Re: [PATCH v4 2/3] sched/fair: Take thermal pressure into account while estimating energy

[Lukasz Luba <lukasz.luba@arm.com>]

On 6/15/21 4:31 PM, Dietmar Eggemann wrote:
> On 14/06/2021 21:11, Lukasz Luba wrote:
>> Energy Aware Scheduling (EAS) needs to be able to predict the frequency
>> requests made by the SchedUtil governor to properly estimate energy used
>> in the future. It has to take into account CPUs utilization and forecast
>> Performance Domain (PD) frequency. There is a corner case when the max
>> allowed frequency might be reduced due to thermal. SchedUtil is aware of
>> that reduced frequency, so it should be taken into account also in EAS
>> estimations.
> 
> It's important to highlight that this will only fix this issue between
> schedutil and EAS when it's due to `thermal pressure` (today only via
> CPU cooling). There are other places which could restrict policy->max
> via freq_qos_update_request() and EAS will be unaware of it.

True, but for this I have some other plans.

> 
>> SchedUtil, as a CPUFreq governor, knows the maximum allowed frequency of
>> a CPU, thanks to cpufreq_driver_resolve_freq() and internal clamping
>> to 'policy::max'. SchedUtil is responsible to respect that upper limit
>> while setting the frequency through CPUFreq drivers. This effective
>> frequency is stored internally in 'sugov_policy::next_freq' and EAS has
>> to predict that value.
>>
>> In the existing code the raw value of arch_scale_cpu_capacity() is used
>> for clamping the returned CPU utilization from effective_cpu_util().
>> This patch fixes issue with too big single CPU utilization, by introducing
>> clamping to the allowed CPU capacity. The allowed CPU capacity is a CPU
>> capacity reduced by thermal pressure raw value.
>>
>> Thanks to knowledge about allowed CPU capacity, we don't get too big value
>> for a single CPU utilization, which is then added to the util sum. The
>> util sum is used as a source of information for estimating whole PD energy.
>> To avoid wrong energy estimation in EAS (due to capped frequency), make
>> sure that the calculation of util sum is aware of allowed CPU capacity.
>>
>> This thermal pressure might be visible in scenarios where the CPUs are not
>> heavily loaded, but some other component (like GPU) drastically reduced
>> available power budget and increased the SoC temperature. Thus, we still
>> use EAS for task placement and CPUs are not over-utilized.
> 
> IMHO, this means that this is catered for the IPA governor then. I'm not
> sure if this would be beneficial when another thermal governor is used?

Yes, it will be, the cpufreq_set_cur_state() is called by
thermal exported function:
thermal_cdev_update()
   __thermal_cdev_update()
     thermal_cdev_set_cur_state()
       cdev->ops->set_cur_state(cdev, target)

So it can be called not only by IPA. All governors call it, because
that's the default mechanism.

> 
> The mechanical side of the code would allow for such benefits, I just
> don't know if their CPU cooling device + thermal zone setups would cater
> for this?

Yes, it's possible. Even for custom vendor governors (modified clones
of IPA)

> 
>> Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
>> Signed-off-by: Lukasz Luba <lukasz.luba@arm.com>
>> ---
>>   kernel/sched/fair.c | 11 ++++++++---
>>   1 file changed, 8 insertions(+), 3 deletions(-)
>>
>> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
>> index 161b92aa1c79..3634e077051d 100644
>> --- a/kernel/sched/fair.c
>> +++ b/kernel/sched/fair.c
>> @@ -6527,8 +6527,11 @@ compute_energy(struct task_struct *p, int dst_cpu, struct perf_domain *pd)
>>   	struct cpumask *pd_mask = perf_domain_span(pd);
>>   	unsigned long cpu_cap = arch_scale_cpu_capacity(cpumask_first(pd_mask));
>>   	unsigned long max_util = 0, sum_util = 0;
>> +	unsigned long _cpu_cap = cpu_cap;
>>   	int cpu;
>>   
>> +	_cpu_cap -= arch_scale_thermal_pressure(cpumask_first(pd_mask));
>> +
> 
> Maybe shorter?
> 
>          struct cpumask *pd_mask = perf_domain_span(pd);
> -       unsigned long cpu_cap =
> arch_scale_cpu_capacity(cpumask_first(pd_mask));
> +       int cpu = cpumask_first(pd_mask);
> +       unsigned long cpu_cap = arch_scale_cpu_capacity(cpu);
> +       unsigned long _cpu_cap = cpu_cap - arch_scale_thermal_pressure(cpu);
>          unsigned long max_util = 0, sum_util = 0;
> -       unsigned long _cpu_cap = cpu_cap;
> -       int cpu;
> -
> -       _cpu_cap -= arch_scale_thermal_pressure(cpumask_first(pd_mask));

Could be, but still, the definitions should be sorted from longest on
top, to shortest at the bottom. I wanted to avoid modifying too many
lines with this simple patch.

> 
>>   	/*
>>   	 * The capacity state of CPUs of the current rd can be driven by CPUs
>>   	 * of another rd if they belong to the same pd. So, account for the
>> @@ -6564,8 +6567,10 @@ compute_energy(struct task_struct *p, int dst_cpu, struct perf_domain *pd)
>>   		 * is already enough to scale the EM reported power
>>   		 * consumption at the (eventually clamped) cpu_capacity.
>>   		 */
>> -		sum_util += effective_cpu_util(cpu, util_running, cpu_cap,
>> -					       ENERGY_UTIL, NULL);
>> +		cpu_util = effective_cpu_util(cpu, util_running, cpu_cap,
>> +					      ENERGY_UTIL, NULL);
>> +
>> +		sum_util += min(cpu_util, _cpu_cap);
>>   
>>   		/*
>>   		 * Performance domain frequency: utilization clamping
>> @@ -6576,7 +6581,7 @@ compute_energy(struct task_struct *p, int dst_cpu, struct perf_domain *pd)
>>   		 */
>>   		cpu_util = effective_cpu_util(cpu, util_freq, cpu_cap,
>>   					      FREQUENCY_UTIL, tsk);
>> -		max_util = max(max_util, cpu_util);
>> +		max_util = max(max_util, min(cpu_util, _cpu_cap));
>>   	}
>>   
>>   	return em_cpu_energy(pd->em_pd, max_util, sum_util);
> 
> There is IPA specific code in cpufreq_set_cur_state() ->
> get_state_freq() which accesses the EM:
> 
>      ...
>      return cpufreq_cdev->em->table[idx].frequency;
>      ...
> 
> Has it been discussed that the `per-PD max (allowed) CPU capacity` (1)
> could be stored in the EM from there so that code like the EAS wakeup
> code (compute_energy()) could retrieve this information from the EM?

No, we haven't think about this approach in these patch sets.
The EM structure given to the cpufreq_cooling device and stored in:
cpufreq_cdev->em should not be modified. There are a few places which
receive the EM, but they all should not touch it. For those clients
it's a read-only data structure.

> And there wouldn't be any need to pass (1) into the EM (like now via
> em_cpu_energy()).
> This would be signalling within the EM compared to external signalling
> via `CPU cooling -> thermal pressure <- EAS wakeup -> EM`.
> 

I see what you mean, but this might cause some issues in the design
(per-cpu scmi cpu perf control). Let's use this EM pointer gently ;)