On 09/10/2019 10:57, Parth Shah wrote:
[...]
>> On 07/10/2019 18:53, Parth Shah wrote:
>>>
>>>
>>> On 10/7/19 5:49 PM, Vincent Guittot wrote:
>>>> On Mon, 7 Oct 2019 at 10:31, Parth Shah <pa...@linux.ibm.com> wrote:
[...]
>>> Maybe I can add just below the sched_energy_present(){...} construct giving
>>> precedence to EAS? I'm asking this because I remember Patrick telling me to
>>> leverage task packing for android as well?
>>
>> I have a hard time imaging that Turbosched will be used in Android next
>> to EAS in the foreseeable future.
>>
>> First of all, EAS provides task packing already on Performance Domain
>> (PD) level (a.k.a. as cluster on traditional 2-cluster Arm/Arm64
>> big.LITTLE or DynamIQ (with Phantom domains (out of tree solution)).
>> This is where we can safe energy without harming latency.
>>
>> See the tests results under '2.1 Energy test case' in
>>
>> https://lore.kernel.org/r/20181203095628.11858-1-quentin.per...@arm.com
>>
>> There are 10 to 50 small (classified solely by task utilization) tasks
>> per test case and EAS shows an effect on energy consumption by packing
>> them onto the PD (cluster) of the small CPUs.
>>
>> And second, the CPU supported topology is different to the one you're
>> testing on.
>>
>
> cool. I was just keeping in mind the following quote
> " defining a generic spread-vs-pack wakeup policy which is something
> Android also could benefit from " (https://lkml.org/lkml/2019/6/28/628)
The main thing is that in case we want to introduce a new functionality
into CFS, we should try hard to use existing infrastructure (or
infrastructure there is agreement on that we'll need it) as much as
possible.
If I understand Patrick here correctly, he suggested not to use uclamp
but the task latency nice approach. There is agreement that we would
need something like this as infrastructure:
https://lore.kernel.org/r/20190830174944.21741-1-subhra.mazum...@oracle.com
So p->latency_nice is suitable to include your p->flags |=
PF_CAN_BE_PACKED concept nicely.
>
> BTW, IIUC that does task consolidation only on single CPU unless
> rd->overload is set, right?
Task consolidation on Performance Domains (PDs) w/ multiple CPUs (e.g.
on a per-cluster PD big.LITTLE system) only works when the system is not
overutilized:
6326 int find_energy_efficient_cpu(struct task_struct *p, int prev_cpu)
6327 {
...
6337 if (!pd || *READ_ONCE(rd->overutilized)*)
6338 goto fail;
...
[...]
