RE: on-ness

RE: on-ness

[Brown, Len]

[-- Attachment #1: Type: text/plain, Size: 1719 bytes --]

> Thinking about the discussion of the ON field. How about Limiter? Then
0
> maps to no limit (max power, max freq, whatever) and any other number
is
> some limit of performance/power, similar to what was decided for Idle.

my scribbles on generic sysfs device directory file names say:

state:
on	- running and available
off	- requires a full device initialization to be usable

idle: # = "how idle"
0 - active, not idle at all eg C0, D0
1 - idle. eg C1, D1
...
n - most power saving, highest latency idle state, eg. Cn, Dn

idle_max
max # that can be in idle file above

speed: # = "how fast"
0 - minimum speed
1 - 
...
n -- highest speed, highest power

speed_max
max # that can be in speed file above

So describing the ACPI states using these:

state = on: online
state = off: offline

idle = 0: C0
idle = 1: C1
idle = n: Cn

(for devices with D-states, replace Cx above with Dx -- since they are
both describing a state where the device is present, but not executing
and with increasing latency before resuming execution)

speed = 0: Pn
speed = 1: Pn-1
...
speed = n: P0

Not immediately obvious how to articulate Throttling states here,
would probably need an additional file similar to "speed", since they
are effectively multiplied.  Maybe simply:

throttle
0 - full speed
n - min speed

Re: state
unclear if on/off is sufficient, or if hotplug wold need any
other states.

---

sounds like you're suggesting the inverse of "speed" where
0 is max performance and max power.  I'd certainly be happy
to have 0 mean P0 for processors -- but "on" and "speed"
are certainly the opposite of what we want to call this.
Maybe "powersave" to capture the concept of an executing
but power saving operating point?

-Len


[-- Attachment #2: Type: text/plain, Size: 0 bytes --]

RE: on-ness

[Richard A. Griffiths]

[-- Attachment #1: Type: text/plain, Size: 2004 bytes --]

see below.

On Mon, 2006-04-17 at 17:43 -0400, Brown, Len wrote:
> > Thinking about the discussion of the ON field. How about Limiter? Then
> 0
> > maps to no limit (max power, max freq, whatever) and any other number
> is
> > some limit of performance/power, similar to what was decided for Idle.
> 
> my scribbles on generic sysfs device directory file names say:
> 
> state:
> on	- running and available
> off	- requires a full device initialization to be usable
> 
> idle: # = "how idle"
> 0 - active, not idle at all eg C0, D0
> 1 - idle. eg C1, D1
> ...
> n - most power saving, highest latency idle state, eg. Cn, Dn
> 
> idle_max
> max # that can be in idle file above
> 
> speed: # = "how fast"
> 0 - minimum speed
> 1 - 
> ...
> n -- highest speed, highest power
> 
> speed_max
> max # that can be in speed file above
> 
> So describing the ACPI states using these:
> 
> state = on: online
> state = off: offline
> 
> idle = 0: C0
> idle = 1: C1
> idle = n: Cn
> 
> (for devices with D-states, replace Cx above with Dx -- since they are
> both describing a state where the device is present, but not executing
> and with increasing latency before resuming execution)
> 
> speed = 0: Pn
> speed = 1: Pn-1
> ...
> speed = n: P0
> 
> Not immediately obvious how to articulate Throttling states here,
> would probably need an additional file similar to "speed", since they
> are effectively multiplied.  Maybe simply:
> 
> throttle
> 0 - full speed
> n - min speed
> 
> Re: state
> unclear if on/off is sufficient, or if hotplug wold need any
> other states.
> 
> ---
> 
> sounds like you're suggesting the inverse of "speed" where
> 0 is max performance and max power.  I'd certainly be happy
> to have 0 mean P0 for processors -- but "on" and "speed"
> are certainly the opposite of what we want to call this.
> Maybe "powersave" to capture the concept of an executing
> but power saving operating point?
I like powersave. Then 0 indicates an ON state with no power
saving.  Good idea.

Richard


[-- Attachment #2: Type: text/plain, Size: 0 bytes --]