DPM Rev.1.1 document. You already mentioned about criticism,
"need single, unified architecture for dynamic power management of every
system component and attached devices", "...Linux to favor the system-wide
power management approaches based on sets of cooperating power managers!",
..."Power management mechanism for many type of devices already exist in
operating systems" - where it is in Linux?

You referenced on document [10], which has complicate structure and
overloaded with objects, trying to cover everything.
I agree strongly about your concept of multiple power managers. However, I
think,  OS should have some tools, architecture features, API, in order to
give ability to create other power managers and a global device power
manager without additional power servers and new definitions of API
functions and descriptors. 
And, analyzing DPM, I think, it might be significantly easier solution , if
you expend similar DPM architecture advantages and API (with some minor
changes) for other HW resources (modules)in order to cover it and to make
global power management, using kernel API entry points... I can provide more
inputs and my thoughts, in a case you have interest.


2.	I have also some methodology notes.
During a reading of the document I had some difficulties in a distinguishing
of some basic terminology. I understand, that you try to simplify the
terminology also, but I consider it from practical utilization it during
power management developments.

I would put a terminology of Power State (in DPM - state), distinguishing it
from operating state, which are based on the regular tasks, and substitutes
it in DPM operating state. Power state usually support multiple operational
states and tasks without changing the specific operating points, which
change power consumption. During power management development we assign to
multiple operating tasks specific power states in order to support it.

The main confusion - a meaning of Task Operating States - should be Power
Tasks, the tasks dedicated to set up next operating point, which should be
defined as Power point. There are multiple tasks, including HW control tasks
(events), but, besides it, there are specific control tasks, which change
the power states or power consumption. And it should be distinguished from
others.
However, this is the minor factors for the total ideas.

Best regards.

Vladimir Sklovsky
Architecture, System Engineering
Motorola, PCS
600 N US Highway #45
Libertyville, IL 60048
(847)523-6183
[email protected]










-----Original Message-----
From: Bishop Brock [mailto:[email protected]] 
Sent: Friday, January 17, 2003 10:41 PM
To: Sklovsky Vladimir-WLVS03
Subject: Re: DPM



Vladimir,

Thanks for the note.  We don't have a website, but it's probably a good
idea.  I expect that soon we will at least have a network presence in the
form of a public source code tree (hosted by MontaVista) that others can
look at and help us work on.  Once we have the source code public, a good
contribution would be for someone from Motorola to offer a DPM
implementation for an embedded processor that you were using or thinking
about using in one of your designs.

Thanks again,

Bishop

IBM Research, Austin Center for Low-Power Computing
11400 Burnet Road    MS/904-6F021
Austin, TX 78758
(512) 838-0149    IBM T/L 678-0149



Sklovsky Vladimir-WLVS03 <[email protected]> on 01/16/2003 04:21:28 PM

To:    Bishop Brock/Austin/IBM@IBMUS
cc:
Subject:    DPM




Hi,  Bishop.
I have received mail  from Victory Sales P. Meiss in regards of your DPM ( I
have had already Version  1.1) and IBM/MontaVista announcement. All these
efforts and results, you showed  to me during the presentation in Motorola
PCS, are very impressive. Your color  plot with reduction to 13X is on my
desk as very nice picture for  me. Please, inform me of  any news in DVFS
development (is there some specific web site?). What kind of  contribution
might be helpful for you and this development?

Best regards and  good luck into DPM development.


Vladimir Sklovsky
Architecture, System  Engineering
Motorola, PCS
600 N US Highway #45
Libertyville, IL 60048
(847)523-6183
[email protected]





-------------------------------------------------------
This SF.NET email is sponsored by:
SourceForge Enterprise Edition + IBM + LinuxWorld = Something 2 See!
http://www.vasoftware.com