Re: [PATCH 2/3] arm64: dts: qcom: sdm845-mtp: Add RPMh VRM/XOB regulators

[David Collins]

Hi,

On 07/10/2018 03:55 PM, Doug Anderson wrote:
> On Tue, Jul 10, 2018 at 3:32 PM, David Collins <collinsd@xxxxxxxxxxxxxx> wrote:
>> On 07/10/2018 03:02 PM, Douglas Anderson wrote:
>> ...
>>> +             vdd-s1-supply = <&vph_pwr>;
>>> +             vdd-s2-supply = <&vph_pwr>;
>>> +             vdd-s3-supply = <&vph_pwr>;
>>> +             vdd-s4-supply = <&vph_pwr>;
>>> +             vdd-s5-supply = <&vph_pwr>;
>>> +             vdd-s6-supply = <&vph_pwr>;
>>> +             vdd-s7-supply = <&vph_pwr>;
>>> +             vdd-s8-supply = <&vph_pwr>;
>>> +             vdd-s9-supply = <&vph_pwr>;
>>> +             vdd-s10-supply = <&vph_pwr>;
>>> +             vdd-s11-supply = <&vph_pwr>;
>>> +             vdd-s12-supply = <&vph_pwr>;
>>> +             vdd-s13-supply = <&vph_pwr>;
>>> +             vdd-l1-l27-supply = <&vreg_s7a_1p025>;
>>> +             vdd-l2-l8-l17-supply = <&vreg_s3a_1p35>;
>>> +             vdd-l3-l11-supply = <&vreg_s7a_1p025>;
>>> +             vdd-l4-l5-supply = <&vreg_s7a_1p025>;
>>> +             vdd-l6-supply = <&vph_pwr>;
>>> +             vdd-l7-l12-l14-l15-supply = <&vreg_s5a_2p04>;
>>> +             vdd-l9-supply = <&vreg_bob>;
>>> +             vdd-l10-l23-l25-supply = <&vreg_bob>;
>>> +             vdd-l13-l19-l21-supply = <&vreg_bob>;
>>> +             vdd-l16-l28-supply = <&vreg_bob>;
>>> +             vdd-l18-l22-supply = <&vreg_bob>;
>>> +             vdd-l20-l24-supply = <&vreg_bob>;
>>> +             vdd-l26-supply = <&vreg_s3a_1p35>;
>>> +             vin-lvs-1-2-supply = <&vreg_s4a_1p8>;
>>
>> I would suggest not specifying any of these regulator parent supplies in
>> device tree.  RPMh will be enforcing all regulator parent-child
>> dependencies.  Therefore, handling the dependencies in Linux is redundant.
>>  It will result in additional RPMh requests as well as more time spent in
>> regulator framework calls.  Overall, it will lead to slightly lower
>> performance.  Note that while specifying the parent supplies results in
>> less efficient runtime behavior, it is not technically wrong so you could
>> keep them in place if you prefer.
> 
> Interesting.  ...so RPMh will automatically turn on parent regulators
> when their children are enabled (assuming that the parent regulator is
> also RPMh controlled)?

Yes, exactly.  RPMh also ensures that the voltage of a parent regulator is
sufficient to meet minimum headroom voltage requirements of all
subregulated child regulators.


> Personally I'd still prefer to see Linux managing its own state and
> relying less on RPMh-automatic stuff, but I'd defer to Bjorn / Andy
> (or others) to override me.

Ok

Take care,
David

-- 
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project