Re: [PATCH v2 04/11] sched: Allow all archs to set the power_orig

[Peter Zijlstra]

On Fri, May 30, 2014 at 03:04:32PM +0100, Dietmar Eggemann wrote:
> On 23/05/14 16:52, Vincent Guittot wrote:
> > power_orig is only changed for system with a SMT sched_domain level in order to
> > reflect the lower capacity of CPUs. Heterogenous system also have to reflect an
> > original capacity that is different from the default value.
> > 
> > Create a more generic function arch_scale_cpu_power that can be also used by
> > non SMT platform to set power_orig.
> > 
> > The weak behavior of arch_scale_cpu_power is the previous SMT one in order to
> > keep backward compatibility in the use of power_orig.
> > 
> > Signed-off-by: Vincent Guittot <vincent.guittot@xxxxxxxxxx>
> 
> As you know, besides uarch scaled cpu power for HMP, freq scaled cpu
> power is important for energy-aware scheduling to achieve freq scale
> invariance for task load.
> 
> I know that your patch-set is not about introducing freq scaled cpu
> power, but we were discussing how this can be achieved w/ your patch-set
> in place, so maybe you can share your opinion regarding the easiest way
> to achieve freq scale invariance with us?
> 
> (1) We assume that the current way (update_cpu_power() calls
> arch_scale_freq_power() to get the avg power(freq) over the time period
> since the last call to arch_scale_freq_power()) is suitable
> for us. Do you have another opinion here?
> 
> (2) Is the current layout of update_cpu_power() adequate for this, where
> we scale power_orig related to freq and then related to rt/(irq):
> 
>   power_orig = scale_cpu(SCHED_POWER_SCALE)
>   power = scale_rt(scale_freq(power_orig))
> 
> or do we need an extra power_freq data member on the rq and do:
> 
>   power_orig = scale_cpu(SCHED_POWER_SCALE)
>   power_freq = scale_freq(power_orig))
>   power = scale_rt(power_orig))
> 
> In other words, do we consider rt/(irq) pressure when calculating freq
> scale invariant task load or not?

I don't think you should. The work done depends on the frequency, not on
other tasks present on the cpu. The same is true for an over-utilized
cpu, a task will run less than the desired amount of time, this is no
different from a RT/irq preempting the task and taking its time.

Attachment: pgpV_EGvdQj0C.pgp
Description: PGP signature