Re: A simple cpufreq(9)

[Jukka Ruohonen <jruohonen%iki.fi@localhost>]

On Thu, Sep 29, 2011 at 03:36:03PM -0500, David Young wrote:
> What's the difference in power savings between changing C-state and
> changing frequency?  Do the power savings from every change in C-state
> dominate the savings from any change in frequency?

Depends on the machine. But generally on x86, C-states appear to be now the
dominant form. But these go side by side. To cut the corners short: the
general (hardware) idea is that while few CPUs are in a deep C-state (i.e. 
idle), a group of other CPUs can enter a high-performance P-state. The net
result should be increase of performance, despite of the power management.

But obviously for instance ARM may do this all differenly, using only
frequency scaling.

> It seems that ultimately we need an API for telling a power-savings goal
> and constraints (latency, throughput, battery life, the screen isn't too
> dark to read) for the system to meet.  Do you hope for someone to build
> that into the kernel on top of cpufreq(9)?

Not really; as I've written before, my opinion is that most of this should
be in the user space. The CPU PM is an exception for obvious reasons. There
could be a more involved API for cpu_idle(9) though (cf [1]).

- Jukka.

[1] The Linux cpuidle-subsystem; http://lwn.net/Articles/384146/