On 10/16/20 5:02 PM, Quentin Perret wrote:

On Friday 16 Oct 2020 at 15:42:57 (+0100), Lukasz Luba wrote:

quoted

Do you mean a new entry in DT which will be always below
'dynamic-power-coefficient' and/or 'sustainable-power' saying the unit
of above value?

Yes, something like that.

quoted

There was discussion with Rob (and Doug) about this. I got the
impression he was against any new DT stuff [1].
We don't have to, I think we all agree that DT will only support mW.

Right, I agree this is a 'nice-to-have'.

quoted

I have agreed to this idea having a 'flag' inside EM [2], which
indicates the mW or bogoWatts. It could be set via API:
em_dev_register_perf_domain() and this new last argument.

I can write that patch. There is only two usage (3rd is on LKML) of
that function. The DT way, which is via:
dev_pm_opp_of_register_em() will always set 'true';
Driver direct calls of em_dev_register_perf_domain(), will have to
set appropriate value ('true' or 'false'). The EM struct em_perf_domain
will have the new bool field set based on that.
Is it make sense?

I had something more complicated in mind, where units are arbitrary
('milliwats', 'scmi-bogowatts', ...) as that would help if units can be
specified in the DT too, but if we don't care about that then yes I
suppose a boolean flag should do.

Thank you Quentin for help in sorting this out.
I'll send the v3.

Regards,
Lukasz

Thanks!
Quentin